[NOTE: this web page has been adapted from the original which can be viewed at

Presented at the 7th Annual Summer Sleep Workshop Multi-Site Training Program for Basic Sleep Research September 16 - 21, 1994, Lake Arrowhead, California

Periodicity of Standardized EEG Spectral Measures across the Waking Day

David A. Kaiser and M.B. Sterman
Veterans Administration Medical Center, Sepulveda, CA, and School of Medicine, UCLA


In the course of conducting numerous quantitative studies of the EEG correlates of human attention and cognitive performance we have amassed a substantial amount of topographic data from several standardized baseline conditions. The most reliable of these, with regard to assessment of background arousal, is the eyes open baseline control. For this test the subject stares at a blank or static video screen for 2-3 minutes while EEG data are collected. Due to the varied nature of these studies and the circumstances affecting subject availability, data were acquired at different times throughout the waking phase of the circadian cycle. It was possible, therefore, to examine these data across a 12 hour time period and to evaluate potential ultradian influences on the neural substrates of EEG rhythm generation.


Quantitative topographic EEG data were collected from 130 subjects during a baseline "eyes open" test condition. Both male (85%) and female (15%) subjects were studied. Mean age was 32, with a range of 19 to 43. Using a standard questionnaire it was determined that none of these subjects had chronic or acute health problems or used any prescription medications. None had used alcohol or other non-proprietary drugs during the preceding 24 hours, although tobacco use was not always assessed. All subjects reviewed and signed an approved institutional consent form.

Referential EEG recordings were obtained using a standardized, pre-measured electrode cap with 17 recording sites estimating placements specified by the International 10/20 System. Each site was referenced to linked earlobes. A ground lead was placed just anterior of FZ. Impedance readings below 5K were required before recording was initiated.
EEG data were fed into a 24 channel quantitative analysis system (Neurosearch 24, Lexicor Medical Technologies). Analog data were subjected to 3 Hz high-pass and 32 Hz low-pass filters, with rolloffs of 12 and 48 dB/octave, respectively. Data were digitized at 128 samples/sec. in 2 sec. epochs and reconstructed EEG traces reviewed for both automatic and visual artifact removal. Epochs containing significant blink or muscle artifact were deleted. Less than 10% of the data were eliminated by this procedure.

EEG files were then exported to custom, seamless FFT software. This software applied a series of 1 sec. tapered windows, each overlapping the next by 75%, for FFT calculations. This method maintained edge protection while preserving an updated temporal resolution of 0.25 sec. Resulting data were transported to spreadsheet files for sorting according to site, frequency band, and time of collection.

Analysis was based on spectral magnitude values in six 2 Hz frequency bands from 3 to 15 Hz. Data were expressed both as absolute magnitude in each band and as percentage magnitude across bands for each site in each subject. Percentage data were used to eliminate individual differences in spectral magnitudes, and to evaluate the relative modulation of activity in these frequency bands across time. An ANOVA and sequential round-robin planned comparison test were performed to determine the statistical significance of this modulation.


Different results were obtained for the absolute vs. relative spectral magnitude analyses. For absolute magnitudes both main effects and interactions (site, frequency, time of day based on 2 hr. intervals) were significant (p = <.05). A topographically generalized modulation was obtained in all frequencies with a dominant peak at approximately 2 PM. An analysis of this modulation in the parietal area (Figure 1) confirmed a significant trough between 10 and 12 AM and a peak at 2 PM in all frequency bands (p = <.05). Visual inspection of these data suggests an additional but weak ultradian modulation with a cycle of approximately 90-120 min.

Figure 1. Smoothed scatter plot showing distribution of EEG spectral magnitudes in four 2 Hz frequency bands from 130 subjects recorded during a standardized eyes open control condition at different times between 8 a.m. and 8 p.m. Date shown here are from the mid-parietal area (Pz).

