From sound change to suppletion: case studies from Indo-European languages

From sound change to suppletion: case studies from Indo-European languages * RONALD I. KIM ABSTRACT: Although suppletion has attracted increasing attention in recent years, there is still no accepted dividing line between suppletion and other types of morphological irregularity. It is argued that even explicitly ahistorical treatments have been biased by diachrony, so that forms which are known to have developed from regular paradigms, e.g. Ancient Greek heîs ‘one’ ~ fem. mía or English think ~ thought, tend to be treated as less “truly” suppletive than those which have no known common source. Such cases are not only to be classified as suppletive on formal grounds, but deserve closer attention than they have heretofore received from historical linguists. Given the widespread view that morphological analogy acts to regularize paradigms which have become opaque as a result of phonological changes, suppletion of this sort may be viewed as the logical end point of sound change acting over significant time depths. Any diachronic typology of suppletion must therefore distinguish between paradigms composed of historically unrelated stems, and those whose stems have diverged through the cumulative effects of phonological change. With their long written records, Indo-European languages furnish numerous examples of the latter type. Key words: suppletion, morphology, inflection, regularity, sound change, analogy, Indo-European 1. Introduction Although recent years have seen an upsurge of interest in suppletion, there is still no consensus whether the phenomenon should be restricted to clear-cut examples such as English present go ~ past went, Spanish infinitive ir ‘go’ ~ preterite fui, and Czech singular člověk ‘person’ ~ plural lidé, or also encompasses cases such as English lead ~ led, Spanish querer ‘want’ ~ quiso, and Czech přítel ‘friend’ ~ přátelé. Should the latter be regarded as equally suppletive, less suppletive, or not suppletive at all, on a par with other morphological irregularities? And just how formally divergent do the signs in a morphological relation have to be in order for the relation to qualify as suppletive? The attempt by Corbett (2007: 14ff.) to establish a canonical type of suppletion marks an important step in this direction, yet he too openly admits that most of the commonly cited instances of suppletion are uncanonical in one or, frequently, multiple respects. The present study is motivated by what has long seemed to this author the strangest aspect of many theoretical treatments of suppletion, namely the assumption that the elements in a suppletive relationship must be etymologically unrelated. This view goes * This paper has benefited greatly from discussion with several colleagues, including Cormac Anderson, Jan Bičovský, Joseph F. Eska, and Reiner Lipp. I also thank Martin Maiden for sending a preprint of Maiden (2014); Nigel Vincent for making available a copy of Börjars & Vincent (2011) and for his comments on a final draft; and two anonymous reviewers for their thoughtful remarks. All errors and opinions remain entirely my responsibility. The research for this publication has taken place within the framework of the project “Diachronic Typology of Suppletion”, financed by the Czech Science Foundation (project no. GA14-10673S). It is my pleasure to acknowledge their support here. 354 Slovo a slovesnost, 77, 2016 back to the classic study of Osthoff (1899: 3ff.), who described suppletive paradigms in Indo-European languages as “unecht-stofflich”, i.e. containing members belonging to different “etymologischer Stoff”. More recently, Rudes (1980: 655ff.) defines verbal suppletion as the “grammatically conditioned alternation of etymologically unrelated forms representing the same morpheme within a paradigm,” distinguishing such cases from “pseudo-suppletive” verbs which arise via “radical phonological changes” (660ff.). He goes on to elaborate the relationship between the two subtypes: “While pseudo-suppletive verbs such as those discussed above differ from TRUE SUPPLETIVE VERBS by consisting of a single etymological root, they also differ from other verbs in the respective languages by their irregularity of forms. Thus, it is not surprising that often, over time, pseudo-suppletive verbs become truly suppletive.” (Rudes 662; emphasis original) Similarly, Markey (1985) states that “[o]n the surface at least, the results of suppletion are construed as deviant, but deviant in such a way that the discrepancy cannot be accounted for by regular phonological or morphological change” (53); he goes on to propose a “general developmental rule of thumb […] that etymological opacity indicates a grammatical ‘catastrophe’ (in the sense of René Thom) and is thus an index of developmental discontinuity” (55). However, historical linguists have long been aware that suppletion can evolve over time through the operation of sound change on inflectionally or derivationally related forms.1 The importance of phonological change as a source of suppletion has been recognized among others by Werner (1977: 273ff., with examples from Germanic languages), Ronneberger-Sibold (1988: 455), and Juge (2000: 184–186), yet many morphologists continue to conceive of suppletion primarily in terms of the merger of two or more originally distinct lexemes, admitting at most a secondary role for sound change. For instance, Spencer (1991: 8) distinguishes between “total” suppletion, involving “absolutely no phonological connection between the two forms” (e.g. good ~ better), and “partial” suppletion, typically the result of phonological change (e.g. think ~ thought; see below, section 3.2, example (4)) or borrowing. This dichotomy recurs in more recent literature, e.g. Haspelmath & Sims (2010: 25; “strong” go ~ went vs. “weak” think ~ thought), Börjars & Vincent (2011; “proper suppletion”, in which “one or more forms in a paradigm are borrowed from another paradigm”, vs. “phonological suppletion” due to “changes internal to a paradigm”).2 As a historical linguist and Indo-Europeanist, I do not propose to solve the problem of defining suppletion and its relationship to other types of morphological phenomena, certainly not in a few pages. My intention is rather to argue for the complete irrelevance of diachronic factors in determining suppletive status at a particular stage of a language: in the words of Mel’čuk (2006: 416), “etymological considerations should 1 See Hoenigswald (1960: 49 fn. 2) and the references in Werner (1977: 273 fn. 7). 2 A similar distinction between “suppletive” and “morphophonemically related” allomorphs was made by Hoenigswald (1960: 49), who however took care to qualify his generalization: “the latter typically arise from sound change (except that syncretism from accidentally similar or even morphologically related sources may produce deceptive instances); the former […] arise from syncretism […] (except insofar as they are merely the last remnant of extreme sound change).” Slovo a slovesnost, 77, 2016 355 play no role in defining a synchronic theoretical concept.”3 After some theoretical preliminaries (section 2), I examine in detail a selection of cases of suppletion from a range of ancient and modern Indo-European (IE) languages, some of them undeniably “strong”, that have demonstrably arisen (almost) entirely by regular sound change.4 The consequences of these facts for larger questions of language change are discussed in section 4, in particular the role of phonological change over longer periods of time in creating suppletive paradigms, and the consequences of such long-term change for language acquisition and the mental representation of the lexicon. 