J. Linguirlics32 (!996), r-17. Copyright @ t996 CambridgeUniversityPress
Associative forms in a typology of number systems: evidence from Yup'ikl G RE V I LLE
G. C OR BE T T
AN D M AR IA N N E
MITH U N
Universitlt of Suney
University of Califumia, Santa Barbara ^nd (Received z5 May r995; revised z7 September r995.;
A g€neraltypology of number systemshas to coDfront the problem of variation both in the number valuesin diffelent larguagesand in the inventoriesofnominals involved. We start from the Smith-Stark Hierarchy and extend this approach to additioaal numbers (such as dual and paucal). Associative plurals appear to undermine this typology, if we tueat them as a tiird mrmber. Either the associativeplual or the ordinary plural proves to be exceptional. The morphology of associativemarking in Central Pomo and in Central Alaskan Yup'ik suggeststhat this is the wrong approach. In Central Pomo the associative contains an original plural segment. More significantly, Central Alaskan Yup'ik provides the ideal combination of three numbers and complex morphology to demonstrate that associatives should not be troated as additional numbers. Associativity atrd number are realized separately,\trhich shows tlat associativesare a separatecategory interacting with number. This allows us to maintain the typology proposed. I. I NT RO DUc T I o N : R E Q U IR E M EN TF So R A TY P oLoGY oF N U MB E R SY S T E M S The attempt to construct a typology ofnumber systemsappears to be greatly complicated by the existence of associative forms. These forms consist of a nominal plus a marker, and denote a set comprised of the referent of the nominal (the main member) plus one or more associated members: Hrngaian Jdnos-ik ("/dzos-ASSOCIATIVE)' John and his family/friends'. We shall show how the typological problem arises and then, with data from Central Pomo and from Central Alaskan Yuo'ik. demonstrate how it can be
[r] We owe a great deal to Edith Moravcsik, who drew our attention to the importance of associativeforms, initiated an email discussion on the topic, atrd provided many helpful comments,also to David Gil for exteNiv€ stimulating discussiotr,to Totry Woodbury-for discussionofthe history ofsome Yup'ik forms, atrd to two allonyroous "/Z rcf€reesfor their remarks. We are also grateful to Frans PlaDl for organizing a sessiooon associativesat the Pre-Inaugrual Meeting of the Association for Linguistic T,?olo$/, Konstanz 29 Nov.I Dec. 1994,where a version ofthis paper was given. A revisedversion wasprcsentedat the Litguistics Association of Great Britain, Essex,r 8-2o Sept. 1995;we wish to tha[k those present at these meetings for helpful discussion. The support of the ESRC (UK) under grant Rooo236o63and of the Academic Senate,University of California, Santa Ba$ara, is gatefully acknowledged.
G REVI LLE G . CO RBETT AND M A R I A N N E M I T H U N
resolved. We shall be concerned with nominal number: verbal number (quantification over events or participants) is also a complex problem, but not one for which associative forms are of direct relevance. Analyses of verbal number are given in Durie (1986)and Mithun (1988a);for nominal and verbal quantificationmore generallyseeGil (1993). In trying to establish a general typology of nominal number, a major difficulty is the number of variables. We find that the number of number values varies dramatically from language to language, from those with singular and plural to those with additional duals, trials, quadrals and paucals.At the sametime, the number of nominalsinvolved in the nunber systemshowsconsiderablevariation too. A reasonableapproachis to usea version of what has been termed an Animacy or Topicality Hierarchy to determine the nominals with the highest involvement in the number system, and from theseto determine the possible values (singular, plural, dual, etc.) of the number category in a given language. The possible inventories of numbervaluesare subjectto a typology of number systems. We therefore outline the hierirchy, in the version proposedby Smith-Stark (1974)in sectionz, beforegoing on to the problemwhich associativeplurals causefor suchan approach (section3). We consider data from Central Pomo which suggesta possiblesolution (section4). Then we move on to even clearer data for resolving the problem from Central Alaskan Yup'ik (section
z. Tn p Srr,I t r n- STARK HI ERARCHY NUMBERS
AN D T H E P R o B L E M o F M I N o R
We wish to establishthe inventory ofcount nouns of particular languagesas a basis for establishing the possible pattems. In part this is an oversimplification, since Allan (198o) has argued that countability is a characteristicofnoun phrasesand not ofnouns. However,ashe points out: Even though countability is characteristic of NP's, not of nouns, it is nonethelessa fact that nouns do show countability preference- insofar as some nouns more often occur in countable NP's, others in uncountable NP's, and still othersseemto occur quite freely in both. (I98o: 566). Thus strictly speaking we are investigating the countability preferencesof nominals. In doing this, Smith-Stark (1974) proposed the following version of the Animacy Hierarchy: (t) The Smith-Stark Hierarchy speaker> addressee> kin > rational > human > animate > inanimate (rst person (znd person pronouns) pronouns)
ASSOCIAT I VE FO RM S: EVI DENCE FRO M Y U P 'I K
