Fortescue - The Relationship of Nivkh to Chukoto-Kamchatkan Revisited

download Fortescue - The Relationship of Nivkh to Chukoto-Kamchatkan Revisited

of 18

Transcript of Fortescue - The Relationship of Nivkh to Chukoto-Kamchatkan Revisited

  • The relationship of Nivkh to Chukotko-Kamchatkan revisited

    Michael Fortescue *

    Institute of Scandinavian Studies and Linguistics, University of Copenhagen, Njalsgade 120, 2300 Copenhagen S, Denmark

    1. Introduction

    The current consensus as regards the interrelationship of the so-called Paleosiberian (or Paleoasiatic) languages isthat there is no such relationship, merely a collection of all the non-Altaic languages of eastern Siberia, as originally putunder the rubric Paleoaziatskij by Schrenck (1883). It was intended to cover Chukotko-Kamchatkan (i.e. Chukotian plusItelmen), Nivkh (Gilyak), Yukaghir and Yeniseian (later sometimes extended to Ainu and Eskimo-Aleut cf. Comrie,1981:239). Nivkh in particular is regarded today as an isolate displaying a number of dialects but with little historicaldepth between them (cf. Mattissen, 2003:4). The unity of the Chukotko-Kamchatkan family (CK) within Schrencksgrouping is on the other hand well established (cf. Fortescue, 2005), despite some lingering doubts as to the exactrelationship of Itelmen to the northern branch on the part of Russian specialists (cf. Maslova, 2001). With Yeniseic nowon its way to being firmly related to Na-Dene (cf. Vajda, 2010), and Yukaghirs probable relationship to Uralic and/orEskimo-Aleut still being the most likely hypothesis proposed (cf. Fortescue, 1998:44ff), it is time to take another look atthe two remaining groups, CK and Nivkh (itself once presumably part of a larger Amuric family1), and reconsider whetherthere might still be a special relationship between them that could salvage at least part of Schrencks grouping asgenetically valid. Any progress in relating these distant languages one to the other or to completely different languagefamilies must be based on the relationship between the individual reconstructed proto-languages concerned, not on

    Lingua 121 (2011) 13591376

    A R T I C L E I N F O

    Article history:

    Received 30 July 2010

    Received in revised form 2 March 2011

    Accepted 3 March 2011

    Available online 8 April 2011

    Keywords:

    Paleosiberian

    Chukotko-Kamchatkan

    Nivkh

    Genetic relationship

    A B S T R A C T

    With the availability today of reliable materials for comparing the languages that in the

    past have been lumped together under the rubric Paleosiberian it has become possible to

    reassess the genetic relationship or lack of it between the individual languages of this

    traditional grouping. It will be demonstrated that the case for the genetic relationship of

    two of the constituent groups, Chukotko-Kamchatkan and the isolated Amuric language

    Nivkh (Gilyak), is actually quite strong, although the rest of the grouping must indeed be

    abandoned as a genetic unit. A case is made for reconstructing a Chukokto-Kamchatkan-

    Amuric proto-language associated with the Neolithic of the Lower Amur and adjacent

    coasts of the Sea of Okhotsk. Emphasis is laid especially on the morphology and shared

    typological features of these languages, but numerous lexical items based on systematic

    sound correspondences are also introduced. A plausible archaeological framework

    elucidating the evident closeness of the two linguistic entities is also sketched.

    2011 Elsevier B.V. All rights reserved.

    * Tel.: +45 3532 3660; fax: +45 3532 8377.

    E-mail address: [email protected] This term is due to Janhunen (1996:225), who suggests a Manchurian origin for the family on the Upper Amur during the Neolithic period. He further

    tentatively proposes that the Kamchukotic homeland was on the coast of the Sea of Okhotsk immediately north of the Lower Amur region (Janhunen,

    1996:237). I shall continue to refer to Nivkh as a family (of which it is the one surviving language), though its historical depth is quite shallow. The incursion

    of Tungusic speakers from both south andwest has undoubtedly confined the territory of the language to amuch smaller area than it once covered, with the

    ensuing demise or merging of earlier dialect and even language varieties.

    Contents lists available at ScienceDirect

    Lingua

    journal homepage: www.e lsev ier .com/ locate / l ingua

    0024-3841/$ see front matter 2011 Elsevier B.V. All rights reserved.doi:10.1016/j.lingua.2011.03.001

  • unanalysed contemporary lexical sources (which vary greatly in accuracy). A direct comparison of this sort is nowpossible between Chukotko-Kamchatkan and Nivkh thanks to the reconstructions for the former in Fortescue (2005) plusthe comparative Nivkh files that I have since assembled, based on all the published lexical sources available to me(Japanese as well as Russian) plus the comparative work of Hattori, Jakobson and Austerlitz, as well as the grammaticaldescriptions of Krejnovic, Panfilov, Gruzdeva and Mattissen. These are all included in the references below. Proto-Nivkhforms can be set up according to the criterion that they should be reflected in at least one northwestern dialect, i.e. themainland Amur dialect (A) plus that of the northwestern shore of Sakhalin opposite (NS), and at least one from theeastern Sakhalin dialects, i.e. East Sakhalin (ES) and South Sakhalin (SS).2 This results in some 1100 lexical sets plus 120common affixes and postpositions.

    This comparativebasis is relatively shallow, arguablyonly involvingdialectsof a single language, and the forms I reconstructfor the proto-language are thus very close to the modern forms in the more conservative south-eastern dialects. Standardcomparative procedures can nevertheless be applied in order to reconstruct earlier stages of the language with Nivkh one atleast benefits from the rich possibilities for internal reconstruction, a point stressed by Austerlitz (1982, 1984, 1990a).A methodological paradox arises here, however: the level of Proto-Nivkh that I reconstruct from comparing contemporaryforms is shallower than that Austerlitz aims at, at least in some of his reconstructions, which must be assigned to variouspre-proto stages on my definition. My hypothetical Proto-CKA (Proto-Chukotko-Kamchatko-Amuric) may or may not liestill deeper than any of these stages.

    Various attempts to relate Nivkh to Paleosiberian and/or other languages have been made in the past, but none ofthem has passed beyond listing a few dozen look-alikes (see Mattissen, 2003:4 for detailed references). The most seriousattempt was undertaken by Tailleur (1960), who limited his focus specifically to Nivkh and Chukotko-Kamchatkan, butthis was premature, being based on unreliable data from individual languages, without any preliminary family-internalreconstruction.3 Later attempts to relate Nivkh to larger groupings such as hypothetical Nostratic were no moreconvincing, culminating in the attempt by Mudrak and Nikolaev (1989) to relate it at one and the same time toChukotko-Kamchatkan and to Greenbergs controversial Almosan-Keresiouan. Although their brief list does citereconstructed proto-forms as far as possible it makes hardly any attempt to sort out detailed sound correspondencesbetween families. Of these earlier lists of look-alikes, many (if not most) must in fact be rejected now that Proto-CK canbe directly compared to Proto-Nivkh. It will be seen in what follows that a much stronger case can be made today forrelating Nivkh with CK than was possible for Tailleur, although the evidence will not be sufficient to declareunequivocally that there is a direct genetic link between the two families comparable to the one that can be establishedfor the relationship between, say, Chukotian and Itelmen. Given that we are dealing here with a very distant relationshipit is too early to claim final proof of genetic relationship what I hope to achieve, however, is a significant increase in thelikelihood of such a relationship, bolstered by the relative perspicacity of the reconstructions presented. This willprincipally involve the area of morphology. Although the lexical evidence is slender, it will be seen to involve basicvocabulary, of the kind unlikely to have been borrowed between neighbouring languages. The list of potentialcorrelates between Nivkh and CK presented in Fortescue (1998) taken mostly from Tailleur can no longer beinterpreted as reflecting simple borrowings. Many of them can in fact be dismissed as chance look-alikes, though, as weshall see, a residue does remain, a residue that can be increased by more convincing parallels not noticed earlier. Afurther justification for this new attempt is the light that the broader perspective sheds on some of the more puzzlingmorphological and phonological details that have remained unexplained in both Nivkh and CK individually, such as thehistorical source (and original semantic purport) of the Nivkh undergoer prefix for unspecified object and the CKinverse prefix of similar shape, and of the areal nature of the remnant vowel harmony system of Nivkh and the strangelyskewed vowel harmony system of CK.

    The earliest homeland of the northeastern Paleoasiatics (i.e. Chukotko-Kamchatkans)was, according to Levin (1963:283),along the northern coasts of the Okhotsk Sea (where dwellings of the Nivkh and Kamchatkan type with winter roof entranceshavebeenexcavated inconnectionwith the OldKoryak cultureof the latterpartof thefirstmillenniumBC).4TheNivkh,on theother hand, are believed tobedescendants of the oldest inhabitants of the territory around themouthof theAmurwhich theynow occupy and further along the adjacent mainland coasts. The later Okhotsk Culture of the 5th to 13th century AD thatspread from the southern shores of Sakhalin to northern Hokkaido and up along the Kurile islands towards Kamchatka is alsoconsidered to have been borne by the Nivkh (Levin, 1963:118). Vasilevskii (1969:153) specifically relates the Old Koryak andOkhotsk cultures and suggests their common origin. In the Conclusion I shall return to the archaeological evidence supportingthe idea of a common origin for contemporary Chukotko-Kamchatkan and Nivkh groups in the Neolithic of the Amur region,

    2 The principal phonological differences between the Amur and the East Sakhalin dialects are: */a/ has reduced to // inmany (but by nomeans all) initialstem syllables and /v/ and /w/ havemerged as /v/ in A; final /r/ (a suffix on nouns) has become /s/ and final // (when representing an original suffix on bothnouns and verbs) has been lost in A. North Sakhalin is intermediate. South Sakhalin further differs from ES bymaintaining a probably original distinction

    between a lenis vs. fortis (or voiced vs. voiceless) initial consonant series as opposed to the unaspirated vs. aspirated one developed elsewhere along with

    the introduction of a new, third voiced series as the result of nasal sandhi (I return to this below). The two series are conflated (as fortes) in clusters (and

    generally later in the word). For a precise account of the secondary origin of initial fricatives in Nivkh see Jakobson (1971:85ff).3 As a single example of a dubious lexical comparison (one that Tailleur makes much of) consider the form lewlew-, which he gives as trick, deceive for

    East Sakhalin, supposedly the same as Chukotian lewlew-et of thatmeaning. The nearest one can actually get to that form in themodern dictionaries is either

    valt- lie, deceive or laulaud argue.4 Compare the possible CK cognates of Nivkh tom(s) smoke-hole in the section on lexical material below.

