Statistical Features of Hypnagogic EEG Measured by a New Scoring System

Transcription

1 Sleep, 19(9): American Sleep Disorders Association and Sleep Research Society Statistical Features of Hypnagogic EEG Measured by a New Scoring System Hideki Tanaka, Mitsuo Hayashi and Tadao Hori Department of Behavioral Sciences, Faculty of Integrated Arts and Sciences, Hiroshima University, Higashi-Hiroshima, Japan Summary: The purpose of this study was to examine the durations of individual occurrences of each of nine hypnagogic electroencephalographic (EEG) stages and the interchange relationship among these stages. Most of the alpha patterns (stages 1, 2, and 3), ripples (stage 5), and spindles (stage 9) tended to last >2 minutes. On the other hand, histograms of the durations of time in EEG flattening (stage 4) and vertex sharp wave (stages 6, 7, and 8) patterns had peaks that lasted <30 seconds. Analysis of the sequences of EEG stage changes demonstrated that shifts to adjacent stages were most common for all stages. A smooth change in EEG stage occurred in the downward or upward direction in the hypnagogic state. This was especially true for the first five stages. EEG stages with vertex sharp waves (stages 6, 7, and 8), however, showed less-smooth changes, with approximately 20% of all changes involving a jump of more than one stage. These results show that the basic EEG activities in the sleep onset period are the alpha, theta, and sleep spindles activities, whereas the activities of vertex sharp waves seem to have a secondary or enhancing role, instead of independent characteristics. Key Words: EEG stages-hypnagogic EEG-Sleep onset period-sleep scoring. The hypnagogic period is one of the most interesting states for psychologists to study because of the occurrence of dream-like mentation (1,2). Because the variations in hypnagogic electroencephalogram (EEG) measurements were remarkable, early investigators (3-5) noted the characteristics of the waveforms and frequencies of hypnagogic EEG and classified them accordingly. With standard sleep criteria (6), however, most of the hypnagogic EEG patterns have been combined into standard sleep stage 1. Foulkes and Vogel (1) observed that reports of dream-like events occurred even during the pre-sleep waking period with an EEG alpha rhythm. Their observations (1) suggested that the onset of the hypnagogic period might precede the onset of sleep stage 1 [as identified by a modification of the criteria of Dement and Kleitman (7)]. On the other hand, longer response times and intermittent response failure occurred at sleep stage 1 (8,9). These observations suggest that the behavioral sleep process may begin in sleep stage 1. Sleep spindles or standard stage 2 sleep have long been used as the objective maker of 'true' sleep (10). Even in this stage, however, discrepancies appear between subjective reports and Accepted for publication July Address correspondence and reprint requests to Tadao Hori, Ph.D., Department of Behavioral Sciences, Faculty of Integrated Arts and Sciences, Hiroshima University, Kagamiyama 1-7-1, Higashi-Hiroshima 739, Japan. 731 polygraphic recordings (1,11,12), indicating that the proportion of true sleep judgments in early stage 2 sleep (first sleep spindle) is poor. These studies suggest that the hypnagogic effects on the subjective process probably continue after the onset of sleep stage 2. Quantitative EEG studies also reveal many contradictions associated with treating the hypnagogic state synonymously with standard sleep stage 1. Hori et al. (13) examined the spatio-temporal behavior of EEG activity in the hypnagogic state using the topographic map of 12-channel EEGs. They reported that the EEG structures of the hypnagogic state were not uniform across scalp areas (13). They also found that the hypnagogic state probably started before the onset of sleep stage 1 and continued for a few minutes after the onset of sleep stage 2 (13). Furthermore, seven characteristic patterns were distinguished in the topograms from the sampled period (13). Hasan et al. (14) also indicated that it is necessary to divide wakefulness and drowsiness into substages for the clinical study of conditions such as narcolepsy. Moreover, the studies in which event-related potential (ERP) was employed indicated the presence of components that increased in amplitude during sleep, although the subject became less attentive toward the evoking stimulus (15-22). These findings suggest that the functions of the attention mechanism during wak-

