Number in Yeniseian Vajda

Edward Vajda 9 Number in Ket (Yeniseian) Abstract: Ket is a critically endangered language spoken in Siberia in the Yenisei river basin and is the last surviving member of the once widespread Yeniseian family whose typological profile is very different from its closest neighbours. Nouns and pronouns distinguish singular and plural number, usually by adding a plural suffix, with singular number left morphologically unmarked. Plural suffixes have distinct allomorphs for inanimate class nouns, kinship terms, and other animate class nouns, but there are many exceptions and irregular forms. Attributive adjectives and other modifiers are normally left unmarked for number, though a few adjectives have a plural suffix. Demonstrative pronouns, however, regularly express plurality when modifying animate class plural nouns. The Ket verb expresses agreement in singular and plural number with its subject and object and also has a variety of morphological means for expressing pluractionality, resulting in various patterns of multiple exponence of number on the verb. Among other topics, this chapter focuses on the relationship between the expression of number and animacy, which variously manifests itself in the morphology of nouns, pronouns and finite verbs. It also explains how certain irregularities in Ket number marking developed and includes comparisons with the extinct Yeniseian languages. 1 Overview Ket is spoken today by a dwindling number of elders, mostly in remote Siberian villages near the Yenisei river or its tributaries in Turukhansk District of Krasnoyarsk Krai, the Russian Federation’s second largest administrative unit. Alongside its extinct sisters – Yugh, Kott, Assan,1 Arin, and Pumpokol – Ket belongs to the Yeniseian (Yeniseic) language family. Substrate river names and 17th century Tsarist fur tax records indicate that languages or dialects related to modern Ket were once spoken across much of south and west-central Siberia, from northern Mongolia and the forests southwest of Lake Baikal westward to the Ob river watershed and northward along the Yenisei to the Arctic Circle (Map 1). Three Ket dialects have survived into the early 21st century: Southern Ket (spoken in several villages, including Kellog, Sulomai, and Alinskoye), Central Ket (Sur- 1 In terms of linguistic similarity, Assan could easily be considered a dialect of Kott. However, as in the case of Yugh, the speakers considered themselves to be ethnically distinct. For this reason, Assan and Kott, like Ket and Yugh, are generally regarded today as closely related pairs of languages. https://doi.org/10.1515/9783110622713-010 308 Edward Vajda Map 1: Language distribution in central Siberia during the 17th century. gutikha and Baklanikha), and Northern Ket (Maduika and Kureika). At the time of this writing, these dialects together have no more than a few dozen fully fluent native speakers, all over the age of 65. There are only small differences in how each Ket dialect expresses number, and most are of a predictable phonological nature. Except where otherwise noted, this chapter cites Southern Ket forms. Number marking in Ket broadly resembles Yugh (Werner 1997a), a distinct but closely related language that disappeared in the 1970s. Most 20th century Soviet scholarship on Yeniseian languages treated Yugh as another Ket dialect, calling it “Sym-Ket” after a river where most speakers lived. By contrast, Southern, Central and Northern Ket were called “Imbat Ket” after Inbak – a 19th century ethno-geographic term for several downriver groups of Kets. The other documented Yeniseian language varieties, all of which disappeared between 1730 and 1850, are more distantly related and sometimes provide deeper insights into the historical development of number marking in the family. The extant material representing the Kott language, which belongs to a different primary branch than Ket and Yugh, is the most useful in this regard because it contains many inflected word forms, including plurals, thanks to a gram- Number in Ket (Yeniseian) 309 mar sketch and dictionary recorded by the Finnish field linguist M. A. Castrén (1858). Yeniseian languages are in many ways structurally unlike the “Ural-Altaic” belt of language families originally spoken across nearly all of the remaining portions of northern and Inner Eurasia. The most striking difference is the Ket polysynthetic finite verb, which is based on a template of ten morpheme classes, most of which, at least historically speaking, are represented in prefixal positions. By contrast, verbs in Turkic, Tungusic, and Uralic languages are exclusively suffixing in their inflectional morphology. Possession in Yeniseian is likewise expressed using markers placed in front of the possessed noun (or noun phrase), rather than by suffixes as in the neighboring languages. This chapter covers all aspects of how Southern Ket expresses number categories. It also considers the diachronic origins – whether genealogical or contactinduced – behind the various formal systems of number marking across the Yeniseian family. In keeping with the volume’s general structure, section 2 provides descriptions of number marking across the language’s major form classes. Subsection 2.2 examines the Ket pronoun system, which manifests all of the language’s core number distinctions in one way or another. It also provides a general overview of how animacy intersects with plural marking patterns in several areas of the morphosyntax. Subsection 2.3 investigates the complex motivations behind how plural suffix allomorphs are distributed across the noun lexicon, and also looks at sporadic examples of dual number marking that arose through morphological reanalysis of certain noun stems. Subsection 2.4 describes morphological techniques used for expressing pluractionality. Section 3 discusses agreement morphology. Subsection 3.1 examines number agreement in demonstrative pronouns and adjectives. Subsection 3.2 examines how adjective and action nominal suffixes have been reanalyzed as plural markers in a number of stems. Subsection 3.3 discusses the number agreement suffixes found on certain lexical classes of subject complements in the predicates of clauses with no finite verb form. Finally, subsection 3.4 turns to the highly complex polysynthetic Ket verb to explain its intricate system of class and number agreement with subjects and objects. Section 4 remarks on what is known about the discourse functions of number marking. Section 5 summarizes the typology of number marking in modern Ket and attempts to sort out which techniques were inherited from Proto-Yeniseian and which were later innovated. The discussion throughout this chapter also points out noteworthy features and categories related to number that are found in other languages but absent in Yeniseian. Much of the material here was covered earlier in Porotova (1990), the only existing monograph devoted to the expression of plurality in Yeniseian. Accessible treatments of most aspects of Ket number marking can also be found in Werner (1997b), Vajda (2004), Georg (2007), and Nefedov & Vajda (2015). 310 Edward Vajda 2 Pronominal, nominal, and verbal number 2.1 Generalities Number marking features shared widely with other Siberian language families include the productive use of suffixes to mark plurality in nouns. The expression of number in Ket and other Yeniseian languages is based on a formal distinction between singular (one) and plural (two or more), and this dichotomy strongly affects the morphosyntax of nouns, pronouns, and finite verbs. Yeniseian also has a noun class (grammatical gender) system based on the contrast between inanimate and animate class, with animate class further subdivided into masculine and feminine gender. Ket noun class (animacy and gender) interacts with number marking in several important ways, with inanimate class nouns showing an overt distinction between singular and plural only in the noun form itself, but not in the possessive markers or verb agreement affixes that cross-reference them. Finally, modern Ket has also developed a typologically unusual system of phonemic tones based on an amalgam of features involving melody, length and phonation type.2 Tone is marginally germane to the topic of number in Ket because a small irregular set of singular vs. plural noun forms are formally distinguished solely by tonal differences. 2.2 Pronominal number 2.2.1 Independent pronouns The system of personal and anaphoric pronouns in Ket offers a good starting point for discussing number marking. These pronouns convey all of the core number distinctions present in Yeniseian morphosyntax (singular vs. plural) and provide clear examples of the most typical formal means used to mark them (plurality expressed 2 The Southern Ket data in this chapter derive from the author’s own fieldwork. Monosyllabic phonological words distinguish four phonemic tones. These are transcribed using a macron for high-even tone (sūl ‘blood’), apostrophe for rising laryngealized tone (su’l ‘white salmon’), double vowel letter for rising-falling tone on a geminate vowel (suul ‘snow sled’), and grave accent for falling tone (sùl ‘cradle hook’). Most polysyllabic phonological words have instead an accent-like pitch on the first syllable, as in qópqun ‘cuckoo bird’. In a much smaller number of polysyllabic words the pitch peak falls on the second syllable, one example being the plural form qopqún ‘cuckoo birds’. Because syllable-initial pitch in polysyllables is far more common, it is left unmarked except in disyllabic noun stems like qópqun ‘cuckoo bird’, where it is included to call attention to its relevance for number marking. Finally, each of the three Ket mid vowel phonemes have two regular allophones. The symbol /ǝ/ is used to transcribe the mid back unrounded vowel, which is realized phonetically as [ɤ̄ˑ] under high-even tone and as [ʌ] elsewhere. In a similar fashion, /o/ transcribes both [ōˑ] and [ɔ] and /e/ transcribes [ēˑ] and [ɛ]. Number in Ket (Yeniseian) 311 through suffixes that normally contain a nasal consonant, either /n/ or /ŋ/). Table 1 shows the system of singular and plural animate personal pronouns in Southern Ket: Tab. 1: Ket personal and anaphoric pronouns. 1sg ād 1pl ǝtn 2sg ū 2pl ǝkŋ 3sg.anim bū 3pl.anim būŋ 3inan tude The 3rd person anaphoric pronouns bū ‘he / she’ and būŋ ‘(animate class) they’ normally refer back to nouns denoting human beings, but sometimes animals as well. Other nouns are generally replaced anaphorically by tude ‘that (one)’, ‘those ones’ (pronounced [tuɾɛ] in Southern Ket), which can also be used adnominally as a demonstrative pronoun.3 The simplicity of this paradigm stems in part from the fact that Yeniseian pronouns do not distinguish dual number. Nor is there any inclusive/exclusive distinction. The 1st person plural pronoun ǝtn ‘we’ can mean ‘you and me’ or ‘you and us’ (inclusive); or it can mean ‘me with another or others’ (exclusive). In contrast to many modern Indo-European languages, the Ket 2nd person plural pronoun ǝkŋ ‘you’ was not traditionally used to address a single individual to show respect or mark social distance, though recent Russian interference influenced such occasional usage among the last generation of speakers. Ket pronoun forms also exemplify the number marking strategy that is overwhelmingly favored across Yeniseian nominal morphology. The core distinction between singular (one) and plural (more than one) is formally expressed by augmenting the singular form with a suffix containing either the velar nasal /ŋ/ or alveolar nasal /n/. There is no grammatical means for marking singular number, aside from the absence of plural marking. Plural suffixation in nouns sometimes entails irregular morphophonemic changes, as seems to be the case with Ket 1st and 2nd plural pronouns as well. Finally, although Yeniseian morphosyntax lacks a grammatical category of dual number, there are sporadic instances where a nasal coda in certain noun stems denoting paired objects such as mittens has been reanalyzed as a dual marker (see Section 4.3). The personal pronoun system likewise affords a convenient introduction to the important role played by animacy in Ket morphosyntax. We have already seen that 3 Possessive markers are considered clitics because they normally attach to the preceding word if one is available rather than to the following possessum noun, as shown here. Substitutions of 3rd person animate class bū for inanimate class tude with reference to inanimate class entities have occasionally been documented, possibly due to Russian interference. 