Right now, you’re blind at one particular part of your visual field – because you have no photoreceptors at the location on your retina where the optic nerve begins its journey to visual cortex. Normally, you’re unaware of this blind spot because of perceptual “filling-in” – a mechanism by which your brain actively fabricates the perceptual data it’s missing.
But this isn’t the only case where cognition manufactures perception. In the case of the Kanizsa triangle, you will sense the presence of a full triangle although none truly exists. In other words, your brain has fabricated “illusory contours” that seem to define the edges of an apparent, but actually nonexistent triangle.
In both these cases, believing is seeing – your brain is effectively determining its own perceptual data. A new article by Maertens and Pollmann investigates how these two phenomena might interact – finding interesting evidence that the brain might fabricate reality just one step at a time.
This illusory contour effect has been surprisingly difficult to localize: electrophysiological work suggests it may originate from primary or secondary visual cortex while neuroimaging indicates a higher-level origin (such as in the lateral occipital cortex). In contrast, the authors advocate a distributed model, such that illusory contour effects begin in higher visual areas, which resulting from their lower spatial resolution may fail to appreciate the minor gaps in the triangle’s contours and mistakenly identify the presence of a full triangle. This activity then spreads to lower visual areas, activating the representations of contours where they should exist, if a triangle were to actually be present.
For the illusory contours to be as crisp as they seem, the high resolution of primary visual cortex must play a pivotal role. To test this idea, the authors trained subjects to discriminate between fine gradations in the curvature of illusory contours (see this image) and simultaneously varied the ease with which illusory contours could be seen (via a quantity known as the “support ratio”).
The results: although the subjects were generally quite good at identifying the curvature of (nonexistent!) contours, they were markedly impaired when these contours passed through their physiological blind spots. In other words, subjects were particularly bad at discriminating the boundary of an imaginary triangle if that boundary passed through their blind spot.
Amazingly, this impairment was worst when the task should have been easiest! In other words, the same illusory contours that are easiest to identify in normal vision were hardest to identify in the blind spot. Why should this be the case?
“These results don’t make sense. Shouldn’t the brain be better at filling in perceptual data in the blind spot, where it has practice doing this all the time?” Anticipating this question, the authors suggest that illusory contours cannot be generated in the blind spot because illusory contours are not sufficiently activated in order to produce “filling in.” *
Cognition can manufacture perception via multiple mechanisms; and in the right circumstances, these mechanisms can be turned against one another. Whereas the fine discrimination of illusory contours seems to require representations in primary visual cortex, the filling-in at the blind spot requires information from higher visual areas. These mechanisms interact such that fine discrimination of illusory curvature is more difficult in the blindspot, and this is more true when the surrounding visual information is real than when it is illusory. So the brain can certainly fabricate reality, but seems to do so just one step at a time.
Perceptual Sampling: The Wagon Wheel Illusion (and a follow-up post)
Reversing Time: Temporal Illusions
Pictures in the Brain
Reichardt Detectors and Illusory Motion Reversal
The Constructive Aspects of Visual Perception (@ machines like us)
* – Although the data are compelling, this explanation falls short. One might think, based on this explanation, that increased support ratios (which normally translates into increased ease perceiving illusory contours) should increase the chances for discrimination of contours across the blind spot, but this is the opposite of what was found. Alternatively, the perceptual “filling in” might fail if illusory contours lie entirely within the blind spot; according to this interpretation, illusory curvature discrimination should be impaired with low contour side length, but not as much with high luminance differences between foreground and background (future research?). In other words, the most counterintuitive finding here could be an artifact of the way the support ratio was changed – by decreasing its denominator rather than increasing its nominator.