2 732 H. TANAKA ET AL. EEG Stages 1: Alpha wave train 2: Alpha wave intermittent (>50%) ~ 3: Alpha wave intermittent «50%) ~ 4: EEG flattening 5: Ripples 6: Vertex sharp wave solitary 7: Vertex sharp wave bursts 8: Vertex sharp wave and incomplete spindles 9: spindles ~ ~ ~ ~~50J1.V 1 sec FIG. 1. Typical EEG patterns during hypnagogic state [following Hori et al. (24)]. EEG stage I, alpha wave train; EEG stage 2, alpha wave intermittent A (0'. ~ 50%); EEG stage 3, alpha wave intennittent B (0'. < 50%); EEG stage 4, EEG flattening; EEG stage 5, ripples; EEG stage 6, vertex sharp wave, solitary; EEG stage 7, vertex sharp wave train or burst; EEG stage 8, vertex sharp wave and incomplete spindle; EEG stage 9, spindles. ing differ from those during sleep. Decreases in the Nl and P300 components, which reflect reduction of attention functions of the waking system, continue into sleep stage l. Increases in the PI, N2, N3, and P3 components, which reflect development of attention functions of the sleep system, start at sleep stage 1. The above observations suggest that the hypnagogic state is neurophysiologic ally and psychophysiologically complex, and even paradoxical, and that the standard sleep stage criteria, especially for sleep stage 1, are too vague to define when the convergence of behavioral, subjective, and polygraphical measures in the hypnagogic state is being studied. Therefore, to examine the variations in hypnagogic EEG in detail, the typical EEG patterns during the waking-sleep transition period were classified into nine EEG stages, as shown in Fig. 1 (23,24). These patterns are similar to those described by Shiotsuki et al. (5), who had made them as a revision of the Gibbs and Gibbs Atlas (4). The rank order of these nine EEG patterns was determined by reaction times (RT) to pure tone (23), and were defined as EEG stages (24). We assessed the convergence of behavioral (RT) (23) and subjective (hypnagogic imagery; HI) (24) measures on the nine EEG stages in the hypnagogic period. The results of these studies suggest that EEG stages were somewhat comparable in relation to the binary information provided by the behavioral and subjective measures. Further topographical analysis of the changes in EEG stages might increase our understanding of the hypnagogic period. However, the durations of each EEG stage and their frequency of occurrence have not been examined under non-stimulus conditions. Examination of the statistical features of EEG stages under non-stimulus conditions will help to clarify the behavior of the hypnagogic EEG. The purpose of the present study was to examine the regulation of the order of appearance of the characteristic EEG patterns, the distribution of durations of each EEG stage, and the interchange relationship between EEG stages. Subjects METHODS Twenty-three healthy male subjects, who were unpaid volunteers, participated in the study. Their ages ranged from 20 to 27 years (mean 22.3 years). They were all undergraduate or graduate students taking a course in behavioral sciences at Hiroshima University. The waking EEG record for each subject was visually examined for alpha activity (S-13 Hz EEG activity, with an amplitude of 2:20 j..l V). The subjects with <50% of their waking record occupied by alpha waves were classified as the low-alpha subjects. The subjects with >SO% of their waking record occupied by alpha waves were classified as the high-alpha subjects. The remaining subjects, whose waking time occupied by alpha waves varied from 50% to SO%, were classified as the intermediate-alpha subjects. All of the subjects in this study were the high-alpha subjects with alpha times of 2:90%. These subjects had previously been adapted to the recording chamber. Apparatus All electrophysiological