Joint inference in pragmatic reasoning: Evaluating probabilistic models of embedded implicatures

Team: Michael Franke (ProComPrag, U Tübingen) and Leon Bergen (MIT)

The question whether and when pragmatic enrichments, like scalar implicatures, can occur in non-matrix position is crucial for understanding pragmatic inferences and processing in general. We focus on scalar enrichments in embedded position and in particular the disambiguation problem, namely, to predict how salient conceivable pragmatic readings are, in a given context. The project has an empirical and a theoretical component. On the empirical side, we would like to further our understanding of the preferred pragmatic disambiguation of complex sentences with potential embedded implicatures in context by a series of experiments utilizing different task-types in parallel. On the theoretical side, we explore the potential of different types of probabilistic pragmatics models to predict the observed gradient patterns of salience of pragmatic enrichments. By performing quantitative comparisons of the different models, we aim to determine whether purely pragmatic accounts can explain embedded implicatures, or whether richer types of reasoning – e.g., reasoning about the logical form or semantic content of utterances – are required to explain these inferences.