صدای بیگانه و گفتگوی درونی: به سوی یک اکانت تکاملی توهم شنوایی کلامی

The phenomenon of auditory verbal hallucinations (AVHs) is one of the most intriguing features of the psychiatric literature. Two alternative models of the development of AVHs in both normal and psychotic populations are proposed. In the disruption to internalisation (DI) model, AVHs result from a disruption to the normal processes of internalisation of inner speech. In the re-expansion (RE) model, AVHs result when normal inner speech is re-expanded into inner dialogue under conditions of stress and cognitive challenge. Both models draw on Vygotsky's (The Collected Works Of L.S. Vygotsky, New York, Plenum Press, 1987) ideas about the development of inner speech. On this view, normal inner speech is considerably abbreviated relative to external speech, and also undergoes some important semantic transformations. In both the DI and RE models, AVHs arise when the subject's inner speech involves inappropriately expanded inner dialogue, leading the subject to experience the voices in the dialogue as alien. The two models may prove useful in explaining some of the social-developmental evidence surrounding the phenomenon, and also make a number of testable predictions which are suggested as priorities for future research.

The phenomenon of auditory verbal hallucinations (AVHs), where subjects report the perception of speech in the absence of any external stimulus, has proved to be one of the most consistently puzzling in the psychiatric literature. AVHs are seen as a primary positive symptom of schizophrenia, and were classified by Schneider (1959) as first-rank symptoms. More recently, a number of developments in psychiatry, psychology and neuroscience have contributed to a renewed interest in AVHs. Firstly, the move towards a non-Kraepelinian, symptom-based approach to the psychopathology of schizophrenia (e.g. Bentall (1990) and Bentall (2003); Frith, 1992) has led to a focus on AVHs as a phenomenon worthy of study in its own right, rather than as a feature of a larger explicandum, namely schizophrenia. Secondly, a growing consensus that hallucinations, verbal and otherwise, are not necessarily signs of pathology (Feelgood & Rantzen, 1994; Johns & van Os, 2001; Pearson et al., 2001; Reese, 1971; Tien, 1991) has contributed to a view of AVHs as potentially part of normal as well as abnormal experience. Thirdly, advances in brain-imaging techniques have meant that researchers have been able to study AVHs in vivo, at the very moment that the subject is experiencing them (e.g. Shergill, Brammer, Williams, Murray, & McGuire, 2000). Finally, a degree of success in the psychological modelling of AVHs (e.g. Bentall, 1990; Frith, 1992; Hoffman, 1986) has led to a new optimism that we might be within reach of a psychological-level account of these experiences. This resurgence of interest in the psychological modelling of AVHs comes at a time of growing interest in the linguistic and verbal qualities of normal as well as abnormal consciousness. Ever since James (1890) made his observations on the ongoing interplay of verbal images that characterises human thought, the ‘stream of consciousness’ has been considered a primarily verbal phenomenon. Not only does this view have a considerable amount of intuitive appeal, it has also proved useful for a growing number of psychologists and philosophers (e.g. Carruthers, 2002; Clark, 1998; Dennett, 1997; Kinsbourne, 2000) in explaining how the functional properties of natural language can augment the pre-existing cognitive capacities of the individual. Another factor in the resurgence of interest in inner speech is the growing popularity of Vygotsky's (e.g. 1934/1987) ideas on the phenomenon. On this view, verbal thought develops through the gradual internalisation of external forms of dialogue, with the result that mature inner speech is irreducibly dialogic in character (Fernyhough, 1996; Wertsch, 1991). My aim in this paper is to situate the study of AVHs within the wider context of the study of inner speech, with particular focus on the normal and abnormal development of inner speech. To date, there have been no developmental-psychological accounts of AVHs. This is partly because the phenomenon has traditionally been seen as part of a broader disorder, schizophrenia, which has proved notoriously resistant to developmental explanation. Consequently, AVHs have come to scholars’ attention as a full-blown, adult symptom requiring explanation in terms of organic deficits and resultant cognitive disturbance. The present view, in contrast, is that any satisfactory theory of AVHs must incorporate an account of how these experiences develop, and particularly their relation to normally developing inner speech. My starting point is the Vygotskian view that inner speech is irreducibly dialogic. I take a non-Kraepelinian, dimensional approach to psychosis in general, on which AVHs are viewed as part of normal as well as abnormal experience. Two alternative developmental models, essentially based on Vygotskian ideas about inner speech, are presented. Any developmental account must of course be answerable to developmental data, and, in evaluating the present models with respect to competing accounts of AVHs, I attempt to show how these models can make sense of a growing body of evidence from both normal and abnormal development.

