Recognition of Temporally Distorted Words by Listeners With and Without a Simulated Hearing Loss

Transcription

1 J Am Acad Audiol 9 : (1998) Recognition of Temporally Distorted Words by Listeners With and Without a Simulated Hearing Loss Andrew Stuart* Dennis P. Phillips' Abstract In separate experiments, word recognition performance of two groups of 12 normal-hearing young adults was investigated as a function of temporal distortion (i.e., time compression ratio or reverberation time) with and without a simulated high-frequency hearing loss (i.e., low-pass filtered at 2000 Hz). Performance decreased significantly as a function of increasing time compression, reverberation, and with the simulated hearing loss (p <.05). A statistically significant interaction between each of time compression ratio and reverberation time with the filtered listening condition was found (p <.05). This finding of an interactive, as opposed to a simple additive, effect of multiple distortions of the speech on word recognition performance suggests that loss of audibility alone cannot account for decrements in word recognition performance with time-altered speech. It is suggested that this multiplicative effect for combined acoustic distortions is a consequence of the functional loss of the high-frequency region of the cochlea. Not only is there a loss of audibility, but there is also a loss in temporal resolution capacity since it is perception mediated through the population of high-frequency auditory channels that has the best temporal resolution? Key Words : Abbreviations : Simulated hearing loss, speech perception, temporal distortion RAU = rationalized arcsine unit he peripheral auditory system has been described as containing a series of bandpass filters with continuously overlap- T ping center frequencies. The basilar membrane of the cochlea is believed to be the basis for these auditory filters (Pickles, 1988 ; Moore, 1997). It is generally accepted that in the normal cochlea, the auditory filters are highly tuned and their shape is reasonably symmetric at moderate intensity levels (e.g., see Moore, 1997). The impaired cochlea, however, is characterized by broad and insensitive auditory filters. The consequence of these broad auditory filters is a loss of audibility and reduced frequency resolution at low stimulus intensities. *Department of Communication Sciences and Disorders, East Carolina University, Greenville, North Carolina ; 'Department of Psychology, Dalhousie University, Halifax, Nova Scotia Repriht requests : Andrew Stuart, Department of Communication Sciences and Disorders, East Carolina University, Greenville, NC The perceptual repercussions are threefold : an augmented susceptibility to interference from noise ; an impaired ability to separate two or more signals occurring simultaneously independent of background noise ; and an inability to resolve spectral peaks and valleys in auditory signals (Moore, 1997). Although broad auditory filters in the impaired cochlea are detrimental for frequency resolution, it would seem intuitive that they would be beneficial for temporal resolution (Buus and Florentine, 1985 ; Moore, 1985). Auditory filters ring after the offset of input and the duration of the ringing is inversely proportional to the bandwidth of the filter. As wider auditory filters exhibit faster decay, they should, as well, exhibit better temporal resolution than narrower filters due to the absence of auditory activity. In other words, "since listeners with cochlea? impairments typically have broader-than-normal auditory filters, the temporal response of their filters should be correspondingly more rapid than normal, and this should lead to improved temporal resolution" (Moore, 1985, p. 199). In 199

2 Journal of the American Academy of Audiology/Volume 9, Number 3, June 1998 practice, however, performance on temporal resolution tasks is often poorer in listeners with sensorineural hearing losses than in normal-hearing listeners. It has been hypothesized that one reason that listeners with high-frequency hearing loss display inferior performance on temporal resolution tasks is a restricted listening bandwidth (Bacon and Viemeister, 1985 ; Buus and Florentine, 1985 ; Shailer and Moore, 1987 ; Moore et al, 1989 ; Moore, 1993 ; Viemeister and Plack, 1993). The functional loss of the high-frequency region of the cochlea results both in a loss of audibility and a loss in temporal resolving capacity. With respect to audibility, if the signal, or part thereof, is inaudible then listening performance should suffer for this reason alone. Bacon and Viemeister (1985) have demonstrated this with modulation thresholds for sinusoidally amplitude-modulated broadband noise. Hearing-impaired listeners, with primarily high-frequency hearing losses, displayed lower sensitivity to amplitude modulation and lower highest detectable modulation frequencies relative to normal-hearing listeners. The effects were level dependent : as the carrier spectrum level decreased, so did sensitivity and the highest detectable modulation frequency. Bacon and Viemeister (1985) attributed poorer performance to a narrower effective internal bandwidth within the hearing-impaired listeners. In other words, higher frequencies were inaudible to the hearing-impaired listeners. In addition, normal-hearing listeners exhibited a similar pattern of performance while listening with a simulated high-frequency hearing loss. In a follow-up study Bacon and Gleitman (1992) evaluated modulation detection thresholds in hearing-impaired listeners who had relatively flat mild-to-moderate losses. At comparable sensation levels (SLs), temporal modulation transfer functions for seven of eight hearing-impaired participants were similar