In recent years researches in the field of cognitive psychology have favored an interpretation of moral behavior primarily as the product of basic, automatic and unconscious cognitive mechanisms for the processing of information, rather than of some form of principled reasoning. This paper aims at undermining this view and to sustain the old-fashioned thesis according to which moral judgments are produced by specific forms of reasoning. As critical reference our research specifically addresses the so called Rawlsian model which hinges on the idea that human beings produce their moral judgments on the basis of a moral modular faculty "that enables each individual to unconsciously and automatically evaluate a limitless variety of actions in terms of principles that dictate what is permissible, obligatory, or forbidden".[25, p. 36] In this regard we try to show that this model is not able to account for the moral behavior of different social groups and different individuals in critical situations, when their own moral judgment disagrees with the moral position of their community. Furthermore, the critical consideration of the Rawlsian model constitutes the theoretical basis for the constructive part of our argument, which consists of a proposal about how to develop a semantic, quasi-rationalistic model to describe moral reasoning. This model aims to account for both moral reasoning and the corresponding emotions on the basis of the information which morally relevant concepts consist of.