Intensity discrimination and detection of amplitude modulation in electric hearing.

Wojtczak and Viemeister [J. Acoust. Soc. Am. 106, 1917-1924 (1999)] demonstrated a close relationship between intensity difference limens (DLs) and 4-Hz amplitude modulation (AM) detection thresholds in normal-hearing acoustic listeners. The present study demonstrates a similar relationship between intensity DLs and AM detection thresholds in cochlear-implant listeners, for gated stimuli. This suggests that acoustic and cochlear-implant listeners make use of a similar decision variable to perform intensity discrimination and modulation detection tasks. It can be shown that the absence of compression in electric hearing does not preclude this possibility.

Observer weighting of monaural level information in a pair of tone pulses.

A correlational analysis was used to assess the relative weight given to the levels of two monaurally presented tone pulses for interpulse intervals (IPIs) ranging from 2-256 ms. In three different experimental conditions, listeners were instructed to discriminate the level of the first pulse, the level of the second pulse, or the difference between the levels of the two pulses. The level of the target pulse was chosen randomly and independently from trial to trial from a Gaussian distribution. The level of the nontarget pulse was either fixed at 75 dB SPL or varied in the same manner as the level of the target. In the tasks in which one pulse was to be ignored, listeners gave increasing weight to the nontarget component as IPI decreased. Listeners weighted the level information in the pulses appropriately only when the IPI approached 256 ms. When the listeners were instructed to compare the pulse levels to one another, two of three listeners weighted the levels optimally at all IPIs, while the third listener did so only at the longest IPI. For the two listeners who weighted the pulses optimally, a minimum in performance was achieved at IPIs around 16-32 ms. Intensity discrimination thresholds were also measured for one pulse in the presence of a second fixed pulse for IPIs of 2-256 ms. Thresholds were higher in all the two-pulse conditions relative to a one-pulse condition, and were dependent on the level of the nontarget pulse but not on IPI. The results indicate that level information is integrated to some extent over fairly long durations, but not in a manner that is consistent with simple temporal integration.

Stochastic properties of cat auditory nerve responses to electric and acoustic stimuli and application to intensity discrimination.

Statistical properties of electrically stimulated (ES) and acoustically stimulated (AS) auditory nerve fiber responses were assessed in undeafened and short-term deafened cats, and a detection theory approach was used to determine fibers' abilities to signal intensity changes. ES responses differed from AS responses in several ways. Rate-level functions were an order of magnitude steeper, and discharge rate normally saturated at the stimulus pulse rate. Dynamic ranges were typically 1-4 dB for 200 pps signals, as compared with 15-30 dB for AS signals at CF, and they increased with pulse rate without improving threshold or changing absolute rate-level function slopes. For both ES and AS responses, variability of spike counts elicited by repeated trials increased with level in accord with Poisson-process predictions until the discharge rate exceeded 20-40 spikes/s. AS variability continued increasing monotonically at higher discharge rates, but more slowly. In contrast, maximum ES variability was usually attained at 100 spikes/s, and at higher discharge rates variability reached a plateau that was either maintained or decreased slightly until discharge rate approached the stimulus pulse rate. Variability then decreased to zero as each pulse elicited a spike. Increasing pulse rate did not substantially affect variability for rates up to 800 pps; rather, higher pulse rates simply extended the plateau region. Spike count variability was unusually high for some ES fibers. This was traced to response nonstationarities that stemmed from two sources, namely level-dependent fluctuations in excitability that occurred at 1-3 s intervals and, for responses to high-rate, high-intensity signals, fatigue that arose when fibers discharged at their maximum possible rates. Intensity discrimination performance was assessed using spike count as the decision variable in a simulated 2IFC task. Neurometric functions (percent correct versus intensity difference) were obtained at several levels of the standard (I), and the intensity difference (delta I) necessary for 70% correct responses was estimated. AS Weber fractions (10 log delta I/I) averaged +0.2 dB (delta IdB = 3.1 dB) for 50 ms tones at CF. ES Weber fractions averaged -12.8 dB (delta IdB = 0.23 dB) for 50 ms, 200 pps signals, and performance was approximately constant between 100 and 1000 pps. Intensity discrimination by single cells in ES conditions paralleled human psychophysical performance for similar signals. High ES sensitivity to intensity changes arose primarily from steeper rate-level functions and secondarily from reduced spike count variability.

