Adelaar Willem, Muysken Pieter. The languages of the Andes

Подождите немного. Документ загружается.

34 1 Introduction

Lule (K65) and Vilela (K66); Chikitano (K70), Bor´oroan (K71), Je (K74), Guat´o (K82)

and nine other Brazilian groups; Kunsa (K99) and Kapishan´a.

In connection with other classiﬁcations, we commented en passant upon some of these

suggested groupings (Lule and Vilela, Panoan and Tacanan, Chibchan and Misumalpa,

Aymara and Quechua). Arguments for a comprehensive Macro-Ge grouping including

Ge, Guat´o and Bororoan, as well as several other language groups, can be found in

Davis (1968) and in Rodrigues (1986, 1999), but do not, as yet, extend to Chiquitano.

The proposal of a special genetic relationship between Esmeralda and Yaruro was ad-

vanced by Seler (1902); see section 2.19. Loukotka located them in the same subgroup

of Chibchan. Doris Payne (1984) has presented evidence for a genetic relationship of

Yaguan (K34) and Zaparoan (K33). Callahuaya (Kaufman’s Kolyawaya) is a profes-

sional jargon composed of roots taken from a Puquina dialect and Quechua endings (see

section 3.5). The unity of the extinct Sechura and the equally extinct Tall´an languages

of the Piura area (K45 Katak´aoan) was proposed by Rivet (1949) but the evidence for

it considered inconclusive by Torero (Torero 1986); see section 3.9.2. The possibility

of a genetic relationship between the Boran and Huitotoan languages, on one hand, and

Andoque, on the other, was considered unconvincing by a leading expert on this lan-

guage (Landaburu 1979). As for the proposed special relationship between the Kunza or

Atacame˜no language and the Brazilian KanoˆeorKapishana, the geographic and cultural

barriers seem formidable, and there would need to be a very strong linguistic case to

support it (see also section 3.7).

1.7.2 Quechuan and Aymaran, Quechumaran

The two dominant language groups of the central Andes, Quechuan and Aymaran,

must be viewed as language families rather than as single languages. Traditionally,

however, Quechuan is more often referred to as the Quechua language. The internal

comparison of Quechuan, a linguistic entity consisting of numerous local varieties,

became an issue in the 1960s when Parker (1963) and Torero (1964) published their well-

known articles about the Quechua dialect situation (see section 3.2.3). Hardman (1975,

1978a, b) introduced the name Jaqi (‘man’, ‘human being’) for the Aymaran family,

which, according to her, has three living members to be treated as separate languages:

Aymara, Cauqui and Jaqaru. For a discussion of the terminology and a justiﬁcation of

our use of the term Aymaran see section 3.1.

Quechua(n) and Aymara(n) have repeatedly been compared to each other, but rarely

to other languages. Harrington (1943) suggested a relationship between Quechua and

Hokan; Dum´ezil (1954, 1955) compared some of the Quechua numerals to those of

Turkish. Following an unconvincing ﬁrst attempt by Swadesh (1967), Liedtke collected

a list of lexical and grammatical resemblances between Quechua and Tarascan, some

1.7 Genetic relations of South American Indian languages 35

of which are quite suggestive (Liedtke 1996). No thorough comparative study has been

carried out, however.

Orr and Longacre (1968) set out to prove Mason’s Quechumaran hypothesis by trying

to reconstruct the phoneme system and part of the lexicon of the proto-language un-

derlying it. Although they apparently achieved their aim, the lexicon they reconstructed

consists almost exclusively of shared vocabulary, which is evidently due to intensive

borrowing between the two languages at an early stage of their development. Given

the virtually identical form of the shared items, the radically different character of the

remainder of the lexicon is left unexplained. The same holds for the grammatical com-

ponents of the two language groups, which show quite a few semantic but hardly any

formal similarities (Davidson 1977). Notwithstanding the lack of proof, the idea of a

Quechumaran genetic unity exclusive of all other languages still has supporters. For

an attempt to revive the Quechumaran hypothesis on a more sophisticated basis see

Campbell (1995).

