Auditory Steady-State Responses and Word Recognition Scores in Normal-Hearing and Hearing-Impaired...

Auditory Steady-State Responses and WordRecognition Scores in Normal-Hearing and Hearing-

Impaired AdultsAndrew Dimitrijevic, M. Sasha John, and Terence W. Picton

Objective: The number of steady-state responsesevoked by the independent amplitude and fre-quency modulation (IAFM) of tones has been re-lated to the ability to discriminate speech sounds asmeasured by word recognition scores (WRS). In thepresent study IAFM stimulus parameters were ad-justed to resemble the acoustic properties of every-day speech to see how well responses to thesespeech-modeled stimuli were related to WRS.

Design: We separately measured WRS and IAFMresponses at a stimulus intensity of 70 dB SPL inthree groups of subjects: young normal-hearing,elderly normal-hearing, and elderly hearing-im-paired. We used two series of IAFM stimuli, onewith modulation frequencies near 40 Hz and theother with modulation frequencies near 80 Hz. TheIAFM stimuli, consisting of four carrier frequencieseach independently modulated in frequency andamplitude, could evoke up to eight separate re-sponses in one ear. We recorded IAFM responsesand WRS measurements in quiet and in the pres-ence of speech-masking noise at 67 dB SPL or 70 dBSPL. We then evaluated the hearing-impaired sub-jects with and without their hearing aids to seewhether an improvement in WRS would be re-flected in an increased number of responses to theIAFM stimulus.

Results: The correlations between WRS and thenumber of IAFM responses recognized as signifi-cantly different from the background were between0.70 and 0.81 for the 40 Hz stimuli, between 0.73 and0.82 for the 80 Hz stimuli, and between 0.76 and 0.85for the combined assessment of 40 and 80 Hz re-sponses. Response amplitudes at 80 Hz were smallerin the hearing-impaired than in the normal-hearingsubjects. Response amplitudes for the 40 Hz stimulivaried with the state of arousal and this effect madeit impossible to compare amplitudes across the dif-ferent groups. Hearing aids increased both the WRSand the number of significant IAFM responses at 40Hz and 80 Hz. Masking decreased the WRS and thenumber of significant responses.

Conclusions: IAFM responses are significantly cor-related with WRS and may provide an objective toolfor examining the brain’s ability to process theauditory information needed to perceive speech.

(Ear & Hearing 2004;25;68–84)

Auditory steady-state evoked responses (ASSRs)evoked by modulated tones have been extensivelyevaluated as an objective means to estimate behav-ioral pure-tone thresholds (reviewed by Picton,John, Dimitrijevic, & Purcell, 2003). Most recentstudies have used modulation frequencies in the 80to 100 Hz range (Cohen, Rickards, & Clark, 1991),since responses at lower frequencies (e.g., near 40Hz) are difficult to record in infants (Stapells, Ga-lambos, Costello, & Makeig, 1988) and in sleep(Jerger, Chmiel, Frost, & Coker, 1986; Linden,Campbell, Hamel, & Picton, 1985). The differencesbetween the ASSR thresholds and behavioral pure-tone thresholds are generally between 5 and 15 dB(Dimitrijevicet al., 2002; Herdman & Stapells, 2001,2003; Perez-Abalo, Savio, Torres, Martin, Rodri-guez, & Galan, 2001; Rance & Rickards, 2002).

Although hearing loss is commonly assessed us-ing pure-tone thresholds, the most debilitating as-pect of a hearing loss is difficulty in speech percep-tion. Speech perception is typically measured bypresenting phonetically balanced monosyllabicwords at 30 to 50 dB above the subject’s pure-tonethreshold and scoring the percentage of words thatare correctly recognized—the word recognition score(WRS) (Brandy, 2002; Martin, 1998). A necessaryfirst step in the perception of a word is to discrimi-nate changes in the frequency and amplitude of asound. For example, the /ga/ and /da/ sounds differonly in the rapid frequency-change of the secondformant at the beginning of the syllable and the /da/and /ta/ sounds differ only in when the amplitude ofthe voicing increases.

The ability of the brain to detect changes infrequency and amplitude can be assessed by record-ing ASSRs to modulations in the frequency andamplitude of supra-threshold tones. The ASSRs maythus provide an objective measurement of the su-prathreshold processes needed in the initial stages

Rotman Research Institute (M.S.J., T.W.P.), Baycrest Centre forGeriatric Care, University of Toronto, Canada; andSchool of Audiology & Speech Sciences (A.D.), University ofBritish Columbia, Vancouver, Canada.

DOI: 10.1097/01.AUD.0000111545.71693.48

0196/0202/04/2501-0068/0 • Ear & Hearing • Copyright © 2004 by Lippincott Williams & Wilkins • Printed in the U.S.A.

68

of speech perception. Behavioral thresholds for de-tecting the amplitude modulation (AM) or frequencymodulation (FM) of tones correspond reasonablywell to the thresholds for recognizing ASSRs evokedby AM or FM in young normal-hearing subjects(John, Dimitrijevic, van Roon, & Picton, 2001; Pic-ton, Skinner, Champagne, Kellett, & Maiste, 1987).

The current study evaluated how suprathresholdASSRs relate to measurements of WRS. Our basicgoal was to gain some objective information aboutthe brain’s ability to discriminate sounds. An objec-tive test of suprathreshold hearing would be benefi-cial in examining a child’s ability to discriminate thechanges in amplitude and frequency needed forspeech perception and in evaluating the perfor-mance of hearing aids.

Our previous work has shown that ASSRs arerelated to WRS in normal-hearing young adults(Dimitrijevic, John, van Roon, & Picton, 2001). Inthat study, ASSRs were elicited by independentamplitude and frequency modulated (IAFM) puretones. An IAFM stimulus consists of a carrier that ismodulated in amplitude and frequency, with differ-ent rates of modulation for the AM and FM. Sincethe AM and FM occur at different rates, the result-ing IAFM stimulus will elicit a separate response foreach type of modulation. Multiple IAFM stimuli canbe presented using the MASTER (Multiple AuditorySTEady-state Response) system (John & Picton,2000). Multiple IAFM stimuli have four differentcarriers, each with the AM and FM componentsmodulated at different rates, resulting in a total ofeight possible responses.

In our initial study, IAFM stimuli were presentedat intensities varying from 20 to 70 dB SPL with thedepth of AM at 50% and the depth of FM at 20%.WRS were correlated with both the amplitude of theIAFM response (r � 0.68) and the number of signif-icant IAFM responses (r � 0.74). The current studysought to extend the previous findings using stimu-lus parameters more closely based on the acousticproperties of everyday speech. We also used stimu-lus modulation frequencies near 40 Hz as well as the80 to 100 Hz of the initial study. Responses elicitedat 80 to 100 Hz arise mainly from the brain stemwhereas responses to modulations near 40 Hz aregenerated in both the primary auditory cortex andbrain stem (Herdman, Lins, van Roon, Stapells,Scherg, & Picton, 2002).

The present study also evaluated the effect ofhearing aids on WRS and IAFM steady-state re-sponses. If increases in WRS and IAFM detectioncan be shown while wearing a hearing aid, IAFMstimuli potentially could be used to evaluate theperformance of hearing aids. It is difficult to evalu-ate the effects of hearing aids using the auditory

brain stem responses (ABR) since the stimuli thatare normally used for ABRs are too brief to fullyactivate the hearing aid’s circuitry (Gorga,Beauchaine, & Reiland, 1987). The ASSR provides apotential solution to this since a steady-state stim-ulus is a continuous sound with no rapid transients.The ASSR can be used to reliably estimate pure-tonethresholds in children when the stimulus is pre-sented through loud speakers and amplified by ahearing aid (Picton et al., 1998). We have also shownthat aided IAFM responses can be recorded in hear-ing-impaired adults and the responses were relatedto WRS (Picton, Dimitrijevic, van Roon, John, Reed,& Finkelstein, 2002). In that study phoneticallybalanced monosyllabic words and multiple IAFMstimuli (50% AM, 20% FM) were presented at themost comfortable listening level (MCL) and at 10and 20 dB below MCL in normal-hearing and hear-ing-impaired adults with and without their hearingaids. The overall correlation between the number ofsignificant IAFM responses and WRS was 0.55.

The objectives of this study were to confirm thatthe relationship between IAFM responses and WRSin normal-hearing and hearing-impaired subjects, toexamine the effects of masking noise on IAFMresponses, and to determine whether the improve-ment in the WRS of hearing-impaired subjects wear-ing hearing aids was related to an improved recog-nition of the IAFM responses.

METHODS

Subjects

Three groups of subjects participated in the ex-periment. A young normal-hearing group consistedof 10 subjects (mean age 28 yr; range 22 to 33; fivemen). An elderly normal-hearing group consisted of10 subjects (mean age 68; range 60 to 82; four men).All these subjects reported normal hearing. Thepure-tone averages (500, 1000, and 2000 Hz) werebelow 25 dB HL for all the young subjects and for 9out of the 10 elderly subjects, with one elderlysubject having a pure-tone average of 43 dB HL.This particular elderly subject was included in theIAFM and WRS correlations but not in other com-parisons (e.g., effects of age). Some of these elderlynormal-hearing subjects had a high-frequency hear-ing loss at 4000 Hz. Given the difficulty of obtainingelderly subjects with “normal” pure-tone thresholdsin the high-frequency range, we felt that this partic-ular subset of subjects were representative of anelderly population without hearing problems. Thenormal-hearing subjects were volunteers obtainedfrom laboratory personnel and an elderly subjectpool. An elderly hearing-impaired group consisted of10 subjects (mean age 75; range 57 to 86; four men)

EAR & HEARING, VOL. 25 NO. 1 69

who used hearing aids. These subjects were re-cruited from a local audiological clinic. Six of thesesubjects had monaural hearing aids. Nine subjectshad analog hearing aids, eight with compressioncircuitry. In conditions where testing was done withthe hearing aid, the subjects were asked to use thenormal everyday settings of their hearing aids. Forthe one subject who normally used a noise-reductionsetting in his hearing aid, this option was turned offfor the experiment. All subjects had typical slopingaudiograms with the greatest loss at high frequen-cies. A summary of the behavioral pure-tone thresh-olds is shown in Table 1.

