Paradoxical' alpha synchronization in a memory task

Cognitive Brain Research 7 Ž1999. 493–501 Research report ‘Paradoxical’ alpha synchronization in a memory task W. Klimesch ) , M. Doppelmayr, J. Schwaiger, P. Auinger, Th. Winkler Department of Physiological Psychology, UniÕersity of Salzburg, Institute of Psychology, Hellbrunnerstr. 34, A-5020 Salzburg, Austria Accepted 8 December 1998 Abstract The results of a specially designed memory search paradigm which maximizes episodic short-term memory ŽSTM. and minimizes semantic long-term memory ŽLTM. demands show that the upper alpha band synchronizes selectively in those conditions and time intervals where episodic STM demands are maximal. This finding of a selective alpha synchronization occurring only in the upper alpha band and during highest task demands is surprising because it is well known that usually alpha desynchronizes during mental activity. Because experiments from our laboratory indicate that desynchronization in the upper alpha band is related to semantic LTM processes, the present finding suggests that a selective synchronization in this frequency band reflects inhibition of semantic LTM. It is assumed that once the capacity limits of STM are reached or exceeded, processing resources are no longer distributed and that potentially interfering, task irrelevant, brain areas or processing systems are inhibited. q 1999 Elsevier Science B.V. All rights reserved. Keywords: Event-related desynchronization; Alpha oscillation; Synchronization; Theta; Memory 1. Introduction During the last few years, alpha oscillations have attracted considerable interest as recent publications indicate Žsee, e.g., Refs. w1,5,19,22x.. The well known finding that alpha desynchronizes during a variety of different tasks, seems to imply that desynchronization reflects a state of active cortical information processing. More recent studies have shown that alpha desynchronization is a local phenomenon that occurs over task relevant brain areas whereas task irrelevant regions show a pronounced synchronization. As an example, in a task where subjects had to judge visually presented items and to give a motor response, Pfurtscheller and Klimesch w21x have found that during visual stimulation alpha desynchronizes over occipital recording sites whereas over the motor cortex a strong synchronization could be observed. During the motor response, the opposite result was obtained. Now, regions over the motor cortex desynchronized and occipital areas synchronized. These and similar findings have led to the hypothesis that in contrast to desynchronization, alpha synchronization indicates a state of ‘idling’ w22x or even a state of cortical inhibition w7x. ) Corresponding author. Fax: q43-662-8044-5126; E-mail: [email protected] The present study was designed to test the validity of the hypothesis that alpha synchronization reflects inhibition of task irrelevant cortical processes. In a series of experiments we have found that upper alpha desynchronization responds to semantic long-term memory ŽLTM. demands, whereas theta synchronization responds to episodic short-term memory ŽSTM. tasks Že.g., Refs. w9– 11x.. These findings which have been replicated by other research groups Že.g., Ref. w2x. suggest that in a task which maximizes episodic STM and minimizes semantic LTM demands, the upper alpha band will start to synchronize, particularly under conditions where episodic demands are very difficult. A memory search paradigm w23x was used to vary episodic memory demands. In each trial, subjects viewed a string of 5 or 10 characters Žletters and numbers; termed ‘memory set’. which they were asked to retain in memory. Then, after a retention interval of 5 s, a single character Žtermed ‘frame’. appeared. If the frame was contained in the memory set, subjects had to respond with ‘yes’, otherwise with ‘no’. Episodic memory demands are varied by two factors, memory load Ž5 versus 10 characters per set. and overall processing demands Žconsistent versus varied mapping condition.. Under the consistent mapping condition subjects knew, which characters the memory set on the next trial will contain, under the varied condition each memory set contained different characters. In the varied condition episodic demands increase because characters of 0926-6410r99r$ - see front matter q 1999 Elsevier Science B.V. All rights reserved. PII: S 0 9 2 6 - 6 4 1 0 Ž 9 8 . 