When people talk, various areas of their brain must activate. However, the function of these areas can be severely impaired after damage to the nervous system. For example, due to amyotrophic lateral sclerosis (or Charcot disease), the muscles used for speech may be completely paralyzed. In other cases, for example after a stroke, the areas of the brain responsible for speech may be affected: this is called aphasia. In many of these cases, however, patients’ ability to imagine words and sentences remains partially functional.
Deciphering inner speech is therefore of great interest to neuroscience researchers. But the task is far from easy, as Timothy Prua, a scientist in the Department of Basic Neurology at the UNIGE School of Medicine, explains: “There have been several studies on deciphering spoken speech, but far fewer on deciphering imagined speech. This is because bound neural signals are weak and variable compared to explicit speech. Therefore, they are difficult to decipher with learning algorithms.”
When a person speaks out loud, they make sounds that are produced at certain points in time. In this way, researchers can relate these tangible elements to the brain regions involved. In the case of imagined speech, the process is much more complicated. Scientists have no obvious information about the sequence and tempo of words or sentences formulated internally by an individual. The areas involved in the brain are also less numerous and less active.
To perceive the neural signals of this particular type of speech, the UNIGE team used a group of 13 hospitalized American patients. They collected data using electrodes implanted directly into the patients’ brains to assess their epileptic disorders. “We asked these people to say the words and then present them. Each time, we analyzed several frequency ranges of brain activity that are known to be associated with speech,” explains Anne-Liz Giraud, professor in the Department of Fundamental Neurology at the UNIGE Faculty of Medicine and recently appointed director of the Institut de l’Audition in Paris.
The researchers observed fluctuations in several types of frequencies produced by different regions of the brain when these patients spoke verbally or to themselves. “First of all, oscillations called theta (4-8 Hz), which correspond to the average rhythm of saying syllables. Then gamma frequencies (25-35 Hz), observed in areas of the brain where speech sounds are produced. Third, beta waves (12-18 Hz), associated with the cognitively more efficient areas that are requested, for example, for anticipating and predicting the development of a conversation. Finally, the high frequencies (80-150 Hz) that are observed when a person speaks,” explains Pierre Megevin, associate professor in the Department of Clinical Neurology at the UNIGE School of Medicine and associate physician at HUG.
Through these observations, the scientists were able to show that low frequencies and the relationship between certain frequencies (particularly beta and gamma) contain important information for decoding imagined speech. Their study also shows that the temporal cortex is an important area for possible decoding of inner speech. Located on the left lateral side of the brain, this specific area of the brain is involved in processing information related to hearing and memory, but it also contains part of Wernicke’s area responsible for the perception of words and language symbols.
These results are a major advance in reconstructing speech from neural activity. “But we are still a long way from being able to decipher imagined language,” the research team concludes.
Read more:
Archaeologists discover long lost temple of Hercules from 9th century B.C.
Check out selfies taken by Chinese orbiter of Mars
Airless tires have emerged: they look like animal cell structures