312 Edward Vajda considerations of animacy determine which third-person anaphoric pronouns are used. Third-person singular bū ‘he / she / him / her’ and plural būŋ ‘they / them’ normally reference animate entities, while the demonstrative form tude ‘that’ is used when referring back to singular or plural inanimate class entities. The biological gender distinction in third-person animate singular bū is covert in the pronoun itself, but surfaces elsewhere in the clause by triggering different masculine and feminine forms of singular possessive clitics (1), subject concord suffixes on predicate adjectives (2), and subject/object agreement affixes in the finite verb complex (3): (1) Animate singular gender distinction in possessive marking: (a) masculine, (b) feminine qu’s a. bu=da4 3sg.pron=m.poss tent ‘his tent’ b. bu=d qu’s 3sg.pron=f.poss tent ‘her tent’ (2) Animate singular gender distinction in predicate adjectives: (a) masculine, (b) feminine a. hīk [hīˑɣ] sel-du man bad-m.sbj ‘The man is bad.’ b. qīm sel-da woman bad-f.sbj ‘The woman is bad.’ (3) Animate singular gender distinction in verb-internal subject/object agreement: a. masculine subject with feminine object, hīk qīm d-i-toŋ man woman 3m.sbj-3.obj-see ‘The man sees the woman.’ 4 The combination bu=da is pronounced [buɾa] in Southern Ket. To keep the underlying forms more transparent, the transcription ignores the allophonic intervocalic lenition of /d/ to [ɾ], /b/ to [v], /k/ to [ɣ] and [q] to [ʁ], as well as voicing of /k/ to [g] next to a voiced consonant. It also ignores word-final devoicing of /d/ to [t] and /b/ to [p]. Note that word-final /p/ in words such as dāp ‘shoulder’ belongs to the same phoneme as word-initial /h/ and not to /b/. The symbols /p/ and /h/ here could have been unified by choosing one or the other in all cases; however, doing so would have given a misleading impression of how these allophones are actually pronounced in modern Ket. Phonetic forms are sometimes provided in square brackets, especially in less transparent cases such as certain Southern Ket high-even tone words and all falling-tone words, where Number in Ket (Yeniseian) b. 313 feminine subject with masculine object qīm hīk da=a-toŋ woman man 3 f.sbj=3m.obj-see ‘The woman sees the man.’5 The agreement contrast triggered by masculine and feminine animate singular forms does not extend to expressions of plurality. However, the difference between inanimate as opposed to animate entities is regularly maintained in the same three environments of possessive constructions (4), subject concord (5), and verb-internal subject or object agreement (6): (4) 3p inanimate (a) and animate (b) possessive clitics a. tu-de=d kīd that-inan=inan.poss price ‘its price’ / ‘their.inan price’ b. buŋ=na kīd they=anim.pl.poss price ‘their.anim price’ (5) 3p plural inanimate (a) and animate plural (b) subject concord in predicate adjectives a. tu-de sel-am that-inan bad-3inan.sbj ‘It is bad.’ / They.inan are bad.’ b. būŋ sel-aŋ they.anim bad-3anim.pl.sbj ‘They.anim are bad.’ (6) Verb-internal agreement with inanimate (a) and animate (b) plural objects a. būŋ tu-de du-Ø-toŋ-n they.anim that-inan 3anim.sbj-3inan.obj-see-anim.pl.sbj ‘They see it.’ / ‘They see them.inan.’ intervocalic lenition persists even though the final vowel has elided: hīk [hīˑɣ] ‘man’, qòq [qχɔ̀ʁ] ‘star’. 5 The feminine class subject marker /da/ is also a clitic that prefers to attach to a preceding word, whenever one is available, just like the possessive markers described earlier. Enclisis is less likely when the preceding word ends in an obstruent, as in this example. All Ket morphemes marked as clitics in this chapter behave in a similar way. 314 Edward Vajda b. būŋ de’ŋ d-aŋ-toŋ-n they.anim people 3sbj-3anim.pl.obj-see-anim.pl.sbj ‘They see the people.’ Examples (1) to (6) provide a good introduction to how animacy and gender (noun class) interacts with the expression of number across Ket morphosyntax. Only singular referents trigger the overt expression of gender, while the contrast between animate and inanimate extends to patterns of plural agreement, as well. The pervasive role of animacy in Ket number marking is further explored in the next section. 2.2.2 Number in possessive constructions One key Yeniseian typological feature absent from other families of northern Eurasia outside of Indo-European is the presence of a noun class system, examples of which have already been provided above. Unlike Russian and other Indo-European languages with grammatical gender, however, Ket agreement classes reflect foremost a distinction between animate and inanimate entities, with animate class nouns secondarily divided into masculine and feminine sub classes in singular forms – a division of much less significance to the grammar overall. The animate superclass subsumes not only humans, animals, birds, fish, reptiles, amphibians, and insects, but also certain body parts, as well as trees species. A few animate class nouns denote culturally salient inanimate objects such as pinecones, tent poles, buttons, and cradle hooks, with the masculine designation typically favoring objects of special positive value in traditional Ket lifeways. Grammatical animacy and gender are usually covert in the noun stem itself. The class division between inanimate and animate has broad ramifications for number marking in several domains of Yeniseian morphosyntax. The next several paragraphs analyze the role of animacy and gender in the shape of possessive markers and various word forms that evolved on the basis of possessed noun constructions. The expression of number in possessive markers intersects with animacy and gender distinctions. The forms in Table 2 illustrate how this pattern is reflected in the possessive forms of Ket personal pronouns: Tab. 2: Possessive forms of personal pronouns. 1sg a=b 1pl ǝtn=na 2sg u=k 2pl ǝkŋ=na 3sg.m bu=da 3pl.anim būŋ=na 3sg.f bu=d 3inan tu-de=d Number in Ket (Yeniseian) 315 Singular possessive markers, shown as enclitics in Table 2, distinguish person and gender of the possessor, except that the 3rd person feminine animate singular and inanimate markers formally coincide. More importantly, the same inanimate class possessive clitic is used for singular and plural number. Clitics only differentiate masculine and feminine and person categories with regard to singular entities, while the plural possessive markers distinguish animacy only. In the absence of a preceding pronoun to disambiguate person, constructions with an animate plural possessor such as na=ām could therefore be understood to mean either ‘their mother’, ‘our mother’, or ‘your.pl mother’. In cases where no possessor is named, the possessive marker can attach either to the following word as proclitic or to the preceding word as enclitic, depending on factors such as sentence structure and intonation. Enclisis is favored whenever a preceding word is available as host, particularly those that end in a vowel or sonorant: (7) Possessive marker as (a) proclitic or (b) enclitic a. da=ām his=mother ‘his mother’ b. ǝqaj=da am-as d-ol-daq formerly=his mother-with 3m.sgj-pst-live ‘He used to live with his mother.’ Several types of Ket morphological structures derive from possessed nouns, including some case suffixes, most postpositional constructions, and a minor lexical class known as directional adverbs. All of them retain the same interplay between noun class and the expression of number displayed in possessed noun constructions. Three Southern Ket case suffixes – dative (to, toward), adessive (at, in, having), and ablative (from, out of)6 – are preceded by a possessive augment expressing the same characteristic gender and animacy distinctions as those associated with possessed nouns. When animate class plural nouns are used with these case suffixes, the resulting word forms exhibit multi-site plural marking: 6 The so-called benefactive case forms of -dita ‘feminine singular or inanimate’, -data ‘masculine singular’, and -nata ‘animate plural’, appear to be idiolectal reductions of adessive case forms. 316 (8) Edward Vajda Adessive (a–b), dative (c–d), and ablative (e–f) case suffixes i. with inanimate nouns a. suul-aŋ-di-ŋten snow.sled-pl-inan-ades ‘in the snow sleds’ ii. with animate plural nouns b. am-aŋ-na-ŋten mother-pl-anim.pl-ades ‘at the mothers’ c. baŋ-n-di-ŋa land-pl-inan-dat ‘to our (your, their) lands’ d. hun-aŋ-na-ŋa daughter-pl-anim.pl-dat ‘to our (your, their) daughters’ e. bes-n-di-ŋal rabbit-pl-inan-abl ‘(made) out of rabbit pelts’7 f. bes-n-na-ŋal rabbit-pl-anim.pl-abl ‘(motion) from rabbits (living animals)’ The markers -di- and -na- in (8) are glossed simply as inan and anim.pl, respectively, since the earlier possessive meaning was lost. Vajda (2013b: 80–84) has argued that the segment /ŋ/ located at the beginning of the modern dative, adessive and ablative case suffixes is a vestige of an ancient possessive connector once present in all possessive constructions. Regardless of whether this interpretation is valid, structures such as (8a), (8b), and (8c) suggest that the three possessive-augmented case suffixes may derive from underlying noun forms or postpositions, though the precise etymologies remain unclear. The other case suffixes – locative, comitative-instrumental (with), prolative (along or though), and caritive (without) – attach to nouns or pronouns without a preceding possessive augment, for reasons that are likewise unclear. Ket postpositions, most of which require a case suffix, form complex constructions that are often preceded by a possessive marker that agrees in number, animacy and gender with the preceding noun or pronoun stem; they preserve all of the same morphosyntactic distinctions expressed in true possessive constructions. Some Ket postpositions transparently derive from body-part nouns, suggesting that postpositional constructions arose via the grammaticalization of possessed nouns. For example, the Southern Ket postposition -ɯn- ‘below’, is cognate with a noun root meaning ‘belly’ or ‘underside’ in the extinct Yeniseian languages. In Ket it no longer expresses its original anatomical meaning, having been replaced by hɯ̄j ‘belly’, though Ket does preserve a cognate noun ɯ’n ‘sled runner’. The examples below use inanimate and animate class clitics to show the homology of the structures displayed by possessed nouns (9) and by the postpositional constructions grammaticalized on the basis of possessed nouns (10): 7 See Werner (1997b: 107) for the full paradigms of animate class be’s ‘rabbit (live animal)’ vs. inanimate class be’s ‘rabbit (pelt)’. The inanimate class compound besiŋolt ‘rabbit pelt’ (with iŋolt ‘pelt’, ‘fur’) also exists. Number in Ket (Yeniseian) (9) 317 Possessed nouns with case suffixes a. suul=d ɯ’n snow.sled=inan.poss runner (< underside) ‘sled runner (one of two long pieces of wood beneath a sled for gliding on snow)’ b. ked=da hɯ̄j person=m.poss belly ‘person’s belly’ (10) Contrast between (a) possessed noun construction and (b) postpositional construction a. ked=da=hɯj-di-ŋal person=m.poss=belly-inan-abl ‘from the person’s belly’ b. ked=da=hɯj=d=ɯn-di-ŋal person=m.poss=belly-inan.poss=under (< underside)-inan-abl ‘out from under the person’s belly Directional adverbs, which express orientation relative to a specific point in space, are built exactly like possessed noun constructions. The two most common directional roots are ikda [igda] ‘downland’, ‘orientation from forest to river’ and aka [aɣa] ‘upland’, ‘orientation from river to forest’. Directional word forms are adverbs that consist of a possessive clitic and directional root, followed by a case suffix. If the required case is ablative or dative, the case suffix must be preceded by the connector -di-, originally the inanimate class possessive marker (the adessive case is not used in directional constructions). (11) Directional adverbs a. assanos-in na=ikda-di-ŋal hunter-pl anim.pl.poss=downland-inan.poss-abl ‘(movement) from a location downland with respect to the hunters’ b. qu-ŋ d=aka-di-ŋa tent-pl inan.poss=upland-inan.poss-dat ‘(movement) toward a location upland with respect to the tents’ Ket possessed nouns, certain case forms, postpositional constructions, and directional adverbs all display multi-site marking of number, animacy, and what is often called “gender” (masculine vs. feminine animate singular agreement). Instances of multiple exponence in all of these structures arose through the coalescence of formerly discrete words into single morphological forms. The expression of noun class in connection with number in certain case forms and postpositional constructions 318 Edward Vajda is unique among the indigenous language families of landlocked northern Asia. The feature of multi-site plural marking in some of these word forms (cf. 8 above) is likewise unique to Yeniseian across this broader geographic area. The effect of animacy on number marking patterns extends to other parts of Ket morphosyntax, as well. Section 2.3 explores how considerations of both referential and grammatical animacy influences the choice of plural suffix allomorphs in Ket noun forms. Section 3.3 analyzes the effect of animacy on the subject concord suffixes required by certain subject complements in linking verb clauses, which lack a finite verb. Section 3.4.1 examines the role played by noun class in assigning subject and object number agreement markers in finite verb forms. 