Intensity discrimination and detection of amplitude modulation.

Thresholds for detection of low-rate sinusoidal amplitude modulation and for detection of intensity increments were measured over a wide range of levels in an examination of the relationship between these fundamental aspects of intensity processing. As expected, thresholds measured with a continuous 1-kHz tone decrease with increasing carrier/pedestal level. For levels between 6 and 85 dB SPL the data are well described by 10 log delta I/I = 0.44.(20 log m) + D(fm), where delta I/I is the Weber fraction for increment detection, m is the modulation index at threshold, and D(fm) depends on modulation rate (fm). The relationship between the psychometric functions for modulation and increment detection is also consistent with this equation. The data indicate a clear relationship between modulation and increment detection and thus provide an important additional consideration for models of modulation processing. No existing models provide an adequate account of this relationship.

The effects of frequency region and bandwidth on the temporal modulation transfer function.

Temporal resolution was examined as a function of frequency region and listening region. The first experiment demonstrated that amplitude- and frequency-modulated tones are not appropriate stimuli to study temporal resolution as a functional of frequency region, due to the availability of other cues in addition to temporal ones. In the other experiments, thresholds for detection of sinusoidal amplitude modulation of a noise band were measured as a function of frequency region, bandwidth, and level of surrounding notched noise masker. Temporal modulation transfer functions (TMTFs) measured in low- and high-frequency regions did not differ in sensitivity or in cutoff frequency, suggesting that initial "critical band" filtering did not affect temporal resolution. When the upper cutoff frequency of the noise was held constant, TMTF sensitivity increased with noise bandwidth, while the cutoff frequency of the TMTF did not show measurable change. These results are consistent with the predictions of an envelope detector model if peripheral filtering in the lower-frequency range is assumed to be approximately twice as wide as that estimated by measuring thresholds for a tone in notched noise. Restricting the listening region with notched noise increased thresholds for low modulation frequencies but not for high. This is consistent with other data showing that upward spread of excitation may increase the effective modulation depth, but only for low modulation frequencies.

Psychometric functions and temporal integration in electric hearing.

Temporal-integration functions and psychometric functions for detection were obtained in eight users of the Nucleus 22-electrode cochlear implant. Stimuli were 100-Hz, 200-microseconds/phase trains of biphasic pulses with durations ranging from 0.44 to 630.4 ms (1 to 64 pulses). Temporal-integration functions were measured for 21 electrodes. Slopes of these functions were considerably shallower than the 2.5 dB/doubling slopes typically observed in acoustic hearing. They varied widely across subjects and for different electrodes in a given subject, ranging from 0.06 to 1.94 dB/doubling of stimulus pulses, with a mean [standard deviation (s.d.)] value of 0.42 (0.38). Psychometric functions were measured for 11 of the same 21 electrodes. Slopes of psychometric functions also varied across subjects and electrodes, and were 2-20 times steeper than those reported by other investigators for normal-hearing and cochlear-impaired acoustic listeners. Slopes of individual psychometric functions for 1-, 2-, 4-, and 8-pulse stimuli ranged from 0.20 to 1.84 log d'/dB with a mean (s.d.) value of 0.77 (0.45). Psychometric-function slopes did not vary systematically with stimulus duration in most cases. A clear inverse relation between slopes of psychometric functions and slopes of temporal-integration functions was observed. This relation was reasonably well described by a hyperbolic function predicted by the multiple-looks model of temporal integration [Viemeister and Wakefield, J. Acoust. Soc. Am. 90, 858-865 (1991)]. Psychometric-function slopes tended to increase with absolute threshold and were inversely correlated with dynamic range, suggesting that observed differences in psychometric-function slopes across subjects and electrodes may reflect underlying differences in neural survival.

Masking of a brief probe by sinusoidal frequency modulation.