0 0 0 5 6 - 1 494 W. Klimesch et al.r CognitiÕe Brain Research 7 (1999) 493–501 the current and preceding memory set must not be confused. On the basis of this design, task difficulty is associated only with episodic STM but not with semantic LTM demands. In the most difficult condition Ž10 characters; varied mapping. subjects have to encode the simultaneously presented character string, keep them in STM for 5 s and to scan STM after the frame is presented. This procedure is repeated every 11 s for a total of 240 trials. Semantic LTM demands are minimized and do not increase with task difficulty because the characters Žthe numbers 1, 2, . . . 8 and the consonants B, C, F, H, K, L, R. are meaningless on the semantic encoding level. This also holds true for combinations or strings of characters Žsuch as F, 3, 7, H, R. which are used as memory sets. The idling and inhibition hypotheses of alpha synchronization lead to different predictions. The inhibition hypothesis suggests that with increasing task difficulty, processing resources are no longer distributed but become concentrated on the episodic STM system and that potentially interfering, task irrelevant brain areas or processing systems are inhibited. Thus, once the capacity limits are reached or exceeded, semantic LTM processes become inhibited. We expect that at least in the most difficult condition Ž10 characters; varied mapping. the upper alpha band will show a task related increase in band power. According to the idling hypothesis task irrelevant brain areas are in a resting state. Thus, there is no reason to assume that with increasing task difficulty alpha synchronization will increase. Episodic memory demands dominate early after the presentation of the memory set, because at least 5 different characters must be encoded simultaneously and retained in STM, whereas only one character is presented for retrieval. Accordingly, we assume that upper alpha synchronization will be significantly larger during encoding than during retrieval. 2. Materials and method 2.1. Subjects A sample of 14 right handed students Ž7 males and 7 females. participated in the experiment. Their mean age was 23.7 years with a range of 19.4 to 34.5 years. Before participating in the experiment, subjects were asked about the hand they use in different tasks such as handwriting, throwing a ball, etc. A subject was considered right-handed if hershe indicated to use the right hand for all of these different tasks. 2.2. Materials Frame set size was kept constant throughout the experiment and consisted of only one character. Memory set size Žmemory load. varied and comprised either 5 or 10 characters. Memory sets were constructed from 15 characters Žthe numbers 1, 2, . . . 8 and the consonants B, C, F, H, K, L, R.. By using the following sampling procedure, 60 memory sets were drawn for each of the two memory load and mapping conditions. In the consistent mapping condition, 6 blocks with 10 different memory sets were used for each memory load condition. Within a single block, the characters and their ordering remained identical for 10 subsequent trials. Between blocks, characters varied. They were drawn randomly with the restriction that each character occurs with equal frequency Žhence the use of 15 characters.. In the varied mapping condition 60 different memory sets were drawn Žfor each of the two memory load condition. by using the same sampling rule. For both mapping conditions, the frames were randomly selected Žfrom the same set of 15 characters from which the memory sets were constructed. by considering the following restrictions. Within each of the four experimental conditions, each frame had to occur with equal frequency and half of the frames had to be contained in the memory set and, thus, required a yes response Žthe other half required a no response and was not contained in the memory set.. 2.3. Design Subjects had to respond by pressing a ‘yes’ response key if the letter or number of the frame was contained in the memory set, otherwise with ‘no’. Each of the four experimental sessions was preceded by a practice session of 10 trials. Each subject was tested under all of the four experimental conditions. Presentation sequence was counterbalanced between subjects. 