Optimal modulation frequency for amplitude-modulation following response in young children during sleep

253

Hearing Research, 65 (1993) 253-261 0 1993 Elsevier Science Publishers B.V. All rights reserved 0378-5955/93/$06.00 HEARES 01872

Optimal modulation frequency for amplitude-modulation response in young children during sleep

following

Masaru Aoyagi, Teruo Kiren, Yoshinori Kim, Yutaka Suzuki, Takeo Fuse and Yoshio Koike Department of Otolaqwgology, Yamagata University School of Medicine, Yamagata, Japan

(Received 15 May 1992; Revision received and accepted 16 September 1992)

In young children, there appears to be no advantage to recording steady-state response (SSR) at a stimulus rate of 40 Hz. To determine the optimal modulation frequency in auditory SSR evoked by sinusoidally amplitude-modulated (SAM) tones (amplitude-modulation following response: AMFR) in children during sleep and compare response patterns of AMFR at different modulation frequencies while awake with those during sleep, AMFR was examined in 10 adults with normal hearing while awake and during sleep and in 10 young children with normal hearing during sleep. The stimulus was a 1000 Hz, 50 dBnHL SAM tone with a modulation depth of 95%. Modulation frequency was varied from 20 to 200 Hz in 20 Hz steps. Response was determined by phase spectral analysis and the S/N ratio calculated by spectral amplitude at the modulation frequency and noise level around the modulation frequency using fast Fourier transform. Although AMFR was clearly evoked only by a modulation frequency of 40 Hz in adults while awake, AMFRs at modulation frequencies of 80 and 100 Hz were detected during sleep, in addition to 40 Hz AMFR. In children, 40 Hz AMFR was difficult to detect, but response could be clearly detected at higher modulation rates, especially at modulation frequencies of 80 and 100 Hz, compared with response in adults during sleep. Modulation frequencies from 80 to 100 Hz would thus appear optimal for detecting AMFR during sleep in children. Amplitude-modulation

following response; Modulation frequency; Age effect; Effect of sleep; Phase spectral analysis

Introduction Steady-state response (SSR) elicited by a sinusoidally amplitude-modulated (SAM) tone (amplitudemodulation following response: AMFR) is considered to be a reflection of the synchronous discharge of auditory neurons phase-locked to the modulation frequency (MF) of tonal stimulation and is expected to be a useful tool for objective audiometry (Kuwada et al., 1986). It has been suggested that AMFR has a close relation to the 40 Hz event-related potential (Galambos et al., 1981) or 40 Hz SSR evoked by clicks and tone bursts. Although the amplitude of 40 Hz SSR is high in adults while awake, the amplitude decreases during sleep (Galambos et al., 1981; Linden et al., 1985). Furthermore, 40 Hz SSR is difficult to obtain and is thought to be insufficiently reliable for determining the low-frequency threshold in young children (Suzuki and Kobayashi, 1984; Stapells et al., 1988; Maurizi et al., 19901, though the final goal of electrical response audiometry is to obtain accurate frequencyspecific hearing patterns in infants, young children and mentally retarded patients, on whom standard behavioral audiometric examination cannot be conducted. Correspondence to: Masaru Aoyagi, Department of Otolaryngology, Yamagata University School of Medicine, Iida-Nishi, Yamagata, 990-23, Japan. Fax: 81 (236) 25-3813.

We applied the phase spectral analysis developed by Fridman et al. (1984) to the automatic detection of AMFR and found that threshold determination by the phase spectral analysis is more sensitive than that by visual analysis of wave form configuration (Aoyagi et al., 1991a). We also reported that the mean threshold difference in 40 Hz AMFR between wakefulness and sleep is less than 10 dB in adults, when detected by the phase spectral analysis (Aoyagi et al., 1991b). In young children, however, the same concerns mentioned above in regard to the 40 Hz SSR may be relevant. In order to determine the optimal modulation frequency for AMFR in young children during sleep, and compare the response patterns of the AMFR in different MFs while awake with those during sleep, AMFR was investigated in adults with normal hearing while awake and during sleep and in young children with normal hearing during sleep. In addition, the usefulness of the phase spectral analysis for detecting AMFR is discussed. Materials and Methods Subjects

