Memory

Introduction

Memory is the process by which information is encoded, stored, and retrieved. Understanding how memory works — and how it fails — is central to cognitive psychology. This section covers the multi-store model, the working memory model, types of long-term memory, theories of forgetting, eyewitness testimony, and the cognitive interview.

Key Concepts

The Multi-Store Model (MSM)

Atkinson and Shiffrin (1968) proposed that memory consists of three separate stores:

1. Sensory register: Briefly holds information from the senses. Each sense has its own register (iconic for visual, echoic for auditory). Information decays within milliseconds unless attention is paid, which transfers it to short-term memory. Capacity is large (all sensory input).

2. Short-term memory (STM): A temporary store with limited capacity (Miller, 1956: 7 ± 2 items) and limited duration (Peterson and Peterson, 1959: approximately 18–30 seconds without rehearsal). Information is encoded primarily acoustically. Maintenance rehearsal (repeating information) keeps it in STM; elaborative rehearsal transfers it to long-term memory.

3. Long-term memory (LTM): A potentially permanent store with unlimited capacity and unlimited duration. Information is encoded primarily semantically (by meaning).

Key features of the model:

• Information flows through the system in a linear fashion: sensory register → STM → LTM.
• Rehearsal is the key process that transfers information from STM to LTM.
• Each store differs in coding, capacity, and duration.

Evidence for separate stores:

• Capacity: Miller (1956) found that STM capacity is approximately 7 ± 2 items. LTM has no known capacity limit.
• Duration: Peterson and Peterson (1959) found STM duration is about 18–30 seconds. Bahrick et al. (1975) found LTM for Spanish vocabulary lasted over 30 years.
• Coding: Baddeley (1966) found STM uses acoustic coding (confusion between similar-sounding words) while LTM uses semantic coding (confusion between similar-meaning words).
• Case studies: HM (Scoville and Milner, 1957) — after bilateral hippocampus removal, HM could not form new long-term memories but his STM was intact. Clive Wearing — viral encephalitis damaged his hippocampus; his STM functioned but he could not transfer information to LTM.

The Working Memory Model (WMM)

Baddeley and Hitch (1974) proposed that STM is not a single store but an active system with multiple components:

1. Central executive: The attentional control centre. It directs attention to particular tasks, determining how resources are allocated. It has limited capacity and processes information from all sensory modalities but does not store information itself.

2. Phonological loop: Deals with auditory information. Subdivided into:

   • Phonological store: Holds auditory information for approximately 2 seconds (the “inner ear”).
   • Articulatory control process: Sub-vocal rehearsal (the “inner voice”) that maintains information by repeating it.

   Evidence: The word length effect — people find it harder to recall a list of long words than short words because long words take longer to rehearse (Baddeley et al., 1975). The articulatory suppression effect — repeating an irrelevant sound blocks the articulatory process, eliminating the word length effect.

3. Visuospatial sketchpad: Stores and manipulates visual and spatial information (the “inner eye”). Limited capacity — Logie (1995) estimated approximately 3–4 objects.

   Evidence: Baddeley et al. (1975) found that a visual task (tracking a spot of light) interfered with another visual task but not a verbal task, demonstrating the separation of visual and verbal processing.

4. Episodic buffer (added 2000): A general store that integrates information from the central executive, phonological loop, visuospatial sketchpad, and long-term memory into a unified episode. It has limited capacity (approximately 4 chunks) and provides a bridge between working memory and LTM.

   Evidence: Patients with amnesia who cannot form new LTM can still recall stories they have just heard beyond the capacity of the phonological loop alone, suggesting an additional temporary store (the episodic buffer).

Types of Long-Term Memory

1. Episodic memory: Personal experiences and events, tagged with time and place (e.g., your first day at school). Involves the hippocampus and prefrontal cortex.
2. Semantic memory: General knowledge, facts, and concepts not linked to personal experience (e.g., Paris is the capital of France). Involves the temporal lobe.
3. Procedural memory: Skills and actions performed automatically without conscious awareness (e.g., riding a bike, typing). Involves the basal ganglia and cerebellum.

