Observers rated 500−msec bursts of wide−band Gaussian noise for presence or absence of a 100−msec, 500−Hz signal tone. The tone was present on half of the 400 trials and was centered in the noise bursts. Bandwidth and integration time estimates were found for each observer by correlating the observer ratings with the output of energy detectors of various bandwidths (20, 40, 100, 250 Hz) and integration times (50, 100, 300, 500 msec) to find the best correlating energy detector. The results, computed separately for signal (SN) and no signal (N) trials, indicate a bandwidth of 40 Hz and an integration time of 100 or 300 msec fit best on both SN and N trials. Energy variations in 25 50−Hz by 100−msec segments of the noise were correlated with the observer ratings and showed negative correlations in frequency and time intervals immediately surrounding the signal region on SN trials, but not on N trials. The results suggest observer monitor detectors sensitive to temporal and spectral changes in the energy of the noise bursts, not just the absolute level of the output of a simple filter−integrator energy detector. The N trial results are accounted for by assuming that the observer is looking for the same pattern of temporal and spectral changes as on SN trials, but is uncertain as to the exact location of the signal tone.
Subject Classification: 65.58, 65.35.