A Password for the Inner Voice

In November 2012, the neuroscientist Adrian Owen of the University of Western Ontario announced that a 39-year-old man named Scott Routley had answered a question. Routley had been injured in a car crash twelve years earlier and had been assessed ever since as being in a vegetative state, awake but unaware. Owen’s team placed him in an fMRI scanner and gave him a code. To answer yes, he was to imagine playing tennis, which activates one region of the brain. To answer no, he was to imagine walking through his house, which activates another. Asked whether he was in pain, Routley answered no. “Scott has been able to show he has a conscious, thinking mind,” Owen told the BBC, as reported in The Scientist. “We have scanned him several times and his pattern of brain activity shows he is clearly choosing to answer our questions.”

The case is usually described as mind-reading. It is closer to the opposite. The problem of other minds begins from the fact that no one observes another person’s experience directly; everyone infers it from what the other person does. Ludwig Wittgenstein compressed the point into a line in his Philosophical Investigations (1953): “An ‘inner process’ stands in need of outward criteria.” For twelve years Routley had produced almost no outward criteria, and the inference ran accordingly: no behavior, no mind. The scanner did not let Owen bypass that inference. It supplied a new outward criterion, and the criterion was an act. Imagining tennis on request is something a person does, deliberately, in answer to a question. Owen did not look into Routley’s mind. Routley found a new way to act in public.

Thirteen years later, a team at Stanford reported a case that ran the other way. In a study published in Cell on August 14, 2025, Erin Kunz, Frank Willett, and colleagues recorded from electrode arrays in the motor cortex of four people whose speech had been impaired by ALS or stroke. Earlier devices of this kind decoded attempted speech, in which a participant tries to move the muscles of speaking. This study found that the same area also carried inner speech, words only imagined, in a pattern closely resembling attempted speech but weaker. During tasks that asked participants to recall a sequence or to count, the decoder picked up some of the numbers and words they were thinking without having been asked to say them.

The researchers treated that as a problem. A device that broadcasts whatever passes through a user’s inner voice, Stanford’s announcement warned, could let private thoughts “leak out.” One remedy they tested locked inner-speech decoding behind a password: the system stayed silent until the user imagined a rare phrase, in the study “Chitty Chitty Bang Bang,” which it recognized more than 98 percent of the time.

Where the signal came from matters philosophically. What the decoder read was not thought in general but inner speech in motor cortex, the part of the brain that plans movement, and the study found that a “motor-intent” dimension was what separated imagined words from attempted ones. The philosopher John Campbell had argued for something very like this in a 1999 paper: that thinking is in part a motor process, produced the way a movement is produced, with its own sense of being the one doing it. Campbell’s evidence was schizophrenic “thought insertion,” in which patients report thoughts that occur in their own minds but feel as if someone else is thinking them. The thoughts are still theirs in one sense, since they occur in their minds. What has failed is the sense of authorship. On Campbell’s account, and on Kunz’s data, inner speech looks less like a sealed inner object than like a quieter version of an action, already halfway to being said.

Philosophers of mind now commonly separate these two senses, which Shaun Gallagher, in a 2000 paper, called the sense of ownership (this thought is happening in me) and the sense of agency (I am the one thinking it). Ordinary minds produce plenty of thoughts that are owned but not authored. In a 2014 study led by the psychologist Adam Radomsky, 93.6 percent of 777 university students in thirteen countries reported at least one unwanted intrusive thought in the previous three months, from nagging doubts to sudden images of harming someone. Harry Frankfurt’s account of the person, which an earlier essay applied to De Quincey’s opium habit, gives the reason this matters. On Frankfurt’s view a person is not identical to every desire that arises in him, only to the ones he identifies with. The same holds for thoughts. The step between thinking something and saying it is one of the places where a person does that identifying.

Set side by side, the two cases sort out what is actually at stake. Routley’s mind was entered, by any reasonable standard, and the result was not a violation; it was the first evidence in twelve years that he was a person who could be asked. What made it acceptable was that every answer passed through an act he authored. The Stanford counting was harmless in content, but it was taken before any such act, from a process that was still on its way to becoming speech. The familiar objection to mind-reading pictures the private mind as a sealed room, an “inner citadel” in the phrase Pierre Hadot borrowed from Marcus Aurelius, and locates the harm in anyone getting inside. These two cases locate it elsewhere. A machine that waits for an act, like Owen’s scanner or a password, is a new way of speaking. A machine that takes inner speech before it is authored attributes to a person words he has not yet made his own.

Routley’s case itself remained contested. His family believed he could respond with small movements of his thumb and eyebrows, while some neurologists who examined him found no such awareness at the bedside. The scanner settled what bedside observation could not, but only because it waited for him to answer.