The Visual Cliff Experiment: How Infants Perceive Depth

Place a crawling infant at the center of a glass-topped table. On one side the floor pattern continues right beneath the glass. On the other side the floor drops away by four feet (1.2 m), leaving the glass suspended above a visual drop. Both sides are equally safe — the glass will hold the infant's weight. Yet 92% of infants refuse to crawl over the drop side, even when their mothers stand on the far edge calling to them. Eleanor Gibson and Richard Walk's 1960 visual cliff experiment used that refusal as a window into one of developmental psychology's oldest questions: how much of what we see do we come into the world already knowing how to interpret?

The Nature-Nurture Debate That Set the Stage

For most of the 19th and early 20th centuries, two camps argued about the origins of spatial perception. The empiricist view, associated with philosophers like John Locke and psychologists like William James, held that the perceptual world is built from experience. Depth perception, on this account, is learned — the infant gradually discovers, through reaching, falling, and touching, which visual patterns correspond to actual distances and drops. James famously described the newborn's experience as a "blooming, buzzing confusion," implying that sensory integration is an achievement of development, not a birthright.

The nativist view, associated with Immanuel Kant and later with gestalt psychologists, held that the basic structure of perception is present from birth. We do not learn that some surfaces are far and others are near — the visual system is already wired to interpret certain cues as depth.

By the late 1950s, neither side had compelling experimental evidence. Infants could not report their perceptions. Animal studies were suggestive but not directly applicable to humans. Gibson and Walk's contribution was to build an apparatus that could ask the question without requiring the subject to speak.

Source: Gibson, E. J. (1969). Principles of Perceptual Learning and Development. Appleton-Century-Crofts.

The Apparatus: Engineering a Safe Precipice

Eleanor Gibson's inspiration for the visual cliff reportedly came from a hiking trip to the Grand Canyon, where she watched visitors edge nervously close to the rim and wondered when, developmentally, that aversion to heights begins.

The apparatus she and Richard Walk designed at Cornell University was structurally simple. A large sheet of thick, transparent Plexiglas was supported horizontally at table height. Directly beneath one half of the Plexiglas, a high-contrast checkerboard fabric was placed flush against the glass — this was the shallow side. On the other half, the same checkerboard fabric was placed on the floor four feet below — this was the deep side. The result was a surface that was physically identical on both sides (both solid glass, both safe) but visually different: one side looked like a normal floor, the other looked like a cliff edge with a four-foot drop below it.

The high-contrast checkerboard pattern was chosen specifically because it provides strong optical texture gradient — a visual cue for depth. As a textured surface recedes into the distance, the elements become smaller and denser. This texture gradient is one of the primary monocular cues humans use to judge distance and surface layout.

In the center of the apparatus was a starting platform where the infant or animal subject was placed. The subject's mother (for human infants) or a food reward (for animals) was placed on the shallow side or the deep side alternately, to see whether the subject would cross in either direction.

Source: Gibson, E. J., & Walk, R. D. (1960). The "visual cliff." Scientific American, 202(4), 64–71.

The 1960 Results: Infants and Animals

Gibson and Walk tested 36 human infants between 6 and 14 months of age — specifically those who were old enough to crawl. The minimum age of crawling was a methodological necessity: the experiment required the infant to choose to cross or not, and that requires the ability to move independently.

The results were unambiguous. When mothers called to their infants from the shallow side, almost all infants crossed without hesitation. When mothers called from the deep side, 27 of the 36 infants refused to cross. Several peered over the edge and retreated. Some patted the glass on the deep side — confirming by touch that it was solid — yet still refused to cross. Nine infants did cross to the deep side, but many showed visible distress before doing so.

The cross-species results were particularly informative because they allowed comparison across animals with very different developmental timelines:

Animal	Age at Test	Result on Deep Side
Human infants	6–14 months (crawling age)	92% refused to cross
Chicks	1 day old	100% avoided deep side
Lambs and kids (goats)	1 day old	100% avoided deep side
Kittens	First day of locomotion	Avoided deep side
Rats	Varies	Crossed freely if raised in dark; avoided if raised in light
Aquatic turtles	Varies	Crossed to deep side at moderate rate

The chick and lamb data were striking. These animals avoid the cliff within hours of birth, before any learning from falling could have occurred. This strongly suggested that for precocial species — animals that are mobile immediately at birth — depth perception is either innate or learned within an implausibly short window.

