Trial count correction changed the mean late theta phase load contrast from +0.022 to −0.006 in 12 hippocampal SEEG patients

We tested theta phase consistency during the last memory delay second in public SEEG. We compared eight letters with four letters. OpenNeuro ds004752 1.0.1 supplied 12 patients, two sessions each, and 44 hippocampal bipolar pairs. The main contrast used 690 correct trials with a zero source artifact flag. The mean squared intertrial phase consistency contrast was +0.02205. All 12 patient contrasts were positive. After trial count correction with pairwise phase consistency, the mean was -0.00599. Its 95% t interval was [-0.02233, +0.01034]. The exact patient sign flip test gave Holm p=0.56738. This test assumes independent patients and sign symmetry under the null. The interval assumes approximately normal patient contrasts. The planned test did not detect a load difference. It does not establish equal phase scores. A higher frequency check gave a positive interaction interval. A known identity accounts for the score change. Independent uniform trial phases predict a +0.02701 uncorrected contrast from the observed counts. This model does not show that all recorded phases were uniform. Adaptive task load limits causal interpretation. The package contains code, coefficients, raw examples, source versions and hashes, patient results, all planned checks, a proof, figures, and repeat instructions. Fresh downloads checked all 2,334 source windows. Independent signal and phase calculations passed. A separate run reproduced all eight numerical result files exactly. This is a new analysis of an earlier sample.

pairwise phase consistency · intertrial phase consistency · theta phase · stereoelectroencephalography · verbal working memory · trial count bias