Yordanova, J., Kolev, V. (1998). Event-related alpha oscillations are functionally associated with P300 during information processing. NeuroReport, 9: 3159-3164. Copyright © 1998 Lippincott Williams & Wilkins 

Event-related alpha oscillations are functionally associated with P300 during information processing

Juliana Yordanova* and Vasil Kolev
Institute of Physiology, Bulgarian Academy of Sciences, Acad. G. Bonchev str., bl. 23, 1113 Sofia, Bulgaria
    *Corresponding author:
      Tel.:  (359) 2-979-37-49  Fax: (359) 2-738-469
      email:  jyord@iph.bio.bas.bg

Recent findings indicate that the electroencephalographic alpha (7-14 Hz) activity is functionally involved in cognitive brain functioning, but the issue of whether and how event-related alpha oscillations may relate to the processes indexed by the P300 component of the event-related brain potentials (ERPs) has not been addressed. The present study assessed the effect of auditory oddball task processing on slow (7-10 Hz) and fast (10-14 Hz) alpha activity from the P300 latency range. ERPs from mentally counted targets (20%) and not counted nontargets (80 %) were recorded at Fz, Cz, and Pz in nine subjects. Single-sweep phase-locking, power of phase-locked, and power of non-phase-locked alpha responses during P300 activity were quantified. The results demonstrated that larger and more synchronized phase-locked fast alpha components at anterior (frontal-central) locations, with reduced non-phase-locked slow alpha responses at the parietal site were produced by targets relative to nontargets. Because the simultaneously recorded P300 and alpha activity manifested a similar sensitivity to the oddball task, event-related alpha appears to be functionally associated with the cognitive processing demands eliciting P300. Also, evidence is provided for the functional involvement of frontally synchronized and enhanced alpha oscillations in task processing. NeuroReport  9: 3159-3164 © 1998 Lippincott Williams & Wilkins
Keywords:  Cognitive alpha, Cognitive processing, EEG, Event-related oscillations, Event-related potentials (ERPs), P300,  Phase-locking

The processing of infrequently occurring target stimulus events activates specific functional mechanisms reflected by the P300 (P3) component of the event-related brain potential (ERP). Typically, larger P300 amplitudes are obtained under task-relevant compared to passive processing conditions.[1] Results from various experiments eliciting P300 have led to the suggestion that P300 generation is associated with cognitive functioning such as memory updating and attention allocation.[2,3]
Electroencephalographic (EEG) activity in the alpha (7-14 Hz) frequency band also has been demonstrated to vary with cognitive brain processes.[4,5] Event-related power changes of the EEG activity referred to as event-related desynchronization (ERD) and synchronization (ERS) [5,6] have revealed that conditions engaging attention and memory produce area- and task-specific reduction of alpha activity within one second or more after stimulation, such that the amount and duration of alpha power suppression increase with increases in cognitive (memory) load,  task relevance, or surprise value.[5,7-9] Further, mental task conditions requiring increased attention and intention [4,10,11] or working memory activation [12] have been observed also to elicit significant enhancements of EEG alpha waves. These findings imply that EEG alpha frequency is functionally associated with the cognitive activation or processes reflected by P300, but the precise nature of this relationship is not known.[13]
In a previous report, P300 amplitude and alpha ERD were found to depend in a similar manner on cognitive load and event rate, despite the distinct latency epochs of P300 and ERD occurrence.[8] If ERD amplitude and P300 are indeed different cortical indices of the same processes [7], it may be expected that when P300 emerges as an objective and time-localized marker of specific cognitive activation, EEG alpha activity within the same time period would also vary with task demands. Therefore, the objective of the present work was to assess the effects of auditory oddball task processing on the event-related alpha activity from the P300 latency range.
EEG oscillations following external stimulation may be tightly or loosely phase-coupled with stimulus.[14] During auditory P300 elicitation, alpha activity has been described to desynchronize [15], but prolonged phase-locked alpha activity also has been reported in auditory tasks.[15,16] In this regard, to analyze precisely P300-related alpha activity, the power of both the phase-locked and non-phase-locked alpha responses was measured in the present study.[17] In addition, to assess whether oddball task processing may affect the stability and repeatability of alpha patterns, the phase-locking to stimulus of single-sweep alpha responses within P300 was quantified independently of amplitude effects.[18-20]

