言語研究の多様性について

全文

(1)Title. 言語研究の多様性について. Author(s). 菅原, 光徳. Citation. 北海道教育大学紀要. 第一部. A, 人文科学編, 17(2): 50-68. Issue Date. 1966-12. URL. http://s-ir.sap.hokkyodai.ac.jp/dspace/handle/123456789/3923. Rights. Hokkaido University of Education.

(2) Vol. 17 No. 2 Journal o£ Hokkaido University of Education (Section I A) June, 1966. On Pluralism in Linguistics. Mitsuho SuaAWARA The Study of English, Asahigawa Branch, Hokkaido University of Education. WM: ismww'^^^ 1. INTRODUCTION Recently there has been a strong tendency to understand or criticize one linguistic approach in terms of another. Some criticisms are offered referring to the theory of. linguistic research, and some referring to specific descriptive devices for individual facts and phenomena of a language. This strong tendency seems to imply one and only one universally accepted linguistic approach. In that approach there exists only one. purpose of linguistic research which can not be violated by any linguist with his own unique theory of language. The purpose is then considered to prescribe the theory, the nature of corpus, and many specific descriptive devices. If we recognize that language is a manifestation of such complicated human behavior, it seems tenable for us to insist that there should be more than one approach to the study of language. In fact, there have been manifold approaches since the beginning of this century. It seems to me that the occurrence of such a tendency toward a. single, well-defined, and generally acceptable linguistic approach is largely due to the disagreement in various aspects of modern linguistics today. The disagreement is in turn raised by a failure to understand different theories in various approaches. By " failure," I mean a lack of an appropriate scale to evaluate a possibility (or possibilities) of the revision and development of a theory in each approach. So far, the most reasonable criticism on various systems would be the one that is offered by transformational grammarians. It is true, however, that there exists another group of grammarians who believe that different grammars are written for different purposes and in different situations. But most of those grammarians do not intend to go so far as to propound the distinction of those different purposes and/or situations in terms of predictive or descriptive nature of linguistics like the distinction made, in. general, between physics (predictive) and classificatory botany (descriptive).!) It should be noted here that if different grammars have different purposes there should have been some rigorous attempts to refine the theory of each approach instead of abandon-. — 50 —.

(3) Mitsuho Sugawara. ing it and setting up another theory. To be more specific, we should reconsider whether or not the idea of parts of speech in pre-stmctural grammars, for example, is really untenable, or whether the notion developed by 1C analysis and the procedure of phonemicization in structural linguistics are quite inadequate or have a possibility to be developed to an adequate model within the whole framework of their original purposes. Followiug Postal,2) we have to come to the conclusion: " proposed systems of descriptive grammar" such as Bloch's Japanese Syntax, Well's 1C Approach, Harris' Morpheme Class Substitution System, Hockett's IA Model, Lamb's Stratification Syntax, Pike's Tagmemics, etc. are all " versions " of the theory of PSG (Phrase Structure Grammar) " which has strong limitations" within genera tive grammars.3) It would thus seem hopeless to maintain our hypothesis that it is possible to develop and revise each model within its own framework, or the hypothesis that each grammar has its own purpose and situation. If Postal is correct, then all the systems of descriptive grammar enumerated above and many others are to be involved in a singlele level of transformational grammars. Is there not even a single reason for the independence of. each model ? If there is, what is it ? It seems possible to find the reason for independent theories among disputable problems of phonemicization, of parts of speech, of 1C analysis, of linguistic elements and of meaning. It would be also relevant to discuss the reason from a different dimension :. What is a linguist's task ? Are not manifold independent theories in linguistic description allowed to exist ? Or should the interpretation of linguistic theory be unanimous among. linguists? If there is any single unsettled problem among them, and if there is any possibility of finding a solntion within the framework of each different apprroach, we may be able to plead for pluralism in linguistics. It would not be very wise for us to abandon a well-defined approach for a particular aspect, and then shift totally to another one which has not been thoroughly completed yet. We may have another reason to plead for pluralism if we are given some principles and/or techniques available to foreign language teaching by presuming the possibilities to develop and 1-efine each linguistic theory and practice. The main purpose of this paper is thus to assert plurality of linguistic approaches.. In particular the paper is designed to maintain the assertion made by James Sledd in 1961.4) It would be interesting for us, language teachers, to examine his claim and to plead for pluralism again by emphasizing a part of his discussion such as parts of speech and by clarifying some new problems in modern linguistics. The first item to be discussed will be on the problems of phonemic procedures,. 2. 1 PHONEMIC PROCEDURES The recognition of phonemic units as important features of language has been largely supported by the taxonomic, the tagmemic and the transformational approaches. However, a very fundamental problem of analytical procedures and some particular problems — 51.

