Memory Systems: From Sensory Memory to Working Memory and the Baddeley–Hitch Model

Working Memory and Related Memory Models (Lecture Notes)

• Context and setup

  • The instructor shifts focus from attention to memory systems, specifically working memory (formerly called short-term memory).
  • A light anecdote: the lecture title idea “Fun and Nonsense” was inspired by a childhood book of nonsense poetry; used as a joking aside about naming and humor.
  • Mention of class data analysis over the weekend: a document will be posted on Canvas after all assignments are handed in; it will not be directly examinable, but may be used as a reference or example in answers where appropriate (e.g., drawing on lab concepts without relying on them as exam questions).
  • Emphasis: lab content may inform answers in a broad sense (e.g., queuing design), but you won’t be tested on lab material as a direct question.
  • The lab in focus examines a real experiment with a unique dataset; students are encouraged to use relevant lab knowledge where appropriate, not as a direct exam prompt.

• Core idea: memory is not a single system

  • Atkinson & Shiffrin proposed a three-system model: sensory memory → short-term memory (now called working memory) → long-term (reference) memory.
  • Short-term vs long-term naming reflects duration but also functional differences; memory systems are active, not passive stores.
  • The three-system model is a simplification; there can be multiple sensory memories (for vision, audition, touch, etc.).
  • Sensory memories are highly brief and modality-specific; working memory can translate across modalities and operate on information in several formats.

• Sensory memory (also called iconic in vision, echoic in audition)

  • Visual sensory memory (iconic): stores brief visual information after the stimulus disappears; duration is very short.
  • Auditory sensory memory (echoic): stores auditory information for a slightly longer period.
  • Typical durations:
  • Iconic memory: $ext{duration} \, \approx \;100\ \text{ms}$
  • Echoic memory: $\text{duration} \approx 2\ \text{s}$
  • Sensory buffers exist for multiple senses (visual, auditory, tactile, taste, smell).
  • Key concept: information fades from sensory memory unless attended to and transferred into a longer-lasting store.
  • Relation to attention: attention selects information from sensory memory to move into working memory; questions about early vs late selection relate to when semantics are accessed relative to attention.

• Attention and binding in perception

  • Feature Integration Theory (Treisman): attention binds features (e.g., color and shape) to form coherent percepts (e.g., a red square).
  • Binding likely occurs in working memory; instructions or targets (e.g., red squares) are held in working memory to guide behavior during a trial.
  • Endogenous vs exogenous attention: the debate centers on how attention interacts with semantics and memory; attention links sensory input to memory representations.
  • The lesson: perception and memory are integrated; attention affects what enters working memory and how features are bound together.

• Working memory: an active, multi-component system

  • Working memory is not a passive storage buffer; it actively processes information and supports manipulation, translation across modalities, and rehearsal.
  • Information can be translated between formats (e.g., sounds to images, images to sounds) within working memory.
  • Rehearsal is a key mechanism that moves information into longer-term storage (long-term memory) over time.
  • The three core questions guiding the model: how information is encoded, stored, and retrieved; how attention interacts with memory; how different modalities can be integrated and maintained.

• Immediate recall vs delayed recall (memory testing)

  • Immediate recall: recall right after presentation (e.g., list of letters shown briefly, then asked to write down as many as possible).
  • Delayed recall: recall after a delay with intervening tasks (e.g., arithmetic, counting) that prevents rehearsal.
  • The exercise described: participants saw a rapid list of letters; they were asked to recall in any order; results show classic serial position effects.
  • Serial position effects:
  • Primacy effect: improved recall for early items due to rehearsal and transfer to long-term memory.
  • Recency effect: improved recall for most recent items due to their presence in short-term/working memory or sensory buffers.
  • Manipulations to dissociate effects:
  • Delaying recall tends to wipe out the recency effect while preserving primacy (if primacy is due to rehearsal into long-term memory).
  • Adding a concurrent task (e.g., repeating a syllable) disrupts rehearsal and primarily reduces primacy, while recency remains if preserved in a sensory/short-term buffer.
  • Classic capacity estimate for working memory with such tasks: typically around $7 \pm 2$ items, depending on complexity and individual differences.
  • The takeaway: recency and primacy reflect different underlying memory processes; their dissociation supports multiple memory components contributing to short-term performance.

