Memory

In short

Memory is the set of processes by which the nervous system encodes, stores and retrieves information, letting past experience change present behaviour. It is the core topic of a Foundations/General Psychology paper and appears in every psychology entrance test (CUET-PG, UGC-NET Psychology Paper II, DU/JNU/TISS entrances) through multi-store models, Baddeley's working memory, forgetting theories and the serial position curve. Applied psychologists use it in learning design, clinical assessment (memory scales in dementia and head-injury work), human-factors design and legal settings where eyewitness reliability is at stake. In placement rounds for HR, L&D, UX research, market research and EdTech content roles, candidates are asked to translate memory findings into training design, survey wording or interface design.

Explained

Definition, basic processes and types of measure

• Memory - the capacity to acquire, retain and use information; studied as three stages: encoding (converting input into a usable code), storage (maintaining it over time) and retrieval (bringing it back into use).
• Information-processing approach: the mind is treated as a system with limited-capacity stores and control processes (attention, rehearsal, search), analogous to but not identical with a computer.
• Three codes are distinguished: visual (iconic), acoustic (phonological) and semantic (meaning-based). Short-term storage is mainly acoustic, long-term storage mainly semantic.
• Measures of retention:
  1. Recall - reproduce material without cues; free recall (any order), serial recall (same order), cued/paired-associate recall.
  2. Recognition - identify material among distractors; easier than recall because the cue is supplied (MCQ versus an essay question).
  3. Relearning / savings method (Ebbinghaus): savings % = (original trials − relearning trials)/original trials × 100 - the most sensitive measure, detecting traces that recall and recognition miss.
  4. Reconstruction and implicit measures (priming, speed gains) tap memory without conscious recollection.
• Explicit (declarative) memory is consciously reportable; implicit (non-declarative) memory shows only in performance - the key dissociation in amnesia research.

The multi-store (Atkinson-Shiffrin, 1968) model

• Three structurally distinct stores in sequence: sensory register → short-term store → long-term store, with rehearsal as the control process moving material onward.
• Sensory memory - a brief, high-capacity, modality-specific buffer holding a literal copy of the stimulus.
  • Iconic (visual): Sperling's (1960) partial-report experiment - a 3×4 letter array shown for 50 ms; whole-report gives ~4-5 letters, but a tone cueing one row immediately gives ~3 of 4, implying ~9-12 available but decaying in about 0.25-0.5 s.
  • Echoic (auditory): lasts longer, roughly 2-4 s, which is why speech can be understood across word boundaries.
  • Information survives only if attention selects it; the rest decays.
• Short-term memory (STM) - conscious, limited store.
  • Capacity: Miller's (1956) "magical number seven, plus or minus two" chunks; later estimates with chunking controlled give about 4 ± 1 items.
  • Duration: about 15-30 s without rehearsal - Peterson & Peterson's (1959) Brown-Peterson task used backward counting to block rehearsal; recall fell to near zero by 18 s.
  • Chunking - recoding items into larger meaningful units (a 10-digit mobile number as 5+5) raises effective capacity.
  • Coding is largely acoustic: phonological similarity effect (B, C, D, T, G confused more than F, K, Q, R) and word-length effect (fewer long words recalled).
• Long-term memory (LTM) - effectively unlimited capacity, duration up to a lifetime, semantic coding, forgetting mainly by interference and retrieval failure.
• Evidence for separate stores: serial position effect (two different causes for primacy and recency), differing coding errors, and double dissociations in patients - H.M. (bilateral medial temporal lobe removal, 1953) had intact STM and intact procedural learning but could form no new declarative LTM; patient K.F. had impaired digit span with intact long-term learning.
• Limitations: rehearsal is not necessary for LTM (incidental learning works), STM is treated as a single passive store, and the flow is too strictly serial.