Ten male adults ranging in age from 25 to 46 years (mean 31.4 years) and 10 young children (7 males and 3 females) ranging in age from 2 years 6 months to 4 years 11 months (mean 3 years 6 months) were studied

254

as volunteers. Consent to participate in the test was obtained from each adult and the parent(s) of each child. The adults were shown to have normal hearing by standard pure-tone audiometry. Although the children were not examined by any behavioral hearing tests, all of them were considered to have normal hearing from the findings of ABR elicited by clicks. ABR thresholds were 20 dBnHL in 3 children and 10 dBnHL in 7 children. Latency-intensity functions were normal in all children. Sleep staging

All adults were examined while awake and during natural night sleep or under sedation induced by Triazolam (0.25-0.5 mg, p.0.). During testing in the awake state, the subjects were seated comfortably on an armchair and asked to read a book or magazine. The children were sedated by the oral administration of monosodium trichlorethyl phosphate (70-80 mg/kg body weight) or by an anal administration of chloral hydrate (250-500 mg). All recording procedures were performed in night in a sound-attenuated electrically shielded room. Although sleep staging was not precisely estimated in the present study, all children were tested in stages 3 and/or 4. In adults, sleep stages were varied from stages 1 to 4 during the examination. Recording procedures were conducted even in REM sleep in some adult subjects. Examinations

A sinusoidally amplitude-modulated (SAM) tone was generated by a combination of two function generators (NF Model FG-163) and attenuated by an audible signal controller (DANA Japan DA502A). The stimulus carrier frequency used was 1 kHz, and the stimulus was presented monaurally at a constant intensity of 50 dBnHL to the subject’s right ear through a headphone (TDH-49). The modulation depth was fixed at 95%, and modulation frequency (MF) was varied from 20 to 200 Hz in 20 Hz steps. Silver cup electrodes for electroencephalography were placed on the vertex (active), ipsilateral earlobe (reference) and nasion (ground). Bioelectric activity from the electrodes was enhanced by the amplifier of an electric response-analyzer (NEC San-ei Signal Processor 7TlSA) at a pass-band of 8 Hz (6 dB/Octave roll-off) to 300 Hz (18 dB/Octave roll-off). AMFR was detected by phase spectral analysis of group averages of scalp potentials (Fridman et al., 1984) and by comparing the spectral amplitude at frequencies near that of the stimulus with background noise level (Batra et al., 1986). The wave form configuration of the response was also estimated visually. Phase spectral analysis

Digitized scalp potentials were collected for a set of 10 groups (group averages). Each group consisted of

100 epoches of 204.8 ms (512 samples, frequency spacing about 4.9 Hz). An average wave form was calculated for each group. After tapering 10 group averages with a rectangular window (raised-cosine rounded over 20 ms) to avoid spectral edge effect in the fast Fourier transform (FFT) routine, the component synchrony measure (CSM) for 10 group averages was calculated from phase variance for the Fourier component according to Fridman’s technique (Fridman et al., 1984). The equation of Mardia (1972) was used to compute the phase variance [var(cp(m)}] of each Fourier component among 10 group averages, and CSM was calculated by the following formula: CSM (m) = 1 -var

{q(m))

If the ‘m’th Fourier component has exactly the same phase in all 10 group averages, CSM(m) is equal to 1, indicating the ‘m’th Fourier component to be timelocked with a stimulus phase. It is thus more likely to reflect response. When the phase of a Fourier component is not time-locked with a stimulus phase and changes randomly from wave form to wave form, CSM(m) approaches 0. The Fourier component is then detected as spontaneous background EEG. Spectral amplitude analysis

The ground average of the data (10 group averages) for a stimulus condition was used to calculate the spectral amplitude according to Batra’s technique (Batra et al., 1986). When the MF of the stimulus was the ‘j’th component of the FFT, the estimate of the noise level was calculated from the amplitudes of the 6 components around the MF of the stimulus: j - 4, j - 3, j - 2, j + 2, j + 3 and j + 4. The criterion for the presence of response was the mean plus 3 SD of noise, termed ‘noise level’ in the present study. The signal/noise ratio (S/N ratio) was calculated from the spectral amplitude of the Fourier component corresponding to MF and the noise level. Visual analysis of wave form configuration

