A voice is not a stranger

This piece might be perceived as a critique of AVATAR therapy, but it is rather a concern about what the therapy tells the affected person about their voice and their condition.

AVATAR therapy is a treatment for distressing voices in psychosis and schizophrenia. The person and a therapist give the voice a computer-generated face, and the therapist speaks through it. The idea is that people who have felt powerless before a hostile voice can face it and answer it back, with the therapist at their side, until the voice concedes.¹ In a trial with 345 participants (opens in a new tab), AVATAR therapy eased the distress caused by voices more than usual care alone.²

The benefit is genuine; the drawbacks lie in the method. Is it constructive? Is the person mastering the voice, or is the therapist achieving it through make-believe? The therapist speaks for the avatar, and in the original protocol the avatar is made to cease its abuse after the first two or three sessions, after which its expression is changed to a smile.³ How durable are the gains? In both large trials, the advantage on the main measure had faded within about six months.^1,2 And how does the experience alter the way a person perceives the voice once the sessions are over?

Where did the idea of giving a voice a stranger’s face come from?

The psychiatrist Julian Leff devised AVATAR therapy in London in 2008.¹ His starting point was helplessness: asked about the worst aspect of hearing voices, people invariably named it. People who manage to hold a dialogue with their voices feel more in control, and Leff cited the Maastricht group of Marius Romme, who encouraged such dialogue.⁴ The obstacle, as he saw it, was that an invisible entity offers none of the cues of conversation, such as facial expression or turn-taking, and repeats the same abusive phrases whatever the person says. Some therapists already asked patients to address an empty chair as though it were their persecutor. Leff’s answer was to make the persecutor visible.³

To build the avatar, the person and the therapist choose a computer-generated face and adjust the therapist’s voice until it resembles the voice the person hears. In sessions, the therapist then speaks as the voice.² A Danish virtual-reality version, tested in the Challenge trial, lets people encounter the avatar in three dimensions through a headset.⁵

The problem with AVATAR therapy: an old misreading in a new form

Whether on a screen or in a headset, a voice that arises in the person’s own mind acquires someone else’s face, and is spoken by someone else. This echoes the oldest misreading of schizophrenia. Eugen Bleuler coined the word in 1908, from the Greek for “mind” and “to split”. He meant that thinking and feeling lose their coherence, not that the mind divides into two people. Within a few decades, the word had come to mean a “split personality”, an idea that still feeds the fear of people with the diagnosis.⁶ An avatar risks conveying the same idea in pictures: there is someone else in your head, and this is what they look like.

In fairness, AVATAR therapy also links the voice back to the person. The 12-session version adds a second phase that connects the voice to the person’s life story, including past trauma.² In Leff’s original method, once the avatar has turned friendly, it tells the person that “the voices say what you think about yourself”.³ The message, then, is that the voice belongs to the person; the messenger is a face that does not.

One voice among many

Most people with psychosis hear more than one voice, 3 to 4 on average. In the first large AVATAR trial, almost 8 in 10 participants heard more than one.¹ In 100 people assessed with the Maastricht method, most heard between two and five voices.⁷

Voices are not uniform. In a survey of people who hear voices, about 1 in 4 of those with a psychiatric diagnosis described voices that were positive or useful, and about 4 in 10 described voices that were abusive or violent.⁸ Since the 1990s, clinicians have classed voices as benevolent or malevolent. In Chadwick and Birchwood’s early study of 26 patients, every one of them experienced their voices as all-powerful and all-knowing. Voices believed to be malevolent provoked fear and resistance; those believed to be benevolent were courted. Yet the belief did not always follow what the voice said: in almost a third of cases, it contradicted the content.⁹

AVATAR therapy embodies only one voice: the dominant one, or the one the person would most like to be rid of.³ That voice alone is given a face, and the face belongs to an adversary.

Why the explanation matters

I think of voices differently: as part of the person, not as strangers. In 100 people assessed with the Maastricht method, most of them diagnosed with a psychotic disorder, therapists could formulate the emotional conflict a voice embodied, such as shame, anger or low self-worth, in 94 cases. In 78, they could identify whom or what the voice represented: a family member, a past abuser, or a disowned aspect of the self.⁷ This is the premise of the Maastricht approach (opens in a new tab), and Talking With Voices (opens in a new tab), which grew from it, addresses voices as parts of the person.

If the response to a voice follows the belief held about it, then the explanation a person is given is itself an intervention. An avatar delivers such an explanation non-verbally. By projecting the voice onto another face, it tells the person that the voice belongs to someone else.

If people come to understand a voice as their own mind at work, it may lose some of its omnipotence. The voice may be an attempt at protection or at communication, a reflection of what the person feels, or an outdated loop that replays someone else’s beliefs. It becomes something to understand, not someone to defeat.

AVATAR therapy should remain available to those it helps. But before a voice is given a face, it is worth asking what effect that face will have on the person who hears the voice.