    M. Fortescue / Lingua 121 (2011) 135913761360

  • with the Tokareva culture of the shore of the Sea of Okhotsk of about 3500 BP as the missing link. In Fig. 1 the approximateextent of these early cultures (in capitals) has been superimposed on the extent of the relevant languages in the region today.

    Given the results of the present investigation, the general conclusion in Fortescue (1998) as to the relationshipbetween CK and the hypothetical Uralo-Siberian mesh needs to be readjusted somewhat.5 I would no longer wish torelate CK directly to that mesh, although I believe that some of the lexical evidence adduced for a link with it will hold upin terms of borrowing/diffusion. The more important morphological and grammatical links of CK to other languageswould now seem, however, to point definitely towards the south, to Nivkh, a possibility that I left open in mymonograph.

    One reason this relationship has not been more apparent in the past is the relative dearth of inflectional morphologyin Nivkh (as opposed to its profusion in CK). The most reasonable starting point for a renewed line of attack on this matteris to assume that Nivkh has been more worn down by contact with more numerous neighbours than has CK, which hasremained more isolated, spread over the sparsely populated tundra and adjacent icy shores to the north. Much of whatfollows will be devoted to demonstrating how there are traces of an earlier, somewhat more complex morphology inNivkh that can be correlated with the relatively better preserved situation in CK. The suffixal morphology of the CKlanguages has also been worn down (to be partially replaced by prefixes), and these languages can be shown to haveinnovated much of their inflectional complexity since Proto-CK times, as reconstructed in Fortescue (2003). There are infact numerous typological traits (some quite unusual) shared by the two families that suggest genesis from a commonstock.

    What are these common traits? First, as regards word-order, Proto-CK almost certainly displayed SOV order (althoughit is fairly free in modern Chukotian, these languages display the verb-auxiliary, adjective-noun and possessor-nounorder expected of this type), as does Nivkh. Unsurprisingly, the morphology in both families is primarily suffixing CKprefixes are demonstrably more recent, corresponding to the expected pre-verbal position of subject, object andadverbial constituents (cf. Fortescue, 2003:85). There are also postpositions/enclitics in both families (bound in Nivkhand now mainly suffixal in CK). More significantly, both show a considerable degree of polysynthesis in general and(arguably) of incorporation in particular, both verbal and nominal (i.e. of nominal adjuncts as well as verbal objects),although not all descriptions of Nivkh (e.g. Panfilov, 1962:25ff) have agreed on the incorporation analysis. The specificdependent-head variety of incorporation that typifies Nivkh according to Mattissen (2003:104ff) can be seen also tocharacterize CK as regards the dominant order of incorporated elements, namely dependent elements (objects, attributesor adverbial converbs) before heads (verbs or nominal phrasal heads).6 Although Georg and Volodin (1999:229) cite thelack of incorporation in contemporary (western) Itelmen as a sign of the lack of unity of that family, its loss can be relatedto the relative newness and transparent productivity of the fully-blown phenomenon in Chukotian and/or theabrasion of Itelmen under Russian influence (the surviving western language has been much influenced by both Russianand Koryak/Alutor).

    [()TD$FIG]

    Eskimo

    Chukchi Kolyma

    Even

    OLD KORYAK TOKAREVA

    Evenki

    Koryak

    Kerek Alutor

    Itelmen

    Amur

    Nivkh (A + NS)

    Sakhalin Nivkh (ES + SS)

    AinuHokkaido

    Kamchatka

    Kuriles

    OKHOTSK

    Sea of Okhotsk

    Fig. 1. Languages and cultures of the Sea of Okhotsk region.

    5 This was proposed to cover Eskimo-Aleut, Yukaghir and Uralic as well as (much more tentatively) CK itself.6 Summing up the structural parallelisms between polysynthesis in Chukchi and other northern languages and Nivkh dependent-head synthesis,

    Mattissen (2003:288) states: Especially between Chukchi and Nivkh, differences seem to be of degree rather than kind.

    M. Fortescue / Lingua 121 (2011) 13591376 1361

  • Other potentially diagnostic traits include the lack of distinction between grammatical case forms (nominative andaccusative also genitive) in both families, although Chukotian (not Itelmen) has developed ergative case marking and bothdisplay locative case systems on nominals7; the lack of a special adjective class (there are quality/quantity verbs rather);auxiliary verbs; dual number (only residual in Nivkh and perhaps not original in CK); indefinite object/antipassive-likeprefixes on transitive verbs; fluctuating nasal (and other consonantal) endings on nominals that may reflect earlier nounclasses (the so-called singulative markers of CK); and a distinctive root retraction (or high/low) form of vowel harmony(with only traces left in Nivkh and Itelmen, but virtually intact in Chukotian). It should be pointed out that this cluster oftraits constitutes a quite different profile from those of Eskimo-Aleut, Yukaghir or Yeniseian. As regards other phonologicaltraits, both families display a rather similar and not particularly complex array of phonological units (see below),although Nivkh appears to have undergone considerably more syncope than CK (with the egregious exception of Itelmen),resulting in more complex consonant clusters and the contraction of what once must have been bisyllabic stems intomonosyllables (internal dialect differences and also distinct prosodic patterns in Nivkh help clarify this cf. Austerlitz,1990a:21).

    As for the major divergences between the two families that require explanation, these include ergativity in CK (asecondary phenomenon, as discussed in Fortescue, 1997, 2003) but not in Nivkh; a complex system of numeral classifiers inNivkh but not in CK; evidential/epistemic mood marking in Nivkh but not in CK; exclusive vs. inclusive 1st person in Nivkhbut not in CK; a reflexive vs. non-reflexive distinction in possessivemarkers in Nivkh but not in CK; reduplication of stems inboth families, but used for different purposes8; distinctive sandhi-like alternations at word as well as morpheme boundariesin Nivkh (phonologically determined but utilized for morphosyntactic purposes)9; circumfixes in CK; and a complex systemof verbal paradigms formood/tense/aspect in CK (including remnants of an inverse system) but not in Nivkh.10 All of thesewill be touched upon in what follows.

    2. The common sound system

    The phonological system of the hypothetical proto-language let us call it Proto-Chukotko-Kamchatko-Amuric (or,somewhat less unwieldy, Proto-CKA) can be set up here as an initial approximation at least. Internal reconstruction forNivkh (e.g. Austerlitz, 1990a) and comparative work on CK (Fortescue, 2005) converge on a symmetrical six vowel system(plus schwa, outside of the vowel harmony system in CK see below) and a consonant system involving a velar/uvulardistinction and a single voiceless (or fortis) vs. voiced (or lenis) plosive series, with corresponding voiced fricatives, liquidsand nasals.11 The combining of a single series of voiced rather than voiceless fricatives with a plosive system like this istypologically rather unusual, but is definitely what is found in CK. The voiceless fricatives and initial fricatives in general are of secondary origin in both families (cf. Fortescue, 2005:7ff for CK and Austerlitz, 1990a:20 for Nivkh). The distinctionbetween two open vowels /a/ and // (back/dominant vs. front/recessive) of Proto-CK is not found in Nivkh, and mayindirectly reflect the decay of vowel harmony in that language, the source I reconstruct for CK front//, namely */L/,becoming // or /a/ in that language. The different reflexes here may reflect position respectively within morphemesoriginally of recessive/high vs. dominant/low harmony.

    The twoplosive series hypothesized on the basis ofNivkhmust have collapsed in CK,12 andmostNivkh dialects developed adistinct voiced series of plosives in connectionwith nasal sandhi or nasal alternation (fromsequences of weak nasals cf.Mattissen, 2003:29 plus plosives, the result remaining even if the nasal subsequently dropped). All dialects like Itelmenwithin CK developed independent voiceless fricatives under certain phonotactic conditions (in Nivkh through thefricativization of plosives including that of */t/ to /r/). Austerlitz (1990a:25) supports the idea of the secondary nature ofthe three-way distinction in plosives of all dialects of Nivkh except South Sakhalin, but goes too far tomymind in suggesting mainly on areal grounds that this might further be reduced to one.

    I am also sceptical of his suggestion (Austerlitz, 1990a:20) that the velar vs. uvular distinction is secondary (at least asregards my CKA level) there are just too many contrasting initial velar/uvular plus (same) vowel forms reconstructible in

    7 Besides its locative case endings, Nivkh has an unmarked nominative case (like Itelmen and corresponding to the Chukotian absolutive), an

    instrumental case and one specialized for standard of comparison, also a causee case -(a)X, called dative-accusative in Saveljeva and Taksami (1965)

    and used in causative constructionswith animate causee (perhaps originally a dative). Chukotian also has an instrumental case (used alongside the locative

    to mark the ergative subject on nouns), two comitative cases, as well as an attributive and a referential case. In both families case endings can to some

    extent be applied to bare verbal stems, e.g. in the Amur Nivkh supine and the CK infinitive, both treated below.8 For repeated, intense or reciprocal action with verbs in Nivkh and for the basic forms of certain nouns that do not take singulative suffixes in CK. There

    are, however, some reduplicated nouns in Nivkh, like kff spider (and reduplication of nouns can also indicate plurality cf. Mattissen, 2003:8).9 I use the word sandhi throughout this paper rather loosely it has been the subject of considerable debate whether the phenomenon is best

    characterized by this term or not. Certainly it goes further than changes at morpheme boundaries between free words, as suggested by Panfilovs use of the

    term (see Mattissen, 2003:65ff for a discussion).10 Most noticeable in this area is the almost total lack of person (subject) marking on verbs in Nivkh: it is only present in the distinct 1s and plural person

    endings in converbs.11 In the South Sakhalin dialect of Nivkh the distinction is between lenis and fortis, overt only in initial position before a vowel Austerlitz writes lenis /d/

    vs. fortis /t/, etc. in that position (corresponding to /t/ vs. /t/ in other dialects), elsewhere neutralized to /t/, etc. The aspirated vs. plain distinction is also rare

    outside of initial position in other dialects (a medial aspirated plosive may reflect the first phoneme in an originally independent word, as in postpositional

    -tx on discussed below).12 Compare Southern Wakashan, which has conflated without trace the original voiced/voiceless plosive distinction preserved in Northern Wakashan

    (Sapir and Swadesh, 1952).

    M. Fortescue / Lingua 121 (2011) 135913761362

  • what I call Proto-Nivkh, though the origin of uvulars may well have been due to flanking vowels (as Austerlitz suggests) at astill earlier stage (pre-Proto-CKA).13 It is certainly true that initial uvulars may only appear before /a/ or /o/ in modern Nivkh(Jakobson, 1971:84) and an initial velar has gone to the corresponding uvular before /a/ in many ES/SS forms that retain thevelar plus reduced // < */a/ in the Amur dialect, such as A kj sail, ES qaj (< *kaj). Original *qa-, by contrast, is kept in bothdialects, as in the stem of this shape meaning go downstream. This may also explain suchmedial correspondences as A cir(< *cigar-?), ES cXar tree, wood, stick, where, as Austerlitz proposes (1994a:229), the uvular in ES and SS is due to thefollowing /a/ (plus a shift of stress in A and conflation of /cX-/ with /cX-/ in ES) compare Chukotian cik()l() pole fortraining reindeer with as a likely cognate, going directly with Nivkh tla shaft of spear, which Panfilov (1962:197) relates tocir plus -la, the numeral classifier ending for poles.

