Some Experiments on Frequency Recognition

Abstract

Four frequencies in two configurations were presented separately to four observers in a random order. In one configuration the signals were separated by steps of 5 cy. In the other there were two pairs of frequencies; the members of each pair were separated by five cycles, but the pairs were separated by 500 cy. In each case the observers' task was to state which one of the four possible tones had been presented on any one trial. Frequency recognition error rates for the two configurations were nearly the same, although the patterns of errors were quite different. In the first configuration, there were frequent confusions between the two middle frequencies. In the second, there were few confusions between pairs but more confusions within the pairs. The experiment was repeated with ten cycle separations. Again there were two groups, four closely grouped frequencies, and two pairs of frequencies 10 cy apart with 500 cy separation between pairs. Again the percentages of errors were nearly the same, while the patterns were different. The results are considered in terms of a scanning model, a multiplex model, a threshold model, and variable band model. (This work was carried out under sponsorship of the Operational Applications Laboratory, Air Force Cambridge Research Center, and Evans Signal Laboratory, Signal Corps Engineering Laboratories, Fort Monmouth, New Jersey.