Five group averages for the same stimulus condition were averaged to obtain an averaged scalp potential for visual analysis of wave form configuration. Consequently, two averaged scalp potentials were obtained for each stimulus condition, and the number of sweeps for each averaged scalp potential was 5 times that of a group average in the phase spectral analysis. Statistical analysis

Wilcoxon’s signed-rank test and Mann-Whitney’s U test were used for CSM, and Student’s t-test was used for spectral amplitude, noise level and S/N ratio.

255

Results

clearly detected by visual analysis at MFs of 80 to 100 Hz during sleep, especially in children.

Wave form configuration Phase spectral analysis

AMFR wave forms and CSMs at each MF in an adult subject while awake and during sleep and in a child during sleep are shown in Figs. 1-3, respectively. Although sinusoidal response wave forms were detected visually at MF of 40 Hz in most adults while awake, no response was detected by visual analysis in 5 of 10 adults during sleep. Response wave forms were

<<

SSR

MF

>>

STIM.

DRTE~l992/02/01

CF=lKHZf50DB OHKI

MF043

As shown in Figs. 2 and 3, CSMs of the Fourier components corresponding to MF were high during sleep, especially at higher MFs, and harmonic contents were observed. In awake adults (Fig. 0, however, high CSMs were observed only at lower (40-60 Hz) MFs. Fig. 4 illustrates CSMs as a function of MF in adults

32Y.

MFIKOTO

NORMRL(WQKE)

tl

RUE=1000

40 HZ

60 HZ

00

100

1 ‘51

. .3

HZ

. .3

HZ

. 1..

1

I 1 I ‘7i

. . 5

120 HZ

.

1..

140 HZ

. . 5

.

1..

160 HZ

. .

100 HZ

200

s

HZ L.........,........., 0 50

. 100

150

msec

1 0

100

200

300

400

Hz

Fig. 1. AMFR wave forms and their CSMs at each modulation frequency in an adult while awake. Sinusoidal response wave forms are detected visually at MF of 40 Hz, and CSM corresponding to the MF shows a high value. At other MFs, no response is detected visually, but the corresponding

CSM of a 60 Hz AMFR

is high.

2.56

while awake and during sleep and in young children during sleep. At MF of 40 Hz, CSMs in adults were high even in the sleep state, whereas CSMs in children during sleep were significantly lower than in adults during awaking and sleep (P < 0.01). On the other hand, CSMs at higher MFs were higher in adults during sleep and in children during sleep than in adults while awake. Particularly at MFs of 80 and 100 Hz, most subjects had very high CSMs during sleep. This was especially true for children as compared with adults in the awaking state (P < 0.01). At MFs higher than 120 Hz, CSM varied among adults during sleep. <<

SSR

MF

))

STIM.

MF051

whereas very high CSM (more than 0.81 were observed in 9 of 10 children. Consequently, when response is detected by the phase spectral analysis, AMFR would be stable at MF of 40 Hz in adults in the waking state and at MFs of 80-100 Hz during sleep, especially in sleeping children.

Spectral amplitude, noise level and S/N

CF=lKHZ/SBDB OHKI,

ratio

Fig. 5 illustrates spectral amplitude and noise level as a function of MF in adults while awake and during sleep and in young children during sleep. In awake DRTEll992/02/03

MQKOTO

32,

M

NORMFIL~

SLEEP

WE=1 000

x.. ..s 31 _

20 HZ

40 HZ

60 HZ

80 HZ

100 HZ

120 HZ

. .3

140 HZ

. 1..

160 HZ

. .

s .

1..

I

II

100 HZ

n

tl

I

. .5

200 HZ 1.. 0

. . 50

. -. 8.. 100

-. . 150

-. I msec

.