that plurality 'splits' a languageif it is a significantopPosition He suggests for certaincategoriesbut irrelevantfor others' (1974:657).Plural numberis distinguishedin some languages,for example,on first and secondperson pronouns and nouns referring to rational beings,including humans' but not those referring to animals or inanimate objects. While languagesmake the split in different places,the claim is that the inventory of items which show plurality (whether through direct marking or indirectly through ageement) must involve sometop portion ofthe hierarchy.Thereis a good deal ofdata supporting his claim, though he also notes a small number of less straightforward cases. Consider now the question of further number values. It is possible for a language to have an additional number value, say the dual, which pattems with the plural (that is, the same 'split' applies to dual and to plural) Typically in sucha languageif a noun has a dual it also has a plural, and vice versa. Smith-Stark assumedthat this was the normal situation in his brief considerationof numbers other than the plural, though he did note a potential problem {t974: 669fn. 6). In fact, there are severalinstancesof such number values which may pattern in ways that are not in accord with the Smith-Stark hierarchy. Theseare found for a relatively small proportion of the nouns of a given language,and we have termed them r"nxon I'uIr,mexs (Corbett 1995,sectionz). To take just one exampleof a minor number, considerthe North-East Caucasian language Avar. It has singular and plural and, for a limited numberofnouns, a paucal.The paucalis usedwhenthe numberof referents is restricted('a lew'), the ordinary plural for larger numbers ('many'). (Sulejmanov(1985),lrom whom the data are taken, calls them'restricted plural' and 'unrestrictedplural'.) Examplesof nouns with three number forms include the following: (z) Paucals and plurals in Avar SINGULAR PAUCAL PLURAL nus-6bi nris-al nus bor6q-al b6rq-al bor6q kritn-al kut6n-al kut6n bfl-dul b6l-al bel i,ul-al iil-al i;c,l
GLOSS daughter-in-law snake Plough sPade brush
Sometimes,as in the last example, the differencebetweenpaucal and plural is marked only by the position of the stress.Sulejmanov says only that this three-wayoppositionis availablefor a restrictedgroup ofnouns; he lists 89 which havethe contrast.Of these,one is a kinterm (nus' daughter-in-law') and a further eight denote non-human animates; the rest are inanimates. It appearsthat the nouns which have a distinct paucal form are not in general high on the Smith-Stark Hierarchy. Avar demonstratesthat of the inventory
GREVILLE
G. CORBET T
AND
M ARIAN N E
MITH U N
of nouns which are within the number system, some delimited Sroup may have a further number opposition. In languageslike Avar the distribution of theseforms is not subject to the Smith-Stark Hierarchy and they constitute a minor number (for further examplesseePlank (1989: 296-3te), Kibrik (I992: I5), Corbett (1995)).Note that theseminor numbersinvolvenominals which have an additional number; they are the converse of the familiar problem of defectivenouns, where a small number of nouns lacks one of the major numbers which they might be expectedto have (thus English scrssors lacks the singular). Though minor numbersrequire a relaxationof the typology of number, they do not vary without constraint.We proposethreecriteria:'? First, they involve a proportion of the nouns of a given languagewhich is relatively small by comparisonto those involved in the major number(s), where being a major number involvessplitting the noun inventory, taking sometop segmentof the Smith-StarkHierarchy. Second,the possibleminor numbersare drawn from the inventory of major numbers.That is, a number which can operateas a minor number will be found in anotherlanguageoperatingas a major number. Third, a languagewith a minor number must have a number systemwhich would match an otherwiseattestedsystemof major numbers(e.g.singular - dual plural as opposedto the unattested* singular- trial - plural) both with the minor number includedand without it. The effectofthese claimscan be illustratedby our Avar example.The Avar situation, with major singular and plural and minor paucal, flts in the tlpology of nurnbersystemsbecause: r. The number of nouns with paucal forms is relatively small by comparisonto thosewith only singularand plural forms. z. The paucal, here functioning as a minor number, is also found as a major number,for instancein Yimas. 3. Major singular,major paucaland major plural is an occurringsystem (asin Bayso),and major singularand major plural is alsoan occurring system(as in English). The relevanceof theseclaims concerning minor numbers for associativesis that the latter are also potential counter-examplesto the extension of the Smith-Stark Hierarchy to constrain the distribution of nominals for all [2] Theseare an improvementon the accountin Corbett (t995).
4
ASSOCI ATI VE FO RM S: EVI DENCE FR O M Y U P 'I K
number values.We needto indicate why the same' escapemechanism' is not and hencewhy the latter apparentlyposemore availablefor associatives, seriousproblemsfor a typology of number systems. 3. AssocIATIvEpLURALs:AppARBNTCouNttl-rx,q.upres Considerfirst Hungarian(a Finno Ugric languagespokenin Hungary,data lrom Edith Moravcsik, personalcommunication): (3) J6nos 'John' (4) Jdnos-ok 'Jobn-pl' (more than one John) $)
Jinos-6k John-Assoc.PL 'John and his group','John'n them' 'John and associates',
In Hungarian, the particular members of the group may vary according to circumstances(family, friends and so on), but they will be human and nearequals. The associativeplural ending can be added mostly to proper names and to nouns for kinterms, titles and occupations; examples formed with ordinary common nouns are strange.aThe associativeplural forms function [] Thus we find: (i)
apa / father 'father'
(ii)
elnitk I president 'president'
api-m-6k father- r . sc . Poss-Assoc . PL 'my father and his group' fr-dk az elnitk-'k / az elti5k DF president Mr-asloc . PL DEFpresident-Aslsoc . PL 'Mr Prcsidentand group' 'the presidentatrd his group'
ap6,-Ek I father-Assoc. pL 'father and his group'
The last form, including lr 'Mr' could be us€d as a form of addtess to the President (asking, for instarce, when the president and assoclateswill do something) but not as a folm of addressto the group. As metrtioned above, nouns for professions caa take the associativeform : (iii)
tanit6 / teaoher 'teacher'
a t^nft6-ek DEFteacher-Assoc.PL 'the teacher alld his group'
However, it is not available for other common nouns: (iv)
ember / man 'man'
'az ember-6k DEFman-ASSOC.PL 'the man and his group'
As in the text, the Hungarian data wete kindly supplied by Edith Moravcsik (pe$onal communication).
