生産文法に於ける文法ルールの研究（英文）

全文

(1)Title. 生産文法に於ける文法ルールの研究（英文）. Author(s). 畑, 雅勝. Citation. 北海道教育大学紀要. 第一部. A, 人文科学編, 17(1): 71-94. Issue Date. 1966-06. URL. http://s-ir.sap.hokkyodai.ac.jp/dspace/handle/123456789/3912. Rights. Hokkaido University of Education.

(2) Journal of Hokkaido University of Education (Section I A). Vol. 17, No, 1. Tune, 1966. A Study on Some Grammatical Rules in the theory of Generative Grammar. Masakatsu HATA The Department of English, Sapporo Branch Hokkaido Educational University. TABLE OF CONTENTS I. n.. INTRODUCTORY. II. 5. Some Inadequacies in PS rules. IH. GRAMMATICAL TRANSFORMATIONS. PHRASE-STRUCTURE RULES H.. 0.. Introduction. III. 0. Introduction : Transformations. n.. 1.. Some Basci Notions from Chomskv. ffl. 1. Kernel Sentences. n.. 2.. Immediate Constituents and Phrase. M. 2. Types of T Rules. Markers. ffl. 3. Some T Rules for English HI. 4. Kernel or Non-Kernel Sentences. n.. 3.. Restriction on Phrase-Structure Rules. n.. 4.. Some Example of Rewrite Rules and. BIBLIOGRAPHY. Derivation. I. INTRODUCTORY I. 0. We aim in this paper to study some grammatical rules which underlie the •sentence structures of English. I. 1. When a native speaker of English hears two utterances and judges the one is. English and the other is not, we say that he has an ablility of judging some utterances to be genuine utterances of his own language. We call his ability of judgement a native speaker's intuition. In other words, we may say that he knows grammatical rules of English, because only by his intuition we can test the adequacy of a grammar proposed for English. I. 2. The fundamental aim, Chomsky says, in the linguistic analysis of a language L is to separate the grammatical sequences which are the sentences of L from the ungrammatical sequences which are not the sentences of L and to study the structure of the grammatical sequences.1 We shall consider briefly the substance of a grammar before taking up grammatical rules. According to Bach,2 the linguist begins his study by segmenting the stream of speech into discrete units of sound. And he continues, "In fact, this attempt has occupied the energies of linguists over the last few decades Noam Chomsky, Syntactic Structures, (The Hague. 1957), p. 13. Emmon Bach, An Introduction to Transformational Grammars, (New York, Chicago, San Francisco: Holt, Rinehart and Winstion, 1964), p. 4.. —71 —.

(3) Masakatsu Hata. and led to a theory of representation called phonemics. Because phonology is closely. related to the physical side of speech, it is easy to fall into the error of imagining that phonology is somehow more "objective" or sure than the other parts of linguistic study." But when we intend to consider the internal states of langaage, these operations such as segmentation and classification are of no use, for we cannot reveal grammatical features, given only the phonetic form of speech. Segmentation and classification of utterances are only the first task of the linguists. Grammar is not a mere description of physical events but a subject of theory. What we must account for is that certain sequences are in the language—say, on the rules of the language—and the others are not.. I. 3. We argued above that the grammar is a subject of theory and it cannot be revealed by merely stating the observable phonetic forms. For instance, the following sentence; The shooting of hunters occurred yesterday. is a grammatical one in English. But it has two meanings in spite of showing only one (phonetic) form. This means that in this sentence structure there are two sentenceformation processes as the following ; ( a ) Hunters shoot something. ( b ) Someone shoots hunters. and these two sources, (a) and (b), were transformed respectively. After this process, the sentence; The shooting of hunters occurred yesterday. has been generated. This generative process exists in the mind of a native speaker, therefore he can understand this sentence. This means that he knows rules of his language completely. As the example above shows, an internal state of language cannot be explained easily by mere representations of phonetic forms only. If we intend to reveal the sentence structures by means of merely observable phonetic forms, we will soon find the limitations of this approach and miss the most important fact, say, generative process of the sentence structures. We do not deny the importance of the study of speech sound which is just outside one end of language. But how much we may study these events, we cannot understand internal states or workings of language. We also cannot use a language without intellectual capacity which gives phonetic forms of language deep meanings. This capacity is within the mind of ours. The working of language is not a physical process but it is based on something more intellectual.3 After we have obtained the ability of using the rules of language, we can comprehend 3 Accordingly, the definition of Bloch-Trager's, "A language is a system of arbitrary vocal symbols by means of which a social group cooperates." is quite an unsatisfactory one, because it emphasizes too much upon the physical phenomenon, but ignores a native speaker's intuitwn. Barnard Bloch and George L. Trager, Outline of Linguistic Analysis, (Baltimore: Linguistic Society of America, 1942), p. 5..

