Papers by Marius Zemp
Comparative Dictionary of Tibetan Dialects (CDTD), 2018
Notes from the field on perspective-indexing constructions
This paper focuses on a specific type of perspective-indexing constructions in Tibetic and neighb... more This paper focuses on a specific type of perspective-indexing constructions in Tibetic and neighboring languages, namely a type of verbal marker that is consistently construed from the perspective of the speaker in statements, the addressee in questions, and the source (= the original/reported speaker) in reported speech clauses. As these markers indicate how one obtained the information profiled in a sentence and may thus be viewed as a type of evidential, they cannot at the same time establish reference to any participant of the current speech act and thus by default reflect the perspective of the ‘informant’ of the respective sentence type. If we define the encountered distinctions in relation to a cause-effect vector in the sense of DeLancey (1986), these languages all contain what we may call an ‘insider’ marker indicating access to the entire vector including its causal origin and an ‘outsider’ marker indicating access only to its effect end. Whereas the insider markers typica...
Open Linguistics
The main purpose of this paper is to elucidate a special type of egophoric markers found in Purik... more The main purpose of this paper is to elucidate a special type of egophoric markers found in Purik and other varieties of Tibetan. These factual evidential markers, deriving from the Written Tibetan existential copula yod, are regularly used in Purik to profile not only events in which the informant participates, but also events which the informant is in the position to describe as facts even if she does not directly participate in them. The factual function of yod is argued here to reflect the indicative function yod served when it was the only existential copula at a stage of the language in which no evidential functions had grammaticalized yet. A comparison of the evidential inventory of Purik with those of other well documented Tibetan varieties reveals that it was in resultative constructions that yod first became contrasted by ’dug *‘was there’, facilitating the reanalysis of two evidentially opposed existential copulas. Hence, the factual meaning of yod formed in contrast to t...
A Grammar of Purik Tibetan
Revue d'Etudes Tibétaines 60, Août 2021 – New Research on Old Tibetan Studies – Proceedings of the Panel Old Tibetan Studies VI – IATS 2019, 2021
The Purik member of the Tibetic language family is spoken in the western periphery of the Tibetic... more The Purik member of the Tibetic language family is spoken in the western periphery of the Tibetic linguistic area. In Purik, two demonstratives, de ‘that’ and e ‘the other’, occur not only pre- and pronominally, but also post- and proverbally, in which case they take scope over the sentence they terminate. The proverbal de, occurring instead of an existential predicate, locates an entity or property in the topical situation (which typically corresponds to the interlocutors’ current one). The postverbal de, occurring after a full-fledged sentence, has the effect of laying out the information conveyed by this sentence, inviting the addressee to retrace it, and implying that it should be clear. By contrast, pro- and postverbal e points to information that requires a shift of attention.
The present paper demonstrates that Old Tibetan (OT) ga re ‘where is (X)?’, clause-linking (s)te ~ de, and V-ta re ‘lest (it) will V’, and other phenomena found in written and spoken Tibetic varieties, are best understood if analysed as traces of the mentioned clause-final demonstratives. The comparative study of spoken Tibetic varieties thus not only contributes to our understanding of particular OT texts, but also sheds light on the development and dispersion of Tibetic during the Imperial Period (7th–9th centuries CE).
This paper focuses on a specific type of perspective-indexing constructions in Tibetic and neighb... more This paper focuses on a specific type of perspective-indexing constructions in Tibetic and neighboring languages, namely a type of verbal marker that is consistently construed from the perspective of the speaker in statements, the addressee in questions, and the source (= the original/reported speaker) in reported speech clauses. As these markers indicate how one obtained the information profiled in a sentence and may thus be viewed as a type of evidential, they cannot at the same time establish reference to any participant of the current speech act and thus by default reflect the perspective of the 'informant' of the respective sentence type. If we define the encountered distinctions in relation to a cause-effect vector in the sense of DeLancey (1986), these languages all contain what we may call an 'insider' marker indicating access to the entire vector including its causal origin and an 'outsider' marker indicating access only to its effect end. Whereas the insider markers typically occur when the informant is the subject and the outsider markers when s/he is not, the present paper discusses the different ways in which Tibetic and neighboring languages deviate from this basic pattern, and argues that these differences reflect the fact that the markers in the latter languages were only secondarily evidentialized in reported speech clauses, likely due to contact with Tibetic.
The main purpose of this paper is to elucidate a special type of egophoric markers found in Purik... more The main purpose of this paper is to elucidate a special type of egophoric markers found in Purik and other varieties of Tibetan. These factual evidential markers, deriving from the Written Tibetan existential copula yod, are regularly used in Purik to profile not only events in which the informant participates, but also events which the informant is in the position to describe as facts even if she does not directly participate in them. The factual function of yod is argued here to reflect the indicative function yod served when it was the only existential copula at a stage of the language in which no evidential functions had grammaticalized yet. A comparison of the evidential inventory of Purik with those of other well documented Tibetan varieties reveals that it was in resultative constructions that yod first became contrasted by 'dug *'was there', facilitating the reanalysis of two evidentially opposed existential copulas. Hence, the factual meaning of yod formed in contrast to testimonial 'dug. The evolution of the factual yod is traced from its first emergence up to its restriction to egophoric contexts in Central Tibetan, and compared with that of egophoric markers in West Himalayish Bunan.
Egophoricity is a cross-linguistically rare grammatical phenomenon. While numerous descriptive st... more Egophoricity is a cross-linguistically rare grammatical phenomenon. While numerous descriptive studies have substantially improved our synchronic understanding of the category in recent years, we are still largely ignorant of the diachronic origins of egophoricity systems. In this article, we address this gap and discuss a diachronic process that transforms person agreement markers into egophoricity markers. Based on evidence from three Tibeto-Burman languages, we reconstruct the diachronic transformation and argue that the process starts out in reported speech clauses once the direct construal of the predicate is generalized. This generalization allows for the functional reanalysis of first and third person markers as egophoric and allophoric markers, while second person markers become functionally obsolete. Once person markers have undergone an epistemization in reported speech clauses, the innovative epistemic system is extended to simple declarative and interrogative clauses, where it gradually replaces the conservative person agreement system.
This article was published in a minimally differing version in Historical Development of the Tibe... more This article was published in a minimally differing version in Historical Development of the Tibetan Languages, Proceedings of the Workshop B of the 17th Himalayan Languages Symposium in Kobe, 6th-9th September 2011 (Journal of Research Institute Vol. 49, 2012), edited by Tsuguhito Takeuchi and Norihiko Hayashi (Research Institute of Foreign Studies, Kobe University of Foreign Studies).
Talks by Marius Zemp
Conference Presentations by Marius Zemp
This paper, presented in the "egophoricity" panel at the SLE in Helsinki, discusses my preliminar... more This paper, presented in the "egophoricity" panel at the SLE in Helsinki, discusses my preliminary diachronic account of the evidential contrasts found in Kutang (see Donohue & Gautam 2019) but not the closely related varieties of Northern and Southern Ghale (Paudel 2008).
This paper briefly retraces what we know about the Tibetic tonogenesis, identifies an important r... more This paper briefly retraces what we know about the Tibetic tonogenesis, identifies an important remaining gap in our diachronic account, and attempts to bridge that gap.
The main development of the tonal Tibetic varieties, that is, those in which tone became phonemically distinctive, is well understood (e.g. Sun 2003) and borne out by crosslinguistic phonetic studies such as Hombert et al. (1979): vowels preceded by an originally voiced onset (Ga) have developed a low tone (Ka̠), whereas vowels preceded by a voiceless onset (Ka) have developed a high tone (Kā).
