<<

Anthropol. Sci. 103(3), 209-225, 1995

Proto-Amerind Numerals

MERRITT RUHLEN 4335 Cesano Court, Palo Alto, CA 94306, U.S.A.

Received December 5, 1995

•ôGH•ô Abstract•ôGS•ô The Amerind family includes all the aboriginal of North and South America, except for those belonging to the Eskimo-Aleut and Na- Dene families. Comparative linguistic evidence from extant (or attested) indicates that Proto-Amerind-the language from which all Amerind

languages derive-used a system of counting in which an obligatory numeral

prefix, •ôUH•ô*•ôUS•ô•ôNH•ône•ôNS•ô-,preceded the numeral root. The first three numerals in Proto-Amerind seem to have been •ôUH•ô*•ôUS•ô•ôNH•ône•ôNS•ô-•ôNH•ôk'•ôUH•ôw•ôUS•ôe•ôNS•ô`1,' •ôUH•ô*•ôUS•ô•ôNH•ône-pale•ôNS•ô`2,' and •ôUH•ô*•ôUS•ô•ôNH•ône-q•ôUH•ôw•ôUS•ôalas•ôNS•ô `3.' A fourth numeral, Proto-Amerind •ôUH•ô*•ôUS•ô•ôNH•ôta-pale•ôNS•ô`4,' combined a reflexive prefix with the Proto- Amerind root for `2' in order to express the number `4.'

•ôGH•ô Key Words•ôNS•ô: anthropological , Amerind languages, numerals

INTRODUCTION

The Amerind , as defined by (1987), includes

all the indigenous languages of the Americas, with the exception of those belonging

to the Eskimo-Aleut and Na-Dene families1. The Eskimo-Aleut family consists of

what are essentially two languages, spread across the northern periphery of North

America from Alaska to Greenland. The Na-Dene family consists of four branches,

three of which-Haida, Tlingit, and Eyak-are single languages. The fourth

branch is the Athabaskan family, consisting of several dozen very similar languages,

which are found in interior Alaska and northwestern Canada, with outliers in

Oregon, on the northern California coast, and in the American southwest (Navajo,

Apache). All of the remaining Native American languages, which probably

numbered between one and two thousand at the time of the arrival of Columbus,

belong to the Amerind family.

Though the antiquity of Proto-Amerind-the language from which all extant

or attested Amerind languages derive-cannot be established on linguistic grounds

by any currently known method, the archaeological record in the

strongly suggests that the Americas were uninhabited by humans until 12,000-

11,000 years ago, at which time both North and South America were quickly

populated (Klein,1989, pp. 389-92). If we may identify the first immigrants to the Americas with the Amerind language family, which seems plausible based on the

distribution of Amerind languages, then the age of Proto-Amerind would be

approximately 11,000 years old.

In the present article I would like to offer a preliminary analysis of the Proto-

1 This study is dedicated to the founder of Amerind linguistics, Joseph H. Greenberg, on the occasion of his eightieth birthday, May 28, 1995. 210 M. RUHLEN

Amerind numeral system. Although few contemporary languages have preserved the system itself-the Algonquian family is an exception-many Amerind languages of both North and South America have preserved fossilized relics of the Proto- Amerind system. Furthermore, it will be shown that certain family-specific prob lems,which have remained either unexplained, or, worse, incorrectly explained, are easily resolved within the broader Amerind perspective.

PROTO-AMERIND `1' Let us begin our study of Proto-Amerind numerals with the proposal by in 1913 that two contiguous-but very different-languages spoken along the northern California coast, Wiyot and Yurok, were most closely related to the widespread Algonquian family that stretched from the Atlantic seaboard to northern Montana. I have already examined in some detail the bitter controversy that this proposal engendered (Ruhlen, 1994c). What I would like to do here is to pursue one thread of Sapir's argument beyond the Algic family that he identified in his seminal article. One of the etymologies that Sapir proposed in support of Algic (=Algonquian +Wiyot+Yurok) was the number `1': Wiyot go't-, Yurok qoot-, and Algonquian forms such as Fox neguti, Ojibwa ningoto-, Natick nequt, and Blackfoot nitukska, which Sapir suggested had arisen from earlier *nituki-, and still earlier *nikuti-, by metathesis. The Algonquian forms differed from the Wiyot and Yurok forms by the presence of an initial ni- or ne-, which appears at the beginning of all Proto- Algonquian numerals from `1' to `8.' To explain this discrepancy Sapir surmised that perhaps "original Algonkin -got- [was] modified to nI-got- by analogy of other numerals with n-" (Sapir, 1913, p. 630). In his celebrated reconstruction of Proto-Central Algonquian, (1946) proposed *nekotw-i, but consideration of the Eastern has led modern scholars to revise this to *ne-kwet-wi, where the initial vowel of the root was actually the diphthong we, reduced to o in the central and western languages. The initial prefix and the final suffix seen on the numeral `1' are also found on the numerals 2-5 in Algonquian (Picard, 1986). With regard to Algic, the modern view is that Wiyot and Yurok have lost the initial prefix, rather than that it spread to the number `1' analogically, as Sapir had guessed. Proulx (1984, p. 184) reconstructs Proto-Algic *ne-kwet-, or *ne-kwets-, `1.' But does the story of this particular numeral end at the borders of Algic, or can traces of this numeral be found elsewhere in the Amerind family to which Algic belongs? I would like to now present evidence that the Proto-Algic numeral `1' just discussed most probably already existed in Proto-Amerind, in only slightly different form, and, furthermore, that it formed part of a system of counting that can be projected back to Proto-Amerind. Evidence from the numerals `2' and `3' will also Proto-Amerind Numerals 211

be adduced in support of the Proto-Amerind system.

Joseph Greenberg (1987, p. 175) proposed that the Algic forms are related to

Proto-Salish *nak', or *nek'-u? (Kuipers, 1970, p. 61), which has given rise to such modern forms as Coeur d' Alene nƒÃk'w-ƒÃ' '1,' nik'w-`be tribe' (i.e. `be one with the tribe, be one of the tribe'), Flathead nk'wu, Kalispel nk'wu?, Wenatchee-Columbian nk'w-, Musqueam *nets'-a?, Lummi nets'e?, Duwamish dec'u?, Twana dec'o-,

Clallam nets'u?, Upper Chehalis nacv'aw, Colville-Okanagan nk'w, Spokane nk'w'u?,

Saanich nets'e?, Samish netsa?, and Tillamook nec'-. Kuipers himself remarked on the "striking" resemblance of the Proto-Salish forms with Kutenai -ok'w(e)- `to be one,' n?ao:k'we: `the one, the other,' na?nka `orphan.' He also noticed the similarity between Shuswap s-ne-nke `orphan' and Kutenai na?nka `orphan' (cf. also Thompson s-nen' eke? 'orphan'), though he attributed this resemblance to "linguistic contacts" rather than to convergence or common origin. Yet an additional remnant of the

Proto-Amerind word for `1' appears to have been preserved in the Kutenai noun class marker -nik? `one of, one belonging to.'

