Icelandic and Faroese: A usage-based cognitive analysis of morphological change

Jón Símon Markússon

Icelandic and Faroese: A usage-based cognitive analysis of morphological change

2024, University of Icelandic School of Humanities

Íslensku-og menningardeild Háskóla Íslands hefur metið ritgerð þessa haefa til varnar við doktorspróf í íslenskri málfraeði Reykjavík, 22. maí 2024 Gauti Kristmannsson deildarforseti The Faculty of Icelandic and Comparative Cultural Studies at the University of Iceland has declared this dissertation eligible for defence leading to a Ph.D. degree in Icelandic Linguistics

Icelandic and Faroese A usage-based cognitive analysis of morphological change Jón Símon Markússon Dissertation towards the degree of Doctor of Philosophy 2024 SCHOOL OF HUMANITIES FACULTY OF ICELANDIC AND COMPARATIVE CULTURAL STUDIES Íslensku- og menningardeild Háskóla Íslands hefur metið ritgerð þessa hæfa til varnar við doktorspróf í íslenskri málfræði Reykjavík, 22. maí 2024 Gauti Kristmannsson deildarforseti The Faculty of Icelandic and Comparative Cultural Studies at the University of Iceland has declared this dissertation eligible for defence leading to a Ph.D. degree in Icelandic Linguistics Doctoral Committee: Þórhallur Eyþórsson, supervisor Hjalmar P. Petersen Katrín Axelsdóttir Icelandic and Faroese: A usage-based cognitive analysis of morphological change. © Jón Símon Markússon Reykjavik 2024 Dissertation for a doctoral degree at the University of Iceland. All rights reserved. No part of this publication may be reproduced in any form without written permission of the author. This thesis was supported by The Icelandic Research Fund, grant no. 174253-015. ISBN 978-9935-9736-4-1 ORCID orcid.org/0000-0002-3280-2652 2 Abstract The current thesis presents three published articles on inflectional change in Insular Nordic (Icelandic and Faroese). Papers I and II deal with change in Icelandic, while Paper III focuses on Faroese. The three articles are related through employment of the usage-based cognitive approach, which views the structure of grammar as emergent from prior linguistic experience, assuming a central role for language use as the mechanism of linguistic innovation and change. Usage-based cognitive studies typically invoke factors such as frequency and schematicity to account for the varying degrees of productivity that inflectional classes exhibit cross-linguistically. Such studies also make recourse to domaingeneral cognitive processes like analogy, categorisation, entrenchment, and statistical learning as determinants in the direction of change. Crucially, the usage-based cognitive approach posits rich memory for language. Thus, the cognitive prerequisites for storage and utility of linguistic experience as informative of usage choices are no different from those which inform our interactions with the wider world generally. Papers I and II account for the limited productivity of the low frequency Xó/æTmicroclass which, before the addition of borrowed blók ‘wretch, non-entity’, contained five Icelandic feminine nouns in nominative/accusative plural -ur: bók ‘book’, bót ‘patch’, brók ‘trousers’, nót ‘fishing net’, and rót ‘root’ only, cf. plural bækur, bætur, brækur, nætur, rætur. Productivity is equated with the rate at which feminine grammatical gender is assigned to masculine nouns, while the motivation for such treatment is considered to be phonetic coherence with varyingly schematic feminine classes in plural -ur. Specifically, Paper I accounts for the different rates at which Icelandic masculine plural forms in with final -ur –– be that sequence an ending or part of the stem etymologically –– undergo reanalysis as feminine. Crucially, around 15% of nouns in Icelandic end in plural -ur: almost 92% of these are feminine, while all others are masculine. Further, syncretism in nominative/accusative plural is relatively rare among masculine nouns, but exceptionless among feminines. Also without exception, the relevant forms in plural -ur are always syncretic, irrespective of a noun’s gender. Interestingly, in the minority of cases, plurals such as masculine eigendur ‘owners’, fætur ‘feet’, and vetur ‘winters’ alternate with overtly feminine definite forms such as pl.def. eigendurnar, fæturnar, veturnar, cf. original and more frequent masc.nom.pl.def. iii eigendurnir, fæturnir, veturnir, masc.acc.pl.def. eigendurna, fæturna, veturna. Additionally, masculine forms in plural -ur sometimes occur with feminine modifiers and determiners. Paper I argues that, given the highly schematic nature of the full set of nouns in plural -ur, reanalysis as feminine might be expected at a rate proportionate to the frequency of corresponding masculine forms –– all other things being equal. However, based on corpus data for Icelandic, Paper I reports a mismatch in frequency between sets of doublets defined in terms of gender. Through employment of Bybee’s network model, with some innovative notational features, Paper I demonstrates that graded phonetic structure of a broader feminine subtype in plural -ur –– as it centres around the Xó/æT-microclass –– impacts the rate of reanalysis by means of a gang effect, which is viewed as a function of analogy, i.e. the process by which existing knowledge is extended to new contexts. In a similar vein, Paper II examines the limited productivity of the Icelandic Xó/æTmicroclass. In the article, productivity is equated with the occasional inflection of feminine blók ‘nonentity’ and forms of neuter kók ‘CokeTM’ according to the morphophonological alternation exhibited by e.g. sg. bók ~ pl. bækur, rót ~ rætur, cf. blók ~ blækur, kók ~ kækur. These new plural forms are taken as evidence for the –– albeit highly limited –– productivity of the microclass. Indeed, doublet forms in both paradigms also pattern with inflection classes of higher type frequency, cf. pl. blókir, kókir, like fem. pl. myndir ‘pictures’, both of which prove more frequent than plural blækur and kækur. According to the usage-based cognitive approach to language, the impact of varyingly large and varyingly schematic classes is indeed expected to correlate with graded degrees of productivity. Some have implied contrastive motivation and, therefore also, distinct cognitive mechanisms for the deduction of plural blækur from sg. blók and of plural kækur from sg. kók. Specifically, the opinion has been expressed that plural kækur belies “real” language use because the form only occurs in humorous contexts. Plural blækur, on the other hand, is considered “real” language use. However, it is clear that both forms are based on the pattern of alternation exemplified by e.g. sg. bók ~ pl. bækur. Therefore, Paper II seeks to dispel the idea that different motivations for deduction are at play as a misunderstanding of analogy. This objective is achieved through reference to schematicity, semantics, and pragmatics, as well as to Icelandic corpus data. By this means, Paper II demonstrates that new membership in the microclass is graded as a function of limited productivity. Finally, Paper II shows that innovative kækur and blækur are deduced by identical means, i.e. via analogy. Paper III deals with analogical change in Faroese, with specific focus on levelling of the intricate vowel alternations which characterised the inflection of Old West Nordic u-stems. iv Crucially, levelling is defined as the extension of a stem variant to a cell in which it did not occur previously, as opposed to one form “changing into” another. The Faroese descendants of Old West Nordic u-stems have undergone significant levelling, as is evident from the paradigm of Far. vøllur ‘field, grassy ledge on a rock face, (sports) pitch, airport’: all cells of the paradigm contain the variant vøll- (< OWN nom.sg., acc.sg., acc.pl., dat.pl. vǫll-), while some also contain vall- (OWN gen.sg., gen.pl. vall-). The variant OWN (dat.sg., nom.pl.) vell- has been completely eradicated (see Markússon 2022b). Conversely, the paradigm of Far. fjørður ‘fjord, inlet/bay, sound/strait’ has retained all stem variants, cf. Far. fjørð- (< OWN nom.sg., acc.sg., acc.pl., dat.pl. fjǫrð-), firð- (< OWN dat.sg., nom.pl. firð-), fjarð- (< OWN gen.sg., gen.pl. fjarð-). Further, it has extended those variants to other cells of the paradigm, cf. innovative dat.sg. fjørði, which exists beside older firði, innovative nom./acc.pl. fjørðir and fjarðir beside older firðir, and innovative dat.pl. fjarðum and firðum, which live alongside older fjørðum. The basic forms of paradigms, i.e. those from which new inflectional forms take their stem, and the factors that establish them have typically been defined according to either of two opposing theoretical approaches. The first attributes basic status on account of so-called ‘markedness’, i.e. the perspective that new forms in a paradigm are likely to be based on existing ones that express semantically “natural” and/or “neutral” values, such as singular and nominative. In other words, such “unmarked” forms serve as basic. The opposing approach posits frequency as the determining factor. Thus, levelling proceeds from the most frequent member(s) of the paradigm, due to a correlation between frequency of use and its impact on the strength of representation in memory. In other words: frequent forms are better represented than less frequent forms and more readily accessible in moments of memory lapse. Therefore, frequent forms are most likely to be used as a base when the “correct” form evades the language user. Paper III utilises Faroese corpus data in order to demonstrate that the basic forms of Far. vøllur and fjørður –– both of which refer to topographical entities and occur as complex place names –– are established on the basis of frequency, rather than semantics. Paper III argues that due to the overall low frequency of forms of Far. vøllur, the most frequent stem variant, i.e. vøll-, was extended to the whole paradigm, while vell- was easily forgotten. Conversely, the high token frequency of dat.sg. firði meant that it was well represented in memory and, therefore, easily accessible in moments of temporary memory lapse. Paper III argues that this property of dat.sg. firði triggered spread of the stem variant firð- to the dative plural through the context [í/á/úr + dat.], where younger dat.pl. firðum takes older fjørðum over in v frequency. Further, presence of the stem variant firð- in all plural cells of the paradigm, cf. also nom./acc.pl. firðir, facilitated association of the form firð- with the meaning plural. Subsequently, an attempt was made to level the singular portion of the paradigm in favour of the variant fjørð- to contrive the formal opposition sg. fjørð- : pl. firð-. However, the new form never took over the role of older dat.sg. firði on account of the high token frequency of the latter. The current thesis demonstrates that factors such as frequency and schematicity impact choices made in on-line language use as a function of stored experience with language. Further, if the course of language change correlates with the distributional properties of these factors in the acquired grammar, it follows logically that the usage events which incrementally facilitate change reflect the linguistic experience whence the grammar emerges. Moreover, the fact that language change is a function of language use demonstrates that the structure of grammar is an emergent and dynamic system, rather than one whose adaptive properties are constrained by genetic endowment and ontogeny. Therefore, in light of the conclusions drawn in Papers I–III, the current thesis also showcases the applicability of usage-based cognitive theory as a means to account for the direction of morphological change. vi Ágrip Ritgerð þessi fjallar um þrjár rannsóknir, sem hver hefur verið gefin út í sér grein í viðurkenndu tímariti um málvísindi. Greinarnar þrjár eru birtar strax á eftir 7. kafla. Í köflum 1–7 er vísað í greinarnar sem Paper I (1. greinin), Paper II (2. greinin) og Paper III (3. greinin). Í fyrstu greinunum tveimur er hugað að beygingarbreytingum og -nýjungum í íslensku en í þeirri þriðju er einblínt á áþekkt efni í færeysku. Í hverri grein var beitt málnotkunarnálgun á viðkomandi efni en slík nálgun gerir áhrifum almennra hugrænna ferla hátt undir höfði, s.s. útvíkkun hliðstæðrar þekkingar (e. analogy), flokkunar (e. categorisation), rótfestu (e. entrenchment) og hlutfallslegs lærdóms (e. statistical learning), á bæði málnotkun og stefnu málbreytinga. Þar að auki skera þættir sem ekki varða málkerfið, s.s. notkunartíðni og formleg líkindi ólíkra eininga, úr um stefnu þessara ferla að töluverðu leyti. Þannig er málkerfið álitið vera ævinlega í mótun á grundvelli fyrri reynslu af málinu. Til þess að málnotkun og málbreytingar endurspegli áhrif umræddra hugrænna þátta er nauðsynlegt að gera ráð fyrir ríkulegu minni fyrir mannlegt mál (e. rich memory for language). Þannig séu hugrænar forsendur fyrir geymslu málfræðilegra upplýsinga þær sömu og eru fyrir geymslu annars konar reynslu í minni. Fyrstu greinarnar tvær fjalla um takmarkaða virkni kvenkyns Xó/æT-örflokksins (e. microclass), sem hefur lága mynsturstíðni (e. type frequency). Á forníslenskum tíma samanstóð flokkurinn af nafnorðunum bók, bót, brók, glóð, nót og rót, sbr. ft. bækur, bætur, brækur, nætur, rætur. Virkni er ákvörðuð á grundvelli þess hve oft málfræðilegu kvenkyni er úthlutað til upprunalegs karlkynsorðs og er ástæða úthlutunarinnar talin vera misjafnlega mikil hljóðfræðileg líkindi við nafnorð úr ólíkum kvenkynsflokkum með ft. -ur. Í þessu sambandi fjallar fyrsta greinin um hversu oft ólíkar karlkynsmyndir með umræddri endingu – – hvort sem sama runa reynist beygingarending eða hluti af stofni –– eru umtúlkaðar sem kvenkynsmyndir. Um 15% nafnorða í íslensku fá endinguna ft. -ur en næstum því 92% þeirra eru kvenkyns; önnur eru karlkyns. Þar að auki eru samhljóða myndir í nefnifalli og þolfalli fleirtölu hjá karlkynsorðum afar sjaldgæfar, en samsvarandi myndir kvenkynsnafnorða eru undantekningarlaust samhljóða (e. syncretic). Enn fremur eru nefnifalls- og þolfallsmyndir með ft. -ur afdráttarlaust samhljóða burtséð frá málfræðilegu kyni. Áhugavert er að karlkynsmyndir eins og ft. eigendur, fætur, vetur víxlast á í beygingu við ákveðnar myndir sem að forminu til eru tvímælalaust kvenkyns, sbr. kvk.ft.mgr. vii eigendurnar, fæturnar, veturnar, í stað upprunalegu karlkynsmyndanna kk.nf.ft.mgr. eigendurnir, fæturnir, veturnir, kk.þf.ft.mgr. eigendurna, fæturna, veturna. Auk þess eru stundum notaðar kvenkynsmyndir af ákvæðisorðum með karlkynsmyndum með ft. -ur. Í ljósi þess að nafnorð með endingunni ft. -ur geta verið hljóðfræðilega mjög ólík er í fyrstu greininni velt vöngum yfir því hvers vegna umtúlkunartíðni karlkynsmynda á ft. -ur sem kvenkyns komi ekki heim og saman við hlutfallslega tíðni karlkynsmyndanna sjálfra. Um þetta er staðhæft að búast mætti við samsvörun að öllu jöfnu. Samkvæmt leitarniðurstöðum úr textasöfnum fyrir íslensku er þó tíðni karlkynsmynda annars vegar og samsvarandi kvenkynsmynda hins vegar hlutfallslega ólík. Í fyrstu greininni er beitt tengslanetsnálgun (e. Network Model) sem byggð er að mestu leyti á umfjöllun Bybee um verðandi (e. emergent) eðli málkefisins. Tengslanetin sem höfð eru í þessari ritgerð lýsa því hvernig hljóðfræðileg gerð kvenkynsundirflokksins í ft. -ur einkennist af samfellu (e. gradation). Þannig stuðla likindi við nafnorðin sem skipa Xó/æTörflokkinn að aukinni umtúlkunartíðni vegna áhrifa svokallaðs orðagengis (e. lexical gang). Áhrif örflokksins (e. microclass) eru álitin stafa af útvíkkun hliðstæðrar þekkingar, þ.e. ferlinu sem útvíkkar þekkingu til nýs samhengis. Í svipuðum anda athugar önnur greinin takmarkaða virkni hins íslenska Xó/æTörflokksins en þar er sett samasemmerki milli virkni og beygingar tökuorðanna kvk. blók og hk. kók samkvæmt beygingarmynstrinu sem t.d. bók fylgir, sbr. et. bók ~ ft. bækur, blók ~ blækur, kók ~ kækur. Talið er að hinar nýju fleirtölumyndir beri vitni um (að vísu mjög takmarkaða) virkni örflokksins, enda koma einnig fyrir fleirtölumyndir eftir beygingarflokkum kvenkynsorða sem hafa hærri mynsturstíðni, sbr. ft. blókir, kókir, eins og ft. myndir. Fleirtölumyndirnar með endingunni -ir eru algengari en blækur og kækur en þetta er álitið sýna að tíðni og virkni haldist í hendur og að sambandið þar á milli ráði miklu um stefnu málbreytinga og fótfestuna sem nýjungar ná í málinu. Önnur greinin bendir á að sumir álíta ólíkar hugrænar forsendur vera fyrir myndun fleirtölumyndanna ft. blækur og kækur. Í þessu sambandi hefur verið látið í veðri vaka að myndun ft. kækur endurspegli ekki „raunverulega“ málnotkun af því að viðkomandi mynd er aðeins notuð í gríni. Hins vegar sé ft. blækur dæmi um „raunverulega“ málnotkun. Gegn þessu viðhorfi sýnir önnur greinin fram á að báðar fleirtölumyndir hafa sama beygingarmynstur að fyrirmynd, þ.e. et. bók ~ ft. bækur. Af þessari ástæðu leitast önnur greinin við að hrekja þá hugmynd að ólíkar forsendur séu fyrir myndun fleirtölumyndanna sem í hlut eiga og að hugmyndin sjálf stafi af misskilningi á því hvernig málfræðileg þekking er útvíkkuð til nýs samhengis. Þessu markmiði er náð með vísun til hljóðfræðilegra gerða viii ólíkra nafnorða, merkingar, málnotkunarfræði, auk niðurstaðna úr textasöfnum fyrir íslensku. Þannig eru nafnorð tekin inn í Xó/æT-örflokkinn með samfelldum hætti (e. in graded fashion) og það er talið endurspegla takmarkaða virkni flokksins. Komist er að þeirri niðurstöðu að bæði ft. blækur og kækur séu myndaðar við útvíkkun hliðstæðrar þekkingar. Þriðja greinin einblínir að útjöfnun (e. levelling) í beygingardæmum færeyskra karlkynsnafnorða, sem á fornvesturnorrænum tíma tilheyrðu flokki karlkyns u-stofna. Hér er útjöfnun talin hafa átt sér stað þegar stofnmynd tiltekins orð kemst inn í annan bás beygingardæmisins þar sem hana var ekki að finna áður. Lykilatriði í þessari skilgreiningu á utjöfnun er að ferlið felur ekki í sér breytingu eldri myndar í nýja, heldur viðbót við þær beygingarmyndir sem fyrir voru. Færeyskir afkomendur fornvesturnorrænna u-stofna hafa sætt mikilli útjöfnun en eins og beyging fær. vøllur ‘völlur’ sýnir er stofnmyndin vøll- (< fvn. nf.et., þf.et., þf.ft., þgf.ft. vǫll-) komin inn í allar beygingarmyndir orðsins þó að sumir básar hafi einnig stofnmyndina vall- (< fvn. ef.et., ef.ft. vall-) sem ekki höfðu hana áður. Aftur á móti er stofnmyndin fvn. þgf.et., nf.ft. vell- með öllu horfið úr beygingardæminu. Aftur á móti hefur beygingardæmi nafnorðsins fær. fjørður ‘fjörður’ haldið öllum eldri stofnmyndunum sínum, sbr. fær. fjørð- (< fvn. nf.et., þf.et., þf.ft., þgf.ft. fjǫrð-), firð- (< fvn. þgf.ft., nf.ft. firð-), fjarð- (< fvn. ef.et., ef.ft. fjarð-). Enn fremur hafa allar stofnmyndir orðsins verið útvíkkaðar til bása innan beygingardæmisins þar sem þær komu ekki fyrir áður, sbr. nýjungarnar þgf.et. fjørði, yngri mynd sem lifir við hlið eldra firði og yngri myndirnar nf./þf.ft. fjørðir og fjarðir sem stundum leysa eldri firðir af hólmi og yngri myndirnar fjarðum og firðum sem eru í notkun ásamt eldri fjørðum. Aðallega hefur verið miðað við tvær ólíkar kenningar um ákvörðun grunnmynda (e. basic forms), þ.e. þeirra mynda sem liggja formlega til grundvallar við myndun nýrra beygingarmynda innan beygingardæmis, og áhrifþættina sem stuðla að vali á grunnmyndum. Fyrri kenningin eignar valið stöðu mynda m.t.t. svokallaðrar mörkunar; samvkæmt því eru nýjar myndir sem til verða við útjöfnun taldar líklegri til að tjá merkingarfræðilega „eðlileg“ og/eða „hlutlaus“ gildi eins og eintölu og nefnifall. Á hinn bóginn gerir önnur kenning ráð fyrir að tíðni stuðli að vali á grunnmyndum, þannig að útjöfnun stefni út frá þeim myndum sem koma oftast fyrir í töluðu máli og/eða textasöfnum. Samkvæmt seinni kenningunni ákvarðar tíðni beygingarmyndar rótfestu (e. entrenchment) hennar, þ.e. hversu vel hugrænn fulltrúi viðkomandi myndar hefur búið um sig í minni. Með öðrum orðum eru algengar myndir rótfastari í minni en sjaldgæfari myndir og eru þær fyrrnefndu þ.a.l. tiltækari til notkunar en þær síðarnefndu þegar minnið bregst málnotandanum (þó ekki nema ix stundarkorn). Þess vegna séu algengustu beygingarmyndir orðs líklegastar til að liggja til grundvallar við myndun nýrra mynda við útjöfnun. Þriðja greinin styðst við innihald textasafna og beitir málnotkunarnálgun til að sýna fram á að grunnmyndir færeysku nafnorðanna vøllur og fjørður –– sem bæði vísa til landslags og koma fyrir í samsettum staðanöfnum –– má ákvarða á grundvelli tíðni, frekar en merkingar. Þar eru færð rök fyrir því að vegna staktíðni (e. token frequency) ólíkra mynda af orðinu vøllur, sem almennt er lág, hafi algengasta stofnmyndin, þ.e. vøll-, verið nýtt til grundvallar við myndun á nýjum beygingarmyndum en stofnmyndin vell- hafi síðan gleymst með tímanum. Aftur á móti varð há staktíðni myndarinnar þgf.et. firði, af nafnorðinu fær. fjørður, til þess að hún bjó vel um sig í minni og reyndist þ.a.l. tiltækari öðrum myndum þegar minnið brást. Af þessari ástæðu gerir þriðja greinin ráð fyrir að há staktíðni þgf.et. firði hafi jafnvel snemma stuðlað að útbreiðslu viðkomandi stofnmyndar til báss þágufallsmyndarinnar í fleirtölu í gegnum skemað (e. schema) [í/á/úr + þgf.]. Eftir að yngri myndin þgf.ft. firðum er orðin til tekur notkun hennar fram úr staktíðni eldri þgf.ft. fjørðum. Stofnmyndin firð- í fleirtöluhluta beygingardæmisins er komin inn í alla fleirtölubása, sbr. nf./þf.ft. firðir, þgf.ft. firðum, og tilvist hennar stuðlar enn fremur að sambandi forms og merkingar þar sem stofnmyndin firð- er gædd merkingunni fleirtölu. Gert er ráð fyrir að málnotendur bregðist síðan við þessu sambandi með tilraun til að jafna eintöluhluta beygingardæmisins út þannig að stofnmyndinni fjørð- sé skotið inn í bás þágufallsmyndarinnar í eintölu, þar sem nf.et. fjørður og þf.ft. fjørð var þegar að finna. Með öðrum orðum sé ætlunin að koma á andstæðunni et. fjørð : ft. firð-. En þrátt fyrir þessa viðleitni reynist eldri þgf.et. firði áfram algengari en yngri fjørði vegna staktíðni fyrrnefndu myndarinnar og þótt sú síðarnefnda lifi enn í málinu er hún enn tiltölulega sjaldgæf. Ritgerð þessi sýnir fram á að þættir á borð við tíðni og líkindi við hljóðfræðilega gerð annarra orðmynda geta mörgu ráðið hvað varðar málnotkun, auk þess að þessir áhrifaþættir eiga sér rætur í fyrri reynslu við málið. Enn fremur eru færð rök fyrir því að haldist stefna málbreytinga í hendur við áðurnefnda þætti þá sé rökrétt að líta svo á að málkerfið sé í stöðugri mótun á grundvelli sömu þáttanna. Með öðrum orðum renna rannsóknirnar sem hér eru til umfjöllunar stoðum undir það sjónarmið að málkerfið sé verðandi (e. emergent) fyrirbæri frekar en að mótun þess sé takmörkuð á grundvelli líf- og erfðafræðilegra þátta. Þar af leiðandi og í ljósi niðurstaðnanna sem greinarnar þrjár varpa fram má líta á þessa ritgerð sem innlegg í innlenda og alþjóðlega umræðu um áhrif almennra hugrænna þátta á málnotkun og stefnu málbreytinga, auk þess að sýna nytsemi málnotkunarnálgunar í rannsóknum á beygingarbreytingum í eyjanorrænu. x Table of contents Abstract .................................................................................................................................... iii Ágrip........................................................................................................................................ vii Acknowledgments ................................................................................................................. xiii 1 Introduction ......................................................................................................................... 1 1.1 Focus and general research questions ........................................................................... 1 1.2 Aims and individual research questions of Papers I–III................................................ 7 1.3 General contents of the thesis ........................................................................................ 8 2 Insular Nordic ...................................................................................................................... 9 3 Usage-based cognitive grammar and language change ................................................. 15 3.1 Two opposing theoretical approaches to language change ......................................... 15 3.2 The usage-based cognitive approach ........................................................................... 22 3.2.1 Rich memory for language............................................................................... 23 3.2.2 Usage and the structure of grammar ............................................................... 29 3.2.3 Schematicity and its interaction with frequency .............................................. 36 3.2.4 Categories and the principle of cognitive economy ........................................ 38 3.2.5 Section summary .............................................................................................. 40 3.3 Analogy and the adaptive nature of categories ........................................................... 40 3.4 Chapter summary ........................................................................................................ 47 4 Methodology ...................................................................................................................... 49 4.1 The methodology employed in Paper I ....................................................................... 49 4.2 The methodology employed in Paper II ...................................................................... 52 4.3 The methodology employed in Paper III ..................................................................... 54 5 Icelandic data: Papers I and II......................................................................................... 57 5.1 Content and focus of Papers I and II ........................................................................... 58 5.2 Prototype structure that centres on the Xó/æT-microclass .......................................... 61 5.3 Paper I: Reanalysis of masculine forms in plural -ur as feminine ............................... 66 5.3.1 Categorisation as a function of statistical learning ........................................ 67 5.3.2 Elaborating on taxonomies of increasing schematicity ................................... 72 5.3.3 Language use and hierarchies in linguistic categories ................................... 78 5.3.4 Productivity: Types and degrees of similarity ................................................. 79 5.3.5 The ‘net effect’ ................................................................................................. 83 5.3.6 Section summary .............................................................................................. 89 xi 5.4 Paper II: Deduction of plural blækur and kækur ......................................................... 89 5.4.1 Focus: Semantics, pragmatics, and schematicity ............................................ 89 5.4.2 Deduction as a two-step process ..................................................................... 91 5.4.3 Different mechanisms of analogy: Ice. kók...................................................... 93 5.4.4 Parallel deduction of plural blækur and kækur ............................................... 96 5.4.5 Section summary .............................................................................................. 98 5.5 Chapter summary ........................................................................................................ 98 6 Faroese: Frequency as the determinant of levelling..................................................... 101 6.1 Paper III: Content and theoretical focus .................................................................... 101 6.2 The direction of levelling .......................................................................................... 103 6.2.1 Levelling in Far. vøllur and fjørður .............................................................. 103 6.2.2 The choice of basic form(s)............................................................................ 105 6.2.3 Section summary ............................................................................................ 108 6.3 Determining the basic forms of Far. vøllur and fjørður ............................................ 109 6.3.1 Far. vøllur ...................................................................................................... 109 6.3.2 Far. fjørður .................................................................................................... 111 6.3.3 Section summary ............................................................................................ 114 6.4 Chapter summary ...................................................................................................... 115 7 Conclusion ........................................................................................................................ 117 Paper I .................................................................................................................................... 125 Paper II .................................................................................................................................. 153 Paper III ................................................................................................................................. 185 Bibliography .......................................................................................................................... 221 xii Acknowledgments Completion of the current thesis would not have been possible without the guidance of Þórhallur Eyþórsson (my supervisor), Hjalmar P. Petersen, and Katrín Axelsdóttir (doctoral committee members). I owe them the greatest debt of thanks for assistance with everything from funding to submission of the final product. I also express profound thanks to former teachers: Daisy L. Neijmann, Guðrún Þórhallsdóttir, Gunnþórunn Guðmundsdóttir, Jón Axel Harðarson, Kristján Árnason, and Margrét Jónsdóttir. I wish to acknowledge the input of my friends and colleagues in the Department of Icelandic as a Second Language (University of Iceland), who have provided support and encouragement, in addition to showing understanding when the call of the thesis required my withdrawal into the shadows. The route to completion would not have been as scenic without the impact of my wonderful friends from Aarhus University: Peter Bakker, Kristoffer Friis Bøegh, Joshua Nash, and Jeroen Willemsen. Thank you for making me feel like I belonged to the department in 2018 and for your continuing friendship. For the same reasons, I am grateful to Turið Sigurðardóttir and all at Føroyamálsdeild, University of the Faroe Islands, for accommodating me on my regular work visits from 2017–2019. I also owe a great debt of thanks to my friend Charles Gittins, for his invaluable assistance with formatting before submission. On a personal note, I wish to express gratitude and love to my Danish family: Margit and Søren, Thomas and Annabel, Rosalina, and Solveig. And to my Icelandic family: Anna, Björg, Oddrún, Óli, Ólöf (Sandra Hermannsdóttir); Ólöf (Matthíasdóttir) and Skúli; Brynja (amma); Erna Björk. I thank my dear friend, Gabríela, for showing in 2015 that all creative work is art. Thank you for accepting me into your lives and for becoming part of mine. I shall ever be lucky to have known nothing but unfaltering love, acceptance, and support from my family: Angela Irving, Mark Cretton, Paul Mariner, Christian Cretton, Rebecca Mariner, John Irving, Beckie Cretton, and bellends 1–4. To my beautiful Jesper. This process would literally not have been possible without you and the bubbas by my side. I dedicate this thesis to the memory of Anne Irving and John & Pat Cretton, my dearly departed grandparents. With love, Jón Símon Work on this thesis was funded by The Icelandic Research Fund, grant no. 174253-015. xiii 1 Introduction This chapter introduces the focus of the thesis, setting out the material chosen for analysis within the context of the theoretical approach employed. The contents of the current chapter are as follows. In Section 1.1, I flesh out the thematic focus and state my objectives. In 1.2, I reiterate the individual research questions and specific aims as stated in the respective studies, which form the basis of the thesis. Section 1.3 presents the general contents of the thesis, with respect to the order of subsequent chapters. 1.1 Focus and general research questions The current thesis elaborates on three original articles, each of which focuses on morphological change in Icelandic or Faroese. Referred to collectively in the current thesis as Insular Nordic, the two languages constitute West Nordic sisters within the North Germanic branch of the Germanic language family (see e.g. Vikør 2001:32–33). Throughout, the articles are referred to as Paper I (PI), Paper II (PII), and Paper III (PIII), respectively. 1 Each article has been published in a well-established, peer-reviewed journal of linguistics. The precise bibliographical entries for Papers I–III are given below. I. Markússon, Jón Símon. (2023a). Accounting for different rates of gender reanalysis among Icelandic masculine forms in plural -ur. Nordic Journal of Linguistics. <https://doi.org/10.1017/S0332586522000166> [Published online 2022.] II. Markússon, Jón Símon. (2022a). Tvær blækur labba inn á bar: On limited productivity as graded membership of an Icelandic microclass. NOWELE, 75(2), 194–222. <https://doi.org/10.1075/nowele.00068.mar> 1 When referring to a specific page in an article, the notation e.g. PI:5 will be used and is to be read ‘On page 5 of Paper I’. When referring to an endnote or footnote within an article, the notation e.g. PII:201, footnote 6 will be used, to be read ‘footnote 6 in Paper II, which occurs on page 201 of the same article’. The reader is reminded that Paper I contains endnotes, rather than footnotes. When referencing a table in an article, the notation e.g. PIII:65, Table 3 will be used and should be read ‘Table 3 in Paper III, which occurs on page 65 of the same article’. 1 Jón Símon Markússon III. Markússon, Jón Símon. (2022b). Um áhrif tíðni á stefnu útjöfnunar: Rannsókn á beygingarþróun færeysku nafnorðanna vøllur og fjørður. Íslenskt mál og almenn málfræði, 44, 53–86. As can be discerned from the bibliographical information above, Papers I and II are written in English, while Paper III is in Icelandic. Each study was conducted in accordance with the principles and axioms of the usage-based cognitive approach to language change. Papers I and II discuss different inflection classes of Icelandic and the limited productivity associated with some of these. In particular, both articles account for productivity associated with the Icelandic ending nominative/accusative plural -ur (henceforth ‘plural -ur’) and feminine subclasses that show that ending. Specifically, Paper I accounts for the different rates at which masculine forms in plural -ur are reanalysed as feminine due to its high rate of dispersion among Icelandic feminine nouns. Throughout, dispersion refers to the distinct morphological contexts in which a given exponent occurs (following Gries and Ellis 2015:232). Paper II deals with graded membership of the borrowings Ice. fem. blók ‘non-entity, wretch’ and. neut./fem. kók ‘CokeTM’ in several feminine classes as a function of minimal schematicity, i.e. a high degree of phonetic similarity to the members of a given class.2 In line with e.g. Bybee (2010), the gauge for productivity is the number of items to which an inflectional pattern is extended. Further, in line with Barðdal (2008:1), it is argued that productivity is in part determined by the number of items attracted to a given inflection class on the basis of phonetic coherence with existing members. In other words, the measure of productivity is also predicated on schematicity, a property of an inflection class that correlates with the extensibility of the relevant inflectional pattern. It has been demonstrated that those inflection classes of a language that contain the most members and whose membership is characterised by high schematicity prove the most productive. This property is considered a function of the ease with which highly frequent patterns are relatively more practiced and, therefore, more easily applied in new contexts. Likewise, a lack of formal constraints on a class of items facilitates extension of membership to new items of practically any form (e.g. Bybee 2010:67). However, morphological productivity is a demonstrably graded phenomenon and Papers I and II are concerned with the limited productivity that can be exhibited by classes 2 Throughout, I follow Bybee’s (2010:67) definition of schematicity, i.e. the degree of formal dissimilarity between the members of a class (see 3.2.3 for an in-depth discussion). 2 Introduction and subclasses that contain fewer members. In accordance with the graded view, limited productivity is considered a function of low to minimal schematicity, i.e. when the members of a given class fit a tight phonetic definition. It is argued that this property of a class can offset the limiting effect of diminutive class size (see above). Specifically, Papers I and II demonstrate that small classes of Icelandic nouns are able to attract new members by means of a gang effect, defined in terms of the attractive force facilitated by a high concentration of common phonetic attributes among the members of a small inflection class (Bybee 2010:69; see 3.2.3; also Axelsdóttir 2015 and Markússon 2021, 2022a, 2023a on the gang effect in Icelandic and Faroese). Paper III accounts for the inflectional development of the Faroese nouns vøllur and fjørður. The former has in the main levelled out alternation between old stem variants, cf. Old West Nordic (OWN) nom./acc.sg., acc./dat.pl. vǫll-, dat.sg., nom.pl. vell-, gen.sg./pl. vall-. As a result of analogical levelling, defined as the process by which a paradigm acquires new forms based on existing stem variants (Carstairs-McCarthy 2017:327), all paradigm cells for Far. vøllur now contain the variant vøll- (< OWN vǫll-), while the variant vall- has also spread to a much lesser extent. Consequently, in the main, vøllur has joined the ranks of a larger class of masculine nouns; one in which stem alternation was rare and whose nominative/accusative plural ends in -ir in Modern Faroese, cf. Far. nom.sg. vøllur ~ nom./acc.pl. vøllir (< OWN nom.pl. vellir), gestur ‘guest’ ~ gestir. Conversely, Far. fjørður has not only retained the reflexes of all its old stem variants, cf. OWN nom./acc.sg., acc./dat.pl. fjǫrð- (> Far. fjørð-), dat.sg., nom.pl. firð-, gen. sg./pl. fjarð-, but also has extended them to other cells of the paradigm. Significantly, the paradigm represents the only Old West Nordic u-stem to exhibit such development in Faroese language history. In light of the different paths of levelling discussed above, Paper III argues that the historical token frequencies of individual forms, in addition to a stem variant’s historical dispersion, determined the direction of change. Further, the different frequency effects elucidated in Paper III give cause to question the approach to levelling associated with socalled ‘markedness’. Individually, Papers I–III account for various types of analogical change. As noted above, the changes discussed here are considered a function of productivity. In this connection, analogy is posited as the cognitive means via which productivity proceeds. For example, Paper II posits analogical extension as the mechanism of change, the motivation for which is perceived phonetic similarity to functionally equivalent relations in other paradigms (Bybee 2015:93–94). Thus, it is argued that phonetic similarity between borrowed 3 Jón Símon Markússon Ice. blók and a microclass that originally contained only bók ‘book’, bót ‘patch, remedy’, brók ‘trousers’, nót ‘(fishing) net’, and rót ‘root’ –– referred to here as the Xó/æTmicroclass –– facilitated categorisation of the borrowing with the latter set, cf. bók ~ plural bækur, bót ~ bætur, brók ~ brækur, nót ~ nætur, rót ~ rætur, thus (borrowed) blók ~ (innovative) blækur.3 Conversely, Paper II argues that, while the innovative alternation Ice. sg. kók ~ plural kækur stems from extension of the pattern exhibited by e.g. bók ~ bækur, occurrence of fem. dat.sg. kók, as opposed to neut. dat.sg. kóki, stems from levelling (see PII:206). Although extension and levelling are viewed as variant mechanisms of proportional analogy, levelling is characterised by an attempt at eliminating “unnecessary” alternation between forms of the same word. In other words, through levelling, stem variants from within an inflectional paradigm typically replace older stem variants. Replacement can occur on occasion or, eventually, one variant can be completely replaced. Crucially, the form of new variants is based on existing forms within the relevant paradigm. As demonstrated by Bybee (1985), the structure of inflectional paradigms across multiple, unrelated languages is often characterised by phonetic similarity among those forms of a word which express a common morphological value of significant semantic weight. This property of paradigms is reflected in the direction of levelling. Among nouns, the values singular and plural (and dual, if the relevant language makes that distinction) hold significant semantic weight relative to each other: A single entity is not the same thing as multiple of the same entity. Thus, it is argued that syncretic fem. nom./acc./dat.sg. kók reflects an attempt to align meaning and form, as syncretism better reflects common semantic content than does distinction between nom./acc.sg. kók and dat.sg. kóki. As the variant dat.sg. kók is based on a form found elsewhere in the paradigm, deduction of the form is considered the result of levelling. However, despite substantive differences between extension and levelling (see above), syncretism in nom./acc./dat.sg. kók accords with a pattern that is highly characteristic of the inflection of strong feminine nouns,4 cf. fem. nom./acc./dat.sg. bók, mynd. Therefore, it is 3 The notation Xó/æT-microclass is actually used in Paper I, while the same set of feminine nouns is referred to as the XóT-microclass in Paper II. I have decided to use the former notation in the extended introduction, as it is more informative as to the pattern of alternation between sg. -ó- and nom./acc.pl. -æ-, according to which members of the microclass conform, cf. e.g. the relations bók ~ bækur, rót ~ rætur. 4 Strong Icelandic nouns are those whose genitive singular ends in a consonant; weak nouns are those whose genitive singular ends in a vowel. This applies to nouns of all three genders (Kvaran 2005:221). 4 Introduction argued that levelling to dat.sg. kók and the form’s substitution for dat.sg. kóki demonstrates external pressures on relations within the paradigm of Ice. kók. In other words, the model of an existing pattern motivates levelling via extension, by which neut. sg. nom./acc. kók ~ dat. kóki is levelled to fem. nom./acc./dat. kók. The distinction between the two mechanisms of proportional analogy is, therefore, not completely clear cut as substantive crossover is often clearly observable. Similarly, Paper III argues that masculine classes of higher type frequency, defined as the number of individual words –– each counted once –– that conform to a pattern, exerted pressure on the paradigms of Old West Nordic u-stems (see above). However, the survival of stem-variants in the development of Far. fjørður is accounted for by reference to other measurements of frequency also. One of these measurements is conducted along the dimension of token frequency, defined as the number of occurrences of an inflectional form –– each counted once –– in spoken language or running text. Thus, for example, as Far. fjørður is a highly frequent latter constituent of complex place names that designate both a fjord and human settlement, e.g. Fuglafjørður, Kollafjørður, it occurs frequently in the dative case, which in Faroese performs the same function as the locative case in e.g. Russian, as well as others (see Þráinsson et al. 2012:164 on the dative case on objects of the prepositions á ‘on’, í ‘in’, and others).5 Conversely, due to attributes of the Faroese landscape, the topographical referents of Far. vøllur tend to be very small and, therefore, not to perform established functions. Indeed, the study reported in Paper III reveals that the dative form of Far. fjørður is of high token frequency, while occurrence of Far. vøllur is characterised by low token frequency in all of its case forms. In other words, language users’ perception of real-world conditions appears to determine the frequency of words and their individual inflection forms (following Haspelmath 2006). From the usage-based cognitive perspective, high frequency is a correlate of entrenchment, defined as the reorganisation of knowledge in memory on account of the impact of continued experience (Schmid 2017:3–4). Thus, the more frequently an individual experience a phenomenon in the world, the more embedded, i.e. entrenched, the cognitive representation of that experience becomes relative to the representations of less frequent phenomena. In turn, entrenchment is viewed as the determinant of lexical strength, i.e. how accessible information pertaining to the relevant experience is from memory (Bybee 5 See also Stolyarova 2016, who discusses and compares the equivalent roles of the locative and dative cases in Russian and Icelandic, respectively. 5 Jón Símon Markússon 1985:117). In this spirit, Paper III argues that the original stem of dat.sg. firði is more entrenched than that of original dat.pl. fjørðum in syntactic contexts such [í/úr FJØRÐURdat.], as place names with -fjørður almost always contain the singular form.6 This property of usage is a function of token frequency, which facilitated greater lexical strength of the stem variant (dat.sg.) firð-, a position supported by the fact that younger dat.pl. firðum, based on the highfrequency dative singular form, is now much more frequent than older dat.pl. fjørðum. Paper III subsequently argues that the deduction of innovative dat.sg. fjørði represents an attempt to contrive a dynamic whereby the stem variant firð- is now synonymous with plural, cf. nom./acc.pl. firðir, innovative dat.pl. firðum, opposing singular forms with fjørð-, cf. nom.sg. fjørður, acc.sg. fjørð, dat.sg. fjørði. It is argued that this formal opposition is motivated by the dispersion of stem variants throughout the paradigm (see above). However, dat.sg. firði has most probably always had far higher token frequency than all other forms of fjørður and its stem variant will have been more accessible than others as a result. This property of use renders the likelihood that the variant firð- be forgotten –– albeit only momentarily –– very unlikely, a position that is supported by the fact that use of younger dat.sg. fjørði is characterised by low token frequency, meaning it has never acquired the degree of entrenchment likely required to usurp its older cellmate’s place in memory. In light of the brief overview above, the general objectives of the thesis are set out in (1). (1) a. To determine which cognitive factors have impacted the structure of the inflectional systems of Icelandic and Faroese. b. To ascertain whether these cognitive factors apply only to language or are domain-general in nature. c. To demonstrate how the impact of these cognitive factors are evident through language change as a function of language use. 6 The prepositions Far. í ‘in’ and úr ‘from’ both assign dative to their objects to convey location. The appropriate forms of Far. fjørður within this context would have been dat.sg. firði and dat.pl. fjørðum before the process of levelling began (see PIII:76–80). 6 Introduction In order to fulfil the objectives stated in (1), subsequent chapters attempt to provide answers to the respective research questions posed in Papers I–III, which are reiterated in the next section. 1.2 Aims and individual research questions of Papers I–III This section sets out the stated aims of Papers I–III and lists the research questions of each study in the context of those aims. The aim of Paper I is first and foremost to examine and account for the extent to which schematicity impacts the rate at which Icelandic masculine forms in plural -ur are reanalysed as feminine. The research statement of Paper I (PI:1) is restated as the question in (2). (2) How do we account for the different rates at which Icelandic masculine forms in plural -ur are treated as feminine due to varying degrees of phonetic and/or semantic similarity to clusters of feminines in the same ending? Paper I argues that due to the high rate at which the ending plural -ur is dispersed among feminine nouns, any masculine form in the same ending can be reanalysed as feminine –– all other things being equal. However, it is also demonstrated that phonetic coherence with the Xó/æT-microclass of feminines in plural -ur boosts the rate of reanalysis. Paper II aims specifically to account for the limited productivity of the Icelandic Xó/æT-microclass, discussed in Section 1.1. The objective of Paper II (see PI:196) is stated in (3). (3) To provide a usage-based cognitive account of the limited productivity of the Icelandic Xó/æT-microclass through recourse to: i. Semantics, ii. Pragmatics, iii. Schematicity, iv. Analogy. In this connection, the productivity of the microclass is betrayed by the addition of two loans from English, namely blók, cf. British/Australian Eng. bloke, and kók, cf. Eng. CokeTM. Paper II argues that assignment of the loans to the Xó/æT-microclass is a function of the gang 7 Jón Símon Markússon effect. However, the analysis demonstrates that a property of limited productivity is graded membership of the microclass, whereby inflection of the loans only sometimes accords with the relative pattern, and sometimes with patterns that characterise other classes. Paper II argues that the respective degrees to which each noun is incorporated in the Xó/æTmicroclass are determined by their individual semantic properties, as well as use in context, i.e. pragmatics. The aim of Paper III is to identify the distinct inflectional forms within the paradigms for Far. vøllur and fjørður from which levelling proceeds. As part of this aim, the analysis also endeavours to ascertain the role of frequency and/or semantics as determinants of the direction of levelling. The research question of Paper III (PIII:54) is reiterated in (4).7 (4) Does frequency determine the direction of levelling? Paper III argues that the direction of levelling is determined by token frequency as a correlate of entrenchment and lexical strength. 1.3 General contents of the thesis Chapter 2 discusses the genetic relationship between the West Nordic languages, Icelandic, Faroese, and –– to a lesser extent –– Norwegian, as well as inflectional change in all three languages since the Old West Nordic period. Chapter 3 fleshes out the usage-based cognitive approach to language change. There, I justify application of the approach to the Icelandic and Faroese data presented in Papers I–III on the basis of comparison with the approach from generative grammar and dual-processing. Chapter 4 delineates the methodology employed in Papers I–III in light of the theoretical tenets of the usage-based cognitive approach, as outlined in the previous chapter. Chapter 5 discusses the specifics of Papers I and II in terms of thematic content, application of the usage-based cognitive approach to the relevant Icelandic data, and the conclusions drawn in each paper. Chapter 6 provides a detailed usagebased cognitive analysis of the Faroese data presented in Paper III and relevant conclusions. Chapter 7 draws general conclusions through reference to the research questions set out in (1) (Section 1). Finally, access is provided to Papers I, II, and III at the end of the extended introduction, under the heading ‘Paper I’, ‘Paper 2’, and ‘Paper III’. 7 Translated from Ice. Er tíðni ráðandi afl í stefnu útjöfnunar? (PIII:54). 8 2 Insular Nordic In this Chapter, I discuss the genetic and historical relationship between the West Nordic languages with particular focus on Insular Nordic and, to a lesser extent, through reference to Norwegian. The discussion is by no means exhaustive but is intended as an overview. Below, I draw comparisons between Old West Nordic and Insular Nordic with the specific intention of demonstrating the different degrees to which the constituent inflectional systems of the latter have diverged from the former. Further, I elucidate the nature of the changes discussed. The content of this section will feed into Chapters 3 and 4, where I discuss two opposing theoretical approaches to language change and the methodology employed in Papers I–III, respectively. Along with the dialects of Western Norway, as well as the now extinct Orkney and Shetland Norn, Icelandic and Faroese belong to the West Nordic branch of the Nordic language family tree.8 Collectively, Icelandic and Faroese furnish the Insular Nordic branch, due to their distribution across geographically distinct Nordic island nations, i.e. Iceland and the Faroe Islands. The Insular Nordic languages both derive from Old Norse through Old West Nordic. From a period spanning roughly the 9th to the 11th centuries, loosely designated as the “Viking Age”, the seafaring, Norse-speaking peoples of Mainland Scandinavia –– particularly Western Norway –– sought out new lands to the South-West and West. The fruits of this exploration were subsequent settlements on Orkney, Shetland, in Scotland and the Hebrides, in the Faroe Islands, Iceland, Greenland, and in North America. In each settlement, a variety of Nordic was the community language (e.g. Vikør 21–23). While initial Nordic presence in America was short-lived, a Nordic community in Greenland survived in coexistence with Inuit until the former seemingly disappears before the early 16th century. Conversely, the West Nordic dialects of Orkney and Shetland survived until the mid-1700s and just prior to 1800, respectively (Barnes 2004:75). Since Norn went extinct, Icelandic and Faroese have stood alone as the only extant descendants of Old West Nordic to the West of Mainland Scandinavia. In the context of the developments just delineated, questions arise as to the extent to which Icelandic and Faroese have each diverged from Old West Nordic and to which each 8 Some dialects of Norwegian, e.g. that spoken in and around Oslo, as well as Danish and Swedish (formerly also Old Gutnish), occupy the East-Nordic branch (e.g. Vikør 2001:33). 9 Jón Símon Markússon language resembles the another. In this connection, change in Insular Nordic is documented to starkly different degrees in the written record. The earliest attestations of Icelandic date from roughly 1150–1200, from which time the language is consistently well attested until the present day. Due to the sheer volume of written sources for Icelandic, historical linguists have been able to plot the chronology of change to a significant degree of precision relative to diachrony. Conversely, Faroese language history is sparsely documented until around 1800 (Þráinsson et al. 2012, e.g. p. 374). Early written sources for the language include two runic inscriptions found in the Faroe Islands: one in the village of Kirkjubøur and the other in Sandavágur, dated to roughly 1000 and 1200, respectively. Probably the most important early Faroese source in terms of comprehensiveness constitutes Far. Seyðabrævið ‘The Sheep Document’, dated to 1298. The document bestows regulations for keeping sheep in the Faroe Islands and the original is believed to contain linguistic features that developed in the Faroes. Further, a transcript of Seyðabrævið from around 1600 is considered to reflect hints of a budding new variety of West Nordic, i.e. Early Faroese (see e.g. Þráinsson et al. 2012:371). It is written documents from the beginning of the 15th century which first show that a clearly distinct Faroese variety of West Nordic has developed. The main sources for this variety are the Far. Húsavíkarbrøvini ‘The Húsavík Letters’, which exist in transcribed form from 1407 and 1479. Here, spellings such as hrentadi for older rentaði suggest that scribes were unsure where to write hr for older initial voiceless r, a West Nordic trait that had likely disappeared from Faroese by the 15th century or was almost lost at the time of transcription (see Þráinsson et al. 2012:372). Such spellings suggest a clear difference between Faroese and Icelandic: voiceless r still occurs word-initially in the latter and is represented by orthographic hr, e.g. Ice. hringur ‘ring’, hreinn ‘clean, pure’, hrópa ‘yell, call’. Due to developments such as those just discussed, historical linguists commonly characterise change in Faroese as more progressive than in Icelandic. In other words, the latter is considered more conservative than the former. A common observation in this regard is that the inflectional system of Modern Icelandic better resembles that of Old West Nordic than does Modern Faroese. In turn, the latter’s inflectional system harkens more so to that of Old West Nordic than does Modern Norwegian and the other Mainland Scandinavian languages, i.e. Bokmål, Danish, and Swedish (e.g. Þráinsson et al. 2012:369). Despite this, it is reasonable to assert that many changes observed in Icelandic have also occurred in Faroese, while change in the latter has led to highly complex intra-paradigmatic dynamics not mirrored in the inflectional system of Icelandic (see below). 10 Insular Nordic While there is a marked difference in the degree of conservatism that Icelandic and Faroese exhibit to Old West Nordic, there are also conspicuous structural similarities between both Insular Nordic languages that distinguish them from Mainland Scandinavian. This is particularly true in the context of morphology but also in (morpho)syntax (e.g. Þráinsson et al. 2012:59, 407, Eyþórsson 2015). For example, while the respective phonological systems of Insular Nordic have diverged from that of Old West Nordic and each other, both Icelandic and Faroese retain a complex and active inflectional system for nouns. One commonly cited non-linguistic reason for this conservatism is the geographical isolation of both island nations through the centuries, while another is the kinds of social dynamics that typically characterise small, insular communities. In terms of degree, then, such factors might both account for the retention of morphological complexity in both languages but also for the fact that changes to the inflectional system of Faroese have led to more complexity than in that of Icelandic (see e.g. Trudgill 2011, who discusses how such socio-linguistic dynamics can affect language structure). In other words, while the Faroese system of inflection has changed more than that of Icelandic (see below), intra-paradigmatic change has not been synonymous with simplification, thus counter to ideas presented by Kiparsky (1974).9 For example, the morphologisation of historically phonological processes such as verschärfung and palatalisation in Faroese has served to complexify intra-paradigmatic dynamics as these pertain to alternation between inflectional forms (Petersen 1992; also Petersen and Voeltzel, to appear 2024). Thus, the paradigm of the Icelandic verb búa ‘live’ bears witness to verschärfung but only in the past plural, cf. pres. 1sg. bý, 2sg. býrð, 3sg. býr; past 1/2/3sg. bjó-; past 1/2/3pl. bjugg-. However, in the paradigm of the Faroese cognate, búgva ‘live’, we see the result of verschärfung in different person/number combinations in the present: pres. 1sg. búgvi, 1/2/3pl. búgva but pres. 2sg. býrt, 3sg. býr; but not in the past, past sg. búði, past pl. búðu. Thus, unlike the Icelandic dynamics delineated above, verschärfung in the Faroese paradigm is not at all associated with the past and not fully with the present, nor with a specific person/number combination (see Weyhe 1996, pp. 81–82, for examples of complexity resulting from the morphologisation of palatalisation). The more conservative nature of the Icelandic inflectional system is best characterised in relation to the degree of change exhibited by the respective systems of Faroese and Norwegian. Thus, for example, while Icelandic retains the four-case system for nominals it 9 See Chapter 3 for further discussion on the notion of intra-paradigmatic change as simplification. 11 Jón Símon Markússon inherited from Old West Nordic –– inflecting for nominative, accusative, dative, and genitive –– Faroese has lost its genitive case.10 As the result of more sweeping change, Norwegian nominals, bar remnants among the personal pronouns, are generally not inflected for case, although genitive -s has been retained as a marker of possession.11 While Icelandic and Faroese retain three distinct grammatical genders for nominals –– masculine, feminine, and neuter –– the distinction between masculine and feminine is no longer maintained in all varieties of Norwegian, yielding a class of common gender nouns that opposes the neuter in a two-gender system (Vikør 2001:38, 207–209). Despite the relative conservatism of Icelandic, the system by which its nominals inflect has by no means been left unaltered. To offer some random examples, the Old West Nordic u-stems eventually lost the ending acc.pl. -u to the i-stem ending acc.pl. -i in the early 16th century (Þórólfsson 1925:22), cf. OWN u-st. acc.pl. fjǫrðu  firði ‘fjords’. Almost all members of the feminine class of ijō-stems have lost the older ending nom.sg. -(u)r due to its association with masculine grammatical gender, e.g. OWN fem. ermr  Ice. ermi ‘sleeve’, although cf. the women’s names OWN/Ice. fem. Auð(u)r, Sigríð(u)r, Þuríð(u)r, also masc. OWN hestr > Ice. hestur ‘horse’ (e.g. Þórhallsdóttir 1997). Levelling has affected disyllabic masculine stems with the suffix -il-, cf. e.g. OWN nom.sg. lykill ‘key’ ~ dat.sg. lukli ~ nom.pl. luklar, Ice. nom.sg. lykill ~ dat.sg. lykli ~ nom.sg. lyklar, although OWN/Ice. nom.sg. ketill ‘kettle’ ~ dat.sg. katli ~ nom.pl. katlar. Many changes to the inflectional system of Icelandic are mirrored in Faroese. Thus, for example, the latter has also generalised the root vowel y throughout the paradigm of Far. lykil ‘key’, cf. dat.sg. lykli, nom./acc.pl. lyklar. However, typically, change of this nature has gone even further in Faroese, cf. Far. nom.sg. ketil ‘kettle’, dat.sg. ketil, nom./acc.pl. ketlar; even eradicating alternation that resulted from Proto-Nordic u-umlaut,12 which has been preserved 10 The genitive is no longer actively assigned in Faroese. Rather, use of old genitive forms appears to be fully conventionalised in phrases such as Far. til Føroya ‘to the Faroe Islands’, where the preposition til assigned genitive in Old West Nordic (and still does in Icelandic). However, synchronically, objects of til in the modern language are accusatives. While genitive forms occur in new compounds, e.g. bendingargrunnur ‘inflection database’, where the element -ar- is the old genitive ending, these are modelled on older genitive compounds such as bendingarfrøði ‘inflectional morphology’ and function as linking elements (see Jónsson and Eyþórsson 2011 for a comparison of the development of genitive arguments in Icelandic and Faroese, and Petersen and Szczepaniak 2018 on genitive forms in the broader Germanic context). 11 Some dialects of Eastern Norway retain dat.pl. -um, while some Western dialects have -o/-å. This is exceptional, however (see Sjekkeland 1997, Eyþórsson et al. 2012). 12 On the nature of u-umlaut in Icelandic and Faroese language history, see Markússon 2017 and sources cited there. 12 Insular Nordic in Icelandic, e.g. OWN nom.pl. katlar ~ dat.pl. kǫtlum, Ice. katlar ~ kötlum, cf. Far. ketlar ~ ketlum. The ending of OWN nom.sg. ermr has also been replaced in Faroese, although by -a rather than -i, cf. Far. fem. erma ‘sleeve’, Ice. fem. ermi. Again, typically, Faroese has taken change to Old West Nordic ijō-stems even further, cf. the Faroese women’s names Sigrið, Turið and their Icelandic cognates, Sigríður, Þuríður.13 Sweeping processes of levelling have occurred in Faroese, which lack equivalents in Icelandic. For example, alternation stemming from Proto-Nordic u-umlaut has in the majority of cases been levelled in the Faroese reflexes of Old West Nordic ōn-stems –– all of which were feminine nouns. Thus, alternation between OWN a and ǫ has mostly been levelled in favour of Far. ø (< OWN ǫ by regular sound change), cf. e.g. OWN nom.sg. gata ‘path, road’ ~ acc./dat./gen.sg. gǫtu, Ice. gata ~ götu, but Far. gøta ~ gøtu ‘street’; OWN saga ‘history, saga, story’~ sǫgu, Ice. saga ~ sögu, but Far. søga ~ søgu. Conversely, some nouns that follow the affixal pattern of Faroese ōn-stems show the stem-vowel a in all forms, e.g. Far. nom.sg. mamma ‘mum, mummy’ ~ acc./dat.sg. mammu, cf. Ice. mamma ~ mömmu (see Markússon 2017). Despite the kind of simplification by levelling exemplified above, the process has also yielded complexity in Faroese paradigms. Here, it is necessary to note that, in essence, levelling does not change one inflectional form into another. Instead, the process is defined as the deduction of a new form based on the (phonetic or morphological) structure of another within the same paradigm. In other words, functionally equivalent doublets that existed before levelling are not suddenly eradicated from the inflectional paradigm when a new form is conceived. Rather, these may co-exist for centuries (e.g. Bybee 2015:95). Examples of complication by levelling can be taken from the paradigms of Faroese nouns containing either a or ø in the stem of the nominative singular. Thus, the paradigms og masc. hvalur ‘whale’ and fem. øksl ‘shoulder’ contain the dative plural forms hvølum and økslum, respectively, where ø is the regular reflex of Proto-Nordic (PNc.) *a by the historical process of u-umlaut, cf. PNc. dat.pl. *hvalum(R) > OWN hvǫlum > Far. hvølum, PNc. dat.pl. *axlum(R) > OWN ǫxlum > Far. økslum. However, in the course of Faroese language history, both paradigms have acquired ø-less forms, cf. dat.pl. hvalum and akslum, based on the stem 13 According to Petersen (2009), change from nom.sg. -r to -a and/or the lack of an ending was the only available choice to speakers of Faroese, whose aim was retaining feminine grammatical gender in e.g. erma and Sigrið. This, Petersen argues, is because Far. nom.sg. -ur (< OWN -r) and -i are overtly masculine markers. Therefore, one could argue that, compared with the developments of Icelandic cognates, the formal parameters for gender distinction in Faroese are far clearer cut. 13 Jón Símon Markússon variants hval- and aksl- from elsewhere within the respective paradigms, e.g. nom.sg. hvalur, nom./acc.pl. akslar/akslir (see Þráinsson et al. 2012:398–399). While the forms with a might constitute instances of simplification, their coexistence with forms with ø add complexity that formerly did not exist in the respective paradigms.14 While much analogical change in Faroese can be characterised as levelling, Icelandic paradigms have undergone the process to a far lesser extent. For example, the paradigms of Icelandic feminine nouns in nom.sg. -a have retained u-umlaut alternation that parallels the relational oppositions characteristic of ōn-stem inflection in Old West Nordic (see examples above). Likewise, while alternation between individual stem variants of Old West Nordic ustems has been all but levelled out in Faroese (see 1.2), the inflection of Icelandic u-stems strongly reflects the pattern of alternation that characterised their Old West Nordic predecessors. This brief overview of change in Insular Nordic nouns demonstrates that Faroese has diverged more from Old West Nordic than has Icelandic, and that the kinds of change witnessed between the two languages are often of a substantively different nature. Thus, change in Faroese provides ample opportunity for research into the nature of levelling. Conversely, Icelandic is more conservative with regard to the retention of ancient vocalic alternation, more often exhibiting change resulting from extension. To date, there is a conspicuous lack of in-depth study that seeks specifically to account for the domain-general cognitive mechanisms which underpin these analogical processes in the Insular Nordic context. In this connection, while the current chapter has provided a brief history of the respective inflectional systems of Insular Nordic, Chapter 3 elaborates on the cognitive underpinnings of analogical reasoning as a property of language use, of which language change is a function. 14 Of course, it is not unlikely that doublets existed immediately after u-umlaut occurred in Ancient Nordic. However, the written record suggests that those in a had disappeared by the time that Insular Nordic is first attested. 14 3 Usage-based cognitive grammar and language change The present chapter accounts for the chosen theoretical approach to the changes examined in Papers I–III. In 3.1, I compare and evaluate the applicability of two opposing theoretical approaches that seek to account for language change, namely the dual-processing model and the usage-based cognitive model. Section 3.2 fleshes out the usage-based cognitive approach through reference to rich memory for language (3.2.1), the impact of frequency on the structure of grammar (3.2.2), schematicity as a property of both linguistic and non-linguistic categories (3.2.3), and the prototype structure of categories as a function of cognitive economy (3.2.4). Subsection 3.2.5 provides a summary of Section 3.2. In Section 3.3, I elucidate the adaptive nature of categories as a function of analogy. Section 3.4 summarises the current chapter. 3.1 Two opposing theoretical approaches to language change In this section, I provide an overview of two opposing theoretical approaches to language change, which, in the broadest terms, fall under usage-based cognitive grammar, on the one hand, and generative and dual-processing models, on the other. The aim of this endeavour is twofold: 1. To argue that language use is the mechanism of language change. 2. To demonstrate in light of 1. that a theoretical approach based on usage provides a more suitable framework to account for language change than one that does not. The usage-based cognitive approach to language change is closely associated with the works of Jóhanna Barðdal (e.g. 2008), Joan Bybee (e.g. 1985, 1995, 2007, 2010, 2015), George Lakoff (1987), Ronald Langacker (e.g. 1987, 1988), and Michael Tomasello (e.g. 2000), to name but a few. Usage-based cognitive grammar places prominence on the impact of language use on transition and acquisition, the emergent structure of grammar, and language change. Therefore, in order to demonstrate the hand of language use in language change, it is first paramount to understand exactly what language use involves. Crucially, the process as conceived of here is not restricted to speaking and/or signing a language, but also includes 15 Jón Símon Markússon the tacit processes of receiving linguistic messages and decoding their content. As demonstrated with examples in what follows, this endeavour relies on prior experience with language that is stored in memory and available for the purpose of comparison, mapping, and extension of existing knowledge to new contexts. Bybee (e.g. 2006) has characterised grammar as the cognitive organisation of the individual’s experience with language. In the broader context of linguistic theory, the usagebased cognitive approach constitutes an abrupt departure from parsimonious, highly structural dual-processing model, itself an offshoot of generative/transformational grammar. The latter view economy as an inherent property of memory for language, by which redundant and predictable information is not stored due to constraints on “space” (e.g. Jakobson 1990:321; Rögnvaldsson 2013:139). In other words, the relevant experience is necessarily discarded from memory. In light of the economic view of memory for language, generative and dual-processing models seek to attribute linguistic ability to generative processes that are functions of genetic endowment, specifically evolved to facilitate (and exhibit) linguistic competence. According to generative grammar, the principles that guide competence constitute a biological faculty referred to as “Universal Grammar” (UG) (see e.g. Chomsky and Halle 1968; Clahsen 1999; Pinker 1999; also, more recently, Chomsky 2017 and Hauser et al. 2002).15 According to the approach from UG, the processes that facilitate language acquisition and linguistic performance are ascribed to evolutionary biology, by which the task of acquiring one’s native language(s) is guided first and foremost by prescribed genetic constraints on the types of possible human language. In other words, proponents view the process as assisted by innate linguistic knowledge, an understanding of which is crucial in order to account for the principles that guide language acquisition and, more generally, the nature of human language (e.g. Chomsky 1965). However, due to the supposed innateness of UG, insight into its inherent linguistic principles can be blurred by various kinds of cognitive impairment. These include temporary states such as intoxication, momentary lapses in 15 Despite the fact that generative and dual-processing models have evolved in terms of the instantiation-specific machinery applied in exposition, a fundamental adherence to parsimony has characterised the model since its inception. Indeed, a hyper-economical approach to memory for language is conspicuous in the Distributed Morphology (DM) paradigm, which assumes no lexicon, i.e. that part of memory supposedly responsible for storing necessary lexical information, such as lexical roots, unpredictable inflectional forms, and affixes. This tenet of DM has yielded the axiom “syntax all the way down”, in allusion to the belief that fully formed inflectional forms result from concatenation of lexical heads (roots) and functional heads (morphemes = anything but a root) (e.g. Embick 2015; see Spencer 2019 for a detailed criticism of the theoretical foundations of DM espoused by Embick 2015). 16 Usage-based cognitive grammar and language change retrieval that result in slips of the tongue, cognitive impairment due to stroke, to name but a few examples. For this reason, generative linguists consider language use a function of linguistic performance, i.e. the physical manifestation of language, an account of which pales in comparative theoretical importance to an understanding of linguistic competence, i.e. the system of grammar supposedly acquired with the assistance of UG (e.g. Chomsky 1965). Therefore, generative and dual-processing models view language use as an imperfect, or filtered-out, representation of linguistic competence. The theoretical basis for envisaging linguistic competence as a function of genetic endowment is at best tentative. This is because the argument for UG is a priori in nature. In simple terms, UG was originally posited as part of a generative-driven endeavour to understand how human beings can know so much given their limited experience, a paradox characterised by Chomsky as ‘Plato’s Problem’. In linguistic-specific terms, the purpose of UG is to account for the acquisition of human language in spite of apparent significant inconsistencies in the linguistic input, i.e. Chomsky’s poverty of the stimulus (e.g. Chomsky 1988:24). However, the so-called language faculty, i.e. the theorised biological host of UG, remains a contentious topic, both in linguistics and beyond (e.g. Pullum and Scholz 2002 for a critique of Chomsky’s position). Indeed, numerous scholars have argued against the existence of UG from fields such as evolutionary biology, linguistic typology, and psychology. For example, as part of an interdisciplinary analysis of language change in the evolutionary biological context, Croft (2002) characterises language as a metaphorical “moving target”, whose structural properties change at such a rate as to evade eventual genetic encodement. In other words, language change occurs at too great a pace for what would be the feasible amount of time necessary to genetically encode any synchronic snapshot of linguistic competence. Further, according to the widely accepted axiom associated with Karl Popper (1959), a scientific theory is most appropriately defined as one that is falsifiable. In this connection, Dąbrowska (2015) points out that there is still no explicit evidence for UG, while Evans and Levinson (2009) have demonstrated that the differences between human languages permeate every level of linguistic function to so profound a degree as to render any theory of UG unfalsifiable. In contradistinction to the innatism approach, usage-based cognitive linguists view linguistic competence as emergent from domain-general cognitive processes predicated on language use, the cognitive mechanism for which is the domain-general process of analogical reasoning (see e.g. Bybee 2010:6–8), defined as in (6). 