Wolfgang Klimesch (1994) The Ignition of Cortical Cell Assemblies: Some Arguments Against. Psycoloquy: 5(58) Brain Rhythms (4)

Volume: 5 (next, prev) Issue: 58 (next, prev) Article: 4 (next prev first) Alternate versions: ASCII Summary
PSYCOLOQUY (ISSN 1055-0143) is sponsored by the American Psychological Association (APA).
Psycoloquy 5(58): The Ignition of Cortical Cell Assemblies: Some Arguments Against

Commentary on Pulvermueller et al. on Brain-Rhythms

Wolfgang Klimesch
University of Salzburg
Department of Psychophysiology
A-5020 Salzburg



It is argued that an increase in gamma band power in response to the ignition of a relevant cell assembly can be expected only if the activation of irrelevant cell assemblies can be selectively suppressed or inhibited. Analyses of EEG band power in other frequency bands, such as the lower, upper alpha and beta band, however, have shown that band power decreases as cognitive demands increase.


brain theory, cell assembly, cognition, event related potentials (ERP), electroencephalograph (EEG), gamma band, Hebb, language, lexical processing, magnetoencephalography (MEG), psychophysiology, periodicity, power spectral analysis, synchrony
1. In their target article, Pulvermueller et al. (1994) question whether gamma-band activity is an indicator of cognitive processes. They base their hypotheses on results and theories of Braitenberg and his group as well as on the Frankfurt MPI group of Gray, Singer, Engel and coworkers. The central idea here is that cortical cell assemblies exhibit a synchronous oscillatory discharge pattern in the gamma band during a state of activation. Pulvermueller et al. call this state "ignition" and predict an increase in gamma band power in response to cognitive processes that lead to the activation of relevant cell assemblies. In two experiments they test the hypothesis whether the presentation of words as compared to pseudowords is indeed correlated with an increase in gamma band power. The results show that during a post-stimulus period of about 400 ms which is related to a strong N400, and during which one may assume semantic encoding of words, gamma band power does not increase for words but is significantly lower for pseudowords as compared to words. The latter result could be replicated by using the MEG instead of the EEG.

2. Whereas the data of both reported experiments are interesting in themselves, their interpretation within the context of the "cell assembly theory" leads to some critical questions. One of these concerns the meaning of the term "ignition". It is particularly interesting to see that with respect to this question neural and cognitive theories come very close to each other. "Ignition of a cell assembly" can easily be translated into "activation of a code in semantic long-term memory". The mathematical analysis of spreading activation theories such as Anderson's ACT or ACT* (Anderson, 1976, 1983) have shown very clearly that special mechanisms to delimit the spreading activation must be assumed. Otherwise it is very difficult to explain why only the relevant parts of a complex memory network are activated during a search or activation process. Basically, two types of assumptions help to explain why the spreading activation process is confined to the relevant parts of the network:

3. One type of assumption is based on active inhibition. Traditional spreading activation theories such as ACT or ACT* use a dampening function to prevent an activation process from spreading ultimately into the entire storage network. In the framework of a neurobiological approach, a dampening function can be understood to reflect strong inhibitory processes that weaken an activation process after the relevant information is activated. The problem with assuming inhibitory processes is that as a consequence of this assumption, regulatory mechanisms must be assumed that control the level of activation in particular parts of the network (cf. paragraph 1 of the target article). Thus, the additional question arises of how regulatory mechanisms that guide inhibitory processes are capable of distinguishing between the relevant and irrelevant part of the network. Because the assumption of a distributed storage is widely accepted and because inhibitory processes would have to operate in very particular parts of the network in order to be effective, the idea of a distributed storage seems to be threatened.

4. A second type of assumption is based on the idea that the activation process for relevant information outruns the activation processes spreading into irrelevant parts of the network (Klimesch, 1994). It can be shown that the explanation of this type of self- organizing spreading activation process does not require inhibitory processes (see chapters 8, 9 and 10 in Klimesch, 1994) and rests on assumptions that conform well with current neurobiological knowledge (cf. chapter 11).

