Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Tlingit phonology in a generative framework : an examination of phonological processes and abstract representation De Wolf, Gaelan 1977

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


831-UBC_1977_A8 D48.pdf [ 8.13MB ]
JSON: 831-1.0094182.json
JSON-LD: 831-1.0094182-ld.json
RDF/XML (Pretty): 831-1.0094182-rdf.xml
RDF/JSON: 831-1.0094182-rdf.json
Turtle: 831-1.0094182-turtle.txt
N-Triples: 831-1.0094182-rdf-ntriples.txt
Original Record: 831-1.0094182-source.json
Full Text

Full Text

TLINGIT PHONOLOGY IN A GENERATIVE FRAMEWORK: AN EXAMINATION OF PHONOLOGICAL PROCESSES AND ABSTRACT REPRESENTATION by GAELAN^ /de WOLF B.A., University of British Columbia, 1971 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF ARTS In THE FACULTY OF GRADUATE STUDIES Department of Linguistics We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA August, 1977 GAELAN de WOLF, 1977 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of Brit ish Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the Head of my Department or by his representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Gaelan de Wolf Department of T.-mpn-i «t-i r s The University of Brit ish Columbia 2075 Wesbrook Place Vancouver, Canada V6T 1W5 Date /V? /ff? - i i -ABSTRACT The Saussurean Paradox described by Labov (1971), in which "the social aspect of language can be studied by the theorist asking himself questions, while the individual aspect can only be studied by a social survey", apparently mirrors a predicament occurring in the structuralist and generative models of linguistics. For, while descriptive and structuralist linguistic models seek to mirror the reality of particular languages, a generative model of linguistics, in a search for universals, attempts to discover the underlying reality of a l l languages. Since an accepted raison d'etre of the current model of linguistic science is to provide an explanatory basis for real language, i t seems self-evident that both kinds of theories of language models are necessary: the inductive decision procedures of the reality-based structuralism, and the deductive discovery hypotheses of the mentalistic generative phonology. In the following Chapters, we shall attempt to explain the phonology of a particular language, which has been previously achieved through decision procedures, while investigating the natural and universal processes which have been hypothesized to occur. Although we expect to make no decisions, or even discoveries, we hope to examine the f i t of the phonology of a transformational generative model to a particular language. And, while testing the model, we hope to explicate the phonology of a particular North American Indian language, Tl i n g i t . - i i i -TABLE OF CONTENTS Page CHAPTER I Tli n g i t : Language and Theory 1 1.1 The Tlingit Language ' 1 1.1.1 Location and Population 2 1.1.2 Dialect Divisions 4 1.1.3 Present and Future Speakers 5 1.1.4 Linguistic and Ethnographic Studies 6 Historical Accounts and Ethnographic Data 6 20th Century Linguistic Studies 7 Comparative Controversy -the Na-Dene Hypothesis 8 1.2 Theories, Models and Traditions 10 CHAPTER II Descriptive Phonology 13 11.1 Characteristics of Tlingit Phonology 15 11.2 Consonants: Features and Arrangements 19 11.3 Vowels 25 11.3.1 Vowel Realization 28 11.3.2 Phonemic Variation 30 11.4 The Syllable 31 II.4.1 Syllable Types 34 11.5 Preferred Syllable Structure 36 - iv -Page. 11.6 Morpheme Structure Conditions 41 11.6.1 Redundancy 43 11.6.2 Constraints 43 11.6.3 Idiolectal Variation 48 11.7 Aufhebung or Neutralization 51 11.7.1 Neutralization in Generative Phonology 52 11.7.2 Aufhebung in Tlingit 53 11.8 Tone 54 11.9 Borrowing - Phonological Inloans 57 11.10 Dialectal Variation 61 11.10.1 Vocalic and Tonal Variation 62 11.10.2 Classifiers 64 n 11.10.3 Idiolectal Variation as a Dialect Feature 66 11.10.4 Other Features 67 11.11 Summary 69 CHAPTER III Phonological Processes 70 111.1 Standard Transformational Generative Theory 72 111.2 Naturalness in Phonology 74 111.2.1 Markedness 76 111.2.2 Natural Phonology 77 111.2.3 Natural Generative Phonology 78 111.3 Consonant Devoicing 80 - V -Page 111.4 Yod-Drop 81 111.5 Pre-Vocalic Stem Voicing 84 111.6 Assimilation: Labialization or Vowel Harmony 85 111.6.1 Labialization 86 After Rounded Segments 87 After Marked Segment /a/ 90 111.6.2 Vowel Harmony - Rounding Assimilation 91 After Rounded Segments 92 After Segment /a/ 93 111.7 Tone Dissimilation 95 111.8 Rule Ordering 96 111.8.1 Linear Ordering 97 111.8.2 Refinements to Rule Ordering 99 111.8.3 No Ordering Hypothesis 101 111.8.4 Application of Rules 103 111.9 Summary 108 CHAPTER IV Application and Approach in Phonology 110 IV.1 Abstractness in Phonology 112 IV.2 Morphophonemic Representation 113 IV.3 Approach 115 IV.4 Approach vs Application" 118 IV.5 Abstractness in Methodology 120 - v i -Page IV.5.1 Root Marker 120 IV.5.2 I n i t i a l Vowel Method 125 IV.5.3 Absolute Neutralization 129 IV.6 Concrete Phonology 132 IV.7 Process Solutions 135 IV.8 Prosody 138 IV.9 Evaluation Criteria 139 IV.9.1 Simplicity 140 IV.9.2 Economy 141 IV.9.3 Naturalness 143 IV.9.4 Alternation 144 IV.10 Vowel Harmony Methodology 145 IV.11 Vowel Harmony Typology 146 IV.12 Abstract Treatment 148 IV.12.1 Root Marker 148 IV.12.2 I n i t i a l Vowel 151 IV.12.3 Absolute Neutralization 153 IV.13 Concrete Treatment 155 IV.14 Process Treatments 157 IV.15 Metatheory - An addendum 159 IV.16 Theoretical Validity 163 IV. 17 Summary 165 Chapter V Summary 167 V. l Summary 167 V.2 Conclusions 169 - v i i -Page REFERENCES 171 APPENDIX 189 Chart I Tlingit Vowel Chart 190 Chart II Distinctive Features Chart 190 Chart III Tli n g i t Vowel Markedness 191 Data Appendix 192 - v i i i - -LIST OF TABLES AND CHARTS Page CHART I 18 CHART II 29 TABLE I Table of Tlingit Phonological Processes 104 - ix -ACKNOWLEDGEMENTS I would like to acknowledge, with thanks, the helpful suggestions of the members of my committee, Dr. R.J. Gregg, Head, Department of Linguistics; Dr. M.D. Kinkade, Graduate Student Advisor, and thesis supervisor; and Mrs. Ingrida Brenzinger, Senior Instructor; and in addition, to recognize the efforts of the typist, Donna Popovic, Department of Economics, Simon Fraser University. - 1 -1.0 T l i n g i t : Language and Theory 1.1 THE TLINGIT LANGUAGE Tlin g i t , s t r i c t l y classified as a language isolate of the proposed Na-Dene phylum in the distributional map of North American Indian languages (Voegelin and Voegelin 1966), is the language of Tlingit-speaking peoples of what is now southeastern Alaska, and the late pre-contact speech of bordering communities in what i s now northwestern British Columbia and the Yukon (Drucker 1965, Naish 1966). According to Wilson Duff's informant, William L. Paul, Sr., the name Tlingit is the contraction of a Tsimshian phrase meaning "from-place-of-tidal-waters-people" (cf. Duff in Oberg 1973, Pinnow 1976). Typologically, T l i n g i t phonology forms a part of the north-west coast linguistic diffusion area which, ranging from Alaska to Central California, includes the languages and superstocks of Athapaskan (Na-Dene phylum), Penutian, Salishan, Wakashan, Chemakuan, Ritwan (including Yurok and Wiyot), and Hokan (Haas 1969). Structur-a l l y , by Lehmann's (1972, 1973) typology (in which, according to principles of verb morphology, the object precedes the verb in simple declarative sentences), Tlingit,, like Athapaskan, is a consistent OV language characterized by agglutination, postpositions, and the placement of restricting elements before restricted elements, occasionally by vowel harmony and tone (v. Greenberg 1963, Tai 1972). Using Sapir's (1921) psycholinguistic terminology, Tlingit can be classified as polysynthetic in concept and of agglutinative-isolating technique. Of the six major grammatical processes list e d - 2 -by Sapir (1921), which include word order, composition, affixation, i n -ternal modification and accentual differences, only reduplication, which is otherwise common in languages of the northern Northwest Coast d i f -fusion area, is generally lacking in T l i n g i t . In terms of verb morpho-logy, detailed by Story (1966), the Tlingit verb complex shows evident similarity to the verb composition of the Athapaskan languages (Swanton 1908, Sapir 1915, Hymes 1956, Pinnow 1964, 1965, Krauss 1965). Interestingly, T l i n g i t forms part of the Pacific Rim isogloss which includes Chinese, Japanese, Korean, Ainu, Aleut, certain Northwestern Amerindian languages and Indonesian languages, and a number of Paleo-Siberian languages in possessing a phonemic system characterized by, at most, a single liquid (Jakobson 1968). In addition to sections of the verb dictionary of Story and Naish (1973), the major grammatical studies of Tl i n g i t include those of Boas (1917), Naish (1966) and Story (1966). According to Krauss (1973), in terms of comparative work including Velten (1939, 1944), Tlingit i s a "relatively bright spot" in the Na-Dene language documentation. I.1.1 Location and Population Situated at the northern end of the rich, sea-oriented Northwest Coast cultural area and linguistic diffusion complex, Tlingit territory, which stretched for approximately 500 miles along the coast of Alaska from Yakutat at 60°N. latitude to Ketchikan at 54°40' latitude, was bounded immediately to the south by the Haida (language isolate) on Prince of Wales Island and the Queen Charlotte Islands, to the southeast by Tsimshian (Penutian phylum), to the east by Athapaskan (Na-Dene), and to the west by Eyak (Eyak-Athapaskan family, Na-Dene phylum), Chugach Eskimo and Aleut (Eskimo-Aleut) (Drucker 1955, 1965, Haas 1969, Oberg 1973). - 3 -Sixty years ago, the ethnographer and linguist Swanton (1911:163) observed of the expansionist T l i n g i t : The Tlingit or Koluschan language is spoken throughout southeastern Alaska, from Dixon entrance and Portland canal to Copper river, with the exception-of .the south end of Prince of Wales Island which is occupied by Haida. An interior tribe of British Columbia, the Tagish, are said to belong to the same linguistic stock, but i t is by no means certain that they have not adopted the language from their Chilkat neighbors. Such a change is said,at any rate,to have taken place in the language of the Ugalakmiut, or Ugalentz, of Kayak Island and the neighboring mainland, who were formerly Eskimo and have now become thoroughly Tlingitized. Pinnow (1976) has identified the Ugalakmiut or Ugalentz with Eyak (v. Krauss 1965). With regard to pre-contact location and population, the ethnographer Averkieva (1971:321) has recently stated: At the time of the f i r s t European contacts, the Tlingit inhabited the southeastern coast of Alaska and the adjoining islands of Baranof, Chichagof, Admiralty, Kupreanof, among others. In addition, a small group of inland Tlingit lived around Lake Teslin. The Tlingit comprised the largest Indian tribe of Russian America, numbering around 12,000. However, following the ravages of several disastrous post-contact smallpox epidemics in 1787, 1836, and 1862, and typhoid fever in 1819, 1848, and 1855 (U.S. Federal Field Committee 1968), the 1880 Census (reported by Krause 1885) revealed a number of Tlingit subdivisions with a total population of 6,763. Nevertheless, according to de Laguna (1960), the northern Tlingit population was evidently never as large as that of neighbouring areas (cf. Powell 1891/ 1966, Borden 1945). - 4 -1.1.2 Dialect Divisions The 1880 Census of Petroff (Swanton 1908) lis t e d the fourteen qwan, or major t r i b a l subdivisions which formed the Tl i n g i t feder-ation, as Tongas (S), Sanya or Cape Fox (S), Henya (Klawak) (S), Kuiu (S), Kake (S), Sitka or Sumdum (N), Stikine (S), Taku (N), Auk (N), Hutsnuwu or Killisnow (N), Huna (N), Chilkat (N), Yakutat (N), and Hehl (Wrangell) (S) (cf. Krause 1885, Pinnow 1966, 1970). According to de Laguna (1960), a north and south cultural frontier and dialectal isogloss of the original t r i b a l groups was formed naturally by the waters of Frederick Sound which separated the Kuius, Kakes and other central tribes from the central Hutsnuwu or Angoon people, and the Sitka and other northern groups. The territory of the Angoon people, the most southerly of the northern T l i n g i t , adjoined that of the southern Kake, the northern Huna and Sitka to the north and west, and on the east and northeast the Taku and Auk (de Laguna 1960). The inland T l i n g i t , considered to be Tlingitized Athapaskans, are divided into three dialect groups, the Tagish, Teslin and A t l i n . According to McClellan (1953, in Landar 1967:70 n. 93): Tagish Indians number about 120. The native language in general use now is a Tlingit dialect which sounds "crooked" to the people of Atlin and Teslin. The Tagish regard i t as a recently learned language which supplanted an Atha-paskan dialect s t i l l preferred by some. This old Tagish speech seems to be close to Southern Tutchone and also to Tahltan. A recent U.S. government survey (Federal Field Committee 1968) has listed the present-day location of t r i b a l groups with their - 5 -native (underlined) and non-native community distribution in Alaska: Yakutat (Yakutat), Chilkat (Klukwan, Haines), Huna (Hoonah), Auk (Juneau), Taku-Sumdum (Douglas), Hutsnuwu (Angoon), Sitka (Sitka), Kaku (Kake), Stikine (Wrangell), Henya (Klawak), Sanya (Ketchikan), Tongas (Saxma'n) . The Kuiu evidently have no descendants and the present location of the Hehl is unknown. Drucker (1958) has discussed the v i a b i l i t y of native Tlingit-language groups in the towns and cities in Alaska. 1.1.3 Present and Future Speakers Various estimates have been given of native speakers of T l i n g i t . Chafe (1962, 1965) suggested a total of 2,000 speakers over the age of twenty in Alaska, British Columbia and the Yukon with a l l but two hundred liv i n g in Alaska, while Pinnow (1966) offered an estimate of 1,000 to 2,000. However, Krauss (1973) has offered a l i k e l y figure of more than one thousand, mostly over the age of twenty. According to Naish (1966), monolingualism of children in English has replaced the bilingualism of young adults on the coast, although Drucker (1958) noted that in Indian-speaking communities like Angoon and Hoonah younger people w i l l more generally speak Tlingit than in the predominantly white communities of Juneau, Sitka and Ketchikan. An informant in the Teslin lake area (Bob Fleming, interview, Toronto, 1971) has stated that most of the inland Tlingit now speak non-standard or "Indian" English. Despite the fact that classes i n the Tlingit language are common in Alaska native schools ,,wi"th a programme "of language learning and - 6 -native language teaching in effect, the future survival of Tlingit is doubtful, given the powerful and opposing forces of cultural assimilation and non-assimilation (Krauss 1973, v. Drucker 1958, 1976). 1.1.4 Linguistic and Ethnographic Studies Historical Accounts and Ethnographic Data Averkieva (1971), following Krause 1885, has noted that historical accounts of contact between Europeans and Tlingit began in the mid-eighteenth century. The Russian Bering-Chiricov exped-iti o n , of 1741, which f i r s t contained reports of Tl i n g i t activity, was followed by Spanish, English or French data-gathering expeditions of Perez 1774, Quadra 1775, Cook 1778, La Perouse 1786, Dixon and Portlock 1787, Marchand 1790, and Vancouver 1792-1794. In the nineteenth century, the Russians Lisianski 1812, Davidov 1810-1812, Lutke 1834, Wrangell and Veniaminov 1840 pub-lished ethnographic materials, while Krause 1885, Niblack 1890, Petroff 1881 and Anatoli 1906 detailed later studies (v. Averkieva 1971). American accounts include anthropological and li n g u i s t i c data of Boas (1917), Swanton (1908, 1909, 1911), and the cultural studies of Drucker (1955, 1965), Garfield 1947", McCleMan 1964; de Laguna (1960, 1964), and the well-known work of the Canadians Barbeau and Jenness 19327N- Thesesref erences"may; be' obtained in Averkieva (1971). - 7 -The additional l i s t of Tlingit references suggested by Pinnow (1976) includes Rezanov 1805, Gallatin 1836, Wrangell 1839, Scouler 1841, 1846, 1849, Schott 1843, 1849, Buschmann 1856, 1857, 1858, 1859, 1860, Dall 1877, Muller 1882, 1884, Grasserie (1902)J-Kelly and Willard 1905, Jones 1914, Goddard (1920), Miller 1931, Harrington 1945, de Laguna (1953, 1954, 1972), McClellan 1953 and Jacobs (1954). To these references, Drucker (1958), Durlach 1928J"Oberg (1937, 1973), Olson 1967 and Oswalt 1966 may also be added~(v. Duff in Oberg 1973). Further recent studies include those of Florendo and Dauenhauer 1971, Leer 1971, 1972, and Dauenhauer (1976) which are noted variously in Krauss (1973) and Pinnow (1976). 20th Century Linguistic Studies Boas' (1917) Tlingit grammar, following the earlier but some-what inconsistent work of Swanton (1911), has been the most sub-stantial account u n t i l the recent grammatical studies of Naish (1966) and Story (1966), although Velten (1939, 1944) published two analyses of a southern dialect (v. Landar 1967). Naish and Story (1963) have also published a useful English-Tlingit Dictionary: Nouns, in addition to primers and religious materials, while the Alaska Native Language Center, University of Alaska, has recently made available Story and Naish 1s (1973) Tlingit Verb Dictionary. Naish (1968) has also introduced a set of beginner's lessons in Tlingit (Krauss 1973). - 8 - Comparative Controversy - the Na-Dene Hypothesis The controversial issue of the relationship of T l i n g i t , Haida and Athapaskan was raised by Sapir (1915). Naming the projected phylum Na-Dene (Tl. na "people", Ath. dene "person, people"), Sapir con-cluded that "correspondences are of so intimate a character that mutual borrowing of words and morphological features seem out of the question". However, Krauss (1973) has noted that the Na-Dene hypothesis was suggested by Europeans in the eighteenth century, much earlier than Sapir (1915). For example, early attempts (Resanov 1805) detailed in Krauss (1964) disclosed similarities among Eyak, Tlingit and Tanaina thought to be the result of borrowing, while Wrangell's editors considered Eyak, T l i n g i t and Athapaskan to be related through widely divergent. Radlov .1858 thought Eyak might be related to both Athapaskan and Tlingit and, with Buschmann, considered i t a link between Haida and Tl i n g i t , as did Boas :1894. Subsequently, Swanton (1908) tentatively suggested a very distant genetic relation-ship between Haida and T l i n g i t , perhaps descended from a common parent language such as Athapaskan, while the important paper of Sapir (1915) raised the issue of a genetic relationship based mainly on morphological rather than phonetic similarity, an approach which Goddard (1920) and Boas (1920, 1929) rejected (Hymes 1956). Since 1950, a number of comparative papers on the subject of Na-Dene have appeared: the work of Hymes (1956), Krauss (1964, 1965, 1968, 1969, 1973), and Pinnow (1962, 1964, 1966, 1968, 1970, 1976). - 9 -Noting that the existence of a Na-Dene relationship and the use of morphological c r i t e r i a for genetic relationship are ... histo r i c a l l y interrelated problems Hymes (1956) published an interesting paper on the positional analysis of categories occurring in verb stems in Athapaskan, Tli n g i t , Haida and Na-Dene which established systematic morpho-logical similarities. Following the series of reasoned articles by both Pinnow and Krauss on the Na-Dene question, Pinnow (1970) concluded that i t follows beyond question from Krauss' (1970) discerning articles that Tlingit and Eyak-Athapaskan are related; the parallels which Krauss brings forth rule out any other conclusion. Krauss, who is more involved in the f i e l d of Athapaskan languages generally, has taken longer to be convinced of any genetic relationship based on the striking typological similarities. Recently, though, Krauss (1975) has allowed f o r c a distant relation-ship of Tlingit to Athapaskan-Eyak^salthough ruling-out a relation-ship with Haid'aii(cf. Levine,. 1973) through lack of phonological correspondence or grammatical similarity. Broader relationships have been suggested for T l i n g i t or.'Na-Dene, including a-postulated Sino-Tibetan-Na-Dene of^Sapir and Morice (Shafer 1952, v.. Swadesh 1952); the Vasco-Dene including Basque, Wakashan, Kutenai, Japanese and Indo-European of Swadesh; and the Athapaskan-Tlingit-Yuchi-Souian of Haas (Krauss 1973). - 10 -1.2 THEORIES, MODELS AND TRADITIONS In discussing linguistic blinkers, which through r i g i d i t y may retard the growth of an entire discipline, Heiberg (1973) has noted that, in a hierarchy of models, the culminating root model provides an orientational framework for an entire discipline, while a theoretical model is merely description couched in meta-language which provides a c r i t i c a l method of examining existing explanation. Kelly (1971), noting parallels of thought in the philosophical tradition of the mediaeval speculative grammarians with Chomsky's transformational grammar, has found that repeatedly "the needs and procedures of linguistics have oscillated between description and explanation". Tracing historical patterns in linguistic thought from the schoolmen to the present day, Kelly (1971: 250-251) has placed the model of generative grammar in a logical progression from structural-ism in the context of growth or philosophical activity versus order or systematization: After a period spent in developing historical linguistics, linguists followed de Saussure in developing annew type of descriptive linguistics which was based on observation of language behavior instead of analysis of written texts and literature. As this stream reached the height of i t s development, Chomsky brought to bear on theffield his own training as a mathematician and philosopher. In discussing the history of descriptive linguistics in America (1900-1950), Teeter (1964: 199-201) has commented on the-logical fallacies evolving from the dogmatism of the post-Boasian and post-Bloomfieldian structuralist traditions. From implications of Boas' work explicit in Sapir has stemmed the tradition of f i e l d work in which practice - I l -ls divorced from theory in a view that "reality can be approached without perceptions", while from the anti-mentalist theories of Bloomfield arose the discreditation of theory per se on the rationale that mind i t s e l f i s limited to input-output conditions. From these biases, since no appeal could be made to general theory, "justification of a given analysis came to mean the specifying of the procedures used to arrive at a given result from the data", and theory was thus equated to practice. Chomsky (1966), however, has suggested that the present choice between a generative grammar and a descriptive grammar is not genuine, since the latter is simply "one aspect of the f u l l generative grammar . . . i t is not a choice between competing systems, but rather a choice between a whole and a part". Though opposing views of mentalistic and anti-mentalistic thought may periodically rule linguistic procedures (v. Kelly 1970), in the present study we intend to use the framework of transform-ational generative grammar as a theoretical model to examine the phonology of Tlingit and, insofar as possible, to provide a description of certain phonological processes. In Chapter II and throughout, we examine Tlingit phonology making use of data and materials from Boas (1917), Naish (1966), Story (1966), and Velten (1939, 1944). With the recently incorporated acceptance of surface phonetic constraints, the phonemic system of surface contrasts is explicated more simply within the framework of generative phonology. In Chapter i l l we investigate the simple - 12 -phonological processes detailed by Boas which, according to Stampe's theory of natural phonology (an outgrowth of transformational gener-ative grammar), are universal and natural. In Chapter IV we consider the issue of abstract representation raised by Kiparsky for generative phonologists, and we investigate the level of morpho-phonemic representation using an interpretation of vowel harmony to examine the unresolved controversy. In Chapter V., we review bri e f l y the work of preceding Chapters, having discussed, without conclusion, certain issues of theoretical interest in phonological theory. - 13 -II.0 DESCRIPTIVE PHONOLOGY In this Chapter, a description of Tlingit phonology, using primary data from Boas (1917) and Naish (1966) and Story (1966) with additional examples drawn from Velten (1939, 1944), w i l l be set out within the theoretical framework of generative phonology. The phonology presented w i l l f i r s t be discussed i n articula-tory and descriptive terminology to make plain on a purely phonemic level the number of surface contrasts occurring especially oh a consonantal plane for, as Schane (1971: 503) has suggested, certain linguistic effects may only be explained satisfactorily in phonemic terms: The phoneme as a viable phonological unit for capturing relevant surface contrasts has been rejected within generative phonology ... The historical phenomena of denasalization in French, depalatalization in Rumanian, delabialization in Romance, palatalization and l a b i a l -ization in Nupe, and palatalization in Japanese are due to surface contrast rather than to statable morphophonemic or phonetic processes. From these observations I conclude that the phoneme must be recognized as a phonological entity. Atchison (1974) also, in a diachronic study, has noted that certain sound changes in pre-classical Greek may be more adequately characterized at the explanatory level by recognizing surface phonetic constraints or "output conditions" in addition to symmetricality of arrangement or pattern congruity as part of a speaker's unconscious psychological reality. Phonetic data used w i l l be drawn mainly from Boas' remarkable pre-phonemic study in the Chilkat dialect, although supplemental - 14 -phonological materials w i l l be based on the more recent work of Naish and Story who use an adapted Bendor-Samuel model in a tagmemic framework. While other references w i l l be cited, any analysis of Tlingit within an adequate linguistic context obliges primary r e l i -ance upon these works. Boas' data, the result of two months' work with his erratic but b r i l l i a n t informant, Louis Shotridge (v. Duff's Foreword in Oberg 1973), unlike the data of John R. Swanton (1908, 1909, 1911), some of whose findings appear in the Handbook of American Indian Languages (1911), are phonetically secure. According to H.V. Velten (1939: 65): For establishing dialectal differences only Boas' description of northern speech can be ut i l i z e d with safety. Earlier material and Swanton's transcription, as well as his grammatical analysis, are often inaccurate. For example, he disregards pitch altogether; he f a i l s to distinguish between the glot-alized voiceless lateral spirant 1' and the corresponding affricative t l ' ; arid he uses a single symbol for the glottal-ized velar stop^', and the glottalized palatal and velar spirants x and x, which are a l l clearly distinguished i n southern as well as northern T l i n g i t . However, Boas includes in his Notes a good deal of Swanton's material in corrected form. Until the recent work of Naish and Story, Boas' work in the f i e l d has offered the most complete and reliable material available (v. Landar 1967: 62 n.6). However, certain flaws in Boas, noted in Pinnow (1968: 205), for example, the failure to establish the extensive root variation which occurs in the verb, have been corrected and extensively amplified in Story. For T l i n g i t , the most recent, thorough-going and consistent work has been that of Naish and Story. Based on the Angoon or central dialect, their work represents the most comprehensive attempt to - 15 -construct a lexicon for Tlingit (1963, 1973), in addition to phono-logical and syntactic processes in the contemporary context. While their 1966 work remains unpublished, extensive notes for this paper have been made from the original manuscripts at the University of London. II.1 CHARACTERISTICS OF TLINGIT PHONOLOGY Briefly, according to Story, the Tlingit phonological system may be characterized as possessing some forty-two consonant phonemes (in some idiolects, forty-four), consisting of: 37 obstruents: stops, affricates and fricatives; 3 sonorants: a nasal and two semi-vowels; and 2 laryngeals: a glottal stop and fricative. In some idiolects, depending on rounding, there are four laryngeals. As Velten (1944: 168) has suggested: The consonant system is rather intricate and unusual. There are no la b i a l stops, spirants or nasals, but on the other hand there exist no less than ten rounded (or labialized) back consonants. However, Velten (1944: 169) has noted: 1 Boas' (1917:9) chart indicates a total of 41 consonants. Although the glottal stop is not marked, a bracketed palatal sonant y occurs. According to Naish (1966:222): "The glottal stop is phonemic in intervocalic position within a stem or in i n i t i a l position in a stem whose f i r s t syllable i s the tonic syllable of the tone group. In other environments the glottal stop is facultative." - 16 -If the number of Tlingit consonants seems, therefore, f a i r l y high, i t should be remembered that a comparison of the relative wealth or poverty of phonological systems is a matter of counting, not the phonemes, but the distinctive features. Actually, the Tlingit con-sonant system has only two more such features than has English. These are the glottalization and rounding, i.e. opposition of the type /k - k/ and /k - k u/. There are eight vowel phonemes which may be marked by high or low phonemic tone. Typologically, the consonantal phonology is remarkable for the complete absence of labials, and for the lack of voiced spirants, a factor originally noted in Boas. In terms of Northwest Coast areal phonology, Mary Haas (1969: 85-9) has stated that the most striking features of the complex consonant system in Tlingit are the series of labi6=velars?cba"Gklve0iars]oglotta'la!zedostop's', notably the occurrence of glottalized spirants, and the appearance of lateral affricates and spirants. In a recent paper, Kinkade (1976, Table II) found as a differentiating feature of the proposed "Na-Dene" from their neighbours immediately to the south the contrast of alveolar and palatal affricate-spirant positions in Tl i n g i t , Eyak, Carrier and Chilcotin. Shown in Chart I (adapted from Naish), the consonantal phonemic system i s remarkable also for the number of stop and manner opposi-tions, the presence of fourteen glottalized phonemes including six strongly glottalized fricatives, an almost f u l l set of laterals, and twenty back consonants, rounded and non-rounded. There is a single nasal which according to Troubetskoy (1957: 189) is noteworthy in that: - 17 -Parmi les langues que nous connaissons seul le t l i n g i t pre"sente un rapport oppositionnel isole 'occlusive-nasale' (d-n), n. £tant i c i l a seule nasale et l a classe de localisation labiale n'existant pas.^ According to data taken from Naish (1966: 19-21) and Story (1966: 14-19), phonemes have the following phonetic realizations: The simple alveolar stops /d, t, t/ are somewhat fronted phonetically. The apical affricates and fricatives /dz, c, c, s, s/ are articulated with the tip of the tongue and have a grooved sibilant release. A v v y • v The laminal affricates and fricatives / j , c, c, s/, phone-t i c a l l y palato-alveolar, are articulated with the blade of the tongue. They have a grooved sibilant release. The lateral affricates and fricatives A , 9t, X, 1, i / , with lateral release-,;-are .not •'noticeablyi' .clear or dark. Uvulars, not far removed from back velar position, are articulated at the back of the soft palate. According to Velten (1939: 66) for the Klawak dialect: "The palatal spirant x is articulated slightly further back than the ch in German ich, the sibilant c_ a t r i f l e further back than the English sh." The simple velars and uvulars are unrounded. The l a b i a l velars and uvulars are articulated with simul-taneous rounding. Nasal /n/ is alveolar in a l l positions. The plain stops are frequently lightly voiced intervocali-cally and elsewhere. The glides are palatal lyl and labiovelar /w/. 2 However, Velten (1944: 169 n.8) suggests: "The opposition /n-d/ is not a correlation since i t has no parallel in T l i n g i t . " CHART I Alveolar Velar Uvular Glottal Position Simple Apical Laminal Lateral Simple Round Simple Round Simple Round Stop Affricate Plain STOP Asp. Glott. V d dz j X t c c A 5 5 5 5 t c c * g g w k k w 5 5 k k w g g w q q w 5 5 q q w ? (? w) Plain FRIC Glott. S S i 5 5 S ± X x w 5 5 X X W X x w . . 5 5 X x w h (h w) Nasal SON Semi Vowel h y W Data taken from Naish (1966: 19-21) and Story (1966: 14-19). - 19 -Plain fricatives are voiceless and unglottalized. At syllable f i n a l , the contrast between plain and aspirated stops and affricates is neutralized. Stops and fricatives that are glottalized have almost simultaneous glottal closure. The simple uvular fricative has an uvular t r i l l variant. The glottalized uvular stop may occasionally have affricate release. The plain fricatives are always voiceless and unglottalized; occasionally aspirated word-initially. In certain environments the plain alveolar fricatives, as noted later, may be syllabic. II.2 CONSONANTS: FEATURES AND ARRANGEMENTS Although there are several possible arrangements of consonants, the Chart adapted from Naish (1966) is particularly neat, showing the three series of phonemically opposed alveolar affricates and the manner range of back and alveolar consonants. There are three pure stop positions, alveolar, velar and uvular, whichHoicleetlt (1955: 103) analyses: 3 pure stops: t - k q 2 colour modified: k w w q 3 affricate positions: c V C A four-way modification of simple stop with three affricated stops, apical, laminal and latera l , occurs at the alveolar 3 Back consonants are termed velars and uvulars by Naish (1966) and Story (1966); palatals and velars by Boas (1917), Swanton (1911), Jacobs (1954), and Velten (1939, 1944). Whether this is a question of terminology or acoustic variation is not made clear in Jacobs, pp. 47-50. See also Kinkade (1976: 3, 9) whose terminology is pre- and post-velar. According to Kinkade, at the phonetic level only, Tlingit (like Haida) has a palatalized k series of pre-velar obstruents. - 20 -point.