17 Jón Símon Markússon (6) Analogical reasoning (analogy) is the cognitive process through which existing knowledge is extended to new contexts. Inherent to the definition in (6) is that linguistic experience is stored in memory and provides the raw material for language use. In contradistinction to generative and dual-processing models, then, the approach from domain-general cognition assumes that, once stored, neither experience with language nor any resultant generalisations –– also those characterised by redundant information –– are discarded from memory but, rather, entrenched there for future reference (Bybee 2010:15; see Goldberg 2006:47–48).16 Further, language change provides empirical evidence for the impact of non-linguistic factors such as frequency and schematicity on language use and the direction of change. One common property of language change that elucidates this relation is the cross-linguistic tendency for highly frequent phrases to undergo phonetic reduction over time. Iterations of English goodbye, cf. g’bye and bye, all of which are incremental yields of gradual contraction from God be with you, provide just such examples. The meaning of this phrase is both highly predictable and discernible from context, meaning contraction is not an impediment to comprehension.17 16 Charles Yang (e.g. 2016) espouses his tolerance principle, according to which high type frequency is the determinant of productivity, while token frequency, schematicity, and semantics are not attributed facilitating roles. In this connection, it is clear that knowledge of type frequency must be based on stored experience with language. However, one would also assume storage of knowledge of these other factors. One wonders, then, why the tolerance principle limits productivity to knowledge of type frequency, particularly when both token frequency and schematicity have been identified as contributing factors (see e.g. Axelsdóttir 2015, Jóhanna Barðdal 2006, and Markússon 2022a, 2022b, 2023a, 2023b). In other words, the tolerance principle seems easily disproved by observable productivity in actual language use and change (see Enger 2022 for a comprehensive criticism of Yang 2016). The likely motivation for Yang’s position is that, were productivity considered a function of token frequency, the model would necessarily have to assume storage of information pertaining to the frequency of occurrence of more than the bare minimum of “regular” forms. It has long been the belief in generative grammar and among proponents of dual-processing models for inflection that language users derive “regular” surface forms via the application of symbolic rules to a single lexical form, i.e. a stripped-down memory representation of a word that inflects according to a “regular” pattern (e.g. Rögnvaldsson 2013:137–139 on Icelandic). Yang is no exception in this regard. However, if information about the token frequency of “regular” forms is stored –– remembering that “regular” is often conflated with “of high type frequency”, which is in turn is often conflated with “productive”, “predictable”, and “no need for storage” (Herce 2019) –– it would simultaneously be necessary to assume richer memory for language than proponents of generative and dualprocessing models are typically comfortable with. 17 See the discussion of the reducing effect in 3.2.1. 18 Usage-based cognitive grammar and language change Conversely, high token frequency can facilitate the preservation of irregularity. For example, the infrequent English strong verbs creep, leap, and weep have developed regular past forms, cf. creeped, leaped, and weeped, beside older crept, leapt, wept, while highly frequent keep, mean, sleep alternate with past kept, meant, and slept only. This is considered a function of entrenchment and its impact on the availability of a linguistic item for use, i.e. that item’s lexical strength. It is, therefore, self-evident that if frequency impacts language use and change, knowledge pertaining to frequency is surely based on prior experience with language. Moreover, the relevant knowledge must be retained in memory, where language users have access to it. Indeed, if language users did not track frequency of use, frequency effects like those exemplified above should not be evident in language change. In other words, for prior experience with language to inform later usage choices –– as it evidently does –– memory of that experience must facilitate extension of the relevant knowledge to new contexts, cf. the definition of analogy in (6). Despite clear evidence for language structure as emergent from the application of domain-general cognitive processes to existing knowledge, the approach from analogy has been challenged. In the historical context, this is perhaps understandable from the perspective of parsimony for mainly two reasons. First, as noted above, UG was posited as a working solution to a theoretical conundrum faced by linguists at a time when much less was known about the nature of linguistic storage. Secondly, the cognitive and articulatory apparatus is highly uniform across our species (Beckner et al. 2009), meaning that it is easy to find similarities across languages if that is what the researcher is looking for and then attribute similarity to UG. For this second reason in particular, proponents of UG have pointed to the apparent futility of recourse to analogy as an explanatory device when logically feasible analogies do not occur. In this connection, Kiparsky (1974) has presented the explanation from analogy as theoretically wanting with his observation that conceivable analogies do not necessarily occur and are unlikely to do so in the future. In support of his point, Kiparsky offers the potential proportion in (7), which has yet not come to fruition. (7) ear : hear eye : X; X = *heye 19 Jón Símon Markússon An initial criticism of the point that Kiparsky’s equation is supposed to make is that proportional analogy is generally considered to occur between forms of words that belong to the same word class: ear is a noun, while hear is a verb. An obvious drawback of Kiparsky’s observation is that Eng. see already exists, is highly frequent and, therefore, is unlikely to be forgotten and replaced (following e.g. Bybee 2015:95). Moreover, while it is true that the form *heye is unlikely to be deduced from Eng. eye on the model of ear : hear, the irony of Kiparsky’s point appears to escape him: the proportion in (7) constitutes a clear instance of analogy based on a linguistic model. Thus, the fact that the example may only have occurred in Kiparsky’s mind by no means discounts recourse to analogy as the mechanism for linguistic innovation, even though it is not disseminated to the wider speech community. It is highly probable that Kiparsky’s observation follows in the spirit of King (1969:235), who characterised analogy as “a terminological receptacle devoid of explanatory power”, a criticism levelled at the Neogrammarian habit of liberally ascribing all non-regular change to analogy. However, due to the very theoretical premise upon which the role of analogy is cast into doubt, criticisms inevitably prove to be strawmen. That is, the parsimonious nature of linguistic storage according to generative and dual-processing models has motivated attempts to reconcile synchrony with diachrony, by which change is characterised as changes to the system of symbolic rules that children supposedly acquire through intergenerational language transfer (e.g. Guðmundsdóttir 2008:27; Lightfoot 2006; see Beckner and Wedel 2010 for a critical view).18 One posited function of this process was 18 The application of symbolic rules to stripped down underlying lexical forms is a necessary means of “generating” the surface forms of “regularly” inflected words because, according to the generative model upon which Kiparsky’s (1965) approach was based, only those forms of a word that are not fully predictable on the basis of other forms are stored in memory. By the same token, then, all predictable forms are derived from lexical form(s) by rule, thus sparing supposedly unnecessary strain on memory (e.g. Rögnvaldsson 2013:137– 139; see above on the yet more stringent parsimony that is characteristic of DM). An example of such an underlying form would be #barn-um#, from which the surface form Ice. dat.pl. börnum ‘children’ is supposedly derived through rounding of the root vowel a to ö, a process that some generative linguists have posited as synchronic u-umlaut (e.g. Þráinsson 2011, 2017; Rögnvaldsson 1981; see Markússon 2012, 2017 for a critical view). The underlying form #bar-num# is considered “stripped down” because it does not convey the rounded root vowel that is always present in the surface form. Such stripping down to the bare essentials is only necessary on a highly constrained view of the memory’s capacity for linguistic storage. In the absence of such a view, it is not clear why a child acquiring Icelandic should derive a surface form from a stored form that it has never heard in the input. The question as to whether the input 20 Usage-based cognitive grammar and language change grammar simplification, whereby the language learner chooses the simplest rule set based on the linguistic input (e.g. Kiparsky 1974). However, a highly valid critique of this position is that much language change does not simplify grammar. For example, the innovative form Ice. fem. nom./acc.pl.def. fæturnar ‘the feet’ occasionally occurs instead of original masc. nom.pl.def. fæturnir and acc.pl.def. fæturna. The fact that the feminine innovation eliminates the need for alternation between the latter pair of forms could of course be viewed as the simplification of grammar. However, such a view would be overly simplistic as use of fem. fæturnar redefines the typical dynamic such that overtly masculine nom.sg.def. fóturinn ‘the foot’ now alternates with overtly feminine fæturnar. This can be considered a complicating development in light of the fact that singular forms of Ice. fótur almost never –– if at all –– occur as feminine (see PI:3 and sources cited there). In the broader context, it is an extremely rare relational dynamic within Icelandic noun paradigms that forms differ in grammatical gender (see PI:7 for an example). Therefore, it is doubtful that motivation for the deduction of fem. nom./acc.pl.def. fæturnar was the simplification of grammar in the broader sense (see Sims-Williams and Enger 2021 for further examples and discussion).19 Further, the claim that language change necessarily occurs during language acquisition is refuted by innovations such as Eng. past creeped, leaped, and weeped, which occur beside older past crept, leapt, and wept (see above). As noted by Bybee (2015:95), the verbs in question are unlikely to occur in child speech and, so, the weak innovations are attributable to adult language users (see Dąbrowska 2008, who shows that acquisition of Polish inflection is emergent and subject to reorganisation into adulthood). In light of these arguments, the generative approach to language change is considered insufficient to guide the analyses presented in Papers I–III, as it constrains the scope for permissible change by largely ignoring the impact of frequency, misunderstanding the motivations for change, and attributing change to the language acquisition phase only. determines the shape of underlying forms or whether an underlying form determines the shape of the relevant surface form is a matter for discussion of the kind of logical fallacy inherent to circular reasoning. 19 Kiparsky (2000) assesses analogical change in Gothic ja-stems and concludes that the relevant outcomes “do not complicate either the phonological or the morphological system of Gothic. The phonological constraints do not pick up any exceptions or morphological conditions. In particular, Sievers’ Law ... continues to operate ... in a fully regular way.” Thus, it appears that maintenance of or a move towards simplicity, characterised in terms of “grammar optimization”, is viewed as validation in Kiparsky’s approach to the nature and function of analogical change. 21 Jón Símon Markússon For the reasons stated above, I adopt a usage-based cognitive approach, which accounts for language change through reference to the application of domain-general cognitive processes to the linguistic input and stored memory representations for language across the language user’s lifespan. This choice is motivated by four main factors. First, the approach has been somewhat lacking –– particularly among Icelandic linguists concerned with synchronic variation, who in recent decades have typically followed the generative tradition (though see e.g. Axelsdóttir 2014, 2015; Ottósson 1992). Secondly, while traditional depictions of analogy suffice to delineate the outcome of innovation and change, these are often allowed to speak for themselves, while the domain-general cognitive specifics are merely implied, cf. proportions such as that in (7). Thirdly, of all the so-called “domains”, morphology poses the greatest challenge to the traditional generative view of productivity (see Bybee 1985:5–8). Indeed, the dual-processing model typically attributes a binary quality to productivity, i.e. linguistic structures are either regular, unmarked, and/or derived or irregular, marked, and/or stored (e.g. Clahsen 1999; Rögnvaldsson 2013; see Herce 2019 on the vagueness of such characterisations of “(ir)regularity”; also Langacker 1987 on what he characterises as the rule/list fallacy). The dual-processing view of productivity is, in turn, motivated by the model’s parsimonious characterisation of linguistic storage, according to which productive patterns are extended through the application of symbolic rules in the absence of stored redundancies (see above). Conversely, non-productive patterns are assumed to proceed via direct retrieval from memory or via analogy (e.g. Clahsen 1999:996). However, as demonstrated by e.g. Barðdal (2008), productivity is a gradient property of language that is mediated by grammar-independent factors such as frequency and schematicity. Fourth and finally, the impact of non-linguistic factors like frequency of use on usage choices, as evident through language change, is measurable. In other words, the usage-based approach enjoys support from beyond the realm of language, suggesting that it is falsifiable on terms that are not internal to the grammar nor the approach alone (see Bybee 2015:102). For the reasons just given, I consider both aims set out above to have been met. In support of adopting the usage-based cognitive approach, I elaborate on its main tenets in the next section. 3.2 The usage-based cognitive approach The current section outlines the tenets, principles, and axioms of the usage-based cognitive approach to language change. In doing so, the current section also serves as a brief state of 22 Usage-based cognitive grammar and language change the art via reference to notable and/or influential works within the field. The terminology associated with the usage-based cognitive approach will be defined through use of Insular Nordic examples where possible. Otherwise, examples are lifted from the works of others and acknowledged appropriately. In 3.2.1, I argue in favour of rich memory representations for language based on observable frequency effects in language change. Subsection 3.2.2 accounts for the impact of frequency as a determinant of the structure of grammar. There, I first delineate the causal relation between frequency of occurrence, the domain-general process of entrenchment, and that property of memory termed lexical strength, before arguing that knowledge pertaining to frequency is a property of both specific and general experience. Subsection 3.2.3 discusses the interaction of schematicity with frequency as a determinant of productivity. In subsection 3.2.4, I delineate the principle of cognitive economy, i.e. the correlation between the hierarchical structure of categories and the breadth of applicability of different labels within the same category as a function of analogical reasoning. Subsection 3.2.5 summarises Section 3.2. 3.2.1 Rich memory for language The current subsection elucidates the factors that support rich memory for language, picking up on arguments forwarded in 3.1 against the parsimonious view of storage associated with generative and dual-processing models. In usage-based specific terms, I identify three parameters for counting frequency –– defined in terms of token frequency, type frequency, and (intra-paradigmatic) dispersion, utilising examples from both Insular Nordic and other languages. It is argued that the measurability of frequency renders the chosen approach testable and, therefore, scientifically viable as a theoretical framework employed to account for language change as a function of language use. Finally, examples are given of automatisation both in language and beyond, demonstrating that frequency provides a means of correlating practice with the overlap of articulatory and other motor gestures. The usage-based cognitive approach to language contrasts starkly with the parsimonious view of memory characteristic of traditional generative and dual-processing models, as discussed in 3.1. Proponents of the latter assert that the capacity for linguistic memory is highly constrained and, therefore, characterised by minimal to no storage of predictable and/or redundant linguistic information. To resolve this issue, the generative and dual-processing position asserts that UG provides endowed mechanisms for linguistic competence that ensure minimal cognitive overload of memory for language (see e.g. Jakobson 1990:321). Further, the supposed means by which UG prevents overload is the 23 Jón Símon Markússon derivation of physical language or “surface forms” by the apparent application of symbolic rules to stripped-down underlying forms.20 Conversely, the usage-based cognitive approach assumes rich memory for language. In other words, the brain’s capacity for storing linguistic experience is impressively large and no more constrained by the nature of such content than prevails for non-linguistic experience. Memories are strengthened by the repeated occurrence of “the same” action or event but may fade into obscurity if they are not enforced through continued experience (Bybee 2006:717– 718). In contradistinction, traditional generative and dual-processing models posit an economically constrained lexicon, where information considered surplus to requirements for successful language processing is absent or, in some cases, even discarded from memory (see Bybee 2010:15; Langacker 1987). However, claiming rich memory for language is an empirically falsifiable proposition, as the impact that repeated use has on the strength of representations for language is tied to frequency of occurrence, which is in essence countable on the basis of recorded speech and text corpora (see below for examples). But what exactly is counted and how? One method for counting frequency is defined in terms of token frequency, which refers to the number of times a linguistic item or phrase occurs in spoken language or text corpora. According to the usage-based cognitive view, the higher the token frequency associated with a linguistic item or phrase, the stronger its cognitive representation in memory. In the words of Bybee: If we metaphorically assume that a word can be written into the [mental] lexicon, then each time a word in processing is mapped onto its lexical representation it is as though the representation was traced over again, etching it with deeper and darker lines each time. Each time a word is heard and produced it leaves a slight trace in the lexicon, it increases in lexical strength. (Bybee 1985:117) Hoffmann (2004), who in the main agrees with the view that frequency of use impacts the strength of memory representations for language, considers Bybee’s characterisation of that impact too deterministic and, therefore, an oversimplification of the dynamics involved. Thus, Hoffmann (2004:189–192) argues for the interaction of frequency with semantic relatedness and saliency, characterised in terms of the preferred realisation of a given 20 See footnote 18. 24 Usage-based cognitive grammar and language change linguistic expression, cf. semantic doublets such as the complex prepositions Eng. in light of and in the light of, where the former is more frequent in American English; the latter in British English. Further, the preferred realisation may be more salient than the non-preferred realisation of another phrase even if the latter is of higher token frequency than the former, accounting for the entrenchment of some low frequency complex prepositions. While I agree with Hoffmann’s view, it by no means leads me to doubt the deterministic nature of the impact of frequency –– all other things being equal. If a particular realisation of a linguistic item or phrase is more frequent than another realisation of that same item or phrase, the former will still be more strongly represented in memory. This view is evidenced by the often-noted correlation between the different token frequencies for inflectional forms of the same word and the direction of intra-paradigmatic levelling. In such cases, the prevalence of levelling in favour of the most frequent member of a paradigm demonstrates that although the same concept can have various realisations, i.e. distinct inflectional forms, the most frequent form has the strongest memory representation, i.e. is most salient, and is the most accessible for use. This property of storage is referred to as lexical strength. Another method for counting frequency is defined in terms of type frequency and is determined by the number of individual items –– each counted once –– that conforms to a particular pattern. For this reason, type frequency can also be characterised as dictionary frequency (following e.g. Barðdal 2009:138), because it is the existence of a linguistic item that contributes to the type frequency of a pattern, rather than that item’s token frequency. Thus, for example, the method for forming the past tense of English verbs such as creep, keep, leap, mean, sleep, and weep reflects a pattern of low type frequency because very few verbs inflect in this way, irrespective of each item’s token frequency (see below). Despite these parametric differences, both token and type frequency contribute to productivity (Barðdal 2006). As alluded to above, it is generally accepted in the usage-based cognitive literature that token frequency determines a linguistic item’s lexical strength and, consequently, the ease with which an item can be accessed for use. Given this tenet of the approach, it is assumed that past forms such as Eng. kept, meant, and slept have significant lexical strength due to their high token frequency, while the past forms crept, leapt, and wept are relatively less well represented. In other words, the token frequency of the former set should correlate with relative ease of access, while the latter set are more likely to be forgotten –– albeit only momentarily (see Bybee 2015:95). It is through such measurable 25 Jón Símon Markússon properties of use that token frequency and type frequency interact as testable determinants of change. The pattern to which the English verbs mentioned above inflect is of low type frequency. In keeping with the correlation between frequency and lexical strength, the relevant pattern is less well represented than that represented by e.g. Eng. pres. talk ~ past talked, where the past stem is identical to that of the present, with the addition of -ed. There are literally thousands of English verbs that form their past tense in this way and, therefore, the usage-based cognitive approach assumes that this method both is well represented in memory and, resultantly, easily accessible at the cost of less common methods. This view is borne out through the interaction between token and type frequency: the less frequent past forms crept, leapt, and wept are occasionally forgotten and the most common method for forming the past tense is employed in such instances, cf. past creeped, leaped, weeped. Conversely, highly frequent past kept, meant, and slept have high lexical strength as a correlate of their token frequency and, therefore, are far less likely to undergo regularisation to past *keeped, *meaned, and *sleeped. Yet a third parameter for determining the impact of frequency involves establishing a linguistic item’s dispersion, which Gries and Ellis (2015) define as the morphological contexts in which a linguistic exponent occurs. This property of use can be exemplified along two dimensions. For example, Paper I is concerned with the first dimension, where dispersion refers to the different rates at which the ending Ice. plural -ur occurs among feminine nouns, on the one hand, and masculines, on the other. The dispersion of the ending in question is highly skewed: Of the 14.92% of nouns in plural -ur, 91.89% of these are feminine; the rest masculine (see PI:2). Given the view that frequency of use impacts the strength of memory representations for language, we might assume that the dispersion of plural -ur should imbue the forms that contain it with cue validity for treatment as feminine rather than masculine –– at least in momentary instances of uncertainty or when a new form containing the ending is encountered.21 This hypothesis is tested in Paper I and shown to be correct. The second dimension is characterised as intra-paradigmatic dispersion (Ice. dreifitíðni), which in Paper III refers to the number of paradigm cells that contains a particular stem variant. In this connection, consider the inflectional paradigm for OWN fjǫrðr ‘fjord’ in (8). 21 Here and elsewhere, I follow Taylor’s (2012:187) definition of cue validity: “The cue validity of feature f with respect to category C is the probability of C given f, i.e. p(Cf).” 26 Usage-based cognitive grammar and language change (8) sg. nom. fjǫrðr acc. fjǫrð dat. firði gen. fjarðar pl. nom. firðir acc. fjǫrðu dat. fjǫrðum gen. fjarða It is determined that the dispersion of the stem variant fjǫrð-, which occurred in four cells of the paradigm, had double the rate of dispersion of the variant firð-, on the one hand, and the variant fjarð-, on the other: both of the latter were dispersed across two cells each. The question arises, then, as to whether the different rates of dispersion for individual stem variants facilitates association of a given variant with a particular morphological value associated with one or a set of paradigm cells. This line of inquiry informs part of the analysis in Paper III. Crucially, intra-paradigmatic dispersion may not contribute to lexical strength. Consider, for example, the fact that the dispersion of the stem variant fjǫrð- is twice as high as that of the variant firð-. However, as shown in Paper III (PIII:71), the latter variant is likely to have attained the greatest degree of entrenchment due to its occurrence in dat.sg. firði, which is the most common form of fjørður as a constituent of composite place names. Evidence for frequency as deterministic of entrenchment and lexical strength can be garnered from the performance of practiced behaviour –– both linguistic and non-linguistic. A correlate of practice is commonly witnessed as different frequency effects in language change, specifically, the reducing effect and the conserving effect. The former is characterised by the gradual phonetic reduction of linguistic sequences –– segments, affixes, words, sentences etc. –– and is attributed to the process of automatisation, i.e. increased integration of repeated sequences of individual articulatory gestures. In other words, the more frequently a routine is practiced, the more integrated become the movements that make up that routine (see e.g. Langacker 2008:16; Bybee 2010:39–40 and sources cited there; also the discussion of Eng. Goodbye in 3.1). An example of linguistic automatisation, characterised more specifically as chunking, can be lifted from Bybee (2010:41), who discusses the disparate degrees of gestural overlap between separate articulations of Eng. I don’t know. Use of the phrase exhibits multiple 27 Jón Símon Markússon articulatory variants, cf. e.g. I dunno, and others that are even further contracted. By comparison, separate instances of I don’t inhale exhibit little articulatory variation. Further, a highly contracted form of the latter would unlikely be understood, even when used in response to the question Do you inhale?, i.e. when use of I don’t inhale might be expected. Similarly, semantically equivalent phrases in other languages undergo automatisation to respectively comparable degrees: Faroese Eg veit tað ikki ‘I don’t know’ often takes the form [eɔiʰtʃː] or even [ɔiʰtʃː], while Far. Ég innhaleri ikki has no form that is contracted to any degree of equivalency. The conserving effect has been described by Bybee and Thompson (1997:381), such that “...high frequency sequences ... resist change on the basis of newer productive patterns ...” This effect was exemplified above through reference to the development of the English strong verbs keep, mean, and sleep, on the one hand, and creep, leap, and weep, on the other. Due to their high frequency, individual forms of the former set are likely to have significant lexical strength. Conversely, forms of the latter set occur far less frequently and, as a direct consequence, are more likely to be forgotten –– albeit momentarily. The examples forwarded above provide direct evidence for rich memory representations for language. Further, as frequency of use can be correlated with specific frequency effects, this property of use also provides a measurable gauge for that correlation as a function of practice. Conversely, generative, and dual-processing models argue that linguistic units are either derived via the application of symbolic rules to stripped down, sometimes highly abstract, and preferably redundancy-free underlying forms (“regular” inflection) or retrieved directly from the lexicon (“irregular” inflection). According to the latter position, then, information about frequency of use is superfluous to requirements and, due to the economically constrained view of memory for language, not easily retained in memory.22 Therefore, traditional generative approaches typically fail to account for the significantly varying degrees of automatisation exhibited among conventionalised linguistic sequences or resistance to regularisation as a correlate of frequency. In order to successfully posit language use as an emergent function of domain-general cognition (see 3.1), it is necessary to demonstrate that the process is impacted by the same factors that facilitate automatisation in non-linguistic sequences. In this connection, 22 This position is obviously the yield of circular reasoning: Does the nature of derivation from a single lexical form ease the burden of storage on memory –– then as a function of genetic endowment –– or does the economic nature of memory for language –– then as a function of genetic endowment –– determine the redundancy-free nature of derivation? 28 Usage-based cognitive grammar and language change comparisons can be drawn with the repeated sequenced behaviour involved in the process of learning to use a keyboard or the performance of a complicated dance routine. Indeed, such processes are characterised by the performance of conventionalised sequences of individual actions that gradually exhibit increasing gestural overlap as a correlate of practice. The same cognitive and anatomical mechanisms that facilitate gestural overlap in non-linguistic sequences, such as dance routines, are considered the same as those that are at play in language use. In other words, the separate actions which constitute both linguistic and nonlinguistic routines become conventionalised through gradual entrenchment of increasing gestural overlap due to repetition (Bybee 2010:34). In terms of rich memory, then, prior experience –– both linguistic and non-linguistic –– must surely be stored as both specific and general memory representations (see Bod, Hay and Jannedy 2003; Bybee and McClelland 2005; Posner and Keele 1968). Otherwise, we should always be starting from the beginning each time a sequence of actions is repeated, with no stored basis for automatisation based on practice. Further, if prior experience was discarded from memory –– a necessary entailment of dual-processing –– we should also fail to break routines down into their individual constituent sequences once automated. Moreover, it is clear that frequency of use effects not only language change but, as part of that process, the structure of grammar also. For this reason, it is necessary to account for the relation between frequency and lexical strength, as well as the means by which both specific and general linguistic information comes to be stored as a function of rich memory. These properties of language use are discussed in the next subsection. 3.2.2 Usage and the structure of grammar The current subsection states the core axiom of the usage-based cognitive approach, before discussing and subsequently demonstrating the impact of frequency on the domain-cognitive process of entrenchment as both a function and facilitator of analogy (as defined in 3.1). I then argue through reference to chunking that the impact of frequency on memory for language is driven by the implicit domain-general cognitive process of statistical learning. Finally, the discussion argues for statistical learning as a determinant of the domain-general process of categorisation and the hierarchical nature of taxonomies for varyingly specific/general schemas. Subsection 3.2.1 argued for rich memory for language in light of observable frequency effects in language change. Indeed, without rich memory, we should not expect to observe correlations between change and frequency of occurrence. The examples forwarded in 3.2.1 29 Jón Símon Markússon are, therefore, taken to demonstrate that experience with language is not discarded from memory once the linguistic system is established but, rather, is stored and informs further usage choices, which pattern with prior experience. In this sense, grammar is emergent over the language user’s lifetime (see 3.1). It is in this context that the core axiom of the usagebased cognitive approach to language change, stated in (9), proves highly relevant (following e.g. Bybee 2010:2; Bybee and McClelland 2005:382; Hay 2002; Hopper and Thompson 1984; Tomasello 2000). (9) Language use determines the emergent structure of grammar, while that emergent structure determines further usage patterns. Through explicit statement of this axiom, the usage-based cognitive approach proves free of the kind of (albeit inadvertent) circularity that characterises derivation from the underlying forms of generative grammar (see Bybee 2010:15; Langacker 1987; also footnote 18). In order to account for the impact of frequency on grammar as emergent from use, as well as for the role of frequency in facilitating lexical strength, the domain-general cognitive process of entrenchment is here posited as a causative intermediary. In domain-general terms, Schmid (2017:3–4) defines entrenchment as the continual (re)organisation of knowledge in memory over the individual’s lifetime, based on the mapping of new experience onto old. According to Gentner and Markman (1997), new experience is mapped onto stored experience(s) to the degree that the former reflects instances of the latter. In other words, the perception of similarity across instances of experience facilitates analogy, i.e. the process that extends existing knowledge to new contexts. As noted in 3.1, through reference to Bybee (1985), each time mapping occurs, it strengthens the representation of the relevant experience in memory. Attributing the degree of entrenchment attained specifically to frequency of occurrence, Langacker asserts: Every use of a structure has a positive impact on its degree of entrenchment, whereas extended periods of disuse have a negative impact. With repeated use, a novel structure becomes progressively entrenched, to the point of becoming a unit; moreover, units are variably entrenched depending on the frequency of their occurrence. (Langacker 1987:59) 30 Usage-based cognitive grammar and language change In Langacker’s view, entrenchment also diversifies variation exhibited across separate memory traces that stem from experience of “the same” or similar phenomena. This function stems from the fact that no two instances of what is otherwise considered “the same” experience are ever likely identical. In other words, mapping necessarily elaborates the emergent memory representation as a function of entrenchment. In turn, as the relevant knowledge is also emergent through entrenchment, diversification as a function of mapping continually increases the extensibility of that knowledge to new contexts. It is this function of rich memory that facilitates the storage of specific information. In turn, as a function of analogy, individual instances considered to represent the same or similar experience are subject to mapping between instances, yielding schemas of more general information, where ‘schema’ is defined as a superordinate concept that specifies the basic outline of multiple more specific concepts (following Tuggy 2007:83). In their capacity as varyingly elaborate cognitive representations of experience, schemas serve as points of reference that are accessed for the purpose of categorising newly encountered phenomena based on similarity to prior experience (Taylor 2003:71). In the context of linguistic experience, then, the frequency with which a linguistic unit is used, i.e. the correlate of token frequency, or the number of items to which an inflectional pattern can be applied, i.e. the correlate of type frequency, determines the degree to which the relevant schema becomes entrenched. It is on the basis of this causal relation between frequency and entrenchment that a correlation between entrenchment and lexical strength can be discerned. Despite the opposing functions of the frequency effects exemplified in 3.2.1, the reducing effect should not be considered at odds with the claim that frequency strengthens cognitive representation. Indeed, as already noted, greater phonetic reduction is the correlate of practice. Automatisation –– the mechanism of the reducing effect –– often facilitates the process of chunking, by which sequences of distinct elements come to be analysed as gestalts (see Bybee 2010:34 and sources cited there). Langacker (1987; also Bybee 2010:44– 46) argues that, as with all experience, the accessibility of such chunks from memory is dependent on the frequency of the relevant sequence as part of prior experience. Therefore, an inherent property of chunks is relatively deeper entrenchment in and greater accessibility from memory.23 23 Note that greater accessibility is also a characteristic property of the conserving effect of frequency, demonstrating that distinct functions of frequency effects are by no means at odds. 31 Jón Símon Markússon Thus, it is possible to posit the causal relation frequency > entrenchment > lexical strength. Given the supporting arguments forwarded above, we should logically expect frequent linguistic units and patterns to be more accessible and, therefore, more frequently accessed, than those of lower frequency (see below). Indeed, Bybee (2015:95), in a similar spirit as Langacker, states that items and patterns of extreme low frequency are in danger of being forgotten and, eventually, disappearing, attributing atrophy to properties of lexical strength. In other words, when multiple linguistic items or patterns are in competition given the appropriate context of use, the most frequent option may usurp the “correct” form or pattern if the latter is temporarily forgotten due to low lexical strength (see the discussion of saliency in 3.2.1). Indeed, the process of levelling repeatedly demonstrates that the most frequent members of inflectional paradigms tend to win out in such circumstances, serving as the model for deduction of innovative forms within the paradigm (e.g. Bybee 1985, 2015:102; Tiersma 1982). Based on the above, the impact of frequency on language structure is a measurable, falsifiable, and, therefore, scientifically sound basis upon which to account for the factors that determine the course of language change. Continuing in this spirit, then, the analyses presented in Papers I–III should support a view of language change as a function of the causal relation frequency > entrenchment > lexical strength. Now, it is necessary to delineate the mechanism for that impact. To this end, let us consider the question in (10). (10) Do language users keep an exact tally of linguistic exemplars in memory or does such information accrue in the form of emergent schemas? In order to answer this question, it is necessary to understand the domain-general process of statistical learning, through which tacit knowledge pertaining to frequency of use becomes entrenched in schematic form (see Jost and Christiansen 2017). In line with the discussion of both specific and general information, above, there is mounting evidence that language users retain memories of specific instances of language use (e.g. Bybee and McClelland 2005:384). However, as long since demonstrated by Posner and Keele (1968), schematic knowledge is accrued and elaborated on the basis of specific instances of prior experience that are considered the same or similar. In other words, the cognitive representation of general experience is predicated on prior mental organisation of specific information (Gerken 2010). 32 Usage-based cognitive grammar and language change Rosch (e.g. 1975, 1978) has demonstrated that categorisation is the domain-general process through which living organisms cognitively organise their environment by assigning structure to the phenomena they encounter in it. The process of categorisation, therefore, serves as a means of uncertainty reduction.24 However, Rosch et al. (1976) show through a series of experiments that categories are by no means defined in terms of the necessary and sufficient features that characterise the classical approach to categorisation: the notion that in order to belong to a category, a given phenomenon must necessarily exhibit a specific set of features and that –– in circular fashion –– if the necessary features are discernible in one and the same entity, these are sufficient for assignment to the relevant category (see Taylor 2012:186). Rather, Rosch and colleagues demonstrated that categories can be elaborated to such a level of abstraction as to contain phenomena which at lower levels would represent distinct, i.e. respectively specific, instantiations of individual category members. In other words, thinkers are able to generalise over instances of experience with specific items considered the same or similar so that a more general schema is elaborated. Crucially, both specific and general information, i.e. schematic knowledge representative of various levels of abstraction, are retained in memory and form hierarchical taxonomies of varyingly abstract schemas. For example, Rosch et al. argue that the category FURNITURE is more abstract than hierarchically subordinate categories such as CHAIR, TABLE, BED. In turn, CHAIR is more abstract than e.g. DINING CHAIR, ARMCHAIR, BEANBAG; TABLE ROOM TABLE, DESK; BED is more abstract than COFFEE TABLE, DINING is more abstract than DOUBLE BED, SOFA BED, HAMMOCK. For examples of such taxonomic structure within linguistic categories, consider alternation of the kind OWN/Ice. masc. nom.pl. -ar ~ acc.pl. -a and nom.pl. -ir ~ acc.pl. -i, cf. fuglar ~ fugla ‘birds’ and gestir~ gesti ‘birds’. While alternation of this kind has been levelled in Faroese, cf. Far. nom./acc.pl. fuglar, gestir, Icelandic has both retained and extended both patterns to new paradigms, cf. the loans Ice. nom.pl. iPadar ~ acc.pl. iPada ‘iPads’, nom.pl. barir ~ acc.pl. bari ‘bars’. Crucial in the context of the current discussion is that extension never yields a mixed pattern: alternation of the kind **nom.pl. -ar ~ acc.pl. -i or **-ir ~ -a never occurs. 24 Uncertainty in the context of inflectional morphology is often defined and quantified in terms of entropy, i.e. the degree of uncertainty that pertains to the predictability of one inflectional form on the basis of others from the same paradigm (see e.g. Cser 2023, who quantifies the entropy of inflectional forms within regular Latin verb paradigms). 33 Jón Símon Markússon Relating this to the emergence of general information from specific information, the fact that these functionally identical yet formally distinct relations are systematically extended to new contexts in Icelandic suggests that language users discern a partial identity relation between the relevant forms within each class. However, it can be argued that, simultaneously, yet more schematic knowledge precludes mixing of the respective inflectional patterns. In other words, while application of each specific pattern proceeds on a word-by-word basis, the general knowledge that the individual patterns are distinct necessarily represents a greater level of abstraction. To demonstrate, consider the “kind of” rule of referral for Icelandic stated in (11).25 (11) If the nominative plural of a masculine noun ends in -ar, its accusative plural form will end in -a (and vice versa). As well as that stated in (12). (12) If the nominative plural of a masculine noun ends in -ir, its accusative plural form will end in -i (and vice versa). The stipulations in (11) and (12) imply more abstract relational knowledge that can be represented by the sister schemas [-V1r]nom.pl. ~ [-V1]acc.pl., where subscript ‘1’ indicates phonetic identity between intra-paradigmatic instances of V, i.e. {a, i}.26 Indeed, were more abstract relational knowledge not at play, use of the relevant forms should not preclude alternation of the kind **nom.pl. -ar ~ acc.pl. -i or **-ir ~ -a to the degree that actual usage patterns show. Now, compare (11) and (12) with the statement in (13). 25 Hansson (2007:92) defines rules of referral as stipulations that account for syncretism within an inflectional paradigm. The “rule” states a purely morphological fact, whereby “the realization of one inflectional form is systematically ‘referred to’ that of another form within the same paradigm”. Above, “kind of” refers to the nearsyncretism stipulated by the “rules” in (11) and (12). Rules of referral are thematically akin to e.g. Blevins’s (2016:105) idea of “implication as uncertainty reduction” and Wurzel’s (1984:208) implicative paradigm structure conditions (PSCs) and are not to be confused with the symbolic rules of generative grammar. 26 Audring (2019) defines the relation between sister schemas in terms of equivalency in the level of complexity conveyed. Such schemas are referenced as a means of checking pertinent semantic and/or formal distinctions between the sets of forms over which the relation abstracts. 34 Usage-based cognitive grammar and language change (13) Alternation of the kind nom.pl. -V1r ~ acc.pl. -V2 (V = {a, i}) never occurs in nouns. The statement in (13) is necessarily more abstract and less conditional than those in (11) and (12) –– even when the latter are stated in the most schematic terms, i.e. through reference to the sister schemas [-V1r]nom.pl. ~ [-V1]acc.pl.. This is because (11) and (12) must first stipulate the kinds of alternation that manifest the relevant schema. The immediately preceding preamble can also be framed in terms of frequency of use and the structure of grammar, as the respective patterns are extended at different rates. Thus, alternation of the kind nom.pl. -ar ~ acc.pl. -a has been far more productive through the centuries than nom.pl. -ir ~ acc.pl. -i. In this connection, Bybee (e.g. 2010:67) notes a correlation between high type frequency and productivity. Crucially, the vast majority of Icelandic masculine nouns exhibits the pattern nom.pl. -ar ~ acc.pl. -a, while those that have nom.pl. -ir ~ acc.pl. -i are relatively few (Svavarsdóttir 1993:105). Therefore, their different rates of productivity should come as no surprise, but clearly support a correlation of the productivity that a given pattern exhibits and the type frequency of that pattern. The preceding discussion necessitates the question: How can language users acquire, retain, and utilise knowledge pertaining to frequency unless they keep count of each instance of use? The answer is: They cannot. However, the frequency with which a pattern’s schema is referenced and instantiated continually entrenches it in memory (e.g. Cordes 2017; Janda 2002). In other words, the process of statistical learning occurs tacitly, continuously reshaping schematically represented knowledge pertaining to frequency of occurrence. As the relevant knowledge becomes further entrenched, schemas are elaborated to accommodate new instances of experience. This process by no means requires that existing knowledge or less abstract schemas be discarded from memory in the fashion of more parsimonious approaches to linguistic storage (see 3.1). It is, therefore, unnecessary to assume that specific memories of every single instance of use are readily accessible from memory, even though some may well be (see above). In other words, language users may categorise via reference to a specific memory, although, as experience accrues, categorisation is increasingly likely to proceed via reference to schemas whose inherent generalisations are informed by the application of statistical learning to prior experience. Further, by its nature, the same process teaches the language user to rely on the cue validity of one inflectional form for the intra-paradigmatic occurrence of another (see 3.2.1). Thus, users of Icelandic know that nom.pl. -ar implies acc.pl. -a (and vice versa) 35 Jón Símon Markússon because alternation of the kind **nom.pl. -ar ~ acc.pl. -i or **-ir ~ -a does not occur and, therefore, does not inform the entrenchment of the relevant schemas.27 With regard to the structure of grammar, it is argued above that this is emergent from the impact of frequency of use on domain-general processes such as statistical learning, entrenchment, generalisation, and categorisation. Thus, in Hoffmann’s (2004) terminology, the most frequent instantiations of a concept or pattern are the most deeply entrenched in memory and, therefore, the most salient also. Due to the causal relation frequency > entrenchment > lexical strength, the relevant schemas for deeply entrenched knowledge are more readily accessible and, therefore, more likely to be applied as a means of uncertainty reduction when categorising previously unencountered phenomena, including the application of inflectional patterns to borrowings, as exemplified above. This then begs the question: Are the factors that facilitate entrenchment strictly deterministic? Providing an answer to this question is the aim of the next subsection. 3.2.3 Schematicity and its interaction with frequency In the preceding subsection, I argued for the causal relation frequency > entrenchment > lexical strength and offered examples from Icelandic which indicate that highly frequent patterns are more likely than less frequent patterns to be applied in the inflection of borrowings. As argued in 3.2.2, this property of usage impacts the structure of grammar. However, it is also necessary to account for those instances in which less frequent patterns are extended. To achieve this aim, I discuss here another non-linguistic factor that drives language change, referred to as schematicity and defined as the degree of phonetic dissimilarity among the members of a class (2010:67). As demonstrated in Papers I and II, the impact of minimal schematicity is most obviously viewed as limited productivity, i.e. a relatively low rate at which a pattern is extended to new contexts (see below). The interaction of schematicity with frequency as a determinant of productivity is the focus of the current subsection. As noted in 3.2.3, high schematicity and high type frequency can interact to facilitate all but unlimited productivity. The class of English verbs in past -ed can be used to exemplify the combined effects of these factors. The class in question contains literally thousands of verbs, the numbers of which are constantly increasing. This means that entrenchment of the relevant pattern is continually enforced and, therefore, ever more easily accessed for 27 See Björnsdóttir (2021, 2023) on the nature of generalizations in grammatical gender and inflectional morphology for Icelandic nouns from a child learning perspective. 36 Usage-based cognitive grammar and language change application. Further, the class lacks clear phonetic definition, meaning that, as well as being the pattern with the strongest representation in memory, it is also applicable to items of basically any shape. Despite the combined effects just discussed, schematicity can be a facilitating factor in productivity in its own right, i.e. in the absence of significant type frequency. Thus, patterns of medium to low type frequency can exhibit limited productivity if members of the relevant class are phonetically and/or semantically similar, i.e. exhibit low schematicity. In this connection, Barðdal (2008:1) characterises the productivity of syntactic construction in terms of their attractive force. Thus, a construction is able to attract items that are similar to others that already fill its open slot(s). Further, Barðdal (2008:9) posits an inverse relation between type frequency and semantic coherence which, she argues, adequately predicts the varying degrees of productivity exhibited by different constructions. On the limited productivity of inflection classes, Bybee (e.g. 2010:69) attributes the mechanism for productivity among low-frequency, low-schematicity classes to the gang effect, by which a high concentration of common phonetic properties among a small class of items suffices to attract new members. An example of the gang effect is betrayed by the development of the formerly weak English verb wear, which now has the standard past form wore. Framing deduction of the past form in the context of interaction between frequency and schematicity, it is noteworthy that the relevant pattern has highly limited productivity. Thus, it originally applied to three verbs only, i.e. bear, swear, and tear, cf. past bore, swore, and tore. Due to the deterministic nature of the causal relation frequency > entrenchment > lexical strength, argued for in 3.2.2, representation of the relevant pattern in memory was likely weaker than that of many other methods of forming the past tense –– especially relative to that which adds -ed to the present stem. We should, therefore, expect the pattern [Xear]pres. ~ [Xore]past to be less accessible for use than many others –– all other things being equal. Further, the class containing bear, swear, and tear was minimally schematic, making it unlikely to attract verbs –– weak or strong –– of different phonetic shapes, such as Eng. go, forget, talk, write, which are phonetically dissimilar both from each other and from members of the bear ~ bore class. Nonetheless, the bear ~ bore class was able to attract weak wear on the basis of phonetic coherence with existing members. Thus, while frequency is a positively deterministic factor in language change, the preclusive impact of extreme low frequency on productivity can be counteracted by minimal to low schematicity in order to facilitate limited productivity by means of the gang effect. This is demonstrated by the examples forwarded above and in the analyses presented in Papers I and II. 37 Jón Símon Markússon 3.2.4 Categories and the principle of cognitive economy The current subsection fleshes out the relation between the taxonomic structure of categories and the applicability of category labels in categorisation, in accordance with a principle of cognitive economy. To this end, I delineate the taxonomic relation between varyingly schematic categories for formally and/or functionally related items and demonstrate the level of abstraction at which a category is most extensible to potential inductees. The level of abstraction that accommodates the most broadly applicable category labels is identified as the level that most suitably applies in neutral reference, constituting the basic level of a three-tier hierarchy. Through a series of experiments, Rosch (1975; see also Rosch et al. 1976) argued in favour of a principle of cognitive economy, by which basic category labels and contents convey maximum information with optimally minimal cognitive cost (see below). In order to demonstrate the domain-general nature of this principle, distinct non-linguistic category labels applied at respective levels of abstraction within functionally continuous taxonomies can first serve as exemplary. For this purpose, I will utilise the category FURNITURE, the taxonomic structure of which was delineated in 3.2.2, showing the applicability of that category label to members represented at different levels of abstraction. Thus, the category FURNITURE betrays abstraction to a high degree of schematicity. This is because no single feature or sets of features suffice to represent all individual category members with minimal cognitive cost. However, at the medially schematic level, features are both general enough as to motivate neutral reference to individual members but specific enough to distinguish them from those of another category label that is representative of the same level of abstraction. Therefore, CHAIR, BED, and TABLE serve to distinguish individual types of the highly schematic category FURNITURE, both from each other and from members of the category DOG, which is a medially schematic subcategory of highly abstract (NONHUMAN) ANIMAL. In turn, at the minimally schematic level, category labels such as DINING CHAIR, ARMCHAIR, and GARDEN CHAIR distinguish members of the medially schematic category CHAIR; COFFEE TABLE, DINING ROOM TABLE, DESK distinguish members of the medially schematic category TABLE; DOUBLE BED, SOFA BED, CAMPER BED distinguish members of the medially schematic category BED. In light of the representational scope of the respective category labels, the categories CHAIR, BED, and TABLE are considered basic level categories. Thus, the category CHAIR is not schematic enough to apply to items such as double beds or dining room tables but nor is it too specific as to encompass dining room chairs only; garden 38 Usage-based cognitive grammar and language change chairs, rocking chairs, and armchairs can also be assigned the category label CHAIR. In this sense, CHAIR, BED, and TABLE occupy “[t]he level used for everyday neutral reference”, a property that Croft and Cruse (2004:83) attribute to basic level categories. The less schematic status of e.g. GARDEN CHAIR and ROCKING CHAIR within the taxonomy dominated by the highly schematic category FURNITURE accounts, then, for their function as single points of reference for categorisation of chairs that are otherwise formally and functionally dissimilar. By the same token, highly schematic FURNITURE is considered superordinate to basic level CHAIR, as the former has “…fewer defining attributes than…” the latter (Croft and Cruse 2004:84). Finally, category labels such as GARDEN CHAIR and ROCKING CHAIR reflect subordinate level categories, characterised as those in which “…members have high mutual resemblance…” (Croft and Cruse 2004:85). Perception of similarity to the members of minimally schematic categories, i.e. those that represent the subordinate level, should provide for the highest number of potential one-to-one formal correspondences across sets of functionally related items. Crucial to Rosch’s characterisation of categories is that they exhibit prototype structure, i.e. some items are considered more prototypical of the category as a whole than others. In this sense, a dining chair might be considered more “chair-like” than an armchair, while the latter might in turn be considered more prototypical than a beanbag (see Rosch 1975 for experimental evidence in support of prototype structure). Thus, the common attributes of e.g. chairs are not determined by necessary and sufficient features (see 3.2.2), but, rather, converge upon the most frequent formal and functional attributes common to existing category members. In turn, the schema that abstracts over the formal and functional attributes common to all chairs informs not future assignment of items to the category, but also the status of those items with respect to prototypicality. Thus, prototypicality is best characterised in terms of family resemblance, by which category members share core attributes, but may differ both in form and the means by which they serve their intended function. The principle of cognitive economy is also a property of the structure of linguistic categories. Accordingly, schemas for such categories arrange themselves in hierarchic taxonomies, where category labels represent varyingly specific and general degrees of abstraction. Further, linguistic categories exhibit family resemblance structure that centres around prototypes. Moreover, prototype effects are observable through the varying rates at which different items are attracted by a given constituent schema of the taxonomy. This property of linguistic categories is discussed in the next section, and in more detail in Section 3.3. 39 Jón Símon Markússon 3.2.5 Section summary The current section has argued for the impact of domain-general cognitive processes on language use, which is viewed as the mechanism of language change. Subsection 3.2.1 argued in support of rich memory for language, identifying some of the different types of frequency effects that are observable through language change. Section 3.2.2 discussed frequency as a determinant of the structure of grammar, arguing for the causal relation frequency > entrenchment > lexical strength, before accounting for the tacit accrual of knowledge pertaining to frequency via statistical learning. Section 3.2.3 discussed the interaction between schematicity and frequency and its impact on the direction of change. Subsection, 3.2.4 delineated the cognitive means by which items are assigned to functionally related categories that constitute different levels of a three-tier taxonomy, defined by increasing/decreasing schematicity. The basic level, i.e. medially schematic categories, are considered those that best encompass all functionally identical but formally different category members and, therefore, is used for everyday neutral reference. Finally, it was argued that the category-internal taxonomies that are a function of the principle of cognitive economy entail prototype structure, by which some members are considered better representatives of the category, as these share different iterations of formal and functional attributes with all other category members. Thus, the structure of categories is characterised by family resemblance, rather than specific sets of necessary and sufficient features. 3.3 Analogy and the adaptive nature of categories In the current section, I posit analogy as the mechanism of domain-general cognitive processes applied to both linguistic and non-linguistic tasks. Before the discussion commences, reiteration of the usage-based cognitive approach’s core axiom –– repeated in (14), below, from (9) –– seems pertinent. (14) Language use determines the emergent structure of grammar, while that emergent structure determines further usage patterns. So, the cycle of use continues as a function of domain-general human cognition (e.g. Beckner et al. 2009). In terms of the cognitive mechanism of language use, the current thesis follows Blevins and Blevins (2009:1), who observe that “[t]here is mounting evidence from work in cognitive psychology that the talent for analogical reasoning constitutes the core of human cognition” 40 Usage-based cognitive grammar and language change (citing Penn et al. 2008), and that “analogy may be a highly domain-independent cognitive process” (citing Halford and Andrews 2007). In what follows, analogy is considered one fundamental mechanism of language change, which, in turn, is viewed as a function of language use.28 Subsection 1.1 alluded to two distinct functions of analogy: extension and levelling. In the context of the current thesis, the former process is defined as the application of an existing pattern of intra-paradigmatic alternation in a paradigm that formerly did not show it. The latter process is characterised as the deduction of an innovative inflectional form containing a stem variant previously found in another cell or cells of the paradigm (see below for examples). Thus, strictly speaking, extension involves the influence of external patterns on the internal relational dynamics of paradigms, while levelling occurs within one and the same paradigm (though see below on functional crossover between the two processes). Examples of analogical extension can be taken from Papers I and II. Thus, for example, Paper I posits the influence of feminine nouns in plural -ur on masculines in the same ending, by which the latter are reanalysed as grammatically feminine due to the ending’s 91.89% rate of dispersion among feminine nouns. Paper I argues that, on account of this property of use, masculine forms such as plural fætur ‘feet’ sometimes alternate with overtly feminine forms, cf. fem.pl.def. fæturnar instead of masc.nom.pl.def. fæturnir, masc.acc.pl.def. fæturna. The model for this pattern of alternation is exemplified by relations such as fem.pl. stelpur ‘girls’ ~ fem.pl.def. stelpurnar. Extension is considered the archetypal manifestation of proportional analogy, as depicted by the proportion in (15). (15) pl. stelpur : pl. fætur : fem.pl.def. stelpurnar fem.pl.def. X; X = fæturnar The function of proportions can be delineated in terms of the schema A : B :: C : D (see e.g. Fertig 2013:28). Thus, (15) suggests that the pattern of alternation between the A- and Bforms, i.e. stelpur and stelpurnar, respectively, is extended to the relevant cells of the paradigm for Ice. fótur ‘foot’, such that the functionally equivalent C- and D-forms, i.e. fætur 28 Bybee and Beckner (2015:503) assert that categorisation is the most fundamental cognitive process in language change. However, as noted by Rosch (e.g. 1975), categorisation is dependent on the perception of common attributes. Further, as Gentner and Markman (1997:48) have argued, the perception of functional similarity is a prerequisite for analogical reasoning. Therefore, the position is taken here that categorisation cannot proceed unless the newly encountered phenomenon exhibits similarity to another that forms part of prior experience. In other words, categorisation is here considered a function of analogy (see below). 41 Jón Símon Markússon and X, come to exhibit the pattern modelled by A and B. While the source of the C ~ Drelation is the C-form, it is formal similarity and functional equivalence between A and C that motivates the extension of the A ~ B relation to C and D, hence X = fæturnar, i.e. the target D-form. Similarly, in levelling, a target comes to occur in the same paradigm as the source upon which the former is based. An example of levelling taken from Paper III is the extension of the stem variant firð- in the paradigm of Far. fjørður to the dative plural cell, cf. innovative dat.pl. firðum, beside older dat.pl. fjørðum. Prior to levelling, the variant firð- occurred in the cells of the dative singular, cf. dat.sg. firði, and the nominative/accusative plural only, cf. nom./acc.pl. firðir.29 However, as noted by Bybee (2015:95), levelling does not involve the transformation of an older form into a younger one. In this connection, it is important to note that older dat.pl. fjørðum still exists alongside its younger cellmate. However, despite the fact that paradigm-internal constraints determine the source of levelling, functional motivation for the process can also be characterised as proportional. In brief, Paper III posits dat.sg. firði as the source of levelling to dat.pl. firðum. Crucially, in Faroese, it is most common that all forms of a given paradigm contain the same stem variant. Indeed, widespread levelling of older ø to a in dative plural stems is suggestive of a general tendency to level vocalic alternation, cf. e.g. nom.sg. hvalur, acc.sg. hval, dat.sg. hvali, nom./acc.pl. hvalir, dat.pl. hvølum/innovative hvalum (Þráinsson et al. 2012:398–399). Thus, the proportion in (16) demonstrates that external influence is also a motivating factor in levelling.30 (16) dat.sg. hva1li : dat.pl. hva1lum dat.sg. fi2rði : X; X = fi2rðum 29 It is necessary to note that the stem variant firð- has also found its way to the cell of the accusative plural, cf. OWN fjǫrðu, Far. firðir. However, this process is most likely the result of levelling that occurred in all masculine paradigms where the plural nominative and accusative forms were once formally distinct (see PIII:73). Therefore, the change in question can be attributed to the impact of type frequency on a smaller inflection class, rather than that of the token frequency of dat.sg. firði. 30 Subscript ‘1’ and ‘2’ in (16) are intended to draw attention to phonetic identity between stem vowels of distinct forms within the respective paradigms. 42 Usage-based cognitive grammar and language change Although the proportion does not specifically state it, (16) is representative of the fact that levelling does not lead to change that is incongruent with existing patterns.31 Therefore, in light of the preceding discussion, the definition of analogy in (6) can be elaborated as in (17) so as to encompass both extension and levelling (following e.g. Gentner 2005; Christie and Gentner 2010). (17) Analogy is the extension of existing relational knowledge to new contexts. In (17), “relational” implies that argument structure is aligned across scenarios that are perceived as functionally equivalent, as delineated below with both linguistic and nonlinguistic examples.32 However, in order to better grasp what the alignment of argument structure entails, it is first necessary to understand the functions of analogy that motivate it. In this connection, Gentner and Markman (1997) have shown that analogy facilitates categorisation, a prerequisite of which is structure-mapping between the relations that cause a set of scenarios to be perceived as similar (see discussion in 3.2.2). Perception of relational similarity is optimised by the identification of parallel connectivity, i.e. the state where the items that constitute individual arguments within separate relations correspond on a one-toone basis according to role (e.g. Gentner and Hoyos 2017:675). To take an example from Kotovsky and Gentner (1996), the argument structure involved in the action of a toddler waving its arm and causing a cup to fall from a table can be aligned with that of a squirrel swishing its tail and causing an acorn to fall from a tree. Thus, the toddler aligns with the squirrel; the cup with the acorn; and the table with the tree on a one-to-one basis. Crucially, there is little formal similarity between a toddler and a squirrel; a cup and an acorn; a table and a tree. However, similarity among relations is dependent on the perception of common relational structure across a set of scenarios and, therefore, formal differences between items do little to interfere with that perception (e.g. 31 Of course, the process of levelling has taken various directions within the paradigm of Far. fjørður, leading to a unique inflectional dynamic among Faroese nouns (see Sections 6.2.1 and 6.3.2). However, it is considered likely that each instance of levelling accorded with an existing pattern, although the sum of different analogical operation has blurred the initial intent for each individual development. 32 See Bulloch and Opfer (2009) for experimental findings that demonstrate the alignment of argument structure across non-linguistic domains among both children and adults. 43 Jón Símon Markússon Christie and Gentner 2007; Gentner 1983, 2005; Kotovsky and Gentner 1996; Krawczyk, Holyoak, and Hummel 2004, 2005; Markman 1997; Markman and Stilwell 2001).33 Scholars have demonstrated repeatedly that one-to-one mapping between intrinsic formal and/or functional attributes facilitates elaboration of a category’s existing schema (see Christie and Gentner 2010; Cordes 2017; Gentner 2005; Gentner and Hoyos 2017; Kotovsky and Gentner 1996). If a new inductee to a category exhibits a degree of formal dissimilarity from existing members, yet all share a functional attribute or set of functional attributes, further abstraction will ensure that the relevant schema is “fully compatible with all the members of the category it defines…” (Langacker 1987:371). Accordingly, the toddler-cuptable and squirrel-acorn-tree scenarios mentioned above could well be assigned to a category such as MOTIONS THAT MAKE OBJECTS FALL FROM SURFACES if they constitute a part of prior experience.34 Thus, as the process of schematisation abstracts further away from all but the common attributes of the category, its schematicity increases, permitting a greater degree of dissimilarity among both current and future category members, all of which nonetheless comply with the evermore abstract, emerging schema. The dynamics that prevail between the different members of a category as a function of schematisation have been characterised by Wittgenstein (1978) in terms of family resemblance. The category PREDATOR provides an example of a highly schematic relational category whose members exhibit family resemblance structure as a property of parallel connectivity. The common attributes of predators are necessarily predicated on functional opposition to prey, rather than alignment of formal attributes common to predators. In other words, while e.g. polar bears and Venus flytraps have little in common in terms of appearance –– indeed, they do not even belong to the same biological kingdom –– both are predatory species. Likewise, there is little formal 33 Gentner and Markman (1997) demonstrate with the proportion 1 : 3 :: 3 : 9 that perception of formal similarity is unnecessary for a successful analogy. Thus, despite the fact that 1 and 3, i.e. the A- and C-forms of the proportion, and 3 and 9, i.e. the B- and D-forms, have nothing in common in terms of form, pronunciation, or meaning, it is the relation between the exemplary A- and B-forms that facilitates analogy, i.e. functionally equivalent items of the proportion exhibit parallel connectivity and can, therefore, be aligned according to role: 1 goes thrice into 3 and, on that basis, the relation between 3 and 9 is understood as functionally equivalent. This is not to say that intrinsic formal attributes, common to a set of items, never assist in establishing parallel connectivity. Indeed, the greater the degree of object similarity perceived between a set of items, the greater the likelihood that their common attributes will motivate comparison, structure-mapping and, ultimately, facilitate alignment. 34 See e.g. Lakoff (2018:43) on categories established “on the fly”, such as THINGS YOU TAKE ON A PICNIC. 44 Usage-based cognitive grammar and language change identity between seals and flies, although both of these can be aligned as prey according to the relationship of each with their respective predator types.35 Gentner characterises relational categories as those: ...whose meanings consist either of (a) relations with other entities, as in predator or gift, or (b) internal relations among a set of components, as in robbery or central force system. (Gentner 2005:245) She continues: Relational categories contrast with object categories ... Of course, object categories contain not only property information but also relational information. For example, that tigers hunt and eat animals is part of our concept of a tiger, along with intrinsic attributes such as their stripes. (Gentner 2005:245) Returning to the significance of relational knowledge for the definition of analogy presented in (17), the degree to which common relational structure can be perceived determines the likelihood that existing relational knowledge is extended to new contexts, simultaneously elaborating the relevant schema. The view of categorisation just outlined can now be delineated with linguistic examples, which can be picked up from the discussion of Old West Nordic and Icelandic masculine nouns in nom.pl. -ar ~ acc.pl. -a, e.g. OWN/Ice. fuglar ~ fugla, cf. borrowed iPadar ~ iPada, and those in nom.pl. -ir ~ acc.pl. -i, e.g. gestir ~ gesti, cf. borrowed barir ~ bari (see 3.2.2). In terms of common formal attributes, it is clear that extension of the relevant patterns to the paradigms of borrowed masc. iPad and bar is minimally dependent on formal similarity. Indeed, fugl and iPad, on the one hand, and gestur and bar, on the other, have little to nothing in common in terms of phonetic structure. However, the borrowings have been 35 Note also that the category has been expanded to include human beings on metaphorical grounds, as exemplified by its application to people –– particularly men –– who take advantage of other people –– particularly women –– for sexual gratification. Likewise, the label is commonly applied to criminals who exhibit “predatory” behaviour, whether that behaviour results in death (e.g. serial killers), severe psychological harm (e.g. abusers, rapists, stalkers), or financial loss (fraudsters, burglars). Further, elaboration of the category has proceeded despite the obvious lack of formal similarity between human beings and e.g. polar bears or Venus flytraps (see Gentner and Asmuth 2019 on the metaphorical expansion of relational categories). 45 Jón Símon Markússon categorised as masculine nouns and are, therefore, expected to function as such through application of the appropriate endings.36 Through this operation, the schema for masculine nouns of the kind Ice. nom.pl. -V1r ~ acc.pl. -V1 was elaborated in the following ways. First, assignment of masculine grammatical gender to iPad and bar elaborated the schema for masculines so that it extended its remit to more nouns in nom.sg. -Ø. The lack of an ending in this form is relatively rare among masculine nouns and is mostly associated with stem-final -l, -n, -r, and -s, e.g. masculine fugl ‘bird’, turn ‘tower’, vír ‘wire’, ís ‘ice’ (see PII:215). Most masculine nouns end in either -ur, e.g. hestur, gestur, or -i, e.g. granni ‘neighbour’, in the nominative singular.37 Secondly, extension of the relational schema [XV1r]nom.pl. ~ [XV1]acc.pl. to include nom.pl. iPadar and barir served to elaborate the formal properties of the relevant schemas. In doing so, the emergent schemas are imbued with even more phonetic scope, further facilitating its extensibility. For this reason, the schema remains highly productive and is able to attract new items such as iPad and bar, facilitating their engagement in the relation type nom./acc.sg. -Ø ~ nom.pl. -ar ~ acc.pl. -a, cf. iPad ~ iPadar ~ iPada, and nom./acc.sg. -Ø ~ nom.pl. -ir ~ acc.pl. -i, cf. bar ~ barir ~ bari. In turn, such developments further confirm and entrench the knowledge that endingless singular forms can alternate regularly with nom.pl. -V1r and acc.pl. -V1. Thirdly, such instances of elaboration further entrench the relevant schema, in turn strengthening its representation in memory and, therefore, facilitating recall when the language user requires a point of reference for the inflection of a new masculine noun. Fourth, further entrenchment of the schema bolsters the cue validity that a form of the kind nom.pl. -V1r has for another of the kind acc.pl. -V1 –– and vice versa. Further, due to the process of statistical learning, the more frequent extension of the alternation nom.pl. -ar ~ acc.pl. -a relative to that of nom.pl. -ir ~ acc.pl. -i is in large part a function of the higher type frequency of the former. Moreover, as the stems of nouns belonging to masculine classes in nom.pl. -ar, acc.pl. -a can be of practically any phonotactically permissible structure, their schemas are all but completely open. As noted in 3.2.3, high type frequency in conjunction with high schematicity can facilitate near limitless productivity. 