2.4. Procedure The string of characters of the memory set and the frames appeared at the centre of the screen of a computer controlled monitor. Characters Žletters or numbers. were 3 cm in height. The length of a 10 character-string was 23 cm. Subjects sat at a distance of 1.4 m from the monitor. Each trial had a length of 11 s. The first 1000 ms were used as reference interval Žsee under ‘event-related desynchronization’ ŽERD., below.. After 1500 ms a brief auditory warning signal ŽWS. was presented Ž3000 Hz, lasting for 250 ms.. The memory set was presented for 3000 ms and appeared 1000 ms after the onset of the WS. The frame was presented for 250 ms and appeared 6000 ms after the onset of the WS. 2.4.1. Apparatus EEG-signals were amplified by a 32-channel biosignal amplifier system Žfrequency response: 0.16 to 30 Hz., subjected to an anti-aliasing filterbank Žcut-off frequency: 30 Hz, 110 dBroctave. and were then converted to a digital format via a 32-channel ArD converter. Sampling W. Klimesch et al.r CognitiÕe Brain Research 7 (1999) 493–501 rate was 128 Hz. The data were processed by the B.E.S.T.-system of Grossegger and Drbal. During data acquisition, EEG signals were displayed online on a high resolution monitor and stored on disk. 2.4.2. Recordings A set of 25 silver electrodes, attached with a glue paste to the scalp was used to record EEG-signals. Thirteen electrodes were placed according to the International Electrode Ž10–20. Placement-System, at F3, F4, Fz, C3, C4, Cz, T3, T4, P3, P4, Pz, O1 and O2. From the remaining twelve electrodes, 4 electrodes were placed over parietooccipital areas ŽPO3, PO1, PO2, PO4., 4 electrodes were placed over centro-parietal areas ŽCP5, CP1, CP2, CP6., and 4 electrodes were placed over fronto-central areas ŽFC5, FC1, FC2, FC6.. In addition to the 25 electrodes described above, two ear lobe electrodes Žtermed A1 and A2., were attached to the left and right ear. The Electrooculogram ŽEOG. was recorded from 2 pairs of leads in order to register horizontal and vertical eye movements. All data were referentially recorded by using a common reference placed on the nose. In order to eliminate the effects of the nose reference as well as other types of artifacts, the EEG recordings were corrected by subtracting the arithmetically averaged ear lobe recordings ŽA1 q A2.r2 from all of the other recordings. This type of recording is also termed ‘referential recording with re-referencing against averaged ear lobes’. All of the epochs were checked individually for artifacts Žeye blinks, horizontal and vertical eye movements, muscle artifacts etc.. by visual inspection. Only epochs with a correct yes or no response were used for data analysis. The percentage of epochs that were excluded from data analysis Ždue to artifacts. for hits and correct rejections ranged from 28% to 42% in the four experimental conditions. 2.4.3. The calculation of ERD Event-related changes in band power were calculated by using the ERD-method which was originally proposed by Pfurtscheller and Aranibar w20x. Before calculating ERD, the data for each epoch and each of the 25 channels were digitally band-pass filtered, squared Žin order to obtain simple power estimates. and averaged separately for each experimental condition and for each subject. Based on these data, ERD is defined as the percentage of a decrease or increase in band power during a test interval as compared to a reference interval: ERD%s wŽband power, reference interval. y Žband power, test interval.xrŽband power, reference interval.4 = 100. As mentioned earlier, the first 1000 ms of each trial were used as reference interval. The test intervals are consecutive time intervals of 500 ms. Positive ERD-values are obtained when power in the test interval decreases. On the other hand, negative ERD-values Žor ERS., indicate that band power increases 495 Žsynchronizes. in the test with respect to the reference interval. 2.4.4. The indiÕidual determination of frequency bands The frequency windows for the theta and alpha bands were determined individually for each subject i by using mean peak frequency f Ž i . of the dominant EEG frequency in the alpha band for all recording sites as an anchor point. Frequency f Ž i . was determined from spectra which were calculated over the entire epoch of 11 s and then were averaged over all epochs and leads. In using f Ž i . as individual anchor point, four different frequency bands with a bandwidth of 2 Hz each were defined: Ž f Ž i . y 6. to Ž f Ž i . y 4.; Ž f Ž i . y 4. to Ž f Ž i . y 2.; Ž f Ž i . y 2. to f Ž i .; f Ž i . to Ž f Ž i . q 2.. The ERD was calculated within these individually