Testing Continuum Models of Psychosis

© 2016 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). In Press: Garrison, J. R., et al., Testing continuum models of psychosis: No reduction in source monitoring ability in healthy individuals prone to auditory hallucinations, Cortex (2016), http://dx.doi.org/10.1016/j.cortex.2016.11.011 CORTEX1882_proof ■ 4 December 2016 ■ 1/11

cortex xxx (2016) 1e11

Available online at www.sciencedirect.com 65 66 1 ScienceDirect 67 2 68 3 69 www.elsevier.com/locate/cortex 4 Journal homepage: 70 5 71 6 72 7 73 8 74 9 Special issue: Research report 75 10 76 11 Testing continuum models of psychosis: No 77 12 78 13 reduction in source monitoring ability in healthy 79 14 80 15 81 16 Q8 individuals prone to auditory hallucinations 82 17 83 18 a,b,1 c,d,1 c 84 19 Q7 Jane R. Garrison , Peter Moseley , Ben Alderson-Day , 85 * 20 David Smailes e, Charles Fernyhough c and Jon S. Simons a,b, 86 21 87 a 22 Department of Psychology, University of Cambridge, UK 88 23 b Behavioural and Clinical Neuroscience Institute, University of Cambridge, UK 89 24 c Psychology Department, Durham University, UK 90 25 d School of Psychology, University of Central Lancashire, UK 91 26 e School of Health and Social Sciences, Leeds Trinity University, UK 92 27 93 28 94 29 article info abstract 95 30 96 31 97 Article history: People with schizophrenia who hallucinate show impairments in reality monitoring (the 32 98 Received 18 July 2016 ability to distinguish internally generated information from information obtained from 33 99 34 Reviewed 2 October 2016 external sources) compared to non-hallucinating patients and healthy individuals. While 100 35 Revised 8 November 2016 this may be explained at least in part by an increased externalizing bias, it remains un- 101 36 Accepted 14 November 2016 clear whether this impairment is specific to reality monitoring, or whether it also reflects 102 37 Published online xxx a general deficit in the monitoring of self-generated information (internal source moni- 103 38 toring). Much interest has focused recently on continuum models of psychosis which 104 39 Keywords: argue that hallucination-proneness is distributed in clinical and non-clinical groups, but 105 40 few studies have directly investigated reality monitoring and internal source monitoring 106 41 Schizophrenia Auditory verbal hallucinations abilities in healthy individuals with a proneness to hallucinations. Two experiments are 107 42 108 Reality monitoring presented here: the first (N ¼ 47, with participants selected for hallucination-proneness 43 109 Internal source monitoring from a larger sample of 677 adults) found no evidence of an impairment or external- 44 110 45 izing bias on a reality monitoring task in hallucination-prone individuals; the second 111 ¼ 46 (N 124) found no evidence of atypical performance on an internal source monitoring 112 47 task in hallucination-prone individuals. The significance of these findings is reviewed in 113 48 light of the clinical evidence and the implications for models of hallucination generation 114 49 discussed. 115 50 © 2016 The Authors. Published by Elsevier Ltd. This is an open access article under the CC 116 51 BY license (http://creativecommons.org/licenses/by/4.0/). 117 52 118 53 119 54 120 55 121 56 122 57 123 58 124 59 125 60 * Corresponding author. Department of Psychology, University of Cambridge, Downing Street, Cambridge CB2 3EB, UK. 126 61 E-mail address: [email protected] (J.S. Simons). 127 62 1 Joint first authors. 128 63 http://dx.doi.org/10.1016/j.cortex.2016.11.011 129 64 0010-9452/© 2016 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons. org/licenses/by/4.0/). Please cite this article in press as: Garrison, J. R., et al., Testing continuum models of psychosis: No reduction in source monitoring ability in healthy individuals prone to auditory hallucinations, Cortex (2016), http://dx.doi.org/10.1016/j.cortex.2016.11.011 CORTEX1882_proof ■ 4 December 2016 ■ 2/11

2 cortex xxx (2016) 1e11

1 Gilbert, Frith, & Burgess, 2008), as well as to structural 66 1. Introduction 2 morphology of the nearby paracingulate sulcus (PCS; Buda, 67 3 68 Fornito, Bergstrom, & Simons, 2011). Patients with schizo- 4 Auditory verbal hallucinations (AVH), or the experience of 69 phrenia show impairments in reality monitoring ability (e.g., 5 hearing a voice in the absence of any speaker, are experienced 70 Anselmetti et al., 2007; Brebion et al., 2000; Waters, Maybery, 6 by a large proportion of individuals with a diagnosis of 71 Badcock, & Michie, 2004), which are associated with 7 schizophrenia, as well as those with other psychiatric di- 72 8 dysfunction in the medial anterior PFC (Subramaniam et al., 73 agnoses such as bipolar disorder and post-traumatic stress 9 2012; Vinogradov, Luks, Schulman, & Simpson, 2008), as well 74 disorder (PTSD), and by approximately 1% of the healthy 10 as to altered PCS morphology (Garrison, Fernyhough, 75 population (Kra˚kvik et al., 2015). Cognitive and neuroscientific 11 McCarthy-Jones, Haggard, & Simons, 2015). Indeed, Garrison 76 studies aimed at understanding the underlying mechanisms 12 et al. (2015) indicated that a shorter PCS was associated with 77 of AVH have compared task performance and/or neural acti- 13 a higher likelihood of hallucinations in patients with schizo- 78 vation between individuals with psychiatric diagnoses who 14 phrenia, with these findings together suggesting that the PCS, 79 15 hallucinate and those who do not (Stephane, Kuskowski, 80 and surrounding anterior medial PFC, may be associated with 16 McClannahan, Surerus, & Nelson, 2010), as well as between 81 both reality monitoring and hallucinations. Considering the 17 groups of individuals with no clinical diagnoses who report 82 18 wider underlying network for AVH, an fMRI study with 83 Q1 differing levels of hallucination-proneness (Laroi, Van der 19 healthy individuals observed increased activation in the area 84 Linden, & Marczewski, 2004). One of the most prominent 20 surrounding the auditory cortices in the superior temporal 85 cognitive models of AVH holds that these symptoms occur 21 gyrus (STG) during the encoding stage of a reality monitoring 86 when internal mental events, such as inner speech, are mis- 22 task, which correlated with measures of hallucination- 87 attributed to an external, non-self-generated source (Bentall, 23 proneness (Sugimori, Mitchell, Raye, Greene, & Johnson, 88 1990; Frith, 1992; Moseley, Fernyhough, & Ellison, 2013). As 24 2014). Both the PCS and STG regions are often observed to be 89 such, research has focused on the question of how we typi- 25 active during the experience of AVH in ‘symptom-capture’ 90 26 cally distinguish between different sources of information, 91 fMRI studies (e.g., Zmigrod, Garrison, Carr, & Simons, 2016). 27 and how these processes might fail. 92 To test the suggestion that reality monitoring deficits play 28 The Source Monitoring Framework addresses how we 93 a causal role in the generation of AVH, research has focused 29 make judgements about the origin (source) of remembered 94 on the behavioural association between atypical source 30 information, using characteristics such as perceptual, se- 95 31 monitoring and the presence or intensity of hallucinations. 96 mantic, or affective content, or the nature of the earlier 32 Two mechanisms have been proposed which might explain 97 cognitive operations (Johnson, Hashtroudi, & Lindsay, 1993). 33 this deficit: an externalizing bias and a general source moni- 98 Source monitoring can be broadly divided into three sub- 34 toring deficit. The idea of externalizing bias stems from the 99 categories depending on the contrasts which must be made: 35 observation made during reality monitoring studies involving 100 1. External source monitoring, where the distinction is be- 36 healthy individuals, that participants often exhibit a greater 101 tween non-self-generated sources of information, such as 37 likelihood of falsely attributing new or internally generated 102 38 whether an image appeared on the left or right side of a 103 items to an external source, than of making the reverse error 39 screen; 2. Internal source monitoring, where a distinction 104 (Johnson, Raye, Foley, & Foley, 1981; see Garrison, Bond, 40 must be made between self-generated sources of information, 105 Gibbard, Johnson, & Simons, 2016, for a discussion). Bentall 41 such as whether a sentence had previously been spoken aloud 106 42 (1990) argued that since hallucinations are internally gener- 107 or internally using inner speech; and 3. Reality monitoring, 43 ated events experienced as external, atypical source moni- 108 involving discrimination between internal and external 44 toring in individuals with AVH is most likely to manifest itself 109 sources of information, such as whether a sentence had been 45 as an enhanced externalizing bias (in which self-generated 110 spoken by the individual or by someone else, or even whether 46 information is more likely to be misattributed as externally- 111 an event had been witnessed or dreamt. Each of these variants 47 generated). Behavioral evidence supports this proposal, with 112 are commonly tested using a source memory paradigm, 48 a recent meta-analysis finding that patients with hallucina- 113 49 requiring the participant to encode stimuli from different 114 tions have a greater tendency to misattribute internal items to 50 sources, and on later re-presentation of the stimuli, to judge 115 external sources than non-hallucinating individuals or 51 the original source of the stimuli. For example, a reality 116 healthy controls (Brookwell, Bentall, & Varese, 2013). 52 monitoring task might present participants with a series of 117 An alternative possibility is that individuals with AVH 53 verbal word-pairs (e.g., bubble and squeak), which are shown 118 54 exhibit general source monitoring deficits, which can be 119 either completed (‘perceived’, that is externally generated, 55 observed in terms of poorer performance across all types of 120 e.g., bubble and squeak) or where the second word must be 56 source memory tasks. Such a deficit might arise in addition to 121 supplied by the participant (‘imagined’, that is, internally 57 an externalizing bias (e.g., the deficit might be explained by 122 generated, e.g., bubble and s____). Reality monitoring ability 58 some variation in the application of criteria used to determine 123 might then be assessed by asking the participant to remember 59 the internal/external nature of mental experience), or may 124 whether the second word of the word-pair had previously 60 itself be related to the generation of the bias (e.g., if the weak 125 61 been perceived or imagined. 126 application of decision-making criteria generally has a greater 62 Reality monitoring ability in healthy individuals is associ- 127 impact on the recognition of self-generated status than of 63 ated with activity in the medial anterior prefrontal cortex, 128 external status). Evidence suggests that as well as deficits in 64 (PFC, e.g., Simons, Henson, Gilbert, & Fletcher, 2008; Simons, 129 65 reality monitoring, patients with schizophrenia do often 130 Davis, Gilbert, Frith, & Burgess, 2006; Turner, Simons, exhibit internal and external source memory deficits when

