عملکرد شناختی در هذیانها: مطالعه طولی

Background this study explored the longitudinal course of the relationship between delusions and different aspects of cognitive functioning. Methods deluded patients were compared to psychiatric and non-clinical controls on three tasks: negative priming, a probabilistic judgement task (the ‘beads’ task), and the pragmatic inference task (PIT). All groups were tested at two time points, once when actively symptomatic, and once when in remission. Results deluded individuals exhibited a ‘jump-to-conclusions’ (JTC) reasoning bias: i.e., they made decisions on the basis of limited evidence and were more likely to revise their estimates when faced with disconfirmatory evidence. This JTC bias remained stable over time, although probability judgments seemed to normalise in remission. No deficits in cognitive inhibition were found on negative priming. The deluded group displayed an excessive self-focus on the PIT at both time points, but did not show a depressive attributional style. Only a small sub-sample, characterised by the “bad-me” type of paranoia [Trower & Chadwick, 1995 Clinical Psychology: Science and Practice, 2, 263–278.], demonstrated depressive schemas when symptomatic, but no longer did so when remitted. Few relationships were found between tasks, suggesting that different areas of functioning are relatively independent. The only measures associated with delusion symptom scores were from the ‘beads’ task. Conclusions overall these findings suggest that the JTC bias is a stable factor associated with delusional thinking, while the depressive attributional style characteristic of a small sub-sample of paranoid patients fluctuates with delusional course.

The recent impetus to study symptoms rather than heterogeneous diagnostic categories such as schizophrenia (Bentall, Jackson, & Pilgrim, 1988) has resulted in a plethora of studies looking at delusions from a psychological perspective. Maher (1992) has proposed that delusions arise from the application of normal reasoning processes to abnormal experiences, i.e., delusions are perception-driven. In contrast, Garety and Hemsley (1994) have suggested that delusions are more than statements of experience, and involve an abnormal evaluative judgment arising from reasoning biases. The “delusions as explanations of experience” theory is supported from a number of sources. First, there is a large body of evidence documenting the disruption of information processing in psychotic individuals, leading to a variety of perceptual disturbances (Hemsley, 1993). Studies using negative priming (Tipper, 1985), specifically, have provided fairly robust evidence for Frith (1979)'s suggestion of deficient ‘cognitive inhibition’ in schizophrenia. The ‘negative priming effect’ refers to the observation that normal individuals show an increase in reaction time (RT) when asked to name a target which has previously been ignored as a distractor (Tipper, 1985). The negative priming effect has been proposed to be a direct measure of cognitive inhibition1 (Tipper, Weaver, & Milliken, 1995). Overall there are now 12 studies showing reduced negative priming in psychotic samples on a variety of tasks (Beech, Powell, McWilliam, & Claridge, 1989; McDowd, Filion, Harris, & Braff, 1993; Salo, Robertson, & Nordahl, 1996; Salo, Robertson, & Nordahl, & Kraft, 1997; Laplante, Everett, & Thomas, 1992; David, 1995; Park, Lenzenweger, Puschel, & Holzman, 1996; Moritz et al., 2001; Hoenig, Hochrein, Muller, & Wagner, 2002; McQueen, Galway, Goldberg, & Tipper, 2003), and in positive symptoms specifically (Williams, 1996; Peters et al., 2000). Although a few recent studies have failed to replicate such findings (Moritz, Jacobsen, Mersmann, Kloss, & Andresen, 2000; Baving, Wagner, Cohen, & Rockstroh, 2001; Hoenig et al., 2002; Roesch-Ely, Spitzer, & Weisbrod, 2003), it is likely that these results can all be explained by the critical factors of critical stimulus durations (CSDs) and inter-stimulus interval (ISIs). Second, there is evidence that delusions occur in a large number of medical and psychological conditions (Maher & Ross, 1984), and that irrational beliefs can be induced in the general population under anomalous environmental conditions (Zimbardo, Andersen, & Kabat, 1981). Maher (1992)'s theory also makes intuitive clinical sense, as for the individual who comes to the unsurprising conclusion that she is being spied upon by aliens, because the voices she hears declare themselves to be aliens employed by the CIA to spy on her. However, why not arrive at the potentially equally plausible explanation that the voices are a manifestation of a mental disturbance that distorts conscious experience (Freeman et al., 2004)? Indeed, aberrant perceptions do not always lead to delusions, nor are all delusions based on perceptual disturbances (Chapman & Chapman, 1988). Furthermore, perceptions do not