comodulation masking release in a forward-masking paradigm

Comodulation masking release in a forward-masking paradigm Dennis McFadden and Beverly A. Wright Department of Psychology and Institute for Neurological Sciences Research, University of Texas, Austin, Texas 78712 •

(Received 26 May 1987; accepted for publication 20 July 1987)

Waveforms that yield comodulation masking release (CMR) when they are presented simultaneously with a signal were used in a standard forward-masking procedure. The signal was a 25-ms sample of a 2500-Hz tone. The masker was a band of noise centered at 2500 Hz, 100 Hz in width, and 200 ms in duration. Presented with the masker were two or four cue bands, each 100 Hz wide and centered at various distances from the masker band. These cue bands either all had the same temporal envelope as the masker band (correlated condition) or their common envelope was different from that of the masker band (uncorrelated condition). In the initial experiments, ( 1 ) detectability of the tonal signal was 7-18 dB better when the masker band was accompanied by cue bands than when it was not--an effect that would be expected from past research on lateral suppression--but further, (2) the signal was about 3 dB more detectable in the correlated conditions than in the uncorrelated conditions. In follow-up experiments, these CMR-like differences between the correlated and uncorrelated conditions were substantially reduced (although not eliminated) by presenting a contralateral, wideband noise that was gated synchronously with the masker and/or cue bands. The implications are that the initial results were attributable in part to the "confusion effects" known to exist in certain temporal-masking situations, and that listeners are able to obtain greater information about the temporal extent of a masker band from correlated cue bands than from uncorrelated bands. It is suggested that the small CMR-like difference that did remain in the face of controls for confusion may be attributable to a form of dynamic lateral suppression, where the magnitude of suppression covaries perfectly with the magnitude of the envelope of the narrow- band masker. It appears, then, that no explanations beyond confusion and lateral suppression are necessary to account for the reported CMR-like effects in forward masking.

PACS numbers: 43.66.Mk, 43.66.Dc [NFV]

INTRODUCTION

Hall and his colleagues (Hall et al., 1984; Hall, 1986) have studied the simultaneous masking of a tonal signal in the presence of two noise bands--one centered on the signal and called the masker band, and one located at a variable distance from that masker band and called the cue band.

When the cue band was given the same temporal envelope as the masker band, signal detectability was anywhere from 6- 10 dB better than when the temporal envelope of the cue band was different from that of the masker band. This differ-

ence in detectability was called comodulation masking release or CMR. In a previous article (McFadden, 1986), we reported that a CMR-like difference can also be obtained using a forward-masking procedure. Were this outcome sub- stantiated, it would be difficult to accommodate within cur- rent views about the basis for the CMR. The original study was motivated in part by the knowledge that masking-level differences can be obtained using forward masking (e.g., Small et al., 1972), also seemingly against otherwise-accepted theories. Here we present work that extends those original findings about the CMR and forward masking.

As in the past (McFadden, 1987), we shall try to avoid imprecision by using the term "masking band" only to refer to that noise band centered at the signal frequency in the CMR conditions, and by using the term "cue bands" to refer

to all additional noise bands presented, even though in some instances these bands may have themselves contributed somewhat to the masking of the signal.

I. GENERAL METHODS

The subjects were nine college students aged 18-24. One crew of seven was used for the initial measurements; in the follow-up experiments, two new subjects replaced two who had left. All subjects had hearing better than 15 dB HL at all of the standard audiometric test frequencies, as determined with a screening audiometer (Rudmose ARJ 4A).

The experimental procedure was adaptive, two-interval forced choice. During the first 200 ms of both observation intervals, a masker band and either two or four cue bands were gated on and off simultaneously; in one of the two intervals•etermined at random--a 25-ms signal was presented soon after the termination of the masker and cue bands. The

silent, interstimulus interval (ISI) elapsing between masker/cue offset and signal onset was 0, 15, or 30 ms, and was varied only across blocks of trials. The signal was a 2500-Hz tone obtained from an oscillator (General Radio 1310A) and gated without respect to phase. The trial-timing sequence was: warning light (400 ms), pause ( 350 ms ), first observation interval and light (225-255 ms, depending upon the ISI), pause (nominally 500 ms), second observation in-

