Event-related potentials provide a non-behavioural technique for study of sleep-related changes in information processing. ERPs collected from alert subjects include components such as N100 and P300 known to be related to fairly specific cognitive processes (Donchin et al. 1986; Näätänen and Picton 1987; Picton and Hillyard 1988). Both of these components have been found to be diminished or absent in sleep (Picton et al. 1974; Näätänen and Picton 1987; Bell and Campbell 1988). The most obvious and most reliable indicator of information processing in the sleep ERP is the N350 component, which has been found to vary with the psychological properties of stimuli (Nielsen-Bohlmann et al. 1991; Harsh et al. 1994; Hull and Harsh 1998). Hull and Harsh (1998), for example, found N350 amplitude following repetitively presented tone stimuli to be inversely related to both the probability and the task relevance of the stimuli. More needs to be known, however, about whether N350 is a direct or indirect reflection of cognitive processes and about the level of information processing reflected. One of the objectives of the present study was to examine N350 changes in relation to the processing of semantically meaningful stimuli, i.e. subject's own name vs. other name. The meaning of a word or name has been found to be related to K-complexes and arousals (Oswald et al. 1960; McDonald et al. 1975). However, whether and how a semantically meaningful stimuli will affect the cognitively relevant N350 component of the sleep ERP has not been determined.
A second objective of this study was to determine whether a subject's personality characteristics are related to how that subject responds to potentially arousing stimuli presented during sleepiness and sleep. There is considerable variability in subjects’ responsiveness to stimuli presented in sleep (Zung and Wilson 1961; Evans et al. 1970; Bonnet and Moore 1982; Harsh and Badia 1989) with little investigation of the determinants of these differences. Miller has identified important differences in the way individuals cope with threatening stimuli in uncontrollable situations (1987, 1990). The Miller Behavioral Style Scale (MBSS) identifies subject differences in information processing styles referred to as Monitoring and Blunting. As the words imply, Monitoring is an information-seeking tendency while Blunting is an information-avoidance tendency. Monitoring has been associated with a heightened level of arousal (Miller 1987), a predisposition to engange in worrying (Davey 1990, as cited in Miller 1992), and proneness to anxiety disorders (MacLeod et al. 1986). The Monitoring–Blunting dimension may be importantly related to how subjects respond to stimuli presented as they are going to sleep and while they are sleeping. It is hypothesized that Monitoring as measured by the MBSS will be associated with heightened responsiveness as reflected by behavioural and ERP (N350) measures.
Two hundred and twenty-seven undergraduate students of the University of Southern Mississippi between the ages of 18 and 36 years completed a questionnaire regarding their preferred style of processing information (MBSS). Twenty-six subjects (16 Monitors and 10 Blunters) were tested in the sleep laboratory of the University of Southern Mississippi. Of these, eight Monitors (6 females, 2 males, median age 21, range 18–25) and eight Blunters (6 females, 2 males, median age 21, range 18–36) were able to go to sleep in the laboratory. Of the eight monitors, two were not able to go to sleep twice in the same night and had to be re-tested on a second night. Subjects were screened for health problems, hearing, medication use and abnormal sleep/wake schedules. Informed consent was obtained and subjects received extra credit for their participation.