The relationship between the Quechuan and Aymaran linguistic families is indeed

unique. When the effects of loan trafﬁc between individual Quechua dialects and the

different languages of the Aymaran family are left aside, a substantial basis of common

lexicon remains (about 20 per cent of the root vocabulary in each group), which can be

traced back to the proto-languages. The phoneme inventories of the two proto-languages

were probably very similar, as most of the existing differences may be explained by

later internal developments in each of the two families. The existence of glottalised

and aspirated consonants in Aymaran and in a number of Quechua dialects (Cuzco,

Puno, Arequipa, north and south Bolivian Quechua) is generally attributed to diffusion

from Aymaran into Quechuan, although its distribution within the latter group is far

from predictable (see section 3.2.5). Morphological and lexico-semantic coincidences

are highly speciﬁc and difﬁcult to ascribe to parallel developments of a typological

nature (see chapter 3 for more details). For a systematic inventory and discussion of

all the coincidences see also Cerr´on-Palomino (1994a). The obvious similarities that

have united Quechuan and Aymaran since the stage of the proto-languages stand in

contrast with differences that are equally impressive. The very characteristic phonotac-

tics and vowel suppression rules of all Aymaran languages (see section 3.3.4) are not

found in Quechuan. The structure of the verbal inﬂection, personal reference marking

in particular, differs considerably between the two language groups, and, of course, a

major part of the lexicon and afﬁxes do not show any systematic formal relationship

at all.

All this leads to the conclusion that Proto-Quechua and Proto-Aymaran were spoken in

contiguous areas, if not in the same area, which were probably situated in central Peru, the

heartland of the Middle Andean civilisation. The bi-directional loan inﬂuence between

the two linguistic families was so intense, that possible surviving correspondences of a

36 1 Introduction

genetic kind became hard to detect. If the languages were not genetically related – and

there is no decisive evidence that they were – at the least one of them must have suffered

a profound structural transformation adopting the phonological and morphosyntactic

model represented by the other. This scenario presupposes a period of intense interaction

and common development prior to the stage of the proto-languages. It may have begun

well before the beginning of our era. Although it is risky to venture a statement on

such a speculative matter, a variety of Aymaran would be the best candidate for having

provided such a model because of the more homogeneous character of Aymaran verbal

and nominal inﬂection in comparison to Quechuan inﬂection.

The remaining languages of the central Andean region do not participate in the same

sort of lexical and grammatical entwining that characterises the relationship between

Quechuan and Aymaran, although lexical borrowing has occurred. Since these languages

have been poorly studied so far, further research may eventually cast additional light on

their relationship with either Quechuan, or Aymaran (or both).

1.7.3 Other proposals for individual language families

As we anticipated, there have been many proposals of genetic connections between

speciﬁc groups which were formulated outside the framework of an overall classiﬁcation.

For the earlier period (before 1960), two scholars, Rivet and Jij´on y Caama˜no, deserve

to be specially commended for the size of their contribution to South American Indian

linguistics, including much classiﬁcatory work. Many of their classiﬁcatory proposals

have been the subject of drastic reconsideration. Therefore, it is not necessary to treat

them in detail here, but the amount of data they brought together and their inﬂuence have

been considerable. Among Rivet’s classiﬁcatory contributions we ﬁnd the proposed

connection of Arawakan and Tacanan (see above), the inclusion of Uru and Puquina

within the Arawakan family (see also above), a rearrangement of the Chibchan family

involving many groups in Ecuador and southern Colombia (see also above), and the

association of the isolated Yurumangu´ı language of the Colombian Paciﬁc coast with

Sapir’s Hokan phylum (Rivet 1942). Well known was also Rivet’s conviction that the

Chon languages (Tehuelche and Ona) of Patagonia were genetically related to languages

spoken by the Australian aborigines (Rivet 1925). Rivet’s comparative methods have

met with much criticism. In the case of Yurumangu´ı, for instance, he compared the

vocabulary of this poorly documented extinct language with that of a wide array of

North American and Mesoamerican languages of supposed Hokan afﬁliation. A lexical

similarity between Yurumangu´ı and any of these languages would be considered evidence

of a genetic relationship.

