An examination of evoked K-complex amplitude and frequency of occurrence in the elderly


Ian M. Colrain SRI International, 333 Ravenswood Avenue, Menlo Park, CA 94025, USA (Tel.: +1 650 859 3915; fax: 650 8592743; e-mail:


The elderly consistently display lower levels of slow wave sleep (SWS), primarily because of the small amplitude of their delta activity. Given that delta electroencephalogram (EEG) and K-complexes are thought to be generated by the same mechanisms, it was hypothesized that K-complex amplitude and rate of production would be lower in the elderly. K-complex amplitude was assessed by averaging K-complex responses to auditory stimuli, and measuring the amplitude of the N550 component of the averaged evoked response. Ten young (five males and five females; mean age 23.10 ± 5.36 years) and 10 elderly adults (six males and four females; mean age=75.60 ± 4.48 years) who were neurologically healthy and free from medication spent two non-consecutive nights in the sleep laboratory. EEG was recorded from six gold plate electrodes (Fz, Fcz, Cz, Cpz, Pz and O2) referenced to A1 + A2. Tone clicks (1000 Hz), of varying intensity from 70 to 100 decibels above measured awake detection threshold, were presented binaurally during stage 2 sleep. Responses were classified according to whether they produced: a K-complex, a vertex sharp wave (VSW), both of these responses or neither response. They were then averaged separately for each response type. The elderly showed a substantially smaller N350 (averaged VSWs) and N550 amplitudes compared with the young subjects. The elderly also showed an augmented but delayed P2 component, followed by a long-lasting positive EEG shift. The smaller amplitude of the averaged K-complex N550 component is consistent with lower delta amplitudes previously reported in the elderly and with the hypothesis that K-complexes and delta activity share the same generator mechanisms. The enhanced P2 component and the long-lasting positive deflection in the EEG in the elderly indicate the existence of age differences other than smaller EEG amplitude.


There is substantial literature describing the evoked electroencephalogram (EEG) response to external and internal stimuli during sleep. It indicates that the arousal/attention related N1 component (Naatanen and Picton 1987) is diminished and often absent (Campbell et  al. 1992). The predominant non-rapid eye movement (NREM) components are negativities at around 300–400 ms (N350) and around 500–700 ms (N550). A positive P2 component is often reported at around 200 ms. A second positivity is seen at approximately 450 ms (P450), although it is unclear as to whether this is an independent component or a partial return to baseline between N350 and N550 produced by the low frequency EEG filter. Thus, this literature provides a clear picture as to the pattern of the evoked response in young adults during NREM sleep. However, none of these studies have specifically evaluated subjects over the age of 45.

The N550 component is observed when K-complexes, evoked by stimuli, are included in the averaged response (Bastien and Campbell 1992, 1994; Campbell et  al. 1990; Colrain et  al. 1999, 2000b; Cote et  al. 1999; Gora et  al. 2001; Harsh et  al. 1994; Niiyama et  al. 1994, 1995; Sallinen et  al. 1994). It has a fronto-central topographic maximum (Colrain et al. 1999, 2000b; Cote et  al. 1996; Gora et  al. 1999, 2001; Niiyama et al. 1995; Webster and Colrain 1998), and in Stage 2 sleep has an amplitude in the range of 60–120 μV. It is apparent in stage 2 and slow wave sleep (SWS), and is larger, with an earlier peak latency, in SWS (Cote et al. 1999). The N350 component now appears to be related to vertex sharp waves (VSW), rather than K-complexes (Colrain et al. 2000b; Gora et al. 2001; Harsh et al. 1994). It has a vertex maximum distribution (Colrain et  al. 2000a, 2000b; Gora et al. 1999, 2001; Webster and Colrain 1998), and an amplitude at Cz in the range of approximately 35–45 μV (in the VSW averages). It is apparent early in sleep onset and can be seen during stage 1 sleep (Ogilvie et  al. 1991), particularly when stimuli are presented during periods of theta activity (Colrain et al. 2000b; Gora et al. 1999). Moreover, Gora et al. (2001), has recently reported an enhanced N550 component in averages of responses that contain both a VSW and a K-complex, relative to K-complexes averaged separately.