Word Recognition Scores (WRS)

WRS was measured using Auditec recordings ofW-22 and NU-6 word lists (Brandy, 2002). Fiftywords from a list were presented at 70 dB SPLthrough two free-field speakers. Subjects wereseated in a reclining chair 2 meters from the speak-ers, which were located at head level, 45° fromstraight ahead. The non-test ear was occluded withfoam earplugs. Subjects wrote out the words, andthe written lists were checked at the end of theexperiment to give the WRS, expressed as the per-centage of words correctly identified.

IAFM Stimuli

An IAFM stimulus has a single carrier that isamplitude-modulated at one rate and frequency-modulated at a different rate (Dimitrijevic et al.,2001). The Multiple Auditory STEady-state Re-sponse (MASTER) system (John & Picton, 2000)allows for simultaneous presentation and evalua-tion of multiple stimuli. In the current study, IAFM

stimuli consisted of four carrier frequencies thatwere AM and FM modulated at different rates,giving eight total responses (four for AM and four forFM). Two series of modulations, 40 Hz IAFM and 80Hz IAFM, were tested separately. The AM and FMparameters of the stimuli were selected as represen-tative of the typical amplitude and frequencychanges that occur during speech. Details of howthese parameters were selected and calibrated aregiven in the appendix. The actual stimuli are de-scribed in Table 2, and their spectra are shown inFigure 1. For the 40 Hz stimuli the AM and FMstimuli alternated in their modulation frequencieswith the FM set 2 Hz higher than the AM of thesame carrier, whereas for the 80 Hz stimuli the rateof the FM was 6 Hz greater than the AM of the samecarrier. The combined stimuli were presented at anintensity of 70 dB SPL through the two free-fieldspeakers.

Recordings

Electrophysiological responses were recordedfrom gold-plated Grass electrodes at the vertex andposterior neck (7 to 8 cm below the inion). Anelectrode placed over the left clavicle served asground. All electrode impedances were under 5kOhm at 10 Hz. Recordings occurred in a soundprooftesting chamber with the subjects lying in a reclin-ing chair. Because sleeping subjects produce lessbackground noise, we asked our subjects to try tosleep during the experimental sessions. The re-sponses were amplified using a Grass P55 battery-powered amplifier with a filter band pass of 1 to 300Hz (6 dB/octave), and a gain of 10,000. The notchfilter was turned off.

The MASTER data acquisition system (John &

TABLE 1. Behavioral pure-tone thresholds

500 Hz 1000 Hz 2000 Hz 4000 Hz

Young normal-hearing 7 � 9 (0–30) 3 � 7 (�5–15) 3 � 7 (�5–15) 7 � 8 (0–20)Elderly normal-hearing 9 � 14 (�5–30) 15 � 11 (0–30) 17 � 12 (�10–25) 22 � 15 (0–45)Elderly hearing-impaired (unaided) 36 � 14 (10–60) 43 � 15 (20–70) 45 � 12 (20–60) 56 � 7 (45–65)Elderly hearing-impaired (aided) 32 � 13 (10–60) 33 � 13 (10–60) 30 � 12 (10–55) 44 � 12 (25–65)

Mean � SD, ranges in parentheses.

TABLE 2. IAFM stimuli

Carrier Frequency(Hz)

Amplitude Modulation Frequency Modulation

AmplitudeFrequency (Hz) Depth (%) Frequency (Hz) Depth (%)

500 39 78 55 41 85 35 401500 43 83 50 45 90 30 102500 47 88 45 49 95 20 24000 51 93 35 53 100 35 1

Amplitude is given in arbitrary units.

70 EAR & HEARING / FEBRUARY 2004

Picton, 2000) used an A-D conversion rate of 1 kHzwith 16-bit precision. Consecutive data epochs of1.024 seconds were linked together to form sweeps of16.384 seconds, which then were submitted to anFFT to produce an amplitude spectrum with a res-olution of 0.061 Hz. Epochs containing electrophys-iological activity over �90 �V were rejected and thenext recorded epoch was used to continue the sweep.Recordings were performed after averaging 48sweeps of data (approximately 13 minutes). Record-ings were subjected to an off-line weighted averag-ing procedure to minimize noise in the recordings(John, Dimitrijevic, & Picton, 2001). In a proportionof our cases (70 out of 180) some of the responseswere significant at the p � 0.1 level but not at the p� 0.05 level after averaging 48 sweeps. In theseinstances the recordings were extended by fivesweeps. If the responses still remained nonsignifi-cant (p � 0.05) after the additional sweeps, therecording was terminated and the response wasdeemed nonsignificant.

The MASTER system evaluated the evoked po-tentials in the frequency domain. To determinewhether the FFT components at the stimulus mod-ulation-frequencies were different from backgroundEEG activity, the amplitude at each of these fre-quencies was compared with the amplitudes in the120 adjacent frequencies (60 bins above and 60below the stimulus frequency, or �3.7 Hz, excludingthose frequencies at which other stimuli were mod-ulated) using an F-ratio (John & Picton, 2000).Comparing this ratio against the critical values for Fat 2 and 240 degrees of freedom gives the probabilityof a response being within the distribution of thebackground noise.

IAFM responses were evaluated in two differentways. One was to consider the number of responses

present in the recording. A response (i.e., signal) waspresent if the response amplitude was significantly(p � 0.05) different than the amplitude at adjacentfrequencies (i.e., noise estimate). IAFM responsesscored in this manner were expressed as a percent-age of the total number of possible response (eightpossible IAFM responses for each of the 40 and 80Hz stimuli). Combining the 80 and 40 Hz responsesgave a total of 16 possible responses.

The second manner of evaluating IAFM re-sponses was by measuring their amplitudes. Com-bined IAFM response amplitudes consisted of 1) thesummed total of all responses (eight amplitudesadded); 2) the summed AM (four carriers added); or3) the summed FM (four carriers added) response.Additionally, summed 80 and 40 Hz IAFM responseswere also calculated. In this case, the 40 Hz IAFMresponses were multiplied by a weighting factor,since 40 Hz responses are typically larger than 80Hz responses and equal weighting would have beenbiased by the 40 Hz responses. The weighting factorwas ratio between the summed 80 Hz IAFM re-sponses and the summed 40 Hz IAFM responses foreach subject group over all of the experimentalconditions (quiet and masking). The 40 Hz responseamplitudes were then multiplied by this factor andadded to the 80 Hz response amplitude for eachcarrier to give a combined IAFM response. EachIAFM response had an associated noise level, whichis the root-mean-square amplitude of the spectrum3.7 Hz above and below signal frequency. The noiselevels reported in the current study are the averageof the eight noise levels for each of the signals.

Experimental Design

All behavioral and IAFM testing was performedmonaurally with the test ear being chosen randomlyfor the normal-hearing groups. An equal number ofleft and right ears were tested in both normal-hearing groups. The test ear in the hearing-im-paired group was the ear with the hearing aid incases where the subject had one hearing aid and atest ear was randomly chosen in subjects with hear-ing aids in both ears (number of left ears � 5). WRSand IAFM responses were evaluated in three condi-tions in young normal-hearing adults and elderlynormal-hearing adults. The first condition was lis-tening in quiet with the stimuli presented at 70 dBSPL. In the second condition, spectrally shapedspeech noise at 67 dB SPL was added. All of subjectswere tested in these first two conditions with thehearing-impaired subjects not wearing a hearingaid. The normal-hearing young and elderly subjectswere tested in a third condition with 70 dB SPLmasking noise to decrease intelligibility. Addition-

Figure 1. Frequency and intensity characteristics of the 40 HzIAFM stimuli (black) and the speech noise (grey). The overallintensity of both was 70 dB SPL with linear weighting.


ally, the hearing-impaired subjects were tested us-ing their hearing aids and no masking.

Sleep Staging

All of the subjects were encouraged to sleep. Sincenot all of the subjects slept through the entireexperiment, we evaluated the state of arousal dur-ing each recording. Sleep reduces the amplitude ofthe 40 Hz ASSR (Jerger et al., 1986; Linden et al.1985). However, sleep also reduces the backgroundnoise in the recordings and can therefore lead to amore favorable signal-to-noise ratio (Dobie & Wil-son, 1998). An off-line analysis was performed toestimate how much of the recordings occurred whilethe subjects slept. Although proper sleep stagingrequires multiple channels (Rechtschaffen & Kales,1968), our single channel allowed us to recognize thecharacteristic markers of sleep. The state of thesubject for a particular recording was determined bythe most common sleep stage during the recording.The waking stage was characterized by alpha waveactivity and overt muscle artifacts. Stage I or drows-iness showed discontinuous alpha activity (less thanhalf the recording) and the presence of theta waves(4 to 7 Hz). Stage II sleep showed occasional deltaactivity, vertex sharp waves, K-complexes or spin-dles. To assess the effects of sleep on our measure-ments, we compared the recordings during wakeful-ness with those during Stage I and II collapsedtogether.

