Anticipation Through Imagination

by Austin Kral

The following essay is an explication of the three dominant approaches taken towards resolving the problem of amodal completion (as articulated by Bence Nanay), namely, the perception-, belief-, and access-accounts, followed by a summary of Nanay’s motivation for developing his own contending theory, the imagery-account, which he offers as a replacement to the other three. It concludes with a commentary on the imagery-account’s apparent reliance on aspects of the perception-, belief-, and access-accounts in turn, and its corresponding failure to render them obsolete.

Keywords: philosophy, mind, amodal, perception, psychology, imagination

One of the long-standing problems faced by philosophers of mind, going as far back as Aristotle’s De Anima, is the extent to which perception and imagination are related. In this paper, I will follow Bence Nanay and use the term imagination to refer exclusively to non-perceptual states that possess phenomenal character. Notably, this excludes propositional attitudes or suppositions (Nanay, 2016, p. 1). Although perception and imagination share many of the same phenomenal features, it is also true that they differ in numerous respects, most notably their causal influence (Nichols & Stich, 2000, p. 122 & 130). While perceptual experiences tend to interact with other mental states (such as beliefs, desires, emotions, etc.) in consistent and predictable ways, the role of the imagination remains undefined. Nonetheless, the phenomenological similarity between the two remains significant enough to raise the following question: are perception and imagination really all that distinct (Nanay, 2010, p. 252)?

One area of research in psychology that has proven especially fruitful in fueling this discussion is that of amodal completion, the perception of partially obscured objects as being whole and continuous despite a deficit of corresponding sensory stimuli. While such perceptual “filling-in” is most frequently mentioned in reference to vision, it is also present in other sense modalities (such as the missing fundamental illusion in hearing, where the occurrence of two tones together causes a third illusory tone to be heard). Amodal completion is relevant to the current inquiry because, as Nanay suggests in his paper “Perception and imagination,” standard conceptions of bottom-up processing are incapable of explaining how sensationally non-existent stimuli are nevertheless experienced as being perceptually present. That is, how do occluded objects factor into our perceptual experience without directly interacting with our senses? In answer to this, top-down models of perception that allow for the involvement of mental imagery generated by the imagination are, prima facie, far more promising (Nanay, 2016, p. 6). Nanay claims that not only do imaginative states have the capacity to augment perception by way of mental imagery, but that such mental imagery is a necessary component of amodal completion. Thus, given the ubiquity of obscurants in the visual field, Nanay’s suggestion amounts to the following: all cases of perception involve mental imagery to a greater or lesser degree (Nanay, 2016, p. 8), wherein mental imagery is characterized as being quasi-perceptual, conscious, and largely inactive (as opposed to visualization) (Nanay, 2010, p. 249). However, before Nanay can establish this conclusion, he must first demonstrate the insufficiency of the existing models, namely the perception-, belief-, and access-accounts (Nanay, 2010, p. 242–249).

According to the perception-account, first formulated by the psychologist James J. Gibson, amodal completion is simply the perception of something as being occluded. That is, there is no need to intentionally represent the occluded stimulus because the extant visual cues (such as binocular disparity1, T-junctions2, and texture deletion3) are sufficient to suggest what is present behind the occluder. Nanay briefly rejects it on the premise that an absence of sensory stimulation is antithetical to perception, citing the filling-in of the blind spot as a parallel case. In both instances, a lack of sensory activation implies a corresponding lack of bottom-up perception (Nanay, 2010, p. 242).

Next, Nanay poses two objections to the belief-account, which states that amodal completion consists entirely of belief-like representations of the occluded parts of objects. Rather than framing amodal completion as a special type of perception, the belief-account claims that knowledge of what should exist behind an occluded region is sufficient to inform a mental representation of it. Nanay’s first objection rests on the apparent involuntariness of amodal completion, which makes it insensitive to the beliefs that are supposedly relied upon to represent it. Referencing the elongated horse illusion4, Nanay points out that rather than seeing two horses as being occluded, the front half of one and the back half of the other, most people perceive one elongated horse. This remains true despite the surrounding pattern and previous knowledge about horses favoring the former interpretation. Therefore, Nanay concludes, there are some instances of amodal completion that not only do not rely on belief but are in direct contradiction with it (Nanay, 2010, p. 243–244).