SWS delta is likely to have a common generator mechanism with K-complexes, which are delta frequency events. SWS is defined by the relative percentages of delta wave activity seen in the EEG, with at least 20% of an epoch displaying 75-μV delta activity (Rechtschaffen and Kales 1968). Delta waves (1–4 Hz) are generated within the thalamus during deep sleep by the interplay between two intrinsic currents of thalamocortical cells: the low threshold transient Ca2+ current and a hyper-polarization-activated cation current (h-current) (McCormick and Pape 1990). Reports indicate that it is unlikely that thalamocortical cells have an intrinsic synchronization mechanism (Steriade and Deschenes 1984). Therefore, an external synchronization structure would be required for the intrinsic delta activity to be reflected in the EEG. Although these structures have not yet been identified, Amzica and Steriade (1998b) have proposed that delta synchronization may occur in the reticular nucleus of the thalamus and/or the cerebral cortex.

Given that cortical and thalamic networks interact so extensively during sleep generation, it is possible that the slow oscillation may act to coordinate the occurrence of corticothalamic rhythms. Amzica and Steriade (1998a) argue that the K-complex is rhythmic at the pace of the slow oscillation and that each depolarizing–hyperpolarizing cycle of the slow oscillation corresponds to a K-complex in the cortical EEG. They suggest that in cats, the intracellular depolarization of cortical neurons corresponds to the large amplitude negative peak of the K-complex and that the intracellular hyperpolarization of cortical neurons corresponds to the slow wave shape of the positive peak of the K-complex. The link between slow oscillations and K-complex-like bursts of magnetoencephalographic activity has also recently been described (Simon et  al. 2000). Consistent with this, K-complexes have been shown to be more likely to precede SWS periods than rapid eye movement (REM) periods (De Gennaro et  al. 2000). This last result led the authors to support a hypothesis originally put forward by Loomis et al. (1939) that K-complexes are the forerunner of delta sleep.

The macrostructure of sleep in the elderly is characterized by a decrease in SWS (Bliwise 1993; Blois et  al. 1983; Carskadon and Dement 1985; Ehlers and Kupfer 1989; Feinberg et  al. 1967; Larsen et  al. 1995; Reynolds et al. 1991; Webb 1982). This probably reflects the end point of a gradual reduction that may commence as early as late adolescence (Ehlers and Kupfer 1989; Larsen et  al. 1995). It is important to note, however, that the reduction in SWS primarily reflects a decrease in the amplitude of delta activity, rather than the absence of slow frequency activity (Bliwise 1993). A number of studies have shown that older women have better-preserved SWS than men (Rediehs et  al. 1990; Wauquier et  al. 1992; Webb 1982) and Ehlers and Kupfer (1997) have recently reported that women are largely spared the dramatic reduction seen in SWS in men between the ages of 20 and 40.

It has been well established that ageing is associated with reduced grey matter (Coffey et al. 1992; Itoh et al. 1990; Laffey et al. 1984; Miller et al. 1980; Pfefferbaum et al. 1994, 1999; Sullivan et al. 1993). Further, men may exhibit more age-related cortical volume loss than women (Blatter et al. 1995; Coffey et al. 1998; Cowell et al. 1994; Gur et al. 1991; Matsumae et al. 1996; Murphy et al. 1996; Pfefferbaum et al. 1994; Raz et al. 1997). It thus seems most parsimonious to assume that the age-related reduction in cortical mass leads to a reduction in delta generation capacity and delta amplitude and thus a reduction in SWS. Given the relationship between K-complexes and delta waves outlined above, it was hypothesized that evoked K-complexes would have lower amplitudes in the elderly and a lower likelihood that a K-complex will be elicited by a stimulus, as compared with younger subjects.

Predictions as to the effect of ageing on VSWs and their associated N350 evoked potential component during stage 1 sleep are less clear. This is because, given the shorter period of the waveforms, they possibly reflect theta frequency activity. Thus, it is less clear whether the proportion of VSWs elicited, or the amplitude of the N350 component would be lower in an aged population. Further, given the recent findings that responses involving both a VSW and a K-complex leads to a larger N550 component in young people (Gora et al. 2001), it was of interest to see if this effect was also present in the elderly.

