Short-Term and Working Memory

Introduction to Short-Term and Working Memory

  • Definition of Memory: Memory is defined as the processes involved in retaining, retrieving, and using information about stimuli, images, events, ideas, and skills after the original information is no longer present.

  • Functions and Importance of Memory:

    • Acts as a mental time travel mechanism to re-experience past events from fractions of a second ago to decades prior.

    • Enables current, moment-to-moment actions including maintaining conversations, solving active problems, making decisions, and performing learned motor or procedural skills.

    • Essential for underlying background functions taken for granted: labeling familiar objects (identifying a book), knowing daily sequences (ordering and paying at a restaurant), and navigating physical spatial layouts (finding a classroom).

  • Case Study: Clive Wearing:

    • Background: A highly respected English musician and choral director who, in his 40s40\text{s}, contracted viral encephalitis.

    • Brain Damage: The viral infection destroyed key structures in his temporal lobe that are critical for forming new long-term memories.

    • Symptomatology: Wearing lives entirely within the most recent 1 to 2 minutes1\text{ to }2\,\text{minutes} of his life. He retains what just occurred but forgets everything else immediately. If a visitor leaves the room and returns after 3 minutes3\,\text{minutes}, Wearing reacts as though he has never met them before. He constantly feels as though he has just gained consciousness for the very first time.

    • Diary Records: His personal diary contains hundreds of repeated entries such as "1:10 I WAKE up for the FIRST time", "2:13 I wake up", "4:32 I am alive", and "5:43 Just woke up". He routinely crosses out prior entries because he has no recollection of writing them, even though he recognizes the handwriting as his own. He describes his state of existence as being "like death".

Clive Wearing's diary entries
  • Role of Expertise in Memory:

    • Experienced chess players store vast networks of board configurations and unfolding game strategies in memory.

    • Research demonstrates that experienced players remember actual, meaningful arrangements of chess pieces on a board significantly better than novice or inexperienced players.

Experienced chess player evaluating a board setup

The Modal Model of Memory

  • Proposed by Richard Atkinson and Richard Shiffrin (19681968).

  • Termed the "modal model" because it synthesized the common structural features of major memory models proposed throughout the 1960s1960\text{s}.

  • Structural Features:

    1. Sensory Memory: Initial stage that holds all incoming sensory stimulation for fractions of a second or a few seconds.

    2. Short-Term Memory (STM): System that stores 5 to 75\text{ to }7 items for approximately 15 to 30 seconds15\text{ to }30\,\text{seconds}.

    3. Long-Term Memory (LTM): High-capacity system that holds vast amounts of information for extended periods, ranging from years to decades.

  • Control Processes:

    • Active processes that can be consciously controlled by an individual and vary according to the specific task demands.

    • Rehearsal: Mechanically repeating a stimulus over and over (e.g., repeating a telephone number) to keep it active in STM.

    • Encoding Strategies: Methods used to make information more memorable, such as associating a phone number with a historic date.

    • Attentional Strategies: Focus mechanisms used to selectively attend to specific high-priority inputs.

  • Dynamic Interactions (Example: Rachel Ordering Pizza):

    • Sensory Memory: All visual text displayed on a computer screen enters sensory memory simultaneously.

    • Selection to STM: Selective attention focuses on the phone number (555-5100555\text{-}5100), transferring it into STM.

    • Rehearsal in STM: Repeating the phone number retains it within STM while placing the call.

    • Encoding to LTM: Active memorization strategies transfer the phone number into LTM storage.

    • Retrieval from LTM: Days later, wanting pizza triggers retrieval of the stored phone number from LTM back into active STM consciousness.

Rachel using memory processes to look up, rehearse, store, and retrieve a phone number

Sensory Memory

  • Definition: Retention, for brief intervals of time, of the physical effects of sensory stimulation.

  • Persistence of Vision:

    • Sparkler Trail Effect: Sweeping a lighted sparkler rapidly through the dark creates an apparent visual trail of light. No physical light exists along the trajectory; the light trail is a creation of the mind retaining the visual perception for a fraction of a second.

