Working Memory II

Foundations of Memory Systems & Sensory Memory

• Historical Evolution of Primary vs. Secondary Memory

  • William James (1842–1910) established the distinction between primary memory ("just-past" conscious awareness) and secondary memory ("memory proper" or long-term retention).

  • The rise of Behaviorism led to a period where primary memory was largely ignored, as subjective consciousness was deemed unscientific.

  • The Cognitive Revolution prompted cognitive scientists to revisit primary memory mechanisms.

  • George Miller (1956) demonstrated that humans can discriminate approximately 7 categories ($2.5\text{ bits}$) of objective sensory information (such as tones or tastes), but can hold $7 \text{ chunks}$ of subjectively meaningful information in immediate memory. Transforming raw data into meaningful units expands conscious holding capacity.

• Multi-Modal Memory Architecture (Atkinson & Shiffrin)

• Sensory Input flows directly into Sensory Memory.

• Sensory Memory: Brief hold of sensory data. Unattended information is rapidly lost.

• Attention: Acts as the filter directing information from Sensory Memory into Short-Term Memory.

• Short-Term Memory: Active temporary holding state. Unrehearsed information is lost.

• Maintenance Rehearsal: Loop process within Short-Term Memory that keeps information active.

• Encoding: Mechanism that transfers information from Short-Term Memory into Long-Term Memory.

• Retrieval: Mechanism that moves stored information from Long-Term Memory back into active Short-Term Memory.

• Long-Term Memory: Permanent or semi-permanent store. Some information may be lost over time.

  • Sensory Memory and Sperling's Partial Report Paradigm

• Sensory memory is the initial store holding raw visual (iconic memory) and auditory (echoic memory) information for a fraction of a second.

• Demonstration via Sperling’s Partial Report Method:

• Stimulus Presentation: A grid of 12 letters is flashed on a screen for 12 seconds:     $\begin{matrix} A & E & B & F \\ W & T & H & M \\ X & G & S & L \end{matrix}$

• Whole Report Condition: When asked to recall as many letters as possible from the entire display, participants averaged $42\%$ accuracy due to rapid trace decay.

• Partial Report Condition: When cued immediately after presentation to report only a single specific row, participants achieved $75\%$ accuracy, demonstrating that virtually the entire visual display is initially captured in iconic memory before fading.

Working Memory Duration and Loss Mechanisms

• Definition and Measurement of Working Memory

  • Definition: A limited-capacity memory system that keeps information consciously accessible for short periods to perform complex tasks.

  • Capacity: Holds approximately $7\text{ chunks}$ of information at a time (Miller, 1956).

  • Standard Assessment Tasks: Digit span task (repeating digit strings) and Corsi block tapping task (repeating spatial sequences).

  • Capacity Modifiers: Working memory capacity varies based on individual differences and item pronounceability.

• Duration of Working Memory (Brown-Peterson Task)

  • First systematically investigated by John Brown (1958) in the United Kingdom and Lloyd and Margaret Peterson (1959) in the United States.

  • Methodology:

  • Participants are presented with target stimuli (e.g., letter trigrams or words).

  • Rehearsal Prevention Task: Immediately upon stimulus offset, participants are given a 3-digit number (e.g., 472) and instructed to count backwards by 3s to the rhythm of a metronome.

  • Recall: After a variable retention interval of a few seconds, participants attempt to retrieve the target stimuli.

• Retention Curve Findings:

  • At $0\,\text{seconds}$ delay: Recall accuracy is approximately $80\%$.

  • At $3\,\text{seconds}$ delay: Accuracy drops to approximately $52\%$.

  • At $6\,\text{seconds}$ delay: Accuracy drops to approximately $43\%$.

  • At $9\,\text{seconds}$ delay: Accuracy drops to approximately $27\%$.

  • At $12\,\text{seconds}$ delay: Accuracy drops to approximately $15\%$.

  • At $15\,\text{seconds}$ delay: Accuracy flattens to approximately $13\%$.

• Theoretical Implications: Information leaves working memory after approximately $15\text{--}18\,\text{seconds}$ in the absence of rehearsal. The baseline residual trace (~$10\%$ accuracy at 15+ seconds) reflects information that successfully transferred into long-term memory.

  • Mechanisms of Forgetting: Decay vs. Interference

• Decay Theory: Information traces naturally fade automatically over time.

• Interference Theory: New incoming information actively displaces existing information.

• Experimental Teasing (Waugh & Norman, 1965):

  • Logic: Keep the total elapsed time constant while varying the number of intervening items to isolate interference from decay.

  • Procedure: Presented lists of 16 digits at either a slow rate ($1\,\text{item/second}$) or a fast rate ($4\,\text{items/second}$).

    • Example List ($1\,\text{item/sec}$): 7, 0, 8, 5, 2, 6, 8, 1, 5, 2, 3, 9, 7, 2, 4, 3

    • Example List ($4\,\text{items/sec}$): 8, 2, 6, 1, 5, 9, 7, 4, 3, 8, 2, 0, 7, 5, 9, 6

  • Test Probe: Participants were presented with a probe digit from the list and instructed to recall the digit immediately following it (e.g., if probe is 2, recall 9).

