CURRENT ISSUESIN LINGUISTIC THEORY r 106

Conference on Historical Linguistics '

Reconstructing the Unidentified

Marianne Mithun University of Califurnia, SantaBqrbara

Homophonous grammatical markers arc not unconunon afiDng languages,but in the absenceof an obvious semanticrelationship, there can be little justification for relating them diachronically. Sometimesthe coherenceof a setof grammatical functio$ is obscured, however, by the loss of a marker that would have provided the semantic link. A generalawareness of recurring clustersof functions expressedby homophonous forms crossJinguistically can provide a useful tool in such situations: it can dilect our attention to the kinds of missing links that we should reach for in order to uncover relationships drat might otherwise escapenotice. Cross-linguistic knowledge can have special value in dre reconsrucdon of languages nith small inventories of sounds,where the probability of chance phontic resemblanceamong morphemesis pafiicularly high. Consider Mohawk, a Northem Iroquoian language spoken primarily in Quebec, Ontario, and New York State. Mohawk contains only frfteen distinctive sounds,but it is highly polysynthetic, v/ith extensiveprcfixation, suffixation, and incorporation. Among the lnefixes in verbs are obligatory pronounsrcferring to all core arguments.Comparethe in8ansitive verbs in (1 ). 0) Mohawkl (Iroquoian): Intransitives

AGENTS katerd:roks l.AGT-watch-HABITUAL 'I watch' saterd:roks 2.AGT-WaICh-HABITUAL 'you watch' mterd:roks M.AGT-waIch-HABITUAL 'he watches'

PATIENTS woke?nik\hrat6:ta?s I.PAT-mind-lay-BEN 'I understand' sa?nih4hray {:m?s 2.PAT-mindJay-BEN 'you unde$tand' ro?nih4hray{:a?s M.PAT-mindJay-BEN 'he understands'

330

MARIANNE MTIIIUN

lqterd:roks F.AGT-waIch-HABITUAL 'she watches'

loko?nikvhroy{:tafu F.PAT-mindJay-BEN 'she understands'

Combinationsof agents and patients are expressedby fused transitive prefixes. Traces of the origiDal agent or patient components can still be seen in some of the foms, but all of the ransitive pronominal prefixes are now acquired and processed as singleunits. (2) Mohawk: Transitive Combinations teNyatVhekt{:nis DUALIC-I,AGT/M.PAT-frighten-HAB 'I frighten him' tehskwatb?nekt6:nis DUALIC-M.AGT/l.pAT-fri .he frishrens m'

ehten_HAB

tekheyatv?nektl:niss teyqkv'at|?nekt4:nis DUALIC-I.AGT/F.PAT-fri ghten-HAB DUALIC-F.AGTn.pAT-fri shren_HAB .she frighrens 'I frighten her' me'
The pronominalprefix systemin Mohawk is elaborate, distinguishingnot only agent and patient case,but also fust, second,and third person;inclusive and ex_ clusive io first person;masculine,feminine,and neutergenderin third person;and singular,dual, and plural number.The total numberof agentprefixes,pauenrpre_ fixes, and tansitive combinations is thus quite large: over sixty. Each prefix, funher_ more, appearsin a variety of forms conditioned by context, 1 Two Mohawk morphemes: w- and waAmong the Mohawk pronominal prefixes are two partialy similar elements. In most contexts, the neuter agent prefix, used to refer b animals, things, and some women, appearsas ,ta-. (3) Mohawk neuteragent:*c- 'it, she'

ka?nyd:k4?s N.SG,AGT-escape-HAB 'it or sheis escaping'

tuhnekthrhn? N,SG.AGT-drink-HABITUAL ,it or she drinks'

Beforevowels-a-, -e-, and-4-, howevel the form w- appears. (4) Mohawkneuteragentlt'. 'it, she'
waterd:roks tewardltas

N.SG,AGT-watch-HAB 'it or shewatches'

DUALIC-N.SG,AcT-run-HAB .it or she is running,

T{ECoNSTRUC.nNG TTB UNDEN.nFED

331

whsaks N.SG.AGT-search-HAB 'it or sheis searching'

w$nVlfitrnyw N,SG.AGT+hink-HAB 'it or sheis thinking'

Numberis markedin the pronominalprefrx systemwith a variety of forms. persons, With fust andsecond themostcommon marter of pluralityis -wc(5) Mohawkplural of first andsecond person: -yatewahnekthrha? yakwahneWhrha? sewahnekthrha? kwd:nt4ks takw,i:Nlks shr1kwd: nr4ks etshisevd:nuks etshisewd:npks kwd:ngks shakvd:ngks tdkwd:nV*s ntgks etshisewd: 'we all (inclusive) drink' 'we all (exclusive) drink' 'you all drink' 'I am calling you all' 'you are calling us all' 'he is calling us all' 'he is calling you all' 'we all (inclusive) are calling him' 'we all (exclusive) arecalling you' 'we all (exclusive) are ca ing him' 'you all are calling me,/us' 'you all are calling him'