In fact, as Kuipers noted, the resemblance would be even more striking if one accepted the alternate Proto-Salish reconstruction, first proposed by

(1949, p. 171), *nek'w-e?, but Kuipers opts for a plain velar ejective-which became labialized before-u?-because the alternate solution, with a labialized velar, is "more complicated for Salish" (Kuipers, 1970, p. 61). In a nutshell, the ambiguity in the reconstruction of this form is due to the fact that in the two basic subgroups of Salish we find reflexes which presuppose a different Proto-Salish reconstruction. Proto-Coast Salish *nac"-(e.g. Tillamook nec'-) would require

Proto-Salish *nak'-, while Proto-Interior *nak'w- (e.g. Coeur d'Alene nƒÃk'w-) indicates rather Proto-Salish *nak'w-. It is for this reason that there has been a dispute among Salish scholars over the character of the final consonant. Who is right?

There are, at least, two possible explanations for this distribution, which might be called the Salish-internal and Salish-external explanations. The Salish-internal explanation sees Proto-Salish *nak'- developing regularly into Proto-Coast-Salish *nac'- , but irregularly into Proto-Interior Salish *nak'w under the influence of a following labializing suffix, -aw'>-u?. The Salish-external explanation would see

Proto-Salish *nak'w- as remaining unchanged in the Interior languages, while undergoing an irregular delabialization of the consonant to *nak'-in Proto-Coast-

Salish, after which it shares the regular outcome of -k' in these languages, leading to *nac'. The difficulty with the external explanation is that it requires an ad hoc delabialization of the consonant in Proto-Coast-Salish; the advantage is that the

Salish reconstruction would agree exactly with the labialized velar seen in the

Wakashan, Kutenai, and Algic words for `1,' which, except for this labialization, are highly similar to the Salish word. The advantages of the internal explanation 212 M. RUHLEN

are that it doesn't require the ad hoc delabialization in Proto-Coast-Salish, and it provides a plausible phonetic environment for the subsequent labialization in the interior languages; the disadvantage of the internal solution is that the Proto-Salish form differs from the Wakashan, Kutenai, and Algic words in lacking labialization and thus, if these forms are all cognate, Proto-Almosan *-k'w would have developed into Proto-Salish *-k', only to revert to Proto-Interior *-k'w at some later stage. It seems to me that neither solution can be ruled out a priori and that either may be correct. But if I had to bet, I would bet on the external solution as the historically reel one. Morris Swadesh (1953, p. 230) first pointed out, over four decades ago, the resemblances between the Salish forms and similar Wakashan forms. Lincoln and Rath (1980, pp. 132-33) cite such North Wakashan forms as Heiltsuk n'hxw-, Haisla n'axw-, and Kwakwala n'axw-. Probably Kwakwala na'xula `alone' is also related, and perhaps also Kwakwala -k'ot `one side.' It would thus appear that this particular numeral has been retained in no less than six of the seven branches of Almosan (all but Chemakuan). (1965, pp. 87-88) pointed out the similarity of the first three numerals in Almosan (=Algic+Kutenai+Salish+Wakashan+Chemakuan) languages and concluded that this (and other) "evidence that has been adduced is too substantial to be explained away as entirely the result of borrowing or accident.... [The similarity of] the first three numerals ... seems to me to be sufficient to indicate that the whole problem is worthy of further study." As the remainder of this paper will try to show, further investigation indicates that the distribution of the numeral '1' goes well beyond Almosan , and is in fact most likely attributable to Proto- Amerind. We shall also find that the Almosan numeral for `3' has an equally broad distribution. The numeral `2' noticed by Haas, however, seems to be an innovation within the Almosan branch of Amerind. Greenberg (1987, p. 117) had, in fact, given another etymology for `1,' Chibchan- Paezan No. 142, whose forms bear a striking resemblance with the Almosan forms just enumerated. In the Paezan branch, in southern South America we have Millcayac negui, virtually identical with the Proto-Algic reconstruction, while on the northwestern coast we find Chimu onko and Eten unik. In the Chibchan branch, on the northern coast of South America, we encounter Ika in?gui, Damana ingwi, and Atanque eingui `other.' The presence of such forms in two widely separated Amerind branches-together with their strong phonetic similarity and semantic identity-suggests that perhaps traces of this root might be found in yet other Amerind branches, and such proves to be the case. An isolated example in Keresiouan-the group most closely related to Almosan may be Adai nancas in the Caddoan family. In the Penutian branch of Amerind we find Siuslaw nik'a `alone,' Sahaptin naxs, Proto-AmerindNumerals 213

