Conflicting cues make depth judgments unstable

What it is

Occlusion, pictorial cues, and binocular disparity normally agree with each other in natural scenes, all pointing to the same depth outcome. But when these cues are artificially made to contradict each other — say occlusion implies A is in front while perspective implies B is — depth judgment doesn't settle on some averaged value. It becomes unstable: either one cue forcibly dominates and the other is ignored, or perception flips back and forth between the two interpretations over time. This phenomenon is depth perception under cue conflict.

This entry is specifically about that conflict state itself, not about how any single cue behaves on its own — a single cue, or several mutually consistent cues, is the normal, stable case.

Why it happens

When the visual system integrates multiple depth cues, it generally weights each one by how reliable it seems in the current scene (something like a reliability-weighted integration process). When cues agree, this integration produces a combined depth judgment more precise and stable than any single cue alone. But when cues strongly contradict each other, this integration process runs into trouble: it implicitly assumes all cues arise from the same real physical scene, and once that assumption is broken, there's no good way to "split the difference." The result shows up either as cue dominance — one normally more-reliable cue, such as occlusion or disparity, overrides the others and determines the final percept — or as bistability, where perception flips back and forth over time between the two contradictory interpretations, similar to classic reversible figures.

Studying it

• Classic cue-conflict paradigms: setting binocular disparity against a pictorial cue (or occlusion against perspective) and measuring which cue the final percept favors, along with judgment reaction time, variability, and whether the percept flips over time.
• Context-dependence of variables and outcomes: this line of research consistently finds that which cue "wins" depends heavily on the specific scene and the magnitude of the conflict — it is not a fixed ranking of cue strength, and the same pair of cues can produce different dominance outcomes at different conflict magnitudes; both experimental design and result interpretation need to specify the exact conditions.
• Typical independent variables: which pair of cues is in conflict, conflict magnitude, presentation duration.
• Typical dependent variables: perceived depth/order judgment outcome, judgment reaction time and variability, rate of perceptual flipping under bistable conditions.

Where it stops holding

• This instability specifically requires multiple cues that would normally agree being made to contradict each other — it doesn't apply to an ordinary scene with only one or two mutually consistent cues, which is the normal case for any single cue on its own.
• A finding of "which cue wins" from one study is a result of that specific cue pairing and conflict magnitude — it is context-dependent and should not be treated as a universal cue-strength ranking that transfers directly to a different design scenario.
• This is different from ambiguity caused by insufficient cues: cue conflict requires each cue to individually convey clear but contradictory information. If a scene simply lacks enough cues altogether (no shadow, no occlusion at all), the resulting ambiguous judgment is a matter of insufficient information, not the conflict discussed here.

Applying it

• When compositing or rendering a scene that mixes real and synthetic depth cues (an AR overlay, a 3D-rendered UI on top of a live camera feed, mismatched shadow direction versus perspective direction in the same view), check that the cues you control are mutually consistent — don't assume users will automatically "average" their way to the correct result; cue conflict produces perceptual confusion, not a smooth compromise.
• When a cue conflict can't be avoided (e.g. an AR object that's physically far away but rendered with near-field disparity), anticipate and test for perceptual instability or misjudgment rather than assuming users will interpret it correctly.
• Verification: have users make relative-depth or front-back-order judgments on the actual, fully composited cue set, and watch for hesitation, errors, or a reported sense that "something's off" — rather than only checking each cue individually in isolation.

Related

• Same group: A1.13.1 Occlusion is the strongest and most stable depth cue · A1.13.2 Shading, size, and perspective provide relative depth · A1.13.3 Binocular disparity is only effective at close range
• Nearby: A1.27 Perceptual constancy
• Search terms: depth cue conflict · cue integration · cue dominance · bistable perception