Abstract
We adapted Ross & Burr (JOV 2010) to examine the relative precision of dot density and area discriminations. A reference contained 60 dots randomly distributed within a circle of fixed area. The number of dots in the test stimulus was varied over trials by an adaptive method (Watt & Andrew's APE). In ‘Density’ blocks of trials the area of the test was kept the same as the reference. In ‘Size’ blocks, the density of the test was held constant and size varied. In “Mixed” trials the two conditions were randomly interleaved and observers indicated whether the stimulus was denser, less dense, larger or smaller. RESULTS: (a) Weber fractions for density were higher than for size when expressed as changes in dot number. This result is not consistent with observers discriminating density/area through changes in number (b) Thresholds were also higher for density in the Mixed blocks (c) On trials when the test and reference had equal dot numbers, observers chose the ‘size’ and ‘density’ responses equally often. DISCUSSION We fit the data from all conditions together with a MAX model in which observers choose the largest deviation of 4 signals from zero, with 2 parameters representing noise for the two tasks. The model fails because it predicts that observers will choose the (noisier) density option more often on mixed trials when signal strength is low A much better model is that the transduced density signal has a lower signal/noise ratio than the area signal. Our findings do not rule out the possibility that there is a numerosity signal, but if so, it has an even lower precision than density.
The Wellcome Trust, EPSRC and the Max-Planck Society.