Introduction
Auditory verbal hallucinations (AVH), experiences of hearing voices without external stimuli, affect 75% of patients with schizophrenia (Waters & Fernyhough, Reference Waters and Fernyhough2017). Within this group, 25–30% of individuals experienced persistent AVH despite antipsychotic medication (Kane, Honigfeld, Singer, & Meltzer, Reference Kane, Honigfeld, Singer and Meltzer1988; Meltzer, Reference Meltzer1992) and 75% of voice hearers experienced a high level of distress associated with their voices (Copolov, Mackinnon, & Trauer, Reference Copolov, Mackinnon and Trauer2004). Patients with schizophrenia also commonly experienced anxiety and depression, with reported rates of 45% (Badcock, Paulik, & Maybery, Reference Badcock, Paulik and Maybery2011; Cosoff & Hafner, Reference Cosoff and Hafner1998; Leff, Tress, & Edwards, Reference Leff, Tress and Edwards1988; Lysaker & Salyers, Reference Lysaker and Salyers2007; Turnbull & Bebbington, Reference Turnbull and Bebbington2001). There is cross-sectional evidence for more frequent and more severe AVH when patients also suffered from depression (Smith et al., Reference Smith, Fowler, Freeman, Bebbington, Bashforth, Garety and Kuipers2006; Soppitt & Birchwood, Reference Soppitt and Birchwood1997).
It has been proposed that negative emotions are not merely consequences of AVH, but also antecedents of AVH (Allen & Agus, Reference Allen and Agus1968; Slade, Reference Slade1976). In a semi-structured interview, patients with AVH recalled experiencing sadness, fear, and anger prior to hallucinations (Nayani & David, Reference Nayani and David1996). A number of longitudinal studies investigating onset and relapse of psychosis also pointed toward a pivotal role of negative emotions in the emergence and exacerbations of positive symptoms (including hallucinations) (see review by Hartley, Barrowclough, & Haddock, Reference Hartley, Barrowclough and Haddock2013). In the prodromal phase of psychosis, there is robust evidence for emotional disturbances to precede transition into psychosis in the majority of cases (see reviews by Birchwood, Macmillan, & Smith, Reference Birchwood, Macmillan, Smith, Birchwood and Tarrier1992; Broome et al., Reference Broome, Woolley, Tabraham, Johns, Bramon, Murray and Murray2005; Yung & McGorry, Reference Yung and McGorry1996).
The temporal relationship between AVH and emotions has been investigated mostly by retrospective methods of measurement, which may be subject to recall bias (Shiffman, Stone, & Hufford, Reference Shiffman, Stone and Hufford2008). Experience sampling methodology (ESM) is a structured diary in the respondents’ daily living environment that captures subjective moment-to-moment experiences (see reviews by Fahrenberg, Myrtek, Pawlik, & Perrez, Reference Fahrenberg, Myrtek, Pawlik and Perrez2007; Myin-Germeys et al., Reference Myin-Germeys, Oorschot, Collip, Lataster, Delespaul and van Os2009, Reference Myin-Germeys, Kasanova, Vaessen, Vachon, Kirtley, Viechtbauer and Reininghaus2018; Scollon, Prieto, & Diener, Reference Scollon, Prieto, Diener and Diener2009). As AVH are discrete phenomena with identifiable points of onset and termination (Oorschot et al., Reference Oorschot, Lataster, Thewissen, Bentall, Delespaul and Myin-Germeys2012) and varying intensity over hours and days (Delespaul, Devries, & van Os, Reference Delespaul, Devries and van Os2002), ESM provides a feasible and reliable tool to address the temporality of AVH and associated mechanisms. Using ESM, Peters et al. (Reference Peters, Lataster, Greenwood, Kuipers, Scott, Williams and Myin-Germeys2012a) found that hallucinatory intensity was associated with an increased level of negative affect (NA) and a decreased level of positive affect (PA) within the same moment. An across-moments approach of analysis, however, revealed inconsistent results. Delespaul et al. (Reference Delespaul, Devries and van Os2002) found an elevated anxiety level before the onset of hallucinations, which subsided at the end of the hallucinatory episode. Oorschot et al. (Reference Oorschot, Lataster, Thewissen, Bentall, Delespaul and Myin-Germeys2012), in contrast, reported no change in NA before or after occurrence of AVH. More data are needed to clarify the temporal interplay between AVH and affect.
