Precision-Stepping Gain Changes: Retention, Transfer and Coordinate Error

Abstract

To distinguish a retained change in the dependence of lateral foot placement on body position from improved placement precision, reduced body-state variation and coordinate error. A nine-coefficient model relates foot placement to lateral position, velocity, target location, forward speed and visual preview. The primary contrast compares the early-preview position gain at immediate, delayed and unfamiliar-order probes with the mean of two pre-practice sessions. Repeat-measurement calibration and Bartlett serial-correlation-adjusted standard errors determine change and equivalence decisions. The Monte Carlo study included eight conditions with 2,000 replications each at 320 steps per session, plus 1,000-replication tests at 80, 160 and 640 steps. A position-gain decrease from 0.80 to 0.40 cm per reference unit at all three probes produced joint detection in 89.2% of replications at 320 steps. Detection was 0.8%, 18.1% and 100.0% at 80, 160 and 640 steps. Joint detection was zero under unchanged gains, precision improvement alone, immediate-only change, change restricted to the familiar order and correctly scaled range contraction. Session-specific scaling and shared stance-reference error produced false joint detections in 54.2% and 6.7%. Nominal 95% delayed-contrast coverage was 93.5--94.5% in the six conditions with intact measurement. Under the specified linear distributions, repeated gain contrasts distinguished retained change from reduced placement error. Fixed coordinate scales and independent checks of stance-reference error were necessary to prevent false gain changes. The findings concern computational gain estimation; participant retention and clinical benefit remain untested.