Please cite this article in press as: Garrison, J. R., et al., Testing continuum models of psychosis: No reduction in source monitoring ability in healthy individuals prone to auditory hallucinations, Cortex (2016), http://dx.doi.org/10.1016/j.cortex.2016.11.011 CORTEX1882_proof ■ 4 December 2016 ■ 3/11

cortex xxx (2016) 1e11 3

1 compared to healthy controls (Achim & Weiss, 2008). by the experimenter, and that word-pairs which had been 66 2 Furthermore, the few studies which have compared source imagined should be judged to have been perceived. If such 67 3 68 monitoring deficits in patients with and without hallucina- effects reflect an externalizing bias, they should be specific to 4 69 tions offer some support for an association between general the reality monitoring task, with no difference observed on 5 70 source monitoring deficits and hallucinations (Franck et al., the internal source monitoring task. Alternatively, a general 6 71 & 7 Q2 2000; Gawe˛da, Woodward, Moritz, Kokoszka, 2013). source monitoring deficit account of AVH would predict that 72 8 Interpreting the results of such empirical comparisons higher hallucination-proneness would be associated with 73 9 between patients and healthy individuals can be affected by lower overall performance on both the reality monitoring task 74 10 the confound of medication status, and by other factors. and the internal source monitoring task. 75 11 Continuum models of psychosis, which suggest that that ex- 76 12 periences such as AVH are distributed throughout the general 77 13 Q3 population, entail that studying non-clinical individuals with 2. Experiment 1: reality monitoring 78 14 a proneness to hallucinate can provide a useful model of 79 15 80 clinical syndromes (van Os & Reininghaus, 2016). Based on 2.1. Methods 16 81 this approach, a small number of studies have investigated 17 82 18 whether individuals with no psychiatric diagnosis who report 2.1.1. Participants 83 19 hallucinatory experiences exhibit the same bias and/or deficit 677 participants were recruited to an on-line questionnaire by 84 20 in source monitoring that has been associated with patients email invitation from volunteer lists maintained at the 85 21 with schizophrenia. This area remains under-researched e in Behavioural and Clinical Neuroscience Institute at Cambridge 86 22 their review, Brookwell et al. (2013) reported three source University, and the Department of Psychology, Durham Uni- 87 23 monitoring studies in non-clinical populations, only one of versity, and from advertisements in the Cambridge and 88 24 which has been published. Larøi et al. (2004) tested under- Durham areas. There was no financial incentive to participate 89 25 graduate students on a reality monitoring task, classifying and ethical approval for the study was obtained from the 90 26 91 participants according to their score on a self-report ques- University of Cambridge Psychology Research Ethics Com- 27 92 tionnaire, the Launay-Slade Hallucination Scale (LSHS). They mittee. An individual's proneness to auditory hallucinations 28 93 found that high hallucination-prone individuals were more was assessed using a modified version (Morrison, Wells, & 29 94 30 likely to misattribute self-generated words as having been Nothard, 2000) of the Predisposition to Auditory Hallucina- 95 31 spoken by the experimenter than those in the low tion subscale of the Revised Launay-Slade Hallucination Scale 96 32 hallucination-prone group, whereas there was no difference (LSHS-R, Bentall & Slade, 1985; see Section 2.1.2). Individuals 97 33 between the groups in other errors, or in recognition memory who had LSHS-R scores in the upper or lower quartile (High- 98 34 for previously presented words. However, in contrast, one LS, or Low-LS) indicating high or low proneness to auditory 99 35 study published since the Brookwell et al. meta-analysis hallucinations were invited for testing using the reality 100 36 found no effect of non-clinical hallucination-proneness on monitoring task in the Department of Psychology at either the 101 37 reality monitoring (Self/Experimenter discrimination; University of Cambridge or Durham University. 102 38 103 McKague, McAnally, Skovron, Bendall, & Jackson, 2012). Twenty-five individuals were tested in the High-LS group 39 104 It thus remains unclear whether the reality monitoring (number of females ¼ 18; mean age ¼ 19.8, SD ¼ 2.8 years; 40 105 ¼ ¼ 41 impairment observed in patients with schizophrenia who mean LSHS-R score 13.2, SD 2.1), and 22 individuals in the 106 ¼ ¼ 42 hallucinate is also present in non-clinical hallucination-prone Low-LS group (number of females 20; mean age 22.9, 107 43 samples. Furthermore, to our knowledge, no study has tested SD ¼ 7.5 years; mean LSHS-R score ¼ 2.1, SD ¼ 1.4). Proneness 108 44 the relationship between hallucination-proneness in a non- to auditory hallucinations, as measured by the LSHS-R, 109 45 clinical sample and performance on internal source moni- differed significantly between these groups: t(45) ¼ 20.973, 110 46 toring tasks, which might support the presence of a general- p < .001. There were no significant differences between the 111 47 ised deficit in source monitoring in the generation of groups in terms of age [t(45) ¼ 1.932, n.s.] or sex (c2 ¼ 2.703, 112 48 hallucinations. Here, we report data from two separate ex- n.s.), all participants reported being native English speakers, 113 49 114 periments conducted with individuals recruited from two and no participants reported any hearing problems. 50 115 university populations, which investigated whether non- 51 116 clinical hallucination-proneness is associated with impair- 2.1.2. Design and procedure 52 117 e 53 ments in source memory performance, and if so, whether this Self-report measures The revised version of the Launay- 118 54 is explained by an externalizing bias and/or a general internal Slade Hallucination Scale (LSHS-R), used to assess predispo- 119 55 source monitoring deficit. Experiment 1 recruited participants sition to hallucinatory experiences in the auditory modality, 120 56 who scored in the top or bottom quartiles of a version of the comprises five questions (e.g., I have had the experience of 121 57 LSHS, and tested for an association between self-reported hearing a person's voice and then found that no-one was there), with 122 58 hallucination-proneness and reality monitoring perfor- each item scored on a five-point Likert scale ranging from 123 59 mance. Experiment 2 tested for an association between ‘never’ (0) to ‘almost always’ (4). Total scores can thus range 124 60 hallucination-proneness and internal source monitoring per- from 0 to 20 with higher scores indicating a greater predis- 125 61 126 formance (overt/covert speech judgements). The external- position to auditory hallucinations. The original scale was 62 127 izing bias model of AVH would predict that, on the reality modified by McCarthy-Jones and Fernyhough (2011) to remove 63 128 64 monitoring task, higher hallucination-proneness should be a question with a low endorsement rate and improve internal 129 65 associated with a greater tendency towards incorrectly reliability, and in testing was found to have satisfactory psy- 130 responding that words spoken by the participant were spoken chometric properties.

