Memory — AQA A-Level Psychology
Test yourself on Memory with AQA A-Level practice questions.
7 days Premium · Then the free plan continues · No card, no charge
Memory explained
Atkinson and Shiffrin (1968) pictured memory as three boxes: information arrives at the senses, attention passes it on, and rehearsal transfers it.
Read the full explanation
The sensory register codes by modality (e.g. iconic, echoic), has a huge capacity (Sperling), but a duration under half a second. Coding: Baddeley (1966) found short-term memory codes acoustically and long-term memory codes semantically. Capacity: Miller (1956) put the short-term span at seven plus or minus two chunks, against an effectively unlimited long-term store. Duration: Peterson and Peterson (1959) found short-term recall collapsed after 18 seconds without rehearsal, while Bahrick (1975) found 70% accuracy on faces after 48 years. Evaluation carries most marks: patient KF had a damaged verbal span but an intact visual one, so the short-term store cannot be one box.
Types of long-term memory: episodic, semantic, procedural.
Tulving (1972) argued the long-term store is not one thing. Episodic memory holds personal events with a time and a place attached and takes conscious effort to retrieve: your first day at school. Semantic memory holds facts and meanings with no time stamp: that Paris is a capital city. Procedural memory holds skills, retrieved automatically and awkward to put into words: riding a bike. The evidence is clinical. Clive Wearing lost almost all memory for events after a viral infection yet could still read music and play the piano, and HM learned new motor skills while denying he had ever practised them. Scanning work by Tulving placed recall of events more in the right prefrontal cortex and recall of facts more in the left. For evaluation, these are unique injuries with no measurement beforehand, and Cohen and Squire argue for two divisions rather than three.
The working memory model: central executive, phonological loop, visuo-spatial sketchpad and episodic buffer. Features of the model: coding and capacity.
Baddeley and Hitch (1974) replaced the single short-term box with a work space of three parts, later adding a fourth. The central executive directs attention and allocates tasks; it is modality-free and stores virtually nothing. The phonological loop codes sound, with a store holding what you hear and an articulatory process rehearsing roughly two seconds of speech. The visuo-spatial sketchpad codes visual and spatial information, holding three or four objects. The episodic buffer binds information, holds about four chunks, and links to long-term memory. Dual-task findings show two visual jobs at once interfere, but a visual and a verbal job do not. However, the central executive is described so vaguely that it cannot be tested.
Explanations for forgetting: proactive and retroactive interference and retrieval failure due to absence of cues.
Two very different accounts. Interference says memories compete, and the damage is worst when the material is similar. Proactive means older learning blocks newer; retroactive means newer learning blocks older. McGeoch and McDonald (1931) taught a word list to perfection, then gave a second list, and recall of the first fell to 12.5% when the new words were synonyms against 37.1% when they were numbers, which shows similarity is the active variable. Retrieval failure says the trace is still there but the cue has gone. Encoding specificity predicts that a cue only works if it was present at learning, and Godden and Baddeley (1975) had divers learn and recall on land or underwater, with recall about 40% worse whenever the two settings mismatched. Carter and Cassaday (1998) showed the same for internal state using antihistamines.
Factors affecting the accuracy of eyewitness testimony: misleading information, including leading questions and post-event discussion; anxiety.
In Loftus and Palmer's (1974) first experiment, 45 students each watched seven traffic-accident clips; mean speed estimates ran from 31.8 mph for 'contacted' to 40.8 mph for 'smashed'. In Experiment 2 (150 students), 32% of the 'smashed' group reported broken glass that was never there against 14% of the 'hit' group, arguing for an altered memory over response bias. Post-event discussion works differently: Gabbert (2003) paired witnesses watching a crime from different angles, and 71% reported unseen details, showing conformity. Anxiety cuts both ways. Johnson and Scott (1976) found 49% identification with a pen but 33% with a bloodied knife, while Yuille and Cutshall (1986) found real witnesses to a shooting accurate months afterwards.
Improving the accuracy of eyewitness testimony, including the use of the cognitive interview.
Fisher and Geiselman (1992) built an interview out of what is known about retrieval rather than out of police habit. Four techniques. Report everything, because a trivial detail can cue a bigger one. Reinstate the context, mentally returning to the weather, the layout and the mood, which exploits context-dependent retrieval. Recall in reverse order, which stops a witness running a tidy expected story. Change perspective, describing the scene as another person present would have seen it, which disrupts the same schema. The enhanced version adds rapport, fewer interruptions and open questions. Evidence: Köhnken (1999) pooled 55 studies and found 41% more correct detail than a standard interview, though incorrect detail rose alongside it, so accuracy is not simply improved.
Your focus
- Describe the three stores in order and name the process that carries information between each pair.
- Explain the coding, capacity and duration of each store, citing Sperling, Baddeley, Miller, Peterson and Bahrick.