    Further differences among the contemporary languages can be rather simply (if speculatively) explained such as theaspirated plosive series inmost Nivkh (but not in CK), which probably just reflects the original voiceless (or fortis) series. Thedistinct palatal series of Nivkh appears also to have been original, collapsed in CK except for the distinct plosive /c/ (=*/t/),but with some further residue (especially in Koryak) utilized for affective purposes. The origin of the glottal stop andpharyngeals found in some CK (also the ejectives of Itelmen) are discussed in Fortescue (1998, 2003). In general, Occamsrazor has been applied in providing a coherent picture of the probable original system. This can be laid out in the followingmanner.

    The specific sound correspondences between Nivkh and CK that lie behind this reconstruction (as reflected in the lexicaland morphological material presented in this paper) can be summarized in Table 1. The complete phoneme inventory forProto-CKA is reconstructed in Table 2.

    I assume that *//generallymergedwith */r/in Nivkh (as in Chukchi) note that/r/is treated as a voiced fricative as regardssandhi alternations there. The lack of a sibilant is also notable in Table 2 in contemporary Nivkh /s/ and /z/ are the result ofthe regular sandhi/transitivizing alternation of */t/ and */d/ (though in some cases medial /z/ appears to have developedfrom */r/)14, and /s/ has developed from earlier /c/ (*/t/) in Chukchi and Alutor; /z/ or /s/ is generally from */j/ in Itelmen(Fortescue, 2005:7ff). I cannot accept Austerlitz (1990a:19) very tentative, typologically based proposal that/h/might havecome from a proto-phoneme */s/. The reconstruction of /h/ in Table 2 requires further comment it is absent in CK althoughcommon in Nivkh (initial only). I shall return to this matter in the section on lexical material. I do follow Austerlitz inassuming that initial consonant clusters in Nivkh reflect the loss of an intermediate unstressed vowel that was lost(Austerlitz, 1990a:21), initial clusters being secondary, much as in the more conservative CK languages Chukchi and Koryak,

    Table 1Sound correspondences between Nivkh and CK.

    CK Nivkh

    p, t, c, k, q p, t, c, k, qa

    p, t, c, k, q

    v, , , R v, r/z, , Rm, n, n, m, n, n, w, j w, j

    l, r l, r/z

    hi, u i, u

    e, o e, o

    a/a a

    a This aspirated series could be written p, t, c, k, q, and the plain series below it as b, d, d, g, G (reflecting the conservative SS dialect). The voiced series of

    the other dialects developed from nasal sandhi, and the voiceless fricative series f, r, s, x, X by ordinary (non-nasal) sandhi from the first series (the result of

    ordinary sandhi of the second series falls together with the voiced fricative series below).

    Table 2Reconstructed phoneme inventory for Proto-CKA.

    i () u p t t k qe L o b d d g G

    a v Rm n n w j h

    l

    r

    13 Note that root retraction vowel harmony usually presupposes uvular or pharyngeal segments as a post-vocalic trigger that may or not then disappear

    (cf. Bessell, 1992:167ff). The origins of the uvular/velar distinction and of root retraction vowel harmony are at all events probably related here.14 As in A cuz-, ES cir- new, comparablewith CK tur- new, and ES kuz- go out, comparablewith CK tkur- go off. There is sporadic inter-dialect variationbetween these two phonemes, as in the first form cited here, also A hurur, ES huzus everywhere, A jaz-d, ES jar-d bite, but also A zozu-d, ES roru-d

    extinguish and A vazud, ES varud sew strips of net.

    M. Fortescue / Lingua 121 (2011) 13591376 1363

  • which still by and large disallow such clusters (and where medial clusters are limited to two consonants). There is alsoconfirmatory evidence for his reconstruction of initial CplCfrV- sequences in Nivkh (where pl = plosive and fr = fricative) asgenerally coming from *CplVCplV- in such correspondences as -tx on going with CK -tk(n) of that meaning. In fact this isone of the examples he gives (reconstructed as *tky, where /y/ is a stressed mid-high vowel corresponding to my schwa).

    As regards the vowels, note that root retraction vowel harmony in CK operates between dominant /e/, /o/ and /a/ andcorresponding recessive /i/, /u/ and // such that if a word has one dominant vowel, all its vowels must be dominant too.This works bi-directionally, as still marginally reflected in the residual cases remaining in Nivkh, where a high series /i/, //and /u/ alternate with a low /e/, /a/, /o/ series e.g. in allative e-rX to him from i(f) (where the case suffix in its full form isdominant -toX/-roX), and, less commonly, the epenthetic vowel following (some) reduplicated stems harmonizes with thatof the stem, as in perper-a-d sways vs. prpr-i-d lies about (Mattissen, 2003:79).15 Notice that it is // that alternates with/a/ in Nivkh, not // as in CK, and that in CK // is largely epenthetic and outside the vowel harmony system. Thereconstructed vowel system in Table 2 presents amore symmetric picture for vowel harmony than in CK itself (where the //alternatingwith /a/ is an extra front vowel16). The replacement of // by a centralized /L/ in Table 2 brings the alternationpattern more in line with the simpler high vs. low system in Nivkh also with neighbouring Tungusic languages. Thesedisplay similar root retraction harmony systems in which soft (i.e. high) // varies with hard (i.e. low) /a/ (the latter infact a lowered schwa like /L/ in Even, which has maintained the purest system). In Tungusic this does not operate bi-directionally as in CK but progressively alone, as in Mongolian (cf. Comrie, 1981:70f). This may well represent an arealphenomenon.17

    There are some good potential cognates containing CK // corresponding to Nivkh //, notably r- mouth of river (andrelated forms cited further on) that appear to go with CKr- flow out, and tk edge of sleeping platform which goes withCK -tn near or at edge of (also the Nivkh terminative case endings, discussed in the following section). Others show /a/in Nivkh, which may simply be indicative of the early break down of the vowel harmony system in Nivkh at this point (thealternation is very peripheral today).18 Note that Amur // comes from */a/ in those forms where the Sakhalin dialects havean /a/ (e.g. A tf, ES taf house cf. Jakobson, 1971:91), but r- is precisely not one of these, having schwa in all dialects.

    Finally, the development of the sandhi-like phenomena atmorpheme boundaries in Nivkhmust be of later origin than thehypothesized Proto-CKA stage. The processes involved all quite natural include the development of a voiced series ofplosives through nasal sandhi and themore general results of themorphophonemic principlewhereby plosive plus plosivesequences became dissimilated to fricative plus plosive or plosive plus fricative, with varying patterns according to thewordclasses involved (Mattissen, 2003:44ff). Various morphophonemic processes have produced alternation in the shape ofmorphemes within CK too, but these are more idiosyncratic, such as the alternation of initial /r-/ (Koryak /j-/) with medial/-n-/ in Chukotian transitive verb stems (from *n-), and at all events are not reconstructible for the proto-language (cf.Fortescue, 2005:10). Since there are no commonmorphophonological alternations at morpheme boundaries reconstructiblefor Proto-CKA, there is no evidence that it had a comparable level of polysynthesis to that of modern Chukotian or evenNivkh. In particular, it may have lacked incorporation as opposed to fixed sequences of independent words altogether.

    3. Nominal morphology

    Let me proceed now with an investigation of the morphology of both families I shall return to a more exactconsideration of sound correspondences in connection with the lexical evidence presented in section 5. First I shall considernominal inflection and derivation, and thereafter via pronominal and participial markers I shall move on to verbalinflection and derivation in a separate section.

    The oldest spatial case markers in CK (reconstructible for the proto-language and found in numerous combinations andfunctions in the modern languages) are general locative -k() and general lative (including allative) - (see Fortescue,2005:426). Both elements would appear to have reasonable correlates in Nivkh. As regards the first compare the locative/ablative case marker -(u)x on nouns (also ES extension -uxe = Amur perlative -ue), and -k in ra-k where (and note hu-g/EShun-x there, CK n-k there, mi-k where). As regards * - (*-(a)?) compare suffix -a fairly close to (with locationalstems) and extension -ajo further from,19 as in ES taz-a place out on the water somewhat removed from the shore/village, taz-ajo place out on the water further removed from the shore (from tas (exact) place out on water near shore Krejnovic, 1986:163). Gruzdeva (2008:185) refers to these as medial and distal respectively in the secondary deicticsystem.

    15 See further in Hattori (1962, 2) for traces of the system within polysyllabic stems.16 Its development from */L/ in CKmay have been triggered by the increasing functional burden of //, outside of the harmony system, whose allophonesin the environment of uvulars would have been difficult to distinguish from /L/.17 An original bi-directional high vs. low system in Tungusic, shared with Nivkh, may have been influenced by the Mongolian system to the west, which

    also infused it with superimposed labial harmony.18 An independent merger between // and /e/ has occurred in Chukchi, undermining its own harmony system. In those dialects of Koryak and Alutorwhere there has been amerger it is between // and /a/ rather. In Itelmen,where vowel harmony broke down long ago except for the i/e and u/o alternationsin certain affixes, *// sometimeswent to /a/ as opposed to /e/ by generalization of the dominant form of amorpheme, and original schwa could go to either/i/, /e/ or /a/ (Fortescue, 2005:12).19 With adverbial suffix -jo, which has comparative (as well as iterative and verbalizing) function (cf. Saveljeva and Taksami, 1970:524). It would seem

    more likely that this would attach to an originally directional (or deictic) than to a static locative base. Note that -jo in Nivkh is a dominant vowel harmony

    form alternating with recessive -ju (cf. Austerlitz, 1994b:259), as expected following -a, which also has a dominant vowel CK - is also dominant.

    M. Fortescue / Lingua 121 (2011) 135913761364

  • The first of these may appear added directly to verbal stems in both language families to produce non-finite/adverbialforms, respectively CK infinitive -k() and Nivkh converb ending -ke when V-ing also participial marker -k (especiallycommon in ES). Consider also the special comparative (standard) case marker -(a)k in Nivkh it is precisely the locativecase that is used for marking the standard of comparison in CK.20 The second element (*-(a)) also appears to have verbalfunction in both languages, namely CK - adverbial formant and Nivkh temporal converb endings -a(n) and -.

