THE JOURNAL OF THE ACOUSTICAL SOCIETY OF AMERICA VOLUME 31. NUMBER 8 AUGUST. 1959
Masking Patterns of Tones
By RICHARD H. EHMER
U. S. Naval Medical Research Laboratory, U. S. Naval Submarine Base, Near London, Connecticut (Received October 9, 1958)
Monaural masking patterns were obtained for pure tones spaced by octaves from 250 cps to 8000 cps at 20-100 db SL on three listeners in an attempt to provide extensive data assessing the relative importance of aural harmonies and cochlear spread of masking tone activity in the extension of masking to frequencies above the masking tone. The masking patterns confirm and extend the results of others, but especially indicate that second peaks do not necessarily occur at the second harmonic of the masking tone. The masking patterns are explained in terms 9f (1) the activity pattern of the masking tone in the cochlea, (2) beats between signal and masking tones, (3) aural harmonics, and (4) suppression of cochlear response to the signal.
THE masking pattern of a tone has been wed, for example, by Wegel and Lane and by Fletcher, to infer the pattern of activity set up by the tone in the cochlea. The masking pattern is not a measure of the cochlear activity but only an indicator because, aside from the fact that psychophysical data are used to infer physiological activity, the ear distorts loud tones, introducing harmonics not present in the stimulus. In addition, the masking tone interacts with the signal tones used to measure its masking pattern, producing beats and difference tones.
Inferences of cochlear activity made from masking patterns must agree with inferences made from other psychophysical data, and especially with more direct observations of the physiological processes inferred. Thus, the common assumption that aural harmonics are entirely responsible for the unsymmetrical extension of masking to high frequencies is brought into question by the demonstration by Egan and Klumpp of the masking of aural harmonics and by physiological evidence of asymmetry in the response of the auditory mechanism to tones as shown by the results of Bekesy on the mechanical movement of the cochlea, of Tasaki, Davis, and Legouix in cochlear microphonics, and of Tasaki6 in the electrical responses of single auditory nerve fibers.
Although the results of Wegel and Lane, which have served as the basis for most discussions of masking, are very extensive, they are not sufficiently detailed to evaluate an alternative mechanism for extended masking. The present experiment attempts to provide sufficiently detailed masking patterns for a representative sample of frequencies and intensities to choose between aural harmonics and cochlear spread as the mechanisms of extended masking, and, in general, to permit a finer analysis of the mechanisms underlying masking.
Monaural masking patterns were obtained for pure tones spaced by octaves from 250 cps through 8000 cps at 20--100 db sensation level (SL) on three listeners. Two of the listeners had normal hearing, and that of the third was normal except for a 15-db dip at 1800 cps.
The signal tones were generated by a B&6sy audiometer,' passed through an electronic switch which turned them on for 200 msec and off for 200 msec without audible transients, and then delivered to a PDR-8 earphone mounted in an lIX-41/AR cushion. The Bekesy audiometer also recorded the frequency and intensity of the signal tones during the experiment. For masking, a constant tone from another oscillator was mixed electrically with the signal tones.
In determining the masking patterns, first the absolute threshold was recorded starting at the low frequency end of the range. Next the listener's threshold for the masking tone was found. Then a series of masked thresholds was obtained with the masking tone set at successively higher levels.
Masking patterns were calculated by drawing smooth curves through the midpoints of the intensity variations on the records and measuring the differences between absolute and masked thresholds at 15 points per octave. The extent of the intensity variations was minimized, and therefore the precision in determining the threshold maximized, by adjusting rates of frequency and attenuation change. Frequency rates of either 2 or 4 min per octave combined with attenuation rates of 1 or 2 db per sec were used except at 100 db SL where 1 min per octave and 3 db per sec were used to minimize hearing loss from exposure to the masking tone.
The listener's criterion of threshold also affects the variability of the results. Variability from this source was reduced by interrupting the signal tones and by the instructions. The listeners were instructed to allow the intensity of the tones to increase until they could just detect them and then to press a button which decreased their intensity until they could no longer follow them clearly. In the masking series the listeners were told to respond to any sounds which maintained the same time pattern.