While earlier research suggested that voice hearers’ behavioral and emotional responses directly arise from the experience of AVH per se (e.g. Benjamin, Reference Benjamin1989), recent developments in cognitive models of AVH posit that behavioral and emotional consequences may be exacerbated by the beliefs that voice hearers held about the voices (Birchwood & Chadwick, Reference Birchwood and Chadwick1997; Chadwick & Birchwood, Reference Chadwick and Birchwood1994, Reference Chadwick and Birchwood1995; Morrison & Renton, Reference Morrison and Renton2001). Specifically, voices appraised to be evil (malevolent) and powerful (omnipotent) provoke distress and are resisted, whereas voices perceived to have good intentions (benevolent) lead to PA and are engaged (Birchwood & Chadwick, Reference Birchwood and Chadwick1997; So & Wong, Reference So and Wong2008). While the association between beliefs about voices and emotional distress has been corroborated across studies (Andrew, Gray, & Snowden, Reference Andrew, Gray and Snowden2008; Mawson, Cohen, & Berry, Reference Mawson, Cohen and Berry2010; Peters, Williams, Cooke, & Kuipers, Reference Peters, Williams, Cooke and Kuipers2012b; van der Gaag, Hageman, & Birchwood, Reference van der Gaag, Hageman and Birchwood2003), these studies focused on level of emotional distress rather than change in distress in response to occurrence of AVH. Whether beliefs about voices moderate the temporal associations between hallucinations and NA in real time has not been formally tested. For example, it is possible that individuals who believe their voices to be malevolent may become subsequently more distressed, whereas individuals who believe their voices to be benevolent may choose to initiate a hallucinatory experience when they are upset. Examining the influence of beliefs on the temporal relationship between voices and distress will provide insights on how beliefs, emotions, and hallucinations interact, thereby consolidating the cognitive model of psychosis.
The present study aimed to elucidate the moment-to-moment associations between the intensity of AVH and the level of NA, and the effect of baseline negative beliefs about voices on the moment-to-moment associations, using ESM among patients experiencing frequent AVH.
The key hypotheses were as follows:
(1) Momentary level of NA will predict the intensity of AVH at the following time point
(2) Momentary intensity of AVH will predict the level of NA at the following time point
(3) Baseline beliefs about voices (malevolence, omnipotence, and benevolence) will moderate the moment-to-moment relationships between AVH and NA
Methods
Participants
Ethical approval for this study was granted by the Joint Chinese University of Hong Kong – New Territories East Cluster Clinical Research Ethics Committee (2015.685-T), Kowloon West Cluster Research Ethics Committee [KW/FR-15-216(94-12)] and New Territories West Cluster Research Ethics Committee (NTWC/CREC/17028). Inclusion criteria were: psychiatric outpatients with present AVH [scoring 3 or above on item P3 hallucinatory behavior of the Positive and Negative Syndrome Scale (PANSS; Kay, Fiszbein, & Opler, Reference Kay, Fiszbein and Opler1987)], a diagnosis of schizophrenia spectrum disorder (based on the SCID interview), and aged 18 or above. Exclusion criteria were patients with drug-induced or organic psychosis, a primary diagnosis of substance misuse, and intellectual disability.
Measures
Clinical interview
The Chinese-bilingual Structured Clinical Interview for DSM-IV Axis I Disorders (CB-SCID-I/P; So et al., Reference So, Kam, Leung, Chung, Liu and Fong2003) is a semi-structured interview-based instrument for obtaining reliable DSM-IV Axis I disorders diagnoses. Good inter-rater and test–retest reliabilities have been reported (So et al., Reference So, Kam, Leung, Chung, Liu and Fong2003).
Severity of schizophrenia symptoms and hallucinations
Severity of schizophrenia symptoms was assessed by using the PANSS (Kay et al., Reference Kay, Fiszbein and Opler1987), with a total score ranging from 30 to 210. Severity of hallucinations was assessed by using the hallucination sub-scales of Scale of Assessment for Positive Symptoms (SAPS; Andreasen, Reference Andreasen1984) and the Psychotic Symptom Rating Scales (PSYRATS; Haddock, McCarron, Tarrier, & Faragher, Reference Haddock, McCarron, Tarrier and Faragher1999).
