BCIwiki (bciwiki.com) — A collaboration involving the University of Cambridge, UK, posted a study to bioRxiv on September 25, 2026, and has not been peer reviewed. The researchers analyzed electrocorticography (ECoG) data from grid electrodes on the surface of the human ventral anterior temporal lobe (vATL) to investigate how multidimensional semantic information corresponds to neural activity at different frequencies.
The study applied representational similarity learning (RSL) to time-frequency power and phase data. It compared three hypotheses: multidimensional semantic structure is represented within a single frequency range; each semantic dimension is represented independently within a different range; or at least some semantic information emerges only when multiple frequencies are considered together, termed "transfrequency" representation.
The researchers found significant decoding of graded, multidimensional semantic information across a broad frequency range of 4-200 Hz, but not from individual frequency bands. The authors interpret this result as evidence for transfrequency coding. This interpretation concerns semantic representations revealed by joint analysis across frequencies; it does not establish that individual bands contain no semantic information.