4 cortex xxx (2016) 1e11

1 Reality monitoring task e The task was adapted from one Each study trial commenced with a screen indicating 66 2 used previously (Simons et al., 2006; Simons et al., 2008) and whether the subject or experimenter should read aloud the 67 3 68 involved the initial presentation of word-pairs followed by a word-pair. The word-pair was then shown, either complete 4 69 test phase. In the test phase, the participant was asked to (perceived trials) or with only the first letter of the second 5 70 indicate whether a word had earlier been presented within an word provided such that the second word needed to be self- 6 71 ’ 7 intact word-pair using the response ‘Perceived , or had been generated (imagined trials). In both cases the subject or 72 8 presented in a word-pair which had needed to be completed experimenter then had 3 s to read aloud the entire word-pair, 73 9 by imagining the missing word, with the response ‘Imagined’. completing the word-pair as necessary for imagined trials. 74 10 Participants were also required to judge whether the word- Each study phase was followed by its corresponding test 75 11 pair had previously been spoken aloud by themselves (‘Sub- phase, consisting of one sub-block for each of the two reality 76 12 ject’ response) or was spoken by the researcher (‘Experi- monitoring conditions. The sub-blocks commenced with a 77 13 menter’ response). Previously unstudied words were also used question screen indicating which condition was being tested, 78 14 in the test phase, requiring a ‘New’ response. The stimuli i.e., for the Perceived/Imagined condition: ‘Was the accompa- 79 15 80 consisted of 216 well-known word-pairs (e.g., ‘Laurel and nying word Seen or Imagined or New?’, and for the Self/Experi- 16 81 Hardy’,‘Bacon and Eggs’), which were pilot tested for the cur- menter condition: ‘Was the accompanying word said by Self or 17 82 ’ 18 rent study to ensure their familiarity among adults in the Researcher or New? These were then followed by a test screen 83 19 target demographic range. The task comprised 6 separate containing the first word from one of the studied word-pairs, 84 20 study and test blocks, with 24 word-pair stimuli in each study or a new word, together with the instruction to provide the 85 21 phase (six word-pairs presented in four combinations of appropriate response. Participants had 4 sec to make their 86 22 Subject/Experimenter Perceived/Imagined conditions; Fig. 1) response. 87 23 and an additional 12 new words included in the test phase. The order of presentation of sub-blocks in the test phase 88 24 alternated across the six full blocks of the task and was 89 25 counterbalanced across participants. The word-pairs assigned 90 26 91 to the Perceived/Imagined and Subject/Experimenter condi- 27 92 tions, as well as new words, were also counterbalanced across 28 93 participants, and the order of presentation of word-pairs was 29 94 30 pseudo-randomized to ensure no run of more than three 95 31 items of the same condition in any study or test phase. 96 32 97 33 2.1.3. Data analysis 98 34 Old/new recognition accuracy was calculated as the adjusted 99 35 item recognition score (hits minus false alarms, with hits 100 36 being defined as the proportion of words correctly recognised 101 37 as previously seen and false alarms the proportion of new 102 38 103 words incorrectly endorsed as old). Reality monitoring accu- 39 104 racy was calculated as the number of accurate source re- 40 105 41 sponses divided by the number of correct responses 106 42 recognising an item as old. 107 43 Misattribution errors were calculated for perceived and 108 44 imagined trials as the number of responses made for the 109 45 alternative reality monitoring response as a proportion of 110 46 total errors made. So for example, ‘Imagined judged 111 47 Perceived’ errors were calculated as the number of perceived 112 48 responses that were made to imagined trials divided by the 113 49 114 sum of perceived and new responses to imagined trials. This 50 115 gives a measure of misattribution error unrelated to overall 51 116 item recognition accuracy for each condition. The proportion 52 117 53 of internalisation errors (Perceived judged Imagined, or 118 54 Experimenter judged Subject) was then compared to the pro- 119 55 Fig. 1 e Stimuli used in the Reality Monitoring Tasks. Note: portion of externalisation errors (Imagined judged Perceived 120 56 Sample stimuli used in the study phase (left) and test phase or Subject judged Experimenter) for each participant, to give a 121 57 (right) of the reality monitoring task. In a 2 £ 2design, measure of externalizing bias. Eight participants made no 122 58 either the subject or experimenter spoke aloud the stimuli, errors for one or more of the study conditions for the 123 59 which were presented either complete (perceived) or Perceived/Imagined task (6 in the High-LS and 2 in the Low-LS 124 60 incomplete (requiring the second word to be imagined). conditions) and 15 (8 in the High-LS and 7 in the Low-LS 125 61 126 Subjects were then presented at test with the first word of a conditions) for the Subject/Experimenter task; these partici- 62 127 word pair, and asked to judge whether the accompanying pants were excluded from the misattribution bias analysis 63 128 64 word had been seen or imagined, or if the presented word only. 129 65 was new; or whether the subject or experimenter had read Preliminary analyses confirmed the absence of significant 130 aloud the word pair, or the presented word was new. effects of potentially confounding variables of participants'

