Haas on historical linguistics

27
Trustees of Indiana University Anthropological Linguistics Mary R. Haas and Historical Linguistics Author(s): Lyle Campbell Source: Anthropological Linguistics, Vol. 39, No. 4 (Winter, 1997), pp. 642-667 Published by: The Trustees of Indiana University on behalf of Anthropological Linguistics Stable URL: http://www.jstor.org/stable/30028503 Accessed: 24/02/2010 17:53 Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/action/showPublisher?publisherCode=tiu. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected]. Trustees of Indiana University and Anthropological Linguistics are collaborating with JSTOR to digitize, preserve and extend access to Anthropological Linguistics. http://www.jstor.org

Transcript of Haas on historical linguistics

Trustees of Indiana University

Anthropological Linguistics

Mary R. Haas and Historical LinguisticsAuthor(s): Lyle CampbellSource: Anthropological Linguistics, Vol. 39, No. 4 (Winter, 1997), pp. 642-667Published by: The Trustees of Indiana University on behalf of Anthropological LinguisticsStable URL: http://www.jstor.org/stable/30028503Accessed: 24/02/2010 17:53

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available athttp://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unlessyou have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and youmay use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained athttp://www.jstor.org/action/showPublisher?publisherCode=tiu.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printedpage of such transmission.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range ofcontent in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new formsof scholarship. For more information about JSTOR, please contact [email protected].

Trustees of Indiana University and Anthropological Linguistics are collaborating with JSTOR to digitize,preserve and extend access to Anthropological Linguistics.

http://www.jstor.org

Mary R. Haas and Historical Linguistics

LYLE CAMPBELL

University of Canterbury, New Zealand

1. Introduction. The aim of this paper, as set out by the editor, is to assess Mary R. Haas's contribution to historical linguistics, while conveying insights into her role based on my personal experiences with her. This personal sort of assessment is appropriate, since I can report her influence on my understanding of historical linguistics confidently, but I am much less certain about what her impact on others may be. That is, I know what I think is important in her work and what I think others should get out of it, but I am less certain of the extent to which others have taken these lessons to heart. Therefore, I point out here what I think is important in her historical linguistic work, hoping that this reflects accurately her role in historical linguistic thought in general. To the extent that some of these lessons found in her work may be significant but not yet generally recognized, I hope that this paper will serve to bring these contributions the recognition they deserve.

Mary contributed to the following areas of historical linguistics: areal lin- guistics (and borrowing in general), historiography of linguistics (and philology), distant genetic relationships, and family-level historical linguistics of several North American Indian language families (especially Muskogean and Algon- quian).' Each of these is considered in turn.

2. Areal linguistics and borrowing. At a meeting in 1976, a few of us were talking informally, and the topic came up of why so many scholars seemed to stumble into unjustified conclusions about genetic relationships. Mary's opinion was that their biggest stumbling block was from unrecognized borrowings and the effects of language contact. Mary's teaching and writings in this area have been very influential. Edward Sapir's approach to the historical study of Amer- ican Indian languages, with its emphasis on genetic relationships, had won the following of most students of Native American languages, but Franz Boas's conception of areal linguistic traits and diffusion was instrumental in the development of the notion of Sprachbund or linguistic area (particularly in the Prague School; see Jakobson 1944), which subsequently made its way back to America to become important in contemporary American Indian linguistic studies through the influence of Murray B. Emeneau and Mary (see Haas 1969b:82-97, 1970a, 1976b, 1977:123-25). With Emeneau's (1956) paper, "India as a Linguistic Area," American attention was called back to areal linguistics (neglected since Boas's time), and Mary's and Emeneau's publications and

642

1997 LYLE CAMPBELL 643

teaching at the University of California at Berkeley brought areal linguistics to the attention of a whole generation of students of Native American languages and, through them, made it one of the most significant areas of contemporary historical linguistic research (see Campbell 1997:330-52).

Areal linguistics deals with the diffusion of structural features across lin- guistic boundaries. A linguistic area is usually characterized by a number of linguistic features (thought to be due to diffusion or to convergent development) that are shared by various languages (either unrelated or from different sub- groups of a language family) in a geographically continuous area. As Mary's work shows, areal linguistics is very important to the historical study of lan- guages: the true goal of historical linguistic investigations should be to find out what really happened, that is, to discover the real history-be it genetic or contact-lying behind traits shared by different languages.

By now [1977] we are ready to agree that Boas was right when he perceived the distribution of areal traits as a historical process. In other words, history does not reside in genetic relationship alone but is equally important in areal distributions. Now, half a century after Boas's insistence on this point and twenty years after Emeneau's [1956] article appeared, one of the most exciting fields in anthropological linguistics is areal linguistics. In fact genetic and areal studies are no longer considered antithetical, but are frequently carried on and discussed side by side. [Haas 1978124-25; see also Bright 1976]

Clearly, as Mary's comment in our conversation mentioned above indicates, in attempts to establish more remote genetic relationships (that is, inclusive language families), the investigation of areal linguistic phenomena plays a key role. Attempts must be made to distinguish similarities among compared lan- guages that are due to contact and diffusion from those that may be inherited from an earlier common ancestor. Mary was right-too often proposals of remote linguistic kinship fail to be convincing precisely because the effects of borrowing and areal linguistics have not been adequately dealt with. In the comparative method, similarities shared among related languages are the basis for postu- lating ancestral forms in the protolanguage. However, areal borrowings exhibit similarities that, if undetected, are sometimes assumed to result from common inheritance and are erroneously reconstructed as features of the parent lan- guage. For example, instrumental prefixes and switch reference were recon- structed erroneously in some Native American language families. These features are borrowed in some western North American families (see Sapir [cited in Golla 1984:105-6, 108]; Jacobsen 1983). Sapir had originally enter- tained the idea that perhaps Yurok (and Wiyot) were related to Salishan, based on what he took to be shared numeral classifiers, but later gave up this idea when it became clear that numeral classifiers were more widespread and

probably involved areal diffusion (Golla 1984:105-6, 108).

644 ANTHROPOLOGICAL LINGUISTICS 39 No. 4

2.1. The Northwest Coast linguistic area. Let us consider a specific case, that of the Northwest Coast linguistic area (NWCLA) and how it interacts with the evidence for the "Mosan hypothesis" of a distant genetic relationship be- tween the Salishan, Wakashan, and Chimakuan language families (in the interest of space, I concentrate on the morphosyntactic traits and evidence). The NWCLA is probably the best known of North American linguistic areas. It includes Eyak, Tlingit, the Athabaskan languages of the region, Haida, Tsim- shian, Wakashan, Chimakuan, Salishan, Alsea, Coosan, Kalapuyan, Takelma, and Lower Chinook. Representatives of twelve of Powell's (1891) fifty-eight families are found here, and this represents more linguistic diversity than any other well-defined linguistic area in North America--or most other areas in the world. The NWCLA is renowned for the linguistic complexity, both phonological and morphological, exhibited by its languages.

Typical shared morphological traits are: reduplication processes signaling various grammatical functions (e.g., iteration, continuative, progressive, plural, collective, etc.); numeral classifiers; alienable-inalienable oppositions in nouns; pronominal plural; nominal plural; verbal reduplication signifying distribution, repetition, etc.; suffixation of tense-aspect markers in verbs; verbal evidential markers; locative-directional markers in the verb; masculine-feminine gender distinction; visibility-invisibility opposition in demonstratives; and nominal and verbal reduplication signaling the diminutive. Verbal aspect is relatively more important than tense. All but Tlingit have passive-like constructions. The nega- tive appears as the first element in a clause regardless of the usual word order. Northwest Coast languages also have lexically paired distinct singular and plur- al verb stems (Thompson and Kinkade 1990:42-44). "Lexical suffixes" are found in, at least, the Wakashan and Salishan languages. (Lexical suffixes designate such familiar objects, normally signaled with full lexical roots in other lan- guages, as body parts, geographical features, cultural artifacts, and certain ab- stract notions.) Wakashan, for example, has some 300 of these lexical suffixes (Thompson and Kinkade 1990:40; Kinkade, Elmendorf, Rigsby, and Aoki in press). In these languages, there is a severely limited role for the contrast between nouns and verbs as distinct categories (cf. Thompson and Kinkade 1990:33).

