​Adelson’s Checkershadow Illusion

Adelson’s Checkers Shadow illusion exploits the mechanisms underlying lightness constancy: our capacity to perceive the lightness (or reflectance) of a surface as invariant, even when the intensity of incident light (the illuminance) is changing at a point or is variable across the surface. Under normal circumstances, lightness constancy allows us to distinguish between brightly lit dark surfaces and dimly lit white surfaces, which helps us to detect edges and forms. It is just one of a number of constancies – including size, shape, colour and roughness – that humans exhibit that are used to allow us to skillfully negotiate our environment. In the general case, perceptual constancy is defined as a subject’s capacity to perceive some property of an object (e.g. lightness, colour, size) as being independent of external conditions (e.g. lighting or distance) (Gilchrist 2010).

The visual system is commonly divided into three levels of information processing: low-level vision focuses on physiological mechanisms in the retina and the resultant neural signals; high-level vision is more cognitive or ‘thought-like’ and involves knowledge and reasoning. Mid-level vision lacks a clear definition, but it is sometimes associated with Gestalt psychology and an emphasis on the organizational structures of perception (Adelson 2000). In the Checker-Shadow Illusion, two simultaneously perceived targets (tile A and tile B) are identical local stimuli but one is seen as lighter than the other. This illusory effect is called simultaneous lightness contrast (SLC) and it has a high-level explanation and a low-level one. We outline both below.

The low-level constancy mechanism is referred to as Hering-type lightness contrast, after German physiologist Ewald Hering (1834-1918). Hering emphasised the explanatory power of localized retinal processes affecting the neurons, retinal ganglion cells, which send signals from the eye to the brain. The retinal ganglion cells receive electrical impulses from the photoreceptive rods and cones contained in the outer retina and send corresponding signals to the brain, producing a measure of the intensity of light hitting the eye (the luminance). For each retinal ganglion cell there is a corresponding receptive field, which may be thought as the small region of rods and cones which can affect the given cell’s firing rate. Cells that are close together will have overlapping receptive fields. Two kinds of retinal ganglion cells exist: an ON-centre cell is excited by a central bright spot in its field but is inhibited by a bright surround; an OFF-centre cell has the opposite arrangement.

The process of turning the retinal stimulus into a neural image is spatial filtering; the inhibitory mechanism operates across the 2-D space of the receptive field to effectively remove the image data wherever there is uniform luminance. Of course, we are not consciously aware of any such blind spots; this could be explained by the well-attested perceptual phenomenon of ‘filling-in’ which takes place in the primary and secondary visual cortices (De Weerd 2010).

These retinal and early cortical processes provide an underlying mechanism for lightness constancy—our visual system is less sensitive to gradual changes in luminance than it is to sudden local contrasts, and uniform regions are ‘filled in’. However, this does not fully explain Adelson’s Checker-Shadow. Other lightness illusions have shown that a simple localized centre-surround model is inadequate (Gilchrist 1979), and depth cues and the presence of transparent overlays can dramatically alter perceived lightness without any significant changes to the light information received at the eye (Gilchrist 1977; Hochberg and Beck 1954). These results have motivated an alternative, high-level approach.

The high-level explanation of lightness constancy is associated with Hermann von Helmholtz (1821-1894), a German physician, physicist and philosopher of science. As a physicist, Helmholtz believed that the eye was an imperfect optical instrument and that the brain must combine raw sensory input and previous experience to arrive at a ‘best guess’ as to what is out in the world. Given his emphasis on the purpose of the visual system in relation to the natural environment, it is interesting to note that Helmholtz was a contemporary of Darwin and spoke approvingly of evolutionary theory (Helmholtz 1873/1865). According to Helmholtz (1867), perception is the result of ‘unconscious inferences’. Lightness constancy, in particular, is achieved by the visual system inferring and then discounting the illuminant. While there is a shadow depicted as falling across the checkerboard, there isn't really a shadow there—for Adelson's Checker-Shadow illusion is just a picture of a shadow. And this is what causes the illusion. Adelson’s Checker-Shadow Illusion is therefore the result of a failure of our inductive inferences to line up with the world. Using premises based on sensory evidence, perceived lighting conditions, and previous experiences, our visual system arrives at a false conclusion.

In some ways this ‘best guess’ hypothesis is plausible. Adelson’s Checkers Shadow Illusion is a 2-D figure but if we were really looking at a 3-D scene containing a single light-source and an ordinary checkerboard then the actual tile A really would be darker than the actual tile B. This is why Adelson (2005) says that this illusion ‘demonstrates the success rather than the failure of the visual system’. The illusion makes manifest the mechanism that normally allows us to see the lightness of objects in different illuminats accurately. However, there is no empirical test for the existence of unconscious inferences, and modern theories of lightness perception incorporate both high-level and low-level mechanisms in explaining our experience—see Kingdom (2011) for a survey.

Both Hering-type and Helmholtz-type accounts of vision have philosophical commitments. For Helmholtz, perception occurs in two parts: an apprehension of raw sensory input, followed by an interpretive act of unconscious inference. This distinction between sensation and mental processes leads Smith (2002) to characterise Helmholtz’s view as indirect realism of the kind espoused by Bertrand Russell in The Problems of Philosophy (1998/1912). In broad terms, indirect realism is the view that the objects of the external world exist independently of the mind, but we can perceive them only via intermediary vehicles in the form of mind-dependent objects. On Smith’s reading, Helmholtz and Russell agree that the objects of our immediate acquaintance are not ordinary physical objects. For Helmholtz the objects of immediate acquaintance are simply the brain states which arise from unconscious inferences, whereas for Russell they are ‘sense-data’ such as a red patch of colour, or a smoothness of texture, insofar as these are things that are grasped by the mind.

The capacity for lightness constancy has also been used to motivate certain philosophical positions. Burge (2010) argues for a direct realist form of representationalism—roughly, the view that perceptual states possess content which represents objects, properties and relations as being a certain way, and in virtue of so doing allow us to be directly aware of objects and properties in the world without being aware of any intermediaries. Burge claims that perceptual constancy, understood as a subpersonal capacity to recognise some attribute of the physical environment as unchanged despite variations in sensory stimulation, is necessary for perception—perceptual constancies are capacities for objectification of sensory stimulus (roughly, the capacity to track objects and properties). For Burge, the objects of the external world are the immediate objects of perception, but our contentful representations of these objects may fail to accurately represent the world (they may fail to be veridical). For a discussion of whether sense-data theory is a form of representationalism see Macpherson (2014).