GREVILLE G . CO RBETT AND M ARI A N N E M I T H U N
as plurals for subject-verb agreement,and in object position they control definite object agreementon the verb. Since nouns for kinterms and some others denoting humans have two forms (associativeplural and ordinary plural), we might expect-to-findthe samewith the personal pronouns. What we actually find is the following: (6) Personalpronounsin Hungarian SINGULAR PLURAL
r6n
mi
2let l
3 6
6k
Each pronoun has only one plural form. The first and secondperson p.orrorro. have suppletive plurals, while the third person pronoun, whose ^position on the Smith-StarkHierarchyis somewhatuncertain(Smith-stark i974: 664),usesthe regularplural marker, asin: nd 'wo man' n6 k'women' 'n -ii t tt tigttt, we have a straightforwardviolation of the Smith-Stark Hierarchy: the associativeplural is availablefor a middle segment(nouns denoting kin and some other humans), rather than a top segment.including the persinal.pronouns.However,Moravcsik (1994)proposesthat in fact the fusf and second person plural pronouns universally illustrate associative plural (group plural) meaning: 'We and you-PLU are semantically group plurals in that we normally means"I and some others" (rather than "more ihan one speaker")and the normal meaningofToa is also "you and some others" ratherthan "more than onelistener".'(The ideathat thesepronouns can have associativemeaninggoesback at least as far as Jespersen(1924: r9z).) Moravcsik further links this interpretation specifically to the SmithStaik Hierarchy, and continues: 'It thereforemakes sensethat a special plural, such as the group plural, that applies to lessthan all nominals should apply'to personal l.ottootts and human nouns, and to proper names and kintirms in particular.' Thus if we treat the associativeplural as a separate number,it is possibleto claim that the top segmentofthe hierarchyis indeed involved. In somelanguagesjust the first and secondpersonpronounsare involved, while in others,like Hungarian, kinterms and even other nouns denotinghumansare involved. Thereare two problems.The first is the lessserious'If the fust and second p"rsoo prono,rrs are associativeplurals,why is this not -clearin their form? instanceswhich iFor a iiscussion of the forms of plural pronouns, including answermight (1964) A simple Myrkin ) haveregularplural morphology,see is some formal there long as so be thathe form need not be transparent, problem serious more The 1 and we. contrast, such as that between English plural forms' I4l Eva Csatd points out Oersonal communication) that, itr the ob-lique '-' ptooou^ of all three peisons include a numbq marker; this is similar to the ordinary Dlural. not to the associativemark€r'
ASSOCIATIVE FO RM S: EVI DENCE FRO M Y U P 'I K
concernsthe ordinary plural. If the first and secondpersonplural pronouns are associativeplurals, then the ordinary plurals becomeexceptional, since they would typically involve some central segment of the hierarchy. We would have to say that the claims of the Smith-Stark Hierarchy relate to associativeplurals. The considerabledata on the distribution of ordiirary plurals,which havebeenpartially understoodby referenceto the hierarchy, would now all be problematic. It might be argued that we have been over-simplifying by viewing the pronouns as being either ordinary plurals or associativeplurals. The first personplural may be an ordinary plural, if rarely, and the secondperson (t924: t9z) statesthat plural is often usedas an ordinary plural. (Jespersen the secondperson plural can be of either type). It could therefore be that the associativeplural is a part of the plural, and the personal suggested pronounscoverthe wholeplural range.But this merelyblurs the issue.Ifthe associativeplural is a part of the plural, why shouldwe not treat the dual as part ofthe plural (since'two'is includedin'more than one')?We cannotdo this becausethe dual (in the languageswhich have a dual) shows a regular form-meaning correspondence,which demonstratesthat it forms a separate number. The Smith-Stark Hierarchy is precisely a typology of formal contrasts.It is the lack ofa formal contrastin the pronouns,ascomparedto formal contrasts like Hungarian Jdnosok (plural) verws Jdnosik (associative),which makesthe pronouns a clear problem with respectto the Smith-StarkHierarchy. If we wish to continue claiming that the associativeplural is a separate number, a last escapemechanismwould be to claim that it is a minor number (like the paucalin Avar). This is unsatisfactory,as our data from Hungarian plural meaningis availablefor the show.First, it appearsthat the associative first and secondpersonplural pronounsaswell as for some'higher' nouns: the problem is the lack of a form-meaning correspondence.And second, recallthat minor numbersalwayscorrespondto a major number; that is to say, any number postulated as a minor number is found in another language as a major number. But no instanceof an associativeplural functioning as a major number parallel to an ordinary plural has been found. Thus, as before, either the associativeplural or the ordinary plural would fail to flt the typology. (Sincethe associativeis not found as a major number, it cannot possibly meet the more stringent criterion for minor numbers, namely that the language with a claimed minor number must have a number system matching an attested systemof major numbers both with the minor number includedand without it.) We try two ways forward. First we look for straightforward morphological evidenceon the relation betweenassociativeand plural. In Central Pomo we find suggestivediachronic evidencethat they are, or rather, were separate. Secondwe look at a more complexnumber system,and herewe ar€ able to confirm the conclusion suggestedby the Central Pomo data.