2.2. Mosan. The "Mosan hypothesis," as mentioned above, proposes a genetic connection between the Salishan, Wakashan, and Chimakuan language fami- lies. Sapir (1929) accepted Mosan as a genetic grouping (part of his more inclu- sive Algonquian-Mosan), and Swadesh (1953a, 1953b) attempted to provide sup- porting evidence. Swadesh (1953a:29-30) lists sixteen shared structural similarities, but, as noted below, most are either general NWCLA traits or typologically expected and unremarkable:

1. "Extensive use of suffixes." (This is a NWCLA trait, and, at the same time, it is typologically unremarkable, being typical of many languages in the Amer-

1997 LYLE CAMPBELL 645

icas, Asia, and elsewhere, which are not assumed to be genetically related.) 2. "Nearly complete absence of functioning prefixes in Chimakuan and Waka-

shan, minor role as compared with the suffixes in Salish." (Again, this is in the list of areal traits characteristic of Northwest Coast languages in gen- eral, and it is not unusual typologically. Moreover, numbers 1 and 2 are not independent of each other, since, from a typological viewpoint, many suf- fixing languages lack or have few prefixes, and prefixing languages may lack or have few suffixes.)

3. "Extensive use of stem reduplication, including initial reduplication..,. and full stem reduplication." (Once again, this trait is typical of the linguistic area and reflects the well-known reduplication processes signaling various grammatical functions. There are other places in the world where extensive reduplication is found, including other parts of the Americas.)

4. "Changes in stem vowel, including lengthening and i-mutation." (Ablaut and vowel mutations are found in certain other languages of the linguistic area as well as elsewhere-as noted by Sapir when he associated Atha- baskan languages with Tibetan and in his typological characterization of Penutian languages. Such vowel alternations (ablauts, umlauts, and mor- phological patterns signaled by different vowels in the root) are seen as markers of identity for Indo-European, Semitic, Jicaque, "Penutian," and others. They are not typologically unusual.)

5. "Insertion of glottal stop .... Glottal insertion (or glottalization of sonants) is often connected with reduplication or lengthening and has in part the same functions, expressing distribution, continuation and diminutive." (From Swadesh's examples, it appears that he means that glottal stop has certain grammatical functions, but this cannot be particularly significant, since almost any consonant can signal some morpheme in some language somewhere. Finding it in connection with reduplication and such functions as "distributive," "continuation," "plural," "iterative," "collective," and the like is not limited to the putative Mosan languages. For example, the most productive plural marker on nouns in Nahuatl is reduplication of the first CV with glottal stop inserted [e.g., ko?-ko:yotl 'coyotes'; cf. ko:yotl 'coyote'], and Nahuatl verbs reduplicate in the same way-with inserted glottal stop-to signal repetition or intensity [e.g., sa?-saka 'to haul many things or transport an object to or from various places or for various people to carry an object'; cf. saka 'to haul'].)

6. "Interchange of stop and spirant, of unlabialized and labialized, and of front and back k-sounds." (Swadesh's examples appear to be, on the whole, non-

productive "vestigial evidence" of old morphophonemic alternations. They are not diagnostic; rather, all are found, for example, in languages of the Uto-Aztecan family, spread across several linguistic areas.)

7. "Aspect, including at least the dichotomy of momentaneous and durative."

(This is a feature of the NWCLA. Such aspect distinctions could also easily

646 ANTHROPOLOGICAL LINGUISTICS 39 No. 4

evolve independently in unrelated languages.) 8. "Tense is an optional category." (That aspect is much more important than

tense in Northwest Coast languages has been mentioned as an areal trait; this is also true for many, many languages throughout the Americas and elsewhere in the world.)

9. "Distributive plural is an optional category. This is very different from the European kind of plurality." (Again, distributive plural is common in the NWCLA, but with restrictions. In fact, in many Native American languages, plurality is signaled formally only on human nouns, and even then often optionally. Identification of plurality often depends, rather, on context or on the morphology of the verb.)

10. "Dichotomy of nonfeminine versus feminine gender shown in demonstra- tives and articles." (Once again, the masculine-feminine gender distinction, while uncommon in Native American languages, is a feature of the NWCLA.)

11. "Numeral classifier notions, shown by suffixes." (Numeral classifiers, too, are a trait of the NWCLA. Moreover, they are found among several lan- guages of the Plateau area and in many other American Indian languages. That they should be grammaticalized as suffixes in suffixing languages is not unexpected.)

12. "Two alternate stems for number." (Northwest Coast (and Plateau) lan- guages have lexically paired distinct singular and plural verb stems as a shared areal trait.)

13. "Local [lexical] suffixes (sometimes called field suffixes), referring to body parts and other space references." (The interpretation of this trait may still be open. While some see it as potential evidence of a genetic relationship, I believe that it has been borrowed among those languages of the area that have it.)

14. "Use of demonstratives or articles to substantivize verbs." (I do not know whether this feature has a broader areal distribution, but it could easily be the result of independent parallel grammaticalizations in these and in other languages that have it. This trait is known in many other languages of the world.)

15. "Predicative use of nouns, with personal [i.e., pronominal] predicative af- fixes added directly. For example, 'I am chief' in Nootka and in Kalispel may be expressed by adding the appropriate personal affix to the word for chief." (This is not a particularly unusual grammatical feature in American Indian languages; for example, it is typical of the Mesoamerican linguistic area, as seen for example, in Nahuatl ni-tla:katl [I-man] 'I am a man' or Q'eqchi' kwinq-at [man-you] 'you are a man'. In fact, in the early history of Native American linguistics, examples such as these provoked the discussion of "incorporation" [Einverleibung] in the work of Duponceau, Humboldt, Pickering, and others [see Haas 1969a].)

1997 LYLE CAMPBELL 647

16. "Demonstrative distinctions such as the present versus absent, or visible versus invisible." (This trait, too, is a general areal feature in Northwest Coast languages. It is also found elsewhere in the Americas [e.g., K'iche' ri: 'the (not visible)' and ri? 'that (not visible)', versus le: 'the, that (visible)'], attesting that the contrast can easily develop independently in unrelated languages.)

The similarities (particularly the structural resemblances) that these lan- guages share suggest areal diffusion (Jacobsen 1979; Thompson 1979). Some of the traits are so common typologically that they could easily have developed independently in the languages that share them. The proposed Mosan grouping has no current support among American Indian linguists; subsequent research has called the Mosan classification into question, and it is now largely aban- doned. Not even Swadesh persisted in maintaining the Mosan hypothesis, since later he (Swadesh 1962) placed Wakashan (but not the other putative Mosan groups) with Eskimo-Aleut (and some Old World tongues).

As the Mosan example shows, Mary was right; too often hypotheses of distant genetic relationship fail to be convincing because they do not take into account the effects of borrowing and diffusion. In this case, many of Swadesh's examples presented as evidence of the Mosan hypothesis turn out to be traits of the NWCLA and hence are probably explicable as the results of diffusion.

2.3. Areal linguistics and subgrouping. Areal diffusion can have implica- tions not only for hypothesized remote relationships, but also for subgrouping and classification within known language families. Mary's (1969b) work on Nootkan provides a good example. The sound correspondences upon which Nootkan subgrouping is based are shown in table 1.

Table 1. Sound Correspondences in Nootkan Languages

MAKAH NITINAT NOOTKA PROTO-NOOTKAN

b b m *m b' b' b' *rn d d n *n d' d' n' *r

Sr p r *q,

qW f f , qW

xW xW h xW

x x h *x

Subgrouping is based on shared innovations. Nitinat and Makah appear to share the innovation that changed nasals to corresponding voiced stops, while

648 ANTHROPOLOGICAL LINGUISTICS 39 No. 4

Nitinat and Nootka appear to share the change of the glottalized uvulars to pharyngeals. (Makah and Nitinat also share the retention of uvular fricatives, which Nootka has changed to a pharyngeal, but shared retentions are not valid evidence for subgrouping.) That is, one innovation (denasalization) suggests a grouping of Makah with Nitinat, with Nootka standing apart, whereas the other innovation (pharyngealization) suggests Nitinat and Nootka as a grouping, with Makah less closely related. This impasse is solved when we take into account the fact that the absence of nasals is an areal feature, shared by several other languages of the area, that diffused into both Makah and Nitinat. The subarea of the Northwest that lacks primary nasals includes Twana and Lushootseed (Salishan), and Quileute (Chimakuan), as well as Nitinat and Makah (both Nootkan, of the broader Wakashan family)(Haas 1969b; Kinkade 1985; Thomp- son and Thompson 1972). Comox (Salishan) was also described as having b and d as optional variants of m and n, respectively, and a similar situation was observed in Sechelt, Clallam, and two dialects of Halkomelem (all three Salishan; Kinkade 1985:479). Boas (1911:565) observed that Lower Chinook pronunciation involved much confusion regarding "surds and sonants" on account of "semiclosure of the nose," and older records of several of the other languages reveal a similar situation (Kinkade 1985:478-79). Kinkade reports that, in fact, "in virtually every littoral language [at least twelve of them] of the Northwest from the 46th to the 50th parallel nasals were sometimes pronounced without full closure of the velum" (1985:478), and that, in recent times, many of the languages that had these sounds intermediate between nasals and voiced stops have settled in favor of one or the other of the sounds, eliminating the intermediate variant (Kinkade 1985:480).

