[Proposal] Should we be more suspicious of the evidence base in paranoia research?

Over the past 10-15 years multiple novel interventions for paranoia (a key symptom of mental health problems such as schizophrenia) have been developed and tested. Many of these interventions have focussed on addressing the cognitive biases that researchers have suggested play a role in the development of paranoia. For example, meta-analytic evidence (e.g., 1) has suggested that biases in (a) how much data a person needs before they make a decision and (b) how willing a person is to change decisions they have made when contradictory evidence is provided are associated with paranoia, and interventions have been developed to address these biases (e.g., SlowMo, see 2; Metacognitive Therapy, see 3). Other interventions (e.g., 4 and 5, respectively) have targeted the atypical social cognition, and the memory/attentional biases towards threatening stimuli, that appear to be associated with paranoia.

However, these interventions have proved – by and large – to not be effective. For example, a recent meta-analysis reported that, when considering only low risk-of-bias studies, Metacognitive Therapy did not have a significant effect on delusions (6). Most recently, a trial of ‘SlowMo’ found that that this intervention did not significantly reduce levels of self-reported paranoia (the study’s primary outcome variable) in comparison to treatment-as-usual (7).

One possibility is that these interventions are targeting biases/processes ‘truly’ involved in the development of paranoid thinking, but that they are modulating these processes too late in the developmental trajectory of a person’s mental health problems. If this is the case, then it is understandable that these interventions have had minimal effects on paranoid thinking, and a sensible plan for future research is to employ these same interventions, but at an earlier point in the trajectory of psychosis.

A second possibility is that the biases/processes these interventions are targeting may not ‘truly’ be associated with paranoid thinking. This is plausible because it is unlikely that clinical psychological science has been immune to the ‘replication crisis’ that has affected psychological science over the past decade, and so much of the evidence base in paranoia research may not be reliable. However, unlike other fields of psychological science, clinical psychological science has been slow to (a) examine the replicability of the findings it has generated and (b) adopt methodological practices that will result in clinical psychological science generating more robust, trustworthy findings.

To help determine which of these possibilities is correct, our proposed study will examine whether we can replicate five key findings from paranoia research, using a large, non-clinical sample. Studies using non-clinical participants are a key part of the evidence base in psychosis research, because if we show that cognitive biases/processes are associated with non-clinical (as well as clinical) paranoia, we can be more confident that these biases/processes play a role in the development of paranoid thinking (rather than being a consequence of the development of psychosis).

Our proposed study will employ a correlational, cross-sectional design. Participants will be a nationally representative sample of UK adults, will have a good understanding of English, and will have no history of mental health problems, neurological problems, or significant head injury.

After providing demographic information, participants will complete five tasks and two questionnaires. The five tasks participants will complete assess the cognitive biases/processes that (i) previous studies have suggested are associated with both clinical and non-clinical paranoia, and (ii) are typical targets of interventions that aim to reduce the severity of paranoia. Reduced data-gathering (AKA hasty decision-making) will be assessed using a standard Beads Task (8). An unwillingness to revise judgements following exposure to contradictory information will be assessed using a standard Bias Against Disconfirmatory Evidence task (9). Social cognition will be assessed using two tasks – one that assesses facial emotion recognition (the Penn Emotion Recognition–40 Test; 10) and one that assesses the ability to keep track of others’ mental states (the Imposing Memory Test; 11). Finally, memory for threat-related stimuli will be assessed using Larøi et al.’s (12) memory-for-faces task, (which examines whether participants preferentially remember angry faces, in comparison to faces expressing neutral/positive emotions). One questionnaire – the Paranoia Checklist (13) – will assess the frequency of paranoid thoughts. The second questionnaire – the negative subscale of the Positive and Negative Affect Schedule (14) – will assess levels of negative affect.

Based on a power calculation using GPower (15), where we assumed a smallest-effect-size-of-interest of r = .16 (equating to us accounting for 2.5% of the variation in paranoid thinking), an alpha of .01 (standard alpha with a Bonferroni correction), and aiming for 95% power, a sample of 603 participants will be required. Given that the study will take 60 minutes to complete, the cost of the study will be £6,654.16.

Our proposed study has been pre-registered at OSF As in our recent work (16), all data, all analysis code, and (wherever copyright rules allow) all testing materials, will be posted at OSF | Cognitive Correlates of Paranoid Thinking. A preprint reporting the study’s findings will be posted on psyarxiv and we aim to publish a peer-reviewed manuscript in PLoS One (where we can publish work open-access free-of-charge, as part of their jisc.ac.uk agreement).

Novel interventions for paranoia, which have targeted a set of specific cognitive biases/process, have had disappointingly small effects on patient outcomes. It is possible that this has happened because the evidence base which has supported the development of these interventions is unreliable. The aim of this proposed study is to examine whether we can replicate five key findings from paranoia research. If we are able to replicate previously-reported effects, it may indicate that the evidence base in paranoia research is reasonably robust, and this may suggest that recently-developed interventions may be effective if offered to patients earlier in the trajectory of their mental health problems. Alternatively, if we are unable to replicate previously-reported effects, it will indicate that the evidence base in paranoia research is unreliable, and this may explain why recently-developed interventions have exerted only minimal effects on the severity of paranoid thinking in patients.

References are available here .