(4) On Pluralism in Linguistics concerning the phonemicization of certain segments are still remained unsolved.5) Espedaily there exists severe controversy on the phonemic procedures of the taxonomic. approach. In this approach it is claimed that every linguistic research should be based on a description of phonological distinctive features. These distinctive features are determined by rigorous analytical procedures 6) and once determined as phonemes, they are always phonemes. These claims are the subjects of severe criticism offered by transformational grammarians. It seems to me, however, that the transformationalists do not reject the very existence of phonemic units nor all the principles of phonemic analysis. Noam Chomsky, for example, believes that the rules of phonemic analysis are reasonable, though he says,7) " there are cases where the principles of phonemic analysis force decisions which lead to unmotivated complication in the grammar." In other cases, he seems to accept phonemic analysis if it leads to the best decision for constructing a grammar.. 2. 1. 1 THE LINEARITY CONDITION Chomsky's partial rejection of phonemic analysis can be discussed most effectively in his arguments on the linearity and the biuniqueness conditions. According to Chomsky,8) the lineanty condition is met in the context that each phoneme should occur in the phonemic representation associating with a particular succession of one or more continuous phones in the phonetic representation. Therefore, if phoheme P precedes phoneme Q in the phonemic representation then, in the phonetic representation, a single phone or the sequence of phones which is manifested by phoneme P must precedes the phone or the phone sequence manifested by phoneme Q. This lineanty condition in structural linguistics is not met in an example of the nasalized vowel as discussed by Chomsky.9) We often find in some dialects the nasalized vowel occurring before un voiced stops in the phonetic representation. Thus ohonemic /k sent/ is phonetic. [kgeO. But notice that phonemic /hsend/ is phonetic Chtsnd]. Given the evidence as such, we cannot say that vowel nasalization is phonemically significant in English. In. the face of this fact, we should disregard the linearity condition. The linearity condition, however, does not seem to be such a strong requirement in phonemic analysis as Chomsky might suppose. In many cases the condition is violated even by Hockett 10) under the pressure of, say, pattern congruity and/or economy. For example, in his argument of /of-/]/ vs. /tg/-/dz/,10 Hockett chooses the former solution. on the ground that if we interpret ^tg] and rdz] as phonemic clusters, /t§/ and /dz/ respectively, then these are the only two word initial consonant clusters in English in which the stop precedes the spirant except /ts/ as in " tsetse (fly) ". The use of structural parallelism, as given in the example above, appears relevant to the argument of the lineai-ity condition. There would be no serious problem in disregarding the. condition when CS83 is assigned phonemically to /sen/ before the unvoiced stop or to /se/ in the same condition, if we have the strong evidence of structural parallelism in. — 52 —.

(5) Mitsuho Sugawara. English.. 2. 1. 2 THE BIUNIQUENESS CONDITION One of the fundamental assumptions in phonemic analysis is the biuniqueness condition with regard to the phonemic level of representation. The condition requires the unique interpretation from both phonemic representation to phonetic representation and vice versa. Some cases of phonemic overlapping violate the principle of biuniqueness. A set of words like 'Betty (person's name)', ' berry ' and 'throw' is a good example given by Chomsky.l2). Betty CbeDiy:! /-t-/ (1) berry [beny] /-r-/ (2). throw [;i92ow] (3) [D~] in (3) being in complementary distribution with both (D~] in (1) and f^v') in (2) can be assigned to /t/ as in (1) or /r/ in (2). This is a case of phonemic overlapping. It would thus seem unreasonable to say that the phoneme /t/ has the allophone CD] in the intervocallic post-stress position, and /r/ has CD] as its allophone after dental spirants, though the interpretation is the only reasonable solution to preserve the principle of biuniqueness. Thus given a phone CD] as in ' throw ' in a phonetic context,. we cannot uniquely assign it to the phoneme /t/ nor to /r/, and it is also impossible for us to determine uniquely the phonetic realization of ' throw', even when the phoneme /t/ as in ' Betty ' and /r/ as in ' berry ' are given. We cannot simultaneously have. both the principles of biuniqueness and partial overlapping in the face of such evidence stated above. If we follow Chomsky again, we may find a redundant representation of phonemic level produced by the principle of biuniqueness. When we preserve the assumption that "a string of formatives specifies a string of phones uniquely, but not conversely", is) we need only two representations, the systematic phonemic l4) and the phonetic. Many representations between these two systematic ones can be obtained as a set of derivation from the systematic phonemic representation. These intermediate representations are no longer systematic. (Examples for this argument are given by Chomsky in his. Current Issues in Linguistic Theory, (The Hague, 1964), pp. 71-75.) The biuniqueness condition not only introduces unmotivated complications for the construction of a grammar, but it violates the assumption that the phonological component of a grammar determines the phonetic form of a sentence ^generated] by the syntactic rules."l5). Compare the following two different models of the description of German /bunt/ 'group,' /bunde/ ' (in) the group', /bunt/ ' colored', and /bunte/ ' colored ones': I6). — 53 —.

(6) On Pluralism in Linguistics. Model A. (D (2). (3). Model B. bunD. / /. bunD ./' \. /\. [bunt?]. ^ . .V /bunt - / /bund - /. \. Cbund -. A. 3. , Cbund-3 .tS] CbuntS]. From a point of view of transformational grammars, as we have discussed above, unnecessary complications are introduced in the biuniqueness approach (Model B) by setting a redundant phonemic level (2) between the systematic phonemic level (morpho-. phonemic level) (1) and the systematic phonetic level (3). It is quite important in this model to show that [t] and [d] belong to different phonemes in German since, in. other cases such as in Cbunte), [t] occurs without following S, though [t] and [d] are in complementary distribution in the given examples; [bund-] and CbuntgQ. In Model B, the phonemicization of Ct3 and Cd] as a single phoneme will bring a complete overlapping: Ct] is sometimes assigned to /d/ and in other cases to /t/ without any apparent condition. In order to avoid such a complete overlapping, this model requires a level of representation between the morphophonemic and phonetic levels, Thus in this model a morphophonemic representation of bunD- cannot have two different phonetic forms, CbuntS] and Cbund-], as immediate underlying forms of it. Nor, in other. words, can Cbuntj?] and [bund-3 be built up directly into bunD- skipping over the intermediate level of the phonemic representation. While in Model A, it is permissible. to set up rules like the following: D -> t in the context: ............ f, D —> d in the context: •••••••••••• Vowel. since this particular morpheme jbund} 'group, union' has been derived from uniqtie PS (Phrase Structure) rules and T (Transformational) rules in transformational grammars. But the description has no relation to {bunt} 'colored'. Furthermore, the description of Model A has nothing to do with the unique interpretation of morphophonemic bunD- from the phonetic data. After all, if it is correct to assume that Chomsky's partial rejection of phonemic analysis is based on inadequacy of the biuniqueness condition on the grounds that (1) it leads to unmotivated complications, (21 it violates a very basic assumption of grammatical description and (3) it represents an inconsistent system between the morphophonemic and phonetic representation, then it would be reasonable to raise the followinng questions. First, is it not impossible to develop further the notion of overlapping within the framework of taxonomic phonology so that one can have a consistent and generally acceptable phological system, without given a stricture " once a phoneme, always a pnoneme " ? Secondly, is it unreasonable for a linguist not to rely totally upon the assumption until it becomes universally acceptable one with rigorous proof ? If it is not, then there would be some facts with which we can suspect general theory of language not to be unanimous at present.TO In any event, " we are [not] willling to tolerate. — 54 —.

