Occlusion is the strongest and most stable depth cue
Aliases: interposition · T-junction
What it is
When one object's contour is interrupted by another overlapping it, the visual system immediately infers that the interrupted object sits behind and the one interrupting it sits in front — this depth cue, based on which contour continues and which breaks, is called occlusion (also interposition). It only tells you the front-to-back order, not how far apart the two objects actually are — it's ordinal depth information, not metric.
Occlusion is unusual among depth cues because it barely depends on anything else — no need to know an object's real size, the direction of lighting, or binocular disparity; a contour interrupted by another is enough. That's why it holds up where other cues relying on lighting direction, true object size, or binocular disparity weaken or fail outright: black-and-white line drawings, monocular viewing, a single static frame all leave occlusion fully intact.
Why it happens
The visual system reads front-back order from a local geometric structure: where two edges meet at a T-junction, the edge that passes through the junction uninterrupted belongs to the nearer surface, and the edge that terminates at the junction belongs to the farther, occluded surface. This judgment only requires analyzing local contour topology — it doesn't need to integrate size, lighting direction, or binocular information, so the processing sits at a low level and runs fast, close to automatic.
Because the judgment relies on a single local feature — a contour interruption — occlusion works under almost any presentation condition: a single static image, a plain line drawing, or monocular viewing (one eye covered) all leave T-junction structure intact, so the occlusion cue still holds. This is unlike cues that depend on assumptions about lighting direction and true object size, or binocular disparity, which requires a specific physiological setup — which is why occlusion is considered the strongest and most stable depth cue.
Studying it
- T-junction manipulation experiments: artificially creating or removing T-junction structure in a figure and observing whether front-back judgments appear or disappear accordingly, confirming that the T-junction itself is what triggers the occlusion judgment.
- Cue-conflict experiments: pitting occlusion against another depth cue (relative size, motion parallax) so the two imply opposite front-back orders, then measuring which cue ends up dominating the participant's judgment — occlusion typically wins such conflicts, since it is the most reliable of the ordinal cues.
- Typical independent variables: presence or absence of a T-junction, whether occlusion conflicts with another cue.
- Typical dependent variables: front-back judgment outcome, judgment reaction time, judgment consistency under ambiguous figures.
Where it stops holding
- Occlusion only conveys order, not distance: knowing A is in front of B says nothing about whether they're a meter or ten meters apart — metric depth still requires other cues.
- Occlusion only applies where there is a visible contour interruption: if two objects are spatially separated with no touching edges, there's no T-junction to judge from, and occlusion provides no depth information at all in that case.
- Stacked UI elements (cards, modals) that only hint at layering through shadow or blur, without any actual edge overlap between elements, give the viewer a noticeably weaker layering cue than a design with real contour overlap — a shadow is a supplement to an occlusion relationship, not a substitute for actual overlap.
Applying it
- Signal front-back layering with real element overlap, not just shadow or blur: a card sitting on top of a background, a modal covering a dimmed backdrop — these designs exploit the strongest available depth cue directly, and are one of the few "3D" cues that survive fully intact on a flat 2D screen.
- Make sure an element meant to read as "on top" actually visually covers the element beneath it, rather than relying only on a drop-shadow gradient or slight blur to imply layering — occlusion is the dominant, most reliable cue, and a layering relationship expressed only through weaker cues is easy to miss or misread without real overlap.
- Verification: strip out every other depth cue in the design (shadow, color, blur) and keep only the overlap between elements, then check whether viewers can still correctly judge which element is on top — if layering stays clear with all other cues removed, the occlusion relationship itself is properly designed.