Nez Perce nagc, Klamath nas, Cayuse na., Molale na_??__??_a,Modoc nas, and, without the numeral prefix, Wintun k'et-e., Patwin ?ete-, and Yokuts yet'. These languages are found along the western coast of the from southern Washington to central California. Aoki (1963, p. 110) connected the Sahaptin, Nez Perce, Molale, Cayuse, and Klamath forms; he later compared the Sahaptian forms with Proto-Salish and Proto-Central Algonquian forms (Aoki, 1975, p. 194). DeLancey, Genetti, and Rude (1988, p. 215) compared the Nez Perce, Sahaptin, and Klamath forms. Notice that the Nez Perce and Wintun forms contain a final-tS or-t, reminiscent of Proulx's Proto-Algic reconstructions for this numeral. In addition to the Sahaptin numeral naxs `1,' the Sahaptin verb-anakw- `to leave, desert,' which may be nominalized as anakwet `deserted spouse,' may be etymologically con nected.In the Gulf branch of Penutian, in the American southeast, Swadesh (1946, p. 126) reconstructed *nik'u `1' for Proto-Atakapa-Chitimacha, with reflexes Eastern Atakapa hanik, Western Atakapa tanuk, and Chitimacha ?ungu. There are also several suffixes in the Gulf group that appear to be grammaticalizations of this root: Chitimacha -nk'is `just, only,' Natchez-neke `like, as if,' Tunica-nahku `like,' and Atakapa hinak `like.' Pamela Munro (1994, p. 183) connects the Natchez, Tunica, and Atakapa forms. Another possible reflex of this root may be contained in the Chinook word for `5,' gwanma. Dell Hymes has suggested (pers. corn.) that the modern form may derive from an earlier *qwet-ma through nasal assimilation. If this is correct, then the original meaning was `one hand.' A final possible cognate is Coast Tsimshian k'wi-daat `1' (used in counting human beings aboard canoes). For the Hokan branch of Amerind, Terrence Kaufman (1988, p. 124) has reconstructed Proto-Hokan *ky'a `1.' Kaufman gives no supporting evidence for his reconstruction, but presumably forms such as Yana k'ai `only, one,' and Kashaya ku-hku `1,' are involved in this reconstruction. Also of interest is the Washo word for `1,' lak'a-. William Jacobsen (1979, p. 78) noted the similarity of this numeral to the Proto-Salish, Proto-Sahaptian, and Proto-Central Algonquian forms of the same numeral. Furthermore, the apparent shift of initial *n- to l- in this numeral finds a parallel in the prevocalic allomorph of the Washo first-person prefix le, which derives from *la-, which in turn derives from *na-, this last form being the general Amerind first-person pronoun (the preconsonantal allomorph di-derives from earlier *ni-). In the Coahuiltecan branch of Hokan, possibly Karankawa natsa `1' and Comecrudo nawis `1' preserve both the numeral prefix and the root in question. For the Central Amerind branch similar forms are strongly attested in the Oto - Manguean family, for which David Rensch (1976, pp. 244-45) reconstructs *(Y)(n)kwe(h)(n) `1.' Reflexes of this reconstruction include Popoloca hnku, Proto- Mazatec *(h)nku-*ngu, Chocho ngu, Popoloco (Oaxaca) nakua, Amuzgo nkwi, Mazahua n?aha, Otomi n?na, South Pame ?na, and Trique ?ngo. In addition to the Chibchan-Paezan forms in South America already examined, 214 M. RUHLEN

one may also point to Chimila kute, Chumulu kue, Gualaca kue, and probably Cuna kwena. In other South American branches of Amerind we find, in Macro-Ge, Opaie enex-ha and Yabuti nici, while in Macro-Panoan, Atsahuaca nikatsu and perhaps Amahuaca naa are related. Another intriguing form within Macro-Panoan is Caduveo nigot nubaarat `5,' in which the second element, nubaarat, is the word for `hand.' It is tempting to consider the first element, nigot, as a word for `1'; however, the word for `1' in Caduveo (oniude) is entirely different, and the meaning of nigot is not explained in the source, so this may be a chance resemblance. In the Zaparoan branch of Andean we have Zaparo nokoaki, Shimigae nikino, Conambo nukaki, Iquito noki, and Cahuapana nuki. In the Carib branch, Yabarana enix-pete recalls the Opaie form cited above, enex-ha, but these two languages are spoken almost 2,000 miles apart and belong to different Amerind subgroups. Within the Equatorial branch of Amerind, perhaps Proto-Maipuran *ki `first' (Payne, 1991, p. 403) belongs here as well. Finally, in the Macro-Tucanoan branch, both the numeral prefix and the root are preserved in various Tucano dialects. In the dialect described by Giacone (1949) there are three morphologically related forms of the number `1,' determined by gender: nike (masculine), niko (feminine), nika (neuter). This Tucano gender system matches exactly the Proto-Amerind gender ablaut system that I have posited on other grounds (Ruhlen, 1994d).

PROTO-AMERIND `2' The forms cited in the previous section indicate that Proto-Amerind must have had a numeral '1' of the general shape *ne-k'we-. However, of all the proposed reflexes of this original root examined above, only in Algic is there synchronic support for a nasal prefix before numerals. In all the other cases, as best I can tell, the nasal prefix has simply become fossilized as a part of the root, as in, for example, Proto-Salish *nak'w-. If, however, Proto-Amerind used a numeral prefix before the numeral `1,' it must have used it before other numerals as well-otherwise there would be no reason to analyze it as a prefix. Let us consider this possibility with regard to the numeral `2.' If the hypothesis sketched above is correct, then we should expect to find traces of the numeral prefix on the Proto-Amerind word for `2.' But what was the Proto-Amerind word for `2'? Greenberg (1987, p. 264) supplies the answer to this question by displaying a wide variety of forms from North and South America that might all be derived from an original *pale. (The precise quality of the second vowel is unclear and might well be represented as V, a vowel of unknown timbre. I will use -e-, however, as this vowel seems perhaps most common and would explain the frequent palatalization of -l- in numerous languages, e.g. Aymara paya, Yahgan pai (nominal dual), Atacama poya, Guambiana pay.) and I have supplied additional evidence for this Amerind root, but we have also shown that this is a root which Proto-AmerindNumerals 215