5. Now, the interesting point is that gamma power as the presumed indicator for the activation of (the relevant) cortical cell assemblies should increase sharply during the semantic encoding of a word, provided that irrelevant background activity is suppressed (e.g., by means of inhibition). However, when assuming a type of spreading activation process like the above, one that does not require the suppression of background activity, there is no obvious need to predict such an increase in gamma band power because cognitive processes as a constantly ongoing activity (see paragraph 26 of the target article) will be reflected by a certain level of gamma band power that does not sharply change with time. As the results of both experiments indicate, there was no increase in gamma band power during the semantic encoding of words. In summarizing my arguments so far, I see a need to postulate inhibitory processes if one proceeds from the idea that gamma band power reflects the ignition or activation of a few relevant cell assemblies. Furthermore, only if one were to assume that inhibition works to suppress background activity is it plausible to predict an increase in gamma band power during the semantic encoding of words.

6. Finally, I want to draw attention to the result that during the encoding of pseudowords, gamma band power is suppressed. This effect reminds one of the typical findings based on the method of event-related desynchronization or ERD (Pfurtscheller & Aranibar, 1977; Pfurtscheller & Klimesch, 1992). Pfurtscheller's findings as well as those from our own laboratory (e.g., Klimesch et al. 1990a; 1990b; 1992; 1993) have consistently shown that during a variety of cognitive tasks power values within the alpha and beta band decrease. It was also found that the extent of band power suppression is related to task difficulty. It is hence tempting to apply a similar interpretation to the findings reported by Pulvermueller et al. Encoding a pseudoword may require more semantic processing (as is also reflected by the pronounced N400) and as a result, gamma activity becomes more desynchronized, which is reflected by a suppression of gamma band power. If one adopted the interpretation underlying ERD, changes in gamma band power could hardly be interpreted in terms of the ignition or activation of a few cell assemblies. Instead, we would have to assume that comparatively large cortical areas become desynchronized. The same conclusion can be derived from theoretical considerations, as outlined above: cortical activation processes that do not require the suppression of background activity lead to the desynchronization of comparatively large cortical areas and thus to a suppression of band power. Both arguments speak against the assumption that the activation or ignition of a few relevant cell assemblies can be detected by an increase in gamma band power.


Anderson, J.R. (1976). Language, memory, and thought. Hillsdale, NJ: Lawrence Erlbaum Associates.

Anderson, J.R. (1983). A spreading activation theory of memory. Journal of Verbal Learning and Verbal Behavior, 22:261-295.

Klimesch, W., Pfurtscheller, G., Mohl, W. & Schimke, H. (1990a). Event-related desynchronization, ERD-mapping and hemispheric differences for words and numbers. International Journal of Psychophysiology, 8:279-308.

Klimesch, W., Schimke, H., Ladurner, G. & Pfurtscheller, G. (1990a). Alpha frequency and memory performance. Journal of Psychophysiology, 4:381-390.

Klimesch, W., Pfurtscheller, G. & Schimke, H. (1992). Pre- and post-stimulus processes in category judgement tasks as measured by event-related desynchronization (ERD). Journal of Psychophysiology, 6:186-203.

Klimesch, W., Schimke, H. & Pfurtscheller, G. (1993). Alpha frequency, cognitive load and memory performance. Brain Topography, 5:1-11.

Klimesch, W. (1994). The structure of long-term memory: A connectivity model of semantic processing. Hillsdale, NJ, Lawrence Erlbaum.

Pfurtscheller, G. & Aranibar, A. (1977). Event-related cortical desynchronization detected by power measurements of scalp EEG. Electroencephalography and Clinical Neurophysiology, 42:817-826.

Pfurtscheller, G. & Klimesch, W. (1992). Functional topography during a visuoverbal judgment task studied with event-related desynchronization mapping. Journal of Clinical Neurophysiology, 9:120-131.

Pulvermueller, F., Preissl, H., Eulitz, C., Pantev, C., Lutzenberger, W., Elbert, T. and Birbaumer, N. (1994). Brain Rhythms, Cell Assemblies and Cognition: Evidence from the Processing of Words and Pseudowords. PSYCOLOQUY 5(48) brain-rhythms.1.pulvermueller.

Volume: 5 (next, prev) Issue: 58 (next, prev) Article: 4 (next prev first) Alternate versions: ASCII Summary