Subjects were tested in a 9 ft × 13 ft (2.74 m × 3.96 m) room, which included a bed and a reclining chair. Ag-AgCl electrodes were used for recording of all electroencephalographic (EEG) activity. The electrodes were referred to linked mastoids with a forehead ground and impedances were kept below 5 kOhm. Electrodes placed at C3 and O1 (International 10–20 Electrode Placement System; Jasper 1958) were used to score stages of sleep and wakefulness. Activity at these sites was amplified and filtered (using settings standard for sleep recordings) with Grass Model 7P511 amplifiers. Midline recordings of ERPs were obtained at frontal (Fz), vertex (Cz) and parietal (Pz) scalp sites and amplified using Colbourn High-gain Bioamplifiers. Eye-movement artefact was monitored using electrodes attached at the outer canthus and supraorbitally to the left eye. For the ERP data, output was taken from the back of the Coulbourn amplifiers to obtain an effective time constant of 1.1 s. High cut-off filtering (10.33 Hz) was accomplished using a zero-phase digital filter (Ruchkin and Glaser 1978). The ERP recordings were digitized (100 samples/s) and stored using an A/D board (Data Translation DT 2821) housed in a Compaq 386/25 computer. Off-line averaging was controlled by a laboratory software program that included a routine for the correction of ocular artefact (Gratton et al. 1983). Tone and name stimuli were recorded and presented using a SUN IPC workstation and audiointerface program (SunOS 4.1.2). Names were uttered by an adult female speaker (standard English), tones were recorded as 1500 Hz pulse tones of 100 ms duration. Sweep onset was controlled by 1000 Hz pulse tones of 120 ms duration at 150 ms preceding the stimulus. The pulse was fed separately to a Lafayette Model 1604 A voice-activated relay to trigger the Compaq 386/25 for data collection. The sequence and interval of stimulus presentation was controlled by a UNIX script. A Bernoulli series determined stimulus sequence. Stimuli were presented 5 s apart at ≈60 dB (SPL) through miniature headphones. Each stimulus had a probability of 33.33%. Reaction time was measured by a response board (the same as in previous research: Harsh et al. 1994) strapped to the subject's preferred hand and connected with solid-state equipment. Reaction time was digitized and recorded on an AT & T 6300 microcomputer.
The MBSS consists of two independent subscales (Miller et al. 1988), the Monitoring and the Blunting subscales. The questionnaire describes four hypothetical stress-evoking scenes of an uncontrollable nature such as experiencing turbulences in an airplane or being held as a hostage. Each scene is followed by eight statements, half of which pertain to an information-seeking (monitoring) coping strategy, and half of which refer to an information-avoiding (blunting) coping strategy. The subject marks all statements that apply to him/her. A median split of the monitoring and blunting subscale is used to divide subjects into high or low Monitors and into high or low Blunters.
Individuals scoring as high Monitors and low Blunters typically seek out information relevant to physical and psychological threats. In contrast, individuals identified as low Monitors and high Blunters typically prefer to avoid threat-related information (Miller 1987, 1992). Data on the reliability and validity of the MBSS can be found in several sources (Miller and Mangan 1983; Miller 1987, 1992; Schumacher 1990).
Subjects reported to the laboratory on the test night at their usual bedtime hour (mean: 22.11 hours, range: 21.00–24.00 hours). Subjects were instructed to sleep-deprive themselves for 1 h the night before testing, and not to nap on the previous day or the day of the test. They were asked to abstain from drugs and alcoholic beverages and to drink no more than the usual amount of any caffeinated beverage on the day before testing.
For each subject, the general procedures were explained, informed consent was obtained, electrodes and earphones were attached, and a hearing test was performed. Subjects were then asked to lay down on the laboratory bed and listen to instructions. Before actual testing, subjects were asked if they had any association with the selected ‘Other Name’. If so, an alternative name was selected. In bed, subjects were asked to go to sleep while performing an oddball task. They were told not to fight off sleep in order to continue responding. They were asked to keep a mental count of and to make a fingerlift response to the designated target stimulus. All subjects were tested under two conditions. In the Tone condition, they were instructed to pay attention to 1000 Hz tones and ignore all names. In the Own Name condition, they were asked to attend to their own name and to ignore all other stimuli presented. The conditions were not counter-balanced; the Tone condition was presented first in all cases, followed by the Own Name condition. This procedure was chosen to prevent carry-over effects that would have been expected if the more salient stimulus (own name) would have been the first target stimulus presented.
For both conditions, subjects stayed in bed until they had reached 10 min of uninterrupted Stage 2 sleep. Between conditions, subjects were awakened and walked around the laboratory for 20 min. Testing was terminated at 05.00 hours for subjects who failed to go to sleep.
Sleep records were scored in 30-s epochs based on a modification of standard criteria (Rechtschaffen and Kales 1968). Stage 1A was added to identify epochs separately just preceding Stage 1 sleep and was characterized by a breaking-up of the alpha rhythm, but with alpha present during 50–80% of the epoch. Stage 1B was assigned to traditional Stage 1 sleep. The first 5 min of Stage 2 sleep was referred to as Stage 2A and the first 5 min of Stage 2 preceded by 5 continuous min of Stage 2 sleep was referred to as Stage 2B. Sleep stages were scored by two independent scorers. Interscorer agreement was greater than 90% for all stages. Disagreements were resolved by discussion.
