Cognitive Psychology: Key Question & Practical Investigation

The Edexcel Cognitive topic closes with two applied requirements that connect the science of memory to the world beyond the laboratory. The key question asks you to identify an issue of societal relevance to which cognitive psychology speaks, and to analyse it using the topic's concepts and studies. The practical investigation asks you to design an ethical memory experiment of your own — stating a hypothesis, defining variables, choosing a design and sample, planning a procedure and ethics, and specifying how the data would be described (mean/median, a suitable graph) and analysed (a named inferential test). This lesson works through both. For the key question we take the reliability of memory as legal evidence; for the practical we model a study of the serial-position effect, with acoustic-versus-semantic coding and context-dependent recall noted as equally valid alternatives.

Key Definition: A key question is a real-world issue of societal importance that psychology can help explain or address; a practical investigation is a small-scale piece of research the student designs and (where possible) carries out, applying the methods studied in the course.

Spec Mapping

This lesson addresses the Edexcel 9PS0 — Paper 1, Topic 2: Cognitive Psychology requirements for (a) a key question of societal relevance analysed with cognitive concepts and (b) a practical investigation applying research methods to a memory topic. It draws together the whole topic — reconstructive memory (Bartlett), the multi-store model's serial-position logic, coding (Baddeley, 1966), retrieval failure — and the research-methods content examined in Paper 3 (hypotheses, variables, designs, sampling, ethics, descriptive and inferential statistics). In assessment-objective terms, you should be able to describe the key question and a well-designed study (AO1), apply cognitive concepts to the issue and to the design decisions (AO2), and evaluate the strength of the psychological analysis and the methodology of the investigation (AO3).

Connects to…

Criminological psychology (eyewitness testimony): the key question here — the reliability of memory as legal evidence — is the applied heart of the criminological topic; reconstructive memory and retrieval failure explain why witness memory is fallible.
Reconstructive memory (Bartlett) and retrieval failure (this topic): the key question is analysed directly through schema-driven distortion and cue-dependent forgetting; the practical applies the multi-store/serial-position or coding concepts.
Research methods (Paper 3): the practical investigation is research methods in action — operationalisation, experimental design, sampling, ethics, and the choice of descriptive statistics and an inferential test with a named significance level.

Part A — The Key Question: How Reliable Is Memory as Legal Evidence?

Why this is a key question of societal relevance

Eyewitness memory is one of the most persuasive forms of evidence in a criminal trial — jurors find a confident witness pointing across the courtroom compelling. Yet if human memory is reconstructive rather than a faithful recording, that confidence may be badly misplaced, and mistaken memory can send innocent people to prison. This is precisely the societal stake: the reliability of memory as legal evidence bears directly on justice, on wrongful convictions, and on how the police, courts and lawmakers should treat what a witness "remembers". Analyses of wrongful convictions later overturned by DNA evidence have repeatedly found that mistaken eyewitness identification was the single most common contributing factor — a sobering, real-world demonstration that the question is not academic. Cognitive psychology's account of memory is therefore of direct public importance, which is what makes this a genuine key question rather than a merely interesting one.

Analysing the question with cognitive concepts

The topic's theories converge to explain why memory is an unreliable form of legal evidence.

Reconstructive memory (Bartlett, 1932). Because remembering is an active reconstruction using schemas rather than a replay of a stored copy, witnesses do not retrieve an exact record of the crime. They rebuild it, filling gaps with schema-consistent detail (confabulation) and smoothing oddities to fit expectation (rationalisation). A witness may "remember" a weapon, a hood or a getaway car that fits their crime schema but was never there. Stereotypes shape reconstruction too — schema-driven expectations about who commits which crimes can distort who a witness recalls.
Retrieval failure and the encoding specificity principle (Tulving & Thomson, 1973). A witness's memory of the crime is encoded in a specific, often high-stress context; at interview, days or weeks later, the cues present at encoding are absent, so accurate detail that is available may be inaccessible. This both explains forgetting and points to a remedy — reinstating context (mentally or physically) can recover detail, the principle behind context-reinstatement interviewing.
The fragility of the memory stores. The multi-store and working memory models remind us that memory has sharp capacity and duration limits and is coded imperfectly (acoustically in the short term). A witness attends to only part of a fast, chaotic event; much never enters memory at all, and what does is subject to decay and interference before it is reported.
Post-event contamination. Because reconstruction continues at every retrieval, each retelling — to other witnesses, to police, under leading questioning — can alter the memory. Repeated questioning and misleading information become woven into the "memory", and the witness cannot tell the reconstructed detail from the original.