1615 J. Acoust. Soc. Am. 82 (5), November 1987 0001-4966/87/111615-06500.80 ¸ 1987 Acoustical Society of America 1615

Redistribution subject to ASA license or copyright; see http://acousticalsociety.org/content/terms. Download to IP: 130.88.90.110 On: Sat, 20 Dec 2014 11:54:29

terval and light ( 225-255 ms ), response interval (nominally 1500 ms), feedback interval and light (350 ms). (The durations of two time intervals are characterized as "nominal"

because the computer disk was interrogated for waveform files at these times, with the result that the durations could be as much as 150 ms longer than the nominal value.) The levels of the masker and cue bands were fixed within a block

of trials, and a computer (DEC PDP 11/73) monitored the responses of each subject separately and adjusted the signal level using the two-down/one-up rule of Levitt (1971), thereby estimating the 71% correct point on the psychomet- ric function. The signal was adjusted in 4-dB steps until two reversals had occurred and in 2-dB steps thereafter. Trials were run in blocks of 50. The estimate of performance obtained on a block of trials was retained only if there were at least 47 responses, if the number of usable reversals was six or greater, and if the standard deviation of those reversals was less than 6 dB. All waveforms were gated through Wil- sonics electronic switches (model BSIT) set for a cosine- squared rise-decay time of 10 ms. TDH-39 headphones in circumaural cushions were used for all listening, which was monotic except where otherwise noted.

All noise bands were computer synthesized ahead of time, stored as files on the computer disk, and delivered to 16-bit digital-to-analog converters during the relevant time intervals in the trial sequence--at an output rate of 20 kHz. Synthesis of these waveforms was accomplished by summing sine waves 1 Hz apart over the frequency range of interest. For each sine wave, an amplitude was randomly sampled from a Rayleigh distribution, and a starting phase angle was randomly sampled from a rectangular distribution. For the experiments reported here, all noise bands were nominally 100 Hz wide, and essentially rectangular, but to maintain consistency with the waveforms used in some previous experiments in this lab, all of our noise bands had a 10-Hz notch surrounding their center frequencies.

Noise bands with different center frequencies but the same temporal envelopes were created by reusing the same sequence of values for amplitude and phase for the succes- sion of 1-Hz steps. (In practice this was accomplished by giving a random-number algorithm the same seed.) Four- teen 250-ms samples, each with an index number (0-13) identifying the set of phase and amplitude values used to generate it, were synthesized for each center frequency of interest. With this procedure, each sample in each series of 14 had a "mate" in every other series, each having the same temporal envelope. Sets of 14 waveforms were synthesized at a number of different center frequencies ranging from 1500- 3500 Hz. The series of 14 waveform samples centered at 2500 Hz was used as constructed as the source for the mask-

er bands. The cue-band waveforms, however, consisted of either two or four noise bands centered at different frequencies; thus their construction required an additional summing step. New series of 14 files were created by combining (of[- line), on a point-by-point basis, two or four series of 14 files having different center frequencies but the same temporal envelopes (the same index numbers). [This was the same procedure used to produce the two- and four-band cue waveforms used in McFadden (1987) and Wright and McFad-

den (1986). ] That is, in the present experiments each waveform used as a cue always consisted of either two or four nonoverlapping noise bands, each having a different center frequency but the same temporal envelope. Specifically, two- band cues were constructed using waveforms having the fol- lowing pairs of center frequencies: 2300 and 2700 Hz, 2000 and 3000 Hz, and 1500 and 3500 Hz. Four-band cues were constructed using waveforms having the center frequencies: 2100, 2300, 2700, and 2900 Hz, and using 1500, 2000, 3000, and 3500 Hz. Which groups of these two- and four-band cues were used for each of the experiments below are made explicit in the relevant results sections and figure captions.

For each observation interval in those blocks of trials

dedicated to the correlated CMR conditions, the computer first chose a masker band at random from the set of 14 avail-

able samples centered at 2500 Hz and then took its "mate" from the other set of 14 for use as the cue waveform. In the

uncorrelated CMR conditions, the computer selected the cue waveform at random from the other 13 samples available to it. Both observation intervals of a single trial contained the same masker or cue bands only by chance, and of course, the decision about which observation interval was to contain the

signal was independent of the decision about which particu- lar waveforms to present.