GRB VI LLE G . CO RBETT AND M A R I A N N E M I T H U N 4 . CENTRAL PO M O
Someinsight into the status ofthe associativedistinction cornesfrom Central Pomo, a languageof the Pomoan family indigenous to Northern California. Central Pomo itself is spoken roo miles north of San Francisco,with one dialecton the coastat Point Arena/Manchester(thenameofone 'rancheria', the California term for a smallreservation)and two dialects5o milesinland, on the RussianRiver near Ukiah. Only a handful of speakersremain.6 Number is specifiedin Central Pomo in multiple areasof the grammar and lexicon.The number distinctionsfound on nouns,pronouns,adjectivesand verbs operate independently of each other and differ in important ways. A full accountofthe systemcan be found in Mithun (I988b).We focushereon nominal number and its relation to associativemarking' Only a small setof nounshaveplural forms, most denotinghuman beings ('man', 'woman', 'child', 'old woman', most kinsmen).Severalof the plural forms are simply irregular, but others can be seento be formed with a suffix -(I)4/. It takesthe shape-Ja/after vowels: (7) m:i'!a m|'!2-!ay woman woman-Pl'women' 'woman' The initial J doesnot appearafter consonants:
(8) 6r.8
66't-ay
person-PL person 'persons' 'person' The plural forms are generallyusedonly when their referentsare significantly individuated.Pronouns,by contrast, systematicallyshow distinct singular and plural forms wheneverthey refer to humans. In addition to the relatively restricted plural marking for nouns, Central Pomo also has an associativemarker -1oya.Its usecan be seenin (9):
(9) Norman Ball:toya (name):6596
l6w-ae:Ya
?e
talk . PL-TMPRF. PL : PE
coP
mu'1. that
'Norman Ball and them were talking about that.'6 The verbal morphology in (9) reflects the plurality of the subject. The verb root l6w-'ta]k' is inherentlyplural, denotingconversationin a group, and the suffix -ai is the plural form of the imperfective aspect. The associativephrase can also fill other grammaticalroles. In (ro) it functionsas the grammaticalpatient of the verb 'discuss'. [5] We are grateful to MIs FrancesJack (Hopland Rancheda), Mrs SalomeAlcantra and Mrs FlorencePaoli (Yokaya Rarcheria) and Mrs Wioifled L€al and Mrs Eileen Oropeza(Point Arena Rancheria) for sharing their krlowledge of the language. [6] Hyphens (, are used here to indicate affx boundaries, and equals signs (=) for clitic boundaries. Ceotral Pomo co4tains a dch inventory of€vidential markers, ahong them the clitic : ld that marks information known from direct pe$onal experience(PE)
ASSOCIATI VE I O RM S: EVI DENCB FRO M Y U P , I K
dan6-'d: a? (Io) Eileen:!oya-l:wa ma (name): AsSoc-pAT:Q 2 . sc talk. SG-IMPRF. sG= Ili.{I\,lED 'You're discussingEileenand them?' The associativeencliticappearswith proper names,asin (9) and (to) above, and with kintermsas in (rr):
(r r) Ma-th6-Io'ya
hinlil
6anri 5?riduw-ay
?i-n
3. poss-mother-AssocIndian word not.know-Pl- be'as 'Becausetheir parentsdon't speakthe Indian language.' The associative also appears with indefinite pronouns, which are used as interrogatives as well: (rz)
mida napn6-w? BL':toya:wa who:.Assoc: e there sit. pl--ppJ 'Who [all] is living therenow?'
The shape of the associative: toya is distinct from that of the noun pluralizer -1ay.If we look at the forms of the pronouns,however,we see evidence suggestingthat the associative: /ola includes an origrnal plural marker, (ry\
Personalpronounsin Cenftal Pomo POSSESSIVE PATIENT AGENT PL PL SG SG SG PL yri'?khe ya yal khe t ?a' 10' mdya '?koe m6yal mkhe 2 ma mdya mI o mu'l mri'Iuya mf'Iu rnri'! uyal mri'kk! rnri'luya'khe1 3 1i'khe! Iiya'k"e! 3R' li'Io liyal Ii' liya The plural segment-/a appearsclearly in the second person and the third person long-distance reflexive pronouns, somewhat less clearly in the fust and basic third person forms due to the historical erosion of certain unstressedinitial syllables. The same plural segment -/d appears in the :lo-ra. associative: The situation in Central Pomo has provided a valuablehint. It suggests that the present associative marker was probably formed from separate associativeand plural segments,which would in turn lead us to believe that number and associativity should be treated as separate categories' It is possible to analyse the Hungarian data in a similar way. The associative suffix -dk might be segmentableinto two components, associative-d- plus a plural element -k- also visible in the nominal plural suffix -ok and third personpronominalplural dk. SeeMoravcsik (forthcoming,section2.r.6) for a discussionof the advantagesand disadvantagesof this approach hl The forms marked '3R' are part of a set of specialempathetic/logophotic/long-distance reflexive Dronouns.