Another feature found throughout the tonal Tibetic varieties is, however, that the second syllable of disyllabic nouns is always, even if it originally began with a voiced obstruent, higher than the first syllable, that is, we find LH (where L = low and H = high) and HH, but neither LL nor HL. This conundrum was solved by Caplow (2009), whose robust statistical analysis of the acoustic correlates of stress in two non-tonal varieties in the opposite peripheries of the Tibetic language continuum shows that disyllabic non-verbs are consistently stressed there on the second syllable, and that this stress is primarily conveyed by fundamental frequency. Assuming that this was already the case in Proto-Tibetan (PT) explains the absence of LL and HL in the tonal varieties (Caplow 2009: 541–2). Evidence from another non-tonal variety will be presented to demonstrate how pervasively the LH-pattern marked disyllabic non-verbs in PT.
This allows us to identify the pivotal question of how this syntagmatically distinctive function of fundamental frequency (as manifested in disyllabic non-verbs) developed into the paradigmatically distinctive tonal contrast (affecting also monosyllabic words). Here, another widely recognized process (e.g. Sun 2003: 38) may have played a crucial role: whereas syllables beginning with unprefixed nasals or liquids (Na) developed a low tone (Na̠), those beginning with prefixed nasals or liquids (x-Na) developed a high tone (Nā). Given the extreme productivity of the LH-pattern in PT, we may hypothesize that it was applied not only to disyllabic but, at some point, also to sesquisyllabic non-verbs such as gnam [xnam] ‘sky’ and gla [xla] ‘salary’, yielding x̱nām and x̱lā. When prefixes were later dropped in these dialects, as happened in all of Tibet and Kham, the only feature distinguishing these words from corresponding words without a prefix, such as nam ‘when’ and la ‘(mountain) pass’, was their tone. As voiceless nasals are quite common in eastern dialects, however, the x- prefix may have also devoiced the following nasal there (leading to xn̥am and xl̥a), and then the vowel following a voiceless nasal came to be pronounced with a higher pitch than that following a voiced nasal.
Caplow, Nancy. 2009. The role of stress in Tibetan tonogenesis: a study in historical comparative acoustics. Dissertation at the University of California in Santa Barbara.
Hombert, Jean-Marie, Ohala, John, and Ewan, William. 1979. Phonetic Explanations for the Development of Tones. Language 55 (1): 37–58.
Sun, Jackson. 2003. Variegated tonal developments in Tibetan. In David Bradley, Randy LaPolla, Boyd Michailovsky and Graham Thurgood (eds.), Language variation: papers on variation and change in the Sinosphere and in the Indosphere in honour of James A. Matisoff, 35–51. Canberra: Pacific Linguistics.
This paper briefly retraces what we know about the Tibetic tonogenesis, identifies an important r... more This paper briefly retraces what we know about the Tibetic tonogenesis, identifies an important remaining gap in our diachronic account, and attempts to bridge that gap.
The main development of the tonal Tibetic varieties, that is, those in which tone became phonemically distinctive, is well understood (e.g. Sun 2003) and borne out by crosslinguistic phonetic studies such as Hombert et al. (1979): vowels preceded by an originally voiced onset (Ga) have developed a low tone (Ka̠), whereas vowels preceded by a voiceless onset (Ka) have developed a high tone (Kā).
Another feature found throughout the tonal Tibetic varieties is, however, that the second syllable of disyllabic nouns is always, even if it originally began with a voiced obstruent, higher than the first syllable, that is, we find LH (where L = low and H = high) and HH, but neither LL nor HL. This conundrum was solved by Caplow (2009), whose robust statistical analysis of the acoustic correlates of stress in two non-tonal varieties in the opposite peripheries of the Tibetic language continuum shows that disyllabic non-verbs are consistently stressed there on the second syllable, and that this stress is primarily conveyed by fundamental frequency. Assuming that this was already the case in Proto-Tibetan (PT) explains the absence of LL and HL in the tonal varieties (Caplow 2009: 541–2). Evidence from another non-tonal variety will be presented to demonstrate how pervasively the LH-pattern marked disyllabic non-verbs in PT.
This allows us to identify the pivotal question of how this syntagmatically distinctive function of fundamental frequency (as manifested in disyllabic non-verbs) developed into the paradigmatically distinctive tonal contrast (affecting also monosyllabic words). Here, another widely recognized process (e.g. Sun 2003: 38) may have played a crucial role: whereas syllables beginning with unprefixed nasals or liquids (Na) developed a low tone (Na̠), those beginning with prefixed nasals or liquids (x-Na) developed a high tone (Nā). Given the extreme productivity of the LH-pattern in PT, we may hypothesize that it was applied not only to disyllabic but, at some point, also to sesquisyllabic non-verbs such as gnam [xnam] ‘sky’ and gla [xla] ‘salary’, yielding x̱nām and x̱lā. When prefixes were later dropped in these dialects, as happened in all of Tibet and Kham, the only feature distinguishing these words from corresponding words without a prefix, such as nam ‘when’ and la ‘(mountain) pass’, was their tone. As voiceless nasals are quite common in eastern dialects, however, the x- prefix may have also devoiced the following nasal there (leading to xn̥am and xl̥a), and as a consequence, the vowels following such voiceless nasals came to be pronounced with a higher pitch than those following voiced nasals.
Caplow, Nancy. 2009. The role of stress in Tibetan tonogenesis: a study in historical comparative acoustics. Dissertation at the University of California in Santa Barbara.
Hombert, Jean-Marie, Ohala, John, and Ewan, William. 1979. Phonetic Explanations for the Development of Tones. Language 55 (1): 37–58.
Sun, Jackson. 2003. Variegated tonal developments in Tibetan. In David Bradley, Randy LaPolla, Boyd Michailovsky and Graham Thurgood (eds.), Language variation: papers on variation and change in the Sinosphere and in the Indosphere in honour of James A. Matisoff, 35–51. Canberra: Pacific Linguistics.
Evidentiality has been shown to regularly grammaticalize in equipollent contrasts in the Himalaya... more Evidentiality has been shown to regularly grammaticalize in equipollent contrasts in the Himalayan region (Zemp 2020). In Amdo Tibetan, for example, *V-tha(l) ‘went past V-ing’ (employing the simple past of tha(l) ‘go past’) became contrasted with V-s-’dug ‘was there, having undergone V’ (employing the simple past of ’dug ‘stay, be there’ after a resultative verb form V-s). Through this contrast the two constructions became conventionalized as direct and indirect evidentials, that is, they respectively came to indicate whether a past event was directly witnessed or inferred from circumstantial evidence, see examples (1) and (2) below.
In western dialects of Tibetan, ’dug ‘was there’ became contrasted with the existential copula yod ‘is there’. While ’dug thereby came to mean that a present state was directly witnessed, yod came to mean that the speaker simply knows that state. In Ladakhi Tibetan, rag ‘was felt’ developed into a third existential evidential meaning that a current state was directly witnessed non-visually. Direct evidential ’dug thereby became restricted to contexts in which something was seen.
Equipollent evidential contrasts are also found in other regions of the world, such as the Caucasus (Khalilova 2011), the New Guinea Highlands (San Roque & Loughnane 2012), and the Americas (e.g. Cherokee in North America, see Aikhenvald 2004: 26–7 and Pulte 1985). Jarawara (a language of the southern Amazone, Dixon 2003) has contrasting sets of endings indicating direct and indirect evidence (and agreeing with the subject for person, number, and gender) for the immediate and the recent past tense. Tuyuca (an unrelated Eastern Tucanoan language of the northern Amazone, Barnes 1984), in addition to sets of direct and indirect endings has a set of direct non-visual endings, which appear to have restricted the other direct endings to visual evidence (including events which the speaker performed herself).