Working memory (Baddeley & Hitch, 1974; four-component version 2000)

• Replaces the unitary STM with an active system that both stores and manipulates information during cognition.
• Central executive - limited-attention controller; allocates resources, switches tasks, inhibits irrelevant material; modality-free; linked to prefrontal function.
• Phonological loop - speech-based; a phonological store (traces fade in ~2 s) plus an articulatory rehearsal process that refreshes them. Explains the word-length and phonological-similarity effects and articulatory suppression (saying "the, the, the" abolishes them).
• Visuospatial sketchpad - holds visual appearance and spatial location; Logie's split into a visual cache and an inner scribe.
• Episodic buffer (added 2000) - limited-capacity, multimodal store that binds information from the subsystems and LTM into integrated episodes; explains why meaningful prose span (~16 words) far exceeds unrelated word span.
• Dual-task logic: two tasks using the *same* subsystem interfere; tasks using different subsystems do not - the main evidence for separable components.
• Working memory capacity correlates strongly with reading comprehension, arithmetic and fluid intelligence, which is why it is tested in CAT/placement aptitude style tasks and in clinical batteries.

Levels of processing and encoding factors

• Craik & Lockhart (1972): retention depends on the *depth* of processing, not on which store or how long rehearsal lasts. Structural (is it capitals?) → phonemic (does it rhyme?) → semantic (does it fit the sentence?) gives progressively better recall.
• Elaboration (richness of the added associations), distinctiveness and self-reference improve the trace; maintenance rehearsal (rote repetition) does little, elaborative rehearsal does much.
• Encoding specificity principle (Tulving & Thomson): retrieval succeeds to the extent that cues at recall overlap with cues present at encoding.
  • Context-dependent memory - external setting (Godden & Baddeley's divers: learning and recall both underwater or both on land beat mismatched conditions).
  • State-dependent (mood or drug state) and mood-congruent recall.
• Transfer-appropriate processing: deep encoding helps only if the test demands meaning; rhyme encoding wins on a rhyme-recognition test.
• Spacing (distributed practice) effect - the same total study time spread over sessions beats massed cramming; testing effect - retrieval practice produces more durable retention than re-reading; both are the empirical basis for spaced-revision apps.
• Consolidation - the stabilisation of a trace over hours to years, hippocampus-dependent at first and later neocortical; sleep, especially slow-wave sleep, supports it. Disrupting it causes retrograde amnesia graded by recency (Ribot's law).

Organisation of long-term memory

• Declarative (explicit):
  • Episodic - personally experienced events tagged with time and place; supports autonoetic (self-knowing) consciousness and mental time travel.
  • Semantic - general facts, concepts and word meanings, context-free. Models: Collins & Quillian's hierarchical network (predicts the category-size effect), spreading activation (Collins & Loftus), and schema/script theories (Bartlett's "War of the Ghosts" showed recall is *reconstructive*, assimilated to cultural schemas).
  • Autobiographical memory - episodic plus semantic self-knowledge; shows the reminiscence bump (disproportionate recall of events from roughly 10-30 years of age) and childhood amnesia before about 3-4 years.
• Non-declarative (implicit): procedural skills and habits (basal ganglia, cerebellum), priming (neocortex), classical conditioning, non-associative habituation.
• Prospective memory - remembering to *do* something later (event-cued or time-cued); the everyday failure behind missed medication and deadlines.
• Flashbulb memory - vividly detailed, confidently held memory of a surprising, consequential public event; vividness and confidence do not guarantee accuracy.
• Neural bases (kept brief - detail belongs to the biological-bases note): hippocampus and medial temporal lobe for encoding and consolidation of declarative material; amygdala for emotional modulation; prefrontal cortex for working memory and source monitoring; cerebellum and striatum for procedural learning. Long-term potentiation (LTP) is the leading synaptic mechanism.

