Sacconi and Grimalda (2002, 2005a, 2005b) introduced a model in which two basic motives to action are understood as different type of preferences and represented by a comprehensive utility function: the first is consequentialist motivation, whereas the second is a conditional willingness to conform with an ideal, or a moral principle, which they call a conformist, or ideal, motive to action. A moral ideal is meant as a normative principle of evaluation for collective modes of behaviours which provides agents with a ranking of states of affairs resulting from strategic interaction expressing a greater or lesser consistency with the ideal. The principle moreover is seen as resulting from a (possibly hypothetical) contract between the agents involved in the interaction in an ex-ante phase. Thus, the normative principle boils down to a social welfare function that measures the consistency of outcomes with the normative prescriptions provided by the ideal. Hence, agents understand their own and any other agent's degree of conformity in terms of their contribution to carrying out the ideal given the others' expected action, and a person's own motivation to act in conformity with the principle increases with others' (expected) conformity. In other words, individual conformity with the principle is conditional on others' conformity with it, as perceived by the agent. This peculiar feature of reciprocity over others' behaviour calls for an extension of the usual equipment of decision theory, which is provided by the theory of Psychological Games (Geanakoplos et al., 1989). In this paper we design an experiment for preliminary exploration of the empirical validity of the conformist preferences model, applying it to a simple non cooperative game (the Exclusion Game) meant as the problem of dividing a sum between two active players and a third, dummy player (passive beneficiary). Results are encouraging. Behaviours dramatically change passing from the simple exclusion game to a three steps game, in which once the players have first played the typical non cooperative exclusion game, and before playing it again, they participate in a middle phase, where they anonymously agree on a principle of division. Having agreed on a principle, even though this agreement does not implies reputation effects nor is externally enforceable, induces a substantial part of players - who acted selfishly in the first step - to conform to the principle in the third phase. The additional condition being that they believe the other players will also conform to the agreed principle (what here does happen, as a mater of fact). These results strictly accord with the prediction of the conformist preferences model, but cannot be accounted for by alternative theories of reciprocity