2. Preliminary remarks Before proceeding, I should make explicit my own assumptions about the nature and scope of suppletion. I leave aside derivational categories such as derived nouns and adjectives of location; this therefore excludes Massachusetts ~ Bay Stater, Rio de Janeiro ~ Carioca, or the less extreme Manchester ~ Mancunian, Oxford ~ Oxonian, which in my view clearly belong to the lexical level. Although such cases can certainly be considered suppletive (cf. Mel’čuk 2006: 417–418, with references going back to Osthoff 1899), it may be questioned whether they belong to the same plane of linguistic structure as inflectional suppletion.5 Somewhat more difficult to exclude are cases of complex, synchronically ummotivated allomorphy such as English Rhode Island ~ Rhode Island-er vs. New Jersey ~ New Jersey-an vs. Wisconsin ~ Wisconsin-ite or Polish Lublin ~ lubel-ski, Leszno ~ leszczyń-ski vs. the regular type of Kraków ~ krakow-ski, Szczecin ~ szczeciń-ski; but these too can be plausibly represented as lexical, and will in any case play no role in the following discussion.6 Pace Mel’čuk (2006: 435–436), I also do not consider allomorphy of inflectional morphemes to be suppletion: aside from theoretical objections, this would require one to recognize massive suppletion in the nominal and verbal inflection of most IE languages. However, I do include gradation of adjectives, which as well known exhibits all sorts of suppletive relationships in older as well as modern IE languages; see the stimulating recent study by Bobaljik (2012), which includes much data from non-IE languages as well. With respect to the concept of “ideal” or “strong” suppletion as involving stems which are synchronically underivable from one another, I do not insist on complete lack of phonetic resemblance. In this regard, I agree with Dressler (1985b: 102) that “there is no ready way of grading (not to speak of quantifying) strength of suppletion”; 3 Cf. Mel’čuk (2000: 514); likewise Veselinova (2006: 14), Corbett (2007: 13). 4 Only after completing a draft of this paper did I discover that Plank has made much the same point in his forthcoming paper (Plank forthcoming: section 2), where he cites some of the examples given below in section 3 (specifically, examples (1), (5), and (6)). 5 See the insightful remarks of Dressler (1985a: 106; 1985b: 102–103) concerning the role of sociolinguistic factors, in particular “the sociolinguistic and pragmatic function of signalling learnedness” through proper use of such unabashed neo-Latinisms as French Bellifontain ‘of Fontainebleau’ or English Haligonian ‘of Halifax’. 6 See the brief but judicious discussion of inflectional vs. derivational morphology in Corbett (2007: 12–13). 356 Slovo a slovesnost, 77, 2016 put another way, “[t]he condition “the members of a suppletive pair have no shared phonemes in their signifiers”, which one sees frequently, is neither necessary nor sufficient” (Mel’čuk 2000: 513). Though it may be useful on some taxonomic level to label English go ~ went or Modern Greek present vlépo ‘see’ ~ aorist í∂–a “strongly suppletive” and English think ~ thought or Greek érxome ‘come’ ~ írθ a “weakly suppletive”,7 there is to my knowledge no probative evidence from acquisitional or psycholinguistic studies that the latter are necessarily represented or accessed differently in the grammar, and in this regard more closely resemble irregular morphological patterns represented by multiple examples such as English sing ~ sang (cf. ring ~ rang, swim ~ swam) or Greek v éno ‘go out’ ~ v íka (cf. béno ‘go in’ ~ bíka, anevéno ‘go up’ ~ anévika). In fact, the Greek verb érxome ~ írθ a descends from an earlier érxome ~ ílθ a (< Ancient Greek érkhomai ~ ē̂lthon) by the sound change of *lC > *rC, which in this case has led to an accidental overlap between the present and aorist stems. Must the present-day paradigm therefore be considered less “genuinely” suppletive than its predecessor, in which the stems had no consonantal segments in common? In short, I am not aware of any convincing proposals to date to quantify degrees of suppletion according to formal (lack of) resemblance, nor of any detailed treatments in the morphological literature of borderline cases, which are far more numerous than usually assumed.8 This lack of clarity is hardly surprising: the fact is that “[t]here is a continuum of decreasing formal suppletivity from formally strong suppletion down to zero suppletion, with an infinity of intermediate cases” (Mel’čuk 2000: 519).9 Partly for this reason, I suspect, etymology is often invoked as a criterion to mark off “genuine” suppletion of the type of go ~ went from cases such as think ~ thought, which would simply be more irregular than sleep ~ slept or lead ~ led. Yet not only is it unjustified to define a synchronic phenomenon with reference to diachrony, but the underlying assumption, that sound change can only produce “weak suppletion” (see above, section 1), is demonstrably false. As the following section will demonstrate, the operation of phonological change over time can give rise to quite “strong” suppletion indeed, to the point that the etymological relationship among the stems involved is no longer recognizable. 7 The distinction goes back at least to Vennemann (1972: 231 fn. 24); cf. Carstairs (1988: 70–72, “gross” vs. “partial” suppletion), Spencer (1991: 8, “total” vs. “partial” suppletion), Mel’čuk (2006: 440–443, formally strong or “genuine” suppletion vs. formally weak or “quasi-”suppletion). 8 The current situation with respect to definitions of suppletion has reminded me more than once of the notorious statement of U.S. Supreme Court Justice Potter Stewart in the case Jacobellis vs. Ohio (1964), who admitted that he could not define “hard-core pornography”, yet insisted: “I know it when I see it.” 9 See also Boyé (2006: 297–298), Mel’čuk (2006: 418–419, 438–443), Haspelmath & Sims (2010: 25). The only major point on which I disagree with Mel’čuk’s otherwise convincing treatment of suppletion concerns the role of morphology in defining regularity of semantic relation; to take his example on pp. 438–440, is the morpheme “× 10” really a “less regular” semantic element in Turkic languages than, say, in Slavic languages? Synchronically, it would seem more justified to treat all the decads as lexically specified, i.e. “more or less” suppletive: hence e.g. Turkish yirmi ‘20’, otuz ‘30’, kırk ‘40’, elli ‘50’, altmış ‘60’, yetmiş ‘70’, seksen ‘80’, doksan ‘90’ (cf. iki ‘2’, üć ‘3’, dört ‘4’, beş ‘5’, altı ‘6’, yedi ‘7’, sekiz ‘8’, dokuz ‘9’), like Russian dvádcat’, trídcat’, sórok, pjat’desját, šest’desját, sém’desjat, vósem’desjat, devjanósto (cf. dva, tri, četýre, pjat’, šest’, sem’, vósem’, dévjat’). For an extreme case of lexical arbitrariness in the higher numerals, cf. Hindi/Urdu, where ‘11’ to ‘99’ are all synchronically underivable. Slovo a slovesnost, 77, 2016 357 3. From regularity to suppletion in Indo-European languages In many fusional and agglutinating languages with longer attested histories, including virtually all IE languages, there exist cases of morphological irregularity which are known to have developed by regular sound change from regular morphological correspondences at an earlier stage (cf. Dressler 1985b: 102, 103). Most of these may be considered instances of “weak suppletion”, as the stems involved retain at least some formal similarity to one another. Cf. the examples mentioned above in section 1: a) English lead [li:d] ~ led [lεd]; b) Spanish quer-er [ker-] ‘want’ ~ pret. quis-o [kis-]; c) Czech přítel ‘friend’ ~ pl. přátel-é. In the case of lead ~ led, the synchronically irregular preterite formation is the result of two historical processes operating on Old English lœ̄dan ~ pret. lœ̄dde, namely Middle English shortening and simplification of root-final dental + preterite marker *-d-. Both are found in meet ~ met and read [i:] ~ read [ε]; the first also recurs in keep ~ kept, sleep ~ slept, etc. Spanish quer-er [ker-] ~ quis-o [kis-] has likewise developed via a long series of mostly regular changes from Latin quaerere ~ quaesīvī < Old Latin *kwais- ~ *kwais-s-; but because the stems have the same initial consonant, this case is usually grouped together with other “irregular” preterites such as ten-er ‘have’ ~ tuv-o, pon-er ‘put’ ~ pus-o, etc. Finally, Czech přítel ~ přátel-é contains a unique vocalic alternation, historically resulting from differing contractions of the stem *prijatel(cf. Russian prijátel’, Polish przyjaciel); but since the two stems are otherwise identical in shape, přátelé is treated in descriptive grammars as a synchronic irregularity, rather than a “truly” suppletive plural like lidé ‘people’.10 In other cases, however, phonological change has given rise to what can synchronically only be called “strong suppletion”, where the allomorphs in question show little or no formal resemblance. The following examples have been selected from ancient and modern IE languages, but there is no reason to believe that comparable cases would not recur in highly inflecting non-IE languages.11 3.1. Nominal suppletion We begin with one of the most spectacular examples of suppletion arisen by a combination of divergent sound changes and ordinary leveling: (1) Ancient Greek hê:s (ε‘˜ıς) ‘one’, neuter hén (ε‘′ν) ~ feminine mía (µ íα) This unique paradigm goes back to one of the reconstructible Proto-Indo-European (PIE) lexemes whose reflexes in the daughter languages have come to be used as the 10 The case of dítě ‘child’, pl. děti, frequently invoked by my Prague colleagues, is somewhat different, since singular and plural belonged to different inflectional classes already in Proto-Slavic (i.e. the noun was “heteroclitic”): the former contained the diminutive suffix *-ęt- (< *-ent-), while the latter had a stem ending in *-ı̆- (< *-i-). 11 See section 4 below on plurals in the Northeast Caucasian language Archi. 358 Slovo a slovesnost, 77, 2016 cardinal numeral ‘one’: masculine/neuter *sém- ~ *sm-ę́, feminine *sém-ih2- ~ *sm-yéh2-. ˚ Both of these belonged to inflectional types universally reconstructed for the protolanguage. The former was a typical “root noun”, with stressed, fully vocalized (“full-grade”) root *sém- in the nominative and accusative alternating with unstressed, vowelless (“zero-grade”) root *sm- and stressed ending in the oblique cases; while the feminine ˚ alternated between a so-called strong stem *sém-ih2- with stressed full-grade root and unstressed zero-grade suffix *-ih2-, and a weak stem *sm-yéh2- with unstressed zerograde root and stressed full-grade suffix *-yéh2-.12 As Table 1 illustrates, the formal divergence between the masculine/neuter and feminine stems is almost entirely the result of Greek-specific sound changes, plus well-known morphological alterations such as the generalization of the gen. sg. ending *-os and leveling of the root vocalism in the feminine. Table 1: Evolution of *sem- from PIE to Classical Greek m. nom. n. nom./acc. m./n. gen. PIE *sḗm *sém *sm-és pre-Proto-Greek → *sem-s > *sem → *sem-ós > > > Proto-Greek *hen-s *hen *hem-ós Classical Greek > heîs [hê:s] > hén → hen-ós13 f. nom. gen. *sém-ih2 *sm-yéh2-s → *smia > *syâ:s > > *hmía *hiâ:s > mía → miâ:s14 This paradigm has remained remarkably robust in the subsequent history of Greek, surviving down into the modern language with only the usual phonological and morphological changes: masc. nom. énas, acc. éna, neut. nom./acc. éna, masc./neut. gen. enós vs. fem. nom./acc. mjá, gen. mjás. (2) a) French œil [œj] ‘eye’, pl. yeux [jø] b) Romanian om [om] ‘man’, pl. oameni [oa' menj] ˆ The French forms œil, yeux are the regular phonological outcomes of their Late Latin sources OCULUM, OCULŌS (< Classical Latin acc. oculum, oculōs); similarly, Romanian om ~ oameni regularly continue Late Latin HOMŌ, HOMINĒS (< Classical Latin nom. homō, nom./acc. hominēs). To be sure, the latter could be interpreted as containing an isolated infix -en- (cf. the Italian cognates uomo, uomini), with the phonologically conditioned alternation of o ~ oa known from hundreds of other nominal and verbal paradigms. Given the almost total lack of resemblance between singular and plural, 12 This latter type is called “proterokinetic” or “proterodynamic” in IE comparative grammar. For more on the reconstructed PIE nominal accent-ablaut paradigms, see Fortson (2009: 119–122). 13 The stem *hem- was altered to *hen- after masc. nom. *hen-s and neut. nom./acc. *hen, where *m > *n by sound change. The older form is attested in Mycenaean dative e-me /hem-éi/, corresponding to classical hen-í. 14 That Proto-Greek had an alternation in the feminine between nom./acc. *hmia- and gen./dat. *hia- is suggested by the distribution of variants in Homer (nom. mía ~ ía, gen. miâs ~ iâs, etc.). The latter stem has been explained as the regular phonological reflex of the weak stem *syā- < *syeh2- < *smyéh2-, with reduction of the consonant cluster *smy > *sy by a (pre-)PIE rule; see Gippert (2004: 160–164), with references going back to Schmidt (1900: 399). Slovo a slovesnost, 77, 2016 359 however, one wonders if the relation between the two is better described, and perceived by native speakers, as suppletive. Other examples of what is typically labeled “weak suppletion” of singular and plural stems may be observed elsewhere in modern IE languages: cf. Italian bue ‘ox’, pl. buoi < Late Latin BOVEM, BOVĒS, or (archaic and dialectal) English cow [ka ], pl. kine [kaIn] < Middle English cow [ku:], kine [ki:n] ← Old English cū, cȳ. These may be seen as special cases of the sort of morphological shuffling and rebuilding responsible for the proliferation of plural formations, and nominal inflection classes in general, in modern IE languages such as Irish or Romanian; yet once again one may legitimately ask at what point