People use language to accomplish things together. They must coordinate at multiple levels, most critically at the level of what the speaker means when producing an utterance and what their addressee takes them to mean. For example, consider the question “What are you doing tonight?” How might the addressee interpret this question? The speaker may be requesting information from the addressee (ex. as a college student calling a high school friend living in a different state) or suggest that they and their addressee meet up that evening (ex. two roommates trying to set up a dinner plan). To add to the complexity of interpreting the speaker’s question, the addressee even needs to infer the meaning of the words “doing” or “tonight”. Does scrolling through Instagram or putting ice cubes in the freezer count as activities implied by “doing” that the speaker would want to know about? Does “tonight” mean after 7? After 9?

As illustrated by an example as simple as the question “What are you doing tonight,” what is said and what is meant depends on what the speaker and addressee take to be the context and goal of their exchange. These complex beliefs are mentally represented (private to the speaker and addressee themselves) and thus inaccessible to inspection.

We explore the idea that people’s verbal behavior reflects their understanding of the activity and the plan (steps they need to take) for advancing the activity.

In order to study how questions are interpreted in conversation, we had participants play a card game with a clearly defined goal. While the goal of the game is clear to us, how the participants interpret each individual question and its role in taking a step towards achieving that goal is mentally represented by the participants. Analyzing their responses to the questions provides insight into the participants’ understanding of the question and the activity at hand (what is their partner trying to achieve with this question?). In this study we study how three different question categories consisting of yes/no questions yield different answer types from participants depending on how the participants interpret the goal of each question type.