(4) A Study on Some Grammatical Rules in the theory of Generative Grammar. the workings of language, I. 4. So far, we have discussed that the workings of language cannot be revealed by mere physical processes. Next, we shall demonstrate that a sentence is not a mere arrangement of morphemes or words. If sentence structures were made clear by observing an arrangement of morphemes, the 1C analysis of sentence structures would offer more clear-cut results. For instance, if we analyze the following phrases;. ( 1 ) the growling of lion ( 2 ) the raising of flowers, the result of the 1C analysis of the two phrases is the same like T-N-P-N. This example shows that the two phrases are analyzed as having the same class-construction, and we miss the important fact between the two sentences. That is to say, we cannot make explicit the difference between the two sentences by adopting the analysis like that, I. 5. The result above shows that the sentence structures are not revealed by observing merely an arrangement of \vords and analyzing the sentence into the immediate constituents. The sentence structures shown as the example above have quite deeper structures. Some sentences have the same structure even though they are composed of totally different forms. Some sentences are understood in a certain way parallel to other sentences of a different structure. Some sentences are related in a definite way to certain other sentences. Some sentences are ambiguous, and so on.4 These facts should be explained by a linguistic theory. There are two layers in the sentence structures. They are surface structures such as phonetic forms, and deep structures. Grammar must describe deep structures. What we wish to reveal is deep structures of sentences. A grammar of the language L is essentially a theory of L." The complete theory will offer grammatical rules and explain structural relations between sentences. To describe sentence structures correctly, we should consider the process of generating sentences, and describe in the shape of grammatical rules. This is an approach to sentence structures. In the following chapters, we shall study phrase-structure rules and transformational rules of English.. II. PHRASE-STRUCTURE RULES II. 0. Introduction. English is not a finite state language. Thus it shows all the characteristics which a natural language contains. First, in Eglish, not all sequences of the elements are permitted. Hence a theory about English cannot be a mere list of the elements; we must show how these elements are combined into sentences. Second, the sentences of English are finite in length, but its members are infinite. Therefore, English is not explained by a simple theory of linguistic structure based on 'Markov process models' 4 E. Bach, An Introduction to Transformational Grammars, p. 12. ° N. Chomsky, Syntaciic Structwes.. — 73 —.

(5) Masakatsu Hata. and the like. Although we stated above that English is infinite sets of sentences, a grammar of the form illustrated will, however, generate only a finite number of terminal strings. Thus, the restriction that a grammar should meet is recnr&iveness. This is an important conception of a grammar. We shall study in this chapter a more powerful theory what we call Phrase-Structure Rules, considering conditions of this grammar.. II. 1. Some Basic Notions from Chomsky Here we shall only introduce Chomsky's notions. Chomsky defines his model for a constituent structure grammar in the following passage.1 "A particular simple assumption about the form of grammars'"would be that each rule be an instruction of the form rewrite </> as <p (symbolically, ^-»^), where ^ and <p are strings of symbols. Given such a grammar, we say the tf1 follows for a if (T = • • • <f> ... and ffl = • •• </>••• (that is, if a1 results from substitution of </> for a certain occurrence. of <1> in <r), where </> -> </> is a rule of grammar, we say that a sequence of strings a\, •••, o-i, is a (A-DERIVATION if </>=ai and for each i, tf,-+i follows from ff,. A (A-derivation. is TERMINATED if its final line contains no substring X such that X—>-w' is a rule. In particular, we will be interested in terminated S S j? derivations ; that is, terminated derivations that begin with the string S S S. Suppose that each syntactic rule <p—></' meets the additional condition that there is a single symbol A and a non-null string w/ such that i^XiAXa and </>=Xi\v /Xa. This rule thus asserts that A can be rewritten w/ (i. e., w/ is of type A) when in the context Xi—Xa, where Xi or Xa may, of. course, be mull. A set of rules meeting this condition I will call a CONSTITUENT STRUCTURE GRAMMAR. If in each rule <f>->^, <js is a single symbol, the grammar (and each rule) will be called CONTEXT-FREE; otherwise, CONTEXT-RESTRICTED."2 II. 2. Immediate Constituents and Phrase Markers The notion of immediate constituents is quite traditional. It is based on the belief that segmentation and classification of utterances can reveal the sentence structures of a languafe. Thus Chomsky says ; Customarily, linguistic description on the syntactic level is formulated in terms of constituent analysis (parsing). We now ask what form of grammar is presupposed by description of this sort.3 And, if we are given a sentence like;. ( I ) The man hit the ball. we find that the sentence as a whole consists of two parts, The man and hit the ball. The man in turn consists of The and man ; hit the ball consists of two parts hit and the ball, and finally the and ball. Ultimately, the sentence consists of as the following 1 N. Chomsky, "On the Notion 'Rule of Gremmar", Structnre of Language and Its Mathematical. Aspects, Proceedings of the 12th Symposium in Applied Mathematics, (Providence : 1961). 2 This passage is quoted from G. H. Harman, "Generative Grammar without Transformational Rules ; A Defence of Phrase-Structure," Language, 39, No. 4. 3 N. Chomsky, Syntactic Structures, p. 26.. — 74 —.

(6) A Study on Some Grammatical Rules in the theory o£ Generative Grammar. discrete elements, The, man, hit, the and ball. And a more adequate analysis would. be The. hit. man. ball. the. This cafcegorization is what we call 1C analysis. But this categorization is inadequate unless it is associated with labellinig.-* In order to represent the labelling which can show the grammatical relations of elements, we write the following diagram,. Diagram ( 1) Sentence. VP. NP. N. NP. N the. hit the ball. man. We shall call such representations as Diagram ( 1 ) 'Phrase Markers' (P-markers). Since it is natural to consider that each sentence has at least one P-marker5, we can write another P-marker, given a sentence like ( 2 ) ;. ( 2 ) A boy likes the girl. and we compare these two P-markers and judge certain elements belong to the same class and some do not. Postal explains as; The fundamental notion in the P-markers is that certain strings of elements are related to certain other single elements by the relation 'is a (mamber of the category)'.". Thus, in Diagram ( 2 ), Diagram ( 2 ) Concerning the labelling, Postal states in Constituent Strnctwe, p. 7 ; "the labelling is necessary, if for no other reason, in order to represent correctly the similarities and differences of elements in different sentences and different parts of sentences." About the coHstructional homonymity, see Bach's Introduction to Tranfsormational Grammars, p. 40.. Paul Postal, Constituent Structure: A Study of Comporary Models of Syntactic Description, (Indiana : The Hague, 1964), p. 7.. — 75 —.