Nanay’s second objection to the belief-account relies on neurological evidence, which suggests that there is significant activation in the primary visual cortex during instances of amodal and modal completion. If, as the belief-account claims, amodal completion is non-perceptual, why are activation patterns typically associated with perceptual experiences present in such cases? This failure of the belief-account to explain the current neurological evidence further counts against it (Nanay, 2010, p. 244–246).

Lastly, Nanay has three objections to the access-account. In its most radical form, the access-account states that amodal completion does not involve representation but instead relies entirely on an understanding of what is perceptually accessible (possibly informed by sensorimotor intuition5). However, the access-account quickly runs into problems, as experience seems to play an important role in how an occluded region is amodally completed. If this is the case, the access-account appears to collapse into the belief-account (Nanay, 2010, p. 247).

Second, the access-account provides no way of discriminating between cases of partially obscured objects and fully obscured objects. For example, the experience of amodally completing a cat’s tail behind a fence is very different from the experience of knowing that there is a cat in the next room. While in both cases the occluded part(s) of the cat are perceptually present, the former has a phenomenal quality that the latter lacks, and the access-account fails to explain this (Nanay, 2010, p. 247–248).

Third, there are cases of amodal completion that involve only two-dimensional stimuli (such as perceiving an incomplete circle as being whole but partially covered). Here, there is seemingly no vantage point that would reveal the hidden part of the circle (short of removing its obscurant), and thus it is for all purposes perceptually inaccessible. Yet, it is precisely such perceptual accessibility that the access-account claims enables amodal completion. Therefore, on each of these three counts, the access-account falls short (Nanay, 2010, p. 248–249).

Having demonstrated the inadequacy of the prevailing theories, Nanay puts forth his own, which he titles the imagery-account (Nanay, 2010, p. 249). According to Nanay, amodal completion involves the use of mental imagery to represent the occluded region(s) of objects. Importantly, the generation of mental imagery does not require active visualization and often takes place passively, which “allows us to localize the imagined object in our egocentric space” (Nanay, 2010, p. 250). Nanay goes on to cite further neuroscientific data that seems to support this view: while retinal cells remain inactive due to lack of sensory stimuli, the occluded or incomplete regions nonetheless cause activity in the first stages of visual processing. This suggests that there is a top-down influence at play, one that operates both consciously and unconsciously—explaining our apparent lack of control over amodal completion—but that is nevertheless sensitive to our beliefs and prior knowledge, thus accounting for the determinacy of amodal perceptions (Nanay, 2016, p. 6). The imagination thus proves to be a promising candidate to play this role, and Nanay concludes that his mental imagery model is more empirically and conceptually plausible than its precursors.

Although I do not take issue with Nanay’s claim that mental imagery does, in some cases, contribute to amodal completion, I believe that his rejection of the perception-, belief-, and access-accounts lacks adequate justification. Rather than rendering these models obsolete, the mental imagery account presupposes their explanatory power. The process of amodal completion is more variable than Nanay’s mental imagery model suggests and is better explained by counterfactual reasoning in general, of which mental imagery is a special kind. Here, I use counterfactual reasoning to refer to a broader range of imaginative states that includes both propositional imaginings and suppositions. While the inclusion of suppositions among imaginative states remains controversial, I include them here due to their clear role in forming counterfactuals (such as “If I were a dog, I would have four legs.”). On my account, both perceptions and beliefs (or intuitions) are needed to appropriately constrain the counterfactual possibilities to only those most relevant to the current circumstances. Without such constraints, it would seem inexplicable why, in the case of amodal completion, the imagination consistently produces veridical mental images.