parameters were recorded simultaneously on an IS-channel electroencephalograph (model 1A97, NEC San-Ei) and a 14-channel FM tape recorder (model SR-50, TEAC). The international 10/20 system was used for electrode placement on the 12 scalp sites: Fpl, Fp2, F7, FS, Fz, C3, C4, Pz, T5, T6, 01, and 02. The electroencephalograph ran at a paper speed of 10 mm/second. The EEG data were recorded by FM tape recorder at a tape speed of 12 mm/second. The sleeper was studied in a 3 X 3 m electrically shielded, sound-attenuated, and air-conditioned bedroom. Temperature readings taken throughout the study showed a mean of 22.5 ± 1.0 C. Procedure Subjects were asked to arrive at the sleep laboratory 2 hours before their usual bedtime. After the subjects changed into their night clothes, electrodes were applied. Sleep monitoring varied slightly from that de- Sleep. Vol. 19. No

3 EEG SCORING SYSTEM FOR THE HYPNAGOGIC STATE 733 ~) scribed by Rechtschaffen and Kales (6). Surface electrodes were placed on 12 scalp areas in the international 10/20 system, referenced to ipsilateral ear lobes (EEG). Two horizontal electrooculograms (EOG) were each referenced to ipsilateral ear lobes for slow eye movement (T = 3.2 seconds). Chin electrodes (mentalis: electromyograms; EMG) were referenced to each other. Electrodes were placed on the forehead for the body ground. All electrophysiological parameters were recorded using silver-silver chloride disk electrodes filled with electrode cream; electrodes were attached with surgical tape or with collodion for scalp placements. Interelectrode impedance was <5,000 ohms. The EEGs were amplified using a high-cut filter setting of 120 Hz and a time constant of 0.3 seconds. The subjects were instructed to close their eyes after the light was turned off. An experimenter awakened the subject 5 minutes after the appearance of the first sleep spindle by calling his name and entering the bedroom. The subject was then requested to report his subjective experiences (i.e. dream, hypnagogic imagery) and mood. Presleep and postsleep questionnaires were used to record unusual events that might influence sleep parameters. Criteria for hypnagogic EEG stages The nine hypnagogic EEG stages according to the criteria of Hori et al. (24) are listed below (see Fig. 1). Each recording, from lights off to 5 minutes after the onset of sleep stage 2, was scored by standard sleep stage criteria (6) and analyzed. EEG stages were scored manually for each 30-second period of the C3 EEG record. EEG stage I.-Alpha wave train: epoch composed of a train of alpha activity with a minimum amplitude of 20 flv. EEG stage 2.-Alpha wave intermittent (A): epoch composed of a train of 2:50% alpha activity with a minimum amplitude of 20 fl V. EEG stage 3.-Alpha wave intermittent (B): epoch composed of a train of <50% alpha activity with a minimum amplitude of 20 fl V. EEG stage 4.-EEG flattening: epoch composed of suppressed waves of <20 fl V. EEG stage 5.-Ripples: epoch composed of lowvoltage theta waves (20 fl V < e < 50 fl V) with burst suppression. EEG stage 6.-Vertex sharp wave solitary: epoch contained one well-defined vertex sharp wave. EEG stage 7.-Vertex sharp wave train or burst: epoch contained at least two well-defined vertex sharp waves. EEG stage 8.-Vertex sharp wave and incomplete spindle: epoch contained at least one well-defined ver- TABLE 1. Average times (AT) in minutes. standard errors (SE). and percentage (%) for nine EEG stages during the course from lights off to 5 minutes after EEG stage 9 (n = 20) EEG stages AT SE % Total tex sharp wave and one incomplete spindle (duration <0.5 seconds, amplitude <20 fl V and> 10 fl V). EEG stage 9.