¹ This statement may astonish some readers, but Chomsky himself has suggested it many times. Cf. ‘It would not be inaccurate to regard the transformational model as a formalization of features implicit in traditional grammars'—Preprints of papers for the Ninth International Congress of Linguists 512 (Cambridge, Mass., 1962); ‘In studying syntactic theory, we assume as a known empirical condition, a partial specification of P-markers of many sentences in many languages'—'On the notion “rule of grammar”’, Proceedings of symposia in applied mathematics 12.8 (Providence, R. I., 1961), emphasis mine.

We may, of course, view generative grammar purely as an abstract algebra. But in practice its symbols (S, NP, VP, etc.) are interpreted as standing for the categories of classical grammar. Generative grammar does not seem to be concerned with redefining these categories. It rather treats them as knowns to be introduced in terms of a linear sequence of genetic rules. As an example, cf. Chomsky's analysis of the pair (3) John is easy to please and (4) John is eager to please: ‘John in (3) is the direct object of please … while in (4) it is the logical subject of please‘—Preprints 518. This is, according to Chomsky, the level of descriptive adequacy. The progress to the level of explanatory adequacy is achieved by developing a linear sequence of genetic rules which contain the categories of classical grammar as primitive terms (cf. Chomsky's next example with the terms VP, Aux, Tense, etc.).

² E.g. ‘Every sentence must contain two parts, a subject and a predicate … The subject of a finite verb is in the nominative.‘—William W. Goodwin, A Greek grammar ³ 196–7 (Boston, 1892).

³ William G. Moulton, Lg. 29.176 (1953).

⁴ Cf. Hansjakob Seiler, Relativsatz, Attribut und Apposition 5–6 (Wiesbaden, 1960).

⁵ But cf. William S. Cooper, Set theory and syntactic description (The llague, 1964), and Emmon Bach, An introduction to transformational grammar, chapter 7 (New York, 1964).

⁶ My formulation is as informal and as easily intelligible as I can make it. The specialist will not need the indoctrination, even in a more formal style. My chief source is Patrick Suppes, Introduction to logic, part 2 (Princeton, 1957).

⁷ In metamathematics, substitution is different. It is the replacement of a variable by a constant. See Stephen C. Kleene, Introduction to metamathematics 78 (Amsterdam, 1952).

⁸ These definitions are developed in greater detail in my paper ‘Erfahrungen bei der syntaktischen Analyse eines englischen Textes’, 1. Kolloquium über Syntax natürlicher Sprachen und Datenverarbeitung, Saarbrücken 1963 43–76 (Wiesbaden, 1964). In spite of the difference of phrasing, my notion of transformation is similar to transformation in metamathematics. This is a one-one relation between sets of formulae (cf. Kleene §19). It need not have a fixed place in a linear sequence of genetic rules (as Chomsky's transformations must). Chomsky would therefore probably call it cooccurrence rather than transformation (cf. Preprints 562–3).

⁹ This definition allows of some latitude as to which specific formal changes are considered inflexional. This is close to analytic practice. For instance, some grammarians consider Fr. -ment in précisément an inflexional ending, others take it as a derivational suffix.

¹⁰ These are the first words of pp. 108–14 in Joshua Whatmough, Language: A modern synthesis (London, 1956).

¹¹ This means that the set of acceptable sentences and, ultimately, linguistic structure cannot be a function of its truth value or of what the speakers of the language concerned take to be its truth value.

¹² They are slightly adapted from Lewis Carroll, Alice in Wonderland 87 (Everyman's library 836) and Rudolf Carnap, Einführung in die symbolische Logik ² 83 (Vienna, 1960).

¹³ Many of the ‘normal sentences’ which grammatical handbooks use as examples occur in live speech as little as the marginal ones.

¹⁴ ab is one element, and abc is one element.

¹⁵ Cf. Leonard Bloomfield, Language 194–6 (New York, 1933); Seiler, Relativsatz 5–18. Bloomfield calls both i and iii endocentric. Seiler uses both endocentric and exocentric for different subclasses of i.

¹⁶ [Except in baseball jargon: Mantle hits in almost every game.—Editor] In the literature the construction is always taken as endocentric. This can be true only on a quite different level of abstraction, namely if a and b are interpreted purely as the word classes ‘verb’ and ‘noun’ of classical grammar. As some verbs need not be in construction with a nominal object, the construction cannot be exocentric for verbs as a class. Therefore it must be endocentric.