Thus, the innovation shared by Nitinat and Nootka of glottalized uvulars changing to pharyngeals is real evidence of subgrouping: Nitinat-Nootka consti- tute one branch of the family, Makah the other branch. The change of nasals to corresponding nonnasal stops in Nitinat and Makah is due to diffusion and is therefore not evidence for subgrouping. Moreover, if we did not know about the areal diffusion, we might be tempted to reconstruct the voiced stops in Proto- Nootkan and postulate a change to nasals in Nootka. Thus, areal linguistic traits can have an impact on how we classify (subgroup) and on how we recon- struct (cf. Haas 1969b: 112).

2.4. Lexical borrowing. While lexical borrowing in North American Indian languages has still not received the attention needed, Mary's work in this area is a significant exception and should inspire future investigations (see Haas 1947b, 1968a, 1968b, 1969c:78-82).

3. History of linguistics. Mary called me her "grandson." By this, she just meant that I was Bill Bright's student and that he, in turn, was her student. She meant nothing particularly sentimental by this, but it did mean a lot to me.

1997 LYLE CAMPBELL 649

Mary's observation about where we all fit into the on-going history of linguistics (in my case, with the academic pedigree of Bright-Haas-Sapir-Boas) was char- acteristic of her interest in and keen sense of the historiography of linguistics, especially of American Indian linguistic research (see Haas 1966a, 1967c, 1969a, 1969c, 1969d, 1970b, 1973a, 1975, 1976a, 1977, and 1978). Mary cites Bloom- field's statement that Boas is "the teacher, in one or another sense, of us all" and adds that "it is fitting to make the same statement in regard to all three of these great men [Boas, Sapir, and Bloomfield] as teachers, in one or another sense, of all who are working in the field today" (Haas 1976a:69). Mary also is the teacher, in one sense or another, of a great many of us working in the field today, and much of the progress in American Indian linguistics from the 1950s to the present comes from her, her Berkeley students, and those whom they have influenced.

Mary's work in linguistic historiography predates the interest that has given rise to the volumes, journals, and societies dedicated to the topic that began to appear in the 1970s. Most notably, her publications on the history of scholarship on North American Indian languages have helped to illuminate on-going areas of investigation and to clarify persistent controversies. Certain of these involve methodology, and, in particular, the respective roles of grammar, sound corres- pondences, and vocabulary as evidence of genetic relationship (e.g., Haas 1969a).

3.1. Regularity and sound correspondences in Native American historical linguistics. To cite a specific instance, Haas (1967c:817) discussed the significance of the Algonquian sound correspondences first pointed out by Roger Williams and John Eliot and later rediscovered by John Pickering. In 1643, Roger Williams, in A Key into the Language of America, specified what, in effect, is an Algonquian sound correspondence involving n, 1, and r in several of the New England Algonquian languages (Haas 1967c:817). In 1666, John Eliot, another famous New England pioneer, observed the same correspondence: "We in Massachusetts pronounce N; the Nipmuck Indians pronounce L; and the Northern Indians pronounce R" (quoted in Haas 1967c:817). In 1833, John Pickering reported these sound correspondences anew with the observation:

An attention to these established differences is indispensable to a just com- parison of the various dialects [languages], and the useful application of such comparisons [is indispensable] to the purposes of philology; and it will enable us to detect affinities, where at first view there may be little or no appearance of any resemblance. [quoted in Haas 1967c:817]

As Mary pointed out, Pickering made it clear that he could extend such "substi- tutions" to other Algonquian languages and that this helped him to identify cognates:

650 ANTHROPOLOGICAL LINGUISTICS 39 No. 4

The letter [sound] R... is a characteristic of the Abnaki dialect; as, for instance in the words arem8s [aremos], a dog, in the Delaware, L is used, and they would accordingly say, n'dalemous, my dog; the n being the inseparable personal pronoun, here signifying my. In Abnaki, mirar8 [miraro] is the tongue; and in the Massachusetts dialect,--which takes N instead of R,-the same word becomes meenan [minan] ... The numeral five, which in Abnaki is barenesk8 [bareneskw], in the Delaware is palenach [palenax]... though at first view their resemblance is not obvious. [quoted in Haas 1967c:817]

What Mary's citations of Williams, Eliot, and, especially, Pickering show is that sound correspondences did indeed play an important role in the history of American Indian historical linguistics, in spite of claims to the contrary. Joseph H. Greenberg has contended that "the notion of sound correspondences as proving relationship" is not true of Indo-European and other families and that sound correspondences played no role in the recognition of Native American language families (Greenberg forthcoming; see also 1987:25-33, 1989, 1990, 1991). However, a number of linguists working with Native American languages utilized this criterion, and it played an important role in American Indian linguistics, particularly during the last third of the nineteenth century and the early twentieth century. For example, Hyacinthe de Charencey (1870) used sound correspondences to classify the languages of Mesoamerica, and his sub- grouping of Mayan languages (Charency 1872, 1883) included several corre- spondence sets and sound changes. Otto Stoll, too, presented a number of sound correspondences and associated sound changes among Mayan languages. His remarks are revealing: "These changes follow regular phonetic laws and bear a strong affinity to the principle of 'Lautverschiebung' (Grimm's law), long ago known as an agent of most extensive application in the morphology of the Indo- Germanic languages" (Stoll 1885:257). "When... it concerns... on which basis ... I proposed the diversification of the Mayan family..,. the following can here be mentioned... One of the most striking differences between the individual groups of Mayan languages is the regular sound shift from one group to the other [several examples of which are given]" (Stoll 1912-13:40). Raoul de la Grasserie (1890:438-39) listed sound correspondences among the criteria that argue against chance that he used to establish genetic relationship, and he utilized sound correspondences in his classifications of several Native American language families. Adrien Gabriel Morice (1891, 1892) established sound corre- spondences among several of the Athabaskan languages, comparing them explicitly to Indo-European in the regularity of their development. The list goes on (see Campbell 1997), including Sapir's and Bloomfield's famous proofs of the regularity of sound change in, and the applicability of the comparative method to unwritten, so-called exotic languages (see section 3.2). Pliny Earle Goddard leaves no doubt concerning the awareness by linguists working with American Indian languages of the importance of sound correspondences and of their role in establishing genetic relationships:

1997 LYLE CAMPBELL 651

Modern linguistic study is based on a belief in phonetic laws which produce uniform results under identical conditions. The one recognized method of establishing genetic relationship is to point out the uniform changes which in the course of time have caused the separation of a uniform linguistic area into dialects and related languages. The claim that regular sound change and sound correspondences played no role in American Indian historical linguistics is mistaken. [Goddard 1920:271]

3.2. Unwritten languages, sound change, and the comparative method. In the past, it was often believed that the comparative method could not be applied successfully to "unwritten" languages, that is, to "exotic" or so- called primitive languages, and even today some expressions of doubt are still encountered from time to time. More specifically, the regularity of sound change in unwritten or exotic languages was doubted by many, and the comparative method relies on regularity to be successful. One of the more significant con- tributions of American Indian linguistics has been its compelling demonstration that sound change in unwritten and so-called exotic languages is regular and that the comparative method can be validly applied to them. This demonstration is rightfully credited to Bloomfield and Sapir, and yet I find Mary's (1969c: 17-24) discussion of the matter more convincing and compelling, since she puts the matter in historical context and then exposes the flaws in the thinking of those who question regularity in unwritten languages (see Campbell 1994 for further discussion).