to normal controls. The findings from both studies suggest that temporal resolution task performance may be dependent on audibility. Turner et al (1995) have demonstrated that when audibility is compensated for, temporal acuity of listeners in terms of speech recognition is not impaired. They investigated the ability of listeners to use temporal cues in speech directly. Five normal-hearing listeners, eight hearing-impaired listeners with sloping losses, and eight hearing-impaired listeners with flat losses served as participants. Participants listened to both natural (i.e., unprocessed) and processed vowel-consonant-vowel nonsense syl- lables. In the processed conditions, the spectral information of the nonsense syllables was removed, resulting in a time-varying speech envelope amplitude modulating a noise carrier. Four processed conditions were employed : the broadband envelope of the speech signal modulating a broadband noise ; a high-pass envelope of the speech signal modulating a high-pass noise ; a low-pass envelope of the speech signal modulating a low-pass noise ; and a combined two-channel signal of the low- and high-pass modulated signals. Participants listened to test stimuli at a level that offered their maximum performance for syllable recognition. The normal-hearing participants significantly outperformed the hearing-impaired listeners with the unprocessed natural speech stimuli. There was no syllable recognition performance difference, however, between the hearing-impaired and normal-hearing listeners for the processed stimuli. In other words, there was no difference in normal-hearing or hearing-impaired listeners' ability to process the temporal cues in speech. This held true whether the processed stimuli were presented in quiet or in a background of steady-state or amplitude-modulated noise with a signal-to-noise ratio of +10 db. The authors concluded that hearing-impaired listeners, when compensated for the loss of audibility, do not exhibit temporal acuity deficits in terms of speech recognition compared to normal-hearing listeners. Temporal resolution performance suffers with the functional loss of the high-frequency region of the cochlea as it is believed that these cochlear output channels have the best temporal resolving ability (Bacon and Viemeister, 1985 ; Buus and Florentine, 1985 ; Shailer and Moore, 1987 ; Moore et al., 1989). That is, listening is confined to low-frequency channels whose temporal resolution, albeit normal, is inherently inferior to that of the high-frequency channels. Consequently, hearing-impaired listeners' performance on temporal resolution tasks is poorer than normal listeners'. This speculation has been given support by a number of researchers who found that the performance of normal listeners with simulated high-frequency hearing losses parallels hearing-impaired listeners performance on forward and backward masking tasks (Zwicker and Schorn, 1982), gap detection (Buus and Florentine, 1985), temporal modulated transfer functions (Bacon and Viemeister, 1985), and detection of tone bursts in modulated noise (Arlinger and Dryselius, 1990). 200

3 Recognition of Temporally Distorted Words/Stuart and Phillips We have employed a word recognition in noise paradigm to demonstrate that noiseinduced hearing-impaired listeners' (Phillips et al, 1994) and presbyacusic listeners' (Stuart and Phillips, 1996) temporal resolution ability is diminished relative to normal-hearing listeners. Normal-hearing and hearing-impaired listeners were presented with Northwestern University Auditory Test No. 6 (NU-6) stimuli at 30 db (Stuart and Phillips, 1996) or 40 db (Phillips et al, 1994) SL relative to their respective speech recognition thresholds (SRTs) in spectrally identical competing continuous and interrupted broadband noise as a function of signal-to-noise ratio. The interrupted broadband noise had a 0.50 noise time fraction and was characterized with noise bursts and silent periods of durations varying randomly from 5 to 95 msec. Word recognition performance was superior in the interrupted noise. That is, all listeners experienced a perceptual advantage (i.e., a release from masking) in the interrupted noise relative to the continuous noise. Since the spectra of the two maskers were virtually identical, we reasoned that this advantage was due entirely to the temporal properties of the noise and the ability of the auditory system to resolve speech fragments in the silent gaps between noise bursts. Hearing-impaired listeners, however, experienced a significantly smaller release from masking than normal-hearing listeners in the interrupted noise. We reasoned that the reduced release from masking displayed by hearing-impaired listeners in the interrupted noise could be attributable to their restricted listening bandwidth. Again, the functional loss of the high-frequency region of the cochlea results in a loss of audibility and a loss in the temporal resolving capacity. Normal-hearing listeners, with a simulated high-frequency hearing loss, have demonstrated a detriment in performance in the interrupted noise akin to that of hearing-impaired listeners (Stuart et al, 1995). Word recognition performance was investigated in 12 normal-hearing young adults in continuous and interrupted broadband noise as a function of signal-to-noise ratio with and without a simulated highfrequency hearing loss (i.e., low-pass filtered at 2000 Hz). It was found that the normalhearing participants with a simulated highfrequency hearing loss displayed poorer performance in the interrupted broadband noise but not in the continuous broadband noise, relative to their normal listening performance. The similarities between normal listeners with a simulated hearing impairment and