As has been noted above, event related potentials (ERPs) have not been studied in the elderly during sleep. Thus, the first aim of the study was to test the hypothesis that the amplitude of the evoked K-complex and its probability of elicitation are lower in the elderly as compared with younger subjects. The second aim was to explore the effect of ageing on the amplitude and probability of elicitation of VSWs.



Ten young (five males and five females; mean age 23.10 ± 5.36 years) and 10 elderly subjects (six males and four females; mean age=75.60 ± 4.48 years) participated in the experimental nights. Older subjects underwent a physical examination and a neurological evaluation, including medical, psychiatric and sleep history. Younger subjects were screened using a questionnaire. All participants were non-smokers, free from respiratory, auditory or sleep-related disorders. None were using medications with a known effect on the central nervous system (CNS) or sleep. Subjects were asked to refrain from alcohol and coffee consumption for at least 12 h prior to participation. Hearing was verified using an audiometer to be within 15 dB ISO at 1000 Hz frequency, and each subject's auditory threshold level was established.

Measurement of EEG and digital recordings

The EEG was recorded from Grass gold-cup electrodes placed according to the International 10–20 system at Fz, FCz, Cz, CPz, Pz and O2 and referenced to linked ears. A horizontal electro-oculogram (EOG) was recorded from electrodes placed at the outer canthus of each eye. A ground electrode was placed on the forehead. Electrode impendence values were maintained below 5 kΩ. The Cz and O2 EEG, EOG and submentalis electromyography (EMG) signals were continuously recorded on a Compumedics automated sleep acquisition system at a sampling rate of 500 Hz. Band pass filters were set at 0.1–40 Hz for EEG, 0.1–40 Hz for the EOG and 3–40 Hz for the EMG. These data were used to determine sleep stage using modified Rechtschaffen and Kales' criteria (Rechtschaffen and Kales 1968).

A Neuroscan system continuously recorded all EEG sites and the EOG signal at a sampling rate of 1000 Hz. The bandpass filters were set to 0.1 and 500 Hz. Data were stored on hard disk for later off-line analysis. A total of 2000 data points were sampled beginning 500 ms prior to stimulus onset and continuing for 1500 ms after it.

Auditory stimuli

Auditory stimuli were 1000 Hz tone pips having 52 ms duration and a rise and fall time of 2 ms. Stimulus intensity was varied randomly at 70, 80, 90 or 100 dB above awake detection thresholds and presented with a variable interstimulus interval (ISI) of 15–30 s. Tones were digitally generated by a 16-Bit Sound BlasterTM (SB-16, US Robotics) sound card attached to an IBM compatible computer. Stimuli were presented binaurally via E-A-RTONETM 3A insert earphones (Auditory Systems, Indianapolis, IN, USA), which ensured consistency of auditory input in the subject despite changes in head or body position. All stimuli were measured and calibrated using a Bruel and Kjaer sound level meter 2260.


All participants underwent a protocol consisting of two non-consecutive nights. Subjects were instructed to arrive at the laboratory approximately 30 min prior to their usual bed time. Upon arrival at the laboratory, electrodes were affixed and the auditory device fitted. Presentation of auditory stimuli began after the commencement of consolidated sleep (i.e. 10 min following continuous stage 2 sleep). A minimum of 400 stimuli were presented in stable stage 2 sleep during each night. When subjects entered SWS or REM sleep or the EEG showed signs of arousal or motor artefact, stimulus presentation was halted and subject awoken and resumed only when stable stage 2 sleep was established.

Data scoring and component definition

Within periods retrospectively confirmed as being stage 2 NREM sleep, single trial ERPs were classified according to whether the stimulus elicited: a K-complex in the absence of a VSW (KC); a VSW in the absence of a K-complex (VSW); both a VSW and a K-complex (VSW/KC); and other responses in which neither a K-complex nor a VSW were elicited (OTHER). Examples of individual responses of each type for a young subject are presented in Fig. 1.

Figure 1.