The partial resemblancebetween the rr- allomorph of the neuter singular agent, and the wc- plural of fust and second person, might in itself arouse little interest, espe.ially in light of the phonologica! inventory of Mohawk. Mohawk contains only nine surfaceconsonants.The chanceof dre sameconsonantappearingin two areasof the vast pronominal prefix systemis extremely high, in fact to be expected. Both of these forms, w- and wd-, appearin all of the other Northern Iroquoian languages(Oneida, Onondaga, Susquehannock,Cayuga, Seneca,Huron, Wyandot, Tuscarora), so they have been in place for a considerable length of time with their modern functions. Comparativevidence does reveal that the two are more similal in form than might be expected fmm the Mohawk forrns alone. In Mohawk, the nonmasculine plural agent prefix, used for groups of women or animals, is most often &Uti. In Senela and Huron, however, the fonn is wati. (6) Seneca(lroquoian): Myrtle Pete$on p.c. watitha?tha?'they (non-masculine) arepounding' Humn (Iroquoian): Potier in Fraser 1920:8 awat ihiato n ' [th e y (n o n -ma s c u l i n e )w r ote]'

Q)

The basic form of the plural marker for third person beforc consonantsin Senecaand Huron, as well as in Mohawk, is +&, The original form of the neuter singular must

MARIANM MITHTN thus have been the remainder of the prefix: ,a-. By regular rule, the vowel a disappears in Nonhem Ircquoian languagesbefore c, e, and 4. These arc the very contexts in which the w- form of the pronoun appears. By the same rule, the vowel of rhe plural marker -wa- disap;rars from the samecontexts.

(8) Mohawkplural wtewafthtats tuwehsaks tew4nphtlny6s

'we all (inclusive) are running' 'we all (inclusive) are searching' 'we all (inclusive)arc thinking'

The neuter singular agent prcfix w-, and the plural marker of first and secondperson w4-, thss had the same basic shapehistorically: /a-. Yet their similarity in shape,a singleconsonant plus vowel, in a languagewith only 9 consonants and 6 vowels,is still not sufficient justification in iself to relate them diachronically.

The unidentified

The behavior of a third grammatical caregory in a number of umelated languages suggests that the two Mohawk forms actually descendfrom a common ancestor. Many languagescontain affixes indicating dre involvement of an otherwise unidentified, unspecified, or indefinite participant- Note tbe contribution of the Central Pomo prcclitic b4 to the meaningsof the verbs below. (9) CenEal Pomo (Pomoan): FrancesJack p.c. ba?61 'to call' bdbaa?,i'l 'to call somebody' klts'iiw bdklts'iiw 10 be afraid' 'to be aftaid of people'

In some cases,the diachronic sourcesof indefinite aflixes are still retrievable. The most common sources are free pronounsmeaning'someone'or 'something',or nouns meaniog 'people' or 'things'. The origin of the Central Pomo proclitic Dais a free pronoun bda 'someone' that is still presentin the language. Indefinite pronouns are likely candidatesfor affixation, since they typically representparticipants whose identity is not especially important to the discourse at hand. Morphologization provides a meansof backgroundingtheir mention. In Mojave, a Yuman languageof Califomi4 the moryhologization of an indefinite obje.t marker ?f 'something' can be senin progress.It still appea$ both as a separaleword and as a prefix (i)d .

RECONSTRUCTING THEUNIDENTIFIED (10) Mojave (Yuman): Munro 1974:260

333

fri n<rut:-m. something 2-eat-INDICATM 'You ate,'

M-ii-ona:+t. g-eat-INDICATIVE 2-somethin 'You ate.'

In Koasati, a Muskogean language of I-ouisiana, the noun for person is d;ti. An obviously related prefrtxa:ti-lo:t-latJa- calj.indicate the panicipation of otherwise unidentified persons. (1 1) Koasati(Muskogean): Kimball 1985:137 6hyak a:tiohittiM:lok all someone-DIS'fR-RECIP-hear-SS 'Everyone heard about it ftom each other' hind:p, atcitna*itlok lo someone-2.STATM.StBJECT-be.afraid-SAME.SUBJECT.FOC 'lo, you are afraid of people.' In some languages, indefinite affixes serve only one function: indicating the i.nvolvementof an unidentifid Frricipa ln a large number of languages,however, the same forms serve additional ftnctions as well. The clusters of functions they serve,furthennore, recur with suprising fiequency aqoss unrclatd languages.

2.I D etr attsitivimtia n In many languages,indefinite pronouns function as detransitiven. As is well known, indefinite subject markers produce an effect similax to that of passivization, detopicalization of agentsor topicalization of patients of tansitives. (Note the useof French on in On m'a dit ... 'l was told'.) With transitive verbs, indefinite direct object markers generally preclude the appearance of additional direct objects, so their effect is akin to detransitivization. The basic Mojave vetb ama:-m 'eat', for example, can appearwith a separatedirect object, but the verb ?ii-atna:-m'eat (something)' does not. Pipil, a Uto-Aztecan language of El Salvador, contains an unspecified object prefix. Campbell repons that the prefix "is employed to show that the (non-human) objcct of a transitive verb is Rot specified. Verbs with this prefix, td-, are translated vrith an object 'something' or 'to be doing' whatever the action of the verb is, without speci$ing what the object offte acdonis; ... rhe 'unspecified objectl ,a- is natural in contexts where the action of the verb is in focus and the object is of little relevance"(1985:?7).