Evidence: Tulving et al. (1994) used PET scans to show that episodic and semantic memories activate different brain regions. Case studies (e.g., HM) show that some types of LTM can be impaired while others remain intact.

Interference Theory

The idea that forgetting occurs because other memories interfere with retrieval.

• Proactive interference (PI): Old memories interfere with the recall of new information (e.g., your old phone number makes it harder to remember your new one).
• Retroactive interference (RI): New memories interfere with the recall of old information (e.g., learning new phone number makes it harder to recall the old one).

Evidence: McGeoch and McDonald (1931) found that the more similar the interfering information, the greater the forgetting. Participants learned a list of words, then learned a second list. Forgetting was greatest when the second list consisted of synonyms of the first list.

Retrieval Failure Theory

Forgetting occurs because the cues available at retrieval do not match the cues present at encoding. This is based on the encoding specificity principle (Tulvin, 1972): a cue needs to be present at encoding and retrieval for it to be effective.

• Context-dependent forgetting: External environmental cues (e.g., Godden and Baddeley, 1975 — divers recalled words best in the same environment where they learned them: underwater-underwater or dry-dry).
• State-dependent forgetting: Internal physiological cues (e.g., Carter and Cassaday, 1998 — participants recalled more when their learning and recall states matched — antihistamine or placebo).

Eyewitness Testimony (EWT)

Loftus and Palmer (1974):

Aim: To investigate the effect of leading questions on eyewitness testimony.

Procedure (Experiment 1): 45 American students watched film clips of car accidents and were asked questions about them. The critical question was “About how fast were the cars going when they ****ed each other?” The verb was manipulated: smashed, collided, bumped, hit, or contacted.

Findings: | Verb | Mean speed estimate (mph) | |---|---| | Smashed | 40.5 | | Collided | 39.3 | | Bumped | 38.1 | | Hit | 34.0 | | Contacted | 31.8 |

Procedure (Experiment 2): Participants watched a film and were asked either “smashed” or “hit” questions. One week later, they returned and were asked “Did you see any broken glass?” (there was none).

Findings: Participants in the “smashed” condition were more likely to report seeing broken glass (32% vs. 14% in the “hit” condition).

Explanation: Loftus proposed two explanations:

1. Response-bias factor: The wording of the question influences the participant’s response without changing the memory itself.
2. Substitution: The misleading information actually alters the original memory, replacing it with new (false) information.

Factors affecting EWT:

• Anxiety: The weapon focus effect (Loftus et al., 1987) — when a weapon is present, witnesses focus on the weapon rather than the perpetrator, reducing identification accuracy. The Yerkes-Dodson law suggests moderate anxiety improves recall, but very high anxiety impairs it.
• Age: Children and the elderly are generally less accurate witnesses. Children are more susceptible to leading questions and post-event misinformation.
• Own-age bias: People are better at recognising faces of their own age group (Anastasi and Rhodes, 2005).

The Cognitive Interview (Fisher and Geiselman, 1992)

A method of police interviewing designed to improve the accuracy and quantity of eyewitness recall. Based on four key principles:

1. Report everything: Witnesses are asked to report every detail, even if it seems trivial or irrelevant. Seemingly unimportant details may be cues that trigger recall of more significant information.
2. Context reinstatement: Witnesses are asked to mentally recreate the environment and their emotional state at the time of the event. This is based on context-dependent memory (cuing).
3. Reverse the order: Witnesses are asked to recall events in reverse chronological order. This makes it harder to rehearse a false narrative and encourages genuine retrieval.
4. Change perspective: Witnesses are asked to recall the event from a different perspective (e.g., from the viewpoint of another person present). This creates new retrieval pathways.

Enhanced cognitive interview: Adds features such as minimising distractions, building rapport, asking open-ended questions, and pausing to allow witnesses time to think.