Statistical Evaluations

WRS, the number of significant IAFM responses,and mean IAFM amplitudes in sleep and wake werecompared across different groups or across differentconditions within a group using ANOVAs or t-tests.Post hoc analyses were performed with Fisher leastsignificant differences. The amplitudes for AM andFM responses were also evaluated using a repeatedmeasures ANOVA (modulation type [AM/FM] �carrier frequency). Greenhouse-Geisser correctionswere used when appropriate. Correlations betweenWRS and IAFM responses were examined usingPearson product-moment correlation coefficients.Multiple linear regression analysis was performed

using averaged AM and FM response amplitudes aspredictor values for WRS.

RESULTS

Word Recognition Scores

The mean WRS in the different groups are pre-sented in Table 3. There was a significant differencein WRS between the three subject groups of subjectsin the quiet condition (F � 10.25; d.f. � 2,27; p �0.001). Post hoc tests showed that this was due tothe decreased WRS of the hearing-impaired sub-jects. A 2-way ANOVA (elderly/young � condition[quiet/67 dB masking/70 dB masking]) showed thatyoung normal-hearing subjects had higher WRSthan the elderly (F � 11.82; d.f. � 1,17; p � 0.005)and that masking significantly decreased WRS (F �285.69; d.f. � 2,34; p � 0.001, � � 0.94). A significantinteraction between subject-group and maskingshowed that elderly subjects were more susceptibleto the effects of masking (F � 7.15; d.f. � 1,17; p �0.05, � � 0.80). Elderly normal-hearing and hearing-impaired (unaided) WRS were compared using arepeated measures ANOVA (subject group � condi-tion [quiet/masking at 67 dB SPL]). Hearing-im-paired subjects had significantly lower WRS (F �22.25; d.f. � 1,17; p � 0.001). Masking significantlydecreased WRS (F � 100.25; d.f. � 1,17; p � 0.001),but there was no significant interaction. In thehearing-impaired subjects, wearing a hearing aidsignificantly increased WRS (t � 2.77; p � 0.05).

IAFM Responses

Most of the normal-hearing subjects showed clearresponses to most of the IAFM stimuli when thesewere presented in quiet. Figures 2 and 3 illustratethe grand mean 40 and 80 Hz IAFM responses in thequiet condition. Figure 4 illustrates sample resultsfrom individual subjects with their correspondingWRS.Sleep Effects • The elderly subjects were less ableto sleep through the recordings that the youngsubjects. The young subjects slept through 62% oftheir recordings whereas the elderly subjects sleptthrough only 43% of their recordings (50% for thenormal-hearing elderly and 37% for the hearing-

TABLE 3. Word recognition scores

Quiet 67 dB Masking 70 dB Masking

Young normal-hearing 97 � 5 62 � 12 38 � 18Elderly normal-hearing 97 � 2 43 � 13 17 � 9Elderly hearing-impaired (unaided) 56 � 30 17 � 10 —Elderly hearing-impaired (aided) 76 � 18 — —

Percentages, mean � SD.


impaired elderly). The effects of sleep on the ampli-tude and the number of significant IAFM responseswere examined for the 40 Hz and 80 Hz IAFMamplitudes in young normal-hearing and elderlynormal-hearing subjects. In the quiet condition, 3out of 10 of the young normal-hearing subjects wereawake and six out of nine of the elderly normal-hearing subjects were awake. The mean responseamplitudes for the young and elderly normal-hear-ing subjects in sleep and wake are given in Table 4.The amplitude of the 40 Hz responses was signifi-

cantly lower during sleep (F � 5.63, d.f. � 1,15, p �0.05) but not significantly different between theyoung and elderly groups. The 40 Hz noise levelsalso decreased significantly during sleep (F � 9.36,d.f. � 1,15, p � 0.01). Signal and noise amplitudesfor the 80 Hz IAFM stimuli did not change signifi-cantly with sleep. For both the 40 or the 80 Hzresponses, the number of significant responses didnot change with sleep or age.Correlations between WRS and the Number ofSignificant IAFM Responses • Figure 5 plots thepercentage WRS versus the percentage of significantIAFM responses (combined assessment of 40 and 80Hz) for all of the subject groups and for each subjectgroup separately. Table 5 summarizes the percent-age of significant IAFM responses for all the groupsand conditions and Table 6 provides the correlationsbetween WRS and number of significant responses.Effects of Age, Hearing Loss and HearingAids • The number of significant 40 and 80 HzIAFM responses in the quiet condition was slightlygreater in the young normal-hearing than in theelderly normal-hearing subjects but this differencefailed to reach significance. The number of signifi-cant 40 and 80 Hz IAFM responses was greater inthe elderly normal-hearing than in the elderly un-

Figure 2. Grand mean vector-averaged 40 Hz IAFM responsesin young normal-hearing, elderly normal-hearing, and elderlyhearing-impaired. AM responses are indicated by squares,and FM responses by triangles. Higher carrier frequenciesalso have higher modulation frequencies. Filled and opensymbols indicate significant and nonsignificant responses,respectively.

Figure 3. Grand mean vector averaged 80 Hz IAFM responsesin young normal-hearing, elderly normal-hearing, and elderlyhearing-impaired. AM responses are indicated by squares andFM responses by triangles. Filled and open symbols indicatesignificant and nonsignificant responses, respectively.

Figure 4. Individual subject IAFM responses. Shown are theWRS and the 40 and 80 Hz IAFM in three subjects, one fromeach of the subject groups: young normal-hearing, elderlynormal-hearing, and elderly hearing-impaired are shown. AMresponses are indicated by squares, and FM responses bytriangles. Filled and open symbols indicate significant andnonsignificant responses, respectively.

TABLE 4. Effects of sleep on the 40 Hz IAFM responses

AM Amplitudes(nV)

FM Amplitudes(nV)

Awake Sleep Awake Sleep

Young normal-hearing 58 35 84 43Elderly normal-hearing 66 48 66 55


aided hearing-impaired subjects (p � 0.05). Thenumber of significant IAFM responses increasedwhen the hearing-impaired subjects wore their hear-ing aids (40 Hz: p � 0.001, t � 7.53; 80 Hz: p � 0.05,t � 2.75).Effects of Masking • Masking decreased the num-ber of significant IAFM responses (Table 5) for boththe 40 Hz stimuli (F � 70.31; d.f. � 1,26; p � 0.001)and the 80 Hz stimuli (F � 44.44; d.f. � 1,26; p �0.001). An interaction between subject group andmasking was observed for both the 40 Hz stimuli (F� 7.73; d.f. � 2,26; p � 0.005) and the 80 Hz stimuli(F � 6.86; d.f. � 2,26; p � 0.005), with the normal-hearing subjects having the greatest reduction inthe number of significant responses.Response Amplitudes • The response amplitudesare tabulated in Tables 7 to 9. Due to the confound-ing effects of sleep on the amplitudes of the 40 Hzresponses, only the amplitudes of the 80 Hz re-sponses were analyzed statistically. The 40 Hz datain these tables should be interpreted with cautionsince these changes in amplitudes might have beenrelated to changes in the state of arousal. This mighthave affected the comparisons between the youngand the elderly but would not have affected thecomparisons within the elderly subjects. Figure 6illustrates mean 80 Hz IAFM response amplitudesacross all three subject groups.

To show an overall relationship between response

amplitudes and WRS, each subject’s AM and FMresponse amplitudes were summed to give an IAFMresponse amplitude. Table 8 provides a summary ofthe correlations between response amplitudes andWRS. Multiple linear regression analysis using 80Hz AM and FM response amplitudes as predictorvariables showed that AM and FM had significantcontributions to WRS (F � 28.1; d.f. � 2,87; p �0.001). The contribution of AM (standardized betacoefficient � 0.36; p � 0.01) was slightly greaterthan FM (standardized beta coefficient � 0.34; p �0.02).

The effects of age were examined by comparingyoung and elderly normal-hearing subjects in thequiet condition (Table 9). FM responses at 80 Hzwere generally larger than AM responses (F �13.87; d.f. � 1,17; p � 0.005) and the amplitudedecreased with increasing carrier frequency(F �17.36; d.f. � 3,51; p � 0.001, � � 0.42). FM responseswere larger than the AM responses for youngersubjects but not for the older subjects (F � 5.85; d.f.� 1,17; p � 0.05). The amplitudes of the 80-Hzresponses were reduced by masking (F � 54.70; d.f.� 1,9; p � 0.001), and there was no interactionbetween age and masking.

The effects of hearing loss were examined bycomparing elderly normal-hearing and hearing-im-paired subjects in the quiet condition. A repeatedmeasures ANOVA (hearing loss � modulation type

Figure 5. IAFM and WRS correlations. Thenumber of significant IAFM responses using acombined assessment of 40 and 80 Hz (per-centage out of 16) are plotted against thecorresponding WRS. Larger symbols indicateoverlapping data points. The upper left graphshows the results for all subjects in all condi-tions. The upper right graph shows the resultsin the young normal-hearing subjects. Thelower graphs show the results for the elderlysubjects with the normal-hearing subjects onthe left and the hearing-impaired on the right.

