Perception, Learning, Memory and Forgetting
2. The senses: vision, hearing, the skin, chemical and body senses
Each sense transduces a different energy. In vision, light passes the cornea and lens to the retina, where rods serve dim light and movement and the cones serve colour and fine detail. Colour vision needs two theories together: the trichromatic (Young–Helmholtz) theory for three cone types at the retina, and opponent-process (Hering) theory for red–green, blue–yellow and black–white channels further along, which explains afterimages. In hearing, sound waves move the eardrum and ossicles and set the cochlear fluid vibrating; the frequency becomes pitch and the amplitude becomes loudness.
- Skin senses carry pressure, warmth, cold and pain. Pain is modulated, not simply transmitted: the gate-control theory (Melzack and Wall) holds that a neural 'gate' in the spinal cord can block or let through pain signals, which is why rubbing a knock eases it.
- Smell (olfaction) and taste (gustation) are the chemical senses. Olfactory receptors respond to airborne molecules and project without a thalamic relay, linking smell strongly to memory and emotion; taste codes sweet, sour, salty, bitter and umami. Much of what we call flavour is actually smell.
- Kinesthesis senses the position and movement of the body's parts through receptors in muscles, tendons and joints; the vestibular sense, housed in the inner ear's semicircular canals and otolith organs, senses head orientation and balance. Conflict between the vestibular and visual signals produces motion sickness.
3. Perception: attention, Gestalt organisation, depth and illusions
Attention selects a small part of the sensory flood for full processing. Broadbent's filter model puts a filter early, before meaning is extracted, which the cocktail-party effect — hearing your own name in an unattended conversation — embarrasses; Treisman's attenuation model answers it by weakening rather than wholly blocking the unattended channel. Attention can be selective or divided, and its limits produce inattentional blindness, missing an obvious event while focused elsewhere.
The Gestalt school (Wertheimer, Köhler, Koffka) argued that the whole is different from the sum of its parts, and named the principles by which the visual field is organised into figures against a ground. The master principle is Prägnanz — we see the simplest, most stable organisation available.
| Principle | We group elements that are... |
|---|---|
| Proximity | close together in space |
| Similarity | alike in colour, shape or size |
| Continuity (good continuation) | fall along a smooth, continuous line |
| Closure | complete a familiar whole across a gap |
| Common fate | move together in the same direction |
Depth perception turns a two-dimensional retinal image into three dimensions using cues of two kinds. Binocular cues need both eyes: retinal (binocular) disparity, the slight difference between the two eyes' images, is the principal binocular cue, and convergence is the inward turn of the eyes for near objects. Monocular cues work with one eye — relative size, interposition (overlap), linear perspective, texture gradient, aerial perspective, relative height and motion parallax. Illusions are systematic misreadings that expose these mechanisms. On the usual textbook explanation, the Müller–Lyer arrows and the Ponzo rails arise from depth cues (linear perspective, size constancy) misapplied to a flat figure; explanations of the moon illusion are still debated.
4. Theories of learning
Classical conditioning (Pavlov) teaches a reflex to a new signal. An unconditioned stimulus (UCS), food, produces an unconditioned response (UCR), salivation; pair a neutral stimulus, a bell, with the food repeatedly and the bell becomes a conditioned stimulus (CS) that alone produces a conditioned response (CR). Key phenomena are acquisition, extinction when the CS is shown without the UCS, spontaneous recovery, generalisation to similar stimuli and discrimination between them.
Operant conditioning (Skinner), built on Thorndike's law of effect (responses followed by satisfaction are stamped in), shapes voluntary behaviour by its consequences. Reinforcement raises the rate of a response and punishment lowers it; each can be positive (add a stimulus) or negative (remove one), so negative reinforcement is not punishment — it strengthens behaviour by removing something aversive. The timing of reinforcement matters: ratio schedules reward a number of responses and interval schedules reward after time, and variable schedules (especially variable-ratio) give the highest, most extinction-resistant response rates.