they cease to be mere “irregularities” and should be regarded as instances of suppletion.15 The following example does one better: under any synchronic analysis, it is unquestionably suppletive.16 (3) Modern Irish bean [bjæn] ‘woman’, genitive plural ban [b n] vs. genitive singular and nominative plural mná [mr :] ~ [mn :] These forms have developed almost purely by sound change from the archaic PIE paradigm of ‘woman’, in which the strong stem *gwén-h2- in the nominative and accusative alternated with the weak stem *gwn-éh2- (*gwn-éh2-) in the remaining cases.17 ˚ Such alternations were productive in the reconstructed PIE system of nominal inflection, but already in the oldest attested IE languages they were greatly reduced by the operation of leveling; in particular, root alternations have been almost entirely eliminated, surviving for the most part only in isolated forms. Old Irish ‘woman’ is among the very few nouns in which the PIE alternation of (stressed) *e vs. (unstressed) *Ø in the root has been preserved in a living paradigm; the only irregular change was the extension of the weak stem from the oblique cases (genitive and dative) to the nominative-accusative plural.18 Table 2: Evolution of ‘woman’ from PIE to Modern Irish nom. sg. gen. nom. pl. gen. PIE *gwén-h2 *gwn-éh2-s *gwén-h2-s *gwn-éh2-oHom ˚ → > → > pre-Old Irish *benā *bnās *bnās *banan > > > > Old Irish ben mná mná ban > > > > Modern Irish bean [bjæn] mná [mr :] ~ [mn :]19 mná [mr :] ~ [mn :] ban [b n] 15 Interestingly, the relationship between cow and kine is generally recognized as suppletive on an orthographic level, as the answer to the question “What is the only English noun whose singular and plural have no shared letters?” 16 Cf. Hoenigswald (1960: 49 fn. 2), where the Scottish Gaelic forms are cited. 17 The allomorph *gwn-éh - with syllabic nasal is due to a variable syllabification rule of initial 2 ˚ sequences of obstruent + sonorant + vowel in PIE (“Lindeman’s Law”; Fortson 2009: 72). See gen. pl. ban in Table 2. 18 And to the accusative singular mnaí, beside archaic Old Irish bein < *benen < *benan < PIE *gwénh2-m . ˚19 The change of [Cn -] > [Cr -] occurs in the dialects of Connacht and Ulster (cf. cnoc [kr -] ‘hill’), as well as in Scottish Gaelic (gen. sg. mnà [mra:], pl. mnathan [mra n]). e 360 Slovo a slovesnost, 77, 2016 3.2. Verbal suppletion Turning to the verb, we find similar cases in which the operation of sound changes has almost entirely obscured the formal relation between paradigmatically associated stems. Rudes (1980: 660) cites Romanian pres. 1sg. ia-u ['ja-w] ~ 1pl. lu-ăm [lu-' m] ‘take’, in which Late Latin *levā- (cf. Latin levāre ‘lift up’, Italian levare ‘take away’) has developed differently depending on the position of stress, producing a unique alternation in the language.20 Another example comes from Modern English, where certain preterites now share only an initial consonant with the corresponding present. e (4) English buy ~ bought, seek ~ sought, think ~ thought, work ~ wrought The ancestors of these verbs belonged to a small inflectional subtype in Old English and other early Germanic languages, in which the pre-Proto-Germanic stem vowel *-iwas syncopated in the preterite but not the present; the root vowel then underwent iumlaut in the present, but remained unaltered in the preterite.21 The changes involved are illustrated below for seek ~ sought: Table 3: Evolution of ‘seek’ from pre-Proto-Germanic to Modern English pre-Proto-Germanic *sōkij-anan > *sōki-dē > Proto-Germanic *sōkijanan > *sōktē [-xt-] > Old English sēc.an [se: an] sōhte [so:xte] Modern English → seek [sik]22 > sought [s t] c pres. infinitive pret. 3sg. An additional argument for regarding this type as suppletive, rather than simply “irregular” as in the case of e.g. tell ~ told, is that the final -t is synchronically unmotivated as a preterite marker. From the viewpoint of present-day English grammar, a preterite such as bought is an indivisible morpheme, and the formal divergence of buy ~ bought is exceeded only by the canonical be ~ was and go ~ went. Farther afield, Latin ferō ‘bring, carry’, preterite tulī, past passive participle lātus provides a familiar Paradebeispiel of complex suppletion, yet the three-way stem contrast is due to a late sound change in the prehistory of the language. Historically, only two PIE stems are involved in the morphology of this verb, *bher- (cf. Sanskrit bhar-, Greek pher-, English bear, Czech ber-) and *telh2- (cf. Latin tollō ‘lift (up), raise’, Old Irish tlenaid ‘takes away, steals’ < PIE nasal pres. *tl-né-h2- ~ *tl-n-h2-′). The former was ˚ ˚ defective in the protolanguage, forming only a present; its preterite in Latin thus came 20 The famous case of Proto-Slavic *xoditi ‘go (frequentative)’ < *sod- vs. perfect active participle *šı̆d-ŭš-, *šı̆d-lŭ < *sed- does not strictly speaking belong here, since the latter synchronically pattern with nonfrequentative *i-ti ‘go’. – Note that the discussion of Hittite te- ~ tar- ‘say’ (e.g. pres. 3sg. te-zzi ['te:tsi], 3pl. tar-anzi [ta'rantsi]) by Rudes (1980: 661) is entirely outdated; see Kloekhorst (2008: 870–871, s.v. ter-). 21 The Old English forms are respectively byc.g.an ‘buy’, pret. bohte; sēc.an ‘seek’, pret. sōhte; enc.an . ‘think’, pret. ōhte; and wyrcan, pret. worhte ~ wrohte; the presents originally contained the Proto-Germanic suffix *-i- ~ *-ja- < PIE *-ye- ~ *-yo-. Cf. also bringan ‘bring’, pret. brōhte, which is exceptional in having a present in *-i- ~ *-a- < PIE *-e- ~ -o- typical of so-called strong presents (Ringe 2006: 251–252). This small group, which appears to be of Proto-Germanic date, was later joined in (parts of) West Germanic by numerous other verbs, including the type of English tell ~ told; for details, see Ringe & Taylor (2014: 71–75, 97–99). 22 The present was replaced in Middle English after 2sg. seekest, 3sg. seeketh < Old English sēcst, sēc- ; the older form survives in be-seech. Slovo a slovesnost, 77, 2016 361 to be supplied by forms of *telh2-, namely (post-)PIE perfect (*te-tólh2- ~) *te-tlh2-′ > ˚ Old Latin pret. tetulī → Classical Latin tulī and deverbal adj. *tl h2-tó- > *tlātos > ˚ Classical Latin past pass. ptcp. lātus.23 (5) Latin pret. tulī ~ past pass. ptcp. lātus The simplification of the initial cluster in *tlātos must have severed the synchronic link between preterite and participle; at any rate, I am aware of no evidence that speakers of Classical Latin still perceived a derivational relationship between the two. The verb was later lost in most Romance languages, but the participle survives as a lexicalized substantive in e.g. Italian lato, Spanish lado ‘side’. A widespread example of suppletion arisen largely through the cumulative effects of sound change is the anomalous present of ‘be’ in most older IE languages, e.g. (6) Latin sum, es, est, pl. sumus, estis, sunt Old English 3sg. is, pl. sind 24 Old Church Slavonic 3sg. jestŭ, pl. so ˛ tŭ In PIE, *h1es- ‘be’ formed an entirely regular root present with stress alternating between root in the singular and ending in the dual and plural, i.e. *h1és-mi, *h1és-si, *h1és-ti vs. pl. *h1s-mós, *h1s-té, *h1s-énti.25 This paradigm is