Secondary Logo

Journal Logo

The Auditory Steady-State Response Is Not a Suitable Monitor of Anesthesia

Pockett, S., PhD; Tan, S. M., PhD

doi: 10.1097/00000539-200211000-00041

Previous studies show that the human 40-Hz auditory steady-state response (ASSR) disappears on induction of general anesthesia, suggesting that it may be a good candidate for a monitor of anesthesia. In this study, we aimed to learn whether all normal alert adults display ASSRs with adequate signal-to-noise ratio. Clicks were presented at a series of frequencies between 35 and 70 Hz and electroencephalographic records taken at the vertex were Fourier transformed. ASSRs were observable as sharp peaks in the electroencephalograph spectrum at the frequency of the clicks. Initial results showed that a discernible ASSR could not be obtained from about half the subjects studied at any click frequency used. Further investigation revealed that in subjects whose ASSR was undetectable in the alert state, induction of a drowsy mental state resulted in appearance of an observable ASSR. This was attributable to an increase in signal in the drowsy state, not to a decrease in noise. We conclude that, because a significant proportion of subjects do not display easily recordable ASSRs when alert, it is not practical to use disappearance of the ASSR as a routine test for adequacy of anesthesia.

Department of Physics, University of Auckland, Auckland, New Zealand

Financial support was provided by Technology New Zealand.

July 9, 2002.

Address correspondence and reprint requests to S. Pockett, PhD, Department of Physics, University of Auckland, Private Bag 92019 Auckland, New Zealand. Address e-mail to s.pockett@auckland.

The 40-Hz auditory steady-state response (ASSR) is an electroencephalograph (EEG) waveform evoked by auditory stimuli presented fast enough so that the responses to individual stimuli overlap. ASSRs are reported (1,2) to disappear during anesthesia caused by most kinds of general anesthetic, although ketamine is reported (3) to cause a slight increase in ASSR amplitude. Because most anesthetics reduce or abolish ASSRs, it has been suggested that these waveforms are possible candidates for use in monitoring the level of consciousness during anesthesia (1).

We investigated the suitability of ASSRs as monitors of anesthesia by determining (a) whether we could record these potentials with a useful signal-to-noise ratio from all normal awake humans, and (b) whether slightly different stimulus rates would be optimal for different subjects. With these aims in mind, we depart from tradition in the present report and present data from single subjects, rather than using the conventional approach of reporting only grand averages over a number of subjects. The rationale for this is that it is data from single subjects that are important in the operating room. The common practice of using grand averages works very effectively to obscure the differences between subjects, as well as the normal fluctuations that occur over time in the evoked potentials of individual subjects.

Back to Top | Article Outline

Materials and Methods


Data were collected from 17 normal adult subjects (5 women), ranging in age from 23 to 68 yr. The Human Ethics Committee of the University of Auckland approved the study, and all subjects gave informed consent. Hearing thresholds were not tested, but all subjects reported easily hearing the stimuli that were presented.

Back to Top | Article Outline


Acoustic stimuli consisting of trains of clicks at different frequencies were presented binaurally through Phillips SBC HL120 headphones. In preliminary experiments on 10 subjects, clicks were presented continuously at 40 Hz. The loudness and length of each click were varied widely, and online analysis was performed. In more structured experiments on a further 7 subjects (reported in the Figures), clicks consisting of 3 cycles of a 1000-Hz tone at 60-dB sound pressure level were used. A series of 7 20-s trains of clicks were presented, each train at a different frequency (35, 40, 45, 55, 60, 65, and 70 Hz), with the order in which the frequencies were presented pseudorandomized. This series was immediately followed by a 2-min period of 10-Hz clicks for the purpose of obtaining transient evoked potential data. The whole sequence (i.e., the 7 different frequencies plus 2 min of 10 Hz) was repeated 5 times while the subject was in a state of relaxed alertness (reading or writing) and then another 5 times immediately after consumption of approximately 30 mL of 37.2% alcohol diluted with tomato juice, with the lights turned down and with no reading matter or other distraction. These experiments were conducted just after lunch. Data were collected with subjects seated in a comfortable office chair, in a room that was quiet internally but with no attempt to screen out noises external to the room.