hearingimpaired listeners suggest that both groups are handicapped by a restrictive listening bandwidth and that poorer temporal performance followed from the availability of solely the lowfrequency cochlear channels. The present study tested the generality of these findings. We adopted the premise that on tasks challenging the temporal resolving capacity of the auditory system, a listener's performance can be governed in part by the functional ability of the perceptual channels emerging from different sectors of the cochlea. Our working hypothesis is that the low-frequency channels support poorer temporal resolution performance. Normal-hearing listeners should outperform high-frequency impaired listeners regardless of the temporal resolution task. This speculation was explored among normal-hearing listeners with and without a simulated hearing loss with two temporal resolution tasks. It was predicted that normal-hearing listeners' word recognition performance would decrease with increasing temporal distortions (e.g., increasing time compression ratio and reverberation time). If listening-with a restricted bandwidth resulted in simply a loss of audibility (i.e., spectral limiting), one would predict that normal-hearing listeners, with a simulated high-frequency hearing loss, would display a constant detriment in performance with increasing temporal distortion. That is, with a simulated high-frequency hearing loss, a normal-hearing listeners' performance should parallel his/her normal listening performance, but at a reduced level. If the performance detriment, with a simulated high-frequency hearing loss, does not parallel the normal listening performance, then audibility alone cannot be the sole consequence of restricted listening bandwidth. The purpose of this study was to investigate this speculation by comparing normal-hearing listeners' word recognition performance as a function of time compression ratio and reverberation time with and without a simulated high-frequency hearing loss. Method Participants EXPERIMENT 1 Twelve young adults served as participants (mean = 25.8 years, standard error = 0.9 ; three males and nine females). All participants presented with normal middle ear function (ASHA, 201

4 Journal of the American Academy of Audiology/Volume 9, Number 3, June ) and normal hearing sensitivity defined as having pure-tone thresholds at octave frequencies from 250 to 8000 Hz and SRTs of < 20 db HL (ANSI, 1996). Apparatus The test stimuli consisted of compact disc recordings of 50 monosyllabic word lists of the NU-6. NU-6 stimuli that were compressed 45 percent and 65 percent were obtained from lists five to eight of the Tonal and Speech Materials for Auditory Perceptual Assessment, Disc 1.0 compact disc (Department of Veterans Affairs, 1992). Uncompressed NU-6 stimuli lists one to four were obtained from the Speech Recognition and Identification Materials, Disc 1.1 compact disc (Department of Veterans Affairs, 1991). Both compact disc recordings of the compressed and uncompressed NU-6 stimuli were generated from the same master analog tape (Wilson, 1993 ; Noffsinger et al, 1994). A double-walled sound-treated audiometric suite (Industrial Acoustics Corporation), meeting specifications for permissible ambient noise (ANSI, 1991), served as the test environment. The recorded stimuli were routed from a compact disc player (Sony Model 608 ESD) through a passive analog filter (Krohn-Hite, Model 3340) to a clinical audiometer (Grason Stadler GSI 10 Model ) and presented to each participant through an insert edrphone (Etymotic Research Model ER-3A). During the simulated hearing impairment condition (i.e., filtered listening), the recorded stimuli were low passed, at 2000 Hz with a roll-off slope of 48 db/octave, to simulate a high-frequency hearing loss. Procedure order of the six stimulus conditions was determined by a digram-balanced Latin square (Wagenaar, 1969). All test stimuli were presented monaurally to the right ear of each participant. Results Participants'responses were scored as total whole word percent correct. Mean scores as a function of listening condition (i.e., simulated hearing loss vs normal listening) and time compression ratio are shown in Figure 1. As evident in Figure 1, participants exhibited poorer performance with increasing time compression and overall poorer performance in the simulated hearing-impairment condition. The participants' proportional scores were transformed to "rationalized" arcsine units (RAUs; Studebaker, 1985) prior to subjecting them to inferential statistical analyses. A twofactor repeated measures analysis of variance (ANOVA) was undertaken to investigate mean word recognition performance as a function of listening condition and time compression ratio. Statistically significant main effects were found for listening condition (F [1, 111 = 83.65, Geisser- Greenhouse p <.0001, w2 =.81) and time compression ratio (F [2, 221 = , Geisser- Greenhouse p <.0001, w2 =.94). That is, word recognition performance was significantly poorer in the simulated hearing impairment condition and decreased with increasing time compression ratio. A statistically significant listening condition by time compression ratio interaction was also found (F [2, 221 = 3.90, p =.035, ci2 =.088). In other words, word recognition performance decreased significantly more with increasing time compression ratio in the simulated hearing loss condition. Orthogonal single-df com- Participants were presented with the identical NU-6 stimuli at a 30 db SL relative to their respective SRTs.I The speech stimuli were presented at time compression ratios of 0 percent, 45 percent, and 65 percent both for the normal listening and the simulated hearing-impairment condition. The presentation of NU-6 lists was counterbalanced while the presentation d U 0 ---r- Normal Listening --0- Simulated Hearing Loss 'This presentation level in both experiments was consistent with the SL employed in our previous temporal resolution paradigm (Stuart and Phillips, 1996). In addition, we have found this level tolerable for hearing-impaired listeners in our ongoing research. Figure 1 Mean percent correct word recognition score as a function of listening condition and time compression ratio (n = 12). Error bars represent -1 standard deviation of the mean. 202