Contrary to level detection models, the thresholds for a brief-duration probe masked by a sinusoidal frequency modulation (FM) masker increases as the modulation index (beta) of FM increases [Zwicker, Acustica 31, 243-256 (1974)]. In this paper the reason for this phenomenon is investigated. In experiment 1, a 10-ms, 1-kHz probe was detected in the presence of an FM masker centered at 1 kHz and sinusoidally modulated at 16 Hz. Thresholds increased by over 15 dB with increasing beta, consistent with Zwicker's findings. In experiment 2, the instantaneous frequency changes of the masker used in experiment 1 were clipped and the resulting thresholds indicated that detection was determined primarily by the masker's total frequency excursion rather than by its instantaneous sweep rate. In experiment 3, the FM maskers from the first two experiments were passed through a roex filter centered at 1 kHz and the resulting envelope was used to amplitude modulate a 1-kHz tone, producing approximately the same effective envelope at 1 kHz as the FM maskers. Threshold functions for the amplitude modulated (AM) maskers were similar to those for their corresponding FM maskers. Thresholds increased by over 15 dB while the total energy of the AM masker decreased by over 10 dB, again contrary to standard level-detection models. The results from these experiments can be explained, at least qualitatively, by a model based on envelope shape discrimination: similarities between the envelopes of the masker alone and masker-plus-probe at the output of an auditory filter centered on the frequency of the probe are primarily responsible for the observed masking, particularly at large beta's.

Intensity discrimination as a function of stimulus level with electric stimulation.

Difference limens (DLs) for changes in electric current were measured from multiple electrodes in each of eight cochlear-implanted subjects. Stimuli were 200-microseconds/phase biphasic pulse trains delivered at 125 Hz in 300-ms bursts. DLs were measured with an adaptive three-alternative forced-choice procedure. Fixed-level psychometric functions were also obtained in four subjects to validate the adaptive DLs. Relative intensity DLs, specified as Weber fractions in decibels [10 log (delta I/I)] for standards above absolute threshold, decreased as a power function of stimulus intensity relative to absolute threshold [delta I/I = beta (I/I0) alpha] in the same manner as Weber fractions for normal acoustic stimulation reported in previous studies. Exponents (alpha) of the power function for electric stimulation ranged from -0.4 to -3.2, on average, an order of magnitude larger than exponents for acoustic stimulation, which range from -0.07 to -0.11. Normalization of stimulus intensity to the dynamic range of hearing resulted in Weber functions with similar negative slopes for electric and acoustic stimulation, corresponding to an 8-dB average improvement in Weber fractions across the dynamic range. Sensitivity to intensity change ¿10 log beta¿ varied from -0.42 to -13.5 dB compared to +0.60 to -3.34 dB for acoustic stimulation, but on average was better with electric stimulation than with acoustic stimulation. Psychometric functions for intensity discrimination yielded Weber fractions consistent with adaptive procedures and d' was a linear function of delta I. Variability among repeated Weber-fraction estimates was constant across dynamic range. Relatively constant Weber fractions across all or part of the dynamic range, observed in some subjects, were traced to the intensity resolution limits of individual implanted receiver/stimulators. DLs could not be accurately described by constant amplitude changes, expressed as a percentage of dynamic range ¿delta A(% DR)¿. Weber fractions from prelingually deafened subjects were no better or worse than those from postlingually deafened subjects. The cumulative number of discriminable intensity steps across the dynamic range of electric hearing ranged from as few as 6.6 to as many as 45.2. Physiologic factors that may determine important features of electric intensity discrimination are discussed in the context of a simple, qualitative, rate-based model. These factors include the lack of compressive cochlear preprocessing, the relative steepness of neural rate-intensity functions, and individual differences in patterns of neural survival.

Cues for discrimination of envelopes.

Thresholds for detection of sinusoidal amplitude modulation at a signal modulation frequency were measured in the presence of a masker modulation frequency, with broadband noise carriers. Broad tuning for modulation frequency was observed. For maskers half or twice the signal frequency, thresholds depended on the relative phases of the signal and masker. These results were used to determine what aspects of envelopes listeners might be using in making decisions. Simulations were performed using an envelope detector model, consisting of bandpass filtering, half-wave rectification, and low-pass filtering. Decisions were based on envelope statistics that have been used to predict other data. These statistics were (1) rms power, (2) ratio of maximum to minimum amplitude (max/min), (3) crest factor, (4) fourth moment, and (5) average slope. The max/min statistic was successful at predicting the major trends in the data, without requiring the presence of channels tuned to modulation frequency.

Intensity discrimination under forward and backward masking: role of referential coding.