Back to Top | Article Outline

EEG Recording and Analysis

EEG signals were obtained by using one gold-plated cup electrode at Cz, referred to one mastoid with a ground electrode on the other mastoid. Interelectrode impedances were always below 8 kohm. The EEG was amplified, AC coupled with a corner frequency of 0.16 Hz, and low-pass filtered at 200 Hz using a fourth-order Butterworth filter (24 dB/octave roll-off). It was sampled at 500 Hz and digitized to 12-bit resolution, with full-scale voltage corresponding to 590 μV. Frames of 2-s duration were recorded in files, which also contained the times at which the stimuli were presented. These files were stored for subsequent processing in Matlab.

To obtain information about ASSRs, the EEG data in each 2-s frame, together with a representation of the stimulus (either 1 or 0, depending on whether the click was present at the time of the sample), were fast Fourier transformed. This produced a spectrum with 1000 points and a frequency spacing of 0.5 Hz for each frame. An online display of the running average of 10 consecutive 2-s frames was produced, in which the ASSR could often be seen as sharp peaks at multiples of the click frequency. Pickup of ambient 50-Hz mains noise routinely caused artifactual peaks at 50, 100, and 150 Hz, which were disregarded. Raw EEG was also displayed during the experiment. Occasional bursts of noise caused by jaw muscle activity were seen and the analog-to-digital converters would occasionally saturate. These artifacts were rejected by computing the standard deviation of the recorded signal in each 2-s frame and discarding any frames in which the standard deviation was >2.5 times or <0.4 times the median standard deviation computed over an entire 20-min session.

Back to Top | Article Outline

ASSR Signal-to-Noise Ratio

To establish the presence of an ASSR, the spectral data within each 20-s period over which a single stimulus frequency was presented were averaged. It was decided to use signal-to-noise ratio as an index of the presence of ASSRs, because this is more easily mea-sured than absolute signal levels and more useful when dealing with equipment that might not be carefully calibrated during routine use. To produce a signal-to-noise ratio, the power in the 0.5-Hz wide bin at which the stimulus was presented was divided by the average of the power in 4 frequency bins at ±1 Hz and ±2 Hz away from the stimulus frequency. Confidence limits for establishing the presence of a signal in the noise were found by using the f (variance ratio) distribution (4). If there is no signal present, the noise at each frequency is independently Gaussian distributed. The expected value of the power is proportional to the variance of the Gaussian distribution, so the signal-to-noise ratio as calculated above is equal to the ratio of the variance of the distribution at the stimulus frequency to the average of the variances of the components in the neighboring spectral bins. The f (variance ratio) distribution thus gives the likely range of the signal-to-noise ratio in the absence of signal. When presenting the graphs of signal-to-noise ratios, we have drawn 4 lines corresponding to the 1%, 5%, 95%, and 99% points of the appropriate f distribution. If the signal-to-noise ratio lies above the 99% line, for example, there is <1% probability that a signal-to-noise ratio of that size could be caused by noise alone.

Back to Top | Article Outline

Auditory Middle Latency Responses

Transient evoked potentials (auditory middle latency responses) were computed by averaging the responses to all of the 10-Hz stimuli presented during a particular arousal condition, i.e., 5 2-min periods of 10-Hz stimuli, making a total of 6000 clicks. The result was filtered with a 50-Hz notch filter to remove mains frequency noise.

Back to Top | Article Outline

γ and θ Power

In an attempt to obtain some objective measure of arousal or drowsiness, γ and θ power in the raw EEG were measured by averaging the squares of the Fourier components (the power spectral density) in the range 4–7 Hz for θ power and 30–70 Hz for γ power. Bins between 49 and 51 Hz, and the bin in which the stimulus was presented, were excluded for the purposes of this calculation. The results were presented as average power spectral densities in each band.

