Arousal and the false memory: how a salient moment reshapes what you remember

A whitepaper on arousal-biased encoding, the New Look tradition in perception, and how the same machinery that makes emotional memories vivid also manufactures confident errors.

The claim

A salient event does not simply mark time. It reshapes memory for the moments around it. It decides what gets stored strongly and what gets dropped, it sharpens the center of attention while blurring the periphery, and under the right conditions it inflates confidence in things that never happened. Memory is not a recording that emotion makes louder. It is a reconstruction that emotion actively edits, starting at the door of perception and continuing through storage.

This paper connects three literatures that are usually kept apart: the neurobiology of arousal and consolidation, the New Look tradition showing that perception itself is shaped by expectation, and the false-memory research on how gist overwrites detail. They describe one continuous process.

Arousal turns up the gain, selectively

Begin with the circuit. As the companion whitepaper on reconsolidation describes, noradrenaline in the basolateral amygdala gates how strongly an arousing event is consolidated (the Roozendaal and McGaugh rule). Arousal is the physical dial on trace strength. This is why you remember where you were during a shock and forget an ordinary Tuesday.

But the gain is not turned up uniformly. The key correction to a naive "arousal boosts memory" story is that arousal is selective, and it trades the center against the periphery.

Arousal-biased competition. Mather and Sutherland's framework holds that arousal amplifies the processing of whatever is already high-priority in a scene and suppresses whatever is low-priority. It does not add a flat bonus to everything. It sharpens the winners and starves the losers. So a high-arousal moment produces better memory for the central, attended object and worse memory for the peripheral details in the same instant.

Cue-utilization and weapon focus. This is the older Easterbrook observation in modern dress: as arousal rises, the range of cues a person encodes narrows. The classic demonstration is weapon focus, where witnesses to an armed crime remember the weapon in vivid detail and the assailant's face poorly. Their memory did not fail. It was narrowed, on purpose, by arousal.

The practical upshot is a prediction with a specific shape. A siren-level trigger should raise memory for the central item it lands on and lower memory for peripheral items in the same window, rather than lifting both. That interaction, arousal times centrality, is the load-bearing prediction, because it cannot be faked by a participant who knows a memory test is coming.

Perception is already reconstructive: the New Look

If you think editing only happens in storage, the New Look tradition of the late 1940s shows it starts earlier, at perception itself. Two experiments from that program are built as interactive tools in this project.

Perception of incongruity (Bruner and Postman, 1949). Flash a playing card briefly, then mask it. Some cards are normal, some are tricks with the color wrong for the suit: a red spade, a black heart. At brief exposures people confidently misreport the trick cards, seeing a normal red heart where a black heart was shown, or perceiving a muddy purple that reconciles the conflict. Expectation fills in what the eyes did not have time to resolve. The mind reports what it predicted, not what arrived. The /perception tool runs this directly.

Perceptual defense (McGinnies, 1949). Flash a word for a fraction of a second and lengthen the exposure until the person can read it. Threatening or taboo words have higher recognition thresholds than neutral ones: the perceptual system is slower to admit them. The interpretation is debated, but the phenomenon endures and the lesson is the same. Perception is not a passive readout of the world. It is shaped by need, value and expectation before it ever reaches memory. The /defense tool runs this one.

The through-line from New Look to arousal-biased competition is direct. The same expectation-and-priority machinery that fills in a masked card, or defends against a threatening word, is the machinery that, under arousal, decides which items in a scene win the competition for a strong trace. Editing does not begin at storage. It begins at seeing, and storage continues it.

Gist wins: how confident false memories are built

Now the third literature. When detail is weak but meaning is strong, memory reconstructs from the gist, and the reconstruction can feel exactly as real as a true memory.

Gist over verbatim. People retain the meaning of an event far better than its surface particulars. Asked later, they reconstruct plausible particulars consistent with the gist. If the gist says "a kitchen," memory may confidently supply a stove that was never there. The error is not random. It is meaning-consistent, which is what makes it convincing.

Gist-consistent lures. This is why a good false-memory probe is not a wild fabrication but a small, plausible swap: the book was open when it was closed, the coat was blue when it was green. A lure that fits the gist draws confident false recognition. The DRM paradigm (Roediger and McDermott, 1995) is the laboratory-pure version: study a list of words all related to "sleep" (bed, rest, tired, night) without ever presenting "sleep," and people later recall and recognize "sleep" with high confidence. The gist, sleep, was never shown, yet it is remembered.

