Distance Perception

Abstract

Distance perception is a form of depth perception: the recovery of absolute egocentric distance — how far a target lies from the observer, scaled to the body so that action can be deployed correctly. Because the two-dimensional retina discards absolute distance, the visual system infers it from a reliability-weighted combination of cues, each dominant over its own range, from binocular disparity near to angular declination far. Gibson's ground theory holds that distance is carried outward along the terrain surface, while blind-walking studies reveal a dissociation between biased verbal judgments and accurate visually directed action. A contested embodied account adds that effort and intent rescale apparent distance, and applied work finds egocentric distance systematically compressed in virtual reality. Three interactive demonstrations let the reader weight cues, read distance off angular declination, and rescale a scene.

Keywords: distance perception, egocentric distance, angular declination, cue combination, depth perception

What Distance Perception Is

Distance perception is the recovery of egocentric distance: the straight-line separation between the observer's vantage point and a target, expressed in a scale the observer can act in. It is a special case of depth perception, which encompasses any judgment about the three-dimensional layout of a scene, but it carries an extra demand. A great many depth judgments are relative and ordinal — this edge occludes that one, so it is nearer; this texture is finer, so it recedes — and these can be made from the image alone, without ever assigning a number. Distance perception proper is absolute and metric: it asks not which of two things is nearer but how far the target is, scaled to the observer's body so that the hand, the foot, or the brake can be deployed correctly. This observer-referenced quantity is distinct from exocentric distance, the separation between two external objects judged independently of the viewer.

That scaling is the crux. The retinal image is a projection onto a two-dimensional surface, and projection discards exactly the information distance perception needs: a small near object and a large far object can cast the identical image. To break this ambiguity the visual system must fold in information beyond the image's instantaneous geometry — the state of the two eyes, the structure of motion over time, the regularities of a terrain surface, and prior knowledge of object sizes. Cutting and Vishton (1995) catalogued these sources of information about depth and, crucially, quantified how their usefulness changes with range, reframing the old list of pictorial cues as a set of partially redundant channels each with its own effective zone.

Depth Cues and Their Effective Range

The classical inventory separates cues into binocular (requiring two eyes) and monocular (available to one), and into pictorial (present in a static image) and dynamic (requiring motion or time). Binocular disparity — the small difference between the two eyes' images — and vergence — the inward rotation of the eyes to fixate a near point — supply strong metric information but only where the eyes' geometry still varies appreciably with distance, which is within roughly two metres. Motion parallax, the differential image motion of near and far points as the head translates, and relative and familiar size operate across a broader middle band. Angular declination and texture-density gradients on a ground surface extend furthest, carrying usable distance information out to the horizon. The extraretinal signals have drawn particular scrutiny: Foley (1980) modelled how binocular distance perception depends jointly on vergence and disparity and why the resulting estimate is systematically distorted, and Mon-Williams and Tresilian (1999) and Tresilian, Mon-Williams, and Kelly (1999) examined how far the vergence angle alone can specify near distance, concluding that it is a weak absolute cue easily dominated by the others.

Cutting and Vishton's (1995) central contribution was to make this range-dependence explicit. They argued that the visual system does not weight cues by a fixed hierarchy but by their reliability at the current distance, so that the dominant cue for a target at arm's length (binocular disparity) is nearly useless for a target across a field, where declination dominates. This reframing anticipated the modern statistical account, in which the brain combines cues in proportion to their reliability — an idea given a formal Bayesian treatment by Scarfe and Hibbard (2025), who model cue integration as reliability-weighted inference and show where the normative scheme predicts the biases that observers actually display.

The Ground Theory of Distance

Gibson (1950) rejected the idea that space is perceived as empty extension punctuated by objects and proposed instead that distance is perceived along surfaces, pre-eminently the ground. On his account the continuously textured terrain running from the observer's feet to the horizon is the substrate that carries distance outward; an object's distance is registered by where it meets the ground, not by any property of the object hanging in a void. This ground theory makes a sharp prediction: distance to an object resting on the terrain should be well judged, while distance to an object detached from the ground — floating, or seen across a gap in the surface — should be judged poorly, because the carrying surface is interrupted.