The present experiments investigated the hypothesis that listeners can code intensity by reference to proximal stimuli in order to improve intensity discrimination performance in conditions of nonsimultaneous masking. The experiments used 30-ms tone bursts as the masker, pedestal, and "proximal burst." The masker level was 80 dB, the pedestal level was 50 dB. In the first experiment the silent interval between the masker and the pedestal was varied. Surprisingly, in both forward and backward masking situations, the Weber fraction decreased as the silent interval was decreased from 100 to 12.5 ms. This is consistent with the referential coding hypothesis: At short intervals performance improves because the level of the pedestal is coded by reference to the proximal masker. In a further set of experiments, the silent interval was 100 ms and an additional proximal burst was presented either 12.5 ms before or 12.5 ms after the pedestal. The proximal burst produced a substantial decrease in the Weber fraction, but only when it was close in frequency to the pedestal, and with a higher intensity. The results are consistent with the auditory system having the ability to produce a robust intensity measure by reference to proximal signals. These findings also provide further evidence that the mid-level elevation in forward masking is not solely the result of processes operating at the level of the auditory nerve.

Intensity discrimination in normal-hearing and hearing-impaired listeners.

Weber fractions (delta I/I) for gated 500-ms tones at 0.3, 0.5, 1, 2, and 3 kHz, and at levels of the standard ranging from absolute threshold to 97 dB SPL, were measured in quiet and in high-pass noise in five listeners with cochlear hearing loss and in three normal-hearing listeners. In regions of hearing loss, the Weber fractions at a given SPL were sometimes normal. When the Weber fractions were normal or near-normal, the addition of high-pass noise elevated the Weber fraction, strongly suggesting the use of spread of excitation to higher frequencies. Inversely, when the Weber fractions were elevated, the addition of high-pass noise produced no additional elevation, suggesting an inability to use spread of excitation. In general, the relative size of the Weber fractions, the effects of high-pass noise, and to a lesser extent, the dependence of the Weber fraction on level, were consistent with expectations based upon the audiometric configuration and the use of excitation spread. There were several notable inconsistencies, however, in which normal Weber fractions were seen at a frequency on the edge of a steep high-frequency loss, and in which elevated Weber fractions were observed in a flat audiometric configuration. Finally, when compared at the same SL, the Weber fraction was sometimes smaller in cochlear-impaired than in normal hearing listeners. This was true even in high-pass noise, where excitation spread was limited, and may reflect the unusually steep rate versus level functions seen in auditory nerve fibers that innervate regions of pathology.

Frequency modulation versus amplitude modulation discrimination: evidence for a second frequency modulation encoding mechanism.

The encoding mechanisms for amplitude modulation (AM) and frequency modulation (FM) were investigated using AM-FM discrimination tasks. In the first experiment, AM and FM were set at equally detectable levels within a trial, and discrimination thresholds were obtained adaptively in a 3IFC task. Here, AM-FM discrimination thresholds were considerably larger than both AM and FM detection thresholds. This is consistent with an encoding system whereby AM and FM are partially encoded by the same mechanism. In the second experiment, performance on AM-FM discrimination is measured with a fixed-level procedure. Psychometric functions obtained for a constant modulation depth of AM were nonmonotonic with FMs modulation index beta and each displayed a single minimum. The nonmonotonic nature of the functions is consistent with a model in which FM is encoded primarily with the same mechanism that encodes AM but also with a second mechanism, probably related to changes in instantaneous frequency, that is independent of the mechanism that extracts AM. The fact that minima in the discrimination psychometric functions increase from d' = 0 as beta increases indicates that the information encoded by the second mechanism becomes more detectable with increasing beta.

Intensity discrimination and increment detection at 16 kHz.

When presented for several seconds, a very high-frequency tone can decay to inaudibility in subjects with normal hearing. The purpose of the present study was to determine how such a tone behaves once it is inaudible. Intensity difference limens (DLs) at 16 kHz were measured for gated (audible) and continuous (inaudible) pedestals over a range of pedestal sensation levels from about 0-60 dB, and were compared with those obtained in the same two subjects at 1 kHz [N. F. Viemeister and S. P. Bacon, J. Acoust, Soc. Am. 84, 172-178 (1988)]. The results at the two frequencies were remarkably similar, indicating, among other things, that a continuous 16-kHz pedestal--despite being inaudible-behaves as if it were audible. In addition, the results suggest that there is little or no relationship between high-frequency tone decay and intensity DLs. The locus of this long-term adaptation effect is presumably peripheral to the site where binaural interactions occur, and may be at the hair cell or auditory nerve. The intensity DLs are more consistent with a multiplicative model of (long-term) adaptation than with a subtractive model, suggesting that the nature of this adaptation is different from that which characterizes short-term adaptation.

