ABSTRACT Explicit and implicit claims of various approaches to contrastive linguistics are considered. Naive CL and general CL are differentiated. It is argued, on the basis of considerable research, that the strong claims of naive CL are generally false where they can be tested and useless where they cannot be made explicit enough to be tested. Four gradients of the contrastive linguistics hypothesis are considered, ranging from the naive assertion that the errors of second language learners are nearly perfectly predictable on the basis of a comparison of the native and target language grammars, to the other extreme suggesting that none of the errors second language learners make can be predicted on the basis of a contrastive analysis. The research favors a moderate CLH which merely claims that interference from the native language is a significant though small factor in second language learning. It is argued that the emphasis of contrastive linguistics in particular and structuralism in general on the surface form of language was perhaps misplaced. Present studies of the deeper generative system (expectancy grammar, interlanguage, or conceptual dependency system) that is being internalized by the learner seem far more promising. It is little wonder, for this reason, that contrastive analysis has lost much of its former appeal. Nevertheless, it is claimed that we can profit greatly by seeing the seriousness of the errors committed in superficial analyses of language. The current trend of placing greater emphasis on the relationship of utterance to experience (that is, the renewal of interest in pragmatics) seems more promising theoretically and more vulnerable empirically. These two qualities augur well for future study.