    Another possibly common element is -r in such and such a location (with locational roots) in Nivkh, goingwith fossilizeddative -ri (with directional adverbs only) in CK, and Nivkh terminative/limitative (up to an edge) -tk/ES -taka(regarded as a case ending in the standard grammars), going with CK -tn near or at edge of cf. Nivkh tk edge ofsleeping platform as a noun (Panfilov, 1962:143). A similar item (somewhere between being a case ending and apostposition) is Nivkh -tx on (no doubt originally an independent word cf. Mattissen, 2003:9f.), going with CKderivational affix -tk(n) of the same meaning, already mentioned.21

    Of particular interest is the relationship of the CK ablative to the Nivkh case system: Tailleur (1960:117) directly links asupposed ablative -nxe as in East Sakhalin tu-nxe from here (his hyphen) to the Koryak ablative -qo and (supposed)Itelmen -nk (-enk is actually the locative). However, the Nivkh ending is in fact the locative/ablative -(u)x/-(u)xe alreadymentioned (and found in all dialects) and added here to base tun- or tu-. The Koryak ablative reflects Proto-CK ablative *-qo(r), probably related to qor hither (added to the lative) and distinct from the Itelmen ablative -x?al, (on personalpronouns -nx?al), which is probably the same as Alutor -qal, consisting of - plus qal(a) side, direction (see Fortescue,2005:434 fordetails).What is interestinghere isnot somuchtheexact formbut the typologically oddassociationof theablativeand the locative in a single case ending in Nivkh (only the Amur dialect has a distinct locative -(u)in).22 For the ablative is alsoan innovation in CK and as can be seen from the above its meaning is curiously intertwined with that of the lative directiontowards meaning (and even seems to contain the old lative case ending). A clue: qor hither, this way, bringme X is used inpointing and to here is in deictic contexts the same as from there.

    As was noted above, neither family has a genitive case, but both have possessive pronouns. In CK these are formed fromthe derivational possessive ending -in (productively applicable to nouns as well as pronouns),23 corresponding to Nivkhpossessive pronounmarker -n (as a nominal root thing, perhaps reflecting the common CKA source). Note also that 3s andplural prefixes/proclitics in possessive usage cause following nasal sandhi, as if an *- or -n has been lost (this is the essentialorigin of this pattern of alternation in the dialects of Nivkh that have it). Most of the other case endings in both families aresecondary and/or relatable to independent stems, e.g. Nivkh instrumental -kir, related to verbal stem i-r-/-kir- use (assuggested by Austerlitz, 1982:85), and allative -toX/roX/doX, which Austerlitz (1990a:28) relates to verb -tXop- touch, andChukotian allative -jt, based on verbal stem jt- go for plus lative - above. The CK comitative will be dealt with inconnection with the dual number below.

    Let us turn to number marking on nominals. Both families have a dual/plural distinction, although it has been lost inChukchi and in Nivkh it is limited to 1st person pronouns. As I have argued in Fortescue (1997:373) the original plural (orperhaps paucal) in CKmay have been -t(i), nowdual in Koryak (but not Chukchi or Itelmen), where the new plural is -wwi< *-gvi from derivational suffix -iniv collection of.24 Note that -wwi on personal names has a clear collective sense (X andfamily/companions). The Itelmen plural is -n (=glottalized/n/).25 In Nivkh the only trace of a plural -t is in the plural ofconverbs (falling together with a different, 1s -t I shall return to below). The regular nominal plural is -kun(u)/-gun(u)/-un(u),which may very well go with CK -iniv above (and hence have a parallel origin to the Koryak plural). Moreover, this is inturn related to the comitativemarkers Amur du. -ke, pl. -ko (ES -kin, -kunu), which can be directly compared to CK comitativeprefix k- from adverbial knml together (andwhich combines with a participial or an instrumental suffix respectively toform two related circumfixes on nouns).

    Beyond thedual andplural inCK, one alsohas to consider the singulative, anunusual feature of nouns in this family:many(perhapsmost)mustbemarkedbya special suffix (or elseby reduplicationof the stem)asbeing singulativeoriginallyperhapsindicating distinct semantic classes (for instance items that typically occur in pairs or groups cf. Fortescue, 2005:433). Thesearedropped (except in Itelmen)whencase endingsare added. Themost commonof these are -, -lnand -n (Itelmenalsohas-c, and -lin and -mi, both ofwhich are especially used for paired objects). Nowanotable inter-dialectal aspect ofNivkh is thatmany nouns in the East (and South) Sakhalin dialects end in a nasal (mainly -) which is lost in themainland (Amur) dialect.26Could these be the remnants of singulative markers of the Chukotko-Kamchatkan kind? There may even be traces of thesemanticclass ofobjects towhich theyonce referred, forexample in the fact thatmanyanimateobjects tend toendwithsuchanending in themorearchaiceasterndialects. The ending -n (alsoanominalizer inall dialects) is common innouns referring tokin

    20 Though Mattissen (2003:82) relates this to verbal stem (j)k- reach.21 Compare also Nivkh -le/laxe next to, up against and CK -lku(n) between, among.22 cf. Levinson (2003:101), where the conflation of location and source (as opposed to goal) is not an attested possibility. Krejnovic (1979:302)

    relates Amur -(u)in to the -n- in ES hunx there, etc. Note that there is a variant -uine (cf. Panfilov, 1962:138), perhaps from *-uine, of which the first partcould be the locative/ablative above.23 Also ingredient in -kin thing associated with and a number of participial suffixes like -lin and -qin, where the -i- is probably a copular element.The source of the prefix may have been an independent word *nik meaning something (> CK *nik cf. Fortescue, 2005:187).24 On verbs the plural is distinguished from the dual in Koryak by the addition of suffix -la- of several subjects/objects of an action (also found in Chukchi)

    before the dual marker -t.25 Probably deriving from CK Class 2 (animate) plural -()nti containing 3rd person demonstrative n- (which I shall return to below).26 This dialect tends in general to showmore syncope than the more archaic eastern dialects according to Krejnovic as a result of the shifting forward of

    word stress from the second to the first syllable in many words in that dialect (Krejnovic, 1979:299).

    M. Fortescue / Lingua 121 (2011) 13591376 1365

  • or animals according to Saveljeva andTaksami (1970:528f.). They alsomention the ending -r found in a fewwords referring tonatural objects that occur in aggregates (Saveljeva and Taksami, 1970:523),which can be compared to CK -rri a group of -s.

    As regards more clearly derivational (deverbal) suffixes, consider Nivkh -f place of -ing (the basic form is actually -v,regularly devoiced in final position, as in mif/miv- land from postposition mi- inside to which compare CK mi-kwhere); this corresponds to CK -nv of the same meaning.27 This is involved in purposive and/or supine forms of theverb in both languages Nivkh has purposive -f-toX, with the allative casemarker -toX following it (on a bare verbal stem thelatter alone produces the so-called supine cf. Gruzdeva, 1998:51f), while in Chukotian the supine formant is -nv-withthe lative marker following -nv (Itelmen -no-ke). Also some important deverbal participial formants apparently commonto both families can be introduced here. In ES Nivkh there is an attributive participial form of verbs in -, which (like thesingulatives above) has been dropped in the mainland dialect. In the SS dialect this varies with -n, so perhaps it derives from*-n and can thus be compared with CK -rn quality/action nominalizer. The 2/3s converb formant -r (plural -t) inNivkh28 could in turn be correlatedwith the first part of CK progressive/present participle -k(n), pl. -kt (I shall return tothis in the following section). The most important of these participial (or rather deverbal) forms in Nivkh is *-nt or *-nc (>Amur -d, ES -nt/nd/d, SS -nt), which is not only a deverbal nominalizer but forms the indicativemood of verbs (same form forall persons), and furthermore verbalizes interrogative and demonstrative roots (seeMattissen, 2003:16 for examples).29 It istempting to relate this to CK verbalizer (of nouns and adjectives) -()t, also a detransitivizer (e.g. in benefactiveconstructions with incorporated object), and together with transitivizing prefix *n- involved in transitivizing verbs.However, there is also a Nivkh verbalizer -t which may be more directly related to the CK suffix, so there is no obviouscandidate for a direct CK correlate to Nivkh *-nt. 30

    Finally, there is a suffix of permanent quality -la in Nivkh which is applied to quality verb stems (combinable withfollowing indicative *-nt). This can be compared with the Itelmen adjective formant -laX, perhaps going with Chukotianparticipial -lRn one who s. Alternatively, it is possible that the original indicative in Nivkh was actually the archaicnarrative mood suffix -qana/-(ja)Rana, later ousted by the nominalizing*-nt suffix, and it is the former that goes withChukotian -lRn.

    We can now turn to personal pronouns and in particular their bound (affixed and cliticized) forms, an area where NivkhandCK appear on the surface to diverge considerably. First let us take the independent pronouns. TheNivkh1pl (inclusive)mer(n)/ESmirn we (incl.) (where the -nmarks plurality as in nominal suffix -kun) is directly comparable toCKmur(i) we (wherethe -i is from plural *-ti via *-i).31 On the other hand, there is no obvious candidate either to match with Nivkh 1s ni or with1p exclusive nin (the plural of ni) in CK, where I is *km. There may be some evidence for a connection in the 1s form of theconverb -t mentioned above compare the CK 1s prefix t- (if both reflect some much reduced combination of a t-initialdemonstrative root plus a 1st person marker *m(V)).32 Panfilov (1962:236) suggests that 1s ni is identical with and thesourceofnumeralni- one,whichhas a reasonable cognate inCK nnn treated in the sectionon lexicalmaterial below. If thisis so, then the reverse order of derivation would seem more likely.33 Better affixal matches with the Nivkh 1st person rootmer-/mir- are the CK 1s optative/imperative prefix m-, pl.mn-, and the Itelmen transitive object suffixes 1s -mi, 1pl -mi(< *-mit?). Whatever the case, there would seem to be somewhat less of a problem as regards the 2nd person Nivkhci (pl. cin), corresponding to CK 2pl. turi you (parallel to muri) and 2s k/kt (on a demonstrative base k- like the1s form above). Note that the prefixed form is c-, alternating in the SS dialect with d- before a dental consonant (the prefixed1 s form is n- alternating with n- Mattissen, 2003:55). This may have influenced the 1 s form in Nivkh (so n(i)- couldoriginally have been from *m(i)r- just as c(i)- could have been from *t(i)r-). There is no prefixed subject form on verbsparalleling the 1s form in CK (1s marking is idiosyncratic in another way in Nivkh, namely as regards the converb suffixmentioned above).