0

100

200

300

Fig. 3. AMFR wave forms and their CSMs at each modulation frequency in a child during sleep. Responses 60-140 Hz, and CSMs corresponding to MFs are extremely high at MFs higher than 60 Hz. Harmonic contents than 40 Hz.

400 Hz are clearly observed at MFs of are also observed at MFs higher

257

adults, spectral amplitude corresponding to MF was significantly higher than the noise level (P < 0.01) and the S/N ratio exceeded 1.0 in all subjects at an MF of 40 Hz. At other MFs, difference between the spectral amplitude of the response and the noise level was not significant. During sleep, difference between spectral amplitude of response and noise level was not significant, and the S/N ratio was less than 1.0 in 3 adults at an MF of 40 Hz. In contrast, spectral amplitudes of response were significantly higher than the noise levels

<<

SSR

MF037

MF

>>

STIM.

(P < 0.05) at MFs of 80 and 100 Hz, and the S/N ratio exceeded 1.0 in 7 and 9 subjects, respectively. In children, the spectral amplitude of the response was significantly lower than the noise level (P < 0.02), and the S/N ratio exceeded 1.0 in only one child at the MF of 40 Hz. At MFs higher than 80 Hz, however, spectral amplitude of response was significantly higher than the noise level (P < 0.05). Noise levels during sleep were significantly higher than during wakefulness at MF of 20 Hz (P < 0.051, and the noise level at MF of 40 Hz

UOL50DB,BFlK,MF20-200

D1?TE:1992/01/29

14-19-92-6

3Y6M

NORMFIL

1 1.. e. s

WE=1000

l.D

..m

20 HZ

a

40 HZ

60 HZ

00 HZ

100 HZ

120 HZ

140 HZ

160 HZ

100 HZ

200 HZ , 0

*. 50

I . . . I 100 150

I msec

m. s.1 1.. .m 31 m. * 1.. .I .1.. 5 .1 ..s1 0

100

200

300

400 Hz

Fig. 2. AMFR wave forms and their CSMs at each modulation frequency in the same adult as in Fig.1, during sleep. Although no response detected visually at any MFs, CSMs corresponding to MFs are high especially at higher MFs, and harmonic contents of CSMs are observed MFs higher

than 100 Hz.

is at

258

Adults : Awake

0

40

CSM

80

120

160

0

200 Hz

Adults : SleeD

0

40

80

120

160

200

800

Adults: Sleep 600

0

0

cm

40

80

Children

120

160

200 Hz

0

40

80

80

120

160

200

40

80

120

160

200

: Sleep

0 0

40

120

160

Modulation Frequency (Hz)

200 Hz

modulation Frequency Fig. 4. CSMs as a function of MF in adults while awake and during sleep and in children during sleep. Dotted lines indicate individual data and thick lines with open circles indicate means for each group.

was almost the same in the 3 groups. At MFs higher than 80 Hz, however, the noise Ievels were significantly lower during sleep than waking (P < 0.05). The means and standard deviations of the S/N ratio as a function of MF in adults while awake and during sleep and in children during sleep are shown in Fig. 6. The same tendency seen in the phase spectral analysis (Fig. 4) was observed in the S/N ratio.

Fig. 5. Means and standard deviations of the spectral amplitude of the Fourier component corres~nding to MF (thick line with open circle) and noise level (dotted line with closed circle) as a function of MF in adults while awake and during sleep and in children during sleep. Although noise levels at MF of 40 Hz are almost the same among the 3 groups, noise levels during sleep at MFs higher than 60 Hz are significantly lower than those while awake.

the modulation frequency (MF) regardless of the carrier frequency and is thought to reflect the synchronous discharge of large populations of auditory

SIN 4.0

T

T1

Discussion Effect of sleep and age on the AMFR

Scalp-recorded auditory evoked potentials elicited by a continuous sinusoidally amplitude-modulated (SAM) tone have been reported by several authors (Rickards and Cfark, 1984; Kuwada et al., 1986; Rees et al., 1986; Dolphin and Mountain, 1992), and termed the amplitude modulation following response (AMFR) (Kuwada et al., 1986) or the envelope following response (Dolphin and Mountain, 1992). AMFR entrains

0

40

Ii0 160 80 modulation Frequency

200 Hz

Fig. 6. Means and standard deviations of the S/N ratio as a function of MF in adults while awake (thick line with open circle) and during sleep (dotted line with closed circle) and in children during sleep (dotted line with closed square). The same pattern was observed as in the phase spectral analysis.