36 As far as I am aware, Ice. bar is the only masculine borrowing to have taken up the pattern nom.pl. -ir ~ acc.pl. -i (see also Kvaran 2005:345). Its assignment to the relevant class is probably due to phonetic similarity with e.g. hvalur ‘whale’, salur ‘hall’, dalur, ‘valley, dale’, staður ‘place’, selur ‘seal’, refur ‘fox’, and others, all of which contain a rhyme in a lax vowel followed by a sonorant consonant and end in -ir in the nominative plural. 37 See Þórhallsdóttir (1997) on evidence from language change for association of the ending nom.sg. -ur with masculine grammatical gender. 46 Usage-based cognitive grammar and language change The arguments forwarded above support the core axiom of the usage-based cognitive approach reiterated in (14), above. Further, it identifies language use and, therefore, language change as functions of analogy. In other words, prior knowledge of relations and the argument structure that cause them to be perceived as similar facilitate extension of the relevant schematic knowledge to new contexts for the purpose of categorisation (see (17), above). Subsequent to extension, the schema is further elaborated to encompass new contexts of use, affecting the structure of schematic knowledge –– linguistic or otherwise. 3.4 Chapter summary The current chapter delineated the theoretical approach applied in Papers I–III in both general and context-specific terms. Section 3.1 compared the merits of and selected between opposing theoretical approaches to language change, i.e. generative and dual processing models, on the one hand, and the usage-based cognitive approach, on the other. In this connection, it was noted that the former posits genetically endowed language competence and assumes language change to be a function of grammar simplification, the mechanism for which is viewed as inter-generational transfer of the discerned grammar during first-language acquisition. In order to identify the oversimplified view of language change as inherent to such claims, I presented examples of change that obviously complicates grammar, while further examples demonstrate that change can be a function of use among adult language users. It was also noted that while traditional generative and dual-processing approaches assume little role for the impact of frequency in language change, this position is incongruent with observable frequency effects. It was subsequently argued that the usage-based cognitive approach converges with the facts, e.g. that change occurs in adult language as well as in child language and that frequency impacts the process. Section 3.2 fleshed out the tenets, principles, and axioms of the usage-based cognitive approach and simultaneously served as a brief state of the art. Terminology associated with the approach was defined in the context of Insular Nordic examples of morphological change. Subsection 3.2.1 argued in favour of rich memory for language through reference to different types of frequency effects in language use and change. Subsection 3.2.2 accounted for the impact of the causal relation frequency > entrenchment > lexical strength on the structure of grammar, before arguing that knowledge pertaining to frequency imbues memory representations for both specific and general experience. Subsection 3.2.3 examined the interaction of frequency with schematicity, where the latter serves to facilitate limited 47 Jón Símon Markússon productivity in the absence of significant type frequency. Subsection 3.2.4 discussed prototype structure as a property of categories, itself a function of cognitive economy. Subsection 3.2.5 offered a summary of Section 3.2. In Section 3.3, I discussed the assignment of linguistic and non-linguistic items to relational categories, the entrenchment of these in memory, and elaboration of relational schemas as a function of analogy. 48 4 Methodology In this Chapter, I delineate the methodology employed for amassing and analysing the data presented in Papers I–III, in line with the theoretical tenets of the usage-based cognitive approach fleshed out in Chapter 3. Section 4.1 outlines the methodology employed for analysing the Icelandic data reported on in Paper I. Section 4.2 delineates the theoretical approach that determined the choice of methodological approach to the Icelandic data analysed in Paper II. Section 4.3 provides a detailed account of the methodological approach employed for analysing the data reported on in Paper III. 4.1 The methodology employed in Paper I As reported in Paper I, according to counts based on the IsTenTen electronic corpus,38 reanalysis of masculine plural fætur ‘feet’ as feminine occurs in the minority of instances. The gauge for the rate of reanalysis followed in Paper I is the number of individual occurrences of fem.pl.def. fæturnar (1,274 instances), masc.nom.pl.def. fæturnir (1,426), and masc.acc.pl.def. fæturna (2,882) (see PI:13–14). To determine the rate, the number of feminine forms is calculated as a percentage of the total number of instances (5,582 instances). Therefore, reanalysis occurs in 22.82% of instances. In this connection, Paper I also identifies an interesting property of reanalysis pertaining to masculine forms in plural -ur: fem.pl.def. fæturnar is far more frequent than e.g. fem.pl.def. veturnar (of masculine vetur ‘winter’). This fact is considered interesting because masc.acc.pl.def. veturna is much more frequent than both masc.nom.pl.def. fæturnir and masc.acc.pl.def. fæturna combined (see PI:14). At first glance this property of reanalysis might be considered understandable in light of the causal relation frequency > entrenchment > lexical strength. Indeed, according to the usage-based cognitive approach, masc.acc.pl.def. veturna is deterministically more entrenched on account of its relatively higher frequency and, therefore, logically has greater lexical strength than both masc.nom.pl.def. fæturnir and masc.acc.pl.def. fæturna. As a correlate of relative ease of access, we should expect the formal properties of frequent forms to be less subject to change than those of less frequent 38 The IsTenTen corpus is an electronic text corpus for Icelandic. It is maintained by Sketch Engine and contains sources including social media. On my reasons for choosing to present results from this corpus rather than those returned from other electronic corpora for Icelandic, see Paper I (PI:23, endnote 18). See the bibliography for the relevant URL. 49 Jón Símon Markússon forms because ease of access implies less need to create a novel form in on-line language processing, which can happen when a given form is momentarily forgotten (see 3.2.1). All other things being equal, frequency counts from text corpora should either confirm or refute such correlation. However, searches of the corpus for the individual forms listed above demonstrate a mismatch between the descending token frequencies of the respective feminine doublets relative to those of corresponding masculine forms for the same word. In other words, e.g. while masc.acc.pl.def. veturna is much more frequent than both masc.nom.pl.def. fæturnir and masc.acc.pl.def. fæturna combined, fem.nom./acc.pl.def. fæturnar is the most frequent feminine doublet. It is in light of this mismatch that schematicity, in addition to token frequency counts, is invoked as an explanatory device. Paper I utilises Bybee’s Network Model (e.g. 1985), with the employment of some innovative notational features. This method for modelling the emergent nature of formfunction correspondences utilises connecting lines between segments or larger elements common to the words that form a given class. The lines are of varying thickness, depending on the nature of the correspondence –– formal and/or functional (see PI:19; also 5.3.5). The greater the number of lines that connects the items in the network, the less schematic the set and the greater the likelihood that it is representative of an easily definable class. Crucially, the greater the number of one-to-one correspondences is shown in Paper I to correlate with the impact of minimal schematicity as a facilitator of limited productivity. Through employment of the model and the innovative features in Paper I, it proves possible to not only demarcate the emergent status of specific sequences as representative of grammatical function, cf. the association of the sequence -ur with the function nominative/accusative plural, but also to model the degree to which other sequences within a word align with prototypical formal attributes of a microclass within a subset of strong feminine nouns in plural -ur. The number of connections denoting a form-function match between individual forms in the network is argued to correlate with the potential for alignment of those forms with a specific schema (see the network in PI:19 and those represented in 5.3.5). Thus, the network presented in Paper I shows that plural fætur is the only original masculine form in -ur that aligns perfectly with the subordinate level schema [XæTur]nom./acc.pl., which was posited for functionally equivalent forms of the feminine Xó/æT-microclass, i.e. Ice. plural blækur ‘nonentities, wretches’, bækur ‘books’, bætur ‘patches, remedies, compensation, financial aid’, brækur ‘trousers’, nætur ‘(fishing) nets’, 50 Methodology rætur ‘roots’. Additionally, plural fætur also aligns with the basic level schema [Xur]nom./acc.pl. on account of the ending plural -ur alone (see 3.2.4 on taxonomic relations between the constituent schemas of categories; see 5.2 and 5.3.2 on the prototype structure and taxonomic structure of the subtype for Icelandic feminine nouns in plural -ur specifically). Conversely, plural vetur aligns at best only tentatively with subordinate [XæTur]nom./acc.pl. (see PI:19–20). Therefore, the network demonstrates that plural vetur is at best formally ambiguous with regard to class membership, even though it is easily associated with feminine grammatical gender due to alignment with basic level [Xur]nom./acc.pl..39 Thus, plural fætur can be specifically attracted by two schemas, each representative of its own level of abstraction. Therefore, it is argued that frequency counts from the IsTenTen corpus provide support for the usage-based cognitive approach for the following reasons. First, in line with the posited impact that frequency has on entrenchment and lexical strength, we should expect the attractive force of a schema that abstracts over very few items to be limited –– all other things being equal. Accordingly, as noted above, plural fætur is reanalysed as feminine in the minority of cases. However, this then begs the following questions: Why is fem.pl.def. fæturnar markedly more frequent than any functionally equivalent feminine doublet of an original masculine form in plural -ur? Given that any form with the ending in question can be aligned with medially frequent [Xur]nom./acc.pl., do the rates of reanalysis reported in Paper I not suggest that low frequency [XæTur]nom./acc.pl. nonetheless exhibits greater productivity? It is in this context that the condition “all other things being equal” proves highly relevant. The answer to the questions just posed brings us to our second reason, namely that low frequency can interact with low schematicity to facilitate the gang effect as a function of prototype structure (see 3.2.3 and 3.2.4). Thus, employment of the network model in Paper I demonstrates that alignment with subordinate level, low frequency [XæTur]nom./acc.pl. facilitates assignment of plural fætur to the Xó/æT-microclass and, therefore, identity with specifically feminine formal attributes beyond plural -ur. For this reason, plural fætur is discernibly far less ambiguously feminine in form than plural vetur. As is argued in Papers I and II, frequency counts also shed light on the nature of reanalysis within the context of the paradigm as a whole. Both papers account for reanalysis as extension, by which formal and/or functional similarity across intra-paradigmatic relations is aligned as a function of parallel connectivity. Thus, it is argued, the cue validity of plural 39 See 5.3.2 on the status of the sequence -ur in forms such as bræður ‘brothers’, fingur ‘fingers’, and vetur. 51 Jón Símon Markússon fætur suffices to motivate deduction of innovative fem.pl.def. fæturnar, in line with existing relations such as feminine plural rætur ~ fem.pl.def. ræturnar. However, frequency counts for other forms of Ice. fótur reveal that singular forms of the word hardly –– if ever –– occur in an overtly feminine form (PI:8). Therefore, this property of usage is considered evidence for reanalysis as a two-step process, by which individual forms have cue validity for the treatment of certain others within the paradigm, as opposed to a wholesale process that affects all forms simultaneously. Following from the above discussion, a methodological approach that employs frequency as diagnostic of productivity supports rich memory for language. As argued in Paper I, it is likely overtly masculine attributes exhibited by singular forms of the noun Ice. fótur that impede reanalysis as feminine. By the same token, plural fætur patterns perfectly with other feminine forms at two distinct levels of abstraction (see above). Therefore, the formal attributes in question, as well as their dispersion over gender-determined contexts, must have representation in memory –– otherwise, we should not expect to find a correlation between attributes’ force in numbers and the rate of reanalysis.40 4.2 The methodology employed in Paper II As noted in Chapter 1, Papers I and II are related in terms of thematic content: Both deal with gender reanalysis and assignment of grammatical gender on account of similarity to feminine nouns in plural -ur. However, unlike Paper I, Paper II seeks to discern the different analogical means by which different forms of the Icelandic borrowings blók ‘wretch, nonentity’ and kók ‘Coke’ are deduced, i.e. by intra-paradigmatic levelling and extension. The different mechanisms of change are accounted for via reference to non-linguistic factors such as frequency and schematicity, on the one hand, and semantics and pragmatics, on the other. The methodology employed in Paper II seeks to highlight the basic forms for levelling in terms of a relation between meaning and form, which, as demonstrated by Bybee (1985), is reflected cross-linguistically in paradigmatic structure. Synchronically, the relation between meaning and form is betrayed by formal coherence among forms that share a semantically highly significant morphological value, such as those defined in terms of tense and aspect among verbs, in terms of number among nouns. By the same token, the tendency towards phonetic coherence as a correlation of shared semantics can lead to formal distinction between forms on different sides of the significant semantic divide within the 40 In this connection, see 3.2.1 on the development of Eng. keep, mean, sleep vs. that of creep, leap, weep. 52 Methodology paradigm. Diachronically, the relation between meaning and form is manifested through levelling as forms that share a significant semantic property are levelled to resemble each other more than those forms that express an opposing value (see e.g. Lahiri 2000; also PIII).41 Informed by this approach to meaning and form, Paper II argues that innovative forms of Ice. kók, which usually shows agreement as neuter in the mass noun sense, are deduced via different analogical means, i.e. extension or levelling. Further, application of either mechanism depends on the degree of semantic relatedness to the source of innovation, i.e. nom./acc.sg. kók. On the basis of numerical evidence, Paper II demonstrates that the formal distinction between nom./acc.sg. kók and innovative feminine plural kækur, when the latter occurs instead of plural kók, aligns with the semantic difference between singular and plural, as mirrored in functionally equivalent relations of the Xó/æT-microclass, cf. e.g. bók ~ bækur, rót ~ rætur, etc. Conversely, levelling to fem. dat.sg. kók, instead of neut. dat.sg. kóki, where the latter is most common in the mass noun sense, reflects the semantic relatedness of the relevant forms one to another, cf. the source nom./acc.sg. kók as used in the count noun sense. Crucially, syncretism in nominative/accusative/dative singular is a highly frequent attribute of strong feminine nouns in Icelandic, cf. e.g. nom./acc./dat.pl. fem. bók ‘book’, mynd ‘picture, form’, skeið ‘spoon’, suggesting the influence of type frequency and dispersion on the direction of levelling. If all members of the paradigm for Ice. kók were reanalysed as feminine simultaneously, we might expect feminine forms to occur at proportionately representative rates compared with neuter doublets. However, this is not the case. Paper II therefore relies on frequency counts from text corpora in order to shed light on the nature of reanalysis as a function of distinct analogical means, both of which presume influence from respective forms of the word (see PII:206). Paper II also seeks to demonstrate that innovative (non-humorous) plural blækur and (humorous) plural kækur are deduced from singular blók and kók, respectively, by identical analogical means. This endeavour is a reaction to the belief expressed by some that plural kækur is a joke and, therefore, not an example of “real” language use, while plural blækur is considered to reflect “real” usage (see an example of this view in PII:195–196, footnote 1). In 41 This relation between synchrony and diachrony exemplifies Bybee’s view that domain-general cognitive mechanisms of language change give rise to “universal” tendencies through language use, although specific outcomes of the process are not universal as they are constrained by the respective grammars of individual languages (Bybee 2008:120–121). 53 Jón Símon Markússon order to demonstrate that the cognitive mechanisms for deduction of the relevant forms are not substantively different, Paper II demonstrates that analogy feeds on both semantics and pragmatics to extend knowledge of contextually appropriate language use to imagined, inappropriate contexts. The result of extension to this new context can seem so absurd as to imbue to the use of language in it humorous. Paper II concludes that, no matter the motivation for extension, language use in new contexts is the result of analogy and that this is the cognitive means by which all innovative forms of Ice. blók and kók are deduced. Therefore, if use of plural blækur represents “real” language use, so does use of e.g. plural kækur. 4.3 The methodology employed in Paper III Evidence from language change suggests time and again that the basic forms of paradigms are also those with the highest token frequency –– all other things being equal. In this connection, Paper III utilises frequency counts from an electronic text corpus for Modern Faroese, Teldutøka tekstasavn Føroyamálsdeildarinnar (TTF),42 to demonstrate that this property of usage facilitates deduction of innovative dat.pl. firðum, which is based on by far the most frequent form of Far. fjørður, i.e. dat.sg. firði. However, while original dat.sg. firði has likely always been the most frequent form (see below), innovative forms containing the variant fjørð-, cf. innovative dat.sg. fjørði, are based on far less frequent forms, such as nom.sg. fjørður, acc.sg. fjørð and/or original dat.pl. fjørðum. At first glance, the point just made might appear to contradict arguments for the impact of the causal relation frequency > entrenchment > lexical strength as a determinant of the direction of levelling. In other words: Why should a stem variant found in a far less frequent form of the paradigm be selected at the cost of a variant found in by far the most frequent form, given that the latter has the greatest lexical strength? In answer to this seeming contradiction, Paper III argues for the impact of two dimensions of frequency on the direction of levelling. First, token frequency is demonstrably deterministic of lexical strength and serves as the catalyst of levelling, given a specific context of use. Crucially, token frequency is blind to intra-paradigmatic dynamics between forms with regard to the relation between meaning and form (see below). Secondly, in the wake of change predicated on token frequency, it is argued that language users relied on the dispersion of stem variants across the 42 See the bibliography for the relevant URL. 54 Methodology singular and plural portions of the paradigm in order to establish a relation between either value and a respective stem variant. In context-specific terms, Paper III argues that the stem variant firð- is basic to the paradigm, due to its high token frequency and the impact of this factor on lexical strength. Further, it is hypothesised that the same form has likely always been basic to Far. fjørður, as its number of referents and their individual meanings have not changed during the centuries since settlement of the Faroe Islands. Therefore, it is concluded that different rates of usage for individual forms of the word have always correlated proportionately with those reflected by corpus data for the modern language, as usage-needs have not changed substantially with regard to the number of definitions. Subsequently, extension of the variant fjørð- was not a function of token frequency, but rather constituted a reactionary attempt to establish opposing relations between meaning and form across the semantic divide singular vs. plural. In other words, the plural was represented by firð-, cf. nom./acc.pl. firðir, innovative dat.pl. firðum (beside original fjørðum) and singular by the variant fjørð-, cf. nom.sg. fjørður, acc.sg. fjørð, innovative dat.sg. fjørði (beside original firði). It is the often-observed course of levelling, which time and again proceeds from the most frequent member, that permits positing this chronology. Thus, frequency counts based on text corpora guide the application of usage-based theory to the available data, rather than the other way around.43 In the absence of written sources for different stages of Faroese language history (see discussion in Chapter 2), Paper III also employs a method of determining historical token frequencies of different forms of the noun vøllur. The reason for this endeavour is that, unlike fjørður, Far. vøllur has innovated with regard to its number of referents since the time of settlement. Therefore, given the usage-based cognitive assumption that frequency is a determinant of the direction of levelling, it proved necessary to ascertain which form(s) most likely served as basic before the process commenced. To this end, Paper III applies a comparison of frequency data obtained from TTF and Íslenskt textasafn (ÍT),44 an electronic text corpus for Icelandic, spanning all periods. The utility of ÍT in estimating historical frequencies for earlier Faroese is obviated by the discussion of Old West Nordic in Chapter 2. In other words, Icelandic and Faroese were 43 Paper III also relies on frequency counts from the same corpus to determine the direction of levelling taken in the paradigm of Far. vøllur. The methodology for achieving that endeavour is treated specifically in section 4.3.1. 44 See references for electronic access to the corpus. 55 Jón Símon Markússon almost indistinguishable in the centuries immediately subsequent to settlement, so use of the search function Fornrit ‘ancient texts’ likely gives an indication of usage patterns for search words in Old West Nordic. Crucially, however, the usage-based cognitive approach acknowledges that language users’ perception of the real-world conditions that prevail in their environment impact usage. Thus, due to both weather and topographical traits of the Faroe Islands, pre-20th century referents of Far. vøllur –– before its use in the sense ‘(sports) pitch’ increased and the sense ‘airport’ came into use –– tended to be small and, therefore, not to perform established functions. Conversely, referents of the Icelandic cognate völlur are typically larger, likelier to perform established functions and, therefore, occur more often. Based on this comparison, the estimation of historical token frequencies for individual forms of Far. vøllur asserted that the word occurred less frequently in the sense ‘(sports) pitch’ in Faroese than in Icelandic. However, I decided to be generous with the estimate and assumed half the token frequency of the Icelandic equivalent for pre-20th century Faroese. Additionally, the number of instances in which Far. vøllur occurred in the sense ‘airport’ according to TTF were subtracted for earlier phases of the language. Use of the word in the senses ‘field, grassy ledge on a rock face’ was considered to proportionally represent historical token frequency, given that the number of referents in these senses have likely changed little over the centuries. On this basis, it is estimated that acc.sg. vøll was the most frequent form of the word in pre-20th century Faroese (see PIII:68, Table 4). Given the relation frequency > entrenchment > lexical strength, Paper III assumes that this conclusion will have had consequences for the choice of basic form of Far. vøllur and determined the direction of levelling within the paradigm. 56 5 Icelandic data: Papers I and II This chapter details the respective studies presented in Papers I and II, both of which account for the productivity of the Xó/æT-microclass –– albeit to varying degrees of focus. Due to the common thematic content, both studies can mostly be delineated within the same sections but are also treated separately where necessary. Section 5.1 fleshes out the contents of Papers I and II in light of theoretical focus. In Section 5.2, I delineate the prototype structure that centres on the feminine Xó/æT-microclass. Section 5.3 fleshes out the analysis presented in Paper I, accounting for the linguistic and non-linguistic factors that facilitate the reanalysis of masculine forms in plural -ur as feminine at different rates of frequency. To this end, subsection 5.3.1 demonstrates that frequency of use impacts categorisation via the domain-general cognitive process of statistical learning, the theoretical basis for which was discussed in 3.2.2. In subsection 5.3.2, I elaborate on cognitive economy as a property of functionally related linguistic categories. Subsection 5.3.3 presents instances of actual language use as evidence for hierarchical relations between varyingly schematic representations of linguistic categories. In subsection 5.3.4, I discuss the effects of minimal schematicity as a facilitator of limited productivity in light of the arguments forwarded in the preceding subsections. Subsection 5.3.5 models the extent to which varying degrees of schematicity facilitate different rates of productivity for functionally equivalent schemas represented at different levels of abstraction, where the notational conventions employed characterise what I call the ‘net effect’. Subsection 5.3.6 offers a summary of Section 5.3. Section 5.4 fleshes out the analysis presented in Paper II. Subsection 5.4.1 presents the content of Paper II. In subsection 5.4.2, I argue for the deduction of plural kækur and blækur from singular forms of the respective paradigms as part a two-step process. Subsection 5.4.3 details the different analogical means by which individual forms of a paradigm can be deduced, i.e. by extension and levelling. Subsection 5.4.4 argues that plural (non-humorous) blækur and (humorous) kækur are deduced by extension, negating the position that one is a better example of “real” language use than the other. Section 5.4.5 provides a summary of Section 5.4. Section 5.5 offers a summary of the current chapter. 57 Jón Símon Markússon 5.1 Content and focus of Papers I and II Paper I accounts for the different rates at which Icelandic masculine forms in plural -ur are reanalysed as grammatically feminine on the basis of schematicity. Specifically, it establishes a correlation between the rate of reanalysis, the dispersion of plural -ur, and the degree of formal similarity that individual masculine forms exhibit to a phonetically coherent prototype in the same ending. The prototype is manifested by the individual members of the Xó/æTmicroclass (following Dressler 2003), i.e. the feminine nouns Ice. blók ‘non-entity, wretch’, bók ‘book’, bót ‘patch, remedy’, brók ‘trousers’, nót ‘(fishing) net’. rót ‘root’. While the Xó/æT-microclass is considered minimally schematic, the masculine forms discussed in Paper I are as phonetically dissimilar beyond plural -ur as e.g. masc. plural bræður, eigendur, fingur, fætur, vetur. The occurrence of reanalysis can be discerned when masculine forms in plural -ur acquire doublets containing the definite article fem. pl. -nar. Examples are bræðurnar, eigendurnar, fingurnar, fæturnar, veturnar, which exist alongside (original/standard) masc. nom.pl.def. bræðurnir ~ acc.pl.def. bræðurna, eigendurnir/-na, fingurnir/-na, fæturnir/-na, veturnir/-na. Crucially, the feminine doublets occur in the minority of cases (see PI:10–11). As explained in Chapter 4, in order to determine the rate of reanalysis, Paper I adds the token frequencies of the masculine and the feminine forms together before the number of feminines is calculated as a percentage of that total. Paper I argues that reanalysis occurs –– all other things being equal –– due to the dispersion of plural -ur: despite the ending occurring in 14.92% of noun paradigms, 91.89% of these are feminine. However, the analysis demonstrates that all things are not equal, highlighting different rates of reanalysis as a function of the degree to which a masculine noun in plural -ur aligns with the functionally equivalent schema for the Xó/æT-microclass. In other words: the higher the number of one-to-one formal and functional correspondences between forms, the greater the likelihood that a masculine form in plural -ur will be reanalysed as feminine on alignment with the minimally schematic schema [XæTur]nom./acc.pl.. Therefore, as discussed in 3.2.3, minimal schematicity correlates with a boost to –– albeit still limited –– productivity as a function of the gang effect. Paper II deals with the limited productivity of the Xó/æT-microclass specifically. Despite low type frequency and low schematicity, the relevant schemas have been extended to three new contexts since the 16th century, i.e. innovative feminine pl.def. fæturnar; feminine dat.sg. kók, gen.sg. kókar, nom./acc.pl. kækur, cf. the borrowing kók ‘CokeTM’, which is mainly neuter in the mass noun sense but feminine in the count noun sense; and the 58 Icelandic data: Papers I and II entire paradigm of borrowed feminine blók. The three paradigms are shown in (18), beside that of rót (for comparison, the nominative and accusative plural forms of the article are included for the relevant forms of feminine rót and masculine fótur). fem. fem. neut./fem.45 masc./fem. nom. rót blók kók fótur acc. rót blók kók fót dat. rót blók kóki/kók fæti gen. rótar blókar kóks/kókar fótar pl. nom.(def.) ræturnar blækur/blókir kók/kækur fæturnir/-nar acc.(def.) ræturnar blækur/blókir kók/kækur fæturna/-nar dat. rótum blókum kókum fótum gen. róta blóka kóka fóta (18) sg. Papers I and II both demonstrate that extension of the Xó/æT-microclass pattern to new contexts is a function of phonetic and semantic coherence between the sources of innovation and the respective feminine targets listed above, cf. plural bræðurnar, eigendurnar, fingurnar, fæturnar, veturnar (Paper I); plural kækur, blækur (Paper II). However, before writing of Paper II commenced, I had become aware through informal conversations with colleagues that some do not view plural kækur as an example of “real language use”. Conversely, the same did not apply to plural blækur. This position appears to be based on the fact that kækur is used in humorous contexts only (see PII:195, footnote 1). Therefore, Paper II seeks to demonstrate through joint appeal to semantics, pragmatics, and schematicity that deduction of plural kækur on the basis of sg. kók by no means belies “real language use”. Rather, it concludes that all linguistic innovation predicated on analogy is a property of language use. In order to account for the feminine innovations discussed, appeal is made in Papers I and II to the domain-general cognitive processes of statistical learning and categorisation, as 45 Although plural kók is almost always qualified by feminine forms, some examples of neuter modifiers can be found, cf. (neut.acc.) tvö kók ‘two CokesTM’. Further, the form pl. kók patterns with most other strong neuters in that it is fully syncretic with the nominative/accusative singular, i.e. kók, e.g. neut. nom./acc.sg., nom./acc.pl. borð ‘table(s)’, hús ‘house(s)’, epli ‘apple(s)’. This suggests that plural kók is phonetically based on the singular mass noun sense of the word, despite being used as an uninflected feminine form in most instances in the plural count noun sense (see PII:207). 59 Jón Símon Markússon guided by analogy (see 3.2.2 and 3.3). In Paper I, such appeal proves a means to demonstrate that a taxonomy of formally distinct, yet functionally continuous schemas has been elaborated via categorisation and entrenchment. Further, both studies argue that categorisation by the minimally abstract sister schemas [XóT]nom./acc./dat.sg. ~ [XæTur]nom./acc.pl. ~ [XæTurnar]nom./acc.pl.def. facilitates perception of one-to-one formal correspondences between inflectional forms, yielding parallel connectivity between functionally equivalent arguments on each side of a relation (see discussion and examples of parallel connectivity in Section 3.3). Papers I and II argue that this process results in alignment with relations represented by the sister schemas above, ultimately motivating assignment of the relevant forms to the Xó/æT-microclass specifically. Further, Papers I and II demonstrate that the greater the extent of one-to-one formal and functional correspondences between aligned forms, the greater the likelihood that an existing form be assigned feminine grammatical gender. In Paper I, the greater likelihood of reanalysis as feminine is reflected by the mismatched descending frequencies between functionally equivalent sets of masculine forms, on the one hand, and innovative, feminine doublets, on the other (see above; also PI:14, Table 1). Ultimately, a greater degree of one-toone correspondence accounts for both the limited productivity of the feminine subtype in plural -ur and the higher rate at which plural fætur is reanalysed as feminine compared with other masculines in the same ending. In Paper II, recourse is made to both extension and levelling in accounting for the deduction of feminine forms of borrowed blók and kók. Further, it is demonstrated that semantic relationships between specific forms within each paradigm, real world referents of each word, and use in context, i.e. pragmatics, motivate the analogical means of deduction. In terms of semantics, for example, Ice. blók can refer to people of any gender, a property of usage that likely facilitates application of feminine grammatical gender. 46 In terms of pragmatics, Ice. kók often occurs in a syntactic context where it is qualified by feminine modifiers, used to qualify an omitted feminine noun that names the container or vessel in which the drink is bought or served (see PII:202). It is argued that the distinct semantic and 46 In this connection, Icelandic does exhibit a tendency to align biological sex and grammatical gender. Consider examples such as bróðir ‘brother’ and móðir ‘mother’. The two words inflect identically in their indefinite forms, a remnant from the Proto-Indo-European gender system that distinguished animate nouns from inanimate (this is a gross simplification of the situation). However, despite following identical patterns of inflection, bróðir is grammatically masculine while móðir is grammatically feminine. For an elaboration on the link between biological sex and grammatical gender in Icelandic (see PII:214–215). 60 Icelandic data: Papers I and II pragmatic factors at play, in conjunction with schematicity, facilitate graded membership of the Xó/æT-microclass. In light of the above, the common themes of Papers I and II are the following: 1. Both papers detail innovation and change pertaining to the gender assignment of Icelandic nouns as a function of analogical reasoning. 2. Both papers demonstrate the influence of minimal schematicity as a facilitator of limited productivity. 3. In both papers, reanalysis as feminine is attributed to the high rate at which plural -ur is dispersed among classes of feminine nouns –– all other things being equal. 4. In both papers, semantic and pragmatic properties of the forms under analysis are shown to impact both the rate of alignment with the Xó/æT-microclass and, consequently, the rate of reanalysis as feminine. 5. In both papers, frequency of use is a gauge for the influence of minimal schematicity as a facilitator of productivity. The rest of the current chapter elaborates on the arguments forwarded, the theoretical approach applied, the methodology employed, the data presented, and the conclusions made in Papers I and II. The next section delineates the prototype structure of the feminine subtype in plural -ur as it centres around the Xó/æT-microclass. 5.2 Prototype structure that centres on the Xó/æT-microclass In subsection 3.2.4, I applied Rosch’s principle of cognitive economy to delineate a taxonomy of functionally related schemas that represent members of the category FURNITURE at different levels of abstraction, cf. the superordinate category label FURNITURE, basic CHAIR, and subordinate DINING CHAIR. In the current section, I delineate the prototype structure of the Icelandic feminine subtype in plural -ur according to the same principle, accounting for its convergence with the members of the Xó/æT-microclass at the subordinate level. At both the basic and the subordinate levels, the taxonomy is defined by varyingly schematic sets of feminines in plural -ur, represented by the schema [Xur]nom./acc.pl., cf. e.g. 61 Jón Símon Markússon plural bækur, geitur, stelpur, and [XæTur]nom./acc.pl., cf. e.g. bækur, nætur, rætur, respectively.47 Like the majority of Icelandic strong feminine nouns (see footnote 4), members of the subtype in plural -ur comprise a single syllable in the syncretic nominative/accusative/dative singular, which alternates in almost all cases with a disyllabic genitive singular in -ar and a syncretic, disyllabic nominative/accusative plural in -ur, e.g. nom./acc./dat.(/gen.)sg. kind(ar) ‘sheep’ ~ plural kindur, geit(ar) ‘goat’ ~ geitur, eik(ar) ‘oak’ ~ eikur.48 Further, these alternate intra-paradigmatically with a definite form in feminine -nar, cf. sg. kind(ar) ~ pl.def. kindurnar, geit(ar) ~ geiturnar, eik(ar/-ur) ~ eikurnar. Additionally, many paradigms exhibit vocalic alternation stemming from Proto-Nordic i-umlaut, by which stem vowels were fronted due to the influence of unstressed *i (which has since been lost from the following syllable): nom./acc./dat.sg. mörk ‘250 gr.’ ~ gen.sg. markar/merkur ~ plural merkur, where PNc. *a was original, sg. kló(ar) ‘claw’ ~ plural klær,49 rót(ar) ~ rætur, where PNc. *ō was original. The feminine subtype in plural -ur can be demarcated according to a continuum of prototypicality, determined in part by the morphophonological attributes of the paradigms to which the relevant schemas have been extended as graded similarity exhibits family resemblance structure (see 3.3; also 5.3.1). In this connection, Papers I and II detail the extension of the relevant schemas to the paradigm of masculine fótur, cf. fem.pl.def. fæturnar, and those of borrowed blók, cf. pl. blækur, and kók, cf. pl. kækur. Thus, it is argued that the nominative/accusative/dative singular forms of a prototypical subset of feminines within the subtype each comprises a single syllable with the diphthong ó [ou:] as its nucleus,50 and a coda in either t or k.51 Further, -ar is the prototypical ending of the genitive singular form, as is the case for the vast majority of Icelandic strong feminines. 47 In the schema [Xur]nom./acc.pl., ‘X’ stands for ‘a stem of any shape preceding the ending plural -ur’. In the schema [XæTur]nom./acc.pl., ‘X’ generalises over the stem-initial consonant (cluster) of members of the Xó/æTmicroclass, i.e. b, br, n, and r, cf. bækur, bætur, brækur, nætur, rætur, (possibly d, cf. dætur ‘daughters’, and m, cf. mæður; see below), before the addition of plural blækur and kækur. Further, the notation ‘T’ represents the consonants k and t, cf. bækur, bætur, brækur, nætur, rætur, blækur, and kækur (possibly also ð; see footnote 51). Thus, ‘XæT’ stands for ‘stems like bæk- and ræt- preceding the ending plural -ur’. 48 The variant gen.sg. eikur exists beside gen.sg. eikar. 49 The ending -r of plural klær is the result of a historical process of contraction from earlier *-ur. 50 The general rule for vowel length in Icelandic is that stressed vowels are long in open syllables, but short in closed syllables. Stress always falls on the initial syllable (see Árnason 2005:135). 51 After publication of Papers I and II, it was brough to my attention that the feminine noun glóð ‘ember’ alternates with plural glæður, as well as plural glóðir (I had only heard and only ever used the latter). Clearly, 62 Icelandic data: Papers I and II Moreover, as is the case for the subtype in plural -ur where applicable, i-umlaut alternation is applied in the inflection of those nouns that constitute the prototype, cf. nom./acc./dat./gen.sg., dat./gen.pl. -ó- [ou:] and nom./acc.pl. -æ- [ai:], e.g. bók- ~ bækur, rót~ rætur. This morphophonological attribute is also associated with the extension of the subtype’s schema(s) to new contexts, cf. nom./acc./dat./gen.sg., dat./gen.pl. blók- ~ plural blækur (see PI:3–5). Further, although the innovative feminine form pl.def. fæturnar does not involve extension of the alternation ó ~ æ –– the [ai:] of plural fætur is the result of historical fronting of PNc. *ō by i-umlaut –– alternation between ó and æ within the paradigm of masculine fótur is likely a motivating factor in the relatively high rate at which plural fætur is attracted by the Xó/æT-microclass. Given this morphophonological definition imposed on the prototype, feminine (borrowed) blók, (native) bók, bót, brók, nót, and rót, cf. sg. blók(ar) ~ nom./acc.pl. (standard) blækur, bók(ar) ~ bækur, bót(ar) ~ bætur, brók(ar) ~ brækur, nót(ar) ~ nætur, rót(ar) ~ rætur, are considered prototypical of the feminine subtype in plural -ur and a model for the analogical extension of its schemas. Further, compounds with Ice. -rót as their final constituent, such as engiferrót ‘ginger’, i.e. engifer-rót, kvaðratrót ‘square root’, gulrót ‘carrot’, piparrót ‘horseradish’, also exist. Despite the obvious association of their latter constituent with rót, such compounds are here considered distinct lexical items that boost the type frequency of the feminine subtype in plural -ur. This is due to the fact that the referent of each compound renders it lexically distinct, both from rót and each other. In other words, one cannot refer to a carrot or ginger as rót alone and expect language users to discern the referent. By the same token, compounds in -bók, such as kirkjubók ‘church book’, i.e. kirkjubók, dagbók ‘diary’, fundarbók ‘book of minutes’, do not boost the type frequency of the subtype: all are types of bók that can be referred to as such without further specification within the appropriate context. For this reason, however, it must be conceded that compounds in -bók contribute to the token frequency of Ice. bók. The same applies to compounds in plural -bætur, cf. atvinnuleysisbætur ‘unemployment benefit’, i.e. atvinnuleysis-bætur, húsnæðisbætur ‘rent rebate’, i.e. húsnæðis-bætur, which are kinds of financial aid; in -brók, alternation of the kind sg. glóð ~ pl. glæður accords with the pattern of alternation exhibited by e.g. bók ~ bækur, rót ~ rætur and, therefore, elaborates the relevant schemas to include one example of stem-final ð (potentially also plural mæður; see below). Elaboration of the schema to include stem-final ð presents no giant feat for generalisation and does not create problems for any of the claims made in Papers I and II. 63 Jón Símon Markússon cf. nábrók ‘necro-pants’, i.e. ná-brók, which are a kind of trousers; -blók, cf. e.g. skrifstofublók ‘pencil-pusher’, i.e. skrifstofu-blók, which is a kind of non-entity.52 A less central member of the subtype relative to the definition of its prototype is feminine nótt ‘night’. The vowel of the singular is short due to the nature of its coda (see footnote 50): Icelandic orthographic <tt> represents [ht],53 cf. nom./acc./dat.sg. nótt [nouht]. Further, the nótt paradigm also contains the form gen.sg. nætur, which is syncretic with the plural form, i.e. plural nætur (both of which are, in turn, syncretic with the plural of Ice. nót ‘(fishing) net’). This is a relatively rare iteration of syncretism. Despite these nonprototypical attributes, Ice. nótt is considered to straddle the cusp of prototypicality as a satellite, due to the nature of vocalic alternation between ó and æ manifested by its paradigm, in conjunction with the ending plural -ur. For the purpose of comparison, the full paradigms for the Xó/æT-microclass members rót and bók, as well as near-prototypical nótt, are shown in (19). (19) sg. nom. rót bók nótt acc. rót bók nótt dat. rót bók nótt gen. rótar bókar nætur pl. nom./acc. rætur bækur nætur dat. rótum bókum nóttum gen. róta bóka nótta Still more distant satellites are feminine geit and eik. Although the coda of each complies with the phonological definition of the prototype, the nucleus of both words consists of the diphthong [ei:]. This vowel does not engage in i-umlaut alternation, cf. sg. geit(ar) ~ nom./acc.pl. geitur, as opposed to e.g. prototypical bók(ar) ~ bækur. Further, the existence of the doublet form gen.sg. eikur, beside eikar, further distances the morphophonological attributes of eik from those of the prototype. The same can be said of nouns such as brík ‘armrest, bracket’, flík ‘item of clothing’, and tík ‘female dog, (derogative) bitch’, which have long had doublets in the genitive singular, cf. bríkur/bríkar, flíkur/flíkar, tíkur/tíkar (see 52 Conversely, as compounds in -blók, -bók, -bót, -brók, and -rót do not comprise a single syllable, they are not considered to boost the type frequency of the prototype. 53 See Árnason 2005:161, 206–207 on phonetic transcription conventions for Icelandic consonants. 64 Icelandic data: Papers I and II Guðmundsson 1922:69), as is also true of flík in the nominative/accusative plural, cf. flíkur/flíkar (see above about gen.sg., plural nætur).54 The feminine noun Ice. vík has gen.sg. ur only (see Jónsdóttir 2020:24 and sources cited there). Less peripheral on formal grounds are plural dætur (of dóttir ‘daughter’) and relatively more peripheral plural mæður (of móðir ‘mother’). While dætur meets the phonological definition of the Xó/æT-microclass in terms of its rhyme, mæður is considered more peripheral on account of the onset of its second syllable, i.e. inter-vocalic <ð> [ð] (though, see footnote 51). Further, it is likely that dóttir and móðir form a more consistent class with other words for familial relations, i.e. bróðir ‘brother’, faðir ‘father’, and systir ‘sister’, on both morphophonological and semantic grounds. In this connection, as discussed in Paper I (PI:15) and Paper II (PII:216), it is worthy of note that a strong semantic association with biological sex can impede the rate at which grammatical gender is projected onto an inflectional form that otherwise exhibits definitive phonetic coherence with the schema(s) for a given prototype (see below). As noted in 3.2.2, schemas are cognitive points of reference that abstract over the formal and functional attributes of a set of items perceived as similar. The process of schematisation takes its cue from factors such as common phonetic structure, semantics, grammatical gender, appropriate contexts of use etc. (following Bybee 2001:27). Therefore, in light of the fact that 91.89% of nouns in plural -ur are feminine, the schemas [Xur]nom./acc.pl. and [XæTur]nom./acc.pl. are necessarily imbued with cue validity for feminine grammatical gender. However, association of inflectional forms in plural -ur with human referents of male biological sex may hinder reanalysis even when the form in question aligns with the relevant schema (PI:15; PII:216). Significantly in this connection, despite the fact that the phonetic structure of masculine plural bræður ‘brothers’ is highly similar to that of plural mæður, the token frequency of innovative fem.pl.def. bræðurnar indicates that plural bræður is very rarely attracted by the schema [XæTur]nom./acc.pl.. Indeed, plural bræður and feður ‘fathers’, unlike e.g. plural fætur, are only rarely –– if at all –– reanalysed as feminine (PI:14). In Paper I, this dynamic is attributed to a strong semantic association of bróðir and faðir with male biological sex, i.e. 54 Knudsen (1967) posits Ice. brík, flík, tík, vík as a feminine microclass on phonetic grounds, e.g. due to the rhyme sequence -ík. Given that the microclass forms a phonetically well-defined subset in plural -ur, its schemas likely constitute another subordinate-level category within the broader feminine subtype. See Jónsdóttir (2020:27), who discusses the inflection of multisyllabic words (borrowings of Greek and Latin origin) in -ík, some of which get gen.sg. -ur and/or -ar. 65 Jón Símon Markússon they refer to men only. Therefore, in those instances where reanalysis of plural bræður and feður does occur, it is arguably semantic and/or formal association to feminine plural dætur, mæður and systur ‘sisters’ –– all of which arguably constitutes a semantic class –– that facilitates reanalysis as feminine, rather than alignment with the schema [XæTur]nom./acc.pl.55 By the same token, Paper II (PII:216) argues that assignment of feminine grammatical gender to the entire paradigm of Ice. blók is attributable to the fact that its use is not constrained by the biological sex and/nor gender identity of the referent. However, pragmatics plays a role in the assignment of feminine grammatical gender to forms of Ice. kók. Thus, the noun is almost always qualified by the feminine forms of modifiers in the count noun sense, i.e. in reference to a portion of the liquid delimited in contexts such as ‘a bottle/can of coke’. Importantly, Icelandic flaska ‘bottle’ and dós ‘can’ are both grammatically feminine. Thus, it can be argued that in the cases of both blók and kók, formal alignment with schemas for the Xó/æT-microclass is further assisted by semantics and pragmatics, i.e. the very factors argued to hinder reanalysis of plural bræður and feður as grammatically feminine. In light of the above, phonetic coherence with a schema can be considered the main facilitating factor in both alignment and reanalysis of an inflectional form –– all things being equal. However, it is clear that semantic and pragmatic factors also play a facilitating role. A combination of these factors likely accounts for the fact that plural fætur is reanalysed as grammatically feminine at a higher rate than any other masculine form in plural -ur, albeit in the minority of cases: plural fætur aligns perfectly with the prototype for the Xó/æTmicroclass in terms of phonetic coherence, cf. [XæTur]nom./acc.pl., while its referents are neutral with regard to biological sex and/or gender identity (see PI:14, Table 1). 5.3 Paper I: Reanalysis of masculine forms in plural -ur as feminine The current section fleshes out the analysis presented in Paper I, accounting for the linguistic and non-linguistic factors that facilitate the reanalysis of masculine forms in plural -ur as feminine at different rates of frequency. Subsection 5.3.1 demonstrates that frequency of use impacts categorisation via the domain-general cognitive process of statistical learning, discussed in 3.2.2. In subsection 5.3.2, I elaborate on cognitive economy as a property of functionally related linguistic categories, as discussed in the previous subsection. Subsection 55 Reanalysis of plural bændur ‘farmers’ as feminine at a higher rate than either plural bræður or feður is perhaps due to the fact that the word bóndi ‘farmer’ can –– perhaps more now than before –– apply to people of both sexes (see PI:15). 66 Icelandic data: Papers I and II 5.3.3 presents evidence from actual language use in support of hierarchical relations between varyingly schematic representations of linguistic categories. In subsection 5.3.4, I discuss the effects of minimal schematicity as a facilitator of limited productivity in light of arguments forwarded in the preceding subsections. Subsection 5.3.5 models the extent to which varying degrees of schematicity can facilitate different rates of productivity for respective schemas represented at distinct levels of abstraction. This model is characterised as the ‘net effect’. In subsection 5.3.6, I present a summary of the current section. 5.3.1 Categorisation as a function of statistical learning In Section 5.2, I delineated the prototype structure of the Icelandic feminine subtype in plural -ur, as it centres around the Xó/æT-microclass. The posited structure of the subtype, it was argued, accords with Rosch’s principle of cognitive economy in terms of the taxonomic arrangement of constituent schemas at distinct levels of abstraction. In the current section, I demonstrate that frequency of use impacts categorisation via the domain-general cognitive process of statistical learning, discussed in 3.2.2. Further, I demonstrate that via this process, we avail ourselves of tacit knowledge pertaining to the skewed frequencies of non-randomly co-occurring attributes which characterise linguistic and non-linguistic structures discerned in our environment (e.g. Taylor 2012:187). Moreover, due to the impact of co-occurring attributes on categorisation, statistical learning is considered to directly facilitate the function of categorisation, i.e. to reduce uncertainty in our environment. To demonstrate the influence of statistical learning on categorisation, let us consider an exceptionless and, therefore, prototypical attribute of feminine and neuter nouns in Icelandic, the schema for which abstracts over the rule of referral in (20). (20) The accusative plural form of all feminine and neuter nouns is the same as the nominative plural. It is important to note that the stipulation in (20) applies to literally thousands of paradigms in Icelandic, meaning that the relevant schema is entrenched at a high level of abstraction across all relevant instances of prior experience.56 However, given that Icelandic has three grammatical genders, mention of two of these in (20) implies that the stipulation is not generally applicable to masculines. Indeed, this is correct, cf. that alternation of the kind 56 In fact, the stipulation in (20) applies to the paradigms of all nominals, i.e. nouns, pronouns, adjectives, numerals, in the nominative/accusative plural feminine and neuter forms. 67 Jón Símon Markússon nom.pl. hestar ~ acc.pl. hesta ‘horses’ and gestir ~ gesti ‘guests’ is applied in the inflection of the majority of Icelandic masculine nouns (see the relevant discussion around the stipulations stated in (11) and (12) in 3.2.2). Despite this fact, adherence to the pattern of alternation nom.pl. -ar ~ acc.pl. -a or -ir ~ -i is not a prerequisite for assignment of masculine grammatical gender, as demonstrated by the relatively few masculine nouns in plural -ur, e.g. bændur ‘farmers’, eigendur ‘owners’, fætur. Crucially, 91.89% of nouns in plural -ur are feminine. Therefore, there is a degree of crossover between the stipulation in (20) –– which applies generally to feminines and neuters only –– and that which applies to the inflection of masculines and feminines in plural -ur, cf. the rule of referral in (21). (21) A nominative plural in -ur has the same form as the accusative plural. Note that (21) makes no reference to grammatical gender. However, as argued by Bybee (2001:27), schemas necessarily contain all information pertinent to the items over which they abstract. In other words, we can assume that the knowledge inherent to (21) is informed by statistical learning, the process by which knowledge of the dispersion of plural -ur and, therefore, its cue validity for assignment of feminine grammatical gender is amassed. Given the above, what answers can we provide to the question posed in (22), pertaining to the function that Rosch attributes to categorisation? (22) How does reference to the stipulations in (20) and (21) facilitate uncertainty reduction? To answer this question, let us first elaborate on this function in domain-general terms: [T]o treat [the] segmentation of the world as originally arbitrary … would be reasonable only if the world were entirely unstructured; that is … if the world formed a set of stimuli in which all possible stimulus attributes occurred with equal probability combined with all other possible attributes. (Rosch 1976:383) In other words, the world is not an unstructured entity. Rather, some phenomena are perceived as similar due to the non-randomly skewed frequencies with which distinctive sets of attributes co-occur. Indeed, it is similarity based on such skewed frequencies that facilitates assignment of items such as dining chairs, garden chairs, rocking chairs, armchairs, 68 Icelandic data: Papers I and II and beanbags to the basic level category CHAIR, rather than, say, DOG, despite the fact that members of both categories typically have four “legs”. Further, it is the varyingly consistent combination of co-occurring attributes such as four wooden legs, a horizontal seat on top of those legs, and, typically, a vertical or almost vertical back attached to the seat that facilitates elaboration of the schema for the relevant basic level category. Note that while rocking chairs, armchairs, and beanbags may not to have legs, they share other formal attributes with dining chairs and garden chairs. Further, all instantiations of the category CHAIR perform a continuous function –– they allow people to take the weight off their feet by sitting. The non-random frequencies with which these formal attributes cooccur to different degrees across functionally related items informs representation of the category at the basic level. Further, Rosch and others appear to view categorisation as a function of the domain-general cognitive process of analogy, via which entrenched knowledge of the world facilitates structure mapping and subsequent alignment between relationally similar and functionally parallel sets of attributes. In other words, the skewed frequencies at which non-randomly co-occurring attributes are integrated into separate phenomena determines the likelihood that a newly encountered item will be assigned to a given category. In turn, the degree to which the newly assigned item shares a random number of attributes with all existing members determines the status within the category, i.e. whether it is considered prototypical (think dining chair) or not (think bean bag) (see 3.3 on family resemblance structure within categories). As a function of this process, categorisation reduces uncertainty, both in terms of category content and in relation to category-external phenomena. In order to demonstrate that linguistic categorisation satisfies the domain-general function of analogical reasoning, it is necessary to show that linguistic categories are also established and expanded through analogy. To this endeavour, Taylor (2012:193–194) argues that language exhibits structural qualities attributed by Rosch (1978:29) to the world more generally: Language, too, does not present itself to us as ‘an unstructured set of equiprobable elements’, whether the elements be sound segments, syllable types, words, or constructions. The elements that we perceive in a language do not occur uniformly and they are not distributed randomly over the speech that we encounter. It is the ‘correlational structure’ of language which makes possible the emergence of linguistic categories, thus ensuring both the viability of the language as a system of communication as well as its learnability. (Taylor 2012:193–194) 69 Jón Símon Markússon Taylor’s argument is, therefore, valid to the extent that non-randomly skewed frequencies of co-occurring attributes have cue validity for assignment to linguistic categories also (see above). In this connection, it can be demonstrated that inflectional categories are characterised by non-randomly co-occurring sets of attributes, by which formal and functional relations between inflectional forms have cue validity for inflection class membership. In other words, as the vast majority of inflection classes in Icelandic contain nouns of a given gender, the nature of intra-paradigmatic relations between forms serves to reduce uncertainty with regard to gender assignment. Therefore, it also assists with agreement in the broader morphological and/or syntactic context, such as indicating the appropriate form of the definite article, or the morphologically appropriate marking of determiners. In domain-general terms, given that non-randomly co-occurring attributes determine perceived structure in our environment, the tracking of skewed frequencies can be viewed as a necessary function of statistical learning. Proceeding on the assumption that language change is subject to constraints imposed by domain-general cognition on language use, it is claimed that knowledge pertaining to the skewed frequencies of co-occurring attributes that distinguish inflectional patterns is entrenched via statistical learning (see Jost and Christiansen 2017). In light of this claim, it proves necessary to demonstrate that the properties of inflection classes noted above also have cue validity for appropriate morphological treatment, i.e. gender assignment (form), and syntactic treatment, i.e. agreement (function). Taylor (2012:187), based on Murphy (2002:215), argues that the cue validity of an inflectional form for class membership triggers the category validity of that form in terms of appropriate usage.57 In other words, assignment of a noun to a given inflection class based on similarity to other class members (cue validity) facilitates further inferences as to how forms of the noun should be treated both within and beyond the context of the paradigm in terms of gender agreement (category validity). The principles that govern categorisation in language are, therefore, the same as those that facilitate assignment of four-legged items with a seat and upright back piece to the category CHAIR, the schema for which also abstracts over functional attributes such as facilitating rest. 57 The term ‘category validity’ does not occur in Papers I–III. This is because both referees for Paper I advised that I collapse of distinction between cue validity and category validity for the sake of expediency. Therefore, I resolved to use the term ‘cue validity’ to convey both senses. 70 Icelandic data: Papers I and II Let us now impose the theme of this discussion onto inflectional categories in Icelandic, specifically those alluded to –– both explicitly and implicitly –– by the stipulations in (20) and (21). For some nouns, instantiation of the plural nominative and accusative is characterised by formal distinction of the kind nom.pl. -ar ~ acc.pl. -a or -ir ~ -i (see above; also PI:11). As formal distinction of this kind is characteristic of the inflection of the vast majority of Icelandic masculine nouns and, crucially, occurs among masculines only, it is considered a prototypical attribute of their inflection. This view is supported by the fact that forms which engage in this relation are hardly ever reanalysed as feminine. Thus, as a function of prototypicality, the non-randomly co-occurring cluster of attributes that defines relations of the kind -ar ~ acc.pl. -a or -ir ~ -i has high cue validity for the assignment of masculine grammatical gender. Conversely, Icelandic masculine forms in plural -ur, e.g. masculine fætur, are occasionally treated as feminine, as betrayed by their occurrence with the feminine form of the article, cf. pl.def. fæturnar, instead of masculine nom.pl.def. -nir, acc.pl.-def. -na (see PI:13–14 on different rates of reanalysis). Expressed in terms of cue and category validity, the skewed dispersion of plural -ur among masculine and feminine nouns, particularly in relative juxtaposition to the high frequency of the relations nom.pl. -ar or -ir and acc.pl. -a or -i among masculines only, means plural -ur has high cue validity for feminine grammatical gender.58 In turn, categorisation as feminine motivates use as such. In terms of prototype structure, despite the fact that forms such as plural fætur are used as masculine in the majority of instances, it can be claimed that they occupy a more peripheral space within the relevant category than do those in nom.pl. -ar or -ir and acc.pl. -a or -i. In other words, masculine forms in plural -ur are best characterised as satellites, occupying border space with the equivalent category for feminine forms, in which the ending plural -ur has more central status. Let us, then, attempt an answer to the question posed in (22), above. To this end, it is important to distinguish between language use as prescribed and language use as described. In other words, if we equate certainty with the “correct” use of language as prescribed by school grammars and the like, characterised as e.g. “The plural nominative/accusative for 58 Markússon (2023b) demonstrates that the same factors motivate use of Far. masculine plural føtur ‘feet’ as feminine. Further, he shows that the skewed dispersion of the Faroese plural ending -ar among feminine and masculine nouns –– overwhelmingly in favour of the masculines –– is responsible for the reanalysis of feminine nouns in this ending as masculine. This is taken to suggest that dispersion perhaps determines the rate of gender reanalysis in the broader context. 71 Jón Símon Markússon fætur is masculine and, therefore, should not occur with the feminine form of the article”, the answer to (22) is: The stipulations in (20) and (21) do not reduce uncertainty but rather facilitate the incorrect use of language. However, we get a different answer if we concede that uncertainty arises due to the exceptionless occurrence of syncretism among the literally thousands of Icelandic feminines, as well as properties of the dispersion and resultant peripheral position of plural -ur within the relevant category for masculines. In other words, those instances in which historically masculine forms in plural -ur occur with the feminine definite article, e.g. bændurnar, eigendurnar, fæturnar, likely reflect momentary uncertainty with regard to appropriate use on the very basis of the stipulations in (20) and (21). Therefore, outcomes in the feminine article -nar are a reasonable attempt at uncertainty reduction, i.e. the main function of categorisation, betraying reliance on the skewed dispersion of plural -ur among masculine and feminine classes. In light of this answer, the analysis presented in Paper I demonstrates that the prototype structure of categories is a function of statistical learning and, thus, supports a view of rich memory for language. 5.3.2 Elaborating on taxonomies of increasing schematicity In the current subsection, I elaborate on cognitive economy as a property of functionally related linguistic categories, as discussed in 5.3.1. Following e.g. Audring 2019, Paper I argues for and posits a taxonomy of formally distinct but functionally continuous schemas, characterised by a hierarchy of abstraction (see PI:12). Each level of the taxonomy is represented by its own schema, which abstracts over inflectional forms in plural -ur. The hierarchical nature of the taxonomy is defined by the number of one-to-one correspondences between a masculine form in plural -ur, such as plural fætur ‘feet’, vetur ‘winters’, fingur ‘finger’, eigendur ‘owners’, bræður ‘brothers’, and the functional attributes conveyed by a given schema. It is argued that the extent of alignment between form and schema determines the rate that a given masculine form in plural -ur is reanalysed as feminine –– all other things being equal. The standard paradigms for Ice. masculine fótur ‘foot’, vetur ‘winter’, fingur ‘finger’, and eigandi ‘owner’ are shown in (23), with the article for the nominative and accusative plural forms in bold. 72 Icelandic data: Papers I and II (23) sg. nom. fótur vetur fingur eigandi acc. fót vetur fingur eiganda dat. fæti vetri fingri eiganda gen. fótar vetrar fingurs eiganda pl. nom.(-def.) fæturnir veturnir fingurnir eigendurnir acc.(-def.) fæturna veturna fingurna eigendurna dat. fótum vetrum fingrum eigendum gen. fóta vetra fingra eigenda As highlighted in 5.1, masculine plural vetur, fingur, and fætur serve as the respective intraparadigmatic sources for fem.pl.def. veturnar, fingurnar, and fæturnar. The overtly feminine definite forms, it is argued, occur when the former set is reanalysed as feminine, i.e. subsequent to alignment with the basic level schema [Xur]nom./acc.pl., which has a 91.89% dispersion rate among feminine nouns. In Paper I, it is argued that language users occasionally analyse the morphological composition of inflectional forms such as plural fætur, on the one hand, and plural vetur, fingur, bræður, on the other, in a way contrary to etymological regularity. The general view is that the sequence -(u)r in vetur and fingur actually belongs to the stem (see below).59 In this connection, the vast majority of strong masculine nouns in nom.sg. -ur have no ending in the accusative singular, cf. nom.sg. fótur ~ acc.sg. fót, also e.g. masc. hestur ~ hest, gestur ~ gest. This morphological attribute is taken to indicate that -ur is an inflectional ending in such paradigms (Kvaran 2005:242, 304; see PI:10–11). Conversely, the sequence -ur is present in the syncretic nominative/accusative singular of e.g. vetur and fingur, as well as in all other forms of these words. This fact lends itself –– quite rightly in my view –– to interpretation of -ur in e.g. nom./acc.sg., plural vetur, fingur, and -r- in e.g. dat.sg. vetri, fingri as part of the stem both diachronically and synchronically. However, below it is argued that the sequence -ur in e.g. plural vetur, fingur, bræður is occasionally analysed as an inflectional ending that is both formally and functionally identical to the same sequence in plural fætur (and e.g. eigendur). Given the structural and 59 The sequence -ur in e.g. vetur, fingur, bræður alternates with -r- in forms whose ending is or begins with a vowel, cf. nom./acc.sg. vetur but gen.pl. vetra. For convenience, this element will henceforth be referred to as -ur. 73 Jón Símon Markússon etymological position endorsed above, what, then, might be the motivation for occasionally attributing the status of ending to the sequence in plural vetur, fingur, bræður? It is argued that an all-or-nothing structural approach proves unhelpful in accounting for the fact that masculine forms in plural -ur –– be that sequence an inflectional ending or part of the stem etymologically –– are occasionally reanalysed as feminine, while other forms of the paradigm are not. In light of this, any claim that analysis of plural -ur as an inflectional ending lacks intra-paradigmatic support can be countered through reference to the reanalysis of e.g. both vetur and fætur as feminine, which occurs despite intra-paradigmatic alternation with forms that are only ever used as masculine. Thus, we get the relations masc.nom./acc.sg.def. veturinn ~ ?fem.pl. vetur ~ fem.pl.def. veturnar; masc.nom.sg. fótur ~ masc.dat.sg. fæti ~ ?fem.pl. fætur ~ fem.pl.def. fæturnar. As noted in Paper I, the rarity or non-occurrence of feminine singular forms of fótur likely stems from the status of nom.sg. ur as an indicator of masculine grammatical gender due to properties of its dispersion. Why, then, do e.g. plural vetur and fætur serve as intra-paradigmatic sources of fem.pl.def. veturnar, fæturnar when the former pair occupies their respective paradigms with masculineonly forms? Paper I argues that this is due to formal identity between the historically stem-final sequence -ur in plural vetur (also e.g. plural fingur and bræður) and the inflectional ending plural -ur (PI:10–11). As already noted, plural -ur has strong cue validity for assignment of feminine grammatical gender on account of its dispersion. Further, as also argued in Paper I, the semantically significant difference between singular and plural in nouns is reflected in the reanalysis of e.g. plural vetur as feminine, despite the fact that the same process never affects singular forms of the word. In other words, reanalysis occurs as a two-step process that affects individual forms rather than all members of the paradigm simultaneously (see Chapter 4). Thus, for the reasons just stated, it is concluded that a lack of support from other members of the paradigm does not suffice to prevent reanalysis of e.g. plural vetur as feminine on account of the ”ending” plural -ur. While 8.11% of masculine nouns shows the historically regular inflectional ending plural -ur, most of these belong to the class of nd-stems, all of which are masculine (see Iversen 1972:66; also PI:12–15). The class in question conforms to the intra-paradigmatic relation [XAndi]nom.sg. ~ [XAnda]acc./dat./gen.pl. ~ [XEndur]nom./acc.pl., e.g. nom.sg. eigandi ~ 74 Icelandic data: Papers I and II acc./dat./gen.sg eiganda ~ pl. eigendur.60 It is argued that the very specific association of the pattern sg. -and- ~ pl. -end- with the inflection of masculine nouns only is likely responsible for the relatively low frequency with which words containing it are reanalysed as feminine, all below 5.5% of instances (see PI:14, Table 1). Indeed, the schema [Xur]nom./acc.pl. should otherwise be schematic enough to attract masculine forms such as plural eigendur, which shares the suffix pl. -end-, as well as plural -ur, with a microclass of feminines, cf. e.g. sg. strönd ~ pl. strendur ‘coastline, beach(es)’ (following Booij 2010:41; Brown and Hippisley 2012:34). As already intimated, actual usage occasionally betrays reanalysis of the nominative/accusative plural forms of nd-stems as feminine, cf. fem.pl.def. eigendurnar ‘the owners’ at a rate of 2.83%, bændurnar ‘the farmers’ at 2.87%, nemendurnar ‘the students’ at 1.62%, instead of original masc.nom.pl.def. eigendurnir, bændurnir, nemendurnir and acc.pl. eigendurna, bændurna, nemendurna. However, as reported in Paper I, the rate at which ndstem sources in plural -ur are reanalysed as feminine is much lower than for plural fætur, fingur, and vetur (PI:14). Further, this lower rate of reanalysis can likely also be attributed in part to the fact that nd-stems have human referents and, therefore, often refer to men. In light of the schematic and semantic factors considered to impact the rate of reanalysis for nd-forms in plural -ur, Paper I focuses particularly on the facts pertaining to the different rates of reanalysis reported for plural fætur at 22.82%, fingur at 15.12%, and vetur at 9.74% of instances, respectively (see PI:14, Table 1). The descending token frequencies of fem.pl.def. fæturnar, veturnar and fingurnar betray a mismatch in the rate of reanalysis relative to the individual token frequencies of original/standard masc.nom.pl.def. fæturnir, veturnir, fingurnir and masc.acc.pl.def. fæturna, fingurna, veturna. Appeal is made to the varying degrees of schematicity that characterise the taxonomy referred to above. In order to provide a convincing account of the effects of schematicity on the different rates of reanalysis reported, a more detailed word on the dispersion of plural -ur among masculine and feminine classes is in order. As discussed in 3.2.2 and 5.3.1, the endings for the plural nominative and accusative forms of the vast majority of Icelandic masculine nouns can be represented schematically as [XV1r]nom.pl. and [XV1]acc.pl., respectively. In the context of intra-paradigmatic alternation, these schemas constitute the sister schema [XV1r]nom.pl. ~ [XV1]acc.pl., cf. nom.pl. -ar ~ acc.pl. 60 Here, ‘X’ refers to the root sequence of such words, e.g. eig- in eig-and-i, nem- in nem-and-i ‘student’. The notation [...A...]sg. ~ [...E...] generalises over the fact that, while the vast majority of nd-stems inflects like eigandi, cf. suffixes in sg. -a-, pl. -e-, a small number does not. Thus, Ice. bóndi engages in alternation between sg. -ó- and pl. -æ-, cf. nom.sg. bóndi ~ acc./dat./gen.sg. bónda ~ nom./acc.pl. bændur. 