Note that all the discussed evidentials have a neatly defined evidential value (i.e. they indicate how the speaker knows the information conveyed in a statement) which is consistent across the contexts in which they occur. The shared tense-aspect values of contrasting evidentials point to the contexts in which their contrasting implications grammaticalized, that is, in which the evidentials became defined against each other.
In stark contrast to this evidence, it has been common practice among linguists working on evidentiality for the past fifty years to attribute a ‘visual evidential’ value to unmarked verb forms in languages in which there are inferential and/or reportative constructions containing additional morphemes. For Desano (another Eastern Tucanoan language), e.g., Kaye (1970) infers that the unmarked verb form in (4) implies “direct observation” from the fact that it lacks the -jo- morpheme which indicates that a past event is inferred in (3), and because (4) may indeed occur in contexts in which the speaker saw the event referred to (beware that ‘E’ in (3) stands for evidential, but has no phonetic substance, even if it stands in the transcription line).
For the majority of languages Aikhenvald (2004: 72ff.) adduces in her discussion of this issue, the ‘visual analysis’ of an unmarked verb form is supported by one example sentence at the most. Consulting the grammatical descriptions of the respective languages, we may thus identify the same recurrent argument reversal: if an unmarked verb form occurs in a context in which the event referred to was likely or could have been directly witnessed, it is assigned a ‘visual evidential’ value. Languages such as Tariana and Hup (Epps 2005), on the other hand, where the unmarked verb form is documented in various contexts, helped establishing the idea that it is common for ‘visual evidentials’ to refer to “generally known facts” or to events for which the speaker “takes full responsibility” (Aikhenvald 2004: 191), but that they rarely actually signal that an event was observed.
This paper proposes to analyze as evidentials only verb forms which consistently convey a specific evidential meaning, to stop attributing evidential values solely based on cross-linguistic parallels, and to re-establish the zero-hypothesis that a construction which is formally unmarked in terms of evidentiality is also functionally unmarked.
Amdo Tibetan (Sun 1993: 950):
(1) ʈʂaɕʰi=kə ʰtæ ɲu=tʰæ
Bkra-shis=ERG horse buy=DIREV
‘(I saw that) Tashi bought a horse.’
(2) ʈʂaɕʰi=kə ʰtæ ɲu=zəg
Bkra-shis=ERG horse buy=INDIREV
‘(It appears that) Tashi bought a horse.’
Desano (Kaye 1970: 34):
(3) igyN widi + á + waa + jo + biN
he leave ~pres.* happen ev. p.e.
‘he has left (I did not see him go but he is not here now)’
(4) igyN widi + á + waa + E + biN
he leave ~pres. happen ev. p.e.
‘he has left (I saw him and he is still away)’
*According to footnote 1 in Kaye (1970: 31), ‘p. e.’ stands for “any personal ending”, and ‘~pres.’ for “non-present tense”.
References:
Aikhenvald, Alexandra. 2004. Evidentiality. Oxford: Oxford University Press.
Barnes, Janet. 1984. Evidentials in the Tuyuca Verb. International Journal of American Linguistics 50: 255–71.
Dixon, R. M. W. 2003. Evidentiality in Jarawara. In Alexandra Aikhenvald and R. M. W. Dixon (eds.), Studies in Evidentiality, pp. 165–88. Amsterdam: John Benjamins.
Epps, Patience. 2005. Areal diffusion and the development of evidentiality: Evidence from Hup. Studies in Language 29: 617–50.
Kaye, Jonathan. 1970. The Desano Verb: Problems in Semantics, Syntax and Phonology. Ann Arbor: UMI.
Khalilova, Zaira. 2011. Evidentiality in Tsezic languages. Linguistic Discovery 9(2): 30–48.
Pulte, William. 1985. The experienced and non-experienced past in Cherokee. International Journal of American Linguistics 51: 543–4.
San Roque, Lila, and Robyn Loughnane. 2012. The New Guinea Highlands evidentiality area. Linguistic Typology 16(1): 111–67.
Sun, Jackson T.-S. 1993. Evidentials in Amdo Tibetan. Bulletin of the Institute of History and Philology, Academia Sinica 63(4): 143–88.
Zemp, Marius. 2020. Evidentials and their pivot in Tibetic and neighboring Himalayan languages. Functions of Language 27(1): 29–54.
Evidentiality has been shown to regularly grammaticalize in equipollent contrasts in the Himalaya... more Evidentiality has been shown to regularly grammaticalize in equipollent contrasts in the Himalayan region (Zemp 2020). In Amdo Tibetan, for example, *V-tha(l) ‘went past V-ing’ (employing the simple past of tha(l) ‘go past’) became contrasted with V-s-’dug ‘was there, having undergone V’ (employing the simple past of ’dug ‘stay, be there’ after a resultative verb form V-s). Through this contrast the two constructions became conventionalized as direct and indirect evidentials, that is, they respectively came to indicate whether a past event was directly witnessed or inferred from circumstantial evidence, see examples (1) and (2) below.
In western dialects of Tibetan, ’dug ‘was there’ became contrasted with the existential copula yod ‘is there’. While ’dug thereby came to mean that a present state was directly witnessed, yod came to mean that the speaker simply knows that state. In Ladakhi Tibetan, rag ‘was felt’ developed into a third existential evidential meaning that a current state was directly witnessed non-visually. Direct evidential ’dug thereby became restricted to contexts in which something was seen.
Equipollent evidential contrasts are also found in other regions of the world, such as the Caucasus (Khalilova 2011), the New Guinea Highlands (San Roque & Loughnane 2012), and the Americas (e.g. Cherokee in North America, see Aikhenvald 2004: 26–7 and Pulte 1985). Jarawara (a language of the southern Amazone, Dixon 2003) has contrasting sets of endings indicating direct and indirect evidence (and agreeing with the subject for person, number, and gender) for the immediate and the recent past tense. Tuyuca (an unrelated Eastern Tucanoan language of the northern Amazone, Barnes 1984), in addition to sets of direct and indirect endings has a set of direct non-visual endings, which appear to have restricted the other direct endings to visual evidence (including events which the speaker performed herself).
Note that all the discussed evidentials have a neatly defined evidential value (i.e. they indicate how the speaker knows the information conveyed in a statement) which is consistent across the contexts in which they occur. The shared tense-aspect values of contrasting evidentials point to the contexts in which their contrasting implications grammaticalized, that is, in which the evidentials became defined against each other.
In stark contrast to this evidence, it has been common practice among linguists working on evidentiality for the past fifty years to attribute a ‘visual evidential’ value to unmarked verb forms in languages in which there are inferential and/or reportative constructions containing additional morphemes. For Desano (another Eastern Tucanoan language), e.g., Kaye (1970) infers that the unmarked verb form in (4) implies “direct observation” from the fact that it lacks the -jo- morpheme which indicates that a past event is inferred in (3), and because (4) may indeed occur in contexts in which the speaker saw the event referred to (beware that ‘E’ in (3) stands for evidential, but has no phonetic substance, even if it stands in the transcription line).
For the majority of languages Aikhenvald (2004: 72ff.) adduces in her discussion of this issue, the ‘visual analysis’ of an unmarked verb form is supported by one example sentence at the most. Consulting the grammatical descriptions of the respective languages, we may thus identify the same recurrent argument reversal: if an unmarked verb form occurs in a context in which the event referred to was likely or could have been directly witnessed, it is assigned a ‘visual evidential’ value. Languages such as Tariana and Hup (Epps 2005), on the other hand, where the unmarked verb form is documented in various contexts, helped establishing the idea that it is common for ‘visual evidentials’ to refer to “generally known facts” or to events for which the speaker “takes full responsibility” (Aikhenvald 2004: 191), but that they rarely actually signal that an event was observed.