334

MARIANNE MITI{UN (12) Pipil (Uto-Azte{an): Campbell 1985:77 ni-it-u:ya I-UNSPECIFIED,OBJ-shc ll 'I am shelling(com)'

ni-k-u:ya I-it-shell 'I shellit'

Often combinations of indefinite markers and tansitive verb roots take on lives of their own as derived intransitive verbs. In Pipil, some verbs with the ro- prefix no longer have counterparts without it. Somecombinationshave developedidiomatic meaningsof their own. (13) Pipil (Uto-Aztecan): Campbell1985:78 latnati 'to imagine, guess' nati 'to know' lane:si nJi lakwi kwi 'to daytn' 'to appear,be seen' 'to borrow' 'lo hke'

2.2 Nomimlization Indefinitepronounsappearin many languages in nominalizations, often functioning as nominalizersthemselves. Nominals,as namesof entities,usually do not refer to specific events: they are generic labels. Food might be characterized as 'somthing pople eat', for example, not a specific object that a certain person ate at a particular time. A gardenermay be 'one who gardens',aII invalid'one who is sick', grain 'something ground', Note the indefinite prefixes in the Koasati nominalizations below. (14) Koasati(Muskogean): Kimball 1985:137 Mllon 'to be dangercus' atholl6 'witch' incdJin ,ta:sincd:ka 2.3 Nunber With surprisingfrequency,the sameforms that mark unidentified pa icipanrs in languages also indicateplurality. Unidentified,nonspecific, or indefinite arguments typically lack sharpindividuality. Pronouns leferring to unidentifiedparricipanrs are 'to write, mark' 'that (which) is wdtten' > 'book, paper,letter'

I{ICONSTRUCTING TI{E UNIDENIIFED

335

usually translated altemately as'one'; they, people', etc., or as 'something' or 'rhings'. Such pronouns are often used to refer genedcally or distributively to various people or objecs at various times or places,rather than to a single specific individual. Sometimesthey refer to a group collectively, a group whose members lack prtinent individuality. In many languages,this semantic overlap is refle.ted in a formal overlap the shapesof indefinite pronouns and number markers coincid. Atakapa, for example, a language formerly spoken in Texas, contains a prcfix c4&- 'someone', a contmction of the noun ic&al 'a person' (Swanton 1929). Tbe same prelix marks plurality of thid Personobjects(15) Aukapa (Gulfl: Swanton1929:126 wi cakitso I someone woke I woke somebody up.' cakwinelem cakhuo cakhLael cakinA 2.4 Specific reference In a number of languages, the same forms that designate unspecified or indefinite participants are also used where the identity of the referents is clear. Compare, for example,the useof Frcnchon in On arriye d A& 'we're coming into Aix'. The nonspecific character of the indefinite affrxes is often exploited for purposes of defocusing direct reference.Sometimesthis defocusing is be used when the identification of referents is not imponant, as in the passivization and detransitivization discussed in section 2.1. above. Sometimes it functions to maintain social distance. Chinook. for example, a language of the Pacifrc Northwest, contains an indefinite ergative pronominal prefix 4-. (16) Chinook (Chinookan): Boas 1911a:584, 589 a4ta'witx 'somebody gave them to them' aqLgEm6'*uit 'somebody stands near it' The same prefix is used for first person exclusive ergatives when combined with s{ond person absolutives. (17) Chinook (Chinookan): Boas 1911a:584 qamyun 'we/thee' qamt- 'we/you two' ydrzt qamc- 'rrelyotuall' lamc 'you fould them' 'I sethem' 'we saw thern' I ask them'

'Vthee' 'yyour two selves' 'Vyou'

336

MARIANNE MITHTIN

Similarly in Caddo, indefinite agent and padent pronouns are used for inclusive fint person, and for relatives by marriage (Chafe p.c.). The diachronic reinterpretation of third person nonspecific markers as general third penon pronouns can be reconstructed in a number of language families from of the Athabaskanfamily, for example,procomparativeevidence. Among Languages nominal plfixes can b reconstructed within the verb referring to first and second person subjects and objects, in singular and plural. No general rcferential third person prefixes can be reconstruded, and in most of the languages,basic third persons are usually unmarked. Cfransitive combinations of two third personsreceive sPecial marking.) Several indefinite pronominal prefixes can be reconstructed for ProtoAthabaskan, however(Hoijer 1971).

