PSYCH 390: Research Methods in Memory

Unit 1: Memory in the Real World and Why It Fails

Memory as a Reconstructive Process

Human memory does not work like a video camera. Rather than storing a perfect, permanent record of experience, it preserves selected features and reconstructs the event at retrieval, filling in gaps with expectation, inference, and general knowledge. This reconstructive character explains why memory is both remarkably adaptive and surprisingly fallible.

A foundational demonstration of this comes from gist-based memory. We tend to retain the general meaning or theme of an experience far better than its precise sensory details. Elizabeth Loftus and colleagues showed that when witnesses to an event are later exposed to misleading post-event information — even something as subtle as a leading question that phrases a yield sign as a stop sign — they frequently incorporate that misinformation into their recollection and report it with confidence. The memory that emerges at test is a blend of what was originally encoded and what was inferred or suggested afterward.

This reconstructive process serves us well most of the time: we do not need a verbatim transcript of every conversation to navigate daily life. But it creates problems whenever precision matters — in eyewitness testimony, medical recall, or any situation where the specific details distinguish one memory from another.

Evolutionary Perspective on Memory

Memory did not evolve to be accurate; it evolved to be useful. Our ancestors benefited from remembering the location of food, the faces of allies and enemies, and the outcomes of past actions. This means human memory is especially strong for emotionally salient, personally relevant, and spatial information, and comparatively weak for verbatim text, arbitrary lists, and perceptual details that carry little survival value.

Recognising this evolutionary bias helps explain a common frustration: students studying from textbooks often feel they understand the material during reading, yet cannot reproduce it on a test. Understanding activates the general schema of an idea, but precise details — dates, definitions, names — require a different kind of effortful encoding.

Neurogenesis and Adult Memory Capacity

Early models assumed the brain was a fixed organ whose neurons could only die, not multiply. We now know that neurogenesis — the birth of new neurons — occurs in the adult hippocampus, the region most critical for forming new memories. This finding implies that the capacity for new learning is not simply fixed at birth, and that experience, exercise, and enriched environments can support the production of new hippocampal cells throughout life.

Flashbulb Memories and the Illusion of Precision

Some memories feel photographically precise: where you were, what you were doing, and who told you when you first heard about a major public event. These are flashbulb memories — vivid, confident, highly detailed recollections of the moment one learned of a surprising, consequential event. They were originally proposed to be laid down by a special neurobiological mechanism triggered by novelty and emotional arousal, producing memories that, unlike ordinary ones, do not fade or distort.

Subsequent research has not supported this special-mechanism account. Studies tracking people’s memories for major events over months and years find that flashbulb memories are just as subject to distortion, forgetting, and inconsistency as other autobiographical memories. What is distinctive is not accuracy but confidence: people believe flashbulb memories are more accurate than ordinary ones and resist updating them when corrected. The subjective sense of vividness — the phenomenology of “remembering” — is not a reliable guide to objective accuracy. This dissociation between felt vividness and factual accuracy is one of the most important lessons from memory science for everyday life.

Unit 2: Taxonomy of Memory Systems

Declarative and Non-Declarative Memory

The most important broad distinction in memory science separates declarative memory, which can be consciously accessed and verbally reported, from non-declarative memory, which influences behaviour without necessarily entering awareness.

Within declarative memory, episodic memory stores personally experienced events embedded in their specific spatial and temporal context — not just that a birthday party occurred, but which birthday party, with whom, and what happened. Semantic memory stores general world knowledge — facts, concepts, and language — stripped of the autobiographical context in which they were originally acquired.

Procedural Memory and Classical Conditioning

Procedural memory supports the learning of skills and habits — riding a bicycle, playing piano scales, or typing. Unlike episodic and semantic memory, procedural learning typically occurs gradually with practice and does not require conscious attention to the rules being acquired. The basal ganglia and cerebellum are critical for this system.

Classical conditioning represents another form of non-declarative memory. After repeated pairing of a neutral stimulus (the conditioned stimulus, or CS) with a biologically significant stimulus (the unconditioned stimulus, or US), the CS comes to elicit a conditioned response (CR) on its own. The fear conditioning paradigm — famously illustrated historically by Watson and Rayner’s “Little Albert” study — demonstrates that emotionally charged associations are acquired through this mechanism. The amygdala is the crucial structure for emotional, affective learning.

Priming

Priming is the facilitation of processing or retrieval produced by prior exposure to an item or a related item, without any requirement for conscious recollection of that prior exposure. There are two principal varieties:

Repetition priming (also called perceptual priming): Prior exposure to a stimulus speeds up or improves performance on a subsequent encounter with the same or perceptually similar stimulus. In a lexical decision task (deciding whether a string of letters is a word), participants respond faster to words they have recently seen, even when they have no explicit memory of having seen them. Repetition priming is preserved in amnesic patients.

Associative priming: Prior exposure to a stimulus facilitates processing of a related stimulus. Seeing bread makes recognition of butter faster, even if bread was not presented in a recent study list. This demonstrates that long-term semantic associations — not just recent perceptual history — influence current processing.

Priming is classified as non-declarative because it does not require, and is not blocked by, frontal or hippocampal damage. It is thought to arise from perceptual or conceptual representation systems that encode the features of stimuli and carry traces of prior activation. These systems are distinct from the episodic memory system and from semantic memory proper.

Implicit Memory and the Mirror-Drawing Task

Implicit memory refers to changes in performance or behaviour that reflect prior experience without requiring conscious recollection of that experience. The mirror-drawing task provides the most striking demonstration: patient HM, who had his hippocampus surgically removed to treat severe epilepsy, could not consciously recall attempting the task from day to day. Yet his performance improved steadily across sessions, just as in healthy participants. Each morning he was convinced he had never tried the task before — and yet his hand knew what to do. This dissociation proves that skill acquisition and declarative memory are supported by different neural systems.

Unit 3: The Neuroscience of Memory Formation

Patient HM and the Role of the Hippocampus

Henry Molaison, known in the literature as patient HM, underwent bilateral hippocampal removal in 1953 to treat intractable epilepsy. The operation was effective at reducing seizures but produced a devastating and permanent anterograde amnesia — the inability to form new long-term memories for events after the surgery. His retrograde amnesia extended approximately a decade back in time, with more distant memories relatively preserved.

HM could hold a conversation, read, and perform many cognitive tasks normally — his working memory and semantic knowledge were largely intact. What he could not do was transfer new experiences into lasting long-term storage. He never learned the names of the researchers who visited him daily, never recognised his own aging face in a mirror, and could not find his way around the nursing home he moved to in later life. The study of HM established the hippocampus as the critical structure for consolidating new declarative memories.

Long-Term Potentiation and Hebbian Learning

At the synaptic level, memory formation is widely believed to depend on long-term potentiation (LTP) — a persistent strengthening of synaptic connections following repeated or intense neural activation.

The conceptual framework underlying LTP traces to Donald Hebb’s 1949 principle: “neurons that fire together, wire together.” When two neurons are repeatedly activated simultaneously, the synapse between them becomes more efficient. Over time, interconnected assemblies of neurons — cell assemblies — come to represent specific memories, such that partial reactivation of the assembly (by a single cue) can reinstate the whole pattern.

Consolidation and the Standard Model

The standard model of memory consolidation proposes that newly encoded memories are initially fragile and hippocampus-dependent, requiring a period of consolidation to become stable. During this window, the memory can be disrupted by head trauma, electroconvulsive therapy, or interference from similar learning. Over time, through repeated retrieval and sleep-dependent replay, the memory is gradually transferred to and becomes independent of distributed cortical networks, eventually residing in a representation that no longer requires hippocampal mediation.