The cue that implements ground-theoretic distance is angular declination: the angle by which the line of sight to a ground-level target falls below the horizon (equivalently, the eye-height-referenced angular declination below the horizon). Given the observer's eye height, declination determines distance by simple trigonometry, and the visual system appears to use it. Ooi, Wu, and He (2001) showed that experimentally manipulating the perceived horizon with prisms shifted judged distance in exactly the direction declination predicts, and Wu, Ooi, and He (2004) demonstrated that when the ground surface is textured only in patches, the visual system integrates distance sequentially across the near-to-far surface — a “sequential surface integration process” — so that a gap in near texture degrades far judgments. Both findings place the ground surface, not the isolated target, at the centre of egocentric distance.

Perception and Visually Directed Action

A target's verbally reported distance and the distance an observer will actually walk to it with eyes closed need not agree. Loomis, Da Silva, Fujita, and Fukusima (1992) found that observers systematically underestimate egocentric distances in explicit verbal or visual matching tasks, foreshortening far distances relative to near ones, yet when asked to view a target and then walk to it blindfolded — blind walking, a visually directed action — they stop with remarkable accuracy out to twenty metres and beyond. The dissociation suggests that the information guiding action is better calibrated than the information available to conscious report, or that the two tap partially separate representations of the same scene. The underlying estimate is also computed quickly: Gajewski, Philbeck, Pothier, and Chichka (2010) found that a glimpse of only a fraction of a second is enough to support accurate blind walking, so the ground-based distance signal is extracted almost as soon as the scene is available.

This result reframed the measurement problem. Because explicit judgments are biased, a verbal estimate is a poor index of what the perceptual-motor system “knows,” and blind walking (and its variants — triangulated walking, blind throwing) became the preferred action measure of perceived egocentric distance. The dissociation is also the empirical wedge for later embodied claims: if effort or intent can move the explicit estimate without moving the action calibration, the two measures come apart in a theoretically diagnostic way.

The Embodied Account

Proffitt (2003) proposed that perceived distance and slant are scaled not only by optical cues but by the observer's bioenergetic state and intended action — an economy-of-action view in which a hill looks steeper when one is tired or carrying a load, and a target looks farther when reaching it will cost more effort. On this account perception is not a neutral readout of geometry but a representation tuned to behavioural relevance, discounting or inflating apparent distance according to the metabolic price of acting on it.

The embodied interpretation has been vigorously contested. Durgin and colleagues have argued that many effort effects are artefacts of experimental demand and of the explicit measures used, reappearing only when observers can guess the hypothesis and vanishing under action measures or better controls. The debate is unresolved and has become a case study in how an explicit-judgment bias can be read either as a genuine perceptual effect or as a response bias — a distinction the perception-action dissociation makes it hard to settle by verbal report alone.

Figure

Interactive Demonstrations

Each distance cue is reliable only over part of the range. Move the target and watch the dominant cue change: disparity near, declination far, motion and size bridging the middle.

Dominant cue at 1.5 m: Angular declination. The combined estimate is a reliability-weighted blend, not a fixed hierarchy — the modern Bayesian reading of Cutting and Vishton's range-dependent weighting.

The first demonstration plots the relative reliability of four distance cues — binocular disparity, motion parallax, relative size, and angular declination — as a function of target distance. Drag the distance slider and watch the dominant cue change: disparity carries the near field, declination the far field, with motion and size bridging the middle. The combined estimate is a reliability-weighted blend, illustrating Cutting and Vishton's (1995) range-dependent weighting and the Bayesian integration formalised by Scarfe and Hibbard (2025).

For a target on flat ground, distance d = h / tanγ, where γ is the angle below the horizon and h is eye height. Inject a horizon error to see the far field foreshorten.

