Pragmatics is about language in use, and, in particular, about the emergence of meaning in interaction between speaker and audience. It is in keeping with the philosophy behind this discipline to test for pragmatic phenomena in interactive and task-based scenarios. We have successfully developed an experimental paradigm in which pragmatic interpretations are inferred from action choices based on a game theoretic model of the experimental situation. In this project, we further develop our methodology and widen its empirical coverage. Our core topic is one of the most complex and controversial phenomena studied in pragmatics today: scalar implicature in logically complex sentences. In the first funding period, we demonstrated that there are natural scenarios in which controversial implicatures are drawn reliably. We concentrated on situations with competent speakers and the all-some scale. In the new funding period, we intend (A) to extend our research to a larger variety of scales (adjectival scales), (B) study violations of the competence assumption (i.e. partial knowledge, uncertainty), and (C) investigate the relevance of the principle of cooperativity in contexts with opposing interests of speaker and hearer. The debate on complex sentences started out with the issue of whether or not implicature can be embedded in compositional semantics. This has been a hotly debated topic for which diverse theoretical models with conflicting claims have been developed, prominently among them feature structural accounts, and game theoretic accounts. Recently, Bayesian models have been proposed that learn production and interpretation strategies directly from experimental data. The advantage of game theoretic and Bayesian models is that they are able to link their theories to interactive experimental settings. In particular, they make predictions for both utterance production and interpretation. In accompanying theoretical work, we will therefore further these theories and develop detailed game theoretic and probabilistic models for all planned experiments.