Audience Design in Multiparty Conversation

How do speakers design what they say in order to communicate effectively with groups of addressees who vary in their background knowledge of the topic at hand? Prior findings indicate that when a speaker addresses a pair of listeners with discrepant knowledge, that speakers Aim Low, designing their utterances for the least knowledgeable of the two addressees. Here, we test the hypothesis that speakers will depart from an Aim Low approach in order to efficiently communicate with larger groups of interacting partners. Further, we ask whether the cognitive demands of tracking multiple conversational partners' perspectives places limitations on successful audience design. We find that speakers can successfully track information about what up to four of their partners do and do not know in conversation. When addressing groups of 3–4 addressees at once, speakers design language based on the combined knowledge of the group. These findings point to an audience design process that simultaneously represents the perspectives of multiple other individuals and combines these representations in order to design utterances that strike a balance between the different needs of the individuals within the group.     

Perceptual Categorization: Connectionist Modelling and Decision Rules

Although it is currently popular to model human associative learning using connectionist networks, the mechanism by which their output activations are converted to probabilities of response has received relatively little attention. Several possible models of this decision process are considered here, including a simple ratio rule, a simple difference rule, their exponential versions, and a winner-take-all network. Two categorization experiments that attempt to dissociate these models are reported. Analogues of the experiments were presented to a single-layer, feed-forward, delta-rule network. Only the exponential ratio rule and the winner-take-all architecture, acting on the networks' output activations that corresponded to responses available on test, were capable of fully predicting the mean response results. In addition, unlike the exponential ratio rule, the winner-take-all model has the potential to predict latencies. Further studies will be required to determine whether latencies produced under more stringent conditions conform to the model's predictions.