- Social learning theory (Bandura) adds learning by observation and imitation of a model, without the learner being reinforced directly. The Bobo doll studies showed children copy aggression they merely watched; vicarious reinforcement (seeing the model rewarded) raises imitation, and attention, retention, reproduction and motivation are its four steps.
- Cognitive learning holds that learning can occur without performance and without reinforcement. Tolman's rats formed a cognitive map and showed latent learning — knowledge that stayed hidden until a reward made it worth using. Köhler's chimpanzees solved problems by sudden insight, a reorganisation of the field rather than trial and error.
5. Memory: encoding, storage, retrieval and information processing
Memory has three operations: encoding (turning information into a storable form, often acoustic in short-term and semantic in long-term memory), storage (holding it over time) and retrieval (bringing it back). The information-processing account, the Atkinson–Shiffrin multi-store model, routes information through three stores: a brief, high-capacity sensory memory; a limited short-term store (STS) holding about 7 ± 2 items (Miller) for seconds unless rehearsed; and a vast, durable long-term store (LTS).
Two refinements followed. Baddeley and Hitch's working memory replaces the passive STS with an active system: a central executive directing a phonological loop (speech-based) and a visuo-spatial sketchpad, later joined by an episodic buffer. The levels-of-processing view (Craik and Lockhart) argues that durability depends not on which store but on depth of processing — shallow (how a word looks) versus deep, semantic processing (what it means), which is remembered far better. Retrieval from LTS is cue-dependent: recall regenerates without cues, recognition only matches, and the encoding-specificity principle says retrieval succeeds when cues match the encoding context.
Long-term memory is reconstructive, not a recording. Bartlett showed, with the story War of the Ghosts, that recall is rebuilt to fit a schema, so unfamiliar detail is dropped or normalised. Loftus's work on the misinformation effect showed that a leading question after an event can alter the eyewitness's later memory report — asking how fast cars 'smashed' rather than 'hit' raised later speed estimates — which is why confident testimony can still be false.
6. Forgetting: encoding failure, interference and trace decay
Forgetting has several causes, and they sit at different stages. Encoding failure means the information never entered long-term memory in a usable form — you cannot recall what you never properly attended to, which is why few people can draw a common coin accurately. Trace-decay theory holds that an unused memory trace fades with time through disuse; it fits sensory memory best, is contested for short-term memory, and is hard to separate from interference in long-term memory, because time also lets interfering events accumulate.
Interference theory attributes forgetting to competition between memories. Proactive interference is older learning disrupting newer (an old phone number intrudes when you try to recall the new one); retroactive interference is newer learning disrupting older (learning Spanish makes an earlier French word harder to recall). A memory may also be present but temporarily inaccessible — the tip-of-the-tongue state is retrieval failure, not loss, since the right cue releases it.
Key takeaways
- The absolute and difference thresholds are defined at 50% detection. Weber's law says ΔI/I = k: the JND is a constant fraction of intensity, so a fraction found at one standard predicts the JND at any other. Fechner's law S = k log I builds sensation on the logarithm of intensity.
- Perception organises sensation: attention selects (Broadbent's early filter, Treisman's attenuation), Gestalt principles group by proximity, similarity, continuity, closure and common fate under Prägnanz, depth uses binocular (disparity, convergence) and monocular cues, and, on the usual textbook account, illusions such as Müller–Lyer and Ponzo expose depth cues misapplied.
- Classical conditioning (Pavlov) links a CS to a reflex; operant conditioning (Skinner, on Thorndike's law of effect) shapes behaviour by consequences, where negative reinforcement strengthens by removing an aversive. Social learning (Bandura) adds observation, and cognitive learning (Tolman's latent learning, Köhler's insight) needs neither performance nor reinforcement.
- The Atkinson–Shiffrin model routes information through sensory, short-term (7 ± 2 items) and long-term stores; working memory (Baddeley) makes the short-term store active, and levels of processing (Craik and Lockhart) make depth, not store, decide durability.
- Long-term memory is reconstructive: Bartlett showed recall is rebuilt to fit a schema and Loftus showed leading questions can reshape eyewitness memory. Forgetting is encoding failure, trace decay, or interference — proactive (old blocks new) and retroactive (new blocks old) — and the tip-of-the-tongue state is retrieval failure, not loss.