-Spindles: epoch contained at least one well-defined spindle of at least 0.5 seconds duration and 20 fl V in amplitude. EEG stages 1 and 2 correspond to sleep stage W (wake) by standard criteria (6); EEG stages 3-8 correspond to sleep stage 1, and EEG stage 9 corresponds to sleep stage 2. Each record was scored entirely by one scorer and was rescored by an independent scorer. A comparison was then made between the two recording analyses on an epoch-by-epoch basis to determine between-scorer reliability. The reliability score was >90% for each subject's record. Statistical analysis For one-way repeated ANOV A, significance levels were determined following the adjusted Greenhouse Geisser (25) approach for repeated observations. As a post hoc test, the Newman-Keuls test (25) was carried out. The frequencies of stage changes were summed up for each direction (upward or downward). The dominant direction of EEG stage changes was identified by using a chi-square test. RESULTS Table 1 displays the average time of each EEG stage in minutes, standard errors (SE), and the percentages of total sleep time for the nine EEG stages during the waking-sleep transition period. The data from three male subjects (3/23, 13.0%) were excluded from the analyses because their records lacked EEG stage 4. The average time of EEG stage 2 was 3.4 minutes; that of EEG stage 9 was 3.9 minutes. The sum of these two stages occupied about half (41.5%) of the total sample time (17.6 minutes). The mean appearance time of EEG stage 4 was 0.9 minutes; that of EEG stage 6 was 0.8 minutes. The ANOV A results indicated a significant effect of the EEG stage [F(3,58) = Sleep. Vol. 19. No

4 734 H. TANAKA ET AL. TABLE 2. Frequency distributions of EEG stage durations (n = 20) EEG stage Duration in seconds Total 121 or more (10.8%) (5.1%) (2.3%) (5.7%) (27.5%) (5.9%) (5.4%) (16.9%) (9.3%) (5.7%) (12.5%) (6.4%) (16.2%) (13.6%) (7.0%) (20.0%) (4.9%) (4.8%) (12.5%) (9.2%) (27.0%) (18.6%) (25.6%) (16.7%) (28.6%) (3.2%) (19.5%) (19.0%) (17.5%) (19.8%) (40.6%) (45.8%) (55.8%) (83.3%) (40.0%) (96.8%) (75.6%) (76.2%) (30.0%) (58.7%) Total () () () () () () () () () () 6.65, E = 0.38, P < 0.01]. Post hoc testing revealed two subgroups [(EEG stages 2 and 9) > (EEG stages 1, 3, 4, 5, 6, 7, and 8)]. Table 2 presents a frequency distribution of the durations of each EEG stage. The total number of EEG stage changes for 20 subjects was 358 (mean 17.9 changes per subject). Because the mean total sample time was 17.6 minutes (Table 1), an EEG stage shift occurred every 0.98 minutes, on average. However, there were only 71 (19.8%) samples with durations of 31 to 60 seconds that corresponded to the mean values (0.98 minutes), whereas there were 210 (58.7%) episodic appearances «30 seconds). Short EEG stage durations «30 seconds) were predominant for each EEG stage, especially EEG stage 6. For EEG stages 1, 2, 3, 5, and 9, there were cases of stage durations of >2 minutes, but the durations of most of the samples in EEG stages 4, 6, 7, and 8 were <30 seconds (83.3%, 96.8%, 75.6%, and 76.2%, respectively). This latter group of EEG stages (EEG stages 4, 6, 7, and 8) was unstable; they changed from one stage to another rapidly. These data suggest that the basic EEG stages in the hypnagogic state are stages 1, 2, 3, 5, and 9 and that the others (stages 4, 6, 7, and 8) are phasic stages. The sequence in which the EEG stages appeared is shown in Table 3. To construct this table, all sampled records were scored by tabulating the frequency with which the 339 changes that were observed followed each other in a 9 X 9 contingency table. This type of analysis shows that the major interchanges occurred among EEG stages 1, 2, and 3. For instance, 41.9% of all examples of EEG stage 3 were followed by EEG stage 2, and 28.8% of all examples of EEG stage 2 were followed by EEG stage 1. As Table 3 shows, there was an almost universal tendency for the stageby-stage change in the downward or upward direction to be smooth. The smooth nature of the transitions during the hypnagogic state is illustrated by the fact that the per- centages of downward shifts show higher values, ranging from 97.3% of EEG stage 1 (maximum) to 46.4% of EEG stage 7 (minimum). The range of percentages of upward shifts was lower, from 41.9% of EEG stage 3 (maximum) to 9.7% of EEG stage 6, with the exception of 61.9% of EEG stage 9. To examine this tendency statistically, the total numbers of both upward changes and downward changes were computed for each stage. For example, the total numbers of upward changes from EEG stage 5 were 9 (2 to stage 2, 3 to