Cambridge Journal of Human Behaviour

RM, externalising errors, and hallucinations
Reality Monitoring (RM) can be defined as the ability to discriminate between internally and externally-generated stimuli (Humpston et al., 2017; Simons et al., 2017; Johnson et al., 1993; Johnson & Raye, 1981). To achieve this, RM uses cognitive faculties to accurately identify sensations derived from experiences within compared to those outside of the self. 

In particular, externalising errors refer to incorrectly attributing an internally generated event as being externally generated (Humpston et al., 2017; Brookwell et al., 2013; Woodward & Menon, 2013; Waters et al., 2012) which can result in hallucinations; for example: an individual reporting they hear voices that are not there because they mistake imagined internal voices, they have generated themselves, for real external voices. 

Hallucinations can be defined as sensory experiences that occur in the absence of corresponding external stimuli which elicit a particular sensory response (Woodward & Menon, 2013; Slade & Bentall, 1988). In other words, hallucinations are sensory misperception—instances of perceiving stimuli that are not actually there—that occur when self-generated stimuli are taken as being externally generated. Additionally, hallucinations can occur in any sensory modality including auditory, visual, olfactory, gustatory, or tactile (Waters et al., 2014; Shergill, 2001; Kopala et al., 1994; Mueser et al., 1990), although, most are auditory or visual (Waters et al, 2014). Even within one particular sensory modality, hallucinations can come in many different forms. For instance, visual hallucinations may result in seeing people, shapes, flashes of light, and colours that are not actually there.

The Source Monitoring Framework
A salient theoretical explanation outlining how RM deficits may lead to hallucinations is the Source Monitoring Framework (Johnson et al., 1993; Johnson & Raye, 1981). This theory posits that decision-making processes performed when remembering are evaluated and attributed to particular sources, thereby allowing us to determine the origin of information as either being internal or external. In this sense, RM is viewed as a product of judgement processes which allow us to encode information about experiences. Examples of this include encoding sensory details, contextual information (spatial and temporal), semantic content (e.g., tendency for internally generated stimuli to relate to the self) and the type of cognitive processes engaged (e.g., imagery; Johnson & Raye, 1981). These characteristics of memories provide a sense of reality and are similar to ongoing perceptions that occur in real-time (Garrison et al., 2017; Johnson & Raye, 1981). 

Essentially, information that we use to conceptualise an experience is woven into a composite of integrated information which makes up a memory, or our perception of an event. Each piece of information has different respective “weights” which are evaluated against criteria to determine the probability of stimuli being self or other generated. However, errors in RM can occur when one source of information is weighted as being greater than another through biassed cognitive processing or due to impairments in cognitive operations (Garrison et al., 2017). Not meeting the ascribed criteria along a single dimension may result in RM deficits when judgement processes suggest that an event is between the “cut-off” for an internally and externally generated stimulus. Therefore, difficulty in determining the origin of a stimulus can lead to RM deficits as this makes it hard to distinguish between internal and external stimuli. Consequently, the Source Monitoring Framework proposes that flexible criteria used in judgement processes are prone to errors which may compromise an individual’s ability to discern the origin of stimuli, leading to RM deficits. 

More recent theories, such as misattribution models, characterise RM deficits leading to hallucinations in a similar way. Woodward and Menon (2013) focus on two important dimensions of hallucinations: (1) the self-generated/non-self-generated and (2) the inner/outer dimension (Stephane et al., 2003). The former dimension relates to the origin of a specific cognitive event, whilst the latter relates to the spatial location of a cognitive event (Woodward & Menon, 2013). This is an extension of the Source Monitoring Framework explanation of hallucinations with the addition of an “inner/outer” spatial dimension, which assumes that hallucinations result from certain cognitive operations leading to externalising error. 

On the other hand, current computational models offer an account of hallucinations which incorporate the role of top-down processes (e.g., the influence of expectations and prior beliefs) to a larger extent. This theory seems to converge with the Source Monitoring Framework in that features used to generate a memory tend to be driven by the interaction between top-down and bottom-up processes, with a failure to determine the origin of stimuli generating hallucinatory experiences (Garrison et al., 2019; Powers et al., 2017). 