still transparent in Sanskrit (ásmi, ási, ásti, smás, sthá, sánti); but in even the closely related Avestan, the change of *s > *h in most positions has created an opacity between e.g. 3sg. asti and 3pl. hanti. In the cases in (6), the reflexes of PIE 3sg. *h1és-ti and 3pl. *h1s-énti have become completely opaque, and though they share the segment [s], it is most unlikely that speakers could have perceived any sort of rule-based relationship between them. The same of course holds true for later stages of the languages in question, e.g. French est [e] ~ sont [s˜ ] or Czech je(st) ~ jsou.26 Finally, a special case of suppletion may be identified in the Old Irish verb, with its peculiar system of double inflection. In Old Irish, simplex verbs take the so-called “absolute” form when initial in a main clause; otherwise they occur in the “conjunct” form. Compound verbs pattern in the same way, appearing in the “deuterotonic” form when initial and the “prototonic” form otherwise. As these terms suggest, the distinction between the two was purely phonological and originally involved the position of stress, which fell on the second component of the verbal complex, i.e. the element after the (first) preverb if nothing preceded (hence “deuterotonic”), but the (first) preverb if preceded by negation or any other element (hence “prototonic”). The stress contrast enc 23 For details, see Meiser (1998: 108, 210; 2003: 192) and Weiss (2011: 432). 24 Beside sint (with final devoicing in unstressed position) and sindon (with -on added from the class of presents traditionally called “preterite-presents”, e.g. wit-on ‘know’, cunn-on ‘are able’; Brunner 1965: 353). 25 Cf. the pattern of root nouns described above under (1). The dual forms have been left out for simplicity; also, the reconstruction of the 1pl. ending is not entirely certain. Note that the 2sg. was realized as *[h1ési] according to a synchronic PIE rule (cf. Sanskrit ási, Hittite ēši, Ancient Greek eî < *ehi < *esi; Fortson 2009: 70). 26 See Werner (1977: 273) on the present of ‘be’ in German and English. Needless to say, this is only one source of copular suppletion in IE languages, which reflect the PIE roots *bhuh2- ‘become’, *h2wes- ‘remain, dwell’, and *steh2- ‘stand’ as well as *h1es-. For discussion of copular suppletion crosslinguistically, see Veselinova (2006: 115–126). 362 Slovo a slovesnost, 77, 2016 tailed a series of divergent phonological developments, so that by the Old Irish period the two allomorphs usually differed considerably: cf. deuterotonic do⋅ beir [do'bjerj] ‘gives’ < *to-s 'bereθ vs. prototonic ní tabair ['t v rj] ‘does not give’ ← *nēs 'to-bereθ.27 Although the phonology of Old Irish involves a fair amount of abstraction under any analysis, and the proto- and deuterotonic stem allomorphs of many, even most verbs could in principle be derived from a single underlying form by the application of synchronic phonological rules such as shortening, syncope, and assimilation (McCone 1997: 4–9 and passim), it is not obvious to me that that was still the situation in Classical Old Irish. Pairs such as the following, which developed by sound change, were probably almost as opaque to speakers as undeniable cases of suppletion, e.g. pres. 1sg. tíagu vs. 3sg. téit ‘go(es)’, fut. regaid, pret. luid.28 e (7) Old Irish (all forms present indicative 3sg.) ad⋅ suidi [a∂–'su∂– ji] ‘details’ < *aθ -es 'so∂–īθ vs. (ní) astai ['asti] ‘does not detail’ < *aθ sθ i < *'aθ -so∂–īθ do⋅ rósc(c)ai [do' ro:ski] ‘excels’ ← *di-s 'ro-(u)ss-skuxīθ vs. (ní) derscaigi ['dersk ji] ‘does not excel’ < *'di-ro-(u)ss-skuxīθ do⋅ sluindi [do'sLunjdji] ‘denies’ < *di-s 'slondīθ vs. (ní) díltai ['di:Lti] ‘does not deny’ < *dihl(n)di < *'di-slondīθ e The relationship between prototonic and deuterotonic continued to increase in complexity and opacity in Middle Irish, until the language eventually eliminated all but a handful of these functionally otiose doublets. The drastic sound changes that took place in the prehistory of Old Irish have given rise to even more remarkable instances of suppletion. Consider the following imperatives: (8) do⋅ goa [do'goa] ‘chooses’, iptv. tog [to ] ad⋅ guid [a∂–'gu∂– j] ‘invokes (as surety)’, iptv. aic(c) [agj] These forms have been convincingly explained by Jasanoff (1986) as phonologically regular reflexes of an archaic PIE imperative in *-si, in origin a haplologized 2sg. subjunctive to aorist stems formed with the suffix *-s-.29 Combined with the divergence between proto- and deuterotonic forms just described, the result is allomorphy of an extreme sort: cf. PIE *gwhédh-s-i ['gwhetsi] → pre-Old Irish *ad-gwessi > *adgwes > *agjgjeh > Old Irish aic(c) [agj]. For speakers of the language, these imperatives, which resembled no other finite forms of their respective paradigms, must have been listed separately in the lexicon.30 27 The replacement of stressed o by a in tabair is irregular, but all the other developments are well established. 28 To consider just one detail, the pre-Old Irish preverbs *di- vs. *to- were preserved under stress in prototonic forms, but merged as do⋅ in deuterotonic forms. Hence do⋅ gní ‘does’ < *di-s 'gnīeθ like do⋅ beir ‘gives’ ← *to-s 'bereθ, but ní dén(a)i ‘does not do’ < *nīs 'di- nīeθ contrasts with (ní) tabair ‘does not give’ ← *nīs 'to-bereθ. A similar confusion among pretonic ad⋅ , as⋅ , and in⋅ may be observed; see McCone (1997: 5, 9). 29 The type is best represented in Indo-Iranian, but has also left traces in Tocharian and possibly Hittite. Cf. PIE *ĝews- ‘taste’, aor. subj. 2sg. *ĝéws-s-e-si > iptv. *ĝéwssi > Vedic jós. i ‘enjoy!’. 30 Iptv. tog ‘choose!’ does resemble the verbal noun togu, but verbal nouns in Old Irish, as in Classical Arabic, must themselves be specified for each individual verb in the lexicon. Cf. the remarks on derivation above in section 2. Slovo a slovesnost, 77, 2016 363 3.3. Adjectival comparison The last set of suppletive relationships to be explored comes from adjectival comparison. As is well known, the great majority of suppletive comparatives and superlatives in IE languages have no conceivable historical connection to the positive and/or to each other, e.g. English good ~ [better ~ best], bad ~ [worse ~ worst] or Latin bonus ~ melior ~ optimus. The lability of suppletive stems is also documented in several IE branches: one need only compare the Slavic comparative forms for ‘good’, e.g. Russian xoróšij ~ lúč-šij vs. Czech dobrý ~ lep-ší vs. Serbo-Croatian dobar ~ bolj-ī. However, instances of weak suppletion have been known to arise by regular sound change alone, e.g. (9) Latin magnus ‘big, great’, comparative major [majjor], superlative maximus31 Magnus is an ordinary adjective formed with the suffix *-no- to the PIE root *meĝh2‘great’, also continued in Sanskrit máhi and Greek mégas. The generalized root shape *mag- served as the basis for the comparative and superlative, but in both forms regular sound changes have obscured the link to the positive: *mag-jōs- > *majjōr- > major [majjor]; *mag-ism mo- > *mag-simo- > *mā̆ksimo- > mā̆ximus.32 One may speculate ˚ that the resulting (weak) suppletion contributed to the breakdown of the paradigm, as magnus was replaced