People often think and act non-rationally when they are required to solve simple reasoning problems. The errors of reasoning depend not only on the form of an inference but also on its content. In the classic Wason four-card test, for example, individuals are more likely to confuse the logical schema of the Modus Tollens with the Fallacy of the affirmation of the consequent when the inference is presented in an abstract form, than when it is endowed with content. Two further factors play a fundamental role: the error of reasoning seems to occur less frequently when the conditional premise has a deontic form and when the content of the conditional is relevant in the context. In this paper we provide experimental evidence showing that, by manipulating the level of relevance between the antecedent and the consequent of a deontic conditional sentence, the expectation of relevance created in the participants elicits different responses in the four-card test. Data collected show that, if the level of relevance between antecedent and consequent of a conditional is low or null, the subjects are led to make the systematic error of reasoning more frequently than when the level of relevance is high.