Depressive symptoms
Calgary Depression Scale for Schizophrenia (CDSS; Addington, Addington, Maticka-Tyndale, & Joyce, Reference Addington, Addington, Maticka-Tyndale and Joyce1992) is a nine-item, four-point (0–3) rating semi-structured interview scale for assessing depressive symptoms in patients with schizophrenia. Each item is anchored by descriptors. CDSS total score ranges from 0 to 27. High internal consistency and strong correlation with other established measures of depression have been reported (Addington et al., Reference Addington, Addington, Maticka-Tyndale and Joyce1992).
Anxiety symptoms
The Chinese version of the Beck Anxiety Inventory (BAI; Beck, Epstein, Brown, & Steer, Reference Beck, Epstein, Brown and Steer1988) is a 21-item, four-point (0–3) rating self-report questionnaire that measures anxiety severity and unique symptom features that are independent of that in depression, including somatic symptoms and subjective experience of panic and anxiety. Good reliability and validity of the Chinese version have been reported (Cheng et al., Reference Cheng, Wong, Wong, Chong, Wong, Chang and Wu2002).
Experience sampling method
ESM, a structured self-assessment diary method, was used to measure momentary levels of affect and symptoms (see reviews above). A mobile application was programmed to display the same ESM questionnaire for 10 times a day over 6 days, using a signal-contingent protocol in parallel to our previous ESM studies (see Chan et al., Reference Chan, So, Mak, Siah, Chan and Wu2019; Leung et al., Reference Leung, So, Kwok, Ng, Chan, Lo and Lee2019; So, Peters, Swendsen, Garety, & Kapur, Reference So, Peters, Swendsen, Garety and Kapur2013, Reference So, Peters, Swendsen, Garety and Kapur2014 for more details). The application evoked a beep signal at a random time of the day to prompt participants to complete a new ESM entry. Each entry was at least 30 min apart. Care was taken to ensure that the signals would occur within the waking hours of the participants, with 10–12 target hours each day. The questionnaire could only be activated within 15 min after the signal to minimize any preemptive responses, and would otherwise be inaccessible by the participants. The questions were displayed one at a time, and each response was time-stamped. Participants were told to respond as soon as possible upon a signal.
The ESM questionnaire included items on NA and PA, psychotic symptoms, and current activity and environment. NA items measured subjective feelings of ‘irritated’, ‘low’, and ‘tense’, while PA items measured subjective feelings of ‘cheerful’, ‘relaxed’, and ‘content’ (Myin-Germeys, Nicolson, & Delespaul, Reference Myin-Germeys, Nicolson and Delespaul2001; Myin-Germeys, Birchwood, & Kwapil, Reference Myin-Germeys, Birchwood and Kwapil2011). Each affective item was phrased ‘Do you feel…right now?’. High internal consistency of NA (Cronbach's α = 0.853) and PA (Cronbach's α = 0.867), based on the data on day 1, was achieved. The AVH item was phrased as follows: ‘Do you hear voices that other people cannot hear right now?’ (Kimhy et al., Reference Kimhy, Wall, Hansen, Vakhrusheva, Choi, Delespaul and Malaspina2017; Varese, Udachina, Myin-Germeys, Oorschot, & Bentall, Reference Varese, Udachina, Myin-Germeys, Oorschot and Bentall2011). Items on NA, PA, and AVH were rated on seven-point Likert scales, which ranged from one (not at all) to seven (very much).
Beliefs about voices and responses to voices
The Chinese version of the Revised Beliefs about Voices Questionnaire (BAVQ-R; Chadwick, Lees, & Birchwood, Reference Chadwick, Lees and Birchwood2000) is a 35-item, four-point (0–3) rating self-report questionnaire that consists of three subscales pertaining to beliefs about voices: malevolence (i.e. voices as evil and persecutory), omnipotence (i.e. voices as controlling and powerful), and benevolence (i.e. voices as helpful), with scores of each subscale ranging from 0 to 18, and two subscales pertaining to emotional and behavioral responses to voices: resistance and engagement (four items for emotion and five items for behavior each), with score ranges from 0 to 27 and 0 to 24, respectively. Good internal consistency and test–retest reliability of the Chinese version have been reported (Wong & Chen, Reference Wong and Chen2015).
Procedures
Upon written consent, participants completed the baseline interview conducted by a graduate-level research assistant and trainee clinical psychologists, closely supervised by the first author, and self-report questionnaires. Toward the end of the assessment, participants were trained to use the ESM application on their personal smartphone, or an iPod touch fifth generation (Model A1509) borrowed from the research team. A practice trial was completed with the guidance of a researcher. Participants were encouraged to maintain their daily routine while carrying the mobile devices with them at all times throughout the assessment period. A follow-up phone call was made on the next day. Additional technical support was provided as required. After 6 days, participants returned the device where applicable, and were debriefed.
