Psychology of Memory: STM, Working Memory & Encoding

Memory (two definitions)

Process of retaining, retrieving and using information after the original is gone; also active any time a past experience affects how you think or behave now or later

Is memory a perfect recording?

No. Memory is a representation of experience that is reconstructive: we store fragments and rebuild them, so we can never be sure which details are accurate

Three stages of memory

Encoding (getting info in), storage (holding it over time; consolidation moves STM to LTM), retrieval (accessing it). Every stage is vulnerable to distortion or loss

DRM paradigm (Deese-Roediger-McDermott)

Study a list of related words (bed, dream, nap...) never containing the lure ("sleep"); people falsely recall/recognize the lure. Shows memory links related concepts automatically and false memories are easy to create

What does the DRM task demonstrate about STM?

Even with immediate recall nobody remembers perfectly: STM is brief and limited in capacity

Episodic memory

Memory for specific personal events/experiences (e.g., what you had for breakfast); involves mental time travel ("remembering")

Semantic memory

General facts and knowledge without the context of learning (e.g., capital of Canada is Ottawa); "knowing"

Procedural memory

Knowing how to do something (skills like riding a bike, typing, playing an instrument); automatic once learned; implicit

Explicit (declarative) memory

Intentionally encoded and deliberately retrieved with conscious awareness (episodic and semantic memory)

Implicit (non-declarative) memory

Learned incidentally and retrieved automatically without awareness; includes procedural memory, priming, classical conditioning

Mental time travel (Tulving)

Re-experiencing a past event; defining feature of episodic memory ("remember"), versus semantic "know" which has no time travel

Recall vs recognition

Recall = produce the items yourself (short answer, free/cued recall). Recognition = identify previously seen items among others (multiple choice); usually easier and relies on familiarity

Word-stem completion

Implicit memory test: fill in missing letters (e.g., CLO_ ) with the first word that comes to mind; tests priming without asking about memory

Lexical decision task

Decide whether a letter string is a real word; used with reaction time to measure priming/implicit memory

Signal detection theory (SDT)

Framework to evaluate accuracy and response bias in memory decisions (hits, false alarms)

Reaction time (RT) as a measure

Speed of retrieval or decision; faster responses indicate easier access (e.g., priming, mental rotation)

Modal model of memory (Atkinson & Shiffrin, 1968)

Flow model: sensory memory -> short-term memory -> long-term memory. Structural features = the boxes; control processes (rehearsal, attention, strategies) are controlled by the person

Control processes

Dynamic, person-controlled processes tied to the structural features of the modal model, e.g., rehearsal, mnemonic strategies, attention strategies

Rehearsal

Repeating a stimulus over and over to keep it in STM; maintenance rehearsal does little for LTM, elaborative rehearsal (using meaning) encodes better

Encoding vs retrieval

Encoding = getting information into LTM; retrieval = bringing info from LTM back into working memory/consciousness

Sensory memory

Very brief retention (fractions of a second to seconds) of the effects of sensory stimulation; holds all incoming information

Persistence of vision

Continued perception of a visual stimulus after it is gone; explains sparkler trails and why we don't see dark gaps between movie frames (24 frames/sec)

Iconic memory

Visual sensory memory (visual icon); lasts less than 1 second (Sperling)

Echoic memory

Auditory sensory memory; lasts a few seconds (about 2-4); e.g., "hearing" what someone said after you say "What?"

Sperling's whole report method

Flash 12 letters (3 rows of 4) for 50 ms; report as many as possible. Average = 4.5 of 12 letters

Sperling's partial report method

Immediately after the flash a tone cues which row to report. About 3.3 of 4 letters (82%) reported, so ~82% of the display was available; the rest faded during whole report

Sperling's delayed partial report

Cue tone delayed. With a 1-second delay only ~1 letter per row. Shows iconic memory decays in under 1 second

Why is partial report better than whole report?

Participants see most letters but sensory memory fades while they report; the cue lets them report from the trace before it fades

Short-term memory (STM)

Stores small amounts of information for a brief time; our "window on the present"; about 15-20 seconds without rehearsal

Duration of STM

About 15-20 seconds or less when rehearsal is prevented (Brown 1958; Peterson & Peterson 1959)

Brown-Peterson paradigm: method

Present 3 nonsense syllables/letters (e.g., FZL), count backward by 3s from a number for 0-18 seconds (prevents rehearsal), then recall

Brown-Peterson: results and conclusion

Recall ~80% after 3 s but ~12% after 18 s. Info fades from STM without rehearsal (trace decay); with rehearsal recall is much better. Counting also adds interference

Capacity of STM: Miller

"Magical Number Seven, Plus or Minus Two": about 5-9 items

Capacity of STM: Cowan / Luck & Vogel

More recent estimate: about 4 items

Digit span

Number of digits a person can remember in order; typical 5-9. KF had a digit span of 2