Forgetting

• Ebbinghaus (1885) used nonsense syllables on himself: retention falls steeply in the first hour, then flattens - the forgetting curve, roughly logarithmic.
• Theories:
  1. Trace decay - disuse weakens the trace; adequate for sensory memory, weak for LTM (sleep spares memory better than equal waking time).
  2. Interference - retroactive (new learning impairs old) and proactive (old learning impairs new); strongest with similar material. Underwood/Melton tradition; also release from PI when the category changes.
  3. Retrieval failure / cue-dependent forgetting - the trace is available but not accessible; demonstrated by tip-of-the-tongue states and cued recall exceeding free recall.
  4. Motivated forgetting - repression (psychoanalytic, poorly supported experimentally) and retrieval-induced forgetting / directed forgetting (laboratory-robust).
  5. Reconsolidation failure and memory distortion - the trace changes rather than disappears.
• Misinformation effect (Loftus & Palmer, 1974): "smashed" versus "hit" in a question raised speed estimates and false reports of broken glass; the basis of modern caution about eyewitness identification. Indian courts treat eyewitness testimony with corroboration requirements, and the Bharatiya Sakshya Adhiniyam, 2023 (in force 1 July 2024, replacing the Indian Evidence Act, 1872) governs how such testimony and expert opinion are admitted.
• Serial position effect - U-shaped free-recall curve: primacy (early items rehearsed more, transferred to LTM) and recency (last items still in STM). A filled delay removes recency but spares primacy; faster presentation reduces primacy but spares recency - a classic double dissociation.
• Amnesias: anterograde (cannot form new memories - H.M., Korsakoff's syndrome from thiamine deficiency), retrograde (loss of pre-onset memories, temporally graded), transient global amnesia, dissociative/psychogenic amnesia (DSM-5-TR; ICD-11 code 6B61).
• Clinical assessment in India: PGI Memory Scale (Pershad & Wig) and the Wechsler Memory Scale are standard; AIIMS/NIMHANS neuropsychological batteries use them in dementia, epilepsy and head-injury evaluation. Testing by qualified professionals is regulated under the Rehabilitation Council of India Act, 1992 for clinical psychologist registration and the Mental Healthcare Act, 2017 for assessment and care standards.

Improving memory - techniques examined

• Mnemonics: method of loci, peg word, keyword method (second-language vocabulary), acronyms and acrostics, story/link method; all work by imposing organisation and imagery.
• Organisation: hierarchies, mind maps, categorising a list before learning (Bower's clustering studies).
• Imagery and dual coding (Paivio): concrete words, imaged twice (verbal + visual), are recalled better than abstract words.
• Spaced retrieval practice, interleaving, self-explanation and sleep after study are the best-evidenced study habits; re-reading and highlighting are the weakest.
• SQ3R (Survey, Question, Read, Recite, Review) and the PQRST method are the textbook study schemes.
• Limitations: mnemonics help arbitrary, unordered material but not conceptual understanding; expertise effects are domain-specific (a chess master's recall advantage vanishes with random board positions).

Key points

• Memory - encoding, storage and retrieval of information; three stages, not three places.
• Sensory memory - modality-specific, large, iconic ~0.25-0.5 s, echoic ~2-4 s; demonstrated by Sperling's partial report.
• STM capacity - 7 ± 2 chunks (Miller), ~4 with chunking controlled; duration 15-30 s (Brown-Peterson).
• Chunking - recoding into meaningful units; the one way to beat the capacity limit.
• LTM - unlimited capacity, semantic coding, lifetime duration; forgetting mainly interference and cue failure.
• Working memory - central executive + phonological loop + visuospatial sketchpad + episodic buffer; stores *and* processes.
• Articulatory suppression - blocks subvocal rehearsal, abolishing word-length and phonological-similarity effects; proves the loop.
• Levels of processing - semantic > phonemic > structural encoding; depth beats duration of rehearsal.
• Encoding specificity - recall is best when retrieval cues match encoding context (context- and state-dependent memory).
• Episodic vs semantic - events with time and place vs context-free facts (Tulving).
• Declarative vs procedural - H.M. learned mirror-drawing without remembering having done it.
• Serial position effect - primacy from rehearsal into LTM, recency from STM; dissociable by delay and presentation rate.
• Interference - retroactive (new on old) and proactive (old on new); greatest for similar material.
• Ebbinghaus forgetting curve - rapid initial loss then a long flat tail; savings is the most sensitive retention measure.
• Reconstructive memory - Bartlett's schemas and Loftus's misinformation effect; confidence is a poor index of accuracy.
• Consolidation - hippocampus-dependent stabilisation over time; sleep aids it, disruption causes graded retrograde amnesia.
• Spacing and testing effects - distributed practice and retrieval practice outperform massed study and re-reading.
• Amnesia - anterograde (new learning) vs retrograde (old memories); organic vs dissociative.