2 .4.I TtE Proto-Athabaskan unspecirtedsubject One prefix, sometinrs called a deictic or indefrnitesubject, can be reconstructedwith the meaning 'someone'. It marks the participadon of an unidentified prson as it is variously uanslatedas subject.Like indefinite pronounsin most languages, 'someone','one', 'a person', 'people', etc, (18) Athabaskanunspecifiedsubjectprefix Biesh dao bee niinae dd 'one shouldn't Play Navajo: (Young & Morgan 1980) with knives' Tututni: (Colla l9?6) Chipewyan (Li 1946) Sarcee: (Cook 1984) Tanaina: Clenenbaum1978) Slave: (Rice 1989) sya{'eni?a ndt'sed.e 'someon gave it to me' 'people are staying' 'somebody will kick him' 'someone is walking' 'someone woke me up'

ts'd?is ch'uyu ts'esets'ey(shr1

Like indefinite pronouns in other languages,constructions with the indefinite human marker ar sometimesused in place of passives. (19) Navajo(Athabaskan): Young & Morgan 1980:187 jint 'pople say,it is said'

REcoNsrRUcrn rc rlIE UMDENflFED

Similarly, the Slave prefix is also usedlir forming nominals. (20) Slave (Athabaskan):Rice 1989:625 ndts'ezd gha deyt 'hunting is hard' In some of the Athabaskan languages,rhis indefinite subject prefix is also used for persons whose identity is inferrable. In the Apachean languagesof the SouthwesL it is used ds a polite form for addressingindirecdy those one should respect, in particulir, siblings, or inJaws of the opposite sex (Hoijer 1946:76, Young and Morgan 1980:187).The rcference thus becomesspecific through usage. In Tanaina and Slave, the prefix is usednot oDly for unidentified personsbut for first personplual as well. (21) Tanaina(Athabaskan): Tenenbaum 1978;62 'we (or someone) saw him' ch'ghiTan (22) Slave (Athabaskan):Rice 1989:624 shAts'uyee 'let's (pl) eat' In the Bearlake dialect of Slave, dle semanticextensionof this prefix hasresulted in a finer numberdistinction.The Fefix rs'e-indicates first personplural, while the odginal frrst person plural prefix td- now marks f[st person dual only. In some of the languages,this indefinite prefix has gained in specificity in another way, Cook notes that in Sarce, tj'i- is "chosennot necessarily because the identity of the subjectin question is unknown or unspecified, but bcausethe identity is inelevant or redundant in a given context" (1984:195). The reference of the prefix is thus still unspecified, but it is not necessarily nonspecific.The verb below is often used when the identity of dre speakeris known. (23) Sarcee (Athabaskan): Cook 1984:195 aas'tnllaut 'someone said' The Apacheanlanguagesshow a similar development, where the cognate prcfix distinguishes multiple thkd persons in discourse. It represents the central character, presumably the personwhoseidentificadonwould be redundantso long as a single point of view is mainrained. Finally in Hupa, a Pacific Athabaskan language,the development of specificity is complete. Goddardreporredin.191l that rhe cognate prefix wasusedwhen speaking of adult Hupa. (A different prefix was used for "very aged people, members of other Fibes and races, and animals" (Goddard 191l:117)). According to Golla (1985) the prefix i'i- now servesas the rcgular third person subject pronoun.

MITHTJN MARIANNE (24) Hupa (Athabaskar): Golla 1985:210,'14) 'he went offl ch'itehsyay 'he hits mc' ch'iwhirkis 2-4.2 The Proto-Athabaskan urspecified hwnar obiect prefa A second indefinite prefix can b reconstructed for Proto-Athabaskan that marks spaces,times, evenls, abstractentities, etc. It also functions as the object counlerpart of the unspecified human subject prefltxesdescribdabove. (25) Athabaskanunspecifiedhumanobject prefix Kiowa Apache: gddAyieel 'he dreams about someone'

(Bitde1963) Sarcee: (Cook1984) grizisxili 'he killed a person'

this indefiniteobjctprefix resembles in manylanguages, Like indefinitepronouns plurality of prefix usedin mostof the languages to emphasize anotherAthabaskan person. third (Athabaskan): (26) Sarcee Crok 1984:201 gfr?ds 'hewill kick them' is also usedin the This indefinite object prefix, like its subjectcounterpart, andto thedeicticcenonerEspects Apachean languages to rcfer indircctlyto Persons into a full It hasalsodeveloped ter of a discourse involving multiple third persons. In Hupa,thecognate fledgedbasicthird person specificpronounin somelanguages. prefix.ro- has becomea fully specificthird personreferentialpronoun:it marks human dirct objects. (27) Hupa(Athabaskan): Golla 1985:44,46 'he hitshim' ch'ixolkis 'I sethin/her (he baby)down' no:mnehhe:n 2.4.3.TheProto-Athabaskan inanirwteprertx unspecilied for Proto-Athabaskan meamng Finally, a third indefiniteprefix canbereconstrueted lhe pai$ of verbsbelow. 'something'.C-ompare

RECONSTRUCTING THEUNIDET{'NFIED (28) Navajo (Athabaskan); Hoijer 1945: 198,201 ?adld 'he's drinking' yidl6 'he's drinkingit' (29) Chipewyan (Athabaskan): Li 1946:416 ?eW 'he is eating (several objects)' yeldel 'he is eating them' (30) Sarcee (Athabaskan): Cook 1984:199, 198 filtsdl 'I'll throw something' dtstsd+ 'I'll throw it' (31) Tanaina (Athabaskan): Tenenbaum 1978:63 chik'datuq 'he killed somthing' chiydalyug 'he killed him' (32) Slave(Athabaskan): Rice 1989:629 nd?enehfu 'I am sewingsomething' ruirchfu 'I am sewing a parka' ?endkee?e