Evidence for this transfer comes from the retrograde amnesia gradient in patients like HM: older, well-consolidated memories are more resistant to hippocampal damage than more recent ones. However, a rival account — the multiple trace theory — argues that the hippocampus is never truly dispensable for episodic memories, because each retrieval creates a new hippocampally mediated trace, and fully context-rich memories remain hippocampus-dependent even decades later. On this view, what becomes independent of the hippocampus over time is semantic, decontextualised knowledge, while genuinely episodic memory retains hippocampal dependence permanently.

The debate between these frameworks has practical implications: the standard model predicts that sufficiently remote memories will survive hippocampal damage intact, while multiple trace theory predicts that all episodic memories, regardless of age, will be impaired.

Neuroimaging Methods

Modern cognitive neuroscience uses several complementary techniques to study the brain bases of memory. Structural MRI images brain anatomy at high resolution, allowing researchers to observe hippocampal volume changes in aging or disease. fMRI (functional MRI) tracks blood oxygen levels as a proxy for neural activity. PET (positron emission tomography) measures regional blood flow or glucose metabolism. TMS (transcranial magnetic stimulation) temporarily disrupts processing in a targeted region, allowing causal inferences about its necessity. EEG records electrical activity through scalp electrodes, providing millisecond temporal resolution. CT (computed tomography) is primarily used to identify structural lesions such as tumours or haemorrhages.

Each technique involves trade-offs between spatial resolution (where) and temporal resolution (when). Converging evidence across methods gives the most reliable picture of memory’s neural underpinnings.

Unit 4: Sensory Memory and the Multi-Store Framework

The Atkinson-Shiffrin Model

The Atkinson-Shiffrin model proposes three distinct memory stores: sensory memory, short-term memory (STM), and long-term memory (LTM). Information enters through sensory buffers, is selectively transferred into short-term memory by attention, and can then be rehearsed into long-term storage. While later models have substantially revised this picture, the three-stage framework remains a useful starting point.

Iconic Memory

Iconic memory is the visual sensory buffer, preserving a brief, high-capacity representation of the visual scene. The classic demonstration comes from Sperling’s (1960) partial report experiments. In the whole report condition, participants tried to recall all letters from a briefly flashed 3×4 matrix; they averaged about 4–5 items — yet reported seeing more than they could name. The icon was fading as they named items.

In the partial report condition, a tone immediately following the display offset cued participants to report only one row. Under these conditions, participants could report nearly all letters in the cued row regardless of which row was signalled — demonstrating that all items were available in the icon immediately after the display. As the delay between display offset and tone increased, partial report performance fell toward whole report levels, showing that the icon decays within about 300 milliseconds.

The capacity of iconic memory is approximately 75% of the display, far exceeding the capacity of short-term memory. Backward masking — presenting a patterned mask immediately after the target display — erases the icon before it can be read out, confirming its sensory, pre-categorical nature.

Echoic Memory

The auditory analogue of iconic memory is echoic memory, which preserves incoming sound for approximately 4–5 seconds. The “three-ear man” procedure delivered simultaneous messages to three ear positions and used a tone cue to signal which stream to report. Unlike iconic memory, echoic memory is relatively language-specific: speakers of a particular language show better echoic storage for utterances in that language, suggesting some phonological categorisation before full decay.

The Serial Position Curve

When participants hear or read a list of unrelated words and immediately try to recall them in any order, recall probability follows a characteristic U-shaped function of serial position: items at the beginning (primacy effect) and end (recency effect) of the list are recalled better than items in the middle.

Critically, the recency effect disappears when recall is delayed by a filled interval (e.g., counting backwards for 30 seconds) before the recall test — consistent with the view that recent items are held in a short-term buffer rather than in LTM. The modality effect refers to the finding that the recency advantage is larger for auditory than visual presentation, attributed to the longer persistence of echoic relative to iconic memory. The suffix effect shows that adding a spoken word (e.g., “zero”) after the final list item abolishes the recency advantage for auditory lists: the suffix enters the echoic store and overwrites the just-presented final items.

Unit 5: Short-Term Memory and Working Memory

Short-Term Memory: Capacity and Forgetting

William James distinguished between primary memory (the immediate present, roughly what we now call short-term memory) and secondary memory (the long-term storehouse). The digit span, averaging 7 ± 2 items, reflects the capacity of this primary memory system.

Chunking dramatically extends effective capacity. When individual items are grouped into meaningful units — area codes, acronyms, syllables — each chunk occupies one “slot,” allowing far more raw information to be held. Memory champions exploit this principle to memorise hundreds of digits by converting digit strings into vivid narratives or images.

Brown and Peterson’s studies of forgetting from STM revealed that even a few seconds of rehearsal prevention produces massive forgetting. Participants heard a consonant trigram, counted backwards by threes (preventing rehearsal), and recalled the trigram after delays up to 18 seconds. Recall dropped steeply, approaching chance by 18 seconds. This was originally interpreted as rapid decay — spontaneous fading of the memory trace. However, subsequent work showed that proactive interference also plays a major role: earlier trigrams contaminate recall of later ones. Wickens demonstrated release from proactive interference by shifting the category of materials mid-experiment — recall recovered dramatically on the first trial of the new category, confirming that PI had been causing the decline, not pure decay.

Evidence against a single, unitary short-term store came from patients like PV, who had near-zero digit span yet could learn new language normally — dissociations that pointed toward a fractionated system rather than a single buffer.

The Episodic Buffer

In 2000, Baddeley proposed a fourth component of working memory: the episodic buffer. This is a limited-capacity temporary store that integrates information across different codes — linking phonological information from the loop, spatial information from the sketchpad, and long-term memory representations — into coherent, multi-dimensional “episodes.” Unlike the loop and sketchpad, the episodic buffer is under the direct control of the central executive and can bind information across modalities and time scales. The episodic buffer explains phenomena that neither the loop nor the sketchpad can account for individually — such as the fact that memory for meaningful prose exceeds what the phonological loop alone could support, because long-term semantic knowledge held in the episodic buffer supplements the phonological code.

Baddeley’s Working Memory Model

In 1974, Alan Baddeley and Graham Hitch replaced the unitary STM concept with a multi-component working memory model. Rather than a passive storage box, working memory is an active workspace that simultaneously holds and manipulates information.

The Phonological Loop

The phonological loop maintains spoken or speakable information. It consists of a phonological store (inner ear) that holds speech-based information for 1–2 seconds before decay, and an articulatory rehearsal process (inner voice) that refreshes the store by sub-vocal repetition.

Evidence for the phonological loop is extensive:

• Phonological similarity effect: Phonologically similar words (e.g., cat, hat, bat) are recalled more poorly than dissimilar words, because the store represents items in phonological code and similar items are more easily confused.
• Word length effect: Short words are recalled better than long words in serial recall. Recall capacity tracks articulation time rather than syllable count, explaining cross-linguistic differences in memory span.
• Unattended speech effect: Irrelevant speech in the background impairs serial recall even when participants cannot understand it, because incoming speech has obligatory access to the phonological store.
• Articulatory suppression: Repeating an irrelevant sound (e.g., “the-the-the”) while studying visual material disrupts both the word length and phonological similarity effects, because the articulatory loop is occupied.

Patient PV, with selective impairment of the phonological store, had very limited digit span and strikingly poor ability to learn new foreign vocabulary — confirming the phonological loop’s vital role in language acquisition.

An important cross-linguistic finding: Welsh words have shorter mean articulation duration than English words, so Welsh-speaking children measured on English-administered tests appear to have larger digit spans — and artificially higher IQ scores that include digit span as a component. This artefact illustrates how the word length effect can contaminate psychometric measurement across languages.