exist independently of their interpretations, since cognitive templates actively influence the perceptual search (so-called “Gestalt” or “top-down” processing, Norman & Bobrow, 1976), and as such delusions may at least sometimes shape abnormal experiences rather than the other way round (Slade & Bentall, 1988). There is now a growing body of evidence demonstrating reasoning and attributional biases in people with delusions (Garety & Freeman, 1999), which challenges Maher's position. Garety and her colleagues (Huq, Garety, & Hemsley, 1988; Garety, Hemsley, & Wessely, 1991) originally demonstrated that deluded people have a “jump-to-conclusions” (JTC) reasoning style on a probabilistic reasoning task (the ‘beads’ task): they require less information before making a decision, and are more likely to revise their hypothesis in the light of disconfirmatory evidence. These results suggest that limited amounts of information represent sufficient evidence for a hypothesis to be accepted, thereby increasing the likelihood of inaccurate beliefs being formed hastily (Garety, 1991). The greater likelihood of deluded patients to revise their hypotheses is particularly interesting, since it shows that, contrary to popular misconception, they are not characteristically incorrigible if faced with neutral material. Rather, the certainty and incorrigibility traditionally ascribed to delusional beliefs are in fact the normal characteristics of any challenged, strongly held belief, such as religious or scientific beliefs (Alloy & Tabachnik, 1984; Maher, 1992). An impressive number of studies have since found a JTC bias, using various modifications of the basic paradigm, in deluded and delusion-prone individuals (see Garety & Freeman, 1999, for a review). In the largest study to date, Garety and her colleagues (In Press) found that a JTC bias was present in approximately half of their sample of 100 deluded participants. Most experiments have compared deluded with non-deluded patients, irrespective of diagnosis. In the few studies which have looked specifically at the specificity of JTC to delusions over and above diagnosis, three studies found specific relationships with delusions, but not diagnosis (Garety et al., 1991; McGuire, Junginger, Adams, Burright, & Donovick, 2001; Moritz & Woodward, in press), while two other studies found a JTC bias in patients with schizophrenia, which was not related to the presence of delusions (Mortimer et al., 1996; Menon, Pomarol-Clotet, McKenna, & McCarthy, in press). However, at least two studies have demonstrated the presence of a JTC bias in patients diagnosed with delusional disorder rather than schizophrenia (Fear & Healy, 1997; Conway et al, 2002). In addition, a recent paper comparing psychotic individuals with and without current delusions demonstrated that the JTC bias was indeed specific to the presence of delusions, since diagnostic comparisons in the same sample did not show significant differences on the same tasks (Peters, Thornton, Siksou, Linney, & MacCabe, under review). Garety et al. (in press) further specified that JTC is related to delusional conviction specifically, but not to delusional distress, preoccupation, or disruption to life. JTC is not a function of impulsiveness, since the psychotic groups adjust the amount of evidence required with a changed probability ratio (Dudley, John, Young, & Over, 1997a; Menon et al. in press; Garety et al., in press). It appears unrelated to a memory deficit (Dudley et al, 1997a; Garety et al., in press), although one study found normalised performance with the presence of a memory aid (Menon et al. in press). Error rates tend to be low, and deluded participants made significantly more errors in only two studies (Fear & Healy, 1997; Young & Bentall, 1997a). The JTC reasoning style appears to be a specific bias in data-gathering rather than a general deficit in reasoning (Bentall & Young, 1996; Young & Bentall, 1997a; Dudley et al., 1997a; Linney, Peters, & Ayton, 1998; Peters et al., under review), and is more pronounced for emotionally salient material (Dudley, John, Young, & Over, 1997b; Young & Bentall, 1997a; McGuire et al., 2001). The specificity of this bias, and the exacerbation of this bias with emotional material, answer two of the criticisms levelled at this body of work by Simpson, Done, & Vallee-Tourangeau (1998), namely that it ignores the fractionation of cognitive abilities and the influence of content on reasoning processes. Another challenge to Maher's (1992) theory is the large body of work demonstrating attributional biases in delusions (Bentall, Kinderman, & Kaney, 1994; Bentall, Corcoran, Howard, Blackwood, & Kinderman, 2001). Kaney & Bentall (1989) originally found that patients with persecutory delusions made excessively external attributions for negative events, but excessively internal attributions for positive events, on the attributional style