The aim of this study was to evaluate the discriminant validity of the RI-48 test, a shorter French version of the Category Cued Recall portion of the Double Memory Test developed initially by Buschke and colleagues (1997), in the diagnosis of mild and very mild Alzheimer disease (AD). The distinctive feature of the RI-48 task is that encoding specificity was increased by adding an immediate cued recall stage at the encoding phase. The results show that the RI-48 task seems to be well adapted to the clinical context and to have good psychometric properties, in particular a lack of a ceiling effect. Moreover, this task appears to be especially well suited for the diagnosis of both mild and very mild AD (sensitivity of 93% and 83.8%). From a more theoretical point of view, this study confirms the importance of optimizing the encoding specificity for the diagnosis of very mild AD, since the more encoding specificity is accentuated, the more discriminating power is increased for the diagnosis of very mild AD.