Key Studies

Study	Researcher(s)	Year	Method	Key Findings	Evaluation
Magical number seven	Miller	1956	Lab experiment	STM capacity is 7 ± 2 items	Influential; later research (Cowan, 2001) suggests 4 ± 1 chunks
Duration of STM	Peterson & Peterson	1959	Lab experiment	STM duration ≈ 18–30 seconds without rehearsal	Well controlled; artificial stimuli (trigrams); interference may be confound
Duration of LTM	Bahrick et al.	1975	Natural experiment	LTM for Spanish lasted 30+ years; recognition better than recall	High ecological validity; no control of rehearsal between tests
Coding in STM and LTM	Baddeley	1966	Lab experiment	STM: acoustic coding; LTM: semantic coding	Clear evidence for separate coding; used word lists (artificial)
Leading questions	Loftus & Palmer	1974	Lab experiment	Verb in question affected speed estimate and false memory of broken glass	Well controlled; artificial film clips; ethical issues (deception)
Working memory model	Baddeley & Hitch	1974	Dual-task experiments	Dual tasks using same component caused interference; different components did not	Strong empirical support; less evidence for central executive
Diver memory	Godden & Baddeley	1975	Field experiment	Best recall when learning and recall contexts matched	Supports context-dependent memory; used word lists (low mundane realism)

Key Terminology

Term	Definition
Encoding	Converting information into a form that can be stored in memory
Storage	Retaining encoded information in memory over time
Retrieval	Accessing stored information from memory
Rehearsal	The process of repeating information to maintain it in STM or transfer it to LTM
Chunking	Grouping individual pieces of information into larger units to increase STM capacity
Acoustic coding	Representing information in memory based on its sound
Semantic coding	Representing information in memory based on its meaning
Visual coding	Representing information in memory based on its visual properties
Proactive interference	When previously learned information interferes with the recall of new information
Retroactive interference	When newly learned information interferes with the recall of previously learned information
Retrieval cue	A stimulus that helps access a stored memory
Context-dependent memory	Improved recall when the external environment matches the encoding context
State-dependent memory	Improved recall when the internal physiological state matches the encoding state
Leading question	A question phrased in a way that suggests a particular answer or contains misleading information
Weapon focus	The tendency for witnesses to focus on a weapon, reducing identification accuracy
Cognitive interview	A police interviewing technique designed to maximise accurate eyewitness recall
Long-term potentiation	The strengthening of synaptic connections through repeated stimulation, underlying learning

Evaluation Points

Strengths of the Multi-Store Model

• Empirical support: Numerous studies demonstrate differences between STM and LTM in capacity, duration, and coding (Miller, Peterson and Peterson, Baddeley).
• Case study evidence: Patients like HM and Clive Wearing show that damage to the hippocampus impairs LTM formation while STM remains intact, supporting the idea of separate stores.
• Clarity and testability: The model is clear and straightforward, generating specific, testable predictions about memory.

Limitations of the Multi-Store Model

• Oversimplified: The model treats STM and LTM as unitary stores. The WMM demonstrates that STM has multiple components, and research shows LTM has multiple types (episodic, semantic, procedural).
• Rehearsal is not the only route: The model claims rehearsal is the primary mechanism for transferring information to LTM, but people form long-term memories without conscious rehearsal (e.g., flashbulb memories).
• Linear processing: The model suggests information flows in one direction, but real memory involves complex, bidirectional interactions between stores.

Strengths of the Working Memory Model

• Empirical support: Dual-task studies demonstrate that tasks using the same component interfere with each other, while tasks using different components do not (Baddeley et al., 1975).
• Clinical evidence: Patients with specific brain damage show selective impairment of certain WMM components (e.g., patient KF had impaired phonological loop but intact visuospatial sketchpad).
• Practical applications: The model has been applied to understanding reading difficulties, working memory deficits in ADHD, and educational strategies.

Limitations of the Working Memory Model

• Central executive is vague: The central executive is poorly defined and difficult to test empirically. Baddeley himself (2003) admitted it is “the most important but least understood component.”
• Limited to STM: The model does not address LTM in detail, despite the episodic buffer’s connection to it.
• Incomplete: Additional components may exist (e.g., an “episodic buffer” was only added in 2000, suggesting the model was initially incomplete).