However, the mistaken assumptions about unwritten or exotic languages and the putative inapplicability of the comparative method have continued to come up, in spite of Mary's convincing rebuttals and Sapir's and Bloomfield's proofs. For example, Boretzky (1982, 1984) has made such claims, and these have been seconded by Miihlhilusler (1989). Boretzky contrasts Aranta (in Aus- tralia) and Kate (in New Guinea) with Slavic and Romance-that is, "exotic" with "European." Boretzky asserts that in exotic languages semantic slots are likely to be filled either by difficult to relate morphs, or, conversely, that the phonological differences are so small (in Arandic) that there is no scope for reconstruction. He thinks that change in the Arandic languages proceeds more by abrupt lexical replacement through borrowing than by gradual phonological change. However, if this were true, it still would not invalidate the comparative method. In particular, extensive borrowing does not invalidate the comparative method for these languages. Dixon shows that "it is quite clear that Australian languages change in a regular fashion, in the same way as Indo-European and other families" (1990:398). In fact, it was through a demonstration of regular changes that Hale (1964, 1976) was able to show that the languages of north- eastern Queensland with many short monosyllabic words, formerly thought to be quite aberrant and unrelated, in fact developed regularly from languages that originally were like other Pama-Nyungan languages in type. Vocabulary may present difficulties in Australia, but lexical borrowings are a fact of

652 ANTHROPOLOGICAL LINGUISTICS 39 No. 4

linguistic life that the comparative method must contend with everywhere, not just in Australia or New Guinea (cf. Johnson 1990:430). In the face of claims such as Boretzky's (1982, 1984), it may be appropriate to drive home Mary's points yet again.

Though in the minds of many, development of the comparative method is associated with written Indo-European languages, comparative linguistics also has a long association with "exotic," unwritten languages. In fact, the compar- ative method was applied to exotic, unwritten languages from its very earliest applications. In 1770, Janos Sajnovics attempted to see whether Hungarian, Lapp, and Finnish were related, and on an astronomical research trip to the Norwegian arctic, he elicited Lapp words and transcribed them in an orthog- raphy that he devised himself. On the basis of these field recordings from unwritten Lapp, he applied the comparative method and demonstrated that Hungarian, Lapp, and Finnish are related. He reasoned, however, that in order to convince skeptics it would be necessary to utilize previously published data. Thus, in his published report (Sajnovics 1770), he employed none of his own recorded Lapp data, where he had approached Lapp as an unwritten language, but selected examples from the only sources available, Knut Leem's textbook (1748) and lexical samples (1768-81)-though these were recorded in an inade- quate Danish orthography with Danish glosses, both of which were major obstacles to adequate comparison. Thus, the issue of data from unwritten lan- guages has been with us from the very beginning of the use of the comparative method. A brief look at language families to which the comparative method has been applied successfully shows that many of these families involve some languages with a written tradition, together with others that are largely unwritten. For example, Semitic includes languages with some of the earliest known written documents, together with Ethiopian relatives that have scarcely been recorded by linguists. Uralic, with the oldest tradition of comparative linguistics (cf. Sajnovics 1770; Gyarmathi 1799), includes Hungarian (with writ- ten attestations from the eighth century onward) and several other "written" languages (e.g., Finnish, Estonian, etc.), but also languages with no real tradi- tion of writing at all (e.g., Vogul, Ostyak, and certain Samoyed languages). Sapir's (1913-19) reconstruction of Uto-Aztecan, which demonstrated conclu- sively and for the first time the validity of this family, was one of the earliest rigorous demonstrations of the applicability of the comparative method to unwritten languages; it also relied on the assumption of the regularity of sound change. It, too, involved a mixture of written and unwritten languages, includ- ing Nahuatl, with abundant written colonial material, and Southern Paiute, only available from Sapir's own field notes. Reconstruction by means of the comparative method for families composed of both "written" and "unwritten" languages is common today. The existence of unwritten languages (that is, of languages with no long written tradition) is not an obstacle to comparative reconstruction.

1997 LYLE CAMPBELL 653

Nevertheless, as Mary points out, the mistaken perception of the role of "written" languages remains a problem:

Since the existence of written languages... was of great strategic importance in the development of our knowledge of Indo-European, some scholars came to believe that the historical and comparative study of languages was impossible without written records of earlier stages of the same or related languages. [Haas 1969c:20]

Although this mistaken perception of the role of writing in Indo-European linguistics persists today, it had already been challenged and abandoned by the first neogrammarians. In part, the prejudice in favor of old written traditions reflects a holdover from an earlier stage of comparative linguistics, where lan- guage change was thought to take place in discrete stages of progress and decay. The languages of so-called savage people were thought to be "primitive" relics that had not yet evolved-progressed through processes of compounding and agglutination-to the state of greater perfection that it was believed that the older written Indo-European languages, in particular Sanskrit, had attained.

Modern languages were typically viewed as merely decayed reflections-due to analogy and sound changes, which were assumed to be operative only in this later phase-of their more perfect ancestors. Therefore, the old written lan- guages, thought to be more perfect, were allotted a special status. By the time of the neogrammarian movement (cf. Osthoff and Brugmann 1878), comparative linguistics had adopted a uniformitarian position, where language change was no longer held to take place in discrete stages of either progress or decay, but where languages were seen to undergo the same kinds of changes at all times throughout their histories. With this reorientation, written language was accorded less of a special status, and attention turned more toward spoken language, in particular to dialects (cf. Osthoff 1883), and attention to dialec- tology promoted the development of phonetics-i.e., of techniques for recording forms of spoken language (cf. Sievers 1876). For example, Berthold Delbriick, writing of the principle that sound laws are without exception in his influential neogrammarian introduction to linguistics, affirmed that:

This natural constitution of language is not manifested in the cultivated tongues, but in the dialects of the people. The guiding principles for linguistic research should accordingly be deduced not from obsolete written languages of antiquity, but chiefly from the living popular dialects of the present day. [Delbriick 1974:61]

And, later in the same volume:

It is of far greater importance to collect further facts from living languages, in order to draw conclusions from them with regard to the ancient languages. [Delbriick 1974:126]

654 ANTHROPOLOGICAL LINGUISTICS 39 No. 4

This reorientation led linguists to view spoken language, not written, as the more valuable for linguistic research. By 1900, Henry Sweet was able to report:

It is now an axiom of scientific philology that the real life of language is in many respects more clearly seen and better studied in dialects and colloquial forms of speech than in highly developed literary languages. [Sweet 1900:791

In short, the persistent association of the comparative method with Indo- European languages with an old written tradition is misinformed, misguided, and obsolete.

Bloomfield (1925) is credited with disproving the presumed challenge to regularity from unwritten or exotic languages. His aim was to disprove defini- tively the assertion that comparative reconstruction could not be successful in the absence of written records of earlier stages of the language (Haas 1969c:22). Bloomfield's (1925, 1928) famous proof of the applicability of the comparative method to unwritten languages was based on the correspondence sets and reconstructions for Central Algonquian given in table 2.

Table 2. Correspondence Sets and Reconstructions for Central Algonquian

Fox OJIBWA PLAINS CREE IMENOMINI PCA

1 hk sk sk ck *ck 2 ik Ak sk sk *sk 3 hk hk sk hk *xk

4 hk hk hk hk *hk 5 sk sk hk hk *gk

NOTE: PCA = Proto-Central Algonquian

He postulated the reconstruction of *gk for set 5, as distinct from the other sets in table 2, on the basis of scant evidence-a single reconstructible morpheme- but under the assumption that sound change is regular and that the difference in this correspondence set (although exhibiting only sounds that occur in different combinations in the other sets) could not plausibly be explained in any other way. Later, the correctness of his decision to reconstruct something differ- ent for set 5 was confirmed when Swampy Cree was discovered to contain the correspondence htk in the morpheme upon which set 5 was based-distinct in Swampy Cree from the reflexes of the other four reconstructions. Bloomfield concluded from this:

As an assumption, however, the postulate [of sound-change without exception] yields, as a matter of mere routine, predictions which otherwise would be impossible. In other words, the statement that phonemes change (sound- changes have no exceptions) is a tested hypothesis: in so far as one may speak of such a thing, it is a proved truth. [Bloomfield 1928:100]

1997 LYLE CAMPBELL 655

Ironically, Bloomfield's Algonquian proof shows how writing not only can be overrated, but can sometimes be an obstacle to reconstruction. He employed mixed "written" and "unwritten" source materials. While he relied, in part, on records from missionaries and traders, he trusted what he considered to be scientific recordings, i.e., his own field records for Menomnini and Cree and the material written down by William Jones, a linguistically trained native speaker of Fox, for Fox and Ojibwa. Jones's recording of Ojibwa is important with regard to the point about written records. Since Fox does not contrast sk and sk, Jones (as a native speaker of Fox) failed to recognize and record this contrast in Ojibwa. Had Jones's recordings of Ojibwa (the written source Bloomfield chose to rely on solely) not failed to represent this contrast, Swampy Cree would not have been the only extant witness to the distinctness of set 5. Bloomfield distrusted the missionary and trader records as less accurate "written" repre- sentations-his stumbling block in this case:

The fuss and trouble behind my note in Language [Bloomfield 1928] would have been avoided if I had listened to O[jibwa], which plainly distinguishes sk (< PA 9k) from &k (< PA Ik); instead, I depended on printed records which failed to show the distinction. [Bloomfield 1946:88]

In this case, it is the materials upon which Bloomfield relied (those from Jones) that were an obstacle to reliable reconstruction by the comparative method- although Bloomfield seems to suggest in the quotation above that it was the later, recorded field data, the usual sources for "unwritten" languages, that

provided the correct solution. However, there is a lesson in Bloomfield being misled by distrust of the older written sources. In fact, older missionary sources on Ojibwa (e.g., Baraga 1878, 1878-80; Cuoq 1886) did correctly distinguish sk and ik. The obstacle to reliable reconstruction was Bloomfield's commitment to

modern field records.2 The moral is that written representations require inter-

pretation, irrespective of whether they are old or recent. Comparative recon- struction, which depends on records, whether old or new, can be no better than our ability to extract from the sources relevant interpretations of the phonology of the languages involved. That is, "written records are a means to an end, and there is no justification for holding them in high esteem, or even in reverence (as is sometimes the case) except as indirect evidence for what one is trying to discover" (Hockett 1970:502). And, as Edward Sapir has pointed out, "if these laws [regular phonetic changes] are more difficult to discover in primitive lan-

guages, this is not due to any special characteristic which these languages possess but merely to the inadequate technique of some who have tried to study them" (Sapir 1949:74).

As is well known, Bloomfield's (1925, 1928) demonstration of the regularity of sound change in unwritten "primitive" or "exotic" languages was in response to misgivings such as those expressed by Meillet and Cohen in their famous book Les langues du monde:

656 ANTHROPOLOGICAL LINGUISTICS 39 No. 4

One may well ask whether the languages of America (which are still for the most part poorly known and insufficiently studied from a comparative point of view) will ever lend themselves to exact, exhaustive comparative treatment; the samples offered so far hold scant promise . . . it is not even clear that the principle of genealogical classification applies. [Meillet and Cohen 1924:91

It was against these sentiments (see also Meillet 1925:vi-vii; Rivet 1925) that Bloomfield directed his famous article:

I hope, also, to help dispose of the notion that the usual processes of linguistic change are suspended on the American continent. (Meillet and Cohen, Les langues du monde, Paris, 1924, p. 9). If there exists anywhere a language in which these processes do not occur (sound-change independent of meaning, analogic change, etc.), then they will not explain the history of Indo-European or any other language. A principle such as the regularity of phonetic change is not part of the specific tradition handed on to each new speaker of a given language, but is either a universal trait of human speech or nothing at all, an error. [Bloomfield 1925:130]

Sapir, who had already engaged in the comparative reconstruction of a number of American Indian language families, seconded Bloomfield:

The methods developed by the Indo-Europeanists have been applied with marked success to other groups of languages. It is abundantly clear that they apply just as rigorously to the unwritten primitive languages of Africa and America as to the better known forms of speech of the more sophisticated peoples . . . . The more we devote ourselves to the comparative study of the languages of a primitive linguistic stock, the more clearly we realize that phonetic law and analogical leveling are the only satisfactory key to the unravelling of the development of dialects and languages from a common base. Professor Leonard Bloomfield's experiences with Central Algonkian and my own with Athabaskan leave nothing to be desired in this respect and are a complete answer to those who find it difficult to accept the large-scale regularity of the operation of all those unconscious linguistic forces which in their totality give us regular phonetic change and morphological readjustment on the basis of such change. It is not merely theoretically possible to predict the correctness of specific forms among unlettered peoples on the basis of such phonetic laws as have been worked out for them-such predictions are already on record in considerable number. There can be no doubt that the methods first developed in the field of Indo-European linguistics are destined to play a consistently important rble in the study of all other groups of languages. [Sapir 1949:78]

Following Sapir's and Bloomfield's work, the assumption that sound change is regular has proved itself useful and valid in case after case involving American Indian, as well as other "exotic," languages, and, of course, it is not to be for- gotten that the assumption was employed fruitfully in many earlier instances (see Haas 1969c, 1976a, 1977).

1997 LYLE CAMPBELL 657

4. Family-level historical linguistics. A very large portion of Mary's histor- ical linguistic output deals with standard problems of change, reconstruction, and subgrouping in specific language families, especially Algonquian and

Muskogean (e.g., Haas 1941a, 1941b, 1941c, 1946, 1947a, 1947b, 1949, 1958c, 1967a, 1967b, 1969b, 1973b, 1979). This work, on the whole, is very solid but is less visible to those who are not specialists in these languages. That is, there is little grand news in the application of standard procedures to specific problems. Nevertheless, a closer look at this portion of her work can be instructive.

For example, her classification and subgrouping of Muskogean (Haas 1941a, 1947a) grapples with the difficulty of areal diffusion and seemingly overlapping innovations. It provides insights that can inform attempts to subgroup other language families. Her reconstruction of Proto-Muskogean *kw, based on the sound correspondences in complicated contexts involving Creek k : other Musko- gean languages b, is a brilliant example of reconstruction and the application of the comparative method. Being aware of it can make the historical linguist more successful when confronted with other difficult cases elsewhere.

While individual applications of historical techniques to little-known lan-

guages seldom have much impact on the field in general, many of Mary's in-

sights from work in these languages have had more impact than otherwise would have been the case because of the publication of her The Prehistory of Languages (Haas 1969c), which might not have come into existence if she had not been asked to write the original version for Current Trends in Linguistics (Haas 1966a). Both versions are general treatments of historical linguistics that make extensive use of her own historical linguistic work in Native American

languages.

5. Haas on remote relationships. Not only did Mary write about the history of American Indian linguistics (see section 3), but her thinking helped to shape, and was shaped by, that history. Her early efforts, like those of Edward Sapir and many of his other students, involved, in addition to straightforward family- level historical linguistic research (see section 4), proposals of remote relation-

ships among various Native American language groups-that is, she also engaged in the "lumping" practice so prevalent at that time. As a student of Sapir's, she recognized this lumping correctly for what it was-preliminary hypotheses to be tested in later work:

Although a number of Sapir's [1929] proposed connections were considered by him to be merely working hypotheses (no more and no less), the various suggested relationships have been taken too seriously by some and perhaps not seriously enough by others. [Haas 1954:571

Just as with Sapir, some of Mary's work involving possible remote relation-

ships was successful, providing positive methodological lessons, while some was not. For example, her (1958a, 1966b) arguments on behalf of Sapir's contro-

658 ANTHROPOLOGICAL LINGUISTICS 39 No. 4

versial Algonquian-Ritwan proposal can be read with methodological profit, even today, in spite of the great amount that has been written in the last few years about procedures and criteria for investigating hypothesized remote relationships. In the case of Algonquian-Ritwan, she presented rather persua- sive evidence, utilizing mainstream methods relying on sound correspondences and morphological evidence. However, in the unsuccessful cases of more inclusive classifications that she proposed, her methods (true to the times) left much to be desired. Consequently, some proposals most closely associated with her name-for example, Gulf, Algonquian-Gulf-and others (see Haas 1951, 1952, 1954, 1956, 1958b, 1959, 1960, 1964, 1965) have since been questioned or abandoned, while others require more research (e.g., Natchez-Muskogean [Haas 1956, 1979]). There are, however, methodological lessons to be learned from seeing why the procedures employed resulted in unconvincing hypotheses of genetic affinity.