    As regards the 3rd person, it is tempting to relate Nivkh 3s if or at least the corresponding prefixed undergoer formi- (j- before vowel) to CK n he, she. The principal evidence for this lies in the fact that i- as a possessor or emphatic/definite

    27 Compare winv track and atnv wound, mentioned in the section on lexical material below.28 ES regularly 2s/3s -r, 1s/all persons pl -t (or -n in the future or imperative combined presumably with intentional suffix -in); Amur -r/-t. The pluralperson form -tmay reflect the plural -tmentioned above in connection with nominal plural marking. I shall deal with the 1s form of this shape under verbal

    morphology below.29 Jakobson (1971:100) analyses it as a thematic verbal element -n- (cf. the attributive form of verbs in --) followed by a nominalizer -c (=-s in theAmur dialect, but -r in ES and SS).30 However, Krejnovic analyses the indicative/nominalizer as semantically someone/something that is /that V-s (Krejnovic, 1979:307f.), as if the final

    -t/d of *-ntwere an indefinite pronoun. This could reflect ES nunt/nud (indefinite) what, with the same ending as in demonstrative tud, Amur td this. Hestates that this form expresses more uncertainty than runt/rud, which is the form going with the Amur equivalent sid (and note also Amur n-d thing, do,mentioned above in connectionwith possessives). So there is in fact a possible link to CK q(ut) what (the last part is ut this bothwill be found in thesection on lexical material below), as in Chukchi renut what, something, and/or CK nik(ut)something. Note extended form nudlu someone/something(Amur sidlu). Nivkh *-nt is at all events a very general kind of nominalizer (also on transitive verbs, as in ES ind food from in- eat), and is in the right

    position vis-a`-vis the verbal stem to reflect an earlier nominal construction in Nivkh.31 The 1st person dual form, in A megi (ES me), is from *men-gin or *me-gin according to Panfilov (1962:238), where -gin is the dual comitative. Hissuggestion that the first part is mi/me two seems unnecessary.32 As I have previously suggested, the 1s and 2s pronouns in CK may actually have come from still earlier forms *t()km and *t()kt, with regular loss ofinitial /t/ in this position (Fortescue, 1997:374). Compare also Chukchi kur- < Proto-CK*tkur- go off.33 Thus nin one (person) (compare the impersonal/regal one in English). Also worth noting is the use of Chukotian antipassive prefix in- someone/something as a 1s undergoer prefix in inverse transitive verb inflections (Tailleur, 1960: 143 directly compares this to the Nivkh 1st personmarker). Cross-

    linguistically, 1st person is the highest candidate for subjecthood, and its function as object/undergoer is therefore prone to being somehow marked.

    M. Fortescue / Lingua 121 (2011) 135913761366

  • undergoer causes nasal sandhi in following plosives.34 The corresponding ES plural forms in(un)/irn they could then berelated to CK ()i they (< *n plus plural -i). The -f of if and the -v or -m of Amuric plural im/iv (also im) are moreproblematic, but they may well contain the place of -ing suffix -f ( It. -()en (comparethe 3pl form -kt, with additional plural -t > It. glottalized -n). In either case the second element could have been a lateraddition, influenced by the aorist. There is also an imperative/optative paradigmwith similar endings but different prefixes,namely 1sm- (It.m-), 1plmn- (It.mn-), 2s/pl q- (It. q-), 3s/pl n- (It. xan-). Georg andVolodin (1999:229) have argued that

    Table 3Chukotko-Kamchatkan verbal inflections.

    Chukotian

    1s t-V-R-k 1pl mt-V-mk2s V-R- 2pl V-tk3s V-R-n 3pl V-R-t

    Itelmen

    1s t-V-k(icen) 1pl n-V-k(icen)

    2s V-c 2pl V-sx

    3s V-()en 3pl V-()e?n

    34 Also independent pronoun if causes nasal alternation with some following endings (-doo, -da < -too, -ta). Its relationship to the ES and SS form ja isproblematic the latter may have a different origin involving interrogative/relative root ja-, as suggested by Austerlitz (1959a), or i/j- as above followed by

    (Amur) a who (cf. Otaina and Nedjalkov, 2007:1742; Panfilov, 1962:240 suggests ha that as the second part rather). Also Jakobson (1971:88) seesprepositive i/j- here. Note also respectful 3rd person forms av (pl. avun), Amur f. These suggest the -v ending in the (ES form of) ethnonym Niv, lit.inhabitant of (somewhere called) Ni, apparently containing -v- from pi-/-vi- be (in/at) and participial/attributive - (Krejnovic, 1986:163).35 nen in ES and SS, which suggests that the ending was originally *-n.36 The same in all persons, except 3pl -kt and 2pl -knitk (see below for the latter).

    M. Fortescue / Lingua 121 (2011) 13591376 1367

  • all the verbal prefixes in Itelmen are borrowings from Chukotian, which I regard as unlikely (cf. Fortescue, 2003:53 forarguments), but this has no great bearing on the comparisonwithNivkh, since it is clear, asmentioned, that the CK prefixes areinanycase later thanthesuffixesandpost-date thehypothetical Proto-CKAstage.Thecorresponding transitiveparadigmswerebuilt upessentiallybyaddinga3rdpersonpronominal suffix to the intransitive forms inbothbranches, extended to1st and2ndperson objects by integrating two different inverse constructions (involving originally antipassive prefix in- or generalinverse prefix n- originally a non-specific 3rd person subject marker). For details see Fortescue (1997). The parting of theway between the two branches occurred according to this model when the ergative construction subsequently developed inChukotian, superimposed on the otherwise thoroughly nominative-accusative morphology of the family.

    The basis from which the comparison between CK and Nivkh verbal morphology should be made then is prior to thebuilding up of transitive paradigms in CK. The essential repertoire at the pre-CK stage would have consisted of a basic two-way distinction between aorist (or perfective) and present (or progressive), the former characterized by suffix -R addedto the verbal stem compare ES Nivkh completive -Rar-/-Xar and resultative -Rare-/-Xare (the equivalent of -t- and-ta- respectively on the mainland).37 The suffix would have been followed by subject person/number markers. Note thatthe 3s ending in CK was (and is still) -()n, which simply represents the demonstrative n that/he/she/it. The present/progressive in -kn, on the other hand, apparently a participial form inflected only for number, may be bi-partite, asdiscussed in the previous section, with the first consonant comparable to Itelmen present tense affix -z- and the Nivkhconverb -r (likewise inflected only for number, apart from the 1s formmentioned above).38 Perhaps its original function wasone of temporal subordination, like the Nivkh converb. As the aorist person endings decayed already in pre-CK, prefixeswereadded for 1st person (1s t-, 1pl. mt-, the latter presumably with plural -t). The present/progressive further distinguished2pl subject by addition of the personmarker -tk following a linking or copular -i-, whichmay once have occurred alsowithother person subjects, but could be an innovation in Chukotian.39

    In CK a whole series of binary pairs of mood/tense forms emerged, respectively perfective and imperfective, as furtherparadigmswerebuiltup, for instance the future (involvingcircumfix- - want to, future cf. verbal stemjin()- desirefor the prefix, comparable to Nivkh (j)an- want) and the conditional (an extension of the optative40). Note that there arenumerousmodal and focal suffixes in Nivkh but these are not organized into inflectional paradigms like in CK (they can followgeneral indicative, converb or imperative suffixes, with the usualminimal person and number distinctions). CK languages alsodisplay numerous independent modal particles that overlap semanticallywith the scalar and focal suffixes/enclitics ofNivkhand include evidential/epistemic items. Althoughmood, aspect and tense do not form tight paradigms as in CK, there is a basicfuture/non-future distinction in Nivkh (future marked by a suffix) and aspect is also distinguished by suffix (including thecompletive suffix above). Moreover, it displays numerous converb suffixes besides -rwhich could in theory have developedfrom independent particles. It is reasonable to suppose, in conclusion, that Proto-CKA had simple intransitive paradigmsmarked by some kind of modal or aspectual suffix combined with person/number subject suffixes, like in (pre-)Proto-CK (asreconstructed in Fortescue, 2003:59) but that these become so attenuated in time inNivkh as to be replaced almost entirely byindependent pronouns (which further became prefixed/proclitic to the verb when used as object markers).

    Another central category of the verbalmorphology of both families is valency-increasing (and decreasing) by affixation. InCK this is a matter of causative/transitivizing prefixes or circumfixes plus an antipassive prefix, in Nivkh of initialconsonant alternation or a single (non-productive) transitivizing suffix plus a productive causativizing suffix. Neither familyhas a passive voice apart from an innovatory paradigm in Itelmen although the CK inverse constructions mentionedabove are very ancient (they are shared by Chukotian and Itelmen). Can these things be compared? Perhaps surprisingly, yes.

    Let us start with the situation in Nivkh. Most intransitive verbs start with a plosive and transitive ones with a fricative(though sonorant-initial ones can be either) there are many pairs of verb forms associated by this alternation. A priori onemight expect this to correlate with the use of the ubiquitous transitivizing prefix *n- in CK (Itelmen n-,< pre-CK *tn-41).In fact there are just a handful of cases where there is additional nasalization present in the resulting Nivkh transitives, as ine-mXaRu- make s.o. younger from paRa- (be) young, and e-mXaqu- shorten s.th. from pXaq- (be) short, which couldreflect the final nasal segment of *tn- compare also the difference between ES e-nra- aim at s.th. and Amur cognate e-zra-.However, all of these contain the (indefinite) 3rd person undergoer prefix i-/e-42 (j- before vowel). This morpheme isrequired in all cases where fricativization of an initial plosive cannot apply (apart from before single fricatives, nasals andliquids, when nothing further happens), i.e. before consonant clusters and vowels.43

    37 For the change of */a/ to / / in A and /Ra/ in ES here compare the discussion of cir/cXar tree in the section on sound correspondences above.38 According to Krejnovic (1979:321) the -r/t converb endings represent relics of person markers.39 The remaining 1pl suffix -mkmay have been influenced by 2pl -tk (dissimilation from *-tt?), but the k in both could reflect the old dual. As regards thelinking -i-, this is probably the same as copula verb i(t)-, which is comparable to ES Nivkh -i-, a verbal ordinal formant after numeral classifiers (Krejnovic,

    1979:305).40 Marked by a prefix (following person prefixes) R-, which, being of adverbial origin, may be compared to the Nivkh conditional converb in -qa/Ra.41 Note that initial fricatives in Chukotian generally derive from corresponding plosives, which are mostly preserved in Itelmen (Fortescue, 2005:7). The

    Itelmen prefix n-may reflect assimilation after loss of the medial schwa in *tn- compare the other form of the transitivizing prefix (appearing in certaincombinations) in-, with // regularly < */t/ next to a sonorant (the /i/ may be intrusive). The CK prefix itself probably comes from an independent auxiliaryverb nt- have (as) (

  • Now, although this prefix is generally described as indicating a 3rd person undergoer (or object), it is often glossable asindefinite (do to) something/someone and although its presence is not itself determined by definiteness it is precisely notpresent when a definite object precedes the verb and is integrated with it. As Otaina and Nedjalkov (2007:1718) put it: ifthere is no explicit direct object, a 3sg object like he/she/it is generally implied unless it is used in a citation form. In otherwords, the prefix is neutral as to definiteness, only implying a definite object in a suitable context where one is understood.Also Jakobson (1971:86) glossed this prepositive pronoun as someone or something. This is at least in part comparable towhat the CK antipassive prefix in-/ena- does to transitive stems, namely render the implicit object indefinite. It may infact be relatable to Nivkh ena- other, stranger (cited under the lexical correspondences below), whichwould be a reasonablesource for both an antipassive and an inverse marker.44 In that case this morphememight havemore than one form andfunction in Nivkh (respectively as stem and as prefix), and the CK inverse might in fact find its source still reflected inNivkh, namely in the common antipassive construction.45

    For a specific 3s object or undergoer there is also a different i- prefix in Nivkh, often called emphatic (as inMattissen,2003:57f.) but perhaps better definite, one that causes following nasal rather than fricativizing alternations,46 just likethe 3s pronominal possessor discussed in the previous section in connection with CK pronominal n, so possibly from *in-.Actually Iwould suggest that the ultimate source is CKA *hn, related toNivkh demonstrative root h-/ES hu-, and it is such aform that can be compared directlywith the CK 3rd person root n-. I shall return to this relationship at the end of the sectionon lexical correspondences in connection with h-initial morphemes in Nivkh in general.