* Using the markedness feature system of Chomsky and Halle (1968), this four-way modification may be distinguished as follows J simple alveolar +ant apical a f f r i c . +ant , laminal -ant lateral + l a t r e l +del alveolar +cor alveolar r e l — —' It should be noted, however, that in distinguishing a similar series of coronal contrasting non-continuants in Nitinat, Klokeid (1975: 90) has used, besides coronal, the five features of [+ delayed release], [+ forward=anterior], [+ lat e r a l ] , [+ high], and [+ distributed]. Lateral affricates and spirants are distinguished by the features [+ continuant], [+ forward], [+ s t i f f vocal cords], and [+ spread g l o t t i s ] . At velar and uvular points, colour modification (rounding) provides two extra positions of labialized velar and uvular. 4 Welsch (1976: 291) distinguishes lateral affricates and spirants in Hydaburg Haida by the use of the features [+ high], [+ back], and [+ continuant]. In the structuralist position, ju s t i f i c a t i o n for treating labio-velars and affricates as simple phonemes is offered by Troubetskoy (1957: 61): "Si dans ces langues des groupes phoniques comme ph, th, kh, ou pf, ts, kx, ou tw, kw, etc., peuvent se trouver a l ' i n i t i a l e , i l est c l a i r alors qu'ils doivent etre consideres comme des realisations de phonemes simples (consonnes aspirees, affriqueesj labialisees, etc.). Cela vaut par ex. pour les groupes ts, dz, ts, dz du t l i n g i t . . . " And Velten (1944: 169): "We have represented the affricatives as single phonemes. For i f they were to be considered as di-phonematic, we should have to state that T l i n g i t possesses the phonemes */z/, /z/ and / l / (occurring only after /d/), a somewhat absurd assumption which would lead to unnecessary complications." - 21 -There is a five-way manner contrast of plain, aspirated and glottalized stop; plain and glottalized fricative at the apical, laminal and lateral alveolar affricate points, and at the velar and uvular (rounded and non-rounded) positions. ~* The three-way stop opposition at the six points of articulation is of plain (unmarked) unaspirated, unvoiced stop vs.aspirated; voiceless stop vs glottalized voiceless stop. Although voicing i s clearly non-distinctive and redundant (Naish 1966: 21), plain stops may occur l i g h t l y voiced i n intervocalic position or occasionally at syllable onset. With regard to the feature voicing, Anderson (1974: 303) has suggested: "The feature [+ voice] occurs in a large number of rules, and i t s interpretation varies from language to language. It provides a parameter distinguishing 'more voiced 1 from 'less voiced' sounds in any given language . . . Thus the voiceless series of one language may be essentially the same as the voiced series of another. Generally some com-bination of glottal width and vocal-cord stiffness i s involved, and some configurations are certainly typical of voiced sounds while others are typical of voiceless sounds." 5 Hockett (1955: 105) has noted that five-way manner contrasts are of two types only: (1) voiceless unaspirated stop (sometimes glottalized), aspirated stop, voiced stop, voiceless spirant, and voiced spirant; and (2) unaspirated stop (sometimes voiced), aspirated stop, glottalized stop, voiceless spirant, and voiced spirant. Tlingit would thus f i t into a sub-type of (2) because of the stop opposition, forming a separate sub-type because of the voiceless vs. glottalized spirantal opposition. - 22 -In terms of distinctive features, however, the opposition could perhaps be captured by using the features aspiration and glottal closure, for example, unmarked /A/ would be: /d/ asp glott clos Itl + asp - glott clos It/ - asp + glott clos Features of [+ tense] and [+ heightened sub-glottal pressure] might be used instead, as suggested by Chomsky and Halle (1968), or the features of [+ s t i f f vocal cords], [+ slack vocal cords], [+ spread g l o t t i s ] , and [+ constricted glottis] u t i l i z e d by Klokeid for the Nitinat feature system. However, using laryngeal features of tenseness and aspiration introduced by Halle and Stevens (1971, in Anderson 1974: 301-2) instead of the features above, the three-way stop opposition could be depicted as: Id/ - spread glottis - constr glottis - s t i f f voc cords + slack voc cords , Itl + spread glottis - constr glottis + s t i f f voc cords - slack voc cords Itl - spread glottis + constr glottis + s t i f f voc cords - slack voc cords Lacking from the symmetry of the Chart is glottalized Is/ at the laminal alveolar position. For this Chart, with the exception of the voiced late r a l , a complete set of laterals occurs. Concerning lateral articulation as a contrastive pointy, Troubetskoy (1957: 157-8) has suggested: En effet du point de vue phonologique 1'articulation laterale ne peut etre consideree comme une particularity de localisation que s i elle est propre a plusieurs phonemes dont les autres marques distinctives sont semblables aux particularites de mode de franchissement - 23 -presentees par les phonemes d'autres series fondamentales (ou apparentees) appartenant au meme systeme (comme c'est le cas par ex. en pedi, en sandave, en t l i n g i t , en chinook, en adyghe, en avar, etc.). Only one nasal, alveolar /n/, occurs. Laryngeals /?K/ and /h w/ are bracketed since these phonemes vary in occurrence with idiolect.'' There are two semi-vowels, palatal and labiovelar /y/ and /w/. A voiced pos.fe«palafcal*aspirantef.kb:tadketed?oin- "Boasv'^tcharli.yas (y), has since disappeared and merged with /y/ (v. Pinnow 1976). The problem of the merging and disappearance of the sole voiced spirant i s an interesting one, noted by both Boas (1917) and Swanton (1908). Thissyoicgdlspiranfed(for which we useothe/IPA symbolt/x/) evidently occurred ihi;thei'SG)utherhadialect-srcircall§llv'bufeewas inetheaprocess of being replaced by the glide /y/. In northern dialects, the merging of the voiced post-palatalaspirarit^and^thetpalafialUiglideahad-already occurred, not a trace appearing in the speech of Boas' Chilkat informant whereas Boas (1917: 9) reports "In 1886 I heard i t distinctly and without any tendency to merge into y when taking down notes from a Stikine A single nasal i s not unknown in the phonology of American Indian languages. See Hockett (1955: 119) where: "Tillamook, and most or a l l of the Iroquoian languages (Seneca, Oneida, Cherokee, Mohawk) have no la b i a l con-sonants, and only one nasal /n/." For laryngeals, see Naish (1966: 21) and Story (1966: 19). Rounded laryngeals, evidently f i r s t noted by Naish and Story, occur i d i o l e c t i c a l l y in only a few words. According to Story: "In some idiolects labialized glottal stop and fricative occur in certain lexical items where the simple glottal phonemes occur in other idiolects." - 24 -Indian." 0 However, according to Krauss (Kinkade, p . c ) , this spirant s t i l l occurs in an unspecified Tlingit dialect. As noted, no labials occur although, according to Krauss (Kinkade, p . c ) , "for some speakers w is in certain positions m, and for some also n i s 1 (not related to the obstruent laterals at a l l . . . ) " . The 9 striking lack of labials has been widely noted in the literature, though T l i n g i t , as noted by Kinkade ( p . c ) , i s not alone in this feature since Tillamook (Salish) and Iroquoian languages also lack labials. 8» Swanton (1908: 396) termed this voiced spirant 'vocalic-velar', while noting (1911: 165) in the pronunciation of younger people i t s exact similitude with the English palatal glide. See also Boas' Tsimshian (1911: 288) wherein he has stated: "It corresponds to the sound in Tlingit which Swanton... writes y, but which I have heard among the older generation of Tlingit distinctly as the same sound as the Tsimshian here discussed. With the assumption that i t was originally the continued sonant corresponding to x of other Pacific Coast languages agrees i t s prevalent u_*tinge." Pinnow (1966: 21) has referred to [x] as a post-palatal, pre-velar spirant (v. n. 3 above). I. Brenzinger (p.c.) notes correctly that the symbol for the IPA voiced palatal fricative is [j] (v. Boas 1917": 9). 9 As Troubetskoy (1957: 135) has stated: "Nous ne connaisons aucune langue sans apicales; les guttura-les manquent par ex. dans quelques dialectes Slovenes de Car-inthie, les labiales en t l i n g i t (Alaska), mais ce sont l a des cas extremement rares: en general les trois series de localisation nominees ci-dessus apparaissent dans toutes les langues du monde." However, according to Jakobson (1939, in Troubetskoy 1957: 370): "Tel est le manque des labiales dans le t l i n g i t (et dans quelques parlers feminins de l'Afrique centrale), marque du a l a mutil-ation a r t i f i c i e l l e des levres, et meme dans ces cas, la classe des labiales tend a etre representee dans le systeme phonologique par des substituts specifiques." Chomsky and Halle (1968) note also that Jakobson (1940: 357-58), in propounding his theory of implicational sound laws, stated that (continued pg. 25) - 25 -II.3 VOWELS Phonemically, there are eight vowels in T l i n g i t . Schematically, from data of Story (1966: 20-21), Naish (1966: 22), and Velten (1944: 168), they may be arranged as front vs. back, close vs. open, and long vs. s h o r t , a l t h o u g h according to Story (1966: 20): The short vowels, with the exception of the short member of the central pair, are more open than the corresponding long vowels. The short central vowel is closer than the long central vowel. 9 continued Tlingit i s one of the few languages in the world without labials, to which is attributed the use of labrets. (Labrets were the bone or wood (rarely stone) l i p ornaments worn after puberty as a sign of wisdom by the aristocratic women of the tribe, v. de Laguna 1960). Chomsky and Halle (1968: 413) reiterate that: "The absence of labials in the speech of women in a few Central A f r i c a l languages is caused by r i t u a l i s t i c mutil-ation of the l i p s , and such mutilation also occurs among the T l i n g i t , where both men and women wear labrets." The above reasoning, however, is not valid on several grounds. Although evidently only high-caste Tl i n g i t women wore labrets, their use was f a i r l y widespread in the northern cultural area. For example, the neighbouring Tsimshian women used labrets (Tsimshian, of course, has a f u l l set of labials), as well as some of the Alaskan Athapaskans, the Haida, and the Babine, a Carrier (Athapaskan) tribe (v. Krauss 1964: 131 n. 6). Krauss has stated that: "The Na-Dene lack of labials, then, i s far stronger evidence in favor of the unity of the consonant systems of Proto-Athapaskan, Eyak, Tli n g i t and Haida, than for any correlation between the use of labrets and the lack of labials." 10 It should be noted that Troubetskoy (1957: 115), using Swanton's (1911) data, has classified both Tl i n g i t and Haida as having a triangular system of two degrees and two classes: a u i Unless Troubetskoy analysed /e,e/ as schwa, the triangular des-cription does not appear to be an apt f i t . - 26 -Front Back Close long i u short I U Open long e a short e A While phonetic length duration ±ntvowelsoi?slonly maintained in tonic syllables, long close vowels are more tense than the short close vowels.''""'" According to Naish (1966: 22) length distinction i n vowels, which also indicates a change in quality, is significantlgram-12 ati c a l l y in indicating the different inflections of the verb stem. Although Krauss (Kinkade, p.c.) is of the view that tense vs. lax i s the best way to state the opposition, which i s "definitely not a close-open opposition", Jones (1964: 40) has suggested: It is generally advisable to apply the terms tense and lax only to the case of close vowels. It i s extremely d i f f i c u l t to determine in the case of the opener vowels whether the sensation of 'tenseness' i s present or not, and there is in regard to some vowels considerable difference of opinion on the subject. In terms of distinctive features, however, since the feature LI According to Kinkade (1976: 4) a contrast of three or more vowel positions i s usual in Northwest Coast areal phonology. 12 Long vowels may occur with ei.therjihiighf orjnlow tone. Velten (1944: 168) noted in the verb system the morphophonemic alteration of open and close vowels, as well as of / A / and /e/. - 27 -length i s evidently not a Chomsky and Halle preferred feature, the suggested long-short feature opposition for both high and low vowels is that of tense-lax. According to Chomsky and Halle (1968: 324) : The feature "tenseness" specifies the manner in which the entire articulatory gesture of a given sound i s executed by the supraglottal musculature. Tense sounds are produced with a deliberate, accurate, maximally distinct gesture that involves considerable muscular effort; nontense sounds .rare produced rapidly and somewhat indistinctly. In tense sounds, both vowels and consonants, the period during which the articulatory organs maintain the appropriate configuration i s relatively long, while in nontense sounds the entire gesture i s executed in a somewhat superficial manner. Tense vowels, marked by greater effort, distinctiveness and longer duration, are articulated with a greater deviation from neutral position than are lax vowels. In an interesting discussion of the problem of feature opposi-tion in low vowels, however, Halle (lecture, U.B.C, 1974) has stated that tense-lax distinctions in low vowels are physiologically impossible since low vowels always occur with the feature [+ constricted pharynx]. He has therefore suggested that low vowels such as /as, a, n/ be distins guished by the feature [+ constricted pharynx], while high and mid lax, as well as low vowels, be distinguished by the use of the feature [- advanced tongue root]. For example, Tli n g i t long-short vowels using Halle's suggested feature approach could be characterized thus: - 28 -- back + back + high + A.T.R. - CP. - A.T.R. - CP. + A.T.R. - CP. A.T.R. CP. - high + A.T.R. - CP. 8 A.T.R. - CP. A.T.R. CP. - A.T.R. + CP. II.3.1 Vowel Realization The vowels indicated in Chart II are realized as follows accord-ing to Naish (1966: 22): i i s a close front unrounded vowel; I is a half-close to close front unrounded vowel; e i s a half-open to half-close front unrounded vowel; £ is an open to half-open front unrounded vowel; u i s a close back rounded vowel; U i s a half-close to close back rounded vowel; a i s an open central vowel; A is a half-open central vowel. Using secondary source material, Pinnow 1966:42 has suggested the following vowel realizations (v. 1966: 19-21) with some qualifications according to source: - 29 -CHART II front back close half-close half-open open - 30 -l±l i s r e a l i z e d as [ i i ; ], In/ as [u u-], [o O-] . 111 i s r e a l i z e d as [I], /u/ as [U] [o o]. lei i s r e a l i z e d as [e e* ], / A / as [ A o] , a f t e r /w/, lul as [o]. lei i s r e a l i z e d as [e], /a/ as [a a-], a f t e r /w/, / w / as [o] Although the back vowels /u, U/ generally tend to be lowered next to velars and especially to uvulars, there i s apparently l i t t l e 13 sub-phonemic variation for the Angoon or central dialect. Naish (1966: .22) has noted, however, that U tends to be lowered to o next to a tvui'a&&whi'ldl'4tnfe^e^i^e*€o-f olo^.e^±^ti&pa^c^£^e^whe^^3: is between two uvulars." For the northern Chilkat dialect, Boas (1917: 11) has noted: The quantitative value of vowels varies considerably. Unaccented syllables tend to have open vowels, which is due to the lack of intensity of movement. Where u and U are in contact with velars, they are apt to assume a less rounded character, and verge on o^, rarely on o_. II.3.2 Phonemic Variation On the basis of Boas' pre-phonemicized data of the northern dialect, Kinkade (personal communication) has raised the possibility of the occurrence of certain lax vowels as phonemic schwa within certain predictable consonantal environments. For example, where Boas' (1917: 17) orthography indicates the high lax back vowel U, contiguous to velars and uvulars, a 'coloured' schwa might be indicated: 13 According to Naish and Story (1963: v i i i ) , "sub-phonemic varia-tions are at a minimum; the phonemic symbols may be taken as representative of the phonetic values also." - 31 -k U q w A 4 /k a q w a 1 I shall hide i t II g U x w X I s i n II I hid i t for myself" /g a x A I s I n/ In support of this possibility i s Boas' (1917: 8) own testimony of his uncertain transcription with regard to the quality and quantity of some of the vowels: There is a considerable amount of uncertainty in regard to the quality of some of the vowels recorded by me, particul-arly in regard to the use of Q.[=/A/], e, and e, which is due to difference of pronunciation in rapid and slow speech. I discovered the significance of some of these differences in the course of my work, and had not the time to revise the whole material. Further support of phonemic schwa may be i t s widespread appearance in the Northwest Coast phonology, particularly among the Salishan lang-uages, e.g., Squamis.h, Upper Chehalis, Thompson and Shuswap, for which reference may be made to discussions in Kinkade (1963) , Davis (1971), Kuipers (1967, 1974) and Grubb (1974). According to Kuipers (1967: 36): The possible reduction of a l l unstressed vowels to schwa and the possible opening of schwa-type vowels to ae, o, etc., in the neighbourhood of uvulars, combined with the fact that reduced forms in many cases have become standardized, causes certain d i f f i c u l t i e s with regard to the phonemic interpretation of individual recordings. II.4 THE SYLLABLE The basic syllable structure of Tlingit i s CV where, in a sample of running text, Story (1966: 14) found 70 per cent open syllables. (In the adapted tagmemic framework used by Story, the syllable i s - 32 -defined as the phonological unit above the phoneme and below the tone group, while syllable division i s defined as phonemic). Tli n g i t syllable structure thus accords with Sapir's (1915: 534) findings for the proposed Na-Dene: The most typical and doubtless historically primary type of stem form found in the Na-Dene languages i s the monosyllabic stem consisting of consonant plus vowel . . . and with the universals of syllable structure discussed by Schane (1973: 52-53). According to Anderson (1974: 253) the syllable as a unit in generative phonology has generally been ignored, and a re-examination of i t s inclusion into the theory should be made, for: The consistent attempt in generative phonology to ignore the syllable as a structural unit-has not been based generally on a refusal to recognize the existence and? potential articulatory and psychological integrity of such elements. A case against the inclusion of syllables in phonological representations has never been made in any detail in print, but i t has generally been assumed that the sort of facts associated with syllable structure can in fact be attributed to the s t r i c t l y segmental representation, and do not require additional elements . . . It would be of some importance i f facts of syllable structure were to turn out to be mechanical consequences of other aspects of phonological representation. This does not, however, appear to be the case . . . Brown (1969: 5), also, has asserted that a striking feature of generative phonology is the non-introduction of the syllable "to account for the distribution of different phonetic realizations of the same systematic phoneme...." Both Hooper (1972: 525-540) and Brown (1969), however, have suggested ways of incorporating the syllable as a unit. - 33 -Hooper would recognize the syllable as a phonological unit in order to gain simplicity, generality and explanatory power. Prosodic features such as stress and tone affect the syllable rather than the segment or morpheme, for example, in tone languages such as Chinese (and Tlingit) where the syllable i s the tone-bearing unit. To Hooper, the syllable might be defined in two ways within the theory: F i r s t , as segment sequences in the lexicon, underlying syllables might be postulated, with syllable boundaries referred to by marking conventions and morpheme-structure conditions. Second (Hooper's choice), the syllable could be defined in terms of segment sequences with syllable boundaries inserted by a universal rather than a phonological rule. For example, a syllable boundary could be inserted by universal convention between a vowel and the following CV thus: 0 » $ / [+ syll] [- syll] [+ syll] Applied to Tl i n g i t , Hooper's definition of the syllable might be very useful in the formation of phonological rules to account for the complex morphology of the verb word with i t s numerous morpho-phcnemic changes and the shifting vowel deletion. Though sufficient data from Story's extensive work is not at hand, i t i s evident from Boas' comments and examples that a syllable explanation such'as Hooper's might lead to greater clarity of explanation in prefixal morphophonemics. For example, Boas (1917: 61) has noted: When a prefix ending in a vowel precedes the nominal subject and the modal element, the pronoun which stands between the modal element and the c l a s s i f i e r loses i t s vowel, i f the - 34 -cl a s s i f i e r i s vocalic. In this case the modal element retains i t s vowel. If the cl a s s i f i e r i s consonantal, the pronoun retains i t s vowel, while the modal element loses it.14 In an alternative method to Hooper's, Brown (1969) using i f -then sequence structure conditions would introduce the syllable into generative theory by having the redundancy rules assign syllabic status, as well as the distribution of allophones, to the phonetic realization. For example, Brown (1969: 5-6) has suggested as a morpheme structure condition: The f i r s t of a sequence of two [- consonantal] segments w i l l be assigned the value [- syllabic] (glide), and the second w i l l be assigned the value [+ syllabic] (syllable nucleus). II.4.1 Syllable Types There are two syllable types in Tlingit: consonantal and vocalic. Vowels are the nucleus of a vocalic syllable. Consonantal syllables consist of a single consonant which may only be the apical, laminal or lateral non-glottalized fricatives /s, s, ±/. According to Story (1966: 23): 1 5 Every vowel constitutes the nucleus of a vocalic syllable. Vocalic syllables are of the structure V, VC, CV, CVC, CVCC ... Consonantal syllables comprise a single consonant. 14 Dauenhauer (1976: 11) evidently concurs that "a series of three open syllables i s unstable in T l i n g i t . " 15 Aecordihghfco Naishr (196.6 :T24) ^ TGVahrid CVC syllables occur in a l l environments. CVCC occurs almost always as a tonic syllable, while VC and V occur only in pre-tonic syllables. - 35 -Except in cases where the morphology i s not known ... i t can be stated that these syllables are congruent with (1) a total grammatical word which i s not a c l i t i c or ... (3) with an extensor when this extensor occurs between 2 consonants. Congruent with a grammatical word, i.e., s "reflexive" or ± "negative", or morphologically as an extensor form /s, s, 4/, the plain alveolar fricatives may occur syllabically either an an i n -dependent word or interconsonantally in the verb word. Naish (1966: 24-25) has stated the environment as follows: (1) between two consonants within a tone group preceding the tonic syllable — kAyiy•s*xidin - 'you (pi.) had written' (2) i n i t i a l t ^ the tone group and preceding either a con-sonant or a vowel which may be potentially preceded by a glottal stop, that i s i n i t i a l to a grammatical word: I'qUstini 'blind person', I'Us^ ' e v i l ' . Exceptions to this rule occur under two conditions: i f , as grammatical words, the alveolar fricatives Is, 4/ are preceded by (1) a non-tonic bearing word, and (2) i f Is, s, 4/ are preceded by a non-tonic bearing word which i s also vowel f i n a l , e.g. ya s« nAd?en - 'he i s dressing'. Only the independent words (s or 4) or the extensor forms that are unglottalized alveolar fricatives undergo the following rule: Consonant S y l l a b i f i -cation ~+ cor + cont ; - glott 'clos n / -> [+ syll] 7 rcx-| # stem Plain alveolar fricatives may syllabify within stated grammatical environments. - 36 -There is a general congruence between word and syllable bound-aries with certain exceptions. For example (Naish 1966: 25), the f i r s t consonant in an i n i t i a l consonant cluster of a head word w i l l belong to the preceding syllable of a vowel-final modifier. Syllabicity of the apical alveolar fricatives in certain grammatical positions has already been noted. Within a word, a single consonant between vowels forms the onset of the following syllable. Two consonants word-medially are divided by syllable boundary. A l l consonants may occur as syllable onsets, although laryngeals (Story 1966: 27) may occur only in stem-initial syllable onsets. II.5 PREFERRED SYLLABLE STRUCTURE Schane (1973: 53-4) has defined a preferred syllable structure as: Any process which takes a more complex syllabic structure and reduces i t to the CV pattern leads to a preferred syllable structure. The effect of such processes i s to break up clusters of consonants or sequences of vowels. A preferredcsyllable structure, which is basically the reduction of a mo more complex syllabic pattern to the preferred CV, may be one result of certain phonological constraints, 1^ such as the surface phonetic 16 Aitchison (1974: 5), in arguing for the psychological reality of surface phonetic constraints, has suggested: If a l l that is required is an apparatus for generating a l l and only the permissible sequences, then no such constraints are needed since the morpheme structure conditions (lexical redundancy rules) and the phonological reules can achieve this. But i f one is interested in the kinds of mechanisms that a speaker is l i k e l y to have and which play a role i n language change, then output conditions become relevant as a template against which a l l new forms are matched. Without them, the phonological rules seem essentially arbitrary and pointless. - 37 -constraints (Shibatani 1973: 87-106) which operate to produce a phonotactically patterned output of the P-rules. According to Shibatani (1973: 87), while such constraints on distribution and combination expressed through allophonic rules and phonotactics were featured in the structuralist framework and phonetic theory, generative theory s does=^ nOits currentiy ipr.ozv-id'e^  jar-.v '.de a . . . direct means to capture phonetic constraints, and generative phonologists seem to bee content twith'-an-.inr--direct account by means of abstract morpheme-structure conditions together with the effects of phonological rules. Within the context of generative theory, syllable structure processes may occur within and across morpheme boundaries. Morpheme structure conditions, which include statements of permissible sequence combinations, handle such constraints within a morpheme, while phonological rules may account for such processes across and within morphological boundaries. For example, Anderson (1974: 287) in-cases of syllable restructuring has noted that: . . . the morpheme structure condition and the phono-logical rule are clearly different aspects of the same fact, and should not be treated as completely separate. If they are simply stated, each in i t s own. component of the grammar, this w i l l not be expressed, however: i t w i l l appear as an accidental fact that some of the phonological rules have the effect of ensuring that constraints on underlying structures are also imposed on derived structures. Although the standard theory of Chomsky and Halle (1968) has generally not recognized the need for some kind of surface constraints, - 38 -various suggestions, briefly reviewed below, such as the 'derivational constraints' and 'conspiracies' of Kisseberth (1970), the 'phono- i l ogical targets' of Haiman (1972), the 'negative target' of Kiparsky (1972) and the 'surface phonetic constraints' of Shibatani (1973) have shown the necessity for incorporating some'surface device into standard theory although, (as in the case of Shibatani's s.p.c.'s) surface phonetic constraints may overlap with morpheme-!Structure conditions).. Kisseberth's derivational constraint, to be examined bri e f l y in T l i n g i t , i s the effect of a set of rules operating together functionally to produce, by virtue of certain rule environment restrictions, a general constraint on the linguistic output. For example, concerning output eohditionstin-Yawelmani, Kisseberth (1970: 293) has stated: There are rather heavy constraints in Yawelmani phonetic representations on the clustering of consonants and of vowels. No vowel-vowel sequences are permitted. Words may neither end nor begin with consonant clusters. Nowhere in a word may more than two consonants occur in a sequence ... But in fact there are a variety of phonological processes which, i t might be said, 'conspire' to yield phonetic representations which contain no word-f i n a l clusters and no t r i l i t e r a l clusters. Kisseberth's conspiracy, then,-,is the result of a related morpheme structure condition and phonological rule which may, for example, have the effect of reducing a consonant cluster. Haiman (1972: 376), has suggested that various languages, such as Papago and Yawelmani, may have CV(C) syllable-structure targets (CV is the maximally unmarked syllable) which are realized by means - 39 -of epenthesis and deletion rules, while Kiparsky (1972: 195) has argued that "the theory of derivational constraints and 'conspiracies' f a i l s to provide an adequate explanatory account of these constraints on phonotactic structure" and that a notion of 'negative target' is needed instead. In borrowing from the theory of autonomous phonemics, Shibatani has suggested the notion of surface: phonetic constraints to explain surface redundancies and phonotactic patterns in generative phonology. In discussing the need for derivational constraints in Yawelmani phonology, Kisseberth (1970: 294) has noted that: Fir s t of a l l , in the underlying representation of Yawelmani morphology, there t r i l i t e r a l clusters. If regular-i t i e s of this sort are to be reflected in the grammar, there must be a language-specific morpheme structure condition which requires the segments on either side of a CC sequence to be vowels. The existence of this morpheme structure condition is not sufficient to account for the absence of t r i l i t e r a l clusters, for i t blocks such clusters only inside the morpheme. Morphological processes of suffixation and special stem formation rules operate so that morphophonemic representations of words do contain t r i l i t e r a l clusters. Specifically, in T l i n g i t , according to Naish (1966: 22), and Story, a similar process of consonant cluster reduction occurs for, while the basic syllable structure is CV, consonant clusters of three or more may be found across grammatical boundaries. According to Boas' northern dialect data (1917: 12-13), conson-ant clusters are rare both i n i t i a l l y and terminally in stems, although clusters may originate with certain consonant suffixes, e.g., the diminutive. Medial clusters which originate through.word composition - 40 -are evidently unrestricted. In such environments, an epenthetic high vowel is inserted between consonants, breaking the cluster, thus leading to a preferred syllable structure. It would appear, therefore, given the general lack of consonant clusters of three or more aGrossomprphemeob'ounda'ries^hthat a derivational constraint i s operating in certain grammaticallyndnducedhphonological environments where a high lax vowel i s inserted automatically. This high lax vowel is inserted by rule under the following conditions: (1) following a stem ending in a f i n a l double consonant cluster and before the preceding i n i t i a l consonant of the modifier; (2) between the f i n a l double consonant cluster of a modifier in an attributive clause and the following head noun beginning with a consonant in a noun0phrase. Story (1966: 24-6) has enumerated the grammatical environments as occurring in structures of: (1) noun plus Cpost-modifier I c l i t i c (2) noun possessive plus locative (3) verb attributive plus noun (4) verb plus auxiliary, as well as across morpheme boundaries of tone-bearing stem plus suffix(es), e.g. guxUx (f:(6m gux+x) s l t i "He's a slave" The epenthetic vowel agrees in roundness with the preceding consonant. A general rule for this epenthetic assimilation (which - 41 -Kinkade (p.c.) has suggested may be a quite general process in Tlingit) might be stated: EPENTHETIC , V [a round] / C # ASSIMILATION + high - tense a round ~~| An inserted high lax vowel agrees in roundness with a preceding consonant over a word boundary (since the derivational constraint operating w i l l not allow inserted vowels to violate morpheme structure conditions which operate generally in preventing contiguity of palatal and labiovelar sonorant forms). II.6 MORPHEME STRUCTURE CONDITIONS Within standard theory, morpheme structure conditions have been established as a part of the readjustment component to account for a speaker's knowledge of the lexical structure of items in the lexicon, and to characterize formally a possible morpheme of a language. According to Chomsky and Halle's (1968: 171) definition, lexical redundancy rules apply " s t r i c t l y within a single lexical entry and... simply f i l l n i n unspecified sequences of phonological matrices, without violating invariance. (Markedness theory has been an attempt to obviate the necessity for morpheme structure conditions. Universal marking rules would postulate the naturalness or unnaturalness of certain configurations in terms of features in a phonological matrix). Developed by Halle (1959), the original function of morpheme structure rules was to f i l l in the unspecified or blank features in - 42 -items in the dictionary representation. Stanley's (1967: 393) original summary of the function of morpheme structure rules was as follows: Thus, a f u l l set of morpheme structure rules for a language w i l l do two things: i t w i l l state, in terms of features, a l l constraints on what sequences of phonemes are possible in morphemes, and i t w i l l allow each morpheme to have a representation in which redundant feature values are omitted.