Sparkler trail generated by persistence of vision
  • Cinematic Film Projection: Movies project 2424 still frames per second. Each frame is flashed on screen when the shutter opens; the shutter then closes while the film moves to the next frame, leaving the screen completely dark. Persistence of vision retains the visual image of the preceding frame during dark shutter intervals, filling in the dark spaces to create smooth, continuous perception (apparent movement). Early movie projectors flashed images too slowly, causing perceived flickering ("flickers").

    • Sperling's Experiment (19601960): Capacity and Duration of Visual Sensory Store:

  • Stimulus: Flashed a 3×43 \times 4 array of 1212 letters on a screen for 50 ms50\,\text{ms} (0.050 seconds0.050\,\text{seconds}).

  • Whole Report Method: Participants were asked to report as many letters as possible from the full matrix.

    • Result: Participants reported an average of 4.54.5 out of 1212 letters (37.5%37.5\%).

  • Partial Report Method: Designed to test if visual perception fades during the act of reporting. Immediately after the 12-letter12\text{-letter} array was extinguished, a tone sounded directing attention to a specific row:

    • High-pitched tone: Top row

    • Medium-pitched tone: Middle row

    • Low-pitched tone: Bottom row

    • Result: Participants correctly reported an average of 3.33.3 out of 44 letters (82%82\%) for whichever row was cued. Because any row could be cued, Sperling concluded participants had registered roughly 82%82\% of the entire array (≈10\approx 10 letters) in visual sensory memory immediately after presentation.

  • Delayed Partial Report Method: Evaluated the fading rate of the visual trace by delaying tone presentation by fractions of a second.

    • Result: When the cue tone was delayed by 1.0 second1.0\,\text{second}, performance dropped to slightly over 11 letter per row (≈4\approx 4 letters total across all three rows), matching the performance of the whole report method.

Results graph of Sperling's partial report experiments showing decay curve over 1 second
  • Subtypes and Functions of Sensory Memory:

    • Iconic Memory (Visual Icon): Brief sensory store for visual stimuli, lasting less than 1.0 second1.0\,\text{second}.

    • Echoic Memory: Brief sensory store for auditory stimuli, lasting for a few seconds (2 to 4 seconds2\text{ to }4\,\text{seconds}; Darwin et al., 19721972).

    • Primary Functions: (1) Collects incoming environmental information to be processed; (2) holds information briefly while initial feature extraction occurs; (3) fills in perceptual gaps during intermittent stimulation.

Short-Term Memory (STM): Duration, Capacity, and Coding

  • Definition: The storage system involved in retaining small amounts of information for brief durations (15 to 30 seconds15\text{ to }30\,\text{seconds} without active rehearsal).

  • Methodological Testing Methods:

    • Recall Test: Participants are presented with stimuli and, following a delay, are asked to reproduce as many stimuli as possible. Evaluated by percentage correct and recall sequence patterns.

    • Recognition Test: Participants identify previously encountered stimuli from among novel distractor items.

  • Duration of Short-Term Memory:

    • Investigated by John Brown (19581958) and Lloyd Peterson & Margaret Peterson (19591959).

    • Procedure: Read three letters (e.g., FZLFZL) followed by a 3-digit number (e.g., 4545). Participant repeats the number and counts backward out loud by 3s3\text{s} (e.g., 45,42,39...45, 42, 39...) to prevent vocal rehearsal until instructed to recall the letters after delays ranging from 3 to 18 seconds3\text{ to }18\,\text{seconds}.

    • Peterson & Peterson Findings: Participants recalled 80%80\% of letter groups after counting for 3 seconds3\,\text{seconds}, but only 12%12\% after counting for 18 seconds18\,\text{seconds}. Attributed loss to memory trace decay over time.

    • Keppel & Underwood (19621962) Reanalysis: Examined performance on Trial 1 only and found almost no performance decline between 3 seconds3\,\text{seconds} and 18 seconds18\,\text{seconds}. Memory loss occurred systematically across later trials.

    • Proactive Interference (PI): Interference occurring when information learned previously interferes with learning or retrieving new information (e.g., memory of an old phone number 521-5100521\text{-}5100 interfering with recalling a new number 522-4100522\text{-}4100).