cortex xxx (2016) 1e11 5

1 age or sex on Old/New recognition, reality monitoring accu- Finally, an analysis of errors was undertaken by calculating 66 2 racy or externalizing bias, all F's(1,44) < 3.291, n.s. a misattribution error rate as a measure independent of reality 67 3 68 monitoring accuracy, to give an indication of the proportion of 4 69 errors that were ascribed to the alternate reality monitoring 5 2.2. Results 70 condition as opposed to a new item (Fig. 2). A mixed ANOVA 6 71 7 There was no difference between the high and low halluci- was undertaken for the analysis of errors on the Perceived/ 72 8 nation proneness groups for Old/New memory, t(45) ¼ .416, Imagined task, with the misattribution error rate as DV, group 73 9 p ¼ .679, d ¼ .115, indicating that the groups had similar as factor, and two within-subject variables of whether the 74 10 recognition memory ability (Table 1). trials has been spoken by subject or experimenter, and 75 11 76 To analyse the reality monitoring data, a mixed ANOVA whether the error direction was internalising (that is, 12 77 with High-LS and Low-LS group as between-subjects factor, Perceived judged Imagined) or externalising (that is, Imagined 13 78 and the reality monitoring condition (Subject/Experimenter or judged Perceived). The analysis revealed no significant group 14 difference for the proportion of misattribution errors made 79 15 Perceived/Imagined) as a within-subject factor, was conduct- 80 overall, F(1, 37) ¼ .051, p ¼ .823, h2 ¼ .001, but with a consistent 16 ed. There was a within subjects effect of task condition: p 81 ¼ < h2 ¼ externalizing bias for both groups, as measured by a greater 17 F(1,45) 64.479, p .001, p .589, indicating that both groups 82 18 were better at judging whether a word-pair had been spoken number of externalisation compared to internalisation errors 83 ¼ < h2 ¼ 19 by the subject or experimenter, compared to distinguishing for each condition: F(1, 37) 59.146, p .001, p .615. This 84 20 whether a word-pair had been perceived or imagined. How- externalising bias was not significantly different for items 85 21 ever, there was no main effect of hallucination-proneness which had been spoken by the subject or the experimenter: 86 ¼ ¼ h2 ¼ 22 ¼ ¼ h2 ¼ F(1,37) 1.276, p .266, p .033. 87 group, F(1,45) .014, p .905, p .000, and no interaction 23 88 between hallucination-proneness group and reality moni- The analysis of variance of misattribution errors in the 24 89 toring condition: F(1,45) ¼ .460, p ¼ .501, h2 ¼ .010, thus giving Perceived/Imagined tasks did reveal a marginal interaction 25 p e 90 no indication of an association between hallucination prone- between group and internal external error direction, F(1, 26 ¼ ¼ h2 ¼ 91 37) 3.838, p .058, p .094, suggesting that participants in 27 ness and reality monitoring ability. 92 the High-LS group might have some tendency to make more 28 To allow a direct comparison with the findings of the 93 externalizing errors than participants in the Low-LS group. 29 similar study by Larøi et al. (2004), the results of the Subject/ 94 30 Experimenter reality monitoring task were then broken down However, when this possibility was tested, there was found to 95 31 for trials which had been spoken by the subject or by the be no overall difference between the groups in either the 96 ¼ 32 experimenter (Table 1). Contrary to the findings of the earlier proportion of externalising errors, [I judged P: t(45) .995, 97 ¼ ¼ 33 study, a mixed ANOVA with Self/Experimenter accuracy as p .326, d .291], or internalising errors, [P judged I: 98 34 ¼ ¼ ¼ 99 DV, group as factor and whether the word-pair had been t(45) .439, p .663, d .127]. Furthermore, the absence of a 35 100 spoken by the subject or experimenter (‘speaker’) as within- three way, group error direction (spoken by subject or 36 ¼ ¼ 101 subjects variable, revealed that while both groups were bet- experimenter) condition interaction, F(1, 37) .654, p .424, 37 h2 ¼ 102 p .017, suggests that any such tendency was not associated 38 ter at the Self/Experimenter discrimination for word-pairs 103 ¼ < with information that had been specifically generated by the 39 spoken by the Experimenter, F(1, 45) 31.744, p .001, 104 h2 ¼ ' subject, as opposed to by the experimenter. 40 p .414, there was no group difference in subjects ability to 105 41 remember that they had previously spoken the word-pair, A similar analysis of variance undertaken for the Self/ 106 42 compared with their memory for experimenter spoken stim- Experimenter task also revealed an externalizing bias for both 107 ¼ < h2 ¼ 43 uli, i.e., no significant group speaker interaction: F(1, groups: F(1, 30) 42.594, p .001, p .587, suggesting that 108 44 ¼ ¼ h2 ¼ participants were more likely to ascribe a word-pair spoken by 109 45) .649, p .425, p .014. 45 themselves to the experimenter than they were a word-pair 110 46 spoken by the experimenter to themselves. This was the 111 47 case regardless of whether the stimulus had earlier been 112 48 Table 1 e Old/new recognition and reality monitoring perceived or imagined at encoding: F(1, 30) ¼ .274, p ¼ .605, 113 Q5 49 accuracy. h2 ¼ 114 p .009. There was no difference between the groups for the 50 115 proportion of misattribution errors made, nor any other sig- 51 Accuracy variable Low-LS High-LS t statistic p 116 (df ¼ 45) nificant main effects or interactions, F(1,30) < 2.231, p > .146, 52 M (SD) M (SD) 117 h2 < 53 p .069. 118 Old/New recognition .85 (.11) .86 (.05) ¡.416 .68 54 The analysis of misattribution errors across both reality 119 Perceived/Imagined .85 (.07) .84 (.08) .385 .70 monitoring conditions therefore gives no evidence of an 55 reality monitoring 120 56 Self/Experimenter .92 (.05) .93 (.05) ¡.319 .75 enhanced externalising bias in individuals with a greater 121 57 reality monitoring proneness to AVH. 122 58 Self/Experimenter: .88 (.08) .90 (.08) .583 .56 123 59 Self generated 124 60 Self/Experimenter: .96 (.04) .96 (.03) .509 .62 3. Experiment 2: internal source monitoring 125 61 Experimenter generated 126 62 127 Note: To aid comparison with the findings of Larøi et al., (2004) the The second experiment used an internal source monitoring 63 128 results of the Self/Experimenter reality monitoring task were paradigm, requiring participants to either read a word-pair to 64 further broken down for trials which had been spoken by the 129 themselves using inner speech (i.e., covert speech), or to read 65 subject or experimenter. 130 a word-pair aloud (overt speech). At a later point, participants