5.1. Algonquian-Gulf. For example, Mary's (1958b, 1960) Algonquian-Gulf proposal once received considerable attention, although today it is largely abandoned. It proposed a relationship between Algic (Algonquian-Ritwan) and the supposed "Gulf" languages (Muskogean, Natchez, Tunica, Atakapa, and Chitimacha). Mary left open the possibility of "additional affinities," claiming Tonkawa as "another likely affiliate" (Haas 1958b:231, cf. Haas 1960:985-86), but remained noncommittal concerning possible connections with Algonquian- Mosan (Algonquian-Ritwan, Mosan, and Kutenai), Mosan (Chimakuan, Waka- shan, and Salishan), Siouan, Hokan, and Hokan-Siouan-all of which were involved in broader proposals (many by Sapir) that involved in some way the various languages of her Algonquian-Gulf proposal. She reported her own surprise at what her comparison of Algonquian and Muskogean had turned up, since she had "assumed for a long time that any resemblances noted to Algonkian were the result of borrowing" (Haas 1958b:235). She presented 132 sets of lexical resemblances, together with tentative sound correspondences for Algonquian-Gulf. However, this evidence includes many sets that must be eliminated based on what are now standard methodological procedures. For example, several of these involve onomatopoeia (e.g., 'beat', 'bee', 'blow', 'breathe', 'crow', 'cry, weep', 'hawk', 'ring [hum, roar]', 'shoot', 'to sound', 'spit', 'split', 'swallow', and 'whistle'); some are nursery formations (e.g., 'older brother', 'daughter' ['daughter, father, mother'], and 'father' [three terms]; see Jakobson 1960); although Mary is generally careful in this regard, some involve liberal semantic associations (e.g., 'brain, hair of head', 'son, father, mother, daughter', 'defecate, stink, rotten', and 'mouth, tongue'); some 28 involve short forms or longer forms with only short corresponding portions (too short to eliminate chance as a possible explanation for the similarities shared); some involve expressive or sound-symbolic forms ('bloom', 'squeeze out juice, milk cow', 'foam', and 'swell'). Several of her sets involve comparisons of Algic forms with forms

1997 LYLE CAMPBELL 659

from only one other language rather than from a fuller range of the putative Gulf languages (e.g., 'big', 'crawfish', 'dry', 'dust', 'ear', 'far', 'father1', 'father, 'father-in-law', 'fear', 'foam', 'hair', 'head', 'hot', 'joined', 'liver', 'male', 'mouth',

'neck2', 'open', 'otter', 'third person pronoun', 'road', 'shake', 'shoot2', 'skin, hide', 'skin [verb]', 'snow', 'son', 'stone', 'swing', 'tapering at base, pear-shaped', 'ten', 'true, good', 'turn around', 'turtle', 'two'). Some so-called pan-Americanisms show up in the list (e.g., 'belly', 'bone', 'cover, spread, wide', 'dog', 'dust', 'foot', 'give', 'hand', 'first person pronoun', 'second person pronoun', 'leg', 'negative', 'wet, wash'; see Campbell 1997:290-319 for further discussion). Diffusion may be involved in 'skunk' (cf. Haas 1963a), 'crawfish', and 'buy'. Here, given the nature of the Southeastern linguistic area, the possibility of borrowing must be kept in mind and appropriate precautions taken. The few remaining forms do not

provide sufficient support to sustain the hypothesis.

5.2. Hokan. Another example involves Mary's work on the Hokan hypothesis. She (1964) compared 92 Yana and Karuk forms, including also forms from other

putative Hokan languages. Many of her forms are questionable, putting the sound correspondences she postulated in doubt. For example, of the 92 look- alikes compared, 13 involve onomatopoeia (e.g., Yana pu-, Karuk fum- 'to blow'); 26 involve forms too short to eliminate chance as a possible explanation for the

similarity (e.g., Yana ni-'one male goes', Karuk in- 'to go' [rare]); 10 forms exhibit too much semantic latitude (e.g., 'snow, to rain'); and 23 involve wide-

spread or pan-American forms. Another 15 have little phonetic similarity; 'digging stick' and 'manzanita berry' are suggestive of borrowing (with similar forms in a number of northern Californian Indian languages), and 'father' involves a nursery formation. Needless to say, when so many forms are in doubt, a number of the proposed sound correspondences cease to be viable. For ex-

ample, p: f is illustrated by only two proposed cognate sets: 'blow', clearly onomatopoetic (see above), and 'excrement', a pan-Americanism. The corres-

pondence b : fis exhibited only by 'frog' (onomatopoetic) and 'manzanita [berry]' (probably borrowed). Of the other sound correspondences, three are illustrated by only a single putative cognate set (and one case never constitutes a legitimate recurring correspondence), and ten by only two examples. Thus, her evidence does not suffice to show a genetic relationship between Yana and Karuk.

Haas (1963b, 1964) proposed nine Proto-Hokan reconstructions based on

phonologically similar lexical sets from a number of putative Hokan languages, all inspired by the assumption that "certain long vowels in Shasta have resulted from the contraction of a Proto-Hokan... *VmV... P[roto-]H[okan] *ama >

Sh[asta] /a./ and... *ima or *ami > Sh[asta] /e./" (Haas 1963b:42). While her charts for these nine forms show similarities, they also leave much room for doubt. For example, without a more fully developed proposal for the historical

phonological developments, one easily suspects in the case of 'ear' that Karuk t<i-v (where < means that the sound is assumed to have undergone a change of

660 ANTHROPOLOGICAL LINGUISTICS 39 No. 4

assimilation), Chimariko -sam, and Chumash tu? may not really be cognate forms in genetically related languages and that they may not derive from her proposed protoform *iamaruk'a - *iamak'aru (Haas 1963b:46). Similarly, Achomawi

owe.> 'liver' is a stretch from the assumed Proto-Hokan *6-imapasi -

imacipasi (Haas 1963b:47), and Chumash top'o, Achomawi alu, and Washo

i>?b 'navel' are a leap from each other and from the proposed Proto-Hokan *imarak'"i - *imak'"ari. The other sets are based on similar comparisons. Other problems involve: the small number of forms (nine) is too few to form the basis of a convincing genetic hypothesis; some of these involve so-called pan- Americanisms; and the alternative reconstructions and the reliance on meta- thesis provide so much leeway in the matchable phonetic space of the compared items that the possibility of accident is greatly increased. As William Bright points out (p.c. 1994), it is strange that 'ear', 'liver', and h'navel', basic body parts, should be four- and five-syllable words. In short, Mary's nine reconstructed forms do not constitute compelling evidence for the relationship.

5.3 Later views. Mary's outlook regarding proposals of wide-ranging genetic relationship apparently changed, in concert with general shifts in the field, over the course of her career. Thus, in 1977, she wrote with regard to the recent history of Americanist historical linguistics:

Others, however, were beginning to proceed in the opposite direction [from proposing new broad classifications], namely the splitting up of some of Sapir's superstocks, at first usually to put the parts together in new combinations. Several people have participated in making proposals for new combinations and not a few are still engaged in this activity. However, the pendulum seems now to be swinging gradually towards an increase in the amount of splitting being done. So we now have the classic opposition of lumpers vs. splitters. In our proposals about the classification of North American languages we have clearly moved in cycles of splitting and lumping, more or less successively .... While splitting appears to be in the ascendancy now, there will in the future be more lumping, but in a different way and after a fresh look at all the proposed components. [Haas 1978:124]

In earlier work, Mary had proposed Gulf (which would group Muskogean, Natchez, Tunica, Atakapa, and Chitimacha together) and the even broader Gulf-Algonquian hypothesis (1951, 1952, 1958b, 1960, 1979). However, later she had doubts about these broad classifications. Her doubt of Atakapa and Chiti- macha as "Gulf" languages is indicated by her use of dotted lines and question marks (Haas 1969c:63), and she expressed reservations concerning Gulf in general (Haas 1979; p.c. 1979). And, indeed, the "Gulf proposal is not upheld by most specialists in the field today (Campbell 1997).

6. Professionalism. Mary once expressed to me strong criticism of a very well known, very learned scholar, who gave frequent papers on various Native

1997 LYLE CAMPBELL 661

American languages and language families of which he had only fleeting and casual experience, at best. Though otherwise friendly to him, she did not condone his "gadfly" conference behavior-flitting from topic to topic with no depth and little substance and not infrequently causing unnecessary confusion. Mary's own work, in contrast, is a model from which others can take inspiration. She had first-hand experience with the languages and language groups she wrote about; these include Algonquian, Biloxi (and Siouan generally), Chipew- yan, Muskogean, Natchez, Nootkan, Thai, Tunica, several California languages, and others. Her work was not superficial or hasty.

I was inspired by Mary Haas's attitude about professional behavior, by her opinions about linguistic careers in general, and by her own career in particular. I don't remember precisely what she said or why she said it, but what I took from one particular conversation was her no-nonsense, down-to-earth beliefs. She said, in essence, that you do what you do to do it right, for the joy of scholarship, and not to try to impress people. And, if you do do the work right, the other aspects of career (like getting a job, getting published, or getting promoted) take care of themselves. She had no inflated sense of herself or of her work; her work contributed much to the field, and therefore she deserves to be

recognized as "the teacher, in one sense or another, of us all."

Notes

1. She also published much on Thai and worked in Tibeto-Burman, but I do not take up these aspects of her work in this article.