Practice questions (16)
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 psychophysics, the absolute threshold for a stimulus is best defined as the:
Show answer
Answer: A — smallest intensity that a person detects on 50% of the trials
The absolute threshold is the intensity detected on half the trials — detection is probabilistic, not an all-or-none edge. The smallest just-noticeable change is the difference threshold (JND), a different quantity. The other options describe a pain ceiling and adaptation, which are unrelated to the absolute threshold.Weber's law (ΔI/I = k) implies that as a standard stimulus grows more intense, the just-noticeable difference (JND):
Show answer
Answer: B — grows in absolute size, keeping a constant ratio to the standard
Since ΔI = kI, the JND rises in proportion to the standard I while the fraction ΔI/I stays fixed — this is exactly why a 2 g difference is enough at 100 g but 20 g is needed at 1000 g. A constant absolute JND would be Weber's law failing, not holding; the JND never vanishes above the absolute threshold.A person can just detect that a 51 g weight is heavier than a 50 g standard weight. Assuming Weber's law holds for lifted weight, what is the just-noticeable difference, in grams, for a 200 g standard weight? Give the number only.
Numerical answer — type the value.
Show answer
Answer: 4
The Weber fraction is k = ΔI/I = (51 − 50)/50 = 1/50 = 0.02. For a 200 g standard the JND is k × I = 0.02 × 200 = 4 g. The usual slip is to keep the 1 g JND unchanged across standards, which contradicts Weber's law.A listener can just detect a change in pitch when a 400 Hz standard tone is raised to 404 Hz. Assuming Weber's law holds for frequency, to what frequency, in Hz, must a 1000 Hz tone be raised before the change is just noticeable? Give the number only.
Numerical answer — type the value.
Show answer
Answer: 1010
The Weber fraction is k = (404 − 400)/400 = 4/400 = 0.01. For a 1000 Hz standard the JND is 0.01 × 1000 = 10 Hz, so the tone must be raised to 1000 + 10 = 1010 Hz. Reporting the JND of 10 Hz alone, or carrying the 4 Hz increment across unchanged, are the usual errors.A shopper can just tell a 515 g bag from a 500 g standard bag by heft. Assuming Weber's law holds, what is the Weber fraction for lifted weight? Give the answer as a decimal to two places.
Numerical answer — type the value.
Show answer
Answer: 0.03
The Weber fraction is k = ΔI/I = (515 − 500)/500 = 15/500 = 0.03. Note the denominator is the standard (500 g), not the larger weight — dividing by 515 gives 0.029, the usual slip.The opponent-process theory of colour vision was proposed to explain a fact that the trichromatic theory alone could not. Which fact is that?
Show answer
Answer: B — negative afterimages, where staring at green yields a red afterimage
Opponent-process theory posits red–green, blue–yellow and black–white channels, so fatiguing the green response leaves the red side unopposed — a red afterimage. The three cone types are exactly what trichromatic theory already explains, and the rod count and acuity gradient concern receptor distribution, not the colour coding opponent-process was built to handle.Which of the following correctly state a Gestalt organizing principle of perception?
Show answer
Answer: A — Proximity: elements near one another in space are grouped together; B — Closure: the visual system completes a familiar whole across a gap; C — Common fate: elements moving together in the same direction are grouped
Proximity, closure and common fate are three of the Gestalt grouping principles, alongside similarity and good continuation, all governed by Prägnanz. The claim about an enlarging retinal image always signalling an approaching object misdescribes perceptual constancy, which is the tendency to see an object's size, shape and colour as stable despite changing retinal images.A one-eyed painter conveys depth on a flat canvas by making a road's edges converge toward a point and painting distant hills in paler, bluish tones. Which depth cues are being used?