The overall level of the masker band was 70 dB SPL

throughout. Within each of the 14 files for the multiple cue bands, the overall level of one band (chosen pseudorandom- ly) was approximately 70 dB, and the levels of the other bands were intentionally varied in a pseudorandom manner in 2-dB steps over a range of 8 dB. The initial intent of this manipulation was to deprive the auditory system of a known spectral profile (see Green, 1983). We have now shown, however, that performance in both the CMR and the CDD (comodulation detection difference) tasks is essentially the same with equal and "scrambled" cue levels (McFadden, 1987)--at least in simultaneous masking conditions.

To place our noise bands into perspective, if a subject were listening to our waveforms using only a "critical-band" filter of the sort detailed by Glasberg et al. (1984) and centered at 2500 Hz, our 1500- and 3500-Hz cue bands would have been attenuated by about 50 dB, our 2000- and 3000-Hz cue bands would have been attenuated by about 20 dB, and our 2300- and 2700-Hz cue bands would have been attenuat-

ed by about 5 dB by this filter.

II. RESULTS: CMR FORWARD MASKING

Figure 1 shows seven-subject means for the CMR conditions having cue bands centered at 2300 and 2700 Hz, and for those having cue bands centered at 2100, 2300, 2700, and 2900 Hz. For reference purposes, detectability is also shown for the 25-ms signal in quiet (no masker or cue bands ) and in the presence of the masker band only. Let us begin by consid- ering just the data obtained with cue bands at 2300 and 2700 Hz. Several points are worth noting. First, the masker band alone--at an ISI of 0 ms--reduced the detectability of the tonal signal by about 45 dB relative to the no-masker condition. Second, the presence of cue bands of either sort, uncorrelated or correlated, improved detectability by 7-10 dB rel-

1616 J. Acoust. Soc. Am., Vol. 82, No. 5, November 1987 D. McFadden and B. A. Wright: Comodulation masking release 1616


ative to the masker-only condition. Such an outcome was predictable from past research on lateral suppression (e.g., Houtgast, 1972; Shannon, 1976). [A corresponding im- provement was not seen in McFadden (1986), presumably because of the close proximity of the single cue band to the masker band. ] Third, when the two cue bands had the same temporal envelope as the masker band, detectability was better by about 3 dB than when the temporal envelopes were different. This advantage for the correlated condition over the uncorrelated condition held over the range of ISis studied, and was present in the individual data of every subject save one. When the data were examined by analysis of vari- ancestwo conditions of listening X three values of ISI, within subjects--the advantage for the correlated condition was statistically significant, F(1,6) = 9.1, p < 0.024, as was the main effect for ISI, F(2,12) = 72.3, p < 0.0001. The interaction was not statistically significant.

Now let us examine the data in Fig. 1 for the CMR conditions in which there were four cue bands present. Again, each point is a mean across the same seven subjects; again, the addition of either uncorrelated or correlated cue bands to the masker interval improved detectabilityinhere by a. bout 14-18 dB; and again, not surprisingly, the main effect for ISI was statistically significant, F(2,12) = 124.2, p < 0.0001. Most importantly, there was again an advantage for the correlated over the uncorrelated condition of about 3

dB across the range of ISis studied, which is about the same value obtained above and earlier (McFadden, 1986). The direction of this difference was the same in the individual

data of every subject, and under a within-subjects analysis of

variance of the same type described above, the main effect for uncorrelated/correlated condition was statistically significant-F(1,6) = 17.7, p < 0.006. Within subjects, these CMR-like differences ranged from about 1.5-8 dB averaged across the three values of ISI. By comparison, CMRs of about 5-7 dB were obtained with these four cue bands in a

simultaneous masking task (McFadden, 1987 ). The interaction was not statistically significant. When least-squares fits were made to the data shown in Fig. 1, the coefficients of determination (t a) ranged from 0.939-0.999.

III. COMMENT

The above results appear to confirm the earlier, preliminary finding of a small CMR-like difference obtained with a forward-masking paradigm (McFadden, 1986). However, an important control condition is absent in the experiments reported above, as well as in the initial report (McFadden, 1986), and this absence makes interpretation of the results difficult.