This paper proposes to analyze as evidentials only verb forms which consistently convey a specific evidential meaning, to stop attributing evidential values solely based on cross-linguistic parallels, and to re-establish the zero-hypothesis that a construction which is formally unmarked in terms of evidentiality is also functionally unmarked.
Amdo Tibetan (Sun 1993: 950):
(1) ʈʂaɕʰi=kə ʰtæ ɲu=tʰæ
Bkra-shis=ERG horse buy=DIREV
‘(I saw that) Tashi bought a horse.’
(2) ʈʂaɕʰi=kə ʰtæ ɲu=zəg
Bkra-shis=ERG horse buy=INDIREV
‘(It appears that) Tashi bought a horse.’
Desano (Kaye 1970: 34):
(3) igyN widi + á + waa + jo + biN
he leave ~pres.* happen ev. p.e.
‘he has left (I did not see him go but he is not here now)’
(4) igyN widi + á + waa + E + biN
he leave ~pres. happen ev. p.e.
‘he has left (I saw him and he is still away)’
*According to footnote 1 in Kaye (1970: 31), ‘p. e.’ stands for “any personal ending”, and ‘~pres.’ for “non-present tense”.
References:
Aikhenvald, Alexandra. 2004. Evidentiality. Oxford: Oxford University Press.
Barnes, Janet. 1984. Evidentials in the Tuyuca Verb. International Journal of American Linguistics 50: 255–71.
Dixon, R. M. W. 2003. Evidentiality in Jarawara. In Alexandra Aikhenvald and R. M. W. Dixon (eds.), Studies in Evidentiality, pp. 165–88. Amsterdam: John Benjamins.
Epps, Patience. 2005. Areal diffusion and the development of evidentiality: Evidence from Hup. Studies in Language 29: 617–50.
Kaye, Jonathan. 1970. The Desano Verb: Problems in Semantics, Syntax and Phonology. Ann Arbor: UMI.
Khalilova, Zaira. 2011. Evidentiality in Tsezic languages. Linguistic Discovery 9(2): 30–48.
Pulte, William. 1985. The experienced and non-experienced past in Cherokee. International Journal of American Linguistics 51: 543–4.
San Roque, Lila, and Robyn Loughnane. 2012. The New Guinea Highlands evidentiality area. Linguistic Typology 16(1): 111–67.
Sun, Jackson T.-S. 1993. Evidentials in Amdo Tibetan. Bulletin of the Institute of History and Philology, Academia Sinica 63(4): 143–88.
Zemp, Marius. 2020. Evidentials and their pivot in Tibetic and neighboring Himalayan languages. Functions of Language 27(1): 29–54.
SLE presentation, 2021
According to Willet (1988:55), evidentiality is “the linguistic means of indicating how the speak... more According to Willet (1988:55), evidentiality is “the linguistic means of indicating how the speaker obtained the information on which s/he bases an assertion”. In short, it may also be said to indicate one’s information source (Willet 1988: 51; Aikhenvald 2004: 3). Most of the Tibetic evidential distinctions much like conjunct-disjunct oppositions in other Himalayan languages seem to be informed by the speaker’s involvement in the event profiled in a statement (DeLancey 1986, 1992; Hale 1980). Both types of distinctions may be said to involve “egophoric” and “allophoric” markers, that is, markers which mainly occur when the speaker respectively is and isn’t the subject in a statement (Tournadre 1991, 2008; Floyd, Norcliffe & San Roque 2018). The present paper shows that a majority of the evidential distinctions found in the Greater Himalayan Region (GHR) contrast ego- with allophoric markers, but that the contrasting markers nowhere strictly keep to their respective domains. Instead, we may distinguish different types of distinctions based on what the markers mean in their “atypical environments.”
The original 20-minutes presentation with voiceover is found here: https://osf.io/zxe25/
Uploads
Papers by Marius Zemp
The present paper demonstrates that Old Tibetan (OT) ga re ‘where is (X)?’, clause-linking (s)te ~ de, and V-ta re ‘lest (it) will V’, and other phenomena found in written and spoken Tibetic varieties, are best understood if analysed as traces of the mentioned clause-final demonstratives. The comparative study of spoken Tibetic varieties thus not only contributes to our understanding of particular OT texts, but also sheds light on the development and dispersion of Tibetic during the Imperial Period (7th–9th centuries CE).
Talks by Marius Zemp
Conference Presentations by Marius Zemp
The main development of the tonal Tibetic varieties, that is, those in which tone became phonemically distinctive, is well understood (e.g. Sun 2003) and borne out by crosslinguistic phonetic studies such as Hombert et al. (1979): vowels preceded by an originally voiced onset (Ga) have developed a low tone (Ka̠), whereas vowels preceded by a voiceless onset (Ka) have developed a high tone (Kā).
Another feature found throughout the tonal Tibetic varieties is, however, that the second syllable of disyllabic nouns is always, even if it originally began with a voiced obstruent, higher than the first syllable, that is, we find LH (where L = low and H = high) and HH, but neither LL nor HL. This conundrum was solved by Caplow (2009), whose robust statistical analysis of the acoustic correlates of stress in two non-tonal varieties in the opposite peripheries of the Tibetic language continuum shows that disyllabic non-verbs are consistently stressed there on the second syllable, and that this stress is primarily conveyed by fundamental frequency. Assuming that this was already the case in Proto-Tibetan (PT) explains the absence of LL and HL in the tonal varieties (Caplow 2009: 541–2). Evidence from another non-tonal variety will be presented to demonstrate how pervasively the LH-pattern marked disyllabic non-verbs in PT.
This allows us to identify the pivotal question of how this syntagmatically distinctive function of fundamental frequency (as manifested in disyllabic non-verbs) developed into the paradigmatically distinctive tonal contrast (affecting also monosyllabic words). Here, another widely recognized process (e.g. Sun 2003: 38) may have played a crucial role: whereas syllables beginning with unprefixed nasals or liquids (Na) developed a low tone (Na̠), those beginning with prefixed nasals or liquids (x-Na) developed a high tone (Nā). Given the extreme productivity of the LH-pattern in PT, we may hypothesize that it was applied not only to disyllabic but, at some point, also to sesquisyllabic non-verbs such as gnam [xnam] ‘sky’ and gla [xla] ‘salary’, yielding x̱nām and x̱lā. When prefixes were later dropped in these dialects, as happened in all of Tibet and Kham, the only feature distinguishing these words from corresponding words without a prefix, such as nam ‘when’ and la ‘(mountain) pass’, was their tone. As voiceless nasals are quite common in eastern dialects, however, the x- prefix may have also devoiced the following nasal there (leading to xn̥am and xl̥a), and then the vowel following a voiceless nasal came to be pronounced with a higher pitch than that following a voiced nasal.
Caplow, Nancy. 2009. The role of stress in Tibetan tonogenesis: a study in historical comparative acoustics. Dissertation at the University of California in Santa Barbara.
Hombert, Jean-Marie, Ohala, John, and Ewan, William. 1979. Phonetic Explanations for the Development of Tones. Language 55 (1): 37–58.
Sun, Jackson. 2003. Variegated tonal developments in Tibetan. In David Bradley, Randy LaPolla, Boyd Michailovsky and Graham Thurgood (eds.), Language variation: papers on variation and change in the Sinosphere and in the Indosphere in honour of James A. Matisoff, 35–51. Canberra: Pacific Linguistics.
The main development of the tonal Tibetic varieties, that is, those in which tone became phonemically distinctive, is well understood (e.g. Sun 2003) and borne out by crosslinguistic phonetic studies such as Hombert et al. (1979): vowels preceded by an originally voiced onset (Ga) have developed a low tone (Ka̠), whereas vowels preceded by a voiceless onset (Ka) have developed a high tone (Kā).