Materials and Methods
Nine healthy 18-30 year-old adults (5 females) were assessed. None reported any neurologic, psychiatric disorders, or hearing problems. Auditory stimuli were presented in an oddball condition with intensity of 60 dB SPL, duration of 1000 ms (r/f 10 ms), interstimulus intervals between 3.5 and 6.5 s. Two tone bursts of 2000 and 1950 Hz (N = 100) were delivered randomly, with P = 0.2 for the lower (target) tones. Subjects kept their eyes closed and counted mentally the rare targets. The EEG data were recorded at Fz, Cz, and Pz with linked-mastoids as a reference using a 0.1-70 Hz band pass, with a sampling frequency of 250 Hz /12 bit. EEG segments contaminated with ocular or muscular activity, or exceeding ±50 µV, were excluded from further analysis. For each subject, 16-18 artifact-free target sweeps were processed. The same number of artifact-free sweeps was chosen randomly from the nontarget ERPs.
Averaged target and nontarget ERPs were obtained to measure the time-domain P300 component. Event-related alpha activity was evaluated in two frequency ranges:  7-10 and 10-14 Hz. Three parameters were measured:  (1) power of phase-locked activity, (2) power of non-phase-locked activity, and (3) single-sweep phase-locking. Digital filtering was performed by using a modified linear pass band filter providing zero phase shift, with filter weights based on binomial coefficients. The filter band width was greater than 5% from the total analyzed frequency band, which was experimentally tested to minimize filtering artifacts. To achieve this ratio, the original signals were downsampled to 125/s, which introduced no distortion in the signal. The exact half-power frequencies of the digital filters were 6.84, 10.25, and 14.16 Hz, referred to as 7, 10, and 14 Hz in the text. The length of the filtered single-sweep epochs was 2048 ms (-1024, +1024 ms), so that possible edge effects did not alter the analyzed epoch.
To obtain phase-locked alpha power, single sweeps were filtered, then averaged and squared. Non-phase-locked alpha power was calculated according to the intertrial variance method [17] based on the following procedure:  From each single sweep band pass filtered in the alpha range, the averaged ERP filtered in the same range was subtracted. The resulting sweeps were then squared and averaged, thus obtaining the instantaneous non-phase-locked power. Only for the sake of presenting alpha reactivity in a manner comparable with other studies, ERS/ERD curves were obtained by means of the Hilbert transform.[21] For quantitative evaluation of single-sweep phase-locking, the single-sweep wave identification (SSWI) method was applied.[18-20] A histogram of the number of phase-locked single-sweep alpha waves (SSWI histogram) was obtained according to the following procedure:  The filtered single sweeps were coded such that the extrema were replaced with (+1) or (-1) for maxima and minima, respectively. The time points not belonging to the extrema were replaced by zero. Thus, for each sampling point, the sum of the identified coded (+1, -1) extrema was calculated for the trial set and the number of the phase-locked waves was determined. The obtained value was represented in a corresponding histogram bar. The histogram was normalized according to the number of single sweeps used for analysis.
For statistical evaluation, mean power values of phase-locked and non-phase-locked activity in the two frequency ranges were measured for the time window 250-600 ms, in which P300 was expressed (Fig. 1). Power values were log10-transformed to normalize the distribution. The sum of the absolute histogram values was calculated for the same time window (250600 ms). Measurements of slow and fast alpha activity were made for each subject, stimulus type, and electrode site. Each parameter was subjected to a repeated-measures analysis of variance with two within-subjects variables:  stimulus (target vs nontarget), and electrode (Fz, Cz, Pz). P300 latency was measured as the latency of the maximum ERP peak within 250-600 ms from stimulus onset. P300 amplitude was measured relative to a prestimulus baseline of 200 ms before stimulus. P300 amplitude and latency values were subjected to 2 stimulus x 3 electrode analyses of variance. The Greenhouse-Geisser correction was applied to the analyses with repeated measures factor electrode. The original df and corrected probability values are reported throughout the text.