connects the Amerind family with numerous other language families from all parts of the world (Bengtson and Ruhlen, 1994, pp. 316-17). Is there any evidence for this hypothetical Amerind numeral *ne-pale `2'? I believe there is, at least in Almosan-Keresiouan and Penutian, and perhaps elsewhere. Within Almosan-Keresiouan, one is immediately struck by Proto-Siouan *nupa `2,' with reflexes such as Woccon numperre, Dakota nupa, Mandan nup, Osage noba, Crow nup, Chiwere nowe, Biloxi nopa, and Ofo nupa (Wolff, 1950). Catawba, the most divergent Siouan language, has napari `2,' while in Yuchi, the language most closely related to the Siouan family, the numeral `2' is nowe. The labialization of the vowel (e>u) before the labial consonant p is a common phonetic shift. In Tutelo the root seems to have been preserved in the word for `4,' paropreh, a reduplicated form of Proto-Amerind `2' (i.e. 2-2). In the Wakashan family Nootka has two suffixes that appear to be grammaticalizations of the Amerind root for `2,' -pala ' ... on each side' and -pal 'half....' Traces of Proto-Amerind *ne-pale `2' are also found in three branches of the Penutian family. In the Plateau subgroup in Oregon we find Sahaptin napu and Nez Perce lepu, on the basis of which Aoki (1962, p. 182) reconstructs Proto-Sahaptian *Nepu . Cayuse, also a member of the Plateau subgroup, has leplin and this form suggests that the consonant of the numeral prefix has assimilated to the l later in the word (*neplin>leplin), not only in Cayuse, but probably in Nez Perce as well, though in Nez Perce the second l was subsequently lost. Two other members of Plateau Penutian, Klamath lab and Molale leepka? show even more eroded forms of this numeral than Nez Perce. DeLancey, Genetti, and Rude (1988, p. 202) connected the Nez Perce, Sahaptin, and Klamath forms. To the south, in California Penutian, the root-but not the numeral prefix- is preserved in several languages. In Wintun we find pale-t `2' and pal winthuh `forty' (literally , `two persons'). In Patwin, a language closely related to Wintun, we find p'an- `twins,' and indeed-as the name Penutian itself attests-in California Penutian the-l has generally become-n, as, for example, in Proto- Maiduan *pen `2.' Further discussion of this point is given below. In the Gulf branch of Penutian, spoken in the southeastern United States, we find Eastern Atakapa hapalst `2,' Western Atakapa wos pe `2' (=`hand'+`2'), and Chitimacha ?upa `2.' On the basis of these forms, Swadesh (1946, p. 125) reconstructed Proto-Atakapa-Chitimacha *pa `2.' However, given the presence of root-final -l in Eastern Atakapa- and in the external context-it seems likely that a more realistic reconstruction would be *pal `2.' In addition, Tunica ?apalkali `doubled' also shows the final -l. The Yuki and Wappo languages, though spoken in northern California, belong to the Gulf subgroup of Penutian. Possibly Yuki nop `twins' is a direct reflex of Proto -Amerind *ne-pale `2,' while Wappo p'ala `twins' merely reflects the numeral root. Munro (1994, p. 205) connected the Tunica and 216 M. RUHLEN

Wappo forms. In the Muskogean family, also a member of Gulf Penutian, we find Alabama pa_??_-pa_??_at-`to split in two.' Here the meaning parallels that found in Huave apool `snap in two,' a Mexican Penutian form cited by Greenberg. Finally, in -pala the Amerind root for `2' has been grammaticalized as a verbal suffix meaning `... again' (to do something again is usually the same as doing it twice). What is particularly striking about the Sahaptian and Chitimacha examples is that both of these geographically distant languages have preserved traces of the numeral prefix with both of the Proto-Amerind numerals `1' and `2,' as seen in Table 1.

Table 1. Preservation of the numeral prefix and numeral root in Penutian languages

Evidence for this root is largely absent in Hokan, which is not surprising inasmuch as the name "Hokan" is based on yet another word for `2' (cf. Atsugewi hoqi).

However, Karl-Heinz Gursky (1968, p. 39) compared Swadesh's Proto-Atakapa-

Chitimacha *pa `2' with Subtiaba apu `2'; Tlappanec, closely related to Subtiaba, shows a•Ep•Eu' `twice.' From the Coahuiltecan branch of Hokan, Gursky adds

Comecrudo pa- `2' in the numerals pa-nawuyi `6' (cf. nawui `4') and pa-makwel `7' (cf . makwel `5'). Sapir (1925) added forms from the Yuman branch of Hokan to these comparisons: Havasupai hopa `4' and Walapai hupa? `4.' Finally, in northern California, Washo bu?-, a dual prefix, probably represents a grammaticalization of the Amerind numeral for `2.' The root *pale `2' is particularly widespread in the South American branches of

Amerind (Greenberg,1987, p. 264; Bengtson and Ruhlen,1994, p. 317); in addition to forms cited in those two works, one may add from the Tupi branch of Equatorial,

Curuaya pora-ka, Monde para-seram, Sanamaica pali-saru, and probably also

Wayampi pali `play area for children,' which is described as follows: "a play area for children inside a house which is made by cutting the space in two by means of a low barrier in the middle of the house" (Grenand, 1989, p. 337). In the Murato branch of Equatorial, we have Murato tsim-boro and Shapra tsim-boro. In the

Moseten branch of Macro-Panoan we find Moseten pora and Chimane pore. As a supplement to the Macro-Tucanoan forms cited by Greenberg, one may add

Chiranga , Tucano pearo, Tuyuca pealo, Wanana pearo, and Uasona peaLo.

To the Chibchan forms given by Greenberg, one may add Xinca bi-al-pi-ar and Lenca pa-pe; in the Yanomama group, Sanema pola-kapi, Yanomami pora- Proto-AmerindNumerals 217 kapi, and Yanomam pora-kapi; the second element in these forms is no doubt a reflex of the Proto-Amerind root *kapa `finger, hand' (Ruhlen 1994e). Within Nuclear Chibchan we find Cuna po-kwa; in the Misumalpan group, Miskito wal `2,' walwal `4,' Matagalpa buyo, Cacaopera burru, Ulua bo, Sumo bu, Twahka bo, and Panamaka bu; in the Talamanca group, Chiripo bor, Terraba kra-bu `2' (cf. kra ra`1,' kra-mia `3'), Bribri bul-bur, Estrella bor, Cabecar bol-bur; and in the Guaymi group, Murire-bu, Muoi-bu, and Move-bu. In the Barbacoan group, within the Paezan branch of Amerind, one may point to Cara pala, Colorado palu-(ga), `2,' um-balu-(ga) `4,' Cayapa palju, Awa pas, Totoro pa, and Guambiano pa. In the Andean branch, in addition to the forms cited by Greenberg, there is Aymara pa `2,' pa-tunka `20' (-tunka `10' is a borrowing from Quechua. In the Quechuan family we find Quechua (Wanka) pula `both' and Quechua (Ancash) pullan `half.' There are also two grammatical suffixes in Quechua which appear to be grammaticalizations of the root under discussion: Proto-Quechua *-pa, a repetitive suffix on verbs that shows the same semantic development as the Totonac suffix mentioned above, and *pas, a nominal enclitic meaning `too, also' (Parker, 1969, pp. 136, 143). Despite this abundance of evidence for the numeral root itself in South American languages, examples of the root with the numeral prefix are scant. The only seemingly clear example that has come to my attention is a series of forms in several closely related languages of the Mataco branch of Macro-Panoan: Chunupi ne-pu, Churupi nape-nepu, and Chulupi napu. The similarity of these forms to the Proto- Sahaptian reconstruction (*Nepu) is more than striking. Another possibly isolated retention, within the Arawakan branch of Equatorial, is Baniva enaba.