5 Recognition of Temporally Distorted Words/Stuart and Phillips parisons were undertaken to further assess the main effect of time compression ratio (Keppel and Zedeck, 1989 ; Keppel, 1991). There was a significant reduction in word recognition performance with time-compressed speech (i.e., 45% and 65% time compression) versus uncompressed speech (i.e., 0% time compression) (F [1, 22] = , Geisser-Greenhouse p <.0001). There was also a significant difference in word recognition performance between both levels of timecompressed speech (F [1, 22] = , Geisser-Greenhouse p <.0001). Further singledf comparisons were undertaken to assess differences between the two listening conditions at each time compression ratio (i.e., normal vs simulated hearing impairment at 0%, 45%, and 65% time compression). All comparisons were significantly different (Geisser-Greenhouse p <.0001). In other words, word recognition performance was significantly better in the normal listening condition regardless of the time compression ratio. Discussion The present finding of word recognition performance being inversely related to time compression ratio among normal-hearing adult listeners is consistent with previous literature (Fairbanks and Kodman, 1957 ; Beasley et al, 1972a, b; Schwartz and Mikus, 1977 ; Grimes et al, 1984 ; Beattie, 1986 ; Gordon-Salant and Fitzgibbons, 1993 ; Bornstein, 1994 ; Wilson et al, 1994 ; Humes et al, 1996). In addition, word recognition performance found in the present study was virtually identical to that reported in previous studies that employed normal-hearing young adults listening to the same compact disc recordings (Wilson et al, 1994 ; Humes et al, 1996). The detriment in performance for the filtered versus normal listening condition in quiet (i.e., 0% compression ratio) is consistent with the effects of low-pass filtering on the intelligibility of speech in quiet (Egan and Wiener, 1946 ; French and Steinberg, 1947 ; Pollack, 1948) and listeners' performance under the same conditions in our previous study (Stuart et al, 1995). The fording of an interaction, as opposed to a simple additive effect, of multiple distortions of the speech (i.e., high-pass filtering and time compression) on word recognition performance has been reported previously for normal-hearing young adult listeners (Lacroix and Harris, 1979 ; Lacroix et al, 1979). The performance of the normal-hearing listeners in the simulated hearing loss condition is similar to that of hearing-impaired listeners. That is, they perform significantly worse on time-compressed speech recognition tasks than normal-hearing listeners (Kurdzeil et al, 1975 ; Grimes et al, 1984 ; Gordon-Salant and Fitzgibbons, 1993, 1995a, b). Further, normal-hearing listeners with a simulated hearing loss (i.e., high-pass filtered listening) perform comparably to hearing-impaired listeners on time-compressed word recognition (Lacroix and Harris, 1979). The fact*that similarities in performance detriments exist between normal listeners with a simulated hearing impairment and hearing-impaired listeners suggests that both groups are handicapped by restrictive listening bandwidths. In other words, they are compelled to use exclusively their lowfrequency cochlear channels, which have inherently poorer temporal performance. Method Participants EXPERIMENT 2 A second group of 12 young adults served as participants (mean = 26.9 years, standard error = 1.1 ; three males and nine females). Different listeners from Experiment 1 were chosen for logistic reasons. All participants presented with normal middle ear function (ASHA, 1990) and normal hearing sensitivity defined as having pure-tone thresholds at octave frequencies from 250 to 8000 Hz and SRTs of <_ 20 db HL (ANSI, 1996). Apparatus The test stimuli consisted of compact disc recordings of 50-word lists one to four of the NU-6 from the Speech Recognition and Identification Materials, Disc 1.1 compact disc (Department of Veterans Affairs, 1991). The same double-walled sound-treated audiometric suite described in the first experiment served as the test environment. The recorded stimuli were routed from the same compact disc player through a multi-effect digital signal processor (Yamaha, Model SPX 1000) to the same clinical audiometer and presented to each participant through the insert earphone. During the simulated hearing impairment condition (i.e., filtered listening), the recorded stimuli were low passed, as described above. Procedure Participants were presented with the identical NU-6 stimuli at a 30 db SL relative to their 203

