Learning to speak in time: Perception and production processes involved in conversational turn-taking

Stanford University, Stanford CA

Investigators

Abstract

Every conversation is made up of turns at speaking. Many people are surprised to learn that speakers take turns with very little time between speakers (gaps average ~200ms) with hardly any overlaps. This is true of conversation in languages around the world. Since there's no single clue to indicate when someone will finish speaking, listeners must track speech in real time and attend to a variety of cues to when a speaker will end. But what information do participants in a conversation rely on? This project investigates how children and adults take turns 'on time' by studying natural variation in how quickly speakers take their turns, and by measuring onlookers' eye movements as they watch videos of two people talking. This project will investigate how much participants' ability to anticipate the end of a turn relies on words vs. intonation and rhythm (prosody) in everyday speech. Words give information about content, and also about the structure of upcoming speech. Prosody is a continuous cue to how a phrase is structured, and also relates to the words in many cases. Because young children are good at differentiating between some prosodic patterns, but don't yet know much about words or syntax, we hypothesize that their use of cues will differ from adults'. This project will study turn-taking in natural conversations in order to identify the developmental path for this conversational skill. Conversation is the gateway for analyzing interaction with other social beings. For children, conversation is a way to learn about language and the world. Trivial skills for adults, such as knowing when to come in and how to ask relevant questions, take children years to develop. Understanding how children develop these skills should also shed light on how they gain access to the information all around them.