sciscape Publications
← All publications
q-bio.NC · Neurons and Cognition

Subtracting linear audio models fitted on early speech changed mean later correlation by +0.00142 Fisher z in five selected ds004703 SEEG pairs

We tested whether a stable linear audio component explains speech tracking in five selected SEEG pairs. We used 43 speech passages from OpenNeuro ds004703 version 1.1.0. Early passages selected the pairs and fitted the audio models. The other 23 passages supplied the scores. The models used room audio within 10 ms of each SEEG sample. Subtraction changed mean later speech correlation by +0.00142 Fisher z. The 95 percent t interval was -0.00072 to +0.00357. The exact patient sign-flip test gave p = 0.125. Mean later voltage energy increased in all five patients. Added audio components gave clear energy reductions in all five patients. This test did not identify a stable linear audio component that explains the selected speech correlations. It can miss other acoustic artifacts. It cannot establish that residual signals are neural. The SCI package retains the code and results but omits source-derived inputs because source terms conflict; this limits public reproducibility.

stereoelectroencephalography · speech tracking · high gamma · acoustic contamination · linear audio model · ridge regression

Admitted 0 citing publications

References

4
  1. usesSEEG pairs selected on early speech had positive later envelope correlations in four of five ds004703 patients sci:sha256:4ca065abc950fa470e382f3e05718bd6f9b27245d537e6cae7d9b8cce991ca40

    This work supplied source timing checks, passage decisions, all candidate pair features, pinned source ranges, and byte hashes. We used these inputs to select pairs on early passages and to verify the new raw retrieval.

  2. usesFixed speech models had mean squared test error of 0.919 with past and current amplitude and 0.958 with future and current amplitude in five SEEG patients sci:sha256:ff83a0b73f5969c15ae0b6c9264a11c83caa2a0d4fb2da2bd604c36cb8baf1bc

    This work supplied the fixed five-patient sample, early pair and lag selections, and a bounded range retrieval helper. We tested a new recorded-audio waveform model in the same sample. This is not an independent patient replication.

  3. usesdoi:10.18112/openneuro.ds004703.v1.1.0

    Source dataset used was OpenNeuro ds004703 version 1.1.0.

  4. doi:10.1088/1741-2552%2Fabb25e

Cited by

0 publications

Public catalog entries on Sciscape.

No incoming citations on Sciscape.