A model for words after annotated gaps of 0.25 to 2 s failed a shared prediction test in six iEEG patients and twenty fMRI adults

We tested a response term for words after annotated gaps of 0.25 to 2 s. The baseline controlled sound level, nonlinear sound functions and ordinary word timing. Separate public studies supplied intracranial high-gamma amplitude and regional BOLD fMRI. Three patients and five listeners supplied development. A local final code lock preceded tests in six other patients and twenty other listeners. Mean added test R2 was 0.0001494 in iEEG and -0.0013959 in fMRI. Their 95% person t intervals were [-1.57229e-05, 0.000314461] and [-0.0027261, -6.57023e-05]. Scores were positive in 6/6 and 4/20 people. The shared rule failed, with conjunction p=0.9796626. Whole-person and common story-block uncertainty, every negative result and fixed descriptive checks remain. Independent solvers of the human data agreed within 1.39e-14 R2 per output, below the 1e-6 tolerance fixed before confirmation. Pre-confirmation numerical failures and repairs remain. A later audit found seven fMRI events that missed a prose-plan exclusion. The code-fixed and stricter post-result outcomes remain. This is a new endpoint in reused cohorts, not a new acquisition replication. Source selection, offline cleaning, clinical sampling, separate stories and measurement filters limit scope. The test does not establish absent pause coding, a common neural process or causality. The package gives permitted numerical inputs, event timings, code, source versions and terms, all outcomes, figures and repeat instructions.

auditory encoding · word timing · intracranial EEG · fMRI · speech pauses · prediction · human neuroscience