(7) B. c. we can represent this relation, say 'is a', as the following. Elements B and C are traceable back to A, so that these elements are said to be an A, or to be constituent of A. A. is said to dominate such strings B and C. Consequently in Diagram ( 1), (NP-VP) is a sentence, and so on.. As explained above, a proper P-marker (when represented in a labelled tree diagram) IS. a topological structure of lines and nodes comforming to the general requirement that a unique path be traceable from the termination of every branch to the point of origin of the whole tree ••••••7. And these diagram shown below are not permitted as a proper P-marker.8. Diagram ( 3 ). Diagram ( 4 ). Diagram ( 5 ). P-markers surely can tell us the grammatical category of words and phrases, but cannot tell us more. That is to say, the grammatical description of a language cannot be a mere list of P-markers, But we think it not fruitless to study the principles which generate the P-markers like above. Next we study what restrictions must be placed on PS rules, 7 E. Bach, An Introduction to Transformational Grammars, p. 71. 8 Ibid., pp. 71. — 76 —.

(8) A Study on Some Grammatical Rules in the theory of Generative Grammar. II. 3. Restrictions on Phrase-Structure Rules Phrase-structure (PS) rules are of the form X-. Y and obey the following restric-. tions.. ( 1 ) They are string-replacement rules (symbolically a-. »b).. ( 2 ) They are expansion rules (symbolically a—^b+c). ( 3 ) No rules of the form a+b—>b+a are permitted. ( 4 ) No rule of the form a—>a+b may be used.. Restriction ( 1) can be shown as; Diagram ( 6 ). x. Restriction ( 2 ) ; If there were in a rule,. C + A—>hit + the + ball then there would appear in a derivation succeeding lines containing the segments. A+C+A A+hit+the+ball. If such were the case, then there would be two ways of relating the lines ;. Diagram ( 7 ). C hit. A ball. the. or,. Diagram ( 8 ). A hit. /. the. ball. in such a derivation, it is impossible to construct a proper P-marker. This is the reason why restriction ( 2 ) is needed in PS grammBr,. Restriction ( 3 ) ; Regarding to restriction, ( 3 ), Postal states as the following ; the P-markers which result from permutations cannot correctly represent language structures because the asymmetry of the relation 'is a(n)' is not preserved. That is, this relation must hold between, say Noun and NP but not NP and Noun.9 ° P. Postal, Constituent Structure : A Study of Comtemporary Models of Syniactic Description, p. 13.. — 77.

(9) Masakatsu Hata. and he explains this restriction using the following instances. The sentences like John. will come and will John come are to be derived by ordered rules which are; S—>Noun+Verb Verb—>Modal+V Noun—>-Modal in Modal. Modal—Woun in Modal Noun—> John Modal—>will V—>come Then the sentence will John come will receive the P-marker ;. Diagram ( 9 ). Noun. Modal Modal. will. John. come. As Postal says, the categorization in Diagram ( 9 ) is linguistically absurd. This is the same to say that John is similar to will and mill is similar to boy, teacher, chalk, etc. Thus whenever permutation a+b—>b+a occurrs, there will be much confusions in PS grammar.. Restriction ( 4 ) ; Finally, restriction ( 4 ) states there must not be rules which change an element into itself plus another element. If there is a rule A—>A+B in a derivation to the left of an element C;. A+C A+B+C we cannot judge whether B is derived from A or from C. II. 4. Some Examples of Rewrite Bales and Derivation In this section, we shall follow the instruction 'rewrite X as Y' and constitute the grammar of some English sentences. At the beginning, let us suppose the following simple grammar. Gi Sentence. rl Sentence—>NP + VP. r2 NP—>T+N r3 VP—>Verb+NP — 78 —.

(10) A Studyon Some Grammatical Rules in the theory of Generative Grammar. r4. T—>tbe. 1-5. N—>cat, rat, etc,. r6. Verb—>caught, took, etc.. Di 1 Sentence. 2 NP+VP 3 T+N+VP 4 T+N+Verb+NP 5 the+N+Verb+NP 6 the + cat + Verb +NP 7 the 4-cat + caught+NP 8 the+cat+caught+T4-N 9 the+cat+caught+the+N 10 the+cat+caught+the+rat We shall call Di a derivation of the sentence The cat caughi the rat. If the last line of the derivation cannot be rewritten by the rules of Gi, we call the line a terminal. string (in this case, the+cat+caught + the + rat), and its symbols are called terminal symbols. Other symbols are non-terminal. Thus, a derivation is a sequence of lines beginning with Sentence, each obtained from the preceding by application of one and only one rule to a single element.10 Now we can write the derivation of (Di) as the following P-marker;. Diagran (10) Sentence. NP. VP. N. Verb. NP. N the cat caught the. rat. But in the grammar (Gi), Now, then, let us consider a more complete type of grammar. We shall take up the subject-predicate form again.". Ga Initial String SSff 10 Ibid., p. 11. u Robert B. Lees says, "Indeed, except possibly for certain 'pathological' colloquial forms, it is quite likely that all English sentences are whether of this form (Subject-predicate form) or else transforms of it." The Grammar of English Nominalwaticns, (Indiana: The Hague, 1963), p. 4.. — 79 —.

