In our laboratory experiment, subjects, in sequence, have to predict the value of a good. We elicit the second subjects belief twice: first (first belief), after he observes his predecessors action; second (posterior belief), after he observes his private signal. Our main result is that the second subjects weigh the private signal as a Bayesian agent would do when the signal confirms their first belief; they overweight the signal when it contradicts their first belief. This way of updating, incompatible with Bayesianism, can be explained by multiple priors on the predecessors rationality and a generalization of the Maximum Likelihood Updating rule. In another experiment, we directly test this theory and find support for it.