259

neurons phase-locked to MF. Because the power spectrum of SAM tone has a narrow band, AMFR is expected to be a useful tool for objective audiometry and to provide frequency-specific information. It has been suggested that AMFR has a close relation to 40 Hz event-related potential (Galambos et al., 1981) or 40 Hz steady-state response (SSR) evoked by clicks and tone bursts and that 40 Hz SSR probably excites the same neural structure as AMFR in a less frequency-specific manner (Kuwada et al., 1986). Although 40 Hz SSR shows robust potential in awake adults, the amplitude decreases by 50-60% during sleep (Galambos et al., 1981; Linden et al., 1985). Furthermore, 40 Hz SSR is difficult to obtain in newborns (Maurizi et al., 1990) and young children (Suzuki and Kobayashi, 19841, and infants show no consistent amplitude peak in SSR evoked by tone bursts at any rate between 9-59 Hz (Stapells et al., 1988). The effects of stimulation rate in SSR and modulation frequency in AMFR has been investigated by many authors, but only a few articles (Rees et al., 1986; Kuwada et al., 1986) describe response at stimulation rates or modulation frequencies higher than 70 Hz. Kuwada et al. (1986) suggested that there are two generators of AMFR. The one evoked by low modulation frequencies (30-50 Hz) has about a 30 ms latency, higher response amplitude and a threshold very close to the behavioral hearing threshold. The other evoked by higher modulations (75-350 Hz) has a latency of about 7-9 ms, smaller response and can only be evoked at higher intensities (70-80 dB HL). The former is thought to be the same response as 40 Hz SSR evoked by clicks or tone-bursts. However, all subjects were adults in Kuwada’s report and the recording procedure was carried out only in the awake state. Since sedation is necessary to obtain reliable bioelectrical scalp potentials in infants and young children, the effects of sleep on AMFR and differences in the function of MF between adults and young children should be investigated. In the present study, 40 Hz AMFR was also robust in adults, even during sleep, but it was difficult to obtain 40 Hz AMFR in young children during sleep. In contrast, at MFs of 80-100 Hz, where detection of AMFRs was not stable among adults in the waking state, response was clearly detected during sleep, especially in young children. Moreover, in young children, AMFRs were clearly detectable also at MFs higher than 120 Hz by the phase spectral analysis or Batra’s technique. These findings are basically the same as Kuwada’s description of the two generators of AMFR. He stated that AMFRs evoked by higher modulations are small and their thresholds are 70-80 dBnHL, as mentioned above. In the present study, however, CSMs of 80 Hz and 100 Hz AMFRs elicited in the sleeping state were very high even at 50 dBnHL. Modulation

frequencies from 80 to 100 Hz would thus appear optimal for detecting AMFR in young children during sleep. This difference in the MF function between waking and sleeping can be clearly explained by differences in noise levels in the two states, as shown in Fig. 5. Although the spectral amplitudes of response at MFs of 80-100 Hz showed almost the same level in the two states, the noise levels at MFs higher than 60 Hz were significantly lower in the sleeping state than in the waking state. Thus, AMFRs evoked by higher MFs, which are hidden by noise in the waking state, appear in the sleeping state. Although carrier frequency was fixed at 1000 Hz in the present study, the most important requirement for electrical response audiometry is to estimate the low frequency hearing threshold exactly, even in an abrupt high frequency hearing loss or low frequency hearing loss. We are investigating the threshold of AMFR elicited by 1000 Hz SAM tones with an MF of 80 Hz in children with normal and impaired hearing during sleep. In many children, 80 Hz AMFR thresholds were 10 dB lower than those of ABR evoked by 1000 Hz tone-pips. These results will be discussed in another article, and further investigation using SAM tones with lower carrier frequencies should be made for the development of a useful tool for the precise estimation of the hearing threshold in the lower frequency range in infants and young children.