(11) Masakatsu Heta. Ri S—>NP+VP Since NP occurrs as a high order constituent of the predicate, objects of transitive verbs and predicate nominals in the predicate type of sentences, the VP must be developed first. Moerover, VP develops sentence patterns in various ways as He runs very fast or He runs a farm.. Rs VP—>Aux+MV Here the symbol of Aux represents the verbal auxiliaries, and MV represents the remaining part of the predicate, i. e., the finite verb and its modifiers, objects, etc.. R., MV—>Verb+NP R.i Aux—>Auxi (Auxa) where Auxi, containing the tense affixes, is obligatory. Often parentheses are used to enclose optionally chosen items. Ro Auxi—>Tns (M) M is shall, will, may, can, etc.. RB Aux2—>(have+En) (be+Ing) En represents a past participle. Ing represents a present participle.. R7 Tns-. Ip res (ent) (past. The import of this rule is to demand a choice between pres ent and past. Here we take present. :. Rg. The import of this rule is to demand a choice between singular and plural for each NP. Here we take NPaing.. (T+N RO NP,,,,,. pro; prop. I etc. Rio Nprop—>-Mary, John, ••• Rn Verb—>write, read, ••• Ri2 T—>the, this, a, ••• R 13 N—> ball, man, letter, •••. The derivation of the rules from Ri to Ris D. 1 Sentence. 2 NP+VP 3 NP+Aux+MV 4 NP+Aux+Verb+NP 5 NP+Auxi+Auxa+Verb+NP 80 —.

(12) A Study on Some Grammatical Rules in the theory of Generative Grammar. 6 NP+Tns+Auxa+Verb+NP 7 NP+Ths+be+Ing+Verb+NP 8 NP+pres+be+Ing+Verb+NP 9 NP»,ng+pres+be+Ing+Verb+NP 10 NP8,ng+pres+be+Ing + Verb+NPsi,,s 11 Nprop+pres+be+Ing+VeA+NPa.ng. 12 Nprop+pres+be+Ing+Verb+T+N 13 John +pres+ be -Hng+ Verb 4-T+N 14 John+pres+be+Ing+write 4-T+N 15 John+pres+be + Ing + write+a+N 16 John + pres + be + Ing + write + a + letter Thus the P-marker is as the following ;. Diagram (11). N'P. prop. Auxi Tns. John. pres be Ing. write a. letter. Thus the 'John + pres + be + Ing + write+a+letter' is a proper terminal string of the derivation. We cannot rewrite its symbols further. If we intend to make progress, much confusion would occur. That is to say, PS grammar would be more complicated and clumsy one. II. 5. Some Inadeeuacies in PS Rules The rules for the phrase-structure of English point up several difficulties in a grammar conforming to the adopted restrictions. As shown in II. 4, one of the difficulties is the treatment of the verbal auxiliaries and we recognize that the treatment is beyond the scope of PS rules. Clearly something besides PS rules must be included in an efficient grammatical theory. Before going on this theory, let us note some of the inadequacies of a phrase structure grammar. First, let us suppose the following two sentences ;12. ( a ) What are you looking for ? ( b ) What are you running for ? 12 Robert B. Lees, "Review of Syntactic Structures," Language, 33, (1957).. — 81 —.

(13) Masakatsu Hata. These two sentences have different sturctures. If we use the simple paraphrase, (a) contains prepositional phrase for what and (b) contains an interrogative ivhat for (=====why), but we have no mechanism for associating a sentence with its paraphrase. Then, the two sentences appear to have identical constituent structures. Secondly, as we stated in I. 2, there are many instances of sentences which are ambiguous in the level of the phrase-structure grammar. The sentence below can be justified as having only one single analysis. Thus the sentences; The stooting of hunters occurred yesterday, is understood in two ways; ( a ) Hunters shoot something. ( b ) Someone shoot hunters. There is no way, however, for assigning two different immediate constituent analysis and there is no word in the sentence which may be said to have two different lexical meanings, and it is more important that we have no way how to account for its ambiguity in the phrase-structure grammar. Aud finally, we have no way to show the following sentences are related each other;. ( a ) John hit hill. ( b ) Bill was hit by John. though they show a different constituent structure. And though the sentences;. ( a ) John hit Bill.. ( b ) Bill hit Jdhn. are not related each other, we analyze them as having the same structure.. III. GRAMMATICAL TRANSFORMATIONS III. 0. Introduction ; Transformations Here we shall consider a transformational (T) rule to be a rule of the form ;. X—>Y In PS rules, X and Y stand for single strings, but this time they do not represent single strings. Thus Bach states as the following; a transformation is a rule which requires us or allows us to perform certain changes in the terminal strings of the PS grammar is, and only if the string nas a certain structure. This structure must be precisely what is represented in the P-marker of this string as derived in a certain way from a particular set of rules.1 As we have seen in chapter II, PS rules could not provide us a mechanism how to analyze sentences successfully. But T rules are believed to provide us a powerful tool for deciding how to analyze sentences. We state transformational rules in the following way. Each rule must be marked E. Bach, An I'ntrocliiction to Transformational Grammars, p. 60. Similary H. A. Gleason states as the following: "A transformation is a statement of the structural relation of a pair of constructions which treats that relation as though it were a process." An Introduction to Descriptive Lwguistics, (Revised ecl.; New York : Holt, Rinehart and Winston, 1961), p. 172.. — 82 —.