There is a temptation to appeal to lateral suppression (e.g., Houtgast, 1972; Shannon, 1976) to explain the large improvements in detectability seen in Fig. 1 when cue bands were presented along with the masker band. However, Moore (see Moore and O'Loughlin, 1986, for example) and Neff (1986), among others, have demonstrated that such improvements can be produced by effects other than suppression. Specifically, these authors have argued that "confusion" about the temporal extents of the masker and signal can produce performance that is spuriously bad in the mask-

55

5O

•45

n- 40 O O

• 35

• 30

LU 25

Z 20

15

10

[]•

CUE BANDS(kHz) UNCORR. CORR. • 2.3 & 2.7 A ß

2.1, 2.3, 2.7 & 2.9 V ß

MASKER ONLY []

X

NO MASKER 0 I

15

INTERSTIMULUS INTERVAL (ms)

FIG. 1. Forward-masking data averaged across seven subjects. The no- masker condition shows the detect-

ability of the 25-ms signal in the quiet. In the masker-only condition, only the 100-Hz noise band centered at 2500 Hz was presented during the (approximately 200-ms) masker intervals. In the other conditions, the masker band was accompanied by 100-Hz cue bands centered at 2300

and 2700 Hz, or at 2100, 2300, 2700, and 2900 Hz. Each point is based upon 5-15 blocks of 50 trials for each subject.



er-only condition and which improves when an additional sound is presented along with the masker. The implication is that the additional sound reduces the listeners' uncertainty about when to listen for the signal by helping to better mark the temporal offset of the masker. In general, confusion effects are of greatest concern when the masker is a narrow- band noise, when the ISI is short or nonexistent, and when the signal and masker are similar in level. A commonly used control for establishing the presence of confusion effects is a contralateral waveform that is gated synchronously with the masker (e.g., Neff, 1986), and wideband noise has been suc- cessfully used for this purpose. Accordingly, a replication of the preceding experiments was performed using a wideband noise (WBN) contralaterally in an attempt to reveal the presence of confusion effects. Of greatest interest, of course, was not the effect on the masker-only condition, but whether the contralateral stimulus would alter the difference between

correlated and uncorrelated conditions observed in the previous experiments.

IV. CONTRALATERAL CUE EXPERIMENT

Five of the previous crew members served in these follow-up experiments, along with two new subjects. All of the basic features of these experiments were the same as in the previous ones. A major difference was that we took the op- portunity to study additional configurations of the cue bands as well as the possible effects of "confusion." Specifically, series of 14 two-band cues were synthesized using bands centered at 2000 and 3000 Hz and at 1500 and 3500 Hz, and a series of 14 four-band cues was synthesized using bands centered at 1500, 2000, 3000, and 3500 Hz. Again, the individual bands were each 100 Hz wide with a 10-Hz notch at the

cenl•er, and, as before, the temporal envelopes of all waveforms in a sample were always the same. In one-half of the test conditions, a wideband noise (WBN; 175-5175 Hz at the 3-dB down points and 135-6200 Hz at the 12-dB down points) was gated simultaneously ( 10-ms rise-decay time) with the masker/cue bands during both observation intervals of a trial, and was presented to the contralateral ear. This noise was obtained from a noise generator and was passed through a passive filter prior to delivery to the ear- phones. The overall level of this contralateral cue was about 50 dB SPL. All subjects received approximately eight hours of practice with the contralateral WBN cue prior to data collection. Owing to time constraints, only two values of ISI were tested for some of the contralateral-cue conditions.

V. RESULTS: CONTRALATERAL CUE EXPERIMENT

In Fig. 2 are shown data obtained with and without the contralateral WBN cue. For the conditions shown, there were cue bands present at 2000 and 3000 Hz, and, as before, the masker-only data are also shown for reference (the no- masker data are shown in Fig. 3). As can be seen, the presence of a contralateral WBN cue did improve performance in the masker-only condition by about 7 and 2 dB at ISis of 0 and 30 ms, respectively. This outcome is in accord with past reports on the effects of contralateral cues in temporal- masking conditions (e.g., Moore and O'Loughlin, 1986; Neff, 1986). When a within-subjects analysis of variance was

• 50

• 45

n- 40 O

,,>, 35

z

• 30

CUE BANDS: 2 0 & 3.0 kHz

No CC CC

[3 ß MASKER ONLY

ß ß CORR.CUE BANDS

i /• V UNCORR. CUE BANDS ß ß

i

0 15 30


FIG. 2.' Similar to Fig. 1 except that there were two 100-Hz cue bands, centered at 2000 and 3000 Hz. In addition, a contralateral, wideband noise was gated synchronously with the masker band in some conditions (designated CC for contralateral cue). Five of the subjects were the same as in Fig. 1, and two were new. The average no-masker data for this crew of listeners are shown in Fig. 3. In some instances, symbols have been displaced lateral- ly for clarity of presentation. Each point is based upon 5-10 blocks of 50 trials for each subject.

performed on the two factors of ISI (three levels) and presence or absence of the contralateral WBN cue for the mask-

er-only data, the main effect for the latter factor was statistically significant, F(1,6) -- 7.0, p < 0.038, as was the main effect for ISI, F(2,12) = 54.5,p < 0.0001, but the interaction was not.