Change detection (Luck & Vogel, 1997)

Flash a display of colored squares (100 ms), blank 900 ms, then a second display; judge same or different. Performance near perfect for 1-3 squares, drops at 4+, so STM holds about 4 items

Chunking

Combining small units into larger meaningful units (words into phrases) to increase how much STM can hold

S.F. (Ericsson et al., 1980)

Runner with average memory (7-digit span) who trained 230 hours to recall 79 digits by chunking digits into running times and ages; used LTM knowledge to make chunks

Alvarez & Cavanagh (2004)

Change detection with objects of varying complexity: capacity 4.4 for colored squares but 1.6 for shaded cubes. More information per item = fewer items stored; supports "amount of information" view

Items vs information view of STM capacity

Items view: limit is a number of items (about 4). Information view: limit depends on amount of detail/complexity per item

STM vs working memory

Same system conceptualized differently: STM emphasizes brief storage; working memory emphasizes active manipulation of information for tasks like comprehension, learning, reasoning (e.g., mental math)

Working memory (definition)

Limited-capacity system for temporary storage AND manipulation of information for complex tasks (Baddeley & Hitch, 1974)

Why the modal model was challenged by Baddeley

Modal model treats STM as a single storage box, but people can read while holding digits in memory, and STM does more than store (math, language). Led to multi-component working memory

Baddeley's working memory model: components

Central executive, phonological loop, visuospatial sketchpad (original 3, Baddeley & Hitch 1974) plus episodic buffer (added 2000)

Central executive

"Traffic cop"/attention controller: coordinates the other components, focuses and divides/switches attention, pulls info from LTM, suppresses distractions; does not store info

Phonological loop

Holds verbal and auditory info: phonological store (few seconds, limited) + articulatory rehearsal process (keeps items from decaying); the "inner voice"

Visuospatial sketchpad

Holds visual and spatial information; used for imagery, puzzles, finding your way; the "inner eye"

Episodic buffer

Added by Baddeley (2000); integrates info from phonological loop, sketchpad and LTM into coherent episodes; extra storage capacity and link to LTM (explains memory for 15-20 word sentences)

Phonological similarity effect (Conrad, 1964)

Errors in recalling letters were based on sound (F confused with S or X) even though letters were seen. Shows phonological coding in the phonological store

Word length effect

Lists of short words are recalled better than long words because long words take longer to rehearse/pronounce; people recall what they can say in about 1.5-2 seconds

Articulatory suppression

Repeating an irrelevant sound ("the, the, the") prevents rehearsal in the phonological loop, reducing memory and eliminating the word length effect

Shepard & Metzler (1971) mental rotation

Participants judged if pairs of 3D shapes were the same; reaction time increased with angular difference (about 2 s at 40 degrees, 4 s at 140 degrees). Supports visual imagery in the sketchpad

Della Sala visual pattern task

Remember filled squares in a matrix; average about 9 squares before errors; far above 4 items so probably chunking into subpatterns; visual imagery

Brooks (1968) "F" task

Mentally trace the letter F: pointing to Yes/No (visuospatial) is harder than saying it (verbal) because pointing overloads the sketchpad. Same-store tasks interfere

Perseveration

Repeating the same action or thought even when it no longer works; seen with frontal lobe damage; breakdown of central executive's attention control

Phineas Gage

1848 railroad accident, tamping rod through frontal lobe; personality/planning/impulse control changes (accuracy of reports uncertain); drew attention to frontal lobe functions

Delayed-response task (Goldman-Rakic)

Monkey sees food in one of two wells, delay, then chooses. Removing the prefrontal cortex drops performance to chance, so PFC holds info across short delays

Funahashi et al. (1989)

PFC neurons fire to a cue location and keep firing during the delay; neural record of working memory for the location

Stokes: activity-silent working memory

Information held during a delay by short-term changes in synaptic connectivity (not continuous firing); lasts a few seconds

Working memory and the brain (distributed)

PFC is central but many areas (visual, parietal, LTM areas) interact; an example of distributed representation

Logie (1990) dual-task study: method

Participants did two tasks at once; each could be phonological (verbal) or visuospatial. IV = same store vs different stores; DV = performance

Logie (1990) results and conclusion

Good performance when tasks used different stores; poor when both used the same store. Supports Baddeley's model: separate verbal and visual subsystems

Operation span task

Letter -> math problem -> letter -> math problem (3-8 sets); must solve math (80% accuracy) and recall letters in order. Measures working memory capacity (storage + processing)

Reading span (Daneman & Carpenter, 1980)

Read sentences and recall the last word of each; span of 2-5 correlated with reading comprehension and verbal SAT. WM capacity is a source of individual differences

Vogel et al. (2005) ERP study

High- vs low-capacity WM groups did change detection with distracting blue bars. Low-capacity participants' ERPs increased more: they couldn't ignore distractors. Central executive differs; high-capacity people "tune out" distractors better

Working memory capacity: what is it related to?