339

(1978;64) Tenenbaum notesfor Tanaina, and Rice for Slave(1989:629) that this prefix is usd additionally for specific objects whose identity is understoodfrom context or generalculture. (33) Slave(Athabaskan): Rice 1989:629 'I undressed bebi kardmewiwa the baby' (clothing) (took unspecified.object off the baby) yek'e?ed6?q 'Vhe ironed it' (iron) (handled3-dimensionalunspec.objct on it)

It is easy to imagine how the use of this prefix to refer to identifiable objecrs could lead to its reinterpretation as a referendal prcnoun. In fact in Sarcee,the prefix is used not only for unspecifiednonhumanobjects, but also for specific singular objectsin someparadigms. (34) Sarcee(Athabaskan):Cook 1984:202 ints?ds 'I'll kick him' intsl 'I called him' The indefinite prefixes of the Athabaskan languages thus resembleindefinire markersin other unrelatedlanguages, servingin somecasesas detransitivizers, as nominalizen, and as pluralize$. Each hasadditionallyevolvedin some,but not all,

MARIANNE MITHUN

of the languagesinto specific referential pronouns, a developmentperhapsfacilitated by the absence of basic third personmarkersfrom the parent language.

Unidentified cognates of Mohawk wa-

Recall that Mohawk contains two prefixes of the form l"a-, one refrring to third person neuter agents,the other marking plurality, In themselves,these ffo might not appear sufficiently similar semantically to be related diachronically. Specific pronouns and plural markers do recur cross-linguistically, howeyer, in clusters of cognate morphemesrelated through indefinite ma*ers. (None of the languagesdescribed in section 2 is demonstsablyrelated to Mohawk.) If tracesof an earlier indefinite pronoun *wa- could be found within Mohawk or its relatives,this form could provide the missing semanticlink betweenthe modem neuter agentand plural markers. Unfortunately, there is litde evidence in Mohawk or any of the modern Iroquoian languages of an earlier indefinite pronoun wd-. But the Iroquoian family is remotely related to arother family: Siouan. In the Siouan languages,verbs contain pronominal prefixes for first and second persons, but not third. Third persons are generally designatedby separare nominals, demonstratives,or nothing at all. Nearly all of the modem Siouan languages contain a prefix that indicates the participation of an otherwise unidentified patient: usually lea-. Comparethe meanings of the two Hidarcavefusin (35). (35) Hidatsa(Siouan): Haris wahkiwa'x uc l-ask-FlNAL 'I askedOim)' & Voegelin1939:193 wu'wahkiwa'x'uc somethingJ-ask-FINAL 'I asked(him) something, questions'

Cognateprefixes throughout the family have similar functions, nra*ing the involvement of unidentified paiients of either inEalsitives or lansitives (36) Crow (Siouan): l-owie 1960:vi '(he) sees' tkyak' ba'tkyak' '(he) seessomething' (37) Mandan(Siouan): Kennard1936:38 sukina ce'rekha wd:xkahanarn4kaha'ene children noisily were chasing something it is said 'The children were noisily chasing something.' (38; Lakhota(Siouan)Sun Redbirdp.c. spqrdye 'I cooked(it)' waip(waye 'I'm cooking,I cooked'

RECONSTRUCTING THEUNIDEN'flFED (39) Winnebago (Siouan): Lipkind 1945:17 nrc' 'to eat (it)' |9oruc' 'to eat' (40) Ioway-Oto(Siouan): Whitman 1947:244 wafulJe'to sew (it)' wawd0t4|e 'to sew (things,something)' (41) Osage(Siouanl:Wolff 1952:234 'you shorus' @lalce

341

wofukrice

'you shotsomthing'

(42) Biloxi (Siouan): Einaudi1976:78 pehe 'pound' apehe 'pound something' As in most languages, the Siouanindefinite patient markersfunction as detlansitivizers, (43) Ioway-Oro(Siouan): Whitman 1947:244 alfuce 'to marry ftim)' wdlixe 'to marry,takea husband' (44) Osage(Siouan): La Flesche1932 tuhd 'to show,display' wabdhl 'to signal' Sometimes the resultingintransitiveverbsake on lives of their own as lexical items whose meaningsare more than the sum of their pans. (45) Osage(Siouan): La Flesche1932

tutd

wabdhi bdlobe wdba!ibe

'to pick,sort' 'to gnze' 'to slice, cut up 'to cut skinintopieces for making moccasins'

Like indefinite markersin other languagefamilies, the Siouan prefixes also serve as nominalizers. They are more productive in some languagesthan othrs, but all languagesretain a large set of nominals containing them. The derived nominals refer to the patient of the base verb, whether it is transitive or intransitive, or sometimes to agents.