K-complexes were scored as large (≥ 200 μV) negative– positive waveforms of at least 0.5 s duration. A K-complex was considered to be evoked by a stimulus if it occurred within 1 s following stimulus onset. Arousals were scored as bursts of alpha activity of at least 2 s duration beginning up to 2 s following stimulus onset. K-complexes were evaluated during Stage 2 sleep and arousals during Stages 1B and 2A, only.
For each subject, averages of event-related potentials (each average was based on at least 10 stimulus presentations) were obtained for stimulus type, condition and sleep stage. The recording epoch of the ERPs was 1500 ms, starting 150 ms prior to the onset of the stimulus and ending 1350 ms following stimulus onset. Waveforms for each subject and for each sleep stage were first filtered and examined on a single trial basis. Trials with movement artefact and with K-complexes were removed. To ensure that the ERP waveform was not confounded by K-complex-like deflections, all ERPs with a large negative–positive deflection exceeding 100 μV were excluded from analysis. A total of 10.08% of all trials had to be removed. The remaining trials were computer-scored, resulting in averages of event-related potentials (each average was based on at least 10 stimulus presentations) for stimulus type, condition and sleep stage. Since name stimuli differed in duration (610–1100 ms), latency was not a reliable criterion and morphology and lead became the determinants of the ERP. Waveforms following the same stimulus but different conditions were measured for amplitude at corresponding latencies and lead. Since latencies varied markedly across subjects and across conditions, the following scoring criteria were employed: for the awake data, the scoring window for P300 was set at 300–600 ms. P300 was evaluated only during wakefulness. The N350 was scored as the largest negative-going waveform that was maximal fronto-centrally within a time window of 200–600 ms. For all waveforms, peak amplitude was measured relative to the prestimulus baseline (the average of the values recorded during the 150 ms prior to stimulus onset), and peak latency was based on the time point at which the amplitude of the waveform was maximal.
Subjects differed in sleep pattern, i.e. not all subjects passed from Stage Wake to 1A, 1B, 2A and 2B. Other subjects passed so quickly from one stage to another that only a very small number of sweeps were collected during that stage. Averaged ERPs with less than 10 sweeps per average were not considered in the analysis and were treated as missing values (46 of 480 values=9.58%). Since most values were missing from Stage Wake (25 of 46), total means (across all subjects) were substituted for the missing P3 data. All other missing values were replaced by group means.
A group (Monitors, Blunters) × condition (attend Tones, attend Own Name) × stage (Wake, 1A, 1B, 2A, 2B) repeated-measures mixed design analysis of variance yielded a significant Stage effect, F (4, 56)=64.77, P≤0.01, ??=0.429, reflecting the steady decrease in responsivity that occurred as sleep progressed. The stage × group interaction approached significance, F (1.72, 24.01)=2.50, P=0.11. Since the purpose of the experiment was to explore differences between Monitors and Blunters, group comparisons of mean percentage responding were made at each sleep stage. In both conditions, Monitors responded more often than Blunters in stages Wake, 1A and 1B (see Fig. 1). This effect only reached significance in stage 1B of the Tone condition.
Of the 16 Monitors tested, eight subjects did not reach sleep Stage 2. Of the 10 Blunters tested, only two subjects were unable to fall asleep. Group comparison using binomial expansion coefficients yielded a significant difference between Monitors and Blunters (P≤0.05). Of the eight Monitors evaluated, two subjects were not able to go to sleep twice in the same night and one subject did not reach sleep Stage 2B.
A group (2) × condition (2) × stimulus (3) repeated-measures mixed design analysis of variance yielded a significant condition × stimulus interaction, F (2, 28)=8.80, P≤0.01, ??=0.813. Table 1 shows the mean arousal frequencies for condition × stimulus collapsed across groups. Newman–Keuls post-hoc procedures revealed that Own Name triggered the most arousals in both conditions. The difference in means reached significance for Own Name vs. Other Name in the Tone condition and for Own Name vs. Other Name as well as Tones in the Own Name condition.
Further, the target stimulus evoked more arousals than the same stimulus when it was not a target, i.e. Own Name in the Own Name condition produced more arousals than Own Name in the Tone condition (P≤0.02) and Tones in the Tone condition produced more arousals than Tones in the Own Name condition (P≤0.05).