Modulation detection and discrimination with three-component signals.

In an attempt to study the processing of amplitude and frequency modulation (AM and FM), detection and discrimination tasks using mixed modulation (MM) signals were performed. Modulation detection thresholds were obtained for three-component signals that span the parameter space between AM and quasi-FM. A single-cue modulation detection model predicts the thresholds with reasonable accuracy. If one assumes that the AM and FM components are extracted separately, thresholds are also well predicted if the d' of the MM signals is equal to the sum of the separate d's of the AM and FM components (two-cue summation model). This could arise from common internal noise that puts the AM and FM information along a single decision axis. A modulation discrimination task was then examined in which the subjects discriminate between signals with both different modulation depths and different modulation types. The single-cue model predicts performance well. In order for the two-cue model to predict the results, the AM and FM cues must be combined into a single statistic before a decision can be made; the listener cannot process the cues separately.

Psychoacoustic equivalence of frequency modulation and quasi-frequency modulation.

Frequency modulation (FM) is known to be reasonably approximated by quasi-frequency modulation (QFM) for small modulation indices, beta, but the range of beta for which this approximation is appropriate is unclear. Thresholds for discrimination between FM and QFM with equal beta's are obtained in order to estimate an upper bound on beta for determining when this approximation is valid psychoacoustically, i.e., when FM and QFM are indiscriminable. At low modulation frequencies (fm < or = 4 Hz), QFM is never a valid approximation to FM at any detectable modulation level since discrimination thresholds are below FM detection thresholds. For modulation frequencies between 8 and 32 Hz, discrimination thresholds are approximately -2.5 dB (20 log beta) and can be accounted for by detection of envelope fluctuations in the QFM signal. For modulation frequencies at 64 Hz and above, discrimination thresholds improve with increasing modulation frequency in the same manner as FM and QFM detection thresholds. Discrimination in this frequency region seems to be mediated by detection of the component 2fm Hz below the carrier, i.e., by the most detectable component in the FM signal which does not occur in the QFM signal.

Is useful speech information carried by fibers with high characteristic frequencies?

It has been proposed that auditory-nerve fibers with characteristic frequencies (CFs) above the speech-frequency range are important in speech perception when the signal-to-noise ratio in the speech frequency range is low [S. Greenberg, J. Phon. 16, 139-149 (1988)]. If this is true, then it might be expected that recognition of speech at low signal-to-noise ratios would worsen with the addition of high-pass noise sufficiently intense to mask information in high-CF fibers. This hypothesis was tested for closed-set recognition of vowels and spondees. Recognition was measured as a function of signal-to-noise ratio in speech-shaped noise, with and without an intense high-pass noise. The addition of high-pass noise did not degrade vowel recognition. Spondee recognition decreased with the addition of the high-pass noise at the highest levels of the speech-shaped noise. However, this same decrease was seen when the spondees were low-pass filtered to simulate downward spread of masking by the high-pass noise. This indicates that the decrease in spondee recognition with high-pass noise was due to masking of information in fibers with CFs in the speech range. Overall, these results suggest that fibers whose CFs are above the speech range are not necessary for speech perception in noise or in quiet.

Suppression and the dynamic range of hearing.

The results from experiments that have examined intensity discrimination in the presence of notched noise indicate that spread of excitation is not necessary for the auditory system to maintain a large dynamic range. In those experiments the notched noise and the pedestal were simultaneously present. It is possible, therefore, that the notched noise suppressed the pedestal, and increased the dynamic range by reducing the excitation level [A. R. Palmer and E. F. Evans, Hear. Res. 7, 305-323 (1982)]. In the experiment described here, spread of excitation was masked nonsimultaneously in order to avoid suppressive effects. The brief sinusoidal pedestal was presented in a 13-ms gap between two bursts of a masking complex. The masking complex consisted of two sinusoids at frequencies of 0.8fc and 1.2fc (where fc was the pedestal frequency), each having a level either the same as, or 10 dB below the pedestal level, and a notched noise with a spectrum level 40 dB below the level of the sinusoids. Detection thresholds were measured to ensure that the complex was effective in masking spread of excitation. Weber fractions were measured at two pedestal frequencies, 1 and 4 kHz, and at eight pedestal levels at each frequency, covering a range of 70 dB. The results indicate that, although the masking complex raised the Weber fraction by up to 10 dB in some conditions, performance was no worse at high levels than at medium or low levels. This suggests that the auditory system can maintain a large dynamic range in the absence of suppression and spread of excitation.