Another feature found throughout the tonal Tibetic varieties is, however, that the second syllable of disyllabic nouns is always, even if it originally began with a voiced obstruent, higher than the first syllable, that is, we find LH (where L = low and H = high) and HH, but neither LL nor HL. This conundrum was solved by Caplow (2009), whose robust statistical analysis of the acoustic correlates of stress in two non-tonal varieties in the opposite peripheries of the Tibetic language continuum shows that disyllabic non-verbs are consistently stressed there on the second syllable, and that this stress is primarily conveyed by fundamental frequency. Assuming that this was already the case in Proto-Tibetan (PT) explains the absence of LL and HL in the tonal varieties (Caplow 2009: 541–2). Evidence from another non-tonal variety will be presented to demonstrate how pervasively the LH-pattern marked disyllabic non-verbs in PT.
This allows us to identify the pivotal question of how this syntagmatically distinctive function of fundamental frequency (as manifested in disyllabic non-verbs) developed into the paradigmatically distinctive tonal contrast (affecting also monosyllabic words). Here, another widely recognized process (e.g. Sun 2003: 38) may have played a crucial role: whereas syllables beginning with unprefixed nasals or liquids (Na) developed a low tone (Na̠), those beginning with prefixed nasals or liquids (x-Na) developed a high tone (Nā). Given the extreme productivity of the LH-pattern in PT, we may hypothesize that it was applied not only to disyllabic but, at some point, also to sesquisyllabic non-verbs such as gnam [xnam] ‘sky’ and gla [xla] ‘salary’, yielding x̱nām and x̱lā. When prefixes were later dropped in these dialects, as happened in all of Tibet and Kham, the only feature distinguishing these words from corresponding words without a prefix, such as nam ‘when’ and la ‘(mountain) pass’, was their tone. As voiceless nasals are quite common in eastern dialects, however, the x- prefix may have also devoiced the following nasal there (leading to xn̥am and xl̥a), and as a consequence, the vowels following such voiceless nasals came to be pronounced with a higher pitch than those following voiced nasals.
Caplow, Nancy. 2009. The role of stress in Tibetan tonogenesis: a study in historical comparative acoustics. Dissertation at the University of California in Santa Barbara.
Hombert, Jean-Marie, Ohala, John, and Ewan, William. 1979. Phonetic Explanations for the Development of Tones. Language 55 (1): 37–58.
Sun, Jackson. 2003. Variegated tonal developments in Tibetan. In David Bradley, Randy LaPolla, Boyd Michailovsky and Graham Thurgood (eds.), Language variation: papers on variation and change in the Sinosphere and in the Indosphere in honour of James A. Matisoff, 35–51. Canberra: Pacific Linguistics.
In western dialects of Tibetan, ’dug ‘was there’ became contrasted with the existential copula yod ‘is there’. While ’dug thereby came to mean that a present state was directly witnessed, yod came to mean that the speaker simply knows that state. In Ladakhi Tibetan, rag ‘was felt’ developed into a third existential evidential meaning that a current state was directly witnessed non-visually. Direct evidential ’dug thereby became restricted to contexts in which something was seen.
Equipollent evidential contrasts are also found in other regions of the world, such as the Caucasus (Khalilova 2011), the New Guinea Highlands (San Roque & Loughnane 2012), and the Americas (e.g. Cherokee in North America, see Aikhenvald 2004: 26–7 and Pulte 1985). Jarawara (a language of the southern Amazone, Dixon 2003) has contrasting sets of endings indicating direct and indirect evidence (and agreeing with the subject for person, number, and gender) for the immediate and the recent past tense. Tuyuca (an unrelated Eastern Tucanoan language of the northern Amazone, Barnes 1984), in addition to sets of direct and indirect endings has a set of direct non-visual endings, which appear to have restricted the other direct endings to visual evidence (including events which the speaker performed herself).
Note that all the discussed evidentials have a neatly defined evidential value (i.e. they indicate how the speaker knows the information conveyed in a statement) which is consistent across the contexts in which they occur. The shared tense-aspect values of contrasting evidentials point to the contexts in which their contrasting implications grammaticalized, that is, in which the evidentials became defined against each other.
In stark contrast to this evidence, it has been common practice among linguists working on evidentiality for the past fifty years to attribute a ‘visual evidential’ value to unmarked verb forms in languages in which there are inferential and/or reportative constructions containing additional morphemes. For Desano (another Eastern Tucanoan language), e.g., Kaye (1970) infers that the unmarked verb form in (4) implies “direct observation” from the fact that it lacks the -jo- morpheme which indicates that a past event is inferred in (3), and because (4) may indeed occur in contexts in which the speaker saw the event referred to (beware that ‘E’ in (3) stands for evidential, but has no phonetic substance, even if it stands in the transcription line).
For the majority of languages Aikhenvald (2004: 72ff.) adduces in her discussion of this issue, the ‘visual analysis’ of an unmarked verb form is supported by one example sentence at the most. Consulting the grammatical descriptions of the respective languages, we may thus identify the same recurrent argument reversal: if an unmarked verb form occurs in a context in which the event referred to was likely or could have been directly witnessed, it is assigned a ‘visual evidential’ value. Languages such as Tariana and Hup (Epps 2005), on the other hand, where the unmarked verb form is documented in various contexts, helped establishing the idea that it is common for ‘visual evidentials’ to refer to “generally known facts” or to events for which the speaker “takes full responsibility” (Aikhenvald 2004: 191), but that they rarely actually signal that an event was observed.
This paper proposes to analyze as evidentials only verb forms which consistently convey a specific evidential meaning, to stop attributing evidential values solely based on cross-linguistic parallels, and to re-establish the zero-hypothesis that a construction which is formally unmarked in terms of evidentiality is also functionally unmarked.
Amdo Tibetan (Sun 1993: 950):
(1) ʈʂaɕʰi=kə ʰtæ ɲu=tʰæ
Bkra-shis=ERG horse buy=DIREV
‘(I saw that) Tashi bought a horse.’
(2) ʈʂaɕʰi=kə ʰtæ ɲu=zəg
Bkra-shis=ERG horse buy=INDIREV
‘(It appears that) Tashi bought a horse.’
Desano (Kaye 1970: 34):
(3) igyN widi + á + waa + jo + biN
he leave ~pres.* happen ev. p.e.
‘he has left (I did not see him go but he is not here now)’
(4) igyN widi + á + waa + E + biN
he leave ~pres. happen ev. p.e.
‘he has left (I saw him and he is still away)’
*According to footnote 1 in Kaye (1970: 31), ‘p. e.’ stands for “any personal ending”, and ‘~pres.’ for “non-present tense”.
References:
Aikhenvald, Alexandra. 2004. Evidentiality. Oxford: Oxford University Press.
Barnes, Janet. 1984. Evidentials in the Tuyuca Verb. International Journal of American Linguistics 50: 255–71.
Dixon, R. M. W. 2003. Evidentiality in Jarawara. In Alexandra Aikhenvald and R. M. W. Dixon (eds.), Studies in Evidentiality, pp. 165–88. Amsterdam: John Benjamins.
Epps, Patience. 2005. Areal diffusion and the development of evidentiality: Evidence from Hup. Studies in Language 29: 617–50.
Kaye, Jonathan. 1970. The Desano Verb: Problems in Semantics, Syntax and Phonology. Ann Arbor: UMI.
Khalilova, Zaira. 2011. Evidentiality in Tsezic languages. Linguistic Discovery 9(2): 30–48.
Pulte, William. 1985. The experienced and non-experienced past in Cherokee. International Journal of American Linguistics 51: 543–4.
San Roque, Lila, and Robyn Loughnane. 2012. The New Guinea Highlands evidentiality area. Linguistic Typology 16(1): 111–67.
Sun, Jackson T.-S. 1993. Evidentials in Amdo Tibetan. Bulletin of the Institute of History and Philology, Academia Sinica 63(4): 143–88.
