Object Recognition

Bottom-up (data-driven) processing: builds recognition from raw sensory features upward — relied on for unfamiliar stimuli. Top-down processing: uses prior knowledge/memory/expectation — relied on for familiar stimuli. They work together.
Depth perception uses a binocular cue (binocular disparity) plus monocular cues: relative size, linear perspective, interposition, shading, motion parallax.
Gestalt psychology (Wertheimer, 1912): the brain perceives organized wholes, not separate parts. 6 key principles below.
Subjective/illusory contours: brain creates edges that were never drawn (e.g., Kanizsa triangle, 1955) — different from closure, which fills in a genuinely missing part of an otherwise-present object.
Law of Prägnanz (simplicity): when an image is ambiguous, the brain defaults to the simplest, least complex interpretation.

Law of proximity: Objects close together are perceived as belonging to the same group.
Law of similarity: Objects sharing color, shape, or size are grouped together, even if not adjacent.
Law of good continuation: The brain prefers smooth, continuous patterns — perceives one continuous line rather than separate pieces.
Subjective (illusory) contours: The brain perceives edges/boundaries that were never actually drawn (e.g., Kanizsa triangle).
Law of closure: The brain fills in missing information to perceive an incomplete object as complete.
Law of Prägnanz: Also called the law of simplicity — the brain favors the simplest, least complex interpretation of an ambiguous image.

Bottom-Up (Data-Driven)	Top-Down
Starts from raw sensory features, builds up	Starts from prior knowledge/expectation, interprets down
Relied on heavily for unfamiliar stimuli	Relied on heavily for familiar stimuli

Binocular Cue	Monocular Cues
Binocular disparity (requires both eyes)	Relative size, linear perspective, interposition, shading, motion parallax (each needs just one eye)

Closure fills in a part of an object that's genuinely missing but otherwise physically present; subjective contours create an edge that was never drawn at all — don't conflate the two.
Proximity groups by spatial closeness; similarity groups by shared visual features regardless of distance — a commonly swapped pair.
Bottom-up and top-down processing aren't competing — they run together, one supplying raw data, the other supplying interpretation speed/efficiency.

Which type of processing is used most heavily for a completely unfamiliar stimulus?

What depth cue requires only one eye and depends on distant objects being covered by closer ones?

What is the key difference between the law of closure and subjective contours?

What does the law of Prägnanz predict when an image is ambiguous?

Biology chapters 1–3 are free. Sign up to unlock every chapter.