75 Jón Símon Markússon -a, e.g. hestar ~ hesta; nom.pl. -ir ~ acc.pl. -i, e.g. gestir ~ gesti. The dispersion of the sister schema is associated with masculine classes only and is of medium type frequency in Icelandic (based on statistical data collated by Svavarsdóttir 1993). However, the functionally equivalent forms of some masculine nouns, e.g. plural fætur, eigendur, demonstrate that conformity to this relation is not a prerequisite for categorisation and use as masculine. Indeed, masculine forms in plural -ur most often alternate with overtly masculine plural definite forms, cf. masc.nom.pl.def. fæturnir, eigendurnir and masc.acc.pl.def. fæturna, eigendurna, i.e. these are more common than their functionally equivalent feminine doublets. Conversely, syncretism in nominative/accusative plural is an exceptionless formal attribute of feminine (and neuter) nouns, e.g. feminine plural myndir ‘pictures, forms’, greinar ‘branches, articles’, stelpur ‘girls’, mýs ‘mice’. In (24), the paradigms of feminine mynd, grein, stelpa and rót are given alongside the plural forms of masculine fótur (the endings of the nominative/accusative plural are in bold). (24) sg. nom. mynd grein stelpa rót acc. mynd grein stelpu rót dat. mynd grein stelpu rót gen. myndar greinar stelpu rótar pl. nom./acc. myndir greinar stelpur rætur fætur dat. myndum greinum stelpum rótum fótum gen. mynda greina stelpna róta fóta As noted in Paper I (PI:11), entrenched knowledge of this pervasive and characteristic formal distinction between masculine and feminine classes is necessarily stored at a high level of abstraction (see e.g. Janda 2002; 2007; also below). However, at the physical level of language use, the distinction is instantiated as in the masculine forms in (25a), on the one hand, and the feminine forms in (25b), on the other. The relevant plural forms of masculine fótur, vetur and fingur, which straddle the masculine-feminine border, are given in (25c). (25) a. [XV1r]nom.pl. ~ [XV1]acc.pl. (masc.) hestar ~ hesta, gestir ~ gesti b. [XVr]nom./acc.pl. (fem.) myndir, greinar, stelpur, rætur c. [XVr]nom./acc.pl. (masc./?fem.) fætur, vetur, fingur (Lifted from PI:12) 76 Icelandic data: Papers I and II Given the view that schemas are not elaborated independently of the formal, functional, and interactional attributes over which they abstract (e.g. Bybee 2001:27; Lakoff 2018:86–87), the notation ‘-V1-’ should be taken to imply the phonologically arbitrary subset {a, i} only. In other words, statistical learning yields knowledge that only the relations nom.pl. -ar ~ acc.pl. -a and -ir ~ -i can instantiate the sister schema [XV1r]nom.pl. ~ [XV1]acc.pl.. Conversely, highly schematic [XVr]nom./acc.pl. in (25b), whose dispersion is overwhelmingly associated with feminine classes, offers a tried and tested point of reference for successful deduction of forms in nom.pl. -ir, -ur and -ar from acc.pl. -ir, -ur, and -ar and vice versa. Therefore, this property of the schema’s dispersion greatly reduces the likelihood that it should project masculine gender onto the inflectional forms referred to it. On the contrary, indeed, Paper I argues that we should expect alignment of e.g. plural fætur with the schema [XVr]nom./acc.pl. to facilitate reanalysis as feminine. With this in mind, let us now consider the schemas to which forms in plural -ur might be referred for the purpose of categorisation. The schema [XæTur]nom./acc.pl. is minimally schematic, i.e. it renders highly specific formal and functional constraints imposed on the set of forms over which it abstracts. For this reason, it is considered subordinate to medially schematic [Xur]nom./acc.pl.. In Audring’s (2019) terms, the former is a daughter of the latter. In turn, the schema [Xur]nom./acc.pl. is posited as basic, i.e. the daughter of highly schematic [XVr]nom./acc.pl.. This view is congruent with arguments expressed by e.g. Audring (2019), Barðdal (2008), Booij (2010) and Bybee (2001:8), and similar to Albright’s (e.g. 2002; 2008; 2009) conception of form-to-form mapping “rules” of varying specificity (also Albright and Hayes 2003). Thus, the hierarchical subordination of basic [Xur]nom./acc.pl. and subordinate [XæTur]nom./acc.pl. to superordinate [XVr]nom./acc.pl. is rendered as the taxonomy in (26). (26) a. Superordinate [XVr]nom./acc.pl. myndir, greinar, stelpur, rætur, fætur b. Basic [Xur]nom./acc.pl. stelpur, vetur, rætur, fætur c. Subordinate [XæTur]nom./acc.pl. rætur, fætur (Based on PI:12) Regarding the representation of inflectional endings in (26), the notation [-Vr]nom./acc.pl. in (26a) abstracts over syncretism in plural -ar, -ir and -ur, thus encompassing the ending represented in both [Xur]nom./acc.pl. and [XæTur]nom./acc.pl.. 77 Jón Símon Markússon Thus, [Xur]nom./acc.pl. is neither schematic enough to abstract over the corresponding forms of all feminine nouns, cf. e.g. plural myndir, greinar, nor is it too specific as to only facilitate assignment of a noun to the Xó/æT-microclass. In this sense, it occupies “[t]he level used for everyday neutral reference” (Croft and Cruse 2004:83; see discussion in 3.2.4). Its status as a basic level category accounts for its function as a single point of reference for categorisation of non-feminine forms in plural -ur that are otherwise formally dissimilar (see the examples in (26b)). By the same token, [XVr]nom./acc.pl. is considered superordinate to [Xur]nom./acc.pl., as the former has “…fewer defining attributes than…” the latter (Croft and Cruse 2004:84). Therefore, [XæTur]nom./acc.pl. reflects a subordinate level category on account of the fact that “…members have high mutual resemblance…” (Croft and Cruse 2004:85). The hierarchical nature of the posited linguistic taxonomy is, therefore, congruent with Rosch’s (1975) principle of cognitive economy. 5.3.3 Language use and hierarchies in linguistic categories Positing the taxonomy in (26) is further justified by the fact that use of historically masculine plural fætur has shown development that is both convergent with and divergent from that of e.g. masculine plural vetur and fingur. In Paper I, it is argued that all three are reanalysed as feminine on account of syncretism in plural -ur –– all other things being equal. In the current subsection, I focus on the different rates of reanalysis for plural fætur, fingur, and vetur, as the rates for nd-stems are largely insignificant –– at 5.47% or less –– likely due to association of nd-stem inflection with male referents (see discussion in 5.3.2). In light of this view, it is argued that the respective rates of reanalysis for plural fætur, fingur, and vetur speak specifically to the influence of varying degrees of schematicity, when all other things are equal. As noted in Chapter 4, acc.pl.def. veturna is the most frequent of the relevant masculine forms, returning 6,913 results from the IsTenTen corpus. Next most frequent is masc.acc.pl.def. fæturna, with 2,882 results. Finally, a search for masc.nom.pl.def. veturnir returned 245 results, making it the least-frequent of the relevant masculine forms. Of their feminine doublets, a search of the same corpus returned more results for pl.def. fæturnar than for pl.def. veturnar and fingurnar combined, with 1,274, 773, and 285 results, respectively (see PI:14, Table 1). As argued in Paper I (PI:14–16), the higher rate of reanalysis for plural fingur as feminine relative to vetur can be accounted for by a combined appeal to two factors. First, to the cue validity that the schema [Xur]nom./acc.pl. has for feminine grammatical gender. 78 Icelandic data: Papers I and II Secondly, to semantic association of fingur with feminine forms such as plural hendur ‘hands’, (plurale tantum) herðar ‘shoulders’, even tær ‘toes’, and lappir ‘legs’, cf. pl.def. hendurnar, herðarnar, tærnar, and lappirnar. In this connection, it must be noted that semantic association with these same feminine forms likely also accounts in part for the reanalysis of plural fætur as feminine. However, Paper I takes the much higher rate of reanalysis for plural fætur compared with fingur to suggest unequal semantic links with the feminine nouns for body parts listed above, meaning that all things might not be equal. Note that a common synonym of Ice. fótur, i.e. Ice. löpp, is feminine, while that of fingur, cf. Ice. putti, is masculine. Returning to the more general picture, the statistical information presented above demonstrates that the mismatch in respective rates of reanalysis for masculine plural fætur and vetur as feminine is significantly greater than those for fingur, on the one hand, and either fætur or vetur, on the other. For this reason, reflecting the path taken in Paper I, the analysis presented in the next subsection will focus mainly on the mismatched descending token frequencies of masculine and feminine doublets containing plural fætur and vetur. Despite this focus, reference will be made to other masculine forms in plural -ur where relevant. 5.3.4 Productivity: Types and degrees of similarity This subsection elaborates on minimal schematicity as a facilitator of limited productivity. Further, I discuss the impact of degrees of similarity on the rate that masculine forms in plural -ur undergo reanalysis as feminine. In this connection, consider the question posed in (27), repeated from Paper I (PI:17). (27) How do we account for the frequency relation between fem.nom./acc.pl.def. fæturnar and veturnar, on the one hand, and masc.nom.pl.def. fæturnir, acc.pl.def. fæturna and veturnir, veturna, on the other? Answers to this question elucidate the impact of schematicity on the rate of reanalysis –– all other things being equal. With the question in (27) in mind, I endeavour below to characterise limited productivity in terms of categorisation, i.e. the domain-general cognitive process that attributes structure to phenomena in our environment. In this connection, deduction of the targets fem.pl.def. fæturnar and veturnar are considered varyingly specific consequences of the gang effect, different functions of which are properties of the prototype structure 79 Jón Símon Markússon delineated in 5.2 for the feminine subtype in plural -ur. The view taken in Paper I is that an appeal to alignment, reanalysis, and extension, as delineated by the proportional equations employed for exposition in historical linguistics, should not be allowed to speak for itself by rendering the outcome of analogical change only. Rather, Paper I seeks to render these underlying cognitive operations as the motivators and/or facilitators of change through innovative notation of proportions. To this end, an understanding of two different kinds of similarity –– each of which motivates analogy –– is key. The type of similarity that exists between e.g. plural rætur, kindur, fætur, and vetur can best be characterised as object similarity (see e.g. Kotovsky and Gentner 1996:2798; see also Gentner 2005, who uses the term “overall similarity”). It is due to object similarity between separate instances of the common formal attribute -ur, itself an expression of the function nominative/accusative plural, that language users might perceive the four forms above as similar. Thus, object similarity based on plural -ur alone facilitates alignment with the basic level schema [Xur]nom./acc.pl.. However, further points of similarity can be discerned between e.g. plural rætur and fætur, motivating alignment with subordinate [XæTur]nom./acc.pl., as well as the basic-level schema. With regard to the degree of productivity exhibited by an inflection class, it has been demonstrated repeatedly that this property is particularly correlated with type frequency and schematicity. Thus, as noted several times, the interaction of high type frequency with high schematicity can facilitate all but limitless productivity. In relative terms, the schema [Xur]nom./acc.pl. is associated with medium type frequency in Icelandic (based on frequency counts by Svavarsdóttir 1993), while its schematicity is also within the medium range (see the discussion surrounding (26) in 5.3.2). Therefore, given this often-observed correlation, we might expect to attribute a higher rate of productivity to medially schematic [Xur]nom./acc.pl. than to minimally schematic [XæTur]nom./acc.pl. –– all other things being equal. As argued in Paper I (PI:11–12), any inflectional form in plural -ur can be aligned with medially schematic [Xur]nom./acc.pl., while only those that show the relevant additional points of similarity can align with the minimally schematic [XæTur]nom./acc.pl.. Thus, it is argued, different rates of reanalysis for masculine forms in plural -ur as feminine are a function of the degree of formal and functional one-to-one correspondence between a schema and an inflectional form that has become distanced –– albeit only momentarily –– from other members of its paradigm due to formal ambiguity. Once the form has been aligned with the relevant schema on formal and functional grounds, projection of morphological content from the latter to the former commences. Thus, Paper I argues, subsequent to alignment, projection 80 Icelandic data: Papers I and II of feminine grammatical gender is the initial function of one-to-one correspondence, as depicted in (28) (see below on the notation employed). (28) [Xur]nom./acc.pl. nom./acc.pl. stelpur  nom./acc.pl. vetur The schema [Xur]nom./acc.pl. is included in (28) to represent the knowledge of form and function that is projected onto plural vetur (or any functionally equivalent masculine form in plural -ur that might be substituted for it). Projection of the morphological content, i.e. the attribute feminine grammatical gender, is represented by the arrow ‘’. Thus, (28) depicts the projection of functional content from feminine forms that align with the schema [Xur]nom./acc.pl. onto plural vetur. Subsequent to projection, once a form has been categorised as feminine, it is highly likely to have category validity for further alignment with its sister schema [Xurnar]nom./acc.pl.def. due to the high rate of dispersion that the schema [Xur]nom./acc.pl. has across feminine paradigms. Therefore, it is this property that subsequently activates the former schema [Xur]nom./acc.pl. as an output schema, i.e. one which has category validity for a given pattern of intra-paradigmatic alternation, in this case according to the sister schema [Xur]nom./acc.pl. ~ [Xurnar]nom./acc.pl.def.. Thus, reanalysis can be posited as the second function of one-to-one correspondence, while deduction of the target form is the third. Perception of object similarity across distinct intra-paradigmatic relations is likewise a prerequisite for the deduction of a target, i.e. the D-form of a proportional equation. However, it is the perception of similarity among relations that informs this process. For this reason, the second kind of similarity to be discussed is best characterised as relational similarity (e.g. Kotovsky and Gentner 1996:2798). Thus, alternation between e.g. fem.pl. stelpur, rætur, and fem.pl.def. stelpurnar, ræturnar, respectively, assists in establishing parallel one-to-one correspondences across relations that already align with the sister schema [Xur]nom./acc.pl. ~ [Xurnar]nom./acc.pl.def.. In turn, a form in plural -ur that is aligned with the output schema [Xur]nom./acc.pl. as a means of uncertainty reduction, i.e. the function of categorisation, subsequently serves as the source for alignment with overtly feminine [Xurnar]nom./acc.pl.-def.. Therefore, the entrenched knowledge represented by sister schemas 81 Jón Símon Markússon facilitates parallel connectivity, by which two relations and their arguments are placed in correspondence according to role. In other words, the relational knowledge that defines a sister schema is projected in order to extend an inflectional pattern by attracting a target via the category validity of the source. Further, it has been demonstrated that parallel connectivity is the prerequisite of structural alignment, i.e. successful alignment between relations, which is considered the hallmark of analogy (e.g. Gentner and Markman 1997; Kotovsky and Gentner 1996). An exemplary depiction of the process that yields structural alignment with the sister schema [Xur]nom./acc.pl. ~ [Xurnar]nom./acc.pl.def. is shown in (29). (29) [Xur]nom./acc.pl.  nom./acc.pl. stelpur [Xurnar]nom./acc.pl.def. ~ nom./acc.pl.def. stelpurnar  nom./acc.pl.def. X; X = veturnar  nom./acc.pl. vetur To delineate in terms of the proportional schema A : B :: C : D, (29) implies that object similarity between the A-form, i.e. stelpur, and the C-form, vetur, as predicated on the common ending plural -ur, facilitates alignment with the schema [Xur]nom./acc.pl.. Therefore, alignment on formal grounds assists in establishing one-to-one correspondence that pertains to both form and function. Subsequently, one-to-one correspondence facilitates projection of morphological content onto the C-form, (now feminine) plural vetur. Intra-paradigmatic alternation as instantiated by the A- and B-forms in (29), i.e. plural stelpur and pl.def. stelpurnar, respectively, activates [Xur]nom./acc.pl. with category validity for intra-paradigmatic alternation with its sister schema [Xurnar]nom./acc.pl.def.. This property of the former schema is represented by the symbol ‘’. Once the morphological content of [Xur]nom./acc.pl. has been projected onto the C-form, i.e. plural vetur, it is likewise imbued with category validity for intra-paradigmatic alternation of the kind [Xur]nom./acc.pl. ~ [Xurnar]nom./acc.pl.def.. Thus, the only logical solution to (29) is fem.pl.def. veturnar. Due to this function of analogical reasoning –– as assisted by the domain-general cognitive process of statistical learning –– functionally equivalent forms of Ice. fótur could easily be substituted for the respective forms of vetur in (29). In that case, the D-form of the relevant proportion would logically be solved as fem.pl.def. fæturnar, based on the C-form, (now feminine) plural fætur. In line with arguments forwarded in Paper I, the following 82 Icelandic data: Papers I and II subsections argue that masculine plural fætur aligns with the formal and functional attributes associated with both basic [Xur]Nom./acc.pl. and subordinate [XæTur]nom./acc.pl. as a means of uncertainty reduction, i.e. when momentarily distanced from other members of the paradigm during a usage event. Alignment at both levels of abstraction is depicted by the proportion in (30). (30) [Xur]nom./acc.pl.  [Xurnar]nom./acc.pl.def. [XæTur]nom./acc.pl.  [XæTurnar]nom./acc.pl.def. nom./acc.pl. stelpur ~ nom./acc.pl.def. stelpurnar nom./acc.pl. rætur ~ nom./acc.pl.def. ræturnar  nom./acc.pl.def. X; X = fæturnar  nom./acc.pl. fætur In (30), one-to-one correspondences based on the ending plural -ur are in bold, demonstrating that this is a formal and functional attribute common to the relevant forms at both levels of abstraction. However, instances of underlined æT convey one-to-one correspondence that is specific to the subordinate level only. Paper I argues that while e.g. plural vetur aligns perfectly at the basic level, it does so more tentatively than plural fætur at the subordinate level. Therefore, at this stage, the following answer to the question posed in (27) is proposed: plural vetur is more likely to escape alignment with subordinate [XæTur]nom./acc.pl. on schematic grounds. Further, though plural vetur were occasionally attracted at the subordinate level, it would at best be peripheral to the Xó/æT-microclass and not assigned to any specific feminine class. Conversely, plural fætur is not only reanalysed as feminine on alignment with [Xur]Nom./acc.pl. but is also specifically categorised as a Xó/æT-microclass form on alignment with subordinate [XæTur]nom./acc.pl.. Therefore, plural fætur is considered less likely to escape reanalysis as feminine than is plural vetur. The next subsection delineates the means employed in Paper I of modelling varying degrees of schematicity and the impact of this factor on the rate of reanalysis. 5.3.5 The ‘net effect’ Paper I employs the metaphor of the ‘net effect’ when modelling the extent to which varying degrees of schematicity facilitate different rates of productivity for respective schemas, each of which represents a distinct level of abstraction (PI:17–18). The net effect of minimal 83 Jón Símon Markússon schematicity is a network model interpretation of the gang effect, by which a network of connecting lines demonstrates the skewed frequencies of formal and/or functional attributes common to sets of forms. This approach to modelling productivity is based on Bybee’s Network Model (e.g. 1985, 2001, 2010). The innovative notations employed show that the greater the degree of one-to-one correspondence across forms, the greater the cue validity that the relevant cluster of attributes will have for assignment of grammatical gender and, potentially, for assignment of a form to a specific inflection class. In turn, the network approach demonstrates that once gender is assigned on this basis, category validity for use according to that assignment is activated. As noted in 5.3.4, Paper I accounts for the mismatch in descending token frequency between fem.pl.def. veturnar and fæturnar, on the one hand, and their respective masculine doublet forms, nom.pl.def. veturnir, acc.pl.def. veturna and fætur-nir, fætur-na, on the other. There, it is argued that alignment of plural fætur with respective schemas that represent distinct levels of abstraction facilitates the relatively general net effect of basic [Xur]nom./acc.pl., while alignment with subordinate [XæTur]nom./acc.pl. nets plural fætur and potentially some instances of peripheral plural vetur closer to the physical level of language use. Thus, as alluded to in 5.3.4, the net effect can be expressed as the rate at which an inflectional form escapes reanalysis relative to the rate at which others do not as a function of the gang effect. Subordinate [XæTur]nom./acc.pl. elaborates on the form of Xó/æT-microclass members proper and satellites (see 5.2), all of which are feminine. Therefore, the combination of its dispersion with its schematicity facilitates reanalysis of plural fætur as a Xó/æT-microclass form specifically. Conversely, equivalent properties as expressed in the schema [Xur]nom./acc.pl. enables it to attract and net any masculine form in plural -ur caught adrift in the sea of usage –– all other things being equal. However, as noted in 5.3.4, this does not lead to assignment and treatment of the relevant form according to a specific feminine class. Thus, plural vetur and fem.pl.def. veturnar might just as well be substituted for fætur and fæturnar in (30), as in (31). (31) [Xur]nom./acc.pl.  [Xurnar]nom./acc.pl.def. [XæTur]nom./acc.pl.  [XæTurnar]nom./acc.pl.def. nom./acc.pl. stelpur ~ nom./acc.pl.def. stelpurnar nom./acc.pl. rætur ~ nom./acc.pl.def. ræturnar  nom./acc.pl.def. X; X = veturnar  nom./acc.pl. vetur 84 Icelandic data: Papers I and II Note that while plural vetur –– also e.g. fingur and eigendur –– share the ending plural -ur with e.g. feminine plural stelpur, there is little reason to assume that any of the former has been assigned to the same feminine class as stelpa. This function of the schema, i.e. to facilitate reanalysis on the basis of grammatical gender but not definitive class assignment, stems from its association with several classes and subclasses of Icelandic feminine nouns, in which the majority of functionally equivalent forms lack clear phonetic definition beyond plural -ur. Considering now the often-observed correlation between productivity and type frequency, we might expect to associate subordinate [XæTur]nom./acc.pl. with a total lack of productivity, given the combination of its low type frequency –– it represents a six-strong prototype and potentially several more with satellites of various proximity –– and is minimally schematic. However, as noted by Bybee (2001:29), a pattern is capable of productivity if it is instantiated by a minimum of six items, while a minimum of two items likely suffices to facilitate the abstraction of a schema (Bybee 2010:64). 61 Therefore, by the same token, we might also expect to attribute greater productivity as a correlate of token frequency to [Xur]nom./acc.pl., relative to that of [XæTur]nom./acc.pl., on account of the former’s higher type frequency and greater degree of schematicity –– all other things being equal. In light of the preceding discussion, then, the mismatch in relative token frequencies between fem.pl.def. veturnar and fæturnar, on the one hand, and their respective masculine doublet forms, on the other, appears to run counter to the view that high productivity correlates with high type frequency and high schematicity. However, we know that crosslinguistically, minimally schematic classes of medium to low type frequency display limited degrees of productivity (see the relevant discussion in 3.2.3). Therefore, Paper I argues that minimal schematicity is the very property that accounts for different degrees of productivity. On account of the schematicity of basic level [Xur]nom./acc.pl., alignment with the schema is predicated on the ending plural -ur only. In other words, word-internal attributes beyond this sequence need not exhibit multiple one-to-one correspondences across paradigms for successful alignment. Conversely, alignment with [XæTur]nom./acc.pl. requires a far greater degree of one-to-one correspondence between individual segments. The nature of alignment between plural stelpur, vetur, fætur, rætur, and the schema [Xur]nom./acc.pl., as well as the schema’s category validity for intra-paradigmatic alternation of the kind [Xur]nom./acc.pl. ~ [Xurnar]nom./acc.pl.def., is depicted in (32) (see below on the notations employed). 61 In support of such claims, it is noteworthy that the six-strong Xó/æT-microclass has demonstrated some –– albeit highly limited –– productivity, while the semantic class that contains Ice. kýr ‘cow’ and ær ‘ewe’, i.e. two items after Ice. sýr ‘sow’ fell out of use, is completely unproductive. 85 Jón Símon Markússon (32) [Xurnar]nom./acc.pl.def. [X ur]nom./acc.pl. s t e l p urnar s t e l p ur v e t v e t urnar ur f æ t f æ t urnar ur r æ t r æ t urnar ur Common formal attributes that conform with schemas in terms of grammatical function are connected by thick lines, cf. connections between instances of plural -ur; the feminine article pl. -nar. Arrow-headed lines symbolise the category validity of the ending plural -ur for intra-paradigmatic alternation with a form in the feminine article plural -nar. Thin lines render arbitrary phonetic identity, which, by itself, is void of both cue validity and category validity. Note that despite the functional role of the vowel æ in fætur and rætur as a marker of plural in members of the Xó/æT-microclass, cf. sg. fót-, rót- ~ pl. fæt-, ræt-, (32) implies that this phonetic attribute is not specified at the basic level. In other words, plural vetur is just as likely to be reanalysed as feminine on analogy with e.g. plural rætur as is plural fætur due to formal identity between instances of plural -ur. Therefore, Paper I argues that at the basic level of the taxonomy posited in 5.3.2, common formal attributes beyond plural -ur do not impact the rate of reanalysis because all other things are equal. To what extent, then, does disproportionate potential for structure mapping justify positing the Xó/æT-microclass as a prototype within the feminine subtype in plural -ur, when the ending plural -ur alone suffices to facilitate reanalysis of masculine forms as feminine? To answer this question, let us now compare the implications of (32) and those inherent to (33). 86 Icelandic data: Papers I and II (33) v e t urnar v e t ur g ei t urnar g ei t ur n æ t urnar n æ t ur f æ t f æ t urnar ur r æ t urnar r æ t ur b æ k b æ k urnar ur [X æ T [X æ T urnar] ur] On inspection of (33), the question likely arises as to why the network implies a greater degree of functional similarity between plural nætur and fætur (and, therefore, also between fætur, rætur, and bækur), on the one hand, than between plural vetur, geitur and nætur, on the other. Below, it is argued that this disparity stems from the prototype structure of the network, as the degree of schematicity decreases on convergence with the schema(s) [XæTur(-nar)]nom./acc.pl.(def.) as a property of family resemblance structure (see 3.3). Thus, while common stem-final -t alone does not suffice to imply a form-function correspondence between plural vetur, geitur and nætur, its co-occurrence with the preceding e [ɛ:] in vetur, the diphthong ei [ei:] geitur, the diphthong æ [ai:] in nætur, and plural -ur potentially facilitates graded membership of the microclass. In other words, the sequence -etin vetur [vɛ:tʏr] bears some similarity to the sequence -eit- of geitur [cei:tʏr] due to the similarity of open-mid front unrounded [ɛ:] and the initial quality of the diphthong [ei:], i.e. 87 Jón Símon Markússon close-mid front unrounded [e]. Subsequently, the latter quality of the diphthong [ei:] shares near-high near-front unrounded [i] with the diphthong [ai:]. Thus, (33) allows for the possibility that [XæTur]nom./acc.pl. is schematic enough to facilitate the reanalysis and use of plural vetur as feminine on account of its (albeit tentative) similarity to the periphery of the Xó/æT-microclass, via the connection of plural geitur to nætur and, thence, fætur, rætur, bækur, etc. However, on account of the high degree of similarity between plural nætur, fætur, bætur, rætur as depicted in (34), Paper I argues that [XæTur]nom./acc.pl. nets the forms that align with it fully in terms of both form and function at a level of abstraction that closely reflects physical manifestation. To incorporate a maritime metaphor again: The greater the number of one-to-one form-function correspondences between a set of forms, the greater the ease with which they are netted by the relevant schema when caught adrift in the sea of usage. Therefore, the relatively greater number of one-to-one connections can be viewed as a metaphor for a net, the lines of which are ideal to both “catch”, i.e. attract, and “release”, i.e. facilitate the use of, inflectional forms as members of the Xó/æT-microclass. In this sense, subordinate [XæTur]nom./acc.pl., unlike its mother, basic level [Xur]nom./acc.pl., is capable of fulfilling the main function of categorisation, i.e. uncertainty reduction, at a minimal level of abstraction. That is, the one-to-one form-function correspondences in (33) that are predicated on structure mapping between instances of plural -ur only are just as strong as their form-function parallels in (32). However, minimally schematic [XæTur]nom./acc.pl. accurately reflects additional points of similarity between a particular set of forms in (33). In light of the above, an answer to the question posed in (27), repeated in (34) for convenience, can be given. (34) How do we account for the frequency relation between fem.nom./acc.pl.def. fæturnar and veturnar, on the one hand, and masc.nom.pl.def. fæturnir, masc.acc.pl.def. fæturna and veturnir, veturna, on the other? Paper I concludes that masculine plural fætur is relatively less likely to escape alignment with the schema [XæTur]nom./acc.pl. on both formal and functional grounds than is vetur or, indeed, any other masculine form in plural -ur. This is due to the relatively higher number of one-toone form-function correspondences between forms that align with prototypical, functionally equivalent forms of the Xó/æT-microclass. The greater the number of one-to-one correspondences, the greater the potential for successful alignment between functionally 88 Icelandic data: Papers I and II equivalent forms. Therefore, the limited productivity exhibited by the low-type frequency Xó/æT-microclass is predicated on the gang effect as a function of minimal schematicity. 5.3.6 Section summary Subsection 5.3.1 demonstrated that frequency of use impacts categorisation via the domaingeneral cognitive process of statistical learning. Subsection 5.3.2 elaborated on cognitive economy as a property of functionally related linguistic categories. Subsection 5.3.3 presented instances of actual language use as evidence for hierarchical relations between the schemas that constitute linguistic categories. In subsection 5.3.4, I argued that minimal schematicity facilitates limited productivity in the absence of significant type frequency. Subsection 5.3.5 modelled the extent to which varying degrees of schematicity facilitate different rates of productivity for schemas represented at distinct levels of abstraction by notational conventions characterised as the ‘net effect’. 5.4 Paper II: Deduction of plural blækur and kækur This section elaborates on the content of Paper II. Subsection 5.4.1 presents the content of Paper II. In subsection 5.4.2, I argue for the deduction of plural kækur and blækur from singular forms of the respective paradigms as part of a two-step process. Subsection 5.4.3 details the different analogical means by which individual forms of a paradigm can be deduced, i.e. by extension and levelling. Subsection 5.4.4 argues that plural (non-humorous) blækur and (humorous) kækur are deduced by extension, negating the position that one is a better example of “real” language use than the other. Section 5.4.5 provides a summary of the current section. 5.4.1 Focus: Semantics, pragmatics, and schematicity This and the following subsections elaborate on the content of Paper II, the focus of which is the assignment of feminine grammatical gender to forms of borrowed Ice. blók ‘non-entity, wretch’ and kók ‘CokeTM’. Inflection of both nouns occasionally follows that of the Xó/æTmicroclass. The innovations that manifest such use are feminine (humorous) plural kækur, beside (rare) feminine plural kókir,62 and (standard) feminine plural blækur, beside feminine blókir. The paradigms of Ice. kók and blók, including doublet forms, are shown in (35), beside feminine bók. 62 As noted by Svavarsdóttir (2014), masculine forms of kók also occurred, mainly from before and around 1960. However, these forms have since fallen out of use. 89 Jón Símon Markússon (35) sg. neut./fem. fem. fem. nom. kók blók bók acc. kók blók bók dat. kóki/kók blók bók gen. kóks blókar bókar pl. nom./acc. kók/kækur, kókir blækur, blókir bækur dat. kókum blókum bókum gen. kóka blóka bóka Paper II makes clear from the outset that plural kækur is a humorous innovation and that some consider its occurrence not to constitute “real language use”, implying that deduction of the form says little of the –– albeit highly limited –– productivity of the Xó/æT-microclass (see PII:195, footnote 1). However, such reasoning fails to account for the innovative relation nom./acc.sg. kók ~ plural kækur, which is clearly modelled on e.g. bók ~ bækur. Further, assuming that humour in some way impedes actual language use begs the question in (36). (36) In what way is the deduction of non-humorous plural blækur from sg. blók(ar) a better example of “real language use” than is the deduction of plural kækur from sg. kók? Paper II seeks to account for the occasional inflection of both Ice. kók and blók on analogy with schemas for the Xó/æT-microclass. To this end, the objective of Paper II (see PI:196) is restated in (37), repeated from (3). (37) To provide a usage-based cognitive account of the limited productivity of the Icelandic Xó/æT-microclass through recourse to: i. Semantics, ii. Pragmatics, iii. Schematicity, iv. Analogy. It is argued that deduction of both plural kækur and blækur is predicated on a high degree of phonetic similarity between both singular kók and blók and functionally equivalent forms of 90 Icelandic data: Papers I and II the Xó/æT-microclass. However, where treatment of individual forms of the respective paradigms differs, explanation refers to the relation between meaning and form. As information pertaining to form and function is considered to be based on stored experience with language, the mechanism posited for innovation through extension of the relevant knowledge is the domain-general cognitive process of analogy.63 5.4.2 Deduction as a two-step process The current subsection argues that deduction of plural kækur and blækur from their respective intra-paradigmatic sources, i.e. singular kók and blók(ar), proceeds as a function of alignment with the Xó/æT-microclass. Further, I justify positing extension as the mechanism for innovation as part of a two-step process that begins with the projection of feminine grammatical gender onto singular forms of the paradigm. Ultimately, deduction of plural kækur and blækur accords with the inflectional pattern represented by the sister schemas [XóT(ar)]sg. ~ [XæTur]nom./acc.pl.. As explained in 3.2, sister schemas abstract over intra-paradigmatic relations that actually occur in the language. Further, due to the category validity inherent to output schemas, which activates one constituent form as implicatory of the formal attributes of another, the relevant relation is extended to new contexts, where its formal attributes come to define a functionally equivalent relation within a new paradigm, e.g. nom./acc.sg. kók ~ plural kækur, sg. blók(ar) ~ plural blækur, cf. bók ~ bækur. In the current context, then, the schema [XóT(ar)]sg. is posited as the output schema, i.e. the source for plural targets of the kind [XæTur]nom./acc.pl.. This characterisation of the process pre-empts questions that may arise as to whether all forms of the respective paradigms are affected simultaneously with assignment of feminine grammatical gender. In other words, is the deduction of plural kækur and blækur part of a simultaneous, i.e. one-step, process or a two-step process like that argued for in Paper I (PI:9–10)? In answer, first, it is important to note that reanalysis is assumed to occur on-line, i.e. during actual usage events (e.g. Rácz et al. 2020). Secondly, it is assumed that establishment of multiple forms as constituents of a single paradigm cell is predicated on repeated reference to the relevant sister schema(s) over time, i.e. constitutes a gradual process. In terms of potential counter evidence provided by other forms of the paradigm, i.e. for wholesale, one-step gender assignment, some forms prove entirely uninformative. For 63 For convenience, the definition of analogy is repeated here (from (6) in 3.1): [T]he cognitive process through which existing knowledge is extended to new contexts. 91 Jón Símon Markússon example, as noted in Paper I (PI:7), the endings dat.pl. -um and gen.pl. -a are all but omnipresent across the paradigms for Icelandic nouns. Further, the endings in question are common to Icelandic nouns of all three genders and, therefore, do not speak to reanalysis.64 Moreover, wholesale reanalysis of kók as feminine should yield syncretic nom./acc./dat.sg. kók and also gen.sg. kókar, an intra-paradigmatic relation of high type frequency among Icelandic strong feminine nouns, cf. e.g. nom./acc./dat.sg. bók ~ gen.sg. bókar, mynd ~ myndar. However, only one example of gen.sg. kókar was returned from a search of the electronic text corpus Tímarit.is,65 along with one example of gen.sg.def. kókarinnar. Therefore, it appears that reanalysis as feminine can and does occur independently of other forms of Ice. kók. In this connection, it has been demonstrated that forms on opposite sides of the singular-plural divide within noun paradigms are semantically less related than forms that share a number value, i.e. either singular or plural. Such intraparadigmatic dynamics are reflected in the relation between meaning and form, which is characterised by a tendency to align common semantics with formal similarity (see Chapter 4). Therefore, on a theoretical basis, it can be argued that the semantic relatedness of neut. dat.sg. kóki and gen.sg. kóks, on the one hand, and fem. plural kók, on the other, to nom./acc.sg. kók does not proceed from an equal footing (PII:207). Paper II argues that this mismatch in relatedness is reflected by the higher token frequency of feminine plural kækur relative to that of fem. nom./acc./dat.sg. kók and/or gen.sg. kókar: plural kækur is disproportionately more frequent than the other two forms relative to the frequency of their neuter doublets. But if the entire paradigm underwent reanalysis as part of a one-step process, we should expect feminine doublets to occur with something like proportionate frequency. Further, if semantic relatedness between functionally distinct forms is skewed, that dynamic should be reflected by a disproportionate effect of the source of change on different targets within the paradigm. As this seems to be the case with different forms of Ice. kók, appeal to reanalysis as a two-step process appears to be justified by statistical data. 64 A notable exception being dative plural forms in -m of some nouns with a stem-final long vowel, cf. skór ‘shoe’, dat.pl. skóm, and weak feminines and neuters in gen.pl. -na, cf. fem. saga ‘story, history’, gen.pl. sagna; neut. auga ‘eye’, gen.pl. augna. 65 A description of the corpus taken directly from its website: “Timarit.is is a digital library where millions of pages in digital format are made available on the Internet. This gives access to the printed cultural heritage that is preserved in newspapers and periodicals of the Faroe Islands, Greenland, and Iceland.” See the bibliography for the relevant URL. 92 Icelandic data: Papers I and II 5.4.3 Different mechanisms of analogy: Ice. kók The current subsection details the different analogical means by which individual forms of a paradigm can be deduced, i.e. by extension and levelling. The motivation for positing different mechanisms rests on the interaction of schematicity and semantics with pragmatics, as these factors facilitate graded membership of the minimally schematic, low-type frequency Xó/æT-microclass. To demonstrate this impact, Paper II provides examples, which are repeated and discussed below, of conventionalised phrases in which Ice. kók occurs as feminine in the count noun sense, a context in which this semantic property interacts with schematicity to facilitate the deduction of plural kækur for humorous effect. Concerning innovative feminine plural kækur, it is possible to point to morphosyntactic relations that facilitate the role of nom./acc.sg. kók as the source of alignment with the sister schema [XóT]nom./acc./dat.sg. ~ [XæTur]nom./acc.pl.. In this connection, a particular morphosyntactic context appears conducive to reanalysis of nom./acc.sg. kók as feminine via the pragmatic interaction of morphophonology with agreement. Thus, the following account also appeals to the use of feminine modifiers with neut. kók and e.g. neut. rauðvín ‘red wine’. As noted in Paper II (PII:202), it is common idiomatic practice when ordering beverages in Icelandic to inflect modifiers of the liquid according to the grammatical gender of the container in which the former is sold. Both neut. kók and rauðvín are typically sold in a flaska ‘bottle’, while kók is also sold in a dós ‘can’.66 The words for both vessels are grammatically feminine in Icelandic and, therefore, sentences like those in (38) occur both naturally and frequently within the appropriate context (see Svavarsdóttir 2014).67 (38) a. Ég ætla að fá eina/ tvær kók I intend to get one-fem.acc./two-fem.acc. Coke™-acc.sg./acc.pl. rauðvín ‘I’ll have one/two Coke(s)™’ b. Ég ætla að fá eina/ tvær I intend to get one-fem.acc./two-fem.acc. redwine-acc.sg./acc.pl. ‘I’ll have one/two red wine(s)’ 66 Since the publication of Paper II, Sigurðsson et al. (2022) have published on agreement with ‘concealed’ or ‘elided’ –– in my terminology ‘omitted’ –– nouns in both what they characterise as ‘restaurant talk’, i.e. ordering food and beverages, and beyond. 67 It is also common practice to order both beverages in a neut. glas ‘glass’, by use of the compounds neut. acc.sg./pl. kókglas/-glös ‘glass(es) of Coke™’, rauðvínsglas/-glös ‘glass(es) of red wine’. In such instances, the modifying numerals always take the neuter form, cf. neut.acc. eitt/tvö ‘one/two’. 93 Jón Símon Markússon Should acc.sg. kók be treated as feminine,68 it might not seem unnatural on schematic grounds that language users should occasionally extend the functionally equivalent schema [XæTur]nom./acc.pl. to the relevant cell of the paradigm, given the morphosyntactic context in (38a). In other words, once acc.sg. kók is formally aligned with the schema [XóT]nom./acc./dat.sg., feminine grammatical gender is projected onto the former and, thus, the latter attracts acc.sg. kók for the anticipated usage event. Subsequently, as functionally equivalent forms of the Xó/æT-microclass already alternate with a plural of the type [XæTur]nom./acc.pl.., cf. feminine plural bækur, bætur etc., it is a short step to deducing plural kækur as the target of the (now) feminine source acc.sg. kók, cf. (39), whether for humorous purposes or otherwise. (39) [XóT]nom./acc./dat.sg. acc.sg. bók  [XæTur]nom./acc.pl. ~ nom./acc.pl. bækur  nom./acc.pl. X; X = kækur  acc.sg. kók Thus, feminine grammatical gender is projected via formal alignment of C with A, both of which align with the schema [XóT]nom./acc./dat.sg.. Subsequent to projection, the C-form also aligns functionally with the schema, which has category validity for alternation with B, i.e. bækur. Finally, C triggers a pattern of alternation analogous to A ~ B, hence, Paper II asserts, ‘X = plural kækur’. The above analysis suggests that the minimal schematicity of the Xó/æT-microclass facilitates a significant degree of structural alignment with the sister schemas [XóT]nom./acc./dat.sg. ~ [XæTur]nom./acc.pl. as a function of parallel connectivity. According to this line of reasoning, the same analogical process accounts for the general assignment of Ice. blók to the Xó/æT-microclass. Indeed, as is argued in the next subsection, categorisation and use of blók as a fully-fledged Xó/æT-microclass member serves to negate the view that deduction of plural kækur from singular kók is a substantively different process from that which yields plural blækur from singular blók(ar). At this juncture, it is important to recognise the formal difference between the source of change, i.e. singular kók, and the target, i.e. plural kækur. As argued in Paper II (PII:203– 204), this outcome is facilitated by a number of factors. First and foremost, from the 68 Note that, in terms of agreement, the mass noun kók is otherwise inherently neuter, cf. a sentence such as Hvar er kókið? ‘Where is the Coke™?’, where kók is clearly neuter due to the form of the definite article, nom./acc.sg. -ið. This is also the case when Ice. kók refers to cocaine. 94 Icelandic data: Papers I and II perspective of schematicity, deduction of the innovative feminine plural form results from alignment of singular kók with the output schema [XóT]nom./acc./dat.sg., as discussed above. However, when considered in the context of semantics and pragmatics, the context of use exemplified in (38a) is conducive to the expression of a semantic distinction between forms of kók as used in the count noun sense and, in turn, the establishment of formal differentiation of the relevant forms based on that semantic distinction. In other words, it is argued that differentiation of forms in a context such as (38a) is a function of the relation between meaning and form. The developments just discussed are juxtaposed in Paper II with an operation imposed on the singular dative form of Ice. kók when used in the count noun sense. Sentences such as vodki/romm í (dat.sg.) kók ‘vodka/rum in (idiom. and) Coke™’ occur relatively commonly, where the preposition í requires a dative object. Therefore, a pertinent question given the theme of Paper II can be stated as in (40). (40) Does dat.sg. kók in the sentence vodki í kók reflect an uninflected form of kók or should it be analysed as fem. dat.sg. kók? Paper II (PII:205–210) argues for the latter reading through reference to schematicity and analogy based on semantic and pragmatic similarity to contexts such as (38a). Thus, Paper II contends that sentences such as vodki/romm í kók likely reflect the mass noun sense of Coke™, i.e. an indeterminate amount of the liquid used as a mixer with a measure of spirit. However, while the mass noun sense is typically expressed through use of neuter forms, a semantic bridge between such usage and that exemplified in (38a) lends itself to the analysis argued for here. In other words, the count noun sense is understood through conventionalised agreement with the feminine words flaska and dós, i.e. as portions of the liquid served in the relevant containers. This reading is obviated by the use of feminine modifiers with acc.sg., acc.pl. kók, as exemplified in (38a), despite the conventionalised omission of feminine flaska and dós. Like flaska and dós, Ice. glas is a container that is used to serve liquid, such as a shot/measure of spirit along with a mixer. In other words, conventionalised language use like the context in (38a) is considered frequent and, therefore, entrenched enough to exert influence on related contexts, such as when ordering Coke™ as a mixer. This analysis is supported by the occurrence of PPs such as í kaldri kók ‘in cold Coke™’, where kaldri is the feminine dative singular form of kaldur ‘cold’. In answer to the question posed in (40), then, 95 Jón Símon Markússon language use of this kind supports the argument that dat.sg. kók is a feminine –– rather than an uninflected form. Given this interpretation, Paper II asks which analogical mechanism best characterises the deduction of fem. dat.sg. kók and which form of the paradigm is the most likely source of change. Paper II (PII:206–207) argues that fem. (nom./)acc.sg. kók, which occurs in the influential context exemplified in (38a), is the source of the innovation on semantic grounds. In other words, when acc.sg. kók is used to express the mass noun as served in a feminine flaska or dós, Ice. kók is treated as feminine as obviated by modifiers such as fem.acc. eina, tvær. Subsequently, contexts of use in which kók expresses the mass noun and omitted neuter glas, i.e. ‘a glass of Coke™’, are often aligned with contexts such as that exemplified in (38a) on schematic, semantic, and pragmatic grounds.69 It is, therefore, through this process of alignment that feminine grammatical gender is generalised to kók in both contexts. Further, this position opens the door for account via reference to the relation between meaning and form: both the source, i.e. fem. (nom./)acc.sg. kók, and target, i.e. fem. dat.sg. kók, express the common value singular. As noted in Chapter 4, the relation between meaning and form is reflected in greater formal similarity to the extent that forms of the paradigm express a common, semantically significant value, such as a number value in nouns. Thus, in light of common semantic and pragmatic motivation for alignment of phonetic structure, the analogical mechanism for deduction of fem. dat.sg. kók is best characterised as levelling. In other words, as pragmatic function is aligned on the basis of common semantics, the phonetic structure of the forms involved converges as an expression of that function. Therefore, the respective analogical mechanisms via which distinct forms of the word Ice. kók are deduced take their cue from the interaction of schematicity, semantics, and pragmatics to express the relation between meaning and form. 5.4.4 Parallel deduction of plural blækur and kækur As noted in Paper II (PII:195, footnote 1), some have suggested that humorous motive for the deduction of plural kækur belies “real” language use. However, Paper II argues that, while humour is indeed the likely motivation for use, schematicity, semantics, and pragmatics played their role in the deduction of singular kók through analogy with the sister schemas [XóT(ar)]sg. ~ [XæTur]nom./acc.pl.. Plural blækur is indubitably deduced from singular blók(ar) 69 See Hoffmann 2004, who argues for entrenchment of complex preposition via semantic analogy with more frequent, conventionalised, and functionally equivalent phrases. 96 Icelandic data: Papers I and II via the very same means but the process is not considered to belie “real” usage. In light of the application of parallel processes in deduction, that of feminine plural kækur from singular kók –– clearly modelled on the Xó/æT-microclass –– demonstrates that humorous motivation for the process is no impediment to “real” language use as a function of analogy. Further, pragmatics may have played a role in the reanalysis of blók as feminine between the source language, i.e. English, and the receiving language, i.e. Icelandic. Indeed, English bloke can only refer to men (see PII:205, footnote 8). In this connection, various factors, such as phonetic structure and the applicability of a word meaning ‘non-entity, wretch’ to lowly people irrespective of gender identity may well have facilitated categorisation of Ice. blók as a feminine noun. While this constitutes conjecture due to the lack of documented change, it is noteworthy that Ice. plural blækur and blókir are both feminine, cf. in this regard that (nom./)acc.sg. kók also occasionally alternates with fem. plural kækur and (rare) kókir, cf. the paradigm in (35), suggesting that schematicity has facilitated parallel usage for both nouns. For the reasons stated above, it is argued that parallel processes of analogical reasoning facilitate the deduction of both plural kækur and blækur from their respective intraparadigmatic sources, i.e. singular kók and blók(ar). First, the formal similarity of blók to original members of the Xó/æT-microclass likely facilitated attraction by the schema [XóT(ar)]sg.. Thus, the proportional equation in (41), which bears a strong phonetic resemblance to the proportion in (39), delineates extension of the Xó/æT-microclass pattern to the paradigm of blók. (41) [XóT]nom./acc./dat.sg. acc.sg. bók  [XæTur]nom./acc.pl. ~ nom./acc.pl. bækur  nom./acc.pl. X; X = blækur  acc.sg. blók Secondly, the fact that the feminine forms of numerals are used in sentences such as (38a), despite the fact that Ice. kók is chiefly neuter in the mass nouns sense, surely interacts with the noun’s formal attributes to facilitate analogy. Therefore, Paper II (PII:216) argues that it is the interaction of these factors, i.e. schematicity, semantics, and pragmatics, which motivates the extension of the Xó/æT-microclass schemas to other parts of the paradigm. Likewise, the fact that Ice. blók can refer to people of any gender despite Eng. bloke only 97 Jón Símon Markússon referring to men, surely interacts with the formal similarity of Ice. blók to functionally equivalent forms of the Xó/æT-microclass, facilitating extension of its schemas to the paradigm. In this connection, then, the question posed in (36) is now repeated in (42). (42) In what way is the deduction of non-humorous plural blækur from sg. blók(ar) a better example of “real language use” than is the deduction of plural kækur from sg. kók? Paper II concludes that while deduction of plural kækur certainly is motivated by humour, the form is just as deducible from (nom./)acc.sg. kók as plural blækur is from singular blók(ar) due to the pragmatic and schematic parallels outlined above. Therefore, deduction of both plural forms is considered a mechanism of language use, itself a function of analogy. Further, this process satisfies the function of categorisation, i.e. to reduce uncertainty in the phenomena we encounter in our environment by attributing structure to it. 5.4.5 Section summary The current section elaborated on the content of Paper II. Subsection 5.4.1 presented the content of Paper II. In subsection 5.4.2, I delineated the process by which plural kækur and blækur are each deduced as part of a two-step process. Subsection 5.4.3 argued for different analogical means for deduction of individual forms within the relevant paradigms. Subsection 5.4.4 argued that plural (non-humorous) blækur and (humorous) kækur are both deduced by extension, negating the position that one is a better example of “real” language use than the other. 5.5 Chapter summary This chapter presented the analyses that constitute Papers I and II, which focus on the productivity of the Icelandic Xó/æT-microclass to varying degrees. Section 5.1 laid out the content of both papers. Section 5.2 delineated the prototype structure of the Icelandic subtype in plural -ur as it centres on the feminine Xó/æT-microclass. Section 5.3 fleshed out the analysis presented in Paper I, accounting for the linguistic and non-linguistic factors that facilitate the reanalysis of masculine forms in plural -ur as feminine at different rates of frequency. In subsection 5.3.1, I argued that frequency of use impacts categorisation via the domain-general cognitive process of statistical learning. Subsection 5.3.2 elaborated on the principle of cognitive economy as a property of functionally related linguistic categories. In subsection 5.3.3, I presented instances of actual language use as evidence in support of 98 Icelandic data: Papers I and II hierarchical relations between constituent schemas of individual linguistic categories. Subsection 5.3.4 argued for the facilitating effect of minimal schematicity on limited productivity. In subsection 5.3.5, I modelled the extent to which varying degrees of schematicity facilitate different rates of productivity according to the ‘net effect’. Subsection 5.3.6 offered a summary of Section 5.3. Section 5.4 fleshed out the analysis presented in Paper II. Subsection 5.4.1 presented the content of Paper II. In subsection 5.4.2, I presented arguments forwarded in Paper II for the deduction of plural kækur and blækur from singular kók and blók by means of a two-step process characterised as analogical extension. Subsection 5.4.3 demonstrated that individual forms of a paradigm can be deduced by different analogical means, i.e. by extension or levelling. Subsection 5.4.4 argued that plural (non-humorous) blækur and (humorous) kækur are deduced by extension, negating the position that one is a better example of “real” language use than the other, when language use is viewed as a function of analogical reasoning. Section 5.4 was summarised in Section 5.4.5. 99 6 Faroese: Frequency as the determinant of levelling The current chapter delineates the objectives, arguments, data, and conclusions presented in Paper III, which is written in Icelandic. The aim of this endeavour is to provide a sufficiently detailed overview of the article’s content for those who do not have Icelandic reading proficiency. Section 6.1 discusses the content of Paper III in light of the theoretical focus. Section 6.2 fleshes out the divergent paths of levelling taken by the Faroese nouns vøllur ‘field, grassy ledge on a rock face, (sports) pitch’ and fjørður ‘fjord, inlet/bay, sound/strait’ (6.2.1), before discussing the impact of common semantics and/or frequency of use on the choice of basic form(s) for each paradigm. Subsection 6.2.3 provides a section summary. Section 6.3 establishes the basic forms for both paradigms based on the corpus data presented in Paper III. Subsection 6.3.1 posits acc.sg. vøll as basic according to an estimation of historical token frequencies for different forms of vøllur. Subsection 6.3.2 details the process of levelling in the paradigm of fjørður, where dat.sg. firði is posited as basic due to the relation frequency > entrenchment > lexical strength. Subsection 6.3.3 offers a summary of Section 6.3. Section 6.4 summarises the current chapter. 6.1 Paper III: Content and theoretical focus As with Papers I and II, the analyses presented in Paper III are guided by the usage-based cognitive approach to language change. Unlike Paper I, and in common with part of the analysis in Paper II, Paper III focuses primarily on the direction of intra-paradigmatic levelling, i.e. the process through which a word acquires new inflectional forms based on existing stem variants of the paradigm. Crucially, levelling should not be viewed as the transition of one form into another. In other words, it would be a mischaracterisation of the process to say that e.g. older Far. dat.pl. fjørðum “changed into” younger dat.pl. firðum. Rather, the latter results from extension of the stem variant firð- from another cell of the paradigm, cf. existing dat.sg. firði, nom./acc.pl. (henceforth ‘plural’) firðir. This characterisation of levelling is supported by the fact that older and resultant doublets co-exist for centuries, cf. that e.g. original dat.pl. fjørðum and innovative dat.pl. firðum ‘fjords’ are still in use. Moreover, Paper III deals with divergent processes of levelling in the paradigms of the Faroese masculine nouns vøllur and fjørður, both of which belonged to the class of u-stems 101 Jón Símon Markússon in Old West Nordic. What Papers I–III do have in common is that each seeks to account for the impact of token frequency, type frequency, and the dispersion of endings and/or stem variants on language change. As discussed in detail in 3.2.2, frequency of use impacts the domain-general process of entrenchment, which, in turn, is the determinant of lexical strength. Thus, by building on the testimony of levelling in other languages, Paper III seeks to determine the basic forms of paradigms, i.e. those forms upon which new ones are based. Given the causal relation frequency > entrenchment > lexical strength, Paper III seeks to answer the question already posed in (4); now repeated in (43). (43) Does frequency determine the direction of levelling? As noted in Chapter 4, in order to answer this question, Paper III utilises modern-day token frequencies for different inflectional forms of the relevant nouns, based on results returned from the electronic corpus for Modern Faroese, Teldutøka tekstasavn Føroyamálsdeildarinnar (TTF) and, in order to establish historical token frequencies for different forms of Far. vøllur, a comparison with data returned from the electronic corpus for Icelandic, Íslenskt textasafn (ÍT). In this connection, Paper III asserts that use of Far. vøllur for topographical referents, i.e. in the senses ‘field, grassy ledge on a rock face, (sports) pitch’, has likely always been rare due to the topography of the Faroe Islands (see PIII:62 and sources cited there). Further, it is demonstrated that Far. vøllur is used far more frequently in the modern language in reference to sports pitches than in any other sense at any period before the 20th century. Moreover, use of the word as a shortened form of flogvøllur ‘airport’ is an obvious nuance of the 20th century. In light of this, Paper III seeks to estimate the historical token frequencies for individual forms of vøllur through comparison with equivalent instances of use in both Modern Faroese and Old Icelandic written sources, where the latter is taken as representative of Old West Nordic more generally (see Chapter 2 on the relative scarcity of written sources for Faroese). This approach also considers real-world conditions such as topography, weather, and luminescence, as such factors would have impacted frequency of use. Conversely, it is demonstrated that Far. fjørður has likely always referred to the same phenomena since before the time of settlement, i.e. ‘fjord, inlet/bay, sound/strait’. This noun is also common in place names, where it refers to one of these natural phenomena. Thus, comparison of the different directions of levelling taken by Far. vøllur and fjørður prove highly informative with regard to the validity of theoretical approaches that seek to account 102 Faroese: Frequency as the determinant of levelling for the choice of basic form(s): both vøllur and fjørður have topographical referents and occur in place names (see PIII:54). Therefore, development of the paradigms in question provide opportunity to evaluate the impact of common semantics on the direction of levelling; a tenet of the approach to levelling associated with so-called markedness, which views semantically “natural”, “neutral”, and/or “default” inflectional categories as diagnostic of basic status (Greenberg 1966). Given this view, we might expect nouns that share a semantic domain to undergo similar changes with regard to the direction of levelling. However, as demonstrated time and again, actual language change is often at odds with the approach from markedness, which, it is argued, is ultimately characterised by circularity (see PIII:58 and sources cited there). In light of this drawback to markedness, I chose to test the impact of frequency because it is a measurable property of usage, while semantics is not (following Bybee 2015:102; Haspelmath 2006:27, 58; Sims-Williams 2022:571). In Paper III, the term frequency (Ice. tíðni) is first and foremost used in reference to token frequency (Ice. staktíðni), which is considered to determine the initial choice of basic form as a function of the causal relation frequency > entrenchment (Ice. rótfesta) > lexical strength (Ice. minnisstyrkur). Further, it is argued that both the type frequency (Ice. mynsturstíðni) of patterns and the dispersion (Ice. dreifitíðni) of stem variants within the paradigm assists in completing the process of levelling as a function of the cross-linguistic tendency to establish a relation between meaning and form (Ice. samband forms og merkingar). Paper III concludes that the divergent processes of levelling discussed stem from the interaction of token frequency, type frequency, and (intraparadigmatic) dispersion. 6.2 The direction of levelling This section delineates the divergent paths of levelling taken by the Faroese nouns vøllur and fjørður (6.2.1), before discussing the impact of inhabiting a common semantic domain, i.e. having topographical references and occurrence in place names, and/or frequency of use on the choice of basic form(s) for each paradigm. Subsection 6.2.3 provides a section summary. 6.2.1 Levelling in Far. vøllur and fjørður In the current subsection, I delineate the divergent paths of levelling taken by Far. vøllur and fjørður. Crucially, both nouns belonged to the class of u-stems in Old West Nordic, meaning that they showed the same endings and exhibited functionally parallel patterns of alternation (see below). The paradigms of Old West Nordic u-stems have undergone extensive 103 Jón Símon Markússon analogical levelling, eliminating all but one older stem variant in the vast majority of cases. However, as discussed in Paper III (PIII:55–56), the direction of levelling in the paradigms of Far. vøllur and fjørður (< OWN vǫllr and fjǫrðr) demonstrates that the process was not uniform across-the-board. In order to gain perspective in the historical context, the paradigms of OWN vøllur and fjørður are shown in (44). vǫllr fjǫrðr acc. vǫll fjǫrð dat. velli firði gen. vallar fjarðar pl. nom. vellir firðir acc. vǫllu fjǫrðu dat. vǫllum fjǫrðum gen. valla fjarða (44) sg. nom. For the purpose of comparison, the paradigms of Far. vøllur and fjørður are given in (45), where the stems of forms that arose through levelling are in bold.70 (45) sg. nom. vøllur fjørður acc. vøll fjørð dat. vølli firði/fjørði pl. nom./acc. vøllir/vallir firðir/fjørðir/fjarðir dat. vøllum/vallum fjørðum/firðum/fjarðum As obviated by its paradigm in (45), Far. vøllur has followed the general path of u-stems, i.e. levelling in favour of the stem variant found in the singular nominative and accusative and those of the plural accusative and dative, i.e. vǫll- (> Far. vøll-).71 The variant vell- has been 70 Although separate genitive forms exist in Faroese, the genitive case is no longer actively assigned, its disappearance having yielded a three-case distinction in the modern language, i.e. nouns are inflected for nominative, accusative, and dative case (Þráinsson et al. 2012:62). 71 Extension of the stem variant vall- appears to be modelled on the pattern of alternation exhibited by e.g. sg. høll ~ nom./acc.pl. hallir ~ dat.pl. høllum/hallum ‘palace(s); centre(s), hall(s)’. In other words, it is probably the 104 Faroese: Frequency as the determinant of levelling completely eradicated from the paradigm. Conversely, Far. fjørður has retained all of the stem variants it inherited from Old West Nordic and has acquired several new forms with the variants fjørð-, fjarð- and firð-. It is this disparity between the two paradigms that motivates the question in (4)/(43) and now repeated in (46) for convenience. (46) Does frequency determine the direction of levelling? The data, their interpretation, and attendant arguments laid out in the following subsections will facilitate the answer to this question. 6.2.2 The choice of basic form(s) In this subsection, I discuss the impact on nouns of sharing a semantic domain and of frequency of use on the choice of basic form(s) for the paradigms of Far. vøllur and fjørður. Focus on the development of these nouns was not a random choice. Rather, as noted in 6.1 and 6.2.1 the nouns both have topographical referents and appear in place names. Further, they formerly exhibited parallel inflectional attributes, both with regard to inflectional endings and alternation between stem variants. First, both nouns are descended from Old West Nordic u-stems, i.e. vǫllr and fjǫrðr, respectively. This means that they showed the same inflectional endings and exhibited functionally parallel patterns of alternation at an earlier stage, cf. the Old West Nordic paradigms in (44). Secondly, Far. vøllur and fjørður have topographical referents that occur both as features of the landscape and in place names, e.g. Harðavøllur, Tórsvøllur; Fuglafjørður, Skopunarfjørður (all place names). This point of commonality is especially significant in light of the theoretical approach to levelling that is associated with so-called markedness. The approach from markedness mainly attributes the selection of basic forms to the influence that semantically “natural”, “neutral”, and/or “default” inflectional values supposedly have on use, as briefly noted in 6.1. In the context of nominals, proponents of markedness consider the nominative case to possess all three qualities and, therefore, to be unmarked compared with other case forms, which are considered marked (e.g. Greenberg 1966). By the same token, singular number is generally considered unmarked against the dual and plural. In other words, nominative and singular represent the supposedly more neutral case and number values, while other values within each category convey extra information. result of extension rather than levelling and, therefore, falls out of the scope of the current focus (see PIII:56, footnote 5). 105 Jón Símon Markússon In metaphorical terms, this extra information often literally “marks” the forms in questions, then requiring additional phonetic material in order to express extra semantic information (see below). According to the approach from markedness, then, the nominative singular form of a “regularly inflected” noun should logically serve as basic, i.e. new forms within the paradigm should be based on the form of the nominative singular (Petersen 2009:89 states specifically that the nominative is the default case for Faroese). However, actual language change demonstrates that levelling does not always proceed from those forms of the paradigm that proponents of markedness consider unmarked. Indeed, while some might attribute this status to nom.sg. vøllur in light of levelling within its paradigm, the course of change in Far. fjørður clearly runs counter to such a view in some respects, cf. (44) and (45). Indeed, as argued in Paper III (PIII:77), innovative dat.pl. firðum suggests that nominative and singular fjørður did not serve as basic with regard to levelling. So, how have proponents of markedness sought to resolve such exceptions to the general applicability of the general theory? In order to impose systematicity on the obvious shortcomings of the approach, scholars attribute the quality of local markedness to paradigms where levelling runs counter to the notion of general markedness, discussed above. For example, Mańczak (1958) noticed that levelling in the paradigms of place names tended to favour locative forms as basic. In Faroese, the most obvious equivalent of the locative case in e.g. Russian is the dative. Indeed, objects of the prepositions Far. í ‘in’ and á ‘on’ occur in the dative to indicate location, e.g. í (dat.) Fuglafirði ‘in Fuglafjørður’. Bernharðsson (2004:25–26) notes that the dative case was especially common in this function in Old Icelandic, where the preposition að with dative object frequently denoted location also, e.g. lét kalla að (dat.) Ökrum ‘named it Akrar’. As this is the case in Old Icelandic, the same dynamic likely also applied in the contemporaneous Old West Nordic dialect spoken in the Faroe Islands (see Chapter 2). Beyond the dative case, use of the accusative in Faroese renders a relation of increasing proximity to a location. In other words, the Faroese accusative is not prototypically locative but indicates movement towards a place (see Þráinsson et al. 2012:164). Its use in this function is also associated with the prepositions í and á but also with Far. til ‘to’, e.g. Eg fari oman í (acc.) býin ‘I’m going downtown’, Vit fara til (acc.) Fuglafjørð ‘We are going to Fuglafjørður’.72 Given the fact that both the dative and accusative are associated with 72 In Old West Nordic, the proposition til governed the genitive case, as its Icelandic reflex does to this day. However, the accusative now frequently marks the object of til in Faroese (see footnote 70). 106 Faroese: Frequency as the determinant of levelling location in Faroese –– albeit to different degrees –– Paper III contends that we might have expected Far. vøllur and fjørður to show more uniform development with regard to levelling in light of the approach from (local) markedness. So, why is this not the case? In advance of the explanation, it should be noted that Haspelmath (2006) considers the term markedness inherently vague, attributing this shortcoming to the generality of the term as applied in the linguistic literature. Thus, he demonstrates, markedness can be defined in at least twelve different ways, so that it means different things to different linguists, depending on the material chosen for analysis and the theoretical approach applied. Further, it appears from Haspelmath’s survey of the term’s application in linguistic inquiry that some scholars are not aware of its different uses (2006:27). In other words, “markedness” is reducible to numerous practical applications in the linguistic literature and, as a result, serves a highly general function, rather than constituting a well-defined term. In an attempt to resolve this issue, Haspelmath (2006) notes that, in most studies where markedness is invoked, the term is reducible to frequency effects in language use and change. In this connection, it should be noted immediately that frequency has itself been considered an indicator of the (un)marked status of forms within paradigms (e.g. Greenberg 1966). However, as pointed out most recently by Sims-Williams (2022:571) and in Paper III (PIII:58–59), invocation of frequency as a diagnostic criterion for markedness renders application of the term subject to circularity. Consider the following axiom: Inflectional values considered semantically “natural”, “neutral”, and/or to be “default” relative to opposing categories are unmarked. Further: Unmarked forms tend to be more frequent than marked forms. However, a logical fallacy inherent to such an approach is obvious if relatively higher frequency is subsequently taken as diagnostic of semantically more natural, neutral, and/or default status within the paradigm. In other words, the approach from markedness proves circular: Higher frequency as a property of use provides evidence of unmarked status > if a form is unmarked it will likely be more frequent than other forms of the paradigm > higher frequency as a property of use provides evidence for unmarked status. And so the cycle continues. Paper III espouses a more falsifiable approach to determining the basic forms of paradigms. By accounting for the process of levelling in terms of the impact of frequency on the cognitive organisation of language, linguists are offered a means of disproving their hypotheses: frequency is a measurable property of language use, while meaning is not. Further, any account that seeks to attribute the status of basic form to semantics must do so on a theory-internal basis. In other words, the theory that a set of inflectional values 107 Jón Símon Markússon determines the general direction of levelling (general markedness) requires that a given word convey a meaning that permits those categories to determine the direction of levelling (local markedness). Conversely, the usage-based cognitive approach to language finds support from beyond the linguistic system, in properties such as frequency of occurrence, schematicity, and context of use, all of which are demonstrable determinants and yields of statistical learning, schematisation, and entrenchment. Support for this view comes from various studies on the determinants of levelling. For example, demonstrating that the direction of levelling is independent of the semantic properties of individual words, Tiersma (1982) noted that basic forms for Frisian nouns were selected on the basis of their individual token frequencies. Levelling of this nature has also occurred in Icelandic, where e.g. the stem OIce. gæs- ‘geese’ from the more frequent plural has replaced OIce. sg. gás-, now sg. gæs- (see Bernharðsson 2006). The Old West Nordic vocalic alternation is preserved in the paradigm of the Faroese cognate, thus Far. sg. gás ~ nom./acc.pl. gæs ~ dat.pl. gásum. It is important to note here that statistical learning is a tacit process, meaning that the gradual process of entrenchment determines the ease with which a given stem variant is retrieved from memory, i.e. its emergent lexical strength (see 3.2.2). Therefore, the direction of levelling is determined by the accessibility of competing linguistic structures and the different patterns that can be applied in order to express a given concept (see discussion in PIII:75). In other words, language users are not necessarily actively aware that they are choosing the most frequent form of the paradigm as basic. Rather, the choice is informed by tacitly accrued, entrenched information pertaining to frequency. 