Two experiments were conducted to determine whether participants have source memory for test stimuli that they cannot identify. Using a paradigm developed to investigate the phenomenon of "recognition without identification" (Peynircioglu, 1990), we found that even when participants could not identify a previously studied item, they nonetheless exhibited above-chance performance on a source discrimination task. Most surprising was that source accuracy for unidentified items was independent of old-new discrimination and not different from that of identified items. These results are interpreted as evidence that source memory is based on a continuous, as opposed to a threshold-like, process and suggest that recollection may, in some circumstances, contribute to the phenomenon of recognition without identification. (Contains 4 tables, 1 figure and 4 footnotes.)