    The oldest layer of transitive verbs in Nivkh could still in theory reflect a pre-Nivkh transitivizing morpheme *tn-, as inpre-CK, but thismay have been obscured by the later diffusion of *inL- (> CK *in-), coming to apply to already transitivizedstems (an initial plosive would automatically become fricativized following the final vowel of *inL- according to the generalrules of sandhi in Nivkh). There is some evidence that this might well have been the case, namely in the relic transitivizingsuffix -u- (as in nok-u- make s.th. narrow from nok- (be) narrow) this corresponds nicely to the suffixed part of the CKtransitivizing circumfix n- -v. The prefixed part (corresponding to initial fricativization in the more productive processdescribed above) could well reflect the loss of *n-, as can be seenwith transitive verbs of this kind starting with plosives, e.g. faz-u- undress from intransitive paz- and Xav-u- heat up from qav-.47 This -u- can also verbalize nouns, which CK-v- also can. Recall that prefixes are generally later phenomena than suffixes in CK, and thus the suffixed part of verbalcircumfixes like this can be taken to be older than the prefix. The use of *inL- to mark an (indefinite) undergoer in Nivkhmight then only have been necessary with stems whose initial consonant or cluster precisely could not be fricativized(after the suffix was no longer productive). The modern 3s undergoer prefix may thus represent the conflation of twodifferent morphemes, definite and indefinite respectively.

    The modern Nivkh forms and processes mentioned so far are all non-productive to form causative (as opposed tomerelytransitive) verbs productively, the suffix-ku- is required (applicable to transitive as well as intransitive verbs). Krejnovic(1979:314)suggestsaconnection to thecomitativeaffix -ku-here,48butperhaps itmaybedirectly comparedto thecommonCKsuffix -tku-, which is essentially a frequentative, but may also indicate protracted or intense activity and is used as a generalantipassive (of potential object) it means act or do with on nominal bases. Finally, the reciprocal prefix u-/w- may becompared to CK uviki body, self, which functions as a reflexive pronoun and may well go with Nivkh vic, ES ut body.49

    5. Lexical material

    Rather than simply presenting a list of lexical look-alikes in the manner of Tailleur (a hazardous approach at best), let mestart by singling out a handful of particularly good candidate cognates among basic items of vocabulary. These relate to thecommon hunting/fishing way of life that goes back to Neolithic times in the region. Consider then the following group offorms, which contains one of the derivational morphemes introduced in the section on nominal morphology:

    Nivkh (Amur) tf,50 ES, SS taf house (stem tav-, ending in suffix -f/v placewhere) can be comparedwith Proto-CK tvanvplace (with the same suffix -nv place of on stem tva- be, live, sit note also Chukotian (t)varat people, from the same

    44 Thus he does/you do something to me would be expressed in CK as he does/you do something to someone else, with *in-/ena- as undergoer (the 1sobject being higher on the animacy/person hierarchy than the subject). Why the dominant form of this hypothetical common morpheme should have

    become lexicalized in Nivkh requires explanation but note that in the correlated Chukotian form ena-ra-lRn neighbour it is due to the dominant vowel offollowing stem-ra- house; perhaps Nivkh *tav of that meaning (as in Amur enadv- another house) was also dominant, as the reconstructed vowelsuggests (ena- is, however, a fixed form today).45 Perhaps better called generalized undergoer construction, since the term antipassive has become rather too closely associated with ergative

    languages. Also non-ergative Itelmen, Yukaghir and Na-Dene languages, among others, have similar morphemes (Fortescue, 1998:61).46 This causes a following plain (non-aspirated) plosive to be replaced by the corresponding voiced plosive in the Amur dialect and is generally triggered by

    a final nasal that may or may not itself have dropped (it is generally retained in the ES and SS dialects).47 Note also si- (fricativized from ci-) put, which can be compared with CK il- of that meaning listed in the section on lexical material below.48 Used before a following converb ending -r it forms adverbials and may contain initial -, the attributive form of verbs in ES.49 If not Itelmen reciprocal prefix lu-. Otaina and Nedjalkov (2007:1742) tentatively suggest *un- as the source of the Nivkh prefix u- (where *-n would

    represent the plural), since it causes nasal assimilation of the same sort as 3s i- above. Compare, however, West Sakhalin ut-azi- be of the same height(as), corresponding to Amur u-zi-with reciprocal u- (Otaina and Nedjalkov, 2007:1728). Note that u- is also a reciprocal prefix in Ainu (just as i- is also a3rd person undergoer prefix) this could represent early borrowing from Nivkh during Okhotsk Culture times in northern Hokkaido.50 Also raf (model) mortuary house (which may reflect an original indefinite possessor form *i-taf according to Jakobson (1971:91). But note also CK ra-

    house (and r- enter, raan entrance), as Tailleur suggested Jakobson (1971:99) sees a possible verbal stem ra-/ta- install here, also reflected (withreflexive prefix) in pr- arrive.

    M. Fortescue / Lingua 121 (2011) 13591376 1369

  • stem plus -rt set (of)).51 So the original house word in both families could have been *dvanv, parallel with Chukotianwi-nv track, mentioned earlier, the stem ofwhich goes perhapswith Nivkhwi- go. Also Nivkhwo village (actuallymore agathering place for the winter, away from the summer fishing sites Austerlitz, 1990b) has a possible CK cognate. AsAusterlitz suggests, this word (beginning, unusually for a noun, with /w/) probably contains the reciprocal prefix w-/u-before a hypothetical stem o- meaning gather, as also reflected in verb w-op-u- of that meaning. It looks as if a finalnominalizing element has been eroded, so why not -f/v place again? It is at all events attractive to relate this to CKomak- together, transitive -umk-v- gather (and Chukchi umek group), so Nivkhwo< *w-umk-()nv, via assimilated*wup-nv?52

    Now Nivkh (and Itelmen) summer houses are typically situated near the mouths of large rivers (Black, 1973:6). Considerthe following group of potential cognates: Nivkh eri, ES and SS i river, CKr- flowout and It. i water.53What could bemorenatural than that a river (*Lr-i) be regarded as water (*i) that flows out (*Lr-)? There is also the word for sea itself,ke-rq (where (he)rq is side, direction cf. Austerlitz, 1959b:220), to which can be compared Chukchi e-curmn shore(< *ke-curmn, where the second elementmeans edge).54 Further derivatives of the *Lr- root in Nivkh are: r, ES r mouthof river and rk, ES rkr shore, bank, which can be compared to CKr-rn andr-i of precisely the same meanings(recall what was said in the section on sound correspondences about reflexes of */L/in Nivkh). The suffixes in CK arerespectively the nominalizer -rn treated under nominal morphology and i below. Probable cognates fordownstream and upstream will be found in Table 4 below.

    Another promising lexical area is that of fish more specifically salmon, with which the rivers of the whole area teem.Thus compare: Nivkh co fish and CK wcu salmon, q(w)cu humpback salmon, It. cuv(aj) salmon (chinook). All ofthese could contain an original stem *wLcu (i.e. *wLtu). This assumes syncope and/or assimilation in both Nivkh andItelmen (which would explain the ejective in the latter cf. Fortescue, 2003:67).55

    As regards inland hunting as opposed to fishing, compare Nivkh -/ES aan- hunt (the ES form contains a- landanimal), a(i)- go for, chase after with. Chukotian -rt catch (It. es- hunt), and ta- go for56, perhaps all from root*ra-. A further basic word, relating to the spoils of the hunt, is Nivkh in-/ni- eat (the i- is the 3rd person undergoerprefix), comparable with CK nu- eat,57 perhaps both from *nu-, with assimilation of /u/ to /i/ in Nivkh.

    In Table 4 below I list most of the reasonably good i.e. more or less regular correspondences that have turned up so farin the hope that they will form the basis for a more far-reaching search in the future. The Proto-Chukotian/Proto-Chukotko-Kamchatkan forms are from Fortescue (2005), and the (Proto-)Nivkh ones are the oldest forms to be gleaned from themodern dictionary and word-list sources (primarily Saveljeva and Taksami, 1965, 1970, Nakagawa et al., 1993, and Tangikuet al., 2008). Basically this means ES if there is any discrepancy from corresponding Amur forms.58 Lack of a gloss underProto-Chukotian indicates virtual identity to the gloss under Nivkh. The older Itelmen forms (in particular those from theextinct eastern and southern languages, E. It. and S. It.) are not always reliable as to form. An apostrophe after aconsonant indicates an ejective in Itelmen, not an aspirated/fortis one as in Nivkh (except for /d/ and /n/, which are palatal inNivkh, as is /t/ in reconstructed Proto-CKA forms).

    Let me now sum up the principal evidence here for the sound correspondences given in Table 1 in the light of thehypothesized sound system of the proto-language presented in Table 2. It would appear, if these correlations hold, that theremust have been considerable contraction/syncope in Nivkh, involving the weakening and disappearance of intervocalicconsonants (especially voiced fricatives and liquids), also the reduction of clusters involving sonorants (nasals, liquids andsemi-vowels). The basic shape of nominal stems in the proto-language seems to have been (C)VC(C)V(C), and of verbal stems(C)V(C)-, with no long vowels and no clusters of more than two consonants (the latter medial only). The vowelcorrespondences are fairly regular, with allowance made for instances of labial and dental assimilation (in particular ofschwa to /u/ and /i/ respectively).