stage 3, and 4 to stage 4) and those of downward changes were 26 (17 to stage 6, 6 to stage 7, 2 to stage 8, and 1 to stage 9). Chi-square analyses were applied to these data sets. The chi-square analyses indicated that the percentage of downward changes was significantly higher than that of upward changes in all EEG stages, with the exception of EEG stage 3 {EEG stage 1 [X 2 (1) = 37.00, P < 0.01], EEG stage 2 [X 2 (1) = 10.59, p < 0.01], EEG stage 4 [X 2 (1) = 13,33, P < 0,01], EEG stage 5 [X 2 (1) = 8.26, P < 0.01], EEG stage 6 [X 2 (1) = 17.06, P < 0.01], EEG stage 7 [X 2 (I) = 5.49, P < 0.05], EEG stage 8 [X 2 (1) = 6,10, P < 0.05], EEG stage 9 [X 2 (1) = 21.00, P < O.OI]}. These data suggest that hypnagogic states were generally descending states. Regarding both upward and downward stage changes, the progression of sleep onset from one stage to another typically occurred in a smooth fashion, moving from one stage to the next, when the changes occurred in EEG stages 1, 2, 3, 4, and 8. The smooth nature of these transitions is illustrated by the fact that EEG stage 1 was followed by EEG stage % of the time, EEG stage 2 was followed by EEG stages 1 and 388.1% (28.8% %) of the time, EEG stage 3 was followed by EEG stages 2 and % (4l.9% + 46,5%) of the time, EEG stage 4 was followed by EEG stages 3 and % (13.3% %) of the time, and EEG stage 8 was followed by EEG stages 7 and % (26.2% %) of the time. On the other hand, >20% of the changes from EEG stages 5, Sleep, Vol. 19, No.9, 1996

5 :,) EEG SCORING SYSTEM FOR THE HYPNAGOGIC STATE 735 l... is followed by this EEG stage TABLE 3. Sequence of EEG stage changes (n = 20) This EEG stage Total l % 61.0% 30.5% 1.7% % 41.9% 3.3% 2.3% 81.4% 9.3% % 59.3% 13.3% 16.7% 66.7% % 46.5% 5.7% 8.6% 62.8% % 7.0% 73.4% 3.2% 6.5% % 6.7% 2.5% % % % 10.9% l7.4% 12.7% 8.8% Total , and 7 involved a jump of more than one stage. The unstable nature of these stages is illustrated by the fact that EEG stage 5 was followed by EEG stages 2, 3, 7, 8, and % (5.7% + 8.6% + l7.1% + 5.7% + 2.9%) of the time; EEG stage 6 was followed by EEG stages 4, 8, and % (3.2% % + 3.2%) of the time; and EEG stage 7 was followed by EEG stages 2, 5, and % (2.4% + 2.4% %) of the time In 8, 6 as Ci5 5 Cl ~ Time (min) FIG. 2. Transition patterns of EEG stages for each subject. The x-axis shows the elapsed time from lights off. The y-axis shows EEG stages (n = 20) % % 7.0% % 13.3% 3.3% % 3.2% 8.6% 3 9.7% 54.8% % 14.6% 43.9% % 58.1% 4.8% 19.0% % 25.8% 2.4% 2.4% % 3.2% 10.3% 9.2% % % % 5.7% % 4.8% 35.5% % 26.8% % 45.2% % 22.0% 70.8% % 69.0% 12.1% 12.4% o % 1.7% % 17.4% o % o % 5.7% % 10.3% o % 12.2% % 31.0% % 6.2% % % % 339 Figure 2 graphically displays the transition patterns of EEG stages for each subject. Although there were individual differences in the amount of time each subject spent in each EEG stage, the progression of EEG stages tended to occur as a function of elapsed time. From these data, the progression of EEG stages in one group started from 4 minutes after lights off, and in another group of subjects the downward stage shifts started at 8 minutes after lights off. The gradients of these two groups were roughly parallel. Thus, the process of sleep onset basically occurred in a smooth fashion, moving from one EEG stage to the next, but the phasic EEG stages gave rise to the irregularity and complexity of the transition patterns of EEG stages. DISCUSSION The stability of the EEG stages Although the distributions of duration for each sleep stage, the number of stages, and stage shifts in allnight sleep are well documented (26,27), hypnagogic EEG patterns have not yet been reported. Recently, Hori et al. (23) using button press reaction time (RT) to pure tone, investigated the typical picture of the EEG during the hypnagogic state. They determined the Sleep. Vol. 19, No.9, 1996

