This study was aimed at investigating functional and neuropsychological dissociations between repetition priming and explicit memory tasks. The explicit and implicit versions of the stem completion task were administered to a group of amnesics and a group of control subjects. In Experiment 1 both the explicit and implicit stem completions were significantly higher when the same presentation modality was used for studying and testing than when a change in modality from studying to testing occurred. Amnesics had normal implicit and deficient explicit completion performance. Experiment 2 revealed an advantage of the semantic over the phonological condition only in the explicit task and only in control subjects. Amnesic patients completed the same percentage of words as normal subjects in the phonological and semantic conditions of the implicit task and in the phonological condition of the explicit task but were deficient in intentionally completing semantically processed words. Possible interpretations of these results are discussed according to theoretical models that distinguish memory tasks along an explicit-implicit dichotomy (multiple memory system theory) or along a perceptual-conceptual dichotomy (transfer-appropriate procedures approach), and alternative theoretical positions are evaluated regarding repetition priming and memory deficits in amnesic patients.