PROTO-AMERIND `3' A numeral prefix that occurs with `1' and `2' is likely to occur with even higher numerals, as this particular one does with Proto-Algonquian numerals up to `8.' Can we detect the Proto-Amerind word for `3' in extant Amerind languages? Greenberg did not propose any pan-Amerind numeral `3,' but he suggested forms for three Amerind subgroups, one of which I would like to explore here. Greenberg identified, for the Andean branch of Amerind found along the western coast of South America, a root for `3' whose general shape was *kala: Jebero kala, Cahuapana kala kara, Mapudungu kyla, Hongote calas, Tehuelche kaas. Far to the north, in western North America, there is evidence of the same root in the Almosan branch of Amerind. In the Chemakuan family, we find Chemakum qwali and Quileute qwa?le; in Kutenai we have ga_??_sa-;and for the Salish family , Kuipers (1982, p. 83) has reconstructed Proto-Salish *ka?_??_as,with reflexes such as Columbian ka?_??_as,Colville ka?_??_is,Shuswap ke_??_es,Thompson ke?_??_es,Lillooet ke_??_es, Sechelt ca_??_as,Coeur d'Alene ci_??_es,Kalispel ce?_??_es,and Upper Chehalis ca?_??_i. 218 M. RUHLEN

Similar forms are attested in all three branches of the Algic branch of Mosan, though in this case the cognacy of the forms becomes apparent only with a knowledge of Algic and Algonquian comparative . Indeed, the usual Proto-Algonquian reconstruction for the number `3,' *ne?_??_wi,does not look particularly promising. However, Picard (1986) shows by means of that this form must be presumed to derive from an earlier Pre- Algonquian form *ne-{p,k}V_??_-wi (V represents a vowel of unknown quality). Furthermore, while the choice of p or k as the first consonant of the root is indeterminate based on the Algonquian evidence, Proulx (1984, p. 181) shows that Algonquian's two closest relatives, Wiyot and Yurok, point unambiguously to velar k as the historically correct choice and Proulx reconstructs Proto-Algic *ni-khl (or *ni-khr) for the number `3.' The velar consonant receives further confirmation when one moves beyond Algic, as we have seen above, and the identificaton of the root finalconsonant as l also becomes clear. In sum, the Proto-Amerind word for `3' is preserved in six of the seven major branches of Almosan: Kutenai, Chemakuan, Salish, Wiyot, Yurok, and Algonquian. Only in Wakashan has it been replaced by a family-specific innovation. Furthermore, the numeral prefix is preserved with the root in all three branches of Algic. Outside of Almosan and Andean, possible reflexes of Proto-Amerind *ne-qwalas `3' are scarce. In North America, Yuchi no-ka `3' (cf, no-we `2') seems to preserve both the numeral prefix and the root for `3.' Other possibilities in North America include Tsimshian gul (in Penutian), Esselen xulep (in Hokan; but note also xulax '2') , and Cuitlatec kali_??_i(in Chibchan). In South America, Opaie nyukwari (in Macro-Ge) appears to have preserved both the Amerind root and numeral prefix. In Ge proper, certain languages also seem to have preserved both the numeral prefix and the root for '3': Canela inkre, Craho nkri, and Meuren enkri. Cayuvava, an Equatorial language, may retain the root in kulapa kurapa.

PROTO-AMERIND `4' In addition to the forms listed above involving the Proto-Amerind root *pale `2,' there are additional forms in which this root is associated with the numeral `4.' This may happen in several ways. One common typological path-reduplication- explains the Miskito forms cited above, wal `2' and walwal `4,' as well as Catawba parapari `4' (cf. napari `2'), and, in the Tucanoan family in South America, Yebamasa babari `4.' A different typological path involves the use of a reflexive (or reciprocal) prefix with the number '2'; accordingly, '{REFLEXIVE}-2'=4.In the Paezan branch we find such an example in Cayapa pallu `2,' taa-pallu `4.' However, this same formation -that is, both the reflexive prefix and the numeral-is also found in California Penutian, in forms such as the Kern Lake dialect of Yokuts: puni `2' and to-pani Proto-AmerindNumerals 219

`4.' (In California Penutian the change pal->pan-has affected most languages; see further discussion on this point below.) In Oregon Penutian, Tfalati taph `4' and Kalapuya taope `4' presumably display the same pattern as in Yokuts, though in these cases the number `2' is different so that there is no synchronic motivation for analyzing these roots into prefix and stem. They are simply fossilized relics identificable only by the broader Amerind context. The most northerly Penutian language, Tsimshian, constitutes a separate branch of Penutian by itself. In Tsimshian the numeral `4' is txaa-lpx. DeLancey, Genetti, and Rude (1988, p. 202) regard the second element as an "old root for TWO,"which they equate with Nez Perce lep- and Sahaptin nap-, discussed in the section on Proto- Amerind `2.' (In both of these languages the numeral `4' is also formed from the numeral `2,' though not with the reflexive prefix here being considered: Nez Perce pi-leep-t `4' and Sahaptin pi-nap-t `4.') I would like to suggest that the first element in the Tsimshian numeral `4,' txaa-, is the same reflexive prefix seen in such California Penutian forms as Kerns Lake Yokuts to-pani `4.' In South America, in the Puinave branch of Macro-Tucanoan, Hupda da?ap `4' may represent a similar fossilized relic, as may Quechua tawa in the Andean branch. Yet a third family that appears to retain the paradigm *pale `2,' *ta-pale `4,' is the Siouan family. Recall that the numeral `2' has been reconstructed as Proto- Siouan *nupa, which I have argued consists of the Amerind numeral prefix and the Amerind root for `2.' The Proto-Siouan numeral `4' has been reconstructed as *topa, which is directly reflected in Biloxi topa.) The broader Amerind context strongly suggests that this form is to be analyzed as *to-pa '(reflexive)-2,' parallel to the Penutian and Paezan cases. Wichita, a Caddoan language, apparently uses the same reflexive prefix to form the numeral `4,' though with a different numeral root: witsha `2 ,' ta-kwitsha `4'; the lack of initial k- in the word for `2' is due to the fact that initial kw- clusters are not allowed in Wichita. There are also a few Oto-Manguean languages, within the Central Amerind branch, that appear to have preserved the reflexive prefix in the number `4' as relics, just as the Oregon Penutian languages, e.g. Miahuatlan Zapotec thapa, Isthmus Zapotec tapa, Mitla Zapotec tahp. Finally, it is possible that certain Yuman languages show the Amerind root for `2' in the number `4.' In some dialects of Yokuts, a language belonging to California Penutian, an additional prefix has become attached to the numeral `4.' While Tulamni to-pani `4' shows a single prefix, Chauchila hotoponoi `4' shows two prefixes: ho-to-ponoi. This latter form is remarkably like the numeral `4' in various Yuman languages of the Hokan branch, e.g., Walapai hopa?, Yavapai hopa, Paipai xopa-k. Margaret Langdon and Pamela Munro (1980, p. 123) reconstruct Proto-Pai *xu-pa `4' and suggest that the first element is a contraction of the Proto-Yuman *xwak `2,' while the second element "is a shortened version of Mojave/Pai *pay 220 M. RUHLEN

`all ,' or, in this case, `both.'" Given the similarity to the Yokuts forms it seems to me more likely that the second element is simply a relic of Proto-Amerind *pale- `2.'

