Memory Models

The Multi-Store Model of Memory

Atkinson and Shiffrin (1968)

The multi-store model (MSM) proposed by Atkinson and Shiffrin is one of the most influential models In cognitive psychology. It conceptualises memory as consisting of three structural stores through Which information flows in a linear sequence.

The Three Stores

Sensory memory is the first store in the sequence. It holds incoming sensory information very Briefly, preserving a near-exact copy of the sensory input for a duration of approximately 200—500 Milliseconds for visual information (iconic memory) and approximately 2—3 seconds for auditory Information (echoic memory). Sensory memory has a large capacity but extremely limited duration.

Evidence for sensory memory comes from Sperling (1960), who demonstrated that participants shown a 3-by-4 grid of letters for 50 milliseconds could report only 4—5 items in a full-report condition But could accurately report any specified row when cued by a tone immediately after the display. This partial-report advantage demonstrates that all 12 items were briefly available in sensory Memory but decayed before they could be reported.

Short-term memory (STM) has a limited capacity of approximately $7 \pm 2$ items (Miller, 1956) And a limited duration of approximately 18—30 seconds without rehearsal (Peterson and Peterson, 1959). Information is maintained in STM through maintenance rehearsal — the passive repetition of Information without necessarily processing its meaning. The capacity of STM can be effectively Increased through chunking, the process of grouping individual items into larger meaningful units (e.g., the sequence 1-9-8-4 is more remembered as the single chunk “1984”).

Long-term memory (LTM) has a theoretically unlimited capacity and potentially unlimited Duration. Information enters LTM through elaborative rehearsal — the active processing of meaning, Associations, and connections between new information and existing knowledge. LTM is not a single Store but comprises several distinct types:

• Episodic memory: Memory for personal experiences and events, situated in a specific time and place (e.g., remembering your first day of school).
• Semantic memory: General knowledge about the world, independent of personal experience (e.g., knowing that Paris is the capital of France).
• Procedural memory: Memory for skills and procedures (e.g., riding a bicycle, typing).

Processes in the MSM

1. Attention: Information from the environment is attended to and transferred from sensory memory to STM. Most sensory information is not attended to and is rapidly lost.
2. Maintenance rehearsal: Information in STM is maintained through repetition. Without rehearsal, information decays within approximately 30 seconds.
3. Elaborative rehearsal: Deep processing of meaning transfers information from STM to LTM.
4. Retrieval: Information stored in LTM can be retrieved back into STM when needed.

Evaluation of the MSM

Strengths:

• The model is supported by clinical evidence from case studies. Patient HM (Scoville and Milner, 1957), who had bilateral removal of the hippocampus to treat epilepsy, could form no new long-term memories (anterograde amnesia) but had intact short-term memory. This double dissociation supports the distinction between STM and LTM as separate stores.
• The model has generated a vast body of research and has been highly influential in shaping the study of memory.
• The serial position effect (Glanzer and Cunitz, 1966) provides behavioural evidence for the distinction between STM and LTM. In a free recall task, participants show primacy (superior recall of items at the beginning of the list, attributed to transfer to LTM) and recency (superior recall of items at the end of the list, attributed to items still being held in STM).

Limitations:

• The model is overly simplistic. It treats STM as a unitary store, but the working memory model (see below) demonstrates that STM consists of multiple components serving different functions.
• The model overemphasises the role of rehearsal in transferring information to LTM. Craik and Lockhart (1972) demonstrated that the depth of processing, not the amount of rehearsal, determines whether information is remembered.
• The model treats LTM as a unitary store, but research has identified multiple distinct types of LTM (episodic, semantic, procedural), each with different properties and neural substrates.
• The linear flow of information is an oversimplification. Information does not always flow in one direction; retrieval from LTM can influence how information is encoded in STM.

The Working Memory Model

Baddeley and Hitch (1974)

Baddeley and Hitch proposed the working memory model as a more detailed and accurate description of Short-term memory. Rather than a single passive store, working memory is conceptualised as an active Processing system consisting of multiple components.

Components of the Working Memory Model

The central executive is the attentional control system that coordinates the activities of the Other components. It allocates processing resources, suppresses irrelevant information, switches Between tasks, and integrates information from different sources. It has limited capacity but is not Itself a storage system. Baddeley acknowledged that the central executive is the least Well-specified component of the model.