With h = 1.6 m and d = 8.0 m, γ = 11.31°. A horizon read 1° too high gives a perceived 8.80 m (+10%); 1° too low gives 7.33 m (−8%). Because d = h/tanγ is nonlinear, the same angular error foreshortens the far field most — the worked example in the text.

The second demonstration implements Figure 1 as a manipulable scene. Set the eye height and move the target along the ground; the demo reports the declination angle and the distance it implies, and can inject an error into the registered horizon to show how a small declination misregistration foreshortens the far field — the mechanism behind ground-based distance compression.

Head-mounted displays compress egocentric distance: observers act as though the scene were a fixed fraction of its intended depth. Set the compression gain and the intended range.

At 75% gain, a 6.0 m target is perceived at 4.5 m, so an observer walking blind stops 1.5 m short. The typical reported gain is 0.7–0.8, and it persists with modern wide-field headsets — the distortion is perceptual, not a matter of display resolution.

The third demonstration contrasts an intended scene geometry with the geometry an observer perceives under a compression gain, the hallmark distortion of head-mounted displays. Adjust the gain and the target range to see how a uniform multiplicative compression (typically 0.7–0.8) shifts walked distance, reproducing the pattern reported across the virtual-distance literature (Renner, Velichkovsky, and Helmert, 2013; Creem-Regehr, Stefanucci, and Thompson, 2023).

Cue Summary

Worked Example

Consider an observer whose eye height is h = 1.6 m viewing a target resting on flat ground at a true horizontal distance of d = 8 m. The angular declination below the horizon is γ = arctan(h / d) = arctan(1.6 / 8) = arctan(0.20) = 11.31°.

Now suppose the visual system misregisters the horizon as 1° too high, so the effective declination used for the distance computation is 11.31° − 1° = 10.31°. The perceived distance becomes d′ = h / tan(10.31°) = 1.6 / 0.1819 = 8.80 m — the target is seen as roughly 10% farther than it is.

Reverse the error — a horizon registered 1° too low, effective declination 12.31° — and the perceived distance collapses to d′ = 1.6 / tan(12.31°) = 1.6 / 0.2182 = 7.33 m, about 8% nearer. The asymmetry matters: because distance varies as h/tanγ and tan is nonlinear, a fixed angular error produces a larger absolute distance error the farther the target lies, which is precisely why ground-based far-distance judgments foreshorten. The DeclinationDemo re-derives these same numbers as the horizon-error slider is moved, so the prose and the demonstration agree.

Discussion

Egocentric distance perception is best understood not as the output of one mechanism but as a reliability-weighted synthesis that shifts its basis with range, scaled by the observer's body and — on some accounts — by the cost of acting. The ground theory supplies the dominant far-field mechanism and explains a signature failure mode: detach a target from the terrain and the carrying surface can no longer deliver its distance. The perception-action dissociation warns that any single measure is partial, and that the biases so easily elicited by verbal report may not reflect the calibration that actually guides behaviour. These threads converge on a view of distance as an action-referenced quantity whose errors are lawful rather than random.

The embodied debate remains the field's sharpest open question, because it turns on whether a measured bias is perceptual or decisional — a question the dissociation between explicit judgment and visually directed action makes genuinely hard to resolve. That same methodological fault line runs through the applied literature, where virtual environments compress distance in ways that both stress-test theories of real-world perception and demand practical correction.

Current Directions

The most active current front is egocentric distance in virtual and augmented reality, where observers walk to targets as though the scene were compressed to roughly 70–80% of its intended depth. Early work established the effect and ruled out several trivial explanations (Thompson et al., 2004); Renner, Velichkovsky, and Helmert (2013) reviewed the accumulated evidence and candidate causes, from display field of view to the quality of the virtual ground plane and avatar self-representation. The effect has proven stubborn: Feldstein, Kölsch, and Konrad (2020) found compression persisting even with wide-field modern headsets, and Kelly (2023) documented that although newer consumer displays have narrowed the gap, systematic underperception remains. Creem-Regehr, Stefanucci, and Thompson (2023) synthesised two decades of this work, arguing that virtual distance perception is a window on the same ground-based and embodied mechanisms that operate in the real world, and that closing the compression gap requires getting the virtual ground plane and self-scaling right rather than simply widening the display. Parallel to the applied work, the Bayesian cue-integration programme (Scarfe and Hibbard, 2025) is sharpening the normative question of how reliability-weighting should combine metric and ordinal cues, and where the brain departs from the optimum.

