Under Student mean assumptions, a fixed dictionary term had gains below 1% of baseline error in seven iEEG and 28 fMRI listeners

We tested whether a count of similar sound forms adds brain response prediction during story listening. A public English dictionary set the count. Separate public studies supplied intracranial EEG (iEEG) and BOLD fMRI. Two patients and six MRI adults supplied development. A local code lock preceded tests in seven other patients and 28 other MRI adults. Each MRI adult heard two stories. The control model had sound, word timing, length, duration, frequency, initial phone, pause and recent repetition terms. The added fit kept the control prediction fixed. One error point is 1% of baseline squared prediction error. Mean gains were -0.006148, -0.302538 and -0.191958 points. Their 95% person Student intervals were [-0.020582, 0.008287], [-0.576452, -0.028625] and [-0.570419, 0.186503]. The shared positive test failed, with Holm p=0.984169. The shared mean-below-one test passed, with Holm p=0.000000632. These mean tests assume independent people and a suitable Student model. Combined person and shared time-block upper limits were 0.0039, 0.2195 and 0.3120 points. These checks held fits fixed. The limit applies to this predictor and fitting rule. It does not bound neural competition or all possible readers. Clinical coverage, an early auditory MRI region, fixed stories, canonical pronunciations and measurement filters limit scope. The package preserves code, permitted compact data, source pins, every outcome, figures and repeat steps. Independent loss reading checked 336 records. A clean repeat matched 355 numerical analysis files. No positive cross-modal neural effect or common cause was established.

human neuroscience · intracranial EEG · fMRI · speech · phonological neighborhood · prediction · negative result