An important feature of human intelligence is the use of symbols. This is s
een in our daily use of language and logical thinking. However, the use of
symbols is not limited to humans. We observe planned action sequences in pr
imate behavior and prediction-based action in higher mammals. For the repre
sentation and operation of symbols by the brain neural circuit, no specific
construction principle or computational theory is known so fa. In this pap
er, we regard the brain as a complex of associative memory and dynamic atte
ntional system, and starting from two hypotheses on information representat
ion and operation in the brain, we propose a model of primitive symbolic be
havior emergence that is consistent with the conventional symbolic processi
ng model. We also describe a computational theory of the symbolic processin
g model in associative memory. Through computer simulation studies on a lan
guage-like memory search and map learning by a moving robot, we discuss the
validity of the model. (C) 1999 Elsevier Science Ltd. All rights reserved.