SUMMARY

2.1 CODING, CAPACITY & DURATION

STM

* Acoustic coding

* Capacity: 5–9 items

* Duration: ~18 sec

LTM

* Semantic coding

* Capacity: unlimited

* Duration: lifetime

Coding – Baddeley

* STM worse with acoustically similar words

* LTM worse with semantically similar words

• Shows STM & LTM are separate

• Artificial stimuli ↓ validity

Capacity

Jacobs: digit span → digits = 9.3, letters = 7.3

Miller: “magic number 7 ± 2”, chunking helps

• Jacobs reliable

• Miller overestimated – Cowan says 4 ± 1

Duration

STM – Peterson & Peterson:

* After 18s → 3% recall

* Artificial → ↓ external validity

LTM – Bahrick (Yearbooks):

* 90% at 15 years

* 70% at 48 years

* High external validity

2.2 MULTI-STORE MODEL (MSM)

Stores: Sensory Register → STM → LTM

Rehearsal transfers STM → LTM

Sensory Register: large capacity, very brief, modality-specific

STM: acoustic, 18 sec, 7±2 items

LTM: semantic, unlimited, lifetime

HM case study: damaged LTM, intact STM

Evaluation

✔ Baddeley supports separate stores

✘ STM not unitary — KF study

✘ Rehearsal type matters (Craik & Watkins)

✘ LTM is not one store — Tulving

2.3 TYPES OF LTM (Tulving)

Episodic: events, time-stamped, conscious recall

Semantic: facts/knowledge, not time-stamped

Procedural: skills, unconscious recall

Clive Wearing:

* Intact: procedural/semantic

* Damaged: episodic

Evaluation

✔ Case studies support separation

✘ Lack control in case studies

✘ Neuroimaging findings conflict

✔ Real applications: improve older adults’ episodic memory

✘ Debate whether episodic is part of semantic

2.4 WORKING MEMORY MODEL (Baddeley & Hitch)

Replaces STM in MSM

Central Executive: allocates attention, limited capacity

Phonological Loop: sound, inner ear + inner voice

Visuo-Spatial Sketchpad: visual/spatial data

Episodic Buffer: integrates info, link to LTM

KF study: impaired verbal, intact visual → supports separate stores

Evaluation

✔ Clinical (KF) & dual-task support

✘ CE unclear and vague

✘ Lab tasks = low external validity

2.5 INTERFERENCE (Forgetting)

When memories compete, especially in LTM

* Proactive: old affects new

* Retroactive: new affects old

(PORN = Proactive Old / Retroactive New)

McGeoch & McDonald: Similar memories = worst recall

Evaluation

✔ Rugby study (Baddeley & Hitch) = real world support

✘ Interference rare — cues may matter more

✔ Cues can overcome interference

✔ Drug studies support interference

✘ Lab studies → artificial tasks

2.6 RETRIEVAL FAILURE (Forgetting)

We forget because of lack of cues

Encoding Specificity Principle (Tulving):

Cue must be present at learning & recall

Context-dependent: e.g. land vs water (Godden & Baddeley)

State-dependent: e.g. sober vs drunk (Carter & Cassaday)

Evaluation

✔ Real-world uses (revision, exams)

✔ Strong lab & field support

✘ Context effects weak in real life

✘ Only affects recall, not recognition

✘ Hard to prove cues were encoded

2.7 EWT – MISLEADING INFORMATION

Leading Questions – Loftus & Palmer:

“smashed” → higher speed estimate & false glass memory

Post-event Discussion – Gabbert:

71% recalled things they didn’t see

Evaluation

✔ Real-world application (police, legal)

✘ Lab studies lack consequences

✔ Central details more resistant (Sutherland & Hayne)

✘ Memory contamination → not just conformity

✘ Demand characteristics likely

2.8 EWT – ANXIETY

Negative: Weapon Focus

Johnson & Scott: Knife = worse recall (33%)

Positive: Fight-or-flight

Yuille & Cutshall: real robbery → high stress = best recall (88%)

Inverted U theory: Moderate anxiety = best recall

Evaluation

✔ Unusualness matters (Pickel)

✔ Field studies support negative effects (Valentine & Mesout)

✔ Field support positive (Christianson & Hubinette)

✘ Lack control over real-life studies

✘ Ignores cognitive/emotional aspects

2.9 COGNITIVE INTERVIEW (CI)

4 techniques:

1. Report everything

2. Reinstate context

3. Reverse order

4. Change perspective

Enhanced CI: builds rapport, reduces anxiety

Evaluation

✔ 41% more accurate info (Köhnken)

✘ Also more incorrect info

✔ Some elements (report everything + context) are best

✘ Time consuming & expensive

✘ Variations make comparison difficult