342

MARIANNE MTHUN (tl6) Hidatsa (Siouan): Hanis & Voegelin 1939:206 winpa'c '(he) lied' va'witaPa'I 'the liar' (47) Crow (Siouan): I-owie 1960 tst 'to carry on back' bo'tst 'a load' (48) Lakhota (Siouan); Stan Redbird p.c. tqp'l 'wear aroundthe neck, wan(p'i 'necklace,choker' (49) Winnebago (Siouan): Lipkind 1945:17 Jt 'be fat' wali' 'faC (50) Ioway-Oto(Siouan): Whitman 1947:244 htr'e 'to eat' valrtJe 'food, a table' (51) Osage(Siouan): La Flesche1932 bmse' 'to sew' waba'tse 'to sew,one who sews,seamstress' (52) Biloxi (Siouan); Einaudi1926 duti 'ear sduri 'food' (53) Tutelo (Siouan): Hale 1883 ketgi 'to dance, wagitei 'a dance'

Proto-Siouan, remotely related to hoto-hoquoian, thus contained the very affix _ neededto link the modem Mohawk third personand plural wc- prefixes: an indefinite patientprcfix *wa-. Like indefinite affixes in many languages, proto-siouan *wa_ functioned as a detransitivircr and as a nominalizer as well. Recall that cross-linguistically, it is not uncommonfor indefinite markersto function as plural markers as well. There is so much variety among plural markers in the modern Siouan languages that no plural affix can be reconstructed for protoSiouan.In a few of them,however,wa- hasnow developed into a marker of plural_ ity of thid person objecrs.

RECONSTRUCTING THEUNIDENTIMD (54) Winnebago (Siouan):Lipkind 1945:22

343

rahe' torahe'

'thou buriest(him)' 'thouburiest thm'

(55) Ioway-Oto (Siouan):Whitrnan i947:2,14 '(she)hasit' di( '(she) hasthem' wdfii (56) Ponca(Siouan):Boas& Swanton1911:915 'I heard (it)' ann?d 'I heardaboutthem' awdna?an (57) Osage(Siouan): La Flesche1932:325 'you see' shton'-be 'you sawthem' va-shtott '-be Traces of the same marker can be found in other distant relatives of Mohawk, The Iroquoianand Siouanfamilies are remotelyrclatedto a third family: Caddoan. The Caddoanlanguages,like the koquoian and Siouan languages,contain agent and patient pronominirlprefixes within their verbs, Seyeralof the Caddoanlanguages also contain affixes with the shape1'a. In Wichita, a distdbutive infix wc- appearsbefore the last consonantof the verb stem. According io Rood, this infix "generally refers to a spreadingout of the stateor event, although it sometimesseemsto imply plurality of the patient of the verb instead"(1976:76). (58) Wichita (Caddoan): Rood 1976:76 'it is getting dry' tikakics 'everythingis geningdry' tikakiciwa:ris [:hkiwas hiri:hkiwa:wlskih

'heloses'
'when he was losing everything'

prefix In a few inu-ansitive verbs,a prefix wa- (in combinationwith the non-singular li?-) mark dual numberof the subject. (59) Wichita (Caddoan): Rood 1976:187 'she got up' ri:hri: his?as hit' dri:hiskws?as 'they two got up' Both Arikara and Pawneecontain cognatedistributive markers wa:. In Arikara, the distributive immediately precedesthe verb stem (Merlan l9?5:64). In Pawnee,the marker is sometimesa prefix, sometimesa suffix, and sometimesan infix. It ex-

MARIANNE MITHUI{

presses variously"ar activity or statedisuibutedover a wide area,iterativeaction, distributive plural object of the verb, and dual and/orplural subjectof the verb" (Parks1976:279). in (60). Compare theverb stems (60)Pawnee (Caddoan): Parks i976:186 'to &ink' ki:tu: 'to drink (various things)' ki:kawa
ariki wa:ariki he:r hewa:

(sg)' 'stand 'stand(du, pl)'

'be good (sg,du)' 'be good (pl)'

Like the other Caddoanlanguages,Caddo contains a variety of number markerl A prefix wa-, which appearsbtween two segmentsof the verb stem, indicates pluraliry of an animnte agentor patient. (61) Caddo(Caddoan): Chafe19?6:69 dibdwnah '(he) saw (him)' diwab6wnah 'they (pl) saw (him)' '(he) saw them (pl)' hak6ywabewnah 'they (pl) saw them (pl)' hakdywabriwnah t'6yb6wnah C6''wabdwnah 'I sawyou (sg)' 'we (pl) saw you (sg)'

Recall that cross-linguistically, it is not unusualfor indefinite pronounsto be used for specific reference,and even to evolve ino generalreferential pronouns, particularly in languagesoriginally lacking third personpronominal forms. Proto-Siouan contained first and secondpenon pronominal prefixes but not third- Proto-Caddoan contained frst and secondpersonpronominal prefixes, ard only indefinite third. The modern Iroquoian languagesdiffer among themselves in their sets of third penon pronquns. It is probable that their common ancestor, Proto-Siouan-Caddoan-Iroquoian, lacked basic third person pronouns. The most likely source of one of these indefinite markeB, the reuter agentrr(a)-, is a Proto-Siouan-Caddoan-Iroquoian marker *wc-, a not unusualdevelopmentcross-linguistically.

