A radiologist is told that a certain tumour is very common in the patients she will see that day. According to signal detection theory, she is most likely to show:

More hits and more false alarms, with no change in d′

Fewer hits and fewer false alarms

More hits and more false alarms, with no change in d′

A higher d′ with fewer false alarms

Witkin's Rod-and-Frame Test and Embedded Figures Test measure:

Field dependence–independence

Assertion (A): People from environments with few rectangular buildings and right angles tend to be less susceptible to the Müller-Lyer illusion. Reason (R): According to the carpentered-world hypothesis, experience with rectangular structures teaches people to read angled lines as corners receding in depth. Choose the correct option.

Both A and R are true, and R is the correct explanation of A

Both A and R are true, but R is not the correct explanation of A

Which of the following are ASSUMPTIONS of signal detection theory? Select all that apply.

Detection involves a decision about where to set a criterion

Background noise is always present

Expectations and payoffs can shift responses without changing sensitivity

There is a fixed absolute threshold above which signals are always detected

Detection involves a decision about where to set a criterion

Background noise is always present

Expectations and payoffs can shift responses without changing sensitivity

Perception II: Constancy and Illusions; Form, Depth and Movement; Motivation and Learning in Perception; Signal Detection Theory; Subliminal Perception, Information Processing, Culture, Perceptual Styles, Pattern Recognition and the Ecological Perspective

The rest of Unit 5's perception block reads: "Perceptual Constancy: Size, Shape, and Color; Illusions. Perception of Form, Depth and Movement. Role of motivation and learning in perception. Signal detection theory: Assumptions and applications. Subliminal perception and related factors, information processing approach to perception, culture and perception, perceptual styles, Pattern recognition, Ecological perspective on perception." This chapter takes the constancies and the illusions that betray them, the cues to depth and the kinds of perceived movement, the New Look evidence that needs and set shape perception, signal detection theory with a worked d′, and then the six approaches listed last — ending with the debate between Gregory's perception-as-hypothesis and Gibson's direct perception.

1. Perceptual constancy and illusions

Perceptual constancy is the tendency to perceive objects as stable although their retinal images change. Size constancy: a person walking away is not seen as shrinking, because the brain takes distance into account (size–distance invariance; Emmert's law says an afterimage looks larger when projected on a farther surface). Shape constancy: a door opening is still seen as rectangular although its image becomes a trapezoid. Colour and lightness constancy: a white shirt looks white in shade and sunlight because the visual system uses the ratio of light reflected by an object to that of its surroundings.

Classic visual illusions
Illusion	What is seen	Leading explanation
Müller-Lyer	The line with outgoing fins looks longer than the line with arrowheads	Gregory's misapplied size-constancy scaling: fins read as the inside corner of a room (farther), arrows as an outside corner (nearer); the carpentered-world hypothesis
Ponzo	Of two equal bars between converging lines, the upper looks longer	Linear perspective makes the upper bar seem farther, so it is scaled up
Ebbinghaus (Titchener circles)	A circle surrounded by large circles looks smaller than one surrounded by small circles	Size contrast with the context
Ames room	People standing in the two far corners look very different in size	The trapezoidal room is assumed to be rectangular, so distance cues fail
Moon illusion	The moon looks larger at the horizon than overhead	Apparent-distance theory: terrain makes the horizon moon seem farther, so the same image is scaled larger
Ambiguous and impossible figures	The Necker cube reverses in depth; the Penrose triangle cannot exist	Two equally good interpretations, or locally consistent but globally impossible cues

2. Perception of form, depth and movement; the role of motivation and learning

Form is built from contours and edges, detected by orientation-selective cortical cells and organised by the Gestalt laws. Depth is recovered from a two-dimensional retinal image by cues. Monocular (pictorial) cues: relative size, interposition (occlusion), linear perspective, texture gradient, aerial (atmospheric) perspective, shading, height in the visual field, and motion parallax (near objects seem to move faster, and opposite to the direction of travel). Binocular cues: retinal disparity (the slightly different images of the two eyes, fused in stereopsis) and convergence (the inward turning of the eyes for near objects). Accommodation, the change in the lens's shape, is an oculomotor monocular cue. Gibson and Walk's visual cliff (1960) showed that most crawling infants of about six months refuse to cross the apparent drop.

