Fig. 19.1
Left panel: mean thresholds across four listeners as a function of precursor/masker notch width for the unenhanced (filled squares) and enhanced (open circles) conditions. Right panel: threshold differences (unenhanced-enhanced) vs. notch width. Error bars represent the standard errors
These results are in good agreement, qualitatively, with those of Nelson and Young (2010), which showed maximum enhancement for a notch width of 0.5 octave. The agreement between the IC data and the present data further suggests that signal enhancement as observed in humans is a low-level process and as such may reflect a basic property of hearing.
The present data also indicate a strong increase in frequency selectivity under conditions of enhancement. Although the data do not permit a compelling derivation of “auditory filter shapes” and bandwidths, they indicate, according to the assumptions underlying the notched-noise technique (see Patterson and Moore 1986), a reduced effective bandwidth under conditions of enhancement: at 0.6 octave, there is 4.5-dB reduction in the effective excitation being passed by the filter, implying an increase in frequency selectivity. Carlyon (1989) using 5-ms, 1-kHz signals also showed significant enhancement effects with noise precursors of varying notch widths. His data are highly variable but are consistent in showing increased frequency selectivity under conditions of enhancement. In contrast, Moore et al. (1987) argued that the derived auditory filter shape and bandwidth did not change significantly with delay for a 20-ms, 1-kHz signal presented at the onset and in the temporal center of a notched-noise masker. Their conditions were similar to those of the present experiment, and indeed their data show approximately 5-dB-lower thresholds for signals presented in the temporal center of the masker (likely an enhancement condition) when the notch width was approximately 0.6 octave. Since the thresholds appear to be identical at the onset and center of the masker when there is no notch, their results are consistent with the present results and thus imply a reduction in filter bandwidth according to the power summation assumptions that underlie critical band/auditory filter estimates. Why their derived filter shapes and bandwidths are similar under conditions of enhancement is unclear. However, their data, the data of Carlyon (1989), and the present data are consistent with an explanation based on adaptation of suppression/inhibition.
4 Experiment 2: Effects of Precursor Level and Precursor Type
In addition to providing additional parametric data, this experiment addresses an issue that has been raised in several contexts, namely, the role of grouping and segregation in enhancement. The argument, essentially, is that the explanation for enhancement is that the precursor and masker form one stream and that when a new component is introduced it is segregated and thus is more salient, shows a lower threshold, etc. In our opinion, this is an inadequate explanation and merely describes the phenomenon. For example, it seems inconsistent with data indicating that a temporal gap between the precursor and masker that clearly segregates the two still produces substantial enhancement. In any case, in the present experiment, two of the precursors were perceptually different from the masker and certainly were not perceived as being grouped with the masker.
In this experiment, the precursors were either (1) an inharmonic complex spectrally identical to the masker; (2) a harmonic complex with a fundamental frequency of 200 Hz, spanning 0.6 to 8 kHz; or (3) a notched noise, spanning 0.5 to 8 kHz. The notch width for all precursors was 0.6 octave centered at 2 kHz. The levels of the precursors were varied over a 60-dB range but fixed within a block of trials. The masker level was fixed at 40 dB SPL per component and the masker and probe durations were 100 ms. An additional listener was employed for this experiment.
The left panel of Fig. 19.2 shows the amount of enhancement as a function of level. For the notched-noise precursor, the levels are the spectrum levels (1-kHz equivalents) in the passbands. The differences between the functions shown in the left panel are reduced by equating for overall level, as shown in the right panel of Fig. 19.2.
Fig. 19.2
Left panel: mean enhancement across five listeners as a function of precursor component level with inharmonic (filled squares), harmonic (shaded triangles), and notched-noise (open circles) precursors. Error bars represent the standard error of the mean, and the vertical dashed line indicates the level of the inharmonic masker. Right panel: the data from the left panel replotted in terms of the overall precursor level
The general characteristics of the data are similar for all three precursors: there is a fairly broad maximum in enhancement at precursor levels close to that of the masker. At very low precursor levels, those approaching a no-precursor situation, there is little enhancement. At high precursor levels, the precursor forward masks the signal and eventually results in negative enhancement.
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