determined frequency bands which are termed theta, lower-1 alpha, lower-2 alpha, and upper alpha. Averaged over the sample of subjects, alpha frequency was 10.2 Hz. Thus, when considering averaged values, the four frequency bands show the following cut-off points: 4.2 Hz–6.2 Hz, 6.2 Hz–8.2 Hz, 8.2 Hz–10.2 Hz, 10.2 Hz–12.2 Hz. 2.4.5. Statistical analyses Three factorial ANOVA’s were calculated for hits and correct rejections, for each frequency band and frontal ŽF3, F4, FC5, FC6., central ŽC3, C4, FC1, FC2, CP1, CP2., parietal ŽP3, P4, CP5, CP6., temporal ŽT3, T4. and occipital ŽO1, O2, PO3, PO1, PO2, PO4. recording sites. For these sites, data were averaged separately for the left and right hemisphere but only for those 1 s intervals of an epoch which represent the first 1000 ms after presentation onset of the memory set and frame. Thus, in an attempt to avoid complex interactions, electrode location and brain area were not used as separate factors. The factors and their levels are ENCODE–RETRIEVE Žmemory set, frame., LOAD Žload 5, load 10. and DEMAND Žconsistent, varied.. Because of the large number of ANOVA’s Ž2 response conditions= 4 frequency bands = 5 recording regions= 2 hemispheress 80 ANOVA’s. only results which are significant at or beyond the 2.5%level will be reported. The degrees of freedom Ž df . for main effects and interactions are 1 Ž df numerator. and 13 Ž df denominator. in all of the cases. For a parsimonious description of the results, the df ’s will not be reported together with significant F-values. 3. Results 3.1. BehaÕioral data Averaged over the sample of 14 subjects, the percentage for hits in the conditions, load 5rconsistent, load 5rvaried, load 10rconsistent and load 10rvaried are: 95.7%, 82.3%, 98% and 72%. Listed in that same order, the percentage for correct rejections are: 98.7%, 90%, 98% and 80%. 496 W. Klimesch et al.r CognitiÕe Brain Research 7 (1999) 493–501 3.2. ERD r ERS 3.2.1. Theta band, hits Main effects were found for ENCODE–RETRIEVE only and here at frontal Žleft and right hemisphere., central Žleft and right hemisphere. and parietal Žleft hemisphere. sites. Listed in this sequence the F-values and their significance levels are: F s 17.11, p - 0.01; F s 6.67, p 0.025; F s 19.49, p - 0.01; F s 9.92, p - 0.01; F s 9.97, p - 0.01.. The respective means indicate that the amount of event-related synchronization ŽERS or negative ERD. is significantly larger during retrieval Žframe. than encoding Fig. 1. Theta band, time course of ERDrERS for the first 8.5 s of the experimental trial Žepoch. representing hits. Data are averaged for both hemispheres. Bold horizontal lines indicate the reference interval and those time periods during encoding and retrieval, which were subjected to statistical data analysis. Faint horizontal lines indicate exposure durations for the WS, the memory set and the frame. At frontal and occipital sites, the increase in ERS with task difficulty is stronger during encoding than retrieval Žcf. the vertical lines.. Note that a large ERS was found in response to the WS particularly at frontal but not at occipital recording sites. At frontal sites, theta ERS Žfor the most difficult condition, Load 10 varied. is particularly strong during retention and after the memory set disappeared. W. Klimesch et al.r CognitiÕe Brain Research 7 (1999) 493–501 Žmemory set.. For frontal sites, this result is depicted in Fig. 1a which shows the time course of ERDrERS over most of the trial. Note that the ANOVA results refer only to the time interval of 1 s after onset of the memory set and frame. These intervals are marked by bold horizontal lines below the respective diagrams. Two significant inter- 497 actions were observed, one at frontal sites over the left hemisphere ŽENCODE–RETRIEVE = DEMAND; F s 6.6, p - 0.025. and another at occipital sites over the right hemisphere ŽENCODE –RETRIEVE = DEMAND = LOAD; F s 6.39, p - 0.025.. The two-factorial interaction which was found at frontal sites can be interpreted on Fig. 2. Upper alpha band, time course of ERDrERS for the first 8.5 s of the experimental trial Žepoch. representing hits. Data are averaged for both hemispheres. Bold horizontal lines indicate the reference interval and those time periods during encoding and retrieval, which were subjected to statistical data analysis. Faint horizontal lines indicate exposure durations for the WS, the memory set and the frame. Note that a large upper alpha ERS was found during encoding in the most difficult condition ŽLoad 10 varied., whereas a pronounced ERD can be observed during retrieval Žpresentation of the frame.. At frontal and temporal sites ERS increases with task difficulty Žcf. the vertical lines in the diagram and the significant interactions reported in Sections 3.2.6 and 3.2.7.. 