Memory is sensitive to several aspects of its contents, including item similarity (Tulving, 1981) and value (Kachergis, Recchia, & Shiffrin, 2011). For the present study, we developed and scaled a new class of blob-shaped stimuli in order to manipulate item similarity. In two experiments, participants studied blob-point value pairs, with categories defined by valence (positive/negative point values) and by perceptual similarity of the stimuli. Using a dynamic 2AFC decision-making paradigm with a response deadline, we recorded response trajectories for different foil types (e.g., similar to the target, or unique). We find similarity effects in recognition memory and semi-paradoxical category size effects in accord with Tulving (1981). Interpretation of the accuracy effects is enhanced by examining the shape and timing of response trajectories, allowing us to measure phenomena such as decision reversals. Finally, we present a general Gaussian process regression framework to analyze such continuous response data (Cox, Kachergis, & Shiffrin, submitted).