Abstract: This paper urges that one of the great, quasi-theological debates in legal theory depends on answers to empirical questions. The debate is whether courts should be "formalistic," that is, whether they should interpret statutory terms in accordance with their literal meaning, or whether they should be permitted to reject literal meaning by reference to "purposes," or canons of constructions, or considerations of policy. Any good answer turns on what approach will minimize decision costs and error costs, and that depends on empirical judgments about the likely performance of courts and legislatures. There is discussion as well of information-eliciting and market-mimicking default rules in the interpretation of contracts and statutes.