Furthermore, Nanay’s description of passive mental imagery depends on what appears to be a form of sensorimotor intuition akin to that which he rejects in the access-account, a misguided dismissal by his own admission (Nanay, 2010, p. 250). In brief, sensorimotor intuition is the representation of an agent’s position in space and its relationships to its surrounding environment. The similarity between this and Nanay’s suggestion that we “localize the imagined object in our egocentric space” (Nanay 2010, p. 250) is readily apparent and suggests that perception is not strictly timebound. In other words, perception is essentially predictive and involves the projection of possible future states onto the present environment. Here, I draw heavily upon Ralf Möller’s behavior-based approach to visual perception, which can be summed up by the tagline “perception through anticipation” (Möller, 1999, p. 171). My suggestion is as follows: through perception, patterns of spatial relationships are identified and become constitutive of the beliefs (i.e., informed predictions) which, in turn, inform the sensorimotor intuitions suggested by the access-account. These intuitions then serve to constrain the counterfactuals presented to the observer. In this way, the four models of amodal completion outlined above are temporally distributed and represent different stages of a multifaceted and qualitatively variable process.

The benefit of including counterfactual reasoning becomes readily apparent when one considers the complexity that characterizes much of human action. While many of the purely reactive behaviors demonstrated by nonhuman animals could be explained without appealing to imaginative mental states, human decision-making is often far more involved and suggests the weighing of multiple potential futures. Similarly, amodal completion depends on predicting what “would be perceived” should the obscurant be removed or the observer repositioned, and the final filling-in is only one of numerous options. While counterfactual reasoning seems like a promising candidate to play this role, it would be insufficiently constrained without the aid of perceptions and beliefs and would consequently fail to account for the general accuracy of amodal completion. Thus, while Nanay’s inclusion of the imagination in amodal perception has merit, his limiting it to mental imagery is unjustified and does not explain cases of amodal completion that are unaccompanied by vivid sensory qualities. Additionally, the dismissal of all three alternate accounts weakens his imagery-model, which cannot independently account for the primarily veridical nature of amodal completion. Some form of statistical inference appears to be necessary to narrow the range of possible outcomes to only those that are most relevant. And such inference requires the incorporation of the perceptual cues described by the perception-account, the informed predictions appealed to by the belief-account, and the sensorimotor knowledge that informs spatial relationships entailed by the access-account. 


[1] Binocular disparity is the difference between the right and left retinal images caused by their horizontal separation (see also stereopsis).

[2] T-junctions are formed when one surface occludes the edge of another surface, causing a T-shaped intersection to appear between them. 

[3] Texture deletion occurs when an object passing in front of the background temporarily obscures the background’s texture.

[4] The elongated horse illusion is caused by slicing the silhouette of a horse in half and placing the two halves on either side of an occluding rectangle. This creates the illusion of a single elongated horse’s body (extending behind the rectangle) rather than of two horses, one with its back half occluded and the other with its front half occluded.

[5] To quote Möller (1999) directly, “Perception of space and shape is based on the anticipation of the sensory consequences of actions that could be performed by the agent, starting from the current sensory situation” (p. 171).


Möller, R. (1999). Perception Through Anticipation. A Behavior-Based Approach to Visual Perception. In: Understanding Representation in the Cognitive Sciences: Does Representation Need Reality? Ed. by Alexander Riegler, Markus Peschl, and Astrid von Stein. Springer, 169–176.

Nanay, B. (2010). Perception and imagination: amodal perception as mental imagery. In: Philosophical Studies, 150(2), 239–254. doi: 10.1007/s11098-009-9407-5.

— (2016). Imagination and Perception. In: Routledge Handbook of the Philosophy of Imagination. Ed. by Amy Kind. Routledge, 1–12.

— (2018). The Importance of Amodal Completion in Everyday Perception. In: i-Perception, 9(4), 1-16. doi: 10.1177/2041669518788887

Nichols, S. & Stich, S. (2000). A cognitive theory of pretense. In: Cognition, 74(2), 115–147. doi: 10.1016/S0010-0277(99)00070-0.

Acknowledgements: Thanks to: Dr. René Jagnow and The Classic Journal reviewers for suggestions which motivated me to clarify portions of the essay.

Citation Style: APA