The phonological loop processes auditory and verbal information and consists of two Subcomponents:

• The phonological store: A limited-capacity store that holds auditory information in a phonological (sound-based) code. Information in the phonological store decays in approximately 2 seconds unless refreshed.
• The articulatory rehearsal process: A subvocal rehearsal mechanism that refreshes the contents of the phonological store by silently “speaking” the information. This component is subject to the word length effect (longer words take longer to rehearse, so fewer can be maintained) and the phonological similarity effect (lists of phonologically similar words are harder to remember because they compete for the same store).

The visuospatial sketchpad processes visual and spatial information. It is responsible for Creating and manipulating mental images and representing spatial relationships. The visuospatial Sketchpad can be dissociated from the phonological loop: participants can perform a verbal task and A spatial task simultaneously with minimal interference, but two tasks of the same type interfere Strongly with each other.

The episodic buffer (Baddeley, 2000) was added to the model to address the problem of how Information from different modalities is integrated. The episodic buffer is a limited-capacity Temporary storage system that integrates information from the phonological loop, visuospatial Sketchpad, and LTM into a coherent, multi-dimensional episodic representation. This component Explains how participants can recall integrated scenes (e.g., “the red car was parked next to the Tree”) that require combining visual and verbal information.

Baddeley et al. (1975)

Baddeley, Grant, Wight, and Thomson investigated the dual-task interference predicted by the working Memory model. Participants performed a verbal reasoning task (judging whether sentences such as “A Follows B” were true or false given the letter sequence “BA”) while simultaneously performing either A verbal secondary task (reciting digits) or a visual secondary task (tracking a moving light).

Results: Performance on the reasoning task was impaired more by the concurrent verbal task than by The visual task. This supports the existence of separate verbal and visuospatial processing systems, As predicted by the working memory model.

Evaluation of the working memory model:

• Strongly supported by experimental evidence from dual-task paradigms, neuroimaging, and case studies of patients with specific working memory deficits.
• The model has been highly influential in understanding cognitive deficits in developmental disorders (dyslexia, ADHD) and in educational contexts.
• The central executive remains underspecified. Subsequent research has attempted to fractionate it into distinct executive functions (inhibition, shifting, updating; Miyake et al., 2000).

Levels of Processing

Craik and Lockhart (1972)

Craik and Lockhart proposed the levels of processing (LOP) framework as an alternative to the Structural models of memory (the MSM and working memory model). The key claim is that memory is not A function of the store in which information is held or the amount of rehearsal it receives, but Rather of the depth of processing applied to the information at encoding.

Three levels of processing:

1. Structural processing: Processing the physical features of a stimulus (e.g., “Is the word written in uppercase?”). This is the shallowest level and produces the poorest recall.
2. Phonological processing: Processing the sound of a stimulus (e.g., “Does the word rhyme with “cat’?”). This is an intermediate level.
3. Semantic processing: Processing the meaning of a stimulus (e.g., “Does the word fit into the sentence ‘The ____ chased the mouse’?”). This is the deepest level and produces the best recall.

Craik and Tulving (1975)

Craik and Tulving provided the seminal evidence for the LOP effect. Participants were shown words One at a time and asked questions requiring structural, phonological, or semantic processing. In a Surprise recall test, participants recalled significantly more words that had been processed Semantically than those processed phonologically or structurally.

Evaluation of the levels of processing framework:

• The LOP framework has been robustly replicated and provides a parsimonious explanation for memory differences.
• However, the framework has been criticised for being vague and unfalsifiable. What counts as “deep” processing is not precisely defined, making it difficult to generate specific, testable predictions.
• Morris, Bransford, and Franks (1977) demonstrated that the effectiveness of processing depends on the type of memory test. Rhyming (phonological) processing produced better performance on a rhyme recognition test, while semantic processing produced better performance on a standard recognition test. This finding, which cannot be explained by the simple LOP account, led to the development of transfer-appropriate processing theory.

Reconstructive Memory and Schema Theory

Bartlett (1932): “War of the Ghosts”

Frederick Bartlett argued that memory is not a passive recording of experience but an active, Reconstructive process. In his most famous experiment, Bartlett asked British participants to read a Native American folktale, “War of the Ghosts,” and then recall it at increasing intervals (from Immediately to several months later).