PROTO-AMERIND `5' I have found no general Amerind term for the numeral `5.' In various languages the Proto-Amerind word for `hand,' *makan, has become the numeral `5,' e.g. Maidu (California Penutian) ma-wyk'y (literally, 'hand-one'), Mixe (Mexican Penutian) magosk, Tequistlatec (Hokan) amake?, Classical (Uto-Aztecan) makwil-li (literally, 'hand-one'), Colorado (Paezan) manta, and Muellama (Paezan) mojo-ba (literally, 'hand-one'). These sporadic occurences seem to me, however, convergent developments to be explained by the world-wide tendency for the word for `hand' to be used as the numeral `5,' just as the word for `finger' often becomes the numeral `1' (cf. Karok (Hokan) tik `finger' and Mangue (Oto-Manguean) tike `1') and the word for `person' becomes the numeral `20' (e.g. Wnntun wintun `person ,' kete wintun [literally, `one person'] '20').

OUTGROUP COMPARISON IN One of the prized methods of resolving taxonomic questions in biology is by appeal to the evidence provided by the nearest kin, or immediate outgroup. In historical linguistics, however, the information provided by outgroups has been underutilized-or even ignored-because twentieth-century historical linguists long ago decided that there were no outgroups, just hundreds of obvious, but distinct, families with no hint of affinity among them. This view has been treated as received dogma throughout most of this century- and is still widely defended-but it now appears to be teetering on the verge of collapse (Ruhlen, 1994a,b). While the investigation of Proto-Amerind is a worthy goal in and of itself (if we are interested in understanding the prehistory of the human species), its investigation also offers rewards for specialists who concern themselves with families whose validity was established decades, if not centuries, ago. In each such family, whether it is Uto-Aztecan, Algonquian, or Penutian, there remain unresolved questions- or ever questions that have been resolved incorrectly. What the larger Amerind context offers is the possibility to resolve unanswered questions, and in some cases to correct erroneous conclusions. Let us first consider an unresolved problem within the Uto-Aztecan family. In their discussion of Southern Paiute numerals, Pamela Bunte and Robert Franklin (1988, p. 16) analyze the word for `5' as ma-nuxi-y 'hand-?-nominative.' The first part, ma-, is obviously the Uto-Aztecan (and Amerind) word for `hand' and the final -y is an Uto-Aztecan marker of the nominative case , but "the remainder of the stem is opaque." In other words, the meaning of this element, -nuxi-, and, of course, its origin, cannot be determined on the basis of evidence internal to the Uto-Aztecan Proto-Amerind Numerals 221

family. Similar constructions occur widely in the Numic branch of Uto-Aztecan;

additional examples include Mono ma-nigi `5,' Shoshone ma-naigi-ti `5,' Ute ma

nigi-ni `5,' Northern Paiute ma-nigi `5,' Kawaiisu ma-nigi-yu `5.' On the basis of

such forms Irvine Davis (1966) reconstructed Proto-Numic *m-n-k `5' (he does not

treat the vowels). Although Uto-Aztecanists seem agreed that the first element of

this numeral is simply the Uto-Aztecan word for `hand,' *ma- (Miller, 1967, p. 69),

the second element has remained unexplained. And yet within the broader Amerind

context outlined above it seems almost obvious that this element is simply a slightly

altered form of the Proto-Amerind word for `1,' *ne-k'we-, remarkably similar to

such South American forms of the numeral `1' as Iquito noki and Cahuapana nuki.

The broader Amerind context can clarify problems that are insurmountable for

specialists who resist looking beyond their family of specialization. Thus, there can

be little doubt that the literal meaning of the Southern Paiute word for `5' is `one

hand,' which contains of course five fingers.

The second example I would like to discuss involves another word for `5,' Proto-

Eastern Algonquian *palene_??_kwi. Frank Siebert (1975, p. 307) divides this form into two parts, pale-ne_??_kwi. The second part, -ne_??_kwi, is alleged to be an analogical

development of the Proto-Algonquian dependent noun *ne_??_ki `hand'; the first part,

Proto-Eastern Algonquian *pale' is, however, "of uncertain meaning," though

Siebert later suggests (p. 311) that perhaps the meaning is `full,' so that the

etymological origin of Proto-Eastern Algonquian `5' is `full hand.'

I would like to suggest that the true origin of Proto-Eastern Algonquian `5' is

less obscure- and less complex-than Siebert envisaged. First, the second part

of the numeral, -ne_??_kwi, can hardly be other than a slightly deformed variant of the

Proto-Algonquian word for `3,' *ne?_??_wi. Specifically, it would appear that between

the stage called Pre-Proto-Algonquian by Picard (1986), at a time when the numeral `3' was pronounced *ne -ka_??_-wi, and the later Proto-Algonquian stage, by which time

the numeral had come to be pronounced the numeral for `5' had

undergone an idiosyncratic metathesis (*-k_??_>*-_??_k) before k and p were reduced

to ? when preceding a full consonant, as in Proto-Algonquian *ne-?_??_-wi.

