Abstract
Attention is essential in selecting task-relevant information to enter capacity-limited visual short-term memory. The Attention Gating Model (AGM, Reeves & Sperling,1986) captures the temporal dynamics of attention shifts and explains a range of attention effects in high item signal to noise conditions (Sperling & Weichselgartner, 1995; Shih & Sperling, 2002). We developed a new model, the attention gating perceptual template model (agPTM), combining the AGM and the perceptual template model (Lu & Dosher, 2008), to account for performance in different stimulus conditions, including low signal contrast, high external noise, unequal item intensity or signal-to-noise ratio. Input stimuli go through contrast-gain, and are then modulated by an attention gate, with internal noise in each stage of processing. The outputs determine the discriminability and memory strength of each item. Two experiments presented a single RSVP stream of 19 letters at fovea @ 10 letters/s. Subjects reported the four earliest letters following a visual cue between the 6th and the 13th letter. Experiment 1 tested six signal contrasts spanning the full range of performance in both zero and high external noise. In experiment 2, all letter contrasts were 0.5 except that two-thirds of the energy of one letter (1, 2, 3, or 4th after the cue) was replaced with external noise by phase-scrambling the same letter. This allowed us to test two potential models of memory strength: proportional to the signal-to-noise ratio or total energy of each item. Data from both experiments were modeled with a single set of parameters. Assuming memory strength proportional to item signal-to-noise ratio provided an excellent account of all the data. In conclusion, a new agPTM accounts for performance in attention switching over a wide range of stimulus conditions. Attention gated memory strength reflects the signal-to-noise ratio instead of total energy of the items in the stimuli.
Meeting abstract presented at VSS 2014