Statistical analysis
Baseline variables were analyzed using SPSS version 24.0. Multilevel linear regression modeling with maximum likelihood estimation was used to evaluate the association between ESM variables (Goldstein, Reference Goldstein1987). All regression models were tested using the XTMIXED command in STATA 11.2. ESM variables were entered as level-1 variables and participant was entered as a level-2 random-effect variable. In order to capture the mean relation between a patient's time-specific deviation in a certain ESM variable, relative to the overall level of that variable for each patient, repeated ESM variables were group-mean centered (i.e. within each patient). In order to compare the effects of baseline variables across patients, baseline variables such as beliefs about voices were grand-mean centered (i.e. within the entire sample) (Curran & Bauer, Reference Curran and Bauer2011; Wang & Maxwell, Reference Wang and Maxwell2015). As in previous ESM studies (Udachina, Bentall, Varese, & Rowse, Reference Udachina, Bentall, Varese and Rowse2017; Vaessen et al., Reference Vaessen, Kasanova, Hernaus, Lataster, Collip, van Nierop and Myin-Germeys2018; Westermann et al., Reference Westermann, Grezellschak, Oravecz, Moritz, Lüdtke and Jansen2017), participants who completed <30% (i.e. 18 out of 60) of ESM observations were excluded from the analysis.
To model between-moment associations of ESM variables, time-lagged analyses were performed. The effect of the independent variable (IV) at a given moment t (IVt) on change in the dependent variable (DV) was modeled by regressing DV at the subsequent moment t + 1 (DVt +1) on IVt, controlling for DV at moment t (DVt). To test the effect of baseline variables (Malevolence, Omnipotence, Benevolence) on the moment-to-moment association between NA and AVH, separate multilevel linear regression models were estimated. With the level of NA as DV, Malevolence and Malevolence × AVH interaction, Omnipotence and Omnipotence × AVH interaction, Benevolence and Benevolence × AVH interaction, along with the level of AVH, were entered as IVs, respectively. With the level of AVH as DV, Malevolence and Malevolence × NA interaction, Omnipotence and Omnipotence × NA interaction, Benevolence and Benevolence × NA interaction, along with the level of NA, were entered as IVs, respectively.
Results
Demographic and clinical data
A total of 54 patients consented to participate in this study. After excluding seven patients who did not meet our inclusion criteria, the final sample consisted of 47 patients, among whom 41 completed at least 18 out of 60 ESM entries and were included in the analysis. There was no significant difference between completers and non-completers in age, gender, years of education, number of psychiatric admissions, or any of the clinical and self-report measures (p > 0.05). Four extreme responders for the AVH item of the ESM questionnaire were identified (three reported score of 1 for all entries and one reported score of 7 for all entries). Results did not differ after removing these individuals. Therefore, we kept them in the analysis, and reported results based on a sample of 41 individuals. In total, 1654 data points were obtained. The consecutive ESM moments within the same day were 18.92 to 172.78 min apart (mean = 71.77, s.d. = 24.26).
Table 1 presents the descriptive statistics of the sample (N = 41). The average baseline rating of SAPS hallucinations scale score was 14.41 (s.d. = 6.59, range 5–29), PSYRATS auditory hallucinations scale score was 25.49 (s.d. = 7.37, range 9–40), and PANSS item P3 (hallucinatory behavior) was 4.12 (s.d. = 0.81, range 3–6), indicating a moderate to moderate–severe level of severity (Kay et al., Reference Kay, Fiszbein and Opler1987). Moreover, the average PSYRATS AVH frequency and duration score were 2.49 (range 1–4) and 2.71 (range 1–4), respectively, indicating an hourly to daily occurrence of AVH that lasted for several minutes to at least an hour (Haddock et al., Reference Haddock, McCarron, Tarrier and Faragher1999).
Table 1. Sample characteristics (N = 41)
CPZ, chlorpromazine equivalent; HAL, haloperidol equivalent; PANSS, Positive and Negative Syndrome Scale; CDSS, Calgary Depression Scale for Schizophrenia; BAI, Beck Anxiety Inventory.