A group (2) × condition (2) × stimulus (3) repeated-measures mixed design analysis of variance yielded a significant effect for stimulus (Own Name, Other Name, Tone), F (2, 28)=52.39, P≤0.01, ??=0.770. This effect is shown in Table 1. Newman–Keuls post-hoc procedures revealed that, as expected, presentation of Own Name resulted in a significantly higher number of evoked K-complexes than Other Name and Tones.
P3 in wakefulness
ERPs recorded in wakefulness are shown in Fig. 2 for all 16 subjects during both conditions and for all three stimuli. Analysis of P3 was based on recordings at the Pz lead because P3 has been shown to be dominant parietally (Donchin et al. 1978; Hansen and Hillyard 1980). Table 2 shows mean P3 amplitudes for Monitors and Blunters to the three stimuli in each condition.
A group (2) × condition (2) × stimulus (3) repeated-measures mixed design analysis of variance resulted in a main effect for condition, F (1, 14)=16.12, P≤0.01 and for group, F (1, 14)=5.81, P=0.03. The stimulus effect approached significance (F (2, 28)=2.79 (Pillais), P=0.1. The overall amplitude of P3 was greater in the Tone condition (M=8.36 μV, SEM=1.36 μV) compared to the Own Name condition (M=5.00 μV, SEM=1.07 μV). Monitors had larger P3 amplitudes than Blunters (Mmon=9.09 SEM=1.72, Mblun=4.26, SEM=1.03). Stimulus comparisons were made because of their relevance to the study. Overall, Tones as well as Own Name evoked larger P3 s than Other Name (P< 0.05).
Analysis was restricted to the central lead because N350 is maximal centrally (Weitzman and Kremen 1965; Campbell et al. 1992; Harsh et al. 1994). A condition (2) × stimulus (3) × stage (5) × group (2) repeated-measures mixed design analysis of variance showed a significant condition × stimulus × group interaction, F (2, 28)=3.60, P=0.05, ??=0.787, as well as a condition × stage effect, F (4, 56)=6.06, P≤0.01, ??=0.748. Table 3 lists means showing the condition × stage interaction collapsed across groups and stimuli. Compared to the Tone condition, the amplitude of N350 was larger in the Own Name condition during stages Wake and 1A and smaller in the Own Name condition in Stage 2B.
Table 4 lists mean N350 amplitudes for Monitors and Blunters following the three stimuli under both conditions collapsed across stage.
Monitors produced a significantly smaller N350 to Tones compared to Own Name in the Tone condition (P≤0.05, Newman–Keuls post-hoc procedures), i.e. the target stimulus in the Tone condition evoked the smallest N350. No such effect was found for the Own Name condition (P> 0.2). Blunters showed the largest N350 to the stimulus attended to in both conditions, i.e. Tones in the Tone condition and Own Name in the Own Name condition. However, between-stimulus comparisons using Newman–Keuls post-hoc procedures provided statistical evidence of an effect in the Own Name condition only. For Blunters in the Own Name condition, N350 to Own Name was significantly larger than Other Name and Tone (P≤0.05). Overall, Blunters displayed larger N350 s to target stimuli than Monitors in both conditions (P≤0.01 in the Tone condition, P≤0.05 in the Own Name condition). Figures 3 and 4 show single subject N350 data for a Blunter and a Monitor, respectively.
Figure 5 shows target-evoked N350 amplitudes for Monitors and Blunters in each sleep stage. During Stage 1 A, Blunters produced a significantly larger N350 to the target stimulus in both conditions (P< 0.01). Further, Blunters had greater N350 amplitudes in the Tone condition during Stages 1B and 2A (P< 0.05).
Information processing during sleep onset
The data collected during wakefulness show that ERPs to semantic stimuli are comparable to ERPs to simple tone stimuli. The failure to find a difference between Own Name and Tone in the Tone condition is attributed to the innate relevance of the ‘Own Name’ stimulus. Berlad and Pratt (1995) found P3 amplitude to a subject's own first name to be significantly larger than to either an irrelevant rare stimulus or an irrelevant frequent stimulus. Berlad and Pratt (1995) used a passive oddball paradigm, which may have led to more distinct differences in P3 amplitude since it allowed subjects to internally switch their attention to the stimulus carrying inherent salience. This interpretation is supported by the finding that both stimuli carrying salience, Own Name and Tone, resulted in significantly larger P3 amplitudes than Other Name.
