The main purpose of this work is to understand some limitations introduced by the classical definitions of cellular automata (CA). To this end, we have defined a new model of CAs ( fuzzy CAs) which allows the observation of interesting “chaotic” properties of elementary CAs. To date neither a formal nor a precise definition of “chaos” in CAs exists; we believe that the proposed model provides a “sharper” tool to detect which properties can be associated to a “chaotic” behavior. We also define a measure ( rule entropy) which gives information about the CA's dynamics solely on the basis of the rule table and provides theoretical explanations to some of the empirical observations.