(14) A Study on Some Grammatical Rules in the theory of Generative Grammar. as either optional or obligatory. The first part of the rule is a structural description (SD) which specifies the class of strings to which the rules apply. Thus let us take the sentence Bill hit John and state a SD of the sentence. The P-marker of the sentence is as the following ;. Diagram (12). VP. N'P. Aux. MV. Tns. Np,o,,. Bin. past hit. prop. John. SD might be as the following; SD: NP-Aux-V-NP The second part of the rule specifies the strnctitral change (SC), The operation is as. the following;. ii. I. i. SD: NP - Aux - V - NP. SC: Xi -. X, -. Xa - X... Then let us consider the passive transformation.2 Passive optional SD: NP - Aux - V - NP SC : Xl — Xa — Xa — X.i. X< - Xs +be+ En- X3 - by+ Xi Thus 'Bill+past+hit+John' may be transformed into the passive transformation;. John+past+be+En+hit+by+Bill The product of a transformation is referred to as a transform. As well known by the examples, the relation between PS rules and T rules are quite interesting. PS rules must be set up in such a way that such and such a string will undergo such and such a transformation. Further it is possible to conceive that a transformation has a power of explaining the relations between structures. Thus it explains similar connections between John is writing a letter and any of the following sentences. N. Chomsky, Syntactic Structures.. — 83 —.

(15) Masakatsu Hata. Who is writing a letter ? Is John writing a letter? What is John writing ? John is writing a letter, isn't he? The most interesting characteristic of T rules is that'they can apply to already transformed structures. Examples are the following;. John hit Bill.. i i i i. Bill was hit by John.. Who was hit by John?. Was Bill hit by John ? Bill was hit by John, wasn't he? III. 1. Kernel Sentences In the preceeding chapter, we studied some grammatical rules in PS grammar. That is, we followed the rewrite rule and obtained the following terminal strings by a stepby-step presentations of a set of rules;. ( 1 ) John + pres + be + Ing + write + a + letter And we also pointed out this step was the limitation of PS grammar so far as we were concerned this grammar. If we intend to make the germinal string to be a appropriate form, we must add at least the following rules; Pi, Af==En, Ing, Tns, •••••• Rig v====M, have, be, Verb. Ru, Af+v—>v+Af But this rule cannot be permitted in PS grammar, so that we set up an auxiliary transformation rule. T,,,,: X-Af-v-Y—>.X-v®Af-Y.3 Thus our terminal string becomes ;. ( 2 ) John + be + pres + write + Ing + a + letter (Taux-two times) Further, to make ( 2 ) a perfect form, we had better consider a word boundary instead of a concatenation sign. Thus let us add the following word-boundary transformation. T,,,b: X-Y=XSY. The effect of the rule is; ( 3 ) S John S be ® pres i? write ® Ing ji a S letter S © denotes an affix-juncture : see Lees' The Grammar of English Nominaliwtions, p. 48.. 84.

(16) A Study on Some Grammatical Rules in the theory of Generative Grammar. The morphophonemic rule will convert ( 3 ) into ; ( 4 ) John is writing a letter.'* The sentence like ( 4 ) is called a kernel sentence. Kernel sentences are defined as. the following by Bach: Sentences which are derived from terminal strings by the application of obligatory transformations only and phonological rules are called kernel sentences." It the kernel sentences are applied by a sequence of one or more transformations, we can generate many and more complex types of sentences." Examples are given by Harris.7. Si Mary has a sad fate. Sa Mary's fate is sad. Sa Mary's fate is a sad one. S< Mary's is a sad fate. These are transformed from some or all of three kernel sentences! Ki Mary has a fate. Ks Fate is a fate. Kg Fate is sad. The following transformations are involved. For Si : Kg overlap with Ki For S-2 : Ki N has N's, overlap with K:i.. For 83 : Ki N has N—>N's N, overlap with Ka (first N) ; Ka Pro-N of second N; Kg overlap with Ka (second N).. For Si : Ki N has N—>N's N overlap with Ka (first N) ; K..J Zero recurrence of first N; Ka overlap with Kz (second) N. Thus the setting up the kernel sentences is an important role for a generative grammar.. III. 2. Types of T Rules As we have studied in II. 3, PS rules contain only two grammatical operations, namely they are string replacement rules (a—>b) and expansion rules (a—>b+c). But T rules can carry out not only these operations but also the other grammatical operations. These operations which are not permitted in PS grammar will be listed below. ' We shall not treat the morphophonemic rule in this study. 6 An Introducion to T'ransformational Grammars, p, 69. " Harris says, "Transformations have some particular effects in the over-all structures of the language. They make possible an unbounded variety and length of sentences out of the stock of kernel sentences, thanks to the unbounded repeatability of various sequential transformations. They give an organized view of complex sentences." ; Co-occurrence and Transformation in Linguistic Structure", Language, 33, 283-340, (1957). 7 1 bid.. - 85 —.

(17) Masakatsu Hata. 1 A deletion rule (symbolically a+b—>b) 2 A reduction rule (symbolically a+b—>c) 3 An addition rule (symbolically a—>a+b) 4 A permutation rule (symbolically a+b—>b+a) The characteristic of T rules is that they may perform any of the six rules in any combination. We shall classify T rules into six types. Before listing each type of them, let us consider the major differnce between PS rules and T rules. As an example, we shall take up a replacement rule (a—>b). Suppose that we have the following P-marker ;. Diagram (13). x a. then, if we follow the PS rule a-. c. •b, the resulting P-marker will be as the following ;. Diagram (14). This fact shows that PS rules apply only to a simple string. On the other hand T rules may be regarded as applying to P-markers rather than simple strings. Therefore, if we apply a T rule a—>b to the P-marker described above, the result will be shown as follows.. Diagran (15). x. x (by the application of T rule). b This is the most important difference between PS rules and T rules. Now let us classify the types of T rules. 1. Deletion T rule SD : X - Y - Z SC ; Xl — Xa — Xg—^Xi — Xa 2. Replacement T rule SD : X - Y - Z. 86.