2.3 Nominal number Yeniseian nouns have no grammatical marker of singular number. Count nouns typically distinguish number by augmenting the bare singular stem with a plural suffix containing a nasal consonant – either /ŋ/ or /n/. There are no nominal plural prefixes. Stem reduplication is absent as a means of expressing plurality or pluractionality, being limited in Yeniseian morphology to a few expressive interjections. Overall, the choice of noun plural suffix shape follows several broad patterns that together form a complex interweave of semantics (animate vs. inanimate class) with certain formal features of stem structure, the most important being whether the noun stem already ends in /ŋ/ or /n/. There are also many irregularities caused by diachronic processes of reductions in the noun stem’s final syllable or coda, some of which have caused the loss of the plural suffix itself. The subsections below discuss the motivations that determine plural noun forms (2.2.1), dual number marking arising sporadically through reanalysis of noun stem coda nasals (2.2.2), as well as pluralia tantum, singularia tantum, singulatives, and a few other number-related lexical categories at work in Ket noun morphology (2.2.3). This section goes into considerable detail because the expression of plurality in nouns is fundamental to number marking throughout Yeniseian morphology, given that nearly all other number marking categories reflect the singular/plural dichotomy inherent to nominal morphology. 2.3.1 The morphology of noun pluralization Animacy plays a fundamental role in how nouns build their plurals in Yeniseian languages. Animate and inanimate class nouns typically have different plural suffix shapes, except where phonological or diachronic factors intervene. Inanimate class nouns normally take the suffix -(V)ŋ. The symbol (V) here and elsewhere represents a phonologically undetermined vowel that is often inserted between stem coda and nasal suffix: Number in Ket (Yeniseian) (12) Examples of inanimate class noun pluralization using -(V)ŋ qobad ‘back’ qobad-aŋ bǝtl ‘bubble’ bǝtl-aŋ sɯ’k ‘bowl’ sɯk-ŋ qoŋloq ‘bell’ qoŋloq-ŋ asl ‘ski’ asl-iŋ àj ‘bag’ aj-éŋ 319 ‘backs’ ‘bubbles’ ‘bowls ‘bells’ ‘skis’8 ‘bags’ The animate class noun plural suffix is -(V)n, with alveolar rather than velar nasal. It too is sometimes augmented by a vowel after a stem-final consonant: (13) Examples of animate class noun pluralization using -(V)n hīk ‘man’ hik-n ostɯ́k ‘Ket’ ostɯ́k-ǝn be’s ‘rabbit’ bes-n qīm ‘woman’ qim-n hǝ́ mka ‘Evenki’ hǝ́ mka-n la’t ‘beaver’ lat-n ‘men’ ‘Kets’ ‘rabbits’ ‘women’ ‘Evenkis’ ‘beavers’ The distribution of the alveolar as opposed to velar nasal consonant in noun plural suffixes is determined by much more than grammatical animacy, however. In fact, this deceptively elegant dichotomy has so many exceptions that recognizing a noun’s inherent class membership based on its surface plural form is unreliable. (Noun class can be detected more reliably from observing the form of possessive markers or agreement morphology). The quality of the vowel before either of these basic noun plural suffixes is often unpredictable, as well. Some of these vowels appear to be epenthetic, which often appears true for [ǝ] or [ɯ]. Others reflect preservation of an original stem vowel lost in the singular form, as tends to be true of many plurals that augment the nasal suffix using [i], [e], or [a]. In other cases, it remains unclear what determines the choice of vowel in noun plural suffixes (or even whether any vowel appears at all). One fairly consistent semantic exception to the basic pattern of animate plurals marked with -(V)n involves nouns denoting kinship. Despite obviously belonging to the animate class, most kinship terms build their plural by adding the suffix form -aŋ: (14) Kinship nouns pluralized with -aŋ ām ‘mother’ ōb ‘father’ hɯ’b ‘son’ hu’n ‘daughter’ am-áŋ ob-áŋ hɯb-aŋ hun-aŋ ‘mothers’ ‘fathers’ ‘sons’ ‘daughters’ 8 The nouns bǝtl ‘bubble’ and asl ‘ski’ both end in a syllabic sonorant and thus are disyllables with rising tone on the initial syllable. 320 Edward Vajda Full vowel forms of the kin noun plural suffix are consistently found in other Yeniseian languages, as well, though not necessarily with the same vowel (cf. Yugh améŋ ‘mothers’, ob-éŋ ‘fathers’. Given that this vowel usually draws the pitch peak onto itself in disyllables, the Ket kinship plural suffix -aŋ is probably etymologically distinct from inanimate class -(V)ŋ, which does not normally attract the accent. However, as with every general pattern of Ket noun plural formation, kin nouns also include a few exceptional plural forms. Some nouns that express kinship take the regular animate class suffix -(V)n instead of the kin noun plural suffix -aŋ. Two examples are qīm ‘wife, woman’ → qim-n ‘wives, women’ and tēd ‘husband’ → tad-n ‘husbands’, which do not seem to be treated as primary kin terms, at least as regards the morphology of plural formation. Conversely, a few animate class nouns that are not kinship-related take -aŋ rather than the expected suffix -(V)n: ēs ‘god’, ‘deity’ → esáŋ ‘gods’, ‘deities’; qān ‘khan’ → qanáŋ ‘khans’. The fact that the suffix vowel attracts the pitch peak in the latter two plurals suggests that they share the same suffix -aŋ used by most kinship nouns. Most Ket personal pronouns (shown earlier in Table 1) also use the velar nasal as a plural ending. It should be mentioned here that Ket lacks dyadic kinship forms altogether. Nouns in (15) take the regular inanimate class suffix -(V)ŋ yet trigger masculine animate or feminine animate agreement. This pattern occurs with some animate nouns that do not denote people or animals, and is observed with certain body part nouns, nouns denoting tools or other important objects, and some nouns denoting lower forms of life, which tend to belong to the feminine subclass: (15) Grammatically animate class nouns that pluralize using inanimate class -(V)ŋ Body part nouns that are grammatically (m) masculine or (f) feminine animate bɯ’s ‘penis’ (m) bɯs-aŋ ‘penises’ huud ‘tail’ (f) hud-aŋ ‘tails’ tɯ̄l ‘navel’ (f) tɯl-aŋ ‘navels’ Other inanimate objects that are grammatically (m) masculine or (f) feminine animate quu ‘pole’ (m) quu-ŋ ‘poles’ álal ‘(boat) seat’ (f) álal-aŋ ‘(boat) seats’ Lower forms of life that are grammatically feminine animate (f) biilt ‘martin’ (f) bilt-aŋ ‘martins’ hankó ‘toadstool’ (f) hankó-ŋ ‘toadstools’ ulól ‘leech’ (f) ulól-aŋ ‘leeches’ Although such nouns are grammatically animate as indicated by anaphoric pronoun usage, possessive morphology, and subject or object agreement, most take the Number in Ket (Yeniseian) 321 inanimate class plural suffix -(V)ŋ rather than animate class -(V)n, except where additional phonological factors such as nasal dissimilation play a role (see below, this subsection). In this case, plural formation is governed by referential considerations of animacy, whereas nominal categorization as reflected in agreement patterns is determined on a lexical basis. There is also a tendency for recent loanwords to take the “inanimate class” plural suffix -(V)ŋ, even where such nouns denote human beings or animals: (16) Examples of animate class loanwords pluralized using -(V)ŋ. bótaj ‘rich man’ (< Russian bogatɨj ‘rich’) bótaj-aŋ ópsa ‘sheep’ (< Russian ovtsa id.) ópsa-ŋ mína ‘pig’ (< Russian svin’ja id.) mína-ŋ ondátɯr ‘muskrat’ (< Russian ondatr id.) ondátɯr-aŋ ‘rich men’ ‘sheep.pl’ ‘pigs’ ‘muskrats’ This pattern seems to have become relatively productive in the last phase of Ket language usage, though exceptions can be found, such as koska ‘cat’ → koska-n ‘cats’, where animate -(V)n pluralizes a 20th century borrowing of Russian koška ‘cat’. In 2008, during a conversation with her mother, fluent speaker V. A. Romanenkova spontaneously used the Russian noun sobolj ‘sable’ in the hybrid Russian/Ket plural form sobolj-aŋ in place of native Ket et-n ‘sables’ with its canonical animate class suffix -(V)n. The common usage of -(V)ŋ to pluralize recent loanwords regardless of their meaning perhaps underscores a more general trend away from the originally fundamental morphosyntactic dichotomy in Yeniseian languages between animate and inanimate class entities. The system of Ket noun plural formation also involves numerous phonological and morphological exceptions that override the underlying triple contrast between generic animate -(V)n vs. kinship -aŋ vs. inanimate -(V)ŋ described above. The most regular among these patterns involves dissimilation of the plural suffix nasal consonant in stems that end in either /n/ or /ŋ/, and sometimes /m/. Noun stems ending in /n/ nearly always take -(V)ŋ even when they belong to the animate class: (17) Pluralization of animate class nouns ending in /n/ bɯ́sten ‘wasp’ bɯ́sten-aŋ tuln ‘lizard’ tuln-eŋ lūn ‘grayling (fish)’ lun-áŋ ‘wasps’ ‘lizard’ ‘graylings’ Exceptions such as kùn ‘wolverine’, kun-en ∼ kunn ‘wolverines’ probably involve the diachronic loss of a final consonant other than /n/ that originally occupied the stem-final position. Through an analogous process of dissimilation, inanimate class noun stems ending in either /ŋ/ or /m/ nearly always require the plural suffix -(V)n even where -(V)ŋ would be expected. 322 Edward Vajda (18) Pluralization of inanimate class noun stems ending in /ŋ/ or /m/ ka’ŋ ‘hole’ kaŋ-en ‘holes’ hǝ’ŋ ‘fishing net’ hǝŋ-en ‘fishing nets’ àŋ ‘rope’ aŋ-en ‘ropes’ qām ‘arrow’ qam-en ‘arrows’ kulém ‘lid’ kulém-n ‘lids’ One exception is bōŋ ‘corpse’ → boŋ-á ŋ ‘corpses’, where dissimilation fails to occur for unknown reasons. This noun belongs to the feminine animate class in terms of agreement, but even if the velar nasal was chosen through semantic influence from the word’s lack of logical animacy, the suffix should still undergo nasal dissimilation to -an, but it does not. Instead, it seems to pattern formally more like a kinship noun, including the syllable-final accent shift. The kin noun plural suffix -aŋ is likewise never affected by nasal dissimilation, and its velar nasal appears even after /m/: ádbam ‘sister-in-law’ → ádbam-aŋ ‘sisters-in-law’. It is also noteworthy that other animate class nouns ending in /m/ preserve the canonical animate class suffix form -(V)n: tēm ‘goose’ → tem-n ‘geese’. The stem coda /m/ has no effect on the form of the animate class plural suffix -(V)n either. A different process, possibly historically a type of assimilation, replaces the inanimate class suffix -(V)ŋ with -(V)n in many though not all nouns with stems ending in the segment /s/: (19) Inanimate class noun stems ending in /s/ pluralized with -(V)n rather than -(V)ŋ kūs ‘ringworm’ kus-en ‘ringworms’ qāks ‘wound’ qaks-en ‘wounds’ ádes ‘iron nail’ ádes-n ‘iron nails’ qóles ‘hoof’ qóles-n ‘hooves’ dǝ́ les ‘willow thicket’ dǝ́ les-n ‘willow thickets’ Other inanimate class nouns with stems ending in /s/, however, form their plurals with the expected velar nasal, sometimes with loss of /s/, rather than by assimilating the nasal plural segment to alveolar /n/: qu’s ‘birchbark tent’ → qu’ŋ ‘birchbark tents’, tɯ’s ‘rock’ → tǝ’ŋ ‘rocks’. Finally, the velar nasal is doubly unexpected in the plural form of the masculine animate class noun ùs ‘birch tree’ → us-eŋ ‘birch trees’ since the animate class suffix -(V)n would have been expected on the grounds of its animacy. Even if inanimate class -(V)ŋ appears in this plural because it does not denote a person or animal, the lack of /ŋ/ to /n/ assimilation after stem-final /s/ remains unaccounted for. The same observations are germane in the case of the plural form of masculine animate class bɯ’s → bɯs-aŋ ‘penises’. The reasons behind all of these phenomena remain at present unelucidated. Number in Ket (Yeniseian) 323 A far more frequent and completely regular departure from the underlying semantic patterning of Yeniseian plural allomorphs affects any noun stem built by adding the universal nominalizing suffix -s. All of these stems are pluralized with the suffix form -in. The vowel /i/ derives from the historic form of the nominalizing suffix *-si, which has eroded to -s in singular stems. Virtually any non-noun form, including postpositional constructions and inflected finite verbs, can be nominalized by adding the word-final suffix -s. The plural of such nominalizations always ends -in, regardless of animacy considerations. (20) Use of -in to pluralize nominalizations in -s a. aq-s aq-s-in rot-nmlz rot-nmlz-pl ‘something rotten’ ‘things that are rotten’ b. d=b-il-bet-in-s 3anim.sbj=inan.obj-pst-makeanim.pl.sbj-nmlz ‘something that they made’ d=b-il-bet-in-s-in 3anim.sbj=inan.obj-pst-makeanim.pl.sbj-nmlz-pl ‘things that they made’ c. ul-di-ŋal-s water-inan-abl-nmlz ‘the one (pulled) from the water’ ul-di-ŋal-s-in water-inan-abl-nmlz-pl ‘the ones (pulled) from the water’ There are no exceptions to this highly productive pattern. In example (23c) the plural suffix -in would be used regardless of whether what emerged from the water was a person, animal, or inanimate object. Other nouns undergo an array of irregular changes when a plural suffix is added. Their full range is too broad to be covered exhaustively here since each exceptional form probably has its own idiosyncratic historical explanation. The difference between the plural