Accounting for nonmonotonic precursor duration effects with gain reduction in the temporal window model

Abstract

The mechanisms of forward masking are not clearly understood. The temporal window model (TWM) proposes that masking occurs via a neural mechanism that integrates within a temporal window. The medial olivocochlear reflex (MOCR), a sound-evoked reflex that reduces cochlear amplifier gain, may also contribute to forward masking if the preceding sound reduces gain for the signal. Psychophysical evidence of gain reduction can be observed using a growth of masking (GOM) paradigm with an off-frequency forward masker and a precursor. The basilar membrane input/output (I/O) function is estimated from the GOM function, and the I/O function gain is reduced by the precursor. In this study, the effect of precursor duration on this gain reduction effect was examined for on- and off-frequency precursors. With on-frequency precursors, thresholds increased with increasing precursor duration, then decreased (rolled over) for longer durations. Thresholds with off-frequency precursors continued to increase with increasing precursor duration. These results are not consistent with solely neural masking, but may reflect gain reduction that selectively affects on-frequency stimuli. The TWM was modified to include history-dependent gain reduction to simulate the MOCR, called the temporal window model-gain reduction (TWM-GR). The TWM-GR predicted rollover and the differences with on- and off-frequency precursors whereas the TWM did not.

Figure 1

Growth-of-masking thresholds with the control precursor and masker (filled squares) for each subject are shown in each panel. Averaged data are shown in the bottom-right panel. The filled circle represents quiet threshold plotted at an arbitrary masker level of 45 dB SPL. In the panel for S3, the open circle shows threshold with a control precursor and a 20-ms delay. The open square shows threshold with just the masker and no control precursor. The masker level (and corresponding signal threshold) fixed in Expt 2 for each subject is indicated by an arrow. The dashed lines show the I/O function fits to the data, and the parameter estimates of these fits are shown within each panel.

Figure 2

Signal threshold as a function of precursor duration. Closed symbols are thresholds with an on-frequency precursor and fixed masker and open symbols are thresholds with an off-frequency precursor and fixed masker. Each row shows data for a different participant (indicated within the first panel of each row). Each column shows data for different precursor levels (indicated at the top of each column). For comparison, the dashed line is the threshold with the fixed masker and control precursor.

Figure 3

A schematic of the models used. The TWM is shown as the black lines. The gain reduction module is represented by the gray lines. The TWM-GR is the combination of the gray and black lines. The “Example Output” dashed box shows an example condition processed through the model. The stimuli in this example are a 150-ms, 50-dB SPL on-frequency precursor followed by a 75-dB SPL masker and signal. The solid black line shows the stimuli processed with the TWM, and the dashed gray line shows the stimuli when processed by the TWM-GR. The “Example Convolution” box shows the stimuli after convolution with the integrating window [Eq. 3].

Figure 4

TWM fits (first and second columns) and TWM-GR fits (third and fourth columns) are shown as solid lines for all data. On-frequency precursor data are plotted as filled symbols and dotted lines in the first column and again in the third column. Off-frequency precursor data are plotted as open symbols and dotted lines in the second and fourth columns. The different symbols represent precursor level (see Fig. 2). As in Fig. 2, the horizontal dashed line in each panel is threshold with the fixed masker and control precursor.

Figure 5

TWM fits (first and second columns) and TWM-GR fits (third and fourth columns) after the I/O function compression slope had been changed are shown as solid, black lines. The data from Fig. 2 are replotted and the original model fits from Fig. 4 are replotted for comparison as light gray lines. For S1 (first and third rows), compression slope in the model was changed from 0.82 to 0.2. For S5 (second and fourth rows), compression was changed from 0.29 to 0.8. The solid black lines in the first and second rows show the fits to S1 and S5 with the same parameters as in Table TABLE I., only the compression slope was changed. The solid black lines in the third and fourth rows show the refits with compression slope value changed and new parameter estimates shown in Table TABLE II..