Formulas and examples

Example 1 - savings method. A student needs 24 trials to recite a 20-syllable list perfectly. After 48 hours she relearns it in 9 trials.

• Savings = (24 − 9)/24 × 100 = 15/24 × 100 = 62.5%.
• Interpretation: 62.5% of the original learning effort was spared, even though free recall before relearning was only 3/20 = 15%. The savings score reveals retention that recall misses.

Example 2 - Ebbinghaus decay equation. A trace has S = 20 hours. Retention after 1 h, 9 h and 24 h:

• t = 1: R = e^−0.05 = 0.951 (95.1%).
• t = 9: R = e^−0.45 = 0.638 (63.8%).
• t = 24: R = e^−1.2 = 0.301 (30.1%).
• The loss in the first hour (4.9 points) is small here because S is large; with S = 2 h, R at t = 1 is e^−0.5 = 0.607, showing the steep early drop Ebbinghaus observed.

Example 3 - chunking and span. A 12-digit code 9 1 4 7 2 0 3 8 5 6 1 2 is presented.

• Unchunked: 12 items against a span of about 7 → expected recall ≈ 7/12 = 58%.
• Chunked as four groups of three (914, 720, 385, 612): 4 chunks, well within span → expected recall near 100%.
• Capacity is measured in chunks, not elements: the same 12 digits became 4 units without losing information.

Example 4 - recognition corrected for guessing and d′. In a word-recognition test, 40 old and 40 new words are presented. The participant says "old" to 34 old words and 8 new words.

• H = 34/40 = 0.85; F = 8/40 = 0.20.
• Raw "accuracy" if you only count hits = 85% - an overestimate, because guessing produced 8 false alarms.
• Corrected P = (0.85 − 0.20)/(1 − 0.20) = 0.65/0.80 = 0.8125 (81.25%).
• d′ = z(0.85) − z(0.20) = 1.036 − (−0.842) = 1.88 - good but not perfect discrimination.

Example 5 - serial position curve. Twenty words are read aloud once; free recall across 30 participants gives these mean recall percentages by position:

• Mean overall recall = (68 + 41 + 33 + 45 + 74)/5 = 261/5 = 52.2%.
• If 30 s of backward counting is inserted before recall, positions 17-20 fall to about 35% while positions 1-4 stay near 68% - recency is STM-based, primacy is not.

Comparison table - the three stores:

Where it is used

• Learning and development / EdTech: spaced retrieval and chunked microlearning in corporate training and products such as BYJU'S, Unacademy, upGrad and Physics Wallah; instructional designers justify module length by working-memory limits.
• Clinical and neuropsychology: memory batteries (PGI Memory Scale, Wechsler Memory Scale) used at NIMHANS, AIIMS and private neuro clinics to assess dementia, epilepsy surgery candidates, traumatic brain injury and disability certification under the Rights of Persons with Disabilities Act, 2016.
• Legal and forensic psychology: eyewitness reliability, line-up construction and interview protocols (cognitive interview) advised to police training academies and courts; relevant to testimony admitted under the Bharatiya Sakshya Adhiniyam, 2023.
• Human factors and UX research: interface design that minimises memory load - OTP length, recognition-over-recall menus in UPI apps, error messages that do not require remembering a prior screen; roles at Flipkart, Zomato, Paytm and design studios.
• Advertising and market research: aided and unaided brand recall, day-after recall tests, and ad-wearout studies run by Kantar, Nielsen and IPSOS India.
• Teaching, HR and assessment design: item writing that distinguishes recall from recognition, interview structuring to limit interviewer memory bias, and training evaluation at Kirkpatrick Level 2 (learning retention).

