Proceedings of the Annual Meeting of the Cognitive Science Society

People choose their questions based on their conversational goals. But how do they decide between questions that would yield equivalent information? This study examines the mechanism by which a questioner chooses between constituent ("wh-") questions of positive vs. negative polarity, e.g., "Who came?" vs. "Who did not come?". Our Rational Speech Acts (RSA) model predicts the following: First, when a partial answer is expected, the question whose polarity aligns with the questioner's goal is preferred over its opposite polarity counterpart. Second, this goal alignment effect decreases when the structure of the questioner's decision problem requires an exhaustive answer. These predictions are confirmed in two experiments: Experiment1 shows that goal alignment emerges in a free completion task, and Experiment 2 shows that it persists in a forced choice task, modulated by decision structure. Our utility-based account thus parsimoniously accounts for a phenomenon unexplained in formal semantic theory to date.