(7) Mitsuho Sugawara. such absurdities as the phonemic representations of /ks»t/, /ra'yD-i-r/, /rayD-i-r/ for ' can't, 'rider', 'writer' and so on, in many cases."!8) We are rather just seeking for a better solution of systematic representation of phonemes without disturbing the original principles of phonemic analysis.. 2. 1. 3 INTONATION There seems to exist another controversial problem in the argument of phonolgy. The problem refers to the relation between syntactic structure and intonation.19) There are some linguists20) who strongly suspect the claim of one-one correspondence between syntactic structure and phonological strncture. Longacre, for instance, criticizes the treatment of phonological construction in generative grammars, saying that " attempts to combine grammar and phonology in one complex set of rules must inevitably result in continued neglect of such units as syllable and stress group."2l) Given a phrase such as "the very fresh milk " or " the shooting of hunters," no native speaker would put a pause between "the" and " veiy " like " the/very fresh milk " or between " the " and " sooting " like " the/shooting of hunters," though the place of the pause indicates the syntactic structure of these phrases. Furthermore, we are given a very interesting pioneer work on intonation of a generative grammar by Stockwell.22). (1) S -^ Nuc + IP (Nuc-^Np + Vp) ( ••• ). (2) IP -^ C + JP In fact, there would be many cases in which two or more different syntactic strings. take a single particular string of IP (Intonation Pattern).. Who? I came before you arrived.. the boy studying in the library They all take the same string of IP, say, (3) C -> Disc (4) Disc -^ 001 j. In the face of the evidence, we have to neglect the description of the relationship between various types of syntactic strings and a single intonation pattern. With the same example, we can raise another question : is it not possible to say that fcwo different sh-ings of IP can be assigned to a single particular matrix of morphemes with the same meaning?. (5) Cont-^i f002 4-. {003 4.. (004 4... When we are to select rule (41 and 002 ^ in rule (5) here, we have a sentence: 2I 3came2/2before you ar3rivedlj, But this is not the only choice given to us. According to Longacre,23) the same sen— 55.

(8) On Pluralism in Linguistics. fence has a continuous intonation (C —>• Cont) on the ground that the sentence is the same type of " I came yesterday." Then: 2I came before you ar3rived1^ In both cases of overlapping characteristics, there seems to be no well-defined acceptable solution. Partial rejection or partial acceptance has been attempted so far, as we have seen above, but we have now no new system which can make any claim to general approval.. 2. 2 HIERARCHY24) Problems of hierarchy in language structure offer another controversy in linguistics. The early works of American structural linguisfcics recognized a single hierarchy from the phonemic level up to the syntactic level. The Trager -Smith system of phonology, for example, was based on this assumption. It seems to me that it is since Hockett's. " Linguistic Elements and Their Relations" in Language XXXVII (1961) that the tripartite system of hierarchy has been remarkably developed. This tripartite system is now the nuclear part of tagmemics.25) As Longacre suggests,26) " morphemes do not necessarily coincide with syllables, nor phonological words (stress group or juncture group27)) with grammatical words. Nor must phonological phrases coincide with grammatical phrases, nor phonological sentences coincide with grammatical sentences." For example, the borders between some units of the phonological hierarchy sometimes do not coincide with the borders of the grammatical hierarchy. If we follow W. Cook28), howBver, we can find some inconsistency in the tripartite system of tagmemics. Pike first considered the hypertagmeme as a tagmeme on a higher level, but it was found not to be a tagmeme sequence but to be a unit acting as the filler class in a higher-level tagmeme. He also uses the terms hyperphoneme and hypermorpheme. (See Footnote 25. ) But the hyperphoneme is a phoneme sequence and is no sense a phoneme at all. The hypermorpheme, likewise, is a morpheme sequence and is no longer a morpheme. It would thus seem quite in order to say that the use of the term hyper- unit is rather inconsistent. Although there appear to be some serious problems of phonemicization in tagmemics and taxonomy, however, the patterning of phonological construction can be systematical-. ly described within the independent level of phonology. While in transformational grammars, the negligence of this tripartite system (phonological, lexical and gram-. matical) would seem to introduce a little difficulty on its morphophonemic level. For instance, in transformational grammars, some exemplified verbs are listed as the terminal symbols of grammatical (syntactical) rules. Those verbs are classified in accordance with their functional behavior. Now remember that the phonological component does function to show the phonetic forms of these terminal symbols given by the syntactic rules. It is very dubious, however, that the phonological component of a grammar can. — 56 —.