Zemp, Marius. 2020. Evidentials and their pivot in Tibetic and neighboring Himalayan languages. Functions of Language 27(1): 29–54.
In western dialects of Tibetan, ’dug ‘was there’ became contrasted with the existential copula yod ‘is there’. While ’dug thereby came to mean that a present state was directly witnessed, yod came to mean that the speaker simply knows that state. In Ladakhi Tibetan, rag ‘was felt’ developed into a third existential evidential meaning that a current state was directly witnessed non-visually. Direct evidential ’dug thereby became restricted to contexts in which something was seen.
Equipollent evidential contrasts are also found in other regions of the world, such as the Caucasus (Khalilova 2011), the New Guinea Highlands (San Roque & Loughnane 2012), and the Americas (e.g. Cherokee in North America, see Aikhenvald 2004: 26–7 and Pulte 1985). Jarawara (a language of the southern Amazone, Dixon 2003) has contrasting sets of endings indicating direct and indirect evidence (and agreeing with the subject for person, number, and gender) for the immediate and the recent past tense. Tuyuca (an unrelated Eastern Tucanoan language of the northern Amazone, Barnes 1984), in addition to sets of direct and indirect endings has a set of direct non-visual endings, which appear to have restricted the other direct endings to visual evidence (including events which the speaker performed herself).
Note that all the discussed evidentials have a neatly defined evidential value (i.e. they indicate how the speaker knows the information conveyed in a statement) which is consistent across the contexts in which they occur. The shared tense-aspect values of contrasting evidentials point to the contexts in which their contrasting implications grammaticalized, that is, in which the evidentials became defined against each other.
In stark contrast to this evidence, it has been common practice among linguists working on evidentiality for the past fifty years to attribute a ‘visual evidential’ value to unmarked verb forms in languages in which there are inferential and/or reportative constructions containing additional morphemes. For Desano (another Eastern Tucanoan language), e.g., Kaye (1970) infers that the unmarked verb form in (4) implies “direct observation” from the fact that it lacks the -jo- morpheme which indicates that a past event is inferred in (3), and because (4) may indeed occur in contexts in which the speaker saw the event referred to (beware that ‘E’ in (3) stands for evidential, but has no phonetic substance, even if it stands in the transcription line).
For the majority of languages Aikhenvald (2004: 72ff.) adduces in her discussion of this issue, the ‘visual analysis’ of an unmarked verb form is supported by one example sentence at the most. Consulting the grammatical descriptions of the respective languages, we may thus identify the same recurrent argument reversal: if an unmarked verb form occurs in a context in which the event referred to was likely or could have been directly witnessed, it is assigned a ‘visual evidential’ value. Languages such as Tariana and Hup (Epps 2005), on the other hand, where the unmarked verb form is documented in various contexts, helped establishing the idea that it is common for ‘visual evidentials’ to refer to “generally known facts” or to events for which the speaker “takes full responsibility” (Aikhenvald 2004: 191), but that they rarely actually signal that an event was observed.
This paper proposes to analyze as evidentials only verb forms which consistently convey a specific evidential meaning, to stop attributing evidential values solely based on cross-linguistic parallels, and to re-establish the zero-hypothesis that a construction which is formally unmarked in terms of evidentiality is also functionally unmarked.
Amdo Tibetan (Sun 1993: 950):
(1) ʈʂaɕʰi=kə ʰtæ ɲu=tʰæ
Bkra-shis=ERG horse buy=DIREV
‘(I saw that) Tashi bought a horse.’
(2) ʈʂaɕʰi=kə ʰtæ ɲu=zəg
Bkra-shis=ERG horse buy=INDIREV
‘(It appears that) Tashi bought a horse.’
Desano (Kaye 1970: 34):
(3) igyN widi + á + waa + jo + biN
he leave ~pres.* happen ev. p.e.
‘he has left (I did not see him go but he is not here now)’
(4) igyN widi + á + waa + E + biN
he leave ~pres. happen ev. p.e.
‘he has left (I saw him and he is still away)’
*According to footnote 1 in Kaye (1970: 31), ‘p. e.’ stands for “any personal ending”, and ‘~pres.’ for “non-present tense”.
References:
Aikhenvald, Alexandra. 2004. Evidentiality. Oxford: Oxford University Press.
Barnes, Janet. 1984. Evidentials in the Tuyuca Verb. International Journal of American Linguistics 50: 255–71.
Dixon, R. M. W. 2003. Evidentiality in Jarawara. In Alexandra Aikhenvald and R. M. W. Dixon (eds.), Studies in Evidentiality, pp. 165–88. Amsterdam: John Benjamins.
Epps, Patience. 2005. Areal diffusion and the development of evidentiality: Evidence from Hup. Studies in Language 29: 617–50.
Kaye, Jonathan. 1970. The Desano Verb: Problems in Semantics, Syntax and Phonology. Ann Arbor: UMI.
Khalilova, Zaira. 2011. Evidentiality in Tsezic languages. Linguistic Discovery 9(2): 30–48.
Pulte, William. 1985. The experienced and non-experienced past in Cherokee. International Journal of American Linguistics 51: 543–4.
San Roque, Lila, and Robyn Loughnane. 2012. The New Guinea Highlands evidentiality area. Linguistic Typology 16(1): 111–67.
Sun, Jackson T.-S. 1993. Evidentials in Amdo Tibetan. Bulletin of the Institute of History and Philology, Academia Sinica 63(4): 143–88.
Zemp, Marius. 2020. Evidentials and their pivot in Tibetic and neighboring Himalayan languages. Functions of Language 27(1): 29–54.
The original 20-minutes presentation with voiceover is found here: https://osf.io/zxe25/
The present paper demonstrates that Old Tibetan (OT) ga re ‘where is (X)?’, clause-linking (s)te ~ de, and V-ta re ‘lest (it) will V’, and other phenomena found in written and spoken Tibetic varieties, are best understood if analysed as traces of the mentioned clause-final demonstratives. The comparative study of spoken Tibetic varieties thus not only contributes to our understanding of particular OT texts, but also sheds light on the development and dispersion of Tibetic during the Imperial Period (7th–9th centuries CE).
The main development of the tonal Tibetic varieties, that is, those in which tone became phonemically distinctive, is well understood (e.g. Sun 2003) and borne out by crosslinguistic phonetic studies such as Hombert et al. (1979): vowels preceded by an originally voiced onset (Ga) have developed a low tone (Ka̠), whereas vowels preceded by a voiceless onset (Ka) have developed a high tone (Kā).
Another feature found throughout the tonal Tibetic varieties is, however, that the second syllable of disyllabic nouns is always, even if it originally began with a voiced obstruent, higher than the first syllable, that is, we find LH (where L = low and H = high) and HH, but neither LL nor HL. This conundrum was solved by Caplow (2009), whose robust statistical analysis of the acoustic correlates of stress in two non-tonal varieties in the opposite peripheries of the Tibetic language continuum shows that disyllabic non-verbs are consistently stressed there on the second syllable, and that this stress is primarily conveyed by fundamental frequency. Assuming that this was already the case in Proto-Tibetan (PT) explains the absence of LL and HL in the tonal varieties (Caplow 2009: 541–2). Evidence from another non-tonal variety will be presented to demonstrate how pervasively the LH-pattern marked disyllabic non-verbs in PT.