    • Conclusion: Drop-off in STM duration over multiple trials is caused by proactive interference from earlier letter lists, not passive decay. When rehearsal is completely prevented, the effective duration of STM is 15 to 20 seconds15\text{ to }20\,\text{seconds}.

Graph comparing Peterson & Peterson's average performance over many trials versus Keppel & Underwood's first trial analysis
  • Capacity of Short-Term Memory:

    • Digit Span: The number of digits a person can reliably recall in correct serial order. Typical digit span is between 5 and 85\text{ and }8 or 5 and 95\text{ and }9 digits.

    • George Miller (19561956): Published "The Magical Number Seven, Plus or Minus Two", proposing STM holds 7±27 \pm 2 items.

    • Modern Estimates: Nelson Cowan (20012001) and Steven Luck & Edward Vogel (19971997) established STM capacity at approximately 44 items.

    • Luck & Vogel Task: Flashed two array displays of colored squares separated by a 900 ms900\,\text{ms} delay. Performance was near 100%100\% accuracy for arrays of 1 to 31\text{ to }3 squares, but dropped rapidly when displays contained 44 or more squares.

    • Chunking:

    • Introduced by George Miller (19561956); process of combining smaller units into larger, meaningful units.

    • Chunk Definition: A collection of elements strongly associated with one another but weakly associated with elements in other chunks (Cowan, 20012001; Gobet et al., 20012001).

    • Sentence structure expands word span from 5–85\text{--}8 unrelated words to 2020 or more words.

    • Unchunked letters (BCIFNCCASICBBCIFNCCASICB) are difficult to recall; chunked acronyms (CIAFBINBCCBSCIAFBINBCCBS) group 1212 letters into 44 meaningful units (CIA,FBI,NBC,CBSCIA, FBI, NBC, CBS).

    • S.F. Case Study (Ericsson et al., 19801980): A college student with an average digit span (77 digits) underwent 230230 one-hour training sessions. Expanded his digit span to 7979 digits by chunking random sequences into running times (3492→3 minutes 49.2 seconds3492 \rightarrow 3\,\text{minutes } 49.2\,\text{seconds}) and historical dates stored in LTM.

    • Chess Master Study (Chase & Simon, 1973a,1973b1973a, 1973b): Chess master (>10,000 hours> 10{,}000\,\text{hours} experience) vs. beginner (<100 hours< 100\,\text{hours} experience) viewed board configurations for 5 seconds5\,\text{seconds}.

      • Actual Game Positions: Master correctly placed 1616 out of 2424 pieces on Trial 1 (reproduced perfectly in 44 trials); beginner placed 44 out of 2424 pieces.

      • Random Positions: Master advantage vanished completely; performed identically to beginners (≈2–4\approx 2\text{--}4 pieces).

      • Conclusion: Master advantage stems from storing 4 to 64\text{ to }6 chunked tactical patterns in LTM, not superior structural STM capacity.

  • Coding in Short-Term Memory:

    • Refers to how information is mentally represented.

    • Auditory Coding:

    • Representing items based on sound.

    • R. Conrad (19641964): Flashed target letters visually on a screen. When participants made recall errors, they misidentified target letters as letters that sounded similar (e.g., target "F" misidentified as acoustic confusions "S" or "X", rather than visual confusions like "E"). Proved STM representations are predominantly auditory.

    • Visual Coding:

    • Representing items as visual imagery.

    • Sergio Della Sala et al. (19991999): Presented visual matrices ranging from 2×22 \times 2 grids up to 5×65 \times 6 grids with 50%50\% of squares shaded black. Participants accurately recalled visual patterns averaging up to 99 shaded squares before making errors, relying on visual sub-pattern chunking.

    • Semantic Coding:

    • Representing items based on meaning.

    • Delos Wickens et al. (19761976): Tested Fruit group vs. Professions group across 4 trials (3 words per trial, 15 seconds15\,\text{seconds} counting backward, then recall).

    • Trials 1–3 built proactive interference due to semantic category repetition, causing recall performance to decline.