(9) Mitsuho Sugawara. well describe all the patternings of, say, the conjugation of irregular verbs in English. Some verbs such as transitive ones are to be listed under different terminal strings (of syntactic rules) from those of other verbs like intransitive ones. It is because of their diffeerent behavior of grammatical functioning in an utterance: transitive vs. intransitive verbs. In this grammar, there seems to exist no claim to describe patternings of phonological behavior which may be in common between some transitive verbs and some intransitive ones. If one is to describe the conjugation of verbs independently, for example, then he has to set up a type of an independent phonological level.. 2. 3 PARTS OF SPEECH An argument of parts of speech offers another problem in terms of linguistic hierarchy. Since, in transformational grammars, the classification of form-classes is not relevant, and since they are not interested in the list of all the members in some form-classes, it is only a partial listing of members in form-classes that one is allowed to describe in the grammars. According to Longacre, however, "a. partial listing of this sort will not allow them to generate all possible sentences. They need an exhaustive cross reference dictionary. "29) He also points out that the negligence of the lexical level. will thus lead any grammatical description to oversimplification of the lexis.30) In fact, LeesSl) suggests forty-seven different classes of nouns. It seems to me that this list simply shows a transformational version of parts of speech classification. James Sledd also refers to Lees' grammar and states that Lees " defines the noun Conly3 by. listing all the words which can appear in the positions where the symbol N appears in the rules."32) Thus, transformational grammars can not avoid the syntactical-lexical relation of levels, since a transformafcional grammar will include a lexicon as in Less' grammar, and since it will embed in itself a much more complicated sub-classification of parts of speech than any other previous grammar. In tagmemics, however, the problem of parts of speech classification has not yet been solved. What is at least relevant to say is that the parts of speech classification in tagmemics is not based on the same criteria as in pre-structural grammars. I do not know whether or not a tagmemist has given a clear definition to, say, (n ' manifesting in some cases both a subject and an object tagmemes. But there seems to have been no attempt to explain the correlation between a symbol (or a string) on the word (or phrase) level of ' Ag ' and ( np ' in the following example:. pCl= + S : mNc/np + pP: Vp ± L : n ± Ag : Ag (1) The flag is waved by the man, (2) Jim was hit. (3) Jim was taken home by Bill. It seems thus clear that a definition of parts of speech is not given in tagmemics and it is irrelevant in transformational grammars. From a point of view of foreign — 57.

(10) On Pluralism in Linguistics. language teaching, we need a concise description of those formal features of the position in which ' n' of ' np' in a subject tagmeme and ' n' of a locative tagmeme appear (see the example above), or N in transformational grammars appear.. 2. 4. 1 IMMEDIATE CONSTITUENT ANALYSIS As far as 1C analysis is concerned, it would be inadequate to assume that there may be some possibilities for success in the development of its theory as well as its practice. in grammar description within the range of the original idea of 1C analysis. According to Chomsky,33) 1C analysis " may be adequate as an account of surface structure," but " it is certainly not adequate as an account of deep structure." This is true since there has not been any discussion on this matter in taxonomy.34) It is no longer. dubious that 1C anaysis would be fully developed in a type of level like Chomsky's PS in his grammar. There are, however, other attempts to reform the theory of 1C analysis. Most of these attempts start with the objection against the theory of 1C aalysis. Tagmemists, for example, criticize the system of binary cutting as being arbitrary.SS) The construc-. tions isolated by such arbitrary dichotomization are thus believed to be falsely described constructions. Furthermore, the order of cutting is considered not to be consistent, while a word or clause is cut in another constituent : levels of constituents are different. In tagmemics, the cutting is not necessarily dichotomous: depending upon the construction, it may be a three-unit cut or a four-unit cut. These cuts are based on the assumption of the function-form composite of a construction, not on the arbitrary assumption of " universal binary cutting." The order of cutting is also assumed to be consistent since in tagmemics there are five different levels in its grammatical hierarchy. In each level of the hierarchy, parallel constituents are to be systematically described. Tagmemists' criticism would be correct as far as the deficiencies of 1C analysis are concerned. Besides their criticism of 1C analysis, tagmemic analysts have claimed a generating power in their model,36) and have implicitly expressed transformations in their analysis. In tagmemics, statements and questions, active and passive constructions are all described.37) However, there has been no attempt in tagmemics to find a convenient way of stating the relationship between or among these constructions.S8) Tagmemists should clarify the fact of relationships between constructions. It would be possible to reach that point in the tagmemic model without stating rules of transformations as in generative grammars. It would be interesting to notice here that in general the syntactic comparison between two languages has not been very successful up to this point. As Chomsky suggests,39) linguistics so far (before 1958) has shown so " many superficial truths, but few deeper ones." In taxonomy or tagmemics, a comparison between two languages has been assumed to be possible if linguistic patterns are carefully and elaborately described for both languages.40) it should be taken for granted to assume that it is. — 58 —.

(11) Mitsuho Sugawara. the careful and elaborate description of patterns that enables us to compare two different languages. But, however elaborately or consistently the variety of facts may be described, the resultant patternings would not necessarily lead to successful comparison if there is no attempt to state the relations between or among constructions in one language in terms of deep structure. A practical study based on this asumption has not been attempted yet, but it seems to me that the theory of deep structure implies a strong possibility for a successful contrastive study between languages.. 2. 4. 2 LINGUISTIC UNIT In the discussion above, we should keep in mind the tagmemists' objection against transformational grammars in terms of identifying units. According to Pike,41) an. attempt to identify grammatical units, for example, will lead to great difficulty if one tries to describe grammatical structure in terms of morpheme classes or morpheme sequences alone as in the attempts of 1C analysis and PSG. It is simply because a grammatical unit is not a single morpheme or matrix of morphemes, but is a functionform composite.42) Without the recognition of a unit as a function-form composite, it is very hard to get the precise behavior of a linguistic unit on which the process of rules can be most effectively formulated. It is no longer dubious that " the shooting of hunters," for example, is an ambiguous unit, but how can one systematically arrive at this unit? Since it is not a focal concern of transformational grammars that the rules at various stages should reflect intermediate linguistic patterns, there seems to be no other ways than the arbitrary presentation of such a unit in those grammars. In general, if we are given a systematic unit (distinctive units) and a device to describe the correlation between or among units, we may be able to have a more successful contrastive study of different languages. It would not be sufficient to have either of them as the basic theoy of contrastive study. The theory of deep structure will be more and more acceptable, even in a contrastive study, if it is provided with a well-described device to arrive at all the linguistic units in a language.. 2. 5 MEANING Some problems of meaning also provoke controversy in linguistics.43) A focal concern here is whether or not a syntactic description is possible without referring to " meaning". Bloomfield, Nida and tagmemists (Pike, Longacre, Elson and Pickett, etc.) believe that the systematic description of language is impossible apart from meaning.44) Particularly, tagmemists are in the extreme position that indicates the importance of form-meaning combination in a grammatical description.45) In tagmemics, the formmeaning combination is regarded to be a uint called " form-meaning composite "46) which is manifested by the term " tagmeme". Theoretically, PikeW assumes that the fact of form-meaning separation would lead to a dualism of form versus meaning, which is correlated with a dualism of verbal versus non-verbal behavior. It is his very fundamen-. — 59 —.

