Recent studies have suggested that the benzodiazepine (BZ) lorazepam (LZ) differs from other BZs in its impairing effects on implicit memory tasks. The present study was designed to assess whether this atypical effect withstood the experimental rigour of Schacter's retrieval intentionality criterion and further, whether it could be reversed by the BZ antagonist, flumazenil (FL). The separate and combined effects of LZ, FL and placebo (PL) were assessed on indices of memory, sedation, and attention in 48 healthy volunteers. LZ disrupted performance on both explicit and implicit memory tasks, induced motor sedation and impaired focussed attention. Fl attenuated LZ-induced attentional deficits but did not affect motor sedation. FL also attenuated LZ-induced impairment on the implicit retrieval task. On the explicit retrieval task FL attenuated LZ-induced impairment for words which had been deeply processed at study but not words which had been shallowly processed. A subsequent recognition test showed LZ impaired recognition memory only when accompanied by recollective experience and flumazenil again attenuated this effect. FL itself lowered performance on several measures, reflecting intrinsic activity of this "antagonist". Assessment of the relationship between the mnestic and other effects of the drugs suggested that attentional effects contribute to, but do not explain, effects on implicit memory tasks. These results imply that the apparent atypical effects of LZ on implicit memory tasks are mediated by the same BZ receptor complex as mediates LZ's other effects.