The present paper has considered two models of the development of AVHs arising from Vygotsky's ideas about the ontogenesis of inner speech. On this view, AVHs are seen as an unusual and occasionally pathological form of dialogic inner speech. On the DI model, AVHs develop in psychotic individuals when there is a disturbance to the normal processes of internalisation of dialogue. Such a disturbance may arise at several points in development. For example, AVHs can develop when there is a disruption to typical patterns of dialogic interaction in infancy and early childhood. Other possible causes of disruption/disturbance to internalisation are speech difficulties which prevent the child from engaging in normal external and private speech, and the failure of syntactic abbreviation processes in internalisation, possibly as a result of the child's diminished experience of external dialogue and private speech. In contrast, the RE model holds that AVHs arise, not from any fundamental disturbance to internalisation, but from problems with the process by which condensed inner dialogue is occasionally transformed back into expanded inner dialogue (Level 3), or dialogic private speech (Level 2). On this view, the sudden re-expansion of dialogue under challenging cognitive conditions results in patients experiencing the resulting inner speech as alien. Under very stressful conditions, something similar may happen to previously unaffected individuals. A strength of the DI and RE models is that they generate clear testable predictions about the developmental course of AVHs. The DI model would be supported by evidence that early experiences that might lead to a disruption of internalisation are associated with the later development of AVHs. To date, only one study (Carlson, 1998) has attempted to relate direct measures of infant–caregiver interaction to later psychopathology, although in this case the presence of AVHs was not directly assessed. Further careful longitudinal research in this area is needed if we are to gain a better understanding of the development of the phenomenon. The RE model makes several testable predictions that would allow it to be distinguished from the DI model. It predicts that AVH-experiencers will experience normal condensed inner speech, but not normal expanded inner dialogue; that such experience will be associated with conditions of stress and cognitive challenge; and that very stressful conditions may lead to previously unaffected individuals experiencing AVHs. Gathering further data on these issues—through, for example, interview-based assessments of the quality of inner speech among patients, healthy voice-hearers and non-voice-hearers—would seem to be a priority for future research. One advantage of both the DI and RE models is that they allow a solution to the paradox that voices in AVHs are often acknowledged simultaneously to be both alien and of the self. This is because, in internalising dialogic exchanges, the individual takes on the voice, and thus the semiotically manifested perspective, of the partner in the dialogue. Normal human thought is thus an ongoing interplay between differing perspectives on reality (Fernyhough, 1996). In the pathological case, the voices in internal dialogue are experienced by the subject in an incompletely abbreviated form, as an inappropriately expanded inner dialogue. As these experiences occur in the absence of external speech input, the voices are experienced as alien. Both models thus allow for continuity in the experience of AVHs between psychotic and normal populations. In the case of the DI model, continuity is ensured by the fact that internalisation is not an all-or-nothing process, and thus that its disruption may be a matter of degree. In very seriously affected individuals, internalisation may be held up at Level 1 or 2. Less seriously affected individuals may experience AVHs because of problems with the syntactic abbreviation that would usually occur in the transition between Levels 3 and 4. On the RE model, continuity is possible because of the proposed relation between re-expansion and cognitive challenge. In normal individuals, conditions of extreme stress or cognitive challenge can result in re-expansion of inner dialogue, which may be further re-externalised as dialogic private speech. In very extreme cases of stress and trauma, the re-expanded utterances in inner dialogue may be experienced as alien. In comparison with current theories of AVHs, the DI and RE models enjoy a number of further advantages. Firstly, they avoid getting caught up on the issue of the ‘intendedness’ of inner speech utterances, as discussed by Stephens and Graham (2000) and Akins and Dennett (1986) in relation to Hoffman's (1986) model. Because inner speech, on these accounts, is fundamentally different to external speech in its syntactic structure and semantic properties, it does not require the same level of discourse planning as external speech. Intended inner speech utterances can occur, of course, but these happen at the level of expanded inner dialogue (Level 3), where discourse-planning demands are presumably similar to those involved in external speech. Secondly, current theories of AVHs are hard pressed to explain the social-developmental evidence surrounding the phenomenon. How, for example, might Hoffman's (1986) model of AVHs explain the evidence for disturbed attachment patterns in individuals who later become psychotic? It is similarly difficult to see how such findings can be accommodated within a neuropsychological theory such as Frith's (1992). That is not to say that there is no common ground between the DI and RE accounts and those theories of AVHs that attribute them to a primary neurological disorder, such as a frontal monitoring deficit (Frith, 1992). Later I consider how such brain systems might be necessary for the internalisation of dialogue, and thus how the DI and RE models can be assimilated into the broader picture of neurological damage in schizophrenia. Thirdly, the models presented here are less reliant than many of their competitors on a simple confusion between internal and external sources of data, with all the dangers of regress and circularity that come with that form of explanation. Rather, the Vygotskian approach allows us to rethink the dichotomies of self/other and inner/outer as they relate to inner speech and verbal thought. On this account, the inner is always at least partly outer: normal inner speech is shot through with alternative perspectives on reality. At the same time, individual thought is seen as a distributed, essentially social process: as much a collaboration between individuals as a solo endeavour. This view thus has much in common with recent attempts to do justice to the ‘extended’ nature of individual mental functioning (e.g. Clark & Chalmers, 1998; Dennett, 1997; Wertsch, 1991). It also means that the Vygotskian approach is equally appropriate to situations where alien voices are recognised to come from within the boundaries of the self, and to those where they are attributed to an entirely external source (see footnote 1). To take the Vygotskian view that the internal is always partly external does not, however, explain why some individuals make systematic errors in explaining the provenance of their own thoughts. For most of us, the fact that our inner speech is shot through with other voices does not lead to our perception of these voices as alien. To put it another way, it is difficult to see how the hearing of voices could be construed as anything other than a problem in monitoring the different sources of experiences. What the Vygotskian models presented here can add to, say, Bentall (1990) and Bentall (2003) source-monitoring account is their ability to explain why external attributions are made for internal events. The key lies in the nature of inner speech, and particularly the expanded form of inner dialogue that Vygotsky hinted at in his discussion of the internalisation of linguistic exchanges. That is, certain types of internal event (expanded inner dialogues) have a form that makes them more likely to be attributed to an external source. It may be that individuals with relatively less accurate source monitoring capacities will, when experiencing expanded inner dialogue under conditions of stress or cognitive challenge, be more likely than their healthy counterparts to attribute elements of that dialogue to external sources. Such a view would entail seeing the Vygotskian and source-monitoring accounts as potentially complementary rather than mutually exclusive. Despite these apparent strengths of the DI and RE accounts, a number of gaps in the theory remain to be filled. Firstly, we need more specificity in the four-level model of inner speech development outlined here. What conditions must be met before movement between levels can occur? Under precisely what circumstances does normal subjects’ inner speech revert from Level 4 to Level 3 or Level 2? What demands do these varieties of inner speech place on the various components of the cognitive system, such as working memory? How important is the presence or absence of external speech input in determining whether a subject will experience AVHs? What is the role of previous trauma in the development of AVHs, perhaps in lowering the threshold of stress or cognitive challenge needed to trigger the re-expansion of inner dialogue? In addition to these issues specific to the DI and RE models, the Vygotskian concept of internalisation will continue to benefit from further clarification, elaboration, and empirical testing. Although some important steps have recently been taken in this regard (e.g. Tomasello, Kruger, & Ratner, 1993), much work remains to be done in specifying the cognitive mechanisms underlying internalisation. Another problem facing the DI and RE models is the need to explain why the psychiatric consequences of disrupted internalisation only become apparent in adulthood. This problem is particularly acute for the DI model, which would seem to hold that disruptions to internalisation leading to AVHs in adulthood should be detectable at earlier stages of development. It is also problematic for the RE model, which would need to explain why the re-expansion of inner dialogue should often have a specific onset in adulthood. My approach to this problem is to note the gaps in our knowledge relating to the developmental course of AVHs, and the continuity in AVHs between normal and clinical populations. For example, it may be that ceasing to view AVHs as necessarily pathological might lead to us seeing evidence for them much earlier in development. There is already