  • Results: Probe recall was markedly worse when more items intervened between the target and probe, even when the absolute time duration ($3\,\text{seconds}$) was identical across conditions.

  • Conclusion: Interference from competing items, rather than passive decay over time, is the primary mechanism pushing information out of working memory.

Baddeley's Multi-Component Model of Working Memory

• Conceptual Shift

  • Working memory is an active multi-component system rather than a passive, unitary short-term store.

• Empirical Milestones in Model Development

  • Step 1: Structural Dissociation of Visual and Auditory Stores (Baddeley & Hitch, 1974)

  • Paradigm: Concurrent dual-task paradigm combining a Digit Span task (holding string of numbers) with a visual/verbal Letter Judgment task (verifying spatial/grammatical letter statements).

  • Findings: Participants performed both tasks simultaneously with negligible performance decrement in either task.

  • Implication: Auditory and visual working memory rely on distinct, non-interfering processing stores.

  • Step 2: Implementation of the Central Executive (Baddeley & Hitch, 1974)

  • Function: Attentional control system regulating what content enters working memory subsystems, managing phonological rehearsal, and directing spatial fixation.

  • Attentional Distraction Effect: Introducing competing attentional demands severely impairs overall working memory processing capacity.

  • Step 3: Addition of the Episodic Buffer (Baddeley, 2000)

  • Theoretical Need: Added to account for cross-modal binding that traditional single-modality stores could not explain.

  • Function: Binds multimodal visual and auditory information into coherent temporal/episodic structures (e.g., linking a visual face to a spoken name).

• Subsystem Architecture and Inter-relationships

  • Visuospatial Sketchpad: Holds visual and spatial visual imagery.

  • Phonological Loop: Holds acoustic and speech-based audio signals.

  • Episodic Buffer: Integrates codes from visual, auditory, and long-term sources into unified representation units.

  • Central Executive: Allocates attentional focus across the three slave systems.

  • Long-Term Memory Interface: All three subsystems (Visuospatial Sketchpad, Episodic Buffer, Phonological Loop) maintain bidirectional pathways with Long-Term Memory (spanning Episodic Memories, Visual Semantics, and Language).

Neurobiology of Working Memory

• Central Executive Localization

  • Anatomical Region: Prefrontal Cortex.

  • Empirical Evidence: Functional neuroimaging (fMRI and PET) reveals prefrontal cortical activation during tasks requiring executive control, such as detecting "oddballs" (unexpected stimuli that violate an established sequential pattern).

• Phonological Loop Localization

  • Anatomical Regions: Left Frontal lobe (specifically Broca’s area) and Left Parietal lobe.

• Empirical Evidence: PET imaging demonstrates elevated blood flow across Broca's area and left parietal structures during silent verbal rehearsal of digits and words.

  • Visuospatial Sketchpad Localization

• Anatomical Regions: Right Occipital lobe, Right Parietal lobe, and Right Frontal lobe.

• Empirical Evidence: PET scans confirm right-hemisphere dominant network engagement while subjects hold visual object representations or spatial locations over discrete delays (e.g., tracking sequential green/pink dot presentations across $1000\,\text{ms}$ intervals).

Serial Position Effects and Memory Architecture

• Phenomenology of Serial Position

  • In immediate free recall of an ordered word list, retrieval probability follows a U-shaped curve where items at the beginning and end of the list are recalled significantly better than middle items.

• Empirical Accuracy Breakdown (15-Item List):

  • Item 1: $100\%$

  • Item 2: $93\%$

  • Item 3: $90\%$

  • Item 4: $75\%$

  • Item 5: $65\%$

  • Items 6–7: $50\%$

  • Item 8: $40\%$

  • Items 9–11: $25\%$

  • Item 12: $50\%$

  • Item 13: $60\%$

  • Item 14: $75\%$

  • Item 15: $90\%$

  • Sub-components: Primacy and Recency Effects

• Primacy Effect: Superior recall for items at the start of the list.

  • Underlying Mechanism: Early items receive undivided cognitive rehearsal, allowing successful encoding and consolidation into Long-Term Memory.

• Recency Effect: Superior recall for items at the end of the list.

  • Underlying Mechanism: Terminal items remain actively present in Working Memory at the moment testing begins.

  • Empirical Dissociation of Primacy and Recency

• Experimental Comparison: Immediate Free Recall (IFR) versus Final Free Recall (FFR).

• Immediate Free Recall (IFR): Produces a classic serial position curve featuring both robust Primacy and sharp Recency effects.

• Delayed / Final Free Recall (FFR): Introducing a filled delay or an interference task prior to recall selectively abolishes the Recency effect, causing memory for end-of-list items to drop precipitously.

• Conclusion: The selective elimination of the recency effect proves that recency is governed by active maintenance in Working Memory, whereas the persistent primacy effect is grounded in Long-Term Memory storage.