Error analyses in vibrotactile code learning

Abstract

A signal detection procedure was used to obtain data from nine groups of Ss in a vibrotactile learning experiment. Each group was required to learn, as the “signal”, one of the nine elements available in a code based on three durations and three intensities of signal. It was postulated that a simple model would account for the frequencies and distributions of most false alarms (FA's). The model states that there is a psychological similarity space (PSS) in which equal distances correspond to equal differences in duration and intensity and the probability of confusing another stimulus with the signal is a decreasing function of distance from the signal. Of the resultant FA's 76.60% were errors of adjacency, i.e. most signals were confused with adjacent stimulus elements. When the same FA's were categorized in terms of intensity and duration there were 111 intensity, 130 duration, and 50 combined intensity‐duration errors.