Second, the initial portion of the root is simply the Proto-Amerind word for `2,' *pale , which has survived in only a few archaic expressions in Algonquian and thus is not recognized by Algonquianists. The true etymology of Proto-Eastern Algonquian `5' is thus `2 -3 ,' not the imaginary `full hand' postulated by Siebert. Another archaic remnant of Proto-Amerind *pale `2' in Algonquian may occur in Arapaho (NawaĮ) niabala? `4' (=reciprocal nia-+bala? `2'), in which the Proto-Amerind reciprocal prefix *na- is used to derive higher numerals from lower monomorphemic ones. Reflexes of this same Proto-Amerind reciprocal prefix are also well attested in Uto-

Aztecan, as Sapir noted early in this century, for example, Southern Paiute pai- `3,' na-pai `6'; Hopi loyo-m `2,' na-loyo-m `4'; Tarahumara na-wo `4' (cf. Proto-Uto- 222 M. RUHLEN

Aztecan *wo '2'). In South America this same reciprocal may be seen in Chacobo cani-na-ka `they converse together' (literally, `speak-REcIPRocAL-they'). The Algonquian numeral system offers another example of how out-group comparison can clarify family-internal problems. Proto-Algonquian formed the numerals for `6, 7, 8' by adding the suffix *-waasika to the numeral roots for `1, 2, 3' (e.g. *ne-kwet-wi '1,' *ne-kwet-waasika '6'). I have shown (Ruhlen, 19940 that this numeral suffix-whose origin and meaning has never been satisfactorily explained-is simply an archaic word for `left hand,' strikingly similar to Zuni wesikk?a `left side,' and it can hardly be accidental that the Proto-Algonquians counted the numerals `1-5' on their right hand and the numerals `6-10' on their left. Furthermore, comparative evidence from the entire Amerind family indicates that the Zuni word and the Proto-Algonquian numeral suffix were originally compounds consisting of a Proto-Amerind word for `hand,' *w-asi, and a Proto- Amerind word for `left (side),' *gets' (Ruhlen, 1995). Yet a fourth example of the relevance of outgroup comparison is found in the Penutian family. In fact, it is found in the very name of the Penutian family, which was coined by Dixon and Kroeber (1913) on the basis of the word for `2,' pen in some languages, uti in others. In the languages characterized by pen there is fluctuation in the final consonant between-n and-l. Since-n is more frequent in California Penutian than-l, the family is called Pen-utian rather than Pal-utian. But the fact that one variant is more common than another does not guarantee historical precedence. In the present case, when we place the two California Penutian variants in a wider Penutian context, then in an even broader Amerind context, and finally, in a global context, we find that virtually all of the outgroups show final-l in this root. The sound change -l>-n is properly viewed as an innovation of part of California Penutian. Let us take a look at the forms, beginning with several dialects of the Wintun language that exhibit both variants. Dixon and Kroeber (1907) described the numeral systems of four dialects of Wintun, Northern, Central, Southern, and Cache Creek. In Northern Wintun the numeral `2' is pale-t, in Central Wintun, pale-l, in Southern Wintun, pampa-ta, and in Cache Creek (which is very similar to Southern Wintun), pampu-ta. (In Southern Wintun and Cache Creek, the formative -ta occurs at the end of all numerals from `1' to `11,' and is thus not part of the numeral itself.) In other branches of California Penutian we find stem-final-n or _??_:Proto- Maiduan *pen `2' (Golla, 1980, p. 62) and Proto-Yokuts *pu_??_oy`2' (Golla, 1980, p. 62; from earlier *pan-w(i)y according to Silverstein, 1975, p. 373). The Proto- Yukuts reconstruction has reflexes such as Wikchamni po_??_oy,Hometwoli pu_??_i(y), Choynimni punoy, Yawelmani ponoy, and Lower San Joaquin podoy. Given the four attested variants of the final consonant within California Penutian -l , n, _??_,m-how can one decide which variant is original? The answer is simple: Proto-AmerindNumerals 223 an examination of the relevant outgroups all show stem-final-l-in Penutian forms outside California Penutian (e.g. Wappo p'ala `twins'), in other branches of Amerind (e.g. Colorado palu- '2'), and even in other language families from Australia (e.g. Proto-Australian *pula `2') to Africa (e.g. Nimbari bala `2'). All of these outgroups indicate that the sound change -l>-n is an innovation within California Penutian and hence -l represents the original consonant, despite the fact that most Penutianists believe that -n was original (Silverstein, 1975; Pitkin, 1985, p. 410). As for the -m attested in Cache Creek and Southern Wintun, this is a secondary change from an earlier -n through assimilation to the following labial consonant (*panpan>pampa). This hypothetical intermediate form, *pan, is in fact attested in the Cache Creek word for `8,' pan-emus-ta (cf. emus-ta `4'). Similarly, the Yokuts forms with-_??_are the result of an assimilation between n and w in certain dialects; other dialects, as we have seen, preserve the original -n. We should note that Silverstein (1975, pp. 371-72) and Pitkin (1980) have proposed that Northern Wintun pale-t derives from an earlier (and unattested) *pan le-t, where -le- is a dual suffix. There are several problems with such an analysis. First, within Wintun there is really no evidence that pale- is, or ever has been, bimorphemic, and Silverstein admits as much when he adds that "the root never occurs without the second morpheme." But the real problem with Silverstein's analysis is that all of the outgroups strongly indicate that the root pale is- and always has been-a single, indissoluble formative with no internal structure. Silverstein's hypothesis would entail that Proto-Amerind *pale became Proto- California Penutian *pane- and then changed back to the original Amerind form, *pale , in some Wintun dialects, by a complex morphological process for which there is no synchronic support. Occam's razor suggests rather that Wintun pale- derives from Proto-Amerind *pale without change.

CONCLUSION The language of the first immigrants to the Americas-Proto-Amerind- appears to have used a numeral prefix in counting. The first three numerals were *ne-k'we `1,' *ne-pale `2,' and *ne-gwa_??_as`3.' A fourth numeral, *ta-pale `4,' seems to have combined a reflexive prefix, *ta-, with the root for `2,' *pale.

ACKNOWLEDGEMENTS I would like to thank Sheila Embleton, Joseph Greenberg, Dell Hymes, William H. Jacobsen, M. Dale Kinkade, Marc Picard, Paul Proulx, Frank Siebert, Michael Silverstein, and the participants of William S.-Y. Wang's seminar on language evolution at the University of California, Berkeley, for their useful comments and criticisms of an earlier version of this paper, not all of which I have heeded. 224 M. RUHLEN