498 W. Klimesch et al.r CognitiÕe Brain Research 7 (1999) 493–501 the basis of Fig. 1a Žcf. the vertical lines in the diagram. which indicates that the increase in ERS from the easiest to the most demanding condition Ži.e., from Load 5, consistent to load 10, varied mapping. is larger during encoding than retrieval. Inspection of the respective data underlying the three-factorial interaction reveals a similar result. As compared to the easiest, in the most demanding condition the increase in ERS is stronger during encoding than retrieval. Although averaged over both hemispheres, the data in Fig. 1c show the tendency of this finding. 3.2.2. Theta band, correct rejections Main effects were found for ENCODE–RETRIEVE at frontal Žfor both hemispheres. and central sites Žfor the left hemisphere only.. The corresponding F-values and their significance levels are: F s 6.74, p - 0.025; F s 16.74, p - 0.01; F s 9.35, p - 0.01.. As for the hits, the extent of synchronization ŽERS. is larger at these sites during retrieval as compared to encoding. Significant interactions were obtained for ENCODE–RETRIEVE= LOAD at central sites Žleft hemisphere: F s 6.98, p - 0.025; right hemisphere: F s 6.96, p - 0.025. and for ENCODE–RETRIEVE= DEMAND at left temporal sites Ž F s 18.01, p - 0.01.. The respective means indicate that during encoding Žas compared to retrieval. the increase in ERS is stronger in the more demanding conditions Žload 10 and varied. at left temporal sites. The opposite holds true for central regions, where increasing memory load leads to a stronger increase in ERS during retrieval. 3.2.3. Lower-1 alpha, hits and correct rejections No significant main effects or interactions were obtained reaching or exceeding the 2.5% level. This holds true for hits as well as correct rejections. 3.2.4. Lower-2 alpha, hits Two significant main effects for ENCODE–RETRIEVE were found over left central Ž F s 7.05, p - 0.025. and left parietal sites Ž F s 7.21, p - 0.025.. The respective means show that during retrieval ERD is much stronger than during encoding. 3.2.5. Lower-2 alpha, correct rejections With the exception of a significant interaction for right frontal recordings ŽENCODE–RETRIEVE= LOAD; F s 11.62, p - 0.01., none of the results reached or exceeded the 2.5% level. This interaction indicates that the extent of desynchronization or ERD is stronger during retrieval than encoding and for load 10 as compared to load 5. 3.2.6. Upper alpha, hits With the exception of occipital areas, for all of the recording sites and both hemispheres factor ENCODE– RETRIEVE reached significance. The F-values for left, Fig. 3. Power spectra for selected recording sites, averaged over all subjects. The spectra Žwith a frequency resolution of 1 Hz. were calculated for the 1-s intervals Žreference, memory set and frame. as indicated by bold horizontal lines in Figs. 1 and 2. Note that at Pz, absolute alpha power reaches a maximum during encoding Žpresentation of the memory set.. W. Klimesch et al.r CognitiÕe Brain Research 7 (1999) 493–501 right frontal; left, right central; left, right temporal and left, right parietal regions are in that order: F s 7.55, p - 0.025; F s 9.25, p - 0.01; F s 15.15, p - 0.01; F s 17.89, p 0.01; F s 15.35, p - 0.01; F s 17.41, p - 0.01; F s 8.23, p - 0.025; F s 9.03, p - 0.01. The respective means show a strong ERD during retrieval but ERS during encoding. Significant interactions were found for ENCODE–RETRIEVE = DEMAND at right frontal Ž F s 7.9, p - 0.025. and right temporal Ž F s 11.08, p - 0.01. sites. With increasing demands, upper alpha ERS increases Ž!. during encoding ŽFig. 2a,b.. 3.2.7. Upper alpha, correct rejections Significant effects were found for ENCODE–RETRIEVE over both hemispheres at central and temporal regions and at left parietal sites Ž F s 11.03, p - 0.01; F s 8.06, p - 0.025; F s 12.49, p - 0.01; F s 8.13, p 0.025; F s 8.03, p - 0.025.. As for the hits, the respective means for correct rejections reveal a strong ERD during retrieval but ERS during encoding. Significant interactions were obtained for ENCODE–RETRIEVE= DEMAND at left frontal Ž F s 7.29, p - 0.025., left temporal Ž F s 8.73, p - 0.025. and left parietal Ž F s 6.82, p - 0.025. sites. These interactions are very similar to those found for hits Žcf. Fig. 2.. 