Carryover Effect¶

The validity threat in crossover and within-subject designs where residual influence from an earlier treatment persists into a later measurement window, biasing the contrast — its magnitude set by the unit's relaxation time against the inter-treatment gap.

Version: v3 · 2026-09-28 · History · 1 correction
Domain-specific #: 138
Domain group: Formal Sciences
Origin domain: Experimental Design & Statistics

Core Idea¶

The carryover effect is the validity threat in crossover and within-subject designs in which residual influence from an earlier treatment persists into the measurement window of a later one, so the response attributed to the later condition is contaminated. It requires two features: the unit carries an internal state the earlier treatment modified that has not returned to baseline (a drug concentration, a practised skill, fatigue, an updated belief), and the design assumes each window samples an independent realisation. The bias is proportional to the surviving residual.

How would you explain it like I'm…

The Lemon-Then-Apple Problem

If you bite a lemon and then taste an apple right away, the apple might taste extra sweet because the lemon taste is still in your mouth. Then you can't tell how the apple really tastes. That leftover from the first thing messing up the second is the carryover effect. Waiting and rinsing your mouth in between helps.

Leftovers From the Last Treatment

In some experiments, the same person tries more than one treatment, one after another, like two medicines or two kinds of practice. A carryover effect happens when the first treatment is still affecting the person when the second one is measured, like medicine still in the blood, a skill they just practiced, or tiredness. Then the results for the second treatment get mixed up with leftovers from the first. Scientists fight this by waiting long enough between treatments (a washout period), by giving different people the treatments in different orders so the leftovers balance out, or by using math to estimate and remove the leftover.

Residual Treatment Contamination

The carryover effect is a threat to validity in crossover or within-subject experiments, where each participant receives several treatments in sequence. It happens when an earlier treatment changes something inside the participant, like a drug level, a learned skill, fatigue, or an expectation, and that change has not faded by the time the next treatment is measured. The measured result for the later treatment is then biased by the leftover from the earlier one, breaking the design's assumption that each period gives a clean, independent measurement. How big the bias is depends on how slowly the effect fades compared with the gap between treatments. Researchers use washout periods (for drugs, often about five half-lives), counterbalanced orders such as Williams designs, and statistical models with a carryover term to handle it.

The carryover effect is residual influence from an earlier treatment persisting into the measurement window of a later one in crossover, repeated-measures, or within-subject designs, contaminating the response attributed to the later condition. It requires two ingredients: the unit has an internal state that the earlier treatment modifies and that has not returned to baseline (uncleared drug concentration, learned skill, fatigue, receptor sensitization, updated beliefs), and the design assumes each period gives an independent, clean realization of the treatment-outcome relation. The resulting bias in the treatment contrast scales with how much of the prior effect persists, set by the state's relaxation time relative to the inter-treatment gap; for example, a drug with a 36-hour half-life after only a 24-hour washout leaves substantial residual concentration. Standard countermeasures are washout periods (for pharmacokinetic carryover often about five half-lives), counterbalancing with a Williams design (a succession-balanced Latin square) so residues distribute symmetrically and cancel in contrasts, and explicit modeling via a lagged prior-treatment term in a mixed-effects model. It matters in crossover clinical trials, where regulatory guidance such as ICH E9 treats it as a major concern, in psychology experiments (practice, fatigue, sensitization, demand characteristics), in sequential A/B tests on the same users, and in sensory panels requiring palate cleansing and order balancing.

Scope of Application¶

The carryover effect lives across the within-subject and crossover contexts of statistics and experimental-design practice, restaging identically across settings that differ only in which internal state carries the residue.

Clinical crossover trials — the home; residual drug biasing the next period (ICH E9, Senn).
Pharmacology and PK/PD — half-life-driven washout, residual receptor occupancy.
Behavioural and psychological experiments — practice, fatigue, sensitisation, demand characteristics.
Educational research — skill or motivation retained from an earlier method.
Sequential A/B tests and sensory panels — learning effects and palate residue.

Clarity¶

Naming carryover isolates a threat casual analysis dissolves into its neighbours: it is a deterministic residue inside the unit, not noise, measurement error, external confounding, or regression to the mean. It forces three otherwise-implicit specifications — which internal state carries the residue, over what timescale it relaxes, and whether the later window opens before it has relaxed — converting a vague worry into a quantitative comparison.

Manages Complexity¶

A heterogeneous list of ways an earlier condition poisons a later one compresses to one diagnosis: an internal state modified by the earlier treatment that has not relaxed before the later window opens. The methodologist tracks three things and reads the verdict off comparing two numbers — a relaxation time-constant against the inter-treatment gap — and the remedy menu becomes a decidable response to that one parameter.

Abstract Reasoning¶

Carryover licenses state-space reasoning about the unit: a diagnostic move locating a biased contrast in a carried internal state, a predictive move comparing relaxation time to inter-treatment gap, three interventionist moves each tuned to that parameter (washout, counterbalancing, explicit modelling), and a boundary-drawing move checking the two structural preconditions and the symmetry of the residue.

Knowledge Transfer¶

Within experimental-design practice carryover transfers as full mechanism — the state-relaxation diagnosis, the timescale inequality, the scoping, and the washout/counterbalancing/modelling menu restage identically across substrates differing only in what state carries the residue. Beyond it, the concept is the design profession's name for Hysteresis appearing in the unit of a crossover design; the state-from-history primitive travels via Hysteresis (magnetic remanence, control-system memory, ecological legacy), which inherits none of the design-validity machinery.

Relationships to Other Abstractions¶

Current abstraction Carryover Effect Domain-specific

Parents (1) — outgoing typed relations

Carryover Effect is a kind of Hysteresis Prime

Carryover is hysteresis specialized to an experimental unit whose prior treatment state persists into a later measurement window.

Hierarchy paths (4) — routes to 4 parentless roots

Carryover Effect → is a kind of Hysteresis → is a kind of Path Dependence → presupposes Dependency

Show alternative paths (3)

Neighborhood in Abstraction Space¶

Carryover Effect sits in a sparse region of the domain-specific corpus (67^th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Clinical Trial Design & Drug Safety (23 abstractions)

Nearest neighbors

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