Concealed beneath the implausibly insentient nature of the intervention implied by Leff et al's study title ^{Reference Leff, Williams, Huckvale, Arbuthnot and Leff1} is in fact a highly relational therapeutic approach for voice hearers of potentially Copernican significance! An example of the kind of paradigm shift in both research and clinical practice recently advocated in the British Journal of Psychiatry (e.g. Bracken et al ^{Reference Bracken, Thomas, Timimi, Asen, Behr and Beuster2} ).

Although only a ‘proof of concept’ study, it is predicated on a very different understanding of psychopathology than conventionally argued for in the pages of this Journal. Not only does the study shun conventional diagnosis in favour of a ‘symptom group’, as Tyrer points out in the issue's editorial coda, but it revives the concept of psychotic symptoms as relational phenomena – both in terms of aetiology and intervention – that our group has recently further argued for. ^{Reference Bracken, Thomas, Timimi, Asen, Behr and Beuster2}

Although a large-scale phase III study is clearly warranted, the early impression of an evidently useful shift in the framing of psychosis potentially opens up readers of this Journal to more serious consideration of a wider range of relationally oriented aetiological factors and therapies already advocated for psychosis and psychotic symptoms in several ‘lower impact’ journals – which as Kingdon points out in his related editorial ^{Reference Kingdon3} – have historically proved to be the principle hotbed of past game changers in psychiatric practice.

Although the Journal has itself recently published several articles acknowledging childhood maltreatment to be significant risk factors for psychosis possibly meditated by changes in the hypothalamic-pituitary-adrenal axis and downstream effects on dopamine systems, the idea that hallucinatory phenomena may themselves represent ‘echoes’ of past abuse brings us closer to dissociative concepts of such phenomena, which by definition points towards relational solutions. Indeed, outside the pages of this Journal the once confident distinction between dissociative phenomena and psychosis has been challenged on various counts, including the following.

1. (a) Experimental studies which have shown that psychological measures of dissociation and psychosis are highly correlated and do not have convincing differential construct validity. ^{Reference Moskowitz, Barker-Collo and Ellson4}

2. (b) Historical analysis of changing diagnostic trends, demonstrating a waning in the popularity of multiple personality disorder at the time that the diagnosis of schizophrenia began to gain ascendance is argued to be no coincidence. ^{Reference Rosenbaum5} That childhood abuse is now suggested by some studies to have a ‘dose-dependent’ relationship with later risk of psychotic symptom development, in particular hallucinations, ^{Reference Read, Os, Morrison and Ross6} also weakens the basis for any presumed aetiological distinction between the two.

3. (c) Psychological modelling of how child maltreatment and trauma may give rise to psychotic symptoms (including negative symptoms). Presumed differences between traumatic flashbacks and ‘hallucinations’ may be based more on whether insight into a link between trauma and symptom is acknowledged by the patient (and psychiatrist). ^{Reference Read, Os, Morrison and Ross6} This becomes harder still when the hallucination is symbolic rather than simply echoic or thematic.

If such a model is correct, then we can begin to take more seriously the claims of such relational therapies as the open dialogue family therapy model for early psychosis in Finland, which claims to have reduced the transformation of new-onset psychosis to chronic schizophrenia to a remarkable degree. ^{Reference Seikkula, Alakare and Aaltonen7} We might also take seriously the ideas of relating therapy for voices and even the more radical, direct voice dialogue advocated by some. ^{Reference Corstens, Longden and May8} The implications for wider practice are also substantial – after all, the difference between voice elimination/repression and integration/transformation cannot be overstated, although clearly some patients are likely to still favour a ‘sealing off’ recovery style.

Julian Leff's team and the editorial board of the British Journal of Psychiatry are to be congratulated for the publication of this paper. Greater insight into how the therapist learns to convincingly embody the patient's persecutory voice, through the avatar, would however be welcome.