The sufficiency of behavioral data supporting reasoning biases was challenged. Our objective was to test if physiological data could significantly support reasoning biases. Experiment 1: When performing a rule discovery task with feedback, participants sytematically gave the same response. This was not compatible with norms referring to a formal logic system, but was insufficient to conclude there was a reasoning bias. Only the knowledge of participants' expectations of feedback would merit this conclusion. Experiment 2: Participants' expectations were indexed by electrodermal activity. The results showed that when participants displayed the behavioral bias, they considered their responses to be correct and hence committed an error of logic. However, this error of logic did not prevent them from solving the task. This argues for a theory according to which human reasoning uses more economical strategies than the simple application of formal logical rules.