6.2.3 Section summary In this section, I discussed the nature of levelling and its determinants according to opposing theories. In 6.2.1, I delineated the different processes of levelling taken by Far. vøllur and fjørður, demonstrating that the former followed the majority of u-stems in Faroese language history, while the inflection of fjørður has become more complex. This disparity was considered interesting in light of semantic properties common to the nouns in question: both have topographical referents and occur in place names. In 6.2.2, I argued against semantics as a basis for positing basic forms due to the vagueness of the term “markedness”. It was concluded that frequency of use provided a measurable and, therefore, falsifiable means of testing the impact of language use on lexical strength and is, therefore, a better-suited gauge for establishing the basic forms of paradigms. 108 Faroese: Frequency as the determinant of levelling 6.3 Determining the basic forms of Far. vøllur and fjørður The current section accounts for the direction of levelling based on frequency data presented in Paper III and in light of the three distinct parameters for frequency counts acknowledged in 3.2.1, i.e. token frequency, type frequency, and dispersion. In 6.3.1, I posit acc.sg. vøll as basic on the strength of the estimation of historical token frequencies for different forms of the word. Further, I argue that dispersion of the stem variant vøll- in the singular portion of the paradigm ultimately determined the direction of levelling. Subsection 6.3.2 details the process of levelling in the paradigm of Far. fjørður. There, dat.sg. firði is posited as basic. Further, levelling in favour of the dative singular is argued to have motivated a subsequent attempt to establish a relation between meaning and form that distinguished singular from plural, a process that takes its cue from dispersion. Subsection 6.3.3 offers a summary of the current section. 6.3.1 Far. vøllur In the course of Faroese language history, the stem variant vøll- (< OWN vǫll-) has been generalised to all cells of the paradigm for Far. vøllur, cf. (44) and (45). As noted in 6.2.1, the development of Far. vøllur can be considered representative of the general path of levelling in Old West Nordic u-stems. This process is typically characterised by extension of the stem variant found in the singular nominative, accusative, and plural accusative and dative to all cells of the paradigm (see PIII:56 for a list of examples). As argued in Paper III (PIII:72), such developments reflect the high type frequency of phonetic uniformity in the stems of individual Faroese masculine nouns: the vast majority contains the same stem variant in all cells of the paradigm. Further, u-stems shared the inflectional ending nom.pl. -ir with the larger class of masculine i-stems. Moreover, formal and functional similarity between the two classes is to some degree increased via one-to-one correspondence based on the ending gen.sg. -ar, which all u-stems shared with a small number of nouns from other masculine classes; mainly i-stems. Therefore, the general direction of levelling among u-stems in Faroese can be accounted for via reference to the interaction of high type frequency of stem uniformity in masculine paradigms and points of formal and functional convergence with a larger masculine class. However, reference to type frequency does not suffice in accounting for the divergent course of levelling apparent from the paradigm of Far. fjørður. This fact is taken to suggest that type frequency alone might not be the only factor that contributes to productivity, suggesting that Yang’s Tolerance Principle is too rigid (see footnote 16). In order to 109 Jón Símon Markússon demonstrate that other parameters of frequency are at play in the developments under discussion, Paper III posits a chronologically ordered course of levelling for Far. vøllur through reference to token frequencies for different forms of the word, which are based on an estimate for earlier phases of the language via comparison with contemporaneous frequencies for forms of the Icelandic cognate -vǫllr/völlr/völlur (PIII:68). Further, the account also avails itself of the dispersion of individual stem variants, which is said to constrain distribution across the singular-plural divide as a function of the relation between meaning and form (Chapter 4). Based on the estimated token frequencies reported in Paper III and that are considered to reflect usage patterns before the 20th century, acc.sg. vøll had the highest token frequency, equivalent to 34 instances, compared with dat.sg. velli/vølli, with the equivalent of ten occurrences (see PIII:68, Table 4). Given that token frequencies for individual forms of Far. vøllur were highly likely to have been rare in centuries past, Paper III (PIII:75) asks whether the difference between 34 instances and 10 suffices to determine the choice of basic form. The answer to this question references the intra-paradigmatic dispersion of different stem variants: when the singular portion of the paradigm is considered as a whole, it becomes clear that older dat.sg. velli had not only to compete with the equivalent of 34 occurrences of acc.sg. vøll but also ten instances of nom.sg. vøllur. Both of the latter contained the stem variant vøll-. Thus, in light of the view that frequency impacts the direction of levelling as a function of statistical learning, Paper III argues that the combined token frequencies of acc.sg. vøll (34 instances) and nom.sg. vøllur (10 instances) interacted with the dispersion of the stem variant vøll- over two of the three cells of the singular.73 The combined token frequency of the forms with the stem variant vøll- accounts for 81.48% of the token frequency for all three singular forms. Further, the same variant enjoys a 66.66% rate of dispersion in the singular portion of the paradigm, meaning that language users are more likely to associate the form vøll- than vell- with the meaning singular. Therefore, given the impact of frequency as a determinant of both entrenchment and lexical strength, it is argued that the stem variant vøll- will have been more easily sought from memory to serve this semantic function. 73 Of course, levelling may have commenced before the genitive was lost as an active case in Faroese, meaning that the singular portion of the paradigm would then have contained four cells with three distinct stem variants, i.e. nom./acc.sg. vøll-, dat.sg. vell-, and gen.sg. vall-. However, the stem variant vall- now only occurs in plural forms of the paradigm and is the likely result of extension based on e.g. sg. høll ~ pl. hallir, as noted in 6.2.1, cf. the innovative alternation Far. sg. bjørn ~ pl. bjarnir, OWN bjǫrn ~ birnir. 110 Faroese: Frequency as the determinant of levelling Paper III also asks why the stem variant vell- was not subsequently generalised to the plural portion of the paradigm, given that it expressed the function nominative plural and possibly accusative plural too, had the latter been levelled to velli(r) before the stem variant vøll- started to spread. Such a dynamic would have yielded a relation between meaning and form, by which vøll- equated with singular and vell- with plural. However, as Paper III (PIII:76) notes, the individual token frequencies for plural forms of Far. vøllur have likely always been relatively low (cf. PIII:63, Table 1 and PIII:68, Table 4). Therefore, given the generally low frequency of all forms of the noun, the variant vell- is unlikely to have been entrenched enough to resist levelling to vøll-. Indeed, intra-paradigmatic pressure from both the singular portion of the paradigm and dat.pl. vøllum, which appears to have been more frequent than nom.pl. vellir and acc.pl. velli(r), likely facilitated levelling in favour of the stem variant vøll-. It should, of course, be acknowledged that it is advisable to err on the side of caution when making claims about the historical frequency of inflectional forms and the chronology of change in the absence of primary written sources. However, Paper III argues that we must make use of the data available to us rather than foregoing an attempt at explanation. On the basis of the method for recalculation delineated in Chapter 4, the hypothesis that frequency should impact the direction of levelling is testable. In other words, the available data is not interpreted in order to fit the theoretical approach but, rather, the approach has been subsequently applied to the estimation of historical frequency based on the recalculation. The method for recalculation was motivated by a preponderance of use as affected by real world conditions and was based on a comparison of relevant –– albeit secondary –– sources. As also noted in paper III (PIII:66, footnote 17), the study is not the first to rely on secondary sources as a basis for estimating historical frequencies. 6.3.2 Far. fjørður As discussed in 6.2.1, the development of Far. fjørður with regard to levelling is atypical of the process among u-stems in Faroese language history. Indeed, while almost all nouns of this class have generalised the stem variant descended from the singular nominative and accusative and plural accusative and dative in Old West Nordic, the paradigm for fjørður contains innovative forms with stem variants from other cells of the older paradigm, cf. (44) and (45). As argued in 3.2.2, token and type frequency are considered the explicit mechanisms of entrenchment, which determines lexical strength. It is important to note in this connection 111 Jón Símon Markússon that while the overall course of levelling in the paradigm of Far. fjørður does not speak to the influence of any particular pattern of inflection, the process as it occurred within the singular and plural portions of the paradigm, respectively, does speak to such influence. Thus, the relation nom.sg. fjørður ~ acc.sg. fjørð ~ (innovative) dat.sg. fjørði aligns with functionally equivalent relations in larger masculine classes, cf. a-st. hestur ~ hest ~ hesti, i-st. gestur ~ gest ~ gesti, where the stem is identical in all forms. Likewise, the relation nom./acc.pl. firðir, (innovative) dat.pl. firðum accords with e.g. gestir, gestum for the same reason. However, it is clear from data presented in Paper III (PIII:69, Table 5) that disparities in token frequency between individual forms of fjørður are a property of its use. It is, therefore, fair to assume a role for the impact of token frequency in the direction of levelling and investigate the hypothesis that this frequency parameter served as the initial catalyst for change. As noted in Paper III (PIII:80), an obvious position from which to mount that investigation is to view highly frequent dat.sg. firði as basic in the paradigm. In this connection, it is important to note that the only other cell to which the stem variant firð- has been extended is that of the dative plural, cf. younger firðum, which now exists alongside older fjørðum. It is, of course, also possible that the stem of this younger form is derived from plural firðir. However, Paper (PIII:77–78) presents two arguments against this proposition. First, motivation to level the plural in this way would be a function only of the dispersion of the relevant stem variant and not the impact of its frequency on entrenchment and lexical strength: dat.pl. fjørðum has higher token frequency than nom./acc.pl. firðir and, therefore, the former should be better entrenched and easier to access than the latter (see PIII:69, Table 5). In other words, while dispersion of the variant firð- across the plural portion of the paradigm may well have facilitated a relation between meaning and form, there is little to suggest that dispersion contributes to entrenchment and, therefore, also lexical strength at the cost of token frequency (see discussion in 3.2.1). Secondly, the dative forms of both numbers occur in a common pragmatic context, i.e. as the objects of the prepositions í/á/úr, cf. the schema [í/á/úr FJØRÐURdat.], e.g. í Fuglafirði, úr Kollafirði. Further, the schema just posited is likely deeply entrenched due to the frequency that it is invoked on account of language users’ perception of real-world conditions. In other words, Fuglafjørður, Kollafjørður, and most other place names ending in Far. -fjørður refer to both fjords and towns, where people live and locate events in reported speech. If it is correct that frequency of use is the mechanism of entrenchment, then it is logical to assume that the impact of frequency on levelling is a function of the relative ease 112 Faroese: Frequency as the determinant of levelling with which a stem form is accessible for use, i.e. that form’s lexical strength, within a particular context. In other words, invocation of the schema [í/á/úr FJØRÐURdat.] is a highly practiced behaviour, which, due to the process of automatisation, becomes more effortless with time (see 3.2.1). One recognised function of automatisation is autonomy, by which a form or lexical item acquires a high degree of independence from the functional constraints of its paradigm or common context of use and, in emergent fashion, becomes less compositional, i.e. analysable in terms of its constituent morphological elements (e.g. Bybee 2010:48). Therefore, it is likely that due to the high frequency with which the schema [í/á/úr FJØRÐURdat.] contains the form dat.sg. firði, the stem variant firð- will have become deeply entrenched in this context, irrespective of its function in relation to other forms of the paradigm. Paper III (PIII:75) also argues that, as a function of rich memory for language, individual form-function units, such as stem variants and inflectional endings, are stored in memory both separately, cf. e.g. [firð-]dat.sg., nom./acc./dat.pl. and [-ir]nom./acc.pl., [-um]dat.pl., and in broader contexts as constituent parts of the word, cf. dat.sg. firði, nom./acc.pl. firðir, or dat.pl. firðum. Indeed, were the individual building blocks of word forms not stored in memory, it should not be possible to attach the ending dat.pl. -um to the stem variant firð-, yielding innovative dat.pl. firðum. Further, were the process not a function of the impact of frequency on use, we should expect to witness equivalent processes for all forms of Far. fjørður, as none would be subject to change as a property of stored experience but, rather, online language processing alone. However, the development of the relevant paradigm suggests that forms such as nom.sg. *firður and acc.sg. *firð do not occur, a property of usage attributable to the fact that there is no crossover between nom.sg. fjørður and acc.sg. fjørð, on the one hand, and dat.sg. firði, on the other, within contexts such as [í/á/úr FJØRÐURdat.]. Given the above discussion, Paper III concludes that younger dat.pl. firðum is the first form of Far. fjørður to result from the process of levelling within the paradigm (PIII:77–78). The reasoning for this position is based on entrenchment, in light of the fact that the human cognitive faculties have been uniform across the species for millennia. In other words: the domain-general mechanisms of language use at play today are the same that have propelled language change in all language communities at all times (Bybee 2003). Thus, it is argued that Far. firði has always been the most frequent member of the paradigm across a period well beyond the living memories of contemporaneous speakers at any time since the Faroe Islands were settled. Thus, in light of the causal relation frequency > entrenchment > lexical 113 Jón Símon Markússon strength, it is also reasonable to assume that the same form has always been more entrenched than other forms of Far. fjørður at all stages in Faroese language history. It is argued that token frequency as a mechanism of levelling facilitated extension of the stem variant firð- from dat.sg. firði to the dative plural form –– especially within the context [í/á/úr FJØRÐURdat.] –– potentially even relatively early in Faroese language history. In light of this position, extension of the stem variant fjørð- to the dative singular cell, cf. younger dat.sg. fjørði, is considered to result from a reactionary operation: a function of the relation between meaning and form, rather than the lexical strength associated with any of the forms that originally contained that variant (see PIII:80). Once dat.pl. firðum comes into use, it soon gains in token frequency due to the degree to which its stem is entrenched in the context [í/á/úr FJØRÐURdat.], facilitating association of the stem variant firð- with the meaning plural, cf. nom./acc.pl. firðir and (innovative) dat.pl. firðum. This process yields an uneven relationship between meaning and form because the stem variant firð- still exists in the singular portion of the paradigm, cf. dat.sg. firði, opposing both nom.sg. fjørður and acc.pl. fjørð. Paper III concludes, therefore, that users of Faroese subsequently attempted to establish a relation by which the stem variant fjørð- was equated with the meaning singular, the logical solution to which was to derive younger dat.sg. fjørði by borrowing the stem variant from nom.sg. fjørður and acc.sg. fjørð. However, it is argued that –– despite language user’s best efforts –– the token frequency of original dat.sg. firði was (and still is) so much higher than that of the other singular forms combined that dispersion of sg. fjørð- does not suffice to usurp the function of dat.sg. firði. This position is supported by the fact that the former has not taken the latter over in terms of token frequency, while younger dat.pl. firðum is far more frequent than older dat.pl. fjørðum, probably due to the high token frequency of dat.sg. firði and the resulting entrenchment of the stem variant firð- in the context [í/á/úr FJØRÐURdat.]. 6.3.3 Section summary The current section provided an account of the process of levelling in the paradigms of Far. vøllur and fjørður as depicted in Paper III, which sought to determine the basic forms of each. Subsection 6.3.1 argued that acc.sg. vøll served as the basic form of Far. vøllur due to its historically higher token frequency. Further, it was argued that dispersion of the stem variant vøll- across two of the three cells of the singular portion of the paradigm served to facilitate levelling in favour of the stem variant vøll- generally. In subsection 6.3.2, I detailed levelling in Far. fjørður. There, dat.sg. firði was posited as basic, while extension of its stem 114 Faroese: Frequency as the determinant of levelling variant was argued to have been countered by an attempt to establish a relation between meaning and form that distinguished singular from plural. The process that derived innovative dat.sg. fjørði is regarded to have taken its cue from the dispersion of different stem variants within each number. 6.4 Chapter summary The current chapter delineated the contents of Paper III. Section 6.1 fleshed out the contents of the article in light of the theoretical focus. Section 6.2 delineated the divergent paths of levelling taken by the Faroese nouns vøllur and fjørður (6.2.1), subsequently discussing the impact of common semantics and/or frequency of use on the choice of basic form(s) (6.2.2). Subsection 6.2.3 provided a summary of Section 6.2. Section 6.3 posited the basic forms for both paradigms based on frequency counts from a text corpus for Faroese. Subsection 6.3.1 argued that acc.sg. vøll is basic to the paradigm of Far. vøllur according to an estimation of historical token frequencies for different forms of the word. Subsection 6.3.2 detailed the process of levelling in the paradigm of Far. fjørður, where dat.sg. firði was posited as basic. Subsection 6.3.3 offered a summary of Section 6.3. 115 7 Conclusion The current thesis sought to investigate and demonstrate the impact of domain-general cognition as a determinant of language change, itself a function of language use. To this end, changes in the inflectional systems of Icelandic and Faroese have been in focus, with the yield of these endeavours being presented in Papers I–III (see 1.1 for bibliographical information). As noted in 3.1, research conducted over the last decades into the grammatical system of Icelandic has primarily been guided by the tenets of generative and dual-processing models. Conversely, application of the usage-based cognitive approach has been mostly implicit, i.e. without statement of specific tenets of the relevant model. In this context, it is hoped that the current thesis motivates further research into the respective inflectional systems of Icelandic and Faroese from the usage-based cognitive perspective. As noted throughout, the current thesis argues for the impact of domain-general cognitive processes such as statistical learning, entrenchment, categorisation, and schematisation as facilitators of both language use and language change. Further, in order to show that these processes apply domain-generally, it was paramount that the approach demonstrate the impact of non-linguistic factors such as frequency and schematicity in language change. Moreover, demonstration of this impact is viewed as evidence for rich memory for language, thus calling the validity of more parsimonious theoretical models, such as dual-processing, into question. Indeed, as noted in 3.2.1, were a robust capacity for the storage of linguistic experience not a property of memory, we should expect to witness far fewer –– if any –– observable frequency effects in language change. In the context of the above, the current thesis set out to satisfy the general objectives stated in (1a–c), which, for convenience, are repeated below in (47a–c). (47) a. To determine which cognitive factors have impacted some details of the structure of the inflectional systems of Icelandic and Faroese. b. To ascertain whether these cognitive factors apply only to language or are domain-general in nature. c. To demonstrate how the impact of these cognitive factors are evident through language change as a function of language use. 117 Jón Símon Markússon In light of the aims reiterated above, the analyses presented in Papers I–III are considered to satisfy the objectives in (1a–c)/(47a–c). For example, in terms of the impact of non-linguistic factors on the structure of grammar, Paper I demonstrated that the 91.89% rate at which the ending Ice. plural -ur is dispersed among feminine nouns motivates the reanalysis of masculine forms in the ending as grammatically feminine. Reanalysis is, therefore, a function of statistical learning, itself a domain-general cognitive process applied to both linguistic and non-linguistic phenomena. However, it is the dispersion of formal attributes such as those associated with the Icelandic masculine class of -nd- stems in plural -ur, in conjunction with the knowledge that all of these can have biologically male referents, that yields different rates of reanalysis among masculine forms in the same ending. Thus, the forms of nd-stems and other masculines in plural -ur align with the medially schematic schema [Xur]nom./acc.pl., which has high cue and category validity for treatment as feminine. However, association of nd-stem referents with male biological sex most often impedes such treatment. By the same token, then, alignment with the minimally schematic schema [XæTur]nom./acc.pl., in addition to medially schematic [Xur]nom./acc.pl. –– both of which have high cue validity for assignment of feminine grammatical gender –– facilitates category validity for treatment as a member of the feminine Xó/æT-microclass specifically. Likewise, the interaction of non-linguistic factors such as schematicity with linguistic factors like semantics are at play in the rate of reanalysis: masculine plural fætur is neutral with regard to biological sex, aligns at both levels of abstraction and, therefore, is reanalysed as feminine at a far higher rate than any other masculine form in plural -ur. Continuing with the impact of domain-general cognition on the rate of reanalysis, Paper I demonstrates that the different rates reported are a function of prototype structure –– a property of categorisation that stems from statistical learning and schematisation. Indeed, as argued in 5.3.1, perception of structure in the world proves a function of the skewed frequencies with which sets of different attributes are integrated into phenomena considered similar or the same. Thus, while all masculine forms in plural -ur are discernible as feminine on alignment with the schema [Xur]nom./acc.pl. –– all other things being equal –– the schema lacks cue validity for specific class assignment. In this regard, [Xur]nom./acc.pl. serves as a basic-level category, whose level of abstraction is equivalent to that of the non-linguistic category FURNITURE, cf. e.g. basic CHAIR, as discussed in 3.2.4. However, as the structure of the subtype for feminines in plural -ur converges around the Xó/æT-microclass, the effect of prototype structure as a mechanism for categorisation 118 Conclusion reveals itself as a correlation of the degree of one-to-one correspondence between the schema [XæTur]nom./acc.pl. and plural fætur and vetur, on the one hand, and the rate of reanalysis, on the other. In other words, knowledge of the skewed frequencies with which prototypical features are dispersed among the members of a category is predicated on prior experience with language as stored in memory. Subsequently, this knowledge is extended to new contexts that are perceived as somehow similar to existing knowledge. This conclusion is considered a clear indication of the impact of non-linguistic factors such as frequency of use, dispersion, and schematicity, as well as processes such as statistical learning, analogy, and categorisation on the structure of grammar as a function of domain-general cognition. In a similar vein, Paper II demonstrates that prior experience with language is stored in memory, even when that experience pertains to low-frequency patterns associated with minimal schematicity. Further, the article indicates that such patterns can display –– albeit limited –– productivity, a property of language use that is at odds with the “either/or” approach that characterises traditional generative models and those that assume dualprocessing. Indeed, the fact that the usage-based cognitive approach assumes graded productivity as a property of language use demonstrates that the theoretical tenets of the approach are in line with the nature of observable language change. Therefore, it is submitted that the approach adopted in the current thesis is more suitable than generative and dualprocessing models at accurately accounting for change as a function of usage. Thus, Papers I and II show that productivity is a graded property of language dependent on mitigating linguistic and non-linguistic factors such as semantics, frequency, and schematicity. For example, the attractive force of the Xó/æT-microclass schemas are mediated by semantics and/or pragmatics. Consider the cline according to which forms of Ice. fótur, kók, and blók are treated as members of the microclass in question. While minimal schematicity understandably facilitates assignment of feminine grammatical gender to forms from all three paradigms, this non-linguistic property conspires with semantics in treatment of the borrowing blók as feminine only. In other words, due to the nature of the real-world referents of Ice. blók, i.e. the noun refers to lowly people of any gender, there is clear grounds for association of the schema [XóT]sg. with a given real-world function, i.e. form and function converge when the grammatical gender of a word correlates with the biological sex of its referent (PII:215–216). This association, it is argued, facilitates the assignment of feminine grammatical gender (linguistic) on the basis of reference to biological sex (non-linguistic). Conversely, the paradigm for borrowed kók manifests a significant degree of grammatical gender fluidity, by which semantics dictates association between specific forms 119 Jón Símon Markússon of the noun, on the one hand, and neuter and/or feminine grammatical gender, on the other. In light of this property of use, then, association between forms of the noun and their grammatical gender also accords with a cline, by which similarity to a given pragmatic context occasionally motivates use of fem. dat.sg. kók instead of neut. dat.sg. kóki (see PII:208–209). Furthest along the cline is masculine plural fætur, which is the only form of Ice. fótur to manifest use as feminine. It is noted in Section 5.2 that this property of use stems from the neutrality of its referent with regard to biological sex, in addition to the overtly masculine marking of most singular forms of the word. It is, therefore, clear that rich memory and the wealth of experience stored therein facilitates categorisation via schema. This domain-general cognitive process is considered the mechanism of language change, itself a function of analogical reasoning, which, in line with Blevins and Blevins (2009), is viewed as the core of all human cognition. The view of productivity as a graded property of language garners further support from the findings of Papers II and III, both of which account for levelling as a mechanism of analogy. Indeed, Paper II argues for graded membership of the Xó/æT-microclass as a function of limited productivity, which, in turn, is a property of low frequency and/or minimally schematic inflection classes. Thus, despite the common treatment of Ice. kók as neuter in the mass noun sense, the frequency with which the noun occurs with feminine modifiers in the count noun sense facilitates levelling of the alternation nom./acc.sg. kók ~ dat.sg. kóki to syncretic fem. nom./acc./dat.sg. kók, where nom./acc.sg. kók is considered basic (PII:209). Syncretism in nominative/accusative/dative singular is an exceptionless attribute of strong feminine nouns in Icelandic, knowledge of which is extracted via the process of statistical learning. The deduction of syncretic fem. nom./acc./dat.sg. kók suggests that gender assignment is a function of categorisation due to similarity of sg. kók to Xó/æTmicroclass members. Likewise, Paper III demonstrates that levelling is not “either/or” in nature but, rather, a graded process. Further, levelling is discernibly determined by those conditions in the world that form part of the language user’s experience. Indeed, use of individual forms of Far. vøllur and fjørður was shown not to be constrained by common semantics, cf. that both nouns have topographical referents and can occur as complex place names. Rather, Paper III argues that functionally different forms of the respective paradigms are chosen as basic and that the choice is determined by frequency, which, in turn, reflects language users’ perception of realworld conditions. Thus, through the centuries, Far. vøllur occurred less frequently in the dative than it does in modern times because its pre-20th century referents tended to be small 120 Conclusion and less suitable for performing established functions of the kind that some of its modern referents do. In other words, Far. vøllur was until recently used relatively infrequently in reference to the location of events on account of topographical features of the Faroe Islands. This, it is argued, is the reason that acc.sg. vøll, rather than dat.sg. velli, was viewed as basic before levelling commenced (PIII:76). Further, as the dispersion of the stem variant vøll- was higher among the singular portion of the paradigm, the combined token frequencies of nom.sg. vøllur and acc.sg. vøll facilitated the entrenchment of a form-function pairing through which the form vøll- expressed the function singular. Such a development is in line with the observable cross-linguistic tendency towards a relation between meaning and form (see Chapter 4). Likewise, the interaction of token frequency and dispersion is evident through levelling in Far. fjørður. There, dat.sg. firði is by far the most frequent form of the paradigm in Modern Faroese. Further, given that the noun has not innovated in terms of referents, Paper III argues that its modern-day frequency also proportionately reflects earlier stages of the language. In other words, dat.sg. firði is likely also the form that was most entrenched when levelling commenced and, therefore, basic within the paradigm. Should this be correct, we would expect dat.pl. firðum to be the earliest yield of levelling. That is, the most entrenched form of the paradigm is more likely to spread to other cells than forms with lower lexical strength. Additionally, we should expect innovative firðum to usurp the function of older dat.pl. fjørðum, at least in terms of token frequency, as the former contains the most frequent stem variant of the paradigm, which is closely associated with the schematic slot of the schema [í/úr FJØRÐURdat.]. Indeed, younger dat.pl. firðum is now the most frequent member of the relevant cell (PIII:69). Further, Paper III demonstrates that deduction and subsequent entrenchment of younger dat.pl. firðum established a relation between the stem variant firð- and the meaning plural, cf. nom./acc.pl. firðir (PIII:80). Paper III subsequently argues that, in reaction, language users established a relation between the stem variant fjørð- and the meaning singular, due to the 2:1 dispersion rate of that variant within the singular portion of the paradigm, cf. nom.sg. fjørður and acc.sg. fjørð. This dynamic, it is argued, yielded innovative dat.sg. fjørði as a function of the relation between meaning and form. However, frequency and its impact on cognition once again determines the outcome of reactionary change: Older dat.sg. firði is so entrenched as a function of its high token frequency that younger dat.sg. fjørði never acquires 121 Jón Símon Markússon the lexical strength needed to take over the function of the former. Indeed, this is supported by the low token frequency of the latter (PIII:69). In light of the objectives set out above, (1a)/(47a) has been satisfied. Thus, frequency and schematicity –– both non-linguistic properties of language use –– have been shown to impact domain-general cognitive processes such as statistical learning, entrenchment, and schematisation as applied in linguistic innovation through e.g. categorisation and schematisation. The latter processes feed the former forward as functions of analogy. In support of this assertion, it is possible to refer to the example provided by the cue and category validity of Ice. plural -ur for assignment of feminine gender and treatment according to that assignment. Further, the disparate rates at which masculine forms in plural -ur are reanalysed as feminine speaks to categorisation by schema, by which entrenchment facilitates the further and gradual elaboration of schemas (PI:11). As noted in 5.3.1, the perception of structure in the world is a function of the skewed frequencies with which different attributes occur among sets of items perceived as similar. Moreover, this function of skewed frequencies gives rise to prototype structure, itself a domain-independent property of categories. In this connection, Papers I–II demonstrate that linguistic categories are expanded via alignment of linguistic items with schemas at one or multiple levels of abstraction within the same taxonomy. In this connection, it has also been demonstrated that as the level of abstraction incrementally decreases within the relevant taxonomy, the degree of convergence between category members and a category prototype increases in graded fashion. This property of categorisation determines the rate at which Icelandic masculine forms in plural -ur are reanalysed as feminine. As this method of expanding categories has been shown to apply in non-linguistic categorisation also, it is concluded that the factors which determine the structure of grammar are domain-general in nature, cf. objective (1b)/(47b). In order to show that the objective in (1c)/(47c) has been satisfied, it is necessary to refer again to objectives (1a–b)/(47a–b). First, both linguistic and non-linguistic factors have been shown to impact the structure of the synchronic grammar. Secondly, these factors guide the domain-general cognitive processes that are applied in language use. Further, Papers I–III demonstrate that non-linguistic factors such as frequency guide the relevant domain-general processes in real time, during on-line language processing, as each instance of use elaborates and strengthens relevant memory representations for linguistic experience. Through this process, schemas are elaborated to include all new instances of use, thereby increasing the inclusive scope of the existing schema, i.e. its schematicity, and, simultaneously, its potential attractive force as a function of alignment. 122 Conclusion In light of the emergent nature of schemas and, by implication, the linguistic categories over which they abstract, analogy applied in real time is considered the mechanism of language use and, therefore, also of change subsequent to entrenchment of the relevant innovation in memory. Thus, without statistical learning, schemas would have no cue validity for class assignment, nor category validity for treatment according to that assignment. Were this the case, Icelandic masculine forms in plural -ur would not occasionally be treated as feminine due to the dispersion of the ending. Indeed, the same applies to the borrowings blók and kók in terms of alignment with the schemas for the Xó/æT-microclass. Further, without statistical learning, we should not expect to witness frequency effects in the direction of levelling, based on the choice of most frequent member of the paradigm as basic. However, such effects are observed time and again, cross-linguistically. It is also possible to posit entrenchment as a function of language use and vice versa. Indeed, positing this give-and-take relation between the two processes is fully in line with the tenet of the usage-based cognitive approach, i.e. that use determines the structure of grammar, while the structure of the grammar informs further use. Thus, it has been demonstrated that statistical learning feeds synchronically into the domain-general cognitive process of entrenchment, yielding ever stronger representation of the relevant experience in memory. On the basis of entrenched knowledge, assignment of items to schematically represented categories occurs. Should categorisation then facilitate reanalyses and/or inflection class assignment (PI–II), or levelling due to entrenchment of a stem variant within a particular context of use (PIII), then language change can be said to have occurred as a function of the domain-general cognitive processes that facilitate language use. Therefore, the current thesis considers objective (1c)/(48c) to have been met also. In conclusion, then, Papers I–III demonstrate individually that language change is predicated on the process of language use, experience of which is stored in memory and is accessible relative to the degree to which the relevant knowledge is entrenched. In light of this property of language, it stands to reason that the processes which drive the cycle of use and change are domain-general cognitive in nature, rather than attributable to innate cognitive machinery specific to language. It follows, then, that the processes in question are impacted by non-linguistic factors that are nonetheless witnessed through language use and change, such as frequency and its effect on the strength of memory representations for prior experience. Indeed, as argued throughout the current thesis, this property of human cognition is also evident through the application of domain-independent experience as utilised when discerning structure in the world as a function of analogical reasoning. 123 Paper I Paper I Accounting for different rates of gender reanalysis among Icela Accounting for different rates of gender reanalysis among Icelandic masculine forms in plural -ur doi:10.1017/S0332586522000166 125 Nordic Journal of Linguistics (2022), page 1 of 26 doi:10.1017/S0332586522000166 RESEARCH ARTICLE Accounting for different rates of gender reanalysis among Icelandic masculine forms in plural -ur Jón Símon Markússon Faculty of Icelandic and Comparative Cultural Studies, University of Iceland, Sæmundargata 2, 101 Reykjavík, Iceland Email: jsm2@hi.is (Received 24 November 2021; revised 4 September 2022; accepted 8 September 2022) Abstract This paper presents a usage-based cognitive approach to the different rates at which Icelandic masculine forms in nominative/accusative plural -ur are reanalysed as feminine. Of the 14.92% of nouns in plural -ur, 91.89% are feminine, others masculine. Syncretism in nominative/accusative plural is exceptionless among feminines, but relatively rare among masculines. Interestingly, plurals such as masculine eigendur ‘owners’, fætur ‘feet’, vetur ‘winters’ occasionally yield the feminine outputs definite eigendur-nar, fætur-nar, vetur-nar, and are sometimes modified by feminine forms of adjectives and determiners. As the full set of forms in plural -ur is highly schematic, we might expect reanalysis – viewed as a property of a schema’s productivity – to correlate proportionately with the frequency of corresponding masculine forms. However, corpus data for Icelandic betray a mismatch. Through a network model approach that emphasises prototype structure, minimal schematicity is shown to impact the rate of reanalysis by means of a gang effect. Keywords: frequency; gang effect; gender; Icelandic; inflectional morphology; productivity; prototype structure; reanalysis; schemas; usage-based 1. Introduction This article deals with different inflectional classes in Icelandic (Ice.) and the moderate PRODUCTIVITY found with some of these.1 Specifically, it examines the different rates at which masculine (MASC) forms in syncretic nominative/accusative plural (NA.PL) -ur are treated as feminine (FEM) due to varying degrees of phonetic and/or semantic similarity to clusters of feminines in the same ending. To account for the reanalyses that underlie this treatment, I employ the machinery of cognitive linguistics and its usage-based approaches to morphology, focusing on the relation between TYPE FREQUENCY, SCHEMATICITY, and productivity. Type frequency and schematicity are recognised determinants of productivity in the established usage-based cognitive literature. Type frequency is equated with the number of items that follow an inflectional pattern, while schematicity is defined by © The Author(s), 2022. Published by Cambridge University Press on behalf of The Nordic Association of Linguists. 2 Jón Símon Markússon the degree of phonetic dissimilarity between them (e.g. Bybee 2007:9, 2010:67). Productivity is measured as the rate at which a SCHEMA ‘attracts’ inflectional forms by ANALOGY, defined here as the process by which existing knowledge is extended to new contexts (e.g. Gentner 2005; Barðdal 2008:1, 9; Bybee 2010:57). Schemas are cognitive generalisations that specify the basic outline common to any number of items perceived as similar (Langacker 1987:132–135; Taylor 2003:67; Tuggy 2007:83). The process of schematisation is gradient, yielding formally continuous schemas in taxonomies of increasing abstraction. This gradience reflects the view that schemas take account of all formal and functional attributes common to a set of items at distinct levels of complexity (e.g. Bybee 2001:27; Audring 2019). Productivity is considered to proceed via reference to schemas. Inflection classes characterised by both high frequency and high schematicity generally prove the most productive, as these impose the least phonetic constraints on membership. However, language change demonstrates that productivity is also gradient: classes of all shapes and sizes show different rates of productivity cross-linguistically (Barðdal 2006, 2008). Low schematicity can mediate the constraints of low type frequency by means of a GANG EFFECT. In other words, while a low frequency class is unlikely to prove highly productive, it may show limited productivity if its members are phonetically coherent (Bybee 2010:69). A well-known example is the extension of the English strong schema [Xear]present ∼ [Xore]past,2 e.g. present bear ∼ past bore, swear ∼ swore, tear ∼ tore, to the paradigm of formerly weak wear, resulting in past wore (see Axelsdóttir 2015 and Markússon 2021, 2022 for examples of the gang effect in Icelandic and Faroese). In this connection, Icelandic exhibits significant correlation between the phonetic bases for inflection class membership and grammatical gender (Berg 2019), a factor generally acknowledged to determine the direction of inflection class shift and, often simultaneously, gender REANALYSIS (e.g. Bjorvand 1972, 1975; Bernharðsson 2004; also Ralli 2002 on Modern Greek). The objective here is to account for the different rates at which masculine forms in NA.PL -ur are reanalysed as feminine on account of this correlation, as betrayed chiefly by use of the feminine definite article (DEF) NA.PL -nar, instead of expected masculine N.PL -nir, A.PL -na, but also (less frequently) by agreement with feminine modifiers. Significantly, reanalysis associated with NA.PL -ur is all but exclusively masculine to feminine, a factor attributed to DISPERSION: the morphological contexts in which an exponent occurs (following Gries & Ellis 2015). Crucially, of the 14.92% of Icelandic nouns in plural -ur, 91.89% are feminine, with most belonging to the largest weak feminine class, e.g. N.SG stelpa ‘girl’ ∼ NA.PL stelpur (see Kvaran 2005:239).3 Conversely, just 8.11% in plural -ur are masculine (Svavarsdóttir 1993:105). Further, syncretism in the nominative/accusative plural is exceptionless among Icelandic feminine nouns but, by contrast, relatively rare among masculines. Therefore it is argued that masculine forms in NA.PL -ur pattern with the corresponding forms of feminines based on the ending alone, which has high CUE VALIDITY for treatment as feminine due to properties of its dispersion.4 It should be noted immediately that masculine forms in plural -ur are treated as feminine in the minority of instances. Based on searches conducted in the isTenTen corpus, which consists of sources including social media, plural fætur ‘feet’ is the Accounting for different rates of gender reanalysis 3 masculine form most frequently reanalysed at a rate of 22.82%.5 However, it is considered interesting that reanalysis occurs at all, given that corresponding forms in other endings hardly ever undergo gender reanalysis (see Þórhallsdóttir 1997, Bernharðsson 2004; also Jónsdóttir 1988–1989, 1993 on the endings NA.PL -ar and -ir, properties of their dispersion regarding gender, and schematicity). Additionally, it is argued that treatment as feminine is significantly influenced by the phonetic similarity that a masculine form bears to clusters of feminines beyond common plural -ur. For this reason, particular attention is paid to the relatively high rate at which masculine NA.PL fætur is attracted by the schemas for a small, phonetically coherent MICROCLASS of just six feminine nouns (see Dressler 2003:35). To demonstrate the impact of phonetic similarity, appeal is made to the NET EFFECT, illustrated as a NETWORK MODEL interpretation of the prototype structure of inflectional classes and resultantly varying degrees of cue validity for treatment as feminine. Section 2 provides a brief overview of Icelandic noun inflection and delineates the prototype structure of a feminine subclass in NA.PL -ur. Section 3 explicates the analogical processes that facilitate the microclass’s productivity. Section 4 reports on different rates of reanalysis for masculine forms in NA.PL -ur with reference to phonetic and semantic constraints imposed by individual schemas. Section 5 defines the net effect and accounts for the data presented in Section 4. Section 6 offers conclusions. 2. Prototype structure of a feminine subclass in NA.PL -ur This section delineates a subclass of strong feminines whose prototype structure centres around the inflection of six nouns: blók ‘non-entity, wretch’, bók ‘book’, bót ‘patch’, brók ‘trousers’, nót ‘(fishing) net’, and rót ‘root’, henceforth referred to as the Xó/æT-microclass. The notation ‘X’ abstracts over the various onset consonants and consonant clusters of the Xó/æT-microclass, although historically only b (bók, bót), bl- (blók), br- (brók), n- (nót), and r- (rót) occur. The notation ‘-ó/æ-’ references vocalic alternation between singular/dative and genitive plural blók-, bók-, bót-, brók-, nót-, rót- and nominative/accusative plural blækur, bækur, bætur, brækur, nætur rætur. Upper-case ‘T’ abstracts over the voiceless stops -t or -k in coda position, where this notation represents the historical fact that before the addition of borrowed blók, stem-final -t occurred in the majority of forms. Assignment of items to a minimally schematic microclass proceeds by analogy with the relevant subclass’s most prototypical schema(s) (see Barðdal 2006, 2008). Experiments by Rosch (1975) indicate that categorisation is a domain-general cognitive process facilitated by perceived similarity to a PROTOTYPE (also Rosch et al. 1976). The prototype is a schematically represented entity which abstracts over features common to all members of a category, though some members share more features with the prototype than others. Crucially, Lakoff (1987) has demonstrated that categorisation by prototype is evident through language use. It should be noted immediately that the current study is not the first to posit a minimally schematic microclass of Icelandic nouns centred around a prototypical phonetic structure. Knudsen (1967) posits the feminine microclass Ice. brík 4 Jón Símon Markússon ‘armrest’, flík ‘item of clothing’, tík ‘female dog, bitch’ on phonetic grounds, e.g. the rhyme sequence -ík, plural -íkur, i.e. plural bríkur, flíkur, tíkur.6 The feminine subclass that centres around the Xó/æT-microclass subsumes Knudsen’s microclass, as is fleshed out below. First, however, a brief overview of Icelandic noun inflection is in order, where members of the feminine subclass may serve as exemplary. Icelandic nouns are inflected for case – nominative, accusative, dative (D), genitive (G) – and number – singular (SG), plural. The inflected definite article is almost always suffixed to the inflected form of the noun. Icelandic nouns are masculine, feminine, or neuter (NEUT). While gender is of major relevance for agreement (Corbett 1991:105), gender classes in Icelandic – as in some other Germanic languages – can be viewed as macro inflectional classes subsuming several micro inflectional subclasses (Enger 2004; Kürschner & Nübling 2011). As noted in Section 1, developments in Icelandic have tended toward correlation between phonetic attributes and inflection class membership, where the morphophonological properties of paradigms often indicate and assist with gender assignment. This is demonstrated in (1) by the paradigms of strong FEM rót, nótt ‘night’, and geit ‘goat’, shown alongside the standard paradigm of MASC fótur (the masculine articles N.PL -nir and A.PL -na are in bold).7 (1) N.SG-DEF A.SG-DEF D.SG-DEF G.SG-DEF N.PL-DEF A.PL-DEF D.PL-DEF G.PL-DEF rót-in rót-ina rót-inni rótar-innar rætur-nar rætur-nar rótu(m)-num8 róta-nna nótt-in nótt-ina nótt-inni nætur-innar nætur-nar nætur-nar nóttu(m)-num nótta-nna geit-in geit-ina geit-inni geitar-innar geitur-nar geitur-nar geitu(m)-num geita-nna fótur-inn fót-inn fæti-num fótar-ins fætur-nir fætur-na fótu(m)-num fóta-nna As is evident from (1), masculine fótur shows a similar phonetic structure and inflection to nouns of the Xó/æT-microclass beyond the ending plural -ur, namely a genitive singular in -ar, where most masculines have -s, along with stem-final -t, and (mostly) parallel alternation between -ó- and -æ-. While Ice. rót, nótt, and geit belong to the broader subclass of feminines in plural -ur (Kvaran 2005:221; see Iversen 1972:68 for historical classification), the inflection of nótt and geit differs from that of rót in graded fashion. The prototypical nominative/accusative/dative singular of strong feminines is syncretic and monosyllabic (though see Þórhallsdóttir 2007). Among members of the feminine subclass, this form normally alternates with a disyllabic genitive singular in -ar and nominative/accusative plural in -(u)r, e.g. singular kind(ar) ‘sheep’ ∼ plural kindur, geit(ar) ∼ geitur, eik(ar, eikur) ‘oak’ ∼ eikur, flík(ar) ∼ flíkur. These occur with the feminine article NA.PL -nar, i.e. kindur-nar, geiturnar, etc. Additionally, many paradigms exhibit i-umlaut alternation, e.g. singular mörk (merkur/markar) ‘250 gr.’ ∼ plural merkur, kló(ar) ‘claw’ ∼ klær, rót(ar) ∼ rætur (Iversen 1972:17–20). Members of the Xó/æT-microclass have a monosyllabic nominative/accusative/ dative singular form with ó [ouː] as its nucleus and a coda in stem-final -t or -k. Similarly, the genitive singular ends in -ar, the nominative/accusative plural Accounting for different rates of gender reanalysis 5 in -ur. The microclass also exhibits i-umlaut alternation between singular and dative/ genitive plural -ó- [ouː] and nominative/accusative plural -æ- [aiː] (see above).9 Feminine nótt is not considered prototypical for the following reasons. The vowel of the singular is short [ou] due to the nature of its coda (see Árnason 2005:135 on vowel length and syllabification in Icelandic). Further, its genitive singular is syncretic with NA.PL nætur, as opposed to exhibiting alternation like rótar ∼ rætur. Despite this, nótt is considered closer to the prototype than other members of the wider feminine subclass, due to parallels in i-umlaut alternation: see (1). Less prototypical are feminine geit, eik, flík, for example. While the coda of each complies with the prototype, the respective nuclei consist of [eiː] (geit, eik) and [iː] (flík), neither of which engages in i-umlaut alternation, e.g. singular geit(ar) ∼ plural geitur, flík(ar) ∼ flíkur. Existence of G.SG eikur beside more prototypical eikar further distances eik from the prototype. Significantly, only stems in singular -ót, -ók take up the Xó/æT-microclass pattern. Note that e.g. NEUT flóð ‘flood’, FEM dós ‘can’, NEUT/FEM bón ‘polish/ request’ never alternate with plural *flæður, *dæsur, or *bænur, respectively, while borrowed FEM blók and NEUT/FEM kók ‘Coke™’ alternate with plural blækur and (occasionally) kækur (see Markússon 2022; also Section 3.1). These factors provide further justification for the prototypical structure of the feminine subclass as it centres around the Xó/æT-microclass. Consider in this connection Ice. NA.PL dætur (of dóttir ‘daughter’) and mæður (of móðir ‘mother’). While the stem of the former meets the phonological definition of the prototypical rhyme, the latter contains stem-final <ð> [ð], meaning that dætur is phonetically closer to the prototype than mæður. Further, dóttir and móðir likely form a more consistent microclass with masculine bróðir ‘brother’ (plural bræður), faðir ‘father’, and feminine systir ‘sister’, on morphophonological and semantic grounds (Iversen 1972:66–67; Kvaran 2005:406). 3. The gauge for minimal schematicity 3.1 The (limited) productivity of the Xó/æT-microclass Due to the acknowledged causal relation between type frequency and schematicity, on the one hand, and productivity, on the other (see Section 1), it perhaps comes as no surprise that the Xó/æT-microclass has shown highly limited productivity associated with two borrowings. According to www.timarit.is, a text corpus for written Icelandic from the early twentieth century onward, feminine plural blækur was first deduced from FEM blók in the 1920s; compare British English bloke (Magnússon 1989: s.v. blók). Subsequently, (humorous) feminine plural kækur was first deduced from NEUT/FEM kók around 1960. Markússon (2022) argues that the relations NAD(G).SG blók(ar) ∼ NA.PL(-DEF) blækur(-nar) and kók ∼ kækur(-nar) are unambiguously based on the model of e.g. bók(ar) ∼ bækur(-nar), as illustrated by the proportion in (2). (2) NAD(G).SG NAD(G).SG bók(ar) : blók(ar) : NA.PL(-DEF) NA.PL(-DEF) bækur(-nar) X; X = blækur(-nar) 6 Jón Símon Markússon The case is made in subsequent sections that most instances of (innovative) feminine plural definite fætur-nar, first recorded in the sixteenth century (Þórólfsson 1925:86), also results from analogy with the Xó/æT-microclass. The paradigm of masculine fótur from (1) is repeated in (3), this time with feminine plural definite forms in -nar (in bold). (3) MASC/FEM N.SG-DEF A.SG-DEF D.SG-DEF G.SG-DEF N.PL-DEF A.PL-DEF D.PL-DEF G.PL-DEF fótur-inn fót-inn fæti-num fótar-ins fætur-nir/fætur-nar fætur-na/fætur-nar fótu(m)-num fóta-nna As discussed in Section 2, Ice. fótur exhibits various points of similarity with e.g. prototypical rót. It seems, then, that the paradigm in (3) offers multiple points of phonetic alignment with the Xó/æT-microclass. 3.2 Fótur: masculine singular vs. (occasionally) feminine plural This section argues that treatment of masculine forms in plural -ur as feminine occurs independently of the rest of the paradigm. This view is based on the fact that singular forms of fótur are masculine only. To demonstrate that such treatment can be an expression of grammatical gender, as well as of inflection class membership (Enger 2004; Kürschner & Nübling 2011), syntactic contexts where plural (definite) fætur(-nar) agrees with feminine modifiers are considered. A search of the isTenTen corpus returned results for adjectival phrases such as feminine stórar fætur ‘big feet’, þreyttar fætur ‘tired feet’, blautar fætur ‘wet feet’, báðar fætur ‘both feet’, fjórar fætur ‘four feet’, and others. In such examples, the attributive adjective in -ar is the feminine nominative/accusative plural form. Corresponding masculine forms end in N.PL -ir and A.PL -a. Additional evidence comes from examples such as those in (4), where the form of both the article and modifiers is overtly feminine. (4) a. Þú ert með báð-ar fætur-nar á jörð-inni You-2.SG are with both-F.ACC.PL feet-DEF.F.ACC.PL on earth-DEF ‘You have both your feet on the ground.’ b. Ég sá ekki á henni fætur-nar, vel má vera að þær I saw not on her feet-DEF.F.ACC.PL well may be that they-F.NOM hafi verið máttlaus-ar have been powerless-F.NOM.PL ‘I didn’t see her feet but it may well be that they were useless.’ c. að hita fætur-nar sín-ar því þær eru to heat feet-DEF.F.ACC.PL her-ACC.PL.REFL because they-F.NOM are alltaf kald-ar always cold-F.NOM.PL ‘to heat her feet because they are always cold’ Accounting for different rates of gender reanalysis 7 Of the 1,274 examples of feminine plural fætur-nar contained in the isTenTen corpus, these agreed with feminine modifiers a total of 85 times, or in 6.67% of instances. While this amount is not highly significant, it is argued that use of feminine fætur-nar in 22.82% of all instances of the definite form suffices to demonstrate at least moderate treatment as feminine. Indeed, the nominative/accusative plural article is inescapably suggestive of grammatical gender irrespective of the form of the preceding inflectional ending: MASC N.PL -nir ∼ A.PL -na, e.g. MASC hestar-nir ∼ hesta-na ‘the horses’, gestir-nir ∼ gesti-na ‘the guests’, fætur-nir ∼ fætur-na; FEM NA.PL -nar, e.g. rætur-nar, myndir-nar ‘the pictures’, greinar-nar ‘the branches, articles’, therefore FEM NA.PL-DEF fætur-nar. Conversely, use of modifiers is not a compulsory property of any syntactic context. In other words, while it is true that both the form of the definite article and agreement provide evidence for the expression of grammatical gender, agreement in just 6.67% of cases does not constitute evidence against use of the plural definite article -nar as an expression of the same (see below). Concerning gender mismatches between different forms of the same word, such dynamics are admittedly rare in Germanic. However, Ice. fræði ‘field of study, (academic) subject(s), fields of study’ offers an example of such a dynamic. Plural forms of fræði are grammatically neuter, as the example in (5) demonstrates (from www.timarit.is). (5) Öll eiga þessi fræði hæli í All-NEUT.N.PL own these-NEUT.N.PL studies refuge in Háskóla-num, enda byggjast þau á university-DEF.MASC.D.SG since build-3.PL.PRES.MID they-NEUT.N on vísindaleg-um aðferð-um scientific-D.PL methods-D.PL ‘All these subjects belong at the university, since they are based on scientific methods.’ The forms Öll, þessi, and þau are neuter plural. Conversely, singular forms of fræði are treated as feminine, e.g. FEM málfræði ‘grammar’, lyfjafræði ‘pharmacology’, hagfræði ‘economics’, and would be referred to collectively as neuter plural fræði, e.g. NEUT Öll : : : þessi fræði in (5). Despite this dynamic, the distinction between the singular as feminine and the plural as neuter is not clear-cut. The nominative plural is considered a principal part in Icelandic. In other words, one of the roles of the nominative plural form, in conjunction with that of the nominative and genitive singular forms, is to indicate inflection class membership and therefore also grammatical gender (Kvaran 2005:221–222; see e.g. Finkel & Stump 2007 on principal parts). However, irrespective of gender, the dative plural of almost all nouns ends in -um, while the most common ending in the genitive plural is -a: see the paradigms in (1).10 Further, the inflection of all modifiers is characterised by syncretism of the dative plural form in all three genders. This applies to the genitive plural too. Therefore dative plural fræðum and genitive plural fræða, whether modified or not, can for all intents and purposes be assigned the same gender as NA.PL fræði. Conversely, gender cannot be assigned on the basis of the dative and genitive plural and, consequently, 8 Jón Símon Markússon these cannot impede gender reanalysis of the plural nominative and/or accusative form(s) of a given noun. Therefore it is also impossible to discern whether some instances of D.PL(-DEF) fótum(-num) or G.PL(-DEF) fóta(-nna) betray reanalysis as feminine. In other words, dative plural -um and genitive plural -a are the antithesis of principal parts for Icelandic nouns. Conversely, the nominative/accusative plural has high cue validity for treatment that accords with the expression of grammatical gender. The regularity of dative plural -um and genitive plural -a appears not to have exerted paradigmatic pressure on gender neutralisation in Icelandic, unlike in some Germanic languages. Indeed, the vocalic elements of the plural endings Ice. -ar, -ir, and -ur are still clearly distinguished (Árnason 2011:66–67). Therefore association of these phonetically distinct endings may account for the general rarity with which the relevant form(s) undergo gender reanalysis (although see Davidson 1990 on the loss of gender distinctions in Swedish, where plural -ar, -or, and -er are still clearly distinguished in the modern language; see also Section 4.2). With this in mind, a search of the isTenTen corpus for the overtly feminine forms A.SG *fót-ina, D.SG *fót-inni, and G.SG *fótar-innar returned no results. This is perhaps surprising considering that the indefinite singular accusative and genitive forms, i.e. fót and fótar, pattern perfectly with the Xó/æT-microclass, e.g. rót and rótar. Indeed, given that fótur is subject to treatment as feminine at all renders the forms in question perfect candidates for participation in the process. However, the cross-linguistic tendency towards a relation between meaning and form, as demonstrated by Bybee (e.g. 1985, 2015:106), may account for the resistance of A.SG fót and G.SG fótar to reanalysis. In this connection, the morphological structure of singular nominative fótur and dative fæti is highly suggestive of masculine grammatical gender. This is demonstrated by comparison with the singular forms of Ice. hestur ‘horse’, a typical masculine strong noun, shown in (6) (inflectional endings in bold). (6) N.SG A.SG D.SG G.SG hestur hest hesti hests Within the relational context that characterises paradigmatic structure, the endings singular -ur and -i are typical of strong masculines.11 Indeed, the same applies to the relation between N.SG -ur, A.SG -Ø, and D.SG -i as in hestur ∼ hest ∼ hesti; see also fótur ∼ fót ∼ fæti. Conversely, G.SG -ar is prototypical for strong feminines: see the paradigms of rót and geit in (1). Theoretically, this property of the ending’s dispersion perhaps renders genitive fótar more susceptible than accusative fót to reanalysis as feminine. However, occasional occurrence of innovative masculine D.SG-DEF fót(i)-num and G.SG-DEF fóts-ins, instead of standard fæti-num and fótar-ins, coupled with the nonoccurrence of overtly feminine singular forms, provides evidence that the singular tends to pattern with other strong masculines. Crucially, i-umlaut alternation in singular N fótur ∼ A fót ∼ D fæti in co-occurrence with a genitive singular in -ar has parallels in the inflection of a Accounting for different rates of gender reanalysis 9 subclass of Icelandic strong masculines (Kvaran 2005:229–230). The singular inflection of Ice. spónn ‘(wood) shaving, spinner, spoon’ and þáttur ‘act (in a play), programme, story/narrative’, fjörður ‘fjord’, and vörður ‘guard’ serve as exemplary for its individual microclasses (7).12 (7) N.SG A.SG D.SG G.SG spónn spón spæni/spóni spónar/spóns þáttur þátt þætti þáttar fjörður fjörð firði fjarðar vörður vörð verði varðar Some nouns that follow the patterns represented in (7) are very common. This is true of þáttur and fjörður, while vörður occurs in the compound dyravörður ‘bouncer’ and by itself. Others include köttur ‘cat’ and völlur ‘field’ (colloquially also ‘airport, sports field’), for example. Common personal names such as Björn, Hjörtur, Hörður, and Örn are also fully inflected according to the same patterns of alternation. The token frequencies of such items serve to entrench the patterns in question, a factor that has been demonstrated by Barðdal (2008:89–96) to contribute to productivity. The factors just listed account for the resistance that singular forms of Ice. fótur show to reanalysis as feminine. How, then, do we account for gender reanalysis of plural fætur in 22.82% of instances? In answer, this form does not pattern with any masculine class to any degree of specificity. Conversely, as argued above, plural dative and genitive fótum and fóta do not impede reanalysis of plural fætur as feminine, meaning that, for all intents and purposes, the plural patterns perfectly with the Xó/æT-microclass (see Section 2). 3.3 Reanalysis as a two-step process: A schematic approach Here I allow for the likelihood that treatment of plural fætur as feminine is motivated by phonetic and semantic alignment with any or all of feminine plural bætur, dætur, nætur, rætur, and/or the relevant schema. Positing this cluster as the trigger for gender reanalysis implies abstraction over its respective onsets. Indeed, a minimum of two items likely provides sufficient basis for generalisation (Bybee 2010:64). In light of this, it should surely also be possible for the emergent schema to encompass stem-final -k, e.g. plural blækur, bækur, brækur, kækur (Markússon 2022; see Ross & Makin 1999 on the compatibility of both exemplar and schematic approaches). Due to the modest degree of generalisation required to abstract schemas for the forms in question, these can be represented as [XæTur]NA.PL and [XæTur-nar]NA.PL-DEF, e.g. rætur ∼ rætur-nar, bækur ∼ bækur-nar (see Section 2 on the notation employed). Given the non-occurrence of feminine singular forms of fótur (see Section 3.2), it is unlikely that these provide the basis for the overtly feminine output fætur-nar. Therefore deduction is attributed to alignment with the SISTER SCHEMA [XæTur]NA.PL ∼ [XæTur-nar]NA.PL-DEF (following Booij & Audring 2018).13 It is argued that alignment of plural fætur with the constituent schema [XæTur]NA.PL triggers reanalysis as feminine, imbuing it with cue validity for alternation with feminine definite fætur-nar as depicted in (8) (see below on the notations employed). 10 Jón Símon Markússon (8) [XæTur ]NA.PL ∼ [XæTur-nar]NA.PL-DEF NA.PL rætur ∼ NA.PL-DEF rætur-nar ↓ NA.PL fætur ∼ NA.PL-DEF X; X = fætur-nar The opposition rætur : rætur-nar serves as an example of the grammatically feminine forms over which the sister schema [XæTur]NA.PL ∼ [XæTur-nar]NA.PL-DEF has been abstracted. On the basis of this knowledge, the grammatical attribute FEMININE is projected from [XæTur]NA.PL onto NA.PL fætur, as rendered by the symbol ↓ (see Gentner & Hoyos 2017:674–675 on projection). Subsequently, analogy facilitates phonetic and semantic alignment of the relevant forms of masculine fótur with those of the Xó/æT-microclass, i.e. the relation fætur ∼ (innovative FEM) fætur-nar.14 4. Schematicity and the rate of reanalysis Sections 4.1 and 4.2 argue for and then posit a taxonomy of formally distinct but functionally continuous schemas in line with Audring (2019), for example. The relation between constituent schemas is characterised by a hierarchy of abstraction: while all constituent schemas abstract over inflectional forms in plural -ur, the gauge for schematicity is the ability to attract masculine forms of varying phonetic structure in that ending. Section 4.2 presents different rates of reanalysis for masculine forms in plural -ur and accounts for these through appeal to schematicity and semantics. 4.1 A taxonomy of graded schematicity 4.1.1 The morphological status of plural -ur in NA.PL vetur, fingur Before a schema’s affiliation with plural forms in -ur can be established, it is necessary to account for the status of the sequence in different nouns. As noted in Section 3.2, the vast majority of strong masculines has a nominative singular in -ur, which is lost elsewhere in most paradigms: see those in (6) and (7).15 However, for a small number of strong masculines, -(u)r is present throughout the paradigm, indicating that the sequence belongs to the stem etymologically.16 The standard paradigms of strong masculine fótur, vetur ‘winter’, fingur ‘finger’, and weak masculine eigandi are shown in (9), where a hyphen indicates a morpheme boundary between the stem and plural -ur as an ending (this notation is employed throughout the remainder of the paper; see note 7). Stem-final -(u)r is in bold. (9) N.SG A.SG D.SG G.SG NA.PL D.PL G.PL fótur fót fæti fótar vetur vetur vetri vetrar fingur fingur fingri fingurs eigandi eiganda eiganda eiganda fæt-ur vetur fingur eigend-ur fótum vetrum fingrum eigendum fóta vetra fingra eigenda Accounting for different rates of gender reanalysis 11 As the paradigms for vetur and fingur show, inflectional endings occur after stemfinal -(u)r, accounting for the general interpretation of the sequence as part of the stem. This is not disputed here. However, on the basis that feminine plural definite forms such as fæt-ur-nar, vetur-nar, fingur-nar, and eigend-ur-nar occur, it is suggested that stem-final -ur in NA.PL vetur, fingur is occasionally reanalysed as the ending NA.PL -ur due to phonetic identity between the two sequences. Otherwise, the fact that masculine forms in plural -ur are frequently reanalysed as feminine irrespective of the sequence’s etymological status would be the result of pure coincidence. In other words, a graded rather than an all-or-nothing view of morphological structure appears to reflect more realistically what is happening in speakers’ minds in real time (e.g. Fertig 2013:8). 4.1.2 More on the dispersion of NA.PL -ur A detailed schematic account is provided below of the dispersion of plural -ur across masculine and feminine paradigms. Alternation between the plural nominative and accusative forms of the vast majority of Icelandic masculines can be represented schematically by the sister schema [X-V1r]N.PL ∼ [X-V1]A.PL. The notation ‘-V1-’ indicates phonetic identity between the vowels of the respective endings, i.e. N.PL -ar ∼ A.PL -a, -ir ∼ -i, respectively; see the plural inflection of Ice. hestur and gestur ‘guest’ in (10), where the values for the relevant cells are in bold. (10) N.PL A.PL D.PL G.PL hestar hesta hestum hesta gestir gesti gestum gesta Crucially, the alternation **N.PL -ur ∼ A.PL -u does not occur in Icelandic and therefore provides no basis for abstraction of the sister schemas just posited. In other words, as schemas are inseparable from the phenomena over which they abstract (Bybee 2001:27; Lakoff 2018:86–87), the notation [-V1-] represents the arbitrary subset {a, i}. The sister schema [X-V1r]N.PL ∼ [X-V1]A.PL is associated with masculine classes only. However, masculines in NA.PL -ur demonstrate that conformity to this relation is not a prerequisite for assignment of masculine grammatical gender. Indeed, this is supported by the fact that masculine forms in plural -ur are reanalysed as feminine in the minority of cases (see Sections 1 and 4.2). Conversely, syncretism in the nominative/accusative plural is an exceptionless attribute of Icelandic feminines, with the vast majority ending in NA.PL -Vr, e.g. myndir, greinar, stelpur; see also syncretic NA.PL kýr ‘cows’, mýs ‘mice’. Cognitive representation of this formal distinction between (the majority of) masculine and (all) feminine classes is necessarily highly abstract (see Janda 2002, 2007; also, below). However, at the physical level of language use, it is instantiated by masculine forms as in (11a), on the one hand, and feminine forms as in (11b), on the other. The relevant plural forms of masculine fótur, vetur, and fingur, which straddle the masculine–feminine border, are given in (11c). 12 Jón Símon Markússon (11) a. [X-V1r]N.PL ∼ [X-V1]A.PL b. [X-Vr]NA.PL c. [X-ur]NA.PL MASC FEM MASC/FEM hestar ∼ hesta, gestir ∼ gesti myndir, greinar, stelpur, rætur fæt-ur, vetur, fingur The schema [X-Vr]NA.PL in (11b) and its daughter [X-ur]NA.PL in (11c) are chiefly associated with feminine classes. The former offers a tried and tested means of deducing forms in N.PL -ir, -ar, and -ur from A.PL -ir, -ar, and ur, respectively, and vice versa. Therefore properties of its dispersion should not facilitate projection of masculine gender on alignment. On the contrary, we should expect alignment of masculine forms in plural -ur with [X-Vr]NA.PL or any daughter to facilitate reanalysis as feminine (see Section 3.3). 4.1.3 Positing a taxonomy of increasingly abstract schemas This section posits a taxonomy of formally distinct but functionally continuous schemas for forms in nominative/accusative plural -ur, as rendered in (12). (12) Highly schematic X-Vr myndir, greinar, stelp-ur, ræt-ur, fæt-ur Medially schematic X-ur stelp-ur, vetur, ræt-ur, fæt-ur Minimally schematic X æ T-ur ræt-ur, fæt-ur Note that the inherent structure of the taxonomy is reminiscent of Albright’s (2002, 2008, 2009) conception of form-to-form mapping ‘rules’ of varying specificity (also Albright & Hayes 2003). The schema [XæT-ur]NA.PL, posited for the relevant forms of the Xó/æT microclass in Section 3.3, imposes highly specific phonetic constraints on alignment and consequently productivity (see Sections 2 and 3.1). Therefore, in line with Audring’s (2019) terminology, it is posited as a daughter instantiation of medially schematic [X-ur]NA.PL, which, in turn, is a phonetically more specific instantiation of its own mother schema, i.e. highly schematic [X-Vr]NA.PL: see (11b–c). 