6 cortex xxx (2016) 1e11

1 66 2 67 3 68 4 69 5 70 6 71 7 72 8 73 9 74 10 75 11 76 12 77 13 78 14 79 15 80 Fig. 2 e Misattribution errors. Note: The two charts refer to items misclassified for each of the reality monitoring tasks, 16 81 17 broken down by the trial conditions. So for example, the first 2 bars in the left chart refer to items which had been imagined 82 18 by the subject, which were then incorrectly judged as perceived (an externalisation error), and the last 2 bars in the right 83 19 chart to items which had been imagined by the experimenter during encoding, but which the subject had later judged to 84 20 have been self-imagined (an internalisation error). 85 21 86 22 87 23 were presented with each of the word-pairs again, and were questions with response options 0e4, which was the basis for 88 24 required to recall whether each had been read silently or the questionnaire used in Experiment 1. However, in a later 89 25 aloud. Given that this task required participants to make a corrigendum, McCarthy-Jones and Fernyhough clarified that a 90 26 91 decision between only two options (covert or overt speech) 1e4 scale had actually been used, and this corrected version 27 92 about each word-pair, the data was analyzed using signal was adopted for our Experiment 2. Therefore, although the 28 93 detection theory to investigate both sensitivity and bias on the questionnaires used in our two experiments consisted of 29 94 30 task. It was hypothesized that, if hallucination-proneness is exactly the same questions, the mean scores are not directly 95 31 associated with a general source monitoring deficit, there comparable. 96 32 should be a significant positive correlation between LSHS-R Internal source monitoring task e In contrast to the task 97 33 score and internal source monitoring task performance (task used in Experiment 1, this source memory task, based loosely 98 34 sensitivity). on that used by Franck et al. (2000), asked participants to 99 35 distinguish between two internally generated sources: 100 36 3.1. Method whether verbal stimuli were spoken aloud using overt speech, 101 37 or said silently to themselves using covert (inner) speech. As 102 38 103 3.1.1. Participants with the reality monitoring task, there were two stages to the 39 104 The sample consisted of 125 participants from the staff and task, involving word-pair completion and subsequent recall. 40 105 41 student population of Durham University, UK. One participant Participants were not informed that they would be asked to 106 42 was excluded from further analysis because their task sensi- remember the word-pairs until immediately before the sec- 107 43 tivity (d') on the source monitoring task was <0, indicating ond stage of the task. 108 44 below chance performance, leaving a final sample size of 124 In the word-pair completion stage, participants were pre- 109 45 (number of females ¼ 96, mean age ¼ 20.7 years, SD ¼ 2.5 sented with a series of 72 word-pairs (for example, ‘gold and 110 46 years). Participants all reported being native English speakers, silver’), 36 of which they were instructed to say out loud (‘overt 111 47 and no participants reported any hearing problems. Mean speech’ trials), and 36 of which they were instructed to say to 112 48 score on the LSHS-R was 8.75 (SD ¼ 2.11). When participants themselves using inner speech (‘covert speech’ trials). To 113 49 114 were categorized into high and low hallucination-proneness manipulate the extent to which the stimuli were self- 50 115 groups, there was no difference in age or sex between generated, and in a similar way to the reality monitoring 51 116 groups [age: t(42) ¼ 1.32, p ¼ .195; sex: c2 ¼ 1.91, p ¼ .167]. The task in Experiment 1, within each condition, 18 word-pairs 52 117 53 high hallucination-prone group scored significantly higher on were fully presented to participants on-screen (e.g., ‘gold and 118 ¼ ¼ ’ ’ 54 the LSHS-R (M 11.85, SD .93) than the low hallucination- silver , viz. ‘perceived trials), while the remaining 18 were only 119 55 prone group (M ¼ 5.61, SD ¼ .61); t(42) ¼ 25.06, p < .001. partially completed (e.g., ‘gold and s_____’, viz. ‘imagined’ tri- 120 56 als). For each trial, the participant was asked to say the full 121 57 3.1.2. Design and procedure word-pair (overtly or covertly); thus, half of the trials required 122 58 The auditory items from the revised version of the LSHS-R participants to produce the words themselves (imagined), 123 59 were again used to assess hallucination-proneness (see Sec- whereas half required the participant to read the word-pair 124 60 tion 2.1.2), although in this experiment response options from the screen (perceived). The word-pairs were informally 125 61 126 ranged from 1 to 4 for each question, with total scores thus tested in a small pilot study, to ensure that they were familiar 62 127 able to range from 4 to 20, compared with 0e20 in Experiment to the large majority of participants. Word-pairs were coun- 63 128 64 1. This difference arose due to an error in the reporting of terbalanced across presentation mode (perceived or imag- 129 65 previous literature; McCarthy-Jones and Fernyhough (2011) ined) and condition (overt/covert). For each trial, the 130 described their revised version of the LSHS as comprising instruction (‘Out Loud’ or ‘Inner Speech’) appeared on the

cortex xxx (2016) 1e11 7

1 screen for 1250 msec, followed by the word-pair for 3250 msec, word-pairs) was therefore conducted with both d' and b as 66 2 followed by an inter-trial interval of 750 msec. If the partici- dependent variables. 67 3 68 pant did not know the correct word to complete a pair, they 4 69 were instructed to indicate this with a button press. 5 3.2. Results 70 After the first stage was completed, participants took a 6 71 15 min break, during which they completed the LSHS-R, as 7 There were no significant correlations between proneness to 72 well as various other self-report questionnaires relating to 8 auditory hallucinations and internal source monitoring 73 inner speech phenomenology (Varieties of Inner Speech 9 performance for either of the conditions (perceived or 74 Questionnaire; McCarthy-Jones & Fernyhough, 2011), and 10 imagined) of the internal source monitoring task, or for task 75 11 anxiety and depression (Hospital Anxiety and Depression 76 performance overall, Spearman'srho .124, all ps > .167 (see 12 Scale; Zigmond & Snaith, 1983). These measures were not 77 Table 2). 13 intended to be linked to the source memory task, but were 78 A2 2 mixed model ANOVA with task sensitivity (d') as the 14 included in association with other task-based measures that 79 15 dependent variable (Table 3) showed a main effect of imag- 80 were completed later in the testing session (to be reported ¼ < h2 ¼ 16 ined/perceived status: F(1, 42) 44.27, p .001, p .513; 81 elsewhere; Moseley et al., in preparation). After 15 min, par- 17 sensitivity was greater for imagined word-pairs (M ¼ 1.68, 82 ticipants were then asked to complete the memory test stage 18 SD ¼ .74) compared with those that had been perceived 83 of the task, in which they were presented with the second part 19 (M ¼ 1.04, SD ¼ .61). There was a marginal main effect of 84 ’ 20 of each word-pair (e.g., ‘silver ), in a random order. Participants ¼ ¼ h2 ¼ 85 hallucination-proneness: F(1, 42) 3.36, p .074, p .074. were asked to try to judge whether they had previously said 21 There was no significant interaction between task condition 86 22 each word out loud or using inner speech, responding with a 87 (Perceived/Imagined) and hallucination-proneness (high/low): 23 button press. This test phase was self-timed with each word ¼ ¼ h2 ¼ 88 F(1, 42) .03, p .862, p .001. 24 presented in the centre of the computer screen until a 89 There was no difference in b between the task conditions, 25 90 response was made. ¼ ¼ h2 ¼ 26 F(1, 42) 1.10, p .299, p .026, or hallucination-proneness 91 ¼ ¼ h2 < 27 groups, F(1, 42) .017, p .896, p .001, and no significant 92 3.1.3. Data analysis 28 interaction between task condition (Perceived/Imagined) and 93 Signal detection measures were used to analyze data from the 29 hallucination-proneness (high/low): F(1, 42) ¼ .073, p ¼ .788, 94 internal source monitoring task, as recommended by h2 ¼ 30 p .002. Thus, the experiment indicated no significant dif- 95 McKague et al. (2012). ‘Hits’ were classified as overtly spoken 31 ferences between hallucination-proneness groups on any 96 words correctly recalled as such, whilst false alarms were 32 measure of internal source monitoring. 97 33 classified as covertly spoken words incorrectly recalled as 98 34 overtly spoken (‘Miss’ and ‘correct rejection’ responses are not 99 35 reported here, since they are, necessarily, directly propor- 100 36 tional to hit and false alarm rates). d', a measure of task 4. General discussion 101 37 sensitivity, was calculated as the standardised hit rate (z-score 102 38 103 of hit rate) minus the standardised false alarm rate (z-score of The two experiments presented above addressed a key pre- 39 104 false alarm rate), with a lower value indicating less ability to diction of continuum models of psychosis: whether source 40 monitoring impairments are associated with hallucination- 105 41 distinguish the source of words. The second dependent vari- 106 b proneness in the non-clinical population, as they are in peo- 42 able was , a measure of response bias, which was calculated 107 ple with clinical diagnoses who hallucinate. Experiment 1 43 as outlined by Stanislaw and Todorov (1999) (with a lower 108 found no difference in accuracy either for Old/New recogni- 44 value indicating a lower criterion for deciding that a word was 109 45 spoken aloud). tion, or for Perceived/Imagined or Subject/Experimenter re- 110 h2 46 In contrast to Experiment 1, where an initial group split on ality monitoring judgments, with effect sizes so low ( p .02) 111 47 hallucination-proneness was used to invite participants for as to preclude a sample size explanation. While there was 112 48 behavioural testing, all participants in Experiment 2 113 49 114 completed both the LSHS-R self-report questionnaire and the 50 115 internal source monitoring task. Therefore, we first computed Table 2 e Correlations between internal source monitoring 51 116 correlations (Spearman's rho) between proneness to auditory task performance and auditory hallucination-proneness. 52 117 53 hallucinations and internal source monitoring performance SMT measure Spearman's rho 118 54 (sensitivity and response bias, for the imagined and perceived 119 d' (overall) .114 conditions). To enable comparison with Experiment 1, we also 55 b (overall) .001 120 56 split participants into high hallucination-proneness (those d' (imagined .076 121 ¼ 57 scoring in the upper quartile on the LSHS-R; N 26) and low b (imagined) .010 122 58 hallucination-proneness (scoring in the lower quartile on the d' (perceived) .124 123 59 LSHS-R; N ¼ 18) groups, and compared performance on the b (perceived) .001 124 60 125 source monitoring task between these groups. There was a Note: d' ¼ task sensitivity; b ¼ task response bias. Higher d' mea- 61 126 significant difference between the high and low groups in sures correspond to greater ability to distinguish between overtly 62 127 hallucination-proneness: t(42) ¼ 25.06, p < .001, as expected. A and covertly spoken words. Higher b values correspond to a more 63 128 2 2 mixed model ANOVA with a between-subjects variable conservative criterion for deciding a word was spoken overtly. 64 None of the correlations were significant at p < .05, even before 129 of hallucination-proneness group (high/low) and a within- 65 correction for multiple comparisons. 130 subjects variable of task condition (imagined/perceived

