Recent studies exploring benzodiazepine memory effects have used the distinction between explicit and implicit tasks. There is now increasing evidence that implicit tasks can be "contaminated" by conscious uses of memory and that unconscious (automatic) use of memory can contaminate explicit tasks, leading to mistaken estimates of their respective influences on memory performance. The aim of the present double-blind, double-placebo study was to assess the memory effects of diazepam and lorazepam using a process-dissociation procedure in a stem-completion task, this procedure providing uncontaminated estimates of conscious and automatic memory processes. The memory task was administrated to 60 healthy volunteers randomly assigned to one of three parallel groups (placebo, diazepam 0.3 mg/kg, lorazepam 0.038 mg/kg). Lorazepam markedly reduced conscious as well as automatic influences of memory. Diazepam also reduced conscious uses of memory, albeit to a lesser extent than lorazepam, but did not decrease the influence of automatic memory. Secondary analyses showed that when the deleterious effect on conscious uses of memory was equated between a diazepam subgroup and the lorazepam group, only lorazepam impaired the automatic use of memory. This study strongly suggests a qualitative difference in the memory effects of the two benzodiazepines. It has some implications regarding the relationships between states of consciousness and memory processes.