The visual field automatically splits into figure and ground
Aliases: figure-ground segregation · Rubin's vase · border ownership
What it is
Looking at any patch of visual input, people don't process it as a uniform whole — it's automatically split into two parts: one read as the figure (has an outline, feels like a "thing," seems closer), and the rest read as ground (gets "cut off" by the figure's boundary, seems to extend behind it, has no shape of its own). This split is figure-ground organization.
The clearest demonstration is a bistable figure that supports two equally valid readings (the classic Rubin's vase, or two profiles versus a vase) — the same boundary line, with nothing else in the image changed, and yet the roles of figure and ground can swap. That shows figure-ground assignment isn't a property the image carries on its own; it's a judgment the visual system makes about the same input.
Why it happens
Which side becomes the figure and which becomes the ground is decided by a handful of empirical regularities: the side with smaller area, the more symmetric side, a convex outline, or a side that's fully enclosed is more likely to be judged the figure. Beyond that, a boundary ultimately gets "assigned" to one of its two sides, and that assignment itself is the crux of the matter — whichever side the boundary is assigned to gains an outline and appears to sit in front, while the other side loses that boundary and is read as extending behind, with no shape of its own.
When the cues supporting one assignment are about as strong as the cues supporting the other, the judgment doesn't settle and flips back and forth between the two readings — which is exactly why bistable figures "flip": boundary ownership never gets uniquely resolved, so the visual system keeps alternating between the two assignments.
Studying it
The classic study of figure-ground uses bistable figures, having observers report which side they perceive as the figure at a given moment and how often and how regularly the flip occurs, to study what makes one side more readily and more persistently read as the figure. A second line of research goes down to the cortical level, recording cells in visual cortex that respond to boundaries, and testing whether their response strength changes depending on which side that boundary is assigned to — a change would indicate that boundary assignment is settled at an early stage of visual processing, not after higher-level object recognition.
Where it stops holding
Regularities like area, symmetry, and convexity give tendencies, not deterministic rules: when they're all present and pointing in different directions (the smaller side is irregular while the larger side is convex, say), which side ends up as the figure depends on the specific combination of cues and can't be settled by applying any single regularity alone.
Figure-ground segregation only answers "which part is figure and which is ground" — it doesn't answer what the figure is or whether there's anything else in the ground. Those are recognition-level questions that occur later, building on the segregation once it has already been resolved.