In his monumental El Ecuador Interandino y Occidental antes de la Conquista Espa-

˜nola (Inter-Andean and Western Ecuador before the Spanish Conquest, 1940–5), Jij´on y

Caama˜no assigned most languages of northwestern South America to a Macro-Chibchan

1.7 Genetic relations of South American Indian languages 37

phylum, which, in its turn, would ﬁt into Hokan–Siouan. His Macro-Chibchan was

more comprehensive than any of the previous proposals concerning Chibchan and its

connections. Jij´on y Caama˜no’s Macro-Chibchan not only included all the languages in

Loukotka’s Chibcha, but also Timote, Cof´an, Murato (Candoshi), Yurumangu´ı, Mochica

(Chim´u), Cholona and the Central American Lenca, Xinca, Jicaque and Subtiaba.

By contrast, Tucano and Huitoto–Bora–Z´aparo are listed as separate phyla. Jij´on y

Caama˜no’s interpretation of the comparative method has been much criticised, inter alia,

for its acceptance of systematic equations of phonetically unrelated sounds. Constenla

Uma˜na (1981) mentions some striking examples of this procedure. Nevertheless, both

Rivet and Jij´on y Caama˜no must be credited with having brought to public attention a

wealth of data on many extinct and poorly documented languages, which until then had

been virtually unknown.

Recent investigations of the Chibchan family have tended to reduce the number of

languages associated with it. In his thorough phonological reconstruction of Proto-

Chibchan, Constenla Uma˜na (1981) found that the Barbacoan, Paezan, Andaqu´ı, Kams´a,

Betoi, Jirajaran and Misumalpa languages are not Chibchan. What is left is a family

based primarily in Central America and represented in Colombia and Venezuela by the

Cundinamarcan Chibchan languages Muisca and Duit, Tunebo, the Arhuacan languages

of the Sierra Nevada de Santa Marta, Bar´ı, Chimila and Cuna. In a revised version of

his reconstruction, Constenla Uma˜na (1989) proposes a Paya–Chibchan family con-

sisting of a Paya branch (represented by the sole Paya language of Honduras) and a

Chibchan branch. The Chibchan branch comprises several subgroups. One of them is a

Colombian Chibchan group which comprises the Arhuacan, Tunebo and Cundinamarcan

Chibchan languages. Cuna is found to belong to a different subgroup with the extinct

Dorasquean languages of Panama. Chimila and Bar´ı remain unclassiﬁed as to subgroup

for lack of data. Several poorly documented languages once spoken in the Colombian

department of Antioquia (Nutabe, Cat´ıo Chibcha) are also classiﬁed as Chibchan. The

linguistic evidence seems to point to a relatively recent arrival of the Chibchan people

from Central America, making it less likely for all proposed South American connec-

tions to be correct. The alleged genetic relationship of Chibchan with Warao (in the

Venezuelan Orinoco delta) and with Yanomama (in the Brazilian–Venezuelan border

lands; Greenberg 1959, 1960a, b; Migliazza 1978a) has been the object of an investi-

gation by Weisshar (1982). Among many other similar proposals, we may mention that

of L´evi-Strauss (1948), who suggested a genetic relationship between Chibchan and the

Brazilian Nambikwara languages (refuted in Constenla Uma˜na 1981).

The reduced Chibchan family, such as proposed by Constenla Uma˜na, is almost the

same as that originally outlined by Uhle (1890). The afﬁnities of the different mem-

bers of the family thus being reconsidered, many languages previously classiﬁed as

Chibchan are again left unclassiﬁed. There is convincing evidence that the Barbacoan

38 1 Introduction

languages Cayapa, Colorado and Cuaiquer are related in a family which also includes

the Coconucan languages Guambiano and Totor´o (Constenla Uma˜na 1991; Curnow and

Liddicoat 1998). Whether the extinct languages of the northern Ecuadorian highlands

and the adjacent highlands in the Colombian department of Nari˜no (Cara, Pasto) also

belonged to the same grouping is a question which deserves further investigation (see

section 3.9.1). Although Greenberg and Kaufman classify Guambiano as Paezan, the

distance between Guambiano and P´aez seems to be greater than that between Guambiano

and the Barbacoan languages. The position of P´aez, Andaqu´ı, Kams´a and Betoi requires

renewed attention.

