Six groups of high-school students were exposed to a second-order matching-to-sample task and generalization tests trials using familiar and unfamiliar stimuli as well as a new matching relation. For two groups correct and incorrect matching responses produced the correspondently feedback according to continuous and intermittent schedules, respectively. Correct responses produced feedback and incorrect responses produced blanks and vice versa for other two groups, respectively. Two additional groups were exposed to similar feedback-blanks combinations but participants were instructed about the “meaning” of blanks before training. Extra-relational generalized matching-to-sample performance with either familiar or unfamiliar stimuli was observed after training conditions in which intermittent right-wrong feedback was scheduled, as well as when incorrect matching responses produced blanks and correct responses produced the correspondently feedback.Instructions about the meaning of blanks produced generalized performances slightly higher to those observed after continuous right-wrong feedback, which in turn were similar to performances observed after the uninstructed right-blank feedback combination condition. Results confirm an initial tendency to treat blanks as if they mean right and suggest a common“detachment” processes between intermittent feedback and the wrong-blanks feedback combination.