Methodology

Memory experiments commonly use:

• Laboratory experiments with standardised word lists, digit spans, and recall tasks. High control but low ecological validity.
• Case studies of patients with brain damage (HM, Clive Wearing, KF). Provide rich data but cannot be generalised.
• Field experiments (Godden and Baddeley’s diver study). Higher ecological validity but less control.
• Neuroimaging (PET scans, fMRI) to identify brain regions involved in memory processes. Objective but correlational.

Common Pitfalls

1. Confusing the multi-store model and the working memory model: The MSM (Atkinson and Shiffrin) describes three sequential memory stores. The WMM (Baddeley and Hitch) describes the internal structure of STM only. They are complementary, not interchangeable.
2. Confusing proactive and retroactive interference: Proactive = old interferes with new (forward in time). Retroactive = new interferes with old (backward in time). A useful mnemonic: “Proactive = Past interferes; Retroactive = Recent interferes.”
3. Oversimplifying EWT research: Loftus and Palmer is not the only study on EWT. Anxiety (weapon focus), age, and the cognitive interview are equally important. Leading questions are just one factor that affects witness accuracy.

Worked Examples

Example 1: 16-Mark Essay

Question: Outline and evaluate the working memory model. [16 marks]

Model Answer:

The working memory model (WMM) was proposed by Baddeley and Hitch (1974) as an alternative to the unitary short-term memory store described in the multi-store model. The WMM suggests that STM is not a single store but an active processing system composed of multiple components that handle different types of information simultaneously.

The central executive is the core component, responsible for directing attention, coordinating the activities of the slave systems, and synthesising information. It has limited capacity but processes information from all sensory modalities. Baddeley (2000) described it as a “supervisory attentional system.”

The phonological loop processes auditory and verbal information. It consists of the phonological store, which holds speech-based information for approximately two seconds, and the articulatory control process, which rehearses information sub-vocally to prevent decay. This explains why we find it harder to recall a list of long words compared to short words — the word length effect — because longer words take more time to rehearse.

The visuospatial sketchpad processes visual and spatial information, such as remembering the layout of a room or navigating a route. Logie (1995) suggested it has a capacity of approximately three to four objects. Baddeley et al. (1975) demonstrated its existence using dual-task paradigms: participants tracking a visual target showed impaired performance on a second visual task but not on a verbal task, confirming separate visual and verbal stores.

In 2000, Baddeley added the episodic buffer, a general store that integrates information from the central executive, the slave systems, and long-term memory into a coherent episode. It addresses the model’s earlier inability to explain how we integrate multiple types of information and how working memory interacts with LTM.

A major strength of the WMM is its strong empirical support from dual-task studies. Baddeley et al. (1975) found that two visual tasks impaired each other, but a visual task and a verbal task did not. This provides direct evidence for separate slave systems processing different types of information. The model’s predictions have been consistently replicated, giving it high reliability.

Clinical evidence also supports the WMM. Patient KF (Shallice and Warrington, 1970) suffered brain damage that impaired his verbal STM (phonological loop) but left his visual STM (visuospatial sketchpad) intact. This double dissociation provides compelling evidence that STM is not a unitary store but consists of independent components, exactly as the WMM predicts.

However, the central executive has been criticised as poorly defined and difficult to test. Baddeley himself acknowledged that the central executive is “the most important but least understood component of working memory.” It is unclear how the central executive allocates resources or whether it might consist of multiple sub-components. This lack of clarity limits the model’s explanatory power and makes it difficult to generate testable hypotheses about the central executive specifically.

Furthermore, the WMM focuses on STM and does not provide a comprehensive account of how memory operates as a whole system. It does not adequately explain the relationship between working memory and long-term memory, although the addition of the episodic buffer was an attempt to address this limitation.

In conclusion, the WMM represents a significant advance over the multi-store model by demonstrating that STM is an active, multi-component system. It is well supported by empirical and clinical evidence. However, the vagueness of the central executive and the limited attention to LTM mean it is still an incomplete account of human memory.

