Sound timing and visual reports in a public behavioral dataset

This paper reanalyses the public Experiment 2b data from Williams and colleagues (2022), which already reported an interaction between sound timing and visual reports. The contribution is an explicit participant-level signed-error contrast, a robustness analysis and a reproducibility audit. Of 95 recorded participants, 90 satisfy the implemented primary cleaning rules and all six timing-by-sound cells. The mean contrast between category-related sounds is 1.3962 morph units during audiovisual overlap and 0.5503 during the response phase; their paired difference is 0.8459 units, with a descriptive 95% bootstrap interval of 0.1305 to 1.6783 and one-sided sign-flip p=0.01571. A sensitivity analysis retaining all 92 sufficiently complete participants and using cell medians gives a difference of 0.7092 units, interval 0.1087 to 1.3967, p=0.01600. The smaller response-only pull is retained. The near-zero response-only visual response-time benefit is measured before that sound begins. A full replay reproduced the original local result JSON, and an independent implementation recovered all 12 participant contrast vectors and nominal test decisions. This is a post hoc reanalysis of the same published data, not a new discovery or independent replication. The behavioral contrasts do not uniquely identify perceptual change rather than timing-dependent memory, decision or report processes. This is an author preprint and has not been peer reviewed. The supplement contains author reanalysis code and aggregate findings, with instructions for fetching the original data under their repository terms; original participant files, task code and stimuli are not redistributed. Source study: https://doi.org/10.1177/09567976221121348. Source data: https://osf.io/85kwv/.