The Visuospatial Sketchpad

The visuospatial sketchpad maintains and manipulates visual and spatial information — the “mind’s eye.” Evidence converges from several sources:

• Spatial equivalence (Kosslyn’s island map study): After memorising a map of an imaginary island, participants took longer to mentally scan between locations that were further apart on the map — exactly as if scanning a real scene. Mental images preserve metric spatial relationships.
• Perceptual equivalence: Imagery and perception share neural substrates; the same cortical regions active during visual perception are active during visual imagery. Single-cell recordings in animals confirm that the same neurons respond both during perception of an object and during its imagination.
• Dual coding theory (Paivio): Concrete, imageable words are encoded both verbally and as visual images, giving two independent retrieval routes and a memorial advantage over abstract words, which have only a verbal code.
• Transformational equivalence: Mental rotation is a continuous, time-consuming process that mirrors physical rotation.

Unilateral spatial neglect — in which patients ignore one half of space following parietal lesions — demonstrates that the sketchpad is not merely a metaphor. Patients asked to imagine a familiar scene describe only the right side; when asked to imagine the scene from the opposite end, they describe what was formerly the neglected left half. This proves that imagery uses the same spatial representations as real-world perception.

The Central Executive

The central executive is the control system that allocates attentional resources among subsystems, coordinates dual tasks, suppresses irrelevant information, and shifts between strategies. Murdoch’s (1965) card-sorting dual-task study elegantly demonstrated this role: when participants sorted cards into 2, 4, or 8 categories while simultaneously encoding a word list, recall decreased in direct proportion to the attentional demands of sorting. Simply turning cards over (no sorting) produced no effect — it was the drawing on executive resources that caused forgetting.

Unit 6: Encoding — Factors that Determine Learning

Distributed Practice

One of the most robust findings in learning research is that distributed (spaced) practice produces better long-term retention than massed practice (cramming). Baddeley and Longman’s (1978) study of postal workers learning keyboard skills demonstrated this in an applied setting. Workers trained with shorter, more frequent sessions showed far better skill retention at follow-up than those trained in intensive daily blocks, despite spending equivalent total time on the task.

The mechanism involves temporal coding in the hippocampus: the hippocampus encodes when an event occurred, and studying material that was last studied some time ago constitutes a novel encoding context, forcing reconstruction and reconsolidation. Successive reconsolidation gradually builds a durable long-term representation.

Expanded Retrieval Practice

Landauer and Bjork showed that spacing retrieval attempts in an expanding schedule — testing yourself after a short delay, then a longer delay, then a longer one still — is even more effective than fixed-interval spacing. The Anki flashcard application implements this expanding retrieval schedule algorithmically, testing items just before they would otherwise be forgotten. This strategy exploits Ebbinghaus’s forgetting curve: each retrieval attempt that requires effortful search reconsolidates the memory and flattens the subsequent forgetting rate.

The critical principle is that active recall — generating the answer yourself — is what matters, not passive re-reading. Professor Sanford memorised a prayer over 5,000 times through passive recitation while reading along, yet could not reproduce it in a lab setting when given only a cue. The BBC advertising study similarly showed that housewives who had heard a frequency-change jingle repeatedly showed only 17% accuracy when tested, because neither passive re-reading nor incidental radio exposure engages active retrieval.

The Role of Attention

Attention is a necessary ingredient for explicit memory formation. The dose-response relationship is clear in the Murdoch dual-task study: the more attentional resources diverted to a secondary task, the worse recall for the primary list. The “moonwalking bear” video dramatically illustrates inattentional blindness — when attention is focused on counting basketball passes, a gorilla walking through the scene goes unnoticed by most observers.

This has direct implications for eyewitness testimony. A victim whose attention is riveted on a weapon (weapons focus effect) may be unable to describe the perpetrator, because the intense attentional spotlight prevented encoding of peripheral details.

An important nuance: implicit memory is largely unaffected by attentional division at encoding. Amygdala-mediated emotional associations, perceptual priming effects, and classical conditioning can proceed with little attentional engagement. The Bechara study elegantly dissociated explicit declarative knowledge from implicit emotional conditioning using patients with selective hippocampal or amygdala lesions. The patient without an amygdala could identify which screen colour predicted the aversive tone (declarative knowledge intact) but showed no conditioned skin conductance response (emotional learning absent). The patient without a hippocampus showed the conditioned response (emotional learning intact) but could not identify the CS verbally. The patient lacking both showed neither.

Sleep and Memory Consolidation

Sleep plays an active and essential role in memory consolidation. During sleep, the hippocampus replays the day’s experiences, reactivating newly formed cell assemblies during sharp-wave ripples — high-frequency oscillatory events. The strength of this reactivation during sleep correlates with later memory performance. Optogenetically disrupting sharp-wave ripples in animal models selectively impairs memory for new environments while leaving familiar ones intact.

Targeted memory reactivation (TMR) exploits this mechanism: playing a cue (e.g., music present during learning) during sleep triggers preferential reactivation of the associated memories and enhances them.

The practical implication is clear: studying material and then sleeping allows the hippocampus to consolidate newly formed traces. Sleep deprivation disproportionately impairs free recall (effortful search) relative to recognition (familiarity judgement), and chronic sleep disruption is associated with measurable memory deficits.

A broader point is that consolidation is not a passive process. Rather than the brain passively “saving” a recording, consolidation is an active reconstruction in which hippocampal replay selectively strengthens some traces while allowing others to fade. Emotional content, novel information, and material rehearsed immediately before sleep all receive consolidation priority. Furthermore, REM sleep appears particularly important for integrating new memories with existing knowledge structures — the associative and creative leaps that often occur in the morning after sleep may reflect the REM-mediated loosening of semantic constraints, allowing distantly related concepts to become linked. Slow-wave sleep (SWS), where sharp-wave ripples are concentrated, appears most important for the hippocampal-to-cortical transfer of episodic memories.

The Role of Organisation

Memory is dramatically improved by imposing meaningful structure on incoming information. Three converging lines of evidence establish this:

Hierarchical structure at encoding (Bower et al.): Presenting words organised into a semantic hierarchy boosted recall from 21 to 73 items out of 112, compared to random presentation of the same words. Simply seeing the hierarchical skeleton before encountering the items was sufficient, because it pre-activates the relevant long-term memory categories into which new items can be slotted.

Cued recall (Tulving): Providing category labels at retrieval produces massive improvements over free recall. This demonstrates a critical distinction between availability (what is stored) and accessibility (what can be retrieved given the right cue).

Incidental learning through sorting (Mandler, 1967): When participants sorted cards by any criterion of their choosing — without being told a memory test was coming — they later recalled the items well above chance. The mere act of searching for organisational relationships among items, even arbitrary ones, constitutes deeper processing that benefits later retrieval.

The bartender study (Beach) compared novice and expert bartenders learning to associate drinks with differently shaped glasses. Novices relied on verbal rote rehearsal; experts relied on an external visual mnemonic — the distinctive glass shape cued the recipe without internal rehearsal. Counting backwards impaired novices but not experts (their phonological loops were occupied). Removing the distinctive glasses impaired experts but not novices (eliminating the external visual cue). Overall experts outperformed novices, but the interaction proved that the strategy chosen determines which type of distraction is disruptive. This pattern illustrates a broader principle: expertise does not merely increase raw memory capacity; it changes which memory subsystems are recruited and what kinds of disruption will impair performance.

The actor study (Noyce and Noyce) examined how professional actors memorise scripts. Rather than rote memorisation (mean endorsement: 8.66%), actors predominantly used interaction — thinking about the emotional and motivational significance of their lines in relation to the other characters and the dramatic arc (mean endorsement: 42.21%). The mnemonic Harry Lorayne, added as a comparison, relied almost exclusively on visual imagery (imagining candles raining for “Kendal Frain”). Both strategies outperform rote memorisation by engaging deeper meaning or richer imagery.