Sources of the AMFR The results of Kuwada’s investigation (Kuwada et al., 1986) and the present study suggest characters of AMFRs at MFs lower than 60 Hz and at MFs higher than 80 Hz to be different. Amplitudes of AMFRs at lower MFs are robust in adults while awake, but reduced during sleep. On the other hand, AMFRs at higher MFs become apparent when the subject is examined during sleep. Latencies (or phase shifts) and thresholds are also different in the two responses. The sources of the two responses would thus appear to be different. It is generally accepted that 40 Hz SSR in adults is predicted reasonably well by overlapping or superimposing waves present in the middle latency response (MLR) (Galambos et al., 1981). However, peak latenties of MLR in young children during sleep are significantly longer than in adults. Sleep causes large variation in the MLR patterns of young children than adults (Okitsu, 1984). In the present study, the amplitude of 40 Hz AMFR was slightly reduced in adults and the spectral amplitude of the 40 Hz frequency component was significantly lower than adjacent frequency components (noise level) in young children. There is thus no doubt about the characteristics of AMFR evoked by MFs around 40 Hz, whereas the anatomical site of the

generator is still controversial. Galambos et al. (1982) suggested that 40 Hz SSR may be generated in the thalamus, and a similar conclusion was drawn based on phase changes of the 40 Hz response in patients with thalamus or midbrain lesions (Spydell et al., 1985). Kuwada et al. (1986) stated that the generator of AMFR at low (25-55 Hz) MFs is probably auditory cortex, because of the similarity of the latency of the response to those observed in cortical neurons. Mlkela et al. (1990) concluded from the results of an experimental study using cats that auditory 40 Hz SSR is generated in the cortex. Regarding the generator of AMFR evoked by MFs higher than 80 Hz, Kuwada et al. (1986) suggested that it is probably generated in the midbrain, because latenties (7-9 ms) are similar to those observed in inferior colliculus (Kuwada et al., 1984). The results that the AMFRs at MFs higher than 80 Hz were clearly detected during sleep rather than while awake, and spectral amplitudes of response were almost the same in the two states indicate that sleep causes no change in AMFRs at higher MFs. AMFRs at higher MFs may thus be generated somewhere in the brainstem including the inferior colliculus and cochlear nucleus. To discuss the generators of AMFRs, however, further examination is needed. Technique for detecting SSR

Because of the simple sinusoidal wave form of SSR, it is fairly easy to detect response visually in background electrical noise. However, if background noise is relatively high compared with the amplitude of response, a more sensitive means for detecting response is needed. Several techniques for detecting SSR based on Fourier transform have been reported: spectral amplitude method mentioned above (Batra et al., 1986; Kuwada et al., 1986), calculating the phase coherence (Fridman et al., 1984; Picton et al., 1987a1, magnitudesquared coherence function (Dobie and Wilson, 19911, Hotelling’s T2 statistic (Hotelling, 1931; Picton et al., 1987a and 1987b1, and Victor’s T&, statistic (Victor and Mast, 1991). T& statistic exploits the relationship between the real and imaginary components of Fourier estimates, which is not exploited by Hotelling’s T2 statistic, and also utilizes amplitude information ignored by phase coherence statistics. Victor and Mast (1991) applied T& statistic as well as T2 statistic and phase coherence to SSR elicited by visual stimulus, and found the T,f,, statistic to be the most efficient for the detection and quantification of SSR. Since we have no experience in the use of T2 and T$, statistics for detecting of SSR or AMFR, we cannot comment on the usefulness of those methods. Spectral amplitude analysis in the present study was first applied in the detection of frequency-following