Intensity discrimination under backward masking.

The Weber fraction was measured for a 25-ms sinusoidal pedestal presented 100 ms before, or 100 ms after, an intense narrow-band noise. Consistent with the finding of Zeng et al. [Hear. Res. 55, 223-230 (1991)], the forward masker caused an elevation in the Weber fraction at medium pedestal levels. Surprisingly, however, a much larger midlevel elevation was observed in the backward masking conditions; in some cases, the Weber fraction was increased by over 20 dB by the backward masker. In both masking conditions, presenting a notched noise simultaneously with the pedestal reduced the magnitude of the midlevel elevation. These results indicate that it is possible to produce large masking effects on intensity discrimination in conditions where there is no possibility of the masker affecting the representation of the pedestal at the level of the auditory nerve. This suggests that there may be "central" processes underlying the original finding of Zeng et al. Despite the similarities in the results, however, it is not certain that the elevations seen in the forward and backward masking conditions were caused by the same mechanisms.

The effects of notched noise on intensity discrimination under forward masking.

Zeng et al. [Hear. Res. 55, 223-230 (1991)] reported that at moderate levels there is an increase in the intensity jnd for 25-ms sinusoidal pedestals presented 100 ms after an intense narrow-band noise. They suggested that this effect is related to the finding that low spontaneous rate (SR) auditory-nerve neurons take a considerable time to recover from adaptation [E. M. Relkin and J. R. Doucet, Hear. Res. 55, 215-222 (1991)]: 100 ms after the noise, the low-SR neurons still have elevated thresholds. Therefore, the intensity of a pedestal falling between the saturation level of the high-SR neurons and the elevated threshold of the low-SR neurons will be poorly represented in neutral firing rates, and the jnd will be high. A problem with this interpretation is that subjects may listen "off frequency." Theoretically, it should always be possible to choose a frequency channel for which the pedestal level is within the dynamic range of the high-SR neurons. In the present study, the experiment of Zeng et al. was replicated but with the pedestal presented in the temporal center of a notched noise to prevent off-frequency listening. Surprisingly, the notched noise substantially decreased the jnd at mid levels, removing or severely reducing the mid-level jnd elevation. This was true for pedestal frequencies of 1 and 6 kHz. It was also found that even if the notched noise was terminated before pedestal onset the jnd elevation was reduced. This suggests that the effect of the notched noise is not due to suppression.(ABSTRACT TRUNCATED AT 250 WORDS)

Temporal integration and multiple looks.

The decrease in detection and discrimination thresholds with increases in signal duration has often been taken to indicate that a process of relatively long-term temporal integration occurs in hearing. Two experiments are reported that suggest that no such process occurs. The first experiment is similar to the two-pulse experiment reported by Zwislocki [J. Zwislocki, J. Acoust. Soc. Am. 32, 1046-1059 (1960)] in which the threshold in quiet for a pair of brief pulses is measured as a function of the temporal separation between them. Our data indicate that power integration occurs only for separations less than approximately 5 ms. For separations larger than 5-10 ms, thresholds do not change with separation and the pulses appear to be processed independently. In the second experiment, brief 1-kHz tone pulses separated by 100 ms are presented during gaps in a wideband noise. The threshold for a pair of pulses is lower than that for either pulse presented alone, indicating that some type of "integration" occurs. However, the threshold for the pulse pair is not affected by changes in the level of the noise during the interval between the pulses. These data are inconsistent with the classical view of temporal integration that involves long-term integration. They are consistent with the notion that the input is sampled at a fairly high rate and that these samples or "looks" are stored in memory and can be accessed and processed selectively. This multiple-look model can account for the data from the present experiment and also can account for the data on temporal integration for tones and noise.(ABSTRACT TRUNCATED AT 250 WORDS)