¹ This work was supported in part by the Transformations and Discourse Analysis Projects (National Science Foundation) under the direction of Zellig S. Harris at the University of Pennsylvania, and in part by a grant from the National Institutes of Health.

I wish to thank Paul Ziff, Hans Herzberger, and Carlota Smith for their help with this work.

² Webster's new international dictionary ² (Springfield, Mass., 1960).

³ Noam Chomsky, Syntactic structures (The Hague, 1957).

⁴ See for example A. A. Hill, ‘Grammaticality’, Word 17.1–10 (1961); and H. Maclay and M. Sleator, ‘Responses to language: Judgments of grammaticalness’, IJAL 26.275–82 (1960).

⁵ For studies by psychologists describing the elicitation and evaluation of introspective reports see J. E. Hochberg, ‘Perception: Toward the recovery of a definition’, Psychological review 63.400–5 (1956); W. R. Garner, H. W. Hake, and C. W. Erikson, ‘Operationism and the concept of perception’, Psychological review 63.149–59 (1956); and L. J. Postman, ‘Perception, motivation and behavior’, Journal of personality 22.17–31 (1953).

⁶ Recent experimental work within what has come to be called psycholinguistics is related to formal linguistic models in at least three ways: (1) evaluation of the theoretical work of linguistics, e.g. studies of recognition of transformed sentences by Miller and Mehler as reported by G. A. Miller, ‘Some psychological studies of grammar’, American psychologist 17.748–62 (1962); (2) the introduction of linguistic concepts into traditional problems and procedures of psychology, e.g. M. Glanzer, ‘Grammatical category: A rote learning and word association analysis’, Journal of verbal learning and verbal behavior 1.31–41 (1962); and (3) the study of children's state of language development in naturalistic settings, e.g. Martin Braine, ‘The ontogeny of English phrase structure: The first phase’, Lg. 39.1–13 (1963), and through the use of nonsense words embedded in a verbal and sometimes also pictorial context, e.g. J. Berko, ‘The child's learning of English morphology’, Word 14.150–77 (1958), R. W. Brown, ‘Linguistic determinism and the part of speech’ in Psycholinguistics, ed. S. Saporta (New York, 1961).

⁷ A list of abbreviations appears at the end of this paper.

⁸ On must be pronounced in both instances with low stress to get this result.

⁹ The word but has further uses that are traditionally called prepositional, e.g. Nobody but John is coming. These are not described in this paper.

¹⁰ The intonation contour of these sentences strongly affects informant judgments. This fact will be reflected in the rules of section II. Assume here that there is a ‘colorless’ (noncontrastive) pattern associated with these sentences, and that this pattern is intended in the examples.

¹¹ It is obvious that the choice of the indefinite article is crucial to informant choices in all these examples. A different outcome is achieved by substituting the, but the same general conclusions hold. See section II.E for a discussion of the relationship between the articles and pronominal replacement.

¹² Contrastive stress is indicated by small caps. Although physically the stress falls on a single syllable, I capitalize the entire word or phrase assumed to receive stress. Section II.H gives a more complete description of the stress patterns.

¹³ See for example the phrase-structure description given by Charles Fillmore in Indirect object constructions in English and the ordering of transformations, Ohio State University Research Foundation, Project on Syntactic Analysis, Report No. 1 (Columbus, 1962).

¹⁴ Semicolons in formulas indicate alternative rewritings; parentheses indicate optional elements; ellipsis (…) indicates that a list is not exhaustive.

¹⁵ R. P. Stockwell, The place of intonation in a generative grammar of English, Lg. 36. 360–7 (1960).

¹⁶ Following Stockwell's description, primary stress in the colorless pattern would fall on the last item that is neither a preposition nor a pronoun.

¹⁷ The symbol * is thus an element of description of the written language (where stress is unmarked) as well as the spoken language. For descriptions of written English (e.g. those of Z. S. Harris), it would seem that equivalent marking of repeated and contrastive elements is necessary.

¹⁸ One always is the first instance, another the second; the same order is (much more weakly) evident with this and that. The choice between this and that is further conditioned by features of the nonlinguistic environment. The point here is that that is as much a repetition of this as this is; both are members of a single morpheme (the).

