You are viewing a free preview of this lesson.
Subscribe to unlock all 10 lessons in this course and every other course on LearningBro.
Every Edexcel Cognitive topic is anchored by a classic study that students must be able to describe and evaluate in depth, and for cognitive psychology that study is Baddeley (1966) — the word-list experiments on acoustic and semantic coding in short-term and long-term memory. Alan Baddeley set out to answer a deceptively simple question: in what form does memory store information — as sounds or as meanings? — and, crucially, whether short-term memory (STM) and long-term memory (LTM) answer that question differently. His finding — that STM is confused by acoustic (sound-based) similarity while LTM is confused by semantic (meaning-based) similarity — became one of the most important pieces of evidence that STM and LTM are separate stores with different codes, directly supporting the multi-store model. This lesson gives the study in full — aim, method, procedure, results and conclusion — then evaluates it methodologically and ethically and situates it in the wider theory.
Key Definition: Coding (or encoding) is the form in which information is represented in memory. Acoustic coding represents information by its sound; semantic coding represents it by its meaning. Baddeley's classic study asked whether STM and LTM code information in the same way.
This lesson addresses the Edexcel 9PS0 — Paper 1, Topic 2: Cognitive Psychology requirement to know a classic study in depth: Baddeley (1966) on the effect of acoustic and semantic similarity on short- and long-term memory for word sequences. You must be able to state its aim, method/procedure, results and conclusion, and to evaluate it methodologically (design, controls, validity, generalisability) and ethically, and to explain its contribution to the multi-store model. In assessment-objective terms, you should be able to describe the study accurately (AO1), apply its findings to explaining coding in memory (AO2), and evaluate its methodology, ethics and theoretical role (AO3).
Connects to…
Before Baddeley, it was known that immediate memory tended to make sound-based errors (recalling "cat" as "cap"), suggesting acoustic coding, but it was unclear whether long-term memory worked the same way. Baddeley reasoned that if a memory store codes information in a particular form, then material that is similar in that form should be harder to remember, because the similar items become confusable. Acoustically similar words (sound alike) should therefore disrupt an acoustic store, and semantically similar words (mean the same) should disrupt a semantic store.
Key Aim: To investigate whether short-term and long-term memory for a sequence of words is affected differently by the acoustic similarity (words that sound alike) or semantic similarity (words that mean the same) of those words — and hence to establish the coding used by each store.
The elegant logic is that the pattern of interference reveals the code: whichever kind of similarity disrupts a store tells you the form in which that store represents information. If STM is disrupted by acoustic similarity but LTM by semantic similarity, the two stores must code differently — and therefore be distinct systems.
Baddeley used a laboratory experiment with an independent-groups design: different participants experienced different word-list conditions, so no participant saw more than one list type (avoiding practice and order effects across conditions). The participants were drawn from the Applied Psychology Unit panel in Cambridge (members of the public who volunteered for studies).
Participants were presented with sequences of five words drawn from one of four carefully constructed lists. The independent variable was the type of similarity of the word list.
| List | Type | Example words | Prediction if store codes this way |
|---|---|---|---|
| A | Acoustically similar | man, cad, cap, can, cat, mad… (sound alike) | Poor recall if the store codes acoustically |
| B | Acoustically dissimilar (control for A) | pit, few, cow, pen, sup… (sound different) | Better recall than List A |
| C | Semantically similar | great, large, big, huge, broad… (mean the same) | Poor recall if the store codes semantically |
| D | Semantically dissimilar (control for C) | good, huge, hot, safe, thin… (unrelated meanings) | Better recall than List C |
Lists B and D are the crucial controls: comparing acoustically similar against dissimilar (A vs B) isolates the effect of acoustic similarity; comparing semantically similar against dissimilar (C vs D) isolates the effect of semantic similarity. The dissimilar lists were also matched for word frequency and familiarity so that any difference could be attributed to similarity, not to the words being harder in some other way.
The genius of the design lies in how STM and LTM were separated. Participants heard the list, then had to recall the words in the correct order (serial recall). To probe STM, recall was tested immediately after presentation. To probe LTM, participants first completed an interference task (learning and recalling other, unrelated word sequences) that filled a delay of around 20 minutes, and were then asked to recall the original list. The reasoning is that immediate recall draws on STM, whereas recall after a filled 20-minute delay must draw on LTM (since the original list can no longer be held in the fragile short-term store).
Participants completed several learning trials and were retested; the LTM measure came after repeated learning trials followed by the delay, so that the material had genuinely entered long-term storage.
flowchart TD
START["Participant assigned to ONE list type<br/>(A, B, C or D)"] --> HEAR["Hear sequence of 5 words"]
HEAR --> IMM["IMMEDIATE serial recall<br/>= tests STM"]
HEAR --> LEARN["Repeated learning trials<br/>+ ~20-min interference task<br/>(other word sequences)"]
LEARN --> DELAY["DELAYED serial recall<br/>= tests LTM"]
IMM --> Q1["Does ACOUSTIC or SEMANTIC<br/>similarity impair STM?"]
DELAY --> Q2["Does ACOUSTIC or SEMANTIC<br/>similarity impair LTM?"]
Key Definition: In an independent-groups design, each participant takes part in only one condition, so different people are compared across conditions. It avoids order/practice effects but introduces the risk that the groups differ in participant variables (addressed here by randomisation and matched control lists).