. Examples of each of the four response types, taken from the same young subject. Panel A shows a K-complex (KC), with the arrow pointing to the most negative peak. Panel B shows a response with both a vertex sharp wave (VSW) and a K-complex (VSW/KC). Panel C shows a response classified as OTHER in which there is neither a K-complex nor a VSW. Panel D shows a VSW. Panels A and B show data from Fz. Panels C and D show data from Cz.

The K-complex was visually identified as a large amplitude waveform with a well delineated negative sharp wave, which was immediately followed by a positive component and which exceeded 0.5 s (Rechtschaffen and Kales 1968). In addition, a latency criterion was also applied. The most negative peak had to occur between 450 and 800 ms after the onset of the stimulus. Bi-phasic K-complexes (those with a double peak at Fz, see Fig. 1, panel A), were included in this definition, as were all other types previously identified (Paiva and Rosa 1991).

Vertex sharp waves were also visually scored and identified as a large negative sharp wave that had a duration of less than 500 ms. The negative wave also had to be preceded and followed by a positive wave (Dutertre 1977; IFSECN 1974). The amplitude had to be at least 50 μV at Cz and be maximal at the vertex (see Fig. 1, panel D). The VSW/KC category contained responses in which there was both a K-complex (measured at Fz) and a VSW (measured as a distinct event at Cz) (see Fig. 1, panel B). Biphasic K-complexes would not meet this definition.

The OTHER category could contain any activity not defined by these criteria. Thus, this category contained responses in which no response to the stimuli was visually discernible, or some other slow or fast wave activity was present which did not meet the criteria outlined above (see Fig. 1, panel C).

ERPs were quantified in terms of peak latencies and amplitudes of the maximum negative or positive deflections within specified latency ranges. The P2 component was initially measured at Cz as the maximum positive peak between 175 and 400 ms poststimulus. The N350 was initially measured at Cz as the maximum negative peak occurring between 250 and 400 ms and the N550 was identified at Fz as the maximum negative peak occurring between 400 and 800 ms. The amplitude of these components were then measured at all electrode sites based on peak-to-peak amplitudes from the previous peak of opposite polarity. For example, the peak amplitude of the N350 component was measured relative to the peak amplitude of the preceding P2 component.

Statistical analysis

Independent sample t-tests were used to examine differences between the two groups in terms of the percentage of phasic responses elicited to auditory tones. Amplitude values for the N350 and N550 component peaks were submitted to a mixed model analysis of variance (ANOVA), with age group as a between-subject factor and tone type and scalp site as within-subjects factors. The N350 component was only analysed on VSW phasic responses, while the N550 was only analysed on K-complex phasic responses. In addition, N350 and N550 amplitudes were further analysed comparing K-complex with VSW/KC responses, and VSW with VSW/KC responses. These comparisons were achieved using three-factor ANOVA models [two (groups) by two (phasic responses) by six (scalp site)].

No previous studies have sought to identify the relationship between the P2 and the presence or absence of K-complexes or VSWs. The P2 component was thus subjected to a mixed model ANOVA with age group as a between-subject factor and phasic response and scalp site as within-subject factors.

Comparison of different scalp distributions is fraught with methodological problems. This is because the usual ANOVA procedure assumes that experimental data is a consequence of a constant being added to (or subtracted from) a baseline condition. With EEG data, experimental effects usually result in a constant being multiplied to the scalp distribution data. McCarthy and Wood (1985) have developed a normalization procedure to correct for these non-additive effects. Amplitude values for each ERP component were therefore scaled using the McCarthy and Wood normalization method. Using the normalized data, scalp topographical differences between age groups was assessed with a two (age group) by six (scalp site) ANOVA model. Differences between distributions are indicated by a significant interaction term between the site and age group factors. Thus, the reported main effects of age and site are based on ANOVAs using the recorded EEG voltage values. The interaction effects between age and site are reported from ANOVAs using McCarthy and Wood transformed data.

Latency values were assessed by a two-way ANOVA (tone type by group). For all analyses, results were considered to be statistically significant at an alpha value of 0.05. Huynh–Felt values were determined, and used when appropriate to correct the degrees of freedom when sphericity assumptions were violated.