The Generation Effect

A striking demonstration of encoding depth is the generation effect: words that are actively generated from a cue (e.g., completing hot and ___ with “cold”) are better remembered than words that are simply read. The extra effort required to generate the target constitutes elaborative processing — the person must access semantic knowledge and verify the candidate word against the cue, creating a richer and more distinctive memory trace. Curiously, the generation effect is absent for implicit memory tests: generated words are not more fluent on a lexical decision test than read words, suggesting the effect is specific to explicit, recollection-based retrieval. This dissociation between implicit and explicit memory benchmarks the generation effect as a genuinely encoding-depth phenomenon rather than a simple exposure-intensity effect.

The Role of Meaning: Levels of Processing

Craik and Tulving’s classic experiment showed that encoding operations requiring processing of the meaning of items produce far better long-term retention than operations focused on surface features. Participants judged whether a word was in uppercase (shallow, perceptual), whether it rhymed with a target (intermediate), or whether it fit meaningfully into a sentence (deep, semantic). Recognition was highest after semantic encoding, lowest after case judgement.

Elaborative encoding — linking new material to pre-existing knowledge, generating examples, considering implications — constitutes the deepest semantic processing. Self-referencing (asking how material relates to one’s own life) recruits additional frontal lobe processing and produces some of the highest retention scores in the laboratory.

A powerful applied illustration: imagine a doctor explaining a complex multi-medication regimen. Drawing on these principles, an effective physician would: (a) state the organisational structure upfront (scaffold); (b) present the most critical medications first (serial position — primacy effect); (c) link each medication to a pre-existing routine such as meals or bedtime; (d) request active recall partway through (“what do you take at breakfast?”); (e) use vivid imagery to make the associations memorable; and (f) employ the expanding retrieval schedule within the appointment by spacing brief reviews at progressively longer intervals.

Unit 7: Retrieval and Context Effects

Context-Dependent Memory

Recall is best when the conditions at retrieval match the conditions at encoding. Godden and Baddeley’s underwater study is the paradigm case: scuba divers learned a word list either on land or underwater, and were tested in the same or opposite environment. Recall was substantially higher when encoding and testing environments matched — a crossover interaction pattern. Critically, this effect was found for recall but not recognition: recognition performance was unaffected by the environmental match.

When one cannot physically return to the original encoding environment, mental reinstatement provides a substitute: imagining oneself back in the encoding context facilitates retrieval. This principle underlies the cognitive interview technique used with eyewitnesses.

The context-dependence of memory is not simply a laboratory curiosity. Soldiers who learn tactical procedures under conditions of extreme stress recall them better when tested under similar stress. Students who study in quiet rooms and are tested in noisy examination halls show a contextual mismatch penalty. The practical lesson — study in conditions that resemble exam conditions — follows directly from the environmental-reinstatement principle. Even internal cues matter: if one prepares a presentation in a calm, reflective state and then delivers it in a state of high arousal, the state-dependent mismatch can make information feel temporarily inaccessible. Practising retrieval under performance-like conditions mitigates this.

State-Dependent Memory

State-dependent memory refers to the finding that recall is better when physiological state at testing matches encoding state — whether that state involves intoxication, arousal from exercise, or caffeine. The classic data show that those who encoded under intoxication recall better when tested in the same state than when sober, even though overall performance is highest for sober-sober encoding/retrieval. Again, this phenomenon applies primarily to recall tests, which require effortful search, rather than recognition.

Mood-Dependent Memory and Mood Congruence

Mood-dependent memory: recall is better when the emotional tone at retrieval matches the emotional tone at encoding. Ike, Macaulay, and Ryan induced happy or sad moods through music, then retested two days later in the same or opposite mood state. Mood-matching facilitated free recall regardless of whether the recalled events were positive or negative — it was the match, not the valence, that mattered.

A clinically important extension is mood congruence: people in a depressed mood more readily recall unpleasant memories, which deepens the depressed mood, creating a self-reinforcing cycle. Cognitive-behavioural therapy attempts to break this cycle by deliberately introducing new cues — activities, environments, thoughts — that activate positive memory traces.

Mood congruence also affects encoding, not only retrieval: when encoding information while in a sad mood, individuals are more likely to attend to and elaborate on negative aspects of neutral events. A self-reinforcing system operates in both directions: sad mood → attention to negative features → negative encoding → negative retrieval → sad mood. Breaking the cycle at any point — through exercise, social engagement, environmental change, or behavioural activation — disrupts the bidirectional maintenance of depressive affect.

The Neural Basis of Context Effects: Cortical Reinstatement

At the neural level, episodic memories are stored as multimodal representations — patterns of activation spanning multiple cortical modules (regions specialised for vision, audition, olfaction, motor activity, emotional tone, etc.). The hippocampus acts as a binding index: it ties together simultaneous activations across these cortical modules during encoding.

During retrieval, partial reactivation of any component — a smell, a sound, a visual cue, an emotional state — can trigger cortical reinstatement: the spreading reactivation of the other linked cortical modules, reconstructing the full memory. This is why a familiar song can suddenly transport one back to a specific time and place. Over repeated retrievals, direct connections among cortical modules strengthen, gradually reducing dependence on the hippocampus — explaining why well-rehearsed memories feel automatic and why remote memories are more resistant to hippocampal damage.

The prefrontal cortex (PFC) adds goal-directed control: it provides the schematic scaffold that frames retrieval attempts in a particular goal state or context. Deterioration of the PFC in aging and dementia reduces the quality of goal-directed retrieval without necessarily eliminating familiarity-based recognition.

Unit 8: Measuring Memory and Models of Retrieval

Recall vs. Recognition

The two most common laboratory measures of memory are recall (producing the target without it being present) and recognition (judging whether a presented item was in the study set). Recognition is generally far superior to recall: participants who recall only 38% of a studied list may recognise 96% of items when given a recognition test.

This disparity reflects different retrieval processes:

• Recall involves a two-stage process: (1) a generative search stage, during which retrieval cues are used to produce candidate items, followed by (2) a verification stage. The search stage is highly sensitive to contextual and state-dependent cues.
• Recognition can be based on a single familiarity signal: the item elicits a feeling of recognition, and the subject sets a criterion for how much familiarity is required to endorse an item as “old.”

The word frequency effect dissociates recall from recognition: high-frequency words (apple, road) are recalled better than low-frequency words (aardvark, antelope), because they are more fluent during the search. But in recognition, low-frequency words are recognised more accurately, because their rarity makes them more distinctive — a recognition signal more clearly attributable to the study list. This opposite pattern proves that the two tests access memory in fundamentally different ways.

Signal Detection Theory and Response Bias

A critical methodological concern in recognition research is response bias: individuals differ in how liberal or conservative a criterion they adopt before endorsing an item as “old.” Two people may have identical memory representations yet produce very different hit rates if one adopts a liberal criterion and the other a conservative one.

The corrected recognition score (hit rate minus false alarm rate) provides a simple adjustment. More powerful is d-prime, which standardises both hit rate and false alarm rate as z-scores before subtracting, expressing memory sensitivity in standard deviation units independent of criterion placement.

For example, suppose 15 of 45 test items are old and 30 are new. Person 1 has 10 hits and 5 false alarms: hit rate = 0.67, FA rate = 0.17, corrected score ≈ 0.49. Person 2 has 14 hits and 16 false alarms: hit rate = 0.93, FA rate = 0.53, corrected score ≈ 0.40 — despite more raw hits. Without correction, Person 2 would seem to have better memory; with correction, Person 1’s genuine discrimination is revealed.

This framework has implications for eyewitness identification: a witness may identify a lineup member not from a confident recollection but because their liberal criterion leads them to endorse anyone with some facial similarity.