This allows us to identify the pivotal question of how this syntagmatically distinctive function of fundamental frequency (as manifested in disyllabic non-verbs) developed into the paradigmatically distinctive tonal contrast (affecting also monosyllabic words). Here, another widely recognized process (e.g. Sun 2003: 38) may have played a crucial role: whereas syllables beginning with unprefixed nasals or liquids (Na) developed a low tone (Na̠), those beginning with prefixed nasals or liquids (x-Na) developed a high tone (Nā). Given the extreme productivity of the LH-pattern in PT, we may hypothesize that it was applied not only to disyllabic but, at some point, also to sesquisyllabic non-verbs such as gnam [xnam] ‘sky’ and gla [xla] ‘salary’, yielding x̱nām and x̱lā. When prefixes were later dropped in these dialects, as happened in all of Tibet and Kham, the only feature distinguishing these words from corresponding words without a prefix, such as nam ‘when’ and la ‘(mountain) pass’, was their tone. As voiceless nasals are quite common in eastern dialects, however, the x- prefix may have also devoiced the following nasal there (leading to xn̥am and xl̥a), and then the vowel following a voiceless nasal came to be pronounced with a higher pitch than that following a voiced nasal.
Caplow, Nancy. 2009. The role of stress in Tibetan tonogenesis: a study in historical comparative acoustics. Dissertation at the University of California in Santa Barbara.
Hombert, Jean-Marie, Ohala, John, and Ewan, William. 1979. Phonetic Explanations for the Development of Tones. Language 55 (1): 37–58.
Sun, Jackson. 2003. Variegated tonal developments in Tibetan. In David Bradley, Randy LaPolla, Boyd Michailovsky and Graham Thurgood (eds.), Language variation: papers on variation and change in the Sinosphere and in the Indosphere in honour of James A. Matisoff, 35–51. Canberra: Pacific Linguistics.
The main development of the tonal Tibetic varieties, that is, those in which tone became phonemically distinctive, is well understood (e.g. Sun 2003) and borne out by crosslinguistic phonetic studies such as Hombert et al. (1979): vowels preceded by an originally voiced onset (Ga) have developed a low tone (Ka̠), whereas vowels preceded by a voiceless onset (Ka) have developed a high tone (Kā).
Another feature found throughout the tonal Tibetic varieties is, however, that the second syllable of disyllabic nouns is always, even if it originally began with a voiced obstruent, higher than the first syllable, that is, we find LH (where L = low and H = high) and HH, but neither LL nor HL. This conundrum was solved by Caplow (2009), whose robust statistical analysis of the acoustic correlates of stress in two non-tonal varieties in the opposite peripheries of the Tibetic language continuum shows that disyllabic non-verbs are consistently stressed there on the second syllable, and that this stress is primarily conveyed by fundamental frequency. Assuming that this was already the case in Proto-Tibetan (PT) explains the absence of LL and HL in the tonal varieties (Caplow 2009: 541–2). Evidence from another non-tonal variety will be presented to demonstrate how pervasively the LH-pattern marked disyllabic non-verbs in PT.
This allows us to identify the pivotal question of how this syntagmatically distinctive function of fundamental frequency (as manifested in disyllabic non-verbs) developed into the paradigmatically distinctive tonal contrast (affecting also monosyllabic words). Here, another widely recognized process (e.g. Sun 2003: 38) may have played a crucial role: whereas syllables beginning with unprefixed nasals or liquids (Na) developed a low tone (Na̠), those beginning with prefixed nasals or liquids (x-Na) developed a high tone (Nā). Given the extreme productivity of the LH-pattern in PT, we may hypothesize that it was applied not only to disyllabic but, at some point, also to sesquisyllabic non-verbs such as gnam [xnam] ‘sky’ and gla [xla] ‘salary’, yielding x̱nām and x̱lā. When prefixes were later dropped in these dialects, as happened in all of Tibet and Kham, the only feature distinguishing these words from corresponding words without a prefix, such as nam ‘when’ and la ‘(mountain) pass’, was their tone. As voiceless nasals are quite common in eastern dialects, however, the x- prefix may have also devoiced the following nasal there (leading to xn̥am and xl̥a), and as a consequence, the vowels following such voiceless nasals came to be pronounced with a higher pitch than those following voiced nasals.
Caplow, Nancy. 2009. The role of stress in Tibetan tonogenesis: a study in historical comparative acoustics. Dissertation at the University of California in Santa Barbara.
Hombert, Jean-Marie, Ohala, John, and Ewan, William. 1979. Phonetic Explanations for the Development of Tones. Language 55 (1): 37–58.
Sun, Jackson. 2003. Variegated tonal developments in Tibetan. In David Bradley, Randy LaPolla, Boyd Michailovsky and Graham Thurgood (eds.), Language variation: papers on variation and change in the Sinosphere and in the Indosphere in honour of James A. Matisoff, 35–51. Canberra: Pacific Linguistics.
In western dialects of Tibetan, ’dug ‘was there’ became contrasted with the existential copula yod ‘is there’. While ’dug thereby came to mean that a present state was directly witnessed, yod came to mean that the speaker simply knows that state. In Ladakhi Tibetan, rag ‘was felt’ developed into a third existential evidential meaning that a current state was directly witnessed non-visually. Direct evidential ’dug thereby became restricted to contexts in which something was seen.
Equipollent evidential contrasts are also found in other regions of the world, such as the Caucasus (Khalilova 2011), the New Guinea Highlands (San Roque & Loughnane 2012), and the Americas (e.g. Cherokee in North America, see Aikhenvald 2004: 26–7 and Pulte 1985). Jarawara (a language of the southern Amazone, Dixon 2003) has contrasting sets of endings indicating direct and indirect evidence (and agreeing with the subject for person, number, and gender) for the immediate and the recent past tense. Tuyuca (an unrelated Eastern Tucanoan language of the northern Amazone, Barnes 1984), in addition to sets of direct and indirect endings has a set of direct non-visual endings, which appear to have restricted the other direct endings to visual evidence (including events which the speaker performed herself).
Note that all the discussed evidentials have a neatly defined evidential value (i.e. they indicate how the speaker knows the information conveyed in a statement) which is consistent across the contexts in which they occur. The shared tense-aspect values of contrasting evidentials point to the contexts in which their contrasting implications grammaticalized, that is, in which the evidentials became defined against each other.
In stark contrast to this evidence, it has been common practice among linguists working on evidentiality for the past fifty years to attribute a ‘visual evidential’ value to unmarked verb forms in languages in which there are inferential and/or reportative constructions containing additional morphemes. For Desano (another Eastern Tucanoan language), e.g., Kaye (1970) infers that the unmarked verb form in (4) implies “direct observation” from the fact that it lacks the -jo- morpheme which indicates that a past event is inferred in (3), and because (4) may indeed occur in contexts in which the speaker saw the event referred to (beware that ‘E’ in (3) stands for evidential, but has no phonetic substance, even if it stands in the transcription line).
For the majority of languages Aikhenvald (2004: 72ff.) adduces in her discussion of this issue, the ‘visual analysis’ of an unmarked verb form is supported by one example sentence at the most. Consulting the grammatical descriptions of the respective languages, we may thus identify the same recurrent argument reversal: if an unmarked verb form occurs in a context in which the event referred to was likely or could have been directly witnessed, it is assigned a ‘visual evidential’ value. Languages such as Tariana and Hup (Epps 2005), on the other hand, where the unmarked verb form is documented in various contexts, helped establishing the idea that it is common for ‘visual evidentials’ to refer to “generally known facts” or to events for which the speaker “takes full responsibility” (Aikhenvald 2004: 191), but that they rarely actually signal that an event was observed.
This paper proposes to analyze as evidentials only verb forms which consistently convey a specific evidential meaning, to stop attributing evidential values solely based on cross-linguistic parallels, and to re-establish the zero-hypothesis that a construction which is formally unmarked in terms of evidentiality is also functionally unmarked.
Amdo Tibetan (Sun 1993: 950):
(1) ʈʂaɕʰi=kə ʰtæ ɲu=tʰæ
Bkra-shis=ERG horse buy=DIREV
‘(I saw that) Tashi bought a horse.’