questionnaire (ASQ; Peterson et al., 1982). This pattern of responding is suggested to represent an exaggerated form of the ‘self-serving bias’ characteristic of the normal population as a means of maintaining self-esteem. The externalising bias for negative events in paranoid patients is a robust finding which has now been replicated in four other studies (Candido & Romney, 1990; Lyon, Kaney, & Bentall, 1994; Fear, Sharp, & Healy, 1996; Sharp, Fear & Healy, 1997), although the evidence for an internalising bias for positive events is weaker. Attributional differences between deluded and other groups are restricted to self-referent tasks (Young & Bentall, 1997b), and Kinderman and Bentall (1997) have further specified that deluded individuals have a personalising bias rather than a general externalising bias for negative events, i.e., they blame other people rather than the situation. Therefore, similarly to the data-gathering bias described above, the attributional bias is of a specific rather than a general nature. Interestingly, an exaggerated self-serving bias has only been found with measures that assess attributions directly or explicitly. If the attribution task becomes opaque by being disguised as a memory task (such as the pragmatic inference task (PIT); Winters & Neal, 1985), thereby giving an implicit measure of attributional style, then some studies have reported that the same individuals who show a self-serving bias using the ASQ change to show a depressive attributional style, i.e., blaming themselves for negative events and attributing positive events to circumstances ( Lyon et al., 1994; Krstev, Jackson, & Maude, 1999). This “flipping over” of attributional style depending on whether the measure was transparent or opaque was specific to the deluded individuals, with the attributional style of the depressed and non-clinical comparison groups remaining stable. These results suggest a discrepancy between overt and covert self-representations in paranoid patients (Bentall et al., 1994). Studies exploring the relationship between self-esteem and persecutory delusions have shown contradictory findings (Freeman et al., 1998; Drake et al., 2004), suggesting the associations are complex and potentially cyclical or dynamic (Kinderman, Prince, Waller, & Peters, 2003), and depend on the type and timing of measurement (Bentall et al., 2001), as well as the sub-samples of patients examined (Freeman et al., 1998). Trower and Chadwick (1995) have indeed postulated the existence of two forms of paranoia, one characterised by high self-esteem, externalisation of blame, and delusions of persecution (i.e., “poor-me” paranoia), and the other by low self-esteem, internalisation of blame, and delusions of deserved punishment (i.e., “bad-me” paranoia), although so far there have been few empirical studies examining this distinction. To summarise, there is evidence for a reduction in cognitive inhibition in individuals with positive symptoms of psychosis, and for the presence of a specific data-gathering reasoning bias in deluded people. There is also some suggestion of a discrepancy between implicit and explicit attributional style, with deluded patients showing an implicit depressive attributional style. Garety, Kuipers, Fowler, Freeman, & Bebbington (2001) have recently proposed a cognitive model of the positive symptoms of psychosis which incorporates these three types of disruptions in cognitive functioning. However, there are two major limitations to this body of empirical research. The first concerns the lack of systematic investigation of the presence of the three forms of pathology in the same patients. This omission is likely to be due to the different theoretical contexts in which each of the above three model has been couched: the negative priming data are part of the large “cognitive deficit” in schizophrenia literature pioneered by Hemsley (1977); the reasoning studies have their roots in the study of decision-making (Kahneman, Slovic, & Tversky, 1982); while the Bentall model originates from the social cognition literature, mostly investigated in depression (Seligman, 1975). However, it is important to explore the connections between the different areas of dysfunction, since they are unlikely to function as isolated factors but may reciprocally influence each other (Bentall, 1999). The second is due to the research designs employed to date which rule out the possibility of inferring causal relationships between delusions and any of the above cognitive biases. Birchwood (1999) has pointed out that the longitudinal dimension is crucial in disentangling which processes are involved in the formation of delusions and which are merely implicated in their maintenance. Most studies have been carried out on people with active delusions, and it is not possible to determine whether the various biases found have led to the emergence of delusions, or