TABLE 5. Percentage of significant responses

Subjects


40 Hz 80 Hz 40 Hz 80 Hz 40 Hz 80 Hz

Young normal-hearing 76 83 41 43 33 31Elderly normal-hearing 76 74 44 39 36 24Elderly hearing-impaired (unaided) 43 28 34 21 — —Elderly hearing-impaired (aided) 70 46 — — — —


[AM/FM] � carrier frequency) showed a main effectof hearing loss for the 80 Hz stimuli (F � 13.33; d.f.� 1,17; p � 0.005) where the hearing-impairedsubjects had lower amplitudes. No interaction be-tween hearing-loss and modulation type wasobserved.

Although 80 Hz IAFM amplitudes were larger inthe aided condition, this overall difference just failedto show significance (p � 0.084). However, an inter-action between hearing aids and carrier frequency(F � 3.83; d.f. � 3,27; p � 0.05, � � 0.80) indicatedthat the hearing aid resulted in greater increases inresponse amplitude for the 1500 Hz carrier fre-quency (40%) than the other carrier frequencies (20to 30%).

DISCUSSION

Behavioral Tests of Speech Discrimination

The most common clinically used test of supra-threshold auditory discrimination is the WRS formonosyllabic words (Martin, 1998). However, thismeasurement may be difficult to interpret for manyreasons. Since the perception of continuous speech isquite different from isolated words, the WRS maynot relate that closely to the subject’s everydayfunctioning. Furthermore, since language and mem-ory interact with auditory perception in the WRS, itis often difficult to determine how much of thereduction in WRS is solely due to auditory problems.The choice of the intensity at which to present thestimuli has never been standardized. The level atwhich one attains maximum WRS may not be pre-dictable from the pure-tone audiogram, or measuressuch as Maximum Comfort Level (Beattie & Zipp,1990; Ullrich & Grimm, 1976). The use of perfor-

mance-intensity (PI) functions has been advocatedsince the plateau of PI functions are often unpredict-able particularly in eighth nerve and central disor-ders (Jerger & Hayes, 1977).

Although our young and elderly normal-hearingsubjects had equivalent WRS in quiet conditions, theelderly group performed significantly worse in thepresence of masking noise. Similar results have beenreported in other studies (Dubno, Dirks, & Morgan,1984, Frisina & Frisina, 1997; Pichora-Fuller, Schnei-der, & Daneman, 1995). The effect of masking waseven greater in the hearing-impaired, confirming pre-vious studies (Beattie, Barr, & Roup, 1997; Dubno etal., 1984; Frisina & Frisina, 1997). One reason for thisdecreased performance in hearing-impaired patientsis that sensorineural patients have proportionally lessspeech signal available for processing (Van Tasell,1993). Additionally, wider auditory filters cause spec-tral smearing making it harder to distinguish thespectral peaks and valleys that are normally presentin speech. The situation is even more difficult in noisyconditions where noise acts to fill the valleys (Leek,Dorman, & Summerfield, 1987).

The word list approach to assessing the ability ofa subject to perceive speech assesses the perceptionof a representative set of phonemes. More specificapproaches might measure the subject’s ability toperceive particular phonemes or types of phonemes.For example, a subject with a high-frequency losswill have particular difficulty with fricatives. Thepresent study did not use such high-resolution cor-relations, but future studies may be able to relateparticular phoneme difficulties to particular audio-metric configurations and/or particular abnormali-ties of the IAFM response.

TABLE 6. Correlations between the number of significant IAFMresponses and WRS

Subjects

Frequency of Stimulation

40 & 80 Hz 40 Hz 80 Hz

Young normal-hearing 0.76 0.70 0.73Elderly normal-hearing 0.85 0.81 0.82Elderly hearing-impaired 0.75 0.71 0.65All subjects 0.78 0.71 0.73

TABLE 7. Mean IAFM response amplitudes (nV)


40 Hz 80 Hz 40 Hz 80 Hz 40 Hz 80 Hz

Young normal-hearing 48 26 22 12 20 10Elderly normal-hearing 59 25 30 12 26 10Elderly hearing-impaired (unaided) 56 15 34 13 — —Elderly hearing-impaired (aided) 78 21 — — — —

TABLE 8. Correlations between IAFM response amplitudes andWRS

40 & 80 Hz 40 Hz 80 Hz

Young normal-hearing 0.74 0.66 0.72Elderly normal-hearing 0.77 0.67 0.73Elderly hearing-impaired 0.48 0.40 0.52All subjects 0.62 0.39 0.62

40 and 80 Hz amplitudes are weighted (see text).


IAFM Responses as a Possible Objective Testof Suprathreshold Discrimination

At the present time, there are no objective audio-logical measures of a subject’s ability to perceivespeech. “Objectiveness” consists of two parts: nobehavioral response should be required from thesubject, and no subjective interpretation should berequired from the experimenter. Previous work fromour laboratory has indicated that ASSRs to IAFMstimuli may be related to WRS (Dimitrijevic et al.,2001, Picton et al., 2002). In these studies the IAFMand speech stimuli were presented at different in-tensities, whereas in the current study, we used amasking paradigm to vary WRS and IAFM.

IAFM responses were clearly recognizable in thenormal-hearing subjects. There were no significantdifferences in the number of significant IAFM re-sponses at 40 and 80 Hz. A consistent finding amongall the subjects was that the lower carrier frequen-cies evoked larger responses than the higher carrierfrequencies. This is likely due to the amplitudes of

our stimuli being larger for the lower carrier fre-quencies. The amplitudes were based on the long-term average speech spectrum, which has greateramplitude at lower frequencies.

Effects of Sleep

Sleep significantly decreased the amplitudes ofthe 40 Hz IAFM responses but did not significantlyaffect the 80 Hz responses. These results are consis-tent with previous studies that have also shown thatsleep decreases the 40 Hz response amplitude(Jerger et al. 1986; Linden et al., 1985), but not theresponses at modulation rates greater than 70 Hz(Cohen et al., 1991; Dobie & Wilson, 1998; Lins &Picton, 1995). The reduction in the 40 Hz responseamplitudes during sleep was greater in the youngersubjects than in the elderly (Table 4). This may bepartly related to difficulty in recognizing the sleepstages in the elderly. Our sleep staging depended onthe occurrence of sleep spindles and K-complexes,

TABLE 9. Mean AM and FM amplitudes


40 Hz 80 Hz 40 Hz 80 Hz 40 Hz 80 Hz

AM FM AM FM AM FM AM FM AM FM AM FM

Young normal-hearing 42 55 17 34 24 21 8 15 20 19 8 12Elderly normal-hearing 60 59 23 27 35 26 11 15 30 22 9 13Elderly hearing-impaired (ua) 63 52 15 15 37 35 10 15 — — — —Elderly hearing-impaired (a) 87 72 21 22 — — — — — — — —

Figure 6. Mean 80 Hz IAFM response amplitudes for all the subject groups in the quiet conditions. Left panel shows the AMresponses and the right shows the FM responses. Error bars are � standard error of the mean.


which are known to be reduced in amplitude anddensity in the elderly (Crowley, Trinder, & Colrain,2002; Crowley, Trinder, Kim, Carrington, & Colrain,2002). The elderly may have had a reduced effect ofsleep on the IAFM responses or they may have beenintermittently awake during the recordings that weclassified as sleep.

The average background noise was also signifi-cantly decreased during sleep. This was expectedsince there is reduced muscle activity during sleep(Rechtschaffen & Kales, 1968). Since both the signalamplitude and noise amplitude of the 40 Hz re-sponses decreased by similar amounts, the numberof significant responses did not change as a functionof sleep. This supports previous steady-state thresh-old findings showing that sleep does not affect thedetectability of the responses since both the signaland noise decrease with sleep (Dobie & Wilson,1998; Linden et al., 1985). We therefore felt justifiedin using the number of significant responses forcomparison across groups, even though the stage ofsleep varied across the groups.

Correlations with WRS

The results of this study agree with our previouswork comparing IAFM and WRS (Dimitrijevic et al.,2001; Picton et al., 2002). In our initial study inyoung normal-hearing adults, 80 Hz IAFM stimuliand WRS were presented at intensities varying from20 to 70 dB SPL. The relationship between WRS andthe number of significant responses was significantwith a correlation coefficient of 0.74. In the presentstudy, the correlation was 0.73 in the young normal-hearing adults (Table 6). These results are verysimilar despite the fact that the experimental pro-tocols were very different. First, the previous studyused six different intensities whereas the currentstudy used only one intensity with two levels ofmasking noise. Second, the first study was per-formed using insert earphones while the currentstudy used speakers presented in free field. Third,the IAFM parameters were different since the cur-rent study used speech modeled IAFM stimuli. De-spite these experimental differences, both studiesindicate a significant relationship between IAFMand WRS. In our second study (Picton et al., 2002),free field 80 Hz IAFM stimuli and WRS were pre-sented to young normal-hearing adults and hearing-impaired adults at three intensities correspondingto most comfortable listening levels (MCL), and 10and 20 dB below MCL. In that study, the correlationbetween WRS and the number of significant IAFMresponses was 0.31 in young normal-hearing adultsand 0.52 in the hearing-impaired group. The lowcorrelation between WRS and the IAFM responses

in the young normal-hearing subjects was likely dueto these subjects performing well on WRS even atthe lowest intensity, leaving only a small WRSrange for the correlations. Although the experimen-tal paradigms in this study were different in theprevious study (Picton et al., 2002), the correlationsfor the hearing-impaired groups are comparable.The hearing-impaired group in the current studyhad a correlation of 0.65. This improvement from0.52 is probably related to the stimulus parametersmore greatly approximating normal speech, al-though we cannot exclude the possibility that thedifference was due to the different stimulus intensi-ties. The average MCL in the previous study was 66dB SPL, compared with the constant 70 dB SPL ofthe present study.