(12) On Pluralism in Linguistics tal claim that we should avoid such a dualism under a unified theory of language and behavior. Practically, the claim of form-meaning composite will lead to a fact that no linguist can decide any distinctive unit before the description of meaning in a given linguistic form or in a sequence of forms. Ironically enough, the last statement is accepted by Chomsky in his statement of objection against the notion of the formmeaning approach.48) Chomsky says,49) " there are utterance tokens that are phonemically distinct and identical in meaning (synonyms) and there are utterance tokens that are phonemically identical and different in meaning (homonyms)." It is important to notice here that Chomsky 50) is not interested in attempting to answer the question, raised at the beginning of this section, of whether or not the constructing of a grammar is possible without an appeal to meaning. It is because " the implication that obviously one can construct a grammar with appeal to meaning is totally unsupported. "50 This statement seems to reveal that Chomsky's definition of meaning does not include referential meaning. But he actually relies upon a referential meaning when he discusses the rules of PS and transformations. For example, he attempts to explain the ambiguity of phrases like " the shooting of hunters " in terms of deep structure. It seems to me that his description is based on the referential meaning of a particualr phrase. If not, how can he say that this phrase has a deep structure ? In the case that one is not familiar enough with the English language and knows the phrase only in the context like _ is prohibited here in this area, it seems obvious that he can not demonstrate the transformation of the structure,52) Yet his syntactic description is completely formal and non-semantic. Then we have to conclude that his term " formal " or " non-semantic " does not exclude the use of reference.53) The argument above seems reasonable, but actually it is not. Notice the following statements: 54) "... the semantic component of linguistic description will be taken to be a projective device... Such a projective device consists of two parts: first a dictionary that provides a meaning for each of the lexical items of the language, and second, a finite set of projection rules. The projection rules of the semantic component assign a semantic interpretation to each string of formatives generated by the syntactic components... " To obtain such semantic interpretations, each lexical item in a string of formatives must receive a meaning on the basis of the semantic information in the dictionary." The order of rules in these statements should be: (1) syntactic component; (2) the. meaning of lexical items provided by a dictionary ; and (3) a finite set of projection rules. Remember that the syntactic component has no semantic property. It generates. — 60 -.

(13) Mitsuho Sugawara. strings of formatives whose lexical items have not yet any meaning provided by a dictionary. It would be thus correct to say that this interpretation simply reveals the rejection of the use of reference in a syntactic description. Suffice it to say here, as a conclusion of this section, that tagmcmics has in general treated " meaning as reflecting response reactions at some level of one of the lexical, phonological, or grammatical hierarchies."55) Transformational grammarians, however, criticize the use of meaning in taxonomy and tagmemics. They state that the interpretation of meaning is so broad in those two models that any response to language is called " meaning . Which approach has an advantage over the other in terms of the argument of meaning ? The discussion of meaning may require several volumes of. books until we come to the final decision that one type of linguistic analysis has the advantage over another. At present I strongly doubt- that we can arrive at a conclusion as such.. 2. 6 THEORY OF LANGUAGE We have discussed some controversial problems of indivisual facts and phenomena of language. There is in modern linguistics another disputable matter concerning the. purpose of linguistic research. What is a linguist' task ? What is the legitimate activity for the linguist? To these questions, structural linguists answer that it is the study of phenomena and their correlations : to discover the signaling devices (the structaral. units) of a language and their patterns of distribution. Tagmemists similarly assert that one of the major goals of grammatical description is " the cataloguing and Cdescribing] of stmcturally contrastive strings."56) Both the views thus assume, in common, discovery procedures as the basic practical purpose of the theory. While in transform ational grammars both the taxonomic and the tagmemic views are rejected on the ground that " there is no reason why the linguist limits himself to '' the study of phenomena and their correlations,' avoiding any attempt "57) to find some general and plausible principles which make clear the reason to set up such linguistic units and their patterns of distributions. This transformationalists' criticism may be very reasonable on the whole. There is, however, some difficulties in their claim, especially in regards to the theory of language learning and the assumption of the theory. As for the theory of language tearing, Morris Halle 58) clearly states that " a complete scientific description of a language must pursue one aim above all: to make precise and explicit the ability of native speaker to produce utterances in [a particular] language." If transformational grammars meet this condition, we can say that a native speaker must know the basic strings before the transformation can be treated: that is, he must know the history of derivation before he produces actual sentences. This implies the inner behavior of a speaker. Here let me raise two questions. First, is it really a linguist's task to observe a speaker's reasoning and make a statement on the. — 61 —.