-'-'' Stanley, pointing out that these blank features, along with positive and negative values, were acting as a third value, suggested the need for a f u l l y specified phonological matrix as well as stated sequence structure conditions. Schane (1973: 43) has described the function of morpheme structure conditions, including the definition of a conceivable morpheme, as follows: The segment redundancies and the sequence redundancies jointly form a set of morpheme structure conditions. The former define the set of possible phonemes in a language and the latter a set of possible morphemes -that i s , possible sequences of phonemes. A conceivable morpheme can now be defined as an arbitrary sequence of bundles of unspecified features which does not violate 17 The difference in theory focus between autonomous and systematic phonemics with regard to the specification of morpheme structure has been pointed out by Anderson (1974: 282) who states: "For the theories known collectively as 'taxonomic' phonemics, the primary focus of attention was not the relation between forms, but the autonomous structure of each individual form. Such theories were concerned to specify the regularities which obtain within the domain of the individual form, and to distinguish those aspects of the structure of forms that serve distinctively from those aspects of their.structure that are due to regularities of language." - 43 -any of the morpheme structure conditions of that language. 11.6.1 Redundancy Morpheme structure conditions thus reflect both segment redund-ancy, the marking of non-distinguishing features in a f u l l y specified phonological matrix, and sequence redundancy, the same across segments. Segment redundancy w i l l not be discussed here. However, there are some sequence structure constraints which can be discussed generally, as well as in terms of syllable structure, if-then, and negative conditions, as noted below. In Boas' (1917: 12-13) data, consonant clusters are rare stem i n i t i a l l y ; rare f i n a l l y in verb stems; and not common in noun stems. Because a l l stem-final consonants are devoiceu naturally, the f i r s t member of a consonant cluster across a morpheme boundary w i l l always be voiceless. Specifically, Story (1966: 14, 20) has noted that in spite of the large inventory of consonant phonemes, many occur rarely in affixes, while a limited number occur in function words. In terms of phoneme frequency in a lexical l i s t i n g , the most common CV(C) verb stem-initial phonemes (25%) are l i s t e d as /?, t, n, x, d/,' and the most common CV(C) noun stem-initial phonemes (25%) as It, h, k, x/. Five phonemes /y, d, w, n, x/ in a small text sample accounted for 50% of the consonants. 11.6.2 Constraints Constraints restricting the pattern of distribution and the com-binations of features are expressed by morpheme structure conditions. - 44 -In structuralist terms expressed allophonically and phone-tactically, these constraints are expressed in current generative theory by three kinds of morpheme structure conditions which may be in the form of syllable structure, if-then, or negative conditions (v. Stanley 1967). In T l i n g i t , an interesting phonological constraint on CVC verb. roots i s the non-occurrence of root i n i t i a l l a t e r i a l consonants with root f i n a l apical or laminal consonants, and of root i n i t i a l apicals with root f i n a l laminals. Thus, according to Story (1966: 71): When verb roots are of the shape CVC, a root i n i t i a l lateral consonant has not been found to occur with a root f i n a l apical or laminal consonant; a root i n i t i a l apical consonant has not been found to occur with a root f i n a l laminal consonant. Interestingly, there i s evidently in these roots a marked preference for the occurrence of root i n i t i a l apical with root f i n a l apical, for root i n i t i a l laminal with root f i n a l laminal, and for root i n i t i a l lateral with root f i n a l l a t e r a l . A further interesting feature of these CVC roots, the combination of phonetic and semantic function, has also been noted by Story (1966: 71) in that pairs of monosyllabic verb roots are "identical in i n i t i a l consonant, in close or open vowel, back or front; vowel, contrast in feature of f i n a l consonant, and occur in forms of similar semantic function." Negative constraints may be expressed by the following negative morpheme structure conditions: Negative Condition (a) Lateral Root ^ + [+ lateral] [ I n i t i a l (CVC Verb Root) (b) Apical Root ^ + I n i t i a l + anterior + coronal [ + coronal - lateral - anterior + coronal + + - 45 -Redundant.features are not specified here. Dots indicate features that would ordinarily be specified, e.g., [+ s y l l ] . In addition.], laryngeals, the glottal stop and fricative, may occur only stem i n i t i a l l y . A negative condition, a morpheme structure constraints which specifies a non-permitted sequence, w i l l indicate ( that laryngeal glide segments are unacceptable stem-finally, e.g. in formal terms.^ I Negative Condition Laryngeal ^ + [+ cons] [+ syll] Glide - cons + son + low + A further sequence structure constraint occurs in the extensor series, where an s-series extensor, for example, does not co-occur with a sibilant or lateral stem consonant. With two exceptions, an s-series extensor does not co-occur with sibilant stems. Story (1966: 85) has noted the co-occurrence restrictions occurring between the 18 In their definition of the three major class features as sonor-ance, vocality and consonance, Chomsky and Halle (1968: 301-3) have defined the laryngeals /h, ?/ on the basis of vocal tract shape as both sonorant and non-consonantal, and therefore to be classified as glides. However, Ladefoged (1971: 108-111), defining sonorance in auditory and acoustic terms, has argued that i t is counterintuitive to classify voiceless sounds as sonorant, and that laryngeal stop and fricative ought to be classified as true consonants. (Aitchison (1974: 10) concurs with reference to the behavior of h in Classical Greek in that: "It seems more plausible to regard h (as i t was regarded trad-itionally) as a true consonant rather than a glide." - 46 -extensor series and the stem consonants of a verb theme. A suggested negative condition for the non-occurrence of the s-series extensor with sibilant or lateral stem consonants i s : Negative Condition S-series ^ + Extensor + ant + cor + strid + exten ([+ syll]) + pi- cor ~I . . . Ft-ccor T J + strid 1 J + strid J (Strident sounds are marked by greater air turbulence than non-strident sounds. The feature strident is thus restricted to obstruents, con-tinuants and affricates, and can be a feature of nonvocalic laterals, v. Chomsky and Halle 1968: 329.) In addition (Story 1966: 14), no vowel clusters are permitted within a grammatical word. This could be indicated by a syllable structure condition of the stem. For example: Syllable + ([- syll]) [-syll] [+ syll] [- syll.]: 0[<- syll]) + Structure Condition To recapitulate: Lexical entries, which include as part of their information f u l l y specified matrices of underlying abstract representations of formatives, are stored in the lexicon of the syntactic component. Unordered mor-pheme structure conditions, applying before the phonological ruiLes, state the language-specific constraints on sequence structure and segment formation within the morpheme. However,,unlike Chomsky and Halle (1968: 382) who find the lexica l redundancy rules seemingly - 47 -'like ordinary phonological rules, in form and function', Brown distinguishes between redundancy and phonological rules by their 19 operations. As Brown (1969: 9) has pointed out: Many examples can be found of regularities which operate both within morphemes and across morpheme boundaries: we have seen some examples in the Lugisu data. A theory which refuses to account for this very general phenomenon, and by i t s refusal allows redundant processes to proliferate, needs modification.^0 19 According to Brown (1969: 16): "The redundancy rules are realization rules: they complete the phonetic specification of the output of the syntactic surface structure within the domain of the phonological word. The formal constraint on their function is that they can only add features. The phonological rules, on the other hand, are exclusively mutation rules. They are divided into two parts, the intra-word rules and the inter-word rules. The intra-word rules operate on the output from redundancy rules which is marked for entry to the phonological rules. They exist only to deal with complications in the phonology ... The inter-word rules, like the intra-word rules, are mutation rules. They specify the phonetic consequences of phonolog-i c a l words occurring in sequence. The formal distinction then between redundancy rules and phonological rules is that whereas redundancy rules add feature but perform no other operations, the phonological rules only perform operations other than simple additioning transformations." 20 With regard to other theories, Brown (1969: 5) has stated: "Non-generative analyses have accounted for data of this kind by allowing certain general rules to operate both within and between morpheme boundaries'* (see, for example, Hockett (1958: 131), Whorf (1956: 223) and the implicit reasoning in Bloomfield (1933: 133)), and by invoking the concept of the syllable (see, for example, Jesperson [sic] (1909: 59)> Bloomfield (1933: 121)), oasaprosbdicpphonologi'sts have always done (paperslin-Bazell retialifitl966caexempM'£y.dthis' approach) ." t .en restrained iroa applying across naorphsais V\>ao. - 48 -Brown has given examples of sequence rules which apply across and within morpheme boundaries, for example: ". . .we have only to look at the familiar English examples of voicing similarity in apse, caps, adze and cads and the constraints on the stop sequences that give us act, apt, cracked, capped but not *atk or *atp." II.6.3 Idiolectal Variation Idiolectal variation, involving a y-idiolect and a w-idiolect, have been noted by Naish (1966) and Story (1966: 31-32). These idiolectal variations may involve syllables or morphemes in each of the two idiolects. There i s , for example, a y-idiolect where the sequence /w + i / or /w + 1/ is not permitted across a morpheme boundary, and a w-idiolect where a sequence /uy/ or /Uy/ (where /y/ is syllable onset) is not permitted. In the y-idiolect, a negative condition w i l l specify that /w + 1/ or /w + 1/ may not occur across a morpheme boundary. In w-idiolects, s y l l a b l e - i n i t i a l /y/ w i l l not appear following a rounded back vowel. Examples: y-idiolect w-idiolect gloss du yid du wid "his son" qUyawAqa quwawAqa " i t was suggested" Except in pronounced y-idiolects, where the sequence /uy$/ may also occur across morpheme boundaries, certain constraints in both idiolects w i l l not allow /w/ to occur before / i / , or /y/ to occur after /u/. Thus, according to Story:(1966: 31): "Within the syllable and - 49 -within the morpheme, in the absence of variant forms, the sequences wi and uy are non-permitted in a l l idiolects." Following Brown (1969, 1971) who, using both Lugisu and English examples, demonstrated the advantages of allowing the predicted feature values established by redundancy rules to be carried over morpheme boundaries (and who, unlike Chomsky and Halle,,distinguishes between redundancy and phonological rules by their functions), we could therefore suggest valid morpheme structure conditions for both y- and w-idiolects. This idiolectal material could be handled in the lexicon in several ways. Firs t , and probably psychologically real to the native speaker of T l i n g i t , there would be, for each of the two idiolects, •'one set each of lexical entries for the affected sequences, with perhaps a general recognition of both sets to the speaker-hearer. Alternatively, the underlying abstract forms might have two phonetic realizations for the two idiolects, with a differing set of morpheme structure conditions for each idiolect. Third, as shown below in Condition 2B, the constraints could be stated in general terms, with phonological rules handling feature changes for each idiolect. Condition I, Parts A and B, illustrates the suggested morpheme structure constraints applying to two separate sets of lexical entries in and across morpheme boundaries, while Condition II shows the suggested constraints for an underlying abstract represent-- 50 -CONDITION I A. Negative Condition Y-idiolect I - cons + round [+ F.B.] + s y l l - back / ^ w + i / The sequence of /w + i / or /w + 1/ is not permitted across a morpheme boundary. Negative Condition W-idiolect 1— + s y l l +• back [+F.B.] $1 - cons - round fa u + $y / The sequence /u$y/ or /U$y/ i s not permitted across a morpheme boundary where y is syllable onset. CONDITION II A. If-Then Condition Strong-Y-idiolect If: [+ syll] + [- cons] Then: [+ back] + [- round] /u + y./ In a strong y-idiolect, a sequence of /u + y$/ or /U + y / may occur across a morpheme boundary. If-Then Condition If: [+ syll] i [- cons] / u w / Both idio- Then: [+ back] [+ round] lects (except strong y- If: [- cons] [+ syll] / y i / idiolect) I Then: [- round] [- back] In both idiolects (with the exception of a strong y-idiolect), glides dodnot contrast before / i / or after /u/. - 51 -ation with two sets of morpheme structure constraints, which also apply innand across morpheme boundaries. The f i r s t constraints are realized as negative (non-permissible sequence) conditions and the second as if-then (cause-result) conditions. It seems that these morpheme structure conditions, which appar-ently may cross morpheme boundaries, capture in a general way some of the structural limitations operating generally in Tlingit by illu s t r a t i n g underlying constraints which optimally prohibit certain sound sequences occurring in certain environments. The suggested Conditions are set out on the preceding page. II.7 AUFHEBUNG OR NEUTRALIZATION Traditionally regarded as the suspension of a bila t e r a l phonemic opposition in certain environments, neutralization has been defined by Troubetskoy as the suspension of distinctive opposition in a correlated pair of phonemes that differ by one feature, the unmarked member (in Prague School usage) normally appearing in the neutralized position. However, many linguists, including R.J. Gregg and Martinet, differentiate between neutralization, which results in an archiphoneme (of the representative of a neutralizable opposition, v. Troubetskoy 1969: 79), and Aufhebung which results in a suspension of distinctive . . . . . 21 opposition m certain environments. 21 'Archiphoneme' has been defined by Troubetskoy (1969: 79) as "the sum of distinctive properties that two phonemes have in common" and the representative of a neutralizable opposition. - 52 -In Tl i n g i t , several examples of an environmentally determined suspension of opposition exist. II.7.1 Neutralization in Generative Phonology In generative phonology, neutralization has been discussed under several different headings. In a perceptive discussion of redundancy, Stanley (1969: 40l64®2f)edefiriednoneik-indfofeneu£raliz-ation as occurring in an environment where "the value of the feature in the environment is determined by a sequential constraint." He has suggested somewhat inconclusively that aspects of redundancy pertaining to neutralization could be included in the morpheme structure rules. However, Schane (1968: 715-7-16), following Chomsky and Halle, would use the universal notion of markedness to obviate the problem of phonological neutralization. Unmarked features would represent the underlying form. Where, for example, in the case of und in German, voicing may be undetermined in the underlying representation, Schane would choose the unmarked feature of voicing (i.e., minus voice) to represent the f i n a l segment, so that: Just as the notion of archiphoneme was indispensable for dealing with phonetic realization, the concept of marked-ness is needed for handling phonemic realization. On the other hand, Schane (1973: 59) would treat neutralization as one among several practical phonological processes, including those of assimilation, syllable structure (deletion, epenthesis and - 53 -coalescence), weakening (syncope, apocope) and strengthening (vowel shift, vowel reduction, diphthongization), and neutralization. In a phonological context, neutralization is thus defined by Schane as: ... a process whereby phonological distinctions are reduced in a particular environment. Hence, segments which contrast in one environment have the same representation in the environment of neutralization. In this section, neutralization as a topic w i l l be treated in the traditional manner as a part of the phonological description. II.7.2 Aufhebung in Tlingit In T l i n g i t , the archiphonemes represented in the phonology would appear to be typical of Troupetskoy's (1969: 80) Case II in that: The representative of the archiphoneme is identical with the realization of one of the opposition members, the choice of the archiphoneme being conditioned 'externally 1. This is possible only in cases where the neturalization of a neutralizable opposition depends on the proximity of some particular phoneme. The opposition member that "bears a closer resemblance or relation" to such a neighbouring phoneme, or is even identical with i t , becomes the repre-sentative of the archiphoneme. This is illustrated in modification (colour-rounding), where there is a suspension of opposition between simple velars and uvulars and rounded v velars and uvulars in three positions: next to rounded back vowels, before close front vowels, and following a rounded consonant in a double consonant cluster. Story (1966: 32) has noted that simple velars and uvulars do not contrast with rounded velars and uvulars "following either a labialized velar or uvular or labiovelar semi-vowel nr a seque:c.C3 of - 54 -or a sequence of /u/ and a velar or uvular." In glides, there i s a general suspension of opposition before close front vowels, and after rounded close back vowels. In manner, opposition of plain and aspir-ated stops and affricates i s suspended at syllable coda, a case of partial complementation (v. Naish 1966: 21, Story 1966: 15). II.8 TONE Tone in Tlingit i s phonemic, and is marked on stem vowels which have a high or low tone, thus furnishing a total of sixteen distinctive 23 syllables. According to Velten (1939: 66)': Tlingit syllables may have high, low, or indifferent pitch. The difference between high and low tone may have a semantic value, as in eq (high tone) 'beach' and eq (low tone) 'cooper', xat 'salmon' and xat 'root'. The distinction between low and indifferent pitch, however, is less essential. Stress, though generally associated with high tone, is not 22 In Shuswap, an Interior Sallish language, Kuipers (1974: 34) dis-tinguishes between 'automatic' and 'inherent' rounding of con-sonants. A l l consonants which are correlated for the rounding features are automatically rounded before and after u, as well as contiguous to rounded consonants (or separated only by schwa). 23 According to Naish (1966: 26), a tone group consists of the tonic syllable, always word f i n a l in a tone group, plus optional syllable(s). However, not a l l Tlingit dialects have tone. Historically, according to Krauss (Kinkade, p . c ) : "... Eyak and PCA [Pacific Coast Athapaskan], instead of de-veloping tone l i k e so much Ath, have instead kept glottal modification of the stem-vowel. Southernmost Tl i n g i t (Ton-gass, under Tsimshian influence) has also done the same, the rest of Tlingit having developed tone instead." - 55 -phonemic. According to Velten (1944: 168), neither quantity nor stress is phonemic, while a non-distinctive difference in lower and neutral tones is apparently based on sentence rhythm. Story (1966: 29), who has identified four contrastive tone patterns, notes: The tonic syllable is generally the stressed syllable though syllables are l i t t l e differentiated by stress, and in tone patterns 1 and 2, the syllable immediately before the tonic syllable, when i t contains a long vowel, may be equally stressed with the tonic syllable. Stem tone is not absolutely stable, since high tone may be lost in certain compounds. Boas (1917: 12) has given the following examples: xuts-nuwu^ "bear's fort" xutsnuwu "Bear Fort" (place name) ?ak w " l i t t l e lake" ?akwqwan "Little-Lake" (tribe) Naish (1966: 29) suggests that a free-form stem with tonic syllable has a variant form with no tonic syllable as the f i r s t element of a compound. Velten (1939: 70) states, in reference to his word l i s t , that nouns may change their pitch when entering into compounds. Velten (1944: 168) notes as well that the opposition of high and low tone is neutralized in certain suffixes, e.g., -de, - y l , yln. According to Naish (1966: 28), certain function words have no associated tone. Tone distinction also differentiates members of a verbal paradigm. Story (1966: 63-5) has postulated a number of hitherto unmentioned patterns of the verbal paradigm. Variable verb stems, always of CV(C) shape, may be inflected by the opposition of high vs. low tone and of long vs.' short stem vowel. For example, with stem-final consonant, Stem Type 2, probably the most common, is - 56 -postulated to occur with four inflected stem forms and the following opposition of tone: 1 2 3 4 Cv:C Cv:C Cv:C Cv C Stem form 1 (with affixes) corresponds to the English pluperfect 2 to the future, 3 in part to a similitudinal clause "as" + perfect, and 4 to the imperative. As identified by Story (1966: 29), there are four contrastive tone patterns in normal tone groups: three with high tone on the fi n a l , penult, or antepenult syllable; and one with low tone on the f i n a l syllable of the tone group. According to Story (1966: 29): The pitch of the pre-contour to the tone group (that i s , the pitch of those syllables^ i f any, preceding the tonic syllable), is on a low-mid level. The f i r s t syllable of the pre-contour is slightly higher than any following syllables in the pre-contour. The post-contour, i f any, following a high tone f a l l s to low i f the vowel of the syllable following the tonic syllable i s long; the pitch of the syllable following the tonic syllable is maintained at a high level i f the vowel is short. Tone is not inherent in a l l morphemes. For example, from Boas' data (1917: 17, 84, 87-91), two suffixes of a neutral tone, - y i with a variety of functions including nominal possession and verb nominal-ization, and -yin indicating past tense, when affixed to a stem, take the tone opposite the stem vowel. Minimal pairs indicate the contrastive differences in high and low tones. High tone is marked by acute accent, low by grave accent. The following data, il l u s t r a t i n g this tonal opposition, are taken from Boas (1917: 11-12).. - 57 -1 * ta "king salmon" 5 r ta '/board" xat "root" xat "salmon" t i l "scar" t i l "shoe" V xas "to cut" (past) V xas '(future) qin "to f l y " (past)' qin (future) han "to stand" (past) han (future) In addition to his examples previously cited, Velten (1944: 168) also cites: tu "inside, interior" tu "mind, thought". II.9 BORROWING - PHONOLOGICAL INLOANS As a means of tracing ^ ofitjaCC-i'-pPeliistory anddesteablishing areas of linguistic diffusion, borrowing has been discussed i n some detail by Haas (1969: 80) who has suggested that " words are as important in tracing historical contacts as cognates are in tracing historical origins." In addition, lexical borrowing offers insights into phonological adaptation when a loanword is adopted from the donor language (DL) to the recipient language (RL). Since Tlingit phonology lacks bilabials and labiodentals as well as the voiced alveolar frictionless continuant /r/ and the vocalic lateral / l / , certain changes are made in adaptation. Although as Swanton (1908: 472) noted, in studying the Sitka dialect with material from Wrangell and Yakutat, the use of m occurs "in a few words imitat-ing natural sounds and in words introduced from other stocks such as Tsimshian; 1, however, is' usually transliterated as n." - 58 -More recently, Dauenhauer (1976: 10) has concurred: The Tli n g i t language has no bilabials or labio-dentals, no p, b, f or v. M appears as a dialect variation of w. Also there is no r, and 1 only appears as a dialect variation of n. There are, however, several examples in Tlingit of inloans from Tsimshian, Athapaskan, English, Russian and French, where loanword phonemes or features have been changed to conform with the existing phonological system. Examples below (Boas 1917: 10, 1-3; Dauenhauer 1976: 10, 4-6) show the adaptation of a foreign labial to a Tl i n g i t labiovelar, since labials are noticeably lacking in T l i n g i t . Donor Tlingit Language Gloss Original Borrowing 1. Tsimshian "Tsimshian" cem-sian c'ucxAn 2. Northern Athapaskan "mountain sheep" tXme t Aw£ 3. Northern Athapaskan "caribou" mAd? WAClX 4. English "machine" math in e washeen 5. English "watchman" 'watchman waachwaan 6. English "commissioner" commissioner kawxshan In terms of major class features, the change is perhaps best characterized as + son + cor - cons + back That class of consonantal sonorant segments marked for the feature coronal, i.e. nasal /m/, backs, rounds, and loses the feature of consonantality when borrowed. Further examples from Dauenhauer (1976: 10) i l l u s t r a t e the ^  adaptation of foreign b i l a b i a l sounds into Tlingit as velars (Dauen-- 59 -hauer's orthography and tone markings are followed in*his foreign examples where double letters indicate long vowels). Tlingit Borrowing gee waa wadooshka gweelats ganaashish nakwneit nadaakw g" w w g g According to Dauenhauer (1976:' OiO) also " . . . i n English words spoken with a Tl i n g i t accent, bilabials are commonly realized as velars. Thus 'tip' becomes 'tick' and 'helicopter' becomes 'helicockter'." Further borrowings (Dauenhauer 1976: 10, 12; Story 1966) i l l u -strate the typical adaptation of foreign liquids to nasal or syllabic segment (tone indicated where given). However, I. Brenzinger (p.c.) notes that for orthographic consistency in Russian examples 8 and 16, the palatalized t of batyushka would be indicated as batushka. Donor Language 7. Russian 8. Russian 9. EEnglish 10. EEnglish 11. French 12. French Russian French English Gloss "beer" "priest" "peanuts" "molasses" "priest" "table" r p :bi p b p m Original 'pivo batyushka 'pi:nats me'lffisoz le'p^et-J la'tab} Tlingit Tlingit Tlingit - 60 -13. 14. 15. 16. Donor Language English English English Russian Gloss "dollar" "gold" "rum" krest ("cross") Tlin g i t Borrowing daanaa goon naaw kaneist Characterizing the change, i t would appear than an adapted liquid retains the major class features of sonorance and consonance in the environment of a consonant, gaining a syllabic feature in the environment of a vowel (depending on the syllable placement). In addition, a further rule to do with tone may be deduced from the examples above. Tone Transfer 1 stress Tlingit + high tone Donor languageepriiiiary stress becomes high tone in T l i n g i t . Illustrative of phonemic adaptation is the evidence of in-loan borrowing, from which a few general rules may be deduced: Donor Language Northern Athapaskan, English English RRussian French Adaptation Source (1) DL m >-Tl w Boas (1917) (2) DL|r-J > ft n <vfcory ( l g U ) (3) DL Cr > T l Cn ffiauenhauer (1976) ( 4 ) ^ f b j > ^ Dauenhauer (1976) - 61 -11.10 DIALECTAL VARIATION Outside of the idio l e c t a l variation (mentioned in section II.6), minor differences in dialect also occur. For example, the ethno-grapher Swanton (1911: 163) noted, and classified geographically, 24 certain dialectal differences observed: Although each town appears to have had certain dialectic peculiarities, i t would appear that the language nowhere varied very widely and that differences were mainly con-fined to the different arrangement and handling of particles; the lexical changes being very few and the structure practically uniform. The greatest divergence is said to exist between the Yakutat people on the one hand and the people of Wrangell and the other southern towns on the other hand - the speech at Sitka, Huna, Chilkat, Auk, Taku and Killisnoo being intermediate. Though very l i t t l e has been published, f i e l d workers appear to have covered the major areas of divergence. There are, evidently, two or three dialect areas. According to Story (1966: 11), who has described the speech of the Angoon, or central, dialect: There are dialectal differences but these appear to affect extensor forms, close vowels and intonational features for the most part, and divergence between dialects is small. Two papers by Velten (1939, 1944) are descriptive of the southern dialect; Boas' work is in the northern dialect. 24 Interestingly, with regard to Swanton's s k i l l s as a Northwest Coast ethnographer, Levine (1973: 28) has noted: "Research on the Haida language and non-linguistic .-..culture was conducted in the f i r s t part of the twentieth century by John Swanton, whose s k i l l as an ethnographer rivals that of Franz Boas, the founder of ethnographic methodology in North America. In addition to a monograph on Haida, which is without question the most detailed, accurate and objective work on Haida ever written, Swanton presented a collection of myths and a grammar, based on the texts of those myths." - 62 -However, i t has not been ascertained definitely whether there are two or three, or more, dialect areas in existence. Swanton (1911) and Story (1966) evidently subscribe to three: northern, southern and central. The ethnographer de Laguna (1960) and Velten (1944: 168 n. 1), following Boas, mention two. According to Pinnow (1970: 26-7), Buschmann (1855, 1857) and Swanton (1908, 1909, 1911) worked in Chilkat (N) and Sitka (N); Swanton also in Yakutat (N) and Stikine (S). Krause (1885) dealt mostly with Chilkat and Sitka. Dall (1877) touched upon Taku (N), Stikine, Sitka and Tongas (S). The most reliable work, however, i s that of Boas (1917) in Chilkat, Velten (1939, 1944) in Klawak (which Velten asserts as the most southern of dialects), and that of Naish (1966) and Story (1966) in the Angoon or central dialectt Velten (1939: 65) has perhaps noted most thoroughly the slight variation in the northern Chilkat and southern Klawak dialects: The main peculiarities of the southern dialect consist in the transformation of certain verb classifiers and in the tendency to open closed vowels of high pitch and to close the open vowels a, e, I, U, especially when they have low pitch. Thus, along with a few minor morphological and lexical changes, the major differences appear to be tonal and vocalic. II.10.1 Vocalic and Tonal Variation According to Velten (1939: 70 n. 1), dialects generally agree on pitch, though with many differences in vowel quality. Velten notes, however, that - 63 -...the rules concerning pitch after certain nouns and prefixes, as set forth by Boas (75-7), do not always seem to hold good in the southern dialect. Although Velten provides no examples for these tonal differences, Boas' (1917: 75-7) northern dialect material shows that certain of the 'locative prefixes require an open vowel and low pitch, when occurring with the past tense form of the verb stem. In addition, some of Velten's (1939: 71-72) examples show a dialectal morphophonemic alternation of e^A (Boas 1917: 92., 104), e.g.: Velten Boas (1917: 92, 104) -ye particle of place yA yex "likeness, in like manner" yAx ' 5 5 kenen 11 to jump" kAn, ken Other sound change has been noted in Velten (1939: 72-73): southern ^ a;. anpc: northern xan anger kan kwgwA- (future prefix) gUgA-taq "inland" d/\q Lexical variation other than sound change has not been noted in Velten's data. Close vowels of low pitch occur more often in the southern dialect. In addition, as previously noted, there is a tendency, in contrast with the northern dialects, for the closed vowels of high pitch to open and for the open vowels, especially those with low pitch, to close. Examples from Velten (1939: 71, 72; 1944: 178) and - 64 -Boas (1917: 107, 123-125) showing vocalic and/or tonal change are: Velten Boas yl t "son" y i t te* "stone, rock" te taq "inland" dXq yax "border, edge, yXx shore" Velten notes that the form /dAq/ may appear only without a prefix in southern speech. Because features of tense vs. lax, or even those suggested by Halle (1974) of [+ advanced tongue root] and [+ constricted pharynx], do not adequately characterize the change from open to close in southern vowels, the suggested feature £or reflecting this change is that of length, e.g.: i ^ [-a long] Southern Dialect V a high tone As a feature of the southern dialect, vowels with high tone become short; low-toned vowels lengthen. That this is the right feature characterization seems confirmed by Naish (1966: 22) who suggests that there is ...some contrast in phonetic duration between vowels... only maintained in tonic syllables. II.10.2 Classifiers As previously noted, an important difference in the extensor forms or classifiers i s evidenced dialectally. 'Classifier' is the - 65 -traditional term used by Boas (1917), Swanton (1911), Sapir (1915), Velten (1939), Krauss (1969) and Pinnow (1970). Krauss (1969: 81 n. l ) , noting that the term 'cl a s s i f i e r ' is a "blatant misnomer", defines classifiers as a "set of morphological segments occurring in prefixal position immediately preceding the stem in the Athapaskan, Eyak and Tlingit verb." 