vs. singular stem phonology in some of these number pairs is drastic enough to warrant being called partial suppletion. The examples in (21) only minimally represent the gamut of exceptional noun plurals that exist in modern Ket: (21) A few examples of irregular stem changes triggered by plural suffixation uu ‘meadow’ oo-ŋ ‘meadows’ qàj ‘elk’ qii-n ‘elk.pl’ sa’q ‘squirrel’ saa-n ‘squirrels’ qǝ’q ‘corner’ qǝ̄ -n ‘corners’ dāp ‘shoulder’ daa-n ‘shoulders’ dīd [dīˑt] ‘spruce grouse’ dek-ŋ ‘spruce grouses’ taal ‘otter’ tak-ŋ ‘otters’ This brief sample suffices to illustrate that irregular plurals involve not only vowel and coda consonant changes, but sometimes an unexplained form of the nasal plu- 324 Edward Vajda ral suffix, as well. The pair sa’q ‘squirrel’ → saan ‘squirrels’ involves the intervocalic loss of uvular /q/, which occurs elsewhere in Ket phonology; however, there are also plurals where this does not occur: tǝ’q ‘finger’ → tǝqin [tʌʁin] ‘fingers’. In the case of qǝ’q ‘corner’ → qǝ̄ -n ‘corners’, where /q/ does elide, the appearance of -(V)n rather than -(V)ŋ is unexpected since this noun belongs to the inanimate class. The choice of -(V)ŋ rather than -(V)n for animate class taal ‘otter’ → tak-ŋ ‘otters’ and dīd ‘spruce grouse’ → dek-ŋ ‘spruce grouse.pl’ is equally unexpected. A possible explanation is that the choice of suffix allomorph originally followed the general rules, and the surface appearance of velar /ŋ/ in the plurals of ‘otter’ and ‘spruce grouse’ as well as /n/ in ‘corners’ arose through some sort of formerly predictable phonological interaction. A full account of such diachronic changes would require a separate study. In contrast to the relatively large number of cases of partial suppletion among Ket count noun singular and plural pairs, full suppletion is limited to the following two examples: (22) Suppletive number pairs ke’d ‘person’ ōks ‘tree’, ‘pole’ de’ŋ a’q ‘people’ ‘trees’, ‘poles’ The origin of these two full suppletive pairs is at present likewise unexplained.9 Some plural nouns do not show a suffix at all, but instead express plurality entirely through irregular tonal alternations, with or without concomitant vowel ablaut: (23) Tonal or vowel ablaut plurals in Ket kɯ̄l ‘raven’ kǝ́ qin ‘fox’ qópqun ‘cuckoo’ tīb ‘dog’ tōk ‘axe’ ēj ‘tongue’ sēs ‘river’ hās ‘shaman’s drum’ kɯɯl kǝqín qopqún ta’b tòk [tɔ̀ɣ] èj sàs hàs ‘ravens’ ‘foxes’ ‘cuckoo birds’ ‘dogs’ ‘axes’ ‘tongues’ ‘rivers’ ‘shaman’s drums’ 9 Porotova (1990: 48) and later Georg (2007: 100) suggested that de’ŋ ‘people’ represents a contraction of a regular plural ked-eŋ. Castrén (1858: 167) in fact did record the form keädeŋ as the plural of ket ∼ kiet ‘person’, alongside djeŋ ‘people’. However, the modern Ket word for person, which is pronounced [kɛ’t], contains a coda that reflects Proto-Yeniseian *d, whereas the onset /d/ of de’ŋ is a reflex of PY *dj, a different phoneme. In any event, there are no other plausible cases of initial syllable collapse in Ket roots. The origin of this suppletive pair thus remains unexplained. Number in Ket (Yeniseian) 325 There is evidence that at least some of these suffixless exceptions arose when one of the canonical plural suffix forms -(V)n or -(V)ŋ was absorbed into the noun root (for reasons as yet unclear), similar to the historical development of English ablaut plurals like mice or teeth. The cognate plural of ‘river’ in some 18th century word lists representing extinct Yeniseian languages appears with the expected inanimate class suffix -(V)ŋ. This interpretation is also supported by the frequency of falling tone in irregular ablaut plurals, since the Ket falling tone [v̀] generally arose through the loss of a final stem element. Zero affixation or internal flection does not appear to have been an original technique of inflection in Yeniseian languages. Nevertheless, some of these irregular plural forms are shared across the family’s primary branches and undoubtedly existed in Common Yeniseian, showing that irregularities in noun plural formation are deeply rooted in the family. The few exceptional cases where the singular and plural noun form are identical can also be explained as resulting from the elision of a plural suffix. In these stems, however, suffix absorption did not cause tonal or vowel ablaut, again for reasons that are not clear: (24) Number syncretism in a few Southern Ket nouns bǝ’n ‘duck’ bǝ’n sūj ‘mosquito’ sūj tɯ̄t ‘black midge’ tɯ̄t ‘ducks’ ‘mosquitoes’ ‘black midges’ Grammatical number in this small set of nouns is covert, appearing only in the morphosyntax as part of verb agreement or possessive marking (compare the expression of number in English this deer – these deer). The other Ket dialects occasionally retain the original plural suffix in a few of these words, as when Central Ket tɯtn is compared to Southern Ket tɯ̄t, both of which are animate class plurals meaning ‘black midges’. This suggests that number syncretism in such pairs arose from plural suffix erosion. Why the loss of a plural suffix triggered tonal or vowel changes in some stems (23) but not in others (24) requires further investigation of the language’s historical phonology. Though exceptions to almost every general rule of Ket noun plural formation abound, it is still possible to provide a cogent overall description. The list in (25) summarizes the various intersecting patterns and exceptions that are at play in assigning plural markers to Ket count nouns, beginning with lexically marked plurals and ending with the most general patterns. (25) Summary of plural formation techniques in Ket count nouns a. A random assortment of lexically marked exceptions to the rules listed below, which involve suppletion, tonal or vowel ablaut, and number syncretism b. Stems ending in /n/ usually pluralize with -(V)ŋ regardless of meaning c. Inanimate class stems ending in /ŋ/ or /m/ and many ending in /s/ pluralize with -(V)n 326 Edward Vajda d. Other inanimate class nouns and many grammatically animate class nouns denoting body parts, inanimate objects, or some lower forms of life pluralize with -(V)ŋ e. Recent loanwords almost always pluralize with -(V)ŋ regardless of meaning f. Most kinship terms pluralize with -aŋ, which typically attracts the pitch peak g. Most other nouns denoting people or animals pluralize with -(V)n h. S-nominalizations always pluralize with -in regardless of meaning or form Irregularities in Ket noun plural formation offer a wealth of evidence for reconstructing the Proto-Yeniseian consonant inventory, which included more phonemes than modern Ket. Unstable final-stem /l/ that truncates in plural forms like sèl ‘reindeer’ → se’-n ‘reindeer (plural)’, and sa’l ‘crucian fish’ → sa’n ‘crucian fish (plural)’ can be reconstructed as a Proto-Yeniseian rhotic (or retroflex) sound rather than a lateral, as supported by its reflexes in Yugh seħr ‘reindeer’ and sa’r ‘crucian (fish)’. Stable /l/ in nouns like būl ‘leg’ → bul-aŋ ‘legs’ and ool ‘container’ → ol-aŋ ‘covers’ traces back to an original lateral, appearing also as /l/ in the Yugh cognates būl ‘legs’ and o’l ‘cover’. Unstable coda /d/ in Ket, which participates in morphophonemic alternations with velar /k/ and corresponds to /dj/ in Yugh and /r/ in Kott, reflects the retroflex stop *dr in Proto-Yeniseian. One example is Kott (feŋ-)čera ‘(female) spruce grouse’ and Ket dīd, Yugh dīdj ‘spruce grouse’, where the velar nasal in the Ket and Yugh irregular plural form dekŋ ‘spruce grouse.pl’ presumably arose because the alveolar nasal of the animate class suffix *-(V)n interacted with the retroflex stem coda. The foregoing discussion does not do justice to the full variety of irregular noun plurals in Yeniseian. The extensive coverage of dialectal forms in (Porotova 1990: 22–61), as well as the lucid presentations of plural allomorphy in Georg (2007: 92– 101) and Werner (1997b: 96–102), consider many more exceptional plurals than could be analyzed here. The most complete reference for Southern Ket noun plurals is the two-volume Comprehensive dictionary of Ket (Kotorova & Nefedov 2015). Porotova’s dialectal noun dictionary (Porotova 2002) is additionally useful because it includes plurals in Yugh as well as from all three Ket dialects. 2.3.2 Dual number marking and apparent pleonastic plural marking in Ket nouns A diachronic perspective also explains how dual number marking arose in a few Ket nouns, and why other nouns appear to add two consecutive plural suffixes. Many Ket monosyllabic singular noun stems result from contractions of disyllabic root compounds or involve elided final-stem consonants. These changes sometimes have implications for understanding number marking in the modern language. Plural noun forms that seem to contain two consecutive plural suffixes include: Number in Ket (Yeniseian) 327 kū ‘opening / mouth (of object)’ → kuniŋ ‘openings’, dɯ’ ‘hat’ → dɯniŋ ‘hats’, and kǝ́ jka ∼ kǝjá ‘head’ → kǝ́ jkeniŋ (alongside kǝ́ jken) ‘heads’. In the case of kū ‘opening’ and dɯ’ ‘hat’, the stem originally ended in *x, which elided in the bare singular form but was preserved through nasalization before the plural suffix -(V)ŋ. The forms dɯ’ ∼ dɯn- ‘hat’ and kū ∼ kun- ‘opening’ are nothing more than irregular root allomorphs. The noun kǝ́ jka ‘head’, on the other hand, is a compound of *kɯj ‘center / inner’ + *gen ‘brain’, so that kǝ́ jken is the original singular stem, from which the final /n/ was later dropped in the singular form after being reanalyzed as a plural suffix. This yielded the truncated stem kǝjká ‘head’, with kǝ́ jken reanalyzed as a plural form alongside the original plural kǝ́ jkeniŋ ‘heads’. Other apparent examples of plural suffix concatenations listed by Porotova (1990: 155–158) or Werner (1997b: 96–97) can be explained in similar fashion. In a similar way, dual marking developed in a few Ket nouns denoting naturally paired objects. In modern Ket usage the noun meaning ‘sleeve’ has a singular form bánno, a dual form bánno-n, and a plural (three or more) form bánno-n-iŋ. Etymologically, ‘sleeve’ is a compound of *ben ‘double’ + *gowx ‘opening’. Its original structure is better preserved in the Yugh cognate bengóu ‘sleeve’. The same process created ókde ‘one ear’, ókde-n ‘pair of ears’, ógde-n-iŋ ‘(three or more) ears’, where the “dual” /n/ probably developed via nasalization of a fricative final-stem coda (analogous to *gowx ‘opening’). Another example is ólta ‘one testicle’, ólta-ŋ ‘pair of testicles’, ólta-ɣ-in ‘(three or more) testicles’ (< ol ‘covered’ + tǝ’ŋ ‘head’). This did not occur in holtǝ́ ŋ ‘button’ → holtǝ́ kin [hɔltʌ́ɣin] ‘buttons’ (< hol(ad) ‘leather’ + tǝŋ ‘head’), since buttons are not frequently paired objects. Only nouns that happened to denote naturally paired objects innovated the expression of dual number via semantic reanalysis of a nasal final coda as a dual marker. The rise of dual number in the Ket noun meaning ‘eye’, however, remains unexplained: dēs ‘one eye’, dès ‘pair of eyes’, and destáŋ ‘three or more eyes’. Nouns expressing naturally paired objects also developed a means to specially emphasize singular number by adding the modifier qóleb ‘half’: qóleb dès ‘one single eye, one-eyed’ (literally ‘half eye’), qóleb būl ‘one single leg (būl)’, ‘one-legged’. Georg (2007: 92) points out that this pattern, which is shared with Uralic, is likely a calque from one of the surrounding Siberian language families, where it is ancient and widespread. A few remarks should be made here about the grammatical expression of number in determiner phrases consisting of an attributive numeral followed by quantified noun, as this topic is also tangentially relevant to the expression of dual number in Ket. Cardinal numeral words exist in two forms in Yeniseian. Forms used as attributive modifiers of nouns are normally unmarked for grammatical number. After qo’k ‘(animate class) one’ and qūs ‘(inanimate class) one’, nouns appear in their unmarked singular form: qo’k ke’d ‘one person’, qūs qu’s ‘one tent’. After ɯ̄n ‘two’, dōŋ ‘three’, sīk ‘four’, qāk ‘five’, and so forth, the plural form is used: ɯ̄n de’ŋ ‘two people’, dōŋ qu’ŋ ‘three tents’, etc. Porotova (1990: 171) observes, however, that 328 Edward Vajda singular noun forms sometimes appear after ɯ̄n ‘two’, and still less often after dōŋ ‘three’ or sīk ‘four’: ɯ̄n / dōŋ / sīk to’q ‘two / three / four steps’; after numbers five and higher, Ket nouns invariably appear in the expected plural form: qāk / à / òn toq-ŋ … ‘five / six / seven steps’, etc. Also, nouns that have developed a dual form show a tendency to use it after the numeral ‘two’, as in ɯ̄n kǝ́ jken ‘two heads’ when compared to qūs kǝ́ jka ‘one head’ and qāk kǝ́ jkeniŋ ‘five heads’. This usage does not appear to be fully regular, however, as dōŋ kǝ́ jken ‘three heads’ (containing the socalled dual form of the noun) have also been recorded. Because heads are not a naturally paired object, perhaps the original stem kǝ́ jken came to be treated as a sort of paucal form by certain individual speakers. It is also possible that interference from Russian, with its special treatment of using the genitive singular rather than the plural after the numbers two, three and four, may have influenced the rise of such forms. Porotova (1990: 120) also lists a few rare examples where the nominalized form of the numeral is used attributively with a plural noun, leading to multi-site marking in the number phrase, including ɯn-aŋ de’ŋ ‘two people’, with animate plural –aŋ on the numeral, in place of the expected ɯn deŋ ‘two people’, where the numeral form is unmarked. Such cases do not appear to represent canonical usage. These rare examples also violate the typically strong head marking profile of Yeniseian morphosyntax. Only attributive forms of the numeral ‘one’ regularly and canonically reflect the grammatical class of the head noun and thus constitute genuine examples of dependent marking in modern Ket. See, however, Section 3.2 below for a description of two more widespread instances of marking on both modifier and noun in Ket noun and determiner phrases. 