(14) On Pluralism in Linguistics. inner configuration of his behavior ? Second, if it is so, is there any proof that the statement is actually the correct description of his inner behavior? Pike obviously gives a negative answer to the latter question, citing J. B. Can-oll's remark: 59) " There is ...no behavioral reason why a child, for example, could not learn an interogative structure before he learns a declarative structure; therefore, one would be surprised to confirm Lees' expectation that the order in which a child learns constructions corresponds to the logical taxis of transformations." (J. B. Carroll, " An Operational Model for Language Behavior," Anthropological Linguistics, 1. l, p. 43.) Carroll's remark would be true, especially in terms of the relations between the two structures as follows: John - eats - an - apple (1) John - ate - an - apple (2) If we are given the two sentences, what makes it justifiable to say, that both come from Np -C - Vt-Np rather than to infer that native speaker's inner behavior corresponds to the notion. that (2) is derived from (1), or (1) from' (21 ? Pike as well as some linguists of the British School are also very doubtful as to the first question which has been a subject of much controversy in terms of a linguistic theory since Bloomfield and Sapir. Robert Dixon,60) for example, insists that linguistic observation should be made of relevant manifestation of our mental reasoning, since any direct observation of the reasoning is impossible. It may be reasonable to assume that we can investigate the deeper configuration of reasoning, provided we have a welldefined device for the description of " relevant manifestation " (language structure). But on the contrary, it would be also reasonable to say that there is no proof in the assumption that the investigation of the reasoning thus made with a well-defined device is really the description of the deeper configuration of reasoning. In fact, there are many other linguists other than those of the tagmemic and the London schools, who believe that the investigation of mental reasoning is the task of other disciplines such as philosophy or psychology.. 2. 7 LINGUISTIC CORPUS In the discussion of linguistic corpus, I should like to begin with the statement: " Failure to generate all the sentences means the grammar does not describe the full language."6!) Theoretically, the statement seems to be valid, but what is to prove the assumption that the generated sentences or those to be generated are all the 62 —.

(15) Mitsuho Sugawara. grammatical sentences of a particular language and nothing left7 It seems to me that. the notion of the possibility underlying the statement is really a gigantic assumption in linguistics today. If it is correct to assume that we are not given any way to calculate the number of sentences of a language, it would be important for us to continue to seek out new structures.62) In fact, the progress and revision of the general theory by the discovery of new facts are not rejected by Chomsky.69) In other words, the truth of the general theory is tentative, no matter how precise it is at a given time. We can thus say that it has not been proved that the grammar based on such a theory describes all the sentences of a language.. The general theory which has been based mainly on a particular language (Fnglish) can also be revised or refined by the discovery of new facts about other languages. Yet the theory is assumed to work on all the languages in the world, the number of which is more than 4,000. It seems to me that such a gigantic assumption is like one's belief that the sun will rise tomorrow again: that is, it is not based on rigorous proof.. 3 CONCLUSION Today many people tend to think that the transformational approach is now the only one well-defined and generally accepted linguistic approach. It seems, however, that the truth in the approach sometimes appears to be very strong.64) Some (or, it may be a very few ) parts of the approach have not been generally accepted as we have seen above. It may be unreasonable to state the similarities or dissimilanties between two different approaches in terms of their specific descriptive devices without the discussion of the respective rigorous theories which prescribe them. But we can assume that there does not exist any one well-defined and generally accepted linguistic approach, if there is no general agreement on such matters as the problem of phonemicization, of parts of speech, of 1C analysis and of meaning. Thus Pike s statement: " ...So it would seem possible that if tagmemics and transform grammar are both developed for enough, that they could come to the point of complete overlap " is very dubious. It is because some of the characteristics in transformational grammars such as ordering of rules, optional vs. obligatory character of rules and possibilities of permutations, etc. are all, as Postal says,66) in conflict with the characteristics in the tagmemic theory. The same is true of the reverse. Some characteristic features in tagmemics are not accepted in transformational grammars. It is thus only by the preservation of unique theories and practices in each approach with elaborate attempts of progress and revision that we can say, "if grammars are produced, an evaluative procedure will result ;n better and better grammars, and if we are to produce better and better grammars, an evaluative procedure is necessary.'W). — 63 —.

(16) On Pluralism in Linguistics. If we look at the plurality of linguistic approaches from a different point of view, such as language teaching, we have then a stronger stand for accepting pluralism in linguistics. Structural and tagmemic grammars do provide us new insights into the principles of language teaching, and they do offer sound materials for foreign language teaching in the schools. Trausformational grammars, on the other hand, may help us with some new insights into our teaching materials: the notion of deep structure will lead us to a more systematic comparison of different syntactic structures between. languages; and the distinction between an optional and obligatory parts does clarify the nature of elaborate and automatic habit-forming practices. Levels of lexis (say, 'idiomatic expression') are presented most elaborately in pre-structural grammars. Inventories of various linguistic elements which we need in language teaching are also well presented in those grammars. If the plurality in linguisfcics is preserved, we can say : "...we do not want teachers who have been falsely indoctrinated. with the idea that by learning a single grammatical system they have learned enough to teach grammar. We should give prospective teachers, not a course in English grammar, but a course in English grammars... We should...teach one of the scholarly non-structural books like Curme's or Jespersen's Essentials.. These works provide much more information about the language than most of their structural rivals do... We should teach some one of our structural grammars, and as soon as we have the necessary competence and the necessary materials, we should introduce [the prospective teachers] to transformational analysis."685 Footnotes 1) Cf. Robert Dixon, Lingtiistic Science and Logic, (The Hague, 1963), especially § 1. 1, 2) Paul Postal, " Constituent Structure," JJAL, XXX (1964), 15. 3) Here I used the term " generative grammar " as being interchangeable with " transformational grammar ", following Jean Malmstrom. See her article " The Workshop of Generative Grammar," Lingiustics Composing and Verbal Learning, (Chicago, 1962), p. 29. Cf. Chomsky's "Some Methodological Remarks on Generative Grammar," Readings in Applied English Linguistics, ed. H. Alien, (New York, 1964), p. 182, and Emmon Bach's An Introduction to Transformational Grammars, (New York, 1964), p. 57.. 4) James Sledd, " A Plea for Pluralism," College English, XXXII (1961), 15-20. 5) See Pike's single vowel analysis vs. Trager-Smith's diphthong analysis such as /i/, /e/, /u/, /o/ /iy/, /ey/> /uw/,/ow/ respectively, and Hockett's /c/, /}/ vs. Bloch-Trager's /ts/, /dz/. (Cf. the discussion of uniqueness in Karl V. Teeter, " A Note on Uniqueness," Language, XXXXII (1966), 475-478.) Notice also the criticism on phonemic procedures offered by transformational grammanans. 6) Recently K. V. Teeter has discussed the phonetic nature of distinctive features in. terms of the criteria for the selection of relevant phonetic facts in the phonemic analysis. See his article, "A Note on Uniqueness," Language XXXXII (1966), 475-478, especially pp. 476-477. 7) Noam Chomsky, in the discussion of " A Transformational Approach to Syntax," The Third Texas Conference on Problems of Linguistic Analysis in English, ed. A. A. Hill, (Austin, 1962), P.. — 64 —.