4.2 The interaction of form and meaning as a determinant of reanalysis This section presents different rates of reanalysis for masculine forms in plural -ur and accounts for these through appeal to the interaction of form and meaning. As mentioned in Section 4.1.2, syncretism in the nominative/accusative plural is an exceptionless attribute of Icelandic feminine nouns. However, if syncretism in plural -ur alone accounted for the reanalysis of all relevant masculine forms, we might expect to see proportionate rates of descending frequency between innovative feminine outputs in NA.PL-DEF -nar and corresponding masculine forms. In fact, corpus data reveal significant mismatches across the two sets. This section focuses on token frequencies for masculine and feminine definite forms of 17 masculine nouns in plural -ur and what these reveal about different rates of gender reanalysis: see the list in (13). (13) áhorfandi ‘spectator, onlooker, PL audience’, áskrifandi ‘subscriber’, bróðir, byrjandi ‘beginner’, bóndi ‘farmer’, eigandi, faðir, fjandi ‘devil, PL enemies’, fingur, fótur, hlustandi ‘listener, PL audience’, kaupandi ‘buyer’, leigjandi ‘tenant, lodger’, lesandi ‘reader’, nemandi ‘student, pupil’, notandi ‘user’, and vetur. Accounting for different rates of gender reanalysis 13 Masculine bróðir and faðir belong to a small class of familial referents, also containing feminine dóttir, móðir, and systir (see Section 2). The five nouns in question are classified as r-stems in the historical–comparative literature and are considered a subclass of the broader inflectional category of consonant stems. Importantly, while assignment of these nouns to a single class is predicated on common semantics and inflectional similarity (see Iversen 1972:66–67), it appears futile to posit any kind of semantic link between class members as a whole and biological gender that might either facilitate or impede the rate of reanalysis. The nouns MASC fingur, fótur, and vetur are also classified by Iversen (1972:67– 68) within the broader class of consonant stems, as ‘other consonant stems’ (Norwegian: ‘andre konsonantstammer’). Iversen bases his classification on the fact that inflection of the three nouns accords for the most part with that of masculine ustems in the singular: see the paradigms in (7). However, they diverge from this point of similarity in the plural, where the relevant nominative/accusative forms show the masculine consonant stem ending/have stem-final -r (> Modern Icelandic -ur) instead of u-stem N.PL -ir, A.PL -i: see the relevant paradigms in (9). Interestingly, while fótur and fingur likely belong to a small – mostly feminine – semantic class of body part referents, fingur and vetur patterned together in both the singular and plural in terms of shared inflectional attributes, to the exclusion of fótur (see below). However, unlike fótur and vetur, Ice. fingur had the variant endings G.SG -s and -ar, i.e. fingurs and fingrar, in Old West-Nordic. In Modern Icelandic, only the s-variant occurs, fully distinguishing the inflectional attributes of the paradigms for fingur, fótur, and vetur one from another. Masculine nouns containing the sequence -nd- and ending in N.SG -i, e.g. eigandi, are referred to in the historical–comparative literature as nd-stems (e.g. Iversen 1972:66). As this etymological label is still descriptive for Modern Icelandic, it will also be used here. The nd-suffix derives agentive nouns from verbs. For example, Ice. eigandi is divisible as eig-and-i, where the verb eiga ‘own, have’ serves as root and -i is the nominative singular ending for weak masculines: see the relevant paradigm in (9). The nd-stems listed in (13) were chosen at random as representative of the largest and, likely, the only schematically definable masculine class in plural -ur. In other words, the form–meaning pairing [VERBAL ROOT-(e)nd-ur]NA.PL is associated with deverbal agentive masculine nouns and therefore provides a gauge for the extent to which that association might impede gender reanalysis. In Table 1 the token frequencies of the masculine and the feminine plural nominative/accusative definite forms are added together for each of the nouns in (13) (TOTAL). The number of feminine forms is then calculated as a percentage of that total to determine the rate of reanalysis for the plural nominative/accusative indefinite form.17 Forms are ordered according to the rate of reanalysis, from highest to lowest, while those forms that do not undergo reanalysis as feminine are listed last and ordered alphabetically. The data presented in Table 1 are based on search results from the isTenTen corpus. Table 1 demonstrates that plural fæt-ur, fingur, vetur, lesend-ur, and áhorfend-ur are all reanalysed as feminine at a rate above 5%. Significantly, plural fæt-ur is most frequently reanalysed at a rate of 22.82%, plural fingur in second place at 15.12%, and vetur third at 9.74%. The nd-stems lesend-ur and áhorfend-ur are reanalysed at 14 Jón Símon Markússon Table 1. Rates of reanalysis for masculine forms in NA.PL -ur MASC N.PL-DEF MASC A.PL-DEF FEM NA.PL-DEF TOTAL Rate of reanalysis fæt-ur-nir 1,426 fæt-ur-na 2,882 fæt-ur-nar 1,274 5,582 fæt-ur 22.82% fingur-nir 555 fingur-na 1,044 fingur-nar 285 1,884 fingur 15.12% vetur-nir 245 vetur-na 6,913 vetur-nar 773 7,931 vetur 9.74% lesend-ur-nir 113 lesend-ur-na 94 lesend-ur-nar 12 219 lesend-ur 5.47% áhorfend-ur-nir 280 áhorfend-ur-na 117 áhorfend-ur-nar 21 418 áhorfend-ur 5.02% notend-ur-nir 150 notend-ur-na 67 notend-ur-nar 7 224 notend-ur 3.12% bænd-ur-nir 1,117 bænd-ur-na 199 bænd-ur-nar 39 1,355 bænd-ur 2.87% eigend-ur-nir 1,125 eigend-ur-na 179 eigend-ur-nar 38 1,342 eigend-ur 2.83% nemend-ur-nir 3,088 nemend-ur-na 272 nemend-ur-nar 63 3,878 nemend-ur 1.62% bræður-nir 4,267 bræður-na 678 bræður-nar 51 4,996 bræður 1.02% áskrifend-ur-nir 11 áskrifend-ur-na 6 áskrifend-ur-nar 0 17 áskrifend-ur 0% byrjend-ur-nir 24 byrjend-ur-na 6 byrjend-ur-nar 0 30 byrjend-ur 0% feður-nir 226 feður-na 37 feður-nar 0 fjend-ur-nir 5 fjend-ur-na 16 fjend-ur-nar 0 21 fjend-ur 0% hlustend-ur-nir 12 hlustend-ur-na 11 hlustend-ur-nar 0 23 hlustend-ur 0% kaupend-ur-nir 192 kaupend-ur-na 33 kaupend-ur-nar 0 225 kaupend-ur 0% leigjend-ur-nir 88 leigjend-ur-na 21 leigjend-ur-nar 0 109 leigjend-ur 0% 263 feður 0% rates of 5.47% and 5.02%, respectively. Conversely, plural bræður and nd-stem notend-ur, bænd-ur, eigend-ur, and nemend-ur undergo reanalysis at rates of 1.02%–3.12%, while plural feður and nd-stems áskrifend-ur, byrjend-ur, fjend-ur, hlustend-ur, kaupend-ur, and leigjend-ur are not subject to reanalysis at all.18 While a top rate of reanalysis at 22.82% may seem only fairly significant on its own, it is considered highly significant when compared with a majority rate of 0– 3.12%. Further, disparities in the rate of reanalysis are reflected by the frequency with which some forms in -nar agree with feminine modifiers. For example, NA.PL-DEF fæt-ur-nar and áhorfend-ur-nar occur with feminine modifiers in 6.67% and 8.33% of cases, respctively, while those below 3.12% do not occur with feminine modifiers at all. Conversely, Table 1 suggests no correlation between descending token frequency and the rates of reanalysis reported (though see Barðdal 2008:89–96). For example, plural definite forms of fótur amount to 5,582 tokens, betraying reanalysis at 22.82%, compared with 4,996 tokens for vetur, with reanalysis at 9.74%, and 1,884 tokens for fingur with reanalysis in 15.12% of instances. These facts are taken to indicate that schematicity and semantics determine the rate of reanalysis. This dynamic is not surprising, given the strong association of the schema [VERBAL ROOT-(e)nd-ur]NA.PL with masculine agentive nouns (see above). Accounting for different rates of gender reanalysis 15 Further, as schemas abstract over semantic function (see Sections 1 and 4.1.2), many nd-stems will have mainly referred to human males until modern times due to the more stringent prescription of traditional gender roles in past centuries. In other words, it may well be that the agency expressed by many nd-stems was mostly associated with male biological gender for a significant period of Icelandic language history. Such supposition is perhaps substantiated by regular use of Ice. bóndakona ‘housewife at a farm’ (lit. ‘farmer’s woman’) since at least the fourteenth century, suggesting that the referent of bóndi has typically been associated with men (although this is likely changing). Similarly, association of Ice. fjandi with both the Christian concept of Satan and human enemies, who would typically have been male in the context of war, may account for the security the word exhibits in its traditional grammatical gender: see Table 1. Likewise, that Ice. bróðir refers exclusively to people of male gender likely accounts for its relatively very low rate of reanalysis despite a high degree of formal similarity to corresponding forms of Ice. móðir, e.g. bræður, mæður, with the same applying to the semantics of faðir. Reanalysis of plural áhorfend-ur at 5.02% may appear to contradict the above assertions to a degree. However, all things are likely not equal, as corpus data reveal that singular forms of áhorfandi are 91.96% less frequent than those of the plural. Based on this disparity, it is conceivable that the plural has gained a significant degree of autonomy from the rest of the paradigm (e.g. Bybee 2015:104), perhaps rendering it less likely to trigger relational links with the overtly masculine singular than are the plural forms of other nd-stems. A similar interpretation likely holds for the nd-stem lesandi, whose singular forms are 75.33% less frequent than those of the plural according to the same corpus. As noted above, plural fæt-ur is most frequently reanalysed as feminine, at a rate of 22.82%. Next, plural fingur and vetur are reanalysed in 15.12% and 9.74% of instances, respectively. In this connection, the notion of ‘all things being equal’ proves highly relevant. Relatively frequent reanalysis of plural fingur can no doubt in part be accounted for by its semantic link to a feminine cluster containing (plurale tantum) herðar ‘shoulders’, hönd ‘hand’, löpp ‘paw, leg’, tá ‘toe’: these refer to body parts that typically come in pairs or ten. However, in terms of schematicity, it is important to note that of the feminine forms just listed, only plural hend-ur shares the sequence -ur with plural fingur. For this reason, while fingur, herðar, hönd, löpp, and tá certainly form a semantic class, it can be argued that plural -ur has only a modest association with the vocabulary for body parts. This situation is in juxtaposition to that of a small Icelandic inflection class whose morphophonological attributes are highly associated with body parts, many of which come in pairs. The class in question contains weak neuters including auga ‘eye’, eista ‘testicle’, eyra ‘ear’, lunga ‘lung’, nýra ‘kidney’, which typically denote a pair. Others include hjarta ‘heart’, milta ‘spleen’, vélinda ‘oesophagus’, which refer to single organs, some not associated with the body, e.g. bjúga ‘(smoked) sausage’, and a scattering of loans, e.g. drama ‘drama’, paradigma ‘paradigm’, pasta ‘pasta’. Crucially, only this small class shows the ending NA.PL -u, e.g. plural augu, eistu, eyru, hjörtu, etc.19 16 Jón Símon Markússon Axelsdóttir (2015) argues that the body-part weak neuters attracted vélinda and milta to the class on semantic grounds.20 Further, she accounts for an innovative nominative/accusative plural form of Ice. hjalt ‘pommel, cross-guard’, i.e. plural hjöltu, vs. older hjölt, via reference to both phonetic similarity to hjarta, plural hjörtu, and semantic similarity with referents that come in pairs. In specific terms, the meaning of Ice. Hjalt, like the lexical gang of weak neuter pair words, is characterised by duality. Specifically, Old Icelandic plural hjǫlt (> Ice. hjölt) referred to both a sword’s pommel and the cross-guard, i.e. to the two extremities of its hilt. Since the Old Icelandic period, hjölt has also been used in reference to the cross-guard alone, i.e. to the part of the hilt between the blade and the grip. Significantly, the crossguard lies right-angled and points in two, i.e. opposing, directions. This is taken as further evidence for the role of schematicity and semantics in the productivity of schemas (see above). When compared with the clear link between the phonetic structure, semantics, and – albeit limited – productivity of the weak neuter class, phonetic similarity of masculine plural fingur to the feminine lexical gang is weak. In other words, the influence of plural herðar, lappir, and tær as conducive to reanalysis of plural fingur is semantic only. But what of the different rates of reanalysis for plural fingur and fæt-ur in light of shared semantics? Semantic association with the same subclass of feminines likely accounts for the reanalysis of masculine plural fæt-ur to a similar degree to that of plural fingur. However, it can be argued that the relative disparity between the rates of reanalysis betrays mismatched formal links to the feminine subclass and other classes, there among the feminine Xó/æT-microclass (see Section 5.2). Further, it is not inconceivable that masculine fótur and fingur exert semantic influence over one another. In this context, all things are likely not equal. Heavy reliance on semantics alone is rendered still less credible by the practically non-existent rate at which plural definite forms of masculine handleggur ‘arm’ and fótleggur ‘leg’, which are clearly semantically related to fingur, fótur, herðar, hönd, löpp, and tá, undergo reanalysis as feminine. While the accusative plural of both words ends in -i, i.e. handleggi, fótleggi, the nominative plural has -ir; see the plural inflection of masculine gestur in (10). Significantly, the dispersion of N(A).PL -ir is high among feminine nouns. Therefore, on the premise that all things are equal, we might expect a rate of reanalysis for plural handleggir and fótleggir as feminine similar to roughly half that for plural fingur or even fæt-ur.21 However, according to the isTenTen corpus, neither undergoes reanalysis, suggesting that all things are not equal.22 In light of the above, the rates of reanalysis for plural fæt-ur and vetur are considered most interesting for two reasons. First, as is also true of Ice. fingur, neither paradigm patterns fully with an established class defined in terms of grammatical gender, shared semantics, and/or phonetic similarity. Secondly, while the masculine plural definite forms of fótur are 42.31% less frequent than those of vetur, feminine vetur-nar is 77.62% less frequent than feminine fæt-ur-nar. This disparity suggests rather profoundly that all things are not equal and colours the analysis presented in the next section. Accounting for different rates of gender reanalysis 17 5. The net effect Given that the interaction of form and meaning impacts the rate at which masculine forms in plural -ur are reanalysed as feminine (see Section 4.2), the question in (14) arises. (14) How do we account for the frequency relation between FEM NA.PL-DEF fæt-ur-nar and vetur-nar, on the one hand, and MASC N/A.PL-DEF fæt-ur-nir/-na and vetur-nir/-na, on the other, through reference to schematicity? In Section 5.2, the analysis presented attempts to provide answers to this question via reference to the net effect, which is first defined in Section 5.1. 5.1 Defining the net effect The term ‘net effect’ is an intentional reference to Bybee’s network model (e.g. 1985, 2001, 2010), which renders emergent grammatical and/or semantic function across otherwise distinct phonetic contexts via a network of connecting lines. The network in (15) renders the emergent meanings of the English prefix un- and suffix -able in unbelievable through alignment with readable, washable, unbelievable and unwarranted, unattractive, unbelievable. (15) readable washable unbelievable believe unattractive unwarranted (from Bybee 2010:23) Similarly, the meaning of the English past suffix -(e)d emerges through alignment with other past forms, e.g. played, spilled, spoiled, banned, and rammed, as in (16). 18 Jón Símon Markússon (from Bybee 2010:23) As is evident from (15) and (16), the network model renders phonetic and semantic similarity at the level of individual words, affixes, and segments. It is this dexterous property of the model that makes it highly suitable for capturing the ‘net effect’, defined as the likelihood that an inflectional form should escape alignment with a schema. Crucially, cue validity for alternation with an overtly feminine form in the plural article -nar correlates with the extent of alignment between individual segments and/or sequences. In Section 5.2, I employ innovative notation to convey this correlation. 5.2 Delineating the net effect The descending token frequencies between masculine vetur-nir/-na and fæt-ur-nir/na, on the one hand, and feminine fæt-ur-nar and vetur-nar, on the other, are representative of the impact of schematicity on reanalysis. The different rates of reanalysis reported in Section 4.2 are attributed below to the extent of alignment between a masculine form in NA.PL -ur and two formally distinct but functionally continuous sister schemas, i.e. medially schematic [X-ur]NA.PL ∼ [X-ur-nar]NA.PL-DEF and minimally schematic [XæT-ur]NA.PL ∼ [XæT-ur-nar]NA.PL-DEF. It is argued that alignment at the minimally schematic level facilitates a gang effect and, as a result, a boost to the rate of reanalysis. Medially schematic [X-ur]NA.PL is able to attract masculine plural forms of any phonetic structure beyond NA.PL -ur (see Sections 4.1.1, 4.1.2, and 4.1.3). However, due to the degree of generality associated with [X-ur]NA.PL, alignment does not assign forms to a specific feminine class, as it abstracts away from several schematically distinct classes (see Sections 1 and 2). Therefore it is perhaps understandable that masculine plural fæt-ur, vetur, and fingur, whose plural forms are formally ambiguous with regard to grammatical gender, are more readily reanalysed as feminine than the plural forms of most nd-stems (see Sections 4.1.2 and 4.2). As stated in Section 1, reanalysis is considered a property of productivity. Therefore, considering the acknowledged causal relation between type frequency, schematicity, and productivity (Barðdal 2008), we might expect schemas for the Xó/æT-microclass to be unproductive (in Dressler’s 2003 sense). However, it is likely that a minimum of two items suffices for the abstraction of a schema (see Section 3.3). Accounting for different rates of gender reanalysis 19 By the same token, we might expect to attribute a higher rate of reanalysis to alignment with [X-ur]NA.PL. The different frequency relations between the masculine and feminine sets mentioned above appear, then, to counter the view that productivity correlates with type frequency and schematicity – on the assumption that all things are equal (see Section 4.2). On the contrary, however, the higher rate at which plural fæt-ur is reanalysed as feminine compared to any masculine form in NA.PL -ur is considered a consequence of full alignment with both medially schematic [X-ur]NA.PL and minimally schematic [XæT-ur]NA.PL. As noted in Section 1, the gang effect is facilitated by a significant degree of phonetic similarity between a set of items. The network in (17) illustrates graded similarity between NA.PL vetur, geit-ur, fæt-ur, and ræt-ur, as these converge incrementally on alignment with minimally schematic [XæT-ur]NA.PL. Cue validity for a plural definite form in -nar correlates with the number and thickness of connecting lines according to prototype structure: see Section 2. The notations employed in (17) are explained below. (17) [X - u r-n a r]NA.PL-DEF [X - u r]NA.PL v e t v e t u r-n a r ur g ei t g ei t u r-n a r ur f æ t f æ t u r-n a r ur r æ t r æ t u r- n a r ur [X æ T [X æ T - u r-n a r]NA.PL.DEF - u r]NA.PL Thin connecting lines render arbitrary phonetic similarity, which alone is void of cue validity, i.e. connections between identical segments across forms of the same word and between instances of stem-final -t in vetur and geit-ur. Note that stem-final -t occurs in all of the forms in (17), irrespective of prototypical status within the feminine subclass. Therefore one-to-one connections between instances of stem-final -t only illustrate graded convergence with [XæT-ur(-nar)]NA.PL(-DEF) in 20 Jón Símon Markússon co-occurrence with other prototypical attributes of the Xó/æT-microclass, i.e. -t-ur or -æt-ur (see below). Phonetic similarity that correlates unambiguously with grammatical and/or semantic function is depicted by thick bold lines, i.e. those between individual instances of nominative/accusative plural -ur, between instances of the feminine plural article -nar, and between instances of -æ-, which is an indication of plural in some stems, e.g. (rót ∼) rætur. Arrow-headed lines convey the cue validity of plural -ur for alternation with a form in the feminine plural article -nar. Functional matches that emerge as a property of graded similarity are represented by thinner bold lines. According to the notation employed in (17), then, the stem-vowel e of plural vetur aligns only tentatively with the Xó/æT-microclass prototype. However, consider the connection between e [ϵ] in vetur and ei [ei] in geit-ur, on the one hand, and that between ei in the latter and the æ [ai] of fæt-ur, on the other. It can be argued that perceived phonetic similarity between [ϵ] and [ei] – the former is front and unrounded and so is the initial element of the latter – may suffice to align plural vetur within the network through similarity of [ei] in plural geit-ur to the [ai] of fæt-ur. Concerning the status of plural geit-ur within the prototype structure, stem-final -t may bolster the form’s position on account of co-occurrence with NA.PL -ur and an [i]-final diphthong (see Section 2). In other words, gradient similarity of the kind NA.PL [vεt]-ur ∼ [keit]-ur - [keit]-ur ∼ [fait]-ur may facilitate alignment of plural vetur with minimally schematic [XæT-ur]NA.PL to some degree, in addition to full alignment with medially schematic [X-ur]NA.PL. In terms of the rate of reanalysis as a correlate of cue validity, there are seven connections between each of the relations in (17). However, the one-to-one connections between plural vetur, geit-ur, and fæt-ur are weaker than those between plural fæt-ur and ræt-ur. This is illustrated by the seven thick bold lines that connect the latter pair, compared with four such connections between plural fæt-ur and geit-ur, while merely three connect geit-ur to vetur. Thus the network illustrates that a form’s cue validity for alternation with another in FEM NA.PL-DEF -nar is proportionate to the degree of formal and functional alignment at the minimally schematic level. Due to the nature of form–function connections between instances of NA.PL -ur, minimally schematic [XæT-ur]NA.PL fulfils the function of both schemas. This is because the sequence NA.PL -ur among feminines is always associated with this same function, irrespective of affiliation with either schema. Conversely, alignment at the minimally schematic level establishes additional points of similarity that strengthen the network on convergence with [XæT-ur]NA.PL and therefore its cue validity for alternation of the kind [XæT-ur]NA.PL ∼ [XæT-ur-nar]NA.PL-DEF. In other words, alignment with minimally schematic [XæT-ur]NA.PL constitutes default alignment with [X-ur]NA.PL. In answer to (14), masculine plural fæt-ur has higher cue validity for use as feminine than any masculine form in NA.PL -ur due to the number and extent of one-to-one form–function connections with [XæT-ur]NA.PL, in addition to alignment with medially schematic [X-ur]NA.PL. Therefore the different rates of reanalysis reported in Table 1 are considered a function of minimal schematicity, which facilitates the gang effect of the Xó/æT-microclass. Accounting for different rates of gender reanalysis 21 6. Concluding remarks The current article has sought to offer a usage-based cognitive account of the different rates at which Icelandic masculine forms in NA.PL -ur, which has a 91.89% dispersion rate among feminine nouns, are treated as feminine. Such treatment is betrayed by the occurrence of traditionally masculine forms in plural -ur with the feminine article NA.PL -nar, e.g. fæt-ur-nar, vetur-nar, eigend-ur-nar. The mechanism for such treatment is gender reanalysis, attributed to the high dispersion of NA.PL -ur among feminines. In Section 2, the prototype structure of a subclass of Icelandic feminines in plural -ur was delineated. The subclass’s prototype was argued to centre on phonetically coherent FEM blók, bók, bót, brók, nót, and rót, with plural blæk-ur, bæk-ur, bæt-ur, bræk-ur, næt-ur, and ræt-ur, referred to collectively as the Xó/æT-microclass. Section 3 detailed the highly limited productivity of the Xó/æT-microclass schemas, which have been extended to three paradigms only. The cue validity of native masculine NA.PL fæt-ur for occasional alternation with feminine NA.PL-DEF fæt-ur-nar was considered a function of proportional analogy with the relevant schemas. As innovation based on the Xó/æT-microclass is not associated with any other Icelandic paradigms, it is considered the gauge for limited productivity based on a minimally schematic, low type frequency inflectional pattern in NA.PL -ur. In Section 4, the interaction of a schema’s type frequency and its association with a semantically determined class of nouns was determined to affect the rate of reanalysis. For example, the Icelandic nd-stems, which constitute the largest single masculine class in NA.PL -ur, are identified as overtly masculine through their agentive referents in unison with the sequence NA.PL -(e)nd-ur, e.g. eig-end-ur. These formal and functional properties of the class account for reanalysis at a majority rate of 0–3.12%, as reported in Section 4.2. In light of these data, the treatment of plural fæt-ur and vetur as feminine was considered most interesting. First, unlike the nd-stems, neither belongs to an easily definable class on both phonetic and semantic grounds. Secondly, while masculine N/A.PL-DEF vet-ur-nir/-na is far more frequent than masculine fæt-ur-nir/-na, feminine fæt-ur-nar is almost twice as frequent as feminine vetur-nar. In order to demonstrate prototype effects on the rate of reanalysis as a means of explaining this inverse disparity, Section 5 employed notational conventions based on Bybee’s network model with some innovative features. The network illustrated varying degrees of cue validity inherent to different masculine forms in plural -ur for alternation with a feminine definite form in -nar. The degree of cue validity was rendered as a network of connecting lines, which, depending on number and thickness, modelled the extent of phonetic and/or functional alignment with distinct schemas. Illustration of alignment by the means just described demonstrated the net effect of schematicity, one function of which is that NA.PL vetur, fæt-ur – and any other masculine plural form in -ur – align with medially schematic [X-ur]NA.PL on account of one-to-one form–function connections between instances of NA.PL -ur alone. Reanalysis occurs once alignment facilitates projection of the grammatical attribute FEMININE onto a masculine in plural -ur. As the schema [X-ur]NA.PL has cue validity for alternation with its sister [X-ur-nar]NA.PL-DEF, alignment facilitates alternation such as NA.PL vetur, fæt-ur ∼ FEM NA.PL-DEF vetur-nar, fæt-ur-nar. 22 Jón Símon Markússon Additionally, however, the schema [XæT-ur]NA.PL attracts plural fæt-ur – and possibly some instances of plural vetur – at the minimally abstract level on account of one-to-one form–function connections beyond NA.PL -ur. Subsequently, reference to the sister schema [XæT-ur]NA.PL ∼ [XæT-ur-nar]na.pl-def, in addition to [X-ur]NA.PL ∼ [X-ur-nar]NA.PL-DEF, provides a relative boost to the cue validity of (now feminine) plural fæt-ur for alternation with feminine fæt-ur-nar. In conclusion, the different rates of reanalysis highlighted are predicated on the degree to which a masculine form in NA.PL -ur aligns with schemas at both the medially and the minimally abstract levels. Therefore the likelihood that a masculine form in NA.PL -ur will escape reanalysis as feminine is viewed as a correlate of minimal schematicity. It is hoped that the current study will motivate further corpus-based study of limited productivity in Icelandic and other inflectional systems in line with the usage-based cognitive approach to linguistic innovation. Acknowledgments. I would like to thank three anonymous reviewers for NJL and the editor, Marit Julien, for invaluable comments and feedback. I am indebted to Katrín Axelsdóttir, Þórhallur Eyþórsson, Erna Björk Gestsdóttir, Gunnar Ólafur Hansson, Alex Murphy, and Hjalmar P. Petersen, who read earlier drafts of the paper and suggested improvements. I would also like to express thanks to Peter Bakker, Kristoffer Friis Bøegh, Joshua Nash, and Jeroen Willemsen for making me feel welcome at the Department of Linguistics, Cognitive Science and Semiotics at Aarhus University, while I was conducting this research there as a guest during 2018. For the same reason, I extend my gratitude to the staff at Føroyamálsdeild, University of the Faroe Islands. This research was funded by the Icelandic Research Fund (Rannís), grant number 174253-013. Notes 1 The abbreviations used in the article are the following: A = accusative; D = dative; DEF = definite; FEM = feminine; G = genitive; Ice. = Icelandic; MASC = masculine; MID = middle voice; N = nominative; NA.PL = nominative/accusative plural; NEUT = neuter; PL = plural; SG = singular. 2 The notation of schemas throughout is based on that employed by Booij (2010). 3 The general rule is that Icelandic strong nouns have a genitive plural in a consonantal ending, most commonly -s or -ar. The corresponding form in weak nouns ends in a vowel. 4 I follow Taylor’s (2012:187) definition: ‘The cue validity of feature f with respect to category C is the probability of C given f, i.e. p(C∣f).’ 5 The corpus is accessible here: https://www.sketchengine.eu/isTenTen-Icelandic-corpus/. 6 Markússon (2022) posits an Icelandic microclass of monosyllabic masculine nouns in final -s(s), e.g. foss ‘waterfall’, grís ‘piglet’, ís ‘ice (cream)’, lax [laks], arguing that their phonetic structure, as well as reference to male biological gender, facilitates reanalysis of neuter Ice. fress ‘tomcat’ as grammatically masculine. 7 While it is not a customary practice of Icelandic orthography, for the sake of clarity, the suffixed article will be separated from inflectional forms (also as these are reflected in the relevant schemas) throughout. 8 Parentheses around the -m- of the dative plural ending are intended to convey the result of historical elision of this sound before the initial n of the suffixed article D.PL -num, i.e. D.PL rótum, but D.PL-DEF rótu-num. 9 Compounds such as skrifstofublók ‘pencil-pusher’, i.e. skrifstofu-blók, dagbók ‘diary’, atvinnuleysisbætur ‘unemployment benefit’, nábrók ‘necropants’, engiferrót ‘ginger’, also occur. Due to identity of the latter constituent of each with a member of the Xó/æT-microclass and the emergent nature of meaning through use of the same form in different lexical contexts (e.g. Bybee 2010:23), these are not considered distinct lexical items. Further, due to syllabification, they are not considered prototypical and might therefore not serve as a basis for productivity. 10 Occasionally, the dative plural of a noun (and other nominals) ends in -m, e.g. D.PL skóm (of masculine skór ‘shoe’), due to contraction of earlier -ó-um to -ó-m (Kvaran 2005:224). The same process has occurred in words of all three genders, e.g. D.PL trjám (of neuter tré ‘tree’), kúm (of feminine kýr). Therefore the presence of either variant gives no indication of grammatical gender. Accounting for different rates of gender reanalysis 23 Further, some weak feminines and weak neuters have -na instead of -a in the genitive plural, e.g. feminine klukkna (of klukka ‘clock, bell’), neuter augna (of auga ‘eye’). However, some members of these classes get G.PL -a (Kvaran 2005:239, 243). Therefore, while genitive plural -na is a good indicator of feminine or neuter gender, nouns of neither class can be categorised as such on the basis of G.PL -a. 11 In support of the association of N.SG -ur with masculine grammatical gender, see Þórhallsdóttir’s comments on assumptions regarding the gender of Ice. (FEM) vættur ‘supernatural being’, æður ‘eider’, which are regularly reanalysed as masculine on account of N.SG -ur (1997:41). 12 Such nouns are the modern-day reflexes of a class referred to as u-stems in historical handbooks (e.g. Iversen 1972:66). 13 Audring (2019) defines the relation between sister schemas in terms of equivalency in the level of complexity conveyed. Such schemas are referenced as a means of checking pertinent semantic and/or formal distinctions between the sets of forms over which the relation abstracts. 14 A reviewer suggests that projection of the grammatical attribute FEMININE is an unnecessary step in the process of reanalysis and that ‘[t]he second step, analogical extension of -ur-nar from one word to another, could very well cover what happens here.’ In response, it can be argued that if formal similarity between individual instances of this sequences were the only grounds for the solution of X in (8), we might expect extension of corresponding sequences, such as FEM -ir-nar or MASC -ar-nir, to be relatively just as frequent. However, as noted in Section 1, such change hardly ever occurs. Therefore properties of the dispersion of NA.PL -ur, in conjunction with formal similarity across forms, appears to trigger outputs in -ur-nar as part of a two-step process. In other words, masculines in NA.PL -ur are first reanalysed as feminine and subsequently alternate with the feminine article NA.PL -nar in some instances. 15 It is important to note that the -ur of N.SG fótur and that of NA.PL fæt-ur are etymologically distinct and therefore unrelated from both the diachronic and synchronic perspectives. 16 The alternant -r- occurs before endings that start with a vowel, e.g. G.PL vetr-a, but G.SG veturs. The same applies to other -Vr- ∼ -r- alternations in disyllabic stems, e.g. alternation between G.PL hamra and G.SG hamars. 17 Specifically, the rate of reanalysis is defined as the occurrence of forms in NA.PL -nar as a percentage of the total of all the occurring forms of a given noun in the nominative/accusative plural definite, i.e. of those in N.PL -nir, A.PL -na, and NA.PL -nar. 18 I also calculated the rates of reanalysis based on the Icelandic Gigaword Corpus (2019), available at https://malheildir.arnastofnun.is/?mode=rmh2019#?lang=isis&stats_reduce=word&isCaseInsensitive& searchBy=word&cqp=%5B%5D. The relevant data concur largely with those presented in Table 1 in terms of the order of descending frequency, although individual rates of reanalysis were generally much lower. For example, reanalysis of plural fætur occurs in 9.20% of instances according to the corpus. This should come as no surprise, however, as the Icelandic Gigaword Corpus consists mainly of sources that are more likely to have been checked for errors before publication or release than those contained in the isTenTen corpus. It should be conceded, however, that the isTenTen corpus is around half the size of the Icelandic Gigaword Corpus. Despite this, data based on the former probably better reflect actual usage (see Section 1). 19 The alternation a∼ö between singular hjarta and plural hjörtu stems from Proto-Nordic u-umlaut and is fully morphologically conditioned in Modern Icelandic, e.g. singular barn ∼ plural börn, where no u follows the plural stem (see Markússon 2012, 2017 and sources cited there). 20 As a reviewer points out, the ija-stem vélendi occurs once in the second grammatical treatise (Codex Wormianus, c. 1350); Norwegian velende also points to ON -i. The ija-stem milti is attested in older manuscripts than milta (latter half of the fifteenth century: see ONP), and milti concurs with Norwegian forms. It seems that both words originated as neuter ija-stems and changed inflection class in Icelandic because they denote body parts. 21 This estimation considers that reanalysis of the accusative plural form would first involve a change from -i to -ir, before the addition of the feminine plural definite article -nar. The relative complexity of the process may reduce the likelihood that the accusative plural form should be reanalysed at the same or a similar rate as the nominative plural. 22 A search of the Icelandic Gigaword Corpus (2019) indicates that plural handleggir is reanalysed as feminine in 0.13% of instances, while plural fótleggir never undergoes reanalysis according to the same corpus. 24 Jón Símon Markússon References Albright, Adam. 2002. Islands of reliability for regular morphology: Evidence from Italian. Language 78, 684–709. Albright, Adam. 2008. Explaining universal tendencies and language particulars in analogical change. In Jeff Good (ed.), Linguistic Universals and Language Change, 144–181. Oxford: Oxford University Press. Albright, Adam. 2009. Modelling analogy as probabilistic grammar. In James P. Blevins & Juliette Blevins (eds.), Analogy in Grammar: Form and Acquisition, 185–213. Oxford: Oxford University Press. Albright, Adam & Bruce Hayes. 2003. Rules vs. analogy in English past tenses: A computational/experimental study. Cognition 90, 119–61. Árnason, Kristján. 2005. Íslensk tunga I: Hljóð [The Icelandic tongue I: Sounds]. Reykjavík: Almenna bókafélagið. Árnason, Kristján. 2011. The Phonology of Icelandic and Faroese. Oxford: Oxford University Press. Audring, Jenny. 2019. Mother or sister? The encoding of morphological knowledge. Word Structure 12, 274–296. Axelsdóttir, Katrín. 2014. Sögur af orðum: Sex athuganir á beygingarþróun í íslensku [Stories of words: Six studies of inflectional development in Icelandic]. Reykjavík: Háskólaútgáfan. Axelsdóttir, Katrín. 2015. Beyging og merking orðsins hjalt [The inflection and meaning of the word hjalt]. Orð og tunga 17, 95–114. Barðdal, Jóhanna. 2006. Predicting the productivity of argument structure constructions. Berkeley Linguistics Society 32, 467–478. Barðdal, Jóhanna. 2008. Productivity: Evidence from Case and Argument Structure in Icelandic. Amsterdam & Philadelphia: John Benjamins. Berg, Ivar. 2019. Gender and declension mismatches in West Nordic. In Claudia Fabrizio & Michaela Cennamo (eds.), Historical Linguistics 2015: Selected Papers from the 22nd International Conference on Historical Linguistics, Naples, 27–31 July 2015, 97–114. Amsterdam: John Benjamins. Bernharðsson, Haraldur. 2004. Um Moldhaugnaháls út í Fjósa og Fjörður [Analogical developments in some plural place names]. Íslenskt mál og almenn málfræði 26, 11–48. Bjorvand, Harald. 1972. Zu den altwestnordischen Pluralendungen -ar, -ir und -r bei femininen Substantiva [On the Old West Norse plural endings -ar, -ir and -r in feminine nouns]. Norwegian Journal of Linguistics 26, 195–215. Bjorvand, Harald. 1975. Altwestnordisch barar/bǫrur, fpl.: Eine Analyse der analogen Verbreitung der Pluralendung -ur der ōn-Stämme in den nordischen Sprachen [Old West Norse barar/bǫrur, fpl.: An analysis of the analogous distribution of the plural ending -ur of the ōn stems in the Norse languages]. Norwegian Journal of Linguistics 29, 101–112. Booij, Geert. 2010. Construction Morphology. New York: Oxford University Press. Booij, Geert & Jenny Audring. 2018. Partial motivation, multiple motivation: The role of output schemas in morphology. In Geert Booij (ed.), The Construction of Words: Advances in Construction Morphology, 59–80. Cham: Springer. Bybee, Joan. 1985. Morphology: A study of the Relation between Meaning and Form. Amsterdam: John Benjamins. Bybee, Joan. 2001. Phonology and Language Use. Cambridge: Cambridge University Press. Bybee, Joan. 2007. Frequency of Use and the Organization of Language. Oxford: Oxford University Press. Bybee, Joan. 2010. Language, Usage and Cognition. Cambridge: Cambridge University Press. Bybee, Joan. 2015. Language Change. Cambridge: Cambridge University Press. Corbett, Greville. 1991. Gender. Cambridge: Cambridge University Press. Davidson, Herbert. 1990. Han hon den: Genusutvecklingen i svenskan under nysvensk tid (Lundastudier i nordisk språkvetenskap A45). Lund: Lund University Press. Dressler, Wolfgang U. 2003. Degrees of grammatical productivity in inflectional morphology. Italian Journal of Linguistics 15, 31–62. Enger, Hans-Olav. 2004. On the relation between gender and declension. Studies in Language 28(1): 51–82. Fertig, David. 2013. Analogy and Morphological Change. Edinburgh: Edinburgh University Press. Finkel, Raphael & Gregory Stump. 2007. Principal parts and morphological typology. Morphology 17, 39–75. Accounting for different rates of gender reanalysis 25 Gentner, Dedre. 2005. The development of relational category knowledge. In Lisa Gershkoff-Stowe & David H. Rakison (eds.), Building Object Categories in Developmental Time, 245–275. New Jersey: Lawrence Erlbaum Associates. Gentner, Dedre & Christian Hoyos. 2017. Analogy and abstraction. Topics in Cognitive Science 9, 672–93. Gries, Stefan Th. & Nick C. Ellis. 2015. Statistical measures for usage-based linguistics. Language Learning 65, 228–255. Iversen, Ragnvald. 1972. Norrøn grammatik [Nordic grammar], 7th edn. Oslo: H. Aschehoug & Co. Janda, Laura A. 2002. Cognitive hot spots in the Russian case system. In M. Shapiro (ed.), Peircean Semiotics: The State of the Art, 165–88. New York: Berghahn Books. Janda, Laura A. 2007. Inflectional morphology. In D. Geeraerts & H. Cuyckens (eds.), The Oxford Handbook of Cognitive Linguistics, 632–649. New York: Oxford University Press. Jónsdóttir, Margrét. 1988–1989. Um ir- og ar-fleirtölu einkvæðra kvenkynsorða í íslensku [On ir- and ar- plurals of monosyllabic feminine nouns in Icelandic]. Íslenskt mál og almenn málfræði 10–11, 57–83. Jónsdóttir, Margrét. 1993. Um ar- og ir-fleirtölu karlkynsnafnorða í nútímaíslensku [On ar- and ir-plurals of masculine nouns in Modern Icelandic]. Íslenskt mál og almenn málfræði 15, 77–98. Knudsen, Trygve. 1967. Kasuslære [Inflection], vol. 1: Innledning, nominative, akkusative. Oslo: Universitetsforlaget. Kürschner, Sebastian & Damaris Nübling. 2011. The interaction of gender and declension in Germanic languages. Folia Linguistica 45(2): 355–388. Kvaran, Guðrún. 2005. Íslensk tunga II: Orð [The Icelandic tongue II: Words]. Reykjavík: Almenna bókafélagið. Lakoff, George. 1987. Women, Fire and Dangerous Things. Chicago: University of Chicago Press. Lakoff, George. 2018. Ten Lectures on Cognitive Linguistics. Leiden: Brill. Langacker, Ronald W. 1987. Foundations of Cognitive Grammar, vol. 1: Theoretical Prerequisites. Stanford, CA: Stanford University Press. Magnússon, Ásgeir B. 1989. Íslensk orðsifjabók [Icelandic etymological dictionary]. Reykjavík: Orðabók Háskólans. Markússon, Jón S. 2012. Eðli u-hljóðvarpsvíxla í íslenskri málsögu [The nature of u-umlaut alternations in Icelandic language history]. Reykjavík: University of Iceland (MA thesis). Markússon, Jón S. 2017. Samband veiklunar og hljóðanvæðingar: Vitnisburður u-hljóðvarpvíxla í frum- og vesturnorrænni málsögu [The relation between weakening and phonemicization: The testimony of u-umlaut alternations in Proto- and West-Nordic language history]. In Sakaris S. Hansen, Anfinnur Johansen, Hjalmar P. Petersen & Lena Reinert (eds), Bók Jógvan: Heiðursrit til Jógvan í Lon Jacobsen á 60 ára degnum, 263–276. Tórshavn: Føroya fróðskaparfelag. Markússon, Jón S. 2021. Undir áhrifum (orða)gengis [Under the influence of a (lexical) gang]. In Katrín Axelsdóttir, Veturliði Óskarsson & Þorsteinn G. Indriðason (eds.), Möggubrár heklaðar Margréti Jónsdóttur sjötugri, 21. mars 2021, 99–104. Reykjavík: Rauðhetta. Markússon, Jón S. 2022. Tvær blækur labba inn á bar: On limited productivity as graded membership of an Icelandic microclass. NOWELE 75(2). Forthcoming. ONP = Dictionary of Old Norse Prose. https://onp.ku.dk Ralli, Angela. 2002. The role of morphology in gender determination: Evidence from Modern Greek. Linguistics 40, 519–551. Rosch, Eleanor. 1975. Cognitive representations of semantic categories. Journal of Experimental Psychology: General 104, 192–233. Rosch, Eleanor, Carolyn B. Mervis, Wayne D. Gray, David M. Johnson & Penny Boyes-Braem. 1976. Basic objects in natural categories. Cognitive Psychology 8, 382–439. Ross, Brian H. & Valerie S. Makin. 1999. Prototype versus exemplar models in cognition. In Robert J. Steinberg (ed.), The Nature of Cognition, 205–241. Cambridge, MA: MIT Press. Svavarsdóttir, Ásta. 1993. Beygingarkerfi nafnorða í íslensku [The inflectional system of nouns in Icelandic]. Reykjavík: Málvísindastofnun Háskóla Íslands. Taylor, John R. 2003. Linguistic Categorisation. Oxford: Oxford University Press. Taylor, John R. 2012. The Mental Corpus. Oxford: Oxford University Press. Tuggy, David. 2007. Schematicity. In Dirk Geeraerts & Hubert Cuyckens (eds.), The Oxford Handbook of Cognitive Linguistics, 82–116. New York: Oxford University Press. 26 Jón Símon Markússon Þórhallsdóttir, Guðrún. 1997. Ylgr, heiðr, brúðr: Saga r-endingar nefnifalls eintölu kvenkynsorða [The history of the ending nominative singular -r in feminine nouns]. In Úlfar Bragason (ed.), Íslensk málsaga og textafræði, 41–56. Reykjavík: Stofnun Sigurðar Nordals. Þórhallsdóttir, Guðrún. 2007. The dative singular of ō-stems in Old Norse. In Alan J. Nussbaum (ed.), Verba Docenti: Studies in Historical and Indo-European Linguistics presented to Jay H. Jasanoff by Students, Colleagues, and Friends, 329–41. Ann Arbor/New York: Beech Stave Press. Þórólfsson, Björn K. 1925. Um íslenskar orðmyndir á 14. og 15. öld og breytingar þeirra úr forn-málinu [On Icelandic word forms in the fourteenth and fifteenth centuries and change from the Old Icelandic period]. Reykjavík: Fjelagsprentsmiðjan. Reprinted in 1987, Reykjavík: Málvísindastofnun Háskóla Íslands. Cite this article: Markússon JS. Accounting for different rates of gender reanalysis among Icelandic masculine forms in plural -ur. Nordic Journal of Linguistics. https://doi.org/10.1017/S0332586522000166 Paper II Paper II Tvær blækur labba inn á bar: On limited productivity as graded membership of an Icelandic microclass doi:10.1075/nowele.00068.mar 153 Tvær blækur labba inn á bar On limited productivity as graded membership of an Icelandic microclass Jón Símon Markússon University of Iceland This article examines the limited productivity of a minimally schematic Icelandic microclass, originally containing bók ‘book’, bót ‘patch’, brók ‘trousers’, nót ‘fishing net’, and rót ‘root’ only. Productivity is betrayed by the addition of feminine blók ‘nonentity’ and forms of neuter kók ‘Coke™’. While plural nominative/accusative blækur and kækur speak to the productivity of the microclass specifically, doublet forms in both paradigms follow multiple inflection classes. Some are of the opinion that plural kækur belies real language use, as it is only used for humorous effect. The current paper applies a usage-based cognitive analysis to the relevant data with two objectives. First, to account for graded membership in the microclass as a function of limited productivity. Secondly, to demonstrate that innovative kækur and blækur are deduced by identical means. It is concluded that productivity is a property of language use, which is posited as a function of analogy. 1. Introduction Two borrowings into Icelandic (Ice.), feminine blók ‘nonentity’, cp. Eng. bloke (Magnússon 1989: s.v. blók), and neuter kók ‘Coke™’, occasionally inflect like a mere five native feminine nouns in stem-final -ót [ou:t] or -ók [ou:k]: bók ‘book’, bót ‘patch’, brók ‘trousers’, nót ‘fishing net’ and rót ‘root’. Henceforth, these are referred to collectively as the XóT-microclass (see Dressler 1997, 2003). The notation ‘T’ represents both stem-final t and k, as three of the five native XóT-stems end in the former consonant. Two innovations speak specifically to the XóTmicroclass’s limited productivity, cp. standard feminine plural nominative/ accusative blækur and (humorous) kækur, while doublet forms in both paradigms follow different inflectional patterns. https://doi.org/./nowele..mar NOWELE : (), pp. –. ISSN - | E‑ISSN - © John Benjamins Publishing Company Graded membership of an Icelandic microclass The paradigms of Ice. blók and kók are shown in (1) alongside those of feminine mynd ‘picture’, the XóT-microclass member feminine bók, and neuter borð ‘table’. (1) Nsg. Asg. Dsg. Gsg. fem. mynd mynd mynd myndar NApl. myndir Dpl. myndum Gpl. mynda fem. bók bók bók bókar fem./fem. blók blók blók blókar bækur blækur/blókir bókum blókum bóka blóka neut./fem./fem. kók kók kóki/kók kóks/kókar neut. borð borð borði borðs kók/kækur/kókir kókum kóka borð borðum borða As (1) shows, the paradigm of borrowed blók contains feminine doublet forms only: one consistent with the model of plural bækur, i.e. blækur, while blókir patterns with plural myndir. The Icelandic dictionary (Íslensk orðabók 2010) mentions blækur only. Conversely, many native speakers of Icelandic consider blókir the appropriate plural form. Some even assert, for unspecified reasons, that they would never use Ice. blók in the plural. However, there is no doubt that both plural forms occur, cf. also skrifstofublækur ‘pencil-pushers’. The paradigm of kók, by contrast, contains both neuter and feminine doublet forms, cp. singular dative neut. kóki, fem. kók, singular genitive neut. kóks, fem. kókar, and plural nominative/accusative neut. kók, fem. kækur, kókir, which either pattern with forms of Ice. borð, bók or mynd (see Svavarsdóttir 2014). Masculine plural forms of kók existed before and around 1960 but have fallen out of use completely. They will not feature in what follows. The developments listed above suggest that additions to the XóT-microclass exhibit graded membership. While the inflection of Ice. blók patterns with feminine classes only, neuter kók is clearly in a state of (grammatical) gender fluidity. Therefore, one objective of the current paper is to account for both graded membership of the XóT-microclass and the grammatical gender fluidity of Ice. kók in light of the limitations of low type frequency on productivity. Further, some consider the humorous motivation for use of plural kækur to belie real language use, implying that it speaks little to the XóT-microclass’s productivity.1 However, the position taken here is that extension of the same inflec- 1. One such opinion has been worded as follows: “Ég hef reyndar aldrei trúað því að víxlin kók – kækur væru dæmi um raunverulega málnotkun – held að þetta hafi verið búið til í gríni.” (‘I have actually never believed that the alternation kók – kækur is an example of real language © 2022. John Benjamins Publishing Company All rights reserved 195 196 Jón Símon Markússon tional pattern to the paradigm of Ice. blók should serve to refute this argument, were it possible to demonstrate that the process proceeds by parallel means in both cases.2 Therefore, the other objective of the current paper is to identify the cognitive mechanism(s) by which feminine forms of prescriptively neuter kók are deduced as a function of natural usage, be it for levity or otherwise. In this vein, the analyses that follow provide a usage-based cognitive account of the limited productivity of the XóT-microclass through recourse to: 1. Semantics. 2. Pragmatics. 3. Schematicity, the gauge of which is a high degree of phonetic dissimilarity between the members of an inflection class (e.g. Bybee 2010: 67). 4. Analogy, defined here as the cognitive mechanism by which existing knowledge is extended to new contexts (e.g. Gentner 2005). It is argued that schematicity, semantics, and pragmatics interact to motivate extension of the XóT-microclass pattern to both paradigms. Viewed from this perspective, similarity of form and analogous contexts of use are the facilitators of change. This entails that the mechanism of productivity is analogical reasoning, which is posited as a property of language use.3 The paper is structured as follows. In Section 2, graded productivity is framed in terms of a cline mediated by degrees of type frequency and schematicity. Section 3.1 accounts for the deduction of plural kækur as a function of proportional analogy. In Section 3.2, I argue that different forms of kók are reanalysed as feminine by respective analogical means. Section 4 argues that humour proves no impediment to extension of the XóT-microclass pattern to the paradigm of Ice. kók. Conclusions are drawn in Section 5. use – (I) think it was created as a joke’ (Eiríkur Rögnvaldsson, personal correspondence; my translation)). I thank Eiríkur for his permission to cite him. 2. In a later personal correspondence (see footnote 1), Eiríkur states, “blók – blækur er hins vegar eðlileg málnotkun” (‘On the other hand, blók – blækur is natural usage’; my translation). 3. Blevins & Blevins’ (2009: 1) observation that “There is mounting evidence from work in cognitive psychology that the talent for analogical reasoning constitutes the core of human cognition” (citing Penn et al. 2008) and that “analogy may be a highly domain-independent cognitive process” (citing Halford & Andrews 2007), serves as a guiding light in the approach adopted here. © 2022. John Benjamins Publishing Company All rights reserved Graded membership of an Icelandic microclass 2. Graded productivity and analogy 2.1 The mediation of type frequency by schematicity Bybee (2010: 67) has pointed out that high type frequency and high schematicity interact to facilitate the near unlimited productivity of an inflectional pattern, i.e. its extensibility to new or existing paradigms at a high rate. Such productivity characterises the method of forming English past forms from the present stem with the addition of -(e)d, e.g. present walk ~ past walked, fire ~ fired, gazump ~ gazumped. The class in question contains thousands of verbs of practically all shapes and is constantly accommodating new members due to the interaction of high type frequency and minimal phonetic constraints on its schema (Barðdal 2008: 23–24; see also Bybee 2007: 14). An inverse relation between type frequency and schematicity is responsible for the graded membership that characterises the productivity of the Icelandic XóT-microclass (see (1), above). Feminine nouns ending in plural nominative/ accusative -ir, e.g. myndir, blókir, constitute the largest and most schematic strong feminine class. This entails that while both the class in plural -ir and the XóTmicroclass can accommodate Ice. blók and kók on schematic grounds, the former’s pattern is generally more readily accessible from memory on account of its higher type frequency (Bybee 2015: 95–6, 102). In this connection, however, patterns associated with minimal type frequency can also demonstrate limited productivity if the members of the relevant class show a high degree of phonetic coherence, i.e. low or even minimal schematicity (Bybee 2010: 67, 69). According to Barðdal (2008: 1, 9), classes characterised by low or minimal schematicity are able to attract linguistic units perceived as similar to existing members or types (see also Gentner and Markman 1997; Barðdal 2011, 2012; Cordes 2017: 278). Therefore, the minimal schematicity of the Icelandic XóT-microclass is surely responsible for the occurrence of fem. plural blækur and kækur. In terms of the mechanism for limited productivity, it is of note that schematicity is posited as a facilitating factor in Bybee and Moder’s (1983) study of the sing ~ sang ~ sung and string ~ strung classes of English strong verbs (the distinction between which is collapsed for expediency). The class in question contains some twenty-five members in Modern English. Importantly, however, around half of these arose after the Old English period (Bybee 2010: 68), e.g. present bring ~ innovative (dialectal) past brung, cf. standard past brought. Such limited productivity is perhaps surprising from dual-processing and generative perspectives. Within such models, linguistic performance is traditionally considered a function of a discreet split between complementary methods © 2022. John Benjamins Publishing Company All rights reserved 197 198 Jón Símon Markússon for producing forms of the language. In more specific terminology: forms of words that follow the “default” pattern are supposedly produced through synchronic application of symbolic rules to underlying representations, while other forms are retrieved directly from memory (see Clahsen 1999 for a dual-processing account of productivity and Rögnvaldsson 2013 for a similar approach to Icelandic morphology). Strikingly, such theoretical frameworks typically overlook the constraints of frequency and schematicity on productivity. Indeed, proponents of dual-processing and generative models would likely argue that addition of -(e)d to the present stem is the only productive method of deriving English past forms by symbolic rule. By the same token, past forms such as sang and strung will be stored in memory. Conversely, when viewed from a usage-based cognitive perspective, we should indeed expect varying degrees of type frequency and schematicity to correlate with graded productivity (e.g. Dressler 1997: 6–9, 2003: 36–43; Barðdal 2008; Gardani 2013: 36). During a series of experiments, Bybee and Moder (1983) demonstrated that test subjects were able to deduce past forms based on nonce present prompts in accordance with schemas posited for the sing ~ sang ~ sung and string ~ strung class, as in past *vung, based on pres. *vin. They took these results to indicate that the class’s productivity stems from the interaction of its relatively high type frequency with properties of its schematicity (Bybee 2010: 69). By implication, the process that produces innovative forms stems from reference to a prototype (Rosch 1978; Rosch & Mervis 1975) instead of direct retrieval from memory. Further, and more specifically, Bybee attributes the productivity of the sing ~ sang ~ sung and string ~ strung class to a gang effect, by which a relatively small class is able to attract more new members than a class of comparable number but whose phonetic structure is less uniform (2010: 69; see Axelsdóttir 2015 and Markússon 2012, 2017, 2021 for accounts of levelling and the gang effect in Icelandic and Faroese). Before forms of both blók and kók agreed with the Icelandic XóT-microclass, its prototype was defined by just five phonetically highly coherent nouns (see Section 1), rendering its concentration of shared phonetic properties much higher than that of Bybee and Moder’s class in relative terms. Thus, the schemas for the latter are necessarily more schematic than those of the former, cp. Eng. sing ~ sang ~ sung [ɪ] ~ [æ] ~ [ʌ], drag ~ (dialectal) drug [æ] ~ [ʌ], shake ~ (dialectal) shuck [eɪ] ~ [ʌ], sneak ~ snuck [i] ~ [ʌ], strike ~ struck [aɪ] ~ [ʌ], on the one hand, and, e.g., Ice. bók ~ bækur, bót ~ bætur, brók ~ brækur, rót ~ rætur [ou] ~ [ai], on the other. While it is true that members of the XóTmicroclass contain either final -t and -k, this difference is minimal and, therefore, conducive to the abstraction of a single schema (see Bybee 2010: 64; also 3.1). © 2022. John Benjamins Publishing Company All rights reserved Graded membership of an Icelandic microclass Due to the degree of phonetic dissimilarity exhibited by present forms of the sing ~ sang ~ sung and string ~ strung class, Bybee and Moder posit product-based schemas as the mechanism for deduction of innovative and nonce past forms. Such schemas encapsulate phonetic coherence perceived between e.g. past and/ or past participle sung, drug, shuck, snuck, struck. According to Croft and Cruse (2004: 301), reference to product-based schemas is preferred when the cue validity of a potential output schema (e.g. Booij & Audring 2018) is comparatively weaker. I follow Taylor’s (2012: 187) definition of cue validity: “The cue validity of feature f with respect to category C is the probability of C given f, i.e. p(C⎪f ).” Output schemas can be defined as a generalisation in terms of the cue validity that a set of inflectional forms provides for output forms within the paradigm, cp. the output past *vung as deduced from the nonce present prompt *vin. Bybee and Moder (1983: 143) posit a product-based schema, which can be represented as [ʌŋ]past/ppart. (notation based on Booij 2010), where [ŋ] represents the velar/ nasal property associated with productivity of the class, cp. past *vung. This single schema, they argue, is more informative than the numerous output schemas that it would be necessary to posit based on the phonetic variety of present forms.4 Although product-based schemas are utilised in absence of the coherent phonetic definition necessary for the abstraction of potential output schemas, a high degree of formal similarity between functionally equivalent forms from multiple paradigms should provide a better basis for the deduction of outputs. Thus, while present tense forms of the sing ~ sang ~ sung and string ~ strung microclass exhibit a fair degree of schematicity, functionally equivalent forms of the XóT-microclass are similar along all parallels with the exception of the limited number of onsets, cp. bók, bót, brók, nót, and rót. Therefore, it is perhaps understandable that on the basis of the pattern fem. singular bók(ar), bót(ar), brók(ar), nót(ar), rót(ar) ~ plural bækur, bætur, brækur, nætur, rætur, fem. singular blók(ar) and neut. singular kók should have cue validity for output forms such as fem. plural blækur and kækur. Despite the different mechanisms of analogical reasoning posited above, deduction based on product-based schemas, on the one hand, and output schemas, on the other, need not be substantively ‘either/or’ in nature. Instead, as the examples discussed above suggest, graded productivity is a function of type frequency and schematicity as disparate properties of distinct inflectional patterns. In other words, deduction of output forms that agrees with an existing pat4. Bybee & Moder (1983) refer to such schemas as ‘source-oriented schemas’, defining the production of outputs in terms of their deduction from a source. Booij (2019) and Booij & Audring (2018) describe equivalent processes in output terms. This terminology is adopted here, in keeping with recent usage-based cognitive literature on schemas. © 2022. John Benjamins Publishing Company All rights reserved 199 200 Jón Símon Markússon tern is a property of productivity, however limited or limitless it may be and by whichever analogical means. 2.2 Analogical reasoning: The mechanism of productivity Analogical reasoning is considered the cognitive mechanism of all productivity (e.g. Croft & Cruse 2004: 301; see footnote 3). Further, it is claimed that language users rely on output schemas to extend knowledge represented by the relation between the A- and B-forms of the standard proportional equation to that represented by the C- and D-forms (see e.g. Fertig 2013: 28). This representation of proportional analogy and the process just alluded to are delineated schematically in (2). (2) A C : : B X; X = D To give concrete examples, the entrenched knowledge that singular bók(ar), bót(ar), brók(ar), nót(ar), rót(ar) alternate intra-paradigmatically with plural bækur, bætur, brækur, nætur, rætur, respectively, facilitates abstraction of the output schema [XóT(ar)]NAD(G)sg. → ~ [XæTur]NApl..5 Thus, any form that aligns with the former constituent schema could represent the A-form (or C-form), while the plural of the same word would substitute B (or D), as in (3). (3) NAD(G)sg. bók(ar) C : : NApl. bækur X; X = D In turn, prior applications of the relational knowledge rendered by (3) facilitates entrenchment of and encourages further reference to the output schema [XóT(ar)]NAD(G)sg. → in novel contexts. Therefore, the C- and D-forms in (3) might well be replaced by singular kók(ar) and the (innovative) output plural kækur (see 3.1). It should be noted immediately that inclusion of the A-, B- and C-forms in a proportion does not imply that recall of all three is necessary to solve X, i.e. the output D-form (Bybee 2015: 94). Instead, the A~B-relation represents the knowledge that the sister schema [XóT]NADsg. ~ [XæTur]NApl. is defined by actual forms of the language (Bybee 2001: 27). Audring (2019) defines the relation between sis5. Although it is probably unnecessary to clarify this point, the symbol → here means ‘implies, entails’, in accordance with notational conventions associated with the presentation of formal logic. This symbol should, therefore, not be understood to mean ‘changes into’ (See e.g. Bybee 2015: 96 on the coexistence of doublet forms; and Axelsdóttir 2014: 204; 2015 for relevant examples from Icelandic language history). © 2022. John Benjamins Publishing Company All rights reserved Graded membership of an Icelandic microclass ter schemas in terms of equivalency in the level of complexity conveyed. Such schemas are referenced as a means of checking pertinent semantic and/or formal distinctions between the sets of forms over which the relation abstracts. 3. Ice. neut. kók as feminine Here I account for the schematic, semantic, and pragmatic motivation for deduction of feminine plural kækur from singular forms of neuter Ice. kók. Section 3.1 accounts for the output form’s deduction as a function of proportional analogy. Section 3.2 argues for reanalysis as feminine by respective analogical means, determined by specific morphosyntactic contexts, on the one hand, and skewed semantic relations within the paradigm, on the other. 3.1 Deduction of fem. plural kækur by proportional analogy The output (humorous) fem. plural kækur appears to result from reference of singular kók to the sister schema [XóT]NADsg. ~ [XæTur]NApl., in sentences where singular (nominative/)accusative kók denotes a count noun modified by feminine forms of numerals or adjectives. The schema is an abstraction of the relevant intra-paradigmatic relation (see De Smet 2017: 91), which has long been instantiated by e.g. feminine singular bók(ar) ~ plural bækur, bót ~ bætur, etc (see 2.2). Thus, extension of the sister schema to other paradigms is a function of the cue validity inherent to the output schema [XóT(ar)]NAD(G)sg. → for outputs of the [XæTur]NApl. type (see 2.2; see also Markússon 2022).6 Further, it is possible to point to the facilitating role that semantics and pragmatics play in alignment of (N)Asg. kók with the sister schema [XóT]NADsg. ~ [XæTur]NApl.. In this connection, a particular morphosyntactic context appears conducive to use of neut. singular kók as feminine on schematic grounds. For this reason, the following account also appeals to use of feminine modifiers with neuter kók and e.g. neuter rauðvín ‘red wine’. It is common idiomatic practice when ordering beverages in Icelandic that modifiers of the liquid agree with the grammatical gender of the container. Both 6. By far the most common endings for plural dative and genitive forms are -um and -a, respectively (see discussion in 3.2). This applies to all noun classes, irrespective of grammatical gender. For this reason, the analogical means by which novel plural dative and genitive forms are derived through reference to sister schemas will be left out of consideration, other than to say that almost all output schemas for nouns hold cue validity for another of the kind [Xum]dat.pl. or [Xa]gen.pl.. © 2022. John Benjamins Publishing Company All rights reserved 201 202 Jón Símon Markússon kók and rauðvín are typically sold in a fem. flaska ‘bottle’ or neut. glas ‘glass’, while kók is also sold in a fem. dós ‘can’. Significantly, however, the container is frequently omitted in this context, a practice that has long been conventionalised in natural usage (see discussion below). Thus, sentences like those in (4) occur both naturally and frequently in Icelandic (see Svavarsdóttir 2014). (4) a. Ég ætla að fá eina /tvær kók I intend to get one-fem.acc. /two-fem.acc. Coke™ -neut.acc. ‘I’ll have one/two Coke(s)™’ b. Ég ætla að fá eina /tvær rauðvín I intend to get one-fem.acc. /two-fem.acc. red wine-neut.acc ‘I’ll have one/two red wine(s)’ Whelpton et al. (2014) claim that agreement mismatches of the kind seen in (4) do not reflect coercion, which they define as the use of a mass noun as a count noun (see also Wiese & Maling 2005). Instead, they hypothesise that in Icelandic, there is implicit syntactic structure: the mass noun continues to be a mass noun and the countable reading is introduced by an elided head noun (Whelpton et al. 2014: 44). This statement appears to imply that mass nouns – which both kók and rauðvín are in essence – retain their gender in examples such as (4a–b), while use of feminine modifiers constitutes reference to an “elided” container. However, on this point, one might ask what attributes of kók and rauðvín identify them as neuter mass nouns in (4a–b). In this connection, it is significant that Icelandic strong feminine and strong neuter nouns typically lack overt marking in the accusative singular, as in the inflection of neuter borð in (1). Thus, the relevant forms in (4a–b) neither support nor refute a neuter mass noun reading, while the presence of feminine modifiers, cp. tvær kók/rauðvín, supports both a feminine and a count noun reading. Further, the existence of overtly feminine definite forms, i.e. singular accusative kókina ‘the Coke™’ (18 examples from the text corpus www.timarit.is),7 singular dative kókinni (11 examples) and singular genitive kókarinnar (a single example), betrays reanalysis as feminine. Moreover, while reanalysis certainly stems originally from reference to (countable) fem. flaska and dós, use of singular kók alone in the count noun sense is long since conventionalised in contexts such as (4a). As already argued, this renders unnecessary the assumption that flaska and dós are synchronically elided. 7. A description of the website www.timarit.is, taken directly from the site itself, states that: “Timarit.is is a digital library where millions of pages in digital format are made available on the Internet. This gives access to the printed cultural heritage that is preserved in newspapers and periodicals of the Faroe Islands, Greenland, and Iceland.” © 2022. John Benjamins Publishing Company All rights reserved Graded membership of an Icelandic microclass Reanalysis as feminine may also stem from omission of the latter (countable) constituents of compounds such as fem. kókflaska, lit. ‘Coke™ bottle’ and fem. kókdós, lit. ‘Coke™ can’, cp. also fem. rauðvíns-/hvítvínsflaska, lit. ‘red/white wine bottle’. Thus, analogy with the conventionalised use of feminine modifiers in sentences such as (4a) accounts for the existence of overtly feminine definite forms of kók in the count nouns sense. Should singular kók be reanalysed as a XóT-microclass form in contexts such as (4a), it might not seem unnatural that language users should extend the sister schema [XóT]NADsg. ~ [XæTur]NApl. to the paradigm when ordering multiple portions of the liquid in similar contexts. Note that the mass nouns kók and rauðvín are otherwise inherently neuter, as in sentences such as Hvar er kókið/rauðvínið? ‘Where is the Coke™/red wine?’, where neuter gender is evident from the form of the definite article, i.e. singular nominative/accusative -ið. Further, Ice. kók ‘cocaine’ only ever occurs as neuter. This fact can be taken to imply that the interaction of semantics and pragmatics, as these factors relate to the morphosyntactic context highlighted in (4a), provides an important motivator for reference of singular kók to the schema [XóT]NADsg. and, therefore, its reanalysis as feminine also. However, note that no overtly feminine definite forms of rauðvín occur. This is likely due to schematicity: by sheer coincidence, rauðvín is less susceptible than kók to the gang effect of a cluster of feminine nouns with a high concentration of similar phonetic properties (see 2.1). This is despite the fact that rauðvín occurs in sentences such as (4b), which is fully analogous to (4a), suggesting that the interaction of schematicity, semantics, and pragmatics is less robust in (4b). Concerning the cognitive mechanism for reanalysis in terms of schematicity, once singular kók is aligned with the schema [XóT]NADsg., the morphological content over which the latter abstracts – including feminine grammatical gender – is projected onto the former (see Gentner & Hoyos 2017: 674–675). Consequently, prescriptively neuter kók is grammatically feminine during the anticipated usage event (see e.g. Cienki 2015). Subsequently, as phonetically similar and functionally equivalent forms of the XóT-microclass already alternate with a plural of the type [XæTur]NApl., it is a short step to deducing fem. plural kækur as the output of (now) fem. singular accusative kók, as delineated in (5) (see 2.2). (5) [XóT]NADsg. (N)Asg. bók ↓ (N)Asg. kók → ~ : [XæTur]NApl. (N)Apl. bækur → ~ (N)Apl. X; X = kækur The output schema [XóT]NADsg. → ~ [XæTur]NApl. is included in (5) to represent prior experience with language, on which deduction of the output D-form is pred- © 2022. John Benjamins Publishing Company All rights reserved 203 204 Jón Símon Markússon icated. The symbol ↓ between A and C signifies the direction in which morphological content is projected on alignment of C with A. Thus, the C-form is aligned both formally and functionally with the same schema as A, which has cue validity for alternation with B, i.e. bækur. Finally, C triggers a pattern of alternation analogous to A ~ B, where C now has cue validity for the output D-form NApl. kækur (see 2.2). It has been suggested to me that delineation of the process as in (5) is rather mechanicalistic and analytical, given that plural kækur was likely invented by “a few drunk dudes in a pub who started using these forms in the 60s, and then they spread.” Such a scenario could well account for the origins of the humorous output kækur. However, this in no way detracts from the likelihood that the form in question was deduced by analogy, on account of the similarity of singular kók to e.g. bók and functionally equivalent relations within XóT-microclass paradigms. Whatever the sentiment behind its origin, I consider the analogical means by which plural kækur is deduced just as worthy of an account as the deduction of any other output form through the application of parallel cognitive processes. Further, should it be possible to posit analogical reasoning as a property of language use, the commonly held notion that humorous plural kækur belies this process is easily refuted (see discussion in Section 4). The above analysis suggests that semantics and pragmatics interact with minimal schematicity to facilitate alignment with the sister schema [XóT(ar)]NAD(G)sg. ~ [XæTur]NApl.. According to this line of reasoning, it can be assumed that a similar analysis should account for the assignment of feminine blók to the XóTmicroclass. Indeed, as is argued in Section 3.2, categorisation and use of blók as a member of the microclass serves to refute the view that the perfectly analogous deduction of plural kækur from singular kók is substantively different from that of plural blækur from singular blók(ar). 