6 Journal of the American Academy of Audiology/Volume 9, Number 3, June 1998 respective SRTs. The speech stimuli were presented with reverberation times of 0 sec, 1.0 sec, and 1.5 sec for both the normal listening and the simulated hearing impairment condition. The reverberation times were controlled by the multi-effect digital signal processor. The presentation of NU-6 lists and six listening conditions was counterbalanced, as in the first experiment. All test stimuli were presented monaurally to the right ear of each participant. Results Participants'responses were scored as total whole word percent correct. Mean scores as a function of listening condition (i.e., simulated hearing loss vs normal listening) and reverberation are shown in Figure 2. As is evident in Figure 2, participants exhibited poorer performance with increasing reverberation times and overall poorer performance in the simulated hearing impairment condition. The participants' proportional scores were transformed to RAUs prior to subjecting them to inferential statistical analyses. A two-factor repeated measures ANOVA was undertaken to investigate mean word recognition performance as a function of listening condition and reverberation time. Statistically significant main effects were found for listening condition (F [l, 11] = , p <.0001, w2 =.88) and reverberation time (F [2, 22] = , p <.0001, w2 =.92). In other words, word recognition performance was significantly better in the normal listening condition and decreased with increasing reverberation time. A statistically significant listening condition by reverberation time interaction was also found (F [2, 22] = 9.61, 0 U 0 loon 75 ~ t Reverberation (s) Normal Listening Simulated Hearing Loss Figure 2 Mean percent correct word recognition score as a function of listening condition and reverberation time (n = 12). Error bars represent ±1 standard deviation of the mean. p =.0010, W2 =.19). In other words, word recognition performance decreased significantly more with increasing reverberation time in the simulated hearing loss condition. Orthogonal single-df comparisons were undertaken to further assess the main effect of reverberation time (Keppel and Zedeck, 1989 ; Keppel, 1991). There was significant reduction in word recognition performance with reverberated speech (i.e., 0.5-sec and 1.0-sec reverberation) versus nonreverberant speech (i.e., 0-sec reverberation) (F [1, 22] = , Geisser-Greenhouse p <.0001). A significant difference in word recognition performance between both levels of reverberation was also evident (F [1, 22] = 11.46, Geisser-Greenhouse p <.0001). Additional single-df comparisons were undertaken to assess differences between the two listening conditions at each reverberation time (i.e., normal vs simulated hearing impairment at 0-sec, 0.5-sec, and 1.0-sec reverberation). All comparisons were significantly different (Geisser-Greenhouse p <.0001). In other words, word recognition performance was significantly poorer in the simulated hearing loss condition regardless of reverberation time. Discussion Concordant with previous research, an inverse relationship between word recognition performance and reverberation time was found (Houtgast and Steeneken, 1973 ; Nabelek and Pickett, 1974 a, b; Gelfand and Hochberg, 1976 ; Finitzo-Hieber and Tillman, 1978 ; Nabelek and Robinette, 1978 ; Nabelek and Robinson, 1982 ; Loven and Collins, 1988 ; Harris and Swenson, 1990 ; Gordon-Salant and Fitzgibbons, 1993). Again, performance differences in the quiet condition (i.e., 0-sec reverberation) between filtered and normal listening conditions are consistent with the effects of low-pass filtering on the intelligibility of speech in quiet (Egan and Wiener, 1946 ; French and Steinberg, 1947 ; Pollack, 1948 ; Stuart et al, 1995) and with our listeners in Experiment 1. The finding of an interaction, as opposed to a simple additive effect, of high-pass filtering and reverberation on word recognition performance has also been previously reported for normalhearing young adult listeners (Martin et al, 1956 ; Harris, 1960 ; Loven and Collins, 1988). Simulated hearing-impaired listeners perform akin to hearing-impaired listeners : they do significantly worse in reverberant conditions than normal-hearing listeners (Nabelek and Pickett, 204

7 Recognition of Temporally Distorted Words/Stuart and Phillips 1974 a, b; Gelfand and Hochberg, 1976 ; Finitzo- Hieber and Tillman, 1978 ; Nabelek and Robinette, 1978 ; Harris and Swenson, 1990 ; Gordon-Salant and Fitzgibbons, 1993, 1995a). As a resemblance in performance detriment exists between normal listeners with a simulated hearing impairment and hearing-impaired listeners, we suggest that both groups are handicapped by restrictive listening bandwidths. That is, the perceptual temporal resolution supported solely through the low-frequency auditory channels is inherently inferior to that supported through the high-frequency channels. GENERAL DISCUSSION he findings of decreasing word recognition T performance with increasing time compression and reverberation were expected. As well, poorer word recognition with the simulated hearing loss condition (i.e., low-pass filtered listening) was anticipated. The significant interaction between time compression and reverberation with the filtered listening condition suggests that loss of audibility alone cannot account for decrements in word recognition performance with increasing time alterations in speech. When different types of distortions to speech are combined, "the cumulative effect on intelligibility will be greater than the sum of the individual effects" (Lacroix and Harris, 1979, p. 236). That is, the effects of time alterations and loss of audibility are not additive but "multiplicative" (Harris, 1960 ; Lacroix and Harris, 1979 ; Lacroix et al, 1979). We submit that this multiplicative effect for combined acoustic distortions is a consequence of the functional loss of the high-frequency region of the cochlea. Not only is there a loss of audibility, but there is also a loss in temporal resolution capacity (since it is the population of high-frequency auditory channels that has the best temporal resolution). The finding of multiplicative effects of combining distortions is not novel. There is a plethora of findings documenting such with reverberation and time compression in combination