One of the novel aspects of the current study isthe use of 40 Hz IAFM stimuli. In general the 80 HzIAFM correlations were slightly larger than the 40Hz IAFM correlations. However, the combined as-sessment of 40 and 80 Hz IAFM data resulted inhigher correlations than either the 40 Hz or 80 Hzalone. This improvement can be interpreted a num-ber of ways. First, the range of data for the correla-tion is larger (8 for 40 Hz or 80 Hz, and 16 for thecombined). Using a range of only eight when corre-lating to a behavioral response that can vary from 0to 100 has limitations since small changes IAFMdetectability will cause relatively large effects in thecorrelation. Furthermore, using two modulationranges may improve the correlations because differ-ent aspects of the modeled speech signal are repre-sented by the high and low modulation rates. Sinceconsonant-vowel transitions may have durations be-tween 5 and 50 msec (i.e., modulation frequencies of100 to 10 Hz) (Blumstein & Stevens 1980; Ohde &Abou-Khalil 2001), incorporating both 80 and 40 HzIAFM stimuli into the measurement may be morerepresentative of various aspects of speech. In addi-tion, since the 80 Hz response is mainly generated inthe brain stem and the 40 Hz response is generatedin both brain stem and auditory cortex (Herdman, etal., 2002), examining both the 40 and 80 Hz re-sponses may allow us to assess both cortical andsubcortical processing.

The current study used multiple stimuli in the 40Hz range. Interactions between stimuli resulting inattenuation of some responses occur at these modu-lation frequencies, but not near 80 Hz (John et al.,1998). These interactions may or may not be greaterin elderly or hearing-impaired subjects.

The stimuli used in ASSR studies have focused onmodulation rates near 40 and 80 Hz where thesignal to noise levels make the recordings moreefficient (Picton et al., 2003). The low signal to noiseratio at modulation rates of less than 30 Hz is


unfortunate since much of the speech envelope is inthis range. For example, Drullman, Festen, andPlomp (1994) showed that low pass filtering of thespeech envelope above 16 Hz had minimal effects onspeech recognition threshold. Moore and Alcantara(1996) found significant effects of modulation phaseon vowel perception when using modulation fre-quencies of 10 to 24 Hz. Studies relating speechperception with ASSRs at slower rates are war-ranted. However, they will require longer times toobtain significant responses. However, they mayrelate more specifically to the actual processes ofphoneme identification. Our present studies likelyrelate more to the processing of the frequency andintensity discrimination needed before phonemeidentification can be performed.

Effects of Age

Aging effects were examined by comparing youngnormal-hearing subjects with elderly normal-hear-ing subjects. Response amplitudes to the 40 HzIAFM stimuli were not fully examined because ofthe confounding effects of sleep. However, the 80 HzIAFM responses showed interesting effects. In theyoung normal-hearing subjects the FM responsewas greater than the AM response. Similar resultswere shown previously in young normal-hearingsubjects where the 50% FM response amplitude wasgreater than a 100% AM response (John et al.,2001). In contrast to the young subjects, the elderlynormal-hearing subjects did not show greater FMresponses compared with AM. This difference mightbe attributed to changes in auditory filter shapescurves with age. Previous work suggests that elderlysubjects with normal or mild hearing loss have poorfrequency resolution (He, Dubno, & Mills, 1998;Matschke, 1990; Patterson, Nimmo-Smith, Weber,& Milroy, 1982), whereas others have shown mini-mal changes in frequency resolution with age (Glas-burg & Moore, 1986). Along similar lines, elderlysubjects with normal hearing have increased thresh-olds for detecting mistuned harmonics as well aspoorer speech perception abilities in noise thanyounger subjects (Alain, McDonald, Ostroff, &Schneider, 2001). A steeply shaped tuning curvewould result in larger FM responses than a broadlytuned curve because of the FM activation pattern inthe cochlea. An FM stimulus will move locus ofbasilar-membrane activation back and forth at therate of modulation. A more sharply tuned cochleawill elicit larger responses since more neurons willbe stimulated by this movement of the activationpattern, than when the tuning curves are broad andshallow.

Previous work comparing the effects of age on the

40 Hz AM responses found no differences for both520 and 4000 Hz carriers at various modulationdepths modulation (Boettchner, Poth, Mills, &Dubno, 2001). In the current study, however, elderly40 Hz AM responses were slightly greater than theyounger subjects (Tables 4 and 9). In contrast to theAM responses, previous work has shown that FMASSRs at modulation depths greater than 40% andat modulation rates near 40 Hz were greater inelderly subjects than young subjects (Boettcher,Madhotra, Poth, & Mills, 2002). In the currentstudy, mean 40 Hz FM responses were slightlylarger in the elderly but this difference was notsignificant. Since frequency difference limens tendto be larger in normal-hearing elderly for lowercarriers (He et al., 1998), one might expect that theFM response should be smaller. One possible expla-nation for this discrepancy is the changes in theauditory system associated with age. Recordings ininferior colliculus to AM tones from aged miceshowed that a larger number of units responded tomodulation frequencies of 0 to 200 Hz comparedwith young adult mice (Walton, Simon, & Frisina,2002). Some of these changes were associated withchanges in the inhibitory neurotransmitter, gamma-aminobutyric acid (GABA) with age (Walton, Frisina,& O’Neill, 1998). However, we cannot rule out theeffects of sleep since the young subjects’ 40 FM re-sponses were greater in the waking state (Table 4).

A significant interaction for age was observedwhere young normal-hearing subjects had largerFM responses than AM responses, an effect notobserved in the elderly normal-hearing subjects.This may reflect differences in the auditory system’sAM and FM processing systems that change withage. One way of examining these differences couldbe the use of an FM/AM ratio. The advantage ofusing this type of analysis is that a ratio couldcompare relative FM and AM processing in the samesubject and would be independent of skull thickness,subject arousal level, and electrode position. Thedata in Figure 6 indicate that the FM/AM ratio issmaller in the elderly and decreases further withhearing loss.

Effects of Masking

The masking noise decreased amplitude of theIAFM responses and made fewer of them recogniz-able. This fits well with the effects of masking on theWRS. Adding masking noise and decreasing inten-sity are independent ways to manipulate WRS. Ourprevious work had shown significant correlationswith intensity (Dimitrijevic et al., 2001). The rela-tionship between WRS and IAFM is more convincing


now that we have shown that it also occurs withmasking.

However, the effects of the masking on the IAFMresponses of the elderly subjects was similar to theeffects on the young subjects. This was not what mighthave been expected since the WRS were much moreaffected by masking in the elderly. Young normal-hearing subjects were better able to cope with maskingbehaviorally than the elderly subjects, although thebrain stem and primary auditory cortex were similarlyaffected by masking. One possible reason for thisdiscrepancy may be related to the fact WRS and IAFMare examining different processes. IAFM examineshow the characteristics of speech stimuli are processedin terms of AM and FM. WRS on the other hand,requires additional processes beyond the primary au-ditory cortex, particularly when signals are presentedin noise. A recent study by Salvi, Lockwood, Frisina,Coad, Wack, and Frisina (2002) showed that in youngnormal-hearing subjects, the addition of noise duringspeech perception recruits more cortical areas than inquiet conditions. The neural networks required todecipher the signal from the noise may be compro-mised or less easily activated by attention in theelderly subjects.

Effects of Hearing Loss

The normal-hearing subjects had greater WRSand a greater number of significant IAFM responsesthan the elderly hearing-impaired subjects. The ma-jor difference between these two groups was that the80 Hz IAFM responses were larger in the normal-hearing group.

The present study used multiple IAFM stimuli.The use of multiple AM stimuli with the modulationrates in the 80 Hz range does not attenuate theresponse compared with responses to single stimuliprovided that the carrier frequencies are separatedby at least one half octave (John et al., 1998).However, multiple stimuli presented in the 40 Hzrange do show interaction with multiple stimuli thatare less than one half octave apart (John et al.,1998). Therefore there may have been more interac-tion between responses with the 40 Hz IAFM stimulithan with the 80 Hz IAFM. The degree of interactionof multiple 40 Hz stimuli has yet to be determined inthe hearing-impaired population. However, usingmultiple stimuli may be more representative ofspeech compared with single stimuli since anyspeech signal has energy at multiple frequencies.Comparing interactions between multiple stimuli innormal-hearing and hearing-impaired subjects maybe useful since intra-speech masking is known oc-cur. For example, hearing-impaired subjects showbetter performance identifying speech sounds when

the intensity of the first formant is reduced (Sum-mers & Leek, 1997). The hearing-impaired subjectshad smaller FM responses at 80 and 40 Hz (Table 9)than their normal-hearing counterparts althoughthis difference was not significant. The mean 80 HzFM responses in the hearing-impaired subjects weresmaller than in the normal-hearing subjects. This isprobably related to decreased frequency resolutionassociated with hearing loss (Moore, 1995). Hearingloss has been shown to cause impaired detection ofFM psychoacoustically (Lacher-Fougere & Demany,1998). Additionally, decreases in frequency resolu-tion have also been shown to be associated withdecreased performance in speech recognition taskssuch as recognition of speech with reduced spectralcues (Turner, Chi, & Flock, 1999), with vowel recog-nition (Turner & Henn, 1989).