The Cariban language family, which has it greatest concentration of speakers in the

Guyanas and eastern Venezuela, is represented in the Colombian–Venezuelan border

mountains, west of Lake Maracaibo, with the Yukpa or Motilones group. Rivet (1943a)

assumed a more generalised presence of Carib-speaking peoples in the Colombian

Andes, by assigning the (extinct) Muzo, Colima, Panche, Pijao, Pant´agora and Op´on–

Carare languages of the Magdalena basin to the same family. He also believed the

Chocoan languages of the Colombian Paciﬁc area to be related to Cariban (Rivet 1943b).

Durbin and Seijas (1973a, b, 1975) have shown that only the Op´on and Carare languages,

located to the northwest of the Cundinamarcan highlands, were demonstrably Cariban.

Of the other languages Muzo and Colima may have been Cariban as well. For the three

remaining languages, however, the lexical similarity with Cariban is not such that it

can provide the assumed relationship with a solid basis (see section 2.11). The alleged

connection of Chocoan and Cariban has been superseded by Greenberg’s proposal relat-

ing the Chocoan languages to Paezan. For similarities between Chocoan and Barbacoan

see section 2.3. Similarly, Constenla and Margery (1991) have published evidence for a

relation between Chocoan and Chibchan.

Genetic connections have been sought between the Mochica language (also known

as Yunga and, erroneously, as Chim´u) and Chibchan (Jij´on y Caama˜no, Greenberg),

and also with the Mayan language family in Mesoamerica (Stark 1972a, 1978). Another

language that has been associated with Mayan is Mapuche (Stark 1970, Hamp 1971).

Olson (1964, 1965) has proposed a genetic relationship between Mayan and the Chipaya

language of the Bolivian altiplano (closely related to Uru). As this theory became widely

accepted – Longacre (1968: 320) considers it proven – Uru–Chipaya came to be included,

along with Mayan, into the North America-based Macro-Penutian phylum in published

classiﬁcations of the North American and Mesoamerican languages (e.g. Voegelin and

Vo e gelin 1965). Campbell (1973) later showed that many of the similarities observed by

Olson between Chipaya and Mayan could have been the result of contact with Quechua or

Aymara. Very little is left of the arguments that seemed to have convinced the Americanist

linguistic community for some time. A serious drawback is the lack of a good grammar

1.7 Genetic relations of South American Indian languages 39

and dictionary of Uru and Chipaya, which makes veriﬁcation a difﬁcult task for the

non-initiated.

Family-internal reconstruction work was carried out for Panoan by Shell (1965, 1975),

and for Tacanan by Key (1968) and by Girard (1971). Another complex of proposals

concerns the connection of Panoan and Tacanan with the Bolivian Moset´en language

(Su´arez 1969) and the relationship of both groups to Uru–Chipaya, to Yuracar´e (also in

Bolivia) and to the Chon languages of Patagonia and Tierra del Fuego (Su´arez 1973,

following Swadesh 1962 in the latter two suggestions). Key (1978) compared Pano–

Tacanan and Moset´en to Araucanian.

The extensive Arawakan or Maipuran family has been the subject of much comparative

and classiﬁcatory work (Shafer 1959, Kingsley Noble 1965, Matteson 1972, Tovar 1986,

Valenti 1986, Payne 1991a). Kingsley Noble includes the Uru and Puquina groups and

Araw´ainhis comparison. Arawakan and Araw´a share rather speciﬁc features of their

gender systems but are quite far apart lexically. Matteson includes Madi (Araw´a) and

Harakmbut, a procedure which is rejected by Tovar. David L. Payne (1991a) does not

include Araw´a and Harakmbut within the Arawakan family. Harakmbut has now been

shown to be related to the Brazilian Katukina family with a suggested further connection

to Macro-Ge (Adelaar 2000).

Although Arawakan as a whole constitutes a closely-knit family with strong lexical

resemblances (Rodrigues 1986: 70), the supposed afﬁliation between Arawakan and a

number of languages in the Andean region has met with reserve. Such languages are

Amuesha (recognised as Arawakan by Tello in 1913), Apolista (shown to be Arawakan

by David L. Payne 1991b), Chamicuro (Parker 1987) and Res´ıgaro. Although Amuesha

(spoken in the Andean foothills of the Peruvian department of Pasco) seemed at the

best a highly divergent member of the Arawakan family, Wise (1976) convincingly

demonstrated that Amuesha is closely related to pre-Andean Arawakan groups, such

as Campa, Machiguenga and Piro, the relationship being obscured by rather unusual

phonetic changes that took place in Amuesha. These changes must have occurred recently

because they have also affected loan words from Spanish. The classiﬁcatory status of

Res´ıgaro is discussed by Allin (1975) and by David L. Payne (1985); see also Aikhenvald

(2001).