¹⁷ One difference between rules of expansion and rules of transformation is that the former apply to a single IC, mapping this on to a larger expression which contains the same IC. A rule of transformation may apply to more than one IC, and the relation between the two elements of the transformational pair need not be endocentric. Another and probably more important difference is that transformational relations are one-one, but expansion relations are many-many. For instance, exactly one passive phrase corresponds to each particular active phrase which is susceptible of the passive transformation, and vice versa, but many different substantives are expanded by many different adjectives.

¹⁸ The difference between the single g in big and the double gg in bigger is a matter of spelling, not of the language.

¹⁹ If they were recursive we could derive *a biggerer car than a Cadillac than a Cadillac from a bigger car than a Cadillac. A well-known recursive expansion rule is for the expansion of nominal constituents by certain subclasses of adjectives, cf. a man, a fat man, a big fat man, a grim-looking big fat man, an awful grim-looking big fat man, etc. There is one example of a recursive application of the comparative expansion in Shakespeare, Measure for measure 5.1.50–1. The phrase this is all as true, as it is strange (type 9 [i]) here forms, in its turn, the comparative expansion of a phrase of type 4: this is truer than that. The passage reads: It is not truer he is Angelo, then this is all as true, as it is strange.

²⁰ More formally, this is the class of elements forming (nontransform) subjects in subjeet-predicate constructions. Besides the substitutives of classical grammar this class comprises substantives with attributes, e.g. poor little John (cf. a smaller boy than poor little John), Joan of Arc (cf. a greater hero than Joan of Arc). The restriction nontransform excludes the subject clauses of classical grammar (e.g. what I know is this).

²¹ Restriction iii also accounts for the inadmissibility of the phrases

*more powerful money than influence

*a smaller man than a midget

*tastier butter than margarine

‘Influence’ is not a subclass of ‘money’, ‘a midget’ not a member of the class of men, ‘margarine’ not a special kind of butter. These examples are Carlota S. Smith's, ‘A class of complex modifiers in English’, Lg. 37.342–66 (1961). She ascribes their ungrammaticalness to the fact that ‘if a noun is indefinite, its modifiers may not precede’. In these examples ‘the first (shared) noun of the comparison is indefinite’ (Lg. 37.359). However, they can be converted into good English by using as n₂ a member of the semantic class n₁, even though the first noun remains as indefinite as it was, e.g.

more powerful money than the dollar

a smaller man than a pygmy

tastier butter than canned butter

The phrase tastier butter than margarine would also be good English if we understand butter as a general term for ‘spread on bread’ (cf. apple butter), thereby making margarine a special kind of butter. Of Smith's other examples more pitiable squalor than this satisfies restriction iii (Smith ascribes its acceptability to the fact that ‘the second noun is definite‘), margarine tastier than butter represents my construction 3 in the same way as would money more powerful than influence, a man smaller than a midget, butter tastier than margarine. The semantic restriction which applies to construction 1 does not apply to construction 3.

²² The forms *red hotter, *mighty less are ruled out by restriction iv.

²³ The distribution of these class words is related to plural formation. One replaces plural-forming nouns (a bigger car than the one my neighbor has). For other nouns kind or amount are used (cf. better wine than the kind you drink, more money than the amount I earn).

²⁴ A relative clause may be defined as a transform of an ordinary main clause. One of the nominal constituents of the latter is moved to the initial position and linked to the remainder by a relative pronoun, e.g. my neighbor has a car → the car which my neighbor has, cf. ‘Erfahrungen’ 51 ff.

²⁵ This was recognized by Smith, Lg. 37.349. The rule is suggested in traditional terms by many grammarians, for instance Wilhelm Franz, Die Sprache Shakespeares ⁴ §215 fn.1 (Halle, 1939): ‘Bei der Vergleichung von zwei verschiedenartigen Qualitäten desselben Dinges ist die Komparativform auf -er jetzt nicht mehr üblich’; Otto Jespersen, MEG 7.§11.38: ‘When two qualities of the same person or thing are compared, the rule is to use more'; H. Poutsma, A grammar of late modern English 2:1A.491 (Groningen, 1914): ‘When two qualities are compared, periphrastic comparison is the rule.’ These grammarians are unable, on the basis of this rule, to account for the use of -er in ‘The wall was thicker than it was high’ (Poutsma's example). This sentence belongs to construction 9, while the rule to use more (not -er) is valid only for constructions 2 and 5.