GREVI LLB G . CO RBETT AND M AR I A N N E M I T H U N 5. CENTRAL AL ASKAN
YUP. I K
To make further progress,a good strategy, as is often the casewith number, is to look at a more complex systeminvolving a third number' If we analyse a languagewith a dual in addition to singular and plural, then theoretically we might find two outcomes: r. Four numbers;singular,dual, ordinary plural, associativeplural' z. Three numbers: singular, dual, plural, plus an additional categoryassociativeness.which we would expect to be available for dual and plural' We now examine a languagewith three numbers and associativeforms' Our data are from Central Alaskan Yup'ik, which is spoken in southwesternAlaska in the Yukon-Kuskokwim Delta and Bristol Bay areas' According to Jacobson(1993: r) there were in 1993approximately2o,ooo Central Alaskan Yup'iks, of whom some ro,5oo spoke Yup'ik' Children learn Yup'ik as theii first language in about half of the Yup'ik villages'8 Central Alaskan Yup'ik is one of the Yup'ik group of languageswithin the Eskimo branch of the Eskimo-Aleut family. Since this languageis vital to our claims, a brief sketch will be helpful. 5.r Sketchof CentralAlaskan Yup'ik Verbs, which may be intransitive or transitive, contain pronominal suffixes referring to their core arguments.They distinguish singular, dual, and plural numbei in all persons. There is no gender, but there is a long-distance reflexive used, roughly speaking, for an argument in certain kinds of deoendentclausesthat is coreferent with the subject of a higher clause' The pronominal suffxes are precededby one of a set of mood markers (indicative, optative,interrogative,subordinative,and variousconnectives)'Becauseso many featuresare distinguishedby the pronominal suffixes,and-becausetheir shapesvary with the mood, the number of pronominal forms is quite large' Nouns also take inflectional suffixes,which mark an elaborate casesystem with a basic ergative/absolutive pattem. The singular, dual, and plural forms of the unmarked absolutivecaseof'kayak' can be seenin (I4)' (t5), and (r6):e f8l We are verv sraleful to George Charles and Liz Ali, both of Bethel, Alaska, for sharing ' ' their intuitiois on tbeir languige, and to Tony woodbury for helpful discussion It is interesting to note that there are very few nouns whic,hhave singular lorms only One [9] '-' 'wat"t'. (Abstact nouns with meadngs like 'happiness' are.generallynot e*a-ple is usea.i uost-"inouus tlat would correspotrdto English massnouns are typically interpreted (sg ) as cotmt nounsin Yup'ik wben dual or plural suffxesare added For instance-'4'4 or ja rs of oil .l4tlt(pl)'thteeormoresealpokesorlats 'oil', r./4rk(dual) lwo sealpokes of oil'. ilelatively few noun; ate defectivein tetms ofnumber in other rays Thele are some which have only the plual, such as niidrSTtsrlutet 'tudio' ' IO
ASSOCIATI VE FO RM S: EVI DENCE FRO M Y U P 'I K
(r4) qayaq 'one kayak' (r5) qayak 'two kayaks' (16) qayat 'three or more kayaks' Possessed nouns contain transitive suffixesthat match those on verbs: the possessoris coded like the ergative,the possessed like the absolutive(as illustrated in the Appendix). The form is then further inflected for case. 5.2 Associative forms in Yup'ik Inflectablewords in Yup'ik, as in other Eskimoanlanguages,consistof an initial root (traditionally termed a 'base'), optionally followed by any number of derivational suffixes('postbases'),then by inflectional suffixes ('endings'). There are hundredsof thesepostbases,many with meanings correspondingto those of roots in other languages('to buy an X'), many with meanings typical of derivation in other languages (nominalization), somewith meaningstypical of inflectionelsewhere('habitually'). Thereis a postbasein Central Alaskan Yup'ik that appearson proper namesonly, to give forms meaning'N and family' or 'N and companion(s)' :r0 (I7) cuna 'Chuna' (name; absolutivesihgular) (t 8.; cunankut
ayag-tu-t
Chuna.Assoc.PL go-rNDrc-3.PL 'Chuna and his family/friends left."' (i)
niicugni-ssuut-et dipe-s-ki go.out-cAusB-oprATnE.2.sc/3.pL listen-rNsrR.Nol,trNALrzER-pL 'Tum of the radiol'
Another example is a//iertet 'one-piece trousers with attached fur socks', also found only in the plual. Such nouns tFically refer to objects consisting of multiple parts. Other nounshavedual and plual but no singular.Thus ls,turat'dog hamess'is dual in form, appearswith dual pronominal suffxes on the verb, and can be used for one dog hamessor two. Three or more dog hamesseswould be urr&rar. There is no singular form *uskuraq.Atasuak'summertrousers'functions similarly.For comparativedata on the dual in Esktno languagesseeHammedch(1959). [ro] The associativeis not available for kinterms. This may be b€causethey would needto be followed by transitive pronominal sufrxes to indicate possessorand possessed(N alrd N's kinsman), which wo[ld lead to considerablemorphological complexity with the associative, especiallyifthe whole had th€n to be iDflectedfurther for case.Note that our data are from the dialect known as General CentEl Alaskan Yup'ik; for similar data on the Chevak dialectseeWoodbury (I98I). [r r] The absenceof an overt tensemarker indicates presetrt and immediate past. II
GRE VI LLE G . CO RBETT AND M A R I A N N E M I T H U N
If it werejust Chuna and one friend, the form would be: (I9)
ayag-tu-k cunankuk Chuna.essoc.ou go-INDIc-3.DU 'Chuna and his friend left.'