Although the 40 Hz IAFM responses were thesame between the two groups, the functional signif-icance of this is hard to assess since the responses inthe elderly may have been less reduced by sleep.This may have obscured another possible explana-tion for the 40 Hz IAFM responses being similar inthese two groups. Cortical plasticity associated withhearing loss may effectively increase the “gain” ofthe auditory system. Central auditory “gains” havebeen observed in chinchillas after inner hair cell(IHC) destruction (Qiu, Salvi, Ding, & Burkard,2000). Recordings in the chinchilla 1 to 2 wk afterIHC destruction caused by carboplatinum resultedin decreased responses in the compound action po-tential and inferior colliculus with an increase inauditory cortex potential. The authors suggestedthat IHC loss resulted in an increase in the “gain” ofthe central auditory system to compensate for thereduced input. Therefore with respect to our data, ifthe 40 Hz responses (cortical) are similar in subjectswith normal-hearing and hearing loss, but the 80 Hz(brain stem) responses are smaller with hearingloss, normal 40 Hz amplitudes may represent ascenario where the cortex has compensated for thereduced input by increasing the gain.

Our subjects with hearing loss were elderly. Wethink it likely that the effects of hearing loss on theIAFM response would be similar in younger hear-ing-impaired subjects, but this will require furtherresearch.

Aided Responses

Wearing a hearing aid resulted in higher WRSand more significant IAFM responses. The increasesobserved in WRS performance while wearing a hear-ing aid were similar to the increases in the numberof significant IAFM responses for both the 40 and 80Hz stimuli. The difference in unaided and aided


WRS was 19%. The number of significant 40 HzIAFM responses increased by 28% wearing a hear-ing aid. The number of significant 80 Hz IAFMresponses increased by 19% wearing a hearing aid.Wearing a hearing aid increased the 80 Hz IAFMamplitudes mainly for the lower carriers, perhapsbecause the IAFM stimuli were loudest at the lowerfrequencies. The hearing aid data suggest that ex-amining the number of significant IAFM responsesin aided and unaided conditions may lead to objec-tive methods of examining with the benefits of ahearing aid. At the present time the presence orabsence of a response seems to be a more reliableindicator of the benefits of amplification than re-sponse amplitudes.

We are not sure why the effects of amplificationwere greater at 1500 Hz than at the other frequen-cies. The aided subjects all had sloping high-fre-quency losses. The aids were adjusted to providecompression and this would have been greater at thehigher frequencies. The amplification at the 1500 Hzregion may have been greater for 70 dB SPL soundsthan for the higher frequencies. The normal re-sponse at 500 Hz shows less of an effect of intensitychange that at the mid-frequencies.

Clinical Implications

The general idea behind using multiple IAFM topredict WRS is that speech contains acoustic infor-mation that varies rapidly in intensity and fre-quency. Hearing impairment leading to problems inperceiving speech can occur at many differentstages. First, speech difficulties stemming from pe-ripheral factors can occur when absolute thresholdsare increased. Second, difficulties can be associatedwith processing or discriminating the fine elementsof speech. This may include deficits associated withspeed of processing (Gordon-Salant & Fitzgibbons,2001), reverberations in speech (Helfer & Wilber,1990), gap detection (Snell, Mapes, Hickman &Frisina, 2002), and discrimination of intensity andfrequency changes (Lacher-Fougere & Demany,1998). Third, difficulties may be encountered athigher levels associated with matching the speechstimuli to remembered templates for words or fittingthese words into the ongoing context of continuousspeech. The use of 80 Hz IAFM assesses the auditorysystems ability in processing intensity (AM) andfrequency (FM) changes at the level brain stemwhile 40 Hz IAFM stimuli evaluates a higher level ofprocessing in the auditory cortex. Recording both 40and 80 Hz IAFM responses results in better rela-tionships with WRS than with just one range ofmodulation frequencies. The results from this studyand the previous two (Dimitrijevic et al., 2001;

Picton et al., 2002) demonstrate that the number ofIAFM responses relates to WRS and that IAFMresponses are reliably recorded while wearing ahearing aid.

SUMMARY

The results of this study show that speech mod-eled IAFM stimuli evoke ASSRs that are related toWRS. IAFM responses were examined in young andelderly subjects without hearing loss and in elderlysubjects with hearing loss, with and without theirhearing aids. WRS was related to the number ofsignificant IAFM responses recorded, when theWRS varied with masking, with hearing-loss, orwith hearing aids. Response amplitudes of IAFMstimuli can vary according to subject state andphysical properties such as head size and skullthickness. The presence or absence of a response isrelatively unaffected by these factors, and givesbetter correlations with WRS.

APPENDIX

Speech Modeled IAFM

In the previous two studies (Dimitrijevic et al.,2001; Picton et al., 2002) the depth of modulation ofthe IAFM stimuli were fixed at 50% AM and 20%FM and the intensities of the four different carrierswere the same. The current study used IAFM stim-

Appendix Table 1 FM data

Formant Fmean % FM

Vowel F1 528 27F2 1513 30F3 2464 12

Consonant-Vowel F1 417 42F2 1596 28F3 2512 14

Fricative 4454 35

Appendix Table 2 AM data

Formant Fmean %AM

Vowel F1 500 50F2 1500 34F3 2500 33F4 4000 21

Consonant-vowels F1 500 52F2 1500 51F3 2500 47F4 4000 50

Fricative F1 500 63F2 1500 57F3 2500 73F4 4000 34


uli more representative of human speech. Six pa-rameters were adjusted to mimic real speech: 1)overall intensity, 2) carrier frequencies, 3) relativeintensity across frequencies, 4) depth of FM, 5)depth of AM, and 6) modulation frequencies. Thephonetic examples used to model the stimuli con-sisted of consonant-vowel (C-V) transitions, vowel-vowel (V-V) transitions, and fricative sounds. Thedevelopment of the stimuli occurred in two steps.First, parameters were derived from published de-scriptions of the acoustic properties of speech. Sec-ond, the final parameters were adjusted slightly tomake the recorded spectra of the IAFM stimulisimilar to that of the speech-masking noise used inthe experiment.

1. Overall Amplitude

The intensity of the IAFM combined stimulus wasset at 70 dB SPL. This value was chosen to estimatenormal conversation levels (Cox & Moore, 1988;Peterson & Barney, 1952). Average maximum com-fort level (MCL) for speech in young controls andaided subjects were 60 and 66 dB SPL, respectively(Picton et al., 2002). Hearing-impaired subjects mayget maximum discrimination at higher intensitylevels than MCL (Posner & Ventry, 1977; Ullrich &Grimm, 1976) but levels that are too high may leadto discomfort.

2. Carrier Frequencies

The carrier frequencies were based on the threemean formant frequencies for consonant-vowel(C-V) transitions (Assmann, 1995; Kewley-Port,1982; Summers & Leek, 1997) and vowels (Assmann& Katz, 2000; Hillenbrand, Getty, Clark, & Wheeler,1995; Peterson & Barney, 1952) and the meanfrequencies for fricatives (Hughes & Halle, 1956;Jongman, Wayland, & Wong, 2000; Pittman & Stel-machowicz, 2000). The carrier frequencies usedwere 500, 1500, 2500, and 4000 Hz (see AppendixTable 1).

3. Relative Amplitudes

We decreased the amplitude of the AM and FMcomponents with increasing carrier frequency toapproximate the shape of the long-term averagespeech spectrum (LTASS). The amplitudes of thecarriers were estimated from the LTASS of Cornel-isse, Gagne, and Seewald (1991). The dB difference(dBdiff) between the 500 and 1500, 2500, and 4000Hz carriers was measured as �8, �15, and �16 dB.For the MASTER system, the amplitude of the 500Hz stimulus (a500) was arbitrarily set to 40 units andamplitudes of the other stimuli (ax) were adjustedaccording to the formula:

ax � a500 � 10dBdiff/20

The amplitudes for the 500, 1500, 2500, and 4000Hz carriers were therefore 40, 16, 7, and 6.

3. Depth of FM

The depth of frequency modulation is the fre-quency range of the modulation expressed as apercentage of the carrier frequency. A 25% FM of a1000 Hz carrier would therefore represent frequencychanges between 875 and 1125 Hz (i.e., �12.5%).Representative frequency transitions in speech weremeasured from C-V and V-V frequency changes. C-Vtransition data were compiled from Kewley-Port(1982), Summers and Leek (1997), and Assmann(1995). The compiled data consisted of the first threeformants for the consonants b, g, d, w, and thevowels /i/, /a/, /u/, /ae/, /er/, /e/, /I/, /o/, /U/. Each C-Vpair consisted of an onset frequency (f1) and a targetfrequency (f2). The %FM in these C-V transitionswas estimated as

%FM � 200 � |f1-f2|/(f1�f2)where |f2-f1| is the absolute frequency

difference.Appendix Table 1 shows the %FM for the first

three formants for all C-Vs used. The values for theV-V frequency changes were based on the mean andstandard deviations from Peterson and Barney(1952), Assmann and Katz (2000), and Hillenbrand,Getty, Clark, and Wheeler (1995) of the first threeformants for all of the vowels. In this case, the fmean� mean formant frequency across the vowels andthe transitions were based on 1 SD. In this case theequation used to calculate %FM was %FM � 1SD/fmean � 100. Similarly, for fricatives, data was takenfrom Jongman et al. (2000), Pittman and Stelma-chowicz (2000), and Hughes and Halle (1956).

