Dissociation of color and figure-ground effects in the watercolor illusion

Abstract

Two phenomena can be observed in the watercolor illusion: illusory color spreading and figure-ground organization. We performed experiments to determine whether the figure-ground effect is a consequence of the color illusion or due to an independent mechanism. Subjects were tested with displays consisting of six adjacent compartments--three that generated the illusion alternating with three that served for comparison. In a first set of experiments, the illusory color was measured by finding the matching physical color in the alternate compartments. Figureness (probability of 'figure' responses, 2AFC) of the watercolor compartments was then determined with and without the matching color in the alternate compartments. The color match reduced figureness, but did not abolish it. There was a range of colors in which the watercolor compartments dominated as figures over the alternate compartments although the latter appeared more saturated in color. In another experiment, the effect of tinting alternate compartments was measured in displays without watercolor illusion. Figureness increased with color contrast, but its value at the equivalent contrast fell short of the figureness value obtained for the watercolor pattern. Thus, in both experiments, figureness produced by the watercolor pattern was stronger than expected from the color effect, suggesting independent mechanisms. Considering the neurophysiology, we propose that the color illusion follows from the principles of representation of surface color in the visual cortex, while the figure-ground effect results from two mechanisms of border ownership assignment, one that is sensitive to asymmetric shape of edge profile, the other to consistency of color borders.

Fig. 1

Examples of the stimuli used to study figure-ground organization in the watercolor illusion. The stimuli consisted of six adjacent compartments outlined by wavy double lines. The compartments that were lined orange on the inside are called ‘watercolor compartments’, the others ‘alternate compartments’. The relative width of the two compartments was varied; a ratio of 2:1 is illustrated. Arrows and numbers at the top served to identify the two types of compartments and were displayed 1s after stimulus onset. Stimuli of type A were used in Exp. 2, stimuli of type B in Expts. 3 and 4. Color matching (Exp. 1) was performed for both types of stimuli.

Fig. 2

The results of the color matching experiment. Dots indicate the contrast of the physical color in the alternate compartments that subjectively matched the illusory watercolor, for each of 17 subjects, including repeats in three subjects (dots connected by lines). Vertical lines connect the means of the reversal points in the matching procedure, that is, the two colors that looked clearly lighter and darker (more saturated), respectively, than the watercolor. Dotted horizontal lines indicate the three colors used in the figure-ground experiments (two sets of colors for two groups of subjects).

Fig. 3

‘Figureness’ (probability of being perceived as figure) of the watercolor compartments plotted as a function of the width ratio between watercolor and alternate compartments. Filled symbols and solid lines represent the results for the three color conditions, one in which the color of the alternate compartments matched the illusory color (squares), one in which it was whiter (upward pointing triangles), and one in which it was darker, or more saturated (downward pointing triangles). Open symbols and dashed lines represent the baseline condition in which the compartments were defined by the dark lines only (without the orange lining). It can be seen that figureness of the watercolor compartments tended to decrease as the width of the alternate compartments was reduced (effect of proximity). Figureness decreased also as the color of the alternate compartments was made more saturated.

Fig. 4

Figureness of the watercolor compartments as a function of the color condition, for three width ratios. Dashed line represents baseline condition. Data replotted from Fig. 3. Note that figureness of the watercolor compartments in the color-matching condition was higher than figureness of the corresponding compartments in the baseline condition.

Fig. 5

Results of a similar experiment as described in Fig. 3, but for displays in which the top and bottom of the compartments were occluded by gray bars (Fig. 1B). The figureness values were slightly higher than without the occluding bars, but the results were otherwise similar.

Fig. 6

Figureness of the watercolor compartments as a function of the color condition for displays in which the top and bottom of the compartments were occluded by gray bars (Fig. 1B). Data replotted from Fig. 5. Note that figureness of the watercolor compartments in the color-matching condition was higher than figureness of the corresponding compartments in the baseline condition.

Fig.7

Figureness produced by color tint. Compartments were defined by single dark lines and alternate compartments were tinted with diluted orange. Filled symbols show the probability of the tinted compartments being perceived as figures as a function of the color contrast. Open symbols show the figureness values for the watercolor displays at the equivalent contrast. It can be seen that the watercolor illusion produced higher figureness values than the corresponding color tint.

Fig. 8

Hypothetical explanation of the figure-ground effect in the watercolor illusion as a result of the neural processing of border ownership. See Discussion for further explanation.