In 22 SEEG patients, errors did not measurably alter the next response or confidence; HGA showed no distributed previous-trial accuracy signal

We tested preregistered post-error effects in the ds006253 random-dot task. Behavioural trial tables from 22 SEEG patients were used to test whether an error on trial t−1 changed response time or confidence on trial t. Models adjusted for current coherence and trial number; the patient was the unit for group tests. The mean previous-error coefficient was −0.0031 for log response time (95% bootstrap CI −0.021 to +0.015; 10/22 positive; sign-flip p = 0.74; Holm p = 0.998) and +0.0044 for confidence (95% CI −0.0030 to +0.0122; 12/22 positive; p = 0.29; Holm p = 0.857). For 70–150 Hz high-gamma activity (HGA), stimulus-window features from 8 patients and 714 bipolar channels yielded 32 channels with p < 0.05, against a Freedman–Lane null mean of 34.5 (p = 0.50; Holm p = 0.998). The secondary pre-movement test was also null (37/714 channels; p = 0.39). Coherence after errors was not lower than after correct trials (mean difference −0.0059; 7/22 patients positive), so the staircase did not make the next trial easier after errors. In this task and sample, the three preregistered tests did not detect post-error slowing, a confidence change, or a distributed previous-accuracy HGA signal. The neural analysis used window means, included only 8 patients with HGA, and its channel-count test can miss focal effects; the patients had epilepsy, limiting generalization.

post-error adjustment · SEEG · random-dot task · confidence · high-gamma activity · HGA · previous-trial accuracy · OpenNeuro ds006253 · MetaRDK