Abstract
PURPOSE: Much evidence (due mainly to Durgin & colleagues) shows that human vision is sensitive to the density of elements in sparse displays. Here we investigate the dependence of density judgments on element intensity. METHOD: In Expt. 1, S's viewed a brief display comprising two sparse fields of dots, separated by a central vertical line and judged (with feedback) which had more dots. Dots varied in grayscale over 9 levels from black to white, with grayscale 5 equal to the background (making dots of grayscale 5 invisible). There were 30 experimental conditions, each stipulating a particular pair of grayscale histograms to be pitted against each other. Each trial in a given condition randomly displayed elements conforming to the prescribed histograms across the left and right portions of the viewing window. The data yield a function giving the impact (in multiples of d') exerted on density judgments by different grayscales. In Expts. 2 and 3, S's judged which side had more black (grayscale 1) dots, and more white (grayscale 9) dots, respectively. RESULTS: In Expt. 1, impacts exerted by grayscales 1, 2, 3, 7, 8 and 9 were approximately equal, and approximately double the impacts exerted by grayscales 4 and 6 (Weber contrasts −0.25 and 0.25), showing that the statistic S's use to sense overall density is largely invariant to dot contrast. When S's focused on just the black dots (Expt. 2), impact decreased linearly from grayscale 1 to 5, and was 0 for all positive polarities. When S's focused on white dots (Expt. 3), performance was worse than for black dots, with grayscales of opposite polarity exerting substantial influence. CONCLUSIONS: Observers can use any of several statistics with differential tuning to graylevel for making judgments about texture density. One such statistic is symmetrically sensitive to positive and negative polarity (Expt. 1); another is sensitive exclusively to dots of negative polarity (Expt. 2).