Cell Assemblies as Building Blocks of Larger Cognitive Structures
J. Eric Ivancich
Department of Electrical Engineering and Computer Science, The University of Mic
higan, Ann Arbor, MI 48109, ivancich@eecs.umich.edu
Christian R. Huyck
The University of Sheffield and Department of Computing Sciences, Middlesex Univ
ersity, London, N11 2NQ, chris@umich.edu
Stephen Kaplan
Departments of Psychology and of Electrical Engineering and Computer Science, Th
e University of Michigan, Ann Arbor, MI 48109-1109, skap@umich.edu
Abstract
Pulvermüller's work in pushing Hebb's theory into the realm of language is excit
ing. However, we feel that what he characterizes as a single cell assembly is ac
tually a set of cooperating cell assemblies that form parts of larger cognitive
structures. The
se larger structures more easily account for a variety of phenomena, including t
he psycholinguistic.
Commentary
Pulvermüller is to be congratulated for his accomplishment in pushing Hebb's cel
l assembly theory into the realm of language and for showing that meaningful and
provable predictions can be made with respect to measurable neural activity. Wi
thout undermini
ng his accomplishments, however, we do feel that adjustments in both the termino
logy and the conceptualization of his work reveals a more powerful and flexible
cognitive system.
In 1949, Hebb, in his landmark book The Organization of Behavior, develo
ped the concept of the cell assembly. In his original conception, a cell
assembly consisted of a number of neurons that responded to a set of similar sti
muli and that c
ould sustain activity for about 500 milliseconds (Hebb 1949).
Cell assemblies, while powerful themselves, are also ideal building blocks for
larger cognitive structures (Kaplan et al. 1990; Holland 1998). The recombinatio
n of subsets of building blocks into larger structures yields a vast number of p
otential combin
ations.
Just as a modified Hebbian synaptic learning rule can be used to form cell asse
mblies, the very same learning rule can be used to create an associative link be
tween two assemblies through a set of mediating connections. Thus, when one asse
mbly becomes ac
tive, another will likely become active due to this link.
When a number of cell assemblies become linked in this way, longer sequences ar
e encoded. When many distinct sequences pass through a given set of assemblies i
n such a way that each assembly can be re-used in multiple sequences, a more com
plex structure
evolves. This structure is similar to the cognitive map that Tolman propo
sed (Tolman 1948). This cell assembly interpretation of Tolman's cognitive map c
oncept has been proposed as a general purpose knowledge structure, useful not on
ly for storing
information used in navigating an environment (Kaplan and Kaplan 1982; Levenick
1991; Chown et al. 1995) but also for encoding algorithmic and story structures
of many kinds. It appears to be equally suitable for the generation and parsing
of language.
Hebb focused on cognitive structures with direct sensory and motor content. Sin
ce a neuron is more likely to be connected to those nearby, a notion of neural d
istance exists. Some neurons are close to the sensory and motor apparatuses, whi
le others are f
urther away. Neurons that are close to the sensory interface are well-placed to
become members of cell assemblies with concrete, sensory content. These assembli
es then become the inputs for neurons further away from the sensory interface, w
here assemblies
with diminished sensory content will form.
From the simple notion of distance emerges both the concept of depth and a hier
archical structure. In such a hierarchy depth equates to abstractness, moving fr
om basic-level categories (e.g., tree) (Rosch et al. 1976) to those so ab
stract they have
little or no sensory content (e.g., justice).
Both the cognitive map and the hierarchy are directly applicable to Pulvermülle
r's work in language. The pattern of neural activity generated by a noun, a verb
, or a grammatical function word that Pulvermüller uses in his argument is not a
result of what
we view as a single cell assembly. Rather we would characterize the pattern as
the result of the activation of a set of strongly associated assemblies.
The phonemic representation of a word in perisylvan cortex likely involves a se
quence of cell assemblies where each represents a phoneme. The same phonemes in
a different order would, after all, generate a different word (e.g., "cat" and "
tack"). As the
sequence of phonemes is recognized, activity may then collect in another assembl
y sitting at the next level of a hierarchy and representing the whole word. Note
that this hierarchical relationship is based on the grouping of a sequence. A c
ategorical hier
archy is also available so that "cat" may be an animate noun, or more traditiona
lly, an animal.
The phoneme-based word cell assembly has strong associations to potentially man
y semantic correlates. By keeping the phonemic and semantic assemblies separate,
yet highly associated, the system can readily deal with homonyms by using conte
xtual cues to r
esolve the competing interpretations.
A set of separate cell assemblies can account for a number of other phenomena a
s well. Many concepts simply do not have simple verbal correlates. People can le
arn and represent nonsense words. Since one can have a word without a concept, o
r a concept wit
hout a word, the assemblies are separate.
If we might be a bit more speculative, we conceptualize a set of world maps<
/i> composed of cell assemblies rich in semantic content, and a set of word m
aps composed of phoneme-based cell assemblies. These two maps are highly ass
ociated and tra
versal between them is often easy and consciously seamless.
By keeping the maps separate, the system has a more powerful and flexible struc
ture. There are links within each map as well as links between the maps. Among o
ther things, the word maps encode the likelihood of connections between phonemes
. If two phonem
es never connect in the language, a word in which they do connect will seem unna
tural (which can explain the "phonological rules of the language" that Pulvermül
ler [section 4.2, 7th paragraph] mentions). In word maps, words can be combined
to form multi-w
ord units whose semantic content is unrelated to the base words, as in "The Big
Apple", "kick the bucket", and other idiomatic expressions. It is difficult to e
xplain how this could be the case if the maps were not separate.
Finally, hierarchical and map structures go a long way towards explaining how a
system as complex as language can be learned in the first place. Preverbal chil
dren obviously have sophisticated representations of the world around them, and
children underg
o an explosive growth in vocabulary. A high school graduate knows 60,000 words (
Pinker 1994) and likely knows most of them by age five. It is difficult to accou
nt for this unless children are, at least in part, building larger structures fr
om already-lear
ned building blocks, both semantic and phonemic.
Pulvermüller's findings are indeed encouraging. We believe it would be worthwhi
le to explore the application of his methods towards these more complex cognitiv
e structures, opening up fascinating possibilities for the explanation of many c
omplex psycholo
gical and psycholinguistic phenomena.
Bibliography
Chown, E., Kaplan, S., and Kortenkamp, D. (1995) Prototypes, location, and asso
ciative networks (PLAN): Toward a unified theory of cognitive mapping. Cognitiv
e Science 19:1-51.
Hebb, D.O. (1949) The organization of behavior. John Wiley and Sons, New York.
Holland, J.H. (1998) Emergence: From chaos to order. Addison-Wesley, Reading,
MA.
Kaplan, S. and Kaplan, R. (1982) Cognition and environment: Functioning in an
Uncertain World.
Praeger, New York. Republished by Ulrich's, Ann Arbor, MI, 1989.
Kaplan, S., Weaver, M., and French, R. (1990) Active symbols and internal mode
ls: Towards a cognitive connectionism. AI and Society 4:51-71.
Levenick, J. (1991) NAPS: A connectionist implementation of cognitive maps. Con
nection Science 3(2):107-126.
Pinker, S. (1994) The language instinct. Penguin Books, London.
Rosch, E., Mervis, C., Gray, W., Johnson, D., and Boyes-Braem, P. (1976) Basic
objects in natural categories. Cognitive Psychology 8:382-439.
Tolman, E.C. (1948) Cognitive maps in rats and men. Psychological Review, 55:1
89-203.
6
BBS Comment Ivancich, Huyck, Kaplan