The Continuum Model of Psychosis
Hallucinations are known to occur in both clinical and non-clinical populations. This suggests that there must be some form of continuum between these populations rather than them being categorically opposed. The dimensional nature of phenomena such as hallucinations can be conceptualised through a continuum ranging between psychosis, characteristic of clinical populations, to the lack of psychosis in non-clinical populations (van Os et al., 2009; Johns & van Os, 2001). That is to say, hallucinatory experiences occur in clinical (Toh et al., 2016; Ditman & Kuperberg, 2005) and non-clinical populations (Powers et al., 2017; Sommer et al., 2010; Serper et al., 2005). However, hallucinations are more likely to be transitory (Johns et al., 2014; Linscott & van Os, 2013; Bartels-Velthuis et al., 2011) and are experienced much less frequently in non-clinical populations (Sommer et al., 2010; Badcock et al., 2008; Paulik et al., 2006). This means that hallucinatory experiences are unevenly distributed throughout the general population (Siddi et al., 2019; Garrison et al., 2017). 

This paper will place an emphasis on individuals with schizophrenia as hallucinations are a key symptom required for diagnosis, as defined by the Diagnostic and Statistical Manual of Mental Disorders (American Psychiatric Association, 2013), affecting between 60% and 80% of patients (Waters et al., 2012). However, there is dispute over the extent to which a pure continuum exists between clinical and non-clinical populations (Baumeister et al., 2017) as the underlying mechanisms for hallucinations may differ between these populations. 

Support for the continuum model of psychosis comes from van Os and colleagues’ (2009) seminal systematic review of psychotic experiences and symptoms in the non-clinical population. The authors presented evidence for the (1) distributional, (2) psychopathological, and (3) epidemiological validity of the continuum model. First, the model of a psychosis continuum has distributional validity because the statistical distribution for a single genetic defect does not directly map onto the distribution of psychotic symptoms. This suggests that hallucinations are caused by multiple interacting factors. Thus, it is most appropriate to take a holistic approach incorporating multiple genes and environmental factors that may contribute to the continuum between health and psychosis rather than looking at these concepts as two opposing categories. Second, there is psychopathological validity to the model as scales used to measure subclinical symptoms (e.g., schizotypy scales) seem to be in parallel with symptoms used to identify whether an individual has schizophrenia, although these tend to be less severe (Lewandowski et al., 2006; Mata et al., 2003). Third, the high epidemiological validity of the model extends from its ability to account for the discrepancy between the (a) prevalence of psychosis-like experiences and (b) number of individuals diagnosed with psychosis in the general population. Specifically, continuum models assume that non-clinical populations are able to experience hallucinations in addition to clinical populations unlike categorical models of “illness” and “health” which do not allow for this. Given these fundamental assumptions, the continuum model can explain population data showing that the number of psychosis-like experiences exceeds the number of patients with psychosis.

Extending from this, recent models of psychosis such as the “Transdiagnostic Psychosis Spectrum” have posited that symptoms can be represented as unique to psychosis (e.g., affective dysregulation) or not (e.g., hallucinations; van Os & Reininghaus, 2016). This is supported by empirical work which has demonstrated that some characteristics of hallucinatory experiences are common between clinical and non-clinical populations (e.g., low-level acoustic features, content, frequency of hallucinations) whilst other characteristics are not (e.g., negative affective responses to hallucinations, less perceived control over hallucinations; Powers et al., 2017). For this reason, there may be utility to incorporating both categorical and continuum models when tailoring interventions to clinical or non-clinical populations facing hallucinations. 

RM and hallucinations in individuals with schizophrenia
RM has been applied to inform our understanding of hallucinations in clinical groups including those with schizophrenia. In this section, I will review the behavioural evidence comparing individuals with schizophrenia who hallucinate to those who do not and then turn to the neural evidence to determine whether RM deficits underlie hallucinations in those with schizophrenia. 

Hallucinations in patients with schizophrenia may arise due to RM deficits which result in externalising errors. A noteworthy meta-analysis by Brookwell and colleagues (2013) examined whether auditory hallucinations stem from externalising errors using data from 15 clinical studies between 1985 to 2012. The researchers found that hallucinating patients with schizophrenia (n = 240) displayed increased tendencies to misattribute internally generated events as having been externally generated compared to non-hallucinating patients (n = 249). This provides evidence that the presence of hallucinations in patients with schizophrenia may be related to impaired RM abilities which result in externalising errors. An important critique of this study, and the research field more generally, is that sample sizes tend to be very small. In this particular meta-analysis, a mean of 23 participants were included in each study (Moseley et al., 2021) which could limit our ability to detect statistically significant results when they are present due to low statistical power. 