The listeners also reported the character or identity of the sounds heard during the masking series. These reports were coordinated with indications placed on the records at the time both were made.
And finally, to compare the results obtained with a Bekesy audiometer with those obtained by conventional methods, such as used by Wegel and Lanes and Egan and Hake thresholds and masked thresholds were obtained for one listener for fixed masked frequencies of 1050, 1100, 1200, 1300, 1400, 1500, 1600, 1700, 1800, 1900, 1950, 2050, and 2100 cps for a masking frequency of 1000 cps at 55, 60, 65, 70, 75, 80, and 85 db SL. The equipment was essentially the same as that described in the foregoing except for the substitution of a fixed frequency oscillator and a discrete 1-db per step attenuator for the B6kAsy audiometer. The listener first determined his absolute threshold for the tonal series by the method of adjustments and then repeated these determinations with the masking tone present at successively higher levels.
RESULTS AND DISCUSSION
The masking patterns are shown in Figs. 1-6. The extent of intensity variations in the records used for computing the masking patterns seldom exceeded 3 db except at 100 db SL where a faster attenuation rate was used. Variation among runs and among listeners was greater, but the average masking curves preserve the characteristic features of the individual records.
Comparison of the curves of Figs. 1-5 shows that 250 cps produces masking patterns somewhat different from those of the other four frequencies. At 250 cps all the curves are relatively smooth and regular and not symmetrical even at the lowest masking level.
At the other frequencies the curves are symmetrical at 20 and 40 db SL, but dearly depart from symmetry at 60 db and above. The asymmetry is brought about by the failure of masking to spread further to lower frequencies while it increases rapidly in both amount and extent at higher frequencies. At 60 db SL there is little or no masking at the harmonic frequencies, but despite this lack, the masking curves are not smooth; the extreme case is the sharp second peak in Fig. 4.
At 80 db there is a rapid increase in extended masking. At this level a second peak is found in all the curves of Figs. 2-5. It has moved closer to the second harmonic, but it does not always coincide with this frequency although this is not altogether apparent in Figs. 2 and 4. The migration of the second peak is most rapid between 60 and 80 db. Only at 100 db do the second peaks fall at the harmonic. At this level peaks may also appear at the third harmonic as in Figs. 1, 3, and 4.
The 100-db curve of 4000 cps (Fig. 5) is somewhat anomalous. It shows downward spread of masking for two octaves and a pronounced second peak at 6000 cps, a frequency which is not a harmonic of the masking tone. The same sort of downward spread occurs for a masking tone of 8000 cps (Fig. 6).
The masking patterns just described are in good agreement for corresponding masking frequencies and intensities, even to the occurrence of second peaks at nonharmonic positions, with the results of other writers using similar methods. In general the masking patterns are similar to those of Wegel and Lane,' except for the positions of the second peaks. The second peak in our results emerges dearly only for masking frequencies of 1000 cps and higher, and thus the only comparable frequency between our study and that of Wegel and Lane is 1200 cps. Their Fig. 4 shows extended masking at 60 db SL, but no second peak either at the second harmonic or at any other frequency. On the basis of our results, a second peak would be expected at about 1600 cps in their data but they made no observations at this frequency.
At 80 db Wegel and Lane do provide sufficient data to fix the second peak at the send harmonic. This is not surprising because in our data 80 db is the level at which the migration of the peak to the second harmonic is nearly complete., For a masking tone of 1000 cps one of our listeners regularly showed a broad flat-topped peak centered at 2000 cps while another showed a peak at about 1800 cps on one occasion and at 1900 cps on another.
That a possible source of difference in results is not the difference between using the Bekesy audiometer and fixed masked frequencies is shown by Fig. 7. Here fixed masked frequencies were used and the curves show clearly the progressive migration of the second peak towards the harmonic as the masking tone is raised from 55 through 85 db SL.
Fig. 1. Average masking patterns for 250 cps based upon three Fig. 4. Average masking patterns for 2000 cps based upon two listeners. The sensation level of the masking tone is attached to listeners. The sensation level of the masking tone is attached to each curve. each curve.