with at least one or more other distortions (e.g., Martin et al, 1956 ; Harris, 1960 ; Moncur and Dirks, 1967 ; Houtgast and Steeneken, 1973 ; Nabelek and Pickett, 1974a, b ; Finitzo-Hieber and Tillman, 1978 ; Lacroix and Harris, 1979 ; Lacroix et al, 1979 ; Nabelek and Mason, 1981 ; Loven and Collins, 1988 ; Harris and Swenson, 1990 ; Helfer, 1992 ; Bornstein, 1994 ; Wilson et al, 1994; Gordon-Salant and Fitzgibbons, 1995a, b). It is per- haps surprising to find little discussion as to what is responsible for such an occurrence. Bornstein (1994) suggests that "multiple distortions may interact with reduced redundancy when there is pathology affecting the auditory system" (p. 95). Several authors have offered that the effects of multiple distortions on speech perception can be explained in terms of spectral limiting (e.g., Harris, 1960 ; Harris et al, 1960 ; Lacroix and Harris, 1979 ; Lacroix et al, 1979 ; Loven and Collins, 1988). In particular, performance is described in terms of *the relative amount of moderate-to-high-frequency spectral information available to listeners. The importance that spectral limiting plays with distortions of reverberation and time compression cannot be ignored : in addition to reverberation causing a distortion in the time domain, because of prolongation and smearing of elements of speech and smoothing the temporal envelope of speech, it causes masking of adjacent phonemes (Houtgast and Steeneken, 1973). Time compression, on the other hand, while removing nonphonemic acoustic elements, decreases the auditory contrast between spectral peaks and valleys and increases the amount of information transmitted per unit time. What we find puzzling is the absence of a satisfactory explanation of the interactive effect when spectral and temporal distortions are combined. Spectral limiting alone cannot account for the performance detriments. Data from Loven and Collins (1988) are a case in point. They examined normal-hearing listeners' speech recognition ability under combinations of four signal modifications (i.e., reverberation time, masking, noise, and presentation level). Of interest to the present discussion is listeners' performance as a function of reverberation time and filtering in quiet (see Loven and Collins, 1988 ; Fig. 5). Nonsense Syllable Test (Levitt and Resnick, 1978) performance was examined under three reverberation times (i.e., 0.0, 0.6, and 1.2 sec) and four filtering conditions : broadband ( Hz), narrowband ( Hz), low pass ( Hz), and high pass ( Hz). The authors found differences in performance for conditions where high-frequency channel capacity was reduced relative to the other two where it was not (i.e., low-pass and narrowband vs high-pass and broadband filtering). Where high-frequency listening was restricted, what was evident was an interactive (multiplicative) effect of filtering and reverberation. For the high-pass and broadband filtering conditions, this did not seem to be the case (or at least the 205

8 Journal of the American Academy of Audiology/Volume 9, Number 3, June 1998 effect size was substantially reduced). Why would an additive effect of filtering and reverberation not be consistent across conditions? We submit that in the conditions where high-frequency channel capacity is preserved, temporal resolution ability is more likely to remain intact as it is the population of high-frequency auditory channels that have the best temporal resolution. In conditions of restricted high-frequency bandwidth listening, more than a loss of spectral limiting must be at play. In modeling temporal resolution, generally a four-stage linear system is described : a bank of filters, each followed by a nonlinear device (e.g., half wave rectification or square wave operation), temporal integrator or window, and a decision device (Rodenburg, 1977 ; Viemeister, 1979 ; Moore, 1993, 1997). The first two stages of these models are conceptualized as occurring in the cochlea, while the latter stages occur after the auditory nerve. Stimuli pass through four stages from the periphery to a central decision mechanism whereby some smoothing of the stimuli occurs such that rapid fluctuations are lost while slower ones are sustained. Considering a four-stage model, it is apparent that temporal resolution may be constrained by four factors : "the shape and bandwidth of the initial filter ; the type of nonlinearity assumed ; the cutoff frequency and slope of the low-pass filter (or the shape of the weighting function describing the temporal window) ; and the nature and sensitivity of the decision device" (Moore, 1997, p. 165). We contend that the findings of these experiments and those from our previous work (Phillips et al, 1994 ; Stuart et al, 1995 ; Stuart and Phillips, 1996) confirmed that restricted low-frequency auditory channel bandwidth listening may, in part, account for deficits in tasks involving temporal resolution. The findings do not, admittedly, address directly what perceptual process(es) may be responsible for impoverished performance among some listeners (Phillips, 1995). It may be the case that inherently longer response times of peripheral low-frequency channel filters, as a consequence of longer ringing after the offset of stimulus input relative to high-frequency channel filters, are responsible for poorer temporal resolution. On the other hand, it may be the case that peripheral manipulation (e.g., low-pass filtering) or peripheral high-frequency hearing loss reveals an inherently inferior, although normal, central mechanism. For example, the shape of the ear's temporal window/integrator increases slightly with decreasing center frequency (Plack and Moore, 1990). Temporal resolution may be inferior due to longer central integration as