The article deals with evolution of rationality assumption as component of the model of man in economic science. In classical political economy this assumption didn't play the key role separating the scope of economic science - this role was performed by the assumption of self-interest. The rationality assumption became important as a result of the Marginal revolution and began to be an object of special research only in the context of expected utility maximization hypothesis of von Neuman and Morgenstern. These developments made it remote from everyday human behavior. In order to make economics a bit more realistic various strategies are used which are briefly surveyed. Within the constant rationality approach we could choose a model of a representative agent characterized by full or bounded rationality. Within the variable rationality approach an additional economic agent may be introduced who will be responsible for observed deviations from rational behaviour, while all the other agents could be described by the familiar rational model. The second variant consists in splitting the economic agent into different subpersonalities, having different, and moreover, conflicting goals. Still another option is to make rationality not only variable but also continuous. In this case it is possible to reach the optimal rationality which differs from the maximum one.