Avatars may help control 'voices' in schizophrenia

"Scientists are examining whether computer-generated avatars can help patients with schizophrenia," The Guardian explains. The headlines report on a small study of a novel therapy technique that attempts to tackle auditory hallucinations, where people hear voices in their head.

Hearing voices is a common symptom in people with schizophrenia. In most cases, the voices are hostile, rude and often frightening, making statements like "you are worthless" or "if you don't do what I say, you are going to die".

The study involved schizophrenia patients who had not responded to medication. The patients created a computer-generated face with a voice (avatar) that they thought was similar to the hallucinated voice. They were then encouraged to confront and challenge the avatar, which was "controlled" by a therapist.

Compared with a group of patients who continued to receive standard treatment for schizophrenia (antipsychotic medication), people who had "avatar therapy" showed greater improvements. These improvements were in the frequency and intensity of their hallucinations and their beliefs about how evil and controlling the hallucinations were.

This was a very small study, but the results are encouraging and, in a few cases, striking. One man, who reported hearing the voice of the devil for more than 15 years, found the voice disappeared after just two sessions, saying the treatment had "given him his life back".

Of course, anecdotes such as this do not provide a sufficient evidence base we can use to evaluate a treatment, so a larger trial is being conducted to evaluate this approach.

Where did the story come from?

The study was carried out by researchers from University College London and the Royal Free and University College Medical School, and was funded by the National Institute of Health Research and Bridging Funding from the Camden and Islington NHS Foundation Trust.

It was published in the peer-reviewed British Journal of Psychiatry.

The research was covered appropriately by both BBC News and The Guardian.

What kind of research was this?

This was a randomised controlled trial (RCT) that tested a new therapeutic technique intended to give schizophrenia patients who experience auditory hallucinations (hear voices) control over their hallucinations. A randomised controlled trial is considered the gold standard in determining the effectiveness of a therapy.

Auditory hallucinations (voices) are often abusive, critical or commanding. When asked, people with schizophrenia consistently report that feeling helpless is the worst aspect of these hallucinations.

Standard therapy often includes advice to ignore the voices and not engage with them. However, some studies have shown that patients who talk to their "voices" tend to feel more in control.

The researchers report that it is difficult to talk to an invisible entity (voice or auditory hallucination) that is continuously abusive. This means that therapists have difficulties "steering" a conversation between the patient and the voice in a way that helps the patient.

The researchers wanted to test if putting a face to the voice may make it easier for patients to communicate with their hallucination and gain control.

This was a small proof-of-concept study, and larger trials are needed in order to validate the findings and assess the effectiveness of the intervention more precisely.

What did the research involve?

The researchers recruited 26 patients who had heard "persecutory" (abusive) voices for at least six months and continued to experience these hallucinations even after treatment with antipsychotic medication. The researchers say that around one in four people with schizophrenia fail to respond to antipsychotic drugs.

Patients were randomly assigned to two groups:

  • 14 patients created a computer-based face and voice to communicate with (an avatar)
  • 12 patients were treated as usual, which consisted of ongoing antipsychotic medication for seven weeks

The patients in the intervention group created an avatar similar to the entity they believed was talking to them, essentially giving a human face to the voice they were hearing. Custom-made voice software was used to create a voice that matched the hallucination.

The therapist was then able to use this real-time voice software to speak through the avatar, with the voice heard by the patient. This was designed to let the patient and the hallucination have a conversation. During the sessions, the therapist and patient were in separate rooms and the therapist was able to talk to the patient directly, as well as through the avatar.

Talking directly to the patient in a traditional way, the therapist encouraged the patient to stand up to their hallucination. During the course of the conversation, the therapist allowed the avatar to increasingly come under the patient's control, and shifted the character of the avatar from abusive to helpful and encouraging.

Patients were then given a recording of these sessions to listen to in order to reinforce their sense of control. Patients could complete up to six 30-minute sessions.


The researchers analysed three main outcomes, which were:

  • Frequency and disturbing qualities of hallucinations – this was measured using the hallucinations section of the Psychotic Symptom Rating Scale.
  • Patient experience related to the voices – this was measured using two subscales of the Beliefs About Voices Questionnaire: the omnipotence scale (which assesses the power that the patient perceives the voice has) and the malevolence scale (which assesses the patient's belief about the evil intentions of the voice). This questionnaire assesses the delusions that patients hold about their hallucinations.
  • Depression (common among people with schizophrenia) – this was measured using the Calgary Depression Scale.

