Preregistration

Preregistration-ready. Not preregistered. Nothing has been submitted to a registry, and no registration URL exists.

The full document is research/preregistration.md. Its commitments, in brief:

Question

Are continuous orientation judgements biased by recent stimulus history after accounting for the current stimulus, response history, cardinal attraction, regression to the running mean, and motor persistence?

Primary hypothesis and outcome

H1: the group-mean lag-1 stimulus-history amplitude from model M3adj_full, fitted per observer, differs from zero. Two-sided; direction is not assumed. Test: observer-level bootstrap, 10,000 resamples, α = 0.05. The amplitude is the peak of the fitted derivative-of-Gaussian tuning curve in degrees.

Secondary hypotheses

H2 tuned versus linear bias; H3 monotone lag decay; H4 larger amplitude under high sensory uncertainty (paired within-observer difference); H5 joint identifiability of stimulus and response amplitudes (a design property, tested by simulation, passed); H6 no false positives from history-free generators (a pipeline property, tested by simulation).

Sample and stopping rule

Target 30 desktop participants after exclusions, 600 main trials each; recruit until 30 usable sessions or 45 collected, whichever first. No interim analysis of the effect. Power curves for a range of amplitudes are on the power page; no single effect size is assumed.

Exclusions

Frozen as exclusion-policy-1.0.0; see methods. Applied before any effect is estimated. No exclusion depends on the effect.

Interpretation rules, fixed in advance

If the response-only model beats the stimulus models on held-out likelihood, the effect is reported as decisional or motor, not perceptual. If no history model beats the no-history baseline, a null result is reported and the analysis is not re-specified. An amplitude that is significant in the naive model and not in the adjusted model is reported as confounded.

Confirmatory versus exploratory

Confirmatory: H1–H4 and the model comparison. Exploratory, labelled as such wherever reported: reaction-time relationships, early versus late blocks, split-half reliability, delay-condition effects, lags 4 and 5.