response by Batra et al. (1986) and applied to AMFR by Kuwada et al. (1986). In this method, a signal-tonoise estimate is made by comparing the spectral amplitude at the frequency of stimulation to those at adjacent frequencies (noise). They established a strict criteria (mean noise level, the mean amplitude calculated from the amplitudes of adjacent 6 frequency components, plus 3 SDS of noise) for the presence of response. However, Picton et al. (1987a) emphasized that it is theoretically difficult to determine the frequency range over which to make a comparison with signal amplitude, because EEG amplitude tends to decrease as frequency becomes higher. The synchrony measure method, a kind of phase spectral analysis, used in the present study was developed by Fridman et al. (1984). This is a method for calculating phase coherence and is essentially an amplitude-free version of the T2 statistic (Picton et al., 1987a). This method is based on the following theory. Small phase variance indicates a particular Fourier component to be time-locked with the onset of a stimulus and thus is more likely to reflect auditory-evoked potentials than spontaneous background EEG. Fridman et al. (1984) applied this method to brain stem auditory evoked potential (ABR) detection in patients with posterior fossa tumor. The authors applied this method to the detection of the threshold of AMFR (Aoyagi et al., 1991a and 1991b) and found the phase spectral analysis to be more sensitive than visual analysis for this determination. In the present study, the phase spectral analysis appeared more sensitive for the detection of AMFR than the spectral amplitude analysis. This means that most signal information is carried by the phase rather than the amplitude of signal at near-threshold intensities (Picton et al., 1987a). The predominance of phase in the detection of ABR threshold has been reported by Greenblatt et al. (1985). Picton et al. (1987a) compared the reliability and sensitivity of spectral amplitude analysis, phase coherence and T2 statistic in the detection of SSR elicited by a 500 Hz tone burst at a stimulus rate of 40 Hz, and stated that SSR is best estimated using either T2 statistic or phase coherence. Linden et al. (1985) conducted the ‘sweep technique of Fourier analysis’ to assess the effects of stimulus intensity and rate on steady-state response during sleep and found phase difference between the reference signal and EEG response during sleep not to significantly differ from that during waking, whereas response amplitude to decrease during sleep. Although Rodriguez et al. (1986) emphasized that this technique may be inaccurate if the intensity relationships of amplitude and phase are not linear. Linden’s description and the results in the present study suggest that phase estimation is a useful technique for detecting AMFR threshold even in the sleeping state.

261

Conclusions

Although the amplitude-modulation following response (AMFR) evoked by a modulation frequency (MF) of 40 Hz at an intensity of 50 dBnHL showed a robust potential in awake adults, amplitude decreased during sleep. However, when detected by phase spectral analysis and spectral amplitude analysis using fast Fourier transform, 40 Hz AMFR was clearly detected even in sleeping adults. In contrast, AMFRs at MFs of 80-100 Hz were more frequently observed during sleep than waking, and the corresponding CSMs of 80-100 Hz AMFR during sleep were significantly higher than during waking. In children, 40 Hz AMFR was not clearly observed even by phase spectral analysis or spectral amplitude analysis. At MFs of 80-100 Hz, however, AMFRs were frequently observed visually and CSMs were extremely high in children during sleep. Modulation frequencies from 80 to 100 Hz would thus appear optimal for detecting AMFR during sleep, especially in children.

References Aoyagi, M., Fuse, T., Yokota, M., Suzuki, T., Kim, Y., Inamura, K. and Koike, Y. (1991a) An application of phase spectral analysis to the steady-state evoked response elicited by amplitude-modulated tone. Audio]. Jpn. 34, 149-157. Aoyagi, M., Fuse, T., Suzuki, T., Kiren, T., Kim, Y. and Koike, Y. (1991b) Auditory steady-state response while awake and during sleep: Automatic threshold detection by phase spectral analysis. Plactica Otologica (Kyoto) Suppl. 51, l-9. Batra, R., Kuwada, S. and Maher, V.L. (19861 The frequency-following response to continuous tones in humans. Hear. Res. 21, 167-177. Dobie, R.A. and Wilson, M.J. (19911 Optimal smoothing of coherence estimates. Electroencephalogr. Clin. Neurophysiol. 80, 194200. Dolphin, W.F. and Mountain, D.C. (1992) The envelope following response: scalp potentials elicited in the mongolian gerbil using sinusoidally AM acoustic signals. Hear. Res. 58, 70-78. Fridman, J., Zappulla, R., Bergelson, M., Greenblatt, E., Mails, L., Morrell, F. and Hoeppner, T. (19841 Application of phase spectral analysis for brain stem auditory evoked potential detection in normal subjects and patients with posterior fossa tumors. Audiology 23, 99-113. Galambos, R., Makeig, S. and Talmachoff, P.J. (1981) A 40-Hz auditory potential recorded from the human scalp. Proc. Nat]. Acad. Sci. USA. 78, 2643-2647.