P300:   Figure 1 illustrates that P300 amplitude was significantly larger (F(1/8) = 34.08, P < 0.001) and P300 latency was significantly longer (F(1/8) = 6.2, P < 0.05) to targets than to nontargets. P300 amplitude demonstrated a parietal maximum (F(2/16) = 22.3, P < 0.001).
Phase-locked alpha activity:   Figure 1 (left panel) displays the time course of phase-locked alpha activity. For 7-10 Hz components, a prominent power increase with a maximum at around 150-200 ms was observed for both stimulus types. For 10-14 Hz activity, a biphasic power increase was detected at anterior locations within 1 s after stimulus on-set. The second enhancement of phase-locked 10-14 Hz responses occurred substantially later for the nontargets. Figure 2 illustrates the effects of stimulus type and electrode on phase-locked alpha activity from the P300 range. The power of phase-locked 7-10 Hz responses did not depend on stimulus and electrode. In contrast, phase-locked 10-14 Hz activity was most pronounced at frontal and central locations (F(2/16) = 4.55, P < 0.05), and was significantly larger for targets than for nontargets (F(1/8) = 6.3, P < 0.05) - effects seen also in Fig. 1.
Non-phase-locked alpha activity:   Figure 1 (middle panel) presents time courses of non-phase-locked alpha activity and shows that at posterior sites (Pz and Cz) targets produced  prominent alpha power decreases reaching maximum at 500-1000 ms. As illustrated in Fig. 2, the slow  alpha (7-10 Hz) variance within P300 latency range was significantly smaller for targets at Pz  (stimulus x electrode, F (2/16) = 4.8, P < 0.05). No statistically significant effects were found for the non-phase-locked 10-14 Hz activity.
Alpha ERS/ERD:   Figure 1 (right panel) illustrates that P300 time window coincided with the transition of ERS to ERD. As argued previously [15,17], ERS corresponded to the enhancement of phase-locked and non-phase-locked responses, while ERD dynamics was underlined by the non-phase-locked responses. Although not quantified in the present study, it can be seen in the figure that at Fz, the ERS of 10-14 Hz activity within P300 was much larger for the targets.
Single-sweep alpha phase-locking:   As depicted in Fig. 2, the phase-locking of 7-10 Hz responses within P300 range did not depend on stimulus type and electrode, while the phase-locking of 10-14 Hz responses was significantly stronger for the targets (F(1/8) = 8.80, P < 0.05) and at anterior than at parietal locations (F(2/16) = 3.92, P < 0.05).
Fig. 1. Grand average ERPs, phase-locked and non-phase-locked alpha power, and ERS/ERD at three electrode locations. Shaded areas designate the time window used for analysis. Stimulus starts at 0 ms and lasts till 1000 ms.
Fig. 2. Effects of stimulus type and electrode on the phase-locked and non-phase-locked alpha power, and on the normalized number of phase-locked single alpha waves in the time window 250-600 ms. 
Values are mean +(-) s.e.