SDT also makes precise predictions about how criterion shifts appear in data. A participant who adopts a more liberal criterion will produce both more hits and more false alarms simultaneously; a conservative participant will produce fewer of each. Reporting only hit rates would make the liberal participant appear to have better memory — a systematic confound that plagued memory research before SDT provided a means to disentangle sensitivity from criterion. This is why modern recognition studies always report both hit rates and false alarm rates and typically express memory sensitivity as d-prime rather than raw accuracy.

Two Processes in Retrieval: Recollection and Familiarity

Converging evidence supports two qualitatively distinct retrieval processes:

1. Recollection: A conscious, detail-rich recovery of a specific encoding episode — not just “I recognise that word” but “I remember it was the third word, right after butter.” Recollection depends on the hippocampus and prefrontal cortex and is sensitive to attentional manipulation at encoding.

2. Familiarity: A signal of prior exposure that occurs rapidly and without contextual detail — the item feels known, but without any episodic tag explaining why. Familiarity depends more on perirhinal cortex and is relatively preserved when encoding attention is divided.

Jacoby’s generation effect study demonstrates the dissociation: words generated from anagrams at encoding were better recognised on an explicit test, but words merely read were processed faster on an implicit perceptual identification test. More effortful encoding boosts explicit, recollection-based memory; perceptual repetition boosts implicit, fluency-based memory.

The Remember-Know-New (RKN) paradigm provides a subjective measure: participants indicate for each recognised item whether they genuinely remember the encoding episode (recollection), merely know the item was on the list without contextual detail (familiarity), or did not encounter it. Dividing attention at encoding selectively reduces “remember” responses without affecting “know” responses — directly demonstrating that recollection, but not familiarity, requires attentional resources at encoding.

Unit 9: Memory Failures, False Memories, and Advertising

Misattribution and the DRM Effect

A central form of memory distortion is misattribution — assigning a feeling of familiarity to the wrong source. The Deese-Roediger-McDermott (DRM) paradigm, demonstrated by Daniel Schacter (Harvard), illustrates gist-based false memory: after studying a list of words all associated with “sweet” (candy, sugar, honey, cake…), most participants confidently “recognise” the word sweet even though it never appeared. Their memory was accurate at the level of theme but wrong at the level of specific perceptual detail.

False Fame: Misattribution in the Wild

Jacoby’s false fame study showed that misattribution has powerful real-world consequences. Participants studied a list of non-famous names under full or divided attention. In a subsequent fame-judgement test, those who encoded under full attention could attribute familiarity with a non-famous name to the study episode and correctly reject it as non-famous. Those who encoded under divided attention misattributed the familiarity to celebrity: non-famous names felt famous. This misattribution effect increased under divided attention — the very condition that characterises most real-world advertising exposure.

This explains the Cadbury gorilla advertisement shown in lecture: by embedding the brand in an emotional, attention-capturing spectacle (a gorilla drumming to Phil Collins), the ad produces encoded familiarity and positive affect without conscious registration that one is being advertised to. Later at the store, consumers feel increased fluency when they see the product, interpret that fluency as genuine preference, and cannot attribute it to the commercial.

Practical implications for advertising strategy include: using emotional distractors to prevent conscious encoding of the advertising context; creating semantic associations between the product and positive concepts; relying on the mere exposure effect by which repeated, even inattentive exposure incrementally raises familiarity; and targeting populations (seniors, children) whose frontal lobe development or integrity limits their capacity for correct source attribution.

The Misinformation Effect in Depth

Elizabeth Loftus’s research on the misinformation effect provides the most direct demonstration that post-event information can alter what is retrieved. In one canonical study, participants watched a film of a car accident, then answered questions containing either “smashed into” or “hit” to describe the cars’ contact. Those who received the “smashed” version estimated higher speeds and were more likely to later report having seen broken glass — which was not present in the film. The verb choice changed both the quantitative estimate and the qualitative memory content.

Three mechanisms have been proposed to account for the misinformation effect:

1. Memory trace impairment: The original memory is overwritten or disrupted by the misleading information. If true, the original trace is gone and cannot be recovered.

2. Source misattribution: The original memory persists, but at test the participant confuses the source — they remember the misinformation, but attribute it to the original event rather than the post-event question.

3. Response bias: Participants know the correct answer but defer to the experimenter’s implicit suggestion. In this case, the memory itself is intact but reporting is influenced by social demand.

Evidence from source monitoring paradigms and neuroimaging studies suggests that source misattribution is the primary mechanism in most cases: misleading suggestions are encoded as separate memory traces alongside the original, and at retrieval the wrong trace is selected. The hippocampus encodes contextual tags that should enable source discrimination, but these tags are not always sufficient, especially under high-interference conditions.

The practical implication is stark for legal proceedings: any information conveyed to a witness between the event and formal testimony — through news coverage, conversations with other witnesses, or police questioning — can alter what the witness later sincerely reports. Sincere is the operative word: the witness is not lying. They genuinely remember what they report, and the subjective experience of remembering is identical whether the memory is accurate or contaminated. Memory contamination produces confident, sincere, wrong testimony.

Implicit Memory, Awareness, and the Failure of Attribution

Amnesic patients with hippocampal damage show preserved implicit memory (faster mirror drawing, priming effects, classical conditioning) while lacking explicit recollection. Their increased fluency on an implicit task is a genuine memory effect — but they misattribute the source. Asked why they draw better on Day 3, they say “I must be naturally skilled” rather than “I practised yesterday.” The frontal-hippocampal system required for episodic recollection is compromised, so the implicit benefit of past experience floats free of any explicit memory of having had the experience.

Unit 10: Memory and Aging

Neurological Changes in Normal Aging

The brain changes substantially across the lifespan, with consequences for memory:

• Neuronal cell loss: Particularly in the frontal lobes and hippocampus. The neurons that remain continue to function effectively; cells that die are replaced by cerebrospinal fluid, reducing overall brain mass.
• White matter degradation: The myelinated axonal fibers carrying signals between brain regions deteriorate, slowing information processing — analogous to slower internet connectivity.
• Reduced blood flow and glucose metabolism: Linked to cardiovascular health, these changes are risk factors for vascular dementia.
• Hippocampal vulnerability: The temporal lobes housing the hippocampus are structurally vulnerable to head trauma (the skull is thinnest here), anoxia from cardiac arrest, and Alzheimer’s pathology.

Preserved and Impaired Memory Functions

Not all memory functions decline equally:

Preserved: Semantic memory (world knowledge and vocabulary) remains stable or even improves with age through the continued accumulation of facts and vocabulary across a lifetime. Procedural memory for well-practised skills is largely resistant to slowing — Arthur Rubinstein continued to perform piano at a high level into old age by adapting his repertoire rather than slowing his tempo. Familiarity-based recognition is relatively spared.

Impaired: Episodic memory for specific, time-tagged personal events declines, particularly for recent events still requiring hippocampal support. Working memory manipulation — tasks requiring active transformation of information — declines due to frontal lobe deterioration. Speed of processing slows, which can be confused with memory deficit if tests do not give seniors adequate encoding time.

Source Memory and the Rahal-May-Hasher Challenge

Classic studies consistently show that seniors have poorer perceptual source memory — they remember the content of what they heard but not who said it or where it appeared. This has been attributed to frontal lobe deterioration.

However, Rahal, May, and Hasher (2002) challenged this by distinguishing perceptual source from conceptual source. In their study, trivia statements were read by a male or female voice, with one voice always telling the truth. Seniors were poorer at identifying which voice spoke a given statement (perceptual source: young 92%, old 85%), but equally good as young adults at judging whether the statement was true (conceptual source: young 88%, old 90%).