enclitic The companion can be included overtly, by using the noun plus the spelled= //2. Thus: (zo) cunankuk amaq=llu Chuna.essoc.ou woman=too 'Chuna and the woman left'
ayag-tu-k go-INDIc-3DU
What is specially relevant about these data is flrst that the associative do not meaning fiis equaly with dual and plural. It seems,.therefore,that we attendant the plural, with associative number' a'new need to'recognize problernsfor-a typology of number.However,instancesof the useof duals in this way are known in the literature While they make the'extra number' that solutionlessplausible,they do not eliminateit lt could be still claimed ordinary dual' associative dual, ordinary singular, numbers: Yun'ik has flve plural, associativePlural. important point about the Yup'ik data-is that the The second,urrd plural) -ore associative morpheme-nku-and the number morphemes(ft dual' -t are realizedseparately.This is made clearin (zr)'r' ht\ Associativeand rutmbermorphologyin Cmtral Alaskan Yup'ik ASSOCIATIVE ORDINARY qaYa-k cuna-nku-k DUAL qaya-r PLURAL cuna-nku-t Associativescan haveseparatemorphology,and so they shouldbe treatedas a diferent 'dimension' or category,not a part of the number category the instancesof specialforms for associativeplurals in other it 13 needs "."fo." languagescan reasonablybe treatedas portmanteaus To show this associative separable a lNl number dual separate a Jonibination of assomorpheme, which is exactly what Yup'ik has This number .and various in interact may which categories, different alu,i'u"n".t' t"pt".ent ways. number From the point of view of our original concern,the typology of problem' pose a longer no forms associative that systems,this result means
lr2lWediscussbelowthedevelopmentofthe_ttk'U-sumx;insynchronictemsitappearst ' ' it is no longer analyzable further. It is not rclated to possessivemarking' nxmber b€cause It might b€ argued that it is impos-sibleto separate.associativity from [rj] ' -' assocladvitvimptiesnon_singular'However.ai Martin Haspelrnathpoints out (per5onal ,h. it"lusilveexclusivedisritrctionsimilarlv implies noo-singularilhis ;;-;t;;;;'ri;;i argue' it .L, considered part of the category of number and' we would iirii-""iiassociativity is a compalable case. l2
ASSOCIATI VE FO RM S: EVI DENCE FRO M Y U P 'I K
since they are not part of the number category. Looking ahead, having establishedthese two separate categories,we should investigate their furteractions. For instance,a reasonablehypothesisis: 'if a languagehas a (major) dual and a plural, associativemeaningswill be equally available to both, whetheror not thereis specialmorphology for expressingassociative meaning'. 6. CoNcr,usron We have seenhow associativesare potentially problematic for a typology of number.A possiblesolutionwas suggested by data from CentralPomo,but it was Central Alaskan Yup'ik, which has segmentabledual, plural and associativemarkers, which allowed us to demonstrate the dissociation of number and associativity. Sincewe are dealing with two distinct categories, even though they are frequently realized together, there is no remaining problem in this regardfor the tlpology of number systems.
APPENDIX
The pronounsof Central Alaskan Yup'ik Yup'ik allows us to clarify our notions of the relation of number to associativity in a striking way. It therefore makes sensealso to examine its pronouns;in this language,if anl,where,we might expectto make progress on the relation betweenassociativeforms and pronouns.And indeed,at first sight, certain pronouns of Yup'ik seemto offer additional insight into how associativeswork. Closer examination,however,showsthat the apparent pattern may well be no more than that. As mentioned earlier, all verbs contain pronominal suffixes referring to their corearguments,onefor intransitives,two for transitives.The absolutive sufixes that appear with intransitive indicative verbs are these: (i)
Absolutive suffixes (intansitive indtuative verbs) in Central Alaskan Yup'ik SINGULAR DUAL PLURAL t -(ng)a -kuk -kut 2 -ren -tek -ci -k -t 3-q
The samenumber suffixesseenearlier on nouns, dual -k and plural -r, appear in the pronouns as well. In the third person pronouns, the numbers appear alone. The second person dual and plural forms are built on the second person singular marker -r. (The alveolar/palatal affricate c results dia-
r3
GREVI LLE G . CO RBETT AND M A R I A N N E M I T H U N
chronically from t followed by f.) The first person dual and plural forms contain a soecialsecmentin addition to the numbersuffixes,however:-/rz-' This segmentis cer'iainly reminiscent of the associativesuffix -zku-' While for simplicity we give only the intransitive forms here, in the transitives, the forms iontaining first person dual and first person plural absolutives also have the same segment-/rz-. In addition to the regular pronominal suffixes,Yup'ik containsa set of independentpronounsusedfor specialemphasisor contrast.The first person forms can be seenin (ii). Thosein the first row are usedfor core arguments (ergativesor absolutives).Obliques are formed on the baseillustrated in the iecond row with the locative case -zi. (Other oblique cases ablative, terminalis, vialis, aequalis- show the same stem but have different endings.) (r1') First personforms in Central Alaskan Yup'ik PLURAL SINGULAR DUAL wangkuta wangkuk ERG/ABS wii/wiinga wangkugni wangkutni wangni LOC We return to the first person forms below. The secondperson forms are as follows: (1lt) Secondpersonforms in Central Alaskan Yup'ik PLURAL SINGULAR DUAL elPetek elPeci ERG/ABS elpet elpetegni elPeceni elPeni LOC The basicthird personforms are: (iv\
Basic third personforms in Central Alaskan Yup'ik PLURAL SINGULAR DUAL ellait elkek ABS ellii elkenka ellaita ERG elliin elkegni ellaitni LOC ellini
There is an additional third person long-distance reflexive form, sometrmes abbreviated 3R. This form is usedfor argumentsthat are coreferent with the subjectof the clauseor of a higher clause. (v) Long distancerefexive forms in Central Alaskart Yup'ik PLURAL SINGULAR DUAL ellmeng ellmek ABS/ERG ellmi ellmegni ellmeggni elknini LOC In most of these forms, traces of the original dual -k and plural -t can be discemed.What is of specialinterest to us here is the shapeof the first personpronouns ((ii) repeatedfor convenienceas (vi)).