There were no significant main or interaction effects in either amplitude or latency for the N550, N350 or P2 components for the different tone intensities. Data were therefore combined across all tone types for both nights for all subsequent analyses.

The number of artefact-free responses available for analysis did not differ between the two groups, t(9)=0.29, P > 0.05. On average, the young subjects had responses to 361 ± 40 stimuli, and the older subjects had responses to 369 ± 20 stimuli available for analysis. Responses involving arousals were extremely rare, with several subjects in each group never arousing to any stimuli.

The young subjects produced significantly more isolated K-complexes (48%) than the older subjects (31%), t(13)=−2.53, P < 0.02. A VSW occurred on 7% of responses for young subjects and 11% for older subjects. This difference was not significant, t(13)=1.21, P > 0.05. Younger subjects produced both a K-complex and a VSW on 25% of responses. This was significantly greater than the older subjects (9%), t(13)=−3.04, P < 0.01. Finally, the younger subjects had many fewer OTHER responses (20%) than the older subjects (49%), t(13)=5.15, P < 0.01.

The N550 component

As illustrated in Fig. 2, the elderly showed a substantially reduced N550 amplitude in the KC condition compared with the young subjects, F1,18=54.65, P < 0.01. There was a significant site effect with the amplitudes largest at Fz, F1,5=156.65, P < 0.01 (Table 1), and based on data that had been normalized using the McCarthy–Woods correction procedure, there was a significant site by group interaction with the site-to-site differences being greater in the younger subjects, F1,5=23.05, P < 0.01. The younger subjects also displayed a significantly earlier N550 peak F1,18=5.55, P < 0.05.

Figure 2.

. Grand mean averages of responses eliciting a K-complex (KC and VSW/KC) in younger and older subjects at each recorded scalp site. Data are collapsed over all stimulus intensities. Thicker lines represent data from young subjects. Thinner lines represent data from the elderly. Waveforms are plotted 500 ms prior and 1500 ms following the onset of the stimulus. In this figure (and in the others), negative polarity is shown as an upward deflection. The difference between scale points on the Y axis is −16 μV in each case. LLP refers to the long-lasting positivity.

Table 1.   The means and standard deviations (SD) of the N550 component in averages of K-complexes (KC), vertex sharp waves (VSW), both responses (VSW/KC) and neither response (OTHER). Data presented are from all sites for both young and older subjects Thumbnail image of

The N550 component was largest in the averaged VSW/KC responses. Figure 3 compares VSW/KC averages with the KC averages. For the N550 component there was a significant main effect of phasic response condition, F1,18=28.75, P < 0.01 with the N550 amplitude being significantly greater in the VSW/KC responses (−111.46 μV) than in the KC responses (−96.74 μV) (values averaged over groups). The interaction of response type by group was in the expected direction of larger amplitudes in the younger subjects but this was not statistically significant.

Figure 3.

. Grand mean waveforms for the VSW/KC and KC response conditions at Fz. The thicker lines represent the young subjects and the thinner lines the older subjects. Black represents KC data. Grey represents VSW/KC data. Waveforms are plotted 500 ms prior and 1500 ms following the onset of the stimulus.

The N350 component

As illustrated in Fig. 4, the younger subjects displayed a significantly greater N350 amplitude than the elderly subjects, F1,18=15.01, P < 0.01. The main effect of site was significant, F1,5=104.69, P < 0.01, with fronto-central scalp sites showing the largest amplitude values (Table 2). The site by group interaction effect was also significant, F1,5=8.09, P < 0.01. The N350 latency was significantly earlier in the younger subjects than in the elderly, F1,18=5.56, P < 0.01.

Figure 4.

. Grand mean waveforms from all scalp sites for young and older subjects. Data plotted are from the vertex sharp wave (VSW) phasic response condition. The thicker line represents the young subjects and the thinner line the elderly subjects. Waveforms are plotted 500 ms prior and 1500 ms following the onset of the stimulus. The difference between scale points on the Y axis is −10.5 μV in each case. LLP refers to the long-lasting positivity.