Real movement is signalled by motion-sensitive cells (area MT/V5) and by eye movements that track an object.
Apparent (stroboscopic) movement — the phi and beta phenomena, the basis of cinema.
Induced movement — the moon seems to race through clouds that are actually moving.
Autokinetic effect — a stationary point of light in a dark room appears to move (used by Sherif to study norm formation).
Motion aftereffect (the waterfall illusion) — after watching motion in one direction, a still scene seems to drift the other way; and biological motion — Johansson's point-light walkers are seen as people from a few moving dots.

Motivation, values and set: the "New Look" of the late 1940s showed that perception is not neutral. Bruner and Goodman (1947) found that poor children overestimated the size of coins more than rich children; McGinnies (1949) found higher recognition thresholds for taboo words (perceptual defence); and Bugelski and Alampay (1961) showed that an ambiguous "rat–man" figure was seen as a rat after pictures of animals — perceptual set. Learning: perceptual learning improves discrimination with experience (Eleanor Gibson's differentiation); adults whose sight is restored after early blindness (Gregory's patient S. B.) struggle to perceive depth and form; Stratton (1896) adapted to lenses that inverted the world; and Held and Hein's kitten carousel (1963) showed that self-produced movement is needed for normal visually guided behaviour.

3. Signal detection theory: assumptions and applications

Classical psychophysics assumed an absolute threshold — a stimulus intensity above which a signal is always detected. Signal detection theory (Tanner and Swets; Green and Swets, 1966) replaced it with two assumptions: (1) there is always noise — random neural and environmental activity — so every observation is a sample from either a noise distribution or a signal-plus-noise distribution that overlap; and (2) detection is a decision made by setting a criterion on that evidence. Performance therefore has two independent components: sensitivity, d′, the separation between the two distributions in standard-deviation units; and response bias, the criterion (β or c), which shifts with expectations, the probability of a signal and the payoffs for each outcome.

The four outcomes of signal detection
Stimulus	Observer says "yes"	Observer says "no"
Signal present	Hit	Miss
Signal absent	False alarm	Correct rejection

Worked. d′ = z(hit rate) − z(false-alarm rate). An observer with a hit rate of 0.84 (z = +1.0) and a false-alarm rate of 0.16 (z = −1.0) has d′ = 1.0 − (−1.0) = 2.0. A liberal criterion raises hits and false alarms together without changing d′; a conservative one lowers both. Plotting hits against false alarms as the criterion moves gives the receiver operating characteristic (ROC) curve; the farther it bows from the diagonal, the greater the sensitivity. Applications: radiologists reading scans, air-traffic controllers and radar operators, eyewitness line-ups, lie detection, quality control and vigilance tasks — anywhere the cost of a miss must be weighed against the cost of a false alarm.

⚠️ Sensitivity is not bias

A radiologist told that a disease is common will say "yes" more often: hits rise but so do false alarms. That is a criterion shift, not better eyesight. Only a change in d′ is a change in sensitivity.

4. Subliminal perception, information processing, culture and perceptual styles

Subliminal perception is the registration of stimuli below the threshold of awareness. James Vicary's 1957 claim that flashing "Eat popcorn" raised cinema sales was later admitted to be fabricated, but laboratory work shows real, small and short-lived effects: masked primes speed related judgements, and subliminal exposure increases liking (Kunst-Wilson and Zajonc, 1980). The related factors are stimulus duration and masking, the strength of the prime, and its relevance to a current need — a subliminal drink brand influenced choice only among thirsty participants (Karremans and colleagues, 2006). Blindsight (Weiskrantz) — accurate guessing about objects in a blind field after visual-cortex damage — is dramatic evidence of processing without awareness.

The information-processing approach treats perception as a sequence of stages — sensory registration, feature extraction, pattern recognition and interpretation — and distinguishes bottom-up (data-driven) from top-down (conceptually driven) processing. Richard Gregory argued that perception is an active hypothesis built from sparse data and knowledge (constructivism); Ulric Neisser's perceptual cycle (1976) combines the two, with schemata directing exploration and information modifying schemata.