Fig. 2. Average masking patterns for 500 cps based upon three listeners. The sensation level of the masking tone is attached to each curve.
Fig. 3. Average masking patterns for 1000 cps based upon three listeners. The sensation level of the masking tone is attached to each curve.
Fig.4 Average masking patterns for 2000 cps based upon three listeners. The sensation level of the masking tone is attached to each curve.
Fig. 5. Average masking patterns for 4000 cps based upon three listeners. The sensation level of the masking tone is attached to each curve.
Fig. 6. Masking pattern for 8000 cps at 100 db SL for a single listener.
Fig. 7. Masking of discrete frequencies by 1000 cps for a single listener.
Auditory Characteristics of the Signal under Mashing
Throughout the frequency range of this experiment the signals when presented alone possess definite tonal quality even when they fall in the atonal interval. whenever they are subjected to an appreciable amount of masking, however, they lose their tonal quality for the listener seeking his threshold. Since the characteristics reported in this experiment differed somewhat from those reported by others, they are given in the following material because they are useful in elucidating the nature of masking. Only the predominant aspects are reported.
At masking levels up to and including 60 db SL, the signals maintain their tonal quality everywhere except in the vicinity of the masking frequency. In this vicinity two kinds of beats are heard: wavering beats and fused beats or roughness. Wavering beats are heard when the signal and the masking tone are serrated by a few cycles per second: the sound heard consists of a loudness modulation of the masking tone, the rate being equal to the frequency difference between the two tones. Fused beats are heard at greater frequency differences: the wavering becomes too fast for the fluctuations in loudness to be analyzed out and the tone sounds rough, sometimes to the point that the roughness may be the predominant aspect of the sound.
The range of wavering beats is the same above and below the frequency of the masking tone. Fused beats are heard for a broader frequency range above the masking tone than below. Above the frequency of the masking tone they are heard as the masked threshold decreases. At 20 and 90 db the transition from beats to tone pitch does hot produce a change in slope of the masking curve. But at 50 and 60 db SL the change in sound from beats to tone is associated with the shoulder or second peak of the masking curve.
At higher levels of the masking tone, the fused beats are not displaced by the tone but by beats with a different basic quality which are interpreted to be beats with the aural second harmonic. There may be a fairly broad transition zone between these two kinds of beats or there may be a quick change from the one to the other. At 70 and 80 db SL of the masking tone, the beats with the harmonic pass through the same phases as those with the masking tone itself as the signal approaches and proceeds beyond the second harmonic, until they in turn are replaced by the signal with tonal quality. Although these beats are heard in the vicinity of the second harmonic, the harmonic is not audible by itself at these levels. Only at 100 db SL does the second harmonic of the masking tone become clearly audible.
The sounds heard in the octave between the masking tone and its harmonics, except in the immediate vicinity of these two frequencies, have been regarded by previous workers'-2 as difference tones. However, in the present experiment difference tones were newer identified below 80 db SL of the masking tone and even at 80 db only when the masked threshold was near its summit close to the masking frequency. Difference tones were heard more often when the masking tone was at 100 db SL but by no means throughout the entire range of masking. Often at this level the opportunity for distinguishing beats and difference tones was afforded by their being present simultaneously. In such cases beats were identified by their wavering or roughness which is similar to those attributes heard at much lower masking levels, while the difference tones were identified by their tonal quality which is similar to low frequency pure tones. When beats and difference tones were heard together their thresholds often were the same, or differed by only a few decibels, so that the masked threshold was only slightly affected by using one or the other as the criterion for the signal.