Of greatest interest is the fact that the presence of the contralateral WBN cue did reduce the differences in detect-

ability between the correlated and uncorrelated conditions; specifically, the 3- to 4-dB advantage of the correlated over the uncorrelated conditions seen in Fig. 2 without the contralateral WBN cue was reduced to about 1.8 dB with the

contralateral cue. A three-factor, within-subjects analysis of variance was conducted--uncorrelated/correlated condi-

tions X presence or absence of contralateral cue X two values of ISI (0 and 30 ms ). The main effect for uncorrelated versus

correlated cue bands was statistically significant, F(1,6)- 35.0, p<0.001, implying that the contralateral WBN cue did not completely abolish the CMR. In this analysis, the main effect for contralateral WBN cue was not statistically significant, F(1,6) = 0.03, p > 0.87, but, of course, the main effect for ISI was, F(1,6)= 266.1, p•0.0001. None of the interactions was statistically significant.

In Fig. 3 are the data for the conditions in which the center frequencies of the comodulation cue bands were 1500, 2000, 3000, and 3500 Hz. The basic outcomes in Fig. 3 are similar to those in Fig. 2 in that the contralateral WBN cue



55

5O

• 45

n- 40 o

• 35

o 30

uJ 25

z (D 20

15

10

CUE BANDS: 1.5,2.0,3.0,3.5 kHz

No CC CC

[] ß MASKER ONLY

ß ß CORR.CUE BANDS

/X •' UNCORR. CUE BANDS

[]

ß

i i I i

NO MASKER 0 15 30


FIG. 3. Similar to Fig. 2 except that there were four 100-Hz cue bands, centered at 1500, 2000, 3000, and 3500 Hz. Again, a contralateral, wideband noise was gated synchronously with the masker band in some conditions (designated CC). In the no-masker condition, the 25-ms signal was presented alone in the quiet. The seven subjects were the same as in Fig. 2. Each point is based upon 5- 10 blocks of 50 trials for each subject.

again reduced the difference between the correlated and uncorrelated conditionsinhere from about 2.0 dB to about 0.5

dB. (Unlike Fig. 2, here the difference was reduced as the contralateral WBN cue made detectability worse for both the correlated and uncorrelated conditions than when no

contralateral cue was present.) When an analysis of variance like that described in the previous paragraph was conducted on these data, the main effect for ISI was again significant, F(2,12) = 7.8, œ < 0.007, and the main effect for uncorrelated versus correlated cue bands was marginally significant, F(1,6) = 6.3, p < 0.046. The main effect for the presence or absence of the contralateral WBN cue missed statistical sig- nificance, F(1,6) = 4.3, p > 0.084, and none of the interactions was statistically significant. When least-squares fits were made to the data shown in Figs. 2 and 3, the coefficients of determination (r 2) ranged from 0.944-0.998.

Data were also collected with pairs of comodulation cue bands centered at 1500 and 3500 Hz, but they are not shown here since there was no CMR evident with these distantly located cue bands, whether or not the contralateral WBN cue was present.

Vl. DISCUSSION

These experiments were designed as follow-ups to some preliminary observations presented in McFadden (1986), where temporal masking appeared to be about 3 dB greater when the masker band was accompanied by an uncorrelated cue band than when accompanied by a correlated cue band. That CMR-like effect was confirmed here using a number of different combinations of cue bands. However, the differ-

ence between correlated and uncorrelated conditions was

greatly reduced or abolished when a contralateral, wideband noise was gated along with the masker and cue bands. The implication is that the the CMR-like differences we have observed in temporal masking (Fig. 1 above; McFadden, 1986) can be explained without reference to mechanisms or factors beyond those already well known to students of temporal maskingmspecifically, "confusion" and lateral suppression. The argument is this: The large reduction, or aboli- tion, of the CMR-like difference produced by introducing a contralateral WBN cue strongly implies that uncertainty about the temporal extent of the masker band was a significant factor in producing that difference. A further implication is that correlated cue bands provide listeners with greater information about the temporal extent of the masker than do uncorrelated cue bandsman intuitive conclusionm

and, thus, that the CMR-like difference obtained when no contralateral cue was present was primarily attributable to the greater informational content of the correlated cue bands.

