NeMO, a Noisy Exemplar Model of recognition memory assumes that two sources of interstimulus similarity combine to predict recognition judgments: 1) the summed similarity between the probe item and each of the list items, and 2) the similarity among all of the list items (Kahana & Sekuler, 2002). Whereas NeMO was originally developed to account for episodic recognition of compound sinusoidal gratings, we now extend the model to account for recognition of higher-order stimuli, realistic synthetic human faces, which were arrayed along orthogonal axes in a metric space.

In each of four experiments, a series of briefly-presented study faces was followed by a probe face, and subjects judged whether the probe face appeared in the study series. Replicating previous findings with compound gratings, NeMO confirmed that both summed similarity, and similarities among study items influenced recognition judgment. Multidimensional scaling (MDS) revealed that perceptual similarity preserved some, but not all of the metric structure present in the physical stimuli. NeMO provided a better fit when inter-item similarity was defined in MDS rather than physical space.