- Evaluate the model using patient KF and the evidence that rehearsal alone does not guarantee a lasting memory.
Show all 21 objectives
- Describe episodic, semantic and procedural memory, giving the feature that identifies each and an example.
- Apply the three types to behaviours in a scenario, stating the feature that decides each classification.
- Evaluate the three-way division using Clive Wearing and HM against Cohen and Squire's declarative split.
- Describe the four components of working memory and state the coding and capacity of each.
- Explain why two visual tasks interfere with each other while a visual and a verbal task do not.
- Evaluate the model using dual-task evidence against the charge that the central executive is too vague to test.
- Explain proactive and retroactive interference in the right direction and name similarity as what makes each worse.
- Describe retrieval failure and separate context-dependent from state-dependent forgetting, with a study for each.
- Apply the two explanations to a scenario, deciding whether the loss is competition between memories or a missing cue.
- Evaluate interference research on artificial word lists, given that forgotten material returns once a cue is supplied.
- Describe Loftus and Palmer's verb manipulation and quote the speed estimates and the broken glass percentages.
- Explain why post-event discussion distorts testimony through conformity rather than through a damaged memory trace.
- Apply weapon focus and the inverted-U account to predict how anxiety would affect a witness in a described incident.
- Evaluate laboratory research on anxiety using Yuille and Cutshall's accurate real witnesses and the lack of consequences.
- Describe the four techniques of the cognitive interview and the memory principle each one exploits.
- Explain what the enhanced version adds through rapport, fewer interruptions and open questions.
- Apply the four techniques to a witness account, choosing which one would recover a particular missing detail.
- Evaluate the cognitive interview using Kohnken's 41% gain in correct detail against the rise in incorrect detail.
Memory exam tips
Marking Points
- Names the three stores in order and the two processes between them, attention into the short-term store and rehearsal into the long-term store.
- Sensory register: codes by modality (e.g. iconic for visual, echoic for auditory), has a very large capacity (Sperling), and a duration of less than half a second.
- Coding stated separately for the other stores, credited with Baddeley (1966): acoustic confusions immediately, semantic confusions after a delay.
- Capacity stated as seven plus or minus two chunks for the short-term store (Miller) and as unlimited for the long-term store.
- Duration stated as roughly 18 to 30 seconds unrehearsed (Peterson and Peterson) against up to a lifetime (Bahrick, 48 years).
- Evaluation: KF had poor recall of spoken digits but normal recall of visual material, so the short-term store is not one unitary thing.
- Evaluation: the single rehearsal route is too simple, since prolonged rehearsal can leave no lasting trace and elaborative processing produces one without it.
- episodic memory defined by its time stamp and its conscious, effortful retrieval, with an example
- semantic memory defined as knowledge of facts, concepts and meanings that has lost the occasion it was learned on, with an example
- procedural memory defined as memory for skills and actions, retrieved without awareness and hard to describe in words, with an example
- in an application question, each behaviour in the scenario matched to a type and the feature that decides it stated
- evaluation: Clive Wearing and HM used to show the types can be lost separately, so they are distinct systems
- evaluation: brain imaging support, or the competing declarative and non-declarative division from Cohen and Squire
- central executive described as a limited-capacity attentional controller that allocates the subsystems and is modality-free, holding virtually no material of its own
- phonological loop split into the phonological store and the articulatory process, coding acoustically, with about two seconds of capacity and the word-length effect as evidence
- visuo-spatial sketchpad described as coding visual and spatial information with a capacity of three to four objects
- episodic buffer described as a limited general store with a capacity of about four chunks that binds information from the other components and links it to long-term memory
- evaluation: dual-task studies, where two tasks competing for one subsystem impair each other and tasks in different subsystems do not
- evaluation: the central executive criticised as vague and unfalsifiable, or KF used to support separable short-term stores
- interference defined as competition between memories, with the direction stated: proactive is older on newer, retroactive is newer on older
- an everyday example each way, and similarity named as the condition that makes the effect worse
- McGeoch and McDonald's synonym condition quoted as evidence that similarity, not merely extra learning, causes the loss
- retrieval failure explained as material that is available but not accessible, with encoding specificity and the missing cue
- context-dependent forgetting separated from state-dependent forgetting, with Godden and Baddeley and with Carter and Cassaday
- evaluation: artificial word lists and short intervals limit generalisation, and interference cannot explain why forgotten material returns once the right cue is given
- a leading question defined as one whose wording suggests an answer, with the verb manipulation and the speed estimates quoted
- the two accounts of the effect given: response bias at the moment of answering against a genuinely changed memory, with the broken glass follow-up as the discriminator
- post-event discussion explained as conformity between witnesses, credited with the Gabbert figure
- anxiety treated in both directions, weapon focus lowering identification and real-world witnesses staying accurate under extreme stress
- evaluation: the inverted-U account used to reconcile the two sets of findings, with the criticism that arousal is treated as one simple quantity
- evaluation: student samples, filmed crimes and no real consequences, so laboratory anxiety is not the anxiety of a real incident
- all four techniques named and explained rather than listed: report everything, context reinstatement, reverse order and change of perspective
- the memory principle behind each one given, such as cueing, context-dependent retrieval and disrupting an expected schema
- the enhanced version distinguished by its social elements, including rapport, minimising distraction and avoiding interruption
- evaluation: Köhnken's meta-analysis of 55 studies, the 41% gain in correct information, and the rise in incorrect detail beside it
- evaluation: practical limits of time and training, and the fact that forces adopt only some techniques, so studies are not comparing one procedure
Examiner Tips
- 💡In extended essays, evaluation carries more weight than description, so keep the tour of the three stores concise to leave time for analysis.