'Extensor' is the term devised by Naish (1966) and Story (1966) to express the derivational and inflectional element which, when affixed to a verb stem, comprises the major component of the verb theme. Of the four series of extensors, one of which w i l l precede the verb stem, one series, in the northern and central dialects, i s of the syllable structure CV, while in the southern dialects these forms are mainly consonantal. Velten (1939: 69-70) has described these changes in syllable structure. The northern classifiers l i - , s i , s i - , d i - , A i - , dzi-, j i -appear in southern speech as T—, S , S—, L —, 7T , C , C—• In addition, there i s an apparent loss of voicing in the southern stop and affricate forms. As a result of these changes in syllable structure, Velten (1944: 70) notes that: ...the approximate distinction between definite and i n -definite classifiers traceable in the northern dialects... has, to a great extent, broken down in southern speech. It must be remembered, however, that the original functions of the classifiers are largely effaced in the whole Tli n g i t language, and that in the consciousness of the native speaker verb stem and cl a s s i f i e r form an indivisible unit. - 66 -Velten (1939: 69) cites the following verbs in the southern dialect (the northern equivalents are given in brackets): V V wU-l-sat (wU-li-srat) "he was carrying" wu-s-ha (wU-si-ha) "he missed him" wU-s-geq (wU-si-geq) "he was stingy" wU-t-9iax (wU-di—?tax) " i t was grey" wU-A-gUt (wU-Ai-gUt) "he was going" qiy-c - t i (qiy-dzi-ti) "thou art born" -. v •. K \ wU-c-xix (wU-ji-xix) "he ran" II.10.3 Idiolectal Variation as a Dialect Feature Idiolectal variation (v. II.6.3) has been noted by Story as a dialect feature. For example, Story (1966: 31))suggests: In IPA terms i t may be said that regressive palatal-ization and progressive labialization occur. When these conflict, which takes precedence, or the degree to which either i s operative, depends on idiolect. Idiolects in which regressive palatalization i s dom-inant are called y-idiolects (a y-idiolect i s standard in the central area) and idiolects in which progressive labialization i s dominant are called w-idiolects. Thus, one of the more interesting dialect features i s the alter-nation next to high vowels of semi-vowels which, depending on relative glide strength in an area, results in either regressive palatalization (y-idiolect), or progressive labialization (w-idiolect). The y-idiolect, standard in the central area, was also evidenced in 'the speech of Boas' (1917: 16) northern Chilkat informant. In treating dialect variation as a phonological process, an - 67 -alternation in form, rather than as an inherent constraint (a much stronger view), idiolectal variation may be handled by two phono-logical rules. If these idiolectal variations are to be regarded as P-rules and thus handled under the heading of dialect variation, the suggested rule for palatalization wand*, a further rule for labialization follow: Rule: Y-idiolect (Central and Northern Area) Regressive Palatalization - cons - s y l l [- round] / V [- round] Rule: W-idiolect Progressive Labialization - cons - s y l l [+ round] / [+ round] II.10.4 Other Features A further feature of the southern dialects cited by Velten (1939: 66 n. 1) is an apparent delabialization, which often occurs word-finally after vowels. For example, Velten (1939: 72, 74; Boas 1917: 126, 128) noted: kux "marten" kux w southern northern naq "devil f i s h " naq w The suggested optional ruHe*- for delabialization in the southern dialect, iapplying to tone-bearing stems, i s : DELABIALIZATION (optional) C > ' [- round] / # SOUTHERN DIALECT A consonant may unround optionaiiylword f i n a l l y . - 68 -For the Angoon or Central dialect, Story (1966: 33), however, has suggested a general rule applying to stem-final consonants and consonantal suffixes where "stem-final velar or uvular, or velar or uvular constituting a suffix, [which are] labialized without a following suffix, are simple when preceding a suffix with i n i t i a l vowel". Examples given by Story (1966: 34) are: xuw "blanket" du xiiwu "his blanket" xuwx "blankets" du xuwxu "his blankets" yag w "boat" du yagu "his boat" yag wk " l i t t l e boat" du yag wku "his l i t t l e boat" However, as Kinkade (p.c.) has pointed out, this apparent neutralization (of consonant preceding a rounded vowel) may be a kind of orthographic self-deception since, phonetically speaking, the labialization of vowels w i l l necessarily affect consonant articulation. As noted by Kuipers (1974: 34) for Shuswap, a back consonant contiguous to a rounded consonant or vowel w i l l most commonly show some rounding whether or not indicated orthographically. As indicated by Kinkade ( p . c ) , neutralization of this kind may frequently and mistakenly be referred to as delabialization. However, in contrast, for the northern dialect, Boas' (1917: 17) example shows rounding of an affix following a rounded consonant: guxw "slave" duis gux wx wu "his father's slave" (<du-is gux w-x-yi) - 69 -II. 11 SUMMARY Using transcribed material from Boas (1917), Velten (1939, 1944), and Naish (1966) and Story (1966), who used respectively pre-phonemic, phonemic and adapted tagmemic models, this Chapter has examined in some detail aspects of Tli n g i t phonology generally within the confines of transformational generative theory. Under various section headings, we have dealt with Tl i n g i t phono-logy using the methodology of generative phonology. At the level of observational adequacy we have found that generative theory accounts satisfactorily for the data, using the proposed phonemic modifications of Schane (1971), though perhaps certain questions are raisedaand not answered in examining morpheme structure conditions. However, having examined the phonological structure in some detail, the next Chapter w i l l proceed to a discussion of phonological processes. - 70 -III.O PHONOLOGICAL PROCESSES This Chapter, based exclusively on Boas' (1917) Chilkat data, w i l l cover a few of the most simple and general phonological pro-cesses in Tlingit within a transformational generative framework considered to include a theory of natural phonology. (It should be noted that the rather extensive morphophonemic complexities of the inflectional aspects of the verb word, recently explained in Story (1966), w i l l not be touched upon.) For, as Dinnsen (1974: 29) has described the goal of linguists: The task of descriptive linguistics i s to provide a correct characterization of natural language. We, as linguists, are therefore concerned with proposing a theoretical model which describes natural language and only natural language. Chomsky and Halle (1968) offered generative phonological theory as the most comprehensively defined model of natural language; and since that time much work has concerned i t s e l f with elaborating on that model. The phonological processes described, such as f i n a l consonant devoicing, intervocalic voicing, and assimilation, w i l l , as discussed by Schane (1973) and others, be simple and natural, and the rules w i l l be unrestricted in application. Although the distinction w i l l not be made here, natural processes, which are innate and universal, have been characterized by Vennemann (1972) as allophonic or phonemic in effect, while natural rules may be considered to be morphophonemic or morphological (cf. Rudes 1976). Although according to Boas (1917: 14) no phonetic processes 'occur consistently in Tlingit because of definite sound combinations, - 71 -phonetic change results from contact of the stem with certain affixes or endings, as w i l l be illustrated throughout i n a number of simple and universal processes which occur.''' While Schane (1972) has suggested that natural or context-free, or universal, rules w i l l f a l l mainly into the three categories of assimilation, preferred syllable structure, and maximum differen-tiation, the phonological processes which have been especially noted in Tlingit are f i n a l consonant devoicing (a natural process, Stampe 1969) , pre-vocalic stem voicing before suffixes (an assimilation process), yod deletion (a syllable structure process), and an assimilation process which includes labialization and a feature to be 2 here interpreted under the heading of vowel harmony. 1 Glottal insertion i s apparently an example of an exceptionless low level phonetic process which occurs with the pre-glottalization of stem-initial vowels. See Boas (1917: 14) where: " I n i t i a l vowels open with a glottal closure". Example: 0 — > ? / # V ?ukw "to b o i l " wu±?ugun " i t has boiled" See also Story and Naish (1973: 264) where':' "... no Tlingit verb stem begins with just a vowel." 2 It should be noted that, although this analysis includes underlying yod deletion and vowel harmony assimilation processes, M. Dale Kinkade (personal communication) has pointed, out that a simpler alternative, and perhaps more l i k e l y , hypothesis would be that of intervocalic glide insertion with vocalic assimil-ation. Kuipers (1967: 36) noted a kind of progressive vowel assimilation in Squamish, perhaps i l l u s t r a t i v e of the kind of process Kinkade suggests. Since this explanation of glide insertion and vowel assimilation i s a highly plausible one, examples of this alternative analysis w i l l be noted where appropriate in the data. - 72 -In order to cla r i f y the basic assumptions of this paper, how-ever, with respect to phonological theory, a brief recapitulation of the transformational generative model w i l l be given f i r s t , along with recent modifications to the theory in the f i e l d of natural phonology. I I I . l STANDARD TRANSFORMATIONAL GENERATIVE THEORY In the standard theory of the 1960's expounded prior to Chomsky and Halle (1968) in The Sound Patterns of English, the phonological component of the grammar was said to consist of morpheme structure rules and phonological rules, the latter accounting for morphophonemic alternations in form in the phonetic representations (v. Lightner 1971: 499-564, Anderson 1974). In theory, the phonetic representations were considered to be segmentally discrete bundles of distinctive features, while the ordered or partially ordered phonological rules were given formal and precise status. Phonological descriptions were evaluated simply by counting the number of features in each, with the lowest in number being the most highly valued. After Chomsky's (1965) Aspects model, the phonological compon-ent in a generative grammar, like the semantic component, was con-sidered to be interpretive of the central syntactic component. Exceptions to phonological rules were marked in the lexicon, and the alpha convention to assist with problems of assimilation and dissimilation was recognized. - 73 -Further formal refinements to the theory in rule and in repre-sentation, though perhaps not widely welcomed, were suggested as the sixties .-progressed: Bach's (1968) neighbourhood convention, L. Anderson's (1967) left-to-right syllabic cycle, McCawley's (1969) iterative rules, and, most important theoretically, Kiparsky's (1968) investigation into the abstractness of underlying represent-ations . As suggested by Dinnsen (1974: 29): Elaborations on the Chomsky-Halle theory of phonology (henceforth, the 'standard theory') have been of two types. On the one hand, i t has been argued that a given descriptive device may be too powerful, in that i t permits statements which cannot be supported by evidence from any natural language. In such cases, a more constrained descriptive device is offered. On the other hand, some elaborations on the standard theory have involved adding a descriptive device which i n -creases descriptive power. In these cases, i t i s argued that there is some fact about natural language which must be accounted for, and which escapes a proper characterization in the standard theory. Major issues in rule ordering were raised in the late sixties and early seventies by Chafe's (1968) persistent '("anywhere") rules, Chomsky and Halle's (1968) marking conventions and linking rules, Lightner's (1968) minor rules, Kiparsky's (1968, 1971) feeding and bleeding relationships, and opaque and transparent rules, Kisseberth's (1970, 1971, 1972) derivational constraints, and cy c l i c a l segmental rules, S. Anderson's (1972) local ordering, Norman's (1972) i n s u f f i -ciency of local ordering, and Koutsoudas, Sanders and Noll's (1974) intrinsic rule order. Problems in underlying feature representation arose through an interest in Stampe's (1969) natural processes and - 74 -natural phonology, Schane's (1968) non-uniqueness of phonological representations, and Vennemann's (1972) natural generative phonology based in part on the serious acceptance of Postal's (1968) Natural-ness Condition. Therefore, based on constraints of, or extensions to, generative theory the major phonological issues of rule order-ing and of underlying representation appeared to dominate linguistic discussion un t i l the mid 1970's. Although i t has been suggested by Dinnsen that further major issues in innovations to the standard theory include phonological conspiracies (Kisseberth 1970), surface-phonetic constraints (Shibatani 1973), and cyclic segmental rules (Kisseberth 1972) , i t seems that stemming directly from the aforementioned issues of rule order and abstract versus concrete representation 'there has arisen a major theme of naturalness in phonology. III.2 NATURALNESS IN PHONOLOGY Arising partly from a recurring interest in the hierarchy of phonological development put forth in the implicational laws of Jakobson's (1940) Kindersprache and continued research in child language acquisition, as well as from an ongoing search for tighter phonological constraints in standard transformational generative theory, naturalness in phonology in the guise of natural segment, natural class, natural system, natural process or natural rules has been a recent goal of linguistic theory. According to Stampe (1969: 540), however, Jakobson's implicational laws, and the markedness - 75 -theory of Chomsky and Halle, unlike Stampe's (1969) theory of natural phonology, are merely a subset of context-free phonological processes. Thus, according to Bach and Harms (1972: 5): A number of linguists have begun to emphasize the serious limitations i n this account of phonological [classical generative phonology] theory. The main criticism has been directed at the lack of sub-stantive assumptions about the context of rules, the particular generalizations that are available for the human (as opposed to the dolphin) communi-cator and so on. Under the headings of "markedness", "naturalness" or "archetypal rules" linguists have begun to search for tighter constraints on phono-logical systems and rules. Characterized by a linguistic criterion of simplicity, natural-ness in phonology has been defined as a search for phonetic plaus-i b i l i t y (v. Hyman 1975, Schane 1972), although Bach and Harms (1972) and Skousen (1972) have recently c r i t i c i z e d naturalness conditions for being essentially diachronic, and subject to the process of de-naturalization through time. While opponents to naturalness may raise interesting points, i t seems important to allow proponents of naturalness an adequate synchronic forum, remembering as well the dictum of Saussure (1915/1959: 99) that ... each language in practice forms a unit of study, and we are induced by force of circumstances to con-sider i t alternately from the historical and static viewpoints. Above a l l else, we must never forget that this unit i s superficial in theory, whereas the diver-sity of idioms hides a profound unity. Whichever way we look at studying a language, we must put each fact in i t s own class and not confuse the two methods. Archetypal rules (cf. Foley 1968) w i l l not be discussed in this paper. Chomsky and Halle's (1968) markedness, Stampe's (1969) theory - 76 -of natural phonology, and Vennemann's (1972) natural generative phonology w i l l be discussed briefly below. III.2.1 Markedness Markedness theory has been the recent proposal of Chomsky and Halle to incorporate within transformational generative phonology the intrinsic feature content of a universal sound inventory in a natural manner. To rectify earlier attempts at capturing "naturalness" with the notion of natural class, Chomsky and Halle (1968: 400) noted that The entire discussion of phonology in this book suffers from a fundamental theoretical inadequacy. Although we do not know how to remedy i t f u l l y , we feel that the outlines of a solution can be sketched, at least in part. The problem is that our approach to features, to rules, and to evaluation has been overly formal. They therefore posited the Praguian concept of a set of universal feature values in thefform of 'markedi'i versus 'unmarked' segments. According to Anderson (1974), these universal marking conven-tions, intended as a substitute for language-specific morpheme-structure conditions, are used to interpret lexical matrices to the desired 'natural' configurations, while the accompanying 'linking' rules applying to these derivations adjust feature values in an attempt to characterize naturalness within the generative framework. Marking conventions have been c r i t i c i z e d , however. While approving the move to universalize phonological theory by removing i t s "arbitrary and formal character", Anderson (1974: 291S2) has - 77 -suggested that marking conventions are in relevant respects "simply a version of Stanley's, with universality posited", and that these conventions s t i l l do not capture naturally certain universal phono-logical processes that occur, such as voicing assimilation in consonant clusters. Marking conventions have been used in this paper as a notational abbreviation characterizing features and rules throughout. III.2.2 Natural Phonology The theory of natural phonology, advanced by Stampe (1969) since 1965, is based on a postulated innate system of a complete sefctof phonological processes in which the. residual l e f t by linguistic experience is the result of a number of mergers of a "potential phonological opposition into that member of the opposition which least tries the restrictions of the human speech capacity." As Stampe (1969: 444) has stated: I assume, then, that in i t s language-innocent state, the innate phonological system expresses the f u l l system of restrictions of speech: a f u l l set of phonological processes, unlimited and unordered. The most extreme processes are usually observable only in infancy: unstressed syllables are deleted, clusters and coarticu-lations are simplified, obstruents become lax stops, Unguals become coronals, vowels merge to a. According to Stampe's theory, these contradictory and apparently binary phonetic restrictions impinging on the phonological processes may be limited in three ways: by suppression, by partial suppression, or by ordered rule application. - 78 -For example, the suppression of a natural process would ensure the appearance of voiced obstruents in a l l contexts though obstruents are, through oral constriction, by nature voiceless. Partial suppression, or limitation, of a process would allow voicing of obstruents in voiced contexts, for example intervocalically through assimilation. Ordered application has reference to changes which result from a different ordering of phonological processes. For example, in the case of obstruents, the devoicing process may, through ordered application, either precede or follow intervocalic voicing. Stampe's system of natural phonology may be compared with Vennemann's (1972) natural generative phonology in that no distinction is made between redundancy rules (morpheme -structure conditions) and phonological rules. Instead, a difference occurs between morpho-phonemic and phonolggical (phonemic) rules, termed respectively phonological rules and phonological processes as noted by Rudes (1976: 142 n. 1) where ... within the framework of natural phonology proposed by David Stampe (1973) [where] phonological process i s used in the sense of our phonological rule or process and phonological rule is used in the sense of our mprphophonemic rule.... III.2.3 Natural Generative Phonology Natural generative grammar has been defined by Vennemann (1972: 110) as - 79 -... a generative theory of language which is character-ized primarily by the absence of a provision for extrinsic rule ordering. The goal of this theory is to establish a system of universal constraints which is powerful enough to free the linguist analyzing a language to construct a grammar which directly models the mental representation of the generative and analytic capacity underlying that language. Arising from efforts to constrain or limit the power of transformational generative phonology (TGP), a theory of natural generative phonology (NGP), based on intrinsic rule order (cf. Koutsoudas, Sanders and Noll 1974), the strong naturalness con-dition (Postal 1968), and the basic category principle, has been developed from Vennemann (1972), and the recent work of Hooper (1975). Differences between transformational generative phonology and natural generative phonology have been discussed by Vennemann (1972: 110-11): In morphology, the constraint that a l l rules are unordered (a term which I w i l l use with 'random sequentially ordered' or 'applying whenever a rule's structure description i s met') leads to a number of consequences which make this portion of a grammar look radically different from a TG phono-logy. For example, there is no difference between lexical redundancy rules and phonological rules (NG phonology shares this property with Stampe's 'natural phonology') ... Instead, there i s a dis-tinction between phonological rules (allophonic and phonemic rules) on the, one hand and morphb-phonemic, morphological, and sandhi rules on the other. Recent articles in Glossa of Hooper and Terrell (1976) and Rudes (1976) have discussed stress assignment, and lexical represent-ation within this theory. According to Hooper and Terrell (1976: 65): - 80 -NGP differs from TGP in that NGP has stronger con-straints on abstractness. In particular, the rules of the grammar are not extrinsically ordered, rather aarule applies each and every time i t s structural description i s met... In discussing natural processes we w i l l , after some discussion, adhere to Vennemann's theory of natural generative phonology in the acceptance of unordered rules. III.3 CONSONANT DEVOICING In Tl i n g i t , voiced obstruents are devoiced word f i n a l l y or at morpheme boundary as shown below. Since there are no examples of consonants with word-terminal voicing, and because no underlying voiced consonants are posited word-finally, this appears to be an exceptionless rule-an example^of. one^of-Stampe-' s' innate and, in this context, unsuppressed phonological processes in which obstruents are devoiced through the natural process of oral constriction (Stampe 1969: 443, 445). The following noun and verb stems il l u s t r a t e obstruent devoicing in word-final position, especially in these rare stems ending in consonant clusters. (Data from Boas ^1917: 13 , whose orthography is given a current 4Amerieanist transcription. Item 14, p. 81 via Swanton 1909, v. Boas 1917:7). - 81 -Noun Stems 1. SAXt "devil's club" 9. cUckw "bird" 2. \ ?ixt' "shaman" 10. •r CAlk "gopher" 3. sAqs "a species of tree for making bows" 11. 5 5 kink "stale salmon-head" 4. gAnc H "leaf tobacco" 12. tlnx "Arctostophylus uva ursi 5. yAx wc "sea otter" 13. ' 'v ' xesx w "bluejay, blue" 6. xlxc "frog" 14. kwAlx "green fern root" 7. *Atk "ground" 15. f SAtX "elder sister" 8. nuskw "wolverine" 16. / taql "hammer" Verb Stems 1. suwq "to laugh" 3. xlxc "to be suspended" 2. nA±x "wealthy" 4. klks "to shake" Rule 1 CONSONANT DEVOICING [+ cons] -> [- voice] / [+ F.B.] Obstruents are devoiced stem f i n a l l y . III.4 YOD-DROP In Boas' (1917:' 15) data, na . palatal voiced continuant (yj)^  appears in certain marked affixes where i t is deleted following a consonant or glide. Examples given (Boas 1917: 15) show the loss of underlying /y/ in the suffixes {-yl} and {-yln}. - 82 1. gas "post" dugasi "his post" < du-gas-yl 2. ?an "town" yxis rani "your father's town" < ?an-yi 3. X W A A "tired" xweXIn "having been tired" < xwe^t-yin In the framework of generative phonology adopted here, there are two ways of representing this s u f f i x - i n i t i a l segment with regard to the underlying representation. The f i r s t method would offer the underlying hi s t o r i c a l and etymological voiced post-palatal continuant as an abstract represen-tation with rule-governed loss or assimilation. The second, and more concrete, method would consider the glide as being morphologically marked for loss in certain environments, for assimilation in certainoothers. Primarily because Boas' orthography is not always consistent with regard to a /y^y/ alternation in words or suffixes, the second method of morphological marking has been arbitrarily chosen. Therefore, the suggested rule of yod drop (in which a necessary stipulation i s that this rule must apply to underlying forms as the structural description i s met, unless rule ordering is presumed) i s : Rule 2 YOD DROP - s y l l - cons [+ cons] -> 0 /)\~- syll"] ( [+ F.B.] - cons The glide /y/ in certain morphologically marked suffixes i s dropped following a consonant across a morpheme boundary. However, although following Boas (1917) this analysis includes processes oftyod deletion and vowel assimilation, - 83 -M. Dale Kinkade (personal communication) has suggested that a simpler, and perhaps more natural, hypothesis involving a commonly occurring phonetic process would be that of intervocalic glide insertion with vocalic assimilation. This appears to be the accepted analysis as well of Story and Naish (1973: 348) where: "... i f a stem ends in a vowel, then the consonant y or w comes between the stem and the suffix, y i f the suffix i s -ee, w i f the suffix i s -oo." For several reasons, however, including the etymological origin of the glide /y/, I have rejected this alternative. F i r s t , there are, in Tlingit, a number of morphophonemic alternations involving consonant change or vowel assimilation which, in a generative analysis, would require the positing of an underlying form to be changed in certain phonological contexts by rule, for a more li n g u i s t i c a l l y general solution. Second, other functional inflectional suffixes of the same type as {-yi} and {-yin}, such as {-iq} and {-it} show.vowel loss, rather than glide insertion, after stem-final vowel, while stem-i n i t i a l vowels, as previously noted, are automatically preceded by glottal stop (cf. Story and Naish 1973: 349). Third, since Boas' data are the result of concentrated effort with his native Chilkat informant, Louis Shotridge, one would assume that Boas, like Sapir in Southern Paiute, would check with the informant for an innate intuition of his mother tongue. For an important, and often f i n a l , question in generative phonology i s : "How would the intuition of a native speaker account for the data?" - 84 -III.5 PRE-VOCALIC STEM VOICING A related, though opposing, process to consonant devoicing is the voicing of stem obstruents before vocalic suffixes. This process, according to Stampe's theory of natural phonology, shows the partial suppression of the devoicing process through limitation. According to-Stampe - (1969: .443)'; phonological processes form opposing sets of conflicting phonetic restriction. In the case of obstruents which are voiceless by oral constriction, thoughi: voiced, for example, through voicing assimilation intervocalically, a confict of processes arises, one of which must then be partially suppressed or limited. Final stem voicing (noted also in entries in Story and Naish 1973, Part II), with the qualifications on voicing noted in Chapter I, i s limited to occur only before vocalic suffixes. According to Boas (1917: 14-15), stem voicing does not occur before glides, nor before a phonemic glottal stop plus vowel. Examples are: 1. V xat "root" -\ \ / iduxadl ."hisiroot" du-xat-yi 2. ?at . "to go":(pl) \ r ?adin "having gone" at-yin 3. yek " s p i r i t " duyegl "his s p i r i t " ^ / du-yek-yi 4. yak w "canoe" \ ^ t duyagu "his canoe" du-yak w-yi 5. ?ukw "to b o i l " wul?ugfin " i t has boiled" wu-l-?uk w-yin 6. wAq "eye" \ \ r duwagl • "his eye" du-waq-yi 7. ke^ c "dog" > \ / dukeXI "his dog" du-ke9t-yi - 85 -8. X W A* " t i r e d " xweAin "he was t i r e d " ,x':, we*-yin 9. xliV "to f i s h xlXa "herring rake" xl*-a with rake" 10. yAq "to p u l l " kAyAga "stern sheets" k A-yAq-a (= p u l l e r ) Rule 3 PRE-VOCALIC rj_ , . . . . r, „ „ . r, STEM VOICING [+ cons] >• [+ voice] / [+F.B.] [+ syll] A consonant may be lightly voiced before a following vowel across a morpheme boundary. As the examples show, consonants are voiced intervocalically, a second natural process. However, a lexical exception to the fore-going rule noted by Boas (1917: 14) i s ?it "place" which in compos-iti o n with a vocalic suffix appears to be morphologically frozen, e.g. » • • . * • ' * . xan ? i t l "fireplace" <xan ? i t - y i III.6 ASSIMILATION: LABIALIZATION OR VOWEL HARMONY Of some interesty perhaps., to linguists has been the suggestion by Boas (1917) of the occurrence of vowel harmony in T l i n g i t . Following Boas' analysis, we have chosen to consider the implications of certain of these harmonic processes within a generative framework and to discuss them in some detail. Although Boas included labialization of glide and consonant as a vowel harmony process, we w i l l find i t somewhat simpler at the beginning to separate the two processes for discussion. - 86 -As the discussion w i l l show, there i s approblem, of course, as to whether to include glide assimilation under labialization or under vowel harmony. If glide assimilation occurs under l a b i a l -ization, the rule becomes somewhat awkward. If, however, glide assimilation occurs under vowel harmony, then c y c l i c i t y , simultan-eity, or persistent rule application is required.. These approaches w i l l be^considered' in turn. • . tu-' Kinkade (p.c.) has suggested that labialization may be a more general occurrence than suggested by Boas' data. For example, on reviewing verb roots in Story and Naish's Tlingit dictionary, Kinkade (p.c;)"has perceptively commented: I note that in ther T-lingit dictionary, only a few roots are listed with f i n a l labialized consonants after jL or e_ (there are many after a^ , but i f , as you say, that may derive [sometimes?] from p_, that can be explained). Their i n i t i a l distribution is even more peculiar: there are only five roots with i n i t i a l labialized consonants before i^ or e_; those before a. are presumably from forms with earlier o_. The fact that they are not written before back vowels i s misleading; i t is unlikely that velars and uvulars are not rounded before back vowels, but the acoustic impression of rounding i s masked by the rounded vowel. Note also that S and N do not write rounded consonants after rounded vowels and rounding (in a root) after a rounded vowel may be blocked by a following front vowel." III.6.1 Labialization Progressive labialization, or rounding of velar and uvular (Boas: palatal and velar) consonants as well as palatal glide, occurs i n certain suffixes and prefixes after a l l labialized sounds-vowel, consonant or glide-and. after certain'words ending, in phonetic - 87 -[a] which may be recognized as underlying /o/. These cases w i l l be discussed separately. III.6.1.1 After Rounded Segments The affixes marked for this process are the velar "k" suffixes, the f i r s t person pronominal prefix, suffixes {-yi} and {-yin}, and the prefix {ya-} which only labializes after the vowel /u/ (Boas 1917: 16-18). The vowels of suffixes {-yi} and {-yin}, like the vowels of other functional inflectional suffixes listed by Story and Naish (1973: 349), are also marked for vowel harmony. However, in a stem, there is no rounding assimilation result-ing from stem-affix composition, as the following examples show: * ' 5 5 1. duyitk "her l i t t l e son" du-yit-k 2. qukit' "to pick berries" qu-kit' 3. wUxix " i t f e l l " wU-xix although an exception to this rule occurs, an example of i d i o l e c t a l variation (a further example (Boas 1917: 160) occurs in the compound: t'iy kUwAt' (<yAt' (v.)) "long-elbowed"): v / j \ / 4. ?AkUwati " i t s length" A-ku-yat'-I A few general examples of labialization follow. -"k" suffixes 1. ±14 inuguq w "do not be sick" 414 i-nuk w-iq 2. gux wx w "slaves" guxw-x 3. yAawkw "a l i t t l e strap" yA-aw-k - 88 -Examples are from Boas (1917: 16). The optative verbal suffix / i q / i s subject to vowel harmony or rounding assimilation, as well as to vowel loss following a f i n a l stem vowel (Naish 1966: 31). 1st person pronoun Although weak labialization of k-sounds after /u/ is found in 1st person pronoun combinations, in slow speech i t tends to disappear: \ \ \ t 1. kUqwA-4A-sIn "I shall hide i t myself" 2. gUx w-Al-sIn "I hid i t for myself" The combinatorial changes of verbal prefixes are covered in detail in Story (1966 Ch. 7). (It has been noted by M. Dale Kinkade (personal communication) that the lax vowel (in examples 1 and 2 above) contiguous to velars might, in a phonemic treatment, be treated as schwa phonetically predictable in rounded environments.) Incorporated noun x "mouth" 1. yuxWAXAtAnk "I am t a l k i n g " yu-XA-XA-tAn-k Verbal prefix yA-The verbal prefix {yA-} is labialized only after /u/ in the morphophonemic combination /U W A/ from /wu-yA/, and is thus restricted grammatically and lexically to labialization only after + s y l l +• round [+ F.B.] 1. xat ?uwAdjAq "he k i l l e d me" xat ?u-yA-djAq - 89 -The nominal suffix {-ya} is not marked for labialization, and does not assimilate. If glide and consonant labialization i s treated as a single process, a suggested rule for labialization in these marked affixes, incorporated nouns, and in pronoun combinations is thus: Rule 4' CONSONANT AND r i n 1 . r, . r, r, _ _ . A t N GLIDE ROUNDING I " ^ 1 1 ! > [+ round] / [+ round] [+ F.B.] (V) _ Consonants and glides are rounded following a rounded segment. However, i f labialization applies only to consonants, and the glide is treated more generally as part of vowel harmony or rounding assimilation, a more acceptable rule is this: Rule 4" CONSONANT r^ , v rj_ , ,, r , _, _ , ROUNDING [ C O n S ^ ' ^ round] / [+ round] [+ F.B.] (V) Consonants are rounded following a rounded segment. Though less general, the rule of consonant rounding allows for the special behaviour of yod (etymologically y) in the yod-drop rule. The bracketed vowel in the rule allows for an epenthetic vowel which also occurs in the Angoon or central dialect of Naish and Story. A further, though unlikely, possibility, other than anaptyxis before the diminutive suffix, i s that this vowel may be an underlying stem-ending, deleted f i n a l l y . (This optional vowel, inserted by rule, occurs automatically in certain stated environments, v. II.5). - 90 -III.6.1.2 After marked segment /a/ In some lexically marked cases, labialization of the affix follows nouns ending in phonetic [a], which in a transformational generative treatment may also be regarded as underlying /o/. In these special cases, following the usage of Hyman (1970), morphemes marked [+ L.R.], i.e., plus l i p rounding, indicate that rounding assimilation of the affix follows. 