(18) A Study on Some Grammatical Rules in the theory of Generative Grammar. SC : Xi - X^ - Xs—>Xi - W - Xa 3. Expansion T rule SD : X - Y - Z SC : Xi - Xs - Xs—>Xi - P + S - Xs 4. Reduction T rule SD:X - R +. s. z. X, — ^V.l—7'Al X., 1 — ^2 ~ A^ ~ 1 — yv.l. 5. Additition T rule SD : X - Y. z. >x, ~- Xa — R — Xa Xa1 ~ X2 ~ XB—^Xi 6. Permutation T rule SD : X - Y - Z 1 ~ As ~ A3—T'A.2 ~ ^Vl ~ AS. III. 3. Some T Rules for English8. In the beginning, let us suppose the following terminal string which is derived by PS rules.. ( 1 ) John + pres + have + En + write + a + letter To make this string pronounceable, we must set up rules as the following. Af===En, Ing, Tns (Tns—> <fPres]>). Lpastj. v====M, have, be, V---"-. (A) X-Af-v-Y—>X-v+Af-Y. (B) X-Y—>X?Y (X+v or, Y+Af) We call (A) an auxiliary transformational rule (henceforth Tuux) and (B) a word boundary transformational rule (henceforth Twb). Both of them are obligatory T rules.. Thus ( 1) is applied by Taiix ; ( 2 ) John + have @ p res + write ® En+a+letter (Taiix-two times). then applied by T,vb ;. ( 3) <? John j? have ® pres S write ® En S a S letter {? Ultimately the morphonemic rules will convert this line of derivation into; ( 4 ) John has written a letter in phonemic transcription. These rules, as the examples shown above, play an important rule in a transformational grammar.. Now let us consider the rule which generates the sentence ( 5 ) John hasn't written a letter. In this section we shall follow Chomsky's approach, accordingly most of the examples are taken from Chomsky's Synlacdc Structures, especially pp. 61-84. 87 —.

(19) Masakatsu Hata. The rule which makes a negation introduces not or n'i into the auxiliary verb phrase. Henceforth we shall state this rule as Tnoi. In the case of a negation, not or n't must be introduced immediately after the second morpheme of the auxiliary phrase if the phrase contains at least two morphemes. If the phrase contains only one morpheme, T,,ot introduces not or n't just after the morpheme. But T,n,i must be applied before Taux. Thus Tnnt operates on the string;. (6) John+pres+have+En+write+a+letter and gives. ( 7 ) John+pres+have+n't+En+write+a+letter (Ultimately, John hasn't written a letter.) Next, let us consider how Tuot operates on the following string;. ( 8 ) John + pres 4- write + a + letter we can write;. ( q ) John+pres+n't+write+a+letter But we defined Tnoi should be applied before Taux, and this string does not contain a sequence Af+v. So that we must add to the grammar the following obligatory T rule which applies after Taux and T,,.i>;. X S Af S Y—>X S do © Af S Y We shall call the rule above Tdo. Thus ( 9 ) becomes. (10) S John f, do © pres+n't j? write S a » letter S (Ultimately, John doesn't write a letter.) Thus we can formulate Tnot as the following; Tnot—optional fNP-Tns-V.--..iNP-Tns+M-----.-. SD : {. NP-Tns+have-••••••. (.NP-Tns+be-••••••. SC : Xi-Xa-Xa—>Xi-X2+n'fc-X3. Next we shall consider the class of 'Yes-or-No' questions such as 'has John written a letter' 'can John write a letter' 'did John write a letter'. This time we set up Tq which operates on strings that are analyzed into three segments in one of the following ways ; NP-Tns-V-----'. NP-Tns+M-----.. NP-Tns+have--..... NP-Tns+be-••••••. Notice these strings are just the same as in the case of T,,oi. T,| has the effect of interchanging the first segment and the second segment of these strings. The requirement. That Tq should meet is that it applies after Tns-. 'pres} past;.

(20) A Study on Some Grammatical Rules in the theory of Generative Grammar. and before Txni.. (11) ( i ) John—pres—write+a+letter ( ii ) John — pres + can - write + a + letter. (iii) John—pret+have-En + write+a+letter (iv ) John — pres + be — In g + write + a + letter 'pres (By Tns—»•<;'"_"';>). jiastj. Then we apply Tq ; (12) (i) pres-John—write+a+letter ( ii ) pres + can — John - write + a + letter. (iii) pres+ have — John — En + write + a + letter (iv ) pres + be — John — Ing+ write + a + letter Next we apply Tnux and Tao;. (13) ( i ) ff do ® pres ff John <? write ff a » letter j? ( ii ) S can CD pres S John S write ff a ff letter {? (iii) S have ® pres {? John S write ® Enff a ff letter S (iv ) <? be ® pres {? John S write ® Ing j? a S letter S Ultimatly applying morphonemic rules, we derive;. (14) ( i ) does John write a letter (ii) can John write a lefcter (iii) has John written a letter (iv) is John writing a letter in phonemic transcription. If we apply Taiix and T,vb to (11) and don't apply Tq, we may derive the following sentences;. (15) ( i ) John writes a letter ( ii) John candwrite a letter (iii) John has written a letter (iv) John is writing a letter Thus (13 i-iv) are the interogative counterparts to (15 i-iv). As the examples have shown, "Transformational analysis brings out the fact that negatives and interogatives have fundamentally the same 'structure".9 Fourthly, we shall take up the problem of irregular behavior of 'have'. The terminal string which underlies the kernel sentence 'John has a book' is as the following;. (16) John+pres+have+a+book In this case 'have' is a fimite verb and not an auxiliary verb. Now let us consider how T,,ot and Tq operate on the terminal string (16). Tnoi applies to the string ;. NP-Tns-V-NP-Tns+have-••••••. 9 N. Chnmsky, Syntaclic Structures, p. 64.. — 89 —.