r4
ASSOCIATIV E FO RM S: EVI DBNCB FRO M Y U P 'I K
(vi) First personforms in Central Alaskan Yup'ik PLURAL SINGULAR DUAL wangkuta wangkuk ERG/ABS wii/wiinga wangkugni wangkutni wangni LOC Theseforms are derived from the proximate restricted deictic demonstrative adverbwa-.It appearswith the locativesuffix-nl, for example,in watti'right here, before me' (Woodbury rg81'. 232-2, Jacobson 1984: 4ro). In the singular, the demonstrative base wa- is followed by the first person pronominal suffix -(zg)4. The velar nasal (ng) in the first form follows a pattern of 'velar dropping' betweensingle vowels. Where there is velar dropping betweentwo original low vowels, the vowels are raised. Thus original wa-nga appearsas ltii, the first altemant of the first person singular pronoun. The secondalternantis the result of a secondaddition of the first person pronominal sttffix: wii-nga. (The velar nasal remains in this form becauseit is precededby two vowels.) The other first person pronouns (wangkuk, $)angkuta,wangni, wangkugni, wangkutni) show a common base wang', consisting of the deictic root lvafollowedby the first personpronoun -(ng)a.The-a- of thispronoun regularly appearsas schwaword-internally,and disappearsin certain phonological contextslike thoseabove.(Sincethe velar nasalis not flankedby two single vowels,it remains.)The endingsofthe dual and plural forms, -k and -ta ate as expected:the -/cis the regularmarker of dual number, and the -ta is an older form of the plural that still appearselsewhereas a regular marker of plural. What is intriguing is the internal segment-kr- that appearsbetween the first person baseand the number markers. This segrnentdoesnot appear in any of the other dual or plural independentpronouns,but it doesappear in the first person dual and plural pronominal suffixes' and strongly resemblesthe associativesufrx -nka-. We thus find an intriguing situation: first person dual and plural pronouns seemto sharea specialassociative segment,distinct from the number markers, and not found in any of the other pronouns. This is what we might have expected,since the natural interpretation of these fust person forms is indeed associative,as we we discussedearlier.Ifjust one pronoun had forms similar to associatives person. would expectit to be the first The two segmentsare not identical.The associativemarker that appears with nouns is -zkr.i-,while the marker that appearsin first personpronouns is -ka-. As Tony Woodbury (personalcommunication)haspointedout to us' the initial -n- of the associativeis a reflex of an original plural marker -/, recognizableby its idiosyncraticjunctural properties.It regllarly appearsas n beforek. As a result,we haveapparentlyfound a higbly significantresult: the associativeconstruction consists of an original plural element -n-, followed by an element particular to the associative,-ku-, then followed in turn by distinct number specification,either dual -k or plural -r. The special
r5
CR EVI LLB G . CO RBETT AND M A R I A N N E M I T H U N
element of the associative,-ft2-, appears as well in both the bound and the independent pronominal paradigrns, but only in the non-singular first person,just the location where we would expect associativemeaning. The apparent coincidenceof forms may, uDfortunately, be lesssignificant than at first appears.The specialsegment-fru- appearsin one other location of Yup'ik morphologyas well: it systematicallyforms non-singularbasesof demonstrative pronouns, a large and elaborate set of forms in the language. Its usecan be seenin the sampledual and plural forms of the demonstrative tamqna'lhat one' (extended,i.e. moving, long, or of large extent; from Jacobson(1984:658,660)). (vri) Sample dual and plural forms of tamata ' thdt one' PLURAL SINGULAR DUAL tamakut tamakuk ABS tamana tamakut tamakuk tamatum ERG tamakuni tamakugni LOC tamatumi Th€ segment -ku- may thus have originated as a simple marker of nonsingular number and been exploited in that capacity independently in the formation of the demonstrativepronouns, the first person pronouns, and the associative.Several alternative scenariosremain possible,however. The segnent may have had an original associative meaning that was applied independently to the demonstratives, the pronouns' and associative nominals.Altematively, it may havehad an original non-singularmeaning that was exploited to form demonstratives, but may have taken on an associativemeaning when applied to either the associativeconstruction or the frst person pronouns, which was then exploited for the formation of the other. Perhapsa closer reconstruction of the histories of the three elements will some day tell us more. R E F ER EN CE S Allan, Keith. (r98o). Nouns and countability.La guage56. 