Conclusion

The partial resemblarce betweenthe Mohawk plural marker wa- and the neurcr singular agent marker w- is probably no coincidence after all. The two are apparently

rHE UNIDENnruD REcoNsrRUcrrNc

345

related diachronically through an earlier marker of unidentified or indefinite participants that still persists with that function in the remotely related Siouan languages. Similar developments, from indefinite to definke markers, and from indefinite to plural markers, are not uncommon cross-linguistically. Tho recognition of the recurring relationshipsamong indefinite, plural, and definite pronominal affixes, has further value even within Mohawk, alerting us to possible diachronic reladves of other third person markers that have previously escapednotice. Recall that the Mohawk neuter agent allomorph w- altemates with ka-, as in kahnekihrha?'it ot she is drinking'. This prefix has cognateswith the in the sole Southem samemeaningthroughoutthe Northem hoquoian languages; language,Cherokee,the cognateform is the generalthird person agent pronoun: kd|wo3ni2he '(s)he's spealdog'(Pulre& Feeling \9'l5.275). A prefix of the same shape also appearsas plural in two other Iroquoian languages,Tuscarora and Cayuga. In Tuscarora,tc- marks plurality of all third persons:compare wa?Vih :ri? 'she ate' and. 'they ate' @lton Green p.c.). In Cayuga,/<a-marks wa?kay?ts-&r:ri? plurality of feminine and rnixed third pe6ons: compared?C:yg?'she arrived' and attieye?'they arrived' (Reginald Henry p.c.). The same form sufaces as a plural marker in the remotely relatedCaddoanfamily. In Pawnee,a prefix ,td- indicates iteration or plural object with a few verbs: tzra 'to make them' (Parks 1976:285). The same prefix appearsas an indefinite marker in both Siouan and Caddoan. The indefinite pronoun 'something'in Biloxi (Siouan) is tawa (Einaudi 1976:142). Wichita (Caddoan)containsan indefinite prefix ,t4.-, as in the pronoun /.a.'-&iri? 'somethilg' or katkiya? 'someone', erc. (Rood 1976:11).A nominalizer of the same shape also appearsin some Siouan languages:note for example Mandan karo're'a speaker'frKennadi936) andld(hotawaktdka 'aklller' . Similar clustersof homophones can be found for the Mohawk feminine pronominalprefixes.Within Mohawk and the other Nonhem Iroquoianlanguages, the agent prefix )e- is used for both women and for unspecified or indehnite persons. A cognateform is usedonly for unidentihedpersonsin Cherokee.A cognaleprefix, *yi- referring only to unidenti{iedor unspecifiedagents,can be reconstructed for (Chafe 1976).The developmentof the modem Mohawk feminine Proto-Caddoan prefix from an original unspecified human agent prefix is clear. In most of the Northem Iroquoian languages, the samepronoun is also usedfor third personcollective goups of people. referents, homoThe modernMohawk femininepatientprefix ya,to- is also suggestivly phonous with severalindefinite, nominalizing, and number markers in remotely rclated languages. Pawnee(Caddoan)containsan indefinite patient proclitic &u-: (Siouan) ku?ista:huras'ifyou find anything' (Parks1976:263). Hidatsa contains an agentiyenominalizeraku-: ukuwa'?dii 'hunter' (Robinett 1955:162). Crow containsan agentive nominalizrrak-: akpd:pd,.xarura All of 'pusher'(Kashube1969:19). the Caddoan languagesalso contain plural markers of third person animate patients 4&-: Pawneeta:hkuras 'I found them (people)' (Parks 1976:169). Whether or not we can ultimately relate theseforms diachronically will of

346

MARIANM MITHUN

course depend upon further comparative work, In any case, an awarenss of the clusters of functions signalted by homophonousmarkerscrossJinguistically can be a valuable tool in linguistic rcconstruction; it can alert us ro lhe possibility of historical rclationships among grammatical forms that might otherwise escapenotice, and play a major role in guiding us to the discovery of missing links.

NOIE
I I am gratefll to fte following speakerswho have generously sharcd their expenise on their lsnguages:lhe lare Mr. Elton Greene,of Lewisron, New York, on Tuscarcra;Mr. Reginald Henry, of Six Nations, Ontario, on Cayuga; Mrs, FrancesJack of Hopladd, Califomia, on Ccntral Pomo, Mrs. MFde Pelerson,of Stamburg,New York, on Seneca,Mr. S6nley Redbird, of Rosebud, SouthDakoh, on Lkhota,and the Naove languagetachors involved in Mohawk linguistics classes at Caughnawaga, Oka, and Akwesasne in Quebec, and the Six NationsReserve in Onlario.