Back to Top | Article Outline


Results from the 10 preliminary subjects showed that discernible ASSRs could be recorded from only 6 of them. In the subjects in whom ASSRs could be recorded, online analysis showed that the signal-to-noise ratio of the ASSRs varied from moment to moment (and occasionally approached zero) for any given subject. Various hypotheses to explain the inter- and intrasubject variability of ASSR signal-to-noise ratio were formed and failed to survive initial tests. Finally, it was inferred that the underlying cause of both kinds of variability may be that, in some subjects, the ASSR was relatively large when the subject was drowsy and small or nonexistent when the subject was alert. The experiments reported below were performed on a further seven subjects, with the aims of (a) determining whether different stimulus frequencies were optimal for different subjects, and (b) testing the hypothesis about the correlation of ASSRs with drowsiness.

Back to Top | Article Outline

Optimal Stimulus Frequency for Individual Subjects

A flattish inverted-U-shaped tuning curve was often seen, with different subjects having different optimal stimulus frequencies, ranging from 40 to 60 Hz (Fig. 1). However, some subjects showed an essentially flat tuning curve over the range of stimulus frequencies tested, with good ASSRs still in evidence at stimulus rates of 70 Hz.

Figure 1

Figure 1

Back to Top | Article Outline

ASSRs and Arousal

During the “alert” period of the experiment, subjects read or wrote material of their own choice. They both appeared objectively to be and reported subjectively being in a state of relaxed alertness. During the post-alcohol period, subjects were asked by the experimenter not to converse and had no reading matter or other diversion. Objectively, they appeared to be drowsy (eyelids drooping or closed, occasional head-nodding). When asked after the experiment, they reported experiencing varying degrees of drowsiness in this period, from none at all to significant drowsiness with brief periods of sleep. When quietly asked their mental status during that part of the experiment, they almost always responded and reported being at least marginally awake. Common replies were “drifting,” “dreaming away,” “not sure,” and (most commonly) “mmm.”

With regard to ASSRs during the conditions of alertness or drowsiness, subjects could be classified into one of two groups. In the first group (subjects 1–3 in Fig. 1), the signal-to-noise ratios for the ASSRs induced by all stimulus frequencies were greater in the drowsy condition than in the alert condition. In most of these subjects, ASSRs could barely be recorded at all in the alert condition, even though the subjects always reported being conscious of the clicks when asked and sometimes said they were actually concentrating on the sound. There was no change in overall θ or γ power between the two conditions (Table 1).

Table 1

Table 1

Subjects in the second group (subjects 4–6 in Fig. 1) had relatively large ASSR signal-to-noise ratios in the alert state. In this group of subjects, there was no change in ASSR signal-to-noise ratio between the two arousal conditions, even when subjective drowsiness was reported to be marked in the low-arousal condition and the occasional sleep spindle was recorded.

In one subject from this group (an individual not included in Fig. 1), a very marked difference was observed during the alert condition between the EEG recorded while the subject was quietly writing equations and that recorded during a brief period while he was talking animatedly to the experimenter. While the subject was animated, his overall γ power increased 10-fold and the ASSR disappeared completely (Fig. 2). As soon as he returned to being quiet, both effects reversed. However, as with subjects in Group 1, there was no significant change in overall θ or γ power between the two arousal conditions, for any of this group of subjects (Table 1).

Figure 2

Figure 2

Back to Top | Article Outline


Effects of Arousal

In the following discussion, the presleep state induced by a small amount of alcohol combined with deliberate removal of external arousing stimuli will be referred to as “drowsiness,” or “low arousal.” As explained in detail in Materials and Methods, the appropriateness of this label is confirmed for all subjects by both behavioral observations and subjective report.