    • On Trial 4, switching the Professions group stimulus category to fruits produced release from proactive interference, causing performance to recover significantly. Proved information in STM is coded semantically.

Working Memory

  • Limitations of the Modal Model:

    • Described STM exclusively as a single, passive short-term storage mechanism.

    • Failed to account for simultaneous processing and dynamic storage during complex tasks (e.g., mental arithmetic: calculating 43×6=25843 \times 6 = 258 requires storing digits while carrying numbers and multiplying).

    • Failed to explain how individuals perform dual-tasks simultaneously (e.g., holding a 4-digit span in mind while reading and comprehending a paragraph).

  • Definition of Working Memory:

    • Proposed by Alan Baddeley and Graham Hitch (19741974).

    • Defined as a limited-capacity system for temporary storage and manipulation of information for complex tasks such as comprehension, learning, and reasoning.

  • Distinctions Between STM and Working Memory:

    1. STM is concerned primarily with storing information for a brief duration; Working Memory is concerned with manipulating information during complex cognition.

    2. STM consists of a single storage component; Working Memory consists of multiple interacting components.

  • Baddeley's Three-Component Model:

    1. Phonological Loop: Handles verbal and auditory information.

    • Phonological Store: Limited capacity component holding speech-based information for a few seconds.

    • Articulatory Rehearsal Process: Active verbal rehearsal mechanism that refreshes items in the phonological store to prevent decay.

    1. Visuospatial Sketch Pad: Holds visual and spatial information (e.g., visualizing landmarks, solving spatial puzzles, mental navigation).

    2. Central Executive: The control center / "traffic cop" of working memory. Stores no information itself; coordinates how information is used by the phonological loop and visuospatial sketch pad, focuses attention, divides attention between tasks, and switches attention.

  • Phenomena of the Phonological Loop:

    • Phonological Similarity Effect: Confusion of letters or words that sound similar (Conrad, 19641964).

    • Word Length Effect: Memory for lists of short words is superior to memory for long words (Baddeley et al., 19841984). Occurs because long words take longer to rehearse and produce during recall. Memory span corresponds to the number of items a person can pronounce in roughly 1.5 to 2.0 seconds1.5\text{ to }2.0\,\text{seconds} (Baddeley et al., 19751975).

    • Articulatory Suppression: Disruption of the phonological loop caused by repeating an irrelevant sound out loud (e.g., "the, the, the…") while trying to memorize a list. Repeating the sound occupies the articulatory rehearsal process, reducing overall recall capacity and eliminating the word length effect (Baddeley et al., 19841984).

  • Phenomena of the Visuospatial Sketch Pad:

    • Visual Imagery: Creating visual images in the mind in the absence of a physical stimulus.

    • Mental Rotation Task (Roger Shepard & J. Metzler, 19711971): Measured reaction times to judge whether 3D line drawings represented identical objects. Decision time scaled linearly with orientation angle difference (40o40^\text{o} difference took 2 seconds2\,\text{seconds}; 140o140^\text{o} difference took 4 seconds4\,\text{seconds}), demonstrating dynamic mental rotation in spatial memory.

    • Spatial Interference Task (Lee Brooks, 19681968): Participants visualized a block letter "F" and navigated its perimeter clockwise in mind, categorizing corners as "Out" (outside corner) or "In" (inside corner).

    • Pointing Response: Pointing to "Out"/"In" on a spatial column matrix was extremely difficult because visualization and pointing both required the visuospatial sketch pad, overloading it.

    • Vocal Response: Saying "Out" or "In" out loud was much easier because vocalizing utilized the phonological loop, avoiding structural resource competition.

  • Central Executive and Frontal Lobe Function:

    • Manages attentional control during complex dual-tasks (e.g., driving while navigating via visual sketch pad and conversing via phonological loop).

    • Frontal Lobe Damage: Destroys executive attentional control, producing perseveration — repeatedly performing the same behavior even when rule changes indicate it is incorrect.

  • The Episodic Buffer:

    • Added to Baddeley's revised model (Baddeley, 2000a,2000b2000a, 2000b) to account for abilities unexplained by the original 3-component model (e.g., retaining long 15 to 20 word15\text{ to }20\,\text{word} cohesive sentences via chunking and grammar rules).