2. I thank Ives Goddard for this observation.

References

Baraga, Friedrich 1878 A Theoretical and Practical Grammar of the Otchipwe Language. Montreal:

Beauchemin and Valois. 1878-80 A Dictionary of the Otchipwe Language. 2 vols. Montreal: Beauchemin and

Valois. Bloomfield, Leonard

1925 On the Sound System of Central Algonquian. Language 1:130-56. 1928 A Note on Sound-Change. Language 4:99-100. 1946 Algonquian. In Linguistic Structures of Native America, by Harry Hoijer

and others, 85-129. Viking Fund Publications in Anthropology 6. New York: Viking Fund.

Boretzky, Norbert 1982 Das indogermanische Sprachwandelmodell und Wandel in exotischen

Sprachen. Zeitschrift fiir vergleichende Sprachforschung 95:49-80. 1984 The Indo-Europeanist Model of Sound Change and Genetic Affinity and Its

Application to Exotic Languages. Diachronica 1:1-51. Bright, William

1976 The First Hokan Conference: Conclusions. In Hokan Studies, edited by Margaret Langdon and Shirley Silver, 361-63. The Hague: Mouton.

662 ANTHROPOLOGICAL LINGUISTICS 39 No. 4

Campbell, Lyle 1994 Linguistic Reconstruction and Unwritten Languages. In Encyclopedia of

Language and Linguistics, edited by R. E. Asher and J. M. Y. Simpson, 3475-80. London: Pergamon Press.

1997 American Indian Languages: The Historical Linguistics of Native America. New York: Oxford University Press.

Charencey, Hyacinthe de 1870 Notice sur quelques familles de langues du Mexique. Havre: Imprimerie

Lepellatier. 1872 Recherches sur les lois phonetique dans les idiomes de la famille mame-

huasteque. Paris: Maisonneuve. 1883 M6langes de philologie et de paleographie am6ricaines. Paris: Ernest

Leroux. Cuoq, Jean-Andr6

1886 Lexique de la langue algonquine. Montreal: J. Chapleau et Fils. Delbriick, Berthold

1882 Introduction to the Study of Language: A Critical Survey of the History and Methods of Comparative Philology of the Indo-European Languages. [English translation of Einleitung in das Sprachstudium: Ein Beitrag zur Methodik der vergleichenden Sprachforschung. Leipzig: Breitkopf and Hirtel, 1880.]

Dixon, R. M. W. 1990 Summary Report: Linguistic Change and Reconstruction in the Australian

Language Family. In Linguistic Change and Reconstruction Methodology, edited by Philip Baldi, 393-401. Berlin: Mouton de Gruyter.

Emeneau, Murray B. 1956 India as a Linguistic Area. Language 32:3-16.

Goddard, Pliny Earle 1920 Has Tlingit a Genetic Relation to Athapascan? International Journal of

American Linguistics 1:266-79. Golla, Victor, ed.

1984 The Sapir-Kroeber Correspondence. Survey of California and Other Indian Languages Report 6. Berkeley: University of California.

Grasserie, Raoul de la 1890 De la famille linguistique Pano. International Congress of Americanists

7:438-49. Berlin. Greenberg, Joseph H.

1987 Language in the Americas. Stanford: Stanford University Press. 1989 Classification of American Indian Languages: A Reply to Campbell. Lan-

guage 65:107-14. 1990 The American Indian Language Controversy. The Review of Archaeology

11:5-14. 1991 Some Problems of Indo-European in Historical Perspective. In Sprung from

Some Common Source: Investigations into the Prehistory of Languages, edited by Sydney M. Lamb and E. Douglas Mitchell, 125-40. Stanford: Stanford University Press.

forthcoming Indo-European Practice and American Indianist Theory in Linguistic Classification. In The Classification and Prehistory of American Indian Languages, edited by Allan Taylor. Stanford: Stanford University Press.

Gyarmathi, Samuel 1799 Affinitas linguae Hungaricae cum linguis Fennicae originis grammatice

demonstrata. GSttingen: Joannis Christan Dietrich.

1997 LYLE CAMPBELL 663

Haas, Mary R. 1941a The Classification of the Muskogean languages. In Language, Culture, and

Personality: Essays in Memory of Edward Sapir, edited by Leslie Spier, A. Irving Hallowell, and Stanley S. Newman, 41-56. Menasha, Wise.: Sapir Memorial Publication Fund.

1941b Noun Incorporation in the Muskogean Languages. Language 17:311-15. 1941c A Popular Etymology in Muskogee. Language 17:340-41. 1946 A Proto-Muskogean Paradigm. Language 22:326-332. 1947a The Development of Proto-Muskogean *kw. International Journal of Ameri-

can Linguistics 13:135-37. 1947b Some French Loan-words in Tunica. Romance Philology 1:145-48. 1949 The Position of Apalachee in the Muskogean Family. International Journal

of American Linguistics 15:121-27. 1951 The Proto-Gulf Word for Water (with Notes on Siouan-Yuchi). International

Journal of American Linguistics 17:71-79. 1952 The Proto-Gulf Word for Land (with a Note on Proto-Siouan). International

Journal of American Linguistics 18:238-40. 1954 The Proto-Hokan-Coahuiltecan Word for 'Water'. In Papers from the

Symposium on American Indian Linguistics Held at Berkeley, July 7, 1951, edited by C. Douglas Chretien, Murray B. Emeneau, Madison S. Beeler, and Mary R. Haas, 57-62. University of California Publications in Lin- guistics 10. Berkeley and Los Angeles: University of California Press.

1956 Natchez and the Muskogean Languages. Language 32:61-72. 1958a Algonkian-Ritwan: The End of a Controversy. International Journal of

American Linguistics 24:159-73. 1958b A New Linguistic Relationship in North America: Algonkian and the Gulf

Languages. Southwestern Journal of Anthropology 14:231-64. 1958c Notes on Some PCA [Proto-Central Algonquian] Stems in /k-/. Inter-

national Journal of American Linguistics 24:241-45. 1959 Tonkawa and Algonkian. Anthropological Linguistics 1(2):1-6. 1960 Some Genetic Affiliations of Algonkian. In Culture in History: Essays in

Honor of Paul Radin, edited by Stanley A. Diamond, 977-92. New York: Columbia University Press.

1963a The Muskogean and Algonkian Words for Skunk. International Journal of American Linguistics 29:65-66.

1963b Shasta and Proto-Hokan. Language 39:40-59. 1964 California Hokan. In Studies in Californian Linguistics, edited by William

Bright, 73-87. University of California Publications in Linguistics 34. Berkeley and Los Angeles: University of California Press.

1965 Is Kutenai Related to Algonkian? Canadian Journal of Linguistics 10: 77-92.

1966a Historical Linguistics and the Genetic Relationship of Languages. In Current Trends in Linguistics. Vol. 3, Theoretical Foundations, edited by Thomas A. Sebeok, 113-54. The Hague: Mouton.

1966b Wiyot-Yurok-Algonkian and Problems of Comparative Algonkian. Inter- national Journal of American Linguistics 32:101-7.

1967a The Development of Proto-Algonkian *-awe-. In Studies in Historical Lin- guistics in Honor of George Sherman Lane, edited by Walter W. Arndt et al., 137-45. University of North Carolina Studies in Germanic Languages and Literature 58:137-45. Chapel Hill: University of North Carolina Press.

1967b The Proto-Algonkian Word for 'Sun'. In Contributions to Anthropology: Linguistics 1 (Algonquian), pp. 60-65. Anthropological Series 78, National

664 ANTHROPOLOGICAL LINGUISTICS 39 No. 4

Museum of Man Bulletin 214. Ottawa: National Museum of Canada. 1967c Roger Williams's Sound Shift: A Study in Algonkian. In To Honor Roman

Jakobson: Essays on the Occasion of His Seventieth Birthday, 11 October 1966. Vol. 1, pp. 816-32. The Hague: Mouton.

1968a The Menomini Terms for Playing Cards. International Journal of American Linguistics 34:217.

1968b Notes on a Chipewyan Dialect. International Journal of American Lin- guistics 34:165-75.

1969a Grammar or Lexicon? The American Indian Side of the Question from Duponceau to Powell. International Journal of American Linguistics 35: 239-55.

1969b Internal Reconstruction of the Nootka-Nitinat Pronominal Suffixes. International Journal of American Linguistics 35:108-24.

1969c The Prehistory of Languages. Janua Linguarum, Series Minor 57. The Hague: Mouton.

1969d Swanton and the Biloxi and Ofo Dictionaries. International Journal of American Linguistics 35:286-90.