Common Misconceptions

Glossary

Key Researchers

Sarah H. Creem-Regehr

(ORCID 0000-0001-7740-1118). Professor of Psychology at the University of Utah, whose work on space perception and spatial cognition spans real and virtual environments and has helped establish virtual distance perception as a window on real-world mechanisms.

Frank H. Durgin

. Elizabeth and Sumner Hayward Professor of Psychology at Swarthmore College, known for careful analyses of the systematic biases in explicit distance and slant judgments and for the methodological critique of embodied effort effects.

James J. Gibson

(1904–1979). Founder of the ecological approach to visual perception; introduced the ground theory of space perception and the gradient of texture density as a source of distance information.

Jack M. Loomis

. Research Professor Emeritus in Psychological and Brain Sciences at the University of California, Santa Barbara, whose blind-walking studies revealed the dissociation between explicit distance judgments and visually directed action.

Teng Leng Ooi

. Professor in the College of Optometry at the Ohio State University, whose work with Zijiang J. He established the angular-declination and sequential-surface-integration account of ground-based distance.

Dennis R. Proffitt

. Commonwealth Professor of Psychology Emeritus at the University of Virginia, author of the economy-of-action account on which perceived distance and slant are scaled by bioenergetic cost and intended action.

Frequently Asked Questions

What is the difference between distance perception and depth perception?

Depth perception is the broad recovery of three-dimensional layout, including the relative, ordinal ordering of surfaces. Distance perception is the narrower, harder problem of recovering absolute, metric egocentric distance — how far a target is, scaled to the observer's body so that action can be deployed correctly.

Why can't we just read distance off the retinal image?

Because projection onto the two-dimensional retina discards absolute distance: a small near object and a large far object can produce identical images. The visual system must combine image information with calibrating sources — eye geometry, motion, terrain structure, and known object sizes — to recover a metric estimate.

Which cue matters most for judging how far away something is?

It depends on range. Binocular disparity and vergence dominate within about two metres, motion parallax and relative size across the middle distances, and angular declination on the ground plane out to the horizon. The visual system weights each cue by its reliability at the current distance.

What is the ground theory?

Gibson's proposal that distance is perceived along the continuously textured terrain surface from the observer's feet to the horizon, rather than as empty space containing objects. It predicts that distance to objects resting on the ground is well judged while distance to objects detached from the ground is not.

How do researchers measure perceived distance if verbal judgments are biased?

Chiefly through visually directed action, especially blind walking: the observer views a target, then walks to it with eyes closed. Because action can be better calibrated than explicit report, these measures reveal accuracy that verbal estimates miss, and they expose the dissociation between judgment and action.

Does being tired really make things look farther away?

Proffitt's economy-of-action account holds that effort and bioenergetic state rescale perceived distance and slant, so a target reached at greater cost looks farther. The claim is contested: critics, notably Durgin, argue that many such effects reflect experimental demand and explicit-measure artefacts rather than genuine perceptual change.

Why do things look closer or farther in virtual reality?

Egocentric distance is systematically compressed in head-mounted displays — observers act as though the scene were 70–80% of its intended depth. Resolution is not the cause; the leading factors are the rendered ground plane, the field of view, and the observer's self-representation in the virtual scene.

Is distance compression in VR getting better with newer headsets?

Partly. Newer wide-field, high-resolution consumer displays have narrowed the gap, but systematic underperception still remains, indicating that the distortion is rooted in perceptual mechanisms — the virtual ground plane and self-scaling — rather than in display hardware alone.