Analysis of relative band differences (percent contribution to total magnitude) presented a quite different picture. The lowest frequency bands (3-5, 5-7, and 7-9 Hz) were consistent in showing little variation over time (Figure 2). Conversely, the higher frequency bands were clearly modulated, and this modulation appeared to differ among bands. Thus, relative activity in the 9-11 Hz band showed an ultradian periodicity with a variable cycle approximating 120 min. at all parietal sites. However, in the 11-13 Hz band temporal modulation was both slower and stronger, with peaks at approximately 9AM, 1 PM, and 5PM. The peak at 1-2 PM was dominant in this 4 hr. cycle. Modulation at 13-15 Hz was similar but attenuated.

Figure 2. Smoothed scatter plot showing distribution of percent EEG magnitudes in the 7-9 & 9-11 Hz and 5-7 & 11-13 Hz bands at three recording sites in parietal cortex from 130 subjects recorded during eyes open control.


The preliminary analysis of these data has disclosed several interesting findings. First, spectral magnitudes during a standardized eyes open test condition appeared to be clearly modulated throughout the waking day. Further, this modulation was consistent across a broad frequency range. Rhythmic activity was reliably suppressed during during mid-morning and early evening hours, and increased from 2-3 PM. Secondly, when the relative contribution of different EEG frequency bands within the 3-15 Hz spectrum was evaluated it was found that certain frequency components were expressed as a stable percentage of total frequency across time, while others were clearly modulated in this regard. Thus, frequencies, below 9 Hz were stable and those above 9 Hz were modulated, and this modulation was greatest but different for the 9-11 and 11-13 Hz bands.

It should be pointed out that this analysis was opportunistic, and based on a cross-sectional analysis of data. It would have been far better to have systematic longitudinal data of this kind from a group of 130 subjects. Nevertheless, the present findings do allow for some interesting speculation.

Recent neurophysiological findings provide compelling evidence for a thalamocortical generation of system-specific rhythmic patterns in the EEG. The thalamic origin of these rhythms is based on the intrinsic properties of thalamic relay neurons, their excitatory modulation by local circuitry, and the initiating influence of inhibitory input from various functional pathways. The amplitude and spread of resulting EEG rhythms appears to depend on factors conditioning cortical excitability. Recent reports suggest that a differential influence on thalamic generating mechanisms is expressed by functional pathways mediating general activation, sensory-motor processing, and cognitive integration. These different inputs have been associated with distinctive EEG frequency components.

This conclusion provides a potential explanation for the unique contribution of different frequency components to total spectral magnitude across the waking day. The lower frequencies (3-9 Hz), which have been associated with general activation, were stable in their contribution over time, perhaps defining the waking state from this perspective. Intermediate activity (9-11 Hz), associated with cognitive integration, showed a 90-120 min. ultradian contribution similar in its periodicity to the Basic-Rest- Activity-Cycle of Kleitman. This may reflect a waking manifestation of the cognitive integration cycle associated with REM/NREM periodicity during sleep. On the other hand, higher frequencies (11-15 Hz), related to sensory-motor processing, expressed a slower 4 hr. cycle. Brainstem pathways associated with the regulation of sensory and motor excitability may also be subject to intrinsic temporal modulation.
The fact that absolute magnitudes in all frequencies were similarly modulated across the day suggests a common organizing influence on the cortical expression of thalamic input. The dominant temporal pattern seen in this case was one of attenuation in the mid-morning and early evening and enhancement at mid-day. The primary influence here may instead be circadian, and involve a non-specific modulation of cortical excitability. This conclusion would be consistent with findings relating to the human sleep-wake cycle. Thus MSLT studies commonly report increased sleep latencies at the beginning and end of the day and the shortest sleep latencies at mid-day, around 2 pm.


Circadian and ultradian rhythms emerge at all topographic regions, not just parietal sites, as shown in Figure 3.

Figure 3. Smoothed scatter plot showing distribution of percent EEG magnitudes in the 11-13 Hz frequency band at recording sites Fz and Cz, respectively, from 130 subjects recorded during eyes open control.

Circadian and ultradian rhythms are prominent in all frequency bands between 3-15 Hz, as shown in Figure 4

Figure 4. Smoothed scatter plot showing distribution of percent EEG magnitudes in the six 2 Hz frequency bands (3-15 Hz) at recording site Pz from 130 subjects recorded during eyes open control.