6 736 H. TANAKA ET AL. order of appearance of the characteristic EEG patterns, their individual duration, and the number of occurrences, but they did not attempt to investigate the EEG patterns under non-stimulus conditions (23). In the present study, we investigated the distribution of durations of each EEG stage and the number of stages of shifts, and we statistically analyzed the characteristics of EEG stages under non-stimulus conditions. The percentages of shifts to the next stage occurred most frequently at each EEG stage (ca. 46.4%- 97.3%). Williams et al. (26) reported that sleep stage 2 [by the criteria of Dement and Kleitman (7)] occupied approximately 50% of sleep time and was more evenly distributed throughout the night. In the present study, the average number of times in which EEG stage 2 (alpha intermittent A: ex ;::: 50%) occurred was significantly higher than those of other EEG stages. These results suggest that EEG stage 2 is one of the key stages in the hypnagogic state. On the other hand, the percentages of jumping shifts increased after EEG stage 5, when theta waves started to appear. The temporal sequences of stage changes became irregular and complex after EEG stage 5. These results suggest that EEG stage 5 is an intermediary stage in the wakingsleeping transition period and that this stage is the principal stage of the hypnagogic state. These findings are consistent with reports (23) in which the stimulus condition was studied. These findings also suggest that EEG stages 2 and 5 are important for an understanding of the variable hypnagogic phenomena, especially for details of sleep gates. Ogilvie and Wilkinson (8,9) introduced the concept of a sleep onset period (SOP), defined as the transition between relaxed, drowsy wakefulness and unresponsive sleep. Although the SOP may have sleep stage 1 at its center, it clearly overlaps into sleep stage Wand sleep stage 2, scored by standard criteria (6). Ogilvie et al. (28) also argued that using RT to define an SOP increases the mean latency or decreases the consistency of the response rate or frequency located at the beginning of the SOp, whereas RT cessation denotes the end of the SOP and the beginning of sleep. Using their concept of an SOp, it could be presumed from the present data that EEG stage 2 acts as an intermediate between the waking period and SOp, and that EEG stage 5 (theta wave burst suppression) also acts as an intermediate between SOP and the sleep period. Regarding both upward and downward stage changes, EEG stage changes were usually smooth, moving from one stage to the next, when the changes occurred from EEG stages 1, 2, 3, 4, and 8, but they were less smooth from EEG stages 5, 6, and 7 (the presence of theta waves). More than 20% of these latter three stages involved jumps of more than one stage at a time. These data suggest that the main EEG ac- Sleep. Vol. 19. No tivities in the hypnagogic state are alpha and theta activity, whereas the activity of vertex sharp waves appears to modify the arousal level of the theta activity in the background EEG. In a study of the relationship between EEG stages and the hypnagogic imagery (HI) recall rate (24), the distribution of HI percentages showed an inverted V-shape, with a peak at stage 5 (theta wave burst suppression). This suggests that theta waves are significant for the occurrence of HI. In the present study, the length of EEG stages was short, usually <30 seconds. This tendency reflects the fact that the hypnagogic EEG is a rapidly changing phenomenon. However, in addition to the short durations, the alpha wave state (EEG stages 1, 2, and 3, especially EEG stage 2) and theta wave state (EEG stage 5) tended to be more stable and of longer duration, ranging up to 2 minutes. The duration of the sleep spindle state (EEG stage 9) was also longer; approximately 30% of all EEG stage 9 durations were >2 minutes. On the other hand, the durations of most (75.6%-96.8%) of EEG stages 4, 6, 7, and 8 were <30 seconds. These four EEG stages were unstable and fragmentary, changing rapidly from one stage to another. Therefore, there are two types of EEG stages in the hypnagogic state. One is the rather stable type (i.e. EEG stages 1, 2, 3, 5, and 9) and the other is the unstable fragmentary type (i.e. EEG stages 4, 6, 7, and 8). These results suggest that alpha waves, theta waves, and sleep spindles are EEG characteristics that define the hypnagogic process, and that vertex sharp