The rat data added nuance. Rats raised in darkness — without visual experience — crossed the deep side readily. Rats raised in normal lighting conditions avoided it. This suggested that for rats, depth perception via visual cues does require some light exposure, though perhaps very little.

The turtle data suggested that depth perception might be calibrated to ecological needs. Aquatic turtles spend most of their lives in water and have less need to fear drops. They showed less cliff avoidance than strictly terrestrial species.

Source: Gibson, E. J., & Walk, R. D. (1960). The "visual cliff." Scientific American, 202(4), 64–71.

What the Infant is Using: The Visual Cues for Depth

The visual cliff works because it exploits how the visual system interprets texture gradients and binocular disparity to perceive depth. Understanding what information the infants were using helps clarify what the experiment actually measured.

Texture gradient — the progressive compression of surface texture elements as a surface recedes from the viewer. A tiled floor viewed at an angle shows large tiles nearby and smaller tiles farther away. The visual system interprets increasing texture density as increasing distance.

Binocular disparity — the slight difference between the images projected onto the two retinas, due to the horizontal separation of the eyes. The brain computes depth from this disparity through a process called stereopsis.

Motion parallax — as the observer moves, nearer surfaces shift across the visual field more than farther surfaces. This relative motion provides depth information even without binocular vision.

The visual cliff's checkerboard pattern provides strong texture gradient information on the deep side: the checkerboard pattern visible through the glass is smaller, denser, and farther away. Infants as young as 6 months appear to be sensitive to this information. But the experiment did not isolate which specific depth cues the infants were relying on. Subsequent research using versions of the apparatus that could be modified to remove specific cues helped answer this.

Research by Richard Held and colleagues in the 1970s found that binocular depth perception, specifically stereopsis, develops between about 3 and 5 months in most infants — earlier than crawling age. This suggested that the perceptual capacity for depth discrimination might be present before infants can demonstrate it behaviorally on the visual cliff.

Source: Held, R., Birch, E., & Gwiazda, J. (1980). Stereoacuity of human infants. Proceedings of the National Academy of Sciences, 77(9), 5572–5574.

The Heart Rate Finding: Fear Versus Wariness

One important limitation of the original Gibson and Walk study was that refusal to cross the deep side could mean two different things: the infant perceives depth, or the infant perceives depth and is afraid of it. These are related but distinct. Depth perception does not automatically require fear. A person in a glass-bottomed skywalk can perceive the drop clearly without experiencing fear, because they trust the surface. The original experiment conflated the two.

Later research by Joseph Campos and colleagues addressed this by measuring heart rate while infants were lowered over the deep and shallow sides without any choice about crossing.

The results divided into an age-related pattern:

• Infants around 2 months old showed a decrease in heart rate when placed over the deep side. Heart rate deceleration is associated with attentional orienting — interest and attention, not fear.
• Infants around 9 months old showed a heart rate increase over the deep side. Heart rate acceleration is associated with fear and arousal.

This was an important refinement. Very young infants discriminate the deep from the shallow side — they respond differently to it, attending more to it. But they do not show the physiological signature of fear. Older infants, who have had months of experience with falling, catching themselves, and observing the consequences of drops, do show fear.

This suggested that depth perception itself might be early and possibly innate, but the emotional response of fear to perceived heights develops through experience. The baby doesn't need to learn to see the cliff, but it may need some experience to learn that cliffs are dangerous.

Source: Campos, J. J., Langer, A., & Krowitz, A. (1970). Cardiac responses on the visual cliff in prelocomotor human infants. Science, 170(3954), 196–197.

Locomotion as the Trigger for Fear

A further finding emerged from studies that compared infants who had begun crawling with age-matched infants who had not yet started. Even controlling for age, crawling experience predicted cliff avoidance. Infants who had been crawling for more weeks showed stronger avoidance of the deep side than infants of the same chronological age who had not yet begun crawling.