whether delusions recruit those biases once they are activated, in a similar vein to some of the cognitive biases involved in depression (Teasdale & Barnard, 1993). It is therefore crucial to obtain longitudinal data tracing patients through their delusional fluctuations to investigate the formation versus maintenance issue. This study attempted to address these limitations. First, each patient was tested on multiple tasks, permitting the exploration of potential relationships between information processing deficits, reasoning biases, and implicit attributional style biases, all previously found in isolation in deluded patients. All tasks were chosen specifically because abnormalities of functioning do not predict worse performance (i.e., reduced cognitive inhibition leads to faster RTs in the negative priming condition; a JTC bias is closer to normative Bayesian reasoning), thereby reducing the confounding effects of the generalised performance deficit commonly found in deluded persons. A follow-up was also included to allow the possibility of making tentative causal inferences, with patients being tested both when their delusions were active and after remission. At baseline, based on the literature reviewed above, it was predicted that deluded individuals would (1) fail to show the usual negative priming effect on a negative priming task; (2) show a JTC bias on the ‘beads’ task i.e., ask to see fewer beads before making a decision as to which jar of beads had been chosen, and be more likely to revise their hypothesis in the light of disconfirmatory evidence; and (3) demonstrate an implicit depressive attributional style on an opaque attributional style task such as the PIT. No a -priori hypotheses could be made regarding the potential relationships between negative priming, JTC bias, and attributional style, since this has not been examined in the literature before. At follow-up, a reversal of the predicted lack of negative priming, the JTC bias, and the depressive attributional style, would suggest that these processes fluctuate with delusional course and may therefore occur as a consequence rather than a cause of delusions. In contrast, the stability of the above cognitive biases with symptomatic remission would be consistent with the conjecture that they may represent stable vulnerability factors potentially implicated in the formation, rather than merely the maintenance, of delusions. The JTC bias has been demonstrated both in healthy delusion-prone (Linney et al., 1998; Colbert & Peters, 2002), and at-risk (Broome et al., 2003) individuals, suggesting that it may be a stable factor. Similarly, a reduction in negative priming has previously been found in schizotypal individuals (Peters, Pickering, & Hemsley, 1994). However, Park, Puschel, Sauter, Rentsch, & Hell (2002) showed that negative priming on a spatial task, absent during the acute psychotic state, was restored at a four-months follow-up. Nevertheless, they also found that a lower negative priming score during the acute psychotic state was associated with increased positive symptoms at follow-up, suggesting that negative priming may be causally related to symptoms. Finally, while Kinderman and Bentall (2000) report some evidence for a link between paranoid ideation in healthy participants and a personalising bias, Martin and Penn (2001) did not find any significant associations between attributional bias and paranoid ideation in a non-clinical population. However, both these studies investigated explicit rather than implicit attributional style, as measured in the present study. Overall, we would therefore predict cautiously that performance on each of the tasks employed in this study would remain stable despite remission in symptoms.

Separate analyses were conducted at baseline and follow-up on each task to maximise power when patient groups were symptomatic. Due to the significant group differences in age and verbal IQ, univariate or multivariate analyses of co-variances (ANCOVAs or MANCOVAs), co-varying out age and IQ together, were first carried out on the dependent variables to determine the impact of these differences on test scores and to check for any interactions with group effects. Where significant effects of covariates occurred the relevant variables were covaried out of subsequent analyses. Where no significant effects of covariates were found, either univariate (ANOVAs) or multivariate (MANOVAs) analyses of variances are reported to maximise the power of the analyses (since there were 3 deluded individuals with missing IQ data). At follow-up 3 (group)×2 (time) ANOVAs were carried out to investigate change over time on the dependent measures. Only the within-subjects and interaction effects are reported in these analyses, since the main group effects are influenced partly by the participants’ performance at baseline, already reported in the baseline sections.