Non-Bayesian updating in a social learning experiment

In our laboratory experiment, subjects, in sequence, have to predict the value of a good. The second subject sequence makes his prediction twice: first (“first belief”), after he observes predecessor's prediction; (“posterior private signal. We find that subjects weigh their signal as Bayesian agent would do when confirms belief; they overweight it contradicts belief. This way updating, incompatible with Bayesianism, can be explained by Likelihood Ratio Test Updating (LRTU) model, generalization Maximum rule. It is at odds another family Full Updating. we directly test LRTU model and support for it.