The replication with face-description pairs showed the same pattern. The implication is not a deficit per se, but rather a shift in encoding priorities: seniors preferentially encode information that is conceptually meaningful (truth, character) and de-emphasise arbitrary perceptual details (voice gender) that have less utility in daily life. This may represent an adaptive strategy when neural resources are limited.

Socio-Emotional Selectivity Theory

Socio-emotional selectivity theory (SST) proposes that motivational goals shift as perceived time remaining in life decreases. When time feels open-ended (young adulthood), goals emphasise information gathering and expanding knowledge. When time feels limited (old age, or terminal illness at any age), goals shift toward emotional regulation and preserving positive well-being.

This shift predicts — and research confirms — a positivity bias in older adults’ memory: they disproportionately attend to and recall positive information, while negative information fades more rapidly than in young adults. An attention-dot-probe paradigm shows this directly: seniors are faster to detect a target dot replacing a positive face than one replacing a negative face, indicating preferential attentional capture by positive stimuli.

Young adults with terminal diagnoses show the same positivity bias as seniors — confirming that it is perceived time, not age itself, that drives the motivational shift. The positivity bias is preserved even in moderate Alzheimer’s disease, suggesting that amygdala-mediated emotional memory is relatively spared by the disease process.

The RKN Paradigm in Young vs. Older Adults

The Remember-Know-New paradigm is a particularly illuminating tool for studying aging because it separates two processes with different neural underpinnings. In young adults, correctly recognised items are classified approximately equally as “remember” (recollection-based) and “know” (familiarity-based) across a range of study conditions. Older adults show a reliable shift: they are disproportionately likely to classify recognised items as “know” rather than “remember” — that is, they recognise that an item was presented but cannot recover the specific episode in which they encountered it.

This finding maps directly onto the neurological changes reviewed above. Recollection-based recognition depends heavily on the hippocampus and prefrontal cortex — precisely the regions most affected by age-related neuronal loss and white matter degradation. Familiarity-based recognition, dependent on perirhinal cortex, is relatively preserved. The practical consequence is that older adults are more vulnerable to misattribution errors: they recognise a name or face as familiar but cannot source that familiarity, making them more susceptible to false fame effects, misidentification errors, and persuasion by fluent misinformation.

Understanding these age-related asymmetries is clinically important. Neuropsychological evaluations of older adults that rely solely on hit rates in recognition tests will overestimate memory function by capturing preserved familiarity while missing the recollection deficit. Including source memory tests and RKN paradigms provides a more sensitive and theoretically grounded assessment.

Unit 11: Alzheimer’s Disease

Neuropathology

Alzheimer’s disease (AD) is the most prevalent dementia, characterised by two hallmark pathological features:

1. Neurofibrillary tangles: Insoluble twisted filaments inside neurons, composed of hyperphosphorylated tau protein from microtubules. When neurons die, tau tangles remain as toxic debris.

2. Beta-amyloid plaques: Extracellular deposits of amyloid-beta peptide that accumulate on the surfaces of surviving neurons, interfering with synaptic communication.

The ApoE gene produces a protein involved in clearing amyloid debris. The ApoE4 allele — particularly in homozygous form — confers elevated AD risk because it encodes a less efficient “vacuum cleaner” for amyloid debris. Unlike normal aging (in which neurons die and leave open, fluid-filled spaces), the Alzheimer’s brain shows neurons physically obstructed by extracellular debris. The shrunken, fluid-filled aging brain may actually show fewer symptoms than an Alzheimer’s brain that retains structural tissue but whose neurons are embedded in a dense matrix of plaques and tangles.

AD pathology begins in the entorhinal cortex (a gateway to the hippocampus) and hippocampus, then spreads through the temporal lobes and eventually across the neocortex. This explains why episodic memory is the first and most severely affected cognitive domain.

Cognitive Signatures of Early AD

Distinguishing early AD from normal aging is challenging because both produce episodic memory decline. The diagnostic key is whether semantic memory is affected:

• In normal aging, vocabulary and general world knowledge are stable. Occasional tip-of-the-tongue failures resolve with time or cues.
• In early AD, word-finding difficulties emerge as recurring failures: the person cannot produce a previously known word (e.g., ladle) and substitutes circumlocutions (“the thing you use to stir soup”). This reflects disruption of the lexical-semantic system housed in the temporal lobes — precisely where amyloid deposition begins.

Other early warning signs include: repeating questions that were answered moments earlier (failure to consolidate new information); difficulty with complex sequential tasks such as managing finances (requiring episodic ordering in time); and reduced awareness of deficit (in contrast to normal seniors who overestimate their memory problems).

Formal diagnosis requires memory impairment plus impairment in at least one other cognitive domain (language, visuospatial ability, executive function) that interferes with daily living.

An important diagnostic asymmetry is that people with normal aging frequently over-report memory difficulties — they notice every lapse and attribute it to declining brain function — while people with early Alzheimer’s disease frequently under-report deficits due to a neurologically based lack of insight (anosognosia). This paradox means that clinical concern is often more warranted when a patient’s family reports problems the patient dismisses, rather than when the patient themselves presents with complaints. Validated screening tools such as the Mini-Mental State Examination (MMSE) and the Montreal Cognitive Assessment (MoCA) provide standardised baselines, but longitudinal tracking — comparing a person’s current performance to their own prior baseline — is far more sensitive than single-point comparisons to population norms.

Pharmacological Treatments

The cholinergic hypothesis holds that depletion of acetylcholine (ACh) in the temporal lobe circuits most involved in memory formation is a primary driver of cognitive decline. Cholinesterase inhibitors (e.g., donepezil, rivastigmine) block the enzyme that degrades ACh in the synapse, allowing the limited ACh that remains to persist longer and more repeatedly bind postsynaptic receptors — supporting LTP. These drugs are effective at stabilising symptoms for several years, though the brain adapts by reducing ACh synthesis, necessitating dose adjustments as the disease progresses.

Antioxidants aim to slow the rate of neuronal death by reducing oxidative stress. Anti-inflammatory agents are under investigation based on evidence that neuroinflammation accelerates amyloid and tau pathology.

Cognitive Reserve and Bilingualism

Cognitive reserve — resilience in the face of brain damage, conferred by a lifetime of intellectually demanding activities — moderates the relationship between pathology and symptoms. Bilingual individuals represent a compelling case: active bilinguals spend decades managing two competing language systems, selecting words in one while suppressing the other. This sustained executive demand builds flexible cognitive strategies for accessing information through multiple routes.

Studies from Baycrest Hospital and York University show that bilinguals with the same degree of Alzheimer’s pathology as monolinguals exhibit approximately three fewer years of visible symptoms. The mechanism is thought to involve richer, more redundant cortical networks: even if one route to a memory is degraded, alternative routes remain navigable. Bilingualism does not protect against brain deterioration itself, but it protects against showing cognitive symptoms as that deterioration accumulates.

Cognitive reserve more broadly — built through formal education, occupationally demanding careers, sustained engagement with complex leisure activities, and social connectedness — similarly delays symptom onset. The clinical implication is significant: activities that challenge executive function and episodic memory (learning a new language, playing a musical instrument, taking on complex unfamiliar projects) appear to be more protective than passive cognitively comfortable activities. Neurobiological research suggests that such stimulation increases synaptic density and promotes the development of alternative neural pathways that serve as redundant memory networks when primary routes are compromised by Alzheimer’s pathology.

A final consideration for Alzheimer’s care is the documented preservation of implicit and emotional memory even when explicit memory is severely impaired. Patients who cannot recall a medical procedure may nonetheless show anxiety responses in anticipation of that procedure on subsequent visits — demonstrating conditioned emotional learning despite absent declarative memory. Caregivers and clinicians who understand this asymmetry can structure the patient’s environment to create positive emotional associations (familiar music, consistent warm interpersonal interactions) that persist even when explicit memory for the caregiver’s name and face does not. Baddeley’s textbook describes cases in which Alzheimer’s patients who reported no memory of pleasant interactions nonetheless reliably showed better mood and cooperation in sessions with familiar versus unfamiliar staff — a manifestation of preserved implicit familiarity driving emotional behaviour.