a consequence of broader temporal windows. Further, it may be the case that an interaction of both peripheral and central factors is responsible for impaired performance. Depressed word recognition performance in temporal resolution paradigms accompanying reduced high-frequency bandwidth must then, inevitably, be viewed as a consequence of either a factor(s) in the peripheral (e.g., ringing in the auditory filters) or central (e.g., broader temporal windows) auditory system and/or an interaction of peripheral and central factors. Finally, these results from this study do not answer whether the same temporal process(es) are involved in these two tasks. For example, the relative within- versus acrosschannel processing (Hall et al, 1988 ; Viemeister and Plack, 1993) contributions may not account for the same performance phenomena observed in the time compression and reverberation tasks. In conclusion, these findings demonstrate that performance on tasks that burden temporal resolving capacities significantly diminishes when a simulated high-frequency hearing loss is imposed on normal-hearing listeners. We contend that performance is dictated by a restricted listening bandwidth : performance reduction occurs because their cochlear output is restricted to the low-frequency channels, which function normally, but are inherently inferior to the highfrequency channels in terms of temporal resolution. We also suggest that similar performance on temporal resolution tasks by cochlearimpaired listeners (who typically demonstrate high-frequency hearing losses) occurs because their cochlear output is restricted as well. Finally, although hearing-impaired and normal-hearing listeners with simulated hearing loss perform in a similar fashion, it is recognized that differences in speech recognition ability are inevitable as elevated audibility thresholds, poor frequency resolution, and a reduced dynamic range usually accompany the former group. Acknowledgment. This paper was presented in part at the American Academy of Audiology Ninth Annual Convention, Fort Lauderdale, FL, April 19, REFERENCES American National Standards Institute. (1991). Permissible Ambient Noise Levels for Audiometric Test Rooms. (ANSI S ). New York : ANSI. 206

9 Recognition of Temporally Distorted Words/Stuart and Phillips American National Standards Institute. (1996). Specification for Audiometers. (ANSI S ). New York : ANSI. American Speech-Language-Hearing Association. (1990). Guidelines for screening for hearing impairments and middle ear disorders. ASHA 32(Supp12) : Arlinger S, Dryselius H. (1990). Speech recognition in noise, temporal and spectral resolution and impaired hearing. Acta Otolaryngol Suppl 469: Bacon SP, Gleitman RM. (1992). Modulation detection in subjects with relatively flat hearing losses. J Speech Hear Res35: Bacon SP, Viemeister NF. (1985). Temporal modulation transfer functions in normal-hearing and hearingimpaired listeners. Audiology 24 : Beasley DS, Forman B, Rintelmann WK (1972a). Intelligibility of time-compressed CNC monosyllables by normal listeners. J Auditory Res 12 : Beasley DS, Schwimmer S, Rintelmann WF. (1972b). Intelligibility of time-compressed monosyllables. J Speech Hear Res 15 : Beattie RC. (1986). Normal intelligibility functions for the Auditec CID W-22 test at 30% and 60% time compression. Am J Otol 7 : Bornstein SE (1994). Time compression and release from masking in adults and children. J Am Acad Audiol 5: Buns S, Florentine M. (1985). Gap detection in normal and impaired listeners : the effect of level and frequency. In: Michelsen A, ed. Time Resolution in Auditory Systems. Berlin : Springer-Verlag, Department of Veterans Affairs. (1991). Speech Recognition and Identification Materials, Disc 1.1 [Compact disc]. Long Beach, CA : Auditory Research Laboratory VA Medical Center. Department of Veterans Affairs. (1992). Tonal and Speech Materials for Auditory Perceptual Assessment, Disc 1.0 [Compact disc]. Long Beach, CA : Auditory Research Laboratory VA Medical Center. Egan JP, Wiener FM. (1946). On the intelligibility of bands of speech. J Acoust Soc Am 18 : Fairbanks G, Kodman F. (1957). Word intelligibility as a function of time compression. J Acoust Soc Am 29 : Finitzo-Hieber T, Tillman T. (1978). Room acoustics effects on monosyllabic word discrimination ability for normal and hearing impaired children. J Speech Hear Res 21 : French NR, Steinberg JC. (1947). Factors governing the intelligibility of speech sounds. J Acoust Soc Am 19 : Gelfand SA, Hochberg I. (1976). Binaural and monaural speech discrimination under reverberation. Audiology 15 : Gordon-Salant S, Fitzgibbons PJ. (1993). Temporal factors and speech recognition performance in young and elderly listeners. J Speech Hear Res 36: Gordon-Salant S, Fitzgibbons PJ. (1995a). Comparing recognition of distorted speech using an equivalent signalto-noise ratio index. J Speech Hear Res 38 : Gordon-Salant S, Fitzgibbons PJ. (1995b). Recognition of multiply degraded speech by young and elderly listeners. J Speech Hear Res 38 : Grimes AM, Mueller HG, Williams DL. (1984). Clinical considerations in the use of time-compressed speech. Ear Hear 5: Hall JW Davis AC, Haggard MP, Pillsbury HC. (1988). Spectro-temporal analysis in normal-hearing and cochlear-impaired listeners. J Acoust Soc Am 84: Harris JD. (1960). Combinations of distortions in speech : the twenty-five per cent safety factor by multiple-cueing. Arch Otolaryngol 72 : Harris JD, Haines HL, Meyers CK. (1960). The importance of hearing at 3KC for understanding speech. Laryngoscope 70 : Harris RW, Swenson DW. (1990). Effects of reverberation and noise on speech recognition by adults with various amounts of sensorineural hearing loss. Audiology 29 : Helfer KS. (1992). Aging and the binaural advantage in reverberation and noise. J Speech Hear Res 35 : Houtgast T, Steeneken HJM. (1973). The modulation transfer function in room acoustics as a predictor of speech intelligibility. Acustica 28 : Humes LE, Coughlin M, Talley L. (1996). Evaluation of the use of the new compact disk for auditory perceptual assessment in the elderly. JAm Acad Audiol 7: Keppel G. (1991). Design and Analysis: A Researcher's Handbook. 