• Atkinson & Shiffrin model: three systems in the classic framework

  • Sensory memory (brief, modality-specific) → short-term/working memory (limited capacity, active processing) → long-term/reference memory (stable, retrievable knowledge).
  • Rationale for multiple memory systems: similar properties but distinct drivers and functions; if they were identical, a single system would suffice.
  • Terminology shift: short-term memory evolved into working memory to reflect its active processing role beyond passive storage.
  • The relationship between attention and memory: semantics can be accessed via attention or directly from sensory input, feeding into working memory depending on the theoretical stance (early vs late selection).

• Baddeley–Hitch working memory model (1980s; extended later)

  • Core components:
  • Central Executive: a control system that directs attention, coordinates subsystems, and manages tasks like updating and switching between displays and tasks.
  • Phonological Loop: processes verbal/auditory information; subcomponents include
    • Phonological Store (inner ear): temporary store for spoken words.
    • Articulatory Rehearsal Process (inner voice): subvocal repetition to refresh the store.
    • The loop can translate visually presented words into their phonological form for rehearsal.
  • Visuospatial Sketchpad: handles visual and spatial information; subcomponents include
    • Visual Cache: stores visual form and color information.
    • Inner Scribe: processes spatial relationships and sequential ordering (spatial layout and movement of items in space).
  • Episodic Buffer (added later): a temporary, multimodal storage that binds information from the phonological loop, visuospatial sketchpad, and long-term memory into coherent episodes; supports a unified representation that can be integrated with long-term memory.
  • Functional implications:
  • The phonological loop is sensitive to word length effects: shorter words are recalled better than longer words due to the time available for rehearsal within the loop.
  • Word length effect studies show that the loop can be taxed by longer phonological sequences; when visual words are presented and subjects must subvocally rehearse them, the loop translates visual input into a “sound” representation for rehearsal.
  • Visual words presented visually can interfere with the loop when concurrently engaged in articulatory tasks (e.g., repeating a syllable) if the task uses the same loop resources; auditory words access the loop directly and are less affected by articulatory suppression, implying a subcomponent structure within the phonological loop and a translational process (
    the articulatory control process) that converts visual input to sound for the loop.
  • Subsystems and evidence:
  • Phonological store vs. articulatory control system: visually presented items require conversion to phonological form for rehearsal; auditory items enter directly into the phonological store.
  • Visuospatial sketchpad subdivisions:
    • Inner Scribe (spatial processing): supports tasks requiring exploration of spatial layouts; interference occurs when a concurrent spatial task is performed (e.g., tapping in spatial locations) because both tasks contend for the same spatial resources.
    • Visual Cache (form and color): supports recognition of form and color; interfering tasks depend on whether the task is more form/color oriented or spatial.
  • Example demonstrations:
  • A Führer exercise: recall a drawn shape from a visualization task; translating language to vision demonstrates non-modality-specific storage—input modality is not the limiting factor; storage modality is.
  • A task pairing color/form discrimination with a spatial tapping task tends to disrupt the inner scribe more than color/form tasks; aligning with the idea of distinct subcomponents for different visual dimensions.
  • Central Executive: the proposed control center that orchestrates rehearsal and attention; involved in switching between tasks, maintaining task rules, and coordinating resources across loops; its precise nature remains a topic of ongoing research.
  • Random generation task as a probe of central executive capacity: higher randomness indicates better executive control; when the executive is taxed (e.g., by other demands), randomization declines and repetitive cycles emerge.
  • Dual-task interference logic (additive factors logic analogue): if two tasks rely on the same subsystem, performance suffers due to bottlenecks; if tasks recruit different subsystems, they can run in parallel with less interference.

• Episodic Buffer and its role

  • Episodic Buffer serves as a temporary, multimodal storage that integrates information from the phonological loop and visuospatial sketchpad with long-term memory to form coherent episodes.
  • It explains why people with amnesia can retain longer sentences for short periods, even though they cannot form long-term memories; meaningful information can be temporarily retained in a buffer that is larger than the phonological loop alone but not as large as long-term memory.
  • The episodic buffer is a bridge between immediate working memory and long-term memory; it supports composite representations that involve both visual and auditory information, enabling richer, semantic-rich memory traces.
  • The episodic buffer remains a topic of ongoing research and is not as universally emphasized in older textbook editions; its inclusion helps account for data not easily explained by the phonological loop and visuospatial sketchpad alone.

• Word length effect and related experiments (phonological loop focus)

  • Findings:
  • Short words are recalled better than long words, regardless of presentation modality (visual or auditory).
  • The effect suggests that the phonological loop has a temporal storage limit that is tied to how long the item can be subvocally rehearsed.
  • Dual-task manipulation: repeating syllables (e.g., "ba ba ba…") during encoding or rehearsal interferes with the recall of visually presented words more than auditory words, implying that:
  • Visual words require translation to phonological form for rehearsal (articulatory control), which consumes loop resources.
  • Auditory words enter the loop directly without translation.
  • The implications:
  • Supports subcomponents within the phonological loop (phonological store + articulatory control/translation) and cross-modal interactions.