(17) Mitsuho Sugawara. 171. 8) N. Chomsky, Current Issues in Lmgnistic Theory, (The Hague, 1964), p. 78. 9) Ibid., p. 88. 10) Charles F. Hockett, "A System of Descriptive Phonology, "Language, XXVIII (1942), 3-21. Pike's " structural pressure " (see his Phonemics) seems to be equivalent to Hockett's " pattern congruity." Notice also his " structural parallelism." 11) C. F. Hockett, A Course in Modern Lmguistics, (New York, 1958), P. 110. 12) N. Chomsky, Cwrmt Issues, p. 88.. 13) N. Chomsky, Aspects of the Theory of Syntax, (Cambridge, 1965), p. 1G. 14) In order to avoid the confusion of phonemic representation, Chomsky seems to. have used the. term systematic", because, according to him, the phonemic representation is. not systematic.. (See his Current Issues.) 15) N. Chomsky, Aspects, p. 16. 16) The data are given by Harris, Structural Lmgw'stics, (Chicago, 1951), pp. 230-231. 17) See R. Dixon, Linguistic Science and Logic. 18) N. Chomsky, Current Issues, p. 88. 19) The term " intonation " is taken here to mean all the features of phonology except those of segmental sounds. These features are stresses, junctures and pitches in my paper. 20) Among them, Longacre, Hockett, Hill, Pike, Elson and Pickett are the extreme examples. 21) Robert Longacre, Grammar Discovery Procedw'es (GO-P, hereafter), (The Hague, 1964), P. 9. 22) Robert P. Stockwell, " The Place of Intonation in a Generative Grammar of English, " Readings. in Applied English Lmgnisfics, pp. 192-200. 23) R. Longacre, GDP, p. 127. 24) I follow the American taxonomic and tagmemic approaches in the use of the term " hierarchy." Notice that the term is used differently by M. A. K. Halliday. See his " Categories of the Theory of Grammar, " Word, XVII (1961), 248. 25) The tripartite system of hierarchy in the tagmemic model. Lexical. Phonological. Grammatical. 1.. Phones. Morphs. Tagmas. 2.. Allophones. Allomorphs. Allotagmas. 3.. Phonemes. Morphemes. Tagmemes. 4.. Hyperphonemes. Hypermorphemes. Hypertagmemes. Meaning units. Grammatical meaning units. 5.. Differencial. sound units. This chart is given by Walter A. Cook, On Tagmemes and Transforms, (Washington D. C., 1964). pp. 37-38. 26) R. Longacre. GDP, pp. 7-8. 27) These stress and juncture groups seem to be micro- and macro-segments respectively, propounded by Hockett in his "Linguistic Elements and Their Relations," Language, XXXVII (1961), 29-53. 28) W. Cook, On Tagmemes and Transforms, pp. 35-36. 29) R. Longacre, GDP, p. 8.. 30) Ibid. 31) Robert B. Lees, "The Grammar of English Nominalizations," IfAL, XXVI, (1960), 22-23. 32) J. Sledd, "A Plea for Pluralism," p. 8. 33) N. Chomsky, Aspects, p. 17. 34) Readers may know Hockett's deep grammar" and surface grammar" in his A Course in Modern Lwguistics, but it seems to me that his two diffeent grammars were not discussed on the same assurmption as Chomsky's.. 35) Benjamin Elson and Velma Pickett, An Introduction to Morphology and Syntax, (Santa Ana, 1964), pp. 61-62. 36) See Longacre, GDP.. — 65 —.

(18) On Pluralism in Linguistics 37) See Longacre, GDP and Elson-Pickett, Introdtiction. 38) In this respect, I agree with Postal. See his " Constituent Structure," especially chapters VII & IX. 39) N. Chomsky, in the discussion of " A Transformational Approach to Syntax," The Third Texas Conference, p. 173. 40) R. Longacre, GDP, p. 14.. 41) Kenneth L. Pike, Language in Relation to a Unified Theory of the Structure of Human Behavior, Part III (Glendale, 1960), pp. 36-37. 42) It would thus seem to be very absurd to show the similarity or dissimilarity of two approaches in such a simple diagram as the following:. Rewrite Rules (F). i). Tagmemic Formulae. S -> N-p + Vp. C1=+S:N +P:V ±0:N. ii) iii). Np -> T + N. N = + Lim : art + H : n. Vp -> V + Np. V = + H : tv. iv) V) vi). T -> the. art = the, etc.. N —> man, ball, etc.. n == man. ball, etc.. tv = hit, take, etc.. V —> hit, took, etc.. (as part of clerivation). (as listed in the lexicon) Clause. Sentence. ./~I\.. S : N P : V 0 : N. .vp. A. .1. ^ \ N V N. Np v. T. ^1 the man hit the. N ball. Lim : art H: n H : tv Lim : art H : n. r .1 .i .1'. the man hit the ball. The data are given by W. Cook. On Tagmemes and Transforms, p. 44. 43) See Chomsky, Syntactic Structures, (The Hague, 1957), Chap. 9, and Pike, Language, Part III, Chap. 16. 44) There are, however, some significant differences between Nida's approach and Tagmemics, since the former is largely based on the notion of 1C analysis (form-class composite analysis). While in the tagmemic approach the basic notion is the function-form composite. 45) See Elson-Pickett, AH Introcltiction. p. 18. 46) K. Pike, Language, Part I (1954), p. 24.. 47) Ibid. 48) See Chomsky's argument on the inadequacy of form-meaning assumption. (Chomsky, SS, Chap, 9, especially pp. 94-97.) 49) N. Chomsky, SS, p. 95. 50) Ibid., p. 93.. 51) Ibid. 52) See M. Halliday, " Categories," Word, XVII (1961), 249.. 53) See ^^,-^.?, ?^Cft®?^, (Tokyo, 1961), Pp. 14-15. 54) Jerrold J. Katz and Paul M. Postal, An Integrated Theory of Linguistic Descriptions, (Cambridge, 1964), p. 12. 55) K. Pike, Language, Part III, p. 35. 56) R. Longacre, "String Constituent Analysis," Language XXXVI (1960), 64. 57) N. Chomsky, Current Issues, p. 99. See also his "A Transformational Approach," (1958), and SS (1957), .especially Chap. 6. 58) Morris Halle, " Phonology in Generative Grammar, Tfie Study of Language, ed. Fodor & Katz,. (1964), p. 344. The assumption can be found in various articles and books on transformational grammars such. 66 —.