Example 2: 16-Mark Essay

Question: Discuss research into the accuracy of eyewitness testimony. Refer to evidence in your answer. [16 marks]

Model Answer:

Eyewitness testimony (EWT) refers to the account given by witnesses to a crime or event. It plays a crucial role in the criminal justice system, yet psychological research has consistently demonstrated that EWT is often unreliable. This essay discusses research into misleading information and anxiety, and evaluates the cognitive interview as a method for improving EWT accuracy.

Loftus and Palmer (1974) conducted two experiments demonstrating the effect of leading questions on EWT. In the first experiment, 45 participants watched film clips of car accidents and were asked about the speed of the vehicles using different verbs (smashed, collided, bumped, hit, contacted). The verb “smashed” produced significantly higher speed estimates (40.5 mph) than “contacted” (31.8 mph). In the second experiment, participants who had been asked the “smashed” question were more likely to report seeing broken glass one week later (32% vs. 14% in the “hit” condition), even though no glass was present. Loftus argued that the misleading information had either altered the original memory (substitution) or created a response bias.

However, Yuille and Cutshall (1986) challenged the ecological validity of laboratory EWT studies. They interviewed 13 witnesses to a real armed robbery in Vancouver four to five months after the event. Witnesses provided highly accurate accounts that were resistant to leading questions, and those who were most distressed at the time were most accurate. This suggests that EWT for real, emotionally significant events may be more reliable than laboratory studies using film clips would suggest.

Anxiety also affects EWT. The weapon focus effect (Loftus et al., 1987) describes the phenomenon where the presence of a weapon draws a witness’s attention away from the perpetrator’s face, reducing identification accuracy. In one study, participants saw a customer holding a gun or a chequebook. Those who saw the gun were less accurate in identifying the customer. This is explained by the Yerkes-Dodson law: moderate arousal improves performance, but very high arousal impairs it.

However, the relationship between anxiety and EWT is not straightforward. Christianson and Hubinette (1993) studied witnesses to real bank robberies and found that victims (who presumably experienced the most anxiety) were actually more accurate than bystanders. This contradicts the weapon focus effect and suggests that very high anxiety can sometimes improve recall, particularly for central details of a personally relevant event.

The cognitive interview (Fisher and Geiselman, 1992) was developed to improve the accuracy of EWT. It uses four techniques: report everything, context reinstatement, reverse order, and change perspective. Meta-analyses (e.g., Stein and Memon, 2006) have found that the cognitive interview produces significantly more accurate information than standard police interviews. The enhanced cognitive interview, which also builds rapport and minimises distractions, is even more effective.

However, the cognitive interview is time-consuming and requires extensive training. Police officers may not have the time or resources to conduct a full cognitive interview, particularly for minor offences. Furthermore, the cognitive interview can increase both accurate and inaccurate recall, as witnesses may report more incorrect details when encouraged to report everything. Kebbell and Wagstaff (1999) also noted that some police officers find the techniques impractical in real investigations.

In conclusion, research demonstrates that EWT is vulnerable to distortion from misleading information, anxiety, and other factors. However, real-world evidence suggests that memory for emotionally significant events may be more reliable than laboratory studies indicate. The cognitive interview offers a practical method for improving EWT accuracy, though it is not without limitations. A balanced approach, combining laboratory control with real-world research, provides the most complete understanding of eyewitness reliability.

Summary

Memory research reveals a complex, multi-component system:

• The multi-store model (Atkinson and Shiffrin) describes three stores (sensory register, STM, LTM) with different capacities, durations, and coding.
• The working memory model (Baddeley and Hitch) describes STM as an active system with four components: central executive, phonological loop, visuospatial sketchpad, and episodic buffer.
• Long-term memory includes episodic (events), semantic (facts), and procedural (skills) types, each associated with different brain regions.
• Forgetting can be explained by interference (proactive and retroactive) and retrieval failure (context-dependent and state-dependent).
• Eyewitness testimony is affected by leading questions, anxiety, and age; the cognitive interview improves accuracy through context reinstatement, reporting everything, reversing order, and changing perspective.