54\-567. (rqgs). Minor number and the plurality split. In Borg, Albert & Plark, Corbett. Corbett, rrbett, Greville Gre\,i[; G. ((1995). Greville Maltese nou noo phrase meets tlpology. (European Sci€nce Foundation Frans (eds.) The Matzse Programme in Laflguage Typology: Theme 7, Noun Phrase Sttucturc: Working Paper 25). ,Rltista di Linguistica. Working Paper VII/25.] Revised version to appear in rlltista IBUROTYP Workig Dude, Mark. (1986). The grammaticization of number as a verbal cat€gory. In Nikiforidou' vassiliki; va;Cby, Mary; Niepokuj, Mary & Feder, Deborah (eds.) Prcceedingsof the Twetfth A4 ual Meetirrg of the Berkeley Linguistics Society: Februar)' 15-17, 1986.Betkeley, Twefuh B.L.S.,Unive$ity of Califomia. 355 37o. Catfornia catfornia:: B.L.s., Gil, David. (1993). Nominal and verbal quantificatiotr. Sprachtypologie tmd Unil)e^alienforschung46.215-317. Himmerich, Louis L. (1959).Wena der Dualis lebendigist. Dre SPrdche5. 16-26. Jacobson, steven A. (rg84. Yup'ik Eskimo dictionary' Fairbanks: Alaska Native Language Center, University of Alaska. Jacobson,StevenA. (r 9 g3).A pnctical gammar of the Cental Alaskan Yup'ik Eskimo laaguage' t"Is., Nitive I-anguigJCenier and Program, Unive$ity of Alaska' Fairbanks, AlaskaJespersen,Otto. (r9z4l. The philosophy of grummar' London: Allen & Unwin' Kibrik, Alexardr E. (1994. befective parudigms: number in Daghestan,a,. (Eulopean Science
r6
ASSOCIATIVE FO RM S: EVI DENCB FRO M Y U P 'I K Fouldation Programme in LadguageTypology: Theme 7, Noun PhraseStructure: Working Paperr6.)IEUROTYPWorkingPaperVll/t6.] Mithun, Marianne. (I988a). Lexical categoriesand the evolution of number rnarking. ln Hammond, Michael & Noonan, Michael (eds) TheorcIicalmorphology: approachesin moden /irgrirtr?r. San Diego: AcademicPrcss.2Il-234. Mith;, Marianne. (I988b). Lexical categodesand numbel ir Central Pomo In Shipley, William(ed.) In honor of Mat! Eaas: From the Eaas Festival Conferenceon Native American //n8rrrri.s.Berlin: Mouton de Cruyter. 517-537. plural or cohort plural. Email document, Mor;vcsik, Edith. (1994).Group plural - associative LINGUIST List: Vol-5-68r. rr June 1994.ISSN: 1068-4875 Moravcsik, Edith. (forthaoming). Inflectional morphology in the Hungarian noun phrase- a q?ological assessment.In Planl, Frans (ed.) Now phrasestructurc in the lang)agesoJEwope' Berlin: Mouton de Gru]4er. Myrkin, V. Ja. (1964). Tipologija lidnogo mestoimenijai voprosy rekonstrukcii ego v indoevropejskom aspektq. Vopros! iazrkoznania no 5. 7816. - Plank, Frais. (1989). bn Humboldt on the dual ln Corigan, Roberta, Eckman, Fred & Noonan, Michael (eds.) Linguistic categorization- (Amsterdam Studies irl the Theory and History of LinguisticScience6I.) Amsterdam:Benjamins.293-333. Smith-S;rk, T. Ceddc.(1974).Th€ plurality split ln La Galy, MichaelW.' Fox, Robert A' & Bruck, Anthony (eds.) Papersfrom the Tenth RegronalMeetihg' ChicagoLinguistic Societ!' Aprit rg-2l, Ig74. Chicago. Chicago Li4uistic Societv. 657 47| disla mnozestvennogo i neograniaennogo Sulejmanov,Ja. d. 1r985).b formax ograniaennogo imen suiiestvitel'nyx-v avarskom jazyke. ln Mikailov, K-S- (ed) KateSotija iisla '\' dagesta skix ja4,kax: sbornik statej. Ma\aakala: Dagestanskij 61ial AN SSS\ I t4-t 19 Woodbury, Anthony. (r98I). Studf of the Chewk dialect of Centlal Yup'ik Eskimo PhD' dissertation,Universityol Califomia, Berkeley. Authorc' addresses: (Corbett) D epatment oJ Littguistic anil lktemational St dies, Unirersity oJ Surre!, GriLlford, Suftey GU2 5XH, U,K. E-mail: g-cofi ett @surreY.ac.uk
r7
(Mithun) Departmenl of Linguistics, Universiry of Califo mia, Santa Barbara, CA 93106, U.S.A. E-mail: mith
[email protected]