INFORMATIONAL PROCESSING OF COMPLEX SOUND .3. INTERFERENCE

In this study, theoretical results from information theory and detection theory are applied to provide a formal analysis of the interaction of target and context uncertainty on the discrimination of brief multitone sequences. The experiments employ a sample-discrimination task in which the performance of an ideal observer is held constant while the relative variability of the target sigma(T) and context sigma(C), is varied. Listener performance in these experiments was less than ideal but increased monotonically with sigma(T)/sigma(C), whether the task was to discriminate the frequency or the intensity of the target. Performance was also constant for a constant sigma(T)/sigma(C) regardless of the target's position in the sequence. The results are consistent with a class of models in which the decision variable is a weighted sum of the tone values comprising the sequence. Weighting functions computed from the trial-by-trial data suggest that limits in performance result from trial-by-trial variability in the weights. The decision variable also serves as the basis for an analysis in which performance is linearly related to the information rate of the context independent of target position or task. In this analysis, target and context uncertainty contribute additively to the overall uncertainty associated with the decision variable.