2.3.3 Other number-related aspects of Ket noun morphology Pluralia tantum nouns are rare in Ket. One example is kǝ’t ‘children of the same mother’; the stem dɯ̄l ‘child (in general)’ is singular to the formally related forms dɯlkat ∼ dɯlkitn ‘children (in general)’, forming a pair that displays the typical singular/plural number contrast, like nearly all Ket nouns denoting countable objects. Two other possible examples are obáŋ, the plural of ōb ‘father’, and amáŋ, the plural of ām ‘mother’, when used in the secondary meaning of ‘parents, ancestors’. Because the singular nouns ōp and ām can only mean ‘father’ or ‘mother’ and not ‘(generic) parent’ or ‘ancestor’, the plural forms obáŋ and amáŋ when used to mean ‘parents’, ‘ancestors’ could be regarded as pluralia tantum nouns. Mass nouns in Ket are often singularia tantum lacking plural forms, as is typical for many languages with a number distinction in their noun morphology: ūl ‘water’, aal ‘broth’, altǝq ‘mud’, ho’q ‘filth’, na’n ‘bread’, īm ‘pine nuts’, hu’s ‘whortleberries’, tulit ‘red currants’, etc. Many of these words derive count nouns by adding a suffix expressing a quantifiable amount of the given item. Such mass-quantifying Number in Ket (Yeniseian) 329 suffixes express specific amounts, shapes and consistencies. The most productive is the singulative suffix -des ∼ -dis, which attaches to mass nouns to express a countable quantity in the form of a droplet of liquid or a single berry or pine nut. The examples in (26) juxtapose the underlying mass noun with its derived count noun and its plural form: (26) Count nouns derived with the mass-quantifying suffix -des ∼ -dis ūl ‘water’, ‘rain’ úldis ‘droplet’, ‘raindrop’ (plural: úldisn) īm ‘pine nuts’ ímdis ‘(single) pine nut’ (plural: ímdisn) eel ‘lingonberries’ éldis ‘(single) lingonberry’ (plural: éldisn) óŋniŋ ‘roe’ óŋndis ‘(single) fish egg’ (plural: óŋndisn) The meaning of a single droplet or single discrete berry, nut, or grain of roe accrues from the suffix’s etymological origin in the anatomical noun root dēs ‘eye’, which accounts for why the countable items it expresses are normally roundish is shape. Only rarely is this suffix used to convert a mass noun into a countable object of radically different shape, perhaps the sole example being daan ‘grass’ → daandis ‘blade of grass (plural: dandisn). There are a number of other, less productive mass-quantifying suffixes in Ket, all derived from roots or combinations of roots. The noun lamtǝ ∼ lamt ‘piece’ can be added to Ket mass nouns to convey a congealed lump or broken off piece, thus deriving another set of countable noun stems: (27) Count nouns derived by adding the suffix -lamt ‘piece’ to mass noun stems kɯ̄d ‘fat’ kɯtlámt ‘lump of fat’ (plural: kɯtlámtaŋ) sūl ‘blood’ sullámt ‘blood clot’ (plural: sullámtaŋ) na’n ‘bread’ nanlámt ‘piece of bread’ (plural: nanlámtaŋ) ōks ‘wood’ okslámt ‘(small) piece of wood’ (plural: okslámtaŋ) qō ‘ice’ qoklámt ‘(small) chunk of ice’ (plural: qoklámtaŋ) Yugh uses the suffix -lap in a similar way: sur-láp ‘blood clot’ (plural sur-láf-ɯn). The Ket noun dápqul ‘heap’, which originated via metathesis involving the roots *daq ‘put / laid down’ + *pɯl ‘growth, mass, quantity’, is added to a few mass nouns to denote countable larger pieces: qokdápqul ‘(large) chunk of ice’ (plural: qokdápqulaŋ). Finally, the Ket suffix -les, of unknown etymology, derives count nouns expressing pieces of flat flexible objects: si’k ‘rawhide’ → síkles ‘piece of rawhide’ (plural sikles-n). Some of these Yeniseian mass-quantifying suffixes have been called ‘singulatives’ (Porotova 1990: 65). Note, however, that they are derivational suffixes and cannot be regarded as grammatical markers of singular number because they remain when the stems they derive are inflected for plural meaning. In this connection, the count noun óŋndis ‘(single) fish egg’, which is derived from óŋniŋ ‘roe’ is 330 Edward Vajda noteworthy because it involves the truncation of the final-stem segments -iŋ, probably via reanalysis as a plural marker. The same process also seems to be at play in hǝ́ naŋdis ∼ hǝ́ ndis ‘grain of sand’, where the final portion of the stem hǝ́ naŋ ‘sand’ is optionally deleted, probably due to its resemblance to a plural marker. The mass noun hǝ́ naŋ ‘sand’ also unusually allows pluralization as hǝnaŋ-an in the meaning ‘expanses of sand’, as recorded by Werner (2002, vol 1: 338). Still, the mass-quantifying ‘singulative’ suffix -dis in hǝ́ naŋdis ∼ hǝ́ ndis ‘grain of sand’ cannot be viewed as a grammatical marker of singular number, since it is compatible with the addition of a plural suffix: hǝ́ naŋdis-n There is some evidence, however, that a genuine grammatical singulative suffix once operated in Yeniseian. Helimski (2016) identifies the /s/ element at the end of several noun stems as a fossilized singulative. This interpretation is supported by the fact that the element in question does disappear when a plural suffix is added. (28) Possible examples of a fossilized singular marker (the “S-singulative”) in Ket tɯ’s ‘stone’, ‘rock’ tǝ’ŋ ‘stones’, ‘rocks’ qu’s ‘birchbark tent’, also ‘house’ qu’ŋ ‘birchbark tents’ qe’s ‘sandbar (in river)’ qedeŋ ‘sandbars’ ōks ‘tree’ a’q ‘trees’ On the other hand, the disappearance of word-final /s/ in these forms might instead be interpreted as simply another irregular stem change in Yeniseian plural formation, with the elision of /s/ due to some as yet unexplained morphophonemic process. In particular, the forms for ‘tree’ are suppletive, which complicates any attempt to argue that /s/ in the singular form is a morpheme separable from the rest of the word. In any event, none of these forms can be used as mass nouns by deleting the final /s/; for instance, there is no bare form like [tɯ’] that means ‘rock’ as a generic, uncountable mass. The existence of an S-singulative in an earlier stage of Yeniseian, along with its possible etymological connection to the highly productive nominalizing suffix *-si, remains plausible but unproven. Solving this problem would require additional diachronic investigation of irregular noun plural morphology. Ket nouns normally have only one plural suffix, with apparent cases of consecutive nasal suffixes explainable as resulting from the loss of a final stem nasal from what was originally the full singular stem. However, genuine pleonastic (multi-site) marking of plurality does occur in some noun + noun compounds made through coalescence of two formerly independent word forms. A few examples, adapted from Porotova (1990: 158), are shown in (29), with hyphens inserted to show the morpheme breakdown: (29) Noun compounds with multi-site number marking búl-dal ‘knee tendon’ < būl ‘leg’ + da’l ‘tendon’ → plural: búl-aŋ-dál-eŋ ál-bes ‘waist-length fur coat’ < āl ‘half’ + be’s ‘rabbit’ → plural: ál-aŋ-bés-n sés-tas ‘suede boot’ < sàs ‘suede’ + tès ‘boot’ → plural: sás-eŋ-tés-aŋ Number in Ket (Yeniseian) 331 The first element in such compounds modifies the second to express a contiguitybased relationship involving parts and wholes. Their internal structure reflects the general syntactic pattern of modifier + head required for all Yeniseian noun and determiner phrases. 2.4 Verbal number The polysynthetic Ket finite verb has a diverse arsenal of morphological means for expressing pluractionality. Only a few examples will be examined here. Many stems use suppletion (full or partial) of the base (a term used by Ketologists to refer to the stem’s rightmost lexical morpheme) in order to signal a contrast between a single event and multiple events affecting multiple entities separately: (30) Base suppletion expressing types of pluractionality a. d-aŋ-i-b-to 1sbj-hanging-pres-3inan.obj-put.once ‘I hang it up.’ ‘I hang them up (all together, in one action)’ b. d-aŋ-i-b-uk 1sbj-hanging-pres-3inan.obj-put.more.than.once ‘I hang them up (two or more separately, perhaps one after the other).’ c. t-a-b-qut down-pres-3inan.sbj-one.lies ‘It lies (there).’ d. t-a-b-damin down-pres-3inan.sbj-many.lie ‘They (many things) lie (there).’ The form (30b) d-aŋ-i-b-uk ‘I hang them up’ expresses multiple events, while the form (30a) d-aŋ-i-b-to ‘I hang it up’ expresses a single event and could also mean ‘I hang them up’ in cases where a handful of inanimate objects are simultaneously placed in a hanging position. The form (30d) t-a-b-damin ‘They (objects) lie there’ denotes a single situation, but with emphasis on more than one object lying in different places. This form is also interesting in that its composite base, -damin, seems to contain a fossilized plural suffix -in, which in all other types of finite verb forms correlates with animate class subjects only. Along with base suppletion, the pluractional (pltc) prefix d- sometimes appears in multiple-action stems. 332 Edward Vajda (31) Example of (a) single- and (b) multiple-action pair involving pluractional da. d-es-a-b-daq 1sbj-surface-pres-3anim.pl.obj-put.once ‘I lay it down.’ / ‘I lay them down (objects placed together in one action).’ b. da=d-a-b-da 3 f.sbj=plct-pres-3anim.pl.obj-put.many ‘I lay it down many times.’ / ‘I lay them down (objects, one after another).’ In another pair, the pluractional (plct) prefix n- distinguishes the multiple-action stem, in addition to base morpheme suppletion: (32) a. d-aŋ-baq 1sbj-3anim.pl.obj-give.once ‘I give them (something) once.’ b. d-aŋ-n-bu 1sbj-3anim.pl.obj-plct-give.many.times ‘I give them (something) many times.’ The pluractional prefix n- is unique to this pair, while d-, though also largely unproductive, is found in many stems, where it sometimes surfaces as t- through phonological merger with adjacent morphemes. Some verbs express pluractionality by using two pluractional affixes as well as a suppletion of the morpheme: (33) a. da=don-ba-h-ol-ted 3 f.sbj=knife-1sg.obj-area-pst-hit.endwise ‘She stabbed me (once).’ b. da=don-an-ba-t-ol-do 3 f.sbj=knife-pl-1sg.obj-plct.area-pst-gouge ‘She kept stabbing me.’ The pluractional suffix form /t/ in (33b) derives from d- merged with the area prefix h-, which remains visible in the single-action form in (33a). Notice also that the plural suffix -an on the incorporated instrument noun do’n ‘knife’ is a pluractional marker in that it expresses multiples acts of stabbing (done by one person or by many people) rather than stabbing with multiple knives. The noun plural suffix in this mildly productive verb stem pattern therefore marks pluractionality in conjunction with the addition of the pluractional prefix d- as well as base morpheme suppletion – though both -ted ‘hit endwise’ and -do ‘cut, gouge, hew’ when used on their own can refer to either single or multiple actions. Number in Ket (Yeniseian) 333 Just as multi-site grammatical agreement is not infrequent in Ket, the lexical category of pluractionality is also frequently expressed using combinations of multiple morphemes in the same verb form (31–33). There is no evidence that Ket pluractionality involves a morphological distinction between a small number events (two or three) and a larger number. 3 Agreement and syntax of number The marking of grammatical categories on modifiers is generally atypical for Yeniseian, being limited to a few types of noun or determiner phrases. As mentioned in 2.2.3, attributive forms of the numeral ‘one’ regularly agree with their head noun in animacy but not gender: animate class qo’k qīm ‘one woman’ and qo’k hīk ‘one man’ vs. inanimate class qūs tɯ’s ‘one rock’. Marking of plural number on both the numeral form and the noun has been documented sporadically in a few phrases with higher numbers like ɯn-aŋ de’ŋ ‘two people’ in place of the canonical ɯ̄n de’ŋ ‘two people’, which shows the typical pattern of plural marking on the noun only. This section examines two additional instances where number is marked on modifiers as well as on the noun. In the first instance, examined in 3.1 below, demonstrative pronouns are grammatically marked for noun class (both animacy and gender) as well as animate plural number, while the nouns they modify are marked only for plural number, but not for class. In the second, a small minority of adjectives and action nominals (the Ket equivalent of participles) take a plural suffix that distinguishes them from their singular forms (3.2). There are no other instances in Ket nominal morphology where modifiers take grammatical affixes. Case suffixes in noun or determiner phrases attach to the head noun only and never appear on modifying words. Adjectives and adverbs that undergo S-nominalization can take case inflections, but only as the head of a noun phrase. Finally, subsection 3.3 examines number marking in the subject complements of locative and existential clauses. 