References

Creem-Regehr, S. H., Stefanucci, J. K., & Thompson, W. B. (2023). Perceiving distance in virtual reality: Theoretical insights from contemporary technologies. Philosophical Transactions of the Royal Society B: Biological Sciences, 378(1869), 20210456. https://doi.org/10.1098/rstb.2021.0456

Cutting, J. E., & Vishton, P. M. (1995). Perceiving layout and knowing distances: The integration, relative potency, and contextual use of different information about depth. In W. Epstein & S. Rogers (Eds.), Perception of space and motion (pp. 69–117). Academic Press. https://doi.org/10.1016/B978-012240530-3/50005-5

Feldstein, I. T., Kölsch, F. M., & Konrad, R. (2020). Egocentric distance perception: A comparative study investigating differences between real and virtual environments. Perception, 49(9), 940–967. https://doi.org/10.1177/0301006620951997

Foley, J. M. (1980). Binocular distance perception. Psychological Review, 87(5), 411–434. https://doi.org/10.1037/0033-295X.87.5.411

Gajewski, D. A., Philbeck, J. W., Pothier, S., & Chichka, D. (2010). From the most fleeting of glimpses: On the time course for the extraction of distance information. Psychological Science, 21(10), 1446–1453. https://doi.org/10.1177/0956797610381508

Gibson, J. J. (1950). The perception of the visual world. Houghton Mifflin. 📖

Kelly, J. W. (2023). Distance perception in virtual reality: A meta-analysis of the effect of head-mounted display characteristics. IEEE Transactions on Visualization and Computer Graphics, 29(12), 4978–4989. https://doi.org/10.1109/TVCG.2022.3196606

Loomis, J. M., Da Silva, J. A., Fujita, N., & Fukusima, S. S. (1992). Visual space perception and visually directed action. Journal of Experimental Psychology: Human Perception and Performance, 18(4), 906–921. https://doi.org/10.1037/0096-1523.18.4.906

Mon-Williams, M., & Tresilian, J. R. (1999). Some recent studies on the extraretinal contribution to distance perception. Perception, 28(2), 167–181. https://doi.org/10.1068/p2737

Ooi, T. L., Wu, B., & He, Z. J. (2001). Distance determined by the angular declination below the horizon. Nature, 414(6860), 197–200. https://doi.org/10.1038/35102562

Proffitt, D. R., Stefanucci, J., Banton, T., & Epstein, W. (2003). The role of effort in perceiving distance. Psychological Science, 14(2), 106–112. https://doi.org/10.1111/1467-9280.t01-1-01427

Renner, R. S., Velichkovsky, B. M., & Helmert, J. R. (2013). The perception of egocentric distances in virtual environments — A review. ACM Computing Surveys, 46(2), Article 23. https://doi.org/10.1145/2543581.2543590

Scarfe, P., & Hibbard, P. B. (2025). A Bayesian framework for the reliability-weighted combination of depth cues. PLOS Computational Biology, 21(9), e1013506. https://doi.org/10.1371/journal.pcbi.1013506

Thompson, W. B., Willemsen, P., Gooch, A. A., Creem-Regehr, S. H., Loomis, J. M., & Beall, A. C. (2004). Does the quality of the computer graphics matter when judging distances in visually immersive environments? Presence: Teleoperators and Virtual Environments, 13(5), 560–571. https://doi.org/10.1162/1054746042545292

Tresilian, J. R., Mon-Williams, M., & Kelly, B. M. (1999). Increasing confidence in vergence as a cue to distance. Proceedings of the Royal Society B: Biological Sciences, 266(1414), 39–44. https://doi.org/10.1098/rspb.1999.0601

Wu, B., Ooi, T. L., & He, Z. J. (2004). Perceiving distance accurately by a directional process of integrating ground information. Nature, 428(6978), 73–77. https://doi.org/10.1038/nature02350