American neurophysiologists have, for the first time, documented that the hippocampus of patients under general anesthesia can distinguish between parts of speech and anticipate upcoming words in a narrative. This discovery challenges the traditional view that consciousness is a prerequisite for complex cognitive processing.
A research team from Baylor College of Medicine conducted experiments on epilepsy patients undergoing surgery while under general anesthesia. Using high-precision Neuropixels probes—never before employed to record hippocampal activity under such conditions—scientists tracked the responses of hundreds of individual neurons in this region, which is typically associated with memory.
The resulting data revealed that neural circuits not only maintain sensitivity to external stimuli but also demonstrate a capacity for learning in an unconscious state. These findings were published in the journal Nature.
In the first series of trials, subjects were exposed to repeated sound sequences, occasionally interspersed with irregular tones. Hippocampal neurons consistently reacted to these anomalies, and over time, the brain became increasingly accurate at identifying deviations from the pattern—a clear sign of neuroplasticity occurring without conscious involvement. In the second experiment, short narrative stories were played to the patients, and the pattern of electrical activity showed a distinct separation in the processing of nouns, verbs, and adjectives. Moreover, the signal characteristics made it possible to detect anticipation of words that had not yet been spoken.
Dr. Samir Sheth, a professor of neurosurgery at Baylor, noted that the brain appears to predict the progression of a narrative without any conscious perception. His colleague, Benjamin Hayden, added that such “predictive coding” is typically associated with wakefulness and attention, yet here it occurs in the absence of consciousness. The authors also drew a parallel to the operation of large language models, which generate text by predicting the next token, highlighting similarities in information processing algorithms between biological and artificial neural networks.
The practical significance of this discovery lies in the potential for developing speech neuroprosthetics: hippocampal signals could be harnessed to restore communication in individuals with severe brain damage. Additionally, the study challenges a key dogma in neuroscience by suggesting that consciousness is not a necessary condition for complex computations but may instead be merely a byproduct of cognitive network activity.
About the Author