• Visual-spatial sketchpad details

  • Visualization tasks demonstrate modality-transcending storage; people can translate language instructions into mental images and then respond with yes/no or pointing.
  • Visual-spatial tasks reveal subcomponents:
  • Inner Scribe: spatial processing and ordering; interferes with spatial responses (e.g., tapping in space) but not with color/form judgments.
  • Visual Cache: stores color and form information; depending on the task, interference patterns differ.
  • Experiments using block-letter figures and rule-based yes/no responses show that faster responses occur when a task relies on speaking than when it relies on pointing for some spatial tasks, indicating orientation of rehearsal modality affects interference.

• Central executive and randomization task (control processes)

  • The central executive is invoked when tasks require monitoring, updating, and coordinating resource use across subsystems.
  • Randomization tasks (e.g., generating random animal names) are used to probe executive functioning; performance improves when memory load is reduced, as the executive has fewer items to manage, allowing more random outputs.
  • A key question in the literature: what exactly is the central executive responsible for? It is often treated as a catch-all for unexplained variance; researchers aim to define its properties more precisely beyond the broad idea of executive control.

• Amnesia, complexity, and episodic buffer rationale

  • Prolonged sentence recall in the presence of amnesia suggests the episodic buffer can hold meaningful, integrated material even when long-term memory formation is impaired.
  • People with very small working memory spans still benefit from meaningful structure, indicating a buffer that can operate with semantically coherent content beyond the phonological loop alone.
  • The episodic buffer is posited to integrate across modalities and to support short-term representations that are meaningful, facilitating encoding into long-term memory when possible.

• Practical takeaways for exams and study practice

  • Working memory is an active, capacity-limited system, with a typical capacity of $7 \pm 2$ items for many tasks.
  • Iconic memory lasts about $\approx 100\ \text{ms}$; echoic memory lasts about $\approx 2\ \text{s}$; both are brief and highly sensory.
  • The primacy effect is largely attributed to rehearsal transferring items into long-term memory; the recency effect is tied to the presence of items in short-term/working memory or sensory buffers.
  • The Baddeley–Hitch model provides a structured framework for understanding how verbal and visual information are processed and maintained, with clear subcomponents and testable predictions via dual-task interference and modality-specific tasks.
  • The episodic buffer explains how meaningful, integrated representations can be maintained briefly even when long-term memory is impaired, bridging working memory with long-term knowledge.
  • Rehearsal (both phonological and visuospatial) is central to maintaining information in working memory and promoting transfer to long-term memory; the central executive oversees and coordinates this process.

• Connections to broader topics and upcoming content

  • The material sets the stage for long-term memory discussions (to be covered next Monday).
  • Attention, memory, and perception form an integrated system where binding, encoding, and retrieval depend on both automatic and controlled processes.
  • The lab data and design discussed will be revisited conceptually in terms of how memory systems interact during tasks, and how to interpret dual-task interference findings in practice.

• Quick glossary of terms used in lecture

  • Iconic memory: visual sensory memory with very brief duration.
  • Echoic memory: auditory sensory memory with brief duration.
  • Working memory: active, short-term memory system that manipulates and maintains information for ongoing tasks.
  • Long-term memory (reference memory): stable store of knowledge and experiences.
  • Primacy effect: enhanced recall for early items due to rehearsal and long-term storage.
  • Recency effect: enhanced recall for later items due to short-term/active memory buffers.
  • Phonological loop: auditory/verbal component of working memory (phonological store + articulatory rehearsal).
  • Visuospatial sketchpad: visual/spatial component of working memory (visual cache + inner scribe).
  • Central executive: control system that coordinates subsystems and memory processes.
  • Episodic buffer: multimodal temporary storage that binds information into coherent episodes.
  • Endogenous attention: voluntary, goal-directed attention.
  • Exogenous attention: automatic, stimulus-driven attention.
  • Whole report vs partial report: methods for testing sensory memory reporting capabilities.

• Final note from instructor

  • Long-term memory and broader cognitive integration will be explored in depth in the next session.
  • Students are encouraged to review these models (Atkinson & Shiffrin; Baddeley–Hitch) and relate them to explicit experimental findings (e.g., primacy/recency, word length effect, dual-task interference) for exam readiness.