We present a model of rule-changing cellular automata (CA) to study dy
namical structure of CA in detail. In our model calculation the edge o
f chaos is observed on the basis of the source rule 22 which yields th
e most interesting behavior. We find that the edge of chaos is generat
ed through competition between two rule-changing behaviors. These rule
-changing behaviors are analyzed quantitatively by using the mutual in
formation as a measure of complexity. It is noted that the mutual info
rmation has the maximum value near the edge of chaos because of compet
ition. Copyright (C) 1998 Elsevier Science B.V.