The functional role of EEG alpha activity in cognitive brain functioning has been emphasized in previous reports [4-9,12], but the issue of whether and how event-related alpha oscillations may reflect the processes activated during and indexed by the endogenous P300 ERP component has not been considered. The present study separately quantified the power of phase-locked, the power of non-phase-locked, and the phase-locking of alpha responses coinciding with P300. The results demonstrated that both the phase-locked and non-phase-locked alpha oscillations during auditory P300 were functionally relevant to the oddball task processing.  These findings provide new evidence for (1) the association of event-related alpha activity with the processes eliciting P300, (2) the functional involvement of synchronized and enhanced frontal alpha oscillations in task processing, and (3) a putative mechanism that may account for the more synchronized and ordered alpha states that accompany cognitive processes activated for target stimulus processing and P300 generation.
Event-related alpha activity and P300:   Previous reports have shown that enhanced and phase-locked alpha oscillations can be consistently observed shortly after auditory and visual stimuli, e.g., in the first 250 ms [4,9,14,19]. These synchronized alpha oscillations, called the alpha response [14], typically coincide with the exogenous ERP components.
A major finding of the present study was that the power and pattern stability of phase-locked 10-14 Hz activity within the P300 range (250-600 ms) were significantly larger for the mental count targets than for nontargets. It is noteworthy that phase-locked alpha components were obtained for a latency period much later than that of the classical alpha response, as noted previously for another auditory task condition.[16] It is not likely that spectral P300 components are responsible for the observed task effects:  First, in line with many previous reports (rev. Ref. 1), P300 was with a parietal maximum whereas the synchronization and power of phase-locked 10-14 Hz activity were maximal at frontal and central locations (Figs. 1 and 2). Second, the dominant spectral components of P300 have been recognized to belong to sub-delta, delta, and theta frequency ranges [15,22,23], which is also implied by the present ERP results (Fig. 1). Thus, although fast alpha (10-14 Hz) energy may participate in the frontal-central portions of P300, the larger and better synchronized 10-14 Hz oscillations appear to reflect a frontally distributed specific state or process activated during the mental count condition. However, the synchronized fast alpha oscillations occurred simultaneously with P300 and, like P300, were functionally relevant to the oddball task processing. Hence, it may be concluded that the process reflected by fast alpha synchronization at frontal locations is functionally associated with the major cognitive demands eliciting P300, e.g., attention allocation and working memory activation.[2,3] This conclusion is supported by previous findings according to which a substantial increase of fast (10-12 Hz) alpha activity is produced over large frontal and central regions by auditory stimulus memorization and retrieval, and by visual attention [12]. In these experiments, fast alpha ERS was maximal at 250-500 ms after stimulation and lasted until 500-700 ms during attention and retrieval from memory, and much longer during memorization.

Furthermore, increased attentional demands in auditory oddball tasks have been recently reported to produce significant increases in fast alpha power of the ERP epoch.[22] The power- independent augmentation of phase-locking to targets as observed here (Fig. 2) additionally indicates that fast alpha synchronization within P300 is associated with stimulus-specific processing rather than with processing of task in general. These findings strongly emphasize the role of enhanced and synchronized alpha oscillations in higher brain functioning [10,4,11].

Concurrently, the parietal non-phase-locked slow alpha (7-10 Hz) activity within P300 was more suppressed to targets than to nontargets, which resulted in ERD expression (Fig. 1, Ref. 15). This observation is consistent with previous ERD reports demonstrating that alpha power is reduced by relevant stimulation, such that alpha attenuation is maximal at posterior sites and increases with cognitive load and stimulus significance [5,7-9,12]. The alpha reduction reported here, though differentiating targets from nontargets during P300, reached maximum much later than P300 peaking (Fig. 1, see also Refs. 7,8). Hence, the mechanisms reflected by the parietal slow alpha suppression may appear secondary to P300. As slow alpha ERD has been related with unspecific attentional processes [5,7], such processes may be triggered by the stimulus evaluation performed during P300.[3] Because no task effects were detected for the non-phase-locked fast alpha, it may be further proposed that the memory processes supposed to be associated with the ERD of fast alpha activity [5] are not identical to those eliciting P300.

According to the present results, during P300 time window, the frontal increase in phase-locked fast alpha activity to targets was accompanied by a parietal suppression of non-phase-locked slow alpha activity. Following the classical interpretation, the simultaneous existence of ERS and ERD in distinct scalp areas is explained by accepting that ERD reflects functionally activated cortical regions, and ERS manifests a temporary inactivity in other cortical fields.[6,5] However, it seems improbable that frontal areas are functionally inactive during the cognitive demands imposed by the mental count task, with targets producing even greater frontal inactivity than nontargets, as may be concluded if phase-locked alpha power increase would manifest a cortical state of rest [24] (for a similar result see Refs. 12, 22). Another possible interpretation of the present observations is that there are anterior and posterior alpha systems in the brain that are as functionally separated as to employ entirely different frequencies (slow and fast) and mechanisms (synchronizing and desynchronizing) during relevant event processing. Although frontal and occipital generators of alpha activity have been proposed to exist [25], the scalp topography of slow and fast alpha ERD reported so far has demonstrated no clear frontal vs. posterior differences between the two frequency bands. Instead, slow alpha ERD was reported to be widely distributed over the scalp, while fast alpha ERD was found to be more localized.[5,9,12] Alternatively, within the concept of diffuse and selectively distributed alpha systems in brain [14,4], the topography-specific coexistence of suppressed non-phase-locked slow alpha activity and enhanced phase-locked fast alpha activity may be regarded as a higher-order alpha state, during which the non-phase-locked (or disordered) alpha is minimized and the phase-locked (or ordered) alpha is maximized. Such a viewpoint permits to regard the well known ERD phenomenon not merely as a reduction of alpha power, but as a mechanism which, in addition to phase-ordering, tends to stabilize alpha brain states in relation to relevant information processing.

