In a paper read at the Tenth International Congress of Linguists, Bucharest, 1967 (Bennett, forthcoming), I discussed the question of synonymy within the framework of the stratificational theory of language, taking examples from the area of English prepositions. The present paper has two aims. On the one hand, it incorporates the findings of more recent work on synonymy. Thus whereas, for instance, in the earlier paper two levels of synonymy were distinguished, it now seems necessary to recognize four or possibly five levels. On the other hand, the paper will attempt to set the discussion of synonymy in a wider framework by sketching the outlines of a semological description of English prepositions. To present such a description would be beyond the scope of the paper. I hope merely to indicate the general shape that the description might take. In addition to the remarks on synonymy there will be some discussion of polysemy, componential analysis and idioms. In particular it is hoped that a clear picture will emerge of the relationship between these various areas.

Beginning with the paper of Chomsky, Halle & Lukoff (1956) a theory of accent has been developed which attempts to explain by certain fairly simple rules the complex pattern of degrees of prominence assigned to any given sentence. In the present paper I will not be concerned with those rules which specify the main stress for stems (or, more precisely, for words containing only one stem), but only with the rules which introduce degrees of prominence in compound words and phrases. Let me call these latter rules Phrase-Accent-rules, or for short, PA-rules. The general conditions on the form and the manner of operation of PA-rules may be stated as follows: (i) PA-rules apply cyclically, beginning with the innermost, or lowest constituents of given final derived phrase markers, proceeding ‘upward’ until the topmost constituent is reached. (ii) PA-rules pick out one of the several primary accents of a constituent to which they apply, and make it the main accent of that constituent, thereby lowering all other accents by one degree. (iii) Formally, PA-rules are strictly local transformations whose structural descriptions recognize only three factors: first, the constituents, or bracketing, of a sentence; second, the categorization of the constituents; and third, the previously assigned accents.

Previous parts of this paper have been concerned with two areas in the syntax of the English clause: Part 1 with transitivity, Part 2 with theme. Transitivity is defined as relating to the experiential component of meaning (or ‘cognitive’, though this term is not really appropriate since all components involve a cognitive stratum); the discussion has been confined to the expression of processes and the participants therein – syntactically, those functions having in general verbal and nominal realizations – although a full treatment of the experiential component in the syntax of the clause would take account of other features, the expression of time and place and other adjuncts to and conditions on the process. Theme is related to the discoursal, or informational, component; under this heading are brought together the principal options whereby the speaker introduces structure into the discourse and (in the ideal case) ensures ‘comprehension’ – the recognition of the text as a text, and its interpretation along predicted lines.

In each of the following pairs of sentences, the verb in (a) is intuitively apprehended as being related in some significant way to that in (b): (1) a. The rice grew b. The fanner grew the rice (2) a. The barge sank b. The storm sank the barge (3) a. The coconut dropped b. The man dropped the coconut (4) a. The cart turned b. The driver turned the cart (5) a. The laws changed b. The Government changed the laws (6) a. The mourners collected b. He collected the mourners (7) a. The boat sailed b. The fisherman sailed the boat (8) a. The meeting began b. The Chairman began the meeting For transformational grammarians working on English, the delineation of the kind of relationship illustrated here is vital for the achievement of (at least) descriptive adequacy in their analysis of the language. Deeper investigation of the relationship has, however, revealed that concealed beneath its apparent simplicity there are certain highly complicated problems with fundamental and far-rseaching implications for the theory of transformational grammar itself.

The present paper is initially concerned with a rather minor problem in comparative Germanic morphology: the shape of the consonant in certain present active indicative endings. There are discrepancies among the early attested dialects in this matter, and the traditional explanation has been that one group of Germanic dialects generalized present forms with Indo-European root accent, the other group forms with Indo-European suffix accent. I show that this solution, which is widely accepted, is deficient in several respects, and I demonstrate that the problem has a correct solution with interesting consequences for historical linguistics provided we concentrate on the rules and their order of application in the grammars of the different dialects, that is, provided we attack the problem from within the framework of generative phonology.

1. It is a well-known fact that in English wh-words occur both in interrogative sentences and in certain complex sentences (e.g. in relative clauses), and that whenever this is the case, the members of each such set of wh-words found are identical in form and have related meanings. For example, the interrogative pronoun in sentence (1) is identical in form to the relative pronoun in sentence (2); furthermore, these pronouns have a related meaning in that both refer to ‘human (subjects)’: (1) Who cut the pie? (2) The man who came yesterday cut the pie