This suggested that the onset of independent locomotion — the experience of moving through space and encountering its hazards — calibrates the emotional response to perceived drops, independent of the perceptual capacity to detect them.

Campos later conducted a longitudinal study in which some pre-crawling infants were given experience moving in wheeled walkers. Infants who had more locomotor experience via walkers showed earlier and stronger avoidance of the deep side than infants with less locomotor experience. The perceptual system appears to use movement through the world to determine which perceived features are actually dangerous.

This is consistent with an ecological view of perception developed by Gibson herself: perception is not a passive recording of the environment but an active process tuned by the organism's engagement with its environment. The infant's perceptual system learns not through abstract contemplation of depth cues but through the ongoing feedback loop of moving, acting, and registering the consequences.

Source: Bertenthal, B. I., Campos, J. J., & Kermoian, R. (1994). An epigenetic perspective on the development of self-produced locomotion and its consequences. Current Directions in Psychological Science, 3(5), 140–145.

Social Referencing: What Happens When Mother Looks Scared

A separate line of research extended the visual cliff paradigm to study social referencing — the use of another person's emotional expression to guide behavior in ambiguous situations.

In 1985, James Sorce and colleagues placed infants on the visual cliff in a situation of ambiguity: the deep side was modified to be shallower than in the original study — about 30 centimeters (12 inches) rather than 1.2 meters — a drop that is genuinely ambiguous in its dangerousness. Infants' mothers were positioned on the far side and instructed to display one of several facial expressions: joy, interest, fear, anger, or sadness.

The results were striking:

Mother's Expression	% of Infants Crossing to Deep Side
Joy	74%
Interest	73%
Sadness	33%
Anger	11%
Fear	0%

When mothers displayed fear, not a single infant crossed. When mothers displayed joy or interest, nearly three-quarters crossed. The infants were using their mothers' emotional signals as information about the safety of the environment — calibrating their behavior to the social cue rather than to their own perceptual assessment of the risk.

This finding became foundational in research on how infants use social information to learn about the world. It showed that the relationship between perception and action is not purely individual — the social environment provides a second layer of information that modifies how perceived features are translated into behavior.

Source: Sorce, J. F., Emde, R. N., Campos, J. J., & Klinnert, M. D. (1985). Maternal emotional signaling: Its effect on the visual cliff behavior of 1-year-olds. Developmental Psychology, 21(1), 195–200.

The Gibson Ecological Framework

Eleanor Gibson's work on the visual cliff was part of a larger theoretical project. She was a leading proponent of what became known as the ecological approach to perception, which she developed in dialogue with (and sometimes in contrast to) her husband, the psychologist James J. Gibson.

The ecological approach holds that perception should be understood not as the brain building an internal model of the world from fragmentary sensory input, but as the organism directly picking up affordances — the possibilities for action that the environment offers. A surface that affords support for locomotion is perceived as solid. A drop that affords injury is perceived as dangerous. The visual system is tuned by evolution and development to detect affordances directly, not to process raw pixels and infer their meaning.

Affordance — a property of the environment defined relative to an organism's action capabilities. A step is an affordance for climbing if the organism can climb it; a drop is a hazard affordance if falling from it would injure the organism.

The visual cliff, from this perspective, is interesting not because it produces fear but because it demonstrates that infants perceive the drop as lacking the affordance for safe locomotion — they perceive, correctly, that crawling over the edge would not be supported. Whether or not they are afraid of it is a secondary question. The primary perceptual achievement is recognizing that the deep side does not offer safe footing.

Source: Gibson, E. J. (1988). Exploratory behavior in the development of perceiving, acting, and the acquiring of knowledge. Annual Review of Psychology, 39, 1–41.

Limitations and Legacy

The visual cliff experiment has several limitations that researchers have identified over the decades.

The most frequently cited is the locomotion requirement. The study could not test infants younger than crawling age because it required voluntary movement. The heart rate studies partially addressed this, but they measure something different — emotional response — rather than the behavioral choice that gave the original experiment its clarity.

The apparatus is also limited to a specific class of depth cues (optical texture gradient and, in some versions, binocular disparity). It cannot isolate which specific cue the infant is using, or determine the relationship between different depth cue systems.

