Decomposing relevance in conditionals

Abstract

In many cases, the use of a conditional is felt to be inappropriate unless the antecedent is relevant to the consequent. A number of authors have recently considered this relevance effect, noting that it is difficult to defeat and concluding that it is part of the conventional meaning of conditionals rather than the pragmatics of their use. However, there are also systematic counter-examples to the relevance requirement, where a conditional is used precisely to convey the irrelevance of antecedent to consequent. I argue that both types of conditionals are better understood in terms of the interaction of a unified interpretation of conditionals that does not make reference to relevance, and a separate process of the establishment of coherence relations among successive clauses in discourse, regardless of whether conditionals are involved. This theory is supported by the distribution of discourse particles such as then and still in conditionals and other sentence and text types. I also show that this theory is consistent with previous experimental studies that have been claimed to support the conventionalist position, and to falsify an account of the relevance requirement based on coherence.