The reason for identifying the interactions of signal and masking tones as difference tones instead of fused beats seems to have been twofold. First, since harmonic distortion was thought to be occurring, it seemed plausible to assume that intermodulation distortion was also taking place, and hence, that any sound heard in the place of the signal tone was an intermodulation product, namely, a difference tone. And second, from the published values'-''4 for the frequency difference limits for beats, it seemed that these limits were exceeded when the interaction sounds could still be heard. However, note that the published values for the limits of beats were obtained only for the case of two tones of equal intensities; it is apparent from the present findings that beats can be obtained for greater frequency differences when the two tones producing them are not of equal intensities, the most favorable condition for beats being when the second tone is higher in frequency and just above its masked threshold. In fact, under these conditions the beats can often be eliminated by raising the second tone considerably above its masked threshold. This finding offers further support for the identification of these sounds as beats: if they were difference tones, then raising the intensity of the second tone would give a stronger difference tone, not eliminate it altogether.
Mechanisms of Masking
The masking patterns must be accounted for in terms of the possible mechanisms of auditory masking. First, what determines the frequency range of masking? Clearly at low masking levels (20 and 40 db SL) the only determinant can be the cochlear activity pattern of the masking tone. Even at 60 db SL, excluding perhaps 250 cps, where there is a clear departure from symmetry and well-developed second peak in some instances, but very little masking at the second harmonic, still the activity of the masking tone itself must be considered as the sole determinant of the extent of the masking pattern. The same argument may be extended to include masking at 80 and 100 db SL. The evidence of the present study is that up to 80 db the second harmonic was not audible even when the height of the masked threshold would seem to require the second harmonic to be 70 db above its threshold (Fig. 3, 80 db). This evidence, along with the findings of Egan and Klumpp' which showed masking of the aural second harmonic by the fundamental, indicates that the activity pattern extends into and beyond this region.
If the cochlear activity pattern of the masking tone determines the extent of masking, then what determines the amount of masking? First, in the vicinity of the masking frequency, what may be called direct masking, the activity of the masking tone is responsible, but the height of the masking pattern does not truly show the amount of activity because, as Wegal and Lane' first reported and Egan and Hake" showed in fine detail, there is interaction between the signal and the masking tones. The range of direct masking extends from some point on the steep portion of the masked threshold curve below the masking tone to the dip or shoulder above the masking tone: this amounts to almost an octave in some instances, as at 80 db SL.
At the conclusion of the region of beats, when the pitch of the signal returns (up to 60 db SL of the masking tone), the masking pattern is thought to reflect accurately the level of masking-tone activity in the cochlea. Even at 80 db masking level, the masked threshold probably depends on the activity level of the masking tone, disregarding the beats between signal tone and aural harmonic as here lowering the threshold, since the second harmonic is masked and becomes audible only through these beats.
Only at 100 db SL, where the aural harmonics are audible by themselves is it thought that the harmonics add to the amount of masking, and then the factors determining masking would be the same as brought out in the discussion above but with respect, of course, to the harmonic and not the fundamental.
The foregoing proposals are compatible with what is known of the physiological activity in the cochlea and auditory nerve. Tasaki, Davis, and Legouix have shown that low tones produce cochlear microphonics throughout the cochlea but that with higher and higher tones there is shrinkage of the response area towards the base of the cochlea. When two tones are combined there is only a partial separation of their microphonics in the turns of the cochlea. The same principles of tuning as seen in the cochlear microphonics are manifest in the responses of single auditory fibers. In addition, Katsuki ed al. have furnished evidence of the nature of masking: a pure tone produces spikes in a single fiber at a characteristic rate, but when a low tone is also presented and raised in intensity, the rate becomes changed over to that characteristic of the low masking tone. It is to be emphasized that inhibition or reduction of response was never observed by either Tasakia or Katsuki el al. so that the mechanism of this masking seems only to be the overriding of the stimulus effects of the signal by the masking tone. Responses to beats were also shown to confirm the observations of Galambos and Davis' at the cochlear nucleus. When beats are slow, several spikes appear during the maximum, while there is silence or reduced response during the minimum; when the beat rate is faster, a single impulse may be evoked during each cycle of the beats.'' Thus, the pattern of impulses indicates the beat rate and not the characteristic frequency of the stimulus, such as is ordinarily the case when a single pure tone is presented.' This finding provides a physiological foundation for the widespread occurrence of beats in our listeners' reports.
