Named for its discoverer, the American psychologist Celeste McCollough (b. 1927), the McCollough effect was the first example of a contingent aftereffect to be discussed in scientific literature (McCollough 1965). The effect is contingent in the sense that after viewing the coloured induction pattern, its manifestation is dependent on the presence and, importatnly, the orientation of the black and white gridlines.
Despite decades of discussion, we do not currently possess a satisfactory physiological explanation of the McCollough effect. Well-understood adaptive mechanisms in the retina such as photobleaching and neural rebound (see the entry for Negative Afterimages) reach equilibrium (and hence their consequences in experience disappear) in 40 minutes or less (Frishmann 2005). However, the McCollough effect has been experimentally shown to persist for several days following a single induction, and regardless of the changing visual stimulus over that period (Jones and Holding 1975).
The long-lasting, persistent and apparently non-retinal nature of the McCollough effect has led some vision scientists to suggest that it is a form of associative learning, and in particular that it is an example of classical conditioning (Siegel and Allan 1992; Brand et al. 1987). Classical conditioning—better known as Pavlovian conditioning—can be understood as a remodelling of the brain’s synapses (the communication sites between the information-carrying neurons) due to the repeated co-occurrence of distinct and unrelated stimuli over a period of time (Proffitt and Kaiser 1998; Bourgeois et al. 2000). However, the McCollough effect seems to be more or less limited to the horizontal and vertical bar stimuli; it is not generated by viewing colours paired with faces, angles or curves (McCollough and Webster 2011). This shows that classical conditioning cannot provide a full explanation of the McCollough effect, and suggests that there must be some useful, discriminatory function of the visual system which is being exploited. One proposal recurring throughout the literature is that the McCollough effect is produced by whatever adaptive mechanism is responsible for adjusting visual input to remove chromatic aberration, the natural fringes of colour produced by the failure of a convex lens to focus all wavelengths of a light ray to the same point (see the figure below). In short, the brain interprets the McCollough induction stimulus as arising from a fault in the visual machinery, and adapts to correct for this fault. A flaw in the lens of the eye would require long-term adaptation, which would explain the duration of ME; the remarkable contingency could be explained by the fact that stimuli in which colour and orientation are so strongly correlated are relatively rare in our natural environment (MacKay 2003; Humphrey 1998).
Chromatic aberration of a convex lens – note the different focal points for red, green and blue light.
It should be stressed that any sense in which the McCollough effect is learned is intuitively different to the learning of, say, a musical instrument, or a new language. As McCollough and Webster (2011) remark in their excellent survey article—the McCollough effect 'is widely seen as an example of perceptual plasticity in which the distinction between sensory adaptation and learning is blurred’. Regular exposure to the inducting grids does not increase the strength or duration of the illusion, i.e. the effect cannot be ‘practised’ (Skowbo and Rich 1982), and the illusion is reported by subjects who (due to cortical damage or otherwise) are unaware that they have undergone the induction process (Mullin et al. 2009). In fact, the McCollough effect occurs after viewing inducing images which alternate at frequencies higher than the threshold of conscious human perception (around 15Hz), so that the green and red grid patterns are perceived as a single fused colour.
This suggests that the neural locus or substrate of the illusion lies early in the visual pathway, in cortical areas V1 to hV4 where individual cells appear to be able to track flicker as high as 60Hz (Vul and MacLeod 2006). However, the precise neural substrate is yet to be pinpointed. Vul and MacLeod (2008), and McCollough in her original paper, suspect primary visual cortex V1. This is the earliest cortical area in the pathway of visual information, and the ‘lowest’ level of processing in a hierarchy defined by upwards information flow from the peripheries of the nervous system (e.g. the retina). Both papers cite the fact that the McCollough effect shows little or no interocular transfer, meaning that if only the right eye has received the induction treatment then viewing the test image with only the left eye will produce no illusory effect. This suggests a locus no later than the point where the individual retinal images from each eye are fused into a single neural image, and this fusion is thought to occur in V1. This is consistent with the contingency of the illusion; orientation selectivity does not occur in the visual pathway until V1, where the rate of firing of some neurons is dependent on whether the stimulus in their receptive field is a horizontal bar or a vertical bar (Frishmann 2005). However, fMRI probes of brain activity (Barnes et al. 1999) and studies involving individuals with damaged cerebral cortices (Mullin at al. 2009) suggest that it is the fusiform gyrus near the area hV4 which is most active while a subject is undergoing the illusion, while V1/V2 are relatively inactive. The area hV4 is strongly affected by both the colour of stimulus and the attentional focus of the subject, while the fusiform gyrus – which is active only in the illusory case - has been linked to synaesthesia (Ramachandran and Brang 2008). According to Barnes et al. (1999) these experiments suggest that ME involves ‘top-down’ processing, meaning that higher order areas of the visual cortex modify the output of those of a lower order.
Human visual cortex
The McCollough effect is rarely explicitly discussed in philosophical literature, although it makes for an interesting case study at the intersection between neuroscience and philosophy. The paradigm of "top-down" versus "bottom-up" influences is used in both disciplines. Psychologist Stephen Palmer says that"
“Bottom-up” processing […] refers to processes that take a “lower-level” representation as input and create or modify a “higher-level” representation as output. Top-down processing […] refers to processes that operate in the opposite direction, taking a “higher-level” representation as input and producing or modifying a “lower-level” representation as output. (1999, pp. 84–85)
One sort of top-down process involves the influence of contentful mental states, such as beliefs or desires, on perceptual processes—see Shea (2013) for a detailed discussion. The McCollough effect is not ‘top-down’ in this sense; regardless of where the activity is in the brain, the process of inducing the McCollough effect is not like forming a belief that colour and orientation of lines are concurrent and then that affecting one's perception.
An inaccurate or non-veridical visual experience of the world can persist even when one believes that is not accurate—we will continue to have an illusory experience of green and red gridlines due to theMcCollough effect even if we believe that those lines are colourless. This is potential evidence for the modularity of mind hypothesis advanced in Fodor (1983), where the author argues that a perceptual system comprises a computational unit or module that is an informationally encapsulated part of the mind. In particular, this means that the visual system does not receive inputs from that part of the mind responsible for belief fixation. However, the existence of illusions is nowhere near enough to establish modularity. Prinz (2006) notes that it may simply be the case that perceptual states trump beliefs when the two come into conflict. Macpherson (2012) points out that the existence of some cases in which there is perceptual information encapsulation does not show that there is always information encapsulation.