(21) Masakatsu Hata. But (16) contains the both structural descriptions. Hence if we apply Tnoi to (16), the results may be ;. (17) (i) John-pres+n't-have+a+book ( ii ) John — pres + have + n't— a + book. applying T,vi, to (17 i) (18) ( i ) John—do+pres+n't—have 4-abook (Ultimately, John doesn't have a book.). applying Taiix to (17 ii) (18) (ii) John—have+pres+n't—a + book (Ultimately, John hasm't a book.) Similary, if we apply Tq on (16), the results may be; (19) ( i ) does John have a book ( ii) has John a book But in the case of the other verbs except 'be', such forms as (18 ii) and (19 ii) are impossible. Thus we can say 'have you a book ?' and cannot say 'read you a book ?' This difference can be explained as the following ; 'have you a book' is derived from a kernel sentence 'you have a book' by applying Tq, bufc even if we apply Tq to a kernel sentence 'you read a book ', we cannot derive 'read you a book' from the kernel sentence. This means that the difference between two sentences is decided whether they are generated from the kernel sentences by any transformational rules or not. And if we consider the following P-marker;. Diagram (16). ss» NP. VP. MV. Aux. prop. A. V NP. Auxi Auxa. Tns / \ John. Verbtr T N. \. pres have En. write a. letter. We can not write ; (20) ( i ) John—pres—have + En + write 4-a + letter. but (20) (ii) John-pres+have-En + write+a+letter The reason is explained by showing that 'have' can not be traceable back to a node V. 90 —.

(22) A Study on Some Grammatical Rules in the theory of Generative Grammar. By this fact, the un-grammatical10 sentence 'John doesn't have written a letter' can not be derived. Now let us take up another types of interogatives. That is to say, we shall set up a ne\v rule which can generate the following sentences.. (21) ( i ) What did John write? (ii ) Who wrote a letter ? These sentences, of course, do not receive Yes—or—No answers.. The terminal string underlies (21 i) and (21 ii) is ; (22) John -Tns- write +a+ letter To generate (21 i) and (21 ii) we must set up a new optional transformation Tw. Concerning T,y, Chomsky explains as in the following; T,y operates on any string of the form X—NP—Y where X and Y stand for any string (including, in particular, the 'null' string-i. e., the first or third position may be empty) T,v then operates in two steps. ( i ) Twi converts the string of the form X—NP—Y into the corresponding string of the form NP—X—Y; i. e., it inverts the first and second segments of X—NP—Y. It has the same transformational effects as Tq. (ii) Twz converts the resulting string NP-X—Y into who—X—Y if NP is an animate NP or into what-X-Y if NP is inanimate."" Moerover he adds the requirement that T,, can apply only to which Tq has already. applied. Thus the order of applying the rules to the terminal string which underlies. (20 i) and (20 ii) is; Tns—> jpres{ =^ T, ==> T,,i ==^ T,.. =^^ '} .pasd ' " ' ^_^_,-_^ ' ^Tdo w. ==^ TW b :==^' Morphophophonemic rules For example, the terminal string. (22) John-Tns-write+a+letter is applied by; Tns—>past •••-> (John—past-write+a+letter) ^. '. ;. .. J.. .... 1. Tq ............... _> (past—John—write+a+letter). ^. ']['„! ..............._> (John—past—write+a+letter). _^. T^ ...............--> (who—past-write+a+letter). ^. Taux •••••••••••• —> (who—write © past+a+letter). J-. T,,b ...............-> (s who ff write ® past S a ft letter S). ^. morphophonemic rule"--> (who wrote a letter). Thus (21 ii) can be given. 10 On the notinn of wtgrammatical and non-grammatical, see C. E. Bazell. "Three Misconceptions. of Grammaticalness," Retport of the Fifteenth Annual Round Table Meeting on Luignistics and Language Studies, (Washington : C. I. J. M. Stuart, 1964), p. 4. 11 Syntaclic Structures, p. 69.. - 91 —.

(23) Masakatsu Hata. But in the case of (21 i), the result will be; (Omitting the initial steps.) Tq ...............-^. (past+John+write-a+letter) T,,i ..............._> ( a+lefctsr—past + John + write) T,^ ..•.•.•........-> (what-past+John+write). (There is no substring Af+v, so that we apply Tdo.) Tdo ...............-> (what-do © past + John + write). Ultimately 'what did John write' can be generated.. As we defined in III. 1, a kernel sentence is derived from a terminal string by applying obligatory T rules only. Thus the sentence;. (23) ( i ) John wrote a letter is a kernel sentence, because (23 i) was derived from the terminal string 'John+past + write+a+leter' by applying Tnu.x and T,,.b. But the following sentences;. (23) ( ii ) did John write a letter (iii) what did John write (iv) who wrote a letter are not kernel sentences. Because they are derived by applying not only obligatory but also optional T rules on the terminal string. The method to decide which sentences are kernel is an interesting problem. Generally speaking, if we require that the theory of grammar be as simple as possible, we had better let kernel sentences be simple, declarative and active sentences. Consequently, all the other sentences may be regarded as transforms of kernel sentences. In this section we have only treated Tnoi, Tq and T,,. all of which are based on obligatory T rules such as Taux, Tivi,, and Tjo. While we have not described T rules in sufficient detail but we can understand the grammar of English becomes much more simple and explicit by these rules. III. 4. Kernel or Non-Kernel Sentences Up to this point we have been concerned primarily with describing part of the apparatus of rules in a grammatical theory. First we provided PS rules for kernel sentences, that is to say, by PS rules we are given the terminal strminal strings which underlie kernel sentences. Thus we omitted. (be + En) from the PS rule Ro Auxs—>(have+En) (be+Ing) If we add (be+En) to R(|, our grammar will be more complicated. If we do so, we must add special rules indicating that in the case of VinirunaitivB.) it cannot have tlie auxiliary (be+En), whereas in the case of Vtrane.tivo, it must have (be+En). Now compare the following two P-markers.. Diagram (17) NPi+Aux+V+NPz (active SD) — 92 —.

