Interaction of context and target uncertainty in the discrimination of random tone sequences

Abstract

Two experiments examined the effects of context and target uncertainty on the discrimination of brief (300‐ms) ten‐tone sequences. On each trial, the listener heard a randomly generated pair of sequences; the target was identified as the sixth tone in each pair with remaining tones as context. The frequencies of the tones were selected at random in the first experiment. Context tones were drawn from a single normal distribution with μc = 2000 Hz and σc = 50, 100, or 200 Hz. The two target tones were drawn from two normal distributions differing in means μt2>μt1 centered around 2000 Hz. The standard deviation of target distributions was σt = 50, 100, or 200 Hz. In the second experiment, the frequencies of the tones were fixed at 2000 Hz and the levels of the tones were selected at random. For context tones, μc = 65 dB SPL and σc = 1, 5, or 10 dB SPL. For target tones, μt2>μt1 = 65 dB SPL, and σt = 1, 5, or 10 dB. In both experiments, the listener's task was to select the sequence for which the target was drawn from the distribution with the greater mean, μt2. Performance of four listeners was compared to that of a theoretical ideal observer which, for all conditions, was constant. Listener performance appeared to vary monotonically with σt/σc, approaching ideal performance near σt/σc = 10. The data are consistent with a model in which listeners weigh multiple observations nonoptimally.