8 cortex xxx (2016) 1e11

1 66 e AVH. Indeed we have demonstrated reality monitoring 2 Table 3 Group performance on the internal source 67 monitoring task. impairment and dysfunction in the medial anterior PFC in 3 68 patients with schizophrenia using a very similar version of the 4 69 Source monitoring measure Hallucination-proneness task to that used in Experiment 1 (Garrison et al., in revision). 5 70 Low High As such, the results are inconsistent with continuum models 6 71 & 7 d' (imagined) 1.88 (.69) 1.61 (.66) of psychosis (van Os Reininghaus, 2016), which assert that 72 8 d' (perceived) 1.16 (.67) .85 (.53) hallucinations are distributed throughout the general popu- 73 9 b (imagined) 1.58 (1.14) 1.62 (1.36) lation, and thus which predict comparable effects in healthy 74 b 10 (perceived) 1.93 (1.32) 1.83 (1.08) individuals who are prone to hallucinations as in patients 75 11 Note: d' ¼ task sensitivity; b ¼ task response bias. Measures shown with schizophrenia who hallucinate. However, there are a 76 12 are mean scores, with SD in parentheses. number of alternative possible explanations for the apparent 77 13 discrepancy in findings relating to the association between 78 14 clear evidence of a general externalizing bias in both reality source monitoring impairment and AVH in clinical and non- 79 15 80 monitoring conditions, consistent with that previously re- clinical groups. 16 81 ported from studies in the healthy population (Johnson et al., Firstly, it is possible that the assessment of hallucination- 17 82 18 1981), this was not found to be significantly enhanced in proneness used in the present experiments was ineffective 83 ' 19 participants in the high hallucination-proneness group. in measuring individuals proneness to AVH in the non- 84 20 A marginal interaction (p ¼ .058) between group and clinical population. However, while the LSHS-R comprises 85 21 internaleexternal error direction in the Perceived/Imagined only five questions which ask about unusual auditory expe- 86 22 reality monitoring task suggested that there might be a ten- riences, the scale in its revised form has been well tested and 87 23 dency for high hallucination-prone individuals to judge a found to have satisfactory psychometric properties (see 88 24 greater proportion of imagined word-pairs as perceived than McCarthy-Jones & Fernyhough, 2011). 89 25 perceived words as imagined. However, this possibility was Alternatively, there may be only some shared overlap of 90 26 91 not supported by analysis of the simple effects, and if such a the mechanisms involved in clinical and non-clinical hallu- 27 92 tendency is related to auditory hallucinations, then it might be cinatory experiences (as suggested by Badcock & Hugdahl, 28 93 expected to be specific to items that were generated by the 2012), which might be especially true for the hallucinatory 29 94 30 subject, which was not found to be the case. Furthermore, experiences reported by the samples employed here. Larøi 95 31 there was no evidence of an enhanced externalizing bias in (2012) proposed a (fuzzy) distinction between participants in 96 32 the Subject/Experimenter reality monitoring task for the high studies of non-clinical hallucinations research, referring to 97 33 hallucination-prone group compared to the low these as Type i non-patients and Type ii non-patients. Par- 98 34 hallucination-prone group. ticipants recruited in the current two experiments would be 99 35 The results from Experiment 1 were supported by the classed as Type i non-patients, who typically report infre- 100 36 findings of Experiment 2, which investigated participants' quent hallucinatory experiences that may be phenomeno- 101 37 ability to discriminate between whether they had overtly or logically distinct from the AVH reported by patients (e.g., brief 102 38 103 covertly spoken a word-pair, which had been either perceived experiences that rarely take the form of complex utterances). 39 104 or imagined during the encoding phase. No significant corre- In contrast, Type ii non-patients often report relatively 40 105 41 lations were found between task sensitivity or response bias frequent hallucinations that are phenomenologically more 106 42 and hallucination-proneness, regardless of whether the similar to the AVH reported by patients, except in terms of 107 43 stimuli had previously been perceived or imagined. Further- emotional valence and perceived controllability (Johns et al., 108 44 more, no differences were detected in task sensitivity and 2014). Thus, a reality monitoring impairment may not be 109 45 response bias between groups of participants split by involved in the hallucinatory experiences reported by Type i 110 46 hallucination-proneness as in Experiment 1. There was a non-patients, but may be involved in the ‘full blown’ AVH 111 47 marginal reduction in overall task sensitivity for the higher reported by Type ii non-patients as well as by patients. 112 48 hallucination-proneness group (p ¼ .074), but this was not A further question relates to how reality monitoring 113 49 114 supported by a significant correlation between task sensitivity impairment might be implicated in the generation of hallu- 50 115 and hallucination-proneness score across the entire sample. cinations. Reality monitoring is defined as a mnemonic ability, 51 116 Indeed, the correlation effect sizes were so small (rho .124) but the cognitive operations involved in monitoring the origin 52 117 53 that statistical power is again unlikely to be an explanation, of retrieved information might overlap with those that 118 & 54 and there was no interaction for task sensitivity between monitor the origin of real time information (Johnson Raye, 119 55 LSHS-R group and the perceived or imagined status of the 1981; discussed in; Garrison et al., 2016), with impairments 120 56 stimuli, as might be expected if the deficit related to the in these operations leading to the generation of hallucina- 121 57 inability to distinguish the source of imagined information tions. However, other mechanisms have been proposed to 122 58 from that which had been perceived. explain the failure to correctly identify the origin of self- 123 59 The results of these two experiments thus offer little or no generated information, which might manifest differently 124 60 support for a deficit in source monitoring ability, and/or of across different groups of individuals. For example, AVH may 125 61 126 enhanced externalizing biases in hallucination-prone in- arise from enhanced perceptual content of self-generated 62 127 dividuals in the healthy population. These findings contrast auditory information due to over-activation of secondary as- 63 128 64 with the findings of behavioral and neuroimaging studies in sociation speech and language cortices, such as the voice- 129 65 patients with schizophrenia, which report associations be- selective auditory regions in the STG (Allen, Larøi, McGuire, 130 tween reality monitoring impairment and the presence of & Aleman, 2008). If activation of these brain areas results in