1'. q'a "man" qawu " i t s man" qa-yi [+ L.R.] 2'. nawun "having died" na-yin [+ L.R.] 5 The diminutive suffix /-k/ is also affected by this rounding. Even when epenthetic / A / , which Boas relates to /a/, i s inserted between two consonants across a morpheme boundary as noted by Boas (1917: 18, 86) in examples 4 and 5 below, labialization occurs. 1. ?ak w "a small pond" ?a-k 2. ?akw "a small thing" t ?a-k 3. t a k w "a small spring salmon" •> ta-k 4. hinAk w " a a l i t t l e water1.! hin-k 5. v "> ?exAkw " a l i t t l e grease" ?ex-k 6. '•> ?asAkw " l i t t l e t ree" ?as-k 7. ?anAk w " l i t t l e town" *- 5 ?an-k Labialization caused by /a/ may be followed by.vowel harmony or round-ing assimilation. The rule'for ..glide and consonant labialization follows. - 91 -Rule 4 a CONSONANT AND GLIDE LABIALIZATION [- syll] -> + back + round / l_+ segj Q+- L.R£| [+ F.B.] Consonants and glides are rounded following marked /a/. Because marked /a/ labializes glides directly, as in examples 1' and 2' above, no discussion re rule treatment is warranted since the need for two separate rules for glides and consonants cannot be ju s t i f i e d . III.6.2 Vowel Harmony - Rounding Assimilation Vowel harmony is a process which occurs when certain of the features of a vowel in one syllable come to agree or assimilate with certain features of a vowel of another syllable as, for example, in Turkish where a high suffix vowel assimilates to the backness and roundness of the preceding stem vowel (v. Schane 1973: 52). This type of progressive assimilation, or vowel harmony, where a high suffix vowel 'assimilates to a preceding morpheme-final high vowel', occurs in Yawelmani (Kisseberth 1969, 1970) as an underlying feature (Schane 1973: 78-80, Hyman 1975: 87), and in most idiolects of Walbiri, a language of Central Australia (Hale 1973: 405). In Tl i n g i t , vowel harmony, or progressive assimilation, occurs in certain contexts as automaticealternation. For example, the palatal vowel / i / and the semi-vowel /y/ in certain affixes are velarized after rounded segments or certain nouns ending in /a/. Application of a phonological rule to special morphemes i s not unknown. - 92 -For example, accidental reference of a harmony rule to a single suffix occurs in Turkish (Zimmer 1970) where, according to Hyman (1975: 182), a single morpheme 'satisfies the conditions of a phonological rule.' While no longer a productive process, the exception to harmony noted by Boas (1917: 18) of dutanu " i t s navel" from /du-tan-yi/, and the rounding assimilation caused by certain segments marked /a/ as well as by the inserted vowel / A / , show that the process of distance assimilation or harmony has been a viable concept, perhaps similar to the process of progressive vowel assimilation noted by Kuipers (1967: 36) in Squamish. III.6.2.1 After Rounded Segments The most common assimilations i n Tlingit are those of the suffixes / - y i / and /-yin/ in which palatal phonemes are labialized after rounded segments, whether derived or original (Boas 1917: 17). For this reason, Boas (1917: 16-18, 85, 88) has treated labialization as part of the vowel harmony process for, as the sampling below indicates, the two processes seem inseparable. Further examples of vowel harmony and rounding assimilation, to which the rule of con-sonant and glide labialization has already applied, il l u s t r a t e the process relationship. 1. nu "fort" dunuwu "his fort" du-nu-yi 2. lu "nose" duluwu "his nose" du-lu-yi 3. l'uk w "king salmon" dul'ugu "his king du-±'ukw-yi salmon" - 93 -4. 5. 6. 7. 8. 9. 10. saxaw " h a i r " guxw "slave" yak w "canoe" r f V hltt'agu ?AyAawuu. ?uxun suwgun v \ v r dusaxawu "his hair" duguxu "his slave" \ \ / duyagu "his canoe" "house timber" " i t s handle" "having blown" "having laughed" du-saxaw-yi du-gux-yi du-yak w-yi f h i t - t ' a q w - y i A-yA-aw-yi f u x w - y l n suwq w-yln III.6.2.2 After Segment /a/ 1. duaku \ \ t 2. dutanu "his l i t t l e pond" " i t s navel" * ' 5 du-a-k w-yi du-tan-yi [+ L.R.] As previously noted, this second example, which must be accounted for in the lexicon, i s quite exceptional with vowel harmony applying through a nasal segment, and would therefore have to be marked [+ L.R.] in the lexicon. Vocalic rounding assimilation, however, which occurs in the above examples, must be accounted for by rule. Therefore, a vowel harmony rule for marked affixes applying after the rule of glide and consonant rounding (Rule 4') i s indicated: • Rule 5' PALATAL VOWEL ROUNDING + s y l l - back -> [+ back] / [+ round] [+ F.B.] A high front vowel i n marked affixes i s backed and rounded following a rounded segment across a morpheme boundary. - 94 -However, for a vowel harmony rule applying both to palatal vowel and glide after the application of Rule 4", the rule of consonant labialization, the rule would be: Rule 5" PALATAL VOWEL AND GLIDE ROUNDING - cons + high -> [+ back] / + back + round [+:'F.B.] In certain affixes, the high front vowel and palatal glide back and round following a rounded segment across a morpheme boundary. Nevertheless, i t seems possible and plausible that these Rules 4', 4" and 4 , respectively for consonant and glide rounding, con-sonant rounding, and consonant and glide rounding following marked a, in addition to the above Rulesa5<l aridfS" fofapalatalcvowel.rounding and palatal vowel and glide rounding, can be collapsed together into one natural and phonetically plausible assimilation rule that w i l l apply cyclically or persistently as the structural description i s met. We shall designate this collapsed or reduced rule as Rule 4*. Rule 4* ROUNDING + cons + back ASSIMILATION + high + round (+ roundjy & s e § _ T ( [+ F.B.] (V) In certain affixes, a high consonant, glide or vowel backs and rounds after rounded or marked round segment across a morpheme boundary. An interesting feature of the above rule, of course, i s the - 95 -addition of the distinctive feature of height, which relates the palatal and velar consonants to the high vowels and glides, corre-lating in this way with features of backness and rounding, character-i s t i c of types of vowel harmony as in Turkish (Schane 1973: 52) where "the high vowels of a suffix agree in backness and rounding with the stem vowel". As can be seen, the rules and processes discussed above have been phonetically plausible, inherently natural assimilation rules which, to some degree, il l u s t r a t e the remarks of Schane (1972: 207): Rules of assimilation provide some of the clearest examples of natural rules. In a l l such rules the values for one or more features of a segment are changed so as to agree with the values for the same features in some other segment. Following from the discussion above, i t w i l l be seen that some comments on rule ordering and/or rule restriction are necessary in order to achieve the desired forms of a phonetic representation. III.7 TONE DISSIMILATION The vowels of a number of suffixes, including {-yi} and {-yin}, have a neutral pitch. When added to a stem with high tone, the suffix takes a low tone; when added to a stem with a low tone, the suffix has a high tone. While tonal processes were not discussed by Stampe (1969) as a natural process, tone dissimilation would qualify under the category of rules for maximum differentiation (Schane 1972: 207). As noted, - 96 -the acute accent indicates high tone, while the grave indicates low tone. Examples from Boas (1917: 11) are given below. 1. dutayl "his king salmon" du-ta-yl 2. dutayl "his board" du-ta-yl 3. \ \ / duxadl "his root" du-xat-yl 4. duxadl "his salmon" du-xat-yl 5. \ \ t d u t i l l "his scar" du-til=yl 6. d u t i i l "his shoe" d u - t i i - y l 7. nugun "having been sick" nuk w-yln 8. \ V t xasm "having cut" xas-yln The tone dissimilation rule for these neutrally marked suffixes is the following: Rule 6 TONE DISSIMILATION high' tone -a V higtf tone stem + (G) .The neutral-toned vowel of a suffix w i l l have the opposite tone to the stem vowel. III.8 RULE ORDERING In the standard theory of transformational generative grammar put forth by Chomsky (1965) and Chomsky and Halle (1968), the phono-logical component, ancillary to the syntactic component of the - 97 -grammar, must operate on the syntactic strings or surface structures generated during the operation of the transformational cycle. This phonological component has been defined by Chomsky and Halle (1968: 9) as the system of rules that applies to a surface structure and assigns to i t a certain phonetic representation drawn from the universal class provided by general linguistic theory. However, since the output of the syntactic component, the syntactic surface structures, may not exactly coincide with the input to the phonological component, the readjustment rules (or morpheme structure conditions) may need to reanalyze the underlying lexical representations of the base into the appropriate phonological representations. The function of the phonological component i s , then, to convert the underlying abstract phonological representations of the phono-logical surface structure to concrete phonetic representations by means of a series of ordered and organized phonological rules which can apply to words, or to strings of formatives which include phrases, words or morphemes. III.8.1 Linear Ordering These principles of rule order have been enunciated by Chomsky and Halle (1968: 18): It is always possible to order the rules in a sequence and to adhere s t r i c t l y to this ordering in constructing derivations without any loss of generality as compared to an unordered set of rules or a set ordered on a - 98 -different principle. Such linear ordering makes i t possible to formulate grammatical processes that would not otherwise be expressible with comparable generality. Because principles underlying theories of natural language are assumed to be divided into two categories of linguistic universals, formal and substantive, the formal universals are considered the organizational or rule component of a transformational generative grammar, while substantive universals deal with the categories, features, and linguistic units. According to Chomsky and Halle (1968: 4): the theory of transformational generative grammar proposes certain formal universals regarding the kinds of rules that can appear in a grammar, the kinds of structures on which they may operate, and the ordering conditions under which these rules may apply. With regard to rule application in a transformational generative grammar, the underlying principle of rule order in the phonological component (Chomsky and Halle 1968: 341) is the convention that a l l rules apply extrinsically in a linear order, with each rule operating on a string modified by " a l l earlier applicable rules". However, because Rule A must apply before Rule B in some examples, while Rule B must apply before Rule A in others, the convention of the phonological transformational cycle (Chomsky and Halle 1968: 349), which applies mainly to prosodic or closely related segmental pheno-mena, has been introduced. - 99 -III.8.2 Refinements to Rule Ordering The convention of the phonological transformational cycle for other than prosodic phenomena at, above, or below word-level, though not a significant factor in recent generative phonology (and rejected by White (1972) and Brame 1972)' has been noted in treatments of several Amerindian languages (v. Zwicky 1976: 267-8). However, instead of the transformational cycle convention, iterative-rule application, through the device of i n f i n i t e schema which would allow a rule to apply to i t s own output, has been the suggested modus operandi of Chomsky and Halle (1968). Anderson (1974: 124-133), in dismissing this solution, has noted that this formal device, by reducing linguistic generality, creates more problems than i t solves. Further modifications to a s t r i c t linear rule ordering have been suggested by Kiparsky (1968, 1971), Chafe (1968), and Anderson (1969). For example, Kiparsky (1968) has created the notion of feeding and bleeding relationships in rules where the unmarked or natural order is a feeding relationship and where, in a bleeding relationship, the other order is the unmarked. If there is no rule relationship, rules are then neutral with respect to each other. A feeding relationship has been defined by Kiparsky (1968: 196) as: "one way in which two rules, A and B, can be functionally related is that the application of A creates representations to which B is applicable." - 100 -A further principle enunciated by .Kiparsky (1971: 623) with regard to rule ordering is the dictum: "Rules tend to be ordered so as to become maximally transparent." The converse of transpar-ency, opacity, is defined by Kiparsky as follows: A rule A } B / C D is opaque to the extent that there are surface representations of the form (i) A in environment C D or ( i i ) B in environment other than C D. Chafe (1968), using examples from Caddo, an American Indian language, proposed that the rules of grammar be organized in several depths, with rules of the greatest depth to be applied simultaneously to underlying forms. A persistent rule would be one that would apply at any depth as the structural description was met, an "anywhere" rule. His additive and subtractive inter-ference are the equivalent of Kiparsky' s feeding and bleeding relationships. As Anderson (1974: 202) has stated, however, in criticism of Chafe's theory: We see, therefore, that the theory of persistent rules including the most l i b e r a l view of the possibility of simultaneous ordering cannot accommodate an example like that of Icelandic u-Umlaut. The reason, of course, is the fact that u-Umlaut behaves differently in i t s relation to different rules...A theory such as Chafe's, which divides rules into f u l l y linear and ful l y 'anywhere', cannot accomodate these facts; even the use of simultaneity does not help, i f the rule with respect to which a putatively persistent rule is restricted cannot be ordered last among the linear rules. - 101 -Anderson (1969) has put forth a theory of local ordering in which rules may be related naturally, as discussed by Kiparsky (1968), or e x p l i c i t l y . Only explicit linear ordering w i l l be stated in the grammar, since natural ordering is predictable by universal principles. Even so, as Norman (1972: 491-2) has suggested with respect to the local ordering hypothesis (LOH): An alternative to the LOH is the complete elimination of extrinsic ordering, with a l l ordering being absolutely determined by a set of universal principles, as proposed by Koutsoudas, Sanders and Noll...The principle that a rule must apply whenever i t s structural description is met is sufficient to account for a l l feeding and counter-bleeding orders and to eliminate the possibility of natural languages rules in the orders of bleeding and counter-feeding. III.8.3 No Ordering Hypothesis Recently, however, in phonological theory, a distinction between extrinsic rule order, that imposed by the data of a specific language, and intrinsic rule order, that governed by the formal properties of the rules, has been discussed. According to Dinnsen (1974: 29): Rule ordering, for example, has been an issue of current concern. The standard theory provides for the proper sequencing of most phonological rules by statements of extrinsic rule ordering, i.e. language-specific conditions in a grammar which assign one linear order to the rules. However, Koutsoudas, Sanders and Noll 1974 claim that such rule ordering provides greater descriptive power than can be shown to be necessary. As a more constrained alter-native to language-specific conditions on rule application, they propose universal principles which determine applica-tional precedence relations between rules. - 102 -Like Vennemann's No-Ordering Principle, the effect of Koutsoudas, Sanders and Noll's Proper Inclusion Precedence, which accounts as well for Kiparsky's feeding and bleeding relationships, is that a rule must apply whenever i t s structural description i s met in random sequential ordering. According to Koutsoudas, Sanders and Noll (1974: 13) with regard to the elimination of extrinsic rule ordering: F i r s t , arguments suggesting the necessity of extrinsic ordering based on a very limited range of facts often f a i l completely as soon as additional facts about the langugage are brought into consideration. Second and more important, in a theory that excludes the possibility of extrinsic ordering, the linguist is forced at the outset to look for general explanatory principles which there would otherwise be l i t t l e reason to look for. Theories of grammar which prohibit language-specific restrictions on the application of phonological rules thus provide a degree of stimulation and direction in the search for significant linguistic generalizations which is lacking in those theories which permit such restrictions. Recently, however, Cathey and Demers (1976) have, using Old Icelandic examples, maintained that Koutsoudas, Sanders and Noll's universal principles are insufficient and that extrinsic ordering should hot be eliminated from a grammar. However, as formally argued by Levine (1976: 115) with respect to the current debate of extrinsic versus intrinsic ordering in syntax, "unless a concept of naturalness of transformations can be incorporated into grammatical theory, the debate seems to be one of conflicting methodology, taste and terminology." - 103 -III.8.4 Application of Rules Questions of application in rule ordering have formed an issue in linguistic discussions recently. According to Anderson (1974: 221): Problems of a substantive nature begin to arise, however, when we consider the application of a rule to a string which contains several instances of substrings satisfying the rule's structural description. When several potential applications of a rule exist in a single string, how i s the change specified by the rule to be carried out? Numerous alternatives can be imagined: apply the rule simultaneously everywhere; apply the rule in one place at a time, in random sequence; apply the rule in only one place, making the choice either randomly or on some universal or language particular basis; etc. Thus, a review of the simple processes of voicing, yod-drop and labialization including vowel harmony assimilation indicates that a few comments are necessary. For example, how or how often does the rounding assimilation rule apply to an underlying form? When does yod-drop apply? For convenience, the rules are summarized and ordered in Table I, with categories designated according to Chafe (1968). Derivations including the underlying representations w i l l be discussed on the accompanying data sheet. Briefly, in reference to the data sheets, examples #1 - 5 indicate derivations resulting from a simultaneous application to one string at a time of linearly ordered rules (Table I). In examples #2 and #4, however, rule 4 must apply to i t s own output in order to obtain the desired representation, thus violating TABEEBIE I Table of Tlingit Phonological Processes Label Rule Categbryegor> Devoicing Yod-drop rPreevocaiic Stem-Voicing [+"cons] -> [- voice] / [F.B.] -> 0 / [+ cons] [F.B.] Obstruent devoicing -one of Stampe's (1969) natural processes. Persistent rule (Chafe 1969). Persistent; sel f -ordered before #4. [+ F.B.][+ syll] Occurs before pre-vocalic suffixes. A natural assimilation process. Vowel Harmony Assimilation Tone Di s similat ion + cons + back t + high_ + round v xhigh tone -a V high tone ^tem + (G) Persistent. Affixes or stems must be lexically marked. Applies to tone-neutral suffixes. - 105 -DATA SHEET Example Rule • 1. UR du-gas-yi "his post" 2 du-gas-i v r , o 5 du-gas-i 2. UR na-yin "having died" 4 nawin 4 nawun 5 nawun 3. UR du-gux w-yi "his slave" 2 du-gux w-i 4 du-guxw-u 5 du-gux-u 4a. 2 du-yak w-i "his canoe" 3 du-yag w-i 4 du-yagw-u 5 du-yag-u 4b. UR du-yakw-yi 2 du-yakw-i 4 du-yakw-u 3 du-y.agw-u \ \ / 5 du-yag-u 5a. UR v > » du-a-k-yi "his l i t t l e pond" [+L.R.] 2 du-a-k-i *< ' 5 4 du-a-kw-i Boas Reference 15 18 88 88 18 4 5 du-a-kw-u V / , V du-a-k-u - 106 -Boas Example Rule Representation Reference 5b. UR du-[a-£]-yi 18 [+L.R.] 4 du-[a-i w]-yi 2 du-a-£w-i 4 du-a-£w-u \ f j \ 5 du-a-k-u In example 5b, rules"must first"apply-within square brackets. A labialization rule has not been offered following Kinkade's perception (v. II.10.4) of indicated.labialization 1following rounded consonants as an orthographic artifact. - 107 -the convention that a rule may not apply to i t s own output except through an extra principle such as the transformational cycle or a costly device such as i n f i n i t e schema. However, the use of the transformational cycle in example #2 would s t i l l necessitate the reapplication of rule 4. Example 5b illustrates the effect of the transformational cycle with linearly ordered rules applying f i r s t to the innermost constituents of a syntactic bracketing, then reapplying after erasure of the innermost brackets. Although the principle of the transformational cycle has been used in a few treatments of segmental phonology in American Indian languages with some success, the cycle does not appear workable in example 2, unless some extra syntactic bracketing i s warranted. The remaining examples w i l l i l l u s t r a t e the application of rules applying in random sequential order as the structural desc-ription i s met, with the result that rule 4 may apply to i t s own output in order to achieve the desired phonetic representation. It should be noted that rule 2, yod-drop, is in a feeding relationship to rule 3, pre-vocalic stem-voicing, and thus, accord-ing to principle, occurs before rule 3 in a natural, unmarked order. Thus, in the few simple processes we have described in Tl i n g i t , (as evidenced in the application of the rules in the examples), rules are best applied when the structural description i s met. No rule ordering, although rules could be ordered as in Chomsky and Halle (1968), or local ordering (Anderson 1972), or even unmarked - 108 -ordering, appears to be necessary, although given a f u l l set of gram-matical processes and a f u l l structural description of Tlingit some theoretical amendments might be required. Similarly, in the limited number of examples given, no notion of c y c l i c i t y appears to be needed and, given the nature of the rules, even seems warranted. Nor need a s t r i c t linear or even local order-ing apply, were cyc l i c i t y a pre-requisite in longer derivations. These language-general rules thus appear well satisfied to apply either in conjunction with Chomsky and Halle's universal markedness conditions, or Stanley's language-specific morpheme structure conditions, or within the theory of natural phonology which includes the Basic Category Principle (v. Rudes 1976). Although a lexical marking is necessary on certain morphemes, e.g., nouns ending in /-a/, in order to stimulate a rule to apply, the un- or non-ordering of the rules given here is entirely consistent with Koutsoudas? (1973) (and with Koutsoudas, Sanders and Noll 1974) enunciated principle: A l l restrictions on the relative order of grammatical rules are determined by universal rather than by language-specific principles. III.9 SUMMARY In this chapter, having briefly discussed the standard theory and recent modifications or constraints in the f i e l d of natural phonology, we have examined a few of the simple language-general phonological processes of Tlingit which, while serving, to i l l u s t r a t e - 109 -the theory, provide as well a minor description of language data in the theory of transformational generative grammar. Thus, with regard to rule ordering of the simple phonological processes in Tlingit, we have arranged our system of rules in accord-ance with the no-ordering hypothesis put forward by Koutsoudas, Sanders and Noll (1974), not only as a matter of "methodology, taste and terminology" as suggested by Levine (1976), but in accordance with the search for universal principles in linguistic theory, which in Chomsky's view, according to Cole (1976: 564) should play a crucial role in the analysis of specific languages. It should provide c r i t e r i a to determine the relative inadequacy of a number of apparently adequate descriptions. A theory of this sort would be in a symbiotic rather than a parasitic relationship . to description: linguistic theory would assist in the refinement of linguistic description by ruling out unacceptable analyses, just as description assists in the refinement of theory by falsifying incorrect theoretical hypotheses. In the following chapter, we w i l l examine morphophonemic representation by using various treatments of vowel harmony assimilation in Tlingit in order to discuss current issues in abstractness. - 110 -IV.0 APPLICATION AND APPROACH IN PHONOLOGY Abstractness in phonology is thought to account for deeper, underlying patterned regularities in a language. Concerning abstract-ness, Hyman (1970:58) has raised two questions for generative phonologists: (1) "To what extent do the underlying forms of the most highly valued phonological system of a language differ from those forms in surface phonetics? (2) "Does a phonology that 'explains' certain processes at an abstract level necessarily coincide with the one explanatorily adequate (that i s , the most highly valued) form of phonology?" In this Chapter, using a preferred interpretation of vowel harmony in Tlingit phonology as examplar (and Lyons' (1962) motto that "the actual cannot be properly described (or recognized) except in the framework of what has been previously envisaged as possible"), possible solutions to the problem of abstractness in generative phonology raised by Kiparsky (1968) and others are set out. In this context, Lyons:'; (1962:127) rationale is worth noting, for in a l l social sciences in recent decades, in a formulative process in which linguistics and i t s particular fields of enquiry have not been immune, methodological questions of the most funda-mental nature have arisen: To those of you who, having heard these views, might feel inclined to say that they are of "only theoretical - I l l -interest" and that the linguist's job is to describe what actually occurs in particular languages without troubling himself about what might occur (for I have heard this said), I would suggest that the history of science is f u l l of examples to support the opinion that the actual cannot be properly described, perhaps not even be recognized, except in the framework of what has previously been envisaged as possible. At the same time, of course, the sphere of what is thought of as possible is being constantly revised under the impact of discoveries made in the description of actual languages. Such is the relation between the theoretical and the applied. And, as a consequence of this, linguistic typologies should be built of a judicious mixture of induction and deduction. This statement of Lyons is particularly relevant as i t an t i c i -pates and conforms to the propositions advanced by Kuhn (1970) and Popper (1972) as applied to the spectrum of pure science and social science. Following Kuhn, linguistics i s susceptible to the process of established, competing and ultimately replacement 'world views'. Adapting Popper, a l l theories in linguistics are provisional (are 'conjectures') u n t i l refuted by a superior (or more complete, or general) theory which is i t s e l f of provisional status. These processes asserted by Kuhn and Popper have been indicated M Ghapiterf IJ 1I_in ^ermscofitheir appMcationtitoolainguistic theory gyerdthenpaSitpdgcadeT^ Of perhaps less fundamental concern, but a source of continuing methodological controversy, is Lyons'- distinction between the 'theoretical' and the 'applied' (descriptive, or the empirical testing of hypotheses). Thus, in this Chapter, Lyons-.,, statement is taken more as a caution than as a touchstone or guide. - 112 -IV.1 ABSTRACTNESS IN PHONOLOGY The issue of abstractness in the relationship of levels, phono-logical and phonetic, was f i r s t raised for generative phonologists by Kiparsky (1968) at the beginning of a b r i l l i a n t and productive essay: What is the form of morphophonemic representations? How far removed are they from the phonetic and phonemic surface? The whole gamut of possible answers to this question has been given In modern linguistics. Characterized as a distinguishing feature of generative phono-logy over more traditional theories, abstractness is the result of absolute neutralization: the appearance in the underlying phonological representation of a non-alternating segment, unrealized phonetically and set up solely to classify and distinguish that segment in order to meet the structural description of a rule. Interestingly, and in this context historically, McCawley (1967) has compared the mentalistic representations of Sapir with the under-lying phonological representations of generative phonology, in that only two levels are considered of relevant significance: an underlying representation representing "the speaker's 'mental image' of the various morphemes and a phonetic representation which corresponds to the actual articulation made by the vocal organs in producing utterances". According to McCawley (1967:106): No systematic significance i s attached to any intermediate representation in the sense of e.g. Bloch and Trager's Outline of Linguistic Analysis, although in both Sapir and transformational grammar, 'underlying' phonological repre-sentations are converted into phonetic representations through several intermediate stages, since some 'rules' - 113 -apply to the output of other ''rules' . That the repre-sentations which Sapir called 'phonemic' or 'phonological' were not 'phonemic' as the term is used in the 'Neo-Bloomfieldian' tradition-6f Bloch and Trager op. c i t . i s apparent from a perusal of almost any of Sapir's grammars; to cite an example, in his Takelma grammar, Sapir gives four forms of the verb to shoot which he observes are a l l pronounced [sak'][but have different phonological repre-sentations. Although Gregg (p.c.) has referred to D. Jones' (1957, 1964, 1967:6-7, 204) discussions of abstract'..vs. concrete sounds, abstractness was apparently not an issue in phonemic theory to the dominant post-rBloomfieldian structuralists who were bound, according to Chomsky (1964), by constraints of bi-uniqueness, linearity, invariance, and local determinancy. How-ever, abstractness has remained a concern in recent theory of transforma^ tional grammar. According to Harms (1973:439) one important unresolved issue "concerns the abstract relationship between the phonological shape of morphemes in the lexicon and their phonetic manifestations". IV.2 MORPHOPHONEMIC REPRESENTATION Traditionally, morphophonemic representation has been regarded as a third, though generally undefined, level in the stratif i e d hierarchy consisting of phonetic, phonemic and morphophonemic levels of structural^ i s t phonemic theory. Anderson (1974), in a highly lucid exposition of morphophonemic representation in both phonemic theory and generative phonology, has argued that the structuralists made no attempt to relate the morpho^ phonemic level to the concrete phonetic basis. According to Anderson (1974:33): Morphophonemes, however, could not be given the kind of operationalist definition that appealed to the positivism - 114 -of the period. It is not possible to recover morpho-phonemic information directly and unambiguously from the speech event i t s e l f . Accordingly, the morphophoneme was never regarded as a legitimate or 'real' entity within taxonomic theory, but rather as a pure descriptive artifact, a convenient fi c t i o n devised by the linguist to shorten his description. Thus, no effect was made to impose any sort of condition of naturalness on the ele-ments of a morphophonemic representation: these could be l i t e r a l l y anything at a l l , with any arbitrary set of rules to specify the correspondence between morphophonemic and phonemic entities. Withinitransf ormatlohMng morphophonemic representation, here used interchangeably with the terms phonological or systematic phonemic representation following the usage of Chomsky (1964), Kiparsky (1968) and Schane (1968), (but not of Chomsky and Halle (1968), has been defined by Chomsky (1964) and Chomsky and Halle C1968) as' the representational level which results from the effect of the readjustment rules or morpheme structure conditions on formative strings generated by the syntactic component; and the level which forms the input to the phonological component of a transformational grammar from which the output - the systematic phonetic level or derived surface representation - then results. And although Schane (1968), in defining a phonological repre-sentation as one more abstract than a narrow phonetic one, though related to phonetic representation by a set of interpretive rules, has argued (1971) that the output of generative phonology in most cases is not the detailed phonetic specification i t purports to be, but rather is akin to a Smith-Trager phonemic representation (in other words, the output of the p-rules on the morphophonemic - 115 -representation is phonemic in nature); within transformational generative theory morphophonemic .