Search

Factors which determine the discriminability of tonal sequences, or patterns, were investigated in five experiments. The patterns were generally sequences of ten 40‐msec tonal components, which ranged in frequency from 256 to 900 Hz, or from 500 to 1500 Hz, in equi‐log intervals. Highly trained (15 to 60 h of training prior to collecting experimental data) listeners’ abilities to detect changes in the frequency of single tonal components in these patterns were measured using a same–different psychophysical method. The just‐detectable values of Δf (d′=1.0) were only slightly larger than for single 40‐msec tones presented in isolation for tonal components at the end of the temporal sequences (a ’’recency’’ effect). Performance was systematically worse for earlier components, the just‐detectable Δf increasing by four to five times from the end of the pattern to the beginning (no ’’primacy’’ effect). The ’’recency’’ effect was interpreted as a matter of later arriving components interfering with frequency res...

This is the second in a series of articles in human listeners’ abilities to discriminate between word‐length tonal sequences, or ’’patterns.’’ The first article reported that frequency resolution, by highly trained listeners, is four to five times more accurate for high‐frequency, late‐occurring components of such sequences than for low−frequency early components [Watson, Kelly, and Benbasset, J. Acoust. Soc. Am. 57, 1175–1185 (l975)]. These effects, which are similar to described as ’’recognition masking’’ or ’’informational masking’’ by other authors, have now been shown to be strongly dependent on the degree of trial‐to‐trial stimulus uncertainty of the psychophysical procedure in which they are measured. When stimulus uncertainty is reduced to its psychophysical minimum, frequency resolution for any component of a tonal sequence is only slightly less accurate than for isolated tones. Previous reports of recognition masking this may reflect limitations imposed by those more dynamic parts of the sensory...

In a recent paper we described the decay of auditory memory for patterns consisting of ten 40‐msec tones played in sequence. On each trial, one of 50 permutations of a ten‐tone catalog was played, and then it was repeated either with or without a frequency increment (Δf) added to one of the ten tonal components. The interval between the two patterns was varied from 0.5 to 32 sec. The frequency increments became only slightly less detectible as the interstimulus interval was lengthened (just‐detectible values of Δf increased by about 1/3). In a new experiment, the same highly trained listeners were able to make almost as precise discriminations without a standard present on each trial. When the patterns consisted of ten‐tone sequences drawn from a 1000‐tone catalog the listeners could no longer perform at better‐than‐chance levels in the remembered‐standard conditions. The results begin to establish the limits of listeners' ability to learn the parameters of a catalog from which acoustic stimuli are sampled. [Supported by Grant No. NS03856, National Institute of Neurological Diseases and Stroke, Department of Health, Education and Welfare.]

In two recent experiments we described the discriminability of acoustic patterns consisting of ten 40‐msec tonal components played in sequence, as measured with a same‐different psychophysical method. Listeners can readily discriminate two such patterns when the only difference is in the frequency of one of the ten tonal components, although values of the just‐detectible Δf increase by as much as an order of magnitude for early components as compared to later ones and for lower as compared to higher frequency components. In a new procedure listeners attempted to match the patterns by controlling the frequency of a single tonal component. They were able to routinely adjust that frequency within ± 10 Hz of its value in a standard pattern (which was presented in alternating sequence with the pattern in which they controlled the frequency of one component), although ± 60‐ to 100‐Hz values of Δf had been required for the same patterns to be discriminated in the same‐different procedure. Possible interpretations of this result are that the adjustment procedure enables extremely precise focusing of auditory attention, and that the site of intrapattern interference is more centrally located than previously suspected. [Supported by Grant No. 03856, National Institute of Neurological Diseases and Stroke, Department of Health, Education and Welfare.]

Watson and Nichols [J. Acoust. Soc. Amer. 39, 1247, 1966(a)] described an experiment in which 10 tonal signals and 10 instances of noise alone were distributed randomly over the course of 5‐min listening periods. Although listeners responded at low rates, distributions of response latencies were estimated for signals plus noise and for noise alone, and from these distributions values of the detectability index, d′, were calculated. We now find that listeners can be trained to respond very frequently in this task when they are allowed to use low‐confidence ratings for faint signal‐like events in the noise background. Relations between ratings and the probability that a response is correct, relations between confidence level and response latency, and also the slopes of rating ROC curves (plotted on normal‐normal coordinates) are all similar to those from experiments with defined observation intervals.