3rd Ed. Englewood Cliffs, NJ : Prentice Hall. Keppel G, Zedeck Z. (1989). Data Analysis for Research Designs. New York: WH Freeman. Kurdzeil S, Rintelmann WF, Beasley DS. (1975). Performance of noise-induced hearing-impaired listeners on time-compressed consonant-nucleus-consonant monosyllables. JAm Audiol Soc 1: Lacroix PG, Harris JD. (1979). Effects of high-frequency cue reduction on the comprehension of distorted speech. J Speech Hear Disord 44 : Lacroix PG, Harris JD, Randolph KJ. (1979). Multiplicative effects on sentence comprehension for combined acoustic distortions. J Speech Hear Res 22 : Levitt H, Resnick SB. (1978). Speech reception by the hearing impaired : methods of testing and the development of new tests. Scand Audiol Suppl 6: Loven FC, Collins MJ. (1988). Reverberation, masking filtering, and level effects of speech recognition performance. J Speech Hear Res 31 : Martin DW Murphy RL, Meyer A. (1956). Articulation reduction by combined distortion of speech waves. J Acoust Soc Am 28:

10 Journal of the American Academy of Audiology/Volume 9, Number 3, June 1998 Moncur JP, Dirks D. (1967). Binaural and monaural speech intelligibility in reverberation. J Speech Hear Res 10: Moore BCJ. (1985). Frequency selectivity and temporal resolution in normal and hearing-impaired listeners. Br J Audiol 19 : Moore BCJ. (1993). Temporal analysis in normal and hearing impaired hearing. In : Tallal P, Galaburda AM, Llinas RR, von Euler C, eds. Temporal Information Processing in the Nervous System: Special Reference to Dyslexia and Dysphasia: Vol Annals of the New York Academy of Sciences. New York : New York Academy of Sciences, Moore BCJ. (1997). An Introduction to the Psychology of Hearing. 4th Ed. London : Academic Press. Moore BCJ, Glasberg BR, Donaldson E, McPherson T, Plack CJ. (1989). Detection of temporal gaps in sinusoids by normally hearing and hearing-impaired subjects. J Acoust Soc Am 85: NabelekAK, Mason D. (1981). Effects of noise and reverberation on binaural and monaural word identification by subjects with various audiograms. J Speech Hear Res 24: Nabelek AK, Pickett JM. (1974a). Reception of consonants in a classroom as affected by monaural and binaural listening, noise, reverberation and hearing aids. JAcoust Soc Am 56 : NabelekAK, Pickett JM. (1974b). Monaural and binaural speech perception through hearing aids under noise and reverberation with normal and hearing impaired listeners. J Speech Hear Res 17: Nabelek AK, Robinette L. (1978). Reverberation as a parameter in clinical testing. Audiology 17 : Nabelek AK, Robinson PK. (1982). Monaural and binaural speech perception in reverberation for listeners of various ages. J Acoust Soc Am 71: Noffsinger D, Wilson RH, Musiek FE. (1994). Department of Veterans Affairs compact disc recording for auditory perceptual assessment : background and introduction. J Am Arad Audiol 5: Phillips DP. (1995). Central auditory processing : a view from auditory neuroscience. Am J Otol 16: Phillips DP, Rappaport JM, Gulliver JM. (1994). Impaired word recognition in noise by patients with noised-induced cochlear hearing loss : contribution of a temporal resolution defect. Am J Otol 15 : Pickles JO. (1988). An Introduction to the Physiology of Hearing. 2nd Ed. London : Academic Press. Plack CJ, Moore BCJ. (1990). Temporal window shape as a function of frequency and level. J Acoust Soc Am 87: Pollack 1. (1948). Effects of high pass and low pass filtering on the intelligibility of speech in noise. J Acoust Soc Am 20 : Rodenburg M. (1977). Investigation of temporal effects with amplitude modulated signal. In: Evans EF, Wilson JP, eds. Psychophysics and Physiology of Hearing. London : Academic Press, Schwartz DM, Mikus B. (1977). Performance of normal hearing listeners on time-compressed modified rhyme test. J Am Audiol Soc 3 : Shailer MJ, Moore BCJ. (1987). Gap detection and the auditory filter : phase effects using sinusoidal stimuli. J Acoust Soc Am 81: Stuart A, Phillips DP. (1996). Word recognition in continuous and interrupted broadband noise by young normal-hearing, older normal-hearing, and presbyacusic listeners. Ear Hear 17 : Stuart A, Phillips DP, Green WB. (1995). Word recognition performance in continuous and interrupted noise by normal-hearing and simulated hearing-impaired listeners. Am J Otol 16 : Studebaker G. (1985). A"rationalized" arcsine transform. J Speech Hear Res 28 : Turner CW, Souza PE, Forget LN. (1995). Use of temporal envelope cues in speech recognition by normal and hearing-impaired listeners. J Acoust Soc Am 97: Viemeister NF. (1979). Temporal modulation transfer functions based on modulation thresholds. JAcoust Soc Am 66 : Viemeister NF, Plack CJ. (1993). Time analysis. In : Yost WA, Popper AN, Fay RR, eds. Human Psychoacoustics. New York : Springer-Verlag, Wagenaar WA. (1969). Note on the construction of digrambalanced Latin squares. Psychol Bull 72: Wilson RH. (1993). Development and use of auditory compact discs in auditory evaluation. J Rehabil Res Dev 30 : Wilson RH, Preece JP, Salamon DL, Sperry JL, Bornstein SP. (1994). Effects of time compression and time compression plus reverberation on the intelligibility of the Northwestern University Auditory Test No. 6. JAm Acad Audiol 5: Zwicker E, Schorn K. (1982). Temporal resolution in hardof-hearing patients. Audiology 21 :