Table 2.   The means and standard deviations (SD) of the N350 component in averages of K-complexes (KC), vertex sharp waves (VSW), both responses (VSW/KC) and neither response (OTHER). Data presented are from all sites for both young and older subjects Thumbnail image of

The N350 component also revealed a significant main effect of phasic response type, F1,18=14.89, P < 0.01 with an average amplitude of −71.75 μV for the VSW/KC responses compared to −62.79 μV for the VSW responses (Fig. 5) (values averaged over groups). There was also a significant condition by group interaction F1,18=8.94, P < 0.01, with younger subjects showing significantly greater N350 amplitudes in the VSW/KC responses than older subjects. The main effect of site was also significant with the largest amplitudes at Cz, F1,18=134.74, P < 0.01, as was the site by group interaction, F1,18=7.18, P < 0.01.

Figure 5.

. Grand mean waveforms for the VSW/KC and VSW phasic response conditions at Cz. The thicker line represents the younger subjects and the thinner lines the older subjects. Black is used to represent VSW data and grey to represent the VSW/KC data. Waveforms are plotted 500 ms prior and 1500 ms following the onset of the stimulus.

The P2 component

Table 3 presents the mean data of the amplitudes of the P2 component at all sites in the four different phasic response conditions. It was largest in the VSW responses, smaller but still prominent in the VSW/KC and KC responses and still observable in the OTHER responses. The main effect of phasic response condition was significant, F1,15=10.38, P < 0.05. As evidenced in Figs 2, 4 and 6, the P2 component was significantly larger in the older subjects relative to the younger subjects, F1,18=8.01, P < 0.01. The P2 component revealed a significant site effect being maximal in the fronto-central scalp sites, F1,18=54.65, P < 0.01 and a significant group by site interaction, F1,5=7.55, P < 0.01. Younger subjects also displayed a significantly earlier P2 peak, F1,18=33.84, P < 0.01.

Table 3.   The means and standard deviations (SD) of the P2 component in averages of K-complexes (KC), vertex sharp waves (VSW), both responses (VSW/KC) and neither response (OTHER). Data presented are from all sites for both young and older subjects Thumbnail image of
Figure 6.

. Grand mean waveforms at all scalp sites for responses containing neither a VSW nor a K-complex (OTHER). The thicker line represents the young subjects and the thinner line the elderly subjects. Waveforms are plotted 500 ms prior and 1500 ms following the onset of the stimulus. The difference between scale points on the Y axis is −5.5 μV in each case. LLP refers to the long-lasting positivity.

Long-lasting late positivity

An unexpected finding was the presence of a long-lasting late positive waveform in the elderly subjects. This was apparent in all averages but most obvious in the OTHER averages (Fig. 6). It was maximal at FCz and Cz where the peak amplitude was approximately 34 μV. It lasts to the end of the epoch and when the data were digitally re-referenced to a DC high frequency cut-off, the waveform did not return to baseline within 10 s.


Consistent with earlier findings of spontaneous K-complexes, older subjects also displayed the ability to produce evoked K-complexes and VSW responses to auditory stimuli during stage 2 NREM sleep. This was despite their low levels of SWS (Bliwise 1993; Blois et al. 1983; Carskadon and Dement 1985; Ehlers and Kupfer 1989; Feinberg et al. 1967; Larsen et al. 1995; Reynolds et al. 1991; Webb 1982). Further, the scalp topography of the N550 response from averaged K-complexes was maximal frontally in older subjects and the N350 from averaged VSWs was maximal centrally as they were in the younger subjects. However, the magnitude of the responses was smaller in the older subjects as was the proportion of responses on which phasic responses were elicited. The P2 was larger in the older subjects and they displayed a long-lasting positivity not seen in the younger subjects' data. Thus, the N550 and N350 data cannot be explained merely as being the result of an overall reduction in the amplitude of EEG in the elderly.

The results of the present study confirm the previously described relationship between K-complexes and the N550 component (Bastien and Campbell 1992, 1994; Campbell et al. 1990; Colrain et al. 1999; Cote et al. 1999; Harsh et al. 1994; Niiyama et al. 1994, 1995; Sallinen et al. 1994). The N550 component was considerably larger in the averages of responses that contained a K-complex than in responses in which a K-complex could not be identified. In addition, the frontal maximum of this component was similar to that reported in other AEP studies (Colrain et al. 1999; Cote et al. 1999). While these morphological features were essentially the same for the older subjects, the amplitude of the K-complex related N550 was substantially reduced.

