A recent study (Brady et al., 2008) claims that the capacity and fidelity of long-term visual recognition memory has been underestimated. After viewing a massive number of images for nearly six hours participants were highly accurate at identifying a previously seen image in a two-alternative forced choice task even when the foil was extremely similar to the target. In present study we hypothesised that this impressive memory performance might be specific to the test format. To investigate the effect of test format on recognition accuracy we showed participants 1700 images for 2.5 hours and then tested them in a forced choice task and a yes/no task, where participants judged only one image at a time. We found that accuracy was relatively high in both test conditions; however, the performance was lower in the yes/no task (75% vs. 86%). A follow-up study exploring delayed testing effects will also be described.