Culture and perceptual styles
Topic	Findings
Culture and perception	Segall, Campbell and Herskovits (1966) found the Müller-Lyer illusion weaker in people from non-"carpentered" environments; Hudson (1960) found that some African samples read pictorial depth cues less readily; Turnbull's BaMbuti guide, raised in dense forest, saw distant buffalo as insects — a failure of size constancy; Nisbett and Masuda found East Asians attend more to context and relations (holistic) and Westerners more to focal objects (analytic).
Perceptual (cognitive) styles	Field dependence–independence (Witkin): measured by the Rod-and-Frame Test, the Embedded Figures Test and the Body Adjustment Test; field-independent people separate an item from its context. Also levelling–sharpening (Holzman and Klein), reflection–impulsivity (Kagan's Matching Familiar Figures Test) and augmenting–reducing (Petrie).

5. Pattern recognition and the ecological perspective

Theories of pattern recognition
Theory	Claim	Problem or evidence
Template matching	An input is compared with stored templates	Cannot cope with the endless variation of size, orientation and style
Prototype theory	An input is compared with an abstract best example (Posner and Keele; Rosch)	People "recognise" an unseen prototype of dot patterns
Feature analysis	Patterns are broken into features — Selfridge's Pandemonium (1959): image demons, feature demons, cognitive demons and a decision demon	Fits feature-detector physiology and letter-confusion data
Recognition by components	Biederman (1987): objects are recognised from about 36 simple volumes called geons	Explains viewpoint-invariant recognition; deleting the joints between geons impairs recognition
Context effects	The word superiority effect (Reicher, 1969): a letter is identified better in a word than alone	Evidence of top-down influence; faces rely on the fusiform face area, and its damage causes prosopagnosia

J. J. Gibson's ecological perspective (1966, 1979) rejects the idea that perception needs inference. The light reaching a moving observer — the optic array — is richly structured, containing invariants such as texture gradients and optic flow (the pattern of expansion as one moves forward, whose centre specifies the direction of heading). Perception is direct: the information is picked up, not constructed. What is perceived are affordances — the action possibilities an environment offers an animal, such as a surface that affords walking or a handle that affords grasping. Gibson's view fits fast perception in natural conditions; Gregory's constructivism fits ambiguous, brief or impoverished displays and illusions.

🧠 Gibson vs Gregory in one line

Gibson: bottom-up, direct, rich stimulus, affordances. Gregory: top-down, indirect, poor stimulus, hypotheses. Neisser's cycle is the usual reconciliation.

Key takeaways

Size, shape and colour constancy keep the world stable; illusions such as the Müller-Lyer, Ponzo, Ames room and moon illusion show constancy mechanisms misapplied.
Depth comes from monocular cues (size, interposition, perspective, texture, motion parallax) and binocular cues (retinal disparity, convergence); movement may be real, apparent, induced, autokinetic or an aftereffect.
The New Look (Bruner and Goodman; McGinnies; Bugelski and Alampay) and studies of restored sight and adaptation show that needs, set and learning shape perception.
Signal detection theory separates sensitivity (d′ = z(H) − z(FA)) from the response criterion, replaces the absolute threshold, and applies wherever misses and false alarms have costs.
Subliminal effects are real but small and need-dependent; pattern recognition is explained by templates, prototypes, features (Pandemonium) and geons; Gibson's ecological view of direct perception and affordances opposes Gregory's perception-as-hypothesis.

Practice questions (10)

Attempt each one before opening the answer. Every explanation names the tempting wrong option as well as the right one, because that is where marks are lost.

