Two previous interval terms did not establish lower later judgment loss in 20 ds004200 v1.0.1 records under fixed logistic models

Two fixed terms for the previous heard interval did not establish lower loss for later yes/no duration judgments. We tested 20 released records from OpenNeuro ds004200 version 1.0.1. The base logistic model used the current target and interval. The added model also used the previous interval code and its interaction with the current interval code. Each fit used early blocks from the other 19 people. The test used 1,080 later trials. Mean log loss increased from 0.489416 to 0.489864 nats per trial. The gain was -0.000448 nats. Its 95% whole-person refit bootstrap interval was -0.006611 to 0.002194. All six fixed sensitivity-check intervals also included zero. The rules were saved before fitting. Independent checks covered all 278 point fits and 120 bootstrap fits. A clean offline rebuild reproduced the numerical outputs and figures. The package preserves source files, hashes, code, trial history, predictions, coefficients, bootstrap weights, results, and instructions. The result assumes independent released IDs and fixed trial sequences. It does not prove absence of history effects or a neural mechanism.

duration judgment · temporal perception · trial history · logistic regression · cross-person prediction · bootstrap