- 💡Use a case study as an argument rather than a story: say what the patient could and could not do, then say what that proves about the model.
- 💡If the command word is 'outline', evaluation earns nothing, so do not spend time on it.
- 💡apply questions hand you one behaviour per type; quote the words from the stem, name the type, then justify it in a clause
- 💡pair each case study with a limitation, such as no baseline measurement or damage unique to that person, rather than listing studies
- 💡if asked to explain a difference, contrast the types on one feature at a time, such as time stamping and then conscious retrieval
- 💡in an application question, take each activity in the scenario, name the component it loads, and say why performance suffers or survives
- 💡a comparison with the multi-store model counts as evaluation only if you state what this model explains that the older one cannot
- 💡quote the word-length effect or the articulatory suppression finding rather than asserting that the loop has limited capacity
- 💡if the question names one explanation, the other earns credit only as a comparison inside the evaluation
- 💡in a scenario, decide first whether a cue is missing or another memory is competing, because everything you write follows from that choice
- 💡quote a figure from one study rather than three study names, since a named result is what separates a top band answer
- 💡a scenario puts the misleading element in the stem, so quote it, name the mechanism, then say what it does to that testimony
- 💡evaluation that reaches the courtroom, such as juries trusting confident witnesses, scores better than a line about students being unrepresentative
- 💡if the question names only misleading information, anxiety research belongs in the answer only as a comparison
- 💡in an application question, take the techniques one at a time and say what the named witness would actually be asked to do
- 💡quantify the evaluation, because a pooled figure from a meta-analysis is worth more than a sentence saying research supports it
- 💡if the question says discuss, plan two strengths and two limitations before writing, since the evaluation carries the larger share of marks
Common Mistakes
- Giving one feature for one store and moving on; the marks are for coding, capacity and duration named for the sensory register, the short-term store and the long-term store.
- Confusing maintenance rehearsal, which holds material in the short-term store, with the prolonged rehearsal that Atkinson and Shiffrin said transfers it.
- Quoting Miller's span as seven items rather than seven chunks, which loses the point that chunking is what makes the span stretch.
- Using KF as evidence against the long-term store when the case is about the short-term store having more than one part.
- giving three definitions with no examples, when the distinguishing example is usually where the second mark sits
- assuming semantic memory cannot come from personal experience; the test is whether the memory still carries when and where, not who learned it
- saying HM 'lost his memory', which throws away the reason he is cited, namely that his skill learning survived
- listing the case studies as description when the question wanted them as evaluation of the three-way split
- listing the four components without their coding and capacity, which is exactly what the specification names as the features to know
- calling the central executive a store, when it directs attention and is modality-free, holding almost nothing
- treating this as a model of all memory rather than of short-term memory, which wrecks any comparison with the multi-store model
- saying dual-task studies 'prove' the model, when they support the separation of the subsystems and nothing more
- reversing the two forms of interference; 'pro' runs forwards, so the older memory acts forwards onto the newer one
- writing that interference destroys the memory, when the studies show a blocked route to it rather than an erased trace
- describing the divers study without saying which conditions matched and which did not, so the finding has nothing to attach to
- offering decay or displacement, which are not the explanations this part of the specification asks for
- quoting Loftus and Palmer with no numbers, when 'smashed produced higher estimates' is worth far less than 40.8 mph against 31.8 mph
- treating anxiety as purely harmful, when the specification expects both directions and some attempt to resolve them
- merging the two Loftus and Palmer experiments, since the speed estimates (Experiment 1) and the broken glass (Experiment 2) came from separate procedures with different sample sizes
- calling post-event discussion a leading question, when the misleading information comes from another witness rather than the interviewer
- naming the four techniques with nothing about how each one helps retrieval, which caps the knowledge marks straight away
- claiming the interview makes witnesses accurate, when it raises the amount of correct detail and the amount of incorrect detail together
- mixing up reverse order, which is the order of events, with change of perspective, which is whose viewpoint
- ignoring the enhanced version when the question asks about improving accuracy in general