waves seem to enhance the stability of the EEG pattern, especially the theta wave-dominant pattern. Application of the EEG stages The electrode montage used for the study of EEG in the SOP may be important. Since the early studies of sleep EEG, it has been reported that many EEG events have their largest amplitude at the vertex (4). In the present study, however, C3 was used for scoring instead of the Cz because the standard scoring system of Rechtschaffen and Kales (6) recommends the use of C3 or C4 EEG for staging. The C3 site was chosen in order to facilitate a comparison of the results of this new system and the standard system. In the present study a 30-second epoch was used because this epoch length is most commonly used for sleep stage scoring. However, it is generally known that the electrophysiological changes related to drowsiness or the SOP vary in a rapid manner. The description of drowsiness or the SOP may require higher temporal resolution than is provided by the standard sleep staging system. In recent studies, epochs of <30 seconds have been used to examine the relationships among the electrophysiological activity, reactivity, and

7 li,l EEG SCORING SYSTEM FOR,THE HYPNAGOGIC STATE 737 If)., ~J performance (14,20,28,29). Most of the latter studies used a 5-second short fixed epoch. Our previous studies also used a 5-second short epoch, which helped to assess the convergence of behavioral (RT) (23) and subjective (HI) (24) measures on the nine EEG stages. Hasan et al. (14), however, used adaptive segmentation, considering that as long as the electrophysiological activity remained constant the same epoch was continued (30). In any case, shorter epochs are expected to describe the EEG behavior of the SOP in greater detail than a 30-second fixed epoch. There were no subjects with extremely low amplitude alpha waves in this study. For low-alpha subjects, discriminations among EEG stages 1, 2, and 3 may be quite difficult using the present scoring criteria. For the records of extremely low amplitude alpha or alpha rare type subjects, quantified EEG data using fast Fourier transformation (FFT) may be useful for the resolution of this problem. The present study confirmed the validity of the rank order in EEG stages by sequence analysis. Recently we examined the topographical characteristics of nine EEG stages during the hypnagogic state. In these studies (31-33), topographic maps demonstrated that the dominant areas of alpha band activity moved from the posterior areas to the anterior areas along the midline of the scalp. The dominant areas of delta, theta, and sigma band activities did not move, however, but their activities developed in their own areas of the scalp during the hypnagogic period. In those three frequency bands, the differences in EEG activities between the focus area and the surrounding areas increased with the progress of EEG stages. It is expected hereafter that the nine EEG stages will be a good objective indicator of the variable phenomena of the hypnagogic state. Moreover, further topographical analysis relevant to changes in the EEG stage might help distinguish the SOP from both wake and sleep and help clarify the psychophysiological structure of the SOP and its function. REFERENCES l. Foulkes D, Vogel G. Mental activity at sleep onset. J Abnorm Psychol 1965;70: Schacter DL. The hypnagogic state: a critical review of the literature. Psychol Bull 1976;83: Davis H, Davis PH, Loomis AL, Harvey EN, Hobart G. Human brain potentials at the onset of sleep. J Neurophysiol 1938; I: Gibbs FA, Gibbs EL. Atlas of electroencephalography, vol. 1. Cambridge: Addison-Wesley, Shiotsuki M, Ichino Y, Shimizu K. Changes in the electroencephalogram during whole night natural sleep. Jpn J Surg Soc 1954;55: Rechtschaffen A, Kales A. A manual of standardized terminology, techniques and scoring system for sleep stages of human subjects. Washington, DC: Public Health Service, U.S. Government Printing Office, Dement WC, Kleitman N. Cyclic variations in EEG during sleep and their relation to eye movements, body motility and dreaming. Electroencephalogr Clin Neurophysiol 1957;9: Ogilvie