The project seeks to provide the empirical basis for the comparison of major theories of scalar implicature in complex sentences. The core idea is to use insights of game theory for designing small computer games such that the behaviour of subjects playing these games shows whether or not certain implicatures are drawn. Expected advantages of this experimental design are: first, the toy game fixes the relevant context, and hence blocks accommodation of arbitrary contexts by the test subject; second the game theoretic underpinning provides criteria for telling which information is relevant at the time of utterance; and third, the interpretation of utterances can be indirectly inferred from the behaviour of subjects. We expect that purely behavioural responses are less affected by the test subject’s meta-linguistic or prescriptive conceptions of language use. As the proposed experimental paradigm is new, methodological issues will play an important role. This means that the behavioural approach has to be compared with experimental paradigms which have previously been used for testing implicatures of complex sentences, in particular, truth value judgement tasks and inference tasks. We hope that he project helps to overcome the deadlock which the theoretical debate about embedded implicature has reached.