The notion of 'accelerated long-term forgetting' has often been attributed to disrupted 'late' memory consolidation. Nevertheless, methodological issues in the literature have left this theory unproven, leading some to suggest such findings may be reflective of subtle acquisition or early retention deficits. This study attempts to address such issues, and also to explore which pathophysiological variables are associated with forgetting rates. Eighteen participants with temporal lobe epilepsy (TLE) and eighteen matched controls completed background neuropsychological measurement of immediate and short-delay memory that showed comparable performance, both on verbal and visual tests. Using two novel experimental tasks to measure long-term forgetting, cued recall of verbal and visuospatial material was tested 30 seconds, 10 minutes, one day, and one week after learning. Forgetting of verbal material was found to be progressively faster during the course of a week in the TLE group. For visuospatial memory, participants in the TLE group exhibited faster early forgetting in the first 10 minutes after learning, as indicated by planned comparisons, with comparable forgetting rates thereafter. Our findings provide evidence for two patterns of disruption to 'early' memory consolidation in this population, occurring either at the initial delay only or continuing progressively through time. Differences in how soon after learning accelerated forgetting was detectable were related to factors associated with greater severity of epilepsy, such as presence of medial temporal lobe sclerosis on MRI and use of multiple anti-epileptic agents.