REFERENCES Aoki, H. (1962) Nez Perce and Northern Sahaptin. Internat. J. Am. Ling. 28, 172-182. Aoki, H. (1963) On Sahaptin-Klamath linguistic affiliations. Internat. J. Am. Ling. 29, 107-112. Aoki, H. (1975) The east plateau linguistic diffusion area. Internat. J. Am. Ling. 41, 183-199. Bengtson, J.D., and Ruhlen, M. (1994) Global etymologies. In On the Origin of Languages: Studies in Linguistic Taxonomy (Ruhlen, M.), Stanford Univ. Press, Stanford, pp. 277-336. Bloomfield, L. (1946) Algonquian. In Linguistic Structures of Native America (Hoijer, H. et al., ed.), Viking Fund Publications in , No. 6, New York, pp. 85-129. Bunte, P.A., and Franklin, R.J. (1988) San Juan Southern Paiute numerals and mathematics. In In Honor of Mary Haas (Shipley, W., ed.), Mouton de Gruyter, Berlin, pp. 15-36. Davis, I. (1966) Numic consonantal correspondences. Internat. J. Am. Ling. 32, 124-140. DeLancey, S., Genetti, C., and Rude, N. (1988) Some Sahaptin-Klamath-Tsimshianic lexical sets. In In Honor of Mary Haas (Shipley, W., ed.), Mouton de Gruyter, Berlin, pp. 195-224. Dixon, R.B., and Kroeber, A.L. (1907) Numeral systems of the languages of California. Am. Anthropol. 9, 663-690. Dixon, R.B., and Kroeber, AL. (1913) New linguistic families in California. Am. Anthropol. 15, 647- 655. Giacone, A. (1949) Os Tucanos e outran tribus do Rio Uaup_??_safluente do Negro-Amazonas: notas etnogr_??_ficase folcl_??_ricas, Sao Paulo. Golla, V. (1980) Some Yokuts-Maiduan comparisons. In American Indian and Indoeuropean Studies (Klar, K., Langdon, M., and Silver, S., ed.), Mouton, The Hague, pp. 57-65. Greenberg, J.H. (1987) Language in the Americas, Stanford Univ. Press, Stanford. Grenand, F. (1989) Dictonnaire wayapi francais, Peeters, Paris. Gursky, K.-H. (1968) Gulf and Hokan-Subtiaban: New lexical parallels. Internat. J. Am. Ling. 34, 21- 41. Haas, M.R. (1965) Is Kutenai related to Algonkian? Canadian J. Ling. 10, 77-92. Jacobsen, W.H. (1979) Gender and personification in Washo. J. of California and Great Basin Anthropology: Papers in Linguistics 1, 75-84. Klein, R.G. (1989) The Human Career: Human Biological and Cultural Origins, Univ. of Press, Chicago. Kuipers, A.H. (1970) Towards a Salish . 26, 46-72. Kuipers, A.H. (1982) Towards a Salish etymological dictionary II. Lingua 57, 71-92. Kaufman, T. (1988) A research program for reconstructing Proto-Hokan: First gropings. In Papers from the 1988 Hokan-Penutian Languages Workshop (DeLancey, S., ed.), Univ. of Oregon, Eugene, pp. 50-168. Langdon, M., and Munro, P. (1980) Yuman numerals. In American Indian and Indoeuropean Studies (Klar, K., Langdon, M., and Silver, S., ed.), Mouton, The Hague, pp. 121-135. Lincoln, N.J., and Rath, J.C. (1980) North Wakashan Comparative Root List, National Museums of Canada, Ottawa. Miller, W.R. (1967) Uto-Aztecan Cognate Sets, Univ. of California Press, Berkeley. Munro, P. (1994) Gulf and Yuki-Gulf. Anthropol. Ling. 36, 125-222. Parker, G.J. (1969) Comparative Quechua phonology and grammar II: Proto-Quechua phonology and . Univ. of Hawaii Working Papers in Linguistics 1.2, 123-147. Payne, D.L. (1991) A classification of Maipuran (Arawakan) languages based on shared lexical retentions. In Handbook of , Vol. 3 (Derbyshire, D.C., and Pullum, G.K., ed.), Mouton de Gruyter, Berlin, pp. 355-499. Proto-Amerind Numerals 225

Picard, M. (1986) On the structure of the lower numbers in Pre-PA. Internat. J. Am. Ling. 52, 72- 77. Pitkin, H. (1980) Two plus two makes two. In American Indian and Indoeuropean Studies (Klar, K., Langdon, M., and Silver, S., ed.), Mouton, The Hague, pp. 199-208. Pitkin, H. (1985) Wintu Dictionary, Univ, of California Publications in Linguistics, Vol. 95, Berkeley. Proulx, P. (1984) Proto-Algic I: Phonological sketch. Internat. J. Am. Ling. 50, 165-207. Rensch, C. (1976) Comparative Otomanguean Phonology, Indiana Univ., Bloomington. Ruhlen, M. (1994a) On the Origin of Languages: Studies in Linguistic Taxonomy,Stanford Univ. Press, Stanford. Ruhlen, M. (1994b) The : Tracing the Evolution of the , John , New York. Ruhlen, M. (1994c) Is Algonquian Amerind? In On the Origin of Languages: Studies in Linguistic Taxonomy (Ruhlen, M., ed.), Stanford Univ. Press, Stanford, pp. 111-126. Ruhlen, M. (1994d) Amerind T' A?NA `child, sibling'. In On the Origin of Languages: Studies in Linguistic Taxonomy, Stanford Univ. Press, Stanford, pp. 183-206. Ruhlen, M. (1994e) Proto-Amerind *KAPA `finger, hand' and its origin in the Old World. (forthcoming) Ruhlen, M. (19940 A note on Proto-Algonquian numerals. (forthcoming) Ruhlen, M. (1995) Proto-Amerind *QETS' `left (hand)' Mother Tongue 24, 69-70. Sapir, E. (1913) Wiyot and Yurok, Algonkin languages of California. Am. Anthropol. 15, 617-646. Sapir, E. (1925) The Hokan affinity of Subtiaba in Nicaragua. Am. Anthropol. 27, 402-435. Siebert, F.T., Jr. (1975) Resurrecting Virginia Algonquian from the dead: The reconstituted and historical phonology of Powhatan. In Studies in Southeastern Indian Languages (Crawford, J. M., ed.), Univ. of Georgia Press, Athens, pp. 285-453. Silverstein, M. (1975) On two California Penutian roots for TWO, Internat. J. Am. Ling. 41, 369- 380. Swadesh, M. (1946) Phonologic formulas for Atakapa-Chitimacha. Internat. J. Am. Ling. 12, 113- 132. Swadesh, M. (1949) The linguistic approach to Salish prehistory. In Indians of the Urban Northwest (Smith, M. W., ed.), New York. Swadesh, M. (1953) Mosan II: Comparative vocabulary. Internat. J. Am. Ling. 19, 223-236. Wolff, H. (1950) Comparative Siouan II. Internat. J. Am. Ling. 16, 113-121.