(systematic phonemic, or phono-logical) and systematic or universal phonetic levels are separated and related only by a system of phonological rules (or p-rules) which form the phonological component of a transformational generative grammar. Since this system of relation leaves a tremendous amount of leeway as to interpretation within generative theory, various con-straints and conditions, previously reviewed, have been proposed to allow for a unique phonological representation, including that of Schane, Postal's naturalness condition, Kiparsky's alternation con-dition, Stampe's natural phonology, and Vennemann's natural gener-ative phonology. IV.3 APPROACH Within, and without, transformational generative theory various viewpoints have been entertained with regard to the reality of the morphophonemic level. For example, Chafe (1968:115) has suggested: In some fashion virtually a l l serious linguists, both now and in the past, have recognized the validity of phono-. logical structures which are more abstract than phonetic -less directly related to physical sound - and i t has usually been held that phonetic structures are in some sense derived from these more abstract forms. Bloomfield-ian linguists, because of their unwillingness to stray very far from observable data, admitted only the validity of 'phonemes', units which differed from the phonetic only by virtue of the avoidance of complete phonetic redundancy. Both before and after the Bloomfieldian period, however, bolder abstractions were and have been posited, many of them far removed from the phonetic realm. - 116 -In his 1968 essay on abstractness, Kiparsky defined three schools of thought regarding the approach to an accepted morpho-phonemic representation in linguistic phonology - those of abstract, concrete, and process theory. According to Kiparsky, in the abstract view, morphophonemes are purely classificatory ideal elements with no phonic function. Abstractionists bamhraaistratifiieationalis't, and Householder, a post-structuralist, as well as Fudge of the London school, and a glossematist Hjelmslev, a Saussurian disciple of the Copenhagen school, exemplify this viewpoint. As described by Fudge (1970:89): The parallel between an 'abstract' view of the phoneme and modern approaches in the philosophy of science is brought out by Shaumjan (1968): while allophones and distinctive features are entities with a physical basis, phonemes and 'differentors' (the abstract entities corresponding to distinctive features, cf. cenemes) are purely abstract and have the status of 'constructs' within the phonological theory - their purpose is to account for the complexities of the observed physical data (utterances by speakers of the language), and not directly to describe them. Furthermore, this insistence on a s t r i c t distinction between the (abstract) phonemic elements and the (at least partially concrete) phonetic elements would appear to be well suited to a psychologically real theory of speech production. On the other hand, linguists favouring a concrete or 'reality' view expected that a morphophonemic representation would furnish a l l the realized forms of a particular morphemic alternation. This stand, by Kiparsky's definition, was that of the post-Bloomfieldian item-and-arrangement structuralists of the 1940's (though not of - 117 -Bloomfield himself) and of the early Prague school of linguists, including Troubetzkoy."*" The third view, called process morphophonemics, of Sapir, Jakobson 1948, Kiparsky (1968), and Chomsky and Halle (1968), presumably derives i t s t i t l e from the item-and-process structuralist school typified by Z e l l i g S. Harris (1951). Process morphophonemics combines the ideas of both 'ideal' and 'real' in that morphophonemes are considered as abstract entities realized on a concrete phonetic level. As Kiparsky (1968:1-2) states: Like f u l l y abstract morphophonemics, i t recognizes that there i s an underlying phonological pattern which i s not necessarily identical with the phonetic pattern (e.g., superficial [q] may function like /ng/)*, which f u l l y concrete morphophonemics is forced to deny. On the other hand, i t recognizes that this pattern, while abstract, i s not arbitrary, but in general is related to the phonetic l e v e l e d Q] i s hardly l i k e l y to function as /o/ or /p/ or / l / ) , a relationship which f u l l y abstract morphophonemics, with inexplicable joy, throws overboard. In adhering to Postal's naturalness condition (i.e., a natural relationship exists between concrete phonetic and abstract phono-logical structures), morphophonemes are largely "but not wholly" determined by their surface phonetic realization. As Postal (1968: 56 n. 3) has noted in comparing the three approaches - abstract, 1 In contrast to this view, i t has been noted that one of the defining c r i t e r i a of generative phonology is the acceptance of absolute neutralization: the setting up of phonetically unrealized phonological distinctions in a morphemic lexical representation - v. Kiparsky (1968:9). Later Prague school linguists inclined to an abstract, functionalist view though with a more or less concrete phonetic realization of contrast and opposition: v. Crystal (1971:179). - 118 -concrete, and process - as exemplified in three theoretical models: It is thus proper to look upon the theory of systematic phonemics as intermediate between autonomous phonemics, which assumes in effect that phonological structure is mechanically determinable from phonetic information plus contrast, and a theory, like that in part approxi-mated by stratificational grammar, in which phonological structure would be an arbitrary code. Systematic phonemics is intermediate in the sense that i t recognizes phonetic structure as providing a substantial, but far from complete, portion of the information relevant for the determination of phonological structure, the rest being provided by grammatical information, i.e. information about word boundaries, syntactic and morphological categorizations, morphophonemic alternations, etc. IV. 4 A P P R O A C H E S V A P E E B S A T J J O N . O N As delineator, however, Kiparsky has not clearly distinguished between approach or school of thought and application or method in the use of abstract, concrete and process terminology. For example, while prosodists of the London school regard morphophonemes as abstract entities, their methodology as illustrated by Fudge also tends toward complete abstraction with a s t r i c t separation of morpho-phonemic and phonetic levels, and with the usage of non-phonetic symbols in both mutation (sound-change) and realization (phonetic level) rules. Approach and method may therefore be called synonymous. Thus, Fudge, (1967:23, 1969) (rebutted in part by Kiparsky 1968), has offered rather stern criticism of the failure of Chomsky (1964, 1967) and Chomsky and Halle (1968) to maintain a "consistent and rational distinction" between levels which appear to be "rooted in actual imprecisions of thought... and lead in turn (as often happens) to further confusion". For example, in Fudge's (1969:24) opinion: - 119 -Obviously the fact that the symbol B within the slant lines is the same as the symbol B within the square brackets (bundles of distinctive features in each case) counts for very much more than the fact that the slant lines are distinct from the square brackets. In other words there is not really stratification - the alleged distinction between systematic phonemics and systematic phonetics is not really drawn. I would suggest (cf. Fudge, 1967:6-7) that this i s empirically unsound: thus i t i s not (or at least not only) Lamb who is arguing on 'conceptual' grounds (Chomsky 1967:107) - Chomsky and Halle's insistence on using distinctive features at both phonemic and phonetic levels is equally aprioristic. And although Aoki (1966), Lightner (1965), Bach (1968), and Chomsky and Halle (1968) and other generativists of the process school, tend also to make use of abstract methods, e.g. root marker or i n i t i a l vowel or absolute neutralization, in their solutions to certain phonological problems, Kiparsky (1968:11) has stated that within present generative theory the decision among the three alternatives is arbitrary "because the evaluation measure assigns no relative weight to rule features vs phonological features". However, in the newer natural generative phonology, where greater constraints are placed upon abstractness, a concrete or "surface" solution i s preferred. Thus, clearly, at the outset, a distinction of approach or school versus application or method in terms of abstract, concrete and process terminology has been necessary. Therefore, using, where possible, examples of vowel harmony in various languages possessing this feature, a theoretical review of methods in generative phonology with solutions, criticisms and discussions according to the three categories of application is now commenced. - 120 -IV.5 ABSTRACTNESS IN METHODOLOGY An abstract methodology using absolute neutralization w i l l permit the use of non-alternating underlying phonological segments which do not appear on the phonetic surface (Hyman 1970). Within current abstractness theory various solutions have been devised using archiphonemes, neutralized segments unmarked for certain matched features; the d i a c r i t i c use of phonological features, e.g., the use of schwa as an underlying phonetically unrealized feature; and the phonological use of di a c r i t i c features, e.g., the use of the feature {a GRAVE} to specify roots (v. Kiparsky 1968). Three solutions w i l l be investigated: the root marker, a prosodic method which regards vowel harmony as a property of the morpheme; the i n i t i a l vowel method, which regards vowel harmony as an assimilation process; and absolute neturalization, which is thought to account for the deeper underlying regularities in the specific grammar of a language (Hyman 1970). These three solutions are now examined in their contexts. IV.5.1 Root Marker The root marker, a morphological solution involving archi-phonemes and the phonologicaibsusefofiidiacritieefieatures^awas intro--duced by Lightner as a solution to vowel and velar consonant harmony in Classical Mongolian. Evidently inspired by Z. Harris' (1951) "Phonemic Long Component" (v. Lightner 1965:247 n.8), the intention - 121 -of the root marker, prosodic, non-assimilating and non-directional, is to capture formally the intuitive notion of harmony (Lightner s 1965:249). By vowel harmony rule, roots are furnished with abstract markers which specify simultaneously every vowel in the root, plus attached affixes, with a harmonizing feature. According to Lightner (1965:247), the rationale for root marker method in vowel harmony is thus.specified: With each root we shall associate an abstract marker specified for the binary feature GRAVE. Each root w i l l thus carry either the marker {+ GRAVE} or the marker {- GRAVE}. This marker is not to be confused with the binary feature grave. The distinctive feature grave i s a phonological feature like vocalic, diffuse, and so forth; distinctive features are properties of sound segments. The marker GRAVE is an abstract idiosyncratic property of roots analogous to the markers ANIMATE, TRANSITIVE and others; the markers ANIMATE, TRANSITIVE have primarily syntactic reflexes, the marker GRAVE primarily phonological reflexes. Two examples of vowel harmony prosody in Finnish and in Classical Mongolian ill u s t r a t e the specified representation and the vowel harmony rule. A. Vowel Harmony - Finnish An example of Finnish vowel harmony from Kiparsky (1968), using the root marker method suggested by Lightner, shows the lexical l i s t i n g using unmarked vowel archiphonemes and the morphological feature {GRAVE} from Halle (1962). / pOUtA / / pOUtA / + GRAVE - GRAVE - 122 -A prosodic rule from Bach (1968) assigns features to each unspecified vowel of the root, marking the vowel for backness as indicated by the morpheme-sized feature. Rule HABMONY [ + s y l l ] > [ a g r a V e ] 1 [ a G R A V E ] A syllable agrees in backness according to the assigned morpheme-feature marking. A further assimilation rule f i l l s in the missing features of the vowel archiphonemes with the resultant phonological words: / poyta / and / pouta /. B. Vowel and Consonant Harmony - Classical Mongolian A second example using root marker illustrates the application of root marker morphemes in describing vowel and consonant harmony in pre-sixteenth century Classical Mongolian. In Classical Mongolian, a l l vowels in a word are either a l l acute (front) or a l l grave (back), except for i which may occur with either type. Velar consonants also agree in graveness with vowels, becoming pre- or post-velar (v. Lightner 1965:245) as marked below. UGUtA "bag" KObAGUn "son, boy" {+ GRAVE} {- GRAVE} A suggested rule for vowel and velar consonant harmony in Classical - 123 -Mongolian has been given by Bach (1968): Rule VOWEL AND CONSONANT HARMONY '[+ syll] + high + obs - strid •>• [a grave] / [a GRAVE] According, to this rule, vowels and velar consonants within a morpheme boundary are specified by the morpheme, feature [GRAVE] to agree in backness with the phonological, feature [grave]. Assimilation rules and / kobegiin / appear. Because the traditional solution to vowel and velar consonant progressive or regressive assimilation, the root, marker method, applicable to other vowel harmonic languages,- has been suggested. Discussion Zimmer (1967) suggested a possible use for root-marker in languages where both prefixes and suffixes have pre-determined vowels, thus requiring progressive and regressive assimilation rules. Citing Igbo, an African language, and Nez Perce, an Amer-indian language, Zimmer admitted that an essentially prosodic statement, would successfully handle the data in each. A further argument cited i s the general economy of statement, with the need not to account for intervening consonants in root marker. However, f i l l in the missing features and the phonological words / uyuta / harmony in Classical Mongolian requires, an arbitrary decision as to - 124 -a decision involving simplicity would be arbitrary in view of the d i f f i c u l t y in equating morphological with phonological features. Noting that directional assimilation rules are more efficient in handling cases such as labial harmony in Turkish, Zimmer (1967: 171), giving qualified support to root marker, has suggested: But i f , as has been suggested above, root-marker rules are in some general sense simpler, then i t is rules of this kind that would be chosen for languages like Mongolian. Such rules could be considered as unmarked; i t would be the directional assimilation type of rules that would require some language-specific reason for adoption. One might therefore after a l l agree with Lightner that a root-marker does in some sense capture the notion of harmony more adequately and is to be considered as the basic format for the description of vowel harmony. Arguing against root marker, Bach suggested the obvious advan-tage for Classical Mongolian of the neighbourhood convention, an abbreviating rule device both capturing and generalizing the order of environment in assimilation processes, which would obviate the need for a decision as to progressive or regressive assimilation, should velar consonant harmony not, indeed, be a universal phonetic rule. However, the main objection to root marker is in the use of the ad hoc morpheme marker which is only spuriously related to the phonological features. According to Bach (1968: 144), the morpho-logical feature {GRAVE} bears no more relation to the phonological feature of gravity than would, say, the NOUN or ANIMATE. Moreover, unlike the latter features i t plays no role in the language outside of just this rule. - 125 -The main thrust of Kiparsky was to restrict the power of generative theory gained through abstractness by the use of the alternation condition which would end the use of devices such as archiphonemes and morphological markers. In c r i t i c i z i n g root marker as a quasi-phonological solution, similar in approach to f u l l y abstract phonology, Kiparsky, while concurring with Zimmer and Bach, specifically noted that the method overgeneralized by recognizing only neutral vowels, thus cutting the total number of vowels in half. IV.5.2 I n i t i a l Vowel Method The i n i t i a l vowel method, with a specified f i r s t vowel in a morpheme occurring with vowel archiphonemes, and a rule of progressive assimilation, is the traditional way of treating vowel harmony (v. Lightner 1965: 245 n.4). Preferred by Troubetskoy (1958) and others in describing vowel harmony in Finnish, Hungarian, Turkish and Mongolian and related languages, the i n i t i a l vowel method has been adopted as well by Zimmer and Bach. Examples A. Vowel Harmony - Finnish An example in Finnish illustrates the l i s t i n g in the lexicon with the f u l l y specified f i r s t vowel and the neutralized, unspecified archiphonemes: / pSUtA / / poUtA / - 126 -Using the neighbourhood convention, the suggested assimilation rule (adapted from Bach 1968) would specify the gravity feature of the f i r s t vowel to following vowels: ASSIMILATION r, 1 i n r.. , , RULE ^ S y ^ ^ g r a v e J ' |+ s y l l a grave' X A syllable agrees in backness with the preceding (and following) syllable. Under the neighbourhood convention, the f u l l y expanded rule applies regressively and vacuously. A lower-level rule would assign specific values to the archiphonemes involved with the same forms resulting as with the root marker method. B. Vowel Harmony - Turkish A second example (Bach 1968: 140) u t i l i z i n g the i n i t i a l vowel method is furnished in Turkish where both gravity vowel harmony and rounding assimilation of high vowels in certain contexts occur. Two forms show the lexical l i s t i n g with a differing specified i n i t i a l vowel: /geldlklErl/ "their having come" /gUldlklErl/ "their having laughed" The progr The progressivesassimilationaruiLe belowb(.from(Bach)Billustrates the transfer of the backness feature to the archiphonemes while a further rule furnishes the realization of the archiphonemes. ASSIMILATION r, ^ r -, , R U L E [+ syll] > [a grave] / + s y l l a grave X - 127 -A syllable agrees in backness with a preceding syllable. Another rule i s necessary to determine-rounding, which, occurs with amiimmediately 'preeeding^higE :y.owel£ , giving the resulting formscrms% /geldikleri/ /guldUkleri/ Discussion Lightner, while setting forth the root marker method, criticized the i n i t i a l vowel method as being arbitrary in a choice of progressive assimilation. Zimmer, noting that progressive assimilation was a natural process, commented that such assimilation could account equally well for consonant harmony in Classical Mongolian and Turkish, while being a more convenient and satisfactory device for Turkish la b i a l harmony. However, in terms of simplicity and economy of statement, the i n i t i a l vowel method might be less successful than root marker. Bach was highly prejudiced in favour of assimilation rules in vowel harmony with the use of the proposed neighbourhood convention to account for both back and front assimilation. While acknowledging the possible existence of languages in which either the f i r s t , or the last, vowel might be independently specified, Bach (1968: 147) suggested that: The use of the neighbourhood convention for situations like vowel harmony, in short, expresses the view that (at least in situations like vowel-harmony) the direction is predominantly progressive, possibly progressive and regressive (when one value i s dominant like in Fogny or - 128 -apparently, Nez Perce, Aoki:1966: or when other lexical elements determine affix elements as in Igbo, Ijo, etc.) but never purely regressive. Cognizant of the d i f f i c u l t y in standard generative theory in having identical constraints in both morpheme structure and phono-logical rules, Bach (1968:142) noted that "within current phono-logical theory i t is impossible to use one rule both to f i l l in redundant specifications and switch features as in Diola-Fogny." The i n i t i a l vowel method, i i keh feheor iO.ot'amaEkeryawasrcriticized by Kiparsky for violating the alternationr.condition, and for being unworkable under markedness theory where vowels must be f u l l y specified in the lexicon (v. Stanley 1967), and where lexical representations must not violate Chomsky and Halle's well-formedness conditions. Moreover, arguing for the necessity of morpheme structure conditions and phonological rules in Finnish, Turkish and Hungarian, where root and affix harmony have different sets of exceptions, Kiparsky (1968:31) has stated that " i t i s impossible to derive both root harmony and affix harmony by a single rule", thus rendering root marker and i n i t i a l vowel methods invalid. Specifically, the i n i t i a l vowel method in Finnish, for example, would require odd morpheme structure conditions where vowels under-lying neutral 111 and /el, i.e., /*/ and /a/ (vowels of absolute neutralization), would be required to appear with at least one occur-rence of archiphonemes A,0,U so that morphemes such as */k±vE/ and */ k o l l / would not occur. - 129 -IV.5.3 Absolute Neutralization A third method, which has been termed concrete (v. Kiparsky 1968), and which certainly i s less 'abstract' than the two previous solutions, but which, following the definition and usage of Hyman, we term abstract, uses phonological features such as d i a c r i t i c markers to set up underlying, phonetically unrealized, representations. Examples A. Vowel Harmony - Nez Perce An example of the absolute neutralization method is furnished by Kiparsky's 'concrete' solution of dominant versus recessive vowel harmony in Nez Perce where an underlying, [+ dominant] /e/, later neutralized phonetically to / i / , is established to distinguish between the asymmetric vowel harmony function of the two kinds of 111. In this system of dominant versus recessive vowel harmony in morphemes, an underlying six-vowel system, considered historically accurate, of dominant vowels /a,o,a/ paired with corresponding recessive vowels /ae ,u,i/ i s distinguished. Two rules, a vowel harmony rule changing feature values of recessive vowels when occurring with a dominant vowel in a morpheme, and a later neutralization rule,aaccount for the change to dominant feature values in Nez Perce. ASSIMILATION RULE + back - high / * # x + back - high A vowel in a morpheme corresponds in dominance to i t s paired dominant - 130 -vowel when a dominant vowel occurs. ABSOLUTE NEUTRALIZATION RULE low round - back + high A low-level surface rule changes /s/ to / i / . B. Vowel Harmony - Nupe Another abstract solution is illustrated in Nupe (Hyman 1970) where, in order to explain nativization of foreign words, redupli-cation of certain forms, and certain distributional restrictions, underlying /o/ and /e/ from /a/ are posited. For example, distribu-tionally in Nupe, the consonant clusters /Cw/ and /Cy/ occur only before /a/ and /a/; so that the forms /egwa/1 "hand" and /twa/1 "to him" /egya/2 Vblood" /tya/ 2 "to be mild" /ega/3 "stranger" / f c^3 " t o t e l 1 " occur. If, as Hyman has reasoned, there aretthree kinds of /a/: one which causes labialization, e.g., underlying one which causes palatalization, e.g., /e/, and one which causes no effect, then the surface clusters are explained by a later rule of absolute neutraliz-ation which account for the surface appearance of the neutralized segments. Thus, according to Hyman, with the underlying forms /ego/, /ege/, and /ega/, surface forms can be accounted for by three rules, - 131 -a labialization rule, a palatalization rule, and a rule of absolute neutralization. ABSOLUTE NEUTRALIZATION, RULE - .; + low + back - round Underlying low vowels are realized as a low back surface vowel, > 1 Discussion The method using, absolute neutralization was c r i t i c i z e d by Kiparsky on the basis of markedness (vowels must be f u l l y specified and thus predictable in the lexicon) and the strong alternation condition. However, absolute neutralization might be acceptable to a weak version of the alternation condition, were rule features not available, and were the linguistic cost not too great. Absolute neutralization in Nez Perce (Kiparsky's 'concrete'), simple and historic by rule, less simple than the process solution, does not satisfy the basic alternation condition. In regarding-an abstract solution of absolute neutralization in Nupe as being more explanatory with regard to "the limitations on surface distributions of the bi-unique phonemes", Hyman intended with manipulation of Nupe data, to demonstrate that abstractness and absolute neutralization were more productive and psychologically real. Harms (1973), in rejecting both the notion of morpheme structure and absolute neutralization in Nupe, while positing the underlying - 132 -vowel clusters / i a / and /ua/ to account for Hyman's neutralization, suggested the recognition of surface-structure constraints to i l l u s -trate skewness of consonant palatalization and labialization before [a]. For those supporting the existence of morpheme structure, however, there is evidently no "non-arbitrary basis" for deciding between the generality of morpheme structure conditions (consonant and glide clusters) over absolute neutralization of /o/ and /e/ in the phonology. Harms also rejected a supposed nativization as offering support for absolute neutralization, while suggesting that Hyman's stated palatal-ization rules are not correct. In responding to criticisms of his analysis of Nupe, Hyman (1973) suggested that The challenge 'of Harms 1973 to my abstract Nupe solution (Hyman 1970) i s not so much a principled account of how Nupe must be analysed as i t i s a negative demonstration of our frequent inability to provide sound empirical evidence concerning the nature of phonological systems. Hyman further reiterated that loan words might indeed provide strong empirical support for absolute neutralization and that economy and naturalness were not satisfactory substitutes for badly needed psycho-logical evidence. IV.6 CONCRETE PHONOLOGY Reviewing the basis of a concrete phonology which, for Nupe, Hyman (1970) defined as phonemic, we seek the roots in phonemic theory. - 133 -In an approach similar to Kiparsky, Fudge (1970) has suggested four main classes of various approaches to the phoneme, including the mentalistic or psychological views of Baudouin de Courtenay and Sapir; the physical approach of Daniel Jones; the functional view of Bloomfield; and the abstract view, again of Sapir, Hjelmslev, and Fudge. According to Stankiewicz (1976:33), in Baudouin de Courtenay's psychological viewpoint a speaker aims phonemically at an ideal acoustic image composed of acoustic and articulatory elements "divorced from the actual speech-sounds". Idiolectal variations, in Fudge's opinion, cannot be accounted for. In the words of Baudouin de Courtenay: The phonemes consist of ultimate psychological (articulatory and acoustic) elements ... [1910, 267] which "are not like separate notes, but like chords composed of several elements [1910, 271] ... (Stankiewicz 1976:32). In the physical view, the phoneme is regarded as a family of sounds (v. Jones 1950) phonetically similar, and in complementary distribution. Minimal pairs differentiate meanings as a corollary of the definition, while similarly pronounced words must be phonemic-ally similar or bi-unique, e.g., German Bund "federation", and bunt "coloured" (phonemically /bunt/). The functional viewpoint, converse of the physical (Fudge 1970), defines the phoneme as a minimal meaningful sound unit composed of distinctive features (Bloomfield 1933). Distinctive opposition is the basis of phonological structure in the work of Troubetskoy (1939) - 134 -and Jakobson (1956). The distinctive feature of voicing, for example, may differentiate meaning in the minimal pair /latter/, /ladder/. In the abstract view (cf Kiparsky 1968), the phoneme is regarded as independent of physical properties, able to account for idiolectal variation by the use of non-phonetic c r i t e r i a , e.g., the grouping together of alveolar 1 and velarized ("dark") ±, on the basis of morphological alternation. Bi-uniqueness is not a requirement of the abstract view. However, stemming from d i f f i c u l t i e s in the physical viewpoint of u t i l i z i n g grammatical information, a separate, and higher, level of morphophonemic representation (unrelated to the morphophonemic represent-ation of generative phonology) was established, though mostly undefined. For example, to the physicalists (v. Fudge 1970:90), German Bund would be morphophonemically /bunD/, where D is It/ word-finally, and /d/ elsewhere. And from the early distinctive feature work of Jakobson (Jakobson and Halle 1956, Jakobson, Fant and Halle 1952) has arisen the modified set of distinctive features and phonological representations, currently in vogue, of Chomsky and Halle. Although the phonemic and generative phonological theories, having their bases in opposing philosophical biases with methods, respectively, of induction (or discovery) vs deduction (or validation or decision) procedures, are not to be construed as similar, a similarity in view-point towards abstract, concrete and process has been noted (e.g. Kiparsky 1968, Fudge 1970) and reviewed. For, according to Gregg (1975: 139): - 135 -whatever our views on the matter of the phoneme - whether we regard i t as a concrete, practical unit, useful in the description of languages and dialects known or hitherto unknown, or whether we consider i t an abstract Platonic idea whose translation to the real world involves a series of Protean adaptations or adjustments to the phonological environment - in either case, the most important factor to be considered is distribution. Following Hyman we take note of the concreteness, which Hyman has not defined, of a phonemic solution to the problem in Nupe, suggesting that a concrete method tends towards a physical basis (v. Crystal 1971:179). IV.7 PROCESS SOLUTIONS The process method, according to Kiparsky (1968), recognizes the abstractness of morphophonemic representation while ascribing "an int r i n s i c representation on the phonetic level'. In treating vowel harmony, Kiparsky is bound by the alternation condition which in i t s strong form forbids absolute neutralization. Adhering to Postal's naturalness condition in which a phonemic representation must be as like as possible the phonetic representation, process morphophonemics, in conjunction with markedness and the use of rule features where an asymmetric functioning of segments to rule may occur, treats the level of morphophonemic representation as being ful l y specified, yet as appearing in i t s maximally unmarked form, that i s , in terms of universal phonetic specification, in i t s most neutral form. - 136 -Examples A. Vowel Harmony - Hungarian Using Hungarian vowel harmony as an example, Kiparsky gives a lexical l i s t i n g , using minus rule features, of two identical vowels in morphemes in which one vowel undergoes vowel harmony, while the other does not. (1) hej "rind" (2) kes "knife" [- Vowel Harmony] The vowel harmony rule affecting suffixes with basic back vowels would thus apply only to form (2) and the following forms would result: (1) hej-am "my rind" (2) kes-em "my knife" B. Vowel Harmony - Nez Perce An example of a process solution i s given by Kiparsky's treatment of Nez Perce. With a five vowel system of / i , 33, a, o, u/ -and the dominant vowels /a, o, i / , the rule changes a vowel in which the features backness and height agree to a plus back, minus high vowel, that i s in morphemes with dominant vowels: RULE V VOWEL a back HARMONY a high + back - high / * # X a back - a high - 137 -A vowel which agrees (plus or minus) in backness and height becomes back and minus high in the environment of a vowel that i s of the same backness, and an opposite degree of height. Discussion For Nupe, Hyman offered a process-type solution, using rule-exception features, to the problem of labialization and palatalization before /a/. Positing two rule-exception features [-LR] (labialization rule) and [-PR] ("palatalization rule), Hyman suggested that a single phoneme /a/ might have the variants /a/ [-LR], /a/ [-PR], and /a/ [-LR -PR]. However, the consonant labialization and palatalization rules would imply that a back vowel implies [+LR] and an unround vowel implies [+ER], which, without rule-exception features on each occurrence of /a/, would mean that a consonant would be both palatalized and labialized (an impossibility in Nupe) before /a/, According to Hyman (1970: 71) "the rule-exception feature forces us into an odd constraint, where every occurrence of post-consonantal /a/ in Nupe is an exception". Hyman also, inddiscussing implications of the alternation condition (one of the supports of process morphophonemics), decided for the greater explanatory power of the abstract view. Recently, theories of natural phonology and natural generative phonology have been discussed which would obviate the use of markedness theory - a further support of the process approach. At the level of descriptive adequacy, process morphophonemics according to Hyman (1970, 1973) apparently does not account satisfactorily for data agreeing with a native speaker's - 138 -intuitions. Sufficient discussion of process morphophonemics is yet to be forthcoming in the literature, although Ingram (1976), in discussing child language acquisition, has recently equated the level of descriptive adequacy or concern for significant generalization with Chomsky and Halle's investigations into markedness, with Kisseberth's functional rules, and the search for conspiracies; and the level of explanatory adequacy with the natural rules and processes suggested by Stampe (1977 forthcoming). IV.8 PROSODY The previous solutions have been within the framework of trans-formational generative phonology. Howeverf,Lyons (1962) has discussed J.R. Firth's London school prosodic approach to vowel harmony. Briefly, in describing data, the prosodist distinguishes a multilevel design of system vs structure, with items described in terms of contrastive phonematic units, and prosodies which are akin to phonetic long components (cf. Lyons 1962). In Turkish vowel harmony, for example, binary prosodic contrasts of front vs back and round vs non-round effect the phonematic units of high / i / and low /a/. In an eight-vowel, two-height system (where / i / = high vowel and /a/ = low vowel), with phonemic vowels / i , e, ii, o, l , a, u, o/, the contrasts are realized prosodically (where F=Front, B=Back, R=Round, and N=Non-Round) as follows: - 139 -FR gazlar and phonemically as /gozler/ FN avlar /evler/ BR k a l l a r / k o l l a r / BN adamlar / adamlar/ The prosody, except for R which a f f e c t s only the f i r s t s y l l a b l e when low, covers the whole word, which i s independently defined by s t r e s s . Lyons suggested the advantage of a prosodic approach to avoid redundancy of statement, and to capture generally the front-back, round-non-round opposition of vowel harmony i n the native l e x i c o n . Although the prosodic statement cannot be judged i n generative terms, prosody o f f e r s a simple, economical and abstract d e s c r i p t i o n of Turkish vowel harmony at the l e v e l of observational adequacy. IV.9 EVALUATION CRITERIA Within generative theory, according to Kiparsky, a choice of an