¹⁹ In Figure 1, the expansions of NP are T, A, N. No intervening node is taken to dominate TA or TN or AN. In fact it would be convenient for the description of conjunction to assume a node (PreN) dominating TA. This would account for informant responses to I saw a tall and a short man; *I saw a tall man and short woman. Such decisions about constituent structure cannot be decided by reference to conjunction alone. Informant data for these marginal cases is in any case insufficient to describe the subtle choices here. For example, in the second instance above, one can generally get the informant to change his decision by asking him to imagine a monster who is at once a tall man and short woman.

²⁰ Noam Chomsky suggested to me the idea of conjoining strings n at a time, rather than two at a time (except, of course, for the case of but, which is noniterative). If conjunction is limited to two strings, a complex phrase structure must be built up to describe sentences with more than one conjoining particle. Not only is this picture of phrase structure counterintuitive, but it creates two problems of description: (1) the rule for deletion of the conjoining particle (comma intonation) becomes difficult, and (2) the distinction between coordinate and subordinate conjunction becomes obscure.

²¹ Considerable subclassification is necessary here. For example, color adjectives and numerals are not conjoinable with but: *I have three but four decisions to make; *I have seven but green apples to sell.

^21a All the following discussion is limited to these three coordinating conjunctions, which are certainly the major ones in English. It is possible to exclude the whole host of subordinating conjunctions from this class on a single criterial basis: subordinating conjunctions allow the pronominalization (or ‘deletion‘) to precede their antecedent noun. For example:

(In order that the description of pronominalization remain simple, it is necessary to assume that permutation of the conjuncts is a feature of the structures exemplified above.) There are, nevertheless, conjoining morphemes not excluded by this criterion that are not to be discussed here. An example is yet.

²² It may be argued that this deletion of * would be unnecessary had we assumed that [xxvi] preceded [xxiv]. In that case it would be necessary to compare for morphological identity of the strings from which the conjoined sentence was generated in order to place the stress, i.e. to make reference to the transformational history of the conjoined string. In principle this is no difficulty, but it requires separate statement of the positions in which pronominal replacement is allowable, and seriously complicates the description of tag sentences.

²³ Stockwell suggests that the lesser fade at the juncture between the two ‘main’ conjuncts can be described as an environmentally conditioned allophone of terminal fade (i.e. a constituent string, which is always marked phrase-structurally for terminal fade, has instead terminal sustain or lesser fade when it ends before the matrix sentence ends).

²⁴ R. B. Lees and E. S. Klima, ‘Rules for English pronominalization,‘ Lg. 39.17–28 (1963).

²⁵ V_reflex is a class of verbs requiring a replica of the subject as their object. Such objects are obligatorily replaced by reflexive forms. A detailed description is given by Lees and Klima.

²⁶ An anomaly is: Organisms reproduce themselves.

²⁷ This point of view is described in more detail in my unpublished master's thesis Conjunction with ‘each other‘ (University of Pennsylvania, 1960).

²⁸ For a more complete description of this class of verbs and their relation to reciprocal conjunction, see my master's thesis, and also Lees and Klima, op. cit.

²⁹ Some notation has been added here. Parentheses indicate optional elements as usual, but 〈 〉 indicate an element that is optional only in case some other 〈element〉 appears in the string. In [lii], W₂ is optional if W₅ occurs, and W₅ is optional if W₂ occurs. Alternatively, [lii] can be written as two rules.

The word one has been described previously as a stressed indefinite article (one ₁), as a pronominal replacing indefinite noun phrases (one ₂), and as an impersonal pronoun (one ₃). The pronominal one ₂ is never in contrast with the morpheme one introduced as a pronoun in [lii]; therefore we do not distinguish between them. The numeral one is taken to be a case of one ₁; thus by [liii]:

³⁰ Carlota Smith, ‘A class of complex modifiers in English’ Lg. 37 (1961).

³¹ For a description of the positioning of negatives, see Fillmore, op.cit. For a description of the positioning of verb affixes, see Chomsky, op.cit. A verb affix that cannot be affixed to its left neighbor (there is no M) is affixed to the verb, if the verb is its right neighbor. If something intervenes between the affix and the verb, this permutation cannot take place. In that case, the word do is introduced in the position of M to carry the affix.

³² Some informants exclude outputs of [lv].

³³ X+ = the prefix X; +X = the suffix X.

³⁴ C. S. Smith has discussed this notion in an unpublished paper.

³⁵ Dwight L. Bolinger, ‘Contrastive accent and contrastive stress’, Lg. 37.83–96 (1961).