Voices are commonly experienced as communication with a personified ‘other’ with ascribed attitudes, intentionality and personality (their own ‘character’). Phenomenological work exploring voice characterisation informs a new wave of relational therapies. To date, no study has investigated the role of characterisation in behavioural engagement with voices or within psychological therapy for distressing voices.

Baseline characterisation (the degree to which the voice is an identifiable and characterful entity) of the dominant voice was rated (high, medium or low) using a newly developed coding framework, for n = 60 people prior to starting AVATAR therapy. Associations between degree of characterisation and (i) everyday behavioural engagement with voices (The Beliefs about Voices Questionnaire-Revised; n = 60); and (ii) interaction within avatar dialogue [Session 4 Time in Conversation (participant–avatar); n = 45 therapy completers] were explored.

Thirty-three per cent reported high voice characterisation, 42% medium and 25% low. There was a significant association between characterisation and behavioural engagement [H(2) = 7.65, p = 0.022, ɛ2 = 0.130] and duration of participant–avatar conversation [F(2,42) = 6.483, p = 0.004, η2 = 0.236]. High characterisation was associated with increased behavioural engagement compared with medium (p = 0.004, r = 0.34; moderate effect) and low (p = 0.027, r = 0.25; small−moderate effect) with a similar pattern observed for the avatar dialogue [high v. medium: p = 0.008, Hedges’ g = 1.02 (large effect); high v. low: p = 0.023, Hedges' g = 1.03 (large effect)]. No differences were observed between medium and low characterisation.

Complex voice characterisation is associated with how individuals interact with their voice(s) in and out of therapy. Clinical implications and future directions for AVATAR therapy and other relational therapies are discussed.