abstract, concrete or process s o l u t i o n i s e n t i r e l y a r b i t r a r y lacking an adequate evaluation measure. Where, i n evaluating phonemic descriptions, the c r i t e r i a of contrast and complementation, phonetic s i m i l a r i t y , pattern congruity, and economy (cf. Hockett 1958) were fundamental, i n early generative phonology notions of s i m p l i c i t y , economy and generality were valued. For example, Halle (1959) suggested that phonological rules should be le s s complicated than morpheme structure conditions, with a conse-quent cost to the grammar should the opposite be necessary. In recent - 140 -years, however, generative phonologists have shifted away from feature-counting and the simplicity metric to the newer concept of naturalness. According to Anderson (1974: 79): It has often been argued (e.g. Halle 1962; Chomsky and Halle 1965; Chomsky 1967) that linguistic theory must provide not only a framework in which a l l possible processes that might be found in a natural language can be expressed, but also some way to decide, given two or more distinct accounts of some linguistic fact or process, which is the most natural, in the sense of embodying the most linguistically.ysignificant general-izations. And as Hyman (1970: 59) suggested in discussing an increasing non-uniqueness of solution: We obviously cannot tolerate such a variety of solutions in a theory that claims explanatory adequacy as i t s goal, since we have assumed that there is only ONE correct grammar. It i s this one grammar that we shall c a l l 'psychologically real'. Therefore, we must find some appropriate means to t e l l us which one of these solutions is the most highly valued in a given set of circumstances. Though Hyman (1970, 1973) has reiterated that psychological reality and psycholinguistic evidence are central in deciding among alternate descriptions, c r i t e r i a in phological description of simplicity, economy and naturalness w i l l be briefly outlined. IV.9.1 Simplicity The simplicity metric (cf. Halle 1961) discussed in Bach (1968: 128; Zimmer 1969) is intended to "reflect generality of statement in - 141 -shortness of rules", while predicting a maximally simple phonology to the language acquisition device in child language acquisition. However, using English, Bach pointed out that the simplicity metric, a measure-ment of the predictability of rule correctness by the element of natural class and feature counting, may not always account for the most significant generalization or the most natural rule. The simplicity metric, as such, has recently been discounted in favour of newer proposals: of markedness theory where universal conventions state the most likely feature change to occur on the basis of universal phonetic mapping; of natural processes (Stampe 1972); of natural generative grammar (Vennemann 1972); and of psychological reality and productivity (Hyman 1973). IV.9.2 Economy Economy in generative phonology (cf. Halle 1962) refers to the frugal counting of features in a rule so that the most highly valued solution is that rule most economic in feature use. However, according to Harms (1966: 602): In evaluating possible analyses of phonemic systems within a distinctive feature framework, the average number of bits per phoneme is often taken as a measure of economy... That such a scale of economy cannot serve to evaluate phonological systems which are integrated into the total grammar of a language has been recognized in principle (Halle 1959: 29-30, 45) but is not always considered in practice. Although Halle (v. Bach 1968: 610) considered that economy of phono-logical rule must take precedence over economy of lexical representation - 142 -without great cost to the grammar, Stanley (p.c. reported by Zimmer 1969) suggested that logically, in this view, i t would be most economic to forego phonological rules entirely and to l i s t a l l the morphemic variants in the lexicon. However, as Harms pointed out, in generative phonology lexical features cannot be rated on the same scale as morphophonemic rule features. According to his interpretation of Halle's proposal i t would appear that neither lexical economy nor morpheme structure economy should be considered at a l l . According to Harms (1966: 610): A reliable answer - and an effective evaluation procedure -can be found only after our present meagre knowledge of individual languages in these respects has been substan-t i a l l y bolstered. Therefore, because of the d i f f i c u l t y in evaluating simplicity and economy effectively (though these are lip-service c r i t e r i a which linguists may s t i l l pay), newer forms of evaluation have been . suggested. For example, Hyman (1973) rejected economy and natural-ness in favour of psychological evidence, while Zimmer (1969: 97) has pointed out that: An evaluation measure for grammars that relies exclusively on feature counting would have made what is clearly the wrong choice -.. primarily because i t would have had no way of recognizing the oddness of the rule ... Thus, these newer c r i t e r i a stress naturalness in rules, processes, or even in the whole generative grammar. - 143 -IV.9.3 Naturalness Naturalness has been a concern of generative phonologists following a shift away from the earlier evaluation measures of simplicity and economy. The naturalness condition, stated by Postal (1968) in a tirade against autonomous phonemics, contains the notion of a natural relationship between phonological and phonetic levels. Anderson (1974: 50), in adhering to the naturalness condition, has noted that a morphophonemic representation occurs in a ful l y specified and maximally natural form, which deviates from the occurring phonetic forms only insofar as such is dictated by the need to optimize predictability of variation and to capture generalizations about phonolggiealsstructure. Markedness theory, with the userof marking conventions and linking rules, has been an attempt to capture naturalness by positing universal values. However, major criticisms of linking rules in application and evaluation have been voiced by Bach and Harms (1969), who suggest (p. 6) that "what we are concerned with is the suggestion that plausibility constraints should be reflected directly in the evaluation metric." Marking conventions, while apparently capturing phonetic naturalness, can not provide an adequate evaluation of simplicity in acquisition. Phonetic plausibility i s often taken as a measurement of natural-ness. For example, Hyman (1975: 97-8), in discussing phonetic plaus-i b i l i t y hypothetically, notes that i t is entirely plausible in a language for alveolar [s] to palatalize to [s] before front vowels - 144 -[i, e ] , far less l i k e l y in an alternate solution for [s] to become [s] following back vowels [u, d> a]. Rule plausibility therefore refers to phonetic naturalness, including ease of articulation and usually unidirectional assimilation. According to Hyman (1975: 161): While the study of rule naturalness is in i t s infancy, i t i t clear that naturalness i s not a binary property. Rules are more or less natural or more or less unnatural. IV.9.4 Alternation Applying to absolute rather than contextual neturalization, Kiparsky's alternation condition has raised yet unresolved issues on abstractness which seemingly hearken back to similar disputes on phonemic realization among the post-Bloomfieldian structuralists. The alternation condition may be strong, the total exclusion in generative theory of absolute neutralization, or weak. The weak alternation condition suggests that not a l l cases of absolute neutral-ization w i l l be excluded: Two conditions, that phonetically identical morphemes must have the same underlying representation (to avoid the method of absolute neutralization); and the converse, that phonetically distinct morphemes must have different underlying representations (to avoid root marker methods), are included. The weak alternation condition may well be, according to Kiparsky (1968: 23), the clause of an evaluation measure which says among other things that absolute neutralization is l i n g u i s t i c a l l y complex. In that form the alternation condition would, in any given case, balance out the general-izations gained by absolute neutralization against some - 145 -fixed cost assigned to i t in phonological theory. The relation of absolute neutralization between underlying and phonetic representations would s t i l l be excluded in cases such as those analyzed above, where rule features do just as well, but i t would be allowed where this is not the case. IV.10 VOWEL HARMONY METHODOLOGY Because of the alternating patterns of roots and affixes in vowel harmony languages, vowel harmony offers an interesting way of testing abstractness. As Kiparsky (1968: 46) has suggested: Vowel harmony systems are of special interest, because many of their apparently quite strange features can be explained by the complex interaction of several rich and detailed phonological universals. Neither marked-ness, nor the alternation condition, were set up with vowel harmony in mind. Yet they jointly lead to just the right solution in many different kinds of cases. As far as methodology in treating vowel harmony in generative phonology is concerned, i t has been generally considered that the same rule which determines vowel harmony in alternating suffixes i s also responsible for accord in the root, with vowels unspecified in the lexicon for the alternating feature. A problem has been the categorizing of roots, whether by a completely abstract marker or by vowel assimilation (v. Kiparsky 1968: 28). In discussing vowel harmony methodology in Tli n g i t in accordance with abstract, concrete and process theory, we w i l l examine each approach in turn. These methods w i l l be described with references, illustrated with examples, and fi n a l l y c r i t i c i z e d as to economy, simplicity, and predictability within generative theory. Under - 146 -abstractness, the most detailed, three methods of application w i l l be discussed: the root marker, the i n i t i a l vowel, and absolute neutralization. Under concrete, the previously discussed Tlingit labial harmony w i l l be featured, while under a process heading the use of rule features and markedness w i l l be discussed. I n i t i a l l y , however (following Aoki 1968), vowel harmony typology w i l l be briefly reviewed. IV.11 VOWEL HARMONY TYPOLOGY In a tentative typological classification of vowel harmony to be used in conjunction with markedness, Aoki (1968) suggested c r i t e r i a of opposing features of total versus partial harmony, symmetry versus asymmetry, alternating versus non-alternating systems, while rejecting c r i t e r i a of neutral vowels and direction-a l i t y . According to Aoki's typology, in total harmony a vowel is specified only for the feature plus vocalic (fully abstract), while in partial harmony the features are partly specified. Unspecified features are provided by p-rules that are either of the root-marker (prosodic) or assimilation type. Sub-types of partial harmony are l a b i a l , horizontal and palatal harmony. Labial harmony refers to rounding assimilation, which generally occurs secondarily in Aoki's typology with either horizontal or palatal harmony. Palatal harmony refers to vowels unspecified with regard to backness. In Tlingit, this feature depends on the analysis - 147 -preferred, e.g., archiphoneme or root marker would need the gravity feature specified. Horizontal harmony refers to height, tension, and tongue root position. Symmetry implies that the harmonizing vowels (e.g., i , u) are of equal power with no system dominant, as in Nez Perce or Koryak. Alternating refers to the number of features which are l e f t unspecified in the lexicon, an example of a non-alternating system being internal harmony in Turkish. Using Aoki's typology of vowel harmony, Tlingit may be character-ized as evincing partial, symmetric (with idiolectal restrictions previously noted), l a b i a l and palatal harmony with respect to certain lexically-marked morphemes. As previously discussed, the effect of this vowel harmony, as described by Boas (1917), is to round and back palatal affix segments following rounded, back segments across a morpheme boundary. Nouns ending in /a/ also may cause rounding of certain suffixes. The following affixes listed by Boas (1917) are affected: /-yi'v.-wu/ ^'possessive" /yutt^ wWA-/'" verbal prefix (this labialized only after /u/) i t w /-yinv-wun/ "past" /-k^-k / " l i t t l e " {diminutive attached to nouns) The prefix /ya-/, of complex morphological conditioning, w i l l not be discussed. The problem is to discover whether any of the theoretical solutions—abstract?., concrete or process—outlihedd in this chapter presents a satisfactory, or even a unique, solution to the problem of vowel harmony in T l i n g i t . - 148 -IV.12: • ABSTRACT TREATMENT IV.12.1 Root Marker In a root marker solution of Tlingit vowel harmony, the affix rather than the root must be marked for harmony. By convention, each segment of the affix w i l l thus be marked for the morpheme-sized feature grave where, in terms of the grammar, lexical items w i l l be termed [-Affix], the rest [+ A f f i x ] , an unmarked universally predict-able feature (v. Bach 1968). There are several variants of root marker method: that of Lightner, reworked by Zimmer; and that of Chomsky and Halle. Their suggestions been adapted for Tli n g i t . Rule Data Example Lightner (1965) - cons + high [a grave] du-t'a - y i ,- GRAVE du-l'uk W-YI + GRAVE Zimmer (1967) - cons [a grave] / [a GRAVE] ?uk^-yln + GRAVE + high -> Affix where [a grave] = [a GRAVE] These similar rules state that a glide or vowel or marked segment /a/ in an affix i s marked to agree in backness with the morpheme marker. - 149 -Discussion Lightner's rule distributes the feature of the ad hoc morpheme marker {GRAVE} to the affix marking each segment [ a grave], e.g., Xn)I(n) [a grave]. [a g r a v e ] With only one nasal, grave feature specification of /n/ is redundant, and variation i s allophonic. In terms of data marking: du-t'a-YI du-±'ukW-YI [- grave] [+ grave] A later rule applying to a l l solutions (1), (2), and (3) would specify the [+ grave] archiphonemes Y as [w] [- grave] Y as [y] [+ grave] I as [u] [- grave] I as [ i ] . Zimmer's improvement formalizes Lightner's convention that each phono-logical segment specified i s associated with the abstract marker (of the affix) and thus would simplify Lightner's rule. Chomsky and Halle (1968) In their treatment of Nez Perce with i t s three dominant vowels, Chomsky and Halle offer a readjustment rule which distributes a [+ dominant] feature to a l l segments of a word occurring with a [+ dominant] morpheme. Using gravity as the distributed feature, and - 150 -archiphonemes, e.g., YI, a morpheme structure condition would specify round segments to agree in backness with the feature round, and thus markethe feature of the affix (v. Kiparsky 1968: 39), e.g.: qtJxWXu-YIn qtJxWAki£-YIn [+ round] [- round] Morpheme Structure Condition - Affix Marker [a grave] / # X [a round] + } Affix a round - cons -> A phonological rule states that gravity and backness agree in high vowels and glides: Rule— Vowel Harmony - cons + high [a back] / [a grave] A third rule is needed to stipulate the realization of the archiphonemes Y I. Comment A Lightner-and-Zimmer kind of solution apparently may be applied to Tlingit with some success i f one is willing to overlook the use of what previously has been termed an ad hoc and unrelated morpheme feature (v. Kiparsky 1968). Problems with root marker have been discussed in detail in the preceding section. A root marker solution in Tlingit captures a certain generality in glide and vowel harmony. - 151 -Chomsky and Halle's suggested solution, actually a morphological marker, entails two or three rules and a certain intricacy in rule-writing which suggests that the rules are none too natural. In addition, the f i r s t rule i s actually a morpheme structure condition which marks the affix for the specified feature. No significant generalization i s captured by the rules. The problem with applying a root marker solution to Tlingit is that the affected affixes are most certainly the result of progressive assimilation, and this can not be formally shown by a prosodic solution. However, in the style of Lightner and Zimmer, a root marker type with harmonic affixes specified in the lexicon would appear to be a workable, and perhaps even psychologically real, solution. IV.12.2 I n i t i a l Vowel An i n i t i a l vowel solution which acts in terms of a progressive assimilation would cause problems in Tlingit vowel harmony in those cases where /y/ is not removed directly following a consonant, although one could, in the s p i r i t of terminological quibbling, choose to change the theory to I n i t i a l Glide and Vowel, in which case (rule 3), a simple assimilation 'rule', would suffice. A persistent rule i s also necessary, e.g., Y > 0 / C + . Rule 1 -> [a back] / [a round] + - 152 -This rule accounts only for vowel agreement in backness and rounding across a morpheme boundary. Rule 2 AGREEMENT [ " S y l l ] > [ a r o u n d ] / [ a round] + _ Where the glide is not omitted, rule 2 must apply, then rule 1. A glide is rounded following a rounded segment across a morpheme boundary. Rule 3 GLIDE OR VOWEL AGREEMENT - cons (+ high) -> [a back] / [a round] + A glide or vowel agrees in backness and rounding across a morpheme boundary. Rule 3 applied persistently would obviate the need of Rules 1 and 2 and would provide a satisfactory solution: Data du-Saxaw-YI "his hair" Persistent rule du-guxW-YI "his slave" Y i 0 / C + du-lu-YI "his nose" du-?ia-iYI "his lake" A later rule would specify the realization in backness of the archi-phonemes. A rule 3 solution in Tlingit i s certainly plausible and natural. However, plausibility of this method would have to be determined in - 153 -terms of productivity, since there might be a more explanatory or even more descriptive solution. Alternatively, an i n i t i a l vowel solution could also have been approached in terms of both progressive and regressive assimilation when the glide is not deleted. Example: Rule 4 [+ syll] > [a back] / [a round/ + Rule 5 [- syll] > [a back] / + • [a back] This solution is obviously less economical than Rule 3 above. IV.12.3 Absolute Neutralization From the Boas data, i t i s evident that the voiced post-palatal continuant /y/ and the voiced palatal glide /y/ have but lately merged in southern Tl i n g i t . Indeed, Boas gives the affixes / y i / and /yin/ as [yi] and [yin], orthographically <yi> and <yin>. Thus the continuant may be assumed to be the underlying form of the glide /y/ in certain marked affixes, although the continuant [y] i s , of course, not the underlying form of a i l palatal glides, nor do a l l affixes with i n i t i a l [y] automatically undergo vowel harmony. The glide /w/ i s tentatively marked as a labialized palatal continuant in Boas' chart. Because the segment /a/ in certain nouns also causes rounding of + s y l l Rule 1(a) CONTINUANT BACKING + voiced + cont •> [a round] / [a round] + - s y l l - 154 -A voiced continuant i s rounded following, a rounded segment across a morpheme boundary. Rule 1(b) ABSOLUTE NEUTRALIZATION + voiced + cont - s y l l -> [- cons] The underlying palatal continuant loses i t s consonance. Rule 2(a) + cons + dim DIMINUTIVE J t The- /-k/ suffix rounds following /o/, -> [+ round] / + s y l l + low + round + Rule 2(b) ABSOLUTE NEUTRALIZATION x. e. o + s y l l + low + round -> a -> [- round] As far as a Tlingit solution i s concerned, the absolute neutral-ization of [y] and of /y/ provides a neat explanatory device for round-ing, b'£. the affected affixes although,, since other affixes may also be postulated to have i n i t i a l [y] (as shown orthographically) and do not round, a phonological explanation i s not the complete answer. The answer possibly is to be found in the particular kind of syntax involved. Though.Kiparsky's criticism of the d i a c r i t i c use of features perhaps may be ju s t i f i e d , in T l i n g i t , u n t i l the recent past, a surface representation of the abstract feature [y] occurred which i s now - 155 -posited for rounding for certain probably gramatically marked forms. With regard to /o/ in Tlingit phonology, there are only two back vowels, with tense and lax manifestations, and i t is entirely possible that an o-type of vowel may previously have been neutralized. Though diachronically the positing of such underlying forms may be correct and well motivated, the assimilation rules, while plausible, are neither economic nor particularly simple, while natural solutions are available. IV.13 CONCRETE TREATMENT A concrete solution to the problem of vowel harmony in Tlingit within the confines of generative phonologys would demand that no underlying forms are posited that are not phonetically realized, and that a l l features be f u l l y specified in the dictionary representation. With the aid of minor rules (v. Lightner 1968) to account for those affixes which do not undergo vowel harmony, a possible solution is offered. Lightner (1968: 70) has described the use of minor rules: The convention for application of a minor rule i s that no form i s subject to application of a minor rule unless the form is specifically marked as undergoing a certain rule. In the case of minor rules, redundancy rules would have to specify that certain groups of morphemes are exceptions in that they undergo minor rules. The use of minor rules (i<?e. + rule features) in a concrete solution i s suggested by Kiparsky's strong alternation con-dition which allows for the categorization of elements into regular and exceptional classes with respect to phonological rules. - 156 -The following, rules would account for the data in a concrete solution where the affix is specified as undergoing vowel harmony •£•*- V.H. ]. Minor rules would account for these exceptional cases without the need of postulating underlying forms. Rule PROGRESSIVE ASSIMILATION + cons + high r JI + back 1 ' + round + High segments back and round across a morpheme boundary following a rounded segment or a segment marked ..for vowel harmony. In addition, nouns ending in /a/ which cause labialization would have to be marked in the lexicon as [+ V.H.]. Underlying Forms yAnAlge*n-yin [+ V.H.] qUx Au-yin [+ V.H.] h i t ' t'aq W - y i [+ V.H.] "he was getting big" "I was dwelling" "house timber" a -k [+VH][+VH] qa '-yi [+VH][+VH] na -yin [+VH][+VH] "a small pond" " i t s man" "having died" The concrete method, provides a neat and economical solution to the problem of vowel harmony in Tl i n g i t . The use of minor rules to mark exceptional forms is economical in Tlingit where only a few forms are so marked. This solution provides naturalness and simplicity, as well as sufficient motivation and generalization, although i t lacks psycho-linguistic insights of a native speaker. - 157 -IV.14 PROCESS TREATMENTS A process treatment of Tlingit vowel harmony requires that every stem vowel and consonant be specified in the lexicon, while the variable vowel and glide ' archiphonemes.' are represented in their " f u l l y specified maximally unmarked" forms according to the theory of markedness, the alternation condition, and naturalness (v. Kiparsky 1968: 39). Rule features of the suggested [-rule feature] type would account for sound change of the unmarked features. The systematic phonemic level is related to universal phonetic features, f i r s t according to universal principles, and second by means of language specific p-rules which may shift one or more mapped features (Postal 1968: 60-2). There is no unmarked high vowel. The use of / i / i s arbitrary. A high segment rounds following a rounded segment.or a minus V.H. rule Rule VOWEL + cons HARMONY + high •> [a round] + segment across a morpheme boundary. Data (1) hit taq W-yi (2) xuc ha^-yi II II house timber grizzly bear's dung' II (3) a-k II a small thing II - 158 -(4) xat hin-yi "salmon creek" i (5) t'a-k "a small spring salmon" As seen from the data, in order for the rule to apply to dimin-utive forms in nouns following, /a/, a [- rule feature] would not be satisfactory or even workable, and a [+ rule feature] or minor rule must be involved instead e.g., qa-yi " i t s man1 [+ V.H.][+ V.H.] Minus rule features are not satisfactory in Tlingit in any case since so many features would have to be marked as [- rule feature], and i t is the exception in this analysis that must undergo the rule. In order to have the rule account for a l l forms, the rule must then be amended as follows: Rule T+ r n n * l 0a roundj} [+ F.B.] (1) + cons + high -> [a round] / + seg + V.H. A process treatment is satisfactory in adhering to naturalness and alternation, although a l l treatments so far have ut i l i z e d marked-ness marking conventions as an abbreviatory device. However, although lyl is the maximally unmarked glide under markedness, there i s no one maximally unmarked high vowel, according, to Chomsky and Halle, and one must arbitrarily assign / i / as the unmarked high vowel i f one rule i s to apply. Outside of this requirement, the rule i s simple and economical, - 159 -though the use of [+ rule feature] as a device rather than [- rule feature] is entailed. If the goal of the theory as set out by Kiparsky (1968:41) for processhmoEpKophdnemics.ais'ltoshandilehso.und change in the directest possible way, without either setting up either d i a c r i t i c features where an equivalent phonetic contrast exists, or setting up underlying phonological distinctions which have to be undone where there is no phonetic contrast then both concrete and process solutions handle the problem equally well. The concrete solution, however, does not violate the assigned values of markedness. IV.15 METATHEORY - AN ADDENDUM Basing his theories on a number of attested language-general assimilation rules, Schachter (1970: 342) suggested that the incor-poration of his universals "into the general phonological model can simplify the description of particular phonological systems and can explain otherwise idiosyncratic features of these systems" by extend-ing the notion of naturalness to the p-rules themselves, defining a metatheoretical model of assimilation universals which might be applied to vowel harmony. Schachter argued that the general theory, using the simplicity metric,, f a i l s to give explanatory adequacy by not predicting normal or natural rules with the same number of features as less natural, or universal, rules. For example, in rules of intervocalic consonant voicing and inter-consonantal vocalic devoicing, the former is not predicted as being a natural process. Schachter (1970: 343) defined - 160 -assimilatory p-rules as those "in which the values of some feature (or set of features) of one segment is changed to agree with the value of that feature (those features) in a neighbouring segment, or in neigh-bouring segments". This adheres to Postal's naturalness condition as well as to the idea of process morphonemics set forth by Kiparsky. Schachter postulated that the criterion of naturalness should thus be extended to certain p-rules and that in particular "a meta-theoretical distinction should be made between natural and unnatural assimilation rules". Two of these universals may be adapted to Tlingit vowel harmony (v. Schachter 19 7 0 : 344>-9). Fi r s t : "Feature values of non-vowels assimilate to those of adjacent vowels, rather than conversely." For example: (1) [- cont] > [+ cont] / [+ syll] (+ cont) [+ sy l l ] (+ cont) A A non-confcinuantiheGomesecont^ * of this type characterizes labialization and palatalization as normal assimilation types, following labial or palatal vowels. Second: "Unmarked feature values assimilate to adjacent marked feature values, rather than conversely." For example, a minus nasal feature assimilates to a marked nasal feature in the environment of a nasal: (2) [- nas] •> [+ nas] / [+ nas] - 161 -However, according to Schachter, because universal (1) (v. below) does not account for a l l cases of assimilation, universal (2) must be considered within the scope of (l),aarid as taking precedence over although not superceding (1). Using data from Tlingit and markedness, we can see readily how these generalizations would apply. With - y i , -yin as base forms, natural assimilation rule (1) correctly predicts rounding for the marked affixes following the feature [+ back]. However, i t does not predict rounding of affix vowels following a [+ round] segment, nor the rounding of base affixes after a [+ round] consonant, a l -though universal (2) w i l l . To allow for the f u l l application of universal (1) in this case, i t is necessary to forego the natural-ness condition in order to account for labializing /a/ as underlying /of.. r+ s y l l + back - high + round Universal assimilation rule (2) suggests that unmarked features assimilate to adjacent marked features rather than conversely. This is true for Tlingit in the case of [+ round] consonants, which are marked, and which cause consequent rounding (and backing) of following sounds. Data ii his nose II du-lu-yi dulu-wu II his fort II du-nu-yi du-nu-wd - 162 -"he never Xe*i-, wuduskti-yin Xii wuduskd-wun was known" I used to quj x AKit-yin qU, x Akit-in pick berries" Schacter's universals are certainly workable in T l i n g i t , a l -though universal (1) must be extended in some way to account for the assimilation of the affix vowels / i / ~ /u/, since in markedness theory neither /u/ nor / i / i s more, or less, marked than the other. Rather than tampering with markedness theory i t would probably be wiser to specify that, as a morpheme structure condition or as a surface phonetic constraint, - cons + high must agree in backness in certain specified affixes. The application of Schachter's metatheoretical model to Tlingit vowel harmony suggests, however, that analysis of /a/ as underlying /o/ might be a natural solution in which, as a result of the bleaching process (Stampe 1972), a restrained form of neutralization which depalatalizes and delabializes, vowels such as lol have been neutralized to /a/ by a completely natural and normal process of sound change. This apparently has happened phonologically in T l i n g i t . As previously discussed, Tlingit phonology manifests a number of natural processes suggested by Stampe (1969), e.g., f i n a l de-voicing, intervocalic voicing, limited clusters, lack of Unguals, (undefined by Stampe, Unguals refer to dentals, alveolars, palatals and velars in traditional terminology or to the distinctive features [+ coronal], [+ high] (v. Chomsky and Halle 1968: 304-305)), and in - 163 -the back vowel series the merging (one assumes for a symmetric chart) of [- high] vowels to /a/. Although recent developments in opposition to markedness, expressed by Vennemann (1971), Hyman (1973) and Stampe (1969), have brought within the scope of generative phonology some of the impli-cations of Schachter's assumptions of assimilation universals, Schachter's naturalness conventions, informally expressed above as universals, would, in the case of Tlingit vowel harmony at least, offer the basis of a more explanatorily adequate treatment and a better basis for judgement and problem solving. IV.16 THEORETICAL VALIDITY While offering no replacing world-view, the sociolinguist Labov (^^jl-o^S^glrainp'ropo'sMgtandItfo-tln'g the Tacksof tlfevgoricept of validity irislinguis.bicesci'enceughvasits^gge'sted that Just as impressionistic phonetics should be calibrated against the readings of various instruments, so the intuitions of the theorists should be matched against observations of the unreflecting speech or ordinary men. In noting the rejection of v a l i d i t y a (the independent judgement of right or wrong according to a measurable, observable basis outside of any theory) s by Chao 1934, Harris 1965 and Chomsky (1965), Labov has suggested that the competence-performance distinction blurs the reality that theories ought to be tested against real language. Noting that a generative grammar according to an internal evaluation is only one model of many that are descriptively adequate, Labov, on the basis of recent studies, has questioned the validity of certain convincing - 164 -linguistic analyses. Skousen (1972: 567) also, in disputing theor-e t i c a l , versus the actual, regularities captured by a native speaker, has confirmed the methodological problems involved in choosing a single solution or a single theory to a linguistic problem: In general, there is no single way to account for a set of li n g u i s t i c data. And given a set of static data, there is no a p r i o r i method to determine in what ways speakers might account for the data. A simple analysis might be better than a complex analysis, but only i f i t i s true that speakers would account for the data by means of the simple analysis. By just looking at static data, there is no way at present for a linguist to determine what regularities speakers w i l l capture. A reasonable goal of phonology would be, in fact, to develop such a theory, one that w i l l predict from a given set of data which regularities speakers would account for.. „ „ Within generative theory, Hyman (1970), in assuming a single, correct and psychologically real grammar at the level of explanatory adequacy to beethe end result of a transformational generative model, has demanded that means be provided to distinguish the most highly valued grammar in "a given set of circumstances". However, Lightner (1971), in an excellent theoretical review of generative phonology, has suggested that the effect of a phonetic representation on the underlying representation is d i f f i c u l t to formulate ex p l i c i t l y in generative theory and that under present theory the abstractness issue i s apparently deadlocked. In advancing the notion of linguistic indeterminancy, Lighner has suggested that one may have to accept the idea of an abstract theoretical grammar which w i l l account for a native speaker's presumed rather than actual competence, with yet undefined limits on the abstractness of grammar. - 165 -IV.17 SUMMARY Although Kiparsky raised the current controversy on the inclusion of abstractness of morphophonemic representation in the grammar, Hyman (1970) questioned the basis of whether explanatory phonological solutions may indeed coincide with a level of explanatory