cortex xxx (2016) 1e11 9

1 unusually vivid internal auditory imagery, such information that I left my body temporarily”). In contrast, the present study 66 2 could be erroneously recognised as external in origin, without focused solely on auditory hallucinations. Using the same task 67 3 68 any deficit in normal source monitoring activity. Consistent across clinical and non-clinical groups, together with tighter 4 69 with this suggestion, speech and language areas are active in control of confounding variables such as participants' age, 5 70 addition to anterior regions such as cingulate cortices during language skills or the presence of general memory deficits, 6 71 7 hallucinations (Zmigrod et al., 2016), and a neuroimaging should help address variation across studies going forward. 72 8 study in healthy individuals indicated the presence of inter- We remain a long way from understanding the brain 73 9 mittent episodes of significantly increased activity in bilateral mechanisms underpinning hallucinations, with many exist- 74 10 primary and secondary auditory cortices, together with ing theoretical models of AVH failing to address the 75 11 associated activations in the anterior cingulate cortex, even complexity and diversity of hallucinations (for example, their 76 12 during periods of silence (Hunter et al., 2006). differing developmental trajectories, or complex interactions 77 13 Alternatively, there may be a distinct impairment in the with the individual). Understanding whether a single theo- 78 14 self-monitoring processes which predict the sensory conse- retical model can be applied to clinical and non-clinical hal- 79 15 80 quences of actions through comparator forward modelling/ lucinations will depend on the flexibility of the framework to 16 81 efference copy mechanisms (Feinberg, 1978). Self-monitoring account for how factors implicated in the generation of these 17 82 18 accounts of reality testing suggest that AVH arise from a perceptual anomalies might interact to explain differences in 83 19 disruption in the capacity to monitor the intention to pro- phenomenological experience between groups, as well as the 84 20 duce inner speech (or other cognitions), resulting in it being variety in experience for a single individual. If such a 85 21 erroneously marked as external (Seal, Aleman, & McGuire, framework can be developed, this would map most closely to 86 22 2004). Such accounts thus provide an explanation for the quasi-dimensional models of schizotypy (Yung et al., 2005), 87 23 external directionality of errors without the need for any which allow for discontinuities in the experience of psychosis 88 24 deficit in a separate source monitoring process, as informa- across the population consistent with separable phenotypic 89 25 tion is assumed to be externally perceived in the absence of expressions of associated factors (Meehl, 1989; 1962), rather 90 26 91 an efference copy signal. However, while there is strong ev- than a fully-dimensional model (Claridge, 1972, 1994) more 27 92 idence for self-recognition deficits in patients with schizo- supportive of an unbroken continuum. Although the work in 28 93 phrenia relating to motor action (Blakemore, Wolpert, & the present study does not support the role of reality- 29 94 & 30 Frith, 2002; Frith, Blakemore, Wolpert, 2000), and some monitoring ability as a factor in non-clinical hallucination- 95 31 support for corollary discharge dysfunction in schizophrenia proneness, this does not rule out more complex models 96 32 (Ford & Mathalon, 2004; Ford et al., 2001), direct evidence for a invoking reality monitoring as an important factor in the 97 33 specific comparator model relating to inner/covert speech or transition between hallucination-proneness and more 98 34 auditory imagery is lacking. Furthermore, theoretical argu- frequent hallucinatory experiences, in clinical or non-clinical 99 35 ments have been raised against the idea that the generation populations. 100 36 of thought has the same physiological consequences as that 101 37 of motor action, with Gallagher (2004) arguing that the 102 38 103 self-monitoring account of hallucinations fails in applying 39 Uncited reference 104 an explanation of motor function to one of cognitive 40 105 experience. 41 Larøi et al., 2012. Q6 106 42 Finally, it should be noted that the findings of the reality 107 43 monitoring study in Experiment 1 are in contrast to those of 108 44 Larøi et al. (2004), who reported significant differences be- 109 45 tween low and high hallucination-prone healthy individuals Acknowledgements 110 46 in the accuracy of self-generated, but not experimenter- 111 47 generated, stimuli. The discrepancy appears not to be JRG was supported by a University of Cambridge Behavioural 112 48 explained by the use of non-parametric statistics to analyze and Clinical Neuroscience Institute studentship, funded by a 113 49 114 the results in the Larøi et al. study, as similar non-parametric joint award from the UK Medical Research Council and the 50 115 analysis of our experiments did not alter our findings. What is Wellcome Trust. PM, BA-D, DS, and CF were supported by a 51 116 clear, however, is that the investigation of reality monitoring Wellcome Trust Collaborative award. JSS was supported by a 52 117 James S. McDonnell Foundation Scholar award. 53 and source monitoring deficits in clinical studies has also Q4 118 54 produced a range of varying results (see Brookwell et al., 2013). 119 55 These might be explained by a wide variation in task design, 120 56 with some tasks using verbal stimuli and others using per- references 121 57 formed or perceived actions, and with some tasks using only 122 58 the Self/Experimenter or the Perceived/Imagined discrimina- 123 59 tions separately. This may be the explanation for the Achim, A. M., & Weiss, A. P. (2008). No evidence for a differential 124 60 discrepancy in findings with the Larøi et al. study, which used deficit of reality monitoring in schizophrenia: A meta-analysis 125 61 126 a Self/Experimenter paradigm but with stimuli varying in of the associative memory literature. Cognitive Neuropsychiatry, 62 13, 369e384. 127 emotional valence and cognitive effort. Furthermore, Larøi 63 Allen, P., Larøi, F., McGuire, P. K., & Aleman, A. (2008). The 128 64 et al. used a version of the LSHS consisting of 16 questions, hallucinating brain: A review of structural and functional 129 65 many of which seem only indirectly related to hallucination- neuroimaging studies of hallucinations. Neuroscience and 130 proneness (e.g., “I have had a sensation of floating or falling, or Biobehavioral Reviews, 32(1), 175e191.