One of the major features of the data presented is that in one group of subjects, the signal-to-noise ratio of the ASSR is very low when the subject’s general state of arousal is relatively high, and increases when the subject’s arousal decreases. This is the opposite of what might have been expected, given that both moderate to large doses of alcohol and frank sleep have been reported to decrease the ASSR (5,6).

The most immediately attractive interpretation of our finding is that when these subjects are aroused, the level of biological “noise” generated by the electrical activity of the brain is high enough to swamp the signal, whereas when they become drowsy, the noise level decreases and the stimulus-evoked signal becomes detectable. If this hypothesis is correct, the second group of subjects (in whom the ASSR signal-to-noise ratio is relatively large in both conditions of arousal) should have low biological noise in the supposedly aroused condition and this should not change as arousal decreases. Based on this idea, the two groups might be redefined as a group who are normally highly aroused (and thus “noisy” in terms of EEG power in the γ frequency range) and a group who are normally less aroused (and thus γ-quieter).

Such an interpretation is initially supported by the example shown in Figure 2, where for a brief period, one subject from the supposedly quieter group became highly aroused, with a marked increase in γ power and concomitant disappearance of the ASSR signal. However, the hypothesis can probably be considered to be disproved by the findings that (a) overall γ power did not change between the two arousal conditions, in either group of subjects, and (b) overall γ power was not significantly different between the two groups of subjects, in either arousal condition.

This lack of decrease in overall γ power with the onset of the state here labeled “drowsy,” together with the lack of increase in overall θ power, might be taken as meaning that the arousal state of the subjects in fact did not change between the two conditions. However, against this are behavioral observations (in the low arousal condition, the subjects’ eyelids drooped or closed and their heads sometimes nodded involuntarily, whereas in the aroused condition, they were obviously awake), subjective reports from the subjects (that they were drifting, dreaming, nodding off to sleep, and generally feeling drowsy in the low-arousal condition but were awake in the alert condition), and in some cases actual observations of sleep spindles in the raw EEG during the putative low-arousal state. Drowsiness is a condition that is notoriously difficult to define from observation of the EEG (7,8), so the absence of a change in overall γ or θ power cannot be considered to provide good evidence one way or the other.

The lack of change in γ power does, however, suggest that in those subjects in whom the ASSR signal-to-noise ratio increased in the low-arousal condition, it was the signal level that changed and not the noise level. This suggestion is supported by the fact that the increase in ASSR signal-to-noise ratio with decrease in arousal was paralleled by an increase in the amplitude (but no change in the latency) of the transient auditory evoked potential (Fig. 3).

Figure 3

Figure 3

The cause of this increase of both transient and steady-state evoked potentials with decreased arousal is not clear. What is clear is that the present findings reinforce the suggestion (9) that general arousal level is a factor that must be considered in the interpretation of EEG data in general and evoked potential data in particular. This may be seen as unfortunate, because general physiologic arousal is a concept that is difficult to define clearly and thus hard to measure, let alone control. However, the finding that some kinds of evoked responses may be simply not there in one waking state of arousal, whereas they appear (with no other change in recording conditions) when arousal changes, is not one that can readily be ignored.

Previous work on the influence of arousal on event-related potentials shows a mixture of results. Some investigators find that the amplitude of late evoked potentials increases with decreasing arousal levels, as in the present work (10). However, different researchers find that somatosensory evoked potentials decrease during drowsiness (11). Yet other workers describe an inverted-U-shaped curve for the readiness potential, with greater amplitudes at medium levels of arousal than at either high or low levels of arousal (12). A fourth group finds no change in evoked potentials with arousal (13). These apparently conflicting findings are probably attributable to the fact that each of the studies examined a different event-related potential and they all used different definitions of arousal.

