1. On paraphrase.

The notion of syntactic paraphrase received formal definition, within standard theory generative transformational models, following the adoption of the Katz-Postal constraint that transformations preserve meaning. This development in the mid-sixties can be traced through Harris's attempts to relate sentences by purely formal operations, without any reference to meaning, and Chomsky's earliest transformational grammar. Despite its formal definition, however, the standard theory notion of paraphrase was not uncontroversial. Different linguists make competing claims about paraphrase relationships. One convenient way of characterising such differences is by considering the semantics of the sentence on three levels: (a) the propositional content of the sentence; (b) sentence modality; and (c) discourse semantics, which refers to the potential effects of context on internal sentence structure. Some linguists have considered sentences which are (a +b)-equivaient to be paraphrases; others have taken into account (c) as well, inevitably making different paraphrase judgments. It is suggested here that any experimental approach to determining paraphrase relations will need to rely on the stricter definition of paraphrase, taking discourse semantics into account. Clefted sentences are proposed as an area of English syntax for which a behavioural test for paraphrase would be illuminating.