In ds005415 v1.0.0, ‘Cue Start’ marks auditory stimulus onset, while ‘Cue’ marks a later event

We examined the relationship between events.tsv and beh.tsv timing in OpenNeuro ds005415 v1.0.0 (OHSU ‘Numbers’ SEEG task), using six subjects (72, 73, 78, 79, 82, 83), one run each, with 146–198 trials per subject. When the clock was fit on visual trials (arabic, dots, written), median residuals were −0.4 to +0.4 ms and maximum absolute residuals were 2–32 ms. For auditory trials, the number of ‘Cue Start’ events matched the number of spoken and beep trials; ‘Cue Start’ was −2.0 to +0.5 ms from mapped beh.tsv stimulus_start, with 92–100% within 20 ms. The ‘Cue’ event was 1037–1303 ms (median about 1197 ms) after mapped stimulus_start for spoken trials. For beep trials, ‘Cue’ was late by 469.9–471.9 ms per beep (r ≥ 0.9998 in all six subjects), so it marks a time near the last beep rather than the first. The earlier 463 ms median residual reflects the event label’s meaning, not clock drift or noise. A weak neural check in one subject, with 40 held-out trials, was consistent with this interpretation: held-out median t was 0.98 when locked to ‘Cue Start’ and 0.18 when locked to ‘Cue’; only one of 19 channels had t > 3. One selected channel was EKG because channels.tsv marked it as SEEG and the channel type filter used that table. For stimulus onset, use ‘Cue’ for visual trials and ‘Cue Start’ for spoken and beep trials, or map beh.tsv stimulus_start using a clock fit on visual trials only. Limits: one run per subject; subjects 84–86 and 80 were not tested; the neural check used one subject and 40 held-out trials, and bipolar reference and additional subjects were not analyzed. The neural result alone is not strong evidence.

SEEG · event timing · stimulus onset · auditory trials · OpenNeuro · Numbers task