almost everywhere in the Romance languages and the reflexes of major and maximus became functionally and pragmatically restricted; see e.g. Maiden & Robustelli (2007: 344, 351–352) on Italian maggiore, massimo vs. regular più grande, grandissimo.33 Another likely case of suppletion arising through regular sound change in adjectival comparison involves Ancient Greek polús ‘much, many’. This adjective, with its uniquely irregular inflection (masc. nom. sg. polús, acc. polún, neut. nom./acc. sg. polú, otherwise polló/ε′:- in fem. nom. sg. pollε′:, masc. nom. pl. polloí, etc.), goes back to a PIE paradigm *pélh1-u- ~ *pl h1-éw- also continued in Celtic (Old Irish il), Germanic ˚ (Gothic filu), and Indo-Iranian (Vedic purú-, Avestan pouru-); the comparative and superlative are the more or less regular plé :n (plé: :n) ← PIE *pléh1-yos-, pleîstos < PIE *pléh1-is-to-, cognate respectively with Latin plūs, Old Norse fleiri, Avestan frāiiah- and Old Norse flestr, Avestan fraēšta-.34 Already by the time of the New Testament, however, plé :n (plé: :n) was facing competition from perissóteros, originally the comparative of perissós ‘more than usual, abundant’ (Dieu 2011: 521–523). c c c c 31 Werner (1977: 275) gives the example of German mehr, meist < Proto-Germanic *maizan-, *maista- < (post-)PIE *meh2-is-, *meh2-is-to-, but the positive is the obviously suppletive viel; cf. English much/many, more, most. A similar example is English better, best < Old English betra, betst < Proto-Germanic *bat-izan-, *bat-ista-. 32 For details, see Meiser (1998: 152–153), Weiss (2011: 359). 33 Note furthermore that the original neuter form *mag-is > magis ‘more’ was decoupled from the paradigm of magnus at an early stage, and served in Classical Latin as an adverb of comparison with verb phrases, adverbs, and occasionally adjectives; it retains this function down to the present day in Spanish más, Portuguese mais and Romanian mai. 34 On the formal problems posed by Latin plūs, superlative plūrimus, see Meiser (1998: 153–154), Dieu (2011: 556–560), Weiss (2011: 360); on Proto-Germanic *maizan-, *maista- and *flaizan-, *flaista-, see Dieu (2011: 549–555). 364 Slovo a slovesnost, 77, 2016 This trend continued over the following centuries, as sound changes led to increasing formal divergence between polús and plé :n (plé: :n): the latter became *pléon (*plíon) by loss of contrastive vowel length and underwent synizesis to *pljón, then loss of final nasal to pljó (spelled plió)35 and cluster simplification to pjó (spelled pió). c c (10) Ancient Greek polús ‘much, many’, comparative plé: :n > medieval polýs, pljó(n) c In present-day Greek, pió is restricted to the formation of analytic comparatives, e.g. pió klasikós ‘more classical’,36 while the comparative of polís ‘much, many’ remains the suppletive perisóteros, alongside the less common variants pió polís and pióteros. 4. Discussion and wider theoretical questions The examples adduced in the preceding section, to which many others could no doubt be added, are important in several respects for understanding the phenomenon of suppletion. First of all, they dispel once and for all the still widely repeated notion that (strong) suppletion necessarily involves etymologically unrelated stems. Since the stem alternations discussed above have long been known outside of specialist circles, I can only infer that there remains a synchronically unjustified tendency to treat such cases as less “genuinely” suppletive than those which have no known (or imaginable) common source, i.e. “pseudo-suppletive” in the terminology of Rudes (1980: 660–663) Although there is no question that suppletion is a relatively infrequent morphological phenomenon,37 its crosslinguistic rarity has been somewhat exaggerated by this tacit assumption that the elements participating in a suppletive relationship must be maximally dissimilar in phonetic content, such that they could never have arisen, in whole or in part, by the operation of sound change. As the examples above make clear, this impressionistic criterion is false in any case: it is entirely possible for phonological developments to give rise to stems sharing no common segments, e.g. Ancient Greek m. hen- ~ f. mia- ‘one’ or Irish dialectal bean [bjæn] ~ pl. mná [mr :] ‘woman’. The point is not that suppletion always, or even usually, develops in this fashion – Dressler (1985b: 105) is probably right to call such examples “rare” – but rather that even “strong” suppletion can evolve over time out of morphological regularity with a minimum of analogy or other irregular change.38 In fact, the kind of suppletion described here is of the greatest importance for the study of language change in general, for the following reason. It is a truism of his- 35 Attested in Greek Renaissance literature, e.g. the Erotokritos (Tonnet 2003: 183, 189). 36 Cf. fn. 33 on the reflexes of Latin magis. 37 At least under most definitions; for Mel’čuk, who includes allomorphy of inflectional morphemes, it would in fact be fairly widespread. 38 As one researcher puts it, “[w]ith suppletion induced by sound change the difference between weak and strong outcomes is essentially one of degree in that over long periods of time, series of sound changes (it does not appear to matter when it is regular or ‘irregular’) can result in a chain of development from a regular paradigm to a weakly suppletive one and ultimately to a strongly suppletive one” (Juge 2000: 186). Cf. also Dressler (1985a: 104), Werner (1987: 597, 602) and Hoenigswald’s remarks on suppletion as “the last remnant of extreme sound change”, cited above in fn. 2. Slovo a slovesnost, 77, 2016 365 torical linguistics that analogical change, by which forms are altered to more closely resemble morphologically related items, acts to undo the cumulative effects of sound change, which would otherwise tend to obscure the relationship among inflectionally or derivationally related forms over time; in the classic formulation of Sturtevant (1947: 109; cf. Campbell 2013: 96), “Phonetic laws are regular but produce irregularities. Analogic creation is irregular but produces regularity.”39 What is not yet sufficiently appreciated is that if paradigmatic analogy (i.e. leveling) does not occur, the logical end point of sound change acting over long periods in inflecting languages will in fact be suppletion, as in the examples in section 3. Paradigms such as Ancient Greek hen~ mia- ‘one’, Irish [bjæn] ~ [mr :] ‘woman’, or Czech je(st) ~ jsou simply illustrate the power of regular, conditioned sound change to give rise to drastically divergent outcomes of forms which once stood in a perfectly productive morphological relationship. It follows that any diachronic typology of suppletion must distinguish at a minimum between paradigms composed of historically unrelated stems, and those whose stems can be demonstrated or plausibly argued to have diverged as a result of the cumulative effects of phonological change (cf. Mel’čuk 2000: 520; 2006: 454–455).40 English go ~ went and Czech jde ~ šel are examples of the first type, while think ~ thought and je(st) ~ jsou are examples of the second – though it cannot be overemphasized that such cases