10 cortex xxx (2016) 1e11

1 Anselmetti, S., Cavallaro, R., Bechi, M., Angelone, S. M., Ermoli, E., Johnson, M. K., Raye, C. L., Foley, H. J., & Foley, M. A. (1981). 66 2 Cocchi, F., et al. (2007). Psychopathological and Cognitive operations and decision bias in reality monitoring. 67 3 neuropsychological correlates of source monitoring American Journal of Psychology, 94(1), 37e64. 68 4 impairment in schizophrenia. Psychiatry Research, 150(1), Kra˚kvik, B., Larøi, F., Kalhovde, A. M., Hugdahl, K., Kompus, K., 69 5 51e59. Salvesen, Ø., et al. (2015). Prevalence of auditory verbal 70 6 Bentall, R. P. (1990). The illusion of reality: A review and hallucinations in a general population: A group comparison 71 7 integration of psychological research on hallucinations. study. Scandinavian Journal of Psychology, 56(5), 508e515. 72 8 Psychological Bulletin, 107(1), 82e95. Larøi, F., Sommer, I. E., Blom, J. D., Fernyhough, C., Hugdahl, K., 73 9 Bentall, R. P., & Slade, P. D. (1985). Reliability of a scale measuring Johns, L. C., & Stephane, M. (2012). The characteristic features 74 10 disposition towards hallucination: A brief report. Personality of auditory verbal hallucinations in clinical and nonclinical 75 e 11 and Individual Differences, 6(4), 527 529. groups: State-of-the-art overview and future directions. 76 e 12 Blakemore, S. J., Wolpert, D. M., & Frith, C. D. (2002). Schizophrenia Bulletin, 38(4), 724 733. 77 13 Abnormalities in the awareness of action. Trends in Cognitive Laroi, F., Van der Linden, M., & Marczewski, P. (2004). The effects 78 Sciences, 6(6), 237e242. of emotional salience, cognitive effort and meta-cognitive 14 79 Brebion, G., Amador, X., David, A., Malaspina, D., Sharif, Z., & beliefs on a reality monitoring task in hallucination-prone 15 80 Gorman, J. M. (2000). Positive symptomatology and source- subjects. The British Journal of Clinical Psychology, 43(3), 16 81 monitoring failure in schizophrenia e An analysis of 221e233. 17 82 symptom-speciﬁc effects. Psychiatry Research, 95, 119e131. McCarthy-Jones, S., & Fernyhough, C. (2011). The varieties of 18 Brookwell, M. L., Bentall, R. P., & Varese, F. (2013). Externalizing inner speech: Links between quality of inner speech and 83 19 biases and hallucinations in source-monitoring, self- psychopathological variables in a sample of young adults. 84 20 monitoring and signal detection studies: A meta-analytic Consciousness and Cognition, 20(4), 1586e1593. 85 21 review. Psychological Medicine, 43(12), 2465e2475. McKague, M., McAnally, K. I., Skovron, M., Bendall, S., & 86 22 Buda, M., Fornito, A., Bergstrom, Z. M., & Simons, J. S. (2011). A Jackson, H. J. (2012). Source monitoring and proneness to 87 23 speciﬁc brain structural basis for individual differences in auditory-verbal hallucinations: A signal detection analysis. 88 24 reality monitoring. Journal of Neuroscience, 31(40), 14308e14313. Cognitive Neuropsychiatry, 17(6), 544e562. 89 25 Claridge, G. (1972). The schizophrenias as nervous types. British Meehl, P. E. (1962). Schizotaxia, schizotypy, schizophrenia. 90 26 Journal of Psychiatry, 121,1e17. American Psychologist, 17(12), 827e838. 91 27 Claridge, G. (1994). Single indicator of risk for schizophrenia: Meehl, P. E. (1989). Schizotaxia revisited. Archives of General 92 28 Probable fact or likely myth? Schizophrenia Bulletin, 20(1), Psychiatry, 46(10), 935e944. 93 29 151e168. Morrison, A. P., Wells, A., & Nothard, S. (2000). Cognitive factors in 94 30 Feinberg, I. (1978). Efference copy and corollary discharge: predisposition to auditory and visual hallucinations. The 95 e 31 Implications for thinking and its disorders. Schizophrenia British Journal of Clinical Psychology, 39(1), 67 78. 96 e 32 Bulletin, 4(4), 636 640. Moseley, P., Fernyhough, C., & Ellison, A. (2013). Auditory verbal 97 33 Ford, J. M., & Mathalon, D. H. (2004). Electrophysiological evidence hallucinations as atypical inner speech monitoring, and the 98 of corollary discharge dysfunction in schizophrenia during potential of neurostimulation as a treatment option. 34 99 talking and thinking. Journal of Psychiatric Research, 38(1), Neuroscience and Biobehavioral Reviews, 37(10), 2794e2805. 35 100 37e46. van Os, J., & Reininghaus, U. (2016). Psychosis as a transdiagnostic 36 101 Ford, J. M., Mathalon, D. H., Heinks, T., Kalba, S., Faustman, W. O., and extended phenotype in the general population. World 37 102 & Roth, W. T. (2001). Neurophysiological evidence of corollary Psychiatry, 15(2), 118e124. 38 discharge dysfunction in schizophrenia. American Journal of Seal, M. L., Aleman, A., & McGuire, P. K. (2004). Compelling 103 39 Psychiatry, 158(12), 2069e2071. imagery, unanticipated speech and deceptive memory: 104 40 Franck, N., Rouby, P., Daprati, E., Dale, J., Marie-cardine, M., & Neurocognitive models of auditory verbal hallucinations in 105 41 Georgie, N. (2000). Confusion between silent and overt reading schizophrenia. Cognitive Neuropsychiatry, 9(1), 43e72. 106 42 in schizophrenia. Schizophrenia Research, 41, 357e364. Simons, J. S., Davis, S. W., Gilbert, S. J., Frith, C. D., & 107 43 Frith, C. D. (1992). The cognitive neuropsychology of schizophrenia. Burgess, P. W. (2006). Discriminating imagined from perceived 108 44 Hove: Lawrence Erlbaum Associates. information engages brain areas implicated in schizophrenia. 109 45 Frith, C. D., Blakemore, S., & Wolpert, D. M. (2000). Explaining the NeuroImage, 32(2), 696e703. 110 46 symptoms of schizophrenia: Abnormalities in the awareness Simons, J., Henson, R., Gilbert, S., & Fletcher, P. (2008). Separable 111 47 of action. Brain Research Reviews, 31, 357e363. forms of reality monitoring supported by anterior prefrontal 112 48 Gallagher, S. (2004). Neurocognitive models of schizophrenia: A cortex. Journal of Cognitive Neuroscience, 20(3), 447e457. 113 49 neurophenomenological critique. Psychopathology, 37,8e19. Stanislaw, H., & Todorov, N. (1999). Calculation of signal detection 114 & 50 Garrison, J. R., Bond, R., Gibbard, E., Johnson, M. K., & Simons, J. S. theory measures. Behavior Research Methods, Instruments, 115 e 51 (2016). Monitoring what is real: The effects of modality and Computers, 31(1), 137 149. 116 52 action on accuracy and type of reality monitoring error. Cortex, Stephane, M., Kuskowski, M., McClannahan, K., Surerus, C., & 117 e 53 1 10. Nelson, K. (2010). Evaluation of speech misattribution bias in 118 Garrison, J. R., Fernyhough, C., McCarthy-Jones, S., Haggard, M., & schizophrenia. Psychological Medicine, 40(5), 741e748. 54 119 Simons, J. S. (2015). Paracingulate sulcus morphology is Subramaniam, K., Luks, T. L., Fisher, M., Simpson, G. V., 55 120 associated with hallucinations in the human brain. Nature Nagarajan, S., & Vinogradov, S. (2012). Computerized cognitive 56 121 Communications, 6, 8956. training restores neural activity within the reality monitoring 57 122 Hunter, M. D., Eickhoff, S. B., Miller, T. W. R., Farrow, T. F. D., network in schizophrenia. Neuron, 73(4), 842e853. 58 Wilkinson, I. D., & Woodruff, P. W. R. (2006). Neural activity in Sugimori, E., Mitchell, K. J., Raye, C. L., Greene, E. J., & 123 59 speech-sensitive auditory cortex during silence. Proceedings of Johnson, M. K. (2014). Brain mechanisms underlying reality 124 60 the National Academy of Sciences of the United States of America, monitoring for heard and imagined words. Psychological 125 61 103(1), 189e194. Science, 25(2), 403e413. 126 62 Johnson, M. K., Hashtroudi, S., & Lindsay, D. S. (1993). Source Turner, M. S., Simons, J. S., Gilbert, S. J., Frith, C. D., & 127 63 monitoring. Psychological Bulletin, 114(1), 3e28. Burgess, P. W. (2008). Distinct roles for lateral and medial 128 64 Johnson, M. K., & Raye, C. L. (1981). Reality monitoring. rostral prefrontal cortex in source monitoring of perceived 129 65 Psychological Review, 88(1), 67e85. and imagined events. Neuropsychologia, 46(5), 1442e1453. 130

cortex xxx (2016) 1e11 11

1 Vinogradov, S., Luks, T. L., Schulman, B. J., & Simpson, G. V. Australian and New Zealand Pournal of Psychiatry, 39(11e12), 10 2 (2008). Deﬁcit in a neural correlate of reality monitoring in 964e971. 11 3 schizophrenia patients. Cerebral Cortex, 18(11), 2532e2539. Zigmond, A. S., & Snaith, R. P. (1983). The hospital anxiety and 12 4 Waters, F. A. V., Maybery, M. T., Badcock, J. C., & Michie, P. T. depression scale. Acta Psychiatrica Scandinavica, 67(6), 361e370. 13 5 (2004). Context memory and binding in schizophrenia. Zmigrod, L., Garrison, J. R., Carr, J., & Simons, J. S. (2016). The 14 6 Schizophrenia Research, 68(2e3), 119e125. neural correlates of hallucinations: A quantitative meta- 15 7 Yung, A. R., Yuen, H. P., McGorry, P. D., Phillips, L. J., Kelly, D., analysis of neuroimaging studies. Neuroscience & Biobehavioural 16 8 Dell’Olio, M., et al. (2005). Mapping the onset of psychosis: The Reviews, 69, 113e123. 17 9 comprehensive assessment of at-risk mental states. The 18