We explore contextual adaptation of referring expressions with respect to referential ambiguity and communicative intention. We focus not only on whether people adapt, but also on how by contrasting lexical specification (e.g., "batter") and syntactic modification (e.g., "man in white pants") when discriminating between objects in natural scenes (e.g., a batter wearing white pants and a referee). There are three main results. First, we replicate that speakers adapt their expressions to avoid ambiguity. Second, communicative intention has an effect: participants tended to use more specific names in a discrimination task than in a descriptive task, even without referential ambiguity in the context. Third, when given the choice, participants tended to prefer more specific words over adding modification - that is, using lexical rather than syntactic means to resolve ambiguity. This suggests that it may be less demanding to increase informativity of referring expressions with lexical specification than syntactic modification.

We investigated how verbal communication with a robot differs from talking to a human in terms of brain activity by analysing an open-source fMRI dataset. We focused on modeling conversational dynamics rather than conversation as a whole, by analysing fine-grained events, in particular turn initiation. The results indicate that turn initiation in a conversation with a human involves higher activation in auditory and visual cortex than turn initiation with a robot. Conversely, listening to the robot showed higher engagement of auditory cortex than listening to a human. We suggest that verbal and non-verbal turn-taking cues provided by the human agent engage more cognitive processing for picking up the turn. On the other hand, listening to a robot agent requires more processing than listening to a human. Both findings suggest that the accurate simulation of appropriate turn-taking cues and behaviors will help robots to establish more natural conversation dynamics and that the use of brain imaging can provide valuable objective measurements for assessing user states in human-robot interaction.