Subsequent research suggests that externalising errors are linked to the experience of hallucinations (Damiani et al., 2022; Simons et al., 2017). One review paper found that patients with schizophrenia who had experienced hallucinations (n = 119) were more likely to misidentify internally generated stimuli as having been externally generated than patients without hallucinations (n = 135; Simons et al., 2017). A range of tasks were used to assess RM abilities including semantic association, word recognition, word-stem completion, and sentence completion tasks. A methodological critique of using multiple measures of RM is that what qualifies as “internally” and “externally” generated stimuli differs across the studies reviewed. This may have resulted in differences in the effect sizes calculated, therefore we must carefully interpret the findings. For instance, the word recognition task (Brunelin et al., 2006) generated a large effect size. Perhaps this happened because for those who already struggle with RM it is harder to remember the source of stimuli that are presented artificially (i.e., words taken from a verbal fluency task with the same emotional valence and length), compared to stimuli in everyday life situations. Converging findings from a very recent systematic review and meta-analysis of 41 studies suggest that hallucinating patients with schizophrenia or who were on the schizophrenia spectrum were more likely to make externalising errors compared to patients who did not report hallucinating, of moderate effect size (Damiani et al., 2022). This provides further evidence to suggest that RM deficits leading to externalising errors underlie hallucinations in those with schizophrenia.

Neural evidence supports the above findings by demonstrating that brain morphology is linked to RM and hallucinations in those with schizophrenia. In terms of hallucinations, reduced length of the left paracingulate sulcus (PCS) has been shown in patients with schizophrenia who hallucinate compared to those who do not (Rollins et al., 2020; Garrison et al., 2015; Yucel et al., 2002). The PCS is a structure that lies in the dorsal anterior cingulate portion of the medial prefrontal cortex (Garrison et al., 2019). Garrison and colleagues (2015), found that left hemisphere PCS length was significantly reduced in patients with schizophrenia who experienced hallucinations compared to patients who did not. These findings had good predictive validity as left PCS length could accurately predict hallucinations in 70.8% of patients with schizophrenia. Specifically, reducing the left PCS length by 1 cm increased the likelihood of hallucinations by 19.9% in patients. A strength of the logistic regression model used to predict hallucinations is that several confounding demographic and clinical variables were controlled for, thereby increasing our confidence in the internal validity of the findings. Support for this comes from Rollins and colleagues (2020) who found that, in a cross-cultural data set of patients with schizophrenia from the UK and Shanghai, those who experienced hallucinations had significantly shorter left PCS lengths than those who did not. This suggests that brain morphology may be linked to hallucinations in those with schizophrenia. To complement these findings, a recent study has linked the PCS to RM. Perret and colleagues (2021) found that reduced length of the right PCS in individuals with schizophrenia was significantly associated with RM deficits including decreased RM accuracy and increased externalising errors. Thus, the PCS may be linked to RM deficits leading to hallucinations in those with schizophrenia. 

In summary, RM deficits which result in externalising errors may underlie hallucinations in patients with schizophrenia because they are more likely to exhibit externalising errors than their non-hallucinating counterparts. Neural evidence seems to confirm the behavioural findings by demonstrating a link between brain morphology, RM, and hallucinations in individuals with schizophrenia. This presents a compelling narrative that RM deficitsunderlie the experience of hallucinations in schizophrenia. 

RM and hallucination proneness in non-clinical populations
RM deficits may elucidate our understanding of hallucination proneness in the non-clinical population given the potential link between RM deficits and hallucinations in the clinical population. Hallucination proneness refers to subclinical symptoms which increase the likelihood of an individual hallucinating. This is most commonly measured using scales such as the Launay Slade Hallucination Scale (LSHS) and the Launay Slade Hallucination Scale Revised (LSHS-R; Launay & Slade, 2012). In this section, I will outline behavioural evidence supporting and refuting the link between RM deficits and hallucination proneness in the non-clinical population. Then, I will discuss how the neural evidence mirrors these mixed findings.

The evidence that RM deficits underlie hallucination proneness in non-clinical populations is mixed. On one hand, empirical studies have shown that hallucination-prone participants are more likely to have deficits in RM such as externalising errors (Brookwell et al., 2013; Larøi et al., 2004). Larøi and colleagues (2004) found that hallucination prone participants (n = 25) tended to misidentify self-generated words as having been spoken by the experimenter more than their non-hallucination prone counterparts (n = 25). However, an important methodological critique of Larøi and colleagues’ (ibid.) study is the lack of experimental control over the type of words internally generated by participants in response to being shown a standard stimulus. Self-generated words with increased salience may become more embedded in memory. Thus, participants who produced more meaningful words in response to a standard stimulus were less likely to produce externalising errors, reducing our ability to detect externalising errors accurately.