The same upper frequency-difference limit may thus be assumed for beats as for the preservation of the periodicity of pure tones in the auditory nerve fiber discharge, namely, about 2000. This supposition permits an explanation of the peculiar occurrence in the 100-ib curve of Fig. 5, the peak at 6000 cps for a masking tone of 4000 cps: the frequency limits for beats between -1000 cps and the signal were exceeded causing the threshold to rise, only to be again depressed as the signal tone came within range of beats with the second harmonic (8000 cps) which is manifestly present at this level. The possibility that the peak arose from an actual 6000-cps tone produced by a subharmonic of the 4000cps masking tone interacting with either the masking tone or second harmonic to produce a summation or difference tone must be discarded because no 6000-cps tone was heard either alone or through beats.
Similarly, a subharmonic cannot be invoked to explain the downward spread of masking from 4000 cps or 8000 cps at 100 db SL since no subharmonic was heard, there were no beats heard near the subharmonic frequency, and there was no peak at this frequency. At present, it cannot be said that this downward spread of masking is a result of spread of the masking-tone activity towards the apex of the cochlea since such has not been observed. It is not remote masking, since this results from detection of the envelope of the stimulus in the cochlea'8 There is just one physiological observation of this effect, the reduction of action potentials to a 500-cps tone pip as a 6950 tone was raised from 100 to 120 db, but this also caused a drop in cochlear microphonics'8 Thus, the mechanism for this effect is different from that of the other masking demonstrated in this paper, and seems to involve interference or inhibition of some sort.
The masking pattern of a pure tone results primarily from the activity pattern of the tone in the cochlea. At low intensities, the cochlear activity is confined to a local region and the masking pattern is narrow. As the masking intensity increases, the cochlear activity spreads only towards the base, while retaining a maximum at the locus of original response, and the masking pattern extends unsymmetrically to high frequencies.
Aural distortion plays a much smaller role in masking than had previously been supposed. Aural harmonics axe not responsible for the extension of masking to the octaves above the masking tone since they are, for the most part, masked themselves by the fundamental tone activity. Only at 100 db SL do these harmonics emerge sufficiently above their masked thresholds to add masking to that resulting from the masking tone itself.
Similarly aural difference tones also have less effect on masking patterns. They do lower the masked threshold, but this again is at 100 db SL and for short ranges at 80 db SL.
For the most part what have previously been called difference tones are really fused beats, and it is these fused beats that are heard in the interval above the masking tone between the end of wavering beats and the return of the characteristic pitch of the signal. These beats depress the threshold by amplitude modulating the masking tone
The mechanism of the masking described in the foregoing results from the overlap of the masking tone and signal in the cochlea. Another kind of masking, downward spread, occurs with high frequency masking tones. This appears to result from a different mechanism, interference or inhibition.
THE JOURNAL OF THE ACOUSTICAL SOCIETY OF AMERICA
VOLUME 31. NUMBER 9 SEPTEMBER, 1959
Masking by Tones vs Noise Bands*
Richard H. EHMER
US Naval Medical Research Laboratory, U. S. Naval Submarine Base, New London, Connecticut (Received November 20, 1958)
In a previous study it was proposed that tonal masking arose mainly from the cochlear activity pattern of the masking tone, modified by the formation of beats between the signal and masking tones. The present study casts further light on these proposed mechanisms by comparing the masking effects of pure tones of 500, 1400, 2000, and 4000 cps at 60 and 80 db SL with 1/3 octave hands of noise of equal intensities and centered at the same frequencies. The results show that the noise bands produce about the same amount of extended masking despite the absence of any possible aural harmonic distortion, but greater direct masking due to the elimination of beats. Furthermore, the noise-mashing curves join the tone-masking curves at the second peak in the latter, providing strong additional support for the proposed mechanisms of auditory masking.
IN a recent study' of masking by tones it was found (1) that above the masking frequency the masking pattern became irregular and unsymmetrical with respect to the low frequency side before any appreciable masking occurred at the second harmonic of the masking tone and (2) that even at higher masking levels where appreciable masking occurred at the second harmonic a secondary peak in the masking curve fell between the masking frequency and its second harmonic.