Galambos, R. (1982) Tactile and auditory stimuli repeated at high rates (30-50 per set) produce similar event-related potentials. Ann. N.Y. Acad. Sci. 388, 722-728. Greenblatt, E., Zappulla, R., Kaye, S. and Fridman, J. (19851 Response threshold determination of the brainstem auditory evoked response: A comparison of the phase versus magnitude derived from the fast Fourier transform. Audiology 24, 228-296. Hotelling, H. (1931) The generalization of Student’s ratio. Ann. Math. Statist. 2, 360-378. Kuwada, S., Yin, T.C.T., Sykam J., Buunen, T.J.F. and Wickesberg, R.E. (1984) Binaural interaction in low-frequency neurons in the inferior colliculus of the cat. IV. Comparison of monaural and binaural response properties. J. Neurophysiol. 51, 1306-1325. Kuwada, S., Batra, R. and Maher, V.L. (19861 Scalp potentials of normal and hearing-impaired subjects in response to sinusoidally amplitude-modulated tones. Hear. Res. 21, 179-192. Linden, R.D., Campbell, K.B., Hamel, G. and Picton, T.W. (1985) Human auditory steady state evoked potentials during steep. Ear Hear. 6, 167-174. Maurizi, M., Almadori, G., Paludetti, G., Ottaviani, F., Rosignoli, M. and Luciano, R. (19901 40-Hz steady-state response in newborns and in children. Audiology 29, 322-328. Mlkela, J.P., Karmos, G., Molnar, M., Cstpe, V. and Winkler, I. (1990) Steady-state response from the cat auditory cortex. Hear. Res. 45, 41-50. Okitsu, T. (19841 Middle components of the auditory-evoked response in young children. Stand. Audio]. 13, 83-86. Picton, T.W., Vajsar, J., Rodriguez, R. and Campbell, K.B. (1987a) Reliability estimates for steady-state evoked potentials. Electroencephalogr. Clin. Neurophysiol. 68, 119-131. Picton, T.W., Skinner, C.R., Champagne, S.C., Kellett, A.C. and Maiste, A.C. (1987bl Potentials evoked by the sinusoidal modulation of the amplitude or frequency of a tone. J. Acoust. Sot. Am. 82, 165-187. Rees, A., Green, G. and ‘Kay, R.H. (1986) Steady-state evoked response to sinusoidally amplitude-modulated sounds recorded in man. Hear. Res. 23, 123-133. Rickards, F.W. and Clark, G.M. (19841 Steady state evoked potentials to amplitude-modulated tones. In: R.H. Nodar and C. Barber (Eds.1, Evoked Potentials II, Butterworth, Boston. pp. 163168. Rodriguez, R., Picton, T.W., Linden, R.D., Hammel, G. and Laframboise, G. (19861 Human auditory steady state responses: effects of intensity and frequency. Ear Hear. 7, 300-313. Spydell, J.D., Pattee, G. and Goldie, W.D. (19851 The 40 Hertz auditory event-related potential: normal values and effects of lesions. Electroencephalogr. Clin. Neurophysiol. 62, 193-202. Stapells, D.R., Galambos, R., Castello, J.A. and Makeig, S. (1988) Inconsistency of auditory middle latency and steady-state responses in infants. Electroencephalogr. Clin. Neurophysiol. 71, 289-295. Suzuki, T. and Kobayashi, K. (19841 An evaluation of 40-Hz event-related potentials in young children. Audiology 23, 599-604. Victor, J.D. and Mast, J. (1991) A new statistic for steady-state evoked potentials. Electroencephalogr. Clin. Neurophysiol. 78, 378-388.