A Model for Generating On-Off Speech Patterns in Two-Way Conversation

l.l Applications of the Model A model for generating on-off speech patterns in two-way conversations may have two uses: (z) It may provide insight on the dynamic processes which determine when a person talks or is silent. For example, the model proposed here allows a person to be in one of six states, depending on whether he is talking, listening, or both conversants are talking, and so on. Each state is associated with a parameter which could be in2445 2460 THE HELL SYSTEM TECHNICAL JOURNAL, SEPTEMBER L ' ) 0 F L terpreted as a "motivation" for either starting to talk or falling silent. As a subject talks over different experimental conditions, changes in the "motivation parameters" might be correlated with changes in subjective opinion of the circuit. (ii) The model may predict the statistical behavior of voice-operated devices (such as echo suppressors, voice-switched amplifiers) as the circuits are changed. One alternative to using a model is to have people talk over different circuits and study the circuit behavior. This is often unsatisfactory because too much data may be required to isolate the effects of a particular circuit change. Another alternative to a model is to record an experimental prototype conversation and then play it over different circuits. This is also usually unsatisfactory because the conversants cannot react to circuit changes; their behavior remains the same. A model has the advantage of keeping the statistical structure of the "conversants" unchanged while allowing them to react as the circuit parameters are varied.