The two similarity manipulations produced strikingly different effects on immediate (STM) versus delayed (LTM) recall.
Interpretation: immediate memory is confused by items that sound alike, indicating that STM codes acoustically.
Interpretation: long-term memory is confused by items that mean the same, indicating that LTM codes semantically.
| Immediate recall (STM) | Delayed recall (LTM) | |
|---|---|---|
| Acoustically similar (vs dissimilar) | Large impairment — STM codes acoustically | Little/no impairment |
| Semantically similar (vs dissimilar) | Small impairment | Substantial impairment — LTM codes semantically |
This double dissociation of coding — acoustic similarity hurts STM but not LTM; semantic similarity hurts LTM but not STM — is the headline finding.
Baddeley concluded that STM and LTM use different codes: information in short-term memory is coded predominantly acoustically (by sound), whereas information in long-term memory is coded predominantly semantically (by meaning). Because the two stores are confused by different kinds of similarity, they must be functionally distinct systems, not merely the same memory examined at different time points.
This conclusion mattered enormously for theory. It provided some of the strongest experimental evidence for the multi-store model's central claim that STM and LTM are separate stores — and specifically that they differ not only in duration and capacity but in coding, the dimension hardest to establish by other means. It also foreshadowed Baddeley's own later work: an acoustically coded short-term store is exactly what the phonological loop of the working memory model would later formalise.
A careful statement of the conclusion notes the word predominantly. Baddeley did not claim STM is purely acoustic or LTM purely semantic — the small semantic effect in immediate recall and the fact that some acoustic information can persist show the coding is a matter of emphasis, not absolutes. Under some conditions STM can use a semantic or visual code, and LTM retains some acoustic detail (we can recall how a familiar voice sounds). The robust claim is that the dominant code differs between the two stores.
A major strength is the study's high degree of experimental control, which gives it strong internal validity. Conducted in a laboratory, it standardised the presentation of the word lists, controlled the timing of immediate versus delayed recall, and — crucially — matched the dissimilar control lists (B and D) to the similar lists for word frequency and familiarity, so that any recall difference could be attributed to similarity rather than to some other property of the words. This matters because it allows a confident causal inference: the type of similarity caused the pattern of impairment. The implication is that the study's core finding — that STM and LTM are disrupted by different kinds of similarity — rests on a rigorously controlled design that isolates the variable of interest, which is precisely why the result has been so influential and replicable.
A second strength is the study's theoretical contribution: it provides some of the clearest evidence that STM and LTM are separate stores with different codes. The finding that acoustic similarity impairs STM but not LTM, while semantic similarity impairs LTM but not STM, is a double dissociation of coding that is very difficult to explain if memory is a single, uniform store. This matters because coding is the dimension on which the stores' distinctness is otherwise hardest to demonstrate — capacity and duration differences could, in principle, reflect one store operating differently over time, but a qualitative difference in code strongly implies two systems. The implication is that Baddeley's study underpins the MSM's foundational architecture and, by establishing acoustic coding in STM, directly seeded the working memory model's phonological loop — giving it a lasting dual role in memory theory.
Set against these strengths, the study has been criticised for low ecological validity. The task — learning sequences of unrelated single words in a laboratory and recalling them in order — bears little resemblance to the meaningful, contextualised information we typically remember in everyday life (conversations, events, stories). This matters because the coding people use may depend on the material: for meaningful real-world material, semantic coding may dominate even in the short term, so a study using decontextualised word lists may tell us more about laboratory memory than about how memory codes information in daily life. The implication is that, while the study demonstrates a real difference in coding for this kind of material, its findings should be generalised cautiously to naturalistic remembering — the acoustic-STM/semantic-LTM distinction is best regarded as a robust tendency established under artificial conditions rather than an exceptionless law of everyday memory.
A more technical limitation concerns whether the delayed test measured LTM purely. The delayed condition involved repeated learning trials as well as a filled delay, so the "LTM" performance reflects a learning process over trials, and the boundary between a well-rehearsed STM trace and a genuine LTM representation is not razor-sharp. This matters because it introduces a degree of ambiguity into exactly which store the delayed measure taps, which could in principle blur the interpretation. The implication is that, although the reversal of the similarity effect across immediate and delayed recall is strong evidence for a coding difference, the operationalisation of "LTM" as post-delay, post-learning recall is a simplification — a caveat that a sophisticated evaluation notes without overturning the study's central conclusion, which the clear pattern of results supports.
The independent-groups design is a double-edged feature worth evaluating explicitly. Its strength is that, because each participant experienced only one list type, there were no order or practice effects across conditions — a participant could not carry over learning from an acoustic list to a semantic one. Its weakness is that different people were compared across conditions, so participant variables (differences in memory ability between the groups) could in principle confound the results. This matters because a critic could argue that the acoustically-similar group simply happened to have poorer memories. The implication is that the design choice was reasonable — avoiding order effects in a study about interference is important — and the risk of participant-variable confounds was mitigated by using reasonably sized groups and matched materials; but a repeated-measures or matched-pairs design would have controlled participant differences more directly, and acknowledging this trade-off is the mark of a strong evaluation.
Subscribe to continue reading
Get full access to this lesson and all 10 lessons in this course.