The work of the visual system in finding patches of intensity, color, and motion, and from them detecting the presence of significant edges and regions must eventually culminate in the grouping of these regions into significant gestalts if anything familiar is to be recognized. Gestalt perception can fail if elements comprising a gestalt are insufficiently salient, if too few of these elements are present, or if the manner in which a gestalt's elements represent its presence is not one with which the visual system can deal. The following figures illustrate these remarks and explore their significance.
We begin with a standard red-green colorblindness test. This figure represents the form of a large '2' in green dots, surrounded by a red-dot background. The intensity of the dots is varied randomly so as to leave color as the only clue to the gestalt's presence. The ordinary ged-green colorblind person (the author of these papers is one such) will be able to identify each of the green dots and distinguish them from the red. However,in the visual system of a colorblind person the perceptual saliency of the red-green distinction is sufficiently reduced for formation of the gestalt to form, and no numeral is seen. (Believe me - I can't see it at all.)
Figure 1. Standard red-green colorblindness test
We can use texture differences to represent this phenomenon to the eye of those not colorblind. The following figure gives the green dots of Figure 1 a texture which differs enough from that given to the red dots of Figure 1 for the gestalt to stand out clearly for most people, including the colorblind. In figure this same textural difference is weakened enough for the gestalt to disappear for most people. Note that in Figure 3 the difference in texture between the dot is easily visible on individual inspection. It is just that the texture difference is insufficiently salient for the gestalt to form. This is exactly what happens to the red-green colorblind when they view Figure 1.
Figure 2. Colorblindness test rendered in strongly distinguished textures
Figure 3. Colorblindness test rendered in less strongly distinguished textures.
Any perceptual difference between the elements of a gestalt and their surround can generate the gestalt perception if sufficiently strong. Our net figure demonstrates that a pure motion cue, a pure motion-difference cue, the difference between coherent and incoherent motion, and the difference between two intensities of coherent motion can do this.
Figure 4. A gestalt rendered in a variety of motion differences
Perceptual differences among two or more separate axes, each too indistinct to create a clear gestalt on its own, can reinforce each other if applied uniformly to the elements of a gestalt, and make the gestalt more perceptible. This is shown in our next figure, which shows the same gestalt '2' as in Figure 1, rendered against a random dot background, in three ways: (a) by a barely perceptible intensity difference; (b) by a barely perceptible motion; (c) by the combination of both. These tree states are tagged by he color of a small square shown inconspicuously near the upper left corner of the figure. This is red (resp. blue) when the '2' is being represented by motion (resp. intensity) alone, and black when both are being used. It will be seen that the combination (c) is more perceptible than either (a) or (b) separately.
Figure 5. Strengthening a gestalt by simultaneous use of two unrelated cues.
A gestalt can be rendered, patchwork fashion, by elements separated from their background along a variety of perceptual axes, as for example a dotted line seen against a grey background but rendered by a mixture of dark and light dots, or dark and textured dots, or dots perceptible by their intensity and internal motion. This is shown in our next figure, which displays a variety of such cases successively. (It may be noted that of all the patchwork combinations shown, the one least comfortable to the eye is the unmoving dark-dots/light-dots combination.
Figure 6. A gestalt rendered by a variety of mixed elements.
We summarize the fact apparent in Figure 6 by stating the rule that disjunctive gestalts are perceptible.
Given the fact just stated, it is interesting that conjunctive gestalts are not perceptible, an observation which we may call Treisman's Rule. This is shown in our next figure, which represents our standard '2' gestalt in several ways, some of which make it entirely invisible, others of which make it clearly or marginally visible. For example, the gestalt is entirely invisible when the dots comprising it are marked only as those which are dark and have a boundary, in a field of dots others of which are either dark or have a boundary but not both. The same is true if we replace 'dark' by red. The gestalt can even be difficult to see if it is defined conjunctively but placed in a field of dots some of which have one but not both of the properties defining the gestalt, as for example when the dots comprising the gestalt are marked as those which are dark and have a boundary, in a field of dots others of which are dark but have no boundary.
Figure 7. Imperceptibility of gestalts defined by various cue conjuncts.
The fact that conjunctive gestalts are not perceptible is further confirmed by the following figure, which renders our standard '2' gestalt in three ways: (i) as those dots which are both darkened and internally moving, in a field of dots others of which are either dark or moving but not both; (ii) as those dots which darkened in a field otheres of which are moving; (iii) as those dots which moving in a field others of which are darkened. The gestalt is plainly visible in cases (ii) and (iii), but imperceptible in case (i).