    First as regards the plosives. Note that the difference between voiced and unvoiced (or lenis and fortis) is overwhelminglyconfined to initial position in Nivkh, and so probably also in Proto-CKA. For the labials (collapsed to a single series in CK, as allplosives) see: *pana, *pki(r)-, *pt-, *bl-, *blm-, *hp-, *ba-, and *bla- above. For the dentals see: *dvanv, *dR-, *diwl-,*tlv-, *pt-, *tem-, *duv, *du-, *dod(o), *til-, *tek(a)-, *tl, *ta(la)-, *tL(q)-, *dLqL, *d(ak)- and *dvi-. For thepalatals see: *wLtu, *mtRal, *t(n)ti(l)-, *dLdi-, *tevar-, *dm(i), *mLtv-, *t-, and *tev-. For the velars see: *pki(r),

    51 cf. Itelmen cognates a- sit, a-nom seat (-nom is the Itelmen reflex of *-nv) and compare in turn Nivkh tiv-/riv- sit on and tifc chair (with anothernominalizer).52 Compare more recent productive form wopuf meeting, containing precisely these three elements.53 The latter form and its relationship to Chukotian presented somewhat of a mystery in Fortescue (2005:398).54 Perhaps also Itelmen qiX sea, (with assimilation to the final uvular /X/?) earlier forms attested for Itelmen are kix, kejaga and kaijan (see Fortescue,

    2005:387).55 Other kinds of salmon one might catch in the area are Nivkhwel summer salmon andweck kind of Siberian salmon, comparable to It. fec/vilc of thatmeaning, and/or fackc loach, golets (the -c is a singulative suffix), also Nivkh va()s kind of salmon, sig and Chukotian wit(wit) small kind of salmon,golets.56 Also perhaps rnik animal and r-oR- give birth (animal) for the second element compare Nivkh qo-/eR- give birth to (an ablauting verb wherethe initial e- is the form of the undergoer prefix required by vowel harmony), as in *qola in the table below.57 Which is transitive and seems to contain the transitivizing prefix *n- discussed under verbal morphology (as reflected e.g. in Koryak ju- < *u-), soperhaps < *t(n)nu-.58 Proto- is in quotations here, reflecting the shallow nature of the comparative base, as discussed in the Introduction. ES forms without attested

    equivalent in A (or vice versa) are marked as such. An oblique separating forms otherwise indicates morphophonological variants.

    M. Fortescue / Lingua 121 (2011) 135913761370

  • Table 4Further potential Nivkh-CK cognates.

    (Proto-)Nivkh Proto-Chukotian/CK Proto-CKA

    rf wounda atnv *atnv?al-r berryb l-u- gather berries *Ll-?alv-erq behindc lv- other (It. elveze- go back)d *LlvL-?ES azmc man (male)e rm() leader, strong *LrmL-?pil- big It. pl- big (Ch. pl- completely) *bl-?pla(q) leaf It. pla(l) *bla-?polm- blind plm- dark from rain/snow, blind *blm-?(he)rq() side, direction q edge *dLqL?rk-/-tk-/ES rak- carve, cut - dig, scratch *d(ak)-?rru-/-tru- untie, release r- *dr-?taX epidemic, cold, influenza tRl- ill *dR-?tvi- end -tvi- become *dvi-?irl-/-tl- pull, drag iwl- (It. timp-) take (across), haul *diwl-?tot (upper) arm E.It. soto/soto armf *dod(o)?

    t-/ES tu- this ut thisg *du-?tuv/ruv elder brother, tum friend (It. tumx *duv?blood relative sibling of same age)

    SS dicm-/ES zicv- tread/A zit- kick tti- tread (Kerek caci-tRu-kick) *dLdi-?com() raft timi *dm(i)?ena- other, stranger ena-ra-lRn neighbourh *ena-?km- scurry back and forth (of insects) kame(cat)- move around (and kmk beetle) *gLm-?kit-/ES kitn- run away ntv-i *gnt-?kur- God, the worldj k(R)ukl creator, Raven, It. kutx legendary creator of Kamchatka *gRu-?qa- go downriver qalelle() downriver, downhillk *Ga(l)-?qalRala-l bright qe-light *Ge-?hawa- open mouth widem awa- open (e.g. mouth) *hawa()-?hup- tie p- be attached, penetrate *hp-?hir stomach of animal ji(ji) gut *hi-?hilx tongue (and jelel-/ jil(jil) tongue *hil?helel- lick)

    him cedar thicket im- dense, frequent *him-?

    ur island ilir(i) *ilur(i)?

    ki/xi- up kwji go upn *kLwji-?klmr navel kil(kil) *kl-?im-/-kim- giveo kmit- take *(L)kmi(t)-?kzm whitethornp kilam, E.It. kerem whitethorn berry *krm?ks-/ES kjru- happy krvi- *kr(vi)-?kr- be hungry t(Rt)- *kt-?le(le) very li really *li?lirq wolf l()Ri(n) (It. Xine) *lRi()?if heartr li(li) *li-?ma(g)- strongs mj- big *mLj-?moc/ES mc breast (womans) macve chestt *mLtv-?ml- tow along shore mu(lRt)- wander, go off faru *mu(l)-?ml wooden figurine in It. mila-cX baby, CK ml- fine, *ml-?commemoration of the deceased small

    mla ear (ES; and m-hearv) vilu(ln) ear (and valom- hear) *mlu-?moq- break into pieces mq- small *mq-?mezla()w rowan mic(Ral)/mc(Ral) (It. mce-) *mtRal?nana recently (naf now) It. nen, S.It. dani now *na(n)-?

    nui in front, first E.It. duk- in frontx *nu-?nonqy young (animal) (u)nn child, baby *nun(L)?(SS baby)

    altr/af side (of body) all on both sides *al-?ar skin (animal) (SS nl(n) skin, hide *Ll-?alr skin) otter ntz *Lt?oi penis oj(n) tail *oj()?pan gaiters, leg coveringaa pana (boots of) leg skin of *pana?

    reindeer

    px- return pkir- come *pki(r)-?pt(u)- split -pt piece of *pt-?qa (noun)/Xa- (verb) name It. Xela()/Xaln *qela-?qau(k) no, not haveab qm/It. qa?m no, not *qa(m)?qav- hot (Xav-u- heat qevja- (let off) steam, scald *qjv(a)-?up) self (It. qejve-/qevu- boil)

    oRla childac Rola boy *qola?za-/-ta- beat tala- *ta(la)-?

    ta-/ra- where, which q what (qu why, *tL(q)-?ad

    (and si-/NS su-/ru-/ tRr how many, It. saq

    M. Fortescue / Lingua 121 (2011) 13591376 1371

  • Table 4 (Continued )

    (Proto-)Nivkh Proto-Chukotian/CK Proto-CKA

    ES ru-/nu- what)ae what kind)

    te-af go up on land It. tekej- get up *tek(a)-?tom(s) smokehole in old It. temes roof (tem- cover), *tem-?

    style house Kamen Koryak tome- stop upsmoke-holeag

    t(lf) far tel(), It. tal-ah *til-?ruv-/-tuv- burn tlv- *tlv-?r door l- crawl into (sleeping compart-ment)(E.It. solo-nac lower door)ai *t(l)-?si-/-ci- putaj il- put (down) *t(n)ti(l)-?cev- spear, pierce cvi- cut, chop *tev-?cevcev- wet It. cufcuf rain *tev(tev)?

    cavr- grey cevaro *tevar-?

    co- melt tl- (It. co-)ak *t-?cuz-/ES cir- new tur- new *tur-?

    u- burn (intr.) uji- make fire *u(ji)-?

    uii- disappear, not be uj (there is) no *uj(L)?any

    vic, ES ut body uviki *uvi-?

    vukvuku- get darkal vulq- dark *vul-?wal- cut, chop (and wa wala(-) knife, carve *wa(la)-?

    sword, sabre)

    avlx lipam wmlk(t) *wLmlkL-?panx soup pa() (drink) soupan *ba-?nin/ES nen() one (person)ao nnn *n-?t()k father tl(n), It. isx *tl?a Containing -f (CK -nv) place of; also reduplicated rvrv- in ES as a verb. Note that initial /t/ becomes /r/ after a vowel in certain circumstances in Nivkh

    (/r/ as all fricatives is rare as the initial segment of non-derived nouns cf. Jakobson, 1971:93).b The -r (A -s) in Nivkh is a nominalizer (Saveljeva and Taksami, 1970:532), and the -u in CK is acquire, consume.c (h)erq is side, direction.d Note also CK javal(a) back, behind and ajval- lee(side) (metathesis?).e also perhaps Amur ar, ES ara male (the latter with a animal); but note also s/z master, owner (and the CK // Nivkh // correspondence

    discussed in the section on sound correspondences).f For the s-/s in It. (via //) compare under*t(q) further down in this table this is not regular, but note that /s/ is the usual reflex of */t/ next to another

    voiceless consonant.g Perhaps with metathesis (and influence from un (that) over there?) cf. lative -. Note also Sedanka W. It tin, E. It tyj this.h -ra- is house and -lRn onewho has/lives in . Compare also Alutor ina-ra-k in a neighbouring house with Amur Nivkh ena-dv-uin in another house

    (-dv- is house).i With suffix -v; note that Nivkh medial /nt/ is sometimes reduced to /t/, as in nominalizer -nt/t/d (cf. Gruzdeva, 1998:11).j Originally in Nivkh mythology there were four spirit masters (of the sea, the sky, the fire and the mountains) referred to by this term (cf. Shternberg,

    1999:158,165). The CK form is a personal name, not the common noun for raven.k Cf. perhaps CK qal(a) area around or beside, also lative ending -.l This presupposes an intermediate stage *qlla in Nivkh. Saveljeva and Taksami (1965) also have qalala-/kalRala- there appears to bemetathesis of a

    kind in the second form; the -la- is a suffix indicating a permanent property (cf. Mattissen, 2003:18), which may in turn have affected the *//in the first

    syllable.m And Amur havahavad have breathing difficulty (with mouth open); cf. also perhaps havaf lung (only North Sakhalin has attested/w/here, so there is

    some doubt about whether /w/ or /v/ is original).n Note also ke- upstream, going perhaps with CK krol(a) on top, upstream (Chukchi roca) for the ending compare CK knlo high.o The second form is with prefix and metathesis compare old South Itelmen form emgatyz I take. For the seemingly contradictory semantics of give

    and take compare the situation in Indo-European discussed by Buck (1988:748) under give in terms of a common meaning stretch out the hand.p But also kelm raspberry (bush).q The -r (A -s) is a suffix typical of animates acc. Taksami (1983:286); the -n in CK is a singulative suffix.r cf. -f from *-nv place of-ing, so perhaps from a verbal root *l(i)i-.s Also man very.t And E. It. ki-msevi-in big breasted; the Chukotian may have dominant /a/ due to original dominant singulative -ln, still found in the Koryak.u And Ch. muul?n (reindeer) caravan.v SS mu-/mo-.w West Sakhalin medlan acc. Glehn (in Schrenck, 18921900).x West It/n/corresponds to East (and sometimes South) Itelmen /d/.y The following uvular perhaps causing the change of */u/ to /o/ (there are parallels in the modern language); but note also neneq a little.z Reduplication of *t.aa Austerlitz (1984:40) relates this to Nivkh stem pa-/fa- put on shoes, socks, footgear.ab Takahashi (1942) has SS kavrnt/kavr (for qavrnt/qavr) from *qarv- or *qarm- by metathesis?.ac A oRla, ES eRla (the - is the suffix lost on the mainland discussed above); cf. qo-/eR- give birth to.ad *t(q)- appears to be weakened in most forms to *q- in Chukotian (jaq in Koryak, taq in Alutor, saq in It.) and (in some forms) to ra- in Nivkh.