Key findings:

• Participants’ recollections became progressively shorter and more distorted with each recall.
• Details that were unfamiliar or inconsistent with British cultural expectations were omitted or altered (e.g., “canoes” became “boats,” “hunting seals” became “fishing”).
• The story became more coherent and conventional, conforming to participants’ existing cultural schemas.

Bartlett introduced the concept of schemas: organised packets of knowledge and expectations About the world that influence how information is encoded, stored, and retrieved. When we encounter New information, we interpret it in terms of our existing schemas, and when we recall information, We reconstruct it using our schemas to fill in gaps and resolve inconsistencies.

Schema Theory

Schema theory, further developed by Brewer and Treyens (1981) among others, proposes that schemas Serve several functions in memory:

1. Encoding: Schemas direct attention to schema-relevant information and influence how incoming information is interpreted.
2. Storage: Schema-consistent information is more likely to be stored than schema-inconsistent information.
3. Retrieval: Schemas provide a framework for reconstructing memories, filling in gaps with schema-consistent details.

Brewer and Treyens (1981)

Participants were taken individually into an office and asked to wait for 30 seconds. They were then Unexpectedly asked to recall everything in the office. Results showed that participants recalled Schema-consistent items (desk, chair, books) highly accurately, recalled some items that were not Present but were schema-consistent (books, which were in the office, but also items like a pad of Paper that were not), and often failed to recall schema-inconsistent items (a skull and a pair of Skis that had been placed in the office).

Evaluation of schema theory:

• Schema theory provides a powerful explanation for memory errors and distortions, particularly in eyewitness testimony.
• The theory is difficult to falsify because schemas are latent constructs that cannot be directly observed. It can explain any pattern of results by postulating different schemas.
• The theory does not specify the precise mechanisms by which schemas influence encoding, storage, and retrieval.

Eyewitness Testimony

Loftus and Palmer (1974)

Elizabeth Loftus’s research on eyewitness testimony has had a profound influence on both cognitive Psychology and the legal system. Her work demonstrates that memory is highly malleable and can be Distorted by the way questions are asked.

In their most famous experiment, Loftus and Palmer showed participants films of car accidents and Then asked them to estimate the speed of the cars. The critical manipulation was the verb used in The question: “About how fast were the cars going when they hit each other?” versus “smashed Each other” versus “collided each other” versus “bumped each other” versus “contacted Each other.”

Key findings:

• The verb used significantly affected speed estimates. Participants in the “smashed” condition estimated speeds of approximately 65 km/h, compared to approximately 50 km/h in the “contacted” condition.
• In a follow-up experiment, participants who had been asked the “smashed” question were more likely (one week later) to report having seen broken glass at the accident scene, even though no glass was present in the film.

Interpretation: The post-event information (the leading question) altered participants’ memory Of the event. Loftus proposed two explanations:

1. Response bias: The wording of the question influences the response without actually changing the memory. Participants adjust their answer to match the expectations conveyed by the question.
2. Memory impairment: The post-event information actually alters the stored memory representation, overwriting or merging with the original memory.

Subsequent research has supported the memory impairment explanation, particularly when the Misleading information is introduced after a delay (when the original memory has begun to fade).

Evaluation:

• Loftus’s research has been highly influential in demonstrating the fallibility of eyewitness memory and has led to changes in police interviewing procedures (e.g., the cognitive interview technique developed by Fisher and Geiselman, 1992).
• The ecological validity of laboratory studies has been questioned: watching a film of a car accident is qualitatively different from witnessing a real accident, which is more emotionally arousing and has more personal consequences.
• Yuille and Cutshall (1986) found that witnesses to a real armed robbery had very accurate memories 4—5 months later and were resistant to misleading questions, suggesting that emotional arousal may enhance rather than impair memory accuracy .

For an overview of cognitive topics, see Cognitive Level of Analysis.

Common Pitfalls

1. Misunderstanding the difference between a stack (LIFO) and a queue (FIFO) in data structure applications.

2. Neglecting to normalise database designs, leading to data redundancy and update anomalies.

3. Writing pseudocode that is too language-specific rather than using standard algorithmic constructs.

4. Confusing an algorithm with a program — an algorithm is a step-by-step procedure, not its implementation in code.

Summary

The key principles covered in this topic are linked in the sub-pages above. Focus on understanding the definitions, applying the formulas or frameworks, and evaluating strengths and limitations of each approach.

Worked Examples

Worked examples demonstrating the application of key concepts are covered in the detailed sub-pages linked above.