3.3. Power spectra at Fz, Cz and Pz Because an event-related increase in alpha power is an unusual result, power spectra were calculated for the most demanding condition Žload 10, varied. where the largest alpha ERS was found. In order to give a brief overview of the results, spectra for Fz, Cz and Pz Žaveraged over the sample of subjects. are shown in Fig. 3. Most interestingly, the spectra for Cz and Pz show that during encoding, power is larger as compared to the retrieval and even the reference interval. 4. Discussion As was predicted by the ‘inhibition’ hypothesis, significant upper alpha synchronization was observed only during that condition where episodic STM demands are maximal Ži.e., during load 10, varied mapping. and only during that time interval where episodic encoding processes dominate Ži.e., during the presentation of the memory set but not the frame.. Within the alpha band, the extent of synchronization was largest for upper alpha, although similar trends could be observed for the lower-1 and lower-2 alpha bands. Inspection of the power spectra in Fig. 3 reveals that synchronization and desynchronization occur in frequency ranges that vary between sites. As an example, at Fz and Cz, the increase in power during the presentation of the memory set is restricted to the upper alpha band and if frequency bands would not have been 499 adjusted to the individual alpha peak Žas was done for the ERD analysis in this study., desynchronization in the lower alpha band will mask and cancel the increase in upper alpha power. This fact and the use of broad frequency bands may be the reason why other ERD studies failed to find similar effects w8,24x. In a memory search paradigm studied in Klimesch et al. w8x, a much broader frequency range for calculating ERD was used. Sterman et al. w24x, on the other hand, did not use individually adjusted frequency bands. The findings of the present study are in good agreement with research reported by Krause et al. w14,15x who used a Sternberg paradigm with acoustic Žinstead of visual. stimuli. They have found pronounced Žbroad band. alpha synchronization during the presentation of the memory set. Because other types of tasks Žas e.g., listening to music. also showed pronounced alpha ERS Že.g., Ref. w16x., the conclusion is that acoustic stimulation may be the main reason for obtaining ERS instead of ERD. A possible interpretation of these findings is that in contrast to occipital regions, electrophysiological changes are difficult to detect by scalp electrodes in the primary auditory cortex, which is rather deep inside the brain. Because other cortical regions should not be directly involved in an auditory perception task, there is no reason to assume that alpha desynchronizes over those cortical regions which are not directly involved in task performance. In a variety of studies we have found that lower alpha ERD reflects attentional demands Žcf. the review in Ref. w7x.. In recent experiments w12,13x, we were able to differentiate between two types of attentional processes Žalertness and expectancy. in two subbands of lower alpha. The lower-1 alpha band desynchronizes in response to any alerting stimulus. In contrast, desynchronization in the lower-2 alpha band was observed when subjects were expecting the presentation of an imperative stimulus Žto which a response had to be given. and during that time period Žpoststimulus. when subjects were comparing the expected with the actually presented stimulus. When considering these results for the interpretation of the present findings, we predict that desynchronization in the lower-1 alpha band is of comparable magnitude during encoding and retrieval Žin all demand conditions., because alertness should be equally important in all cases. The complete lack of significant findings for the lower-1 alpha band supports this view. Unlike alertness, expectancy should differ during encoding and retrieval. Only during retrieval, expectancy might play a role in the sense that subjects may await the appearance of a particular frame. Thus, ERD in the lower-2 alpha band should be somewhat stronger during retrieval as compared to encoding. This is indeed the case as the results reported in Sections 3.2.4 and 3.2.5 