(19) Mitsuho Sugawara. as Katz-Postal's Integrated Theory, p. 173 & Chap. 5, and Chomsky's Aspcts, 1. §8, etc. 59) K. Pike, Language, Part III p. 36. 60) R. Dixon, Linguistic Science and Logic, p. 88. 61) P. Postal, " Constituent Stucture," p. 4. 62) R. Longacre, GDP. p. 10. 63) N. Chomsky, SS, p. 50.. 64) See Halliday, " Categories," Word, XVII (1961), 241. 65) K. Pike, Language, Part III, p. 36. 66) P. Postal, " Constituent Structure," p. 51.. 67) Archibald A. Hill, (ed.) The Third Texas Conference, (Austin, 1958), p. 184. 68) J. Sledd, "A Plea for Plusalism," pp. 19-20. References Consulted Bach. Emmon. An Introduction to Trmisrfomatienal Grammars. New York : Holt Rinehart and Winston, Inc., 1964. Bloch, Bernard. " Phonemic Overlapping," Readings in Linguistics, ed. Martin Joos. Washington D. C.: ACLS, 1958. Chomsky, Noam. Syntactic Strncures. The Hague : Moutoa and Co., 1957. Second Printing, 1961. " A Transformational Approach to Syntax," The Third Texas Conference on Problems of Linguistic Analysis in English, May 9-12, 1958. Austin, Texas : The University of Texas, 1962. Current Issues in Lwgnistic Theory. The Hague : Mouton and Co., 1964. " Some Methodological Remarks on Generative Grammar," Readings in Applied English Lmgtnstics, ed. Harold B. Alien. New York: Appleton-Century-Crofts, 1964. Aspects of the Theory of Syntax. Cambridge, Massachusetts : The M. I. T, Press,. 1965, Cook, Walter A. S. J. On Tagmemes and Transforms. Washington D. C. : Georgetown University Press, 1964. Dixon, Robert M. W, Linguistic Science and Logic. The Hague : Mouton and Co., 1963. Elson, Benjamin and Pickett, Velma. An Introduction to Mofpholog', and Syntax. Santa Ana, California : Summer Institute of Linguistics, 1962. Halle, Morris. " Phonology in Generative Grammar," The Structure of Language : Readings in the Philosophy of Language, ed, Fodor and Katz. Englewood Cliffs, New Jersey : Prentice-Hall, Inc., 1964. Halliday, M. A. K. " Categories of the Theory of Grammar," Word, XVII (1961), 241-292. Harris, Zellig S. Structural Lmgnislics. Chicago: The University of Chicago Press, 1951. Sixth Impression, 1963. Hockett, Charles F. "A System of Descriptive Phonology," Language, XVIII (1942), 3-12. A Course in Modern Lingnistics. New York: MacMillan and Co., 1958. " Linguistic Elements and Their Relations," Language, XXXVII (1961), 29-53. Katz, Jerrold J, and Postal, Paul M. An Integrated Theory of Linguistic Descriptions. Cambridge, Massachusetts: The M. I. T. Press, 1964. Lees, Robert B. " The Grammar of English Nominalizations," IJAL, XXVI (1960), Part II. Longacre, Robert E. "String Constituent Analysis," Language, XXXVI (I960), 63-88. Grammar Discovery Procedures: A Field Manual. The Hague: Mouton and Co.,. 1964. Malmstrom, Jean. " The Workshop on Generative Grammar," Lmgnitics Composing and Verbal Learning. Chicago : College Composition and Communication, 1962.. 4'^A^. ?i^:)S®?^. Tokyo, Kenkyusha, 1963. Pike, Kenneth L. Plionemics: A Technique for Reducing Languages to Writing. Ann Arber, Michigan : The University of Michigan Press, 1947. Language in Relation to a Unified Theory of the Structure of Human Behao'wr.. Glendale, California: Summer Institute of Linguistics, I (1954), II (1955), III (1960).. — 67 —.

(20) On Pluralism in Linguistics Postal, Paul M. " Constituent Structure: A Study of Contemporary Models of Syntactic Description,". IJAL, XXX (1964), Part III. Stockwel), Robert P. " The Place of Intonation in a Generative Grammar of English," Language,. XXXVI (1960), 360-367. Sledd, James. " A Plea for Pluralism," College English, XXXII (1961), 15-20. Teeter, Karl V. "A Note on Uniqueness," Language, XXXXH (1966), 475-478. Trager, George L. and Smith, Henry Lee Jr. An Otltline of English Structure. Studies in Linguistics. Occasional Papers No. 3, Norman, Okla., 1951.. — 68 —.

(21)