The fact that Uru–Chipaya and Puquina are by no means closely related invalidates

their inclusion into Arawakan as a subgroup, a hypothesis which is nevertheless defended

by de Cr´equi-Montfort and Rivet (1925–7), Kingsley Noble (1965) and Greenberg

(1987). It does not, however, preclude the possibility that one of the two languages,

i.e. Puquina, exhibits a remote Arawakan afﬁnity. There are, in fact, similarities in the

lexicon and, above all, in the pronominal system (cf. section 3.5). The ongoing inves-

tigation of the internal relations within the Arawakan family, which once spread over

40 1 Introduction

much of South America and the Caribbean, holds great promise for the unravelling of

the continent’s linguistic puzzle.

David L. Payne (1990) has pointed at some very striking similarities concerning the

formation of possessive nouns in four South American families, Arawakan, Araw´a,

Cariban and Candoshi, which are difﬁcult to explain through borrowing. On the other

hand, Rodrigues (1985a) has presented well-documented lexical evidence relating the

Cariban family to the Tupi stock (tronco Tup´ı ), a huge genetic construct attaining its

maximum differentiation in the Brazilian Madeira basin. Lexical similarities between

Tupi and Macro-Ge (tronco Macro-Jˆe)were already noticed by Davis (1968, 1985) and

conﬁrmed in Rodrigues (1985b). The suggested genetic relationship of Tupi, Cariban

and Macro-Ge is supported by typological similarities (a relatively loose morphological

structure and a lack of polysynthesis), as observed by Doris Payne (1990b); see also

Rodrigues (2000).

David L. Payne (1981) investigated the alleged relationship of the Jivaroan languages

and Candoshi (proposed in both Greenberg’s classiﬁcations). Although he found lexical

similarities, these lay in the sphere of ﬂora and fauna and seemed to point at bor-

rowing. David L. Payne (1990: 84–5) no longer considers the evidence for a J´ıvaro–

Candoshi grouping convincing, but he mentions some grammatical similarities between

Arawakan, Candoshi and Cariban. In Su´arez’s classiﬁcation (1974), Jivaroan is linked to

Cahuapanan, another small language family of the northern Peruvian foothills. Kaufman

(1990) considers this a possible relationship. Among several other suggestions, Kaufman

(1994: 63) offers an interesting new proposal concerning a genetic relationship between

Candoshi, Omurano (Mayna), and Taushiro (all in the Peruvian Amazon).

Anew proposal concerning a possible genetic relationship between two linguistic

groups that had never been associated before has been made by Croese and Payne

(Croese 1990). They observed rather striking lexical similarities between Araucanian

(Mapuche) and the Arawakan family. The matter requires further investigation.

For the Panoan languages Migliazza (1978a, mentioned in Migliazza 1985) proposes

a rather close genetic relationship with Yanomaman, based on a cognate number of about

40 per cent, and a more remote relationship with Chibchan (see above).

The Guaicuruan language family (including Toba as its principal representative in

the pre-Andean space) and the Matacoan language family have, together with a third

group, Lengua–Mascoy, their centre of gravity in the Gran Chaco. Tovar (1981) found

considerable lexical resemblance between Toba and the Matacoan languages. Whether

this is due to a common genetic origin or borrowing is an issue awaiting further in-

vestigation. Tovar also observes lexical similarities between Matacoan and Arawakan.

In Greenberg’s classiﬁcation Matacoan, Guaicuruan and Lengua–Mascoy are taken

together. For Charr´uan, a group of extinct languages once located in Uruguay and

1.7 Genetic relations of South American Indian languages 41

Argentinian Entre R´ıos, Arawakan, Matacoan, Lule–Vilelan and Guaicuruan connec-

tions have been proposed by Perea y Alonso (1937), Ferrario (Ms), Rona (1964) and

Su´arez (1974), respectively. See for the three ﬁrst proposals the discussion in Longacre

(1968: 353–4), who seems to give most credit to Ferrario’s arguments for the Matacoan

connection, cited in Censabella (1999: 61). Suˇsnik (1978: 94) appears to favour a

Guaicuruan connection.