The overall amplitude of P3 in the Own Name condition was much smaller compared to the Tone condition, which is interpreted as an order effect, since Own Name was the second condition tested for all subjects. Subjects may have been less attentive at the time of the second testing. The diminished P3 amplitude may have resulted in reduced sensitivity and consequent absence of significant differences between Own Name and Tone stimuli.
Evaluation of arousal and K-complex data suggest that stimuli were differentiated according to their semantic properties while the subjects were asleep. Own Name produced the greatest number of K-complexes and arousals relative to Other Name and Tones. This effect is attributed to the inherent salience of one's own name. This finding replicates and extends the results reported in the Oswald et al. study (1960), in which subjects were instructed to awaken to their own name. The results of the present study are opposite those reported in the McDonald et al. study (1975), in which a higher frequency of K-complexes was found in response to Other Name compared to Own Name. The discrepant findings of the McDonald et al. study (1975) are likely to be due to as yet unidentified methodological factors.
Further evidence of active processing of stimuli during sleep is provided by analysis of arousal frequency to Own Name and Tone stimuli. Own Name in the Own Name condition triggered more arousals than Own Name in the Tone condition. Similarly, Tones in the Tone condition were followed by more arousals than Tones in the Own Name condition. K-complexes were also more frequent, although not significantly, following Own Name in the Own Name condition relative to Own Name in the Tone condition. The absence of any indication of a comparable effect for tones may be due to tones having less innate relevance.
Individual differences in responsiveness and information processing during sleep
Monitors showed a larger overall P3 amplitude than Blunters suggesting that Monitors were more attentive to the experimental stimuli. Interpretation of the mean percentage of stimuli responded to provides some support for this conclusion. During the transition phase from wake to sleep, Monitors were more likely to respond to targets than Blunters. These differences were maximal in Stage 1B. The heightened attention level found in Monitors can be interpreted within the theoretical framework of the Monitoring/Blunting hypothesis. According to this hypothesis (Miller 1990), if placed in an uncontrollable situation, Monitors will remain alert and seek information, even though this behaviour is counterproductive in terms of successful coping. In the present study, subjects were not able to control stimulus presentation and, considering that subjects were supposed to go to sleep, successful coping implied directing attention away rather than toward the experimental stimuli. By being highly attentive to the experimental stimuli, Monitors engaged in a behaviour that was counterproductive to sleep. This interpretation is supported by the fact that, of the 16 Monitors who participated in the experiment, only eight were able to go to sleep in the laboratory. Of those eight subjects who did fall asleep, two could not go to sleep twice in the same night and had to come back a second night. Further, one Monitor never reached sleep Stage 2B, although he spent the entire night in bed. In contrast, only two Blunters were unable to go to sleep in the laboratory.
Differences in information processing style were not reflected in the arousal or K-complex data. Considering the difficulties that Monitors had in falling asleep compared to Blunters, this finding is particularly interesting. Assuming that arousals and K-complexes are both related to the maintenance of sleep rather the induction of sleep, the K-complex and arousal data suggest that, once asleep, the preservation of sleep is no longer influenced by information processing style but primarily by stimulus salience.
Across stimuli, the overall amplitude of N350 during Stage 1A was smaller in the Tone condition relative to the Own Name condition. Considering that P3 amplitude during wakefulness showed the reverse tendency, this observation supports the idea of a relatedness of N350 and P3, as expressed by Harsh et al. (1994), who reported that N350 emerged as the amplitude of P3 decreased. One possible interpretation of these results is that P3 and N350 reflect opposing processes such as ‘attention toward’ (P3) vs. ‘attention away from’ (N350) the stimulus environment. Support for this interpretation includes the observation that Blunters tended to have a larger N350 to target stimuli than Monitors (Fig. 5). If going to sleep requires processes similar to blunting and if blunting is reflected in the N350 waveform, then group differences in N350 amplitude to targets may reflect successful blocking of sensory processing or perhaps attenuation of the CNS activation that accompanies sensory processing. During wakefulness, blunting is associated with decreased levels of arousal (Miller 1987; Efran et al. 1989). Considering that a large number of Monitors were not able to go to sleep in the laboratory, and considering the greater responsivity for Monitors in Stage 1B as well as the heightened attentiveness for Monitors during Wakefulness as reflected in P3 amplitude, the interpretation that N350 reflects stimulus ‘blunting’ seems reasonable.