It is possible, although unlikely, that the smaller N550 in the older participants was because of auditory system effects. The amplitude of N550 is relatively independent of the physical parameters of the stimulus. For example, Bastien and Campbell (1992) have noted that the amplitude of N550 did not vary with manipulation of stimulus intensity or stimulus frequency. Thus, while auditory thresholds increase with ageing, this should not affect the amplitude of N550. Moreover, in the present study, stimulus intensity was set relative to individual's threshold for hearing. It is possible that N550 habituated rapidly in the elderly. Again, there is little evidence that N550 is subject to habituation (Bastien and Campbell 1994). In young adults, the refractory period of N550 is very long, between 10 and 20 s and thus the reduced amplitude of N550 in the elderly might be explained by a substantially longer refractory period. However, the differences that Bastien and Campbell have reported in the amplitude of N550 as a consequence of the rate of stimulus presentation, were much smaller than the between group differences reported in the present study.

It is also possible that the group differences could be because of latency jitter. The peak of N550 might have been quite variable in the elderly. If this were the case, the averaged N550 will spread out in time (N550 would be less peaked) and the average amplitude of the individual trials would be smaller. However, in order for latency jitter to account for the very large difference in N550 amplitude between the younger and older subjects, there would have to be considerable trial-to-trial variability in its peak latency. This would result in a very broad negative waveform. Inspection of the KC and VSW/KC waveforms indicates that this is clearly not the case.

The generators of the K-complex are not known. Given its very high amplitude, it likely reflects the activity of highly synchronized cortical sources. Further, in view of its topography it is likely that frontal and prefrontal cortical grey matter are intrinsically involved. Ageing is associated with smaller grey matter volumes throughout the cortex (Pfefferbaum et al. 1992, 1994, 1999; Sullivan et al. 1995), and longitudinal analysis within subjects reveals greatest grey matter loss in the prefrontal cortex (Pfefferbaum et al. 1998). Thus, differences in frontal and prefrontal cortical grey matter might explain the large decrease in the amplitude of N550 with age.

The role of the N550 remains controversial. Two opposing processes have been suggested. One is that it reflects an arousal process, allowing subjects to awaken to highly relevant stimuli in the environment (Halasz 1993). The other is that it may serve to protect sleep (Wauquier 1995). In support of the latter view, Bastien et al. (2000) indicate that cortical arousal following stimulus presentation may be prevented if the stimulus elicits a K-complex. Within this context, the decrease in the proportion of K-complexes and the reduced amplitude of N550 in the elderly, are consistent with a failure of a sleep protection system. This would also explain the lowered thresholds for awakening and fragmented sleep in the elderly.

The present data also indicated a clear relationship between VSWs and the N350 component. The N350 was maximal in the averages of responses that included a VSW, although it was also present in the averages of K-complexes and OTHER responses. These results, in conjunction with the prominent vertex distribution, are consistent with other AEP studies (Colrain et al. 2000b; Gora et al. 2001; Harsh et al. 1994). Similar to the N550 component, the elderly subjects produced a clear VSW related N350 component, but of significantly smaller amplitude than that seen in the younger subjects.

Previous AEP research has indicated that the N350 may be part of the K-complex (Cote et al. 1999; Niiyama et al. 1995; Sallinen et al. 1997). This has been based on reports that the N350 was larger following the averaging of responses containing a K-complex, than following the averaging of responses not containing a K-complex. In the present study, the amplitude of both the N350 and N550 components were larger in responses that contained both phasic events (VSW/KC condition) than in averages of either event separately. These results, in addition to their different scalp topographies (vertex for the N350 and fronto-central for the N550) suggest that the N350 component in the VSW/KC responses added a degree of source strength (that is, more negativity) to the N550 component. However, it should be noted that in some cases the earlier negative peak of bi-phasic K-complexes, may also contribute to an N350 peak. This is a possible explanation for the small N350 seen in the K-complex average waveform (see Figs 1 and 2). Another possible explanation of the N350 small peak in the K-complex waveforms, the small N550 in the VSW waveforms and the very small N350 and N550 in the OTHER waveforms is the incorporation of responses that were either misclassified or in the case of the N350, consisting of small VSWs that were less than 50 μV.