Interview questions

• Q: What are the three basic processes of memory? A: Encoding (converting information into a usable representation), storage (maintaining it over time) and retrieval (accessing it when needed); a failure at any one looks like forgetting.
• Q: How does working memory differ from short-term memory? A: Short-term memory is a passive store of limited capacity; working memory is an active multi-component system (central executive, phonological loop, visuospatial sketchpad, episodic buffer) that holds *and* manipulates information during thinking.
• Q: What does the serial position effect prove? A: That different mechanisms underlie primacy and recency - primacy from rehearsal into long-term memory, recency from items still in short-term store - supporting more than one memory system.
• Q: Why is recognition usually easier than recall? A: The test item itself supplies the retrieval cue, so it needs only a matching decision, whereas recall requires generating the material from internal cues.
• Q: State the encoding specificity principle with an example. A: Retrieval is best when cues present at recall match those at encoding; divers who learnt words underwater recalled more underwater than on land.
• Q: What did H.M.'s case establish? A: That declarative and procedural memory are separate systems and the medial temporal lobe is needed to form new declarative memories - he improved at motor tasks with no recollection of practising.
• Q: Is forgetting loss of the trace or failure to find it? A: Both occur, but much everyday forgetting is retrieval failure - tip-of-the-tongue states and the gain from cues show the information is available yet temporarily inaccessible.
• Q: Distinguish retroactive from proactive interference. A: Retroactive: later learning disrupts recall of earlier material; proactive: earlier learning disrupts recall of later material; both are worst when the two sets are similar.
• Q: Why are flashbulb memories not necessarily accurate? A: Emotional arousal raises vividness and confidence but not fidelity; repeated retelling introduces distortion, so confidence is a poor index of accuracy.
• Q: As an instructional designer, how would you apply memory research to a two-hour training module? A: Chunk content into 5-7 item units, use elaborative and visual encoding, build in spaced retrieval quizzes rather than re-reading, and schedule a follow-up booster after a day or two.
• Q: What is the misinformation effect and why does it matter in court? A: Post-event information, including the wording of a question, can alter a witness's memory (Loftus & Palmer); it is why leading questions and suggestive line-ups undermine testimony.
• Q: Which single study habit has the strongest evidence, and why? A: Spaced retrieval practice - actively recalling material at intervals strengthens and diversifies retrieval routes far more than re-reading, which only increases familiarity.

Common mistakes

• Treating Miller's 7 ± 2 as items rather than chunks, and quoting it as a fixed, universal capacity when controlled estimates are nearer four.
• Describing working memory as just "a bigger STM" and omitting the central executive and episodic buffer, or forgetting the episodic buffer was a later (2000) addition.
• Confusing the directions of retroactive and proactive interference in exam answers - always write the design (Learn A → Learn B → Recall which?) before answering.
• Equating depth of processing with *time spent*; long maintenance rehearsal still produces poor retention.
• Mixing up episodic/semantic (a content distinction within declarative memory) with explicit/implicit (an awareness distinction).
• Saying sensory memory "holds information for a few seconds" without separating iconic (fractions of a second) from echoic (a few seconds).
• Writing retention percentages without stating the measure used - recall, recognition and savings give very different numbers for the same learning.

Related topics

Learning · Sensation and perception · Biological bases of behaviour · Intelligence · Schools of psychology · Cognitive processes: thinking and problem solving