1970a Consonant Symbolism in Northwestern California: A Problem in Diffusion. In Languages and Cultures of Western North America: Essays in Honor of Sven S. Liljeblad, edited by Earl H. Swanson, Jr., 86-96. Pocatello: Idaho State University Press.

1970b Review of New Views of the Origin of the Tribes and Nations of North America, by Benjamin Smith Barton. International Journal of American Linguistics 36:68-70.

1973a American Indian Linguistic Prehistory. In Current Trends in Linguistics. Vol. 10, Linguistics in North America (pt. 1), edited by Thomas A. Sebeok, 677-712. The Hague: Mouton.

1973b The Southeast. In Current Trends in Linguistics. Vol. 10, Linguistics in North America (pt. 2), edited by Thomas A. Sebeok, 1210-49. The Hague: Mouton.

1975 Problems in American Indian Philology. In Language and Texts: The Nature of Linguistic Evidence, edited by Herbert H. Paper, 89-106. Ann Arbor: Center for the Coordination of Ancient and Modern Studies, Univer- sity of Michigan.

1976a Boas, Sapir, and Bloomfield. In American Indian Languages and American Linguistics: Papers of the Second Golden Anniversary Symposium of the Linguistic Society of America, Held at the University of California, Berke- ley, on November 8 and 9, 1974, edited by Wallace L. Chafe, 59-69. Lisse: Peter de Ridder.

1976b The Northern California Linguistic Area. In Hokan Studies: Papers from the First Conference on Hokan Languages, Held in San Diego, California, 23-25 April 1970, edited by Margaret Langdon and Shirley Silver, 347-59. Janua Linguarum, Series Practica 181. The Hague: Mouton.

1977 Anthropological Linguistics: History. In Perspectives in Anthropology 1976, edited by Anthony F. C. Wallace et al., 33-47. Special Publication of the American Anthropological Association 10. Washington, D.C.: American Anthropological Association.

1978 Linguistics and History. In The Scientific Study of Language: The Role of the Linguistic Society of America, edited by Anwar S. Dil, 136-47. Abbotta- bad, Pakistan: Linguistic Research Group of Pakistan.

1979 Southeastern Languages. In The Languages of Native America: Historical and Comparative Assessment, edited by Lyle Campbell and Marianne

1997 LYLE CAMPBELL 665

Mithun, 299-326. Austin: University of Texas Press. Hale, Kenneth

1964 Classification of the Northern Paman Languages, Cape York Peninsula, Australia: A Research Report. Oceanic Linguistics 3:248-65.

1976 Phonological Developments in Particular Northern Paman Languages, and Phonological Developments in a Northern Paman Language: Uradhi. In Languages of Cape York, edited by Peter Sutton, 7-49. Canberra: Aus- tralian Institute of Aboriginal Studies.

Hockett, Charles F. 1948 Implications of Bloomfield's Algonquian Studies. Language 24:117-31.

Jacobsen, William H., Jr. 1979 Chimakuan Comparative Studies. In The Languages of Native America:

Historical and Comparative Assessment, edited by Lyle Campbell and Marianne Mithun, 762-802. Austin: University of Texas Press.

1983 Typology and Genetic Notes on Switch-Reference Systems in North American Indian Languages. In Switch Reference and Universal Grammar, edited by John Haiman and Pamela Munro, 151-83. Amsterdam: John Benjamins.

Jakobson, Roman 1944 Franz Boas's Approach to Language. International Journal of American

Linguistics 10:188-95. 1960 Why "Mama" and "Papa"? In Perspectives in Psychological Theory, edited

by Bernard Kaplan and Seymour Wapner, 21-29. New York: International Universities Press.

Johnson, Steve 1990 Social Parameters of Linguistic Change in an Unstratified Aboriginal

Society. In Linguistic Change and Reconstruction Methodology, edited by Philip Baldi, 419-33. Berlin: Mouton de Gruyter.

Kinkade, M. Dale 1985 More on Nasal Loss on the Northwest Coast. International Journal of

American Linguistics 51:478-80. Kinkade, M. Dale, William W. Elmendorf, Bruce Rigsby, and Haruo Aoki

in press Plateau Languages. In Handbook of North American Indians. Vol. 12, Plateau, edited by Edward E. Walker, Jr. Washington, D.C.: Smithsonian Institution.

Leem, Knut 1748 En Lappisk grammatica efter den dialect, som bruges af Field-Lapperne

udi Porsanger-Fiorden, sainamt et register over de udi samme grammatica anferte obervationers indhold. Kiebenhavn.

1768-81 Lexicon Lapponicum bipartitum: Lapponico-Danica-Latinum and Danico- Latino-Lapponicum. Nidrosie.

Meillet, Antoine 1925 La methode comparative en linguistique historique. Paris: Champion.

Meillet, Antoine, and Marcel Cohen 1924 Les langues du monde. Paris: Champion.

Morice, Adrien Gabriel 1891 The Dene Languages, Considered in Themselves and in Their Relations to

Non-American Idioms. Transactions of the Canadian Institute 1:170-212. 1892 Dene Roots. Transactions of the Canadian Institute 3:145-64.

Miihlhiusler, Peter 1989 On the Causes of Accelerated Linguistic Change in the Pacific Area. In

Language Change: Contributions to the Study of Its Causes, edited by Leiv

666 ANTHROPOLOGICAL LINGUISTICS 39 No. 4

Breivik and Ernst Jahr, 137-72. Berlin: Mouton de Gruyter. Osthoff, Hermann

1883 Schriftsprache und Mundart. Hamburg: Richter. Osthoff, Hermann, and Karl Brugmann

1878 Morphologische Untersuchungen auf dem Gebiete der indogermanischen Sprachen. Leipzig: Hirzel.

Powell, John Wesley 1891 Indian Linguistic Families of America North of Mexico. In Seventh Annual

Report of the Bureau of American Ethnology for the Years 1885-1886, 1-142. Washington, D.C.: Government Printing Office.

Rivet, Paul 1925 Les Australiens en Amerique. Bulletin de la Societe Linguistique de Paris

26:23-63. Sajnovics, Johannis [Janos]

1770 Demonstratio idioma Ungarorum et Lapponum idem esse. Copenhagen: Typis Collegi societatis Iesu.

Sapir, Edward 1913-19 Southern Paiute and Nahuatl: A Study in Uto-Aztecan. Journal de la

Societe des Americanistes 10:379-425, 11:443-88. 1929 Central and North American Languages. In Encyclopaedia Britannica.

14th ed. Vol. 5, pp. 138-41. London and New York: Encyclopaedia Britanica Company.

1949 The Concept of Phonetic Law as Tested in Primitive Languages by Leonard Bloomfield. In Methods in Social Science: A Case Book, edited by Stuart A. Rice, 297-306. Chicago: University of Chicago Press.

Sievers, Eduard 1876 Grundziige der Lautphysiologie zur Einfiihrung in das Studium der

Lautlehre der indogermanischen Sprachen. Leipzig: Breitkopf and Hiirtel. Stoll, Otto

1885 Supplementary Remarks to the Grammar of the Cakchiquel Language. Edited by Daniel G. Brinton. Proceedings of the American Philosophical Society 22:255-68.

1912-13 Zur Psychologie der indianischen Hochlandsprachen von Guatemala. Festschrift der Geographisch-ethnographischen Gesellschaft in Ziirich 1912-13:34-96.

Swadesh, Morris 1953a Mosan I: A Problem of Remote Common Origin. International Journal of

American Linguistics 19:26-44. 1953b Mosan II: Comparative Vocabulary. International Journal of American

Linguistics 19:223-36. 1962 Linguistic Relations across the Bering Strait. American Anthropologist

64:1262-91. Sweet, Henry

1900 The History of Language. London: J. M. Dent and Co. Thompson, Laurence C.

1979 Salishan and the Northwest. In The Languages of Native America: Historical and Comparative Assessment, edited by Lyle Campbell and Marianne Mithun, 692-765. Austin: University of Texas Press.

Thompson, Laurence C., and M. Dale Kinkade 1990 Languages. In Handbook of North American Indians. Vol. 7, Northwest

Coast, edited by Wayne Suttles, 30-51. Washington, D.C.: Smithsonian Institution.

1997 LYLE CAMPBELL 667

Thompson, Laurence C., and M. Terry Thompson 1972 Language Universals, Nasals, and the Northwest Coast. In Studies in

Linguistics in Honor of George L. Trager, edited by M. Estellie Smith, 441-56. The Hague: Mouton.