This brief note concerns Kempson's review of Uhlenbeck's Critical comments on transformational-generative grammar (Kempson, I975; Uhlenbeck, 1973). My purpose is not to defend Uhlenbeck but rather to point out that Kempson does not understand him in one or two vital points. Kempson quotes Uhlenbeck's contention that, 'as far as the analysis [by Chomsky] of the sentence What disturbed John was being disregarded by everyone is concerned, one has not to do with one sentence with two different interpretations, but with two sentences each with a quite different syntactic structure', then to complain that Ehlenbeck's observation is a 'non-argument', since what Ehlenbeck observes 'is precisely what the formalism of deep structure predicts', Kempson then calls the entire issue a 'terminological squibble' (i i8). Not quite. If one says that a certain 'sentence' is ambiguous one is talking about a sentence token and not a sentence type. The term 'ambiguity' can obviously be used only in application to a token, whereas 'homonymy' holds only between types. Strictly speaking, as a 'sentence' must always be a type, it cannot, but only some concrete utterance of it can, be a token. However, it is surely harmless to use the expression 'an ambiguous sentence', provided one is not lead astray by the inconsistency of this expression. Uhlenbeck's point is, in context, quite clearly that of fighting the notion of deep structure. In Chomsky's theory any deep structure is only a type, and any surface structure is only a token, at least for the purposes of this discussion. There is, then, more at stake here than something merely terminological. When Uhlenbeck says that the above quoted sentence represents two different syntactic structures, he does not mean what Chomsky would mean by this, i.e. that the token in question can be traced back to two deep structures; rather, Uhlenbeck means that the difference in syntactic structure is found in two sentences, each one reading like the example quoted, but each one different in structure from the other, while no deep structure postulate is needed to bring out that difference. It might not be sufficiently worth while to make these observations, if it were not for the circumstance that Kempson gets the impression Uhlenbeck is a behaviourist (i I9). On the contrary, his position that one may recognize different structures in what would appear behaviouristically as identical entails unconditional rejection of behaviourism. I would not be surprised if Uhlenbeck