REFERENCES
Bitdg, William E. t963. "Kiowa-Apache"- Studiesin the AthapaskanLanguaSesC University of California Prnicotions in Un8uistics,29),1G101. Berkeley:University o[ Califomia Press. Boas, Fmnz,l9l la.'chinook". Boas19llb:559-677. Haadbookof Noflh Americar lndian languages,Pa l.Washing0on: Golemmenr ----, en '9l'!.Officc. hinaing l9ll. "Siouan". Boasl9llb:875-966. ---- & JohoR. Swanton. Campbell,Lyle. 1985.Ire Pipil LanguaSe of EI Saryodor.Berlin: Mourcn. Chrfg WallaceL. f976. The Caddool, Iroquoian, and Siouanladguages.The Hagle: Mouron. Co*, Euog-Do, 1984.A SarceeGrammar.Vurcouver:University ofBritish Columbiap.ess. Einaudi,+aufaFenis. 1976,A Grammart Biloi. New York Garland. Fras., Alexande.. 1920.Fifteenth Report of the Bureau of Archives for the prcvince of Ontario: .1918-.19,19. Toronto: 'Ihe tgislarive Assemblyof Onlario. Goddard, Pliny Earle 19l l. "Arhapascar(Hupa)".Boas191rb:85-158. Golla, Viclor. 1976."Tutumi (OrgonAthapaskan)", Intenntional Journal of American Linluistics

42.2n-2n-

-----.1985. A ShortPractical Cr@mar of Hupa. Hoopa:HoopaValley Tribe. Hale, Horatio. 1883.'"fhe Tutelo Tribe andLangrPl$e{'.PnceedinSsof theAmerican philosophical Societt 2il-45. Harris, Zelfig & C.F. Voegeli'I.. 1939. Hidatsa Texts Collected b, Robe H. Lowie. prchistry Research Series,IndianaHistorical Society.Volumg I, Number6. Hoijer, I|arry. 1946.'ChiracahuaApache".LinSr,rj'ic Structures of Native America (= Viki,r8 FMd Publicationr in AnthrottoloS], 6), by Hary Hoijer et al., 55-84,New york: Thc Viking Fund. ----._l??-1. 'Alraqaskan Morphology", Jrtdt:ej i^ An|f..ricatt Indion Lan|uages (= Ilntue\itJ of Califonb P\blications ia Lhgn,rricr, 65), I 13-148.Be*eley: Universiry of Califomia press.

RECoNSTRUCTING TFIE UNIDENnFIED

347

Kashube,DorotheaV. 1969.Structural Elementsof the ktnguage of the Crow Indians of Montana. Boulder:Universityof Coloradopress. KennArd, Edward.1936."M6ndanGmmmar".,lnt? rnationatJournal of AmericanLinguktics 9.143. Kimball, Geoffrey.1985. A Descriptine Grammar of Koasati.ph,D,Dissertarion, TulaneUnivcr_ sity. la Fle,sche, Francis. 1932.A Dictionory of the OsagelanSuage. (= Bweau of Ame can EthnoloSy Srlrrair, 109.)WashiDglon, D.C.: Smirhsonian. Li, Fang-Kuei. 1946."Chipewyan". IirguiJtjc Structuresof Natiye America (=Viking Fund publ._ york: The Viking Fund. . cationsin AnthtopoloSy,6), by Harry Hoijer et al., 398-423,New Lipkind, Wiliam. 1945 . WinnebagoGran rar. New york: King' s Crown press. Irwie, Roben H. 1 0. Cro|| Word Lists.Berkeley:Universily of Califomia pre,ss. Merlan, Fmncesca.1975-Noun-VerbRelationshipsin Arikaro SJntax. ph.D. DisserEtion, Universily ofNew Mexico. Munro,Pamela. 1914.Topicsin MojoveJ)rlar. ph.D. Dissertation, Universityof California, San Diego. Par*s,DouglasR. 1976.A Gtunnar of Pa$,aee. New york Carland. Pulle,William & DurbinFeeling.19?5. "Ourlineof Cherokee Gramm u,,. Cherokce-Dnptish Dic_ ,iorary. Tahlequah: Chroke Nation of Oklahoma.235-355. Rice, Keren. 1989.A Grantmarof Slave. Betlin: MouFn. Robinett, Fforcnce M. 1955."Hida6a II: Affixes".International Journalof AmericanLintuistics 2t.t&-77. Rood, David S. 1916.Wichio Gtunnat New Yorlc Garland. Swanton, 'The Tunics Iaguage" , InternarianalJournal of American Linluistics 2,1John R. 1E23. 39. "A Skerch ----. 1929. of the Alala!6 t&guage".Inte at{onalJournal of AmericanLinSuistics

5.tzt-r49.

TenenbalJm, Joan Marsha. 1978-Morphologt and Semantics of the Tanaina yerr. ph.D. Disscaa_ tion, Columbia Univeisity. WhiEna4 William. 1947."Descriptive Crammarof loway-Oto',./rtet@tiona! Jounal of American Linguistics 13, 23348. Wolfft l{ans. 1952."OsageII: Morphology". Intenational Journal oI American Linguistics 1g.231Young,RobenW. & William Morgan.1980. ThaNayajoLan8ud8e. Albuquerque: Univcrsityof New Mexico Press.