In addition to the possibly greater informational content in the correlated-cue conditions, a second mechanism may have been operating in these experiments. Namely, the small, CMR-like difference that sometimes remained even in the face of the contralateral WBN cue may be attributable to the operation of a kind of "dynamic lateral suppression." Note that when the cue bands are correlated with the masker

band, the envelope fluctuations in the latter are matched perfectlyrain amplitude and timeruby corresponding fluctuations in the cue bands. This means that the amount of

lateral suppression being delivered to the masker band is



covarying perfectly with its own fluctuations in instanta- neous amplitude, and thus with its fluctuations in effective- ness as a masker. It is possible that this perfect covariation in masker and suppressor(s) aids the listener slightly relative to the situation in the uncorrelated condition, and that it results in the small CMR-like difference in det6ctability observed in some situations even in the presence of the contralateral WBN cue. It should be noted that, if it exists, this dynamic lateral suppression also ought to be operative in simultaneous masking situations involving noise bands having the same temporal envelopes.

The upshot of these measurements, then, is that reports of substantial CMR-like effects in temporal-masking para- digms (McFadden, 1986) deserve to be greeted with great skepticism. No new phenomena or mechanisms appear necessary to explain such findings.

ACKNOWLEDGMENTS

This work was supported by research Grant NS 15895 awarded by the National Institute of Neurological and Com- municative Disorders and Stroke. We thank E.G. Pasanen

for technical assistance and for helpful comments on a draft of this article, and H. S. Plattsmier for aid with data reduction.

Glasberg, B. R., Moore, B.C. J., Patterson, R. D., and Nimmo-Smith, I.

(1984). "Dynamic range and asymmetry of the auditory filter," J. Acoust. Soc. Am. 76, 419-427.

Green, D. M. (1983). "Profile analysis: A different view of auditory intensi- ty discrimination," Am. Psychol. 38, 133-142.

Hall, J. W. (1986). "The effect of across-frequency differences in masking level on spectro-temporal pattern analysis," J. Acoust. Soc. Am. 79, 781- 787.

Hall, J. W., Haggard, M.P., and Fernandes, M. A. (1984). "Detection in noise by spectro-temporal pattern analysis," J. Acoust. Soc. Am. 76, 50- 56.

Houtgast, T. (1972). "Psychophysical evidence for lateral inhibition in hearing," J. Acoust. Soc. Am,/51, 1885-1894.

Levitt, H. (1971). "Transformed up-down methods in psychoacoustics," J. Acoust. Soc. Am. 49, 467-477.

McFadden, D. (1986). "Comodulation masking release: Effects of varying the level, duration, and time delay of the cue band," J. Acoust. Soc. Am. 80, 1658-1667.

McFadden, D. (1987). "Comodulation detection differences using noise- band signals," J. Acoust. Soc. Am. 81, 1519-1527.

Moore, B.C. J., and O'Loughlin, B. J. (1986). "The use ofnonsimultaneous masking to measure frequency selectivity and suppression," in Frequency Selectivity in Hearing, edited by B.C. J. Moore (Academic, New York), pp. 179-250.

Neff, D. L. (1986). "Confusion effects with sinusoidal and narrow-band noise forward maskers," J. Acoust. Soc. Am. 79, 1519-1528.

Shannon, R. V. (1976). "Two-tone unmasking and suppression in a forward-masking situation," J. Acoust. Soc. Am./59, 1460-1470.

Small, A.M., Boggess, J., Klich, R., Kuehn, D., Thelin, J., and Wiley, T. (1972). "MLDs in forward and backward masking," J. Acoust. Soc. Am./51, 1365-1367.

Wright, B. A., and McFadden, D. (1986). "Comodulation detection differences and comodulation masking release with multiple cue bands," J. Acoust. Soc. Am. Suppl. 1 80, S60.



comodulation masking release in a forward-masking paradigm

Documents