A photo is flat. How can you still see depth in it with one eye shut?

Sensing versus perceiving

• Sensation is your sense organs detecting physical energy, e.g. light falling on the retina at the back of each eye.
• Perception is your brain organising and interpreting that sensory information so it means something.
• The image on each retina is flat (2-D), yet you perceive a world with depth (3-D).
• The brain works out distance from depth cues. Monocular cues need only one eye; binocular cues need both.

Depth cues for one eye

• Texture gradient: surface detail (grass, pebbles) looks finer and more tightly packed further away.
• Monocular cues work with one eye and in flat pictures, which is why artists use them.

Depth cues for two eyes

• Your eyes are a few centimetres apart, so each retina gets a slightly different image. This is retinal disparity. The closer the object, the bigger the difference.
• The brain uses the difference to judge depth.
• Combining the two images into one view with depth is stereopsis.
• Binocular cues work best for objects close to you.

Two equal lines; one looks longer even when you know. Why?

When perception gets it wrong

• A visual illusion is when what you perceive does not match the physical stimulus.
• Ambiguity (ambiguous figures): one image fits two interpretations, so perception flips. The Necker cube flips which face is at the front; Rubin's vase flips between a vase and two faces.
• Fiction: you perceive something that isn't there. In the Kanizsa triangle you see a bright white triangle with edges, but no edges are drawn.

Size constancy and the Ames room

• Size constancy: an object looks the same size even when its retinal image shrinks with distance.
• The brain scales up anything it judges to be far away, so a friend walking off doesn't seem to shrink.
• The Ames room looks rectangular through a peephole, but one back corner is much further away.
• Assuming a normal room, the brain sees both people as equally far: one looks tiny, one huge.

Depth cues misread

• Müller-Lyer: of two equal lines, the one with fins pointing outwards looks longer.
• Gregory: fins out look like a room's inside corner (far); arrowheads, a building's outside corner (near).
• Size constancy then scales up the "far" line.
• Ponzo: converging lines act as linear perspective, so the upper of two equal bars seems further and bigger.
• These are misinterpreted depth cues.
• Size or length illusions like these are distortions.

Shape and colour constancy

• A visual constancy means you perceive an object as unchanged even though the image reaching your eye changes.
• Shape constancy: a door opening towards you casts a trapezium-shaped image, but you still see a rectangular door.
• Colour constancy: a banana still looks yellow in dim evening light.
• With size constancy, these keep the world stable. Illusions show them applied in the wrong place.

A pilot lands by watching the ground rush past. How is that enough?

Perception without guesswork

• Gibson argued that perception is direct: the pattern of light reaching your eyes, the optic array, holds all the information you need.
• No inference and no past experience are needed. Processing is bottom-up: it starts from the senses.
• Sensory input: the optic array is rich enough on its own to show depth, distance and movement.

Information in movement

• Moving changes the optic array in a regular way, and the change is information.
• Motion parallax: as you travel, near objects stream past fast and far objects slowly, so speed across your view shows distance.
• Optic flow: moving forward, the scene streams outwards from the point you're heading for, which stays still. Faster flow, faster movement.
• Gibson's ideas grew from his wartime work on films for training pilots.

Invariants and affordances

• Invariants are features of the optic array that don't change as you move, so they give reliable information, e.g. texture gradient and the horizon.
• Affordances are what an object offers you for action, picked up directly from its shape: a chair affords sitting, a handle affords grasping, a gap affords walking through.
• You can see what an object affords without knowing its name.

Born ready to perceive?

• Gibson's theory takes the nature side: picking up information from the optic array is innate, not learned.
• In the visual cliff studies (Gibson and Walk, 1960), crawling babies would not cross a glass floor over an apparent drop, even when their mothers called them. Young animals of many species avoided it too.
• But crawling babies are already several months old, so some learning can't be ruled out.

Your eyes give your brain half a picture. What fills the gaps?

Perception as a best guess

• Gregory argued that sensory input is often incomplete or ambiguous.
• The brain makes inferences: it goes beyond the input, using visual cues and past experience (prior knowledge).
• It forms a perceptual hypothesis, a best guess about what is out there, and tests it against new input.
• Processing is top-down: the brain constructs a model of reality from knowledge.

Illusions as evidence

• If perception is a hypothesis, a wrong hypothesis should give an illusion.
• Necker cube: the drawing fits two hypotheses equally well, so perception flips between them.
• Müller-Lyer: the brain infers depth from the fins and applies size constancy, so one line looks longer.
• Hollow face: a hollow mask is seen as a normal face sticking out, because every face you've seen does.

Gibson versus Gregory

• Each may be partly right: clear, rich viewing suits Gibson; brief, unclear or ambiguous input suits Gregory.

One shape reads as H in THE and A in CAT. What decides which?

Ready to see

• Perceptual set is a readiness to perceive certain things and not others. You notice some parts of the sensory input and ignore the rest.
• It works top-down, as in Gregory's theory: what is already in your head shapes what you see.
• So the same stimulus can be perceived differently by different people, or by you at different times.
• Four factors affect it: expectation, motivation, emotion and culture.

Four factors that set you

• Set usually helps you perceive quickly, but it can make you miss things or see what isn't there.

Big ideas

• Sensation is detecting energy with the sense organs; perception is the brain interpreting it.
• Monocular cues (height in the plane, relative size, superimposition, linear perspective, texture gradient) need one eye and work in pictures.
• Binocular cues need two eyes: retinal disparity lets the brain combine two views into depth (stereopsis).
• Illusions happen when perception doesn't match the stimulus: fictions, ambiguous figures and distortions are the three kinds.
• Size constancy keeps a walking-away friend the same size; misread depth cues make it misfire (Müller-Lyer, Ponzo, Ames room).
• Shape, colour and size constancy keep objects stable; Haber and Levin (2001) found size is judged mainly from knowing the object.
• Gibson: perception is direct and bottom-up; the optic array holds enough information, with no inference.
• Key ideas: sensory input (the optic array), optic flow, invariants and affordances.
• It explains fast, accurate real-world perception but not illusions or the effects of experience.
• Gregory: perception is top-down; sensory input is incomplete, so the brain makes inferences from visual cues and past experience to construct a model of reality.
• Key ideas: sensory input, perceptual hypothesis, inferences and prior knowledge.
• It explains illusions and perceptual set but struggles with fast real-world perception and newborn abilities.
• Perceptual set is a readiness to perceive some things and not others; it works top-down.
• Expectation, motivation, emotion and culture all shape it.
• Carmichael, Hogan and Walter (1932) showed that a word given with a figure shapes how it is reproduced.