3.2 Accounting for the (grammatical) gender fluidity of Ice. kók Following the analysis in 3.1, it is clear that the paradigm of Ice. kók is still in a state of (grammatical) gender fluidity, while, by contrast, all forms of borrowed blók are grammatically feminine in all instances (see Section 1). This no doubt raises questions pertaining to potential constraints on use of feminine forms of kók. Therefore, the question in (6) seems pertinent. (6) Why does the entire paradigm of borrowed Ice. kók not reflect use as feminine at all times? In addressing this question in the current section, I argue that treatment of kók as grammatically feminine is highly constrained by use in and semantic association © 2022. John Benjamins Publishing Company All rights reserved Graded membership of an Icelandic microclass with contexts such as (4a). Further, it is argued that the relevant forms of the noun are reanalysed as feminine by distinct analogical means, largely as a function of the cross-linguistic tendency towards a relation between meaning and form (e.g. Bybee 1985). It is important to clarify immediately that reanalysis is assumed to occur online, i.e. during actual usage events (e.g. Rácz et al. 2020), in contexts such as (4a), where the count noun sense of Ice. kók is conventionalised (see 3.1). Thus, by implication, feminine grammatical gender is not considered an inherent property of any form of kók. Rather, use of feminine forms is viewed as contingent on the process of reanalysis delineated in (5), as facilitated by the semantic and pragmatic features of contexts such as (4a) and the interaction of these with the minimal schematicity of the XóT-microclass. Further, it is assumed that the establishment of multiple doublet forms or even wholesale reanalysis of a paradigm occurs through repeated reference to the relevant schema(s) over time. Indeed, this may well have been the case with Ice. feminine blók, which only refers to men in the lending language, i.e. English,8 and it likely did in the context through which it was borrowed (see Section 4). Thus, while it is conceivable that Ice. blók acquired feminine grammatical gender immediately on account of phonetic similarity to e.g. Ice. bók, brók, etc., this is by no means a given. In other words, multiple forms of Ice. kók are still (grammatically) gender fluid, while also bearing a strong phonetic resemblance to forms of the XóT-microclass. Some forms of prescriptively neuter kók prove entirely uninformative in terms of evidence for wholesale reanalysis. For example, as can be discerned from the paradigms in (1), the endings Dpl. -um and Gsg. -a are omnipresent, cp. myndum, bókum, blókum, kókum, borðum, and mynda, bóka, blóka, kóka, borða, being common to the vast majority of Icelandic nouns irrespective of gender (see footnote 6). For this reason, use of the relevant endings on a borrowed noun does not demonstrate gender reanalysis, even when the forms in question are modified by attributive numerals or adjectives: the dative plural form of a given modifier is identical for all three genders, which goes for the genitive plural form too. Further, we should expect wholesale reanalysis of kók as feminine to yield syncretic fem. singular nominative/accusative/dative kók, as well as fem. singular genitive kókar, a relation of extremely high type frequency among the strong fem- 8. The Oxford English Dictionary confirms that this is indeed a semantic property of Eng. bloke (OED, s.v. bloke). Further, according to Quinion (2004/2011), the earliest recorded use of this word in English, on the 9th April 1829, was in reference to a man: http://www .worldwidewords.org/qa/qa-blo3.htm). Finally, as a native speaker of Cockney from Stepney, in the East-End of London, I can assert that bloke only refers to men in Modern British English. © 2022. John Benjamins Publishing Company All rights reserved 205 206 Jón Símon Markússon inine nouns of Icelandic, e.g. bók ~ bókar, mynd ~ myndar, laug ~ laugar ‘hot spring, swimming pool’ (see Þórhallsdóttir 2007 on exceptions). However, while 11 examples of overtly feminine definite singular dative kókinni were returned from a search of www.timarit.is, only one was found of overtly feminine definite singular genitive kókarinnar, as well as one conclusive example of overtly feminine (indefinite) Gsg. kókar. Conversely, 711 results for singular dative kók were returned from a search of the prepositional phrase í kók ‘in Coke™’ (idiom. ‘and Coke™’, cp. in results such as vodki í kók ‘vodka and Coke™’). While the form in question certainly accords formally with dative singular forms of the XóT-microclass, as in fem. bók, bót, brók, nót, rót, few of the syntactic contexts returned for singular dative kók contain feminine modifiers. Nonetheless, four such examples were found: yfir/í kaldri kók ‘over a/in (idiom., ‘and’) cold Coke™’, where both prepositions require the dative case of their objects in this context (see Table 1). Table 1. Search results for PPs containing overtly feminine Dsg. kók; with and without the qualifying adjective (Dsg.fem.) kaldri ‘cold’ PP with Dsg. kók Translation Fem. modifier No. of examples í (Dsg.) kók ‘in Coke™’ Ø í (Dsg.) kaldri kók ‘in cold Coke™’ Dsg.fem. kaldri 1 yfir (Dsg.) kaldri kók ‘over a cold Coke™’ Dsg.fem. kaldri 3 711 The form kók in examples with yfir likely reflects the count noun sense, i.e. ‘over a (bottle/can of ) cold Coke™’. Thus, it can be argued that the use of singular dative kók in this context is semantically associated with that of accusative kók in (4a), which frequently co-occurs with feminine modifiers. However, dative kók in examples with í refers to the mass noun, i.e. an indeterminate amount of Coke™ used as a mixer and served in a glass. In light of this, recourse to the gender of omitted neut. glas as a means of accounting for feminine modifiers after í seems illogical. It appears, then, that there are two feasible means of accounting for Dsg. kók in sentences with the preposition í. On the one hand, it may be that neut. kók is uninflected in such sentences. Indeed, this has probably always been the case when í kók means ‘at Coke™’, i.e. at the company that produces the drink in Iceland.9 However, as unambiguously neuter singular dative kóki exists in con9. A search of www.timarit.is returned results such as Steini í kók, in reference to a former CEO of the Icelandic company that produces Coke™ locally. Interestingly, not a single result for neut. singular dative kóki was returned in this context, suggesting that the company name is not © 2022. John Benjamins Publishing Company All rights reserved Graded membership of an Icelandic microclass texts such as vodki í kóki, substitution of uninflected and, therefore, grammatically ambiguous kók as qualified by overtly feminine attributives is considered unlikely. On the other hand, fem. singular dative kók may have been deduced on analogy with the common idiomatic practice of using feminine modifiers with singular kók in contexts such as (4a). Glasses, like bottles and cans, are countable containers of liquid, a semantic link which provides a plausible bridge across which fem. singular (nominative/)accusative kók is extended to the dative singular. Additionally, the semantic relation between dative kók(i) and nominative/ accusative kók, on the one hand, and that of each form to plural kækur, on the other, does not proceed from an equal footing due to the disparate frequencies with which these refer to Coke™ in the count noun sense directly. In view of this, it is worth considering whether and in what ways skewed semantic relations between the forms of the paradigm determine the degree of phonetic coherence between the innovative outputs fem. singular dative kók, singular genitive kókar, and plural kækur of singular nominative/accusative kók. Bybee (1985) argues that forms sharing a common value engage in stronger semantic relations to one another than do corresponding forms across the value divide. In other words, the formal structure of paradigms often reflects important semantic distinctions between the forms of opposing values, such as the singular and plural in nouns, or the present and past in verbs. Semantic distinctions of the type just discussed are reflected crosslinguistically in a tendency towards greater formal distinction of the inflectional material closest to a word’s stem or, indeed, the stem itself across the value divide rather than between material on either side of it (Bybee 1985, 2015: 105). Thus, the degree of formal coherence between singular nominative/accusative kók, fem. singular dative kók, genitive kókar, on the one hand, and singular nominative/ accusative kók and fem. plural kækur, on the other, likely reflects the highly significant semantic distinction between singular and plural. This is a relatively common dynamic within Icelandic paradigms, which clearly characterises the assignment of new members to the XóT-microclass.10 Another cross-linguistic manifestation of the relation between meaning and form is the tendency towards zero-exponence in the expression of the unmarked inflected by convention. This is perhaps understandable, as kók refers directly to the American brand in this context rather than to the liquid or portions of it. 10. There are, however, notable exceptions to this. For example, the nominative/accusative singular of Icelandic strong neuter nouns is always syncretic with the corresponding forms of the plural, cp. singular and plural nominative/accusative borð ‘table’, unless the singular stem contains the root vowel a, cp. singular land ~ plural lönd ‘country’. Although syncretism of this kind is far less frequent among masculine and feminine nouns (as to border on nonoccurrence), it does exist, cp. masc. singular nominative skór ~ plural nominative skór ‘shoe(s)’; fem. singular kýr ~ plural kýr ‘cow(s)’. © 2022. John Benjamins Publishing Company All rights reserved 207 208 Jón Símon Markússon value (Bybee 1985), e.g. the singular of nouns or the present tense of verbs.11 While this is by no means a defining characteristic of Icelandic paradigms, deduction of the outputs fem. singular dative kók, genitive kókar, on the one hand, and plural kækur, on the other, from singular kók can be interpreted in terms of varying degrees of semantic relatedness between the respective sets of forms. For this reason, it is assumed that skewed semantic relations within the paradigm, in addition to pragmatic constraints on the applicability of the count noun sense of Ice. kók, motivate respective analogical means for deduction of the feminine output forms. Thus, extension of singular nominative/accusative kók to the dative singular is best characterised as analogical levelling, i.e. an attempt towards elimination or simplification of morphophonological alternation within the paradigm, but specifically within the value singular. The reason for taking this position is that syncretism of the relevant forms follows the existing high frequency inflectional pattern among Icelandic strong feminine nouns exemplified by paradigms in (1). Indeed, this is a change that facilitates the well documented phenomenon of inflection class shift, avoiding the establishment of new inflectional patterns on a regular basis. Of course, positing influence from external patterns implies a role for proportional analogy in the deduction of fem. singular dative kók. However, while this should certainly prove the most plausible means of extension were the process motivated by schematicity alone, the semantic and pragmatic factors outlined in 3.1 and above suggest that extension is constrained by association with the occurrence of the form kók in contexts such as (4a). Hence, the direction of levelling from singular nominative/accusative kók to dative kók is determined intraparadigmatically by the semantic link between instances in which forms of Ice. kók represent the liquid as served in a container, as depicted in (7). (7) NAsg. (neut./fem.) kók ‘mass noun’/ ‘count noun’ Dsg. (neut.) kóki ‘mass noun’ ↔ (fem.) kók ‘count noun’ The double-headed arrow ↔ suggests that neut. singular dative kóki and its feminine cellmate kók engage in a division of semantic and pragmatic labour. As pointed out by Bybee (e.g. 2015: 95), levelling is not defined as an older form changing into a younger form. While one form may eventually concede the relevant paradigm cell to the other(s), all can coexist for centuries. 11. Haspelmath (2006) has demonstrated that the notion of ‘markedness’ is better defined through reference to factors such as the different token frequencies of distinct case forms. Thus, the term ‘unmarked’ as used above probably refers to the relatively greater frequency of singular forms relative to plurals in languages where this distinction is systematically coded, as is the case in Icelandic. © 2022. John Benjamins Publishing Company All rights reserved Graded membership of an Icelandic microclass Note that the relation between the relevant singular forms of Ice. kók as used in the mass noun sense is also reflected in the paradigm of neuter borð (see (1)), cp. singular nominative/accusative borð ~ dative borði, kók ~ kóki. Conversely, syncretism in singular nominative/accusative/dative kók in the count noun sense agrees with the inflection of e.g. feminine bók (see (1)). Therefore, mismatches in semantic relatedness and association with contexts where Ice. kók occurs as a count noun appear to determine the means of analogical change, i.e. due to levelling (syncretic singular nominative/accusative/ dative kók) or proportional analogy (singular nominative/accusative kók ~ plural kækur) (see 3.1). Consequently, the (rare) deduction of the output feminine (definite) singular genitive kókar(innar) appears to run counter to arguments based on the relation between meaning and form, as deduction is defined by the addition of feminine -ar rather than the loss of neuter -s from singular genitive kóks by levelling. This development can be accounted for as follows. First, a search of www.timarit.is returned no results where neut. singular genitive kóks unambiguously refers to a countable portion,12 but rather results such as áhrif kóks ‘the effects of Coke™’, i.e. the mass noun sense. Thus, it appears that semantic and pragmatic factors render extension of singular kók to the genitive singular through levelling unlikely, as the latter seems not to occur in sentences such as (4a) nor in semantically related contexts. Rather, as there is no pragmatic association with other singular forms of the kind that motivates levelling through such contexts, deduction of fem. singular genitive kókar is considered a function of proportional analogy, through extension of the sister schema [XóT]NADsg. ~ [XóTar]Gsg.. Syncretic fem. singular nominative/accusative/dative kók, once established across contexts such as (4a) and others perceived as semantically analogous, motivates alignment with functionally equivalent syncretism in other paradigms. As already noted, syncretism of this kind is highly frequent among the classes of Icelandic strong feminine nouns, which have exhibited a moderate degree of productivity over the centuries (Kvaran 2005: 345). Also, neut. singular genitive kóks is quite rare relative to the frequency of feminine singular kók, meaning that the latter, which contains the bare stem, is better represented in memory and more readily accessible for use than the former (see Section 2). Thus, innovation characterised by alternation between a syncretic form in -Ø and another containing both the bare stem and the singular genitive ending -ar follows this frequent and moderately productive pattern. 12. It is questionable whether instances such as the following refer to the drink in the count noun sense: skárra í formi epla og holls brasks en td. prins pólós og (Gsg.) kóks ‘better in the form of an apple and healthy enterprise than, for example, (a?) Prince Polo™ and (a?) Coke™’. © 2022. John Benjamins Publishing Company All rights reserved 209 210 Jón Símon Markússon Concerning the rarity of fem. singular genitive kókar (see above), this fact should come as no surprise. First, the form is rare anyway, particularly in the count noun sense. Secondly, feminine grammatical gender is not considered an inherent property of the neuter noun kók, but rather is contingent on reanalysis in the count noun sense or by semantic association. Therefore, as the singular genitive form does not occur in such contexts, the otherwise neuter genitive singular form is susceptible to (grammatical) gender fluidity on schematic grounds, particularly as a function of the gang effect of the minimally schematic XóT-microclass (see Section 2). The process of alignment suggested above is delineated in (8). (8) [XóT]NADsg. NADsg. bók ↓ NADsg. kók → ~ : [XóTar]Gsg. Gsg. bókar → ~ Gsg. X; X = kókar To sum up, singular dative neut. kóki and fem. kók; singular genitive neut. kóks and fem. kókar; plural neut. kók and fem. kækur are posited as doublets that belong to a single paradigm, i.e. that of neuter kók. As argued above and in 3.1, particular contexts of use motivate reanalysis of forms of the neuter mass noun kók as feminine, i.e. when the count noun sense is conveyed. This takes place on analogy with semantically and pragmatically similar contexts or, in the absence of these factors, due to pressure from patterns of higher type frequency and minimal schematicity. The response to the research question posed in (6), then, is that reanalysis as feminine is highly restricted to rather specific contexts. 4. Extension of the XóT-microclass pattern: That’s not funny! As mentioned in previous sections, some are of the opinion that the output humorous plural kækur belies real language use and, by implication, the productivity of the XóT-microclass, while plural blækur is considered to reflect natural usage (see footnotes 1 and 2). However, based on 3.1 and 3.2, it can be argued that while humour is the likely motivation for use of plural kækur, other factors – formal, semantic, and pragmatic – motivate the deduction of both outputs by proportional analogy. Hence, the question in (9) surely arises. (9) Does the deduction of an output form count as language use? If the answer to (9) is yes, in what ways does plural kækur count as a less authentic example of language use than plural blækur, when both are unarguably based on © 2022. John Benjamins Publishing Company All rights reserved Graded membership of an Icelandic microclass the existing model of the XóT-microclass (see 3.1, Markússon 2022, also below)? It is for those who answer no to the question posed in (9) to defend their position. When determining the model for extension of the XóT-microclass schemas to either paradigm, chronology proves highly significant. According to a search of www.timarit.is, Ice. plural blækur first occurs in 1927, while forms of Ice. kók in reference to Coke™ occur after the middle of the 20th century. This suggests that schemas for the XóT-microclass were first extended to the paradigm of blók on the model of five nouns, i.e. bók, bót, brók, nót, rót. Significantly, the fact that blók is feminine in all instances, coupled with the rarity of plural blókir, makes it likely that the noun was already a wholesale member of a six-strong XóT-microclass when the output fem. plural kækur was first deduced from singular kók. Further, it is unlikely that humour played any role in the deduction of plural blækur because blók is used derogatively in Icelandic rather than for humorous effect. Consequently, initial extension of the Icelandic XóT-microclass pattern was unlikely motivated by humour. Therefore, consideration of whether humour should, for any logical reason, impede the deduction of fem. plural kækur must necessarily inform any answers to (9). This consideration will now be discussed from a usage-based cognitive perspective on productivity. A usage-based cognitive approach to language assumes the impact of domaingeneral cognitive processes, such as comparison, analogical reasoning, and categorisation, on productivity (Clausner & Croft 1997; Croft & Cruse 2004; Barðdal 2008; Bybee 2010: 6–8; Cienki 2015). From this perspective, it can be argued that motivation for a negative answer to the question in (9) would prove ad hoc for two reasons. First, the objective of the accompanying argument would likely be to preserve the integrity of dual-processing and generative models rather than to allow for graded productivity (see discussion in Section 2.1). Secondly, staunch rule-or-retrieval approaches fail to allow for the effect of similarity to an existing morphophonological pattern that had already exhibited – albeit highly limited – productivity as a facilitating factor in the deduction of innovative Ice. kók ~ kækur. Importantly, the usage-based cognitive approach posits language use as a mechanism of language change, through which both are propelled by the same anatomical and cognitive constraints at all times, in all languages (e.g. Bybee 2010: 6, 105, 118; 2015: 9). Relating this to the notion that humorous motive belies language use, scholars have in fact long assumed a positive relation between these factors as facilitators of change. Houghton states: Since satire and humor are generally thought to be critical and corrective, educated people probably consciously or unconsciously feel that a humorous-critical © 2022. John Benjamins Publishing Company All rights reserved 211 212 Jón Símon Markússon use of certain expressions by people with education and prestige is one way of discouraging the use of these expressions by those less sensitive to language […] I suggest that any use of any word or expression may well have the effect of establishing that word or expression more firmly in the language. In language matters, familiarity breeds not contempt but acceptance, and new words or expression thrive on publicity, even bad publicity. (1968: 1178 [emphasis original]) To frame Houghton’s comments in the usage-based cognitive context, language use determines the structure of the grammar, while the grammar, so constructed, influences further use as part of a piecemeal but continuous process of change (e.g. Barlow & Kemmer 2000; Croft 2000; Tomasello 2000; Bybee & McClelland 2005: 382; Bybee 2010: 2; Winters 2020). Thus, due to the impact of continual language use, grammar is an emergent phenomenon whose structure is in constant flux, rendering language a dynamic, complex adaptive system (Beckner et al. 2009). For this reason, in initial answer to (9), above, the process that yields the humorous output fem. plural kækur and facilitates its substitution for older forms of Ice. kók is best characterised as language change as emergent through language use. Dissemination of plural kækur may well have proceeded in a way that is congruent with Houghton’s view, by which a novel form is intentionally deduced for humorous effect and, subsequently, repeated within similar contexts, gradually entrenching it more firmly in the language. Over time, the innovation is accepted and conventionalised within contexts reminiscent of those in which it had already occurred. Thence, language users continue to use plural kækur for humorous effect. There are additional ways in which humour is employed in language that betray analogical reasoning as a property of use. For example, joke telling offers a means of actively highlighting potential instances of language use deemed absurd relative to more context appropriate statements. Needless to say, the motivation for joke telling is the elicitation of a humorous reaction. In this connection, consider the ambiguity highlighted by the well-known joke in (10). (10) Diner: Waiter, what’s this fly doing in my soup? Waiter: Why, madam, I believe that’s the backstroke! (The formulation of the joke is lifted directly from Bybee 2010: 28) Anyone who is familiar with the dynamic depicted in (10) is aware that the waiter’s response is inappropriate. Such intuition is based on prior experience which informs the listener that more appropriate responses are both possible and, indeed, to be expected when a diner complains to staff about the food the latter has served and that the former is expected to pay for. © 2022. John Benjamins Publishing Company All rights reserved Graded membership of an Icelandic microclass The frequency with which listeners will have discharged the role of diner establishes this as common knowledge, i.e. entrenching it more firmly in conscience. Assuming this is the case, the experienced diner is likely to find the waiter’s response in (10) funny due to its absurdity: It breaches all expectations based on personal experience. As the joke teller has likely accrued similar experience over time, they are justified in assuming such familiarity on the part of the listener and, therefore, a humorous response also. Factors other than humour certainly yielded the potential for the pragmatic ambiguity highlighted by the joke in (10) (see Kay & Fillmore 1999). However, it would be farfetched to imply that the humorous motivation behind the joke itself belies real language use. Rather, our knowledge that potential non-humorous responses, such as I’ll bring you another soup, madam!, exist facilitates comparison. According to results presented by Bulloch & Opfer (2009), comparison of relationally similar scenarios is itself a facilitator of analogical reasoning in adults, while it facilitates ontogenetic development of the domain-general ability to reason analogically in children. As suggested throughout, analogical reasoning is the mechanism by which the outputs fem. plural blækur and kækur are deduced within their respective paradigms. Thus, because productivity is considered a function of analogical reasoning, by implication, deduction of outputs according to existing patterns – for humorous effect or otherwise – is considered a property of language use. The only way the language community at large can be made aware that a novel output has been deduced is through its occurrence in an actual usage event. In other words, if an output is never used, it does not exist beyond the grammar of the individual who deduced it, suggesting that deduction is a necessary constituent in the process of linguistic conventionalisation (see above). Given what has been argued above, some aspects of the analogical process that deduced plural kækur from singular forms of Ice. kók appear to reflect the earlier deduction of plural blækur from singular blók(ar). Both processes are predicated on comparison, motivated by phonetic coherence between singular forms of Ice. blók and kók and those of the XóT-microclass. In turn, comparison motivates reference to the relevant sister schemas, which have cue validity for outputs of a given phonetic structure. The proportion in (11) delineates extension of the XóT-microclass schemas to the paradigm of Ice. blók. Note that the C- and D-forms of (11) bear a strong formal resemblance to their equivalents in (5), i.e. singular blók(ar), plural blækur and kók, kækur, respectively (see Section 3.1 on the notation employed and Section 2.2 on the proportional schema A: B :: C : D). © 2022. John Benjamins Publishing Company All rights reserved 213 214 Jón Símon Markússon (11) [XóT(ar)]NAD(G)sg. NAD(G)sg. bók(ar) ↓ NAD(G)sg. blók(ar) → ~ : [XæTur]NApl. NApl. bækur → ~ NApl. X; X = blækur Both schematicity and the applicability of a word meaning ‘nonentity’ to lowly people of any gender may well have conspired to facilitate the assignment of Ice. blók to the feminine XóT-microclass. While this constitutes conjecture due to the lack of documented change, it is noteworthy that both plural doublet forms of Ice. blók, i.e. blækur and blókir, are feminine, cp. in this regard that neut.(/fem.) singular nominative/accusative(/dative) kók still alternates with fem. plural kækur and kókir, while its respective doublet forms were associated with all three genders before 1960 (see Section 1). Other developments in Icelandic support the general argument that interaction of schematicity and semantics does indeed facilitate reanalysis on the basis of grammatical gender. For example, historically neuter fress ‘tomcat’ is occasionally used as masculine. Evidence for reanalysis can be garnered from overtly masculine definite forms such as nominative singular fress-inn, nominative plural fressar-nir – -ar is chiefly associated with the inflection of masculine nouns while -nir is the masculine plural definite article (see Markússon 2022) – as opposed to prescriptive neut. definite singular nominative/accusative fress-ið, plural nominative/accusative fress-in. Further, the personal pronoun Ice. 3sg.masc. hann ‘he/it’ rather than 3sg.neut. það ‘it’, is often used in reference to fress. It might be argued that reanalysis of fress as masculine occurs due to semantic association with the masculine synonym högni ‘tomcat’.13 However, here it is important to recognise that use of the two words tends towards complementary distribution along geographical parameters: högni is mainly used in North Iceland, while fress is more common elsewhere (Sveinsson 2001: 14). I have asked southerners about their use of both words in this connection. The majority answers along the lines that while högni is familiar, fress is their default word for tomcat, while the converse has been the most common response from northerners. Therefore, while it is conceivable that reanalysis of neut. fress as masculine may to some degree occur on account of semantic association with masc. högni, it is questionable whether the latter should affect the treatment of the former in the speech of southerners, among whom högni is relatively uncommon. Significantly, reanalysis of högni as neuter does not occur in the speech of southerners, despite more frequent use of prescriptively neuter fress and the fact 13. I thank an anonymous reviewer for drawing this possibility to my attention. © 2022. John Benjamins Publishing Company All rights reserved Graded membership of an Icelandic microclass that a sizeable subclass of neuter nouns has -i in stem-final position, e.g. neut. epli ‘apple’, kaffi ‘coffee’, svæði ‘area, district’. The semantic and pragmatic motivation for reanalysis of Ice. neut. fress as masculine is considered obvious: it refers to the male of the species. Reanalysis also aligns with a commonly represented correlation in Icelandic between the biological gender of an animal and the grammatical gender of the word that denotes it. Examples of this correlation are masculine rakki ‘male dog’, hrútur, ‘ram’ vs. feminine tík ‘female dog’, ær ‘ewe’, respectively; even (karl)maður ‘man’ vs. kona ‘woman’, strákur ‘boy’ vs. stelpa ‘girl’. Significantly, Ice. læða ‘female cat’ is grammatically feminine. Just as the schematicity of the feminine XóT-microclass suffices to attract forms of neuter kók, a small number of Icelandic masculine nouns in word-final -s(s), e.g. masc. foss ‘waterfall’, grís ‘piglet’, ís ‘ice (cream)’, lax [laks] ‘salmon’, likely facilitates use of fress as masculine. It is also significant that foss, grís, ís and lax all lack the typically strong masculine ending Nsg. -ur, cp. more prototypical masculine bátur ‘boat’, hestur ‘horse’, staður ‘place’ (see Þórhallsdóttir 1997: 41). Therefore, it can be argued that masculine nouns in final -s(s), in addition to e.g. masc. fugl ‘bird’, vagn ‘wagon’, turn ‘tower’, all of which lack -ur in the nominative singular, facilitate assignment of neut. fress to the microclass in question (cp. also borrowings such as masc. kross ‘cross’). Supposed counterexamples do exist, cf. neuter hross ‘horse’, hnoss ‘treasure’, hrís ‘brushwood’, sax ‘large knife’, as well as feminine flís ‘splinter, tile’.14 Perhaps the most problematic for the account offered above is hross, which, like fress, refers to living creatures, some of which are male. However, a search of the digital library www.timarit.is returned 45 examples of singular nominative/accusative hrossinn, betrayed as overtly masculine by the form of the definite article, i.e. -inn. Such examples are unlikely the fault of printing or scanning errors, as the definite article in the prescriptive form is neut. -ið, meaning that an accidental typing or reading of -inn would require several coincidental slips of the finger or major issues with the OCR. Additionally, the same search returned examples such as hvíti hrossinn ‘the white horse’. Attributive hvíti is a masculine form of the adjective hvítur ‘white’, which corresponds with masculine singular nominative hrossinn in case, number, and gender. It must be conceded here that highly common, semantically related masculine hestur ‘horse’, in addition to final -ss in hross, likely provides motivation for reanalysis. The same can be said of the potential influence of masculine hnífur ‘knife’ on the occasional reanalysis of sax as masculine. 14. This was drawn to my attention by the same reviewer (see footnote 13). © 2022. John Benjamins Publishing Company All rights reserved 215 216 Jón Símon Markússon The same corpus also returned 50 examples for masc. hrísinn, as well as gender agreement in syntactic contexts such as og er hrísinn ræktaður ‘and the brushwood is grown’, where ræktaður is the overtly masculine past participle of rækta ‘cultivate, grow’. Forms with the masculine article -inn were also returned for prescriptively non-masculine hnoss and flís. As this treatment is unlikely due to the influence of masculine synonyms or semantically related words, it seems that final -s(s) is suggestive of masculine grammatical gender, due to phonetic resemblance to e.g. foss, grís, kross, lax, and (occasionally) fress. It appears, then, that the interaction of schematicity, semantics, and pragmatics as a facilitator of limited productivity is not confined to the productivity of the Icelandic XóT-microclass. Further, it is considered likely that the productivity of low frequency inflectional patterns is not only facilitated by minimal schematicity but is dependent on it (e.g. Barðdal 2008; also, Section 2). This dependency can be attributed to the relatively compromised accessibility of low frequency patterns in on-line language processing, rendering competition with high frequency patterns futile, all other things being equal (e.g. Bybee 2010: 67). As discussed in Section 2, a high degree of phonetic coherence is the property of low frequency patterns that enables them to attract new items. Additionally, semantic and pragmatic factors conducive to reanalysis likely serve to boost the basis for analogical reasoning provided by minimal schematicity. Concerning the schematicity of the XóT-microclass, the innovative relations singular blók(ar) ~ plural blækur and kók(ar) ~ kækur are unarguably based on the model of e.g. bók(ar) ~ bækur, brók(ar) ~ brækur, etc. In terms of semantic and pragmatic properties, Eng. blók has likely only ever had male referents and, given the context in which it was borrowed (see Section 1), this likely also applied in Icelandic immediately after borrowing. However, it is considered likely that Ice. blók attained feminine gender on account of the interaction between schematicity, semantics, and pragmatics. In other words, the applicability of Ice. blók to women (as well as men) can only have boosted its association with feminine grammatical gender on account of its phonetic similarity to fem. bók, bót, brók, nót, rót, cp. analogous factors at work in the reanalysis of Ice. neut. fress and hross as masculine. Similarly, Ice. kók has exhibited (grammatical) gender fluidity since the time it was borrowed. Thus, its use in the mass noun sense, i.e. to denote an indeterminate amount of Coke™, is associated with neuter gender, while its use in the count noun sense, i.e. to denote countable portions of Coke™, tends to be manifested by grammatically feminine forms (see 3.2). Therefore, it is argued that the interaction of schematicity, semantics, and pragmatics also facilitates reanalysis of Ice. kók as feminine in the count noun sense, where words for countable containers, i.e. feminine flaska, dós and, by analogy, also neuter glas, are omitted. © 2022. John Benjamins Publishing Company All rights reserved Graded membership of an Icelandic microclass To sum up, while it is true that humour is the motivating factor for use, it can be argued that plural kækur is just as deducible from neuter singular kók by analogical reasoning as plural blækur is from singular blók(ar). Considering the objectives set out in Section 1, this has been demonstrated through reference to schematicity, semantics, and pragmatics as these motivate analogical reasoning in both cases. Therefore, in answer to the question posed in (9), the outputs kækur and blækur are considered a function of the domain-general cognitive processes that facilitate the extension of morphophonological patterns as a property of language use. 5. General summary The objective of the current paper was twofold. First, to account for the graded membership of Ice. blók and kók in the Icelandic XóT-microclass as a function of limited productivity. Secondly, to refute claims that innovative fem. plural nominative/accusative kækur belies real language use due to its origins in humour, on the grounds that the same morphophonological pattern was extended to the paradigm of borrowed feminine blók, irrespective of such considerations. The following conclusions can be drawn: 1. The interaction of schematicity, semantics, and pragmatics facilitates analogical extension of the XóT-microclass inflectional pattern to both paradigms, based on the model of e.g. fem. singular bók(ar) ~ plural bækur, bót(ar) ~ bætur, etc. In both cases, extension constitutes part of a process that can be characterised as language use. 2. Productivity is observably a graded property of language that exhibits correlation with degrees of type frequency and schematicity. This is demonstrated by the limited productivity of both English and Icelandic microclasses, whose schematicity serves to mediate the constraints of low type frequency on productivity. 3. The interaction of schematicity, semantics, and pragmatics facilitates the deduction of output forms such as plural kækur by analogical reasoning. Due to the long conventionalised idiomatic practice of omitting the grammatically feminine words for containers in Icelandic sentences such as (4a–b), feminine forms of modifiers qualify neut. kók when it is used in the count noun sense. 4. The circumstances described in 3, above, motivate the reanalysis of neut. kók as feminine through reference to the sister schema [XóT]NADsg. ~ [XæTur]NApl.. This process deduces the output feminine plural kækur from neuter singular kók. © 2022. John Benjamins Publishing Company All rights reserved 217 218 Jón Símon Markússon 5. The disparate rates at which distinct forms of kók are reanalysed as feminine correlate with their occurrence in the morphosyntactic contexts identified in Section 3. Thus, interaction of schematicity, semantics, and pragmatics determines the proportional means by which feminine singular genitive kókar and plural nominative/accusative kækur are deduced from singular nominative/ accusative kók, on the one hand, and of levelling to feminine singular dative kók based on the same source, on the other. 6. Humour is a facilitator of language use, which, according to the usage-based cognitive approach, is the driving force of language change and, therefore, linguistic innovation. 7. Conventionalisation of humorous language exhibits parallels to that of novel language use generally. 8. In light of 7, above, it is unnecessary to assume any significant substantive difference between the cognitive processes that deduce the outputs feminine plural blækur and humorous plural kækur from singular blók(ar) and kók, respectively. A usage-based cognitive approach accounts for the limited productivity of and graded membership in the Icelandic XóT-microclass. While the scope of the current study is microscopic in nature, it is hoped that the approach can be applied in future studies of both partial and wholesale reanalyses and of graded productivity as a correlate of class membership more generally. The productivity of minimal frequency inflectional patterns is far from constrained by humour; schematicity, pragmatics, and humour facilitate productivity by analogy. Funding This research was funded by the Icelandic Research Fund (Rannís), grant number 174253-015. Acknowledgements I thank the editor and two anonymous reviewers for feedback and assistance. I am indebted to Katrín Axelsdóttir, Þórhallur Eyþórsson, Gunnar Ólafur Hansson, Alex Murphy, and Hjalmar P. Petersen for reading earlier drafts of the paper and/or individual chapters, as well as to Erna Björk Gestsdóttir for lively discussions on the topic of the paper. I am also grateful to Peter Bakker, Kristoffer Friis Bøegh, Joshua Nash and Jeroen Willemsen, whom I met whilst a visiting scholar at the Department of Linguistics, Cognitive Science and Semiotics at Aarhus University in 2018, when a significant portion of this research was conducted. Helpful comments and suggestions from all of those just listed served only to sharpen focus and argumentation. Any remaining errors require the author’s head alone. © 2022. John Benjamins Publishing Company All rights reserved Graded membership of an Icelandic microclass References Audring, J. 2019. Mothers or sisters? The encoding of morphological knowledge. Word Structure 12. 274–296. https://doi.org/./word.. Axelsdóttir, K. 2014. Sögur af orðum. Sex athuganir á beygingarþróun í íslensku. Reykjavík: Háskólaútgáfan. Axelsdóttir, K. 2015. Beyging og merking orðsins hjalt. Orð og tunga 17. 95–114. https://doi.org/./ordogtunga.. Barðdal, J. 2008. Productivity: Evidence from case and argument structure in Icelandic. Amsterdam: Benjamins. https://doi.org/./cal. Barðdal, J. 2011. The rise of dative substitution in the history of Icelandic: A diachronic construction grammar approach. Lingua 121(1). 60–79. https://doi.org/./j.lingua... Barðdal, J. 2012. Predicting the productivity of argument structure constructions. Berkeley Linguistics Society 32 (2006). 467–478. https://doi.org/./bls.vi. Barlow, M. & S. Kemmer. 2000. Usage-based models of language. Stanford, CA: CSLI. Beckner, C., R. Blythe, J. Bybee, M. H. Christiansen, W. Croft, N. E. Ellis, J. Holland, J. Ke, D. Larsen-Freeman & T. Schoenemann. 2009. Language is a complex adaptive system: Position paper. Language Learning 59. 1–26. https://doi.org/./j.-...x Blevins, J. P. & J. Blevins. 2009. Analogy: An introduction. In J. P. Blevins & J. Blevins (eds.), Analogy in grammar: Form and acquisition, 1–12. Oxford: Oxford University Press. https://doi.org/./acprof:oso/.. Booij, G. 2010. Construction morphology. New York: Oxford University Press. Booij, G. 2019. The role of schemas in construction morphology. Word Structure 12(3). 385–395. https://doi.org/./word.. Booij, G. & J. Audring. 2018. Partial motivation, multiple motivation: The role of output schemas in morphology. In G. Booij (ed.), The construction of words. Advances in construction morphology, 59–80. Cham: Springer. https://doi.org/./----_ Bulloch, M. J. & J. E. Opfer. 2009. What makes relational reasoning smart? Revisiting the perceptual-to-relational shift in the development of generalization. Developmental Science 12(1). 114–22. https://doi.org/./j.-...x Bybee, J. 1985. Morphology: A Study of the relation between meaning and form. Amsterdam: Benjamins. https://doi.org/./tsl. Bybee, J. 2001. Phonology and language use. Cambridge: Cambridge University Press. https://doi.org/./CBO Bybee, J. 2007. Frequency of use and the organization of language. New York: Oxford University Press. https://doi.org/./acprof:oso/.. Bybee, J. 2010. Language, usage and cognition. New York: Cambridge University Press. https://doi.org/./CBO Bybee, J. 2015. Language change. Cambridge: Cambridge University Press. https://doi.org/./CBO Bybee, J. & J. L. McClelland. 2005. Alternatives to the combinatorial paradigm of linguistic theory based on domain general principles of human cognition. In N. A. Ritter (ed.), The role of linguistics in cognitive science. Special Edition of The Linguistic Review 22(2–4). 381–410. https://doi.org/./tlir...-. © 2022. John Benjamins Publishing Company All rights reserved 219 220 Jón Símon Markússon Bybee, J. & C. L. Moder. 1983. Morphological classes as natural categories. Language 59. 251–270. [Reprinted in Bybee 2007] https://doi.org/./ Cienki, A. J. 2015. Spoken language and usage events. Language and Cognition 7. 499–514. https://doi.org/./langcog.. Clahsen, H. 1999. Lexical entries and rules of language: A multidisciplinary study of German inflection. Behavioral and Brain Sciences 22. 991–1060. https://doi.org/./SX Clausner, T. C. & W. Croft. 1997. Productivity and schematicity in metaphors. Cognitive Science 21(3). 247–282. https://doi.org/./scog_ Cordes, A. K. 2017. The roles of analogy, categorization, and generalisation in entrenchment. In H. Schmid (ed.), Entrenchment and the psychology of language: How we reorganise and adapt linguistic knowledge, 269–88. Washington, DC: American Psychological Association. Croft, W. 2000. Explaining language change: An evolutionary approach. Harlow: Longman. Croft, W. & D. A. Cruse. 2004. Cognitive linguistics. Cambridge: Cambridge University Press. https://doi.org/./CBO De Smet, H. 2017. Entrenchment effects in language change. In H. Schmid (ed.), Entrenchment and the psychology of language: How we reorganise and adapt linguistic knowledge, 75–99. Washington, DC: American Psychological Association. https://doi.org/./- Dressler, W. U. 1997. On productivity and potentiality in inflectional morphology. CrossLanguage Aphasia Study Network (CLASNET) Working papers 7. 3–22. Dressler, W. U. 2003. Degrees of grammatical productivity in inflectional morphology. Italian Journal of Linguistics 15. 31–62. Fertig, D. 2013. Analogy and morphological change. Edinburgh: Edinburgh University Press. https://doi.org/./ Gardani, F. 2013. Dynamics of morphological productivity: The evolution of noun classes from Latin to Italian. Leiden: Brill. https://doi.org/./ Gentner, D. 2005. The development of relational category knowledge. In L. Gershkoff-Stowe & D. H. Rakison (eds.), Building object categories in developmental time, 245–275. New Jersey: Lawrence Erblaum Associates. Gentner, D. & C. Hoyos. 2017. Analogy and abstraction. Topics in Cognitive Science 9. 672–93. https://doi.org/./tops. Gentner, D. & A. B. Markman. 1997. Structure mapping in analogy and similarity. American Psychologist 52. 45–56. https://doi.org/./-X... Halford, G. S. & G. Andrews. 2007. Domain general processes in higher cognition: Analogical reasoning, schema induction and capacity limitations. In M. J. Roberts (ed.), Integrating the mind: Domain general versus domain specific processes in higher cognition, 213–232. New York: Psychology Press. Haspelmath, M. 2006. Against markedness (and what to replace it with). Journal of Linguistics 42(1). 25–70. https://doi.org/./S Houghton, D. E. 1968. Humor as a factor in language change. The English Journal 57. 1178–1181–1186. https://doi.org/./ Íslensk orðabók. 2010. Reykjavík: Forlagið. Kay, P. & C. J. Fillmore. 1999. Grammatical constructions and linguistic generalizations: The what’s X doing Y? construction. Language 75(1). 1–33. https://doi.org/./ Kvaran, G. 2005. Íslensk tunga II: Orð. Reykjavík: Almenna bókafélagið. Magnússon, Á. B. 1989. Íslensk orðsifjabók. Reykjavík: Orðabók Háskólans. © 2022. John Benjamins Publishing Company All rights reserved Graded membership of an Icelandic microclass Markússon, J. S. 2012. Eðli u-hljóðvarpsvíxla í íslenskri málsögu. Reykjavík: University of Iceland (MA thesis). Markússon, J. S. 2017. Samband veiklunar og hljóðanvæðingar: Vitnisburður u- hljóðvarpví xla í frum- og vesturnorrænni málsögu. In S. S. Hansen, A. Johansen, H. P. Petersen, L. Reinert (eds), Bók Jógvan: Heiðursrit til Jógvan í Lon Jacobsen á 60 ára degnum, 263–276. Tórshavn: Føroya fróðskaparfelag. Markússon, J. S. 2021. Undir áhrifum (orða)gengis. In K. Axelsdóttir et al. (eds.), Möggubrár heklaðar Margréti Jónsdóttur sjötugri, 21. mars 2021, 99–104. Reykjavík: Rauðhetta. Markússon, J. S. 2022. Accounting for different rates of gender reanalysis among Icelandic masculine forms in plural -ur. Nordic Journal of Linguistics. https://doi.org/./S OED = Oxford English Dictionary. Oxford: Oxford University Press. www.oed.com Penn, D. C., K. J. Holyoak & D. J. Povinelli. 2008. Darwin’s mistake: Explaining the discontinuity between human and nonhuman minds. Behavioral and Brain Sciences 31. 109–178. https://doi.org/./SX Quinion, M. Bloke. World Wide Words, 16th June 2004 (updated 22nd January 2011). http:// www.worldwidewords.org/qa/qa-blo3.htm (accessed 17 April 2021). Rácz, P., C. Beckner, J. B. Hay & J. B. Pierrehumbert. 2020. Morphological convergence as online lexical analogy. Language 96(4). 1–36. Rosch, E. 1978. Principles of categorization. In E. Rosch & B. Lloyd (eds.), Cognition and categorization, 27–48. Hillsdale, NJ: Erlbaum. Rosch, E. & C. B. Mervis. 1975. Family resemblances: Studies in the internal structure of categories. Cognitive Psychology 7. 573–605. https://doi.org/./-()- Rögnvaldsson, E. 2013. Hljóðkerfi og orðhlutakerfi íslensku. [Electronic edition] Reykjavík. https://notendur.hi.is/eirikur/hoi.pdf Svavarsdóttir, Á. Er virkilega hægt að drekka kók í þremur kynjum? Vísindavefurinn, 11 December 2014. http://visindavefur.is/svar.php?id=68642 (accessed 12 June 2020). Sveinsson, S. 2001. Íslensk málsaga. Reykjavík: Iðunn. Taylor, J. R. 2012. The mental corpus. Oxford: Oxford University Press. https://doi.org/./acprof:oso/.. Tomasello, M. 2000. First steps toward a usage-based theory of language acquisition. Cognitive Linguistics 11 (1/2). 61–82. Whelpton, M., D. Trotter, Þ. G. Beck, C. Anderson, J. Maling, K. Durvasula & A. Beretta. 2014. Portions and sorts in Icelandic: An ERP study. Brain and Language 136. 44–57. https://doi.org/./j.bandl... Wiese, H. & J. Maling. 2005. Biers, kaffi and schnaps: Different grammatical options for restaurant talk coercions in three Germanic languages. Journal of Germanic Linguistics 17. 1–38. https://doi.org/./S Winters, M. E. 2020. Historical linguistics: A cognitive grammar introduction. Amsterdam: Benjamins. https://doi.org/./z. Þórhallsdóttir, G. 1997. Ylgr, heiðr, brúðr: saga r-endingar nefnifalls eintölu kvenkynsorða. In Ú. Bragason (ed.), Íslensk málsaga og textafræði, 41–56. Reykjavík: Stofnun Sigurðar Nordals. Þórhallsdóttir, G. 2007. The dative singular of ō-stems in Old Norse. In A. J. Nussbaum (ed.), Verba docenti: Studies in historical and Indo-European linguistics presented to Jay H. Jasanoff by students, colleagues, and friends, 329–41. Ann Arbor, MI: Beech Stave Press. © 2022. John Benjamins Publishing Company All rights reserved 221 222 Jón Símon Markússon Address for correspondence Jón Símon Markússon Faculty of Icelandic and Comparative Cultural Studies University of Iceland Sæmundargata 2 101 Reykjavík Iceland jsm2@hi.is https://orcid.org/0000-0002-3280-2652 Publication history Date received: 14 May 2021 Date accepted: 18 April 2022 © 2022. John Benjamins Publishing Company All rights reserved Paper III Um áhrif tíðni á stefnu útjöfnunar: Rannsókn á beygingarþróun færeysku nafnorðanna vøllur og fjørður. Íslenskt mál og almenn málfræði, 44, 53–86. Paper III 185 jón símon markússon Um áhrif tíðni á stefnu útjöfnunar Rannsókn á beygingarþróun færeysku nafnorðanna vøllur og fjørður 1. Inngangur Hér verður gerð grein fyrir ólíkum stefnum útjöfnunar í færeysku nafnorðunum vøllur og fjørður, sem bæði eru afkomendur fornvesturnorrænna u-stofna, sbr. fvnorr. vǫllr og fjǫrðr.1 Eins og flokkunin gefur til kynna tilheyrðu orðin sama beygingarflokki í fornvesturnorrænu og sýndu því sömu beygingarendingar.2 Einnig voru fjölskrúðug stofnsérhljóðavíxl einkennandi fyrir beygingu u-stofna, sbr. t.d. fvnorr. nf./þf.et., þf./þgf.ft. vǫll-, fjǫrð- ~ þgf.et., nf.ft. vell-, firð- ~ ef.et. og ft. vall-, fjarð-. Hér er útjöfnun skilgreind þannig að beygingarmynd öðlist hliðarmynd sem að forminu til er byggð á annarri stofnmynd sama orðs (t.d. Carstairs-McCarthy 2015:327). Þó er brýnt að átta sig strax á því að útjöfnun breytir ekki eldri mynd í yngri, heldur eru yngri myndir viðbót við beygingardæmið (Bybee 2015:95). Enn fremur geta myndirnar sem til verða við útjöfnun lifað öldum saman við hlið eldri myndanna, en þó er algengast að ein hliðarmynd sigri hina(r) á endanum (sjá t.d. Harald Bernharðsson 2004, 2005; Jón Símon Markússon 2017, 2021, 2022a, 2022b; Katrínu Axelsdóttur 2015). Stig af stigi hefur stofnmyndin vøll- verið alhæfð um allt beygingardæmi orðsins vøllur, en stofnmyndin vell- er með öllu horfin. 1 Ég þakka þremur ónafngreindum ritrýnum og ritstjórum Íslensks máls fyrir góðar ábendingar. Einnig fá Bartal Torkilsson Kamban, Eivind Weyhe, Erna Björk Gestsdóttir, Hjalmar P. Petersen, Katrín Axelsdóttir, Lena Reinert og Uni Johannesen þakkir fyrir yfirlestur á fyrri gerðum greinarinnar og/eða aðstoð við fyrirspurnir mínar um orðanotkun í færeysku. Archie Cretton skulda ég þakkir fyrir aðstoð við tölfræðina. Höfundur ber ábyrgð á öllum villum sem eftir kunna að standa. Árið 2018 bar ég hugmyndina að greininni undir Anfinn Johansen, lektor við Føroyamálsdeild Fróðskaparseturs Færeyja, fyrstan manna. Ég fékk aldrei tækifæri til að þakka honum í eigin persónu fyrir hvatningu og aðstoð við rannsóknina áður en hann lést 10. janúar 2022. Í staðinn tileinka ég honum þessa grein. Rannsóknin var styrkt af Rannís, styrknúmer 174253-015. 2 Fyrir utan fvnorr. kvk. hǫnd og hk. fé eru u-stofnar karlkynsorð í norrænu. Hér verður aðeins fengist við beygingarsögu karlkyns u-stofna. Íslenskt mál 44 (2022), 53–86. © 2022 Íslenska málfræðifélagið, Reykjavík. 54 Jón Símon Markússon Að auki hafa tvær nýjar myndir með stofnmyndinni vall- bæst við. Aftur á móti hefur orðið fjørður haldið gömlum stofnsérhljóðavíxlum sínum fram til nútímans, auk þess að bæta við sig nýjum myndum. Markmið þessarar rannsóknar er fyrst og fremst að bera kennsl á þær ólíku beygingarmyndir sem útjöfnun hefur stefnt úr, þ.e. grunnmyndir (e. basic forms) beygingardæmanna (t.d. Bybee 2015:91–102), sem og hugrænu forsendurnar fyrir þróuninni sem varð. Ekki verður þó leitast við að tímasetja einstakar breytingar, enda færeyskar málheimildir fyrri alda af skornum skammti. Hugrænar forsendur útjöfnunar eru álitnar vera þær sömu í öllum málsamfélögum á öllum tímum og því taldar skipta meira máli til skilnings á eðli viðkomandi breytinga en nákvæm tímasetning þeirra (t.d. Beckner o.fl. 2009). Málfræðingar hafa aðallega nálgast útjöfnun úr tveimur ólíkum áttum. Samkvæmt eldri nálguninni ráða þættir á borð við merkingu orðs, tíðni einstakra mynda og endingarleysi ferðinni og er þá vísað í svokallaða mörkun (e. markedness). Í seinni tíð hefur málnotkun eins og hún endurspeglast í textasöfnum verið gert hærra undir höfði: áhrif tíðni á almenn hugræn ferli (e. domain-general cognitive processes) skeri þá úr um ákvörðun grunnmynda(r), óháð merkingu orðs eða notkunarsviði þess. Beygingarþróun fær. vøllur og fjørður er því tilvalin til athugunar þar sem orðin tilheyrðu sama flokki í eldra máli og vísa hvort til síns staðfræðilega fyrirbæris. Auk þess koma bæði orðin fyrir sem seinni liðir í samsettum örnefnum. Því kallar ólík þróun orðanna á spurninguna í (1). (1) Er tíðni ráðandi afl í stefnu útjöfnunar? Með hliðsjón af tíðni í nútímafæreysku samkvæmt rafrænu textasafni, áætlaðri tíðni miðað við færeysku fyrri alda og sögulegri dreifingu ólíkra stofnmynda í beygingardæmum verður sýnt fram á að tíðni hefur ráðið stefnu útjöfnunar í fær. vøllur og fjørður. Enn fremur verða færð rök fyrir því að útjöfnun á grundvelli tíðni geti raskað gamalli afstöðu beygingarmynda hverra til annarra með þeim afleiðingum að merkingarmunur á eintölu og fleirtölu stuðli að sambandi tiltekinnar stofnmyndar við ákveðna merkingu. Jafnvel má rekja það til tíðni þegar slíks sambands forms og merkingar verður ekki vart. Efnisskipan greinarinnar er eins og hér segir. Í 2. kafla verða beygingardæmi fvnorr. vǫllr og fjǫrðr og færeyskra afkomenda þeirra borin saman. Í 3. kafla er fjallað um mismunandi tilraunir til að skýra stefnu útjöfnunar, þ.e. annars vegar tilraunir sem vísa í mörkun og/eða sérmörkun og hins vegar tilraunir með skírskotun til málnotkunar, þ.e.a.s. tíðni Um áhrif tíðni á stefnu útjöfnunar 55 mismunandi beygingarmynda í textaheimildum. Í 4. kafla verður rýnt í orðabókarskilgreiningar fær. vøllur og fjørður. Í 5. kafla verður kynnt rannsókn á tíðni beygingarmynda beggja orða í nútímamálinu og gerð tilraun til að áætla tíðni myndanna í eldri færeysku svo að hægt sé að ákvarða grunnmyndirnar á þeim tíma þegar útjöfnun hófst. Í 6. kafla er gerð grein fyrir stefnu útjöfnunar í hvoru beygingardæmi fyrir sig með hliðsjón af rannsókninni í kaflanum á undan. Niðurstöður eru teknar saman í 7. kafla. 2. Beygingarþróun u-stofna í færeyskri málsögu Eins og bent var á í inngangi eru eldri færeyskar málheimildir af mjög skornum skammti (sjá Guðvarð Má Gunnlaugsson 2000; einnig Barnes 2001:228–229). Hér verður því farið að dæmi Höskuldar Þráinssonar o.fl. (2012:407–408) og forníslensk dæmi höfð um beygingu fornvesturnorrænna nafnorða, enda lítill munur á fornvesturnorrænu málunum (Kristján Árnason 2011:4). Í (2) er sýnd beyging fvnorr. u-st. vǫllr og fjǫrðr (sjá Iversen 1972:61–2 og 17–21 um hljóðferlin sem hér eru nefnd). (2) et. nf. þf. þgf. ef. ft. nf. þf. þgf. ef. vǫllr vǫll velli vallar vellir vǫllu vǫllum valla fjǫrðr fjǫrð firði fjarðar firðir fjǫrðu fjǫrðum fjarða Eins og sést í (2) einkenndist beyging fvnorr. vǫllr og fjǫrðr af ólíkum stofnsérhljóðavíxlum, en þetta réðst af því hvert frumnorræna stofnsérhljóðið var.3 Í u-st. vǫllr er fornvesturnorræna stofnsérhljóðið ýmist ǫ, e eða a, en frn. *a lá til grundvallar. Þannig er kringt físl. ǫ komið af ókringdu frn. *a við u-hljóðvarp, sbr. t.d. frn. nf.et. *wallur > fvnorr. vǫllr, fyrir kringingaráhrif frn. *u á frn. *a. Stofnsérhljóðið fvnorr. e er tilkomið vegna i-hljóðvarps, sbr. t.d. frn. þgf.et. *wallijē > fvnorr. velli, fyrir frammælingaráhrif frn. *ij á frn. *a. Myndir með a í stofni erfðu frn. *a óbreytt. 3 Til voru u-stofnar sem sýndu önnur stofnsérhljóðavíxl við beygingu, t.d. fvnorr. ust. þttr, sbr. nf./þf.et./þf./þgf.ft. þtt- ~ þgf.et./nf.ft. þætt- ~ ef.et. og ft. þátt-. Einnig voru til þríkvæðir u-stofnar, t.d. fvnorr. fǫgnuðr (seinna líka fagnaðr), mánaðr, dǫgurðr, skilnaðr (sjá Iversen 1972:61–2). Beygingarþróun þessara orða er ekki til skoðunar hér. Jón Símon Markússon 56 Hjá u-st. fjǫrðr hefur stofninn ýmist að geyma jǫ, i eða ja, en frn. *e lá til grundvallar. Þannig er fvnorr. jǫ komið af frn. *e við u-klofningu, sbr. t.d. frn. nf.et. *ferður > fvnorr. fjǫrðr. Stofnsérhljóðið fvnorr. i er tilkomið vegna i-hljóðvarps, sbr. t.d. frn. þgf.et. *ferðijē > fvnorr. firði. Svo á fvnorr. ja í eignarfallsmyndunum rætur sínar að rekja til a-klofningar af frn. *e, sbr. frn. *ferðar > fjarðar. Í (3) er beyging fær. vøllur og fjørður sýnd til samanburðar við beygingu fornvesturnorrænu fyrirrennara þeirra.4 (3) et. nf. þf. þgf. ft. nf./þf. þgf. vøllur vøll vølli fjørður fjørð firði/fjørði vøllir/vallir5 firðir/fjørðir/fjarðir6 vøllum/vallum fjørðum/firðum/fjarðum Eins og gildir um vøllur hafa gömul stofnsérhljóðavíxl fornra u-stofna iðulega verið jöfnuð út í tímans rás. Beyging orðanna børkur, gøltur, høttur, knørrur, knøttur, løgur, trøstur, vøttur, vøkstur; hjørtur, mjøður, stjølur ‘endinn á kornknippi’ eru til dæmis um þetta. Aftur á móti hefur fær. fjørður bætt við sig tveimur nýjungum, þ.e. þgf.et. fjørði og þgf.ft. firðum, auk þess að viðhalda gömlum víxlum.7 Ferlið hefur því ekki verið til einföldunar (sjá Petersen 2020:118). Eins og fær. fjørður hefur fær. táttur ‘þáttur’ varðveitt gömul stofnsérhljóðavíxl, en þó er víxlunum öðruvísi háttað. 4 Þótt gamlar eignarfallsmyndir sé að finna í færeysku, t.d. Vallarland, er eignarfallið horfið sem virkt fall í nútímafæreysku (sjá Höskuld Þráinsson o.fl. 2012:61–62). 5 Eins og Hjalmar P. Petersen hefur bent mér á má telja líklegt að nf./þf.ft. vallir sé tilkomið vegna áhrifa frá víxlum á borð við et. høll ~ nf./þf.ft. hallir, sbr. et. vøll- ~ nf./þf.ft. vallir. Því verður ekki gerð frekari grein fyrir tilurð umræddrar myndar. 6 Stofn beygingarmyndanna þf.ft. fjarðir og þgf.ft. fjarðum var e.t.v. fenginn að láni frá fvnorr./eldri fær. ef.et. fjarðar og/eða ef.ft. fjarða þegar eignarfallið var enn virkt fall. Þó þykir mér líklegra að beygingin et. fjørð- ~ nf./þf.ft. fjarðir hermi eftir t.d. fær. kvk. bjørn ‘björn’ ~ bjarnir. Ekki verður fjallað um myndir með stofninum fjarð- m.t.t útjöfnunar. 7 Í 5.2 verður gerð stuttlega grein fyrir notkun hliðarmynda m.t.t. mállýskubundinnar dreifingar þeirra. Um áhrif tíðni á stefnu útjöfnunar 57 3. Ráðandi öfl í stefnu útjöfnunar 3.1 Stefna útjöfnunar frá sjónarhóli (sér)mörkunar Þeir sem hafa viljað alhæfa um hvaða myndir séu líklegastar til að gegna hlutverki grunnmynda hafa aðallega nálgast útjöfnun frá tveimur ólíkum sjónarhornum. Samkvæmt eldri nálguninni eru einkenni á borð við merkingarfræðilegan „einfaldleika“ og/eða „hlutleysi“, hlutfallslega hærri staktíðni (e. token frequency) og formleg einkenni eins og endingarleysi til marks um ómarkaða (e. unmarked) stöðu orðmynda. Enn fremur þykja ómarkaðar myndir merkingarfræðilega „sjálfsagðari“ og líklegri til að koma fyrir en aðrar myndir á tilteknu notkunarsviði. Ómarkaðar myndir eru taldar líklegastar til að gegna hlutverki grunnmynda innan beygingardæmis (t.d. Jakobson 1939). Hvað beygingu nafnorða varðar taldi Greenberg (1966) til dæmis eintölu ómarkaða gagnvart fleirtölu; nefnifall ómarkað gagnvart öðrum föllum. Eiríkur Rögnvaldsson tekur í sama streng um íslenska beygingarkerfið á grundvelli beygingarlegra og setningafræðilegra einkenna mynda (t.d. 2013:147, 149–150, 162). Tölur og föll í íslensku og færeysku gegna svipuðu merkingar- og setningarhlutverki og því má ætla að eintala og nefnifall teljist líka ómörkuð í síðarnefnda málinu.8 Sé það rétt hefði mynd nefnifalls eintölu að öllu jöfnu stöðu grunnmyndar í færeyskum nafnorðum. Þó bendir ýmis beygingarþróun til að hugmyndir um mörkun séu alls ekki algildar (Sims-Williams 2016:331). Til dæmis benti Mańczak (1958) á að beyging örnefna ætti til að þróast með öðrum hætti en hjá samnöfnum almennt og skýrði það með svokallaðri sérmörkun: Vegna merkingar sinnar eru staðarheiti einatt notuð um hreyfingu til staðar og dvöl á stað og því eru staðarfallsmyndir þeirra „sjálfsagðari“, algengari og þar með ómarkaðri en til dæmis nefnifallsmyndir (sjá einnig Harald Bernharðsson 2004:25). Í færeysku eru þolfall, notað um ferð á stað/til staðar, og þágufall, notað um hreyfingu frá eða dvöl á stað, staðarföll og þar með talin ómörkuð þegar örnefni eiga í hlut, þ.e. í ljósi merkingar þeirra. Auk merkingar eru endingarleysi og hærri staktíðni álitin vísbending um ómarkaða stöðu beygingarmyndar (t.d. Greenberg 1966:100). Til dæmis telja talsmenn mörkunar að frekar megi búast við endingarleysi í eintölumyndum en fleirtölumyndum. Í færeysku eru fleirtölumyndir einmitt 8 Petersen (2009:89) telur nefnifall vera sjálfgefið (e. default) fall í íslensku og færeysku. 58 Jón Símon Markússon oftar merktar sérstakri endingu en eintölumyndir, auk þess sem eintölumyndir eru oft styttri en fleirtölumyndir, t.d. et. búð en ft. búðir, et. granni en ft. grannar. Tiltölulega stutt form er talið merki um ómarkaða stöðu gagnvart lengri myndum í beygingardæminu, en þó eru samsvarandi myndir eintölu og fleirtölu stundum jafnlangar, t.d. fær. nf.et. hundur, nf./þf.ft. hundar, þar sem ekki er gott að ákvarða ómarkaða stöðu á grundvelli endingarleysis eða lengdar myndanna. Bybee (t.d. 2010:20, 2015:38) hefur rætt almenna tilhneigingu í tungumálum sem felst í því að orðmyndir og aðrar málfræðilegar einingar séu styttri eftir því sem þær koma oftar fyrir, án þess þó að merking hafi nokkur áhrif á ferlið. Sem dæmi um slíkt má nefna beygingu orðsins fvnorr. tǫnn, sbr. fvnorr. nf./þf.ft. tenn, tennr eða teðr, einnig fær. tonn ~ nf./þf.ft. tenn. Þannig má skýra endingarleysi ft. tenn og alhæfingu þeirrar myndar í færeysku með því að hún er miklu algengari en eintölumyndin (sjá einnig Sims-Williams 2022), þrátt fyrir að endingarleysi hjá fleirtölumyndum kvenkynsorða sé sjaldgæft. Hér skal einnig bent á að notkun et. tonn má telja sjálfsagðari en notkun fleirtölumynda þegar um eina tönn er að ræða og því ekki gott að byggja hugmyndir um svokallaða mörkun á tíðni og endingarleysi einstakra mynda í bland við merkingu orðs.9 Um tíðni sem ráðandi afl í stefnu útjöfnunar má benda á þróun sem Tiersma (1982) lýsti hjá frísneskum nafnorðum, en hann tók eftir að t.d. frís. et. poel ‘laug’ skiptist á í beygingu við ft. pwollen og (yngri) poelen, þar sem síðarnefnda fleirtölumyndin er byggð á stofninum úr eintölunni, þ.e. poel-. Við frekari athugun kom í ljós að eintölumyndir orðsins voru tíðari en fleirtölumyndir þess, en svipaðar forsendur virðast fyrir víxlunum frís. et. hoer ‘hóra’ ~ ft. hworren/(yngri) hoeren o.fl. Aftur á móti skiptist t.d. frís. ft. jermen ‘handleggir’ á við tvímyndirnar et. earm og (yngri) jerm, þar sem yngri myndin á upptök í ft. jermen; sbr. sömu þróun hjá frís. et. kies/kjizze ‘tönn’, en yngri kjizze byggist á ft. kjizzen. Hvað yngri dæmin varðar eru fleirtölumyndir tíðari en eintölumyndir. Hér virðist því sem stefna útjöfnunar endurspegli upplifun málnotenda af aðstæðum í raunheiminum (e. real-world conditions) (Haspel9 Þetta viðhorf er almennt áberandi í skrifum þeirra sem nálgast útjöfnun frá sjónarhóli málnotkunar. Sem dæmi má vitna í ummæli Sims-Williams (2022:571) um áhrif tíðni á ákvörðun grunnmynda grískra sagna: „I chose to test token frequency rather than markedness for several reasons. The concept of markedness is vague (see Haspelmath 2006), and it is not clear how it can be measured without circularity. Lack of overt marking generally coincides with high token frequency, and it is likely that markedness effects are reducible to frequency effects (as claimed already by Greenberg 1966).“ Um áhrif tíðni á stefnu útjöfnunar 59 math 2006:45): af tilteknum orðum er t.d. algengara að nota fleirtölu en eintölu af því að orðin vísa til einhvers sem kemur í tvenndum eða settum eins og handleggir og tennur. Þó ræður grunnmerking slíkra orða ekki tíðni einstakra mynda þeirra, heldur tilefnið sem aðstæður í raunheimi gefa til notkunar hverju sinni. Með öðrum orðum er eðlilegt að eintölumyndir af orðinu handleggur séu notaðar um einn handlegg jafnvel þó að flest fólk sé með tvo handleggi (sjá einnig hér að framan um notkun eintölumynda af orðinu tonn). Eins og Haspelmath (2006, t.d. bls. 27, 58) og Bybee (t.d. 2015:102) hafa bent á er tíðni mælanlegri og því afsannanlegri mælikvarði en mörkun. Sé tekið tillit til orðanna fær. vøllur og fjørður sérstaklega er athyglisvert að beygingarþróun þeirra skuli hafa verið svo ólík í ljósi merkingarfræðilegs snertiflatar þeirra; því er e.t.v. ástæða til að draga algildi þess sem felst í hugmyndum um (sér)mörkun í efa. Í því sem á eftir fer verður því farið að tillögu Haspelmaths (2006) og kannað hvort skýring á stefnu útjöfnunar sem byggist á ólíkri tíðni einstakra mynda dugi ein sér. 3.2 Málnotkun: Samband tíðni, rótfestu og minnisstyrks Hér verður gerð grein fyrir stefnu útjöfnunar í ljósi áhrifa frá almennum hugrænum ferlum. Meðal þeirra hugrænu ferla sem helst koma til álita er rótfesta (e. entrenchment) þekkingar, sem felst í stöðugri uppfærslu á þeirri reynslu sem maðurinn öðlast á lífsleiðinni og geymir í minni sér (Schmid 2017). Með öðrum orðum búa minningar um hegðun stöðugt betur um sig í minni því meira sem hegðunin er æfð og það sama gildir um reynslu af málnotkun: því oftar sem beygingarmynd kemur fyrir í málinu þeim mun rótfastari verður hún (Beckner o.fl. 2009; Bybee 2010:8, 2015:95, 97). Af þessu leiðir að því rótfastari sem orðmynd er í minni því tiltækari er hún til notkunar, þ.e.a.s. því meiri minnisstyrk (e. lexical strength) býr hún yfir. Hún er því líklegri en sjaldgæfari mynd(ir) til að gegna hlutverki grunnmyndar (t.d. Bybee 2015:96–97, 101–102; Langacker 1987:59; Schmid 2017:9). Frá sjónarhóli málnotkunar skiptir tíðni því sköpum fyrir stefnu útjöfnunar (sjá einnig umræðuna í 3.1). Í þessu sambandi skal kynnt hugtakið dreifitíðni stofnmynda (e. intraparadigmatic dispersion). Hún er ákvörðuð á grundvelli þess aðgreinda merkingar- og/eða beygingarfræðilega samhengis sem tiltekin málfræðileg eining kemur fyrir í (sjá Gries og Ellis 2015). Með öðrum orðum ræðst dreifitíðni stofnmyndar af fjölda aðgreindra bása innan beygingardæmisins þar sem hana er að finna. Til dæmis má nefna að í elstu fornvestur- 60 Jón Símon Markússon norrænu innihéldu fjórir básar í beygingardæmi orðsins fjǫrðr stofnmyndina fjǫrð- (nf./þf.et., þf./þgf.ft.), tveir stofnmyndina firð- (þgf.et., nf.ft.) og tveir fjarð- (ef.et. og ft.), sbr. (2). Dreifitíðni stofnmyndarinnar fjǫrð- var því tvisvar sinnum hærri en dreifitíðni firð- og tvisvar sinnum hærri en dreifitíðni fjarð-; með öðrum orðum, stofnmyndin fjǫrð- hafði sömu dreifitíðni og báðar síðarnefndu myndirnar samanlagt. Há dreifitíðni getur líka stuðlað að rótfestu, en búi tiltekin stofnmynd yfir bæði hárri stak- og dreifitíðni má gera ráð fyrir að hún verði tiltækari úr minni en aðrar stofnmyndir. Einnig ber að nefna mynsturstíðni (e. type frequency), en í beygingarfræðilegu samhengi ræðst hún af því hve mörg orð beygjast eftir ákveðnu beygingarmynstri (t.d. Bybee 2007:14). Í færeyku er mikill fjöldi karlkynsorða a- og an-stofnar og því er mynsturstíðni þeirra mun hærri en hjá þeim orðum sem beygjast eins og fjørður. Jóhanna Barðdal (2008) og Bybee (t.d. 2010:67) hafa sýnt fram á sterk tengsl milli mynsturstíðni og þess hversu oft víxlamynstur verður útvíkkað, þ.e. hve virkt (e. productive) mynstrið reynist. Í færeyskri málsögu hafa karlkyns a- og an-stofnar sýnt mikla virkni, en hana má því tengja við háa tíðni viðkomandi mynstra og þar með einnig rótfestu þeirra í minninu (sjá Jost og Christiansen 2017; Frost o.fl. 2015; sjá einnig umræðu í 6.1.1). Loks má nefna að málnotkunarnálgunin gerir ráð fyrir skemum (e. schemas) sem taka tillit til tíðni og eru í stöðugri uppfærslu vegna rótfestu (sjá hér að framan). Skema er rótföst alhæfing sem gerir málnotanda kleift að reiða sig á vísigildi (e. cue validity), sem á rætur að rekja til tölfræðilegrar dreifingar málfræðilegs fyrirbæris (sjá t.d. Tuggy 2007:83; Cordes 2017: 281).10 3.3 Samband forms og merkingar Í 3.1 var ákveðið að fylgja Haspelmath (2006) og vísun til (sér)mörkunar við að skýra stefnu útjöfnunar var talið ofaukið. Því fer þó fjarri að merking orðmynda hafi engin áhrif á ferlið, enda hægt að sýna fram á tilhneigingu til meiri samræmingar á formi mynda sem tjá tiltekið beygingargildi eftir því sem munur á því og andstæðu gildi endurspeglar mikilvægan merkingarmun (Bybee 1985, 2015:105). Sem dæmi má nefna að meiri merkingarmunur felst í aðgreiningu á gildum beygingarþáttarins tölu, 10 Hér fylgi ég skilgreiningu Taylors (2012:187): „The cue validity of feature f with respect to category C is the probability of C given f, i.e. p(C | f).“ Um áhrif tíðni á stefnu útjöfnunar 61 þ.e. eintölu og fleirtölu, en á gildum beygingarþáttarins falls, þ.e. nefnifalli, þolfalli, þágufalli og eignarfalli. Með öðrum orðum er meiri merkingarmunur á myndunum fær. nf.et. fuglur og nf./þf.ft. fuglar en á myndunum fær. nf.et. fuglur og þf.et. fugl. Í sögulegu samhengi má nefna beygingarþróun fe. fōt til glöggvunar, sbr. síðfornensku andstæðuna et. fōt- ‘fótur’ : ft. fēt-, en í snemmfornensku komu hins vegar báðar stofnmyndirnar fyrir í eintölu og fleirtölu. Síðfornensku útkomuna telur Lahiri (2000:7) sýna að munurinn á eintölu og fleirtölu skipti meira máli fyrir merkingu en munur á falli innan sömu tölu, til samræmis við fimmta „lögmál“ Kuryłowicz (1945–1949).11 Samband forms og merkingar felst því í tilhneigingu til að eigna tilteknu formi ákveðna merkingu, þ.e. fe. fōt- = eintala, en fe. fēt- = fleirtala (sjá Katrínu Axelsdóttur 2014:145 o. áfr.). 4. Merking orðanna vøllur og fjørður 4.1 Merking orðsins vøllur Føroysk orðabók (1998) gefur eftirfarandi þrjár skilgreiningar á orðinu vøllur. (4) a. Grasgrógvin slætti, vallað fløta; b. (bjarg.) vallrók, torva; c. (ítr.) leikvøllur, ítróttavøllur. Orðið vøllur vísar þá í a.m.k. þrennt: grasi gróinn flöt í (4a); grasi gróna syllu í bjargi í (4b); leik- og/eða íþróttavöll eins og í (4c). En auk þessara merkinga er vøllur einnig notað sem stytting á fær. flogvøllur ‘flugvöllur’, oftast með vísun til flugvallarins í Vogum, sem er eini flugvöllur landsins. Leit að orðinu vøllur sem örnefni á vefnum Føroyakort skilar aðeins einni niðurstöðu, þ.e. Harðavøllur.12 Einnig er örnefnið Tinghúsvøllur til, en vefurinn skilaði engu dæmi um þetta orð. Leit í færeyska textasafninu Teldutøka tekstasavn Føroyamálsdeildarinnar (TTF) skilaði einni niðurstöðu fyrir leitarorðið Harðavøllur en engri fyrir Tinghúsvøllur.13 Áhuga11 Hér má einnig nefna að Eiríkur Rögnvaldsson (2013:143) gerir greinarmun á formdeildum eftir eðli þeirra og telur tölu vera grunnþátt í nafnorðum en fall aðlögunarþátt. 12 Á vefnum <https://kort.foroyakort.fo/kort/> er hægt að sjá staðsetningu á Færeyjakorti með því að fletta upp staðarheitinu í leitarreitnum. Einnig gefur leitin upp öll staðarheiti með viðkomandi leitarorði. 13 Textasafnið er að finna hér: <http://www.teldni.fo/tekstasavn/index>. Við leit á vefnum er gefin upp staktíðni viðkomandi orðmyndar samkvæmt notkun hennar í ýmsum miðlum, þ.á m. vefsíðum og bókum. Jón Símon Markússon 62 vert er að hvorugur vefurinn hefur að geyma samsetningu með ft. -vøllir að seinni lið, sambærilega við t.d. ísl. Þingvellir. Þó hefur Weyhe (pers. uppl.) bent mér á staðarheitið Álavøllir, sem er í grennd við byggðina Sumba á Suðurey. Einnig má finna nokkur dæmi um staðarheiti með ft. -vøllir hjá Matras (1933). Að sögn Weyhe eru vellir í Færeyjum yfirleitt mjög litlir og því sjaldnast merktir á kortum. 4.2 Merking orðsins fjørður Føroysk orðabók (1998) gefur eftirfarandi þrjár skilgreiningar undir flettunni fjørður. (5) a. (Langur) vágur el. vík inn í land (vanl. minni til breiddar enn til longdar); b. breitt sund, sjógvur millum oyggjar; c. lendi kring fjørð. Orðið fjørður vísar þá í þrennt: ílanga vík í (5a); opna sjávarleið milli eyja, þ.e. sund í (5b);14 og landið meðfram firði í (5c). Orðið fjørður kemur einnig fyrir í fjölda samsettra staðarheita, t.d. Fuglafjørður, Kollafjørður, Skopunarfjørður. Ólíkt því sem gildir um vøllur er aðeins eitt færeyskt örnefni með fleirtölumyndinni firðir: Firðirnir í Vestmanna.15 4.3 Tengsl -vøllur og -fjørður og ósamsettu orðanna vøllur og fjørður Í ljósi orðabókarskilgreininganna í 4.1 og 4.2 og þeirrar staðreyndar að bæði fær. vøllur og fjørður mynda samsett örnefni vaknar spurningin: Tilheyra fær. -vøllur og -fjørður í samsettum örnefnum beygingardæmum fær. vøllur og fjørður eða eingöngu beygingardæmi viðkomandi örnefnis? Með öðrum orðum: Er t.d. -vøllur í Tórsvøllur dæmi um notkun orðsins vøllur eða ekki? Í þessu sambandi er litið svo á að forsendan fyrir því að málnotendur tengi fær. -vøllur og -fjørður í samsettum örnefnum við merk14 Ritstjóri spyr hvort merkingin í (5b), þ.e. ‘sund’, sé nýjung í færeysku. Á vefnum <http://norskstadnamnleksikon.no/> kemur fram að „[fjord] tyder opphavleg truleg ‘gjennomfart (inn mot land), innfart.’“ Mörg örnefni með nafnorðinu -fjord í merkingunni ‘sund’, þ.e. opinni sjávarleið milli eyja, er að finna við vesturströnd Noregs, t.d. Gåsværfjorden, Mesøyfjorden, Stabbfjorden, Støttfjorden o.fl. Slík örnefni benda til að notkun no. -fjord og fær. fjørður í sömu merkingu eigi upptök í máli norrænna manna sem námu land í Færeyjum. 15 Þetta staðarheiti fannst þó ekki með leitarorðinu „firðirnir“ á vefnum Føroyakort. Um áhrif tíðni á stefnu útjöfnunar 63 ingarnar í 4.1 og 4.2 sé sú að í hverju einstöku samsettu orði með -vøllur og -fjørður að síðari lið svari merkingin til merkingar samsvarandi ósamsetts orðs (sjá t.d. Bybee 1985, 2010:22–23). Þar af leiðandi eru liðirnir -vøllur og -fjørður í samsettum orðum hér taldir heyra til orðanna vøllur og fjørður. 5. Rannsókn á tíðni einstakra beygingarmynda fær. vøllur og fjørður 5.1 Staktíðni einstakra mynda af orðinu vøllur 5.1.1 Nútímafæreyska Orðið vøllur í merkingunum í (4a–b) er sjaldgæft í færeysku. Við þessu hefði mátt búast, enda náttúrulegir vellir fáir og oftast fyrirferðarlitlir í Færeyjum vegna landshátta og því e.t.v. lítið tilefni til að ræða hversdagslega um þá (sjá umræðuna í 3.1). Tíðni orðsins vøllur stafar einkum af notkun þess í merkingunni í (4c), þ.e. sem styttingar á leik- og ítróttavøllur, og sem styttingar á flogvøllur. Í töflu 1 er sýnd tíðni allra beygingarmynda fær. vøllur eins og hún birtist í TTF miðað við notkun þess utan samsetninga, en alls skilaði textasafnið 448 niðurstöðum. Beygingarmyndir nf.et. þf.et þgf.et. nf.ft. vøllur vøll vølli vøllir vallir þf.ft. vøllir vallir þgf.ft. vøllum vallum Tíðni (alls 448) Hlutfall (%) 77 93 205 11 9 18 5 26 4 17,19% 20,76% 45,76% 2,46% 2,00% 4,02% 1,12% 5,80% 0,89% Alls (et./ft.) 375 73 Tafla 1: Tíðni einstakra beygingarmynda orðsins vøllur og hlutfall þeirra af 448 dæmum úr færeysku nútímamáli. Af töflu 1 má ráða að eintölumyndirnar eru langtum tíðari en fleirtölumyndir, eða 375 á móti 73. Þágufallsmyndirnar eru tíðari en aðrar myndir í báðum tölum — 205 (et.) á móti 30 (ft.) — eða 52,46% af fjölda allra mynda orðsins. Innbyrðis tíðni eintölumynda hækkar eins og sýnt er í (6), þar sem úr örinni má lesa ‘kemur sjaldnar fyrir en’. Jón Símon Markússon 64 (6) nefnifall (77) → þolfall (93) → þágufall (205) Hækkandi tíðni milli samsvarandi mynda fleirtölunnar fylgir sömu röð, sbr. 20 (nf.), 23 (þf.) og 30 (þgf.). Tíðni fallmynda innan hvorrar tölu, þar sem þágufallið er algengast, kemur væntanlega ekki á óvart, enda er völlur staðfræðilegt fyrirbæri. Þar að auki liggja leik- og/eða íþróttavellir og flugvellir venjulega á sléttum flötum og því eðlilegt að nota viðkomandi orð í tengslum við dvöl á stað, þ.e.a.s. í þágufalli, auk hreyfingar á stað, þ.e.a.s. í þolfalli (sjá hér að neðan). Hvað notkun orðsins í mismunandi merkingu snertir hefur 51 af 77 dæmum um nf.et. vøllur ótvírætt merkinguna ‘leikvöllur’ og 17 merkinguna ‘flugvöllur’.16 Auk þess falla níu dæmi undir merkingarliðina í (4a– b), þ.e. grasi gróinn flöt eða grasi gróna syllu í bjargi. Leit að þf.et. vøll skilaði alls 93 dæmum, þar af 50 í merkingunni í (4c), auk níu í merkingunni ‘flugvöllur’ og 34 niðurstöðum í merkingunum í (4a–b). Um þgf.et. vølli fundust alls 205 dæmi, þar af 189 um merkinguna í (4c), tíu í merkingunni ‘flugvöllur’, en sex sem flokkast eftir merkingunum í (4a–b). Þessar upplýsingar eru teknar saman í töflu 2, þar sem hlutfallstölurnar miðast við alls 375 dæmi. Mynd (4c) flogvøllur (4a–b) Alls nf.et. vøllur þf.et. vøll þgf.et. vølli Samanlagt Hlutfall (%) 51 50 189 290 77,33% 17 9 10 36 9,6% 9 34 6 49 13,07% 77 93 205 375 Tafla 2: Tíðni mynda nefnifalls, þolfalls og þágufalls eintölu af orðinu vøllur í mismunandi merkingu auk hlutfalls notkunar eftir merkingu. Hvað notkun í samsetningum varðar sýnir leit í TTF að tíðni orðsins vøllur snarhækkar í því samhengi (sjá umræðuna í 4.3). Sem dæmi voru skoðaðar myndirnar nf. Tórsvøllur, þf. Tórsvøll og þgf. Tórsvølli. Um nefnifallsmyndina er að finna 139 dæmi, 218 dæmi um þolfallsmyndina, en 215 um þágufallsmyndina. Enn og aftur kemur væntanlega engum á óvart að þolfalls16 Ég taldi setningar sem innihalda orðin fær. leikur, ítrótta-, lið, fótbólts- auk annarra vísbendinga, s.s. nafnsins á fótboltaliði, til samhengis sem staðfestir að um merkingu (4c) hefði verið að ræða. Um áhrif tíðni á stefnu útjöfnunar 65 og þágufallsmyndirnar reynist tíðari en nefnifallsmyndin, enda um afmarkaðan stað að ræða sem hægt er að fara á (fs. í/á + þf.), yfirgefa (fs. av/úr + þgf.), og staðsetja hlut eða verknað (fs. í/á + þgf.). Einnig voru skoðaðar eintölumyndir af samsetningunni fótbóltsvøllur, en staktíðni þeirra reyndist tiltölulega lág: 24 dæmi um nf.et. fótbóltsvøllur; 25 um þf.et. fótbóltsvøll, en aðeins þrjú dæmi um þgf.et. fótbóltsvølli. Að auki skilaði leit að þgf.ft. fótbóltsvøllum sex dæmum. Samkvæmt töflu 2 eru eintölumyndir af vøllur í merkingunni ‘flugvöllur’ ekki sérlega algengar, en við leit að mismunandi beygingarmyndum samsetta orðsins flogvøllur kemur annað í ljós: nf.et. flogvøllur kemur fyrir 96 sinnum, þf.et. flogvøll 205 sinnum, en þgf.et. flogvølli 75 sinnum. Tíðni þolfallsmyndarinnar kemur væntanlega ekki á óvart, enda eru flugvellir oftast nefndir í umræðum um hvernig og hvenær fólk fer þangað. Hins vegar snýst tölfræðin svolítið við þegar umræddar myndir eru notaðar með ákveðnum greini en þá skilaði leit 132 dæmum um nf.et.mgr. flogvøllurin, 446 um þf.et.mgr. flogvøllin en 651 fyrir þgf.et.mgr. flogvøllinum. Það á einnig við um ósamsetta nafnorðið vøllur að ákveðnar myndir þess eru langtum tíðari en samsvarandi óákveðnar myndir: nf.et.mgr. vøllurin 152; þf.et.mgr. vøllin 871; þgf.et.mgr. vøllinum 883. Í langflestum tilvikum hafa ákveðnu myndirnar af vøllur merkinguna í (4c) eða ‘flugvöllur’. Leit að myndum fótbóltsvøllur með ákveðnum greini skiluðu eftirfarandi tölum: um nf.et. fótbóltsvøllurin fundust aðeins níu dæmi, 92 niðurstöður um þf.et. fótbóltsvøllin, en 72 dæmi var að finna um þgf.et. fótbóltsvøllinum. Einnig hefur TTF að geyma tíu dæmi um nf./þf.ft. fótbóltsvøllir, sbr. töflu 3. Orð/staðarheiti nf.(et.) þf.(et.) þgf.(et.) Alls Hlutfall/4683 vøllur fótbóltsvøllur Tórsvøllur flogvøllur vøllurin fótbóltsvøllurin flogvøllurin Alls Hlutfall/4683 77 24 139 96 152 9 132 629 13,43% 93 25 218 205 871 92 446 1950 41,63% 205 3 215 75 883 72 651 2104 44,94% 375 52 572 376 1906 173 1229 4683 8,01% 1,11% 12,21% 8,03% 40,70% 3,70% 26,24% Tafla 3: Tíðni eintölumynda viðkomandi orða eftir falli (dálkar); eftir orðum (raðir) Jón Símon Markússon 66 Á grundvelli tölfræðinnar í töflu 3, sem byggist á upplýsingum úr TTF, má draga eftirfarandi ályktanir: 1. Í merkingunni ‘flugvöllur’ koma beygingarmyndir fær. vøllur langoftast fyrir með ákveðnum greini. 2. Tíðni fær. vøllur er almennt hæst þegar orðið er seinni liður í samsetningu. 3. Þó að tíðnimunur á myndum þolfalls og þágufalls sé ekki gríðarlega mikill þá er þgf.et. (-)vølli (með og án greinis) algengari en aðrar fallmyndir eintölunnar. Í ljósi þriðja liðarins væri e.t.v. réttast að ætla að þgf.et. vølli væri grunnmynd beygingardæmisins, miðað við það sem segir í 3.1 og 3.2. Þó er rétt að kanna hvort sú mynd hafi alltaf verið algengust, enda kom merkingin ‘flugvöllur’ ekki til sögunnar fyrr en á 20. öld. 5.1.2 Færeyska fyrri alda Heimildir um eldri málstig tungumála eru afar misjafnar. Hvað varðar íslenska málsögu er staðan t.d. allgóð; til er fjöldi ritaðra heimilda frá öllum tímabilum. Sambærilegum vitnisburði er ekki til að dreifa um þróun færeysku fram á seinni hluta 18. aldar (t.d. Guðvarður Már Gunnlaugsson 2000; einnig Barnes 2001:228–229; Höskuldur Þráinsson o.fl. 2012:370– 372). Þó hefur oft verið sýnt fram á að varðveisla eldri heimilda á öðrum fornvesturnorrænum málum getur komið að góðum notum þegar rýnt er í færeyska málsögu, enda margt líkt með málunum fram eftir öldum (Kristján Árnason 2011:4). Á grundvelli þess verður gerð tilraun til að áætla staktíðni orðsins vøllur í eldri færeysku miðað við tíðni þess í íslenskum heimildum. Í þessu sambandi ber auðvitað að hafa fyrirvara á þegar um áætlaða tölfræði er að ræða, en aftur á móti ber að styðjast við öll þau hjálpargögn sem í boði eru til að komast sem næst sannleikanum.17 Hvað sögulega tíðni orðsins vøllur varðar er merkingin ‘flugvöllur’ vitanlega yngri en merkingarnar í (4), þ.e. (4a) grasi gróinn flötur; (4b) grasi gróin sylla í bjargi; og (4c) leik- og/eða íþróttavöllur. Þar að auki fullyrðir Eivind Weyhe (pers. uppl.) að þegar hann var barn í Færeyjum — um og upp úr 1950 — hafi orðið plena verið notað um það sem heitir í 17 Aðrir hafa gert tilraunir til að áætla tíðni ýmissa málfræðilegra fyrirbæra og borið saman tíðni samkvæmt eldri málheimildum og innihaldi textasafna fyrir nútímamál til að geta gert grein fyrir málbreytingum. Sem dæmi má nefna Dobson 1973 og Kodner 2019. Um áhrif tíðni á stefnu útjöfnunar 67 dag ítróttavøllur eða fótbóltsvøllur.18 Aftur á móti er merkingin í (4c) alls ekki ný af nálinni, enda var samsetta orðið leikvǫllr notað þegar í fornvesturnorrænu, til dæmis í Bærings sögu fagra, íslenskri fornaldarsögu frá fyrri hluta 14. aldar. Því má telja sennilegt að Færeyingar til forna hafi líka þekkt orðið. Þó má af ýmsum ástæðum velta vöngum yfir því hversu oft orðið hafi komið fyrir í fornvesturnorrænu almennt. Í þessu sambandi færir til dæmis Gardeła (2012:239–240) rök fyrir því að utanhússíþróttir hafi síður tíðkast hjá norrænum mönnum fyrri alda en nú. Meðal ástæðna nefnir hann veður, kulda og skort á birtu yfir vetrarmánuðina. Þessu til stuðnings má nefna að samkvæmt fornum heimildum, t.d. Hálfdánar sögu Eysteinssonar (frá fyrri hluta 14. aldar), notuðu norrænir menn fornaldar stundum sali til íþróttahalds, en e.t.v. var þetta aukahlutverk salarins.19 Ætla má því að notkun fær. vøllur í merkingu (4c) hafi áður tíðkast í minna mæli en nú á dögum í ljósi þess að þar til gerðir leik- og/eða íþróttavellir voru fáir sem engir í Færeyjum fram til ársins 1911; þá var fyrsti viðurkenndi fótboltavöllur landsins lagður í Gundadal (Þórshöfn) (sjá Arge 1994).20 Því má ætla að tíðni allra beygingarmynda orðsins vøllur hafi verið lægri í eldra máli en nú er. Auk þess liggur í augum uppi að tíðni fær. vøllur í merkingunni ‘flugvöllur’ og öll dæmi af samsetningunni fótbóltsvøllur dragast frá ef tekið er mið af færeysku fyrri alda. Til þess að fá hugmynd um tíðni beygingarmyndanna fyrr á öldum verður stuðst við tölfræðina í töflu 2 til viðmiðunar. Þegar öll dæmi um orðið í merkingu (4c) og sem styttingar af orðinu flugvøllur hafa verið dregin frá standa eftir 49 dæmi sem gætu hafa átt sér samsvörun í eldri færeysku, þ.e. í merkingunum í (4a–b): níu um nf.et. vøllur, 34 um þf.et. vøll, en sex um þgf.et. vølli. Enn er eftir að áætla tíðni orðsins leikvøllur í eldri færeysku, en við mat á heimildum frá öðrum málsvæðum þyrfti líka að taka tillit til ólíkra aðstæðna í raunheimi, s.s. veðurs og landslags í Færeyjum (sjá hér að framan). 