Unit 12: Research Methods in Memory Science

Why Research Methods Matter

PSYCH 390 is named for research methods, not merely for memory content. Understanding how memory researchers design studies, operationalise constructs, and interpret data is as important as knowing what the data say. Memory is notoriously susceptible to confounding: task demands, instructions, timing, and participant characteristics can each produce performance differences that mimic or mask true memory effects. Methodological literacy is what allows a scientist — or a careful reader — to distinguish a genuine discovery from an artefact.

Experimental Paradigms in Memory Research

The controlled laboratory experiment remains the gold standard for establishing causal relationships in memory science. The core logic involves manipulating one or more independent variables while measuring a dependent variable under conditions that hold other factors constant. The most common experimental designs in memory research include the following.

Study-test paradigms present material during a study phase, then assess memory during a test phase. Manipulations at study (e.g., divided vs. full attention, shallow vs. deep encoding) probe encoding. Manipulations at test (e.g., cued recall vs. free recall vs. recognition) probe retrieval processes. The delay between study and test — the retention interval — is itself a critical variable: most memory processes that matter (consolidation, sleep reactivation, interference) operate over hours to days, so laboratory experiments using very short intervals can miss important phenomena.

List-learning designs expose participants to a series of items and assess which items are recalled or recognised. The classic word list, used since Ebbinghaus, allows tight control of item properties (frequency, imageability, concreteness, length) and serial position. Randomising list order across participants controls item-level confounds.

Naturalistic paradigms trade experimental control for ecological validity by testing memory in more realistic settings — learning medication instructions, eyewitnessing a staged crime, or reading a passage of text. Godden and Baddeley’s underwater study is an exemplar: by training real divers in real environments, they ensured the context-dependent effect they observed would generalise to actual cognitive demands.

Neuropsychological case studies exploit naturally occurring brain damage to make causal inferences about the necessity of specific structures for memory. The logic is straightforward: if a patient lacks a structure and also lacks a specific memory function, the structure may be necessary for that function. But the logic has limits — lesions are rarely precise, brain damage recruits compensatory mechanisms, and a single case may not generalise. The HM case, replicated across decades of testing, and eventually in post-mortem anatomical verification, provides the strongest possible evidence.

Neuroimaging designs (fMRI, PET, EEG) pair cognitive paradigms with brain measurement to localise and characterise neural correlates of memory encoding, storage, and retrieval. A common approach is the subsequent memory paradigm: present many items, measure brain activity at encoding for each, then correlate that activity with whether the item is later remembered or forgotten. Items subsequently remembered tend to show greater activation in the hippocampus, left inferior frontal gyrus, and other regions — even though memory success is unknown at the time of scanning. This design elegantly avoids post-hoc reverse inference.

Measuring Memory: Choosing the Right Test

No single memory test captures all of memory. The researcher’s choice of test determines which process is assessed, and a failure to appreciate this leads to contradictory-seeming findings in the literature.

Free recall requires participants to produce studied items without cues. It is sensitive to organisational strategies, semantic associations, and the availability of self-generated retrieval cues. The serial position curve is cleanly observable in free recall. Its weakness: low sensitivity, especially in clinical populations, leading to floor effects.

Cued recall provides a portion of the studied episode — a category label, the first letter of a word, or the pairing partner in an associate — as a prompt. Substantial improvements over free recall confirm that much information “unavailable” to free recall is actually stored but inaccessible without adequate cues. Tulving’s distinction between availability and accessibility rests on this comparison.

Recognition presents studied and unstudied items intermixed and asks participants to endorse studied items. It is sensitive to familiarity processes, subject to response bias, and generally shows much higher performance than recall — raising the question of whether recall failure reflects retrieval difficulty rather than absence of stored information. Recognition data must always be corrected for false alarms.

Implicit memory tests — fragment completion, perceptual identification, lexical decision — assess memory without asking participants to think back to the study episode. Priming effects on these tests are typically unaffected by study orientation or attentional division, dissociating them from explicit recall and recognition. The demonstration that amnesics show normal priming on implicit tests while showing grossly impaired explicit memory was theoretically transformative.

Neuropsychological batteries (e.g., Wechsler Memory Scale, California Verbal Learning Test) standardise testing to allow comparison across individuals and patient populations. These batteries include story recall, word list learning, and visuospatial tasks, and provide normative data for different age and education groups. Understanding how individual laboratory findings map onto clinical assessment is an important translational skill.

Individual Differences and Their Sources

Memory is not a single number. A person may have excellent recognition memory but poor free recall; superb semantic memory but fragile episodic encoding; extraordinary visuospatial imagery but limited phonological span. Understanding the sources of these individual differences informs both basic science and clinical practice.

Working memory capacity (WMC), measured by complex span tasks that interleave storage with processing (e.g., operation span, reading span), is one of the strongest predictors of higher-order cognitive abilities, including fluid intelligence, reading comprehension, and multi-step arithmetic. WMC correlates with the efficiency of the central executive — specifically, the ability to maintain goals in the face of interference and to suppress task-irrelevant information.

Prior knowledge is perhaps the most powerful determinant of new learning. A chess master’s superior recall of chess positions compared to a novice arises not from superior general memory capacity but from recognising meaningful patterns. The expert chunks a board position into a small number of meaningful configurations; the novice sees an unrelated array of pieces. This chunking mechanism explains why domain experts learn new information in their domain far faster than novices, and why building foundational knowledge accelerates subsequent learning non-linearly.

Age, as discussed in Unit 10, produces asymmetric effects across memory systems. When designing studies with mixed-age samples, researchers must attend to cohort differences in education, health, and technology exposure that can confound apparent age effects. Pure cross-sectional comparisons conflate aging with cohort differences; longitudinal designs control cohort but are subject to attrition bias as healthier individuals remain in the study. The best evidence comes from cohort-sequential designs that track multiple birth cohorts simultaneously over time.

Stress and emotional arousal interact with memory in nuanced ways. Moderate arousal benefits encoding of the central, emotionally significant features of an event (Yerkes-Dodson inverted-U). The stress hormone cortisol also directly modulates hippocampal function, and chronic stress-induced cortisol elevation is neurotoxic to hippocampal neurons. Acute stress at retrieval, however, can impair access to neutral memories while relatively sparing emotional ones.

Ethical Considerations in Memory Research

Memory research touches on some of the most sensitive areas of psychology because it involves (a) implanting or altering memories, (b) studying vulnerable populations (the very young, the very old, patients with dementia or trauma), and (c) producing findings with implications for legal proceedings.

Deception and debriefing: Many memory paradigms require withholding the fact that a memory test is coming (incidental learning design) or presenting misleading information (misinformation paradigm). Ethical guidelines require that participants be fully debriefed as soon as the deception is no longer necessary, and that they understand they can withdraw data.

False memories and the law: Elizabeth Loftus’s misinformation work, and later implanted-memory research, revealed that false memories can be experimentally induced with relative ease. This finding transformed expert testimony in legal cases involving recovered memories, hypnotically refreshed testimony, and child witness testimony. Researchers and practitioners must navigate carefully between acknowledging the genuine possibility of false memories and not wholesale dismissing reports of genuine traumatic experience.

Clinical participants: Patients with Alzheimer’s disease, amnesic syndromes, or psychiatric disorders cannot always give fully informed consent. Research ethics boards scrutinise consent capacity carefully and require surrogate consent from family when appropriate. Researchers have obligations both to the individual participant and to the patient community that may benefit from research findings.