Within each group, researchers calculated the difference in scores from the beginning of the trial to seven weeks after treatment began, and statistically compared these differences between the avatar treatment and usual care groups.

This was a small trial. However, it was reportedly powered to detect a clinically meaningful reduction in the omnipotence score. This calculation assumed a 25% dropout rate among the participants. The researchers didn't report whether the trial was powered to detect differences in the other outcome measures.

What were the basic results?

Most of the participants in the trial were unemployed (54%), had heard voices for more than 10 years (58%), and were fully compliant with their planned drug treatment (85%). There were no significant differences between the groups in the three outcome measures at the beginning of the trial.

Five patients from the avatar group dropped out of the study and were excluded from the analysis.

Compared with the usual care group, the avatar therapy group showed significantly greater improvement at the end of treatment in:

  • the frequency of their hallucinations
  • the disturbing qualities of their hallucinations
  • delusions about their hallucinations

There was no significant difference in depression scores between the groups.

The effect size of the therapy was quoted as 0.8. Effect size is a standardised way of measuring the average difference between two groups. A result of 0.8 is usually interpreted as a large effect.

How did the researchers interpret the results?

The researchers concluded that the reductions seen in the frequency and intensity of the hallucinations, and the patients' beliefs about the omnipotence and malevolence of the voices, "are clinically important considering that the patients' hallucinations had failed to respond to many years of the most effective antipsychotic drugs available".


This study suggests that avatars may have a therapeutic role in the treatment of auditory hallucinations. As the patients enrolled in the trial continued to hear voices despite medication, this new therapy could be an exciting option for a number of patients and their families.

It is important to remember, however, that this was a small proof-of-concept trial, and that the results will need to be replicated in larger – and preferably longer term – trials.

There were several limitations of the study, many of which were discussed by the authors in their published article.

Effects of therapists' intervention

Comparing the avatar therapy with treatment as usual did not allow the researchers to control for the time and attention the patient received during the sessions. It could be the case that it was the experience of being treated – in the sense of regular interaction with the therapist, rather than the treatment itself – that improved symptoms. This may be a type of placebo effect that improved patients' self-esteem, making them better equipped to deal with their voices. The researchers suggest that further studies should include an active control in order to consider this potential confounder.

Could other therapists replicate these results?

The therapy was delivered by a single therapist with an intimate knowledge of the concepts underpinning the study. This raises the question of whether it would be possible to teach the skills required to achieve similar positive results and, if so, how long the training would take.

Patients who dropped out were excluded from the results

The analyses only included those patients who completed the therapy as well as the questionnaires. This could potentially bias the results if the patients who dropped out were less likely to improve. Future trials would ideally undertake an intention-to-treat analysis (where all results of all participants, whether or not they dropped out, are considered) and attempt to account for any missing data. This is particularly relevant, as there was a high drop-out rate of just over a third in the avatar group. As the researchers discuss, it would seem that avatar therapy is not suited to all patients.

Some patients were unable to focus on the single avatar and voice due to simultaneously hearing multiple voices, while other patients were unable to complete the therapy due to the fear instilled by their voices. Part of the benefits of a pilot study is that the selection of suitable patients can be refined for a larger trial, also ensuring that all aspects of the intervention are ideal. Knowing why five people did not complete the study will be important information for future research.

The researchers also discussed an unexpected positive result – three patients stopped hearing the hallucinations altogether. One patient had heard a voice for more than 16 years, and another, who had heard a voice for more than three years, reported that, "It was as if she left the room".

Whether this amounts to a permanent recovery from auditory hallucinations in the long term is not clear. And it is also not clear how common recovery from auditory hallucinations is with other forms of therapy. It will be interesting to see if future studies are able to assess this.

A further phase III trial is being developed in an effort to independently further test the effects of the avatar system on auditory hallucinations, measuring outcomes in more detail and refining exactly what parts of the treatment work best. The results of this trial, whether positive or negative, should make for interesting reading.  

NHS Attribution