The previous nominal duration term gave no clear prediction gain for total reproduced duration on omitted runs in 23 adults from ds006720

We tested one previous duration term in fixed personal linear models for total reproduced duration. The test used 6,236 adjacent valid pairs from 23 adults in OpenNeuro ds006720 version 1.0.0. Whole runs were omitted from model fitting. The base model adjusted for the full current condition and task progress. Relative prediction gain was 0.60%. The 95% participant bootstrap interval was -0.08 to 1.38%. The tested previous nominal duration term gave no clear group prediction gain on omitted runs. With previous response in both models, gain was 0.07% with an interval of -0.43 to 0.66%. The package preserves source events, a fixed retrieval recipe, trial selection, all predictions, coefficients, and bootstrap draws. Numeric event checks, independent regression checks, simulation, and an exact numerical repeat passed. The evidence concerns the tested linear term and recorded labels. It does not identify a neural process.

duration reproduction · serial dependence · duration history · cross-validation · behavioral timing · prediction · working memory