Learning to Identify Nonverbal Sounds

Abstract

Research is under way to determine whether or not the techniques of automated teaching (active participation of the learner, immediate feedback, presentation of items conditional upon the learner's recent past performance, etc.) will significantly facilitate learning to identify nonverbal sounds. We suspect that parametric study of the lesson variables will define a teaching procedure that produces results surpassing the limitations on identification ability disclosed by previous research. This paper describes the research design and apparatus, and the results of the experiments conducted to date. In this study, a PDP-1 digital computer is used to realize a large number of teaching machines—the program permits the experimenter to set up a lesson in any format desired. The computer exercises simultaneous control over independent lessons for several subjects, and produces a complete record of the lesson as well as analyses of the data at the completion of the lesson. The computer also generates the sounds to be learned; it constructs any of a very large number of sounds with an average access time of 0.2 sec. Variables under study include the procedure for selecting the sound to be presented on a given trial, the reinforcement probability, the kind of feedback, the nature of the sounds, and the mode of response. (The research reported here is supported by the U. S. Navy Training Devices Center, Port Washington, New York, under contract for a Human Engineering project.)