The present study aimed to verify the functional relation between accurate self-reports and differential consequences by means of a reversal design. Four children with a history of learning acquisition difficulties performed, individually, computer activities which consisted, initially, in reading a word presented on the screen, and later, in selecting a green square to report correct reading responses and a red square to report incorrect reading responses. Three experimental conditions were implemented: Baseline (A), with no differential consequences for reports of correct or incorrect responses; Correspondence Training (B), in which accurate reports of hits and errors produced points; and Reinforcement of Hits Reports (C) in which only hit reports were followed by points. The Correspondence Training was effective in producing high levels of accurate reports of hits and errors for all participants. The reversal design allowed, with great experimental control, the training effects throughout the programmed reversions, confirming the correspondence training efficacy as a tool to produce correspondent reports.