Converging findings from Brookwell and colleagues’ (2013) meta-analysis demonstrated that hallucination prone individuals were more likely to mistake internally generated events as having been externally generated compared to non-hallucination prone individuals. However, Brookwell and colleagues’ meta-analysis included 2/9 studies that had not been made available in the public domain (“unpublished observations”) and only 2/7 of the remaining studies used a source monitoring task to directly assess RM, with other measures being indirect (ibid.). This suggests that we should approach these results with caution, given such limitations reduce our confidence in the internal validity of the findings. 

On the other hand, empirical studies have shown that RM deficits may not underlie hallucination proneness in non-clinical populations (Moseley et al., 2021; Alderson-Day et al., 2019; Garrison et al., 2017; McKague et al., 2012). In both McKague and colleagues’ (2012) and Garrison and colleagues’ (2017) studies, participants were made to distinguish between whether auditory and visual information had been self or other generated. A synthesis of the data reveals no significant differences in externalising errors exhibited by hallucination prone participants in comparison to their non-hallucination prone counterparts, thus providing consistent evidence that RM abilities are intact in hallucination prone individuals. A strength of Garrison and colleagues’ study is that, in contrast to Larøi and colleagues’ (2004), there was tighter control over confounding variables such as participants’ age, language skills, and the presence of general memory deficits which increases our confidence in the internal validity of the results. Most recently, Moseley and colleagues’ (2021) large-scale pre-registered multisite study (n = 1,394) found no relationship between RM and hallucination proneness in the non-clinical population. A major strength of this study is the large sample size used, which provides greater statistical power for the analysis thereby increasing our confidence in the internal validity of the results. 

Neural evidence linking brain morphology findings to RM and hallucination proneness in the non-clinical population is also mixed, mirroring the behavioural evidence (Garrison et al., 2019; Buda et al., 2011). Garrison and colleagues (2019) found no significant difference in the length of the PCS in either hemisphere for hallucination prone (n = 50) and non-prone (n = 50) controls. This suggests that in the non-clinical population, differences in brain morphology are not linked to hallucination proneness. Turning now to findings on RM, in their seminal paper Buda and colleagues (2011) classified a small sample of 52 healthy volunteers into 4 groups based on the presence or absence of the PCS in the left or right hemisphere. The researchers found that participants with an absent PCS in both hemispheres showed significantly reduced RM performance compared to all other participants for recollection of self/experimenter status, but not for perceived/imagined items. This suggests that brain morphology may be related to RM in the non-clinical population. Taking the evidence together, there are mixed findings on whether RM deficits underlie hallucination proneness in non-clinical populations. Arguably, studies which have suggested that RM deficits do not underlie hallucination proneness in non-clinical populations (Moseley et al., 2021; Garrison et al., 2017; McKague et al., 2012) have improved on the methodological limitations of previous studies to the contrary (Brookwell et al., 2013; Larøi et al., 2004). However, further replication is needed to confirm whether inconsistent findings result from methodological differences or chance. Therefore, RM deficits may or may not underlie hallucination proneness in the non-clinical population. 

Taking the evidence together, there are mixed findings on whether RM deficits underlie hallucination proneness in non-clinical populations. Arguably, studies which have suggested that RM deficits do not underlie hallucination proneness in non-clinical populations (Moseley et al., 2021; Garrison et al., 2017; McKague et al., 2012) have improved on the methodological limitations of previous studies to the contrary (Brookwell et al., 2013; Larøi et al., 2004). However, further replication is needed to confirm whether inconsistent findings result from methodological differences or chance. Therefore, RM deficits may or may not underlie hallucination proneness in the non-clinical population. 

The current study
The present study aims to expand the current literature assessing the association between RM and hallucination proneness in the non-clinical population. To achieve this, an existing dataset collected by previous Part II Psychological and Behavioural Sciences students at the University of Cambridge was used. The analysis used a non-directional hypothesis in line with the mixed findings of previous research. 

First, the research will examine whether there is a relationship between RM accuracy and hallucination proneness in a non-clinical sample of university students. Here, RM accuracy refers to one’s ability to correctly determine the origin of a source as having been perceived or imagined. Second, the study will identify whether externalising errors are related to hallucination proneness in a non-clinical sample of university students. In particular, externalising errors will be assessed by focusing on items presented as “imagined” that are incorrectly reported by participants as “perceived”. 

Two alternate hypotheses will be addressed through the present study. This is because RM deficits are indicated by low RM accuracy and/or high externalising error scores. In the case of low RM accuracy, individuals have difficulty distinguishing between internally and externally generated stimuli whilst high externalising error suggests a tendency to mistake internal stimuli for external stimuli, both of which indicate RM deficits.

• H1: There is a relationship between RM accuracy and hallucination proneness in a non-clinical sample of university students.
• H2: There is a relationship between externalising errors and hallucination proneness in a non-clinical sampel of university students.