The paper is concerned with the interaction between two agents: an expert, announcing his probability that a particular state of the world will occur, and a non-expert decision-maker, who takes action according to his posterior beliefs. The decision-maker considers the expert an experiment of uncertain reliability and takes the received messages as the outcomes of such an experiment. The model of the expert in the decision-maker’s mind bears no relation with any measure of the expert’s actual information. The paper shows that messages will be biased, notwithstanding solidarity between the agents. However, the longer the interaction, the less severe will be the bias.