News & Events

Explaining relevance and irrelevance in conditionals

Dan Lassiter

University of Edinburgh

Lattimore 201

Assertions like 'If the Democrats won, Bill was angry' strongly suggest that 'the Democrats won' is and 'Bill was angry' are relevant to each other. If it turns out that both statements are true, but Bill is angry about something unrelated, such conditionals may be felt to be false or misleading. Most semantic theories of conditionals do not account for this 'relevance effect'. Recent experimental and theoretical work has used this as evidence for a new approach – "Inferentialism" – that encodes relevance directly into the semantics of 'if'. Numerous experiments using truth and probability judgments have been used to motivate Inferentialism, but there are also systematic and unexplained exceptions to the relevance effect. I argue that discourse coherence does a better job of explaining both the effect and its exceptions. I also present experimental results showing that discourse coherence can influence probability judgments, explaining  data previously taken as evidence for Inferentialism. These data motivate a theory of indicative conditionals in which the semantics is relevance-neutral, but certain contexts favor a coherence relation that requires relevance.