^* This work was supported in part by the U. S. Army Signal Corps, the Air Force Office of Scientific Research, and the Office of Naval Research.

¹ The generative grammar that provides the context for this discussion was developed primarily by Noam Chomsky, who has given me much helpful criticism and advice. See his monograph Syntactic structures ('s-Gravenhage, 1957) and other papers by Chomsky; also Robert B. Lees, The grammar of English nominalizations = IJAL 26:3 (1960).

The grammar has three different parts, each consisting of an ordered series of obligatory and optional rules. The first, constituent structure, generates strings of words and symbols, not necessarily in order. The second, transformation, imposes structural changes on the strings; at this level the order of words and/or symbols may be changed, deletions made, and constants added. The third part, morphophonemics, converts the abstract representation produced at the second level to an utterance. This paper will deal with problems of relating sentences structurally and will therefore be concerned only with transformational rules.

² ‘Containing’ is used here to mean a dominating sentence, and ‘contained’ a subordinate sentence. (For the sentence The ball that he has is blue the containing sentence is The ball is blue, the contained is He has a ball.) These terms correspond roughly to ‘matrix’ and ‘constituent’ sentence in Lees's terminology, Grammar of English nominalizations 55.

³ A distinction must be made between restrictive and appositive relative clauses. Restrictive relatives are directly postnominal, while appositive relatives are separated from their nouns by commas or comma intonation. The following discussion applies only to restrictive relative clauses. An adjective or adjective phrase can always occur after a noun (that is, with wh- is deleted) if it is separated from that noun by a comma. In other words, when an adjectival relative clause is appositive, the relative markers can always be deleted. We will find that deletion of the relative markers is not always possible when an adjectival relative clause is restrictive.

⁴ The distinction that makes this sentence ungrammatical, while the previous sentence is grammatical, seems to be marginal for many speakers, although it is clear in my own speech. When a determiner more definite than the is substituted in sentence 6a, the resulting sentence is generally heard as ungrammatical:

Moreover, when a longer adjective phrase directly follows a noun with a definite determiner, as in sentence 8a, the sentence is generally heard as ungrammatical.

⁵ An additional general restriction further limits the occurrence of adjectives with verbal complements as direct modifiers to nouns. Certain adjectival modifiers (however, so, too, as) occur before the article when they are in prenominai position; these words may be embedded to prenominai position only if the article is not zero, as the following examples indicate:

As we have shown, adjectives with verbal complements may be embedded to prenominai position only when their nouns have indefinite determiners. Together, the two generalizations result in the following restriction: adjectives with verbal complements that occur prenominally before the article cannot occur prenominally when the article is zero. The need for such a restriction was pointed out to me by Edward S. Klima.