adequacy. In reviewing theoretical viewpoints or approaches to the issue, along with practical methods, applications, and solutions, we have also explored some of the possible solutions to the problem of vowel harmony in T l i n g i t , using Kiparsky as mentor and Hyman as touchstone. In discussing theoretical approaches to the issue of abstractness, given the present evaluation c r i t e r i a , we have been able to perceive, like phonemicists decades earlier, no one right perception in the present framework of generative grammar. Although Roberts (1976) has termed the abstractness controversy the result of ideological limits in linguistics, u n t i l , or unless, the constraints and conditions of natural generative grammar and of natural processes more widely replace the current standard generative model, the issue may remain at a standoff. Practically, in applying various theoretical methods to a solution of vowel harmony in Tli n g i t (the acceptance of which in i t s e l f has been arguable), we have discerned under the present c r i t e r i a of economy, simplicity and naturalness that certain solutions are perhaps more appealing, natural, normal, economic orpproductive than others. However, lacking the measure of explanatory adequacy which must be the unique solution to the acquisition of grammar in a - 166 -child, in this case of those rare Tlingit-speaking children, while possessing the intuitive insight of Boas' native speaker with regard to our interpretation of /-yi~-wu/, /-yin~-wun/ alternations, we can offer no preferred solution except methodologically in terms of the previously discussed evaluations of economy, simplicity and natural-ness. Therefore, factually and theoretically, what we can offer is in terms of our own analysis of data, our own interpretation of theory, our own determination of observational adequacy. Like the Saussurean paradox (v. Labov 1971), where the individual 'parole' must realize the social and ideal 'langue', because we are unique as individuals, we seek our unique solutions, while adhering to the ideal that there must be, of course, only one. - 167 -V.1 SUMMARY Methodologically, using pre-phonemicized data from Boas (1917), phonemic data from Velten (1939, 1944), and material from the tagmemic grammars of Naish (1966) and Story (1966), we commenced our analysis by investigating in the second Chapter the descriptive phonology of Tlingit in the framework of a transformational grammar. In this process, we noted and discussed certain problems arising from the application of derivational constraints and morpheme structure con-ditions to the yi~wu alternation in Tlingit phonology. In the third Chapter, we examined some of the simpler phono-logical processes of Tlingit according to a -framework of natural phonology and natural processes advanced by Stampe (1969), an analy-t i c a l approach we perceived to be particularly appropriate. We found that, unlike the undiscussed morphological complexities of verbal compositions (for which, v. Story and Naish 1973), the phono-logical processes cited in Tlingit are indeed simple and language-universal. In the fourth Chapter, while using the theoretical issue of abstractness to discuss approaches to morphophonemic representation according to abstract, concrete and process theory, we applied these various models to the problematical interpretation of vowel harmony in T l i n g i t , using formal c r i t e r i a of economy, simplicity, and natural-ness. In this context, we discovered certain problems applicable to linguistic science as a whole. With present evaluation measures, - 168 -and lacking the psychological reality and validated hypotheses of language acquisition deemed necessary, a non-unique solution oriented to the investigator's own interpretation of available data (no matter what the specific language at issue) and based on the constraints of the current grammatical model, as discussed, remains the inevitable option. To expand the range of discussion vis-a-vis this more general problem, we incorporated brief treatments of prosodic and metatheoret-i c a l approaches in order to give these issues further point. Theoretically, we discovered in the second Chapter that, while a phonemic framework featuring surface phonetic constraints provides a good descriptive f i t for the phonology, certain distinctive features described yield analytical insights into the opposition of features, while offering the necessary basis for the operation of the feature-changing phonological rules. In the third Chapter, we determined that Tlingit conforms to a natural framework of natural processes which in reality describes the language-universal processes involved. No claim to explanatory or even descriptive adequacy has been put forward at this stage, lacking any basis respecting a native speaker's insights into Tlingit, a matter clearly requiring specific and previously unattempted f i e l d research. In the fourth Chapter, we discussed theoretical problems of morphophonemic representation in generative theory with internal evaluation c r i t e r i a . In applying various theoretical approaches to methodological solutions we have found satisfactory, or better, - 169 -solutions, though by present theory no notion of validity has been incorporated. Here, we raised what might be termed for linguistics the validation problem to which (unlike certain companion social sciences) insufficient attention, in our derivations^ has yet been addressed. V.2 CONCLUSIONS More than forty years have passed since Chao 1934 f i r s t raised the problem of non-uniqueness in phonemic theory. Not only has this problem been unresolved, but the problem of non-uniqueness in generative grammar, s t i l l lacking in psychological evaluation c r i t e r i a , remains to perplex. As discussed in some detail, a number of theor-etical constraints and conditions have been proposed in recent years to deal with specific cases of non-applicability and thus indetermin-ancy. Thus, practical problems of workability and non-uniqueness frequently are jointly associated. The problem i s , of course, that given the limitations of the current, and generally established and accepted theory, no unique solution has been presented, or may even be possible, within the theory; for i t has appeared evident, given the options in the range of approaches, that certain solutions are more to be preferred than others in terms of the c r i t e r i a examined. In this context, Skousen (1972) has argued that "nearly a l l of the arguments for the theoretical machinery of generative phonology" have been based on the assumption that speakers capture natural, phonetically statable rules. Skousen <(19^2^§75')'rasserts further: - 170 -Yet a l l of the particular arguments for these notions crucially depend upon the assumption that speakers are actually capturing the regularities that linguists claim they are. If speakers are not really capturing these regularities, then any argument for a particular theor-e t i c a l notion that is based on such regularities is without foundation . . . A proper theory of phonology can only be based on rules that speakers actually capture. Skousen's injunction notwithstanding, matters cannot adequately be l e f t merely with some perhaps clearer definition of the problem of indeterminancy, for i t is entirely possible (though, intuitively, we would tend to doubt this would be the case) that the incorporation of psychological reality more firmly and consistently intoothe treatments we have discussed would s t i l l leave gaps to be closed in linguistic analysis, or create further areas which are resultantly opened up. While these matters have been indicated perhaps only implicitly, i t has been clear in the discussion throughout that linguistics (at least in the context we have elaborated) is in the process of funda-mental, perhaps reformative, attempts at synthesis. For this reason, we conclude this summary discussion with a suggestion of Lyons (1962: 131 n. 16) with respect to certain implications for linguistics generally which the analysis in Chapter 4 has perhaps indicated: Prosodists seem generally to prefer the 'hocus-pocus' philosophy of language . . . It seems clear, however, that on this question (as on so many theoretical questions of linguistics) i t is possible to adopt an intermediate position. Where the structure of the language is clearly determinate and a l l linguists can reach agreement on the 'facts', one might claim to be dealing with 'God's truth', but where the language is indeterminate and does not compel one analysis rather than another, the linguist may choose arbitrarily one of the alternative solutions, providing that he makes i t clear what he is doing. - 171 -REFERENCES AITCHISON, JEAN 1974 Phonological change: some causes and constraints. In John M. Anderson and Charles Jones, eds. Historical Linguistics II: Theory and Description in Phonology. Amsterdam: North Holland ANDERSON, STEPHEN R. 1972 On nasalization in Sudanese. Linguistic Inquiry 3.253-268. 1974 The Organization of Phonology. New York: Academic Press AOKI, HARUO 1966 Nez Perce vowel harmony and Proto-Sahaptian vowels. Language 42.759-767. 1968 Toward a typology of vowel harmony. UAL 34.142-145. AVERKIEVA, JULIA 1971 The Tlingit Indians. In Eleanor Burke Leacock and Nancy Oest-reich'Lurie, eds. North American Indians in Perspective. New York: Random House BACH, EMMON 1964 An Introduction to Transformational Grammar. New York: Holt, Rinehart and Winston 1968 Two proposals concerning the simplicity metric in phonology. Glossa 2.128-149. BACH, EMMON AND ROBERT T. HARMS 1968 Eds. Universals of Linguistic Theory. New York: Holt, Rinehart and Winston 1972 How do languages get crazy rules? In' Robert^t"; Stockwell and Ronald KJ S. jiMacaulay, leds. iLinguistic-Change and^Generative Theory: Essays from the UCLA Conference on Historical Linguistics in the Perspective of Transformational Theory 1969. Bloomington and London: Indiana University Press BLOOMFIELD, LEONARD 1933 Language. New York: Holt, Rinehart and Winston - 172 -BOAS, FRANZ 1911 Introduction. Handbook of American Indian Languages. Bulletin 40, Part 1, Bureau of American Ethnography. Washington, D.C.: Government Printing Office. 1911 Tsimshian. Handbook of American Indian Languages. Bulletin 40, Part 1, Bureau of American Ethnography. Washington, D.C.: Government Printing Office. 1917 Grammatical notes on the language of the Tlingit Indians. University of Pennsylvania Museum Anthropological Publications 8.1-179. 1920 The classification of American languages. American Anthropologist 22.367-376. 1929 The classification of. American Indian languages. Language 5.1-7. BORDEN, C.E. r 1945 Distribution,, culture, and origin of the indigenous.' population of V Brifisk-Gblumbia-. Transactions of the Seventh' Annual British Columbia Natural Resources Conference 7.186-197. BRENZINGER, INGRIDA - 1976.• Personal communication BROWN, GILLIAN 1969 Syllables and redundancy rules in generative phonology. Journal of Linguistics 6.1-17. 1971 On the function of redundancy rules. Journal of Linguistics 7.85-90. CAIRNS, CHARLES E. 1969 Markedness, neutralization and universal redundancy rules. Language 45.863-875. CANADA. DEPARTMENT OF INDIAN AFFAIRS AND NORTHERN DEVELOPMENT 1970 Linguistic and Cultural Affiliations of Canadian Indian Bands. Indian Affairs Branch. Ottawa: Queen's Printer. CATHEY, J.D. AND R.A. DEMERS 1976 On establishing linguistic universals: a case for in-depth synchronic analysis. Language 52.611-631. - 173 -CHAFE, WALLACE L. 1962 Estimates regarding the present speakers of North American Indian languages. UAL 28.162-171. 1965 Corrected estimates regarding speakers of Indian languages. UAL 31.345-346. 1968 The ordering of phonological rules. UAL 34.115-136. CHOMSKY, NOAM 1964 Current issues in linguistic theory. In Jerry'A. Fodor'.and Jerrold" J. Katz ,"eds. ...The^l&r^ttecextif; Langw'age. 50-118. Englewood C l i f f s , N.J.: Prentice-Hall 1965 Aspects of the Theory of Syntax. Cambridge, Mass.: M.I.T. Press -11966 The current scene in linguistics: present directions. College English 27.587-595. Reprinted 1967, Owen Thomas, ed. The Structure of Language. Indianapolis and New York: Bobbs-Merrill 1967 Some general properties of phonological rules. Language 43.102-108. CHOMSKY, NOAM and MORRIS HALLE 1968 The Sound Pattern of English. New York: Harper and Row COLE, PETER 19'76 The interface of theory and description: notes on Modern Hebrew relativization. Language 52.563-583. CRYSTAL, DAVID 1971 Linguistics. Harmondsworth, Middlesex: Penguin Books DARNELL, REGNA 1971 The Powell classification of American Indian languages. Papers in Linguistics 4.71-110. 1971 The revision of the Powell classification. Papers in Linguistics 4.233-258. DARNELL, REGNA and JOEL SHERZER 1971 Area linguistics in North America: a historical perspective. UAL 37.20-28. - 174 -DAUENHAUER, RICHARD 1976 Yuwaan Gageets: a Russian fairy tale in Tlingit oral tradition. Paper presented at the Northwest Coast Studies Conference. Burnaby, B.C.: Simon Fraser University. May 12-14. DAVIS, JOHN H. 1971 Notes on Mainland Comox phonology. In James E. Hoard and Thos. M. Hess, eds. Studies in Northwest Indian Languages. Sacramento: Sacramento Anthropological Society 11.23-25. DERWING, BRUCE L. 1973 Transformational Grammar as a Theory of Language Acquisition: A Study in the Empirical, Conceptual and Methodological Found-ations of Contemporary Linguistic Theory. Cambridge: University Press. DINGWALL, WILLIAM ORR (Ed.)', 1971 A Survey of Linguistic Science. College Park, Md.: University of Maryland Press DINNSEN, DANIEL A. 1974 Constraints on global rules in phonology. Language 50.29-51. DRUCKER, PHILIP 1955 Indians of the Northwest Coast. American Museum of Natural History. New York: McGraw-Hill. 1958 The Native Brotherhoods: Modern Intertribal Organizations on the Northwest Coast. Washington, D.C: Bureau of American Ethnology Bulletin 168. 1965 Cultures of the North Pacific Coast. Scranton, Pa.: Chandler Publishing Co. 1976 Tlingit and Haida, Inc.: a success story. Paper presented at the Northwest Coast Studies Conference. Burnaby, B.C.: Simon Fraser University, May 12-14. EASTMAN, CAROL M. 1976 Language re-introduction on the Northwest Coast as a factor of ethnicity and modernization: a suggestion for language policy. Paper presented at the Northwest Coast Studies Conference. Burnaby, B.C.: Simon Fraser University, May 12-14. - 175 -FLEMING, ROBERT 1971 Personal communication re: Tlingit speakers in the Teslin Lake, Y.T., area. Toronto. Dec. 9. FUDGE, E.C. 1967 The nature of phonological primes. Journal of Linguistics 3.1-36. 1969 Mutation rules and ordering in phonology. Journal of Linguistics 5.23-38. 1970 Phonology. In John Lyons, ed. New Horizons in Linguistics. Harmondsworth, Middlesex: Penguin Books GODDARD, PLINY EARLE 1920 Has Tlingit a genetic relation to Athapascan? UAL 1.266-279. GRASSERIE, RAOUL DE LA 1902 Cinq Langues de la Colombie Britannique: Haida, Tsimshian, Kwagiutl, Nootka, et Tlinkit. Paris: Bibliotheque Linguistique Americaine Tome 24. GREENBERG, JOSEPH H. 1953 Historical linguistics and unwritten languages. In A.L. Kroeber, ed. Anthropology Today. Chicago: University of Chicago Press. 265-286. 1957 Essays In Linguistics. Chicago: University of Chicago Press 1963 Some universals of grammar with particular reference to the order of meaningful elements. In J.H. Greenberg, ed. Universals of Language. Cambridge, Mass.: M.I.T. Press GREENE, JUDITH 1972 Psycholinguistics: Chomsky and Psychology. Part I. Harmondsworth, Middlesex: Penguin Books GREGG, ROBERT 1975 The phonology of Canadian English as spoken in the area of Vancouver, British Columbia. In J.K. Chambers, ed. Canadian English: Origins and Structures. Toronto: Methuen 1976 Personal communication - 176 -GRUBB, DAVID McC. 1974 The position of /a/ in the phonology of Kwakw'ala and i t s use in the orthography together with a linguistic key to the alphabet. Papers from the Ninth International Conference on Salishan Languages. Vancouver: University of British Columbia GUDSCHINSKY, SARAH C. 1967 How to Learn an Unwritten Language. New York: Holt, Rinehart and Winston HAAS, MARY R. 1969 The Prehistory of Language. The Hague: Mouton [Janua Linguarum, series minor, 57] HAIMAN, J. 1972 Phonological targets and unmarked structures. Language 48. 365-377. HALE, KENNETH 1973 Deep surf ace canonical disparities. In Thomas A. Sebeok, ed. Current Trends in Linguistics 11. HALLE, MORRIS 1958 On the bases of phonology. Reprinted in J. Fodor and J. Katz, eds. The Structure of Language. Englewood C l i f f s , N.J.: Prentice-Hall 1962 Phonology in a generative context. Word 23.54-72. 1974 Lecture. University of British Columbia. March.1974 HARMS, ROBERT T. 1966 The measurement of phonological economy. Language 42.602-611. 1968 Introduction to Phonological Theory. Englewood C l i f f s , N.iJ..: Prentice-Hall 1973 How abstract i s Nupe? Language 49.439-446. HARRINGTON, JOHN P. 1945 Boas on the science of language. 13AL 11.97-99. - 177 -HARRIS, ZELLIG S. 1951 Structural Linguistics. Chicago: University of Chicago Press HEIBERG, L.R. 1973 Notes on linguistic blinkers. Linguistics 97.72-94. HENDERSON, MICHAEL 1976 Redundancy, markedness, and constraints in phonology. Language 52.314-325. HILDEBRANDT, H.F. and G.L. STORY 1974 A historically oriented study of Babine (Northern Carrier) phonology, (unpublished manuscript) HOARD, JAMES E. and THOMAS M. HESS (Eds.) 1971 Studies in Northwest Indian Languages. Sacramento, Calif,: Sacramento Anthropological Society Paper 11. HOCKETT, CHARLES F. 1955 A manual of phonology. UAL, Memoir 11. Baltimore: Waverly Press 1958 A Course in Modern Linguistics. New York: Macmillan HODGE, FREDERICK WEBB 1913 Handbook of Indians of Canada. Appendix 10th Report of the Geo-graphic Board of Canada, Ottawa. Reprinted by C.H. Parmelee HOOPER, JOAN B. 1972 The syllable in phonological theory. Language 48.524-540. 1975 The archi-segment in natural generative phonology. Language 51.536-560. HOOPER, JOAN B. and TRACY TERRELL 1976 Stress assignment in Spanish: a natural generative analysis. Glossa 10.64-110. HOWREN, ROBERT 1971 A formalization of the Athabaskan D-effect. 13AL 37.96-113. - 178 -HYMAN, LARRY M. 1970 How concrete i s phonology? Language 46.58-76. 1973 The feature [Grave] in phonological theory. Journal of Phonetics 1.329-337. 1973 Nupe three years later. Language 49.447-452. 1975 Phonology: Theory and Analysis. New York: Holt, Rinehart and Winston HYMES, D.H. 1956 Na-Dene and positional analysis of categories. American Anthro-pologist 58.624-638. INGRAM, DAVID 1976 Phonological analysis of a child. Glossa 10.3-27. JACOBS, MELVILLE 1954 The areal spread of sound features in the languages north of California, !h Papers from the Symposium on American Indian Linguistics, University of California Publications in Linguistics. 10.46-56. JAKOBSON, ROMAN 1939 Les lois phoniques du langage enfantin. In N.S. Troubetskoy, Principes de Phonologie. TransJ J.^Gantineau. Paris':-Klincksieck 1968 Child Language, Aphasia and Phonological Universals. The Hague: Mouton [Janua Linguarum, Series Minor, 72] JAKOBSON, ROMAN, GUNNAR FANT and MORRIS HALLE 1952 Preliminaries to Speech Analysis: The Distinctive Features and their Correlates. Cambridge, Mass.: M.I.T. Press JAKOBSON, ROMAN, and MORRIS HALLE 1968 Fundamentals of Language. The Hague: Mouton JENSEN, JOHN T. 1974 How abstract i s abstract? Glossa 8.247-260. - 179 -JONES, DANIEL 19:64 An. Outline: of English...Phonetics.. 9th ed.. Cambridge: W. Heffer and Sons -. .. •" 1967. The Phoneme: • Its-'Nature and.-Use.-' 3rd ed. Cambridge: W. Heffer and Sons 1969 The Pronunciation of English. Cambridge: Cambridge University Press KELLY, L.G. 1971 De modis generandi: points of contact between Noam Chomsky and Thomas of Erfurt. Folia Linguistica- 5.225-252. KENSTOWICZ, MICHAEL and CHARLES KISSEBERTH 1970 Rule brdering and the asymmetry hypothesis. Papers from the Sixth Regional Meeting of the Chicago Linguistics Society 6.504-519. KINKADE, M. DALE 1963 Phonology and morphology of Upper Chehalis. 13AL 29.345-356. 1974 Personal communication 1976 Personal communication 1976 Areal features in the Northwest. Paper presented at the North-west Coast Studies Conference. Burnaby, B.C.: Simon Fraser University, May 12-14 1976 The Salishan languages. Paper presented at the Northwest Coast Studies Conference, Burnaby, B.C.: Simon Fraser University, May 12-14 KIPARSKY, PAUL 1968 How Abstract is Phonology? Bloomington::Indiana University Linguistics Club 1971 Historical linguistics. In William Orr.Dingwall, ed. A Survey of Linguistic Science. College Park, Md.: University of Maryland Press 1972 Explanation in phonology. In Stanley Peters, ed. Goals of Lin-guistic Theory. Englewood C l i f f s , N.J.: Prentice-Hall - 180 -KISSEBERTH, CHARLES W. 1970 On the abstractness of phonology: the evidence from Yawelmani. Papers in Linguistics 1.248-282. 1970 On the functional unity of phonological rules. Linguistic Inquiry 1.291-306. 1971 A global rule in Klamath phonology. In B. Kachru, ed. Papers in Linguistics in Honor of Henry and Renee Kahane. Edmonton: Linguistic Research 1972 Cyclical rules in Klamath phonology. Linguistic Inquiry 3.3-24. KLOKEID, TERRY J. 1975 The Nitinat feature system: a reference paper. Lektos [special issue]. Papers from the 10th International Conference on Salishan Languages, Robert St. Claimed. 1974. 81-93. KNUDSON, LYLE M. 1975 A natural phonology and morphophonemics of Chimalape Zoque. Papers in Linguistics 8.283-346. KOUTSOUDAS, ANDREAS 1973 Extrinsic order and the complex NP complaint. Linguistic Inquiry 4.69-81. KOUTSOUDAS, ANDREAS, GERALD SANDERS, and CRAIG NOLL 1974 The application of phonological rules. Language 50.1-28. KRAUSE, AUREL 1956 The Tlingit Indians: Results of a Trip to the Northwest Coast of American and the Bering S t r a i t s . Trans. Erna Gunther. Seattle: University of Washington Press [Die Tl i n k i t Indianer, Jena, 1885] KRAUSS, MICHAEL E. 1964 Proto-Athapaskan-Eyak and the problem of Na-Dene I: the phonology. UAL 30.'118-131. 1965 Eyak: a preliminary report. Canadian Journal of Linguistics 10.167-186. 1965 Proto-Athapaskan-Eyak and the problem of Na-Denelll: The morphology. UAL 31/18-28. - 181 -1968 Noun-classification systems in Athapaskan, Eyak, Tlingit and Haida verbs. IJAL 34.194-203. 1969 On the classification [classifiers] in Athapaskan, Eyak and T l i n g i t verbs. Indiana University Publications Anthropological and Linguistic Memoirs 24.49-83. 1973 Na-Dene. Current Trends in Linguistics 10.903-978. 1974 Map. Native Peoples and^Languages of Alaska. Alaska Native Language Center. Fairbanks: University of Alaska 1975 Lecture. University of Calgary. March 1975 KUHN, THOMAS S. 1970 The Structure of Scientific Revolutions. Chicago: University of Chicago Press KUIPERS, AERT H. 1967 The Squamish Language I. The Hague: Mouton 1974 The Shuswap Language: Grammar, Texts, Dictionary. The Hague: Mouton KURODA, S.SY. 1967 Yawelmani Phonology. Cambridge, Mass.Y M.I.T. Press LABOV, WILLIAM 1971 Methodology. In William Orr Dingwall, ed. A Survey of Ling-uistic Science. College Park, Md.: University of Maryland Press LADEFOGED, PETER 1971 Preliminaries to Linguistic Phonetics. Chicago: University of Chicago Press 1975 A Course in Phonetics. New York: Harcourt, Brace Jovanovich LAGUNA, FREDERICA de 1960 The Story of a Tlingit Community: A Problem in the Relation-ship between Archaeological, Ethnological and Historical Methods. Bureau of American Ethnology, Bulletin 172. Washington, D.C. - 182 -LANDAR, HERBERT 1967 Two Athapaskan verbs of "being". In John W.M. Verhaar, ed. The Verb 'Be' and Its Synonyms. Foundations of Language. Supplementary Series. 1.40-74. LANGACKER, RONALD W. 1967 Language and Its Structure: Some Fundamental Concepts. New York: Harcourt, Brace & World LEHMANN, W.P. 1972 The comparative method as applied to the syntactic component of language. Canadian Journal of Linguistics 17.167-174. 1973 A structural principle of language and i t s implications. Language 49.47-66. LEVINE, ARVIN 1976 Why argue about rule ordering? Linguistic Analysis 2.115-124. LEVINE, ROBERT 1973 Notes on a Haida Text: I. The Charlottes: A Journal of the Queen Charlotte Islands. Queen Charlotte Islands Museum Society 2.28-32. 1975 Seminar on Haida phonology. University of British Columbia. March 1975. LI, FANG-KEUI 1946 Chipewyan. In Harry Hoijer, ed.LLinguistic Structures of Native North America. Viking Fund Publications in Anthropology 6.398-423. LIGHTNER, THEODORE M. 1965 On the description of vowel and consonant harmony. Word 21.244-250. 1968 On the use of minor rules in Russian phonology. Journal of Linguistics 1.69-72. 1971 Generative phonology. In William Orr Dingwall, ed. A Survey offLinguistic Science. College Park, Md.: University of Mary-land Press 1976 On deglottalizationiin Klamath. UAL 42.14-16. - 183 -LYONS, JOHN 1962 Phonemic and non-phonemic phonology: some typological reflections. IJAL 28.127-133. 1 1968 Introduction to Theoretical Linguistics. London: Cambridge University Press 1970 Chomsky. London: Fontana/Collins 1970 Ed. New Horizons in Linguistics. Harmondsworth, Middlesex: Penguin Books McCAWLEY, JAMES D. 1967 Sapir's phonologic representation. IJAL 33.106-111. MILLER, PATRICIA D. 1972 Vowel neutralization and vowel reduction. Papers from the Eighth Regional Meeting of the Chicago Linguistics Society 8.482-489. NAISH, CONSTANCE 1966 A Syntactic Study of Tlingit, Master's Thesis, University of London NAISH, CONSTANCE and GILLIAN STORY 1963 English-Tlingit Dictionary: Nouns. Fairbanks: University of Alaska, Summer Institute of Linguistics NORMAN, LINDA 1972 The insufficiency of local ordering. Papers from the Eighth Regional Meeting of the Chicago Linguistics Society 8.490-504. OBERG, KALERVO 1973 The Social Economy of the Tlingit Indians. Vancouver: J.J. Douglas Ltd. PALMER, FRANK 1971 Grammar. Harmondsworth, Middlesex: Penguin Books PEI, MARIO 1966 Glossary of Linguistic Terminology. New York: Doubleday, Anchor Books / - 184 -PERCIVAL, W. KEITH 1976 The applicability of Kuhn's paradigms to the history of l i n -guistics. Language 52.285-294. PINNOW, HEINZ-JURGEN 1959 Zwei Probleme der historischen Lautlehre der Na-Dene-Sprachen. Zeitschrift fur Phonetik und allgemeine Sprachwissenschraft 11. 128-159. Abstracted 1962 in IJAL 28.61-66. 1964 On the historical position of T l i n g i t . IJAL 30.155-164. 1966 Grundz'&ge einer historischen Lautlehre des Tlingit: ein Ver-such. Weisbaden: Otto Harrassowitz 1968 Genetic relationship vs. borrowing in Na-Dene. IJAL 34.204-211. 1970 Notes on the classi f i e r s in the Na-Dene languages.. 'IJAL 36. 63-67. 1976 Geschichte der Na-Dene Forschung. Contributions to Ethnology and Linguistics, Archaeology and Physical Anthropology of Indian America, Supplement 5. Berlin: Indiana, Iberian-American Institute POPPER, KARL R. 1972 Objective Knowledge: An Evolutionary Approach. London: Oxford University Press POSTAL, PAUL M. 1968 Aspects of Phonological Theory. New York: Harper and Row POWELL, J.R. 1891 Indian Linguistic Families of America.'North of Mexico. Preston Holder, ed. Lincoln: University of Nebraska Press. Reprint 1966 RIGSBY, BRUCE and MICHAEL SILVERSTEIN 1969 Nez Perce vowels and Proto-Sahaptian vowel harmony. Language 45.45-59. ROBERTS, E. WYN 1976 Phonological theory, absolute neutralization and the case of Nupe. Glossa 10.241-287. - 185 -RUDES, BLAIR 1976 Lexical representation and variable rules in natural generative phonology. Glossa 10.111-150. SAINT-JACQUES, BERNARD 1967 The quest for a sc i e n t i f i c study of language. Word 23.476-493. SAPIR, EDWARD 1915 The Na-Dene languages, a preliminary report. American Anthro-pologist 17.534-558. 1921 Language: An Introduction to the Study of Speech. New York: Harcourt, Brace and World 1923 The phonetics of Haida. IJAL 2.143-158. 1949 The psychological reality of phonemes. In D.G. Mandelbaum, ed. Selected Writings of Edward Sapir. Berkeley: University of C a l i -fornia Press SAUSSURE, FERDINAND de 1959 Course in General Linguistics. Charles Bally and Albert Seche-haye, eds. Trans. Wade Baskin. New York: McGraw-Hill SCHACHTER, PAUL 1970 Natural assimilation rules in Akan. IJAL 36.342-355. SCHANE, SANFORD A. 1968 On the non-uniqueness of phonological representations. Lan-guage 44.709-716. 1971 The phoneme revisited. Language 47.503-521. 1972 Natural rules in phonology. In Robert P. Stockwell and Ronald K.S. Macaulay, eds. Linguistic Change and Generative Theory. Bloomington: Indiana University Press 1973 Generative Phonology. Englewood C l i f f s , N.J.: Prentice-Hall SHIBATANI, MASAYOSHI 1973 The role of surface phonetic constraints on generative phono-logy. Language 49.87-106. - 186 -SHOURUP, LAWRENCE C. 1974 Phonological conspiracies. Glossa 8.198-207. SKOUSEN, ROYAL 1972 On capturing regularities. Papers from the Eighth Regional Meeting of the Chicago Linguistics Society 8.567-577. 1973 On limiting the number of phonological descriptions. Glossa 7.167-178. STAMPE, DAVID 1969 The acquisition of phonetic representation. Papers from the Fifth Regional Meeting of the Chicago Linguistics Society 5.443-454. 1972 On the natural history of diphthongs. Papers from the Eighth Regional Meeting of the Chicago Linguistics Society 8.578-590. STANKIEWICZ, EDWARD 1976 Baudouin de Courtenay and the Foundations of Structural Linguistics. Lisse: The Peter de Ridder Press STANLEY, RICHARD 1967 Redundancy rules in phonology. Language 43.393-436. STORY, GILLIAN 1966 A Morphological Study of Tlingit, Master's Thesis, Univer-sity of London STORY, GILLIAN and CONSTANCES. NAISH 1973 Tlingit Verb Dictionary. Fairbanks: University of Alaska, Alaska Native Language Center SWADESH, MORRIS 1952 Review of Robert Shafer, 'Athapaskan and Sino-Tibetan.' IJAL 18.178-181. 1960 On interhemispheric linguistic connections. In Stanley Diamond, ed. Culture and History, Essays in Honor of Paul Radin, New York: Columbia University Press - 187 -SWANTON, JOHN R. 1908 Social condition, beliefs and linguistic relationship of the Tlingit Indians. In Annual Report. Bureau of American Ethnology. Washington, D.C. 26.391-485. 1909 Tlingit Myths and Texts. Bureau of American Ethnology, Bulletin 39, Washington, D.C. 1911 T l i n g i t , •. Handbook of American Indian Languages. Bureau of American Ethnology, Bulletin 40, Part 1.159-204. 1952 Indian Tribes of Alaska and Canada. Seattle: Shorey Book Store [1969] TAI, JAMES H.Y. 1973 Chinese as a SOV language. Papers from the Ninth Regional Meeting of the Chicago Linguistic Society 9.659-671. TEETER, KARL V. 1964 Descriptive linguistics in America: t r i v i a l i t y vs. irrelevance. Word 20.197-206. TROUBETSKOY, N.S. 1957 Principes de Phono'logie. Trans. J. Cantineau. Paris: Klincksieck 1969 Principles of Phonology. Trans. Christiane A.M. Baltaxe. Berkeley and Los Angeles: University of California Press U.S. FEDERAL FIELD COMMITTEE FOR DEVELOPMENT PLANNING 1968 Alaska Natives and the Land. Anchorage, Alaska and Washington, D.C.: Government Printing Office VAGO, ROBERT 1973 Abstract vowel systems and Uralic vowel harmony. Language 49. 579-605. VELTEN, H.V. 1939 Two southern Tlingit tales. IJAL 10.65-74. 1944 Three Tlingit stories. IJAL 10.168-180. VENNEMANN, THEO 1972 Phonological uniqueness in natural generative grammar. Glossa 6.105-115. - 188 -VOEGELIN, C.F. and F.M. 1966 Map of North American Indian Languages. American Ethnological Society. Rand McNally & Co. WELSCH, ROBERT L. 1976 Features and processes in Hydaburg Haida. Working Papers for the Eleventh International Conference on Salishan Languages. Seattle: University of Washington. 11.283-312. WHEELER, MAX 1972 Distinctive features and natural classes in phonological theory. Journal of Linguistics 8.87-102. WHITE, ROBIN B.W. 1972 Klamath Phonology. Ph.D. dissertation, University of Washington ZIMMER, KARL E. 1967 A note on vowel harmony. UAL 33.166-171. 1969 On the evaluation of alternative phonological descriptions. Journal of Linguistics 6.89-98. ZWICKY, 1976 ARNOLD M. Bibliography on cyclical segmental rules. UAL 42.267-268. - 189 -APPENDIX Data and Charts - 190 -CHART I Tlingit Vowel Chart Tli n g i t Vowels - back + back + tense - tense - tense + tense + high i I U u - high e e A a CHART II Distinctive Features Chart Feature Concrete Representation Underlying Abstract i I u U e a A o o s y l l + + . + + + + + + + + high + + + + - - - - -back - - + + - - + + + + tense + - + - + - + - + round - - + + - - - - + + v. Chomsky and Halle 1968: 176. - 191 -CHART III Tlingit Vowel Markedness (after Chomsky and Halle 1968) Feature Value i I u U e e a A o 0 low u u u u u u u U u m high u u u u m m u m m u back - - + + - - u ik + u tense u m u m u m u m u m round u u u u u u u m u m complexity 1 2 1 2 2 3 0 4' 2 3 - 192 -DATA APPENDIX burtac^ Underlying Manifestation Gloss Form I. -yin Verbal Forms 1. ?uxun "having blown" ?ux w-yin 2. V \ / suwgun "having laughed" suwqw-yin 3. ?adin "having gone" at-yin 4. wu±?ugun " i t has boiled" wul?uk w-yin 5. xweAin "he was tired" "having been tired" xwe*-yin 6. yAnAigenin "he was getting big" yAnAlgen-yin 7. xAt l A c i n i n "I used to be strong" XAt l A c i n - y i n 8. TVEI wuduskuwun "he was never known" kil wudusku-yin 9. qUxwAuwun "I was dwelling" qUxwAU-yin 10. qUx wAkitin "I used to pick berries" « > / » qUx wAkit-yin II. - y i Pronominal Possession 1. duxadi "his root" du-xat-yi 2. duyegi "his s p i r i t " du-yek-yi 3. \ ' duyagu "his canoe" du-yak w-yi 4. duwagi "his eye" du-waq-yi 5. \ \ , dukeXi "his dog" du-keft-yi 6. dutayi "his king salmon" du-ta-yi 7. \ . r \ dutayi "his board" du-ta-yi 8. \ r \ du?ayi "his lake" f du-a-yi 9. dunuwu "his f o r t " du-nu-yi 10. v x , duluwu "his nose" ( i . e . , of someone else) du-±u-yi - 193 -11. dulugu "his king salmon" 5 du-luk w-yi 12. d u t i l i "his scar" d u - t i l - y i 13. dutxli "his shoe" d u - t i l - y i 14. dusaxawu "his hair" du-saxaw-yi 15. duguxu "his slave" du-gux w-yi 16. duyagi "his mussel" du-yak-yi 17. du?Axayi "his paddle" du-Axa-yi III. Nominal Possession 1. hit tagu "house timber" hit-taq w-yi 2. xuc haXi "grizzly bear's dung" xuc haA-yi 3. xat hini "salmon creek" xat-hin-yi IVa. Nouns ending in /a/ - with diminutive ending 1. ?ak w "a l i t t l e pond" a-k 2. duaku "his l i t t l e pond" > du-a-k-yi 3. ?akw "a small thing" a-k 4. du?aku "his l i t t l e thing" > du-a-k-yi 5. tak w "a small spring salmon" ta-k IVb. Exceptions - labialization without diminutive 6. qawu " i t s man" qa-yi 7. nawun "having died" na-yin 


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items