The present findings clearly demonstrate that event-related alpha activity is associated with the processes eliciting auditory P300 ERP component, such that, during P300 generation, frontal alpha oscillations are increased and synchronized, and parietal alpha activity is suppressed to targets than to nontargets. Thus, new evidence is provided for the functional involvement of synchronized and enhanced frontal alpha oscillations in task processing.

ACKNOWLEDGEMENTS:  Research was supported by the National Research Fund by the Ministry of Education and Science, Sofia, Bulgaria (Project B-703/97). Thanks are due to Dr. John Polich for comments and suggestions.

[1] Polich J. J Clin Neurophysiol 15, 14-33 (1998).
[2] Donchin E and Coles MGH. Behav Brain Sci 11, 357-374 (1988).
[3] Polich J and Kok A. Biol Psychology 41, 103-146 (1995).
[4] Basar E, Schürmann M, Basar-Eroglu C et al. Int J Psychophysiol 26, 5-30 (1997).
[5] Klimesch W. Int J Psychophysiol 26, 319-340 (1997).
[6] Pfurtscheller G and Klimesch W. Event-related synchronization and desynchronization of alpha and beta waves in a cognitive task. In:  Basar E and Bullock TH, eds. Induced Rhythms in the Brain. Boston:  Birkhäuser, 1992: 117-128.
[7] Boiten F, Sergeant J and Geuze R. Electroenceph Clin Neurophysiol 82, 302-309 (1992).
[8] Sergeant J, Geuze R and van Winsum W. Psychophysiology 24, 272-277 (1987).
[9]  Sterman MB, Kaiser DA and Veigel B. Brain Topogr 9, 21-30 (1996).
[10] Ray W and Cole H. Science 228, 750-752 (1985).
[11]  Shaw J. Int J Psychophysiol 24, 7-23 (1996).
[12] Krause C, Lang H, Laine M et al. Brain Topogr 8, 47-56 (1995).
[13] Polich J. Electroenceph Clin Neurophysiol 104, 244-256 (1997).
[14] Basar E. EEG Brain Dynamics:  Relation between EEG and Brain Evoked Potentials. Amsterdam:  Elsevier, 1980.
[15] Van Dijk J, Caekebeke J, Jennenkens-Schinkel A et al. Electroenceph Clin Neurophysiol 83, 44-51 (1992).
[16] Kolev V and Schürmann M. Int J Neurosci 67, 199-213 (1992).
[17] Kalcher J and Pfurtscheller G. Electroenceph Clin Neurophysiol 94, 381-384 (1995).
[18] Kolev V and Yordanova J. Biol Cybern 76, 229-235 (1997).
[19]  Yordanova J and Kolev V. Electroenceph Clin Neurophysiol 99, 527-538 (1996).
[20] Yordanova J and Kolev V. Psychophysiology 35, 116-126 (1998).
[21] Clochon P, Fontbonne J-M, Lebrun N et al. Electroenceph Clin Neurophysiol 98, 126-129 (1996).
[22] Kolev V, Demiralp T, Yordanova J et al. NeuroReport 8, 2061-2065 (1997).
[23] Spencer K and Polich J. Psychophysiology (in press).
[24] Fuster J. The Prefrontal Cortex. Anatomy, Physiology, and Neuropsychology of the Frontal Lobe. New York:  Raven Press, 1989.
[25] Inouye T, Shinosaki K, Yagasaki A et al. Electroenceph Clin Neurophysiol 63, 353-360 (1986).

Received 1 July 1998;
accepted 23 July 1998