4. Depth of AM

The depth of amplitude modulation is the ampli-tude range of the envelope expressed as a percent-age. This was calculated using the formula:

%AM � 100 (amax � amin)/(amax � amin)where amax and amin are the maximum and min-

imum amplitude, respectively.Consonant vowel amplitude spectra were taken

from Halle, Hughes, and Radeley (1957) and Klien,Plomp, and Pols (1970) and vowel amplitudes weretaken from Stevens (1980) and fricative amplitudeswere taken from Hughes and Halle (1956). Thedepth of AM could be assessed by the peak to troughranges seen in these spectra. Peaks and troughswere measured for each of the selected carrier fre-quencies: 500, 1500, 2500, and 4000 Hz. Using theabove formula, amax and amin were taken to beamean�1SD and amean�1SD. Where amean is the


average amplitude value (dB) for C-V, V, and frica-tives. The %AM was calculated for C-V, V, andfricatives, then averaged across phonemes. AM pho-nemic data are given in Appendix Table 2.

6. Modulation Frequencies

Modulation frequencies that are typically usedare near 40 Hz and 80 Hz since the ASSR ampli-tudes are generally greater in these two ranges(reviewed by Picton et al., 2003). The durations ofvarious phonemes are can be changed to modulationrates if one considers a half cycle (peak) to be theduration of the phoneme. Stop consonant durationcan be short as 5 msec and as long as 50 msec,equivalent to 100 Hz and 10 Hz. Similarly, subjectscan identify C-V lasting 10 msec range with reason-able accuracy (Blumstein & Stevens 1980; Ohde &Abou-Khalil 2001) resulting in a full cycle at 50 Hz.

Calibration and Adjustment

The IAFM and speech noise spectrum were mea-sured using a Bruel and Kjaer Sound Pressure Levelmeter model 2230 (microphone cartridge 4155 for freefield) whose output was to a digital oscilloscope (Tek-tronix TD210) and frequency analyzed using commer-cial software (WaveTek). Since a speech noise maskerwas used to mask the IAFM responses and speechstimuli, the IAFM stimuli needed to have the samespectral shape as the noise, otherwise the maskerwould not have masked all frequencies equally. Usingthe IAFM parameters derived from Tables 1 and 2 ina free field sound environment resulted in a spectrumthat showed slightly greater energy at 2500 and 4000Hz. Therefore, the amplitudes of these carriers weredecreased. Additionally, the depth of modulation of forthe 2500 Hz AM and FM was increased by 5% toincrease the energy in this region wider and moresimilar to the speech noise. The final parameters thatwere used are shown in Table 2 of the main article.The speech noise used was created by the GrasonStadler Audiometer (GS16). It conformed to the ANSIstandards (1996), showing a flat spectrum from 100 to1000 Hz, and decreasing at 12 dB/octave for 1000 to6000 Hz. Figure 1 of the article shows the frequencyspectrum of the final IAFM stimuli together with thespeech noise spectrum for comparison.

ACKNOWLEDGMENTS:This research was supported by the Canadian Institute of HealthResearch. The authors also appreciate the support of the JamesKnowles and Baycrest Foundation. The authors would also like tothank Patricia van Roon for technical support. Some of these resultswere presented at the XI International Symposium on AudiologicalMedicine and the XXVII National Congress of the Italian AudiologySociety in Padova/Abano Terme in October 2002.

Address for correspondence: Andrew Dimitrijevic, School of Au-diology & Speech Sciences, University of British Columbia, 5804Fairview Ave., British Columbia, V6T 1Z3, Canada. E-mail:[email protected].

Received May 8, 2003; accepted October 29, 2003

REFERENCES

Alain, C., McDonald, K. L., Ostroff, J. M., & Schneider, B. (2001).Age-related changes in detecting a mistuned harmonic. Jour-nal of the Acoustical Society of America, 109, 2211–2216.

American National Standards Institute (1996). Specifications foraudiometers. ANSI 3.6–1996. New York: American NationalStandards Institute.

Assmann, P. F. (1995). The role of formant transitions in theperception of concurrent vowels. Journal of the AcousticalSociety of America, 97, 575–584.

Assmann, P. F., & Katz, W. F. (2000). Time-varying spectralchange in the vowels of children and adults. Journal of theAcoustical Society of America, 108, 1856–1866.

Beattie, R. C., Barr, T., & Roup, C. (1997). Normal and hearing-impaired word recognition scores for monosyllabic words inquiet and noise. British Journal of Audiology, 31, 153–164.

Beattie, R. C., & Zipp, J. A. (1990). Range of intensities yieldingPB Max and the threshold for monosyllabic words for hearing-impaired subjects. Journal of Speech and Hearing Disorders,55, 417–426.

Blumstein, S. E., & Stevens, K. N. (1980). Perceptual invarianceand onset spectra for stop consonants in different vowel envi-ronments. Journal of the Acoustical Society of America, 67,648–662.

Boettcher, F. A., Madhotra, D., Poth, E. A., & Mills, J. H. (2002).The frequency-modulation following response in young andaged human subjects. Hearing Research, 165, 10–18.

Boettcher, F. A., Poth, E. A., Mills, J. H., & Dubno, J. R. (2001).The amplitude-modulation following response in young andaged human subjects. Hearing Research, 153, 32–42.

Brandy, W. T. (2002). Speech audiometry. In J. Katz (Ed.),Handbook of Clinical Audiology. 5th Edition. Philadelphia:Lippincott Williams & Wilkins.

Cohen, L. T., Rickards, F. W., & Clark, G. M. (1991). A compar-ison of steady-state evoked potentials to modulated tones inawake and sleeping humans. Journal of the Acoustical Societyof America, 90, 2467–2479.

Cornelisse, L. E., Gagne, J. P., & Seewald, R. C. (1991). Ear levelrecordings of the long-term average spectrum of speech. Earand Hearing, 12, 47–54.

Cox, R. M., & Moore, J. N. (1988). Composite speech spectrum forhearing aid gain prescription. Journal of Speech and HearingResearch, 31, 102–107.

Crowley, K., Trinder, J., & Colrain, I. M. (2002). An examinationof evoked K-complex amplitude and frequency of occurrence inthe elderly. Journal of Sleep Research, 11, 129–140.

Crowley, K., Trinder, J., Kim, Y., Carrington, M., & Colrain, I.(2002). The effects of normal aging on sleep spindle andK-complex production. Clinical Neurophysiology, 113, 1615–22.

Dimitrijevic, A., John, M. S., van Roon, P., & Picton, T. W. (2001).Human auditory steady-state responses to tones independentlymodulated in both frequency and amplitude. Ear and Hearing,22, 100–111.

Dimitrijevic, A., John, M. S., van Roon, P., Purcell, D. W.,Adamonis, J., Ostroff, J., Nedzelski, J. M., & Picton, T. W.(2002). Estimating the audiogram using multiple auditorysteady-state responses. Journal of the American Academy ofAudiology, 13, 205–224.


Dobie, R. A., & Wilson, M. J. (1998). Low-level steady-stateauditory evoked potentials: effects of rate and sedation ondetectability. Journal of the Acoustical Society of America, 104,3482–3488.

Dubno, J. R., Dirks, D. D., & Morgan, D. E. (1984). Effects of ageand mild hearing loss on speech recognition in noise. Journal ofthe Acoustical Society of America, 76, 87–96.

Drullman, R., Festen, J. M., & Plomp, R. (1994). Effect oftemporal envelope smearing on speech reception. Journal ofthe Acoustical Society of America, 95, 1053–1064.

Frisina, D. R., & Frisina, R. D. (1997). Speech recognition in noiseand presbycusis: relations to possible neural mechanisms.Hearing Research, 106, 95–104.

Glasberg, B. R., & Moore, B. C. (1986). Auditory filter shapes insubjects with unilateral and bilateral cochlear impairments.Journal of the Acoustical Society of America, 79, 1020–1033.

Gorga, M. P., Beauchaine, K. A., & Reiland, J. K. (1987). Com-parison of onset and steady-state responses of hearing aids:Implications for the use of the auditory brainstem response inthe selection of hearing aids. Journal of Speech and HearingResearch, 30, 130–136.

Gordon-Salant, S., & Fitzgibbons, P. J. (2001). Sources of age-related recognition difficulty for time-compressed speech. Jour-nal of Speech, Language, and Hearing Research, 44, 709–719.

Halle, M., Hughes, G. W., & Radeley, J. P. (1957). Acousticsproperties of stop consonants. Journal of the Acoustical Societyof America, 29, 107–116.

He, N., Dubno, J. R., & Mills, J. H. (1998). Frequency andintensity discrimination measured in a maximum-likelihoodprocedure from young and aged normal-hearing subjects. Jour-nal of the Acoustical Society of America, 103, 553–565.

Helfer, K. S., & Wilber, L. A. (1990). Hearing loss, aging, andspeech perception in reverberation and noise. Journal ofSpeech and Hearing Reserach, 33, 149–155.

Herdman, A. T., Lins, O., van Roon, P., Stapells, D. R., Scherg,M., Picton, T. W. (2002). Intracerebral sources of humanauditory steady-state responses. Brain Topography, 15, 69–86.

Herdman, A. T., Picton, T. W., & Stapells, D. R. Place specificityof auditory steady state responses (2002). Journal of theAcoustical Society of America, 112, 1569–1582.