Despite these limitations, the visual cliff remains one of developmental psychology's canonical studies. It has been replicated hundreds of times, across species, cultures, and methodological variants. The basic findings have held: crawling human infants avoid the deep side, the avoidance correlates with locomotor experience, heart rate data distinguish early perception from later fear, and social referencing modulates the behavioral response.

More practically, the visual cliff gave developmental psychologists a reliable, non-verbal method for studying perceptual capacities in pre-linguistic infants — a methodological template that has been extended to study a wide range of perceptual and cognitive capacities in early development.

Source: Adolph, K. E. (2000). Specificity of learning: Why infants fall over a veritable cliff. Psychological Science, 11(4), 290–295.

Frequently Asked Questions

Q: How could the experiment be safe if it involved a four-foot drop?
A: The drop was entirely visual, not physical. The apparatus consisted of a thick sheet of transparent Plexiglas covering the entire surface. The checkerboard fabric was placed four feet below the Plexiglas on the deep side, creating the visual impression of a drop while the glass surface remained flat and solid throughout. No infant was in any physical danger.

Q: Does the visual cliff show that depth perception is innate?
A: It provides evidence consistent with early depth perception, particularly in species like chicks and goats that avoid the cliff within hours of birth. For human infants, the evidence is more nuanced: the perceptual capacity for depth discrimination appears to develop in the first few months of life, earlier than the behavioral avoidance measured by the cliff. Fear of heights appears to require locomotor experience. So the answer is: depth perception emerges early and may have significant innate components, but the emotional and behavioral response to it is calibrated through experience.

Q: Why did Gibson and Walk use a checkerboard pattern specifically?
A: The high-contrast checkerboard provides a strong optical texture gradient — as the surface recedes, the check elements appear smaller and closer together. This gradient is one of the most powerful monocular depth cues. It ensures that the visual information for depth is available even to subjects who may lack fully developed binocular disparity processing.

Q: What does the social referencing finding tell us about how infants learn about danger?
A: It tells us that infants do not assess risk purely on the basis of their own perceptual processing. From around 8 to 12 months, infants actively seek information from caregivers' emotional expressions when they encounter ambiguous situations, and they use that information to guide their behavior. A mother's fear expression can prevent crossing even when the infant's own perceptual assessment might not clearly signal danger.

Q: Why could rats raised in darkness cross the visual cliff without fear?
A: Rats reared in complete darkness lack visual experience but also lack the specific visual experience of surfaces and drops that would calibrate their depth perception system. Their ability to use optical texture gradient information to detect depth appears to require some light exposure, suggesting a role for early visual experience in developing this specific perceptual capacity.

Q: How does the visual cliff relate to theories of perceptual development?
A: It is central to the debate between nativist and empiricist accounts. The evidence from precocial animals (chick, lamb, goat) supports nativist claims. The heart rate data showing that very young human infants detect but do not fear the deep side, combined with evidence that locomotor experience calibrates cliff avoidance, supports a position between the two poles: some perceptual capacity is present early, but full behavioral calibration of risk requires experience.

Q: Did Gibson ever work with infants too young to crawl?
A: Directly measuring behavior on the cliff required locomotion, so the apparatus itself could not be used with pre-crawling infants. However, the heart rate paradigm developed by Campos and colleagues allowed testing of younger infants by passively lowering them over deep and shallow sides. This approach extended the visual cliff research to infants as young as 2 months.

The Takeaway

The visual cliff gave developmental psychology something rare: a physical apparatus that could ask a deep question and get a clear answer. The question was whether infants perceive depth before experience has taught them to. The answer, refined across decades of follow-up research, is that the perceptual capacity for depth discrimination develops early — possibly partly from innate mechanisms, certainly before the infant can demonstrate it behaviorally. The fear of the drop builds more slowly, through the experience of locomotion and the feedback loop of moving through a world that pushes back. And once locomotion begins, the infant does not navigate risk alone: the social environment, especially the caregiver's face, becomes a second source of information about what is dangerous and what is safe.

---

Read next:

• The False Belief Task
• The Little Albert Experiment
• The McGurk Effect Experiment