Fantini D A, Moore B C
MRC Institute of Hearing Research, University Park, Nottingham, England.
J Acoust Soc Am. 1994 Apr;95(4):2180-91. doi: 10.1121/1.408678.
Experiment 1 examined the ability to compare relative level across frequency (profile analysis) for stimuli that were dynamically varying over time. The task was to detect an increment in level of a narrow band of noise (the target) in the presence or absence (reference condition) of four flanking bands (FBs). The envelopes of the FBs were either the same as that of the target (correlated condition), independent of that of the target but the same as each other (co-uncorrelated condition), or all independent (all-uncorrelated condition). The overall level of the stimuli was either fixed or randomly varied from one stimulus to the next. The results showed that subjects can make effective use of spectral-shape cues even for stimuli whose amplitudes vary markedly over time. In the correlated condition, the threshold for detecting an increment in the level of the target band was decreased (relative to the reference condition) both when the overall level was fixed and when it was varied randomly from stimulus to stimulus. In the uncorrelated conditions, the FBs did not lead to better performance when the overall level was fixed; rather they produced a small interference effect. When the overall level was randomized, the presence of uncorrelated FBs produced thresholds that decreased with increasing bandwidth and, for a bandwidth of 64 Hz, produced an improvement in performance (relative to the reference condition) that was almost as large as that produced by the correlated FBs. It seems that the more rapid fluctuations in the wider bands of noise were smoothed by the auditory system, enabling information about the long-term spectral shape to be extracted effectively. Experiment 2 used similar stimuli, but the task was to detect the target, rather than to discriminate its level. Detection thresholds in the presence of FBs were lowest in the co-uncorrelated condition, higher in the correlated condition, and highest in the all-uncorrelated condition. Thus the presence of correlated FBs improved discrimination thresholds in the profile analysis task, but impaired performance in the detection task. Reasons for the discrepancy between the effects of correlated FBs in the two tasks are discussed in the context of the cues available to the listener.
