I have attempted to argue that we need to reconceive the notion of what an illusion is. In the context of the traditional line drawings used over a hundred years ago to illustrate the shortcomings of vision, illusionsm have begun to misguide our thinking about normal perception. Illusionsm do not indicate the error-prone nature of visual perception. On the contrary, they tend to be small compared to the many illusionsd that go unnoticed on a regular basis. To illustrate that this is the case, I have used two examples from the domain of complex relational properties. This choice was based on the conviction that perception of everyday objects always necessarily includes judgment (be it in terms of Helmholtzian unconscious inference, or be it in terms of private models that may or may not become transparent to the perceiver). The notion that illusionsm should be of interest because they reveal the workings of how the visual system derives percepts from simple sensations is not useful. It is not useful because an illusionm only becomes manifest by a comparison process that is at least as fraught with cognition as is the perception of everyday relational properties. We have used the classical stick in the water and the equally classical Ebbinghaus illusion to illustrate that illusionsm only become manifest if a cognitive operation is performed (i.e., a perception-inference-cycle when moving the stick or comparing the circle to a reference circle known to be of identical size).
It is also impossible to investigate illusions as merely phenomenal problems. And it is ill-conceived to limit the study of visual perception to seemingly simple phenomena that end up requiring cognition after all. Perceiving is to make perceptual judgments, be they explicit (e.g., by saying which of two objects is larger), or be they altogether implicit, or merely amenable to consciousness by an act of attention (e.g., by determining hand-aperture when grasping an object). It is thus impossible to investigate illusions as purely perceptual errors. Instead, illusions always have a cognitive component in the sense that they require an act of comparison or inference. This holds for all illusionsm, even if they may not be amenable to consciousness. To take illusions as a discrepancy between what we see and what there is, is doubly mistaken. First, there is always a discrepancy (illusiond) between a visual percept and the object in the world to which it refers, namely the stimulus. And second, only in rare and simple cases do we notice this discrepancy (illusionm). The discrepancy is owed to the underspecification problem (UP), the qualitative information gap between the two-dimensional retinal image and the richer three-dimensional percept. The UP puts the perceptual system in a position from which it has to draw additional information from memory, from inference, or from internalized structures that have been acquired throughout evolution. Such structures have been suggested to include that objects are three-dimensional, that light comes from above, that gravity acts along the main body axis when standing or walking, or that the brightest patch in the visual field is usually "white". Internalized structures gain particular weight if the stimulus is poor. This is the case when looking at simple line drawings and it is all the more the case when looking at relational properties. The quality of solutions to the UP differ greatly as the function of the task demands, but not necessarily as a function of the complexity of the stimulus. On the one hand, the perceptual system achieves performance that seemingly approaches perfection where precise motor action is required in personal space. On the other hand, in more remote action or vista space (for a very useful taxonomy of space see e.g., Grüsser 1983) some blatant errors are made. Our perception often defies the most basic laws of physics. More often than not do these errors go unnoticed. To illustrate how crudely our perceptions approximate reality even in personal space, we have explored errors in balancing objects and judging the slipperiness of surfaces. When it comes to these relational properties, our perception falls far from the truth. It appears that the errors tend to be as large as they can be without interfering with the perception–action cycle required for adequate or acceptable action. The evolutionary fine-tuning would minimize error until it is no longer relevant for survival. In this sense, normal perception (i.e., the illusiond) is a satisficing solution. The magnitude of the perceptual errors many observers make is in the league of errors associated with probability judgments (see e.g., Kahneman et al. 1982) and syllogistic reasoning, as opposed to the much smaller errors typically associated with perceptual illusionsm.
Our perception, just like our cognition, has developed to find solutions to problems that suffice. When reaching for an object, perception is accurate enough not to knock it over but to grasp it (most of the time). When judging a surface, it is accurate enough that we do not slip (most of the time). These examples are noteworthy because they do not relegate perceptual error to remote vista space, where precision would not matter. Toppling over an object or falling on a slippery slope concern us in personal space.