are by definition usually identifiable only from records of earlier stages of the language and/or comparison with related languages, a point to which we shall return below. The number of the second type goes up considerably if one includes irregular weakenings, contractions, and other sound changes originating in fast speech and/or informal styles, and often observed for ‘be’, ‘have’, ‘do’, ‘say’, and other high-frequency verbs. From a synchronic perspective, then, one may classify as suppletive all isolated alternations which have arisen through the “blind” operation of sound change within a language, even if those alternations involve only parts of stems. Such cases are commonplace in the inflectional morphology of IE languages: one need only think of examples of verbal inflection from Romance languages such as French pres. 3sg. boit [bwa] ~ 1pl. buv-ons [byv-] ‘drink(s)’, veut [vø] ~ voul-ons [vul-] ‘want(s)’ or Spanish pres. 1sg. hago ~ 3sg. hace ‘do(es), make(s)’, 1sg. quepo [kep-o] ~ 3sg. cabe [kaß-e] ‘fit(s)’.41 The English present/preterite pairs in (4) obviously belong to this category 39 I have long used the imagery of a “tug-of-war” when describing to my students the diachronic relationship between sound change and analogical leveling. 40 I will not elaborate here on the paradigmatic and semantic processes by which stems originally belonging to separate lexemes come to be associated in a single paradigm, though it seems clear that “pseudosuppletion” of the kind defined by Mel’čuk (2000: 520–521; 2006: 458–460) must be a frequent, if not the most frequent path of change. For interesting case studies of inflectional suppletion in adjectives, see Börjars & Vincent (2011) on the evolution of ‘small’ in the mainland Scandinavian languages, and Maiden (2014) on two examples from Megleno-Romanian. 41 Cf. also Romanian ia- ~ lu- ‘take’, cited above in section 3. The reader will observe that I disagree on this point with Mel’čuk (2000: 514–515; 2006: 419), who seems to imply that the formal relation between suppletive morphs must be “absolutely unique” (i.e. that suppletion is “an absolutely irregular alternation”; Mel’čuk 2006: 440), yet balks at considering [vø] ~ [vul-] on a par with [al-] ~ [i-] ‘go’. 366 Slovo a slovesnost, 77, 2016 as well, as do the present paradigms of ‘be’ discussed under (6). To dismiss such stem relationships as simply “irregular” and not truly suppletive because e.g. both [bwa] and [byv-] begin with the same consonant seems quite arbitrary, yet such has long been the accepted practice of descriptive grammars and much of modern morphological theory.42 Expanding the definition of suppletion to admit such “irregular” stem alternations in turn opens up important questions for the typology of morphological change. For instance, do suppletive relations arise by sound change more often in certain morphosemantic categories? Put another way, do some morphosemantic categories “tolerate” phonologically derived suppletion more than others? And if so, why? We are still far from being able to provide more than a preliminary answer to these questions, though the compilation of an online database by the Surrey Morphology Group constitutes a welcome and necessary first step in this direction.43 Within IndoEuropean, it is noteworthy that suppletion which has developed by sound change seems to be quite rare for the category of number in nominal stems; the examples in (2) and (3) are exceptional in this regard.44 This is not necessarily so in other languages, however: Corbett (2007: 34) adduces the suppletive forms bič’ni ‘corner of a bag, tassel’ ~ pl. boždo and biqI’ni ‘pier of a bridge’ ~ pl. boRIdo from the Northeast Caucasian language Archi, observing that they “arose from sound changes that are no longer productive” (34 fn. 45). The situation is somewhat different for relations among tense and aspect stems of verbs in IE languages. Indeed, if one expands the definition of suppletion to include inflectional morphemes (including both segmental morphemes and processes), examples of suppletion are easy to identify; cf. the person-number markers for presents, aorists, and perfects in Sanskrit or Ancient Greek, or the various ways of forming aorist stems in those languages. Still, cases of highly suppletive stems which have arisen largely or entirely by sound change are rare in the historical record. Modern Greek offers over a dozen examples of fossilized present-aorist pairs which continue Ancient Greek stem patterns, e.g. méno ‘stay’ ~ é-min-a, fév -o ‘leave, go away’ ~ é-fi -a, and the already mentioned v éno ‘go out’ ~ v íka, but none of these is as robustly suppletive as e.g. lé( )o ‘say’ ~ ípa or vlépo ‘see’ ~ í∂–a, both of which go back to suppletive paradigms in the ancient language. The same may be said for stem formation in Slavic languages, where the cumulative effects of sound change have not for the most part produced outcomes more opaque than Russian vz’a-t’ ‘take’ ~ pres. 1sg. voz’m-ú or 42 Interestingly, Haspelmath & Sims (2010: 25; cf. section 1 above) classify English buy ~ bought, etc. as “weakly suppletive”, but in a later discussion of relics of Verner’s Law alternations in English and German, they admit that “[f]or contemporary speakers, the relation between war ‘was’ and gewesen ‘been’ is probably as suppletive and non-systematic as the relation betwen bin ‘am’ and war ‘was’” (218). 43 See <http://www.smg.surrey.ac.uk> and especially the Surrey Suppletion Database at <http://www. smg.surrey.ac.uk/suppletion/>. 44 Old English root nouns such as bōc [bo:k] ‘book’, plural bēc. [be: ] or āc [ :k] ‘oak’, plural ǣc. [æ: ] could be added to this type, though the otherwise well paralleled alternations of back and front vowels (“umlaut”) and of [k] and [ ] do justify grouping them together with e.g. mann ‘man’, pl. menn. (Note that over 20 nouns of this inflectional type are attested; cf. Brunner 1965: 226–229.) Slovo a slovesnost, 77, 2016 367 Polish umrze-ć ‘die’ ~ pret. umar-ł.45 Whatever one’s analysis of these verbs, they are clearly not (yet) on the same level as aspectual pairs of the type of Russian govorít’ ~ skazát’ ‘say’ or Czech brát ~ vzít ‘take’. Future research promises to reveal much more about the diachronic patterning of suppletion, and the extent to which suppletion does and does not arise through the cumulative operation of sound change. Here as so often, the IE languages with their relatively long historical records will play a central role. Even without knowledge of the diachronic background, one could reasonably assume that English think ~ thought or French bois [bwa] ~ buv-ons [byv-] descend from once unitary stems. But it would hardly be conceivable to identify Greek m. én-as ~ f. mjá ‘one’, Irish bean [bjæn] ~ pl. mná [mr :] ‘woman’, or Czech je(st) ‘is’ ~ jsou ‘are’ as (almost) entirely regular phonological reflexes of formerly nonsuppletive, regular prehistoric paradigms without the evidence of older stages and comparison with related languages. For languages without written records, and for language isolates, one should therefore reckon with the likelihood that some instances of “strong” suppletion have arisen in a similar fashion: through the cumulative operation of sound change across generations, unimpeded by analogical alteration, until the synchronic relationship between stems could no longer be captured in any but lexical terms.46 REFERENCES BOBALJIK, J. D. (2012): Universals in Comparative Morphology: Suppletion, Superlatives, and the Structure of Words. Cambridge, MA, London: The MIT Press. BÖRJARS, K. & VINCENT, N. (2011): The preconditions for suppletion. In: G. Tsoulas, A. Galani & G. Hicks (eds.), Morphology and its Interfaces. Amsterdam, Philadelphia: John Benjamins, 239–267. BOYÉ, G. (2006): Suppletion. In: K. Brown (ed.), Encyclopedia of Language and Linguistics, 12. Amsterdam: Elsevier, 297–299. BRUNNER, K. (1965): Altenglische Grammatik nach der Angelsächsischen Grammatik von Eduard Sievers. Tübingen: Niemeyer. CAMPBELL, L. (2013): Historical Linguistics: An Introduction. Edinburgh: Edinburgh University Press. CARSTAIRS, A. (1988): Some implications of phonologically conditioned suppletion. In: G. Booij & J. van Marle (eds.), Yearbook of Morphology, 1. Dordrecht: Kluwer, 67–94. CORBETT, G. G. (2007): Canonical typology, suppletion, and possible words. Language 83 (1), 8–42. DIEU, É. (2011): Le supplétisme dans les formes de gradation en grec ancien et dans les langues indoeuropéennes. Genève: Droz. 45 Rudes (1980: 661) cites Polish cią-ć ‘cut’ ~ pres. 1sg. tn-ę as an example of suppletion developing by sound change alone, but the allomorphy C’ęą- ~ Cn- is also known from pią-ć (się) ~ pn-ę (się) ‘ascend’ and za-cząć ~ za-czn-ę ‘begin’. For an analogous example from Czech, cf. hnát ‘chase’ ~ žen-u, which given the well-established status of the alternation h ~ ž (< Proto-Slavic *g ~ *ž) is readily identifiable as belonging to the same small class as brát ‘take’ ~ ber-u. 46 There is also the reverse danger, that stems that partly resemble one another will be mistaken for irregularities arising from the operation of more or less recent sound changes, rather than instances of “true” suppletion; cf. Modern Greek pres. érxome ‘come’ ~ aor. írθ a, mentioned above in section 1. In practice, however, it seems far more likely that descriptive grammars will underestimate rather than overestimate the possibility of suppletion arising by sound change. 368 Slovo a slovesnost, 77, 2016 DRESSLER, W. U. (1985a): Inflectional suppletion in Natural Morphology. In: B. F. Elson (ed.), Language in Global Perspective: Papers in Honor of the 50th Anniversary of the Summer Institute of Linguistics 1935–1985. Dallas: Summer Institute of Linguistics, 97–111. DRESSLER, W. U. (1985b): Suppletion in word formation. In: J. Fisiak (ed.): Historical Semantics – Historical Word-Formation. Berlin: Mouton de Gruyter, 97–112. FORTSON, B. W., IV. (2009): Indo-European Language and Culture. Malden, Oxford, Chichester: Wiley-Blackwell. GIPPERT, J. (2004): Ein Problem der indogermanischen Pronominalflexion. In: A. Hyllested, A. R. Jørgensen, J. H. Larsson & T. Olander (eds.), Per Aspera ad Asteriscos: Studia Indogermanica in honorem Jens Elmegård Rasmussen sexagenarii Idibus Martiis anno MMIV. Innsbruck: Institut für Sprachen und Literaturen der Universität Innsbruck, 155–165. HASPELMATH, M. & SIMS, A. D. (2010): Understanding Morphology. London: Hodder Education. HOENIGSWALD, H. (1960): Language Change and Linguistic Reconstruction. Chicago, London: The University of Chicago Press. JASANOFF, J. H. (1986): Old Irish tair ‘come!’. Transactions of the Philological Society 84 (1), 132–141. JUGE, M. L. (2000): On the rise of suppletion in verbal paradigms. In: S. C. Chang, L. Liaw & J. Ruppenhofer (eds.), Proceedings of the Twenty-Fifth Annual Meeting of the Berkeley Linguistics Society, 12–15 February 1999. Berkeley: Berkeley Linguistics Society, 183–194. KLOEKHORST, A. (2008): Etymological Dictionary of the Hittite Inherited Lexicon. Leiden, Boston: Brill. MAIDEN, M. (2014): Two suppletive adjectives in Megleno-Romanian. Revue Romane 49 (1), 32–52. MAIDEN, M. & ROBUSTELLI, C. (2007): A Reference Grammar of Modern Italian. Abingdon, New York: Routledge. MARKEY, T. (1985): On suppletion. Diachronica 2, 51–66. MCCONE, K. (1997): The Early Irish Verb. Maynooth: An Sagart. MEISER, G. (1998): Historische Laut- und Formenlehre der lateinischen Sprache. Darmstadt: Wissenschaftliche Buchgesellschaft. MEISER, G. (2003): Veni Vidi Vici: Die Vorgeschichte des lateinischen Perfektsystems. München: C. H. Beck. MEL’ČUK, I. (2000): Suppletion. In: G. Booij, Ch. Lehmann, J. Mugdan, W. Kesselheim & S. Skopeteas (eds.), Morphologie: Ein internationales Handbuch zur Flexion und Wortbildung / Morphology: An International Handbook on Inflection and Word-Formation, 1. Berlin, New York: Walter de Gruyter, 510–522. MEL’ČUK, I. (2006): Aspects of the Theory of Morphology. Berlin, New York: Mouton de Gruyter. OSTHOFF, H. (1899): Vom Suppletivwesen der indogermanischen Sprachen. Heidelberg: Hörning. PLANK, F. (forthcoming): Vom Suppletiv(un)wesen, in Beziehung zu Paradigmenstruktur und in besonderer Rücksicht der historischen Natur beschränkter Möglichkeiten. In: A. Bittner & K.-M. Köpcke (eds.), Regularität und Irregularität. Available at: <http://ling.uni-konstanz.de/pages/home/plank/ for_download/publications/Plank_Suppletiv_forthcoming.pdf>. RINGE, D. (2006): A Linguistic History of English, 1: From Proto-Indo-European to Proto-Germanic. Oxford: Oxford University Press. RINGE, D. & TAYLOR, A. (2014): A Linguistic History of English, 2: The Development of Old English. Oxford: Oxford University Press. RONNEBERGER-SIBOLD, E. (1988): Entstehung von Suppletion und Natürliche Morphologie. Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 41 (4), 453–462. RUDES, B. A. (1980): On the nature of verbal suppletion. Linguistics 18, 655–676. ‚ SCHMIDT, J. (1900): Das Zahlwort µíα, ı′α. Zeitschrift für Vergleichende Sprachforschung 36 (3), 391–399. SPENCER, A. (1991): Morphological Theory. Oxford, Cambridge: Blackwell. STURTEVANT, E. (1947): An Introduction to Linguistic Science. New Haven: Yale University Press. TONNET, H. (2003): Historie du grec moderne : La formation d’une langue. Paris: Langues & Mondes – L’Asiathèque. Slovo a slovesnost, 77, 2016 369 VENNEMANN, T. (1972): Rule inversion. Lingua 29, 209–242. VESELINOVA, L. N. (2006): Suppletion in Verb Paradigms: Bits and Pieces of the Puzzle. Amsterdam, Philadelphia: John Benjamins. WEISS, M. (2011): Outline of the Historical and Comparative Grammar of Latin. Ann Arbor: Beech Stave. WERNER, O. (1977): Suppletivwesen durch Lautwandel. In: G. Drachman (ed.), Akten der 2. Salzburger Frühlingstagung für Linguistik. Tübingen: Narr, 269–283. WERNER, O. (1987): The aim of morphological change is a good mixture – not a uniform language type. In: A. Giacalone Ramat, O. Carruba & G. Bernini (eds.), Papers from the 7th International Conference on Historical Linguistics. Amsterdam, Philadelphia: John Benjamins, 591–606. Institute of Comparative Linguistics, Faculty of Arts, Charles University nám. Jana Palacha 2, 116 38 Prague 1, Czech Republic Department of Older Germanic Languages, Faculty of English, Adam Mickiewicz University in Poznań al. Niepodległości 4, 61-874 Poznań, Poland <ronald.kim@yahoo.com> 370 Slovo a slovesnost, 77, 2016