Reading comprehension, academic performance, graduating high school, emotion control, creativity, cognitive control

Proactive interference (PI)

Older information interferes with learning/remembering new information (calling the new professor by the old professor's name)

Retroactive interference (RI)

New information interferes with remembering older information (calling the old professor by the new professor's name)

Brown-Peterson and interference

In lecture: counting backward disrupts the nonsense syllables, so forgetting involves retroactive interference as well as decay

Release from proactive interference (Wickens)

Recall drops over trials with same-category words (fruits), then rebounds when category changes (fruits to professions or reverse). Shows semantic coding in STM; interference depends on semantic similarity

Serial position curve

Plot of recall by position in list: best for first items (primacy) and last items (recency), worst in middle

Primacy effect

Better memory for early list items: more rehearsal time, so transferred to LTM

Recency effect

Better memory for last list items: still in STM at test

Rundus (1971)

Participants rehearsed aloud; number of rehearsals per word matched the primacy part of the serial position curve. Supports the rehearsal/LTM explanation of primacy

Glanzer & Cunitz (1966)

30 seconds of backward counting before recall eliminated the recency effect but not primacy. Recency is due to STM

Serial position: fast presentation

Primacy disappears (no time to rehearse into LTM) but recency remains

Serial position: delay with counting

Recency disappears (STM disrupted) but primacy remains

Serial position curve: conclusion

STM and LTM are separate systems: primacy reflects LTM, recency reflects STM

Coding in STM vs LTM

STM: mostly auditory/phonological (also visual, semantic). LTM: mostly semantic (also visual, auditory)

Sachs (1967)

Recognition memory for sentences in a passage: people remember the meaning but forget the exact wording. Semantic coding in LTM

Auditory coding example in LTM

Replaying a song in your mind

Visual coding example in LTM

Visualizing your fifth-grade teacher's face

HM (Henry Molaison)

Hippocampus removed (1953) to treat epilepsy; severe anterograde amnesia (no new episodic memories) but intact STM, semantic and procedural learning (mirror drawing improved with no memory of practice)

KF

Parietal lobe damage; poor STM (digit span 2, reduced recency) but intact LTM

Double dissociation (STM vs LTM)

HM: STM OK, LTM impaired. KF: STM impaired, LTM OK. Shows STM and LTM rely on different mechanisms

KC

Motorcycle accident damaged hippocampus: lost episodic memory but semantic memory largely intact; also could not imagine his future; learned sorting books (procedural)

LP

Encephalitis: lost semantic memory (couldn't recognize famous people or know facts) but episodic memory intact

Double dissociation (episodic vs semantic)

KC: semantic OK, episodic poor. LP: semantic poor, episodic OK. Supports separate mechanisms

Clive Wearing

Musician with viral encephalitis; hippocampal damage; severe anterograde amnesia, episodic memory impaired, but semantic and procedural (plays piano) intact; showed implicit learning (anticipating repeated videos); distress and disorientation

Anterograde vs retrograde amnesia

Anterograde = cannot form new memories after injury. Retrograde = loss of memories from before injury; graded (worst for events closest to injury)

Amnesia word-pair experiment (lecture)

Unrelated pairs (Cloud-Field); explicit encoding = just see them; implicit = make a sentence. Test cued recall (explicit) or word-stem completion (implicit). Amnesic patients poor on explicit but fine on implicit. Evidence for two memory systems

Ranganath & D'Esposito (2001)

fMRI: hippocampus activity rose while holding novel faces over a 7 s delay (not familiar faces). Hippocampus also contributes to short-term maintenance of novel info

Neuroplasticity (lecture, HM)

Brain can adapt; other areas can take over some functions after damage

Episodic vs semantic: neuroimaging (Levine)

Diary study with fMRI: episodic and semantic autobiographical memories activated overlapping but different brain regions

Interaction: knowledge affects experience

Semantic knowledge (e.g., baseball rules) guides attention and what becomes an episodic memory; the baseball double-play example

Autobiographical memory

Memory for specific experiences from your life; includes episodic components (relived events) and semantic components (personal semantic facts)

Personal semantic memory

Facts associated with personal experiences (e.g., that table by the window is our favorite)

Autobiographically significant semantic memory

Names of public figures you have personal experience with are recalled better (Westmacott & Moscovitch); advantage vanishes with episodic memory loss

Familiarity vs recollection

Familiarity = seems familiar without details (semantic, "know"). Recollection = remembering the specifics of the experience (episodic, "remember")

Remember/know procedure

Respond remember (recollect circumstances), know (familiar only) or don't know; distinguishes episodic from semantic components