These general findings together with others in the same study and in the work of others seemed to require a revision of the theory of masking. 'therefore, it was hypothesized that the masking pattern of a tone was determined primarily by two mechanisms, the activity pattern set up by the masking tone in the cochlea and auditory beats resulting from the superposition of the test tone upon the masking tone in the cochlea. It was thought that the activity pattern was the primary determinant of the extent (in frequency) and amount (in intensity) of masking, and that beats, where they occurred, served to lower the amount of masking. The hypotheses minimized the role of aural distortion (harmonics and difference tones), for the most part relying on the extension of the activity pattern of the masking tone toward the base of the cochlea as shown in physiological experiments" for an explanation of the rapid advance of masking upon high frequencies, and relying on the widespread occurrence of beats to explain some of the details of the obtained masking patterns.
Since direct tests of these hypotheses are not feasible, further evidence bearing upon them must come from indirect tests. Such indirect tests can be made by using a nonperiodic masking stimulus such as a band of noise and comparing its masking patterns with those produced by tones. With a noise band of the same intensity as a tone, the energy at each frequency within the band will be much lower than that of the tone, thereby producing much less, if any, harmonic distortion. Under these conditions, the classical view would expect no extended masking, and the masking pattern of the noise should be symmetrical. Of course, the situation is not that simple, being complicated by the fact that the cochlea is not a perfect frequency analyzer, but if the noise band is broad enough to include a few critical band widths its energy should be sufficiently spread out in the cochlea to minimize the production of aural harmonics due to overload.
On the other hand, the hypotheses proposed. here, attributing extended masking to spread of cochlear activity, would still predict extended masking to occur as a result of a noise band, and to about the same extent as for a tone at the same overall SPL. In addition, these hypotheses predict that there would be more masking at the frequencies in the noise band than with the single tone, but that the noise-masking curve would join the tone-masking curve at the latter's secondary peak due to the elimination of beats between the tonal signal and the masking stimulus. Of course, it is expected that the noise-masking curve would extend somewhat farther to low frequencies simply because the noise band extends to frequencies lower than the masking tone.
Certain evidence already available encourages the expectations embodied in the preceding paragraph. Egan and Hake in determining the masking pattern of a simple auditory stimulus, compared the masking patterns of a 400-cps tone and a band of noise 90 cps in width centered at 410 cps, both at 80 db over-all SPL. Their Fig. 7 portrays almost exactly what the present hypotheses predict. It was thought, however, that a broad sampling of frequencies and levels should be studied to subject these hypotheses to a more rigorous test.
The equipment and procedure were the same as used in the previous study' except for the noise bands which were produced by passing the output of a gas-tube noise generator through the one-third-octave filters of a Bruel and Kjaer Type 2109 Audio Frequency Spectrometer. Thresholds and masked thresholds were obtained by a B&6sy audiometer for two listeners with normal hearing; masking stimuli were tones at 500, 1000, 2000, and 4000 cps at 60 and 80 db sensation level (SL) and I octave bands of noise centered at the same frequencies and set to the same over-all SPLs.
RESULTS AND DISCUSSION
Typical results for one of the subjects are portrayed in Figs. 1-8. In Figs. 1-4, which give the results at 60 db SL, the tone-masking curves are unsymmetrical, but the noise-masking curves are all symmetrical and in addition coincide with or approximate the tonemasking curves at the latters' shoulder or second peak. This agreement between hypotheses and results, while gratifying, may be due more to the width of the noise band than to the correctness of the hypothesis because, the noise-masking curve is about equally far from the filter characteristic curve both above and below the noise band. And since the band of masked frequencies is so much wider than the width of the masking stimulus (even the one-cycle-wide masking tone produces a broad peak in the masking curve), the upper frequencies in the noise band could have pushed the masked threshold beyond the limits of the second descending portion of the tone masking curve in Figs. 2 and 4.