In a detection experiment an observer's hit rate is 0.84 (z = +1.0) and false-alarm rate is 0.16 (z = −1.0). What is the observer's sensitivity d′? Type a number.
Numerical answer — type the value.
Show answer
Answer: 2
d′ = z(H) − z(FA) = 1.0 − (−1.0) = 2.0: the signal-plus-noise distribution lies two standard deviations above the noise distribution.
A radiologist is told that a certain tumour is very common in the patients she will see that day. According to signal detection theory, she is most likely to show:
1. Fewer hits and fewer false alarms
2. A lower absolute threshold
3. More hits and more false alarms, with no change in d′
4. A higher d′ with fewer false alarms
Show answer
Answer: C — More hits and more false alarms, with no change in d′
Expecting the signal lowers (liberalises) her criterion, so she says "yes" more often: hits and false alarms rise together while sensitivity is unchanged. SDT does not assume an absolute threshold at all.
Which of the following are MONOCULAR cues to depth? Select all that apply.
1. Motion parallax
2. Retinal disparity
3. Linear perspective
4. Texture gradient
Show answer
Answer: A — Motion parallax; C — Linear perspective; D — Texture gradient
Linear perspective, texture gradient and motion parallax are all available to one eye. Retinal disparity requires comparing the images of the two eyes and is binocular, like convergence.
Bruner and Goodman (1947) found that poor children overestimated the size of coins more than rich children. The study is a classic demonstration of:
1. Field independence
2. The influence of needs and values on perception
3. Size constancy
4. Perceptual defence against taboo words
Show answer
Answer: B — The influence of needs and values on perception
The coin study launched the New Look by showing value and need distorting perceived size. Perceptual defence is McGinnies's taboo-word finding, and field independence is Witkin's perceptual style.
In Gibson's ecological theory, the action possibilities that an object or surface offers an animal are called:
1. Geons
2. Schemata
3. Prototypes
4. Affordances
Show answer
Answer: D — Affordances
Gibson held that we perceive affordances directly — a chair affords sitting, a handle grasping. Geons are Biederman's volumetric primitives, schemata belong to Neisser and the constructivists, and prototypes to pattern-recognition theory.
Witkin's Rod-and-Frame Test and Embedded Figures Test measure:
1. Size constancy
2. Field dependence–independence
3. Signal-detection sensitivity
4. Reflection–impulsivity
Show answer
Answer: B — Field dependence–independence
Both tests assess how far a person's perception is dominated by the surrounding field; field-independent people can align the rod with true vertical and find the hidden figure. Reflection–impulsivity is measured by Kagan's Matching Familiar Figures Test.
Selfridge's Pandemonium model of pattern recognition is an example of:
1. Recognition by components
2. Direct perception
3. A feature-analysis theory
4. A template-matching theory
Show answer
Answer: C — A feature-analysis theory
In Pandemonium, feature demons respond to lines and angles, cognitive demons "shout" in proportion to the features that match their letter, and a decision demon picks the loudest — a feature-analysis account. Recognition by components uses geons, not features of letters.
Assertion (A): People from environments with few rectangular buildings and right angles tend to be less susceptible to the Müller-Lyer illusion. Reason (R): According to the carpentered-world hypothesis, experience with rectangular structures teaches people to read angled lines as corners receding in depth. Choose the correct option.
1. Both A and R are true, and R is the correct explanation of A
2. Both A and R are true, but R is not the correct explanation of A
3. A is true, but R is false
4. A is false, but R is true
Show answer
Answer: A — Both A and R are true, and R is the correct explanation of A
Both are true and R explains A: Segall, Campbell and Herskovits attributed the weaker illusion in non-carpentered cultures to the absence of that learned depth interpretation of the fins.
A stationary point of light in a completely dark room appears to move. This is the:
1. Autokinetic effect
2. Motion aftereffect
3. Phi phenomenon
4. Induced movement
Show answer
Answer: A — Autokinetic effect
The autokinetic effect arises without any frame of reference, which is why Sherif could use it to study how group norms form. The phi phenomenon needs two alternating lights, the aftereffect follows prolonged motion, and induced movement is caused by a moving surround.
Which of the following are ASSUMPTIONS of signal detection theory? Select all that apply.
1. There is a fixed absolute threshold above which signals are always detected
2. Detection involves a decision about where to set a criterion
3. Background noise is always present
4. Expectations and payoffs can shift responses without changing sensitivity
Show answer
Answer: B — Detection involves a decision about where to set a criterion; C — Background noise is always present; D — Expectations and payoffs can shift responses without changing sensitivity
SDT assumes constant noise, overlapping distributions, a decision criterion and independence of sensitivity from bias. It explicitly rejects the fixed absolute threshold of classical psychophysics.