Show answer
Answer: C — Linear perspective and aerial perspective, both monocular cues
Converging edges are linear perspective and hazy bluish distance is aerial perspective; both are monocular, so a one-eyed painter on a still canvas can use them. Retinal disparity and convergence are binocular and need two eyes on a real scene, and motion parallax needs movement, so none of those can be shown on a flat, static painting.According to the usual textbook explanation, the Müller–Lyer illusion, in which two lines of equal length look unequal because of the arrowheads at their ends, is best understood as:
Show answer
Answer: D — a misapplication of depth cues to a flat two-dimensional figure
On the usual textbook account, illusions like Müller–Lyer arise when the perceptual system applies depth cues (the arrowheads resemble near and far corners) to a flat drawing, so equal lines are scaled unequally. The retina transduces both lines faithfully; the error is in interpretation, not in sensation or adaptation.A dog learns to salivate to a bell that has been paired with food. If the bell is now sounded many times without any food, salivation to the bell gradually stops. This phenomenon is:
Show answer
Answer: A — extinction of the conditioned response
Presenting the CS (bell) repeatedly without the UCS (food) weakens and eventually stops the CR — this is extinction. Spontaneous recovery is the opposite: the reappearance of the CR after a rest; generalisation is responding to similar stimuli; higher-order conditioning uses an established CS to condition a new one.A child watches an older sibling praised for tidying up and, without ever being rewarded herself, begins to tidy up too. Within the standard theories, this is best explained as:
Show answer
Answer: C — Bandura's social learning through a model and vicarious reinforcement
The child learns by observing a model and seeing the model rewarded — vicarious reinforcement — which is Bandura's social learning theory, and crucially she herself is never reinforced. Operant conditioning would require direct reinforcement of the child; classical conditioning links reflexes to signals, not goal-directed tidying; insight is a sudden solution to a problem, not imitation.Which of the following statements about operant reinforcement are correct?
Show answer
Answer: A — Negative reinforcement strengthens a response by removing an aversive stimulus; B — A variable-ratio schedule produces high, steady and extinction-resistant responding; D — Thorndike's law of effect anticipated operant conditioning
Negative reinforcement removes an aversive and so strengthens behaviour; variable-ratio schedules give the highest, most persistent responding; and Thorndike's law of effect — responses followed by satisfaction are stamped in — is the forerunner Skinner formalised. Punishment and negative reinforcement do not share an effect: punishment lowers a response rate while negative reinforcement raises it.In the Atkinson–Shiffrin multi-store model of memory, the short-term store is best characterised as:
Show answer
Answer: D — a limited store of about 7 ± 2 items held for seconds unless rehearsed
The short-term store holds roughly seven items (Miller's 7 ± 2) for a matter of seconds, extended only by rehearsal. The unlimited durable store is long-term memory, and the very brief high-capacity register holding raw input is sensory memory; no store is immune to loss.Which of the following statements about memory models and processes are correct?
Show answer
Answer: A — Baddeley's working memory includes a phonological loop and a visuo-spatial sketchpad under a central executive; B — The levels-of-processing view holds that deeper, semantic processing yields more durable memory; C — The encoding-specificity principle says retrieval improves when cues match the encoding context
Working memory's components, the depth effect of levels of processing, and the encoding-specificity principle are all stated correctly. The claim that recognition always demands more active regeneration than free recall is backwards: recognition supplies the item and asks only for a match, so it is usually easier than free recall, which must regenerate the item without a cue.A student who has just memorised a new locker combination keeps producing last year's combination instead. Which account of forgetting fits this case most precisely?
Show answer
Answer: B — Proactive interference, where the older combination disrupts recall of the newer one
The old (prior) learning is intruding on recall of the new material, which is proactive interference — the old acts forward in time. Retroactive interference would be the reverse, new disrupting old. Encoding failure is ruled out because the new combination was just memorised, and decay is ruled out because the intrusion is a competing memory, not simple fading over time.You cannot recall the exact design of a common coin you have handled thousands of times. Which cause of forgetting does this most directly illustrate?
Show answer
Answer: A — Encoding failure, because the fine detail was never attended to and stored
The coin's details were handled but never deeply attended to, so they never entered long-term memory in a usable form — encoding failure, not loss of a stored trace. Interference requires competing learned material, spontaneous recovery is a return of an extinguished response, and motivated forgetting concerns painful content, none of which applies to an ordinary coin.