In demonstrating the presence of cell assemblies for word recognition, Pulvermueller et al. (1994) have taken a large step forward. However, because of the extreme simplicity of their examples, they have avoided dealing with the central issue of the cell assembly theory: the question of knowledge representation.
2. Pulvermueller et al. (1994) propose a new approach to identify cell assemblies in the brain: they look for rapid, correlated periodic activity in large numbers of neurons. This is a simple yet powerful methodology which can be expected to yield interesting results beyond the domain of word meanings in which they have applied it. However, the restriction to periodic activity is a serious limitation, which the authors do not appear to recognize. Periodic activity implies correlated, assembly-like behavior; but the converse is not true. In the language of dynamical systems theory (Devaney, 1988), periodicity is only one of the simplest kinds of attracting behaviors. In general, one can expect cell assemblies to demonstrate all sorts of "strange attractor" behaviors, some of which may, like the famous Lorenz attractor, look vaguely periodic, and some of which may not look periodic at all.
3. In other words, I believe Pulvermueller et al. overstate the logical relationship between: (1) the cell assembly theory of cognition, and (2) the presence of periodicity in cell assemblies. A statement like "assembly ignition implies fast periodic and correlated activity of a large number of assembly neurons" (par. 30) is really not justified. The word "implies" here is much too strong; all that really exists is a suggestion. The arguments of paragraphs 8 and 9, which argue for periodic behavior in all cell assemblies, are merely heuristic. The paper of Schuster and Wagner (1990), which is used to bolster this conclusion, is based on a mean field approximation and makes many auxiliary simplifying assumptions; it is more suggestive than conclusive. So, where correlated periodicity is found, the cell assembly theory is confirmed; but where it is not found, the cell assembly theory is not disconfirmed.
4. This may seem to be a quibble, but I would argue that it is more than that. To borrow a term from artificial intelligence, the question is one of knowledge representation. Symbolic AI programs represent knowledge explicitly and locally; connectionist AI programs represent knowledge implicitly and in a distributed fashion. Cell assembly models are neither thoroughly connectionist nor thoroughly symbolic; they are modular connectionist architectures, the design of which is motivated by symbolic issues. The question thus arises: how, in a cell assembly network, is knowledge stored? Pulvermueller et al. advocate a localized model of cell assembly memory, in which each word (more generally, each item stored in memory) gets its own assembly. But this is not the only possibility. For example, Pribram's (1991) holonomic model of memory is likewise based on cell assemblies, but it stores items in an almost completely nonlocal way. And in Goertzel (1994), it is proposed that some knowledge inheres in the structure of the strange attractors of cell assemblies.
5. Empirical study of cell assemblies is absolutely essential to the development of cognitive science. Pulvermueller et al. are to be congratulated for taking a first step in this direction. However, one must guard against the urge to identify the cell assembly theory with its simplest, most easily testable manifestations.
Devaney, R. (1988) Chaotic Dynamical Systems. New York: Addison-Wesley.
Goertzel, B. (1994) Chaotic Logic: Language, Thought and Reality from the Perspective of Complex Systems Science. New York: Plenum.
Hebb, D.O. (1949) The organization of behavior. A neuropsychological theory. New York: John Wiley.
Pribram, K. (1991) The Holonomic Brain. New York: Elsevier.
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.
Schuster, H.G. & Wagner, P. (1990) A model for neuronal oscillations in the visual cortex: 1. Mean-field theory and derivation of the phase equations. Biological Cybernetics 64:77-82.