RD, Wilkinson RT. The detection of sleep onset: behavioral and physiological convergence. Psychophysiology 1984;21 : Ogilvie RD, Wilkinson RT. Behavioral versus EEG-based monitoring of all-night sleep/wake patterns, Sleep 1988;11: , 10, Webb WB. The natural onset of sleep. In: Popoviciu L, Asgian B, Badia G, eds. Sleep 1978: fourth European congress on sleep research. Basel: Karger, 1980: II. Bonnet MH, Moore SE, The threshold of sleep: perception of sleep as a function of time asleep and auditory threshold. Sleep 1982;5: Kamiya 1. Behavioral, subjective and physiological aspects of drowsiness and sleep. In: Fiske OW, Maddi S, eds. Functions of varied experience. Homewood, IL: Dorsey Press, 1961: Hori T, Hayashi M, Morikawa T. Topography and coherence analysis of the hypnagogic EEG. In: Horne 1, ed. Sleep '90. Bochum: Pontenagel Press, 1990: Hasan 1, Hirvonen K, Varri A, Hakkinen V, Loula P. Validation of computer analyzed polygraphic patterns during drowsiness and sleep onset. Electroencephalogr Clin Neurophysiol 1993;87: Broughton Rl. Evoked potentials and sleep-wake states in man. In: Horne J, ed. Sleep '88. Stuttgart: Gustav Fischer Verlag, 1989: Campbell K, Bell I, Bastion C. Evoked potential measures of information processing during natural sleep. In: Broughton Rl, Ogilvie RD, eds. Sleep, arousal and performance. Boston: Birkhauser, 1992: Campbell KB, McGarry PA, Bell I. Information processing during sleep: the effects of high stimulus intensity. In: Koella WP. Obal F, Schulz H, Visser P, eds. Sleep '86. Stuttgart: Gustav Fischer Verlag, 1988: Harsh J, Voss U, Hull 1, Schrepfer S, Badia P. ERP and behavioural changes during the wake/sleep transition. Psychophysiology 1994;31: Noldy N, McGarry PA, Campbell KB. Late auditory evoked potentials as indicators of sleep onset. In: Koella WP, Obal F, Schulz H, Visser J, eds. Sleep '88. Stuttgart: Gustav Fischer Verlag, 1988: Ogilvie R, Simons I, Kuderian R, MacDonald T, Rustenburg 1. Behavioral, event-related potential, and EEG/FFT changes at sleep onset. Psychophysiology 1991;28: Ujszaszi 1, Halasz P. Late component variants of single auditory evoked response during NREM sleep stage 2 in man. Electroencephalogr Clin Neurophysiol 1986;64: Ujszaszi 1, Halasz P. Long latency evoked potential components in human slow wave sleep. Electroencephalogr Clin Neurophysiol 1988;69: Hori T, Hayashi M, Kato K. Changes of EEG patterns and reaction time during hypnagogic state. Sleep Res 1991 ;20: Hori T, Hayashi M, Morikawa T. Topographical EEG changes and the hypnagogic experience. In: Ogilvie RD, Harsh JR, eds. Sleep onset: normal and abnormal processes. Washington, DC: American Psychological Association, 1994: Winer Bl, Brown DR, Michels KM. Statistical principles in experimental design, 3rd edition. New York: McGraw-Hill, Williams RL, Agnew HW Jr, Webb WB. Sleep patterns in young adults: an EEG study. Electroencephalogr Clin Neurophysiol 1964;17: Williams RL, Agnew HW Jr, Webb WB. Sleep patterns in young adult female: an EEG study. Electroencephalogr Clin Neurophysiol 1966;20: Ogilvie RD, Wilkinson RT, Allison S. The detection of sleep onset: behavioral, physiological and subjective convergence. Sleep 1989;12: Badia P, Wright KP lr, Wauquier A. Fluctuations in single hertz Sleep, Vol. /9, No.9, /996

8 738 H. TANAKA ET AL. EEG activity during the transition to sleep. In: Ogilvie RD, Harsh JR, eds. Sleep onset: normal and abnormal processes. Washington, DC: American Psychological Association, 1994: Barlow JS, Creutzfeldt OD, Michael D, Houchin ], Epelbaum H. Automatic adaptive segmentation of clinical EEGs. Electroencephalogr Clin NeurophysioI198l;51: Tanaka H, Hayashi M, Hori T. Topographical analysis of the hypnagogic EEG. In: Memoirs of the Faculty of Integrated Arts & Sciences, Hiroshima University IV. 1993;19: [In Japanese with English abstract.] 32. Tanaka H, Hayashi M, Hori T. Topographical characteristics of the hypnagogic EEG. Jpn J Psychiatry Neurol 1994;48: Tanaka H, Hayashi M, Hori T. The analysis of topographic structure of the hypnagogic EEG activity. Jpn J EEG EMG 1995;23: [In Japanese with English abstract.] Sleep, Vol. 19, No.9, 1996