Among the languages of the southernmost part of South America, a grouping

was recognised as early as 1913 by Lehmann-Nitsche. It consisted of the Patagonian

languages (Tehuelche, Tehues) and the languages of Tierra del Fuego’s main island

(Selk



nam or Ona, Haush). This grouping was called Tshon or Chon,adenomination

that includes elements of the words ‘Tehuelche’ and ‘Ona’. For many languages once

spoken in Argentina it will probably never be possible to even approximately determine

a genetic afﬁliation because the populations in question were exterminated before their

languages could be recorded. Viegas Barros has proposed a genetic relationship between

Kawesqar and Yahgan (Censabella 1999: 88).

A ﬁnal word about possible trans-Paciﬁc genetic connections. Although there have

been many proponents of such connections (Rivet 1925; Imbelloni 1928; Ibarra Grasso

1958), no valid arguments were brought forward to support them. The search for them,

however, has shown at the least two lexical items shared by Polynesian languages and

languages in South America. One of them is the name of a plant domesticated in the

New World, the sweet potato (Ipomoea batatas), Easter Island kumara,Hawaiian

uala,

which is found as k’umar or k’umara in Quechua and Aymara. The second word is toki,

Easter Island ‘stone axe’, Mapuche ‘stone axe’, ‘military chief (the holder of the axe)’;

compare also Yurumangu´ı totoki ‘axe’ (Jij´on y Caama˜no 1945). Although the former

case constitutes near proof of incidental contact between inhabitants of the Andean

region and the South Paciﬁc, the latter is not nearly as convincing but certainly deserves

attention. Apparently, there were sporadic contacts that led to an occasional interchange

of words, not to migrations of entire populations that could have brought along their

languages.

1.7.4 The Greenberg (1987) proposal

The appearance of Greenberg (1987) brought the discussion about the origins of lan-

guage, both in the Andes and elsewhere in the New World, into a new phase. Rather than

merely containing the long expected factual justiﬁcation of Greenberg’s earlier proposal,

which, in part, it did, the terrain of comparison was widened to include the whole of

native America. It also brought a revision of the classiﬁcation of the South American

Indian languages proposed before. In short, Greenberg (1987) contains the following

new elements:

42 1 Introduction

A. All the languages of the New World belong to three families: Eskimo–Aleut, Na–

Dene and Amerind; Eskimo–Aleut and Na–Dene are limited to the Arctic and parts of

North America. Consequently, all South American and Mesoamerican Indian languages,

as well as most North American Indian languages are related. They belong to a single

family: Amerind. The tripartite division of the native American languages is associated

with three consecutive waves of migration, the ﬁrst of which is represented by speakers

of Amerind. Support for this hypothesis is sought from physical anthropology (blood

groups, dental structure) and archaeology; a ﬁrst outline of it had already been published,

before the appearance of Greenberg (1987), in Greenberg, Turner and Zegura (1985,

1986).

B. The South American languages are divided into seven subgroups: Macro-Ge,

Macro-Panoan, Macro-Carib, Equatorial, Macro-Tucanoan, Andean and Chibchan–

Paezan. As in Greenberg’s earlier classiﬁcation, one language, Yurumangu´ıisassigned

to a North American subgroup, Hokan. In addition, Macro-Ge, Macro-Panoan and

Macro-Carib are said to form a group at an indermediate level, which corresponds

to the Ge–Pano–Carib of Greenberg’s earlier classiﬁcation; the same holds for Equa-

torial and Macro-Tucanoan, which formed part of Andean–Equatorial in the earlier

classiﬁcation. Andean–Equatorial as such is abandoned. So there are four groups at

the intermediate level between Amerind and the seven subgroups just enumerated: Ge–

Pano–Carib, Andean, Equatorial–Tucanoan and Chibchan–Paezan (the former Macro-

Chibchan). For Amerind as a whole, including North America and Mexico, Greenberg

posits eleven subgroups and six groups at the intermediate level. As Swadesh (1962)

did before him, Greenberg ﬁnds more genetic diversity in South America than in

North America (except for the presence of non-Amerind Na–Dene and Eskimo–

Aleut).