3.1 Class and number agreement in demonstrative pronouns The most frequently encountered and straightforward exception to the general pattern of Ket head marking appears in determiner phrases, where demonstrative pronouns agree in class and number with their head noun. Ket demonstrative roots express a three-way contrast between proximal ki- (close to speaker), medial tu(farther from speaker but earlier mentioned or in the general vicinity) and distal qa(distant from speech situation). When modifying a singular masculine animate class noun, all three demonstrative roots take the suffix -d (pronounced as the rhotic flap [ɾ] in southern Ket). When modifying a feminine singular animate class or inanimate 334 Edward Vajda Tab. 3: Class and number agreement forms of demonstrative pronouns. -d kīd / tūd / qād hīk ‘this / that (near) / that (far) man’ -de kide / tude / qade qīm ‘this / that (near) / that (far) woman’ kide / tude / qade qu’s ‘this / that (near) / that (far) tent’ -ne kine / tune / qane hikn ‘these / those (near) / those (far) men’ kine / tune / qane qimn ‘these / those (near) / those (far) women’ -de kide / tude / qade qu’ŋ ‘these / those (near) / those (far) tents’ class noun of either number, they take the suffix -de (pronounced [ɾɛ] in Southern Ket). When modifying any plural animate class noun, they take the suffix -ne. Plural marking in demonstratives that modify animate class nouns represents another case where grammatical marking appears on the modifier as well as its head noun. Inflection for class (masculine animate, feminine animate, and inanimate) in singular forms of demonstrative modifiers constitutes dependent marking, however, since these categories are covert in the head nouns themselves. Vajda (2013a: 87–88) argued that the agreement suffixes in modern Ket demonstrative pronouns are vestiges of ancient possessive morphology, which explains why they exactly mirror the grammatical categories expressed by 3rd person possessive clitics. 3.2 Plural marking in adjective and action nominal forms While most Ket adjectives keep the same form regardless of whether they apply to a single or plural entity, about a dozen mark plurality by adding the suffix -(V)ŋ (Bibikova 1976: 91). Although this suffix, with its epenthetic vowel and velar nasal, resembles that used on inanimate class plural nouns, these exceptional adjective plural forms modify inanimate as well as animate class nouns, as can be seen in these attributive phrases. Example (34) provides an exhaustive list of Ket adjective forms that typically occur in different singular and plural forms: (34) Examples of plural agreement in adjectives qà ke’t ‘big person’ qē-ŋ de’ŋ ugd būl ‘long leg’ ugd-eŋ bulaŋ ho’l qa’d ‘short coat’ hol-aŋ qadaŋ bo’l dɯlkit ‘fat child’ bol-aŋ dɯlkitn ka’t ke’d ‘old person’ kat-aŋ de’ŋ sīn ōks ‘old / rotten tree’ sin-aŋ a’q ‘big people’ ‘long legs’ ‘short coats’ ‘fat children’ ‘old people’ [human age] ‘old / rotten trees’ [objects] Number in Ket (Yeniseian) dǝ́ qta qɯ̄t uul la’m ǝǝl hīk èt ko’p ēt ìt bɯ̄d àŋ to’t de’ ‘fast wolf’ ‘smooth board’ ‘unmarried man’ ‘living chipmunk’ ‘sharp tooth’ ‘strong rope’ ‘shallow lake’ dǝ́ qta-ŋ qǝtn uul-aŋ lemiŋ ǝǝl-aŋ hikn et-iŋ koon et-iŋ iteŋ bɯd-eŋ aŋen tot-iŋ dēŋ 335 ‘fast wolves’ ‘smooth boards’ ‘unmarried men’ ‘living chipmunks’ ‘sharp teeth’ ‘strong ropes’ ‘shallow lakes’ Using internal reconstruction, Vajda (2013b: 20–22) demonstrated that the element -(V)ŋ in these words was originally an adjective-deriving suffix that was reanalyzed as a marker of plurality through false analogy with the common noun plural suffix -(V)ŋ. For reasons that are not entirely clear, some adjectives absorbed the suffix into their roots in plural as well as singular forms, which explains the unusually high prevalence of falling tone in monosyllabic Ket adjectives. Other adjectives retain the suffix in all contexts, suggesting it was originally part of the stem rather than a plural marker. Two examples of the latter type of adjective stems are Ket sokŋ ‘thick’, údokŋ ‘lazy’, which modify either singular or plural nouns: údokŋ ke’t ‘lazy person’, údokŋ de’ŋ ‘lazy people’. Other adjectives show free variation between forms with or without the suffix. One example is Yugh súrbes ∼ surbèːħs ∼ súrbesiŋ ‘red’ < *sū r ‘blood’ + *wes ‘resemble’ + *(V)ŋ ‘adj suffix’, with any one of these variants capable of modifying either singular or plural nouns. The Ket cognate súlem ‘red’, which shows a more radical reduction, also occurs with either singular or plural nouns – súlem tɯ’s ‘red stone’, súlem tǝ’ŋ ‘red stones’, providing yet another indication that the final nasal element was originally part of the stem. Recognizing this semantic reanalysis of the adjective derivational suffix *-(V)ŋ, which appears to be an innovation that is limited to the Ket-Yugh branch of the family, makes it unnecessary to posit adjective number agreement as an original feature of Yeniseian morphology and also explains the origin of this highly idiosyncratic trait in a strongly head marking language like Ket. An identical process affected the stem phonology of action nominals, a lexical class of modifiers that serves a participial as well as infinitival function in Ket. Action nominal derivation originally involved attaching a nasal suffix to a finite verb root or complex lexical stem. This suffix is homonymous with the adjective-deriving suffix -(V)ŋ and may be etymologically identical to it; in any event, both adjectival and action nominal -(V)ŋ underwent a parallel evolution in the historical development of Ket. In some action nominals this suffix remained unchanged in all forms: bágdeŋ ‘pulling’, táseŋ ‘getting up’, ániŋ ‘playing’, qódeŋ ‘dying’, íliŋ ‘eating’, éjiŋ ‘going’, háleŋ ‘wrapping up’, etc. In other stems it was absorbed into the root, usually creating falling tone: bǝ̀ k [bʌ̀ɣ] ‘finding’, tàd [tàɾ] ‘beating’, tòs ‘raising’, kàl ‘fighting’, dàq [dàʁ] ‘laughing’, dùd [dùɾ] ‘burning’, èj ‘killing’, etc. Less often, absorption of the suffix into the verb root yields another tone: kīt ‘rubbing’, ǝ̄ n ‘cooking’, kǝ’j ‘going hunting’, dǝ’q ‘living’, etc. A full explanation for why the action nominal and 336 Edward Vajda adjective suffix was sometimes retained and sometimes absorbed into the stem would require a detailed excursion into diachronic phonology, similar to what would be required to explain the origin of exceptional ablaut noun plurals. Paralleling the development of a number contrast in a few adjective stems, this suffix was similarly reanalyzed as a marker of plurality or pluractionality in a minority of action nominal stems. This has resulted in a sporadic contrast between a contracted form used with reference to single actions or singular entities, and a suffix-augmented form that refers to multiple actions or modifies plural nouns. Example (35) shows three Ket action nominals that typically use distinct singular and plural forms. (35) Number marking in action nominals suulbèd ‘sled-making (one event)’ suulbed-eŋ ‘sled-making (in general)’ tǝǝl i’ ‘freezing day’ tǝl-iŋ ekŋ ‘freezing days’ ūs i’ ‘warm day’10 us-eŋ ekŋ ‘warm days’ The same subset of adjective and action nominal stems that show a number contrast when used as modifiers in noun phrases carries this opposition into the stem morphology of nominalizations. Example (36) shows compound words containing adjective suffixes that have been reanalyzed as plural markers: (36) a. noun phrase bol-eŋ lon-eŋ thick-pl lip-pl ‘thick lips’ b. noun phrase ukd-eŋ bul-aŋ long-pl leg-pl ‘long legs’ compound adjective bol-eŋ-lon-eŋ ke’d thick-pl-lip-pl person ‘thick-lipped person’ S-nominalization bol-eŋ-lon-eŋ-s thick-pl-lip-pl-nmlz plural of S-nominalization bol-eŋ-lon-eŋ-s-in thick-pl-lip-pl-nmlz-pl compound adjective ukd-eŋ-bul-aŋ assel long-pl-leg-pl animal ‘long-legged animal’ S-nominalization ukd-eŋ-bul-aŋ-s long-pl-leg-pl-nmlz plural of S-nominalization ukd-eŋ-bul-aŋ-s-in long-pl-leg-pl-nmlz-pl ‘one with long-legs’ ‘ones with long legs’ ‘ones with thick lips’ ‘ones with thick lips’ In complex nouns or noun phrases containing more than one plural marker, it is the number category at the noun’s rightmost edge that signals whether a singular or plural entity is being referred to. The retention of this type of multi-site number marking in complex stems, however, is highly lexicalized, as can be seen from the forms in (37): (37) singular phrase ukd a’d long bone long bone plural phrase ukd-eŋ ad-eŋ long-pl bone-pl ‘long bones’ S-nominalization ukd-ad-s long-bone-nmlz ‘one with big bones’ plural of S-nominalization ukd-ad-s-in long-bone-nmlz-pl ‘ones with big bones’ 10 Though unclear from the English translation ‘warm day’, the morpheme us is a verb root that heads finite forms like b-il-us ‘it turned warm’, in addition to appearing in the action nominal ūs ∼ useŋ ‘turning / being warm’. Number in Ket (Yeniseian) 337 Here the use of plural marking in both modifier and noun fails to carry through to the complex stem form nominalized by -s, in contrast to the presence of precisely two lips or precisely two (or four) legs. The NP ‘long bone’ therefore acts more like a mass noun in the complex nominalizations shown in (37). At the same time, note the lack of plural adjective form in ukde-hǝŋn-eŋ ke’d ‘long-armed man’, a compound recorded by Werner (2002, vol. 2: 322). This type of variation probably follows from the originally derivational nature of the adjective suffix -(V)ŋ, whose grammatical function of marking plural agreement developed only later through semantic reanalysis and remained sporadic. This secondary singular/plural distinction in the same subset of adjective and action nominal stems also shows up when these forms are incorporated into finite verbs (3.4.1) or are used as subject complements in copular clauses (3.3). 3.3 Number marking in the subject complements of linking verb clauses While most Ket clauses are headed by a finite verb, some copular clauses contain a subject complement and no tense-bearing verb form at all. The predicates in clauses of this type have three different kinds of subject complements: 1) a bare noun or noun phrase; 2) an S-nominalization derived from another part of speech; or 3) a qualitative adjective or locational stem inflected with a subject agreement suffix. Bare nouns used as subject complements may be singular or plural in form but lack agreement in person with their clausal subject:11 (38) Nouns used predicatively as subject complements a. tu-de tɯ’s that-inan rock ‘That is a rock’ b. tu-de tǝ’-ŋ that-inan rocks-pl ‘Those are rocks.’ c. tū-d qòj ben ke’d that-m bear really person ‘That bear is really human.’ 11 During his 19th century fieldwork with Ket and Yugh speakers, Castrén (1858: 100–103) recorded noun forms like uob-di ‘I am a father’ and uob-du ‘He is a father’ with subject concord suffixes, but such forms are no longer used in any of the three modern Ket dialects (Werner 1997b: 306). 338 Edward Vajda d. ǝtn de’ŋ we people ‘We are people.’ The second morphological type of subject complement in copular clauses with no finite verb form is the S-nominalization, which can be derived from any non-noun. Used predicatively, S-nominalizations regularly show number agreement with the clause subject; however, similar to bare nouns they lack person agreement. Adjectives like qà ∼ qēŋ (or qàŋ) ‘big’ that display a morphological number contrast when used attributively, maintain this formal contrast when nominalized by -s: (39) S-nominalizations used predicatively as subject complements in copular clauses a. tɯ’s qà-s rock big-nmlz ‘The rock is big.’ b. tɯ’-ŋ qe-ŋ-s-in rock-pl big-pl-nmlz-pl ‘The rocks are big.’ c. ke’d qà-s person big-nmlz ‘The person is big.’ d. de’ŋ qe-ŋ-s-in people big-pl-nmlz-pl ‘The people are big.’ S-nominalizations show plural agreement with inanimate as well as animate class subjects. In addition, forms like qe-ŋ-s-in ‘they (either animate or inanimate class) are big’ display multi-site plural marking, while S-nominalizations derived from non-variable stem adjectives (the majority of forms) mark plurality word-finally only: ke’d udokŋ-s ‘the person is lazy’, de’ŋ udokŋ-s-in ‘the people are lazy’, at aqta-s ‘I am good’, ǝtn aqta-s-in ‘we are good’. The third type of non-finite predicate contains an adjective or locational adverb that takes a suffix agreeing in person, number, and class (animacy and gender) with the clause subject, as shown in (44): (40) Concord suffixes on predicate adjectives used as subject complements a. tɯ’s qa-am [qaɣam] rock big-inan.sbj ‘The rock is big.’ Number in Ket (Yeniseian) 339 b. tǝ’-ŋ qe-ŋ-am rock-pl big-pl-inan.sbj ‘The rocks are big.’ c. at qa-di I big-1sg.sbj ‘I am big.’ d. ǝtn qe-ŋ-dǝŋ we big-pl-1pl.sbj ‘We are big.’ In contrast to S-nominalizations, there is no singular/plural differentiation in the inanimate class forms, except in the minority of stems where adjectival -ŋ has been reanalyzed as a plural marker. The concord suffixes that appear on predicate adjectives, locational adverbs, or action nominals12 used as subject complements mark person and class, in addition to number. Just as in other realms of Yeniseian morphology, the masculine and feminine animate singular classes merge into a single animate plural class. The examples in Table 4 show the full array of predicate concord suffixes on Ket and Yugh forms of the adverb *qapǝ, which originally meant something like ‘inside one’s tent’.13 The Yugh forms are adapted from Werner (1997a: 209). Vajda (2019: 68) argued that modifying words with predicate concord suffixes were originally finite verbs constructed from an incorporated modifier followed by a Tab. 4: Paradigm comparing Ket and Yugh predicate concord suffix forms. 1sg ‘I am at home / in my tent.’ 2sg ‘You are …’ 3m.sg ‘He is …’ 3 f.sg ‘She is …’ 1pl ‘We are …’ 2pl ‘You (all) are …’ 3anim.pl ‘They (people, animals) are …’ 3inan ‘It is …’ ‘They (things) are …’ Southern Ket Yugh qaˑ-di qaˑ-ku qaˑ-du qaˑ-dǝ qaˑ-dǝŋ qaˑ-kǝŋ qaɣ-aŋ qaɣ-am χaˑb-di’ χaˑp-ku’ χaˑb-du’ χaˑb-da’ χaˑb-dǝːħŋ χaˑp-kǝːħŋ χaf-eːħŋ χaf-e’ 12 Subject concord suffixes on action nominals express that the subject is able to perform the given action’: il-di ‘I can sing’ (from i’l ‘singing’, ‘to sing’). 13 The phonemic form of this adverb in modern Ket is /qap/, which could also be transcribed as /qah/, since word-final /p/ and word-initial /h/ represent a single phoneme in the transcription used in this article. The forms in (45) show the surface allomorphs [qaˑ] and [qaɣ] rather than the underlying phonemic form. 340 Edward Vajda subject agreement prefix on an archaic word-final verb root meaning ‘be’ or ‘having become’, probably shaped something like *eŋ. Phonological reductions in this agreement prefix + verb root sequence produced the modern concord suffix forms, in which the original verb root was reanalyzed as a nasal plural agreement suffix. This interpretation also explains why predicate concord in modern Ket is suffixing, whereas subject and object agreement in finite verbs (described in the next section) follows a mostly prefixing template. If subject complements with predicate concord do indeed derive etymologically from prefixed finite verb forms, their adjective or locational adverb portions were once incorporated modifiers exactly like those still found in many finite verb forms today. 3.4 Number marking in finite verb forms Ket verb morphology is based on a discontinuous stem that allows the incorporation of certain types of modifying elements (Vajda 2017a). Agreement affixes interdigitated between the stem’s lexical morphemes index the subject and direct object arguments. Other affixes in the verb complex distinguish past and non-past tense. Finite verb clauses in Ket are exclusively head marking. While subjects and direct objects are indexed verb-internally, any noun or pronoun arguments appearing in the finite verb clause remain unmarked for grammatical relations. Yeniseian nominal morphology lacks any type of nominative, accusative, absolutive or ergative case markers altogether. As earlier described with reference to possessive clitics and other morphologies evolved from them, plural number agreement in finite verbs is grammatically marked only with reference to animate class subjects or objects. Inanimate class subjects or objects trigger the same affixal form regardless of whether they are singular or plural. The discussion below skirts around much of the verb’s internal polysynthetic complexity to concentrate on features relevant to the topic of number marking. Subsection 3.4.1 examines how the finite verb template expresses grammatical subject and object number agreement. The main point is that plural agreement is normally limited to animate class subjects and objects. Subsection 3.4.2 draws on the description given earlier in 3.2 of how number marking in adjectives and action nominals arose via reanalysis in order to examine features of verb morphology originally unconnected with number that later acquired plural or pluractional meaning based on coincidental homonymy with the common noun plural suffix -(V)ŋ. Plural markers that arose in Ket via reanalysis of derivational or aspectual affixes differ from the inherited Yeniseian agreement system in occasionally reflecting the plurality of inanimate class subjects. 3.4.1 Verb-internal class, person, and number concord Subject/object concord in finite verb forms is accomplished using verb-internal affixes that express the same grammatical categories of agreement in person, class 341 Number in Ket (Yeniseian) and number found in possessive constructions. Intransitive verb forms normally require subject agreement, while transitive verbs agree with their subject and direct object. The following template underlies every Ket finite verb form: Tab. 5: Ket finite verb template, with agreement slots shaded.14 8 7 sbj incorperson porated noun, modifier, or action nominal 6 5 4/3 2 1 0 -1 obj or sbj thematic consonant prefix(es) 3 sbj or obj (originally in slot 4) now fused with tensemood aspect (n ∼ l) sbj or obj BASE (verb root or lexical aspect marker) anim pl sbj suffix conjugation marker (s/i ∼ a/o) (originally in slot 3) Each stem lexically chooses which of the shaded slots must be filled to express subject or object agreement. No verb form fills all of the template’s five (shaded) agreement slots, though multi-site subject marking involving two or even three of these slots filled simultaneously in single verb forms is not uncommon. There are five productive intransitive agreement configurations and three productive transitive configurations, in addition to a smattering of unproductive types – a system first explained in Vajda (2001) and Vajda (2004: 44–76) and now most succinctly described in Nefedov & Vajda (2015: 38–48). Vajda (2017b) explains in detail the diachronic processes that gave rise to this typologically unusual system. Table 6 shows basic allomorphs found in each agreement slot.15 Of the positions left blank in Table 6, slots 7, 5, and 0 contain lexical morphemes, while slot 2 is used to mark tense, mood and aspect. Slots 6, 4, and 1 are used variously to express class, person, and number concord. By contrast, the wordinitial marker in slot 8 expresses person and class agreement only, since it works in tandem with the template’s only productive suffix, located word-finally, which expresses plurality of any animate class subject in verbs that use prefixal slot 8 for subject person and class agreement: 14 Subject complements with predicate concord suffixes of the type examined in 3.3 above, are composed of morphemes that once occupied slots 7-1-0, which reflects an intransitive configuration still productive today for several other kinds of finite verb stems. 15 See Nefedov & Vajda (2015: 38–40) for more detail on agreement marker allomorphy, including rules for when P8 markers surface as prefixes and when they surface as clitics. 342 Edward Vajda Tab. 6: Agreement marking series in the Ket finite verb template. 8 7 sbj di- 1person ku- 2person du- 3m.sg dǝ- 3f.sg 6 5 obj or sbj ba- ∼ bo- 1sg ku- 2sg a- ∼ o- 3m.sg i- ∼ u- 3f.sg Ø- ∼ u- 3inan dǝŋ- 1pl ku- 2pl aŋ- ∼ oŋ3anim.pl 4/3 2 3 sbj or obj (d)a- 3m.sg16 (d)i- 3f.sg (d)aŋ- 3anim.pl b- 3inan 1 sbj or obj di- 1sg ku- 2sg ǝ- 3anim.sg or 3inan dǝŋ- 1pl kǝŋ- 2pl ǝŋ- 3anim.pl 0 –1 anim pl sbj suffix -in (41) a. di-k-a-qut [diɣaʁut] 1sbj-up-pres-walk ‘I ascend / go upward.’ b. di-k-a-qut-n 1sbj-up-pres-walk-anim.pl.sbj ‘We ascend / go upward.’ c. ku-k-a-qut 2sbj-up-pres-walk ‘You.sg ascend / go upward.’ d. ku-k-a-qut-n 2sbj-up-pres-walk-anim.pl.sbj ‘You.pl ascend / go upward.’ e. du-k-a-qut 3sbj-up-pres-walk ‘He ascends / goes upward.’ f. du-k-a-qut-n 3sbj-up-pres-walk-anim.pl.sbj ‘They (animate, either gender) ascend / go upward.’ g. dǝ-k-a-qut 3f.sg.sbj-up-pres-walk ‘She ascends / goes upward.’ 16 The prefixes d- (anim) and b- (inan) once filled slot 4. Today b- fills slot 3, having metathesized rightward ahead of the a-conjugation marker. The innovative Ket/Yugh gender contrast between masc.sg a- and fem.sg i- has eclipsed generic anim d-, which survives vestigially in only a small minority of forms. Number in Ket (Yeniseian) 343 In this common intransitive agreement pattern, the marker P8 dǝ- ∼ da= indexes 3rd person singular feminine animate class subjects.17 It is never used to index animate plural subjects, even those specifically denoting groups of females. Instead, the prefix du- appears in verb forms with animate plural subjects. The position 8 marker du- also indexes a 3rd person masculine animate subject. The rest of this subsection examines number marking differences between animate and inanimate class arguments, as well as several cases of multi-site animate plural agreement. Animacy plays just a pervasive role in Ket finite verb as elsewhere in Ket morphology. Despite this fact, Yeniseian lacks true semantic (active) alignment: roughly synonymous verbs often require different agreement configurations, while active and inactive intransitives with completely different semantics often follow an identical agreement pattern. Animate singular nouns display a formal dichotomy between masculine and feminine subclass, as has already been observed in predicate concord suffixes and other areas of the morphosyntax. Inanimate class subjects and objects, on the other hand, require a mostly different set of agreement markers that lack any number contrast between singular and plural. Finally, it is worth mentioning that inanimate class entities rarely serve as the subjects of transitive verbs in Ket. The sentences in (42) show examples of verb-internal subject concord in clauses containing arguments of different noun classes and grammatical numbers. The verb stem in these examples is a linking verb that incorporates its subject complement and uses slot 6 to express subject agreement: (42) Intransitive verbs showing 3rd person animate class number agreement a. qīm baam-i-tonoq woman old.woman-3 f.sbj-became ‘The woman became (very) old.’ b. hīk baat-a-tonoq man old.man-3m.sbj-became ‘The man became old.’ c. qim-n baam-aŋ-aŋ-tonoq woman-pl old.woman-pl-3anim.pl.sbj-became ‘The women became old.’ d. hik-n baat-aŋ-aŋ-tonoq man-pl old.man-pl-3anim.pl.sbj-became ‘The men became old.’ 17 The P8 marker dǝ- ∼ da= is also used to index an inanimate class subject in the few verbs of this agreement configuration that allow inanimate subjects, such as qām hīk da=kasonam ‘an arrow took [= killed] the man’. 344 Edward Vajda Just as in possessive morphology, the masculine and feminine animate distinction disappears in the plural: the same animate class plural affix -aŋ is used for plural nouns that denote males as well as females. This set of forms also illustrates how nouns that fulfill the role of subject complement retain their logical expression of number, which in plural subject forms like (42c) and (42d) produces yet another example of multi-site plural marking in Ket. Inanimate class subjects and objects, by contrast, show no grammatical number distinction, as shown in (43d) which sets them apart from the rich array of animate class object agreement markers in (43a), (43b), and (43c): (43) Class and number marking of objects in one set of Ket transitive verb forms a. ku-a-ted [kuɣatɛt] 2sbj-3m.obj-hit ‘You hit him.’ b. ku-i-ted [kuˑtɛt] 2sbj-3 f.obj-hit ‘You hit her.’ c. k-aŋ-a-ted 2sbj-3anim.pl.obj-pres-hit ‘You hit them (people or animals).’ d. ku-b-ted 2sbj-3inan.obj-beat ‘You hit it.’ / ‘You hit them (inanimate objects).’ e. k-aŋ-a-teɣ-in18 2sbj-3anim.pl.obj-pres-hit-anim.pl.sbj ‘You (plural) hit them (people or animals).’ f. ku-b-teɣ-in 2sbj-3inan.obj-beat-anim.pl.sbj ‘You (plural) hit it.’ / ‘You (plural) hit them (inanimate objects).’ Other verbs require multi-site subject marking, with both person and number agreement marked twice in different slots in the verb form: (44) a. Example of multi-site subject marking in an intransitive verb d-aka-dǝŋ-t-l-aq-in 1sbj-upland-1pl.sbj-to-pst-go-anim.pl.sbj ‘We made a quick round trip from river to forest.’ 18 The verb base -ted is replaced by the allomorph -tek whenever the animate class plural subject agreement suffix -in, a combination that is pronounced [tɛɣin]. Number in Ket (Yeniseian) 345 b. Example of multi-site subject marking in a transitive verb d-ǝla-dǝŋ-a-qos-n 1sbj-outside-1pl.sbj-3m.obj-take-anim.pl.sbj ‘We take him outside.’ Several reasons for the origin of multi-site agreement markers in Ket verb stems have been identified (Vajda 2017a), including semantic reanalysis of an incorporate element that happens phonologically to resemble an agreement marker, but the origin of the phenomenon in examples (44a) and (44b) remains unclear. All of the example verbs given so far are single-action stems. The next subsection addresses how the Ket verb expresses pluractionality. 