A balanced answer: memory is not always unreliable

A sophisticated analysis resists the caricature that memory is worthless as evidence. As the reconstructive-memory lesson stressed, distortion is greatest for unfamiliar, ambiguous, briefly-glimpsed, high-stress events — exactly the conditions of many crimes — but memory for familiar, distinctive, well-attended information can be accurate and durable. The responsible conclusion is therefore conditional: memory is a fallible form of legal evidence whose reliability depends on the encoding conditions, the delay, the questioning method and the witness (recall the individual-differences finding that age and cognitive profile modulate accuracy, and that confidence is a poor guide to accuracy). This is why psychology's contribution is not "never trust witnesses" but "understand when and why witness memory errs, and interview in ways that minimise distortion and maximise accurate recall" — a genuinely applied, evidence-based answer that has shaped real police practice.

The societal impact of the psychological answer

The cognitive account has changed the real world: it underpins improved interviewing techniques designed to avoid leading questions and to reinstate context, informs guidance to juries about the limits of eyewitness confidence, and features in the reform of identification procedures (how line-ups are constructed and administered to reduce misidentification). This demonstrable influence on the justice system is the clearest evidence that the key question matters and that cognitive psychology has something valuable to say about it.

Part B — The Practical Investigation

The practical requires you to design an ethical memory experiment. We model an investigation of the serial-position effect — the finding (from the multi-store model lesson) that in free recall of a word list, the first items (primacy) and last items (recency) are recalled better than the middle. The design principles below transfer directly to the two alternative topics noted at the end.

Aim and hypotheses

Aim: To investigate whether the position of a word in a list affects the likelihood of it being recalled — specifically, whether words from the start and end of a list are recalled better than words from the middle.

For a clean, testable design we operationalise this as a comparison of recall for early/late (primacy + recency) versus middle list positions.

Alternative (recency-focused) framing: compare recall of the last few items under immediate recall versus recall after a 30-second distractor task. The MSM predicts the distractor abolishes the recency effect (Glanzer & Cunitz, 1966). This makes an even sharper single-variable experiment.

Directional (one-tailed) experimental hypothesis (H₁): Participants will recall significantly more words from the start and end of the list (primacy and recency positions) than from the middle of the list.

Null hypothesis (H₀): There will be no significant difference in the number of words recalled from the start-and-end positions compared with the middle positions; any difference is due to chance.

A directional (one-tailed) hypothesis is justified because prior theory and evidence (the serial-position curve; Glanzer & Cunitz) predict the direction of the effect in advance.

Variables

Variable	Definition (operationalised)
Independent variable (IV)	The serial position of the words: "start-and-end" positions (first 5 + last 5 of a 20-word list) versus "middle" positions (the central 10)
Dependent variable (DV)	The number of words correctly recalled from each position category, in free recall (order not required)
Key controls	Same word list for all participants; words matched for frequency, length and familiarity, and not semantically related (to avoid organisation/chunking confounds); constant presentation rate (e.g. one word every 2 seconds); same recall time (e.g. 90 seconds); standardised instructions; quiet, distraction-free setting

Design

A repeated-measures design is appropriate: each participant contributes a recall score for both the start-and-end positions and the middle positions from the same list, so every participant serves as their own control. This eliminates participant variables between conditions (the two scores come from the same person) and needs fewer participants. Because both conditions are measured from a single list presentation, the usual repeated-measures problems of order effects and the need for counterbalancing largely fall away here — a neat feature of the serial-position paradigm.

Key Definition: In a repeated-measures design, the same participants take part in all conditions, so their scores in each condition can be compared directly; it controls participant variables but (in most studies) risks order effects.

Sampling

Target population: adults (for example, sixth-form or college students, given the school setting).
Sampling method: an opportunity sample — recruiting people who are readily available (e.g. students free during a study period) — is realistic and practical for a student project, though it is not representative and may be biased (a limitation to acknowledge). A volunteer sample (advertising for participants) is an equally acceptable alternative.
Sample size: a reasonable number for a student practical — for example, around 20–30 participants — enough to run the inferential test meaningfully while remaining feasible.

