Skip to content

An inner speech Decoder reveals some mental privacy issues

    But it struggled with more complex sentences.

    Push the border

    Once the mental privacy protection was present, the team first started testing their inner speech system with quoted words. The patients were in front of the screen that showed a short sentence and had to imagine that they would say it. The performance varied and achieved 86 percent accuracy with the best -performing patient and on a limited vocabulary of 50 words, but dropped to 74 percent when the vocabulary was extended to 125,000 words.

    But when the team continued to test whether the prosthesis could decode unstructured inner speech, the limitations of the BCI became quite clear.

    The first unstructured inner speech test included looking at arrows that point up a screen up, right or left in a series. The task was to repeat that series after a short delay using a joystick. The expectation was that patients would repeat sequences such as “Up, Right, Up” in their heads to remember them – the goal was to see if the prosthesis would catch it. It did a bit, but the performance was just above chance.

    Finally, Krasa and his colleagues tried to decode more complex sentences without explicit instructions. They asked the participants to think of the name of their favorite food or to remember their favorite quote from a movie. “This didn't work,” says Krasa. “What came out of the decoder was a bit of a gig.”

    In his current state, Krasa thinks, the inner speech -neural prosthesis is a proof of concept. “We didn't think this would be possible, but we did it and that is exciting! However, the error percentages were too high for someone to use it regularly,” says Krasa. He suggested that the most important limitation could be possible in hardware – the number of electrodes that have been implanted in the brain and precision, with which we can register the signal of the neurons. Inner speech representations can also be stronger in other brain areas than in the motor cortex.

    The Krasa team is currently involved in two projects that came from the neural prosthesis of the inner speech. “The first is to ask the question [of] How much faster an inner speech BCI would be compared with an attempt at speech alternative, “says Krasa.

    Cell, 2025. DOI: 10.1016/J.Cell.2025.06.015