Effect of stimulus width on simultaneous contrast

Abstract

Perceived brightness of a stimulus depends on the background against which the stimulus is set, a phenomenon known as simultaneous contrast. For instance, the same gray stimulus can look light against a black background or dark against a white background. Here we quantified the perceptual strength of simultaneous contrast as a function of stimulus width. Previous studies have reported that wider stimuli result in weaker simultaneous contrast, whereas narrower stimuli result in stronger simultaneous contrast. However, no previous research has quantified this relationship. Our results show a logarithmic relationship between stimulus width and perceived brightness. This relationship is well matched by the normalized output of a Difference-of-Gaussians (DOG) filter applied to stimuli of varied widths.

Keywords: Brightness; Edges; Illusion; Simultaneous contrast; Size; Width.

Figure 1. Effect of stimulus width on simultaneous contrast.

The two bars are the same shade of gray throughout, but they appear lighter on the top (against a dark background) than on the bottom (against a light background). The narrow bar has a stronger effect on perception than the wide bar.

Figure 2. Psychophysical design.

(A) Monitor display during the time course of a single trial. (B–D) Three different stimulus presentations of the brightness discrimination task (out of 1,584 possible combinations, see Methods section for details). The cross indicates the fixation point, and vertical red lines indicate the points to compare in the comparator and standard stimuli. Drawings not to scale.

Figure 3. Psychophysical results.

(A) Psychometric functions for the different stimulus widths are displayed in different colors. The conditions in which the comparator appears bright are indicated by triangles, and the conditions in which the comparator appears dark are indicated by squares. (B) Perceived enhancement of the PSEs for the different stimulus widths, with respect to the physical luminance of the comparator. The perceived enhancement of brightness (triangles) and darkness (squares) perception decreases as the stimulus width increases, approximately following a logarithmic function. Error bars in (A) and (B) represent the ± SEM for all subjects in each condition.

Figure 4. Computational simulations with a DOG filter.

The filter parameters were chosen to match physiological center–surround receptive fields at the eccentricity used in the psychophysical experiments (7°). (A) Top: Examples of stimuli analyzed in the simulations (top half of dark-to-bright gradients). The six different comparator widths are illustrated. These stimuli were equivalent to the comparators presented in the psychophysical experiment. The white dashed circles denote the regions of comparison during the psychophysical experiments. Bottom: Predicted responses from a DOG filter. Convolving the DOG filter with the stimuli at the top simulates the output of an array of center–surround neurons. (B) Normalized responses, at the point of discrimination in the psychophysical experiment, for each width. For each data point, the response was divided by the bar width. (C) Spread of the response for each stimulus width. We added the responses along the width of each bar, at the height of discrimination, and divided the total by the bar width. Data points indicate the widths used in the psychophysical experiment. Black lines indicate the fits to the logarithmic function used to fit the data in Fig. 3B.

Figure 5. Influence of the relative weights of the DOG model’s center and surround on data fitting.

(A) Goodness of fit (R²) as a function of relative weight (w). The red vertical line indicates equal weight of center and surround (w = 1). Note that only the results corresponding to w ≥ 1 are biologically relevant. We show the results for w < 1 for mathematical exploration. There was no correlation between spread and logarithmic function for w ≤ 0.5. (B) Examples of data fitting for different w values. Upper Left: spread and fitted curve when the weight of the center is 60% of that of the surround (w = 0.60). Upper Right: equal weights of center and surround (w = 1, i.e., Fig. 4C). Lower Left: the weight of the surround is 85% of that of the center. Lower Right: the weight of the surround is 60% of that of the center.