In essence, the UP is solved with remarkable accuracy for simple properties of objects within our domain of interaction. However, as soon as the perceptual properties become more complex and involve the relation between two or more objects, the perceptual system can no longer solve the UP with any degree sophistication that goes beyond the level of medieval physics. But rather than giving up and seeing astounding illusions everywhere, the system degrades gracefully and builds theories that suffice for the purpose at hand. Their deviation from reality is not experienced. These perceptual theories may be thought of as more or less universal tools for upholding a meaningful world (in the sense of Shepard 1994); however, it might make more sense to think of them as universal tools with a private touch that accommodates individual perception-action requirements. A hockey player or a juggler will for instance have developed private models, be they unconscious or amenable to introspection, about friction or balancing that are more sophisticated than the layperson’s. Note that these models need not be explicit, in the sense of a perceptual process, of which the cognitive elements cannot be separated out.
Such private adjustments and elaborations when solving the UP need not be made in the case of classical geometric-optical illusionsm. I hope the above examples and case studies have shown that ilusionsd, such as the Luther illusion, do not require detection, and illusionsm that become manifest, such as the Ebbinghaus illusion, can be upheld because their limited magnitude makes them irrelevant for action.
This raises the questions why illusionsm arise at all. Illusionsm might arise as mere epi-phenomena or as meaningful warning signs for the system to signal that a perceptual fine-tuning is needed. The epi-phenomenon interpretation would suggest that the juxtaposition of two contradictory percepts is a fluke and happens per-chance every once in a while. Optical illusionsm are merely collections of such flukes. The warning-sign interpretation would see in them the purpose of fine-tuning the perceptual system. If the perceptual system subserves action, it would ideally minimize error (illusionsd), and one mechanism to do so would be the experience of illusionsm. It is unclear, however, why illusions would have to become conscious for this fine-tuning to work. Would the necessary re-direction of attention require the experience of an illusionm? Be this as it may, the system does not even notice error—let alone attempt such fine-tuning—when it comes to perceiving relational properties. Even an approximate veridical perception of relational properties is out of reach of the perceptual system. The system merely arrives at the first solution that satisfies our action needs. A flashy epi-phenomenon or a warning system, as indicated by manifest illusionsm, is not useful here, as the discrepancy between percept and reality is too large.
Now, one might ask about cases where the error is exceedingly large and a warning may indeed be in place. These cases are rare; but they do, however, result in manifest illusionm, and hence are compatible with the purpose of illusionm that we suggest. Take for instance the perception of pain in a phantom limb. Here the sufferer does notice the illusionm. How can pain be so vividly felt in a limb that is no longer there? The warning function of this manifest illusionm is obvious. For instance, learned reflexes involving the absent limb need to be extinguished and reprogrammed. A more interesting case is the infamous rubber-hand illusion (Botvinick & Cohen 1998) or the full-body illusion that can be created in most observers by synchronizing their actions and perceptions with those of an avatar seen in a VR (Virtual Reality) presentation (see e.g., Blanke & Metzinger 2009; Blanke 2012; Botvinick & Cohen 1998; Lenggenhager et al. 2007). Only in such extreme cases does the error manifest itself in a complex relational case. We feel that we are someone or somewhere else and at the same time feel that we are not. It seems to take such extreme cases before we find a sizable illusiond+m that deserves the name “illusion”.
In most cases, we can adjust perceptions once we notice that they are erroneous, be they ball trajectories or balancing properties. However, this adjustment process is painfully slow and may have to draw on early stages of perceptual and cognitive development. It does not take center stage, and some theoreticians would claim that the adjustment process converges on a veridical understanding of the world (Gibson 1979 calls this “attunement”). Others claim that many perceptions are useful precisely because they do not match or converge on the world (e.g., the multimodal user interface theory of perception, Hoffman 2010). The satisficing nature of private perception may not require a perfect solution of the UP in many cases, as long as the slips and falls remain limited to a tolerable number.
Acknowledgements
Markus Homberg, Cornelius Mülenz, and Mana Saadati helped collect and analyze the balance data; Daniel Oberfeld provided valuable input for the experimental designs; Elsa Krauß, Markus Landgraf, and Laura Längsfeld carried out the friction experiments.