C. Chibchan–Paezan receives extensions in North America and elsewhere in the

Americas. Its Chibchan division is made to include Tarascan and Cuitlatec, two lan-

guage isolates located in Mexico; the Paezan division now includes Timucua, a lan-

guage isolate once to be found in Florida. Huarpean (Allentiac), originally classiﬁed

as Ge–Pano–Carib, has been reassigned to Paezan, where it ﬁnds itself together with

Atacame˜no. J´ıvaro–Candoshi and the languages associated with it (Esmeralda, Yaruro

and Cof´an), originally a separate division of Andean–Equatorial, have been reclassi-

ﬁed as Equatorial. Quechua and Aymara are both classiﬁed as Andean but no longer

treated as a unit. Greenberg (1987: 100) admits ‘that Aymara appears relatively isolated

within Andean’.

Greenberg’s new classiﬁcation comes at a point in time when historical linguists tend

to be increasingly reluctant to accept distant genetic relationships if not accompanied

by solid proof (see, in particular, Campbell and Mithun 1979). Greenberg shows him-

self highly critical of the current methods of obtaining such proof, ‘the use of sound

1.7 Genetic relations of South American Indian languages 43

correspondence tables and asterisked reconstructed forms’ (1987: 1). Instead, Greenberg

advocates the search for lexical and grammatical similarities that become apparent from

a comparison of many languages at the same time (‘multilateral comparison’). He does

not pursue phonetic exactitude and considers it premature to look for regular sound cor-

respondence lest other signiﬁcant similarities should be missed. Greenberg also allows

for a substantial amount of error in his data: ‘the method of multilateral comparison

is so powerful that it will give reliable results even with the poorest of materials. In-

correct material should have merely a randomizing effect’ (1987: 29). As a matter

of fact, the data included in Greenberg (1987) are riddled with errors. For instance,

Cochabamba Quechua ‘to see’ is given as ruk, instead of rik

u-, and quechuologists

are puzzled about identifying the ‘Huanacucho dialect’ (probably a designation for

Ayacucho affected by a confusion with the name of some other dialect, such as Hu´anuco

or Huanca). Old misunderstandings are perpetuated and even reinforced. For instance,

the alleged unity of Uru–Chipaya and Puquina is not only defended, but Olson’s work is

quoted as evidence for it (Greenberg 1987: 84). In reality, Olson (1964) merely observed

that some Chipaya call their language Puquina, which says nothing about the undeni-

able fact that there also existed a Puquina language quite distinct from present-day

Chipaya. Identical namegiving is no proof of identity. On the other hand, possible close

connections, such as Guambiano and Barbacoan, or Harakmbut and Katukina, were

missed.

It is not surprising that Greenberg’s work has met with vigorous criticism; see, for

instance, the discussion in Current Anthropology 28: 647–67, Kaufman (1990), and the

reviews by Adelaar (1989) and Matisoff (1990). Nevertheless, not all his proposals should

be dismissed lightly. Some of the proposed genetic links will undoubtedly turn out valid,

even though the factual basis is still insufﬁcient. Greenberg also gives an inventory of

grammatical elements that are widespread in the Amerindian languages. Of some cases

Greenberg was not the ﬁrst to have noticed them (see, for instance, Swadesh 1954).

The grammatical elements in question are not merely cases of typological resemblances

because they concern the formal aspects of morphemes. One of the best-known cases

is a pattern consisting of n,oranother non-labial nasal, for reference to ﬁrst person, in

combination with m for reference to second (both usually followed by a vowel). One

may be tempted, for instance, to investigate the possibility of a genetic link between,

say, Araucanian and Californian Penutian on the basis of Araucanian n

i ‘my’ and mi

‘your’, on the one hand, and Wintu ni ‘I’ and mi ‘you’, on the other, only to ﬁnd out later

that many other Amerindian languages exhibit similar or related patterns of personal

reference. Whatever the origin of such resemblances may be, they can hardly be due to

borrowing.

In table 1.8, we summarise the classiﬁcation proposed in Greenberg (1987) insofar as

it concerns the languages and language groups located in the Andean region. Alternative