The enhanced P2 component and the long-lasting positive deflection in the EEG of the elderly indicate that age differences exist other than reduced EEG amplitude. Our P2 findings are in general agreement with the wake ERP literature, which has shown the existence of an enhanced P2 component in older subjects. Garcia-Larrea et  al. (1992) reported a positivity around 250 ms, which appeared in response to non-targets in an auditory oddball task, compared with the ERPs obtained to the same stimuli, but in a neutral condition. The author related this component to an `excess of central processing of non-targets, relative to neutrals', which they believed reflected an intramodal focusing of attentional resources on the stimuli which belonged to the same modality as the target. Novak et  al. (1992) described a positivity in the difference waveforms resulting from subtracting the ERPs obtained in a simple reaction time task from the ERPs elicited by frequent tones in a pitch discrimination task. The authors interpreted this positivity as reflecting an attention-modulated process required for the performance of an auditory discrimination task. These results have been thought to indicate the existence of an ageing-related progressive deficit in the capacity to withdraw attentional resources from stimuli to which the task requires only marginal attention to be paid (Amenedo and Diaz 1998, 1999).

Pfefferbaum et  al. (1984) also reported an increased P2 component in the elderly, but interpreted these findings as a reflection of decreased negativity because of structural brain changes which accompany ageing. Interpreting P2 amplitude differences is difficult as it is overlapped by a number of other components. During sleep onset and within NREM sleep, the amplitude of an earlier negative wave, N1, decreases while P2 increases in amplitude. Campbell et al. (1992) have attributed this to the removal of a long-lasting negative wave, the Processing Negativity (PN) (Naatanen and Picton 1987). PN reflects that additional processing that occurs when subjects attend to acoustic input. When these stimuli are ignored, PN is reduced. For sleep to occur, the processing of all but the most relevant of sensory input must be inhibited. This results in a marked reduction of PN (Campbell et al. 1992; Cote et  al. 2000; de Lugt et  al. 1996). PN overlaps and summates with both N1 and P2. The removal of this negativity causes N1 to decrease and P2 to increase in amplitude. A reasonable interpretation of the increased P2 in the elderly is therefore that it reflects additional inhibition of auditory processing. Such an interpretation is, however, difficult to reconcile with the fact that the elderly have decreased thresholds for awakening (Zepelin et al. 1984). It could, of course, be argued that the need for greater inhibition in the elderly is because of the fact that they are more easily aroused.

The general pattern of results in this study suggest that the overall sleep AEP waveform in the elderly is characterized by a reduction in the negative components and an increase in the positive components. There is evidence to suggest that a reduction in cortical mass, especially frontal cortical mass, should lead to a reduction in delta generation capacity and reduced delta amplitude. There is also a growing body of literature indicating that K-complex activity and delta activity reflect similar neurophysiological processes (Amzica and Steriade 1997, 1998a; De Gennaro et al. 2000; Steriade et  al. 1993). Therefore, the existence of the age-related changes in K-complex elicitation and amplitude are most likely because of the diminished capacity of the ageing brain to generate thalamocortical delta generation because of the substantial loss of cortical tissue associated with normal ageing. While this argument fits well with the reduction in the negative components, the extent to which it can be used to explain enhancement of the positive components remains moot.

In conclusion, the components of the sleep recorded auditory evoked potential change with age. This is particularly apparent with the K-complex averaged N550 component, with the elderly producing significantly fewer K-complexes in response to auditory stimuli than young adults and of substantially smaller amplitude. It is also apparent with the presence of an enhanced P2 and long-lasting positive waveform in the older subjects that is not seen in the young subjects. We have speculated that the decreased negativity and increased positivity seen in the ERPs in the elderly represent structural brain changes that accompany the normal ageing process. This seems the more reasonable explanation of our findings, but will require structural magnetic resonance imaging data to be collected before this can be directly assessed.