(24) A Study on Some Grammatical Rules in the theory of Generative Grammar. sss NPi. NPi. Tns Vt NP.. Diagram (18) NP^+Aux+be+En+V+by+NPi (Passive SD). ss% VP. NPi. PPh. MV. NPi. Auxi Aux2. NPs. Tns be. Vt. by. NPi. Only by comparing these two P-markers, we are forced to recognize the complexity of passive sentences. Accordingly, we chose actives as kernel sentences instead of passives.12 Passives can be derived from actives by applying a passive transformation, A passive transformation is formulated as the following. Passive-optional SD : NP - Aux - V - NP SC : Xi — Xa — X3 — x,i. »X< - X, + be + En. Xs — by. Xi. By doing this, we can simplify the description of English. To summarize, given, a terminal strings by PS rules, then we apply obligatory T rules on it and gain a kernel 12 If we chose passives as kernel sentences, there arise many difficulfies. Chomsky explains as the following ; The active transformations would have to apply to strings of the form. (100) NPi-Aux+be+En-V-by+NPa Converting them to NPa—Auz—V—NPi. For example, it would convert. (101) the wine was drunk by the guests into "the guests drank the wine',••••••It thus appears that it is the simplest system of phrase structure for English; the sentence. (102) John was drunk by midnight is also based on an underlying terminal string that can be analyzed in accordance with(lOO)""" But application of the 'active' transformation to (102) does not give a grammatical sentence.. — 93 —.

(25) Masakatsu Hata. sentence. If we apply proper T rules on this kernel sentence we can generate grammatical sentences. Thus all sentences belong to either kernel sentences or transforms. Accordingly, if we combine PS rules and T rules properly we can gain all the grammatical sentences of English and reject ungrammatical sentences.. BIBLIOGRAPHY AnTAL LAsZLO. "A Review of Harris, String Analysis of Sentence Structure", Lingtlislics I, 97-104. Bach, Emmon. An Introduction to Tranceformatwnal Grammars. New York, Chicago, San Erancisco : Holt, Rinehart and Winstion, 1964. Bazell, C. E. "Three Misconceptions of Grammaticalness", Report of the Fifteenth Annual Round Table Meeting on Linguistics and Language Studies. Washington : C. I. J. M. Stuart, 3-9, 1964. Bloch, Barnard, and Trager, Goei-ge L. Olltlme of Linguistic Analysts. Baltimore: the Linguistic Society of America, 1942. Bloomfield, Leonard. Language, London : George Alien & Unwin LTD, 1935. Chomsky, Noam. Synlactic Structures. Indiana: The Hague, 1957. -. "On the Notion 'Rule of Grammar'", Strncture of Language and Its Mathematical Aspects,. Proceedings of the 12th Symposium in Applied Mathematics. Providence, 1961. Gammon, E. R. "On Representing Syntactic Structure", Language 39. 369-397. Gleason, H. A. Jr. An Introduction to Descripfive Lmgttistics. New York : Holt, Rinehart and Winstion, 1955, Revised ed., 1961. Harman, G. H. "Generative Grammar without Tsansformation Rules: A Defence of Phrase Structure", Language 39. 597-616. Harris, Zellig S. "Co-occurrence and Transformation in Linguistic Structure", Language 33, (1957). 283-340. -. "From Morpheme to Utterance", Readings in Lmgiustics. Washington : Martin Joos, 142-153.. 1957. -. String Analysis of Sentence Structure. 2d ed. The Netherlands by Mouton & Co.: The Hague,. 1964. -. Structural Linguistics. Chicago: The University of Chicago Press, 1961. Hill, Archbald A. Introduction to Linguistic Structures. New York : Harcourt, Brach & World, INC,. 1958. Hiz, H. "The Role of Paraphrare in Grammar", Report of the Fifteenth Annual Round Table Meeting on Lingutstics and Language Studies. Washington : C. I. J. M. Stuart, 97-104, 1964. Hockett, Charles F. Problems of Morphemic Analysis", Readings in Linguistics. Washington : Martin Joos, 229-242. Hockett, Charles F. "Two Models of Grammatical Description", Reading in Lingnistics. Washington: Martin Joos, 286-339. 1957. Lees, Robert B. The Grammar of English NommaUwtwn. Indiana : C. F. Voegelin, 1963. -. "Review of Syatactic Structures" Language 33, (1957). 375-408. Nida, Eugen A. Morphology : The Descriptive Analyiss of Words, 2d ed. Chicago : Ann Arbor, 1949. Postal, Paul. .Constituent Structure: A Study of Contemporary Models of Syntactic. Indiana University : C. P. Voegelin, 1964. Wells, Rulon. "Immediate Conttituents", Redmgs in Lrugnisties. Washington: Martin Joos, 186-207.. 1957.. 94 —.

(26)