Arousal inflates the false alarm. Put the two together. Arousal strengthens the gist and narrows encoding of the periphery, which is exactly the recipe for confident gist-consistent errors. Near a high-arousal trigger, false alarms to gist-consistent lures should rise. The emotional charge that makes a memory feel true is not evidence that it is accurate. Vividness and accuracy are separate variables, and arousal drives them apart.

The instrument: turning this into data

The trigger and encoding experiment in this project (see docs/trigger-encoding-protocol.md and the /experiment console) is built to measure these effects with the demand characteristics designed out. The logic is worth stating because it is what separates a measurement from a suggestion.

Participants sit in a furnished room while a scripted sequence of events unfolds, some delivered by a live actor and some automated. At scripted moments salient auditory triggers fire (a notification ping, a phone ring, a doorbell, a dog bark, a siren) at controlled offsets before, during or after target events. Everything is stamped against one monotonic session clock, so every trigger-to-event offset is exactly computable rather than estimated.

Two design choices make the results hard to fake.

A cover task suppresses strategic encoding. Participants are told they are evaluating the comfort and acoustics of the space, not that memory is being tested, so they do not deliberately rehearse. The manipulation is debriefed fully afterward.

The real dependent variables are errors, not hits. A participant who knows it is a memory test can still inflate their hit rate by trying hard, but they cannot fake a binding error (misremembering which sound co-occurred with which event) or a confidently-false lure. Loading the analysis onto binding accuracy and gist-lure false alarms is what makes the arousal-times-centrality and arousal-inflates-false-memory predictions survive demand.

The counterbalancing closes the last loophole. Which trigger lands on which event rotates across participants through a Latin square, so any memory effect is attributable to the trigger and its offset, not to a particular event being intrinsically memorable. And the log is append-only and exported whole, so nothing is quietly summarized or dropped before analysis. The test phase doubles as the reveal: each participant is shown their own hits, misses and confident errors, which is both the payoff and the point. The distortion is not described to them. It is demonstrated on their own memory.

How this closes the loop with reconsolidation

The three papers form a single argument. Arousal, through the basolateral-amygdala noradrenaline gate, decides at encoding what gets written strongly enough to matter later. Reconsolidation is the same lever seen from the other end: the moment a strong trace is recalled, it becomes labile, and what re-stabilizes can be edited. The dynamical-systems picture supplies the reason a small, well-timed input can be decisive rather than washing out.

Put together: emotion writes the trace strongly and selectively, imperfectly and with confident errors baked in. Recall reopens that trace. And the labile window is where a memory that was written with distortion can be guided to settle differently. The false memory is not a bug at the edge of the system. It is evidence that the trace is reconstructive, malleable material, which is exactly the property the reconsolidation work depends on.

What to take away

• Arousal does not boost memory uniformly. It amplifies the central and high-priority and suppresses the peripheral (arousal-biased competition, weapon focus).
• Editing starts at perception. The New Look showed perception is shaped by expectation and value before memory is involved (the trick-card and perceptual-defense experiments).
• Confident false memories are built from gist. Meaning-consistent lures draw confident false recognition, and arousal makes this worse.
• Vividness is not accuracy. The emotional charge that makes a memory feel true is a separate variable from whether it is true.
• These effects are measurable with demand designed out, by loading the analysis onto binding errors and gist-lure false alarms rather than hit rates.

Sources

• Mather and Sutherland (2011), arousal-biased competition in perception and memory. Perspectives on Psychological Science.
• Easterbrook (1959), the effect of emotion on cue utilization and the organization of behavior. Psychological Review.
• Bruner and Postman (1949), on the perception of incongruity: a paradigm. Journal of Personality. (Basis of the /perception tool.)
• McGinnies (1949), emotionality and perceptual defense. Psychological Review. (Basis of the /defense tool.)
• Roediger and McDermott (1995), creating false memories: remembering words not presented in lists. Journal of Experimental Psychology: Learning, Memory, and Cognition.
• Roozendaal and McGaugh, basolateral-amygdala noradrenergic gating of consolidation (corpus meta-analysis: noradrenergic-signaling-in-the-basolateral-amygdala). See also whitepaper 01.
• Project instrument: docs/trigger-encoding-protocol.md and the /experiment operator console.