Statistics

There were three questions of interest related to subjects’ performance on the non-manipulated trials: (1) Did subjects improve their performance on Day 1? (2) Did they retain what they learned on Day 2? (3) Did they continue to improve on Day 2? The dependent variables included the absolute angular error, movement time, and antagonist co-activation. To answer Question 1, a paired t-test was performed between the first and last 10 trials on Day 1. To answer Question 2, a paired t-test was performed between the last 10 trials on Day 1 and the first 10 trials on Day 2. Finally, to answer Question 3, which had six time points, a one-way repeated measures analysis of variance (ANOVA) was performed on the average performance on the first and last 10 trials of the initial 60-trial Day 2 adaptation period and the last 10 trials of each of the four pre-manipulation Day 2 baseline blocks. Post hoc comparisons were made using Bonferonni corrections.

Paired t-tests were performed to determine whether the biceps and triceps CVs, noise factors (σ), and smoothing factors (α) were different, whether the slopes of linear fits to the log-transformed linear envelope magnitude vs. STD data were different, and whether the coefficients of determination (R²) for these linear fits was different between the muscles.

The statistical analysis for the noise manipulations tested the main hypotheses that: (1) humans adapt to increased motor noise by increasing antagonistic co-activation, which reduces noise-induced performance decrements and (2) humans adapt to decreased motor noise by decreasing antagonistic co-activation, which improves performance. Performance was again quantified in terms of error, movement time, and co-activation. However, these tests were performed on relative performance measures, i.e., performance with the noise manipulations relative to performance on the most recent non-manipulated trials. To this end, the average performance for the last 10 trials in each of the four manipulation blocks was subtracted from the average of the last 10 trials in each of the immediately preceding baseline blocks. Analyzing the relative change controlled for any continued performance improvements present on the non-manipulated trials. To determine whether the noise manipulations had an effect on performance (i.e., was the change with a manipulation different from zero?), one-sample t-tests were performed on the difference measures (manipulation minus preceding baseline) separately for each noise manipulation. To assess whether subjects adapted to the noise manipulations, paired t-tests were performed comparing the first and second manipulation difference measures (i.e., was the change in error for the first noise increase different from the change in error for the second noise increase?).