Crespi Effect¶

Read operant behavior after a reward change as a discrepancy-driven transient rather than an absolute-value response: animals shifted up overshoot same-reward controls (elation) and animals shifted down undershoot them (depression), calibrated to the signed gap between delivery and a recalibrating expectation.

Version: v2 · 2026-09-19 · History
Domain-specific #: 222
Domain group: Social Sciences
Origin domain: Psychology & Behavioral Sciences
Subdomain: Learning and Reinforcement → Psychology & Behavioral Sciences

Core Idea¶

The Crespi effect names the contrast-based overshoot in operant behavior after a sudden change in reward magnitude. Animals shifted to a higher reward respond above the stable-high control (positive contrast, "elation"); animals shifted lower respond below the stable-low control (negative contrast, "depression"). Behavior is keyed not to the absolute reward but to the signed discrepancy between delivery and a prior expectation, overshooting the new steady state before recalibrating.

How would you explain it like I'm…

The Surprise Treat Bounce

If a rat always gets one treat for running, and suddenly gets lots of treats, it runs extra fast for a while, even faster than rats that always got lots. If it suddenly gets fewer, it runs extra slow. After a while it calms down and runs like the others.

Extra Happy, Extra Grumpy

Scientists noticed this with rats running down a path to get sugary treats. Rats that suddenly got many more treats than before ran even faster than rats that had always gotten many treats. Rats that suddenly got fewer treats ran even slower than rats that had always gotten few. The rats were reacting to the surprise, the difference from what they expected, not just to how many treats they got. After a few tries, they settled down to match the other rats, and the "less" surprise usually lasted longer than the "more" surprise.

Reward Shift Contrast Overshoot

The Crespi Effect is an overshoot in behavior after a sudden change in reward size. Animals switched to a bigger reward respond more strongly than animals that always had the bigger reward (positive contrast, the "elation effect"). Animals switched to a smaller reward respond more weakly than animals that always had the smaller reward (negative contrast, the "depression effect"). This shows behavior tracks the gap between the reward and what was expected, not just the reward's absolute size. The overshoot is temporary: with repeated exposure, the new amount becomes the expectation and behavior matches the no-shift group. Leo Crespi showed this in 1942 with rats running for sucrose pellets, and the negative effect was typically bigger and lasted longer than the positive one.

The Crespi Effect is a contrast effect in operant behavior following a sudden shift in reward magnitude. After an upshift, subjects respond above the level of controls trained on the higher reward all along (positive contrast, or elation effect); after a downshift, they respond below controls trained on the lower reward all along (negative contrast, or depression effect). Crespi established this in 1942 with rats running a straight runway for sucrose pellets: rats shifted from 1 pellet to 16 ran faster than the stable-16 group for several sessions before converging, and rats shifted from 16 to 1 ran slower than the stable-1 group, again transiently. The structure has four steps: a reward history sets an expectation; a sudden shift creates a signed discrepancy between expectation and delivery; that discrepancy drives a transient response amplified beyond what the new absolute magnitude sustains; and repeated exposure recalibrates the reference so the amplification decays. Behavior is thus calibrated to the discrepancy rather than to absolute reward value. The asymmetry, with negative contrast typically larger and more persistent, anticipates the loss-aversion asymmetry later formalized in prospect theory.

Scope of Application¶

The effect lives across reward-driven-behavior subfields wherever a system represents reward magnitude and forms a recalibrating expectation about it.

Animal learning — the founding case: Crespi's runway rats shifted between pellet magnitudes.
Human operant and motivational research — the same transient wherever reward shifts against a baseline.
Workplace psychology — pay-cut depression effects deeper than the new wage predicts; pay-raise elation.
Consumer behavior — discount-then-restore pricing and post-upgrade/downgrade satisfaction.
Behavioral economics — the elation/depression asymmetry as the ancestor of loss aversion.

Clarity¶

The effect makes legible a distinction absolute-value accounts blur: response to how much reward now versus response to how it compares with what was expected. Without it, the runner who outpaces a same-reward control looks like noise; with it, the gap reads as behavior keyed to the signed discrepancy. Crespi's two-group design forces this apart, holding current reward fixed while varying only the preceding history.

Manages Complexity¶

An absolute-value account leaves unexplained residue around every reward change — gaps that appear, fade, and differ by direction, looking like a tangle of separate findings. The effect compresses this by relocating the controlling variable to a signed discrepancy against a recalibrating expectation. The analyst then tracks four quantities — prior expectation, sign of shift, size of discrepancy, recalibration rate — and reads post-shift behavior off them as a settling transient, not a new steady state.

Abstract Reasoning¶

The effect licenses predicting an overshoot from the sign of the discrepancy (up predicts elation, down predicts depression, zero predicts baseline), diagnostic inference of the reference point from a control-relative anomaly, and predicting the transient's decay from the recalibration rate. It also supports predicting amplitude and asymmetry from discrepancy size and direction (negative contrast runs deeper and longer), and boundary-drawing on where contrast governs — around change points — versus where the absolute level suffices, at the recalibrated steady state.

Knowledge Transfer¶

Within reward-driven behavior the effect transfers as mechanism, since its precondition — a system representing reward and forming a recalibrating expectation — is met across animal learning, human motivation, workplace pay studies, and consumer pricing; only the reward currency and response measure change. Beyond cognitive substrates, the genuine cross-domain structure — response calibrated to a reference point with contrast overshoot — belongs to the parent primes reference_point and Contrast, not to Crespi's named machinery. The step-response overshoot of an underdamped physical system is a shape-only look-alike, not transfer.

Relationships to Other Abstractions¶

Current abstraction Crespi Effect Domain-specific

Parents (3) — outgoing typed relations

Crespi Effect is a kind of Transient Response Prime

The Crespi Effect is a Transient Response specialized to a reward-magnitude step whose history-set prediction error drives temporary control-relative over- or under-shoot.
Crespi Effect includes Reward Prediction Error Prime

Reward Prediction Error is a constituent of the Crespi Effect because the signed gap between received and expected reward drives both the transient response and recalibration of the expectation.
Crespi Effect is a decomposition of Contrast Prime

The Crespi Effect decomposes to Contrast because behavior is organized by the emphasized difference between delivered reward and its history-set expectation rather than by absolute reward alone.

Hierarchy paths (3) — routes to 2 parentless roots

Crespi Effect → is a kind of Transient Response → is a kind of Temporal Dynamics → presupposes Time

Show alternative paths (2)

Neighborhood in Abstraction Space¶

Crespi Effect sits in a moderately populated region (56^th percentile for distinctiveness): it has near-neighbors but no dense thicket of look-alikes.

Family — Named Cognitive & Behavioral Effects (32 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-10-25