    Jakobson (1971:92) sees the prepositive indefinite pronoun i-/j- as lying behind this ta- > ra- alternation (so perhaps also behind A si- what?).ae Also ES tamdid how. The what forms are problematical, but note that /s/ alternates with palatal/c/in Nivkh.; the forms with /u/ could be influenced

    by demonstrative root tu- this, here. The nu- form in ES is described by Krejnovic (1979:306) as specifically completely indefinite and thus may contain

    n- thing cf. E. It nike/nakc what, something.af The form given by Gruzdeva (2008:187) for ES compare A tea- run up on sandbank (boat).ag Chukotian top- cover, close - compare also Nivkh to(rf) semi-subterranean house.ah The Chukotian contains dominant allative suffix -() (and the Nivkh contains spatial suffix -f).

    M. Fortescue / Lingua 121 (2011) 135913761372

  • *kLwji-, *kt-, *kl-, *gLm-, *gRu-, *-tk(n), *kr(vi)-, *(L)kmi(t)-, and *dk-. And for the uvulars: *qjv(a)-, *Ge-, *Ga(l)-,*tL(q)-, *qela-, and *qa(m).

    As regards the corresponding voiced fricatives, compare: *dvanv-, *dvi-, *qev-, *vul-, *v-, *tev-, *tevar-, *LlvL-,*dvi-, *uvi-, *dr-, *Ge-, *qm, *lRi(), *li, *mu(l)-, *tl, *l(r), *qola, and *t-. And the correspondingnasals: *matv, *mtRal, *mlu-, *mu(l)-, *mq-, *mLj-, *mLtv-, *ml-, *tem-, *dm(i), *him-, *nun(L), *ena-, *nu-,*nu-, *na(n)-, *n, *al-, *ra-, *Lt, *oj(), *du-, *Ll-, *qal-, and *li-. And the semi-vowels and liquids (including */r/): *wLtu, *wi-, *wa(la)-, *hawa()-,*oj(),*uj(L), *lRi(), *mtRal, *diwl-, *ml-, *tel-, *vul-, *wa(la)-, *qal-, *LlvL-,*bl-, *bla(), *hil, *tavar, *dr-, *krv(i)-, *Lr-, and *ilur(i).

    There remainsmore to say about the problematical /h/. By far themost important rootwith this initial segment inNivkh isdemonstrative h-/ES hu- that and related proverb/auxiliary ha- do thus59, which appear to go somehowwith CK in(in)such (< *hin-?). Compare also Alutor ina over there, Itelmen xe(j)nin such and (em)xenin the same. Thereconstructed CK form is related to CK i now and the ending -in goes with demonstratives like an(in) that and unover there compare the latter with ES hu- that, parallel with tu this going with CK ut from *du on Table 4. Itelmenforms like xejnin show a new voiceless fricative deriving from // in initial position (after loss of initial vowel). In Nivkhinitial /x/ as opposed to /h/ is also innovative, deriving from /k/ under sandhi/transitivizing conditions, but this isobviously not the source of /h/ here.60 This leaves two possibilities (if the correlation is genuine): either the proto-formwas*hin-, the initial being preserved in Nivkh and lost in CK, or the/h/in Nivkh derives from *i- before a vowel with loss of /i/and an idiosyncratic reflex of //.

    To decide which of these possibilities (if either) is correct we need to consider other examples of correspondences in CKwith Nivkh /h/ in Table 4 above, namely hup- tie, hawa- open mouth wide, hilx tongue, him cedar thicket, and hirstomach. There is no obvious way the CK equivalents could come from *i- here (there are plenty of words beginning withthis sequence, though it is rare in Nivkh), so we are left with the first possibility: CK probably lost (initial) /h/, a very commonphenomenon in the worlds languages (it is also dropped in Nivkh following prefixes). Another possible h-initial transitiveverb in Nivkh corresponding to a CK vowel-initial one is hupu-/jupu- dip or plunge s.th. into compare Chukotian up-/jup-push (in), stick in (also related ojp- stick through). As regards the tongue word (which always occurs reduplicated jiljil or with a suffix in Chukchi), compare the Itelmen cognate cel (and other older forms in Fortescue, 2005:115), whichlacks the usual reflex of initial */j/ (/s/ or /z/), so this could be secondary, due to the reduplicated form once the initial */h/ waslost. The stomach word hir stomach of animal (with suffix -r) corresponds to Chukotian ji(ji) gut, which is alsoreduplicating but does have the regular reflex of */j/ in initial position in Itelmen (zives/siis). This is problematical, but itcould reflect original *hji or the like, just as tongue could be from *hjil (and note helel-/jelel- lick, suggestive ofreduplication also in Nivkh).

    Let us return now to hypothetical *hin-. When dealing with demonstrative/pronominal items like this which have highfrequency reflexes in all the languages being compared, it is important to look at awhole range of related forms to seewhat isgoing on. Thus Krejnovic (1979) has for Nivkh, besides the forms given, Amur ho()-/hoRo-, NS hpr-, and ES hmci-/hmra-do thus. Further, Nivkh root h-/ES hu- actually appears as hd/hud61 when functioning as an independent demonstrative,and this should also not be seen in isolation, for there is also t(d)/ES tu-/tud this, which is comparable with CK ut thisunder *du- above (contrasting with un (that) over there). Note also Nivkh hur-/ES hus- as the dummy object of certainverbs like hur-tov- tie to and hur-tiv- sit somewhere (cf. Gruzdeva, 1998:29), also reduplicated hurur everywhere,hurmiin, ES huzmi inside (from mi- inside), Amur huin, hug/hur, ES hunx/huz there, Amur tur/tuin/tug, ES tus/tunk/turhere. The same locative base may be involved in verb stem hum-/ES hunv- be, live.62 Moreover, there is Amur aehd, ES

    ai Secondary way out of the semi-subterranean house. It is the second element that means entrance (traditionally through the roof) in Itelmen. Note

    Chukotian tll (< reduplicated *tltl) door, but also Nivkh r-/-t- push in or through. Otherwise Jakobson (1971:94), who suggests a possible earlierform of the word *kr.

    aj si- is the regular fricative-initial form of this transitive stem (and ci- = *ti- the form occurring with suitable incorporated direct objects) whereas CK

    il- contains transitivizing prefix *t(n)- discussed in the section on verbal morphology.ak The Chukotian presupposes intermediate *tl- or the like. Note that the Nivkh formwith -- rather than -R as in Saveljeva and Taksami (1965) is from

    Austerlitz for SS (1984:45).al But wlwl- dark, black. There seems to be entanglement between two stems here (both reduplicated).am ES also amavli.an And pa(t)- boil, cook; also CK p(pt)- boil, It. paxpa boiled fish.ao The final -nmay, as discussed above, represent an earlier animate class marker, whereas the -nn of the Chukotian could be from ln simply, only.

    There is little else obviously in common between the numerals of Nivkh and CK, but nor is there between Aleut and Eskimo (just one and four), which are

    known to be related. Possibly Nivkh ca-qr three and n-kr four (the forms used for various objects) could be related to CK ro(q) (It. coq) three and ra(q) (It. caq) four the classifier -qr in the Nivkh forms could have undergone metathesis after loss of the medial vowel. The Chukotian roots are actually

    ro- and ra- (compare also Nivkh co-r and nu-r of animals, with an animate classifier -r Panfilov, 1962:202f.).

    59 Note also -hagin, ES -zi any (one/thing), haan (Taksami, 1983 has hoRar) then (ES ha), with recognizable converb endings.60 Note that verb stems in Nivkh that begin with /h/ quite regularly replace this by /j/ (the undergoer prefix) when used transitively - except when the

    undergoer is an incorporated noun (Mattissen, 2003:124f.).61 Krejnovic (1979:305) has ES hnt, hunt that, with what looks like the same indicative/ nominalizer suffix -d/nt seen on verbs.62 Though the suffix may be -m-/-nv-, which derives verbs from local roots. The -v- could reflect pi- be if not jiv- have, be.

    M. Fortescue / Lingua 121 (2011) 13591376 1373

  • aixnt that over there (distant) (with a deictic prefix a-) Krejnovic (1979:305) has ES exnt, ehd, ahd that (a bit further off)(=Amur ad), exnt (Amur ahd) (still further off), and aixnt (Amur aehd) that far away, all of which need to be comparedwith *h-/hu. There is ample scope for explanation of the interrelated forms here (e.g. by way of metathesis, assimilationand analogy), but this is a complicated matter and calls for detailed research, which I can only sketch here (see Panfilov,1962:240ff for a discussion of the relevant forms). The same applies to the interrogative forms listed in Table 4 (see Panfilov,1962:253ff).

    One important outcome of this discussion of /h/ is corroboration of the suggestion that CK 3rd person demonstrativen could be from *hn, going with Nivkh h-/hu-, as discussed briefly in the section on verbal morphology. Note that theChukchi forms that go with CK in(in) are actually nin, not thus. In fact it is only Koryak and Alutor that have thefull reflex inin (from or influenced by i now). Koryak also has nnohe, she, it (< *n-ot, with an expressivevariant of ut there) alongside nRan thus, and Alutor has pronoun nno besides demonstrative n-in(a) that(=Koryak nnin, It. nnu), which further appears to contain possessive ending *-in. So the forms to compare with Nivkhh-/hu may well be CK nin and nut. This (reflecting a common root *hn-) would better explain the forms like Nivkhhu-g/ES hun-x, CK n-k there mentioned under nominal morphology in connection with the old CK locative casemarker.