UCLA

Regular, everyday conversations do not occur in a vacuum. The physical and social context in which a conversation takes place is utilized through the course of that conversation—it is situated. The ways in which a context can be used vary. Sometimes we implicitly and unconsciously utilize context in interpreting others. While at a restaurant, you would effortlessly interpret “What are you going to have?” not as a general question disconnected from the current situation, but as a question about that particular restaurant’s menu. Other times we overtly utilize context to aid in communication. While out on a walk in the city, I might point to an ice cream cone that has fallen onto the street and say “Somebody’s having a bad day”, and it would be clear what I am talking about.

I argue in this dissertation that the psychological category of attention is key for understanding how situated conversations operate. In contrast to viewing communication as a process of pure information exchange, I argue that communication is helpfully analyzed as a process of attention management. Our utterances, gestures, and facial expressions do not serve to merely convey information, but serve to coordinate our attentional states. I begin by focusing on a particular communicative phenomenon that I call rich demonstration. It is, roughly, a deictic (pointing) gesture that functions to communicate an entire thought, in contrast to fixing the reference of a demonstrative expression such as ‘that’. Just by making something salient, one can seemingly say something. I argue that rich demonstration is a speech act in its own right, and provide a formal semantic framework for modeling its communicative effect in discourse.

I then proceed to analyze salience itself. The notion has proven useful for understanding a number of communicative phenomena, including reference resolution, certain kinds of pragmatic inference, quantifier domain restriction, and more. Focusing on cases of reference, I argue that the linguistic notion of salience is best understood in terms of the psychological notion of mutual attention. I discuss competing options, including mutual knowledge and varieties of joint attention that are weaker than mutual attention, and argue that only mutual attention suffices for establishing conversational salience. The picture that emerges is one on which conversational moves function in part to coordinate the attentional states of interlocutors.I take advantage of these considerations to argue that discourse information states are genuinely structured—they represent different types of information that can be updated independently. I argue that such structural models of conversation cannot be reasonably reduced to unstructured models, such as the Common Ground model, which flatten discourse information into one kind.

This dissertation therefore develops a view of conversation as attention management. This is a perspective on situated conversation on which communicative acts are both restricted by facts about attention and function to coordinate the attentional states of interlocutors.