Abstract
Koch and Ullman (1985) proposed a 2D salience map that took features-map outputs as its input and represented the relative importance "saliency" of the features as a scalar (real number). The computation on the map, "Winner-Take-All," was used to predict action priority. We propose that the same or a similar map is used to compute centroid judgments, the center of a cloud of diverse items. Sun, Chu, and Sperling (2021) demonstrated that following a 250 msec exposure of a 24-dot array of 3 intermixed colors, subjects could accurately report the centroid of each dot color, thereby indicating that these subjects had at least three salience maps. Here, we use a partial-report paradigm to determine how many more salience maps subjects might have. Procedure. Subjects viewed arrays of 28 to 32 items, each item had one of M different features. An array was flashed for 300 msec followed by a 50 msec blank field, a 100 mask, and a cue to report the centroid of items of just the cued feature. Features were either M different dot colors, or M shapes, or a mixture of colored dots and shapes. In 11 experiments, M ranged from 3 to 8. Results. Subjects performed reasonably well in all experiments, utilizing at least 17 stimulus items in their responses for M=3, to utilizing 12 items for M>5. We consider two theories to account for the performance loss with increasing M: (1) subsampling, attending to only a subset of features; (2) attending to all features but experiencing a general loss due to increased task complexity. To distinguish between (1) and (2), we determined how performance with M-1 features could predict performance with M. The analysis indicates that of the three subjects, two have at least 5 salience maps, one has at least 4.