Procedure (standardised)

Brief each participant and obtain informed consent; explain they may withdraw at any time.
Seat the participant in a quiet room. Read the standardised instructions: "You will see a list of 20 words, one at a time. Afterwards, write down as many as you can remember, in any order."
Present the same 20-word list at a constant rate (one word every 2 seconds), using matched, unrelated words.
Immediately after the last word, give the participant 90 seconds to write down all the words they can recall (free recall).
Debrief the participant, explaining the serial-position effect and the aim, and answer any questions.
Score each participant's recall: count how many of their recalled words came from the start-and-end positions and how many from the middle positions.

flowchart TD
    A["Brief + informed consent"] --> B["Standardised instructions"]
    B --> C["Present 20-word list<br/>(1 word / 2 s; matched, unrelated words)"]
    C --> D["Immediate free recall<br/>(90 s to write words)"]
    D --> E["Debrief"]
    E --> F["Score recall by position:<br/>start-and-end vs middle"]
    F --> G["Analyse: descriptive + inferential stats"]

Ethics

The practical must be ethical, and a memory experiment makes this straightforward — but the safeguards must still be planned and stated.

Ethical issue	How it is addressed
Informed consent	Participants are told the general nature of the task and agree before taking part (full details of the hypothesis are given at debrief to avoid demand characteristics).
Right to withdraw	Participants are told they may stop at any time and may withdraw their data afterwards.
Protection from harm	The task is low-risk — recalling words causes no more than mild effort; no distress is expected.
Confidentiality / data protection	Data are anonymised (no names on recall sheets); results are stored securely and reported only in aggregate.
Debriefing	A full debrief explains the true aim and the serial-position effect, and offers the chance to ask questions and withdraw.
Deception (minimal)	Only the precise hypothesis is withheld (to prevent participants deliberately rehearsing middle items); this minor, justified withholding is disclosed at debrief.

Describing the data: descriptive statistics

Once scored, the data are summarised before being tested.

Measure of central tendency. Because the DV (number of words recalled) is interval-level data that is usually reasonably symmetrical, the mean ( $\bar{x}$ ) is the appropriate average — for example, the mean number of start-and-end words recalled versus the mean number of middle words recalled. The mean is defined as $\bar{x} = \frac{\sum x}{n}$ where $\sum x$ is the sum of scores and $n$ the number of scores. If the recall scores were skewed (for example, if a few participants recalled almost nothing), the median (the middle value when scores are ordered) would be the more robust average, because it is not distorted by extreme scores. Reporting a measure of dispersion — the standard deviation (or the range) — alongside the mean shows how spread out the scores are.
Suitable graph. A bar chart is the appropriate graph, because the IV (position category: start-and-end vs middle) is categorical/discrete. The bar chart would show two bars — mean words recalled from start-and-end positions, and mean words recalled from the middle — making the predicted difference visible at a glance. (If, instead, we plotted mean recall against every serial position 1–20, a line graph would display the classic U-shaped serial-position curve — the primacy and recency "humps" with the middle trough — which is a compelling way to present the full pattern.)

Serial position band	Expected mean recall	Serial-position interpretation
Start (positions 1–5)	High	Primacy — rehearsed into LTM
Middle (positions 6–15)	Low	Neither rehearsed enough nor still in STM
End (positions 16–20)	High	Recency — still active in STM at recall

Analysing the data: a suitable inferential test

To decide whether the difference between the two conditions is statistically significant (unlikely to be due to chance) rather than merely a difference in our sample, an inferential test is used. Choosing the correct test follows three questions: what are we looking for, what design, what level of data?

We are testing for a difference (not a correlation).
The design is repeated measures (each participant gives a score in both conditions — the two scores are related).
The DV is interval-level data (number of words recalled).

For a test of difference, with a related (repeated-measures) design and interval data, the appropriate parametric test is the related (paired) t-test. (If the interval-data assumptions for a parametric test were not met — for example, markedly non-normal scores — the equivalent non-parametric test for a difference with a related design would be the Wilcoxon signed-ranks test, which uses ordinal ranking and is the safer choice for a small student sample.)

Test-choice summary: difference + related/repeated-measures design + interval data → related t-test (parametric); the non-parametric equivalent is the Wilcoxon signed-ranks test.

The test yields a test statistic and an associated probability (p). The conventional significance level in psychology is $p \le 0.05$ meaning we accept up to a 5% probability that a result this extreme could have arisen by chance. If the calculated probability is ≤ 0.05, the result is significant: we reject the null hypothesis and accept the (directional) experimental hypothesis that start-and-end words are recalled significantly better than middle words. If $p > 0.05$ , we retain the null hypothesis. Because the hypothesis is directional, a one-tailed test is used (the critical value is looked up accordingly).

Cognitive Psychology: Key Question & Practical