evidence that a later proclivity to develop AVHs can be detected in adolescence (e.g. Escher, Romme, Buiks, Delespaul, & van Os, 2002). There is also some suggestive recent evidence that normal children can experience what we would otherwise describe as AVHs, particularly in the context of their dealings with imaginary companions. For example, Pearson et al. (2001) reported that exposing 9- to 11-year-old children to ambiguous perceptual stimuli led to hallucination-like experiences, especially when subjects reported imaginary companions. Much more work remains to be done to link children's experiences of AVHs with the developmental course of private and inner speech. Similarly, future research into developmental precursors of AVHs in particular, and psychosis in general, would do well to consider how individual differences in the emergence of private speech, and its subsequent internalisation into inner speech, are related to later susceptibility to hallucinations. It is important to note, however, that what may be pathological in adulthood may not necessarily be pathological in childhood. Indeed, the Vygotskian approach to AVHs implies that, at the stage when children are internalising external dialogue and engaging in conversations with imaginary companions, experiences that would otherwise be classified as AVHs might be the norm rather than the exception. These experiences should, perhaps, only be seen as pathological if they continue into adulthood.2 The DI model holds that AVHs may be a developmental consequence of failures in this process of internalisation, while the RE model holds that AVHs result from abnormalities in the process whereby internalisation is temporarily reversed. It seems likely that, through learning much more than we currently know about the mechanisms of internalisation, we might gain some important insights into what happens when this crucial developmental process goes wrong. Another challenge for the present model is to account for the overwhelming evidence that the main disorder associated with AVHs, schizophrenia, is a biological disorder with a strong genetic component. How can we square any developmental psychological account of AVHs with such evidence? Firstly, we can point out that these Vygotskian models of AVHs are not models of schizophrenia. Rather, they attempt to explain a symptom which occurs frequently in schizophrenia, but also in bipolar disorder (Potash et al., 2001), and in normal adult and child populations. A second point is that the neurological systems whose breakdown is implicated in schizophrenia may well have a role to play in the process of internalisation. For example, internalising dialogue must minimally involve the integration of one's own contribution to the external dialogue with the contribution of one's interlocutor, thus drawing on the same kind of monitoring capacities whose breakdown has been implicated in schizophrenia (e.g. Frith, 1992). It may also be that the cognitive resources that are furthest stretched in conditions of stress and cognitive challenge—the executive functions—are responsible for controlling the re-expansion and re-externalisation of inner speech, and thus that the disturbance to these processes in subjects with AVHs has a fundamentally executive cause. One further question arises from the Vygotskian approach to AVHs. The foregoing discussion has focused primarily on Vygotsky's ideas about syntactic abbreviation in internalisation. What of the semantic abbreviation processes described by Vygotsky? One might argue that the processes of semantic transformation described by Vygotsky actually amount to a fairly good description of disorded speech in schizophrenia. If this characterisation of ‘schizophrenese’ is accurate, we would have to conclude that the external language of patients with speech disorder undergoes the same semantic transformations as normal subjects’ inner speech. We could perhaps take this as evidence that the inner speech of such individuals is similarly transformed. What seems certain is that there is something inappropriate in the way these features are not confined, as they are in normal subjects, to inner speech. If schizophrenic speech is like normal inner speech spoken out loud, it is interesting to consider what cognitive features of the disorder might be responsible. It seems plausible that this inappropriate externalisation of inner speech is related to the mentalising and perspective-taking deficits reported in such individuals (see Bentall, Corcoran, Howard, Blackwood, & Kinderman, 2001, for a review). Two final points need to be made. Firstly, contra Slade & Bentall (1988), the present account has nothing to say about hallucinations in other modalities. Secondly, we should consider whether the DI model has any implications for therapy. I arrive at this question against the background of some recent successes in enabling patients with disturbing AVHs to engage with their voices (e.g. Davies, Thomas, & Leudar, 1999). It may be that encouraging patients to engage with their voices allows them to correct abnormalities in the normal processes of internalisation and re-expansion. In this way, a troubling experience of alien voices might become a true inner dialogue: condensed, abbreviated, semantically transformed, and indistinguishable from normal inner speech.