Discrimination of partially correlated temporal sequences

Abstract

How do observers determine whether two equitone sequences have the same or different temporal patterns? On each trial of this experiment, observers were presented with two sequences of n tones (n = 8, 12, or 16; tone duration = 35 ms; frequency = 1000 Hz; level = 71 dB). The sequences were played successively and the observers had to report whether the temporal pattern of intertone gaps was the same or different in the two sequences. On half of the trials, the sequences of gaps were identical (perfectly correlated, ρ = 1), and on half of the trials the sequences were partially correlated (0 < ρ < 1). The gaps were determined by summing two random processes, one common to each sequence and one unique; this procedure allowed control of the correlation, mean, and standard deviation of the gap sequences. Observer performance (d′) increased with the magnitude of the gap variance and decreased with the correlation between the sequences. A model based on computation of the sample correlation, limited by an internal temporal variability of approximately 15 ms, predicted observer performance in a variety of conditions. [Work supported by AFOSR.]