Unit 13: Applied Memory Science — Education, Law, and Medicine

Memory and Education

The most practically consequential finding in the memory literature for students is the testing effect (also called retrieval practice effect): taking a practice test produces substantially better long-term retention than re-reading the same material for an equal time. The benefit is not restricted to items that are correctly retrieved on the practice test; even failed retrieval attempts boost memory for the correct answer when feedback is provided.

The testing effect is theoretically explained by elaborative retrieval: the act of generating an answer forces reconstruction of the memory trace, strengthening its storage and creating new retrieval routes. Passively reading activates a recognition-like familiarity signal that does not require reconstructive effort and thus does not consolidate the trace as strongly.

Interleaving — mixing problems from different categories rather than practising one category at a time — feels harder and produces worse performance during practice, but substantially improves transfer to novel problems at test. The difficulty is the mechanism: interleaving prevents participants from solving problems by matching surface features to the most recently practised template and forces them to identify the relevant principle distinguishing each problem type. The feeling of difficulty is a signal that learning is occurring.

Elaborative interrogation — asking “why?” about each studied fact — activates explanatory frameworks that connect the new information to prior knowledge, producing precisely the kind of deep semantic encoding that Craik and Tulving’s levels-of-processing framework predicts will enhance retention.

Memory and the Law

The legal system relies heavily on eyewitness testimony, yet the memory science reviewed in this course consistently shows that eyewitness memory is fallible in ways that are difficult to detect by jurors. The most important findings include the following.

The misinformation effect: After witnessing an event, exposure to misleading post-event information (in media, conversations with other witnesses, or leading police questions) can alter the witness’s memory of the event. Loftus’s studies showed that changing a single word in a question — “Did you see a broken headlight?” vs. “Did you see the broken headlight?” — changed the witness’s subsequent report. The presupposition embedded in “the” implied a broken headlight existed, and many witnesses incorporated this presupposition into their memory.

Weapon focus and attention: When a weapon is present, witnesses tend to fixate on it, reducing encoding of the perpetrator’s face and other peripheral details. The weapons focus effect, while robust, is moderated by the duration of exposure and the witness’s prior familiarity with weapons.

Lineup construction: The absolute judgment strategy — assessing whether each lineup member looks like the perpetrator — leads to more false identifications than the relative judgment strategy — choosing who looks most like the perpetrator relative to the others. When the actual perpetrator is absent from the lineup, relative judgment leads witnesses to choose the most similar-looking filler rather than correctly rejecting the lineup.

The cognitive interview: Developed by Fisher and Geiselman, the cognitive interview applies memory principles to eyewitness retrieval. It includes: (a) mental reinstatement of context (recreating the encoding environment mentally); (b) encouraging the witness to report everything, including apparently unimportant details; (c) recalling events in multiple temporal orders (reverse chronological order disrupts scripted reconstruction); and (d) perspective change (imagining viewing the event from a different position). Relative to standard police interviews, cognitive interviews reliably elicit more correct information without a corresponding increase in false details.

Memory and Medicine

The patient adherence problem is one of medicine’s most persistent challenges: patients frequently forget to take medications, misremember dosages, or confuse instructions from multiple providers. The memory science reviewed in this course suggests several evidence-based strategies:

• Written vs. verbal instructions: Verbal instructions alone are poorly retained; combining written materials with verbal explanation doubles recall rates.
• Serial position management: State the most critical information both at the beginning (primacy) and the end (recency) of the appointment.
• Chunking and categorisation: Group medications by when they are taken (morning, evening, with meals) rather than presenting an undifferentiated list.
• Vivid associations: Link each medication to a distinctive cue — the medication bottle’s colour, a meal eaten alongside, or a brief memorable phrase.
• Active recall within the appointment: Ask the patient to repeat back key instructions; this both tests memory and constitutes a first retrieval practice trial.
• Attention and emotional state: Patients who receive difficult diagnoses during the same appointment are less able to process and retain subsequent medical instructions. Staging information across multiple appointments, or asking a family member to attend, compensates for the attentional narrowing caused by emotional arousal.

Unit 14: Summary — Principles for Memory and Learning

The PSYCH 390 course integrates neurobiological, cognitive, and developmental perspectives into a coherent account of how memory works and why it sometimes fails.

Architecture: Memory is not a unitary system. There are distinct systems for sensory registration, short-term maintenance, long-term declarative storage (episodic and semantic), and non-declarative memory (procedural, conditioning, priming), each depending on partially dissociable neural substrates.

Encoding: What is retained depends critically on how material is encoded. Attention, semantic depth, meaningful organisation, imagery, and distributed practice all boost retention. Rote rehearsal without active recall is one of the least effective strategies despite feeling productive.

Retrieval: Memory is reconstructive, not reproductive. Retrieval is context-sensitive (matching environmental, physiological, and mood states with encoding conditions improves recall), depends on available retrieval cues, and can be biased by response criterion. Recall and recognition engage at least partially dissociable processes — recollection and familiarity.

Individual differences: Memory performance varies systematically with attention resources (age, task load), prior knowledge (chunking, schemas), emotional state (mood congruence, positivity bias), and brain integrity (hippocampus, frontal lobes, white matter). Understanding these sources of variation is essential for scientific interpretation and practical application.

Application: The science of memory has direct implications for education, medicine, advertising, law, and public health. Clinicians can structure patient consultations to exploit serial position effects, hierarchical organisation, and imagery. Educators can design curricula around spaced retrieval. Lawyers and psychologists can design interview and lineup procedures that minimise misattribution and response bias. Understanding memory is not merely academic — it shapes the quality of decisions made every day.

Methods: The science of memory is only as reliable as the methods used to study it. Researchers must carefully choose their measures — distinguishing recall from recognition, recollection from familiarity, explicit from implicit — and must design studies that isolate the variables of interest while controlling potential confounds. The most powerful inferences come from converging evidence: behavioural experiments, neuropsychological case studies, neuroimaging, and computational modelling each contribute a different lens, and conclusions that withstand scrutiny from multiple approaches are the most trustworthy.

Ongoing questions: Despite a century of systematic research, important debates remain unresolved. The extent to which hippocampal involvement in old memories persists (standard model vs. multiple trace theory). The precise mechanism of the misinformation effect (overwriting vs. source confusion vs. response bias). The degree to which consciousness is necessary for learning and for the expression of memory. The relationship between working memory capacity and long-term memory success. And — most practically — the design of interventions that durably improve memory in educational, clinical, and legal contexts. PSYCH 390 equips students to engage critically with these debates, understand the research designs that advance the field, and apply its findings responsibly.

Unifying theme: Across all units, the same lesson recurs: memory is not a passive recording system. It is a dynamic, reconstructive process that serves the needs of the organism in its current context. Encoding is selective, shaped by attention and prior knowledge. Storage is active, shaped by consolidation and sleep. Retrieval is constructive, shaped by cues, context, state, and goals. Failures — forgetting, distortion, misattribution — are not bugs but features of a system optimised for adaptive behaviour rather than verbatim archival. Recognising this is the first step toward using memory more wisely, designing better systems around its limits, and building a science that genuinely helps people in their daily lives.

A note on Baddeley’s textbook: The course follows Baddeley’s Memory (3rd edition) as its primary text. Baddeley’s career spans the full arc of the topics covered here: from early work on STM capacity and forgetting, through the development of the working memory model that bears his name, to applied studies in industry, sport, and medicine. Reading Baddeley is thus not merely encountering a textbook but following the intellectual journey of one of the field’s most productive researchers — a model of how rigorous laboratory work, creative theorising, and sustained engagement with real-world problems can jointly advance understanding.