Removing the true sound file kept best scores unchanged in 33 of 38 six second queries from ten SEEG listeners with ±2 second template search

We tested whether a best match score could detect a sound file missing from a template library. The test reused ten fixed SEEG readers and 38 six second listening queries from OpenNeuro ds004703, version 1.1.0. Each query was scored with its true file present and absent. With a search from -2 to +2 seconds, 33 scores stayed exactly equal. Equal person weighting gave 87.5% unchanged scores. Each condition had equal probability. A proof limits any detector using only the best score to 56.25% balanced accuracy in this paired sample. The top two score gap gave a bound of 62.5%. Across 1,000 fixed library settings, mean best score presence AUC was 0.551 with timing search. The settings reused the same neural queries. Their spread describes library sensitivity. This is a test of missing candidates in recorded listening data with supplied source times. It does not establish a communication decoder or a writing method. The SCI contribution retains figures, the proof, MIT code, source version and range receipts, and full result scores. Source-derived feature and target inputs, source metadata, and small source-sample audits are omitted because the ds004703 records list CC0 metadata while the versioned source README restricts commercial use and participant identification. Result tables and figures remain. Their inputs cannot be replayed offline from this SCI package; the retained extraction scripts and external source link describe how to rebuild them. Fresh source extraction and independent arithmetic checks passed in the archived local run.

stereoelectroencephalography · brain-computer interface · sound envelope · candidate rejection · paired removal test · presence detection · template search