Back to Top | Article Outline

ASSR Signal-to-Noise Ratio and Conscious Auditory Perception

Despite the presently perceived importance of γ oscillations in consciousness, there remains a lack of clarity in our understanding of what the “40-Hz” ASSR represents in terms of auditory perception. The response disappears during sleep (14) and also during anesthesia with most kinds of anesthetic (1,2), so at first blush it might be considered to be a neural correlate of auditory consciousness. However, unconsciousness (or at least lack of responsiveness) induced by the dissociative anesthetic ketamine is actually accompanied by a slight increase in the amplitude of the ASSR (3). Also, there is no evidence that the ASSR is affected by selective attention (15), and masking studies show that changes in the ASSR often fail to predict changes in a listener’s subjective reports (16). Thus, the correlation between the ASSR and conscious auditory perception is far from perfect. The present results support the conclusion that ASSR signal-to-noise ratio does not covary with auditory perception, because at least in some subjects, the ASSR was bigger when the subject was drowsy and reportedly not particularly aware of the sound of the clicks. It was also notable that in some subjects, the signal-to-noise ratio was minimal in the alert condition, even when the subjects reported that they were perfectly aware of the clicks and in fact concentrating on the sound.

These results suggest that the ASSR is not an ideal candidate as the basis of an anesthesia monitor.

We thank Harry Oudenhoven, who constructed the EEG amplifier and assisted with its programming.

Back to Top | Article Outline


1. Plourde G, Picton TW. Human auditory steady state response during general anesthesia. Anesth Analg 1990; 71: 460–8.
2. Plourde G, Villemure C. Comparison of the effects of enflurane/N2O on the 40-Hz auditory steady state response versus the auditory middle-latency response. Anesth Analg 1996; 82: 75–83.
3. Plourde G, Baribeau J, Bonhomme V. Ketamine increases the amplitude of the 40-Hz auditory steady state response in humans. Br J Anaesth 1997; 78: 524–9.
4. Abramowitz M, Stegun IA. Handbook of mathematical functions. Mineola, NY: Dover Publications, 1965.
5. Cohen LT, Rickards FW, Clark GM. A comparison of steady-state evoked potentials to modulated tones in awake and sleeping humans. J Acoust Soc Am 1991; 90: 2467–79.
6. Jaaskelainen IP, Hirvonen J, Saher M, et al. Dose-dependent suppression by ethanol of transient auditory 40-Hz response. Psychopharmacology 2000; 148: 132–5.
7. Santamaria J, Chiappa KH. The EEG of drowsiness in normal adults. J Clin Neurophysiol 1987; 4: 327–82.
8. Santamaria J, Chiappa KH. The EEG of drowsiness. New York: Demos, 1987.
9. Pockett S. The nature of consciousness: a hypothesis. New York: Iuniverse Inc., 2000.
10. Michida N, Ebata A, Tanaka H, et al. Changes of amplitude and topographical characteristics of event-related potentials during the hypnagogic period. Psychiatry Clin Neurosci 1999; 53: 163–5.
11. Beydoun A, Morrow TJ, Shen JF, Casey KL. Variability of laser-evoked potentials: attention, arousal, and lateralized differences. Electroencephalogr Clin Neurophysiol 1993; 88: 173–81.
12. Masaki H, Takasawa N, Yamazaki K. Human movement-related potentials preceding voluntary movements in different arousal states monitored with skin potential level. Percept Mot Skills 2000; 90: 299–306.
13. Khachaturian ZS, Gluck H. The effects of arousal on the amplitude of evoked potentials. Brain Res 1969; 14: 589–606.
14. Suzuki T, Kobayashi K, Umegaki Y. Effect of natural sleep on auditory steady state responses in adult subjects with normal hearing. Audiology 1994; 33: 274–9.
15. Linden RD, Picton TW, Hamel G, Campbell KB. Human auditory steady state evoked potentials during selective attention. Electroencephalogr Clin Neurophysiol 1987; 66: 145–59.
16. Galambos R, Makeig S. Physiological studies of central masking in man. II. Tonepip SSRs and the masking level difference. J Acoust Soc Am 1992; 92: 2691–7.
© 2002 International Anesthesia Research Society