Herdman, A. T., & Stapells, D. R. (2001). Thresholds determinedusing the monotic and dichotic multiple auditory steady-stateresponse technique in normal-hearing subjects. ScandinavianAudiology, 30, 41–49.

Herdman, A. T., & Stapells, D. R. (2003). Auditory steady stateresponse thresholds of adults with sensorineural hearing im-pairment. International Journal of Audiology, 42, 237–248.

Hillenbrand, J., Getty, L. A., Clark, M. J., & Wheeler, K. (1995).Acoustic characteristics of American English vowels. Journalof the Acoustical Society of America, 97, 3099–3111.

Hughes, G. W., & Halle, M. (1956). Spectral properties of fricativeconsonants. Journal of the Acoustical Society of America, 28,303–310.

Jerger, J., Chmiel, R., Frost, J. D., & Coker, N. (1986). Effect ofsleep on the auditory steady state evoked potential. Ear andHearing, 7, 240–245.

Jerger, J., & Hayes, D. (1977). Diagnostic speech audiometry.Archives of Otolaryngology, 103, 216–222.

John, M. S., Dimitrijevic, A., & Picton, T. W. (2001). Weightedaveraging of steady-state responses. Clinical Neurophysiology,112, 555–562.

John, M. S., Dimitrijevic, A., van Roon, P., & Picton, T. W. (2001).Multiple auditory steady-state responses to AM and FM stim-uli. Audiology & Neuro-Otology, 6, 12–27.

John, M. S., Lins, O. G., Boucher, B. L., & Picton, T. W. (1998).Multiple auditory steady-state responses (MASTER): Stimulusand recording parameters. Audiology, 37, 59–82.

John, M. S., & Picton, T. W. (2000). MASTER: A Windowsprogram for recording multiple auditory steady-state re-sponses. Computer Methods and Programs in Biomedicine, 61,125–150.

Jongman, A., Wayland, R., & Wong, S. (2000). Acoustic charac-teristics of English fricatives. Journal of the Acoustical Societyof America, 108, 1252–1263.

Kewley-Port, D. (1982). Measurement of formant transitions innaturally produced stop consonant-vowel syllables. Journal ofthe Acoustical Society of America, 72, 379–389.

Klein, W., Plomp, R., & Pols, L. C. (1970). Vowel spectra, vowelspaces, and vowel identification. The Journal of the AcousticalSociety of America, 48, 999–1009.

Lacher-Fougere, S., & Demany, L. (1998). Modulation detectionby normal and hearing-impaired listeners. Audiology, 37, 109–121.

Leek, M. R., Dorman, M. F., & Summerfield, Q. (1987). Minimumspectral contrast for vowel identification by normal-hearingand hearing-impaired listeners. Journal of the Acoustical So-ciety of America, 81, 148–154.

Linden, R. D., Campbell, K. B., Hamel, G., Picton, T. W. (1985).Human auditory steady state evoked potentials during sleep.Ear and Hearing, 6, 167–174.

Lins, O. G., & Picton, T. W. (1995). Auditory steady-state re-sponses to multiple simultaneous stimuli. Electroencephalog-raphy and Clinical Neurophysiology, 96, 420–432.

Martin, M. (1998). Speech Audiometry. 2nd Edition. San Diego:Singular Publishing.

Matschke, R. G. (1990). Frequency selectivity and psychoacoustictuning curves in old age. Acta Oto-Laryngologica. Supplemen-tum, 476, 114–119.

Moore, B. C. (1995). Perceptual Consequences of Cochlear Dam-age. Oxford: Oxford University Press.

Moore, B. C. J., & Alcantara, J. I. (1996). Vowel identificationbased on amplitude modulation. Journal of the AcousticalSociety of America, 99, 2332–2343.

Ohde, R. N., & Abou-Khalil, R. (2001). Age differences for stop-consonant and vowel perceptions in adults. Journal of theAcoustical Society of America, 110, 2156–2166.

Patterson, R. D., Nimmo Smith, I., Weber, D. L., & Milroy, R.(1982). The deterioration of hearing with age: Frequency selec-tivity, the critical ratio, the audiogram, and speech threshold.Journal of the Acoustical Society of America, 72, 1788–1803.

Perez-Abalo, M. C., Savio, G., Torres, A., Martin, V., Rodriguez,E., & Galan, L. (2001). Steady state responses to multipleamplitude-modulated tones: An optimized method to test fre-quency-specific thresholds in hearing-impaired children andnormal-hearing subjects. Ear and Hearing, 22, 200–211.

Peterson, G., & Barney, H. (1952). Control methods used in astudy of the vowels. Journal of the Acoustical Society ofAmerica, 24, 175–184.

Pichora-Fuller, M. K., Schneider, B. A., & Daneman, M. (1995).How young and old adults listen to and remember speech innoise. Journal of the Acoustical Society of America, 97, 593–608.

Picton, T. W., Dimitrijevic, A., van Roon, P., John, M. S., Reed, M.,& Finkelstein, H. (2002). Possible roles for the auditory steady-state responses in fitting hearing aids. In R. C. Seewald, J. S.Gravel (Eds.), A Sound Foundation Through Early Amplifica-tion 2001. Proceedings of the 2nd International Conference (pp.63–73). Basel: Phonak AG.

Picton, T. W., Durieux-Smith, A., Champagne, S. C., Whitting-ham, J., Moran, L. M., Giguère, C., & Beauregard, Y. (1998).Objective evaluation of aided thresholds using auditory steady-state responses. Journal of the American Academy of Audiol-ogy, 9, 315–331.


Picton, T. W., John, M. S., Dimitrijevic, A., & Purcell, D. W.(2003). Human auditory steady-state responses. InternationalJournal of Audiology, 42, 177–219.

Picton, T. W., Skinner, C. R., Champagne, S. C., Kellett, A. J., &Maiste, A. C. (1987). Potentials evoked by the sinusoidalmodulation of the amplitude or frequency of a tone. Journal ofthe Acoustical Society of America, 82, 165–178.

Pittman, A. L., & Stelmachowicz, P. G. (2000). Perception ofvoiceless fricatives by normal-hearing and hearing-impairedchildren and adults. Journal of Speech and Language HearingResearch, 43, 1389–1401.

Posner, J., & Ventry, I. M. (1977). Relationships between com-fortable loudness levels for speech and speech discrimination insensorineural hearing loss. Journal of Speech and HearingDisorders, 42, 370–375.

Qiu, C., Salvi, R., Ding, D., & Burkard, R. (2000). Inner hair cellloss leads to enhanced response amplitudes in auditory cortexof unanesthetized chinchillas: evidence for increased systemgain. Hearing Research, 139, 153–171.

Rance, G., & Rickards, F. (2002). Prediction of hearing thresholdin infants using auditory steady-state evoked potentials. Jour-nal of the American Academy of Audiology, 13, 236–245.

Rechtschaffen, A., & Kales, A. (1968). A Manual of StandardizedTerminology, Techniques, and Scoring System for Sleep Stagesof Human Subjects. Washington: US Government PrintingOffice.

Salvi, R. J., Lockwood, A. H., Frisina, R. D., Coad, M. L., Wack,D. S., & Frisina, D. R. (2002). PET imaging of the normalhuman auditory system: Responses to speech in quiet and inbackground noise. Hearing Research, 170, 96–106.

Slawinski, E. B. (1994). Acoustic correlates [b] and [w] producedby normal young to elderly adults. Journal of the AcousticalSociety of America, 95, 2221–2230.

Snell, K. B., Mapes, F. M., Hickman, E. D., & Frisina, D. R.(2002). Word recognition in competing babble and the effects ofage, temporal processing, and absolute sensitivity. The Journalof the Acoustical Society of America, 112, 720–727.

Stapells, D. R., Galambos, R., Costello, J. A, & Makeig, S. (1988).Inconsistency of auditory middle latency and steady-stateresponses in infants. Electroencephalography and ClinicalNeurophysiology, 71, 289–295.

Stevens, K. N. (1980). Acoustic correlates of some phoneticcategories. Journal of the Acoustical Society of America, 68,836–842.

Summers, V., & Leek, M. R. (1997). Intraspeech spread ofmasking in normal-hearing and hearing-impaired listeners.Journal of the Acoustical Society of America, 101, 2866–2876.

Turner, C. W., Chi, S. L., & Flock, S. (1999). Limiting spectralresolution in speech for listeners with sensorineural hearingloss. Journal of Speech, Language, and Hearing Research, 42,773–784.

Turner, C. W., & Henn, C. C. (1989). The relation between vowelrecognition and measures of frequency resolution. Journal ofSpeech and Hearing Research, 32, 49–58.

Ullrich, K., & Grimm, D. (1976). Most comfortable listening levelpresentation versus maximum discrimination for word dis-crimination material. Audiology, 15, 338–347.

Van Tasell, D. J. (1993). Hearing loss, speech, and hearing aids.Journal of Speech and Hearing Research, 36, 228–244.

Walton, J. P., Frisina, R. D., & ONeill, W. E. (1998). Age-relatedalteration in processing of temporal sound features in theauditory midbrain of the CBA mouse. Journal of Neuroscience,18, 2764–2776.

Walton, J. P., Simon, H., & Frisina, R. D. (2002). Age-relatedalterations in the neural coding of envelope periodicities.Journal of Neurophysiology, 88, 565–578.


Auditory Steady-State Responses and Word Recognition Scores in Normal-Hearing and Hearing-Impaired...

Documents

Transcript of Auditory Steady-State Responses and Word Recognition Scores in Normal-Hearing and Hearing-Impaired...