Sensory, Short-Term, and Working Memory: Models and Mechanisms

Early Research into Memory: The Span of Apprehension

Historical Context: Introspectionism * Early psychologists (introspectionists) were fundamentally interested in the limits of human consciousness, specifically questioning how much information could be held in consciousness at a single moment. * Jevons (1871) Bean Study: * This study provided an early estimate of what is known as the "span of apprehension." * Findings: * Subjects achieved $100\%$ accuracy when presented with $4$ beans or fewer. * Subjects achieved approximately $50\%$ accuracy when presented with $8$ beans.
Evolution of Research: * Work on the span of apprehension continued into the early part of the 20th century. * Despite consistent span estimates, researchers felt a "nagging feeling" that the experiments were failing to capture the full picture of sensory processing; it seemed as though subjects saw more than they could report.

Sperling’s Experimental Paradigm (1960)

Whole Report Method: * Participants were shown an array of stimuli (e.g., a $3 \times 4$ grid of letters and numbers) for a very brief duration ( $50\,ms$ ). * Instruction: Report as many stimuli from the entire array as possible. * Results: Subjects typically reported only $4$ items correctly, representing roughly $33\%$ of the array.
Partial Report Method: * Participants were shown the same array for $50\,ms$ . * Instruction: An immediate tone cue was played to indicate which specific row (top, middle, or bottom) the subject should report. * Results: Subjects typically got $3$ items correct from the cued row, representing $75\%$ of that row. This implies that for a brief moment, $75\%$ of the entire array was available in memory.
Delayed Partial Report: * Participants were shown the array for $50\,ms$ , but the tone cue was delayed by $1\,second$ . * Results: Performance dropped significantly; subjects reported only $1$ item correct ( $25\%$ of the row).
Forgetting via Masking: * Sensory memory is fragile and can be disrupted. Masking involves presenting a second visual stimulus (like a grid of symbols: %% * & # @ % # @ & % $ *) immediately after the target stimuli to "overwrite" or interfere with the iconic memory trace.

The Distinction Between Short-Term and Long-Term Memory

The Case of Henry (H.M.) and Dr. Milner: * Henry suffered from severe memory deficits. A famous interaction with Dr. Brenda Milner illustrates the distinction between short-term storage and the inability to form new long-term memories. * The Rehearsal Anecdote: * Dr. Milner asked Henry to remember the numbers $5, 8, 4$ . * Twenty minutes later, Henry successfully recalled the numbers: "Five, eight, four." * When asked how he did it, Henry explained his complex rehearsal strategy: "Well, five, eight, and four add up to seventeen. Divide by two, you have nine and eight. Remember eight. Then five—you’re left with five and four—five, eight, four. It’s simple." * The Break in Continuity: Immediately after this, Henry could not remember Dr. Milner’s name. When she asked if he still remembered the number, he replied, "Number? Was there a number?"
Describing Henry’s Problem: * Henry’s short-term storage and retrieval mechanisms were intact (he could maintain information via active rehearsal). * The failure occurred in the transfer/storage into long-term memory. Once his attention was diverted (a "break" in rehearsal), the information was lost forever.

The Modal Model of Memory

Structure and Flow: 1. Environment: External stimuli provide the input. 2. Sensory Memory: Large capacity but extremely brief duration. Information is lost if not attended to. 3. Attention: Acts as a filter to move information into short-term storage. 4. Short-term Memory (Storage): Limited capacity storage where information is active. * Rehearsal: An active process to maintain information in STM. * Forgetting: Information is lost from STM if not rehearsed or transferred. 5. Long-term Memory: Large capacity, long-term storage of information. * Retrieval: The process of bringing information from LTM back into STM for use.

Working Memory: Model #2

Defining the Shift: * Short-term memory is primarily concerned with storage. * Working memory is concerned with storage plus thinking (processing).
Baddeley’s Components of Working Memory: * Phonological Loop: Deals with auditory and speech-based information. * Phonological Store: Holds approximately $2\,seconds$ worth of auditory information. Information can enter from the environment or the articulatory control process. * Articulatory Control Process: The process of "talking to yourself" (inner speech) to refresh information in the phonological store. * Visuo-spatial Sketchpad: Stores visual and spatial information (related to mental imagery). * Episodic Buffer: Acts as a "crosstalk" mechanism between the visuo-spatial sketchpad and the phonological loop. It handles integrated representations and allows for semantic coding. * Central Executive: The coordinator of cognitive processes. It helps maintain information in working memory stores and directs attention.

The Phonological Loop: Predictions and Coding

Key Predictions of the Model: * Capacity and Speed: Since the store lasts $2\,seconds$ , people who can speak faster (rehearse faster) generally have a larger working memory capacity. * Word Length Effect: Capacity is smaller for long words because they take longer to rehearse, filling the $2\,second$ store more quickly. * Acoustic Confusion: Because the store is auditory, people often confuse words that sound alike (e.g., "cap," "cat," "can") rather than words that look alike or have similar meanings. * Articulatory Suppression: If the articulators are kept busy (e.g., saying "blah, blah, blah" repeatedly), the articulatory control process cannot refresh the phonological store. This forces the brain to use alternative coding methods, causing the aforementioned effects (like acoustic confusion) to disappear.

Semantic Coding and the Episodic Buffer

Evidence for Semantic Code in WM: * Proactive Interference (PI): Occurs when previously learned information interferes with the ability to learn new information. * Release from Proactive Interference Study: * Participants were given trials of items from the same category (e.g., different types of meat). Performance typically drops over successive trials. * If the category is suddenly shifted (e.g., from Meats to Fruits), performance recovers significantly. * Conclusion: This recovery indicates that working memory also utilizes a semantic code (meaning), which is facilitated by the Episodic Buffer.
Summary of Coding in WM: There are three primary ways information is coded in working memory: 1. Acoustically (Sound) 2. Visually (Sight) 3. Semantically (Meaning)

Measuring and Correlating Working Memory

Measuring Capacity: * Reading Span: A task requiring participants to read sentences and remember the last word or a specific unrelated number while answering comprehension questions (e.g., "The bishop saw the bouncing red ball in the park. Seven. Saw what?"). This tests simultaneous storage and processing. * The N-Back Game: A task where subjects must identify if the current stimulus matches the one presented $n$ steps earlier ( $1$ -back, $2$ -back, etc.).
Cognitive Correlates: * Working memory capacity is highly correlated with Fluid Intelligence ( $G_f$ ). * Study: Chuderski & Neecka (2012): * This study looked at the Correlation Matrix for Hit/False Alarm Rates and Intelligence test scores (Raven’s Matrices and Analogies). * Raven's Matrices: A non-verbal intelligence test. * Analogies: A component of the Wechsler IQ test. * Results showed significant correlations (e.g., $r = .49$ between $k$ value and Raven's; $r = .40$ between $2$ -back hits and Raven's) between working memory performance and these intelligence measures.