More satisfactory for evaluating the proposed hypotheses of masking are the results obtained at 80 db SL, shown in Figs. 5-8. In these figures, although the masking curve slopes off below the noise band close to the filter characteristic curve, above the band limits the extended masking curves run well above the high frequency cutoffs. Moreover, in these figures the agreement between data and hypotheses is excellent. Within the frequency limits of the noise band there is much greater direct masking than is produced by the pure tone; above the band limits the noise masking curve falls off rapidly, intersecting with the tonal curve at the second peak and thereafter coinciding with it for perhaps half an octave (in Figs. 5-7) before again diverging. The lone exception occurs in Fig. 8 where the noise curve never does quite reach the tone curve; this departure may be due to the less satisfactory dropoff of the filter high-frequency response.
Although the general agreement is excellent, two minor features require comment. The irregularities in the noise-masking curves and the fact that, where the tone and noise curves of extended masking diverge, the noise curve always runs below the tone curve.
The irregularities in the noise-masking curves, seen as the humps at 1600 cps in Fig. 6, 3000 cps in Fig. 7, and 6000 cps in Fig. 8, appear sufficiently often in about the same locations that they may not be regarded as fortuitous. They have also appeared in the data of Egan and Hake' (Fig. 2) and Bilger and Hirsh (Figs. 2 8). They do not appear at the right frequencies to be regarded as aural harmonic distortion, and they occur when the level of the energy per cycle of the noise is too low to produce harmonic distortion. Thus, the possibility that the extended masking is a result of harmonic distortion is ruled out. This leaves the possibilities (1) that the hump is indicative of the activity pattern of the noise band in the cochlea and, somewhat overlapping, or (2) that the hump is a product of the interaction of the signal tone and masking noise similar to the processes postulated for the second peaks in the tonal masking curves.
Regarding the first possibility, that the hump is a function of the form of the noise pattern in the cochlea, nothing can be said because of a lack of direct evidence. Regarding the second possibility, the interaction between tone and noise, the change in phenomenological characteristics of the signal tone at threshold when its frequency lies within or dose to the band limits of the noise can be mentioned: The signal itself becomes noiselike. Similar observations were reported by Egan and Hake" for the masking by a band of noise 90 cps in width and centered at 410 cps. These same investigators also obtained less direct masking within the noise band than for a broad band noise (0-1000 cps) at the same spectrum level. Thus, it appears that a narrow band of noise retains some tone-like properties. Perhaps the noise band maintains a somewhat periodic discharge in the auditory nerve fibers, a discharge modulated by the superposition of a tone of appropriate frequency. Such a possibility was suggested in the case of tonal masking. Again, direct evidence is lacking.
It is not easy to explain why, at frequency regions above their coincidence, the noise-masking curve runs below the tone-masking one. Were it not for the long run of coincidence, it might be thought that, since much of the energy in the noise is below the frequency of the tone, there should be less extended masking; but then it would also be expected that the entire extended masking curve of the noise band would run below the tone. The difficulty of explanation is further complicated by the return to coincidence toward the high-frequency end of masking. Nevertheless, these results cannot be satisfactorily explained even by reverting to hypothetical aural harmonics.
At this point the discrepancy must be merely acknowledged as a difficulty that requires explanation. The same result also appeared in the findings of Egan and Hake (their Fig.7) in consequence of an even narrower noise (less than 2 critical bands wide) than used in the present experiment.
It is concluded that the present results add further support to our hypothesis that spread of activity in the cochlea, and not aural harmonic distortion, underlies extended masking.
Fws. 1-8. Masked thresholds for tones us noise bands. The intensity scale is arbitrary but is the same for all figures. In each graph the lowest curve is the absolute threshold. Of the upper curves, the continuous one is the masked threshold for pure tone masker, the dashed one is for the noise band, and the dash double dot gives the frequency characteristics of the filter settings used. The filter characteristics are not plotted at intensities relative to the thresholds but merely to facilitate comparison of the masked threshold with the spectrum of noise producing it. The actual frequency limits and SPL's of the noise are given in the legends to the individual figures.
Further reading: Residual masking at low frequencies , Cyril Harris 1959