3.4.2 Number marking in finite verbs arising through reanalysis The number distinction developed in some adjectives and action nominals via reanalysis of the derivational suffix -(V)ŋ as a plural marker, as earlier described in 3.2, is maintained when these words are incorporated into finite verb forms. In verb forms with inanimate class subjects or objects, which do not mark plural agreement grammatically, the plural marker in the incorporated adjective is the sole marker of plurality, as in (45b) and (46b): (45) Plural marking of incorporated adjective that correlates with inanimate class subject a. qa-d-a-b-qan big-transition-pres-3inan.sbj-become ‘It becomes big.’ b. qe-ŋ-d-a-b-qan big-pl-transition-pres-3inan.sbj-become ‘They (inanimate objects) become big.’ (46) Plural marking of incorporated adjective that correlates with an inanimate class object a. d-ukde-t-a-p-sin 1sbj-long-cause-pres-3inan.obj-be ‘I lengthen it.’ b. d-ukde-ŋ-t-a-p-sin 1sbj-long-pl-cause-pres-3inan.obj-be ‘I lengthen them (inanimate objects).’ In verbs where such adjectives correlate instead with animate class subjects or objects, which do regularly express grammatical plural agreement, the extra plural 346 Edward Vajda suffix in the incorporated adjective form results in multiple exponence of plural marking, sometimes in tandem with other stem changes like base suppletion that likewise express plurality of the argument or pluractionality, as in (47b). (47) Plural marking of incorporated adjective that correlates with an animate class subject a. t=qa-d-a-qan 3m.sbj=big-transition-pres-become ‘He becomes big / grows up / matures.’ b. t=qe-ŋ-d-a-set-n 3anim.sbj=big-pl-transition-pres-many.become-anim.pl.sbj ‘They (people or animals) become big / grow up / mature.’ Generally, adjectives that lack a number contrast outside the verb do not vary to express plurality when incorporated into a finite verb stem either. (48) Examples of incorporated adjectives that lack a number distinction a. d-aqta-a-qan [daqtaʁan] 3m.sbj-good-pres-become ‘He recovers.’ b. d-aqta-a-set-n 3anim.sbj-good-pres-many.animate.become-anim.pl.sbj ‘They (people or animals) recover.’ Like variable adjective stems, variable forms of action nominals incorporated into finite verbs also signal a number contrast reflecting an intransitive subject or transitive object. Action nominal forms that retain the suffix -(V)ŋ tend to appear in multiple action verbs and contribute to the expression of the idea of pluractionality (plct), while this element usually is absent in verbs expressing single events: (49) Variable action nominal forms used to distinguish single from multiple events a. d-ulad-q-a-it 1sbj-petting-into-3m.obj.pres-put.once ‘I pet / start petting him (male dog, once).’ b. d-ulad-iŋ-q-a-da 1sbj-petting-pl-into-3inan.obj.pres-put.many.times ‘I pet him (male dog, on many different occasions).’ The use of the reanalyzed action nominal suffix -(V)ŋ as a marker of pluractionality is fairly sporadic. It sometimes remains in Southern Ket single-event verbs as well, and there is even more variability across the Ket dialects. For example, Central Ket Number in Ket (Yeniseian) 347 d-uladiŋ-q-a-it ‘I pet him once’ and Yugh d-uljadɯŋ-χ-á j-it ‘I pet him once’ show that the suffix is retained in single-event verb forms. One final instance of a morpheme reanalyzed as a plural (pluractional) marker through coincidental homonymy with the common plural marking -(V)ŋ involves the archaic change-of-state (inchoative) suffix -(V)ŋ. In a few dozen stems, this old aspect suffix survived only in forms with plural subjects, where it was reinterpreted as a subject plural marker. With the base -teel ‘freeze’, it is sporadically retained when referring to plural entities, including inanimate class plurals. (50) b-in-teliŋ 3inan.sbj-pst-freeze (also recorded as b-in-teel, with loss of suffix causing root vowel lengthening) ‘They (inanimate class things) froze (once).’ Example (50) shows that plural markers arising through reanalysis can reference inanimate as well as animate class entities, while canonical plural agreement in finite verb forms only indexes animate class subjects or objects. Most other stems where it survives, however, are generally restricted to animate class subjects, as is the case with the base -doq ‘fly’. (51) Animate plural marking expressed by a reanalyzed inchoative suffix a. d-in-doq 3anim.sbj-pst-flew ‘He flew.’ b. d-in-aŋ-doq-ŋ 3anim.sbj-pst-3anim.pl.sbj-flew-pl ‘They flew (together as a group).’ c. d=t-ol-aŋ-doq-ŋ 3anim.sbj=plct-pst-3anim.pl.sbj-flew-pl ‘They flew (one after another).’ The use of reanalyzed adjective, action nominal, or aspect suffixes as plural or pluractional markers in Ket finite verb forms is sporadic and unpredictable. As already mentioned, such reanalyzed affixes differ most strikingly from canonical agreement morphology inherited from Proto-Yeniseian in occasionally expressing plurality of an inanimate class subject. 4 Semantics and discourse The semantics of number as category values have been described in the relevant sections above. There is no additional data currently available to discuss the pragmatics of number in Ket or its role in Ket discourse. 348 Edward Vajda 5 Summary and conclusions The preceding sections have covered all facets of number marking in modern Ket, offering observations on the historical evolution of plural marking patterns to help distinguish ancient features from more recent innovations. The core distinction between singular and plural in Ket encompasses most nouns and pronouns and is typically expressed by augmenting the bare singular stem with a suffix containing one of two nasal consonants: /n/ or /ŋ/. There is no evidence that plurals were ever productively generated by other means. Although some irregularities in plural formation, including ablaut and odd morphophonemic stem alternations, trace back to the proto language, they appear to be morphophonological or otherwise secondary. The grammatical dichotomy between animate and inanimate class nouns influences the choice of plural suffix allomorph and also dictates the form of 3rd person anaphoric pronouns. Because Yeniseian personal pronouns are the etymological source of possessive markers and verb agreement affixes, these systems likewise formally distinguish between inanimate and animate class entities. Due to its intrinsic connection with number marking, the grammatical factor of animacy thus casts an influence over virtually all areas of the morphosyntax in one way or another. By contrast, the subdivision within animate class nouns between masculine and feminine manifests itself in agreement morphology associated with singular nouns and pronouns only and may represent a Ket and Yugh innovation not traceable back to Common Yeniseian. Both genetic and areal factors in the linguistic history of Yeniseian peoples enhance the value of Ket data for a cross-linguistic study of number marking. Yeniseian languages are genealogically isolated from all other families of northern Asia, and some features of Ket number marking are not found in any of the neighboring languages. This includes, in particular, the pervasive formal contrast evident throughout Ket morphology between how animate class and inanimate class nouns express and reflect number – a trait not observed anywhere else among the indigenous languages of Siberia. Other features, such as the use of plural suffixes on nouns is found widely in other indigenous language families of northern and Inner Eurasia. In addition, the language’s genealogical isolation makes contact-induced changes a particularly interesting topic of investigation. Ket speakers were the last hunter-gatherer-fishers in landlocked northern Asia. They continued to move about seasonally in small kin-based groups long after all of their neighbors had shifted to reindeer herding or, in parts of south Siberia, to stockbreeding. For centuries, demographic pressure induced the Ket hunting bands to take brides from pastoral groups as marriage partners who presumably had to learn the language as young adults. The influence exerted on Ket linguistic forms by young L2 speaking mothers from Turkic, Ugric, Samoyedic, and Tungusic speaking groups could possibly be implicated in the unusually high number of cases where plural marking developed Number in Ket (Yeniseian) 349 via semantic reanalysis in areas of Ket morphology where it had originally been absent. This includes the sporadic creation of dual number forms of nouns, the rise of multi-site subject plural marking on the basis of a moribund change-of-state affix in finite verb forms, pluractional marking developed from a reanalyzed action nominal suffix, as well as plural marking in a small subset of adjectives – a lexical class originally entirely devoid of inflections of any kind within the strongly head marking Yeniseian family. The expression of pluractionality in Ket could conceivably have been influenced by Selkup, with its rich inventory of Aktionsart forms that relate to quantification of events. Analogous features need not have been present as such in the neighboring languages originally spoken by Ket spouses for these L2 speakers to have influenced the spread of new patterns via semantic reanalysis. Among languages in contact with Yeniseian, adjective agreement, for example, is present only in Russian, and these patterns seem to have been established in Ket and Yugh already before the arrival of Indo-European languages in central Siberia. Finally, it need be stressed that although it is now possible to describe Ket number marking in great detail and to account for many intricate aspects of its diachronic evolution, some facets of its morphological development still remain unclear. This particularly concerns the motivations behind certain irregular noun plurals, as well as the reason why only certain adjectives and action nominal stems developed number marking while most did not. The origin of multi-site subject marking in some Ket verb types also remains unclear. Abbreviations abl ades anim dat f inan m nmlz obj pl pltc poss pres pron pst sbj sg ablative case adessive case animate class dative case feminine inanimate class masculine nominalizer object plural pluractional possessive present tense pronoun past tense subject singular 350 Edward Vajda References Bibikova, Vera S. 1976. O kategorii chisla v sisteme imeni prilagatel’nykh v ketskom jazyke [Number marking in Ket adjectives]. In Andrej P. Dul’zon & Tel’mina I. Porotova (eds.), Jazyki i toponimija Sibiri [Languages and toponymy of Siberia] 6, 90–96. Tomsk: TGPI. Castrén, M. A. 1858. Versuch einer jenissei-ostjakischen und kottischen Sprachlehre. St. Petersburg: Imperatorskaja Akademija Nauk. Georg, Stefan. 2007. A descriptive grammar of Ket, part 1: Introduction, phonology and morphology. Kent: Global Oriental. Helimski, Eugene. 2016. S-singulatives in Ket. Journal of language relationship 14(3). 157–163. Kotorova, Elizaveta & Andrey Nefedov. (eds.). 2015. Comprehensive dictionary of Ket. (2 vol.) Munich: Lincom Europa. Nefedov, Andrey & Edward Vajda. 2015. Grammatical sketch of Ket. In Elizaveta Kotorova & Andrey Nefedov (eds.), Comprehensive dictionary of Ket. Munich: Lincom Europa. Porotova, Tel’mina I. 1990. Kategorija mnozhestvennosti v enisejskikh jazykakh [The category of plural in Yeniseian languages]. Tomsk: Tomsk State University. Porotova, Tel’mina I. 2002. Slovar’ govornykh form ketskikh sushchestvitel’nykh (s formami mnozhestvennogo chisla) [Dictionary of dialectal forms of Ket nouns (with plural forms)]. Tomsk: TGPU. Vajda, Edward. 2001. The role of position class in Ket verb morphophonology. Word 52(3). 369– 436. Vajda, Edward. 2004. Ket (Languages of the world/materials 204). Munich: Lincom Europa. Vajda, Edward. 2013a. Metathesis and reanalysis in Ket. Tomsk journal of linguistics and anthropology 1(1). 14–26. Vajda, Edward. 2013b. Vestigial possessive morphology in Na-Dene and Yeniseian. In Sharon Hargus, Edward Vajda & Danny Hieber (eds.), Working papers in Athabaskan (Dene) languages 2012, 79–91. Fairbanks: ANLC. Vajda, Edward. 2017a. Ket incorporation. In Michael Fortescue, Marianne Mithun, & Nicholas Evans (eds.), Handbook of polysynthesis, 906–929. Oxford: Oxford University Press. Vajda, Edward. 2017b. Patterns of innovation and retention in templatic polysynthesis. In Michael Fortescue, Marianne Mithun, & Nicholas Evans (eds.), Handbook of polysynthesis, 363–391. Oxford: Oxford University Press. Vajda, Edward. 2019. Ket lexical typology. In Mikhail Kotkin (ed.), Wortschätze und Sprachwelten (Sprach- und Kulturkontakte in Europas Mitte: Studien zur Slawistik und Germanistik 10), 59–72. Berlin: Peter Lang. Werner, Heinrich. 1997a. Das Jugische. Wiesbaden: Harrassowitz. Werner, Heinrich. 1997b. Die ketische Sprache. Wiesbaden: Harrassowitz. Werner, Heinrich. 2002. Vergleichende Wörterbuch der Jenissej-Sprachen. Wiesbaden: Harrassowitz.