    • Function: Stores integrated information (providing auxiliary capacity) and serves as a direct interface connecting working memory components with Long-Term Memory.

    • Alternative Model (Nelson Cowan, 1988,1999,20051988, 1999, 2005): Views working memory as the temporarily activated portion of long-term memory controlled directly by attentional focus.

Working Memory and the Brain

  • Key anatomical memory structures include the Prefrontal Cortex (PFC), Frontal Lobe, Hippocampus, and Amygdala.

  • Effect of Prefrontal Cortex Damage:

    • Delayed-Response Task in Monkeys (Goldman-Rakic, 19921992): A monkey observes a food reward placed in one of two food wells. Both wells are covered, a screen is lowered for a delay period, and then raised. The monkey must uncover the correct well to receive the reward.

    • Normal, intact monkeys perform accurately after training.

    • Monkeys with bilateral prefrontal cortex removal drop to chance performance (50%50\% correct).

    • Infant Development Connection: Human infants under 8 months8\,\text{months} old exhibit "out of sight, out of mind" behavior because prefrontal cortex circuits are not fully developed.

Delayed-response task setup administered to a monkey
  • Single-Cell PFC Recording Studies:

    • Shintaro Funahashi et al. (19891989): Recorded single PFC neurons in monkeys during a spatial delayed-response task. A visual cue square flashed at a specific screen position while the monkey maintained gaze on a central fixation marker.

    • Neural Delay Response: Specific PFC neurons fired briefly during cue exposure, maintained high firing rates throughout the multi-second delay when the visual cue was absent, and stopped firing once the monkey executed a eye movement (saccade) to the target location.

    • Conclusion: Sustained neural firing during delay periods represents the physiological memory trace for spatial locations.

  • Distributed Brain Activation in Working Memory:

    • Brain imaging (PET and fMRI) demonstrates that working memory is distributed across extensive cortical and subcortical networks.

    • Key regions include the Prefrontal Cortex, visual cortex (Super et al., 20012001), parietal cortex, temporal cortex (Jonides et al., 20052005), and cerebellum (Cabeza & Nyberg, 20002000).

  • Individual Differences in Executive Control Efficiency:

    • Edward Vogel et al. (20052005): Measured event-related potentials (ERP) in humans during a visual working memory task.

    • Design: High working memory capacity vs. low working memory capacity participants viewed target red rectangles with or without blue distractor rectangles (which participants were instructed to ignore).

    • Results:

    • When presented with 22 red target rectangles, both high- and low-capacity groups displayed identical baseline ERP responses.

    • When 22 blue distractor rectangles were added, high-capacity participants showed almost no increase in ERP response (effectively filtering out distractors).

    • Low-capacity participants showed a sharp increase in ERP response, demonstrating that irrelevant blue distractors entered their working memory.

    • Conclusion: High-capacity individuals possess more efficient central executive filtering mechanisms for suppressing irrelevant environmental noise.

Individual Differences in Working Memory Capacity

  • Research demonstrates that individuals with higher working memory capacities achieve superior performance on reading comprehension tests, complex reasoning tasks, and standardized intelligence tests (Conway et al., 2003,20072003, 2007).

  • Reading Span Task (Meredyth Daneman & Patricia Carpenter, 19801980):

    • Developed to assess both dynamic storage and dynamic processing functions of working memory simultaneously.

    • Procedure: Participants read a series of sentences out loud (e.g., 2 to 5 sentences), cover each sentence after reading, and then recall the final word of each sentence in correct serial order.

    • Results:

    • Participants with a reading span of 22 correctly answered an average of 13.613.6 out of 2424 comprehension questions.

    • Participants with a reading span of 4 to 54\text{ to }5 correctly answered an average of 19.719.7 out of 2424 comprehension questions.

    • Reading span scores correlated directly with verbal SAT scores.

  • Core Mechanism: Superior cognitive performance is driven not merely by physical storage size, but by the executive control efficiency of allocating attention to goal-relevant items while filtering out irrelevant distraction (Awh & Vogel, 20082008; McNab & Klingberg, 20082008).