(2) ʈʂaɕʰi=kə ʰtæ ɲu=zəg
Bkra-shis=ERG horse buy=INDIREV
‘(It appears that) Tashi bought a horse.’
Desano (Kaye 1970: 34):
(3) igyN widi + á + waa + jo + biN
he leave ~pres.* happen ev. p.e.
‘he has left (I did not see him go but he is not here now)’
(4) igyN widi + á + waa + E + biN
he leave ~pres. happen ev. p.e.
‘he has left (I saw him and he is still away)’
*According to footnote 1 in Kaye (1970: 31), ‘p. e.’ stands for “any personal ending”, and ‘~pres.’ for “non-present tense”.
References:
Aikhenvald, Alexandra. 2004. Evidentiality. Oxford: Oxford University Press.
Barnes, Janet. 1984. Evidentials in the Tuyuca Verb. International Journal of American Linguistics 50: 255–71.
Dixon, R. M. W. 2003. Evidentiality in Jarawara. In Alexandra Aikhenvald and R. M. W. Dixon (eds.), Studies in Evidentiality, pp. 165–88. Amsterdam: John Benjamins.
Epps, Patience. 2005. Areal diffusion and the development of evidentiality: Evidence from Hup. Studies in Language 29: 617–50.
Kaye, Jonathan. 1970. The Desano Verb: Problems in Semantics, Syntax and Phonology. Ann Arbor: UMI.
Khalilova, Zaira. 2011. Evidentiality in Tsezic languages. Linguistic Discovery 9(2): 30–48.
Pulte, William. 1985. The experienced and non-experienced past in Cherokee. International Journal of American Linguistics 51: 543–4.
San Roque, Lila, and Robyn Loughnane. 2012. The New Guinea Highlands evidentiality area. Linguistic Typology 16(1): 111–67.
Sun, Jackson T.-S. 1993. Evidentials in Amdo Tibetan. Bulletin of the Institute of History and Philology, Academia Sinica 63(4): 143–88.
Zemp, Marius. 2020. Evidentials and their pivot in Tibetic and neighboring Himalayan languages. Functions of Language 27(1): 29–54.
In western dialects of Tibetan, ’dug ‘was there’ became contrasted with the existential copula yod ‘is there’. While ’dug thereby came to mean that a present state was directly witnessed, yod came to mean that the speaker simply knows that state. In Ladakhi Tibetan, rag ‘was felt’ developed into a third existential evidential meaning that a current state was directly witnessed non-visually. Direct evidential ’dug thereby became restricted to contexts in which something was seen.
Equipollent evidential contrasts are also found in other regions of the world, such as the Caucasus (Khalilova 2011), the New Guinea Highlands (San Roque & Loughnane 2012), and the Americas (e.g. Cherokee in North America, see Aikhenvald 2004: 26–7 and Pulte 1985). Jarawara (a language of the southern Amazone, Dixon 2003) has contrasting sets of endings indicating direct and indirect evidence (and agreeing with the subject for person, number, and gender) for the immediate and the recent past tense. Tuyuca (an unrelated Eastern Tucanoan language of the northern Amazone, Barnes 1984), in addition to sets of direct and indirect endings has a set of direct non-visual endings, which appear to have restricted the other direct endings to visual evidence (including events which the speaker performed herself).
Note that all the discussed evidentials have a neatly defined evidential value (i.e. they indicate how the speaker knows the information conveyed in a statement) which is consistent across the contexts in which they occur. The shared tense-aspect values of contrasting evidentials point to the contexts in which their contrasting implications grammaticalized, that is, in which the evidentials became defined against each other.
In stark contrast to this evidence, it has been common practice among linguists working on evidentiality for the past fifty years to attribute a ‘visual evidential’ value to unmarked verb forms in languages in which there are inferential and/or reportative constructions containing additional morphemes. For Desano (another Eastern Tucanoan language), e.g., Kaye (1970) infers that the unmarked verb form in (4) implies “direct observation” from the fact that it lacks the -jo- morpheme which indicates that a past event is inferred in (3), and because (4) may indeed occur in contexts in which the speaker saw the event referred to (beware that ‘E’ in (3) stands for evidential, but has no phonetic substance, even if it stands in the transcription line).
For the majority of languages Aikhenvald (2004: 72ff.) adduces in her discussion of this issue, the ‘visual analysis’ of an unmarked verb form is supported by one example sentence at the most. Consulting the grammatical descriptions of the respective languages, we may thus identify the same recurrent argument reversal: if an unmarked verb form occurs in a context in which the event referred to was likely or could have been directly witnessed, it is assigned a ‘visual evidential’ value. Languages such as Tariana and Hup (Epps 2005), on the other hand, where the unmarked verb form is documented in various contexts, helped establishing the idea that it is common for ‘visual evidentials’ to refer to “generally known facts” or to events for which the speaker “takes full responsibility” (Aikhenvald 2004: 191), but that they rarely actually signal that an event was observed.
This paper proposes to analyze as evidentials only verb forms which consistently convey a specific evidential meaning, to stop attributing evidential values solely based on cross-linguistic parallels, and to re-establish the zero-hypothesis that a construction which is formally unmarked in terms of evidentiality is also functionally unmarked.
Amdo Tibetan (Sun 1993: 950):
(1) ʈʂaɕʰi=kə ʰtæ ɲu=tʰæ
Bkra-shis=ERG horse buy=DIREV
‘(I saw that) Tashi bought a horse.’
(2) ʈʂaɕʰi=kə ʰtæ ɲu=zəg
Bkra-shis=ERG horse buy=INDIREV
‘(It appears that) Tashi bought a horse.’
Desano (Kaye 1970: 34):
(3) igyN widi + á + waa + jo + biN
he leave ~pres.* happen ev. p.e.
‘he has left (I did not see him go but he is not here now)’
(4) igyN widi + á + waa + E + biN
he leave ~pres. happen ev. p.e.
‘he has left (I saw him and he is still away)’
*According to footnote 1 in Kaye (1970: 31), ‘p. e.’ stands for “any personal ending”, and ‘~pres.’ for “non-present tense”.
References:
Aikhenvald, Alexandra. 2004. Evidentiality. Oxford: Oxford University Press.
Barnes, Janet. 1984. Evidentials in the Tuyuca Verb. International Journal of American Linguistics 50: 255–71.
Dixon, R. M. W. 2003. Evidentiality in Jarawara. In Alexandra Aikhenvald and R. M. W. Dixon (eds.), Studies in Evidentiality, pp. 165–88. Amsterdam: John Benjamins.
Epps, Patience. 2005. Areal diffusion and the development of evidentiality: Evidence from Hup. Studies in Language 29: 617–50.
Kaye, Jonathan. 1970. The Desano Verb: Problems in Semantics, Syntax and Phonology. Ann Arbor: UMI.
Khalilova, Zaira. 2011. Evidentiality in Tsezic languages. Linguistic Discovery 9(2): 30–48.
Pulte, William. 1985. The experienced and non-experienced past in Cherokee. International Journal of American Linguistics 51: 543–4.
San Roque, Lila, and Robyn Loughnane. 2012. The New Guinea Highlands evidentiality area. Linguistic Typology 16(1): 111–67.
Sun, Jackson T.-S. 1993. Evidentials in Amdo Tibetan. Bulletin of the Institute of History and Philology, Academia Sinica 63(4): 143–88.
Zemp, Marius. 2020. Evidentials and their pivot in Tibetic and neighboring Himalayan languages. Functions of Language 27(1): 29–54.
The original 20-minutes presentation with voiceover is found here: https://osf.io/zxe25/
(This lecture forms part of the phonology course I held at the Institut für Sprachwissenschaft of the University of Bern in the fall of 2019.)