indicate. Considering the nature of the task, we have to expect a large theta ERS during the presentation of the memory set which should increase with episodic STM demands. Al- 500 W. Klimesch et al.r CognitiÕe Brain Research 7 (1999) 493–501 though theta ERS was obtained, the strength of this effect was much smaller than expected. One possible reason may be due to the fact that in response to the warning signal, an extremely strong theta synchronization was obtained Žsee Fig. 1a.. As we now know from a recent study by Doppelmayr et al. w3x, large theta power preceding the presentation of the relevant stimulus leads to a suppression of theta synchronization. Event-related band power measures such as ERD reflect non phase locked as well as phase locked Žor ‘evoked’. EEG activity. Thus, it remains an open question to what extent an increase in band power is influenced by certain components of event-related potentials ŽERP’s.. In order to differentiate between these two types of EEG activity, a new measure Žtermed induced band power or IBP. was developed which reflects band power changes that do not contain phase locked EEG activity w12x. Results from our laboratory indicate that on the one hand, theta ERS is significantly influenced by evoked activity but that on the other hand a highly significant induced band power component ŽIBP. remains w12,13x. Thus, it is unlikely that the strong theta ERS which was found in response to the warning signal Žcf. Fig. 1a,b. represents only evoked activity. For the alpha frequency range, we have found that evoked EEG activity is practically absent w12,13x. The additional fact that the upper alpha ERS—obtained in this study—can be observed for a period of more than 3000 ms Žcf. Fig. 2a,b. makes it extremely unlikely that these results are influenced by ERP components. Lopes da Silva has found that alpha activity is generated Žor induced. in thalamo-cortical reentrant loops as well as in cortical networks Žw17x and the review in Ref. w7x.. More recent evidence for this view comes from studies using magnetoencephalography w6,26,27x. Theta activity on the other hand, probably is induced into cortical networks via hippocampo-cortical reentrant loops w7,18x. Support for this view is provided by an interesting study from Gevins et al. w4x, who used a new method to spatially sharpen the EEG with magnetic resonance imaging-based finite element deblurring. These authors found a frontal midline theta rhythm which increased with increasing memory load. Most interestingly, dipole models localized this signal to the region of the anterior cingulate cortex which is part of the Papez circuit and, thus, is linked with the hippocampal formation via complex reentrant loops. Because hippocampo-cortical reentrant loops are comparatively denser in more anterior regions, whereas the opposite holds true for thalamo-cortical loops, we would expect that theta ERS is generally larger at anterior and alpha ERD at posterior recording sites. This is in principle the case, if we accept the interpretation Ždiscussed above. that the strong theta ERS in response to the warning signal Žwhich is maximal at frontal sites. has reduced the extent of theta ERS during encoding Žcf. Fig. 1.. In addition, it is important to note that the increase in theta ERS—in response to task difficulty—is most consistent at frontal recording sites. On the other hand, upper alpha ERD Žwhich was found during retrieval Žcf. Fig. 2. is larger at occipital as compared to frontal regions. Finally, when proceeding from the proposed hypothesis that upper alpha ERS reflects the inhibition of semantic LTM, the strength of this effect should be larger over frontal regions which are known to be related to semantic memory processes w25x. In good agreement with this notion is the finding that upper alpha ERS is particularly large over more anterior recording sites, but practically absent at occipital regions. Acknowledgements This research was supported by the Austrian ‘Fonds zur Forderung der wissenschaftlichen Forschung’ P-13047. ¨ References w1x E. Basar, Towards a renaissance of ‘alphas’, Int. J. Psychophysiol. 26 Ž1997. 1–3. w2x A. Burgess, J.H. Gruzelier, Short duration synchronization of human theta rhythm during recognition memory, NeuroReport 8 Ž4. Ž1997. 