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.

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.

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) — more general patterns this builds on

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 → Hysteresis → Path Dependence → Dependency

Show alternative paths (3)

Neighborhood in Abstraction Space¶

Carryover Effect sits in a sparse region of the domain-specific corpus (88^th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Unclustered & Miscellaneous (309 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-07-12