18 Þó er elsta skráða dæmið um notkun fær. fótbóltsvøllur frá 1943, samkvæmt leit á Timarit.is. 19 Hér er þó rétt að minna á að knattleik utanhúss er lýst í 40. kafla Egils sögu Skallagrímssonar. 20 Fyrsti leikurinn á Gundadalsvellinum fór fram 18. júní 1911. Fyrir þann tíma voru tveir óviðurkenndir vellir, annar í Hoydölum frá u.þ.b. 1904 og hinn, sem er frá seinni hluta 19. aldar, í Þvereyri. Ég þakka Brynjari Hlöðverssyni og Tormóði Peturssyni Djurhuus fyrir að veita mér upplýsingar um sögu fótboltavalla í Færeyjum og fyrir að benda mér á heimildir um þetta. Jón Símon Markússon 68 Samkvæmt leit í Íslensku textasafni (ÍT) kemur völlr/völlur fram í ýmsum myndum alls 307 sinnum í íslenskum heimildum.21 Hins vegar kemur samsetningin leikvǫllr/-völlr/-völlur fyrir í ýmsum myndum aðeins átta sinnum. Hér skal haft í huga að samsetningar á -vǫllr/-völlr/-völlur hafa alltaf komið langtum oftar fyrir í íslensku en færeysku. Til dæmis skilar leit í veforðabókinni/seðlasafninu ONP 26 dæmum um orðið völlur sem seinni lið í samsetta staðarheitinu ísl. Þingvǫllr. Þar sem vellir í Færeyjum eru hins vegar oftast mjög litlir eru þeir venjulega ekki sýndir á kortum (sjá 4.1). Þótt þetta merki auðvitað ekki að litlir vellir berist ekki í tal mætti hugsa sér að viðkomandi staðir væru nefndir á kortum ef þeir væru tengdir reglulegu íþróttahaldi eða gegndu öðru fastmótuðu hlutverki. Að auki er hugsanlegt að smæð færeyskra valla geri þá óhæfa sem leik- og íþróttavelli, en eins og segir í 5.1.1 bregður orðinu vøllur oftast fyrir í þeirri merkingu í nútímafæreysku, eftir að þartilgerðir vellir litu dagsins ljós á seinni öldum. Með hliðsjón af ofansögðu, að viðbættum þeim takmörkunum sem landfræðilegar og veðurfræðilegar aðstæður á Norðurlöndum höfðu á íþróttahald utanhúss, má gera ráð fyrir að samsetningar með -vǫllr/-vøllur hafi komið sjaldnar fyrir í færeysku fyrri alda en samsvarandi samsetningar í öðrum fornvesturnorrænum málum (einkum íslensku). Gefum okkur þó að tíðni fær. leikvøllur hafi ekki verið nema helmingi lægri miðað við dæmin átta úr ÍT, sem hefði þá samsvarað fjórum færeyskum dæmum. Áætluðu dæmin fjögur um leikvøllur fyrir færeysku fyrri alda bætast við dæmin sex um þgf.et. vølli úr töflu 2. Einnig er eina dæmið sem fannst um nf. Harðavøllur látið hífa upp tíðni nefnifallsmyndarinnar (sjá 4.1). Miðað við þessa útreikninga má áætla að samsvarandi tíðni nefnifalls-, þolfalls-, og þágufallsmynda í eldri færeysku hafi verið eins og í töflu 4. Beygingarmynd Nútímafæreyska Færeyska fyrri alda nf.et. vøllur þf.ft. vøll þgf.et. vølli 77 93 205 10 34 10 375 54 Alls Tafla 4: Samanburður á tíðni eintölumynda u-stofnsins vøllur í nútímafæreysku og færeysku fyrri alda. 21 Í Íslensku textasafni er hægt að fletta upp tíðni allra orðmyndanna sem safnið hefur að geyma í þeim textum sem valdir eru til athugunar. Textasafnið gefur einnig upp staktíðni Um áhrif tíðni á stefnu útjöfnunar 69 Ólíkt því sem gildir um nútímafæreysku benda áætluðu tölurnar í töflu 4 til þess að þolfallsmyndin hafi verið tíðari en hinar eintölumyndirnar samanlagt fyrr á öldum (sjá töflur 1, 2 og 3 í 5.1.1). Hækkandi tíðni milli mynda er sýnd í (7). (7) nefnifall (10)/þágufall (10) → þolfall (34) Í 6.1.2 verða áhrif bæði stak- og dreifitíðni metin í ljósi beygingarþróunar fær. vøllur. 5.2 Staktíðni einstakra mynda af orðinu fjørður Beygingarmyndirnar nf.et. fjørður, þf.et. fjørð, þgf.et. firði/fjørði, nf./þf.ft. firðir, þgf.ft. fjørðum/firðum/fjarðum koma fyrir alls 571 sinni í TTF. Tíðni einstakra orðmynda er sýnd í töflu 5. Beygingarmyndir nf.et. fjørður þf.et fjørð þgf.et. firði fjørði nf.ft. firðir22 þf.ft. firðir fjarðir þgf.ft. fjørðum firðum fjarðum Tíðni (alls 571) Hlutfall (%) 34 120 34 20 55 95 1 16 169 27 5,95% 21,02% 5,95% 3,50% 9,63% 16,64% 0,18% 2,80% 29,60% 4,73% Alls (et./ft.) 208 363 Tafla 5: Tíðni einstakra beygingarmynda fær. fjørður og hlutfall þeirra af heildarfjölda dæma (571). Hér er rétt að taka fram að stundum fer eftir mállýskum hvaða hliðarmyndir koma fyrir. Í þessu sambandi hef ég eftir Weyhe (pers. uppl.) að einstakra orðmynda auk beinnar tilvísunar í heimildina þar sem viðkomandi mynd er að finna. Til að komast að sögulegri tíðni beygingarmynda völlr/völlur í ÍT var hakað við valmöguleikann „Fornrit“. Leitin skilaði þessum niðurstöðum: völlur (og allar beygingarmyndir orðsins) = 235 dæmi; nf.et. völlr = 12 dæmi; þf.ft. völlu = 43 dæmi; nf.et.mgr. völlrinn = 1 dæmi; þf.ft. mgr. völluna = 16 dæmi; alls = 307 dæmi. Bein slóð er <http:// corpus.arnastofnun.is>. 22 Ekkert dæmi fannst um myndina fær. nf.ft. *fjarðir í textasafninu. 70 Jón Símon Markússon í Suðurey komi stofninn firð- alls ekki fyrir; þar hafi stofnmyndin fjørðverið alhæfð í eintölunni, sbr. nf.et. fjørður, þf.et. fjørð og þgf.et. fjørði, en fjarð- í fleirtölunni, sbr. nf.(?)/þf.ft. fjarðir, þgf.ft. fjarðum. Þessi þróun rennir stoðum undir hugmyndir um samband forms og merkingar, enda einkennist dreifingin hjá Suðureyingum af fyllidreifingu á grundvelli merkingar (sjá 3.3). Samkvæmt upplýsingum frá Weyhe kemur aftur á móti ýmist þgf.et. firði eða fjørði fyrir norðan Skopunarfjörð, t.d. munu Fuglfirðingar sjálfir nota þágufallsmyndina Fuglafirði. Enn fremur segir hann að Skálfirðingar geri greinarmun á þágufallsmynd staðarheitisins Skálafjørður eftir því hvort um fjörðinn sjálfan er að ræða, sbr. þgf.(mgr.) Skálafjørði(num), eða byggðina í kring, þá þgf. Skálafirði. Þetta merkir að á sumum stöðum eru tvímyndir notaðar í tali heimamanna, en e.t.v. í aðeins ólíkri merkingu. Eins og sagði í 4.2 kemur ft. firðir aðeins fyrir í einu staðarheiti í Færeyjum, þ.e. Firðirnir í Vestmanna. Það kann því að koma lesendum á óvart að fleirtölumyndir orðsins fjørður eru tíðari en myndir eintölunnar samkvæmt töflu 5, eða 208 (et.) á móti 363 (ft.), en þegar betur er að gáð hafa ýmsir þættir áhrif á niðurstöðurnar. Í fyrsta lagi eru færeyskir málnotendur líklegri til að nota fjarðarheiti í heilu lagi en að stytta það í Fjørður.23 Samsett örnefni á -fjørður er ekki að finna í töflu 5, en hefði verið tekið tillit til tíðni slíkra samsetninga væru eintölumyndir samt langtum fleiri en fleirtölumyndirnar (sjá hér að neðan). Þannig gefur tafla 5 e.t.v. töluvert skakka sýn á tíðni eintölumynda fær. fjørður þegar samsett örnefni eru talin með. Miðist tíðni eintölumynda orðsins fjørður einnig við samsett örnefni bætast öll dæmi um Kollafjørður, Fuglafjørður o.fl. við, enda ná skilgreiningarnar í (5a–c) utan um staðarheiti (sjá 4.3). Þó að aðeins tíðni staðarheitisins Fuglafjørður sé athuguð tekur staktíðni eintölumynda þá þegar langt fram úr dæmunum um fleirtölumyndir í töflu 5. Niðurstöður eftir leit í TTF eru svohljóðandi: 376 dæmi fundust um nf. Fuglafjørður, 122 dæmi um þf. Fuglafjørð, auk 1352 dæma um þágufallsmyndirnar, þ.e. 1335 fyrir Fuglafirði en 17 fyrir Fuglafjørði. Það sama gildir um beygingarmyndir örnefnisins Kollafjørður: nf. Kollafjørður skilaði 396 niðurstöðum, þf. Kolla23 Þótt það sé óvíst að upplýsingar um þetta komi fyrir í textasafninu, enda eru slíkar styttingar trúlega talmálsfyrirbæri fyrst og fremst, hef ég borið þetta undir færeyska móðurmálshafa og málvísindamenn til staðfestingar. Ég hef þó eftir Lenu Reinert: „Tað gjørdu tey fyrr her í Kaldbak um Kaldbaksbotn/-fjørð. Tey søgdu „at fara inn í Fjørð“, havi eg fingið fortalt. Tá búði eingin har.“ Um áhrif tíðni á stefnu útjöfnunar 71 fjørð 157, þgf. Kollafirði 986, Kollafjørði 27. Nákvæm sundurliðun er gefin í töflu 6.24 Orð/staðarheiti nf.(et.) fjørður Fuglafjørður Kollafjørður Samanlagt Hlutfall Hlutfall/3987 34 376 396 806 22,24% 20,22% þf.(et.) þgf.(et.)* 120 34/20 122 1335/17 157 986/27 399 2355/64 11,01% 64,98%/1,77% 10,01% 59,07%/1,61% Alls Hlutfall 208 1850 1566 3624 5,74% 51,05% 43,21% Hlutfall/ 3987 5,22% 46,40% 39,28% 90,90% 90,90% Tafla 6: Tíðni eintölumynda eftir falli (dálkar); eftir orðum (raðir); og heildarhlutfall (%) (eintölu)mynda (3624) fær. fjørður til samanburðar við samanlagðar tölur fyrir bæði eintölu- og fleirtölumyndir (3987). * Tölurnar í dálknum „þgf.(et.)“ eru fyrir hliðarmyndirnar firði (vinstra megin) og fjørði (hægra megin). Samkvæmt töflu 6 eru dæmi um eintölumyndirnar af fær. fjørður 3624 af 3987, miðað við samanlagða tíðni eintölunnar og fleirtölunnar (sjá töflu 5). Þegar orðið fjørður kemur fyrir utan samsetningar eru eintölumyndirnar 208. Leit að tíðni mynda í samsetningunum Fuglafjørður og Kollafjørður skilaði 3416 dæmum. Samanburður á tíðni einstakra eintölumynda sýnir stigveldið í (8). (8) þolfall (399) → nefnifall (806) → þágufall (2419) Eintölumyndir fær. fjørður eru því 90,89% af tíðni beggja talna samanlagðra, en þgf.et. firði er miklu algengari en allar aðrar myndir orðsins. 5.3 Samantekt Í 5.1.1 var skýrt frá staktíðni ólíkra mynda af fær. vøllur samkvæmt TTF. Í ljós kom að mynd þágufalls eintölu væri algengust í nútímamáli og að sú mynd væri oftast notuð í merkingunum ‘leik- og/eða íþróttavöllur’ og 24 Hér skal tekið fram að ég athugaði tíðni beygingarmynda hjá fleiri samsettum staðarheitum með -fjørður. Skopunarfjörður er eitt fárra dæma um samsett örnefni með -fjørður sem koma oftar fyrir í þolfalli en þágufalli. Helst virðist talað um ferðir yfir eða um Skopunarfjörð, enda ekki nafn á bæ heldur á sundi, eða nafnið á firðinum notað í tengslum við stað í nánd, sbr. t.d. Sunnan fyri Skopunarfjørð búgva 7000 fólk. 72 Jón Símon Markússon ‘flugvöllur’. Í 5.1.2 var sýnd áætluð staktíðni orðmyndanna í færeysku fyrri alda, miðað við að merkingin í (4c) hefði verið mun sjaldgæfari í eldra máli en nú er og að öll dæmi í merkingunni ‘flugvöllur‘ væru dregin frá. Umreikningurinn bendir til að þf.et. vøll hafi verið algengasta mynd orðsins í eldra máli. Í 5.2 var sagt frá staktíðni mismunandi beygingarmynda af orðinu fjørður ósamsettu samkvæmt TTF og reyndust þágufallsmyndir þess í fleirtölu tíðastar. Síðan var mæld staktíðni ólíkra beygingarmynda af samsettum örnefnum á -fjørður og henni bætt við tíðni fær. fjørður almennt. Í ljós kom að þgf.et. (-)firði væri langalgengust þegar einnig er tekið mið af samsetningum. 6. Ákvörðun grunnmynda og stefna útjöfnunar Hér verður gerð grein fyrir útjöfnun í beygingardæmum orðanna vøllur og fjørður í ljósi tíðnirannsóknarinnar í 5. kafla. Fyrst verður þó fjallað um áþekka beygingarþróun langflestra u-stofna í færeyskri málsögu (6.1.1). Síðan verður varpað ljósi á hugrænar forsendur fyrir alhæfingu stofnsins vøll- í beygingardæmi orðsins vøllur (6.1.2). Loks er sett fram skýring á beygingarþróun orðsins fjørður m.t.t. áhrifa frá staktíðni (6.2.1) og sambandsins milli forms og merkingar (6.2.2). 6.1 Alhæfing einnar stofnmyndar 6.1.1 Almenn beygingarþróun færeyskra u-stofna Eins og gildir um beygingarþróun orðsins vøllur hefur stofnmynd nefnifalls eintölu, þolfalls eintölu og/eða þágufalls fleirtölu yfirleitt gegnt hlutverki grunnmyndar hjá gömlum u-stofnum, sbr. fvnorr. nf./þf.et./þgf.ft. vǫll- (> fær. vøll-). Þannig var stofnsérhljóðavíxlum nær allra færeyskra u-stofna útrýmt í aldanna rás. Ljóst er að u-stofnar lágu vel við höggi með tilliti til útjöfnunar vegna þeirra fjölskrúðugu stofnsérhljóðavíxla sem einkenndu beygingu viðkomandi flokks öfugt við beygingu annarra karlkynsorða þar sem sérhljóðavíxl voru lítil eða engin. Hér er nauðsynlegt að taka fram að auk staktíðni einstakra beygingarmynda getur mynsturstíðni algengra beygingarvíxla líka ráðið miklu í stefnu útjöfnunar (sjá Bybee 2010:67, 2015:97; Fertig 2013:72; Garrett 2008:142). Enn fremur er litið svo á að fjöldi orða sem beygjast eftir tilteknu mynstri haldist í hendur við virkni þess (sjá 3.2). Þetta hefur til dæmis í för með sér að ef beyging stórs beygingarflokks einkennist ekki af stofnsérhljóðavíxlum geta minni flokkar byrjað að jafna út stofnsér- Um áhrif tíðni á stefnu útjöfnunar 73 hljóðavíxl sín fyrir áhrif frá þeim fyrrnefnda.25 Í þessu sambandi er rétt að nefna að karlkyns a-, an- og i-stofnar voru alltaf fleiri en u-stofnar, auk þess sem beyging þeirra fyrrnefndu hefur ekki einkennst af stofnsérhljóðavíxlum að neinu marki.26 Þetta sést á fornvesturnorrænu beygingardæmunum í (9). (9) a-st. et. nf. hestr þf. hest þgf. hesti ef. hests an-st. granni granna granna granna i-st. sauðr sauð sauð sauðar u-st. vǫllr vǫll velli vallar u-st. fjǫrðr fjǫrð firði fjarðar ft. nf. þf. þgf. ef. grannar granna grǫnnum granna sauðir sauði sauðum sauða vellir vǫllu → velli(r) vǫllum valla firðir fjǫrðu → firði(r) fjǫrðum fjarða hestar hesta hestum hesta Þótt sumt hafi verið ólíkt með beygingu i- og u-stofna er ljóst af (9) að umræddir flokkar áttu ýmislegt sameiginlegt. Til dæmis má nefna endinguna fvnorr. nf.ft. -ir, sbr. kk. i-st. nf.ft. sauðir, u-st. nf.ft. vellir, firðir. Auk þess deildu u-stofnar endingunni fvnorr. ef.et. -ar með tiltölulega litlum fjölda karlkynsorða, einkum i-stofna, sbr. kk. i-st. ef.et. sauðar, ust. ef.et. vallar, fjarðar. Aftur á móti höfðu u-stofnar endinguna fvnorr. þf.ft. -u á elsta skeiði og líklega eitthvað fram eftir öldum (sjá hér á eftir), en samsvarandi mynd karlkyns i-stofna hafði -i. Hvað áhrif mynsturstíðni varðar skal bent á að þegar mynd þolfalls fleirtölu féll saman við nefnifallsmyndina hjá karlkyns a-, an- og i-stofnum í færeysku tóku samsvarandi myndir u-stofna þátt í ferlinu og fengu endinguna nf./þf.ft -ir. Þannig féllu fvnorr. a-st. nf.ft. hestar og þf.ft. hesta saman í fær. nf./þf.ft. hestar; fvnorr. i-st. nf.ft. sauðir og þf.ft. sauði féllu saman í fær. nf./þf.ft. seyðir;27 myndir á borð við fær. nf./þf.ft. vøllir, firðir benda því til samfalls þf.ft. -i við nf.ft. -ir. 25 Sjá þó Jón Símon Markússon (2022a, 2022b) um takmarkaða virkni lítils undirflokks kvenkynsnafnorða í íslensku sem einkennist af stofnsérhljóðavíxlum milli eintölu og fleirtölu af taginu nf./þf./þgf.(/ef.)et. bók(ar) ~ nf./þf.ft. bækur, rót ~ rætur. 26 Hér eru u-hljóðvarpsvíxl í þágufalli fleirtölu viðkomandi orða undanskilin, sbr. t.d. þgf.ft. a-st. ǫrmum, i-st. stǫðum, an-st. grǫnnum (sjá Jón Símon Markússon 2012, 2017). 27 Fvnorr. au > fær. ey við óskilyrta hljóðbreytingu, mögulega fyrir 1400 (sjá Höskuld Þráinsson o.fl. 2012:394). 74 Jón Símon Markússon Aðeins karlkyns i-stofnar höfðu þf.ft. -i í elstu fornvesturnorrænu, en formleg líkindi á grundvelli sameiginlegu endingarinnar nf.ft. -ir hafa lokkað þf.ft. -i yfir til u-stofna.28 Í ljósi þess að stærri karlkynsflokkarnir sýndu ekki flókin stofnsérhljóðavíxl á við u-stofna reynast þau almennt afar sjaldgæf hjá karlkynsflokkum nafnorða. Því má telja enn meiri þrýsting en ella hafa verið á ustofnum til að gefa stofnsérhljóðavíxlin upp á bátinn. Enn fremur var ekki einsleitt víxlamynstur hjá u-stofnum, heldur aðallega um tvo undirflokka að ræða (sjá 2.1). Þar af leiðandi má líta á útjöfnun hjá u-stofnum sem skref í áttina að meiri regluleika með hliðsjón af mynsturstíðni stofnsérhljóðavíxla (sjá t.d. Fertig 2013:72). Í því samhengi helst regluleiki í hendur við samband tíðni, rótfestu og minnisstyrks (sjá 3.2). Áhrifum stærri karlkyns flokkanna á beygingu u-stofna mætti lýsa með hlutfallsjöfnunni í (10).29 (10) nf.et. seyiður : þgf.ft. seyiðum : nf./þf.ft. seyiðir/seyiði(r) nf.et. vøjllur : þgf.ft. vøjllum : nf./þf.ft. X; X = vøjllir/vøjlli(r) Úr jöfnunni má lesa að alhæfing stofnmyndarinnar vøll-, auk lánsins á endingunni þf.ft. -i(r), stafi af áhrifum frá karlkyns i-stofnum, sem nafnorðið seyður er fulltrúi fyrir. 6.1.2 Fær. vøllur: útjöfnun á kostnað stofnsins vellHér verður gerð grein fyrir alhæfingu stofnsins vøll- í beygingardæmi orðsins vøllur og þar með einnig hvarfi stofnsins vell-. Þeir sem aðhyllast málnotkunarnálgunina gera ráð fyrir ríkulegu vinnsluminni fyrir mannlegt mál (e. rich memory for language) (t.d. Bybee 2010:14, 18–19), þ.e. að beygingarmyndir, berir stofnar og beygingarendingar geymist í minni, bæði ósamsett og samsett. Rök fyrir því að beygingarmyndir séu geymdar í heilu lagi eru til dæmis byggð á reynslu málnotenda við málnotkun: 28 Hér skal bent á að Hammershaimb (1891) nefnir ekki gömlu endinguna -u í ummælum sínum um samfall þolfallsmynda fleirtölu við nefnifallsmyndina á meðan ferlið var enn í gangi á seinni hluta 19. aldar — einkum á Suðurey. Því má telja líklegt að þolfallsmyndir fleirtölu á -i af gömlum u-stofnum hafi þá verið að falla saman við nefnifallsmyndir fleirtölu á -ir: annars hefði Hammershaimb þótt tilefni til að nefna þf.ft. -u. Til viðmiðunar má benda á að endingin þf.ft. -u er horfin úr íslensku um 1500 (Björn K. Þórólfsson 1925 [1987]). 29 Hér beinir ‘ey ’ athygli að því að sama stofnsérhljóð er í öllum beygingarmyndunum i af fær. seyður sem tilfærðar eru í (10). Síðan gegnir ‘øj’ sama hlutverki í viðkomandi myndum fær. vøllur í neðri línunni. Um áhrif tíðni á stefnu útjöfnunar 75 stofnmyndir eru oft bundnar við ákveðnar beygingarmyndir og þar með tengdar endingum þeirra. Enn fremur er litið svo á að t.d. beygingarendingin fær. þgf.ft. -um — sem gegnir sama hlutverki hjá næstum öllum nafnorðum í málinu — sé tákn þágufalls fleirtölu óháð tengslum við tiltekna stofnmynd, en þá þekkingu má tákna sem skemað [Xum]þgf.ft. (sjá t.d. Janda 2007),30 þar sem ‘X’ táknar ‘hvaða stofnmynd sem er’. Hið sama gildir þá um stofnmyndir sem geymdar eru í minninu en er þó hægt að kalla fram án endingar, sbr. t.d. skemað [firð-]þgf.et., nf./nf./þgf.ft.. Með öðrum orðum er gert ráð fyrir að auk heilla orðmynda séu stofnmyndir og beygingarendingar einnig rótfastar í minninu, óháð tilvist hver annarrar. Öðruvísi væri ekki hægt að púsla saman nýrri mynd úr stofnmynd og beygingarendingu sem áður mynduðu ekki orðhlutafræðilega heild. Dæmi eins og nýjungin fær. þgf.ft. firðum, sem samsett er af stofnmyndinni firð- og beygingarendingunni þgf.ft. -um, benda einmitt til þess, að ólíkum orðhlutum geti verið púslað saman eftir að hafa verið sóttir hver í sínu lagi. Í (11) eru sýnd beygingardæmin fyrir fvnorr. vǫllr og afkomanda þess, fær. vøllur, sem eru endurtekin úr (2) og (3). Stofnsérhljóðin í yngri hliðarmyndum á báðum málstigum eru feitletruð (úr örinni í (11) má lesa ‘öðlaðist tvímyndina’). (11) et. nf. þf. þgf. ef. ft. nf. þf. þgf. ef vǫllr vǫll velli vallar vøllur vøll vølli vellir vǫllu → velli(r) vǫllum valla vøllir/vallir vøllir/vallir vøllum/vallum Frá sjónarhóli málnotkunar kemur ekki á óvart að stofnmynd fvnorr. þf.et. vǫll (> fær. vøll) hafi legið til grundvallar, enda hefur sú mynd verið algengust í færeysku fyrri alda miðað við umreikninginn í 5.1.2. En er tíðnimismunurinn á þolfallsmynd eintölu og þágufallsmyndinni — 34 dæmi á móti tíu — nægilega mikill til að ákvarða grunnmyndina? Í þessu sambandi skal bent á að áætluð staktíðni stofnmyndarinnar vell- í eintölu hefur verið u.þ.b. 23% miðað við færeysku fyrri alda, sbr. 44 dæmi um 30 Við framsetningu skema fer ég eftir Booij (2010). 76 Jón Símon Markússon nf./þf.et. vǫll-/vøll- á móti tíu um þgf.et. vell- (sjá töflu 4). Bæði stak- og dreifitíðni stofnmyndarinnar vøll- hefur því verið hærri en vell- í eintöluhluta beygingardæmisins (sjá hér að neðan). Hvað tíðnimuninn á eintölu- og fleirtölumyndum snertir hefur fleirtalan að öllum líkindum alltaf verið hlutfallslega jafnsjaldgæf í merkingunum í (4a–b) og nú. Því má efast um að samanlögð staktíðni mynda með stofnmyndinni vell- hafi dugað til að vinna gegn því að stofnmyndin vøllkæmist inn í þágufallsbás eintölunnar, einkum fyrir áhrif frá stak- og dreifitíðni nf.et. vøllur og þf.et. vøll. Enn fremur eru síðarnefndu myndirnar einmitt merkingarfræðilega náskyldar þgf.et. velli þar sem allar merkja þær eintölu andspænis myndum fleirtölunnar (sjá 3.3 og ummæli um andstöðuna et. fjørð- : ft. fjarð- í 5.2). Í þessu sambandi má svo færa rök fyrir því að hærri dreifitíðni stofnsins vøll- í eintölu hafi fyrir tilurð yngri þgf.et. vølli stuðlað að sambandi formsins vøll- við merkinguna eintölu, enda hafi stofnmyndin vell- e.t.v. hvorki búið yfir stak- né dreifitíðni í eintöluhluta beygingardæmisins til að sporna gegn þeirri þróun (sjá 5.1.2). En hvers vegna stuðlaði alhæfing stofmyndarinnar vøll- í eintölu ekki að andstæðu forms og merkingar af taginu et. vøll- : ft. vell-? Þetta má fyrst og fremst skýra með vísun til staktíðni allra mynda orðsins með stofnmyndinni vell-. Með öðrum orðum hefur minnisstyrkur umræddra mynda verið tiltölulega lítill og þess vegna ekki dugað til að sporna gegn samkeppni við langtum rótfastari stofnmyndina vøll-. Í þessu sambandi skal aftur bent á að staktíðni þgf.ft. vøllum er almennt séð tiltölulega lág, en þó hefur viðkomandi mynd hæstu staktíðni fleirtölumyndanna í nútímafæreysku (sjá töflu 1). Þar sem engin ástæða er til að halda að sú hafi ekki einnig verið raunin í færeysku fyrri alda má e.t.v. gefa sér að þgf.ft. vøllum hafi veitt útbreiðslu stofnmyndarinnar vell- um fleirtöluhluta beygingardæmisins mótspyrnu. Þar að auki er ekki óhugsandi að þolfallsmynd fleirtölunnar hafi enn haft stofnmyndina vøll- þegar ferlið hófst (sjá (2)), en þá hefur dreifitíðni viðkomandi stofnmyndar meðal fleirtölumynda einnig getað spornað gegn útbreiðslu vell- um þann hluta beygingardæmisins. 6.2 Fær. fjørður 6.2.1 Tilurð þgf.ft. firðum Hér verður grunnmynd fær. fjørður ákvörðuð á grundvelli staktíðni og gerð grein fyrir stefnu útjöfnunar á þeim forsendum. Í ljósi þess að tiltölulega mikill fjöldi örnefna hefur -fjørður að seinni lið er skiljanlegt að þágufallsmyndir orðsins séu algengari en aðrar myndir, enda staðir þar Um áhrif tíðni á stefnu útjöfnunar 77 sem fólk á heima oft nefndir í tengslum við dvöl (sjá töflu 6; einnig nmgr. 24). Samkvæmt málnotkunarnálguninni ákvarðar tíðni rótfestu orðmyndar í minni og því einnig minnisstyrk hennar (sjá 3.2), en því mætti eiga von á að þgf. -firði byggi yfir mestum minnisstyrk allra beygingarmynda orðsins fjørður og lægi þ.a.l. útjöfnun innan beygingardæmisins að baki. En þó blasir þetta e.t.v. ekki við af samanburði á beygingardæmunum í (12) (áður í (2) og (3)). Stofnsérhljóðin í yngri hliðarmyndum á báðum málstigum eru feitletruð. (12) et. nf. þf. þgf. ef. ft. nf. þf. þgf. ef. fjǫrðr fjǫrð firði fjarðar fjørður fjørð firði/fjørði firðir fjǫrðu → firði fjǫrðum fjarða firðir/fjørðir/?fjarðir firðir/fjørðir/fjarðir fjørðum/firðum/fjarðum Ljóst er af (12) að stofnmyndirnar fjørð-, firð- og fjarð- hafa breitt úr sér (sjá 5.2). Ef til vill þykir líklegt að þgf.et. fjørði hafi fengið sína stofnmynd að láni frá öðrum eintölumyndum, en þgf.ft. firðum úr hinum myndum fleirtölunnar. Á hinn bóginn mætti líka hugsa sér að eldri myndirnar þgf.et. firði og þgf.ft. fjørðum hafi hvor um sig fengið hliðarmynd að láni frá hinni. Nú verður reynt að ákvarða hvor yngri hliðarmyndanna þgf.et. fjørði eða þgf.ft. firðum hefur komið upp fyrr. Til að átta sig á tímaröðinni má byrja á að bera tíðni myndanna saman innan setningafræðilegs samhengis þar sem þágufallsmyndar er að vænta. Í færeysku koma þágufallsmyndir örnefna með -fjørður gjarnan fyrir sem andlög forsetninganna í og úr, sbr. t.d. í Kollafirði (þgf.) og úr Fuglafirði (þgf.). Eins og áður segir bendir allt til þess að orðið fjørður hafi vísað til sama landfræðilega fyrirbæris frá því að land var numið í Færeyjum (sjá 4.2, einkum nmgr. 14) og því er ekkert tilefni til að halda að þgf.et. firði hafi ekki alltaf verið algengari en aðrar beygingarmyndir. Þess vegna hefur sú mynd líklega alltaf verið auðsóttari en aðrar myndir orðsins og því alltaf gegnt stöðu grunnmyndar (sjá t.d. Beckner o.fl. 2009; einnig 3.2 og 4.2). Aftur á móti er hin langtum sjaldgæfari mynd þgf.ft. fjørðum mun líklegri til að gleymast í því setningafræðilega samhengi sem skemað [í/úr (-) fjørðurþgf.] alhæfir um vegna hlutfallslega minni minnisstyrks (Bybee Jón Símon Markússon 78 2015:102).31 Þar af leiðandi stafar veikari staða stofnmyndarinnar fjørð- af lélegri rótfestu þgf.ft. fjørðum í umræddu setningafræðilegu umhverfi, auk stífrar samkeppni við stofnmyndina firð- þegar sækja á beygingarmynd í eyðuna [í/úr_______þgf.]. Af þessari ástæðu má ætla að vísigildi þessa setningafræðilega samhengis fyrir stofnmyndinni firð- hafi verið mjög hátt í gegnum aldirnar og jafnvel dugað snemma til að ýta stofnmyndinni fjørð- úr því samhengi. Hér er því litið svo á að þgf.ft. firðum sé eldri mynd en hliðarmyndin þgf.et. fjørði. Þó ber einnig að nefna að myndin þgf.ft. firðum kemur ekki fyrir hjá Hammershaimb (1891), en í þessu sambandi virðist hann miða við fordæmi forníslenskra beygingardæma víða í verki sínu (sjá Höskuld Þráinsson o.fl. 2012:385). Þó er auðvitað hugsanlegt að Hammershaimb hafi aldrei heyrt viðkomandi mynd. Í ljósi þess sem segir hér að framan er gert ráð fyrir að beygingardæmið hafi litið út eins og í (13) rétt fyrir tilurð þgf.et. fjørði. (13) et. nf. þf. þgf. (ef. ft. nf. þf. þgf. (ef. fjørður fjørð firði fjarðar) firðir firði(r) fjørðum/firðum(/fjarðum) fjarða) Tilurð þgf.et. fjørði er til umræðu í næsta undirkafla. 6.2.2 Tilurð þgf.et. fjørði Hér verða færð rök fyrir því að tilurð hliðarmyndarinnar þgf.ft. firðum hafi orðið hvatinn að sambandi milli stofnmyndarinnar firð- og merkingarinnar fleirtölu, sem stuðlaði síðan að tengslum stofnmyndarinnar fjørð- við eintölu. Í þessu sambandi skal minnt á að rótfesta er almennt hugrænt ferli og því er minnisstyrkur beygingarmyndar óháður merkingu. Þar af leiðandi getur stofnmynd langtíðustu beygingarmyndar orðs farið úr eintöluhluta beygingardæmisins yfir í fleirtöluhlutann (eða öfugt) 31 Skemað [í/úr (-)fjørðurþgf.] alhæfir um þau tilvik þegar málnotandinn hefur notað orðið fjørður — ósamsett sem og í samsettum örnefnum — í þágufalli sem andlag forsetninganna í og úr. Um áhrif tíðni á stefnu útjöfnunar 79 án tillits til þess hvort ferlið raski formlegri afstöðu beygingarmynda hverra til annarra (sjá Bybee 1985). Með þessa tilgátu að leiðarljósi má velta fyrir sér hvort þágufallsmyndir eintölu og fleirtölu hafi einfaldlega lánað hvor annarri sína stofnmynd (sjá 6.2.1). Þó er ýmislegt sem mælir gegn þessu þegar betur er að gáð. Í fyrsta lagi má efast um að staktíðni þgf.ft. fjørðum hafi tryggt myndinni nægan minnisstyrk til að stofnmynd hennar yrði sótt í stað (þgf. et.) firð-, sérstaklega í setningafræðilega umhverfinu sem skemað [í/úr (-)fjørðurþgf.] alhæfir um (sjá 6.2.1). Í öðru lagi er stofnmynd þgf.ft. fjørðum ekki nægilega rótföst miðað við þgf.et. firði til að hún kæmist yfir merkingarfræðileg mörk eintölu og fleirtölu án milligöngu hárrar staktíðni. Þannig er talið ósennilegt að stofnmyndin fjørð- í þgf.et. fjørði sé fengin að láni beint frá þgf.ft. fjørðum. Því er rétt að kanna líkurnar á að stofnmyndin í yngri myndinni þgf.et. fjørði eigi upptök í nf.et. fjørður og/eða þf.et. fjørð. Í 6.1.2 voru einmitt færð rök fyrir því að útbreiðsla stofnsins vøll- til þgf.et. vølli stefndi út frá nf.et. vøllur og þf.et. vøll. Þó er lykilmunur á aðstæðunum sem stuðluðu að því ferli og afstöðu samsvarandi mynda fær. fjørður hverrar til annarrar í eldra máli. Í fyrsta lagi, samkvæmt umreikningnum í 5.1.2, er staktíðni eldri þgf.et. velli (tíu dæmi) ekki nema 22,72% af samanlagðri tíðni myndanna nf.et. vøllur (tíu dæmi) og þf.et. vøll (34 dæmi), sbr. töflu 4. Af þessum sökum hefur stofnmyndin vell- ekki veitt dreifitíðni stofnmyndarinnar vøll- innan eintöluhluta beygingardæmisins verulega mótstöðu m.t.t. sambands forms og merkingar (sjá 6.1.2). Í öðru lagi hefur minnisstyrkur nf.ft. vellir og þf.ft. velli(r) ekki dugað til að sporna gegn útbreiðslu stofnmyndarinnar vøll- um allt beygingardæmið frá eintöluhluta þess, auk þess sem há mynsturstíðni víxlaleysis í stærri karlkynsflokkunum stuðlar að útjöfnun (sjá 6.1.1). Aftur á móti er þgf.et. firði langtum tíðari en nf.et. fjørður og þf.et. fjørð samanlagt þegar samsett örnefni eru talin með (tafla 6). Þess vegna leikur vafi á því að minnisstyrkur stofnmyndarinnar fjørð- hafi getað veitt þgf.et. firði verulega samkeppni þegar sækja átti þá síðarnefndu úr minni, eins og segir hér að framan. Eigi yngri myndin þgf.et. fjørði upptök sín í eintöluhluta beygingardæmisins þá stafa þau af merkingarfræðilegum tengslum við hinar eintölumyndirnar. Ein vísbending um að sameiginleg merking frekar en staktíðni myndanna nf.et. fjørður og þf.et. fjørð stuðli að tilurð (yngri) þgf.et. fjørði er sú að síðarnefnda myndin er langtum sjaldgæfari en þgf.et. firði. Með öðrum orðum: tiltölulega lág staktíðni yngri þgf.et. fjørði endurspeglar staktíðni Jón Símon Markússon 80 þeirra beygingarmynda þangað sem nýjungina má rekja, eða hví ætti staktíðni stofnmyndarinnar að vera hærri í þgf.et. fjørði en í myndunum þaðan sem hún er fengin að láni, sem annars búa yfir hlutfallslega lágri staktíðni (sbr. töflu 6)? Á hinn bóginn má líta svo á að tiltölulega há staktíðni yngri þgf.ft. firðum miðað við eldri þgf.ft. fjørðum bendi til þess að fyrrnefnda myndin eigi upptök í algengari mynd innan beygingardæmisins, sbr. tíðni myndarinnar þgf.et. -firði samkvæmt töflu 6, enda er litið svo á að algengari einingar séu þeim mun rótfastari í minni og hafi þar af leiðandi meiri minnisstyrk (sjá 3.2). Því er hér talið líklegra að eftir tilurð og rótfestu yngri þgf.ft. firðum hafi tilvist eldri og langtum algengari þgf.et. firði skekkt nýja sambandið milli formsins firð- og merkingarinnar fleirtölu. Sem viðbrögð við þeirri skekkju hafi svo verið leitast við að tengja merkinguna eintölu við stofnmyndina fjørð- — sem þegar hafði hærri dreifitíðni í þeim hluta beygingardæmisins. Þrátt fyrir þessa viðleitni hefur staktíðni nf.et. fjørður og þf.et. fjørð ekki dugað til að eldri þgf.et. firði léti undan þrýstingi frá yngri básfélaga sínum að neinu marki. Ástæðan fyrir því að yngri þgf.et. fjørði og eldri þgf.ft. fjørðum lifa enn er talin sú að staktíðni stofnmyndarinnar fjørð- sé talsverð þótt hún hafi átt lítið í staktíðni þgf.et. firði. Annað gildir um sögulega staktíðni stofnmynda með vell- og því var henni útrýmt (sjá 6.1.2). 7. Niðurstöður Í greininni hefur ólík stefna útjöfnunar í fær. vøllur og fjørður verið til skoðunar og voru orðin valin af tveimur ástæðum: Bæði eru þau afkomendur gamalla u-stofna og bæði vísa til staðfræðilegs fyrirbæris. Eins og segir í 1. kafla var markmið rannsóknarinnar fyrst og fremst að bera kennsl á grunnmyndir orðanna. Í þessum tilgangi var reynt að svara rannsóknarspurningunni í (1), hér endurtekin í (14). (14) Er tíðni ráðandi afl í stefnu útjöfnunar? Í 3. kafla var rædd beygingarþróun í ýmsum málum sem bendir til þess að tíðni sé ráðandi afl í ákvörðun grunnmynda og stuðst við dæmi sem sýna að vísun til (sér)mörkunar bætir óþörfu þrepi við skýringu sem dugir ein sér, enda er tíðni mælanleg, en merking og notkunarsvið orða ekki (3.1). Enn fremur var bent á að merking orðs hefði lítið um tíðni beygingar- Um áhrif tíðni á stefnu útjöfnunar 81 mynda þess að segja, heldur var talið líklegra að upplifun málnotenda af aðstæðum í raunheiminum skæri úr um hversu oft ólík fyrirbæri bærust í tal, til samræmis við hugmyndir Haspelmaths (2006). Í ljósi þessa var ákveðið að hafna hugmyndum um áhrif (sér)mörkunar á stefnu útjöfnunar og leita skýringa sem miðuðu við ólíka tíðni einstakra beygingarmynda í mismunandi merkingum orðanna samkvæmt Føroyskri orðabók (4. kafli). Í 5. kafla var svo sagt frá tíðnirannsókn sem studdist við innihald færeyska textasafnsins Teldutøka tekstasavn Føroyamálsdeildarinnar. Rannsóknin leiddi eftirfarandi í ljós: 1. Myndin þgf.et. vølli er algengasta myndin af fær. vøllur, bæði ósamsettu og í samsettum orðum, en í báðum tilvikum vísar orðið oftast í leik- og/eða íþróttavöll, en næstoftast til flugvallarins í Vogum (5.1.1). 2. Á grundvelli umreiknings á sögulegri tíðni orðsins miðað við notkun þess í forníslenskum málheimildum er þf.et. vøll aftur á móti talin hafa verið algengust í færeysku fyrri alda. Við umreikninginn var nauðsynlegt að draga staktíðni orðsins vøllur í merkingunni ‘flugvöllur’ alfarið frá tíðni orðsins í nútímafæreysku. Auk þess var tekið tillit til sérfæreyskra aðstæðna eins og landshátta, veðurfars og áhrifa þessara þátta á utanhúss íþróttir fyrr á tímum, áður en sérstakur íþróttavöllur var lagður í Færeyjum árið 1911 (5.1.2). 3. Í nútímafæreysku er þgf.ft. firðum tíðasta mynd orðsins fjørður utan samsetningar. Í samsetningum er myndin þgf.et. -firði aftur á móti langalgengust, enda tilhneiging að nefna firði og bæi fullu nafni frekar en stytta það í fjørður. Gera má ráð fyrir að þannig hafi það alltaf verið, enda eru örnefni með -fjørður gamalgróin (5.2). Á grundvelli þessara niðurstaðna var stefna útjöfnunar rakin í ljósi málnotkunar í 6. kafla. Grunnmyndir beygingardæmanna voru því ákvarðaðar með vísun til áhrifa stak- og dreifitíðni á rótfestu og minnisstyrk. Einnig var gerð grein fyrir þróun miðað við tilhneigingu til að samræma tiltekna stofnmynd beygingardæmis við ákveðna merkingu. Eftirfarandi niðurstöður komu fram um stefnu útjöfnunar í beygingardæmi fær. vøllur (6.1.1 og 6.1.2). 4. Beygingarþróun nær allra gamalla u-stofna, þ.á m. fær. vøllur, bendir til áhrifa karlkyns i-stofna, auk skorts á stofnsérhljóðavíxlum hjá stærri karlkynsflokkum almennt. 82 Jón Símon Markússon 5. Vegna samanlagðrar stak- og dreifitíðni myndanna nf.et. vøllur og þf.et. vøll miðað við umreikninginn í 5.1.2 varð stofnmyndin vøllrótfastari en stofnmyndin vell- í eintöluhluta beygingardæmisins. Færð voru rök fyrir því að rótfesta fyrrnefndu myndarinnar leiddi til tengsla milli hennar og merkingarinnar eintölu. 6. Vegna tiltölulega lágrar staktíðni fleirtölumynda með stofnmyndinni vell- breiddist stofnmyndin vøll- út um beygingardæmið allt og sigraði fyrrnefndu myndina á endanum. Eftirfarandi ályktanir voru dregnar um tilurð myndarinnar þgf.ft. firðum (6.2.1): 7. Vegna hárrar staktíðni þgf.et. firði sem síðari hluta samsettra örnefna hefur viðkomandi mynd rótfest verulega í samhenginu [í/úr (-)fjørðurþgf.]. Þar sem fleirtölumyndir orðsins koma sjaldan fyrir í málinu náði eldri þgf.ft. fjørðum aftur á móti aldrei verulegri rótfestu í því samhengi. 8. Af þessum sökum færði stofnmyndin firð- úr þágufallsbás eintölunnar yfir í bás samsvarandi fleirtölumyndar, enda er staktíðni myndar óháð merkingu hennar. 9. Yngri þgf.ft. firðum hefur að öllum líkindum tekið snemma fram úr staktíðni eldri þgf.ft. fjørðum, en þannig mynduðust tengsl milli stofnmyndarinnar firð- og merkingarinnar fleirtölu: nf.ft. firðir, þf.ft. firði(r), (yngri, algengari) þgf.ft. firðum/(eldri, sjaldgæfari) fjørðum. Um tilurð myndarinnar þgf.et. fjørði var komist að eftirfarandi niðurstöðum: 10. Tilurð yngri þgf.et. fjørði er svar við rótfestu þgf.ft. firðum og sýnir tilhneigingu til að greina eintöluna frá fleirtölunni með formlegum hætti, enda hafði stofnmyndin fjørð- hærri dreifitíðni um eintöluhluta beygingardæmisins en stofnmyndin firð-, sbr. nf.et. fjørður og þf.et. fjørð gegn þgf.et. firði. Svipuð þróun á beygingardæmi orðsins fjørður finnst á Suðurey, þar sem eintölumyndir hafa stofninn fjørð- en fleirtölumyndir stofnmyndina fjarð-. 11. Þó hefur hvorki staktíðni né dreifitðni stofnmyndarinnar fjørðdugað til að sigra alfarið í eintöluhluta beygingardæmisins vegna hárrar staktíðni þgf.et. firði. Af þessum sökum er staktíðni yngri þgf.et. fjørði hlutfallslega mjög lág. Um áhrif tíðni á stefnu útjöfnunar 83 Í ljósi þess sem að framan segir er svarið við spurningunni í (1)/(14) talið vera: Tíðni er hið ráðandi afl í stefnu útjöfnunar. Þó geta raskanir á eldri afstöðu beygingarmynda hverrar til annarrar, sem verða vegna áhrifa frá tíðni, stuðlað að tilhneigingu málnotenda til að koma á reglu innan beygingardæmisins með því að samræma form tiltekinna beygingarmynda við tiltekna merkingu. Áhugavert yrði því að athuga hvort málnotkunarnálgunin ætti eftir að leiða svipaðar niðurstöður í ljós hvað varðar útjöfnun í fleiri og jafnvel stærri beygingarflokkum í færeysku eða öðrum málum. heimildir Arge, Jógvan. 1994. Havnarmenn í Gundadali: Havnar Bóltafelag 1904–1954. Havnar Bóltafelag, Þórshöfn. Barnes, Michael. 2001. Faroese Language Studies. Studia Nordica 5. Novus Forlag, Osló. Beckner, Clay, Richard Blythe, Joan Bybee, Morten H. Christiansen, William Croft, Nick. E. Ellis, Joan Holland, Jinyn Ke, Diane Larsen-Freeman og Tom Schoenemann. 2009. Language is a complex adaptive system: Position paper. Language Learning 59: 1–26. Björn K. Þórólfsson. 1925. Um íslenskar orðmyndir á 14. og 15. öld og breytingar þeirra úr fornmálinu. Fjelagsprentsmiðjan, Reykjavík. [Ljósprentun 1987: Rit um íslenska málfræði 2. Málvísindastofnun Háskóla Íslands.] Booij, Geert. 2010. Construction Morphology. Oxford University Press, New York. Bybee, Joan. 1985. Morphology: A Study of the Relation between Meaning and Form. John Benjamins, Amsterdam. Bybee, Joan. 2007. Frequency of Use and the Organization of Language. Oxford University Press, New York. Bybee, Joan. 2010. Language, Usage and Cognition. Cambridge University Press, New York. Bybee, Joan. 2015. Language Change. Cambridge University Press, Cambridge. Carstairs-McCarthy. 2015. Paradigmatic Structure: Inflectional Paradigms and Morphological Classes. Andrew Spencer og Arnold M. Zwicky (ritstj.): The Handbook of Morphology, bls. 322–334. John Wiley and Sons, Hoboken, NJ. Cordes, Anne K. 2017. The roles of analogy, categorization, and generalisation in entrenchment. Hans-Jörg Schmid (ritstj.): Entrenchment and the Psychology of Language: How we reorganise and adapt linguistic knowledge, bls. 269–288. American Psychological Association, Washington, DC. Dobson, A. 1973. Estimating time separation for languages. Isidore Dyen (ritstj.): Lexicostatistics in genetic linguistics, bls. 56–63. Mouton, The Hague. Eiríkur Rögnvaldsson. 2013. Hljóðkerfi og orðhlutakerfi íslensku. [Rafræn útgáfa.] Reykjavík. [Bein slóð: <https://notendur.hi.is/eirikur/hoi.pdf>. Sótt 14. janúar 2022.] Fertig, David. 2013. Analogy and Morphological Change. Edinburgh University Press, Edinburgh. Frost, Ram, Blair C. Armstrong, Noam Siegelman og Morten H. Christiansen. 2015. Domain generality versus modality specificity: The paradox of statistical learning. Trends in Cognitive Sciences 19:117–125. 84 Jón Símon Markússon Føroyakort. Umhvørvisstovan. <www.foroyakort.fo>. [Sótt 9.7.2021.] Føroysk orðabók. 1998. Føroya Fróðskaparfelag og Fróðskaparsetur Føroya, Þórshöfn. Gardeła, Leszek. 2012. What the Vikings did for fun? Sports and pastimes in medieval northern Europe. World Archaeology 44:234–247. Garrett, Andrew. 2008. Paradigmatic uniformity and markedness. Jeff Good (ritstj.): Linguistic Universals and Langauge Change, bls. 125–143. Oxford University Press, Oxford. Greenberg, Joseph. 1966. Language universals: with special reference to feature hierarchies. Mouton, The Hague. Gries, Stefan Th., og Nick C. Ellis. 2015. Statistical Measures for Usage-Based Linguistics. Language Learning 65:228–255. Guðvarður Már Gunnlaugsson. 2000. Færeyskar málheimildir. Magnús Snædal og Turið Sigurðardóttir (ritstj.): Frændafundur 3, bls. 91–105. Reykjavík, Háskólaútgáfan. Hammershaimb, Venceslaus Ulricus. 1891. Færøsk Anthologi I–II. Kaupmannahöfn. Haraldur Bernharðsson. 2004. Um Moldhaugnaháls út í Fjósa og Fjörður. Íslenskt mál 26:11–48. Haraldur Bernharðsson. 2005. Ég er, ég vill og ég fær: Þáttur úr beygingarsögu eintölu framsöguháttar nútíðar. Íslenskt mál 27:63–101. Haspelmath, Martin. 2006. Against Markedness (and what to replace it with). Language 42:25–70. Höskuldur Þráinsson, Hjalmar P. Petersen, Jógvan í Lon Jacobsen og Zakaris Svabo Hansen. 2012. Faroese: An Overview and Reference Grammar. Føroya Fróðskaparfelag, Þórshöfn. Iversen, Ragnvald 1972. Norrøn grammatik. 7. útg. H. Aschehoug & Co, Osló. Íslenskt textasafn. Stofnun Árna Magnússonar í íslenskum fræðum. <http//corpus.arnastofnun.is>. Jakobson, Roman. 1939. Signe zéro. Roman Jakobson, Selected Writings, bls. 211–219. Mouton, The Hague. Janda, Laura. 2007. Inflectional Morphology. Dirk Geeraerts og Hubert Cuyckens (ritstj.): The Oxford Handbook of Cognitive Linguistics, bls. 632–649. Oxford University Press, New York. Jost, Ethan, og Morten H. Christiansen. 2017. Statistical Learning as a Domain-General Mechanism of Entrenchment. Hans-Jörg Schmid (ritstj.): Entrenchment and the Psychology of Language: How we reorganise and adapt linguistic knowledge, bls. 227–44. American Psychological Association, Washington, DC. Jóhanna Barðdal. 2008. Productivity. Evidence from Case and Argument Structure in Icelandic. John Benjamins, Amsterdam. Jón Símon Markússon. 2012. Eðli u-hljóðvarpsvíxla í íslenskri málsögu. MA-ritgerð í íslenskri málfræði, Háskóli Íslands, Reykjavík. <http://hdl.handle.net/1946/11410>. Jón Sí́ mon Markússon. 2017. Samband veiklunar og hljóðanvæðingar: Vitnisburður u-hljóðvarpvíxla í frum- og vesturnorrænni málsögu. Zakaris Svabo Hansen, Anfinnur Johansen, Hjalmar P. Petersen og Lena Reinert (ritstj.): Bók Jógvan: Heiðursrit til Jógvan í Lon Jacobsen á 60 ára degnum, bls. 263–276. Føroya fró ðskaparfelag, Þórshöfn. Jón Símon Markússon. 2021. Undir áhrifum (orða)gengis. Katrín Axelsdóttir, Veturliði Óskarsson og Þorsteinn G. Indriðason (ritstj.): Möggubrár heklaðar Margréti Jónsdóttur sjötugri, 21. mars 2021, 99–104. Rauðhetta, Reykjavík. Um áhrif tíðni á stefnu útjöfnunar 85 Jón Símon Markússon. 2022a. Accounting for different rates of gender reanalysis among Icelandic masculine forms in plural -ur. Nordic Journal of Linguistics (birt á netinu í október 2022). <https://doi.org/10.1017/S0332586522000166>. Jón Símon Markússon. 2022b. Tvær blækur labba inn á bar: On limited productivity as graded membership of an Icelandic microclass. NOWELE 75(2):194–222. <https://doi.org/10.1075/nowele.00068.mar>. Katrín Axelsdóttir. 2014. Sögur af orðum. Sex athuganir á beygingarþróun í íslensku. Háskólaútgáfan, Reykjavík. Katrín Axelsdóttir. 2015. Beyging og merking orðsins hjalt. Orð og tunga 17:95–114. <https://doi.org/10.33112/ordogtunga.17.6>. Kodner, Jordan. 2019. Estimating child linguistic experience from historical corpora. Glossa: a journal of general linguistics 4(1):1–14. Kristján Árnason. 2011. The Phonology of Icelandic and Faroese. Oxford University Press, New York. Kuryłowicz, Jerzy. 1945–1949. La nature des procès dits „analogiques“. Acta linguistica 5:15–37. Lahiri, Aditi. 2000. Introduction. Aditi Lahiri (ritstj.): Analogy, Levelling, Markedness, bls. 1–14. Mouton de Gruyter, New York. Langacker, Ronald. 1987. Foundations of Cognitive Grammar: Theoretical Prerequisites, 1. bindi. Stanford University Press, Stanford CA. Mańczak, Witold. 1958. Tendances générales des changements analogiques. Lingua 7:298–325, 387–420. Matras, Christian. 1933. Stednavne paa de færøske Norðuroyar. Aarbøger for Nordisk Oldkyndighed og historie. Kaupmannahöfn. ONP (Ordbog over det norrøne prosasprog). Ritstjórar Aldís Sigurðardóttir o.fl. Københavns Universitet. <https://onp.ku.dk/onp/onp.php>. [Sótt 18.1.2022.] Petersen, Hjalmar P. 2009. Gender Assignment in Modern Faroese. Verlag Dr. Kovač, Hamborg. Petersen, Hjalmar P. 2020. Føroysk mállæra 1. Kyn, orðmyndan og bending. Nám, Þórshöfn. Schmid, Hans-Jörg. 2017. A Framework for Understanding Linguistic Entrenchment and Its Psychological Foundations. Hans-Jörg Schmid (ritstj.): Entrenchment and the Psychology of Language: How we reorganise and adapt linguistic knowledge, bls. 227–44. American Psychological Association, Washington, DC. Sims‐Williams, Helen. 2016. Analogical levelling and optimisation: The treatment of pointless lexical allomorphy in Greek. Transactions of the Philological Society 114:315– 338. Sims-Williams, Helen. 2022. Token frequency as a determinant of morphological change. Journal of Linguistics 58:571–607. Taylor, John R. 2012. The Mental Corpus. How Language is Represented in the Mind. Oxford University Press, Oxford. Teldutøka tekstasavn Føroyamálsdeildarinnar. Føroyamálsdeild, Fróðskaparsetur Føroya. <http://www.teldni.fo/tekstasavn/index>. [Sótt 9.7.2021.] Tiersma, Peter. 1982. Local and general markedness. Language 58: 832–849. Tímarit.is. Samstarfsverkefni milli Landsbókasafns Íslands – Háskólabókasafns, færeyska landsbókasafnsins (Føroya landsbókasavn) og grænlenska landsbókasafnsins (Nunatta Atuagaateqarfia). <www.timarit.is>. [Sótt 17.5.2022.] Jón Símon Markússon 86 Tuggy, David. 2007. Schematicity. Dirk Geeraerts og Hubert Cuyckens (ritstj.): The Oxford Handbook of Cognitive Linguistics, bls. 82–116. Oxford University Press, New York. Weyhe, Eivind. 2012. Eivindaródn. Greinar 1979–2011. Fróðskapur, Þórshöfn. Lykilorð: hliðstæðuútvíkkun, færeyska, tíðni, útjöfnun, málnotkun Keywords: analogy, Faroese, frequency, levelling, usage-based grammar summary ‘On frequency as the determinant of levelling Tracing inflectional changes in the nouns vøllur and fjørður through Faroese language history’ The intricate vowel alternations exhibited by Old West-Nordic u-stems have mostly been levelled in Faroese, as evident from the development of Far. vøllur. However, Far. fjørður has both retained its old stem alternants and extended them within the paradigm, cf. e.g. innovative dat.sg. fjørði and dat.pl. firðum. Typically, basic forms for levelling are identified on either of two theoretical premises: one places prominence on the effects of socalled markedness, while the other posits the impact of frequency on memory as the determining factor. Faroese corpus data demonstrates that basic forms for levelling in the paradigms of Far. vøllur and fjørður, both of which refer to topographical entities and occur as complex place names, are established on the basis of frequency and not common semantics. It is concluded that the lack of any significant frequency disparity between distinct forms of vøllur facilitated levelling in favour of (nom./acc.sg., dat.pl.) vøll- to the whole paradigm. Conversely, the high token frequency of dat.sg. firði triggered spread of the stem alternant firð- to the dative plural through the context [í/á/úr + dat.]. Subsequently, association of firð- with the plural, cf. nom./acc.pl. firðir, motivated an attempt to align the meaning singular with the form fjørð-. Jón Símon Markússon Hugvísindasviði Háskóla Íslands jsm2@hi.is Bibliography74 Albright, Adam. (2002). Islands of reliability for regular morphology: Evidence from Italian. Language, 78, 684–709. Albright, Adam. (2008). Explaining universal tendencies and language particulars in analogical change. In Good, Jeff (ed.), Linguistic Universals and Language Change, pp. 144–181. Oxford: Oxford University Press. Albright, Adam. (2009). Modelling analogy as probabilistic grammar. In Blevins, James P., and Juliette Blevins (eds.), Analogy in Grammar: Form and Acquisition, 185–213. Oxford: Oxford University Press. Albright, Adam, and Bruce Hayes. (2003). Rules vs. analogy in English past tenses: A computational/experimental study. Cognition, 90, 119–61. Arge, Jógvan. (1994). Havnarmenn í Gundadali: Havnar Bóltafelag 1904–1954. Havnar Bóltafelag. Audring, Jenny. (2019). Mother or sister? The encoding of morphological knowledge. Word Structure, 12, 274–296. Axelsdóttir, Katrín. (2014). Sögur af orðum. Sex athuganir á beygingarþróun í íslensku. Reykjavík: Háskólaútgáfan. Axelsdóttir, Katrín. (2015). Beyging og merking orðsins hjalt. Orð og tunga, 17, 95–114. Árnason, Kristján. (2005). Íslensk tunga I: Hljóð. Reykjavík: Almenna bókafélagið. Árnason, Kristján. (2011). The Phonology of Icelandic and Faroese. Oxford: Oxford University Press. Barðdal, Jóhanna. (2006).75 Predicting the productivity of argument structure constructions. Berkeley Linguistics Society, 32, 467–478. Barðdal, Jóhanna. (2008). Productivity: Evidence from Case and Argument Structure in Icelandic. Amsterdam/Philadelphia: John Benjamins. Barðdal, Jóhanna. (2009). The development of case in Germanic. In Barðdal, Jóhanna, and Shobhana L. Chelliah (eds.), The Role of Semantic, Pragmatic, and Discourse Factors in the Development of Case. Amsterdam/Philadelphia: John Benjamins. Barðdal, Jóhanna. (2011). The rise of dative substitution in the history of Icelandic: A diachronic construction grammar approach. Lingua, 121(1), 60–79. Barlow, Michael, and Suzanne Kemmer. (2000). Usage-based Models of Language. Stanford, CA: CSLI. 74 The bibliography contains all works cited in Papers I–III, in addition to references included in the extended introduction, i.e. Chapters 1–7 of the current thesis. 75 Included in Paper II as ‘Barðdal, Jóhanna. 2012.’ at the request of a reviewer. 221 Barnes, Michael P. (2001). Faroese Language Studies. Studia Nordica, 5. Oslo: Novus Forlag. Barnes, Michael. P. (2004). Norn –– the one-time Scandinavian language of Orkney and Shetland. Íslenskt mál og almenn málfræði, 26, 49–82. Beckner, Clay, and Andrew Wedel. (2010). The roles of acquisition and usage in morphological change. In Kwon, Iksoo, Hannah Pritchett, and Justin Spence (eds.), Proceedings of the Thirty-fifth Annual Meeting of the Berkeley Linguistics Society February 14-16, 2009, 1–12. Berkeley, CA: Berkeley Linguistics Society. Beckner, Clay, Richard Blythe, Joan Bybee, Morten H. Christiansen, William Croft, Nick C. Ellis, John Holland, Jinyun Ke, Diane Larsen-Freeman, and Tom Schoenemann. (2009). Language is a complex adaptive system: Position paper. Language Learning, 59, 1–26. Berg, Ivar. (2019). Gender and declension mismatches in West Nordic. In Michela Cennamo, and Claudia Fabrizio (eds.), Historical Linguistics 2015: Selected Papers from the 22nd International Conference on Historical Linguistics, Naples, 27–31 July 2015, 97–114. Amsterdam/Philadelphia: John Benjamins. Bernharðsson, Haraldur. (2004). Um Moldhaugnaháls út í Fjósa og Fjörður. Íslenskt mál og almenn málfræði, 26, 11–48. Bernharðsson, Haraldur. (2005). Ég er, ég vill og ég fær: Þáttur úr beygingarsögu eintölu framsöguháttar nútíðar. Íslenskt mál og almenn málfræði, 27, 63–101. Bernharðsson, Haraldur. (2006). Gás, gæs og Gásir, Gásar: Brot úr hljóðsögu og beygingarsögu. Orð og tunga, 8, 59–91. Bjorvand, Harald. (1972). Zu den altwestnordischen Pluralendungen -ar, -ir und -r bei femininen Substantiva. Norwegian Journal of Linguistics, 26, 195–215. Bjorvand, Harald. (1975). Altwestnordisch barar/bǫrur, fpl.: Eine Analyse der analogen Verbreitung der Pluralendung -ur der ōn-Stämme in den nordischen Sprachen. Norwegian Journal of Linguistics, 29, 101–112. Björnsdóttir, Sigríður Mjöll. (2021). Discovering Gender and Inflection. A view from Icelandic. Ph.D. dissertation. UIT The Arctic University of Norway. Björnsdóttir, Sigríður Mjöll. (2023). Predicting ineffability: Grammatical gender and noun pluralization in Icelandic. Glossa: a journal of general linguistics, 8(1), 1–40. Blevins, James P. (2016). Word and Paradigm Morphology. Oxford: Oxford University Press. Blevins, James P., and Juliette Blevins. (2009). Analogy: An introduction. In Blevins, James P., and Juliette Blevins (eds.), Analogy in grammar: Form and acquisition, 1–12. Oxford: Oxford University Press. Bod, Rens, Jennifer Hay, and Stefanie Jannedy. (2003). Introduction. In Bod, Rens, Jennifer Hay, and Stefanie Jannedy, Probabilistic Linguistics, 1–11. Cambridge, MA/London: MIT Press. Booij, Geert. (2010). Construction Morphology. Oxford: Oxford University Press. 222 Booij, Geert. (2019). The role of schemas in construction morphology. Word Structure, 12(3), 385–395. Booij, Geert, and Jenny Audring. (2018). Partial motivation, multiple motivation: The role of output schemas in morphology. In Booij, Geert (ed.), The Construction of Words: Advances in Construction Morphology, 59–80. Cham: Springer. Brown, Dunstan, and Andrew Hippisley. (2012). Network Morphology: A Defaults-based Theory of Word Structure. Cambridge: Cambridge University Press. Bulloch, Megan J., and John E. Opfer. (2009). What makes relational reasoning smart? Revisiting the perceptual-to-relational shift in the development of generalization. Developmental Science, 12(1), 114–22. Bybee, Joan. (1985). Morphology: A Study of the Relation between Meaning and Form. Amsterdam/Philadelphia: John Benjamins. Bybee, Joan. (1995). Regular Morphology and the Lexicon. Language and Cognitive Processes, 10, 425–455. [Reprinted in Bybee 2007.] Bybee, Joan. (2001). Phonology and Language Use. Cambridge: Cambridge University Press. Bybee, Joan. (2003). Cognitive processes in grammaticalization. In Tomasello, Michael (ed.), The New Psychology of Language: Cognitive and Functional Approaches to Language Structure, Volume II, 151–174. New York, NY: Psychology Press. Bybee, Joan. (2006). From usage to grammar: The mind's response to repetition. Language, 82(4), 711–733. Bybee, Joan. (2007). Frequency of Use and the Organization of Language. Oxford: Oxford University Press. Bybee, Joan. (2008). Formal universals as emergent phenomena: The origins of structure preservation. In Good, Jeff (ed.), Linguistic Universals and Language Change, 108–121. Oxford/New York, NY: Oxford University Press. Bybee, Joan. (2010). Language, Usage and Cognition. Cambridge: Cambridge University Press. Bybee, Joan. (2015). Language Change. Cambridge: Cambridge University Press. Bybee, Joan, and Clay Beckner. (2015). Language use, cognitive processes and linguistic change. In Bowern, Claire, and Bethwyn Evans (eds.), The Routledge Handbook of Historical Linguistics, 503–518. Oxon/New York, NY: Routledge. Bybee, Joan, and James L. McClelland. (2005). Alternatives to the combinatorial paradigm of linguistic theory based on domain general principles of human cognition. The Linguistic Review, 22, 381–410. Bybee, Joan, and Carol Lynn Moder (1983). Morphological classes as natural categories. Language, 59, 251–270. [Reprinted in Bybee 2007.] 223 Bybee, Joan, and Sandra Thompson. (1997). Three frequency effects in syntax. Berkeley Linguistics Society 23, 65–85. Carstairs-McCarthy, Andrew. (2017).76 Paradigmatic structure: Inflectional paradigms and morphological classes. In Spencer, Andrew, and Arnold M. Zwicky (eds.), The Handbook of Morphology, 322–334. Hoboken, NJ: John Wiley and Sons. Chomsky, Noam. (1965). Aspects of the Theory of Syntax. Cambridge, MA: MIT Press. Chomsky, Noam. (1988). Language and the Problems of Knowledge. Cambridge, MA: MIT Press. Chomsky, Noam. (2017). The Galilean challenge: Architecture and evolution of language. Journal of Physics: Conference Series, 880(1). <https://doi.org/10.1088/17426596/880/1/012015> Chomsky, Noam, and Morris Halle. (1968). The Sound Pattern of English. New York, NY: Harper and Row. Christie, Stella, and Dedre Gentner. (2010). Where hypotheses come from: Learning new relations by structural alignment. Journal of Cognition and Development, 11(3), 356– 373. Cienki, Alan. (2015). Spoken language and usage events. Language and Cognition, 7, 499– 514. Clahsen, Harald. (1999). Lexical entries and rules of language: A multidisciplinary study of German inflection. Behavioral and Brain Sciences, 22, 991–1060. Clausner, Timothy C., and William Croft. (1997). Productivity and schematicity in metaphors. Cognitive Science, 21(3), 247–282. Corbett, Greville. (1991). Gender. Cambridge: Cambridge University Press. Cordes, Anne-Kristin. (2017). The roles of analogy, categorization, and generalisation in entrenchment. In Schmid, Hans-Jörg (ed.), Entrenchment and the Psychology of Language Learning: How we Reorganise and Adapt Linguistic Knowledge, 269–288. Washington, DC: American Psychological Association. Croft, William. (2000). Explaining Language Change: An Evolutionary Approach. London: Longman. Croft, William. (2002). The Darwinization of linguistics. Selection, 3(1), 75–91. Croft, William, and D. Alan Cruse. (2004). Cognitive Linguistics. Cambridge: Cambridge University Press. Cser, András. (2023). Sound patterns, frequency and predictability in inflection. Acta Linguistica Academica, 70(3), 397–409. Davidson, Herbert. 1990. Han hon den: Genusutvecklingen i svenskan under nysvensk tid (Lundastudier i nordisk språkvetenskap A45). Lund: Lund University Press. 76 Mistakenly included in Paper III as ‘Carstairs-McCarthy, Andrew. 2015’. 224 Dąbrowska, Ewa. (2008). The later development of an early-emerging system: the curious case of the Polish genitive. Linguistics, 46(3), 629-650. Dąbrowska, Ewa. (2015). What exactly is Universal Grammar, and has anyone seen it? Frontiers in Psychology, 6, 1–17. De Smet, Hendrik. (2017). Entrenchment effects in language change. In Schmid, Hans-Jörg (ed.), Entrenchment and the Psychology of Language Learning: How we Reorganise and Adapt Linguistic Knowledge, 75–99. Washington, DC: American Psychological Association. Dobson, Annette J. (1973). Estimating time separation for languages. In Dyen, Isidore (ed.), Lexicostatistics in Genetic Linguistics, 56-63. Boston: Mouton. Dressler, Wolfgang U. (1997). On productivity and potentiality in inflectional morphology. Cross-Language Aphasia Study Network (CLASNET) Working papers, 7, 3–22. Dressler, Wolfgang U. (2003). Degrees of grammatical productivity in inflectional morphology. Italian Journal of Linguistics, 15, 31–62. Embick, David. 2015. The Morpheme: A Theoretical Introduction. Berlin: Mouton. Enger, Hans-Olav. (2004). On the relation between gender and declension. Studies in Language, 28(1), 51–82. Enger, Hans-Olav. (2022). Type frequency is not the only factor that determines productivity, so the Tolerance Principle is not enough. Beiträge zur Geschichte der deutschen Sprache und Literatur, 144(2),161–187. Evans, Nicholas, and Stephen C. Levinson. (2009). The myth of language universals: Language diversity and its importance for cognitive science. Behavioral and Brain Sciences, 32(5), 429-448. Eyþórsson, Þórhallur. (2015). The Insular Nordic experimental kitchen: Changes in case Marking in Icelandic and Faroese. In Whelpton, Matthew, Guðrún Björk Guðsteinsdóttir, Birna Arnbjörnsdóttir, and Martin Regal (eds.), An Intimacy of Words/Innileiki orðanna. Essays in Honour of Pétur Knútsson/Festschrift til heiðurs Pétri Knútssyni, 328–352. Reykjavík: Stofnun Vigdísar Finnbogadóttur í erlendum tungumálum/Háskólaútgáfan. Eyþórsson, Þórhallur, Janne Bondi Johannessen, Signe Laake, and Tor A. Åfarli. (2012). Dative case in Icelandic, Faroese and Norwegian: Preservation and non-preservation. Nordic Journal of Linguistics, 35(3), 219–249. Fertig, David. (2013). Analogy and Morphological Change. Edinburgh: Edinburgh University Press. Finkel, Raphael, and Gregory Stump. (2007). Principal parts and morphological typology. Morphology, 17, 39–75. Frost, Ram, Blair C. Armstrong, Noam Siegelman, and Morten H. Christiansen. (2015). Domain generality versus modality specificity: The paradox of statistical learning. Trends in Cognitive Sciences, 19, 117–125. 225 Føroyakort. Umhvørvisstovan. <foroyakort.fo>. [accessed 9.7.2021.] Føroysk orðabók. (1998). Føroysk Fróðskaparfelag og Fróðskaparsetur Føroya. Gardani, Francesco. (2013). Dynamics of Morphological Productivity: The Evolution of Noun Classes from Latin to Italian. Leiden: Brill. Gardeła, Leszek. (2012). What the Vikings did for fun? Sports and pastimes in medieval northern Europe. World Archaeology, 44, 234–247. Garrett, Andrew. (2008). Paradigmatic uniformity and markedness. In J. Good (ed.), Linguistic Universals and Langauge Change, 125–143. Oxford: Oxford University Press. Gentner, Dedre. (2005). The development of relational category knowledge. In GershkoffStowe, Lisa, and David H. Rakison (eds.), Building Object Categories in Developmental Time, 245–275. Mahwah, NJ/London: Lawrence Erlbaum Associates. Gentner, Dedre, and Jennifer Asmuth. (2019). Metaphoric extension, relational categories, and abstraction. Language, Cognition and Neuroscience, 34(10), 1298–1307. Gentner, Dedre, and Christian Hoyos (2017). Analogy and abstraction. Topics in Cognitive Science, 9, 672–93. Gentner, Dedre, and Arthur B. Markman. (1997). Structure mapping in analogy and similarity. American Psychologist, 52, 45–56. Gerken, LouAnn. (2010). Infants use rational decision criteria for choosing among models of their input. Cognition, 115(2), 362–366. Goldberg, Adele E. (2006). Constructions at Work: The Nature of Generalization in Language. New York, NY: Oxford University Press. Greenberg, Joseph H. (1966). Language Universals: With Special Reference to Feature Hierarchies. Berlin/New York, NY: Mouton. Gries, Stephan Th., and Nick C. Ellis. (2015). Statistical measures for usage-based linguistics. Language Learning, 65, 228–255. Guðmundsdóttir, Margrét. (2008). Málbreytingar í ljósi málkunnáttufræði. Íslenskt mál og almenn málfræði, 30, 7–52. Guðmundsson, Valtýr. (1922). Islandsk Grammatik. Islandsk Nutidssprog. Copenhagen: H. Hagerup Forlag. Gunnlaugsson, Guðvarður Már. (2000). Færeyskar málheimildir. In Snædal, Magnús and Turið Sigurðardóttir (eds.), Frændafundur, 3, 91–105. Reykjavík: Háskólaútgáfan. Halford, Graeme S., and Glenda Andrews. (2007). Domain general processes in higher cognition: Analogical reasoning, schema induction and capacity limitations. In Roberts, Maxwell J. (ed.), Integrating the Mind: Domain General versus Domain Specific Processes in Higher Cognition, 213–232. New York, NY: Psychology Press. Hammershaimb, Vencenslaus Ulricus. (1891). Færøsk Anthologi I–II. Copenhagen: S.L. Møllers Bogtrykkeri. 226 Hansson, Gunnar Ólafur. (2007). Productive syncretism in Saami inflectional morphology. In Toivonen, Ida, and Diane Nelson (eds.), Saami Linguistics, 91–135. Amsterdam: John Benjamins. Haspelmath, Martin. (2006). Against markedness (and what to replace it with). Language, 42, 25–70. Hauser, Marc, Noam Chomsky, and William Tecumseh Fitch. (2002). The faculty of language: What is it, who has it, and how did it evolve? Science, 298(5598), 1569– 1579. Hay, Jennifer. (2002). From speech perception to morphology: Affix ordering revisited. Language, 78, 527–555. Herce, Borja. (2019). Deconstructing (ir)regularity. Studies in Language, 43(1), 44–91. Hoffmann, Sebastian. (2004). Are low-frequency complex prepositions grammaticalized? On the limits of corpus data — and the importance of intuition. In Lindquist, Hans, and Christian Mair (eds.), Corpus Approaches to Grammaticalization in English, 171–210. Amsterdam: John Benjamins. Hopper, Paul J., and Sandra A. Thompson. (1984). The discourse basis for lexical categories in universal grammar. Language, 60, 703–52. Houghton, Donald E. (1968). Humor as a factor in language change. The English Journal, 57, 1178–1181–1186. isTenTen: Corpus of the Icelandic Web. Scetch Engline. <https://www.sketchengine.eu/isTenTen-Icelandic-corpus/> [Accessed 18.05.2021]. Iversen, Ragnvald. (1972). Norrøn grammatik. Oslo: H. Aschehoug and Co. Íslensk orðabók. (2010). Árnason, Mörður (ed.). Unabridged edition. Reykjavík: Forlagið. Íslensk orðtíðnibók. See Pind et al. (1991). Íslenskt textasafn. Stofnun Árna Magnússonar í íslenskum fræðum. <http//corpus.arnastofnun.is> [Accessed 30.07.2022]. Jakobson, Roman. (1939). Signe zéro. Roman Jakobson, Selected Writings, 211–219. The Hague: Mouton. Jakobson, Roman. (1990). Some questions of meaning. In Waugh, Linda R., and Monique Monville-Burston (eds.), On language: Roman Jakobson, 315–323. Cambridge, MA: Harvard University Press. Janda, Laura A. (2002). Cognitive hot spots in the Russian case system. Shapiro, Michael (ed.), Peircean Semiotics: The State of the Art, 165–88. New York, NY/Oxford: Berghahn Books. Janda, Laura A. (2007). Inflectional morphology. In Geeraerts, Dirk, and Hubert Cuyckens (eds.), The Oxford Handbook of Cognitive Linguistics, 632–649. New York, NY: Oxford University Press. 227 Jost, Ethan, and Morten H. Christiansen. (2017). Statistical learning as a domain-general mechanism of entrenchment. In Schmid, Hans-Jörg (ed.), Entrenchment and the Psychology of Language Learning: How we Reorganise and Adapt Linguistic Knowledge, 227–44. Washington, DC: American Psychological Association. Jónsdóttir, Margrét. (1988–1989). Um ir- og ar-fleirtölu einkvæðra kvenkynsorða í íslensku. Íslenskt mál og almenn málfræði, 10–11, 57–83. Jónsdóttir, Margrét. (1993). Um ar- og ir-fleirtölu karlkynsnafnorða í nútímaíslensku. Íslenskt mál og almenn málfræði, 15, 77–98. Jónsdóttir, Margrét. (2020). Epík, keramík og klassík: Gerð og beyging fleirkvæðra orða sem enda á -ík. Orð og tunga, 22, 19–37. Jónsson, Jóhannes Gísli, and Þórhallur Eyþórsson. (2011). Structured exceptions and case selection in Insular Scandinavian. In Wiese, Heike, and Horst Simon (eds.), Expecting the Unexpected: Exceptions in Grammar, 213-241. Berlin: Mouton de Gruyter. Kay, Paul, and Charles J. Fillmore. (1999). Grammatical constructions and linguistic generalizations: The what’s X doing Y? construction. Language, 75(1), 1–33. King, Robert D. (1969). Historical Linguistics and Generative Grammar. Englewood Cliffs, NJ: Prentice-Hall. Kiparsky, Paul. (1965). Phonological Change. Ph.D. dissertation. Massachusetts Institute of Technology. Kiparsky, Paul. (1974). Remarks on analogical change. In Anderson, John Mathieson, and Charles Jones (eds.), Historical Linguistics, 257–276. Amsterdam: North-Holland. Kiparsky, Paul. (2000). Analogy as optimization: ‘exceptions’ to Sievers’ Law in Gothic. In Lahiri, Aditi (ed.), Analogy, Levelling, Markedness: Principles of Change in Phonology and Morphology, 15–46. Berlin/New York, NY: Mouton de Gruyter. Knudsen, Trygve. (1967). Kasuslære, vol. 1: Innledning, nominative, akkusative. Oslo: Universitetsforlaget. Kodner, Jordan. (2019). Estimating child linguistic experience from historical corpora. Glossa: a Journal of General Linguistics, 4(1), 1–14. Kotovsky, Laura, and Dedre Gentner. (1996). Comparison and categorization in the development of relational similarity. Child Development, 67(6), 2797–2822. Krawczyk, Daniel C., Keith J. Holyoak, and John Hummel. (2004). Structural constraints and object similarity in analogical mapping and inference. Thinking and Reasoning, 10, 85– 104. Krawczyk, Daniel C., Keith J. Holyoak, and John Hummel. (2005). The one-to-one constraint in analogical mapping and inference. Cognitive Science, 29, 797–806. Kvaran, Guðrún. (2005). Íslensk tunga II: Orð. Reykjavík: Almenna bókafélagið. Kuryłowicz, Jerzy. (1945–49). La nature des procès dits „analogiques“. Acta Linguistica, 5, 15–37. 228 Kürschner, Sebastian, and Damaris Nübling. (2011). The interaction of gender and declension in Germanic languages. Folia Linguistica, 45(2), 355–388. Lahiri, Aditi. (2000). Introduction. In Lahiri, Aditi (ed.), Analogy, Levelling, Markedness, 1–14. Berlin: Mouton. Lakoff, George. (1987). Women, Fire and Dangerous Things. Chicago: University of Chicago Press. Lakoff, George. (2018). Ten Lectures in Cognitive Linguistics. Leiden: Brill. Langacker, Ronald. (1987). Foundations of Cognitive Grammar: Theoretical Prerequisites, vol. 1. Stanford, CA: Stanford University Press. Langacker, Ronald. (1988). A usage-based model. In Rudzka-Ostyn, Brygida (ed.), Topics in Cognitive Linguistics, 127–161. Amsterdam/Philadelphia: John Benjamins. Langacker, Ronald. (2008). Cognitive Grammar: A Basic Introduction. New York: Oxford University Press. Lightfoot, David. (2006). How New Languages Emerge. Cambridge: Cambridge University Press. Magnússon, Ásgeir Blöndal. (1989). Íslensk orðsifjabók. Reykjavík: Orðabók Háskólans. Mańczak, Witold. (1958). Tendances générales des changements analogiques. Lingua, 7, 298–325, 387–420. Markman, Arthur B. (1997). Constraints on analogical inference. Cognitive Science, 21, 373–418. Markman, Arthur B., and C. Hunt Stilwell. (2001). Role-governed categories. Journal of Experimental and Theoretical Intelligence, 13, 329–358. Markússon, Jón Símon. (2012). Eðli u-hljóðvarpsvíxla í íslenskri málsögu. MA dissertation. University of Iceland. Markússon, Jón Símon. (2017). Samband veiklunar og hljóðanvæðingar: Vitnisburður uhljóðvarpvíxla í frum- og vesturnorrænni málsögu. In Hansen, Zakaris Svabo, Anfinnur Johansen, Hjalmar P. Petersen, and Lena Reinert (eds.), Bók Jógvan: Heiðursrit til Jógvan í Lon Jacobsen á 60 ára degnum, 263–276. Tórshavn: Fróðskapur. Markússon, Jón Símon. (2021). Undir áhrifum (orða)gengis. In Axelsdóttir, Katrín, Veturliði Óskarsson and Þorsteinn G. Indriðason (eds.), Möggubrár heklaðar Margréti Jónsdóttur sjötugri, 21. mars 2021, 99–104. Reykjavík: Rauðhetta. Markússon, Jón Símon. (2022a). Tvær blækur labba inn á bar: On limited productivity as graded membership of an Icelandic microclass. NOWELE, 75(2), 194–222. Markússon, Jón Símon. (2022b). Um áhrif tíðni á stefnu útjöfnunar: Rannsókn á beygingarþróun færeysku nafnorðanna vøllur og fjørður. Íslenskt mál og almenn málfræði, 44, 53–86. Markússon, Jón Símon. (2023a). Accounting for different rates of gender reanalysis among Icelandic masculine forms in plural -ur. Nordic Journal of Linguistics, 46(3), 331–356. 229 Markússon, Jón Símon. (2023b). Um hugrænar forsendur fyrir útvíkkun beygingarvíxla: Vitnisburður færeyskra kvenkynsnafnorða með ar-fleirtölu. Ritið, 23(3), 203–236. Matras, Christian. (1933). Stednavne paa de færøske Norðuroyar. Copenhagen: Aarbøger for Nordisk Oldkyndighed og historie. Murphy, Gregory L. (2002). The Big Book of Concepts. Cambridge, MA: MIT Press. OED (Oxford English Dictionary). Oxford: Oxford University Press. ONP (Ordbog over det norrøne prosasprog). <https://onp.ku.dk/onp/onp.php> [Accessed 18.1.2022]. Ottósson, Kjartan G. (1992). The Icelandic Middle Voice: The Morphological and Phonological Development. Lund: Department of Scandinavian languages, Lund University. Penn, Derek C., Keith J. Holyoak, and Daniel J. Povinelli. (2008). Darwin’s mistake: Explaining the discontinuity between human and nonhuman minds. Behavioral and Brain Sciences, 31, 109–178. Petersen, Hjalmar P. (1992). Skerpingin í føroyskum. In Snædal, Magnús, and Turið Sigurðardóttir (eds.), Frændafundur, 11–19. Reykjavík/Tórshavn: Háskólaútgáfan. Petersen, Hjalmar P. (2009). Gender Assignment in Modern Faroese. Hamburg: Verlag Dr. Kovač. Petersen, Hjalmar P. (2020). Føroysk mállæra 1. Kyn, orðmyndan og bending. Tórshavn: Nám. Petersen, Hjalmar P. and Renata Szczepaniak. (2018). The development of non-paradigmatic linking elements in Faroese and the decline of the genitive case. In Ackermann, Tanja, Horst J. Simon, and Christian Zimmer (eds.), Germanic Genitives, 115-149. Amsterdam/Philadelphia: John Benjamins. Petersen, Hjalmar P., and Laurence Voeltzel. (2024). Faroese Phonetics andPhonology. Berlin: Mouton de Gruyter. (To appear.) Pind, Jörgen (ed.), Friðrik Magnússon, and Stefán Briem. (1991). Íslensk orðtíðnibók. Reykjavík: Orðabók Háskólans. Pinker, Stephen. (1999). Words and Rules: The Ingredients of Language. New York, NY: Basic Books. Popper, Karl. (1959). The Logic of Scientific Discovery. London/New York, NY: Routledge. Posner, Michael I., and Stephen W. Keele. (1968). On the genesis of abstract ideas. Journal of Experimental Psychology, 77, 353–363. Pullum, Geoffrey K., and Barbara Scholz. (2002). Empirical assessment of stimulus poverty arguments. The Linguistic Review, 19(1–2), 9–50. Quinion, Michael. Bloke. World Wide Words, 16.6.2004 (updated 22.1.2011). <http://www.worldwidewords.org/qa/qa-blo3.htm> [accessed 17.4.2021]. 230 Rácz, Péter, Clay Beckner, Jennifer Hay, and Janet B. Pierrehumbert. (2020). Morphological convergence as on-line lexical analogy. Language, 96(4), 1–36. Ralli, Angela. (2002). The role of morphology in gender determination: Evidence from Modern Greek. Linguistics, 40, 519–551. Rosch, Eleanor. (1975). Cognitive representations of semantic categories. Journal of Experimental Psychology: General, 104, 192–233. Rosch, Eleanor. (1978). Principles of categorization. In Rosch, Eleanor, and Barbara B. Lloyd (eds.), Cognition and Categorization, 27–48. Hillsdale, NJ: Erlbaum. Rosch, Eleanor, and Carolyn B. Mervis. (1975). Family resemblances: Studies in the internal structure of categories. Cognitive Psychology, 7, 573–605. Rosch, Eleanor, Carolyn B. Mervis, Wayne D. Gray, David M. Johnson, and Penny Boyes Braem. (1976). Basic objects in natural categories. Cognitive Psychology, 8, 382–439. Ross, Brian H., and Valerie S. Makin. (1999). Prototype versus exemplar models in cognition. In Steinberg, Robert J. (ed.), The Nature of Cognition, 205–241. Cambridge, MA: MIT Press. Rögnvaldsson, Eiríkur. (1981). u-hljóðvarp og önnur a~ö víxl í nútímaíslensku. Íslenskt mál og almenn málfræði, 3, 25–58. Rögnvaldsson, Eiríkur. (2013). Hljóðkerfi og orðhlutakerfi íslensku. [Electronic edition.] Reykjavík. <https://notendur.hi.is/eirikur/hoi.pdf.> Schmid, Hans-Jörg. (2017). A framework for understanding linguistic entrenchment and its psychological foundations. In Schmid, Hans-Jörg (ed.), Entrenchment and the Psychology of Language Learning: How we Reorganise and Adapt Linguistic Knowledge, 227–44. Washington, DC: American Psychological Association. Sigurðsson, Einar, Finnur Ágúst Ingimundarson, and Matthew Whelpton. (2022). Samræmi við hulin nafnorð. Orð og tunga, 24(1), 21–56. Sims‐Williams, Helen. (2016). Analogical levelling and optimisation: The treatment of pointless lexical allomorphy in Greek. Transactions of the Philological Society, 114, 315–338. Sims-Williams, Helen. (2022). Token frequency as a determinant of morphological change. Journal of Linguistics, 58, 571–607. Sims-Williams, Helen, and Hans-Olav Enger. (2021). The loss of inflection as grammar complication: Evidence from Mainland Scandinavian. Diachronica, 38(1), 111–150. Skjekkeland, Martin. 1997. Dei norske dialektane. Kristiansand: Høyskoleforlaget. Spencer, Andrew. Manufacturing consent over Distributed Morphology. (2019). Word Structure 12(2), 208–259. Stolyarova, Natalia. (2016). Greining á samanburði fallkerfa rússnesku og íslensku. BA dissertation. University of Iceland. 231 Svavarsdóttir, Ásta. (1993). Beygingarkerfi nafnorða í íslensku. Reykjavík: Málvísindastofnun Háskóla Íslands. Svavarsdóttir, Ásta. (2014). Er virkilega hægt að drekka kók í þremur kynjum? Vísindavefurinn 11.12.2014. Archived URL: http://visindavefur.is/svar.php?id=68642 (accessed 12.6.2020). Sveinsson, Sölvi. (2001). Íslensk málsaga. Reykjavík: Iðunn. Taylor, John R. (2003). Linguistic Categorisation. New York, NY: Oxford University Press. Taylor, John R. (2012). The Mental Corpus. How Language is Represented in the Mind. New York, NY: Oxford University Press. Teldutøka tekstasavn Føroyamálsdeildarinnar. Føroyamálsdeild, Fróðskaparsetur Føroya. <http://www.teldni.fo/tekstasavn/index>. [Accessed 9.7.2021.] Tiersma, Peter. (1982). Local and general markedness. Language, 58, 832–849. Tímarit.is. <www.timarit.is>. [Accessed 17.5.2022.] Tomasello, Michael. (2000). First steps toward a usage-based theory of language acquisition. Cognitive Linguistics, 11(1/2), 61–82. Trudgill, Peter. (2011). Sociolinguistic Typology: Social Determinants of Linguistic Complexity. Oxford/New York, NY: Oxford University Press. Tuggy, David. (2007). Schematicity. In Geeraerts, Dirk, and Hubert Cuyckens (eds.), The Oxford Handbook of Cognitive Linguistics, 82–116. New York, NY: Oxford University Press. Vikør, Lars S. (2001). The Nordic Languages: Their Status and Interrelations. Oslo: Novus Press. Weyhe, Eivind. (1996). Bendingarmunur í føroyskum málførum. Íslenskt mál og almenn málfræði, 18, 71–118. Weyhe, Eivind. (2012). Eivindaródn. Greinar 1979–2011. Tórshavn: Fróðskapur. Whelpton, Matthew, Drew Trotter, Þórhalla Guðmundsdóttir Beck, Curt Anderson, Joan Maling, Karthik Durvasula, and Alan Beretta. (2014). Portions and sorts in Icelandic: An ERP study. Brain and Language, 136, 44–57. Wiese, Heike, and Joan Maling. (2005). Biers, kaffi and schnaps: Different grammatical options for restaurant talk coercions in three Germanic languages. Journal of Germanic Linguistics, 17, 1–38. Winters, Margaret E. (2020). Historical Linguistics: A Cognitive Grammar Introduction. Amsterdam/Philadelphia: John Benjamins. Wittgenstein, L. (1978). Philosophical Investigations. [English translation by Gertrude Elizabeth Margaret Anscombe]. Oxford: Blackwell. Wurzel, Wolfgang U. (1984). Flexionsmorphologie und Natürlichkeit. Berlin: Akademie Verlag. 232 Yang, Charles. (2016). The Price of Linguistic Productivity: How Children Learn to Break the Rules of Language. Cambridge, MA: MIT Press. Þórhallsdóttir, Guðrún. (1997). Ylgr, heiðr, brúðr: Saga r-endingar nefnifalls eintölu kvenkynsorða. In Bragason, Úlfar (ed.), Íslensk málsaga og textafræði, 41–56. Reykjavík: Stofnun Sigurðar Nordals. Þórhallsdóttir, Guðrún. (2007). The dative singular of ō-stems in Old Norse. In Nussbaum, Alan J. (ed.), Verba Docenti: Studies in Historical and Indo-European Linguistics Presented to Jay H. Jasanoff by Students, Colleagues, and Friends, 329–41. Ann Arbor,MI/New York, NY: Beech Stave Press. Þórólfsson, Björn K. (1925). Um íslenskar orðmyndir á 14. og 15. öld og breytingar þeirra úr fornmálinu. Reykjavík: Fjelagsprentsmiðjan. [Reprinted 1987. Reykjavík: Málvísindastofnun Háskóla Íslands.] Þráinsson, Höskuldur. (2011). Um dauðans óvissan tíma. U-hljóðvarp lífs og liðið. Íslenskt mál og almenn málfræði 33, 85–107. Þráinsson, Höskuldur. (2017). U-umlaut in Icelandic and Faroese: Survival and death. In Bowern, Claire, Laurence Horn and Raffaella Zanuttini (eds.), On Looking into Words (and Beyond), 99–113. Berlin: Language Science Press. Þráinsson, Höskuldur, Hjalmar P. Petersen, Jógvan í Lon Jacobsen, and Zakaris Svabo Hansen.(2012). Faroese: An Overview and Reference Grammar. Tórshavn: FøroyaFróðskaparfelag. 233

Log In

Icelandic and Faroese: A usage-based cognitive analysis of morphological change

Related papers

Related papers

Related topics