BCIwiki (bciwiki.com) — According to Journal of Neural Engineering, the journal published a study by Tang Y, Corballis PM, and Hallum LE on October 5, 2026, examining emotion decoding during overt and imagined vocalisation and the contribution of facial muscle activity to EEG recordings. Addressing emotion information in brain-computer interfaces (BCIs), the study found that involuntary facial muscle activity during imagined vocalisation produces emotion-related EEG contamination patterns.
Two experiments involved 14 participants performing overt vocalisation and 21 performing imagined vocalisation across five categories: anger, happiness, neutral, sadness, and pleasure. The researchers recorded 64-channel EEG throughout, extracted time-frequency features, and used a logistic-regression classifier to assess emotion decoding accuracy. In five participants, they also recorded facial surface electromyography (sEMG) during imagined vocalisation to examine its contamination of EEG.
Single-trial EEG emotion decoding accuracy was 78.1% for overt vocalisation and 36.4% for imagined vocalisation, against a 20% chance level for the five-category task. The high-gamma band, from 50 to 100 Hz, and lateral EEG channels T7, T8, and nearby channels were important for decoding. The sEMG analysis indicated that involuntary facial muscle activity contributed to these spectral and spatial patterns during imagined vocalisation.
The researchers identified a prominent muscle-induced pattern over lateral EEG channels, termed the "railroad cross-tie pattern," with greater activation during imagined happy vocalisations. The authors conclude that involuntary facial muscle activity is associated with certain emotion categories, including happiness, and contributes to single-trial EEG emotion decoding. This does not establish that all classification information comes from muscles, but it prevents attributing the full decoding accuracy to brain-derived emotion signals.
The authors also note that the observed EMG patterns closely resembled EEG patterns commonly reported in emotion decoding studies, raising concerns about involuntary facial muscle activity in such work. This is a methodological caution arising from their findings, not proof that other studies have the same degree of contamination.