1039–1042. w3x M. Doppelmayr, W. Klimesch, T. Pachinger, B. Ripper, The functional significance of absolute power with respect to event-related desynchronization, Brain Topogr., in press, 1998. w4x A. Gevins, M.E. Smith, L. McEvoy, D. Yu, High-resolution EEG mapping of cortical activation related to working memory: effects of task difficulty, type of processing, and practice, Cerebral Cortex 7 Ž1997. 374–385. w5x J.H. Gruzelier, New advances in EEG and cognition, Int. J. Psychophysiol. 24 Ž1997. 1–5. w6x L. Kaufman, S. Curtis, J.-Z. Wang, S.J. Williamson, Changes in cortical activity when subjects scan memory for tones, Electroenceph. Clin. Neurophysiol. 82 Ž1992. 266–284. w7x W. Klimesch, Memory processes, brain oscillations and EEG synchronization, Int. J. Psychophysiol. 24 Ž1996. 61–100. w8x W. Klimesch, H. Schimke, G. Pfurtscheller, Alpha frequency, cognitive load, and memory performance, Brain Topogr. 5 Ž1993. 241– 251. w9x W. Klimesch, H. Schimke, J. Schwaiger, Episodic and semantic memory: an analysis in the EEG-theta and alpha band, Electroencephalogr. Clin. Neurophysiol. 91 Ž1994. 428–441. w10x W. Klimesch, M. Doppelmayr, H. Russegger, T. Pachinger, Theta band power in the human scalp EEG and the encoding of new information, NeuroReport 7 Ž1996. 1235–1240. w11x W. Klimesch, M. Doppelmayr, T. Pachinger, H. Russegger, Eventrelated desynchronization in the alpha band and the processing of semantic information, Cogn. Brain Res. 6 Ž2. Ž1997. 83–94. w12x W. Klimesch, H. Russegger, M. Doppelmayr, T. Pachinger, Induced and evoked band power changes in an oddball task, Electroenceph. Clin. Neurophysiol. 108 Ž1998. 123–130. w13x W. Klimesch, M. Doppelmayr, H. Russegger, T. Pachinger, J. Schwaiger, Induced alpha band power changes in the human EEG and attention, Neurosc. Lett. 244 Ž1998. 73–76. w14x C.M. Krause, H.A. Lang, M. Laine, M.J. Kuusisto, B. Porn, ¨ Cortical Processing of vowels and tones as measured by event-related desynchronization, Brain Topogr. 8 Ž1. Ž1995. 47–56. w15x C.M. Krause, A.H. Lang, M. Laine, M. Kuusisto, B. Porn, ¨ Event-related EEG desynchronization and synchronization during an auditory W. Klimesch et al.r CognitiÕe Brain Research 7 (1999) 493–501 w16x w17x w18x w19x w20x w21x w22x memory task, Electroencephalogr. Clin. Neurophysiol. 98 Ž1996. 319–326. C.M. Krause, B. Porn, ¨ A.H. Lang, M. Laine, Relative alpha desynchonization and synchronization during speech perception, Cog. Brain Res. 5 Ž1997. 295–299. F.H. Lopes da Silva, T.H.M.T. Van Lierop, C.F. Schrijer, W. Storm van Leeuwen, Organization of thalamic and cortical alpha rhythms: spectra and coherences, Electroenceph. Clin. Neurophysiol. 35 Ž1973. 626–639. R. Miller, Cortico-Hippocampal Interplay and the Representation of Contexts in the Brain, Springer, Berlin, 1991 G. Pfurtscheller, Event-related synchronization ŽERS.: an electrophysiological correlate of cortical areas at rest, Electroencephalogr. Clin. Neurophysiol. 83 Ž1992. 62–69. G. Pfurtscheller, A. Aranibar, Event-related cortical desynchronization detected by power measurements of scalp EEG, Electroencephalogr. Clin. Neurophysiol. 42 Ž1977. 817–826. G. Pfurtscheller, W. Klimesch, Topographical display and interpretation of event-related desynchronization during a visual–verbal task, Brain Topogr. 3 Ž1990. 85–93. G. Pfurtscheller, A. Stancak Jr., C. Neuper, Event-related synchronization ŽERS. in the alpha band—an electrophysiological correlate w23x w24x w25x w26x w27x 501 of cortical idling: a review, Int. J. Psychophysiol. 24 Ž1r2. Ž1996. 39–46. W. Schneider, R.M. Shiffrin, Controlled and automatic human information processing: I. Detection, search and attention, Psychol. Rev. 84 Ž1977. 1–66. M.B. Sterman, D.A. Kaiser, B. Veigel, Spectral analysis of event-related EEG responses during short-term memory performance, Brain Topogr. 9 Ž1. Ž1996. 21–30. E. Tulving, S. Kapur, F.I. Craik, M. Moscovitch, S. Houle, Hemispheric encodingrretrieval asymmetry in episodic memory: positron emission tomography findings, Proc. Natl. Acad. Sci. USA 91 Ž1994. 2016–2020. S.J. Williamson, L. Kaufman, S. Curtis, Z.-L. Lu, C.M. Michel, J.-Z. Wang, Neural substrates of working memories are revealed magnetically by the local suppression of alpha rhythm, In: I. Hashimoto, Y.C. Okada, S. Ogawa ŽEds.., Visualization of Information Processing in the Human Brain: Recent Advances in MEG and Functional MRI ŽEEG Suppl. 47., Elsevier, Amsterdam, 1996, pp. 163–180. S.J. Williamson, L. Kaufman, Z.L. Lu, J.Z. Wang, D. Karon, Study of human occipital alpha rhythm: the alphon hypothesis and alpha suppression, Int. J. Psychophysiol. 26 Ž1997. 63–76.