²⁶ other is not a comparative form in the same sense as (say) bigger, since the phrase a car other than comfortable is not an expansion of *an oth car.

²⁷ I use the subscript n₂ (rather than n₁), since the class n₂ has, in this construction, the same members as n₂ (not n₁) of construction 1.

²⁸ Transitive verbs are distinguished from nontransitive verbs by the passive transformations, cf. John likes Jack → Jack is liked by John, but John resembles Jack *Jack is resembled by John. Some nontransitive verbs are not intransitive in the sense in which this term is usually understood, e.g. cost, depart, get, perjure in this costs one shilling, he departed this life, you'll get it, he perjured himself. Some transitive verbs consist of several ICs, i.e. call on in John called on Jack → Jack was called on by John.

²⁹ Word 14.258–9 (1958).

³⁰ Cf. this is nastier than I can stand, but hardly without the auxiliary °this is nastier than I stand.

³¹ This rule is the same as in construction 1.1 above.

³² Adverbial elements are recognized by the fact that they expand not the subject or the predicate, but the whole group ‘subject + predicate’, cf. ‘Erfahrungen’ 62 f. By this definition elements like very are not adverbial.

³³ This rule does not operate in the sentence type represented by it is nice to know John, cf. it is nicer to know John than it is to know Jane, but not *… than it is to Jane.

³⁴ A characteristic response is ‘Oh, you'll hear uneducated people say that.‘

³⁵ I bought a more expensive car than John sold is constructionally ambiguous. It belongs either to 9 or to 1.1; cf. I bought a more expensive car than the one which John sold.

³⁶ The full text of the passage reads: ‘And men will then see what difference there is between the magnanimity of a triennial Parliament, and that jealous haughtiness of prelates and Cabin Counsellors that usurped of late, whenas they shall observe ye in the midst of your victories and successes more gently brooking written exceptions against a voted Order than other Courts, which had produced nothing worth memory but the weak ostentation of wealth, would have endured the least signified dislike at any sudden Proclamation.‘—Areapagitica (Everyman's library 795.3).

³⁷ Derived through substitution from the car is bigger than is comfortable (6).

³⁸ Derived through substitution of bigger for more big from the car is bigger than it is comfortable (9).

³⁹ George O. Curme's example, Syntax 302 (Boston, 1931).

⁴⁰ MEG 7.§11.42.

⁴¹ Lyly's passage reads: ‘that our bodies might the better be governed, our seasons the daylier give their increase’ (Everyman's library 402.100).

⁴² Jespersen (2.§14.13, 7.§10.53, §11.18) partially recognizes the distinction between the two sets of forms, but does not follow it up in the body of his treatment, cf. 2.§9.41, 7.§11.13, §11.49, §11.77.

⁴³ Cf. Tauno F. Mustanoja, Middle English syntax 281 (Helsinki, 1960); Jespersen 7.§11.11.

⁴⁴ Cf. fn. 21 above.

⁴⁵ Lg. 37.344 fn.4.

⁴⁶ The Times 1 Aug. 1964, p. 7b (London).

⁴⁷ Several such phrases have already been discussed; see especially in §2 and §9.

⁴⁸ Shakespeare, Richard II 5:5.

⁴⁹ For attempts to define ‘normal’ and ‘emphatic intonation’ see Robert P. Stockwell, ‘The place of intonation in a generative grammar of English’, Lg. 36.360–7 (1960), and my paper ‘Erfahrungen’ 63 f. My hypothesis is that certain ‘normal’ intonation characteristics are inherent in particular syntactic constructions and that emphatic forms are derivable therefrom by specific rules of transformation. It is being tested on extensive material by Henning Wode (Freiburg) in his doctoral dissertation.

⁵⁰ I use Pike's notation, The intonation of American English (Ann Arbor, 1945).

⁵¹ Edward Sapir, quoted by Chomsky, Preprints 525.