Abstract
We used a novel psychophysical technique to determine whether a series of visual stimuli is remembered as a single prototype or as distinct exemplars. On each trial of our experiment, participants saw a series of three briefly-presented gratings, two Study items followed by a Probe item. Each grating comprised a vertical and a horizontal sinusoidal component; within a trial, the spatial frequency of the vertical component varied. Relying on memory of the Study items' vertical component, participants judged whether the Probe was the same as one of the Study items or differed from both. Individual discrimination thresholds of spatial frequency were used to scale the spatial frequency of the gratings. Between blocks of trials, the difference in spatial frequency between the two Studies assumed one of four values: 1, 2, 4, or 8 threshold units. Over trials, the Probe's spatial frequency varied relative to that of the Study items, thereby testing recognition over a range of perceptual distances between Probe and Study items. The results were summarized by “mnemometric functions,” which relate memory performance to inter-stimulus perceptual distance. With Study items separated by 4 or 8 threshold units, the mnemometric functions' bimodality, shape and peaks indicate that Study items were represented in memory as distinct, noisy exemplars that are centered on the spatial frequencies of the Study items. The mnemometric functions were well fit by a three-parameter, signal detection model, whose representation of the first Study item was twice as noisy as the representation of the second Study item. The model also shows that individual discrimination thresholds closely approximate decision criteria in episodic recognition, which represents a near-optimal rule for deciding whether a Probe had been in a Study set. The quantitative links between visual stimuli and memory performance demonstrate that multiple items are stored as noisy exemplars in episodic short-term memory.