Average baseline ratings of subscales of BAVQ-R were as follows: Malevolence = 6.32 (s.d. = 4.85, range 0–18), Omnipotence = 7.54 (s.d. = 3.85, range 2–18), Benevolence = 5.02 (s.d. = 5.20, range 0–18), Resistance = 14.83 (s.d. = 6.09, range 0–25), and Engagement = 6.37 (s.d. = 5.77, range 0–19). Given that the assumption of normality is violated, Spearman's ρ was examined. Malevolence and Omnipotence were positively associated with each other [rs(39) = 0.50, p = 0.001] and with Resistance [rs(39) = 0.53, p < 0.001 and rs(39) = 0.38, p = 0.015], respectively, whereas Benevolence was positively associated with Engagement [rs(39) = 0.82, p < 0.001] and negatively associated with Resistance [rs(39) = −0.41, p = 0.008]. Omnipotence was also positively associated with Engagement [rs(39) = 0.33, p = 0.037].
The average ESM compliance rate was 67.20 (s.d. = 20.47, range = 30.00–98.33). The average ratings of ESM items, collapsed across 6 days, were as follows: NA = 2.81 (s.d. = 1.45, range 1–7), PA = 4.11 (s.d. = 1.65, range 1–7), and AVH = 3.00 (s.d. = 2.12, range 1–7). There was no significant change in momentary level of NA (β = −0.013, s.e. = 0.012, p = 0.290), level of PA (β = −0.015, s.e. = 0.012, p = 0.193), and intensity of AVH (β = −0.008, s.e. = 0.014, p = 0.568) over 6 days.
Same-moment associations between NA and AVH
Multilevel linear regression showed that NAt was significantly and positively associated with AVHt (β = 0.109, s.e. = 0.022, p < 0.001). Conversely, PAt was significantly and negatively associated with AVHt (β = −0.060, s.e. = 0.021, p = 0.004). Results did not differ after controlling for baseline CDSS score, BAI score, age, and gender.
Across-moment associations between NA and AVH
Time-lagged analysis showed that NAt was significantly associated with AVHt +1 (β = 0.130, s.e. = 0.034, p < 0.001). After controlling for AVHt, the association between NAt and AVHt +1 remained significant (β = 0.077, s.e. = 0.034, p = 0.022). Therefore, the level of NA predicted both the intensity of and increase in AVH at the following time point. Results did not differ after controlling for baseline CDSS score, BAI score, emotional content of AVH (on PSYRATS), age, and gender.
Time-lagged analysis showed that AVHt was significantly associated with NAt +1 (β = 0.091, s.e. = 0.026, p < 0.001). After controlling for NAt, the association between AVHt and NAt +1 remained significant (β = 0.060, s.e. = 0.026, p = 0.018). Therefore, the intensity of AVH predicted both the level of and increase in NA at the following time point. Results did not differ after controlling for baseline CDSS score, BAI score, emotional content of AVH (on PSYRATS), age, and gender.
Moderating effect of baseline beliefs about voices (malevolence, omnipotence, and benevolence) on the moment-to-moment associations between NA and AVH
Regression analyses revealed no main effect of Malevolence, Omnipotence, or Benevolence on either NAt or AVHt (p > 0.05). Main effects of Resistance and Engagement were also non-significant (p > 0.05). This indicated that none of the BAVQ-R scores was associated with momentary levels of NA and AVH.
In the model with NAt as DV, there was a significant interaction effect between Malevolence and AVHt (β = 0.016, s.e. = 0.005, p < 0.001), and Omnipotence and AVHt (β = 0.017, s.e. = 0.006, p = 0.004). However, there was no significant interaction effect between Benevolence and AVHt (β = −0.007, s.e. = 0.004, p = 0.067). This indicated that beliefs about voices as malevolent and omnipotent exacerbated the relationship between AVH and NA within the same moment.
In the model with NAt +1 as DV and NAt as covariate, there was no significant interaction effect between Malevolence and AVHt (β = −0.008, s.e. = 0.005, p = 0.151), Omnipotence and AVHt (β = −0.009, s.e. = 0.007, p = 0.197), or Benevolence and AVHt (β = 0.003, s.e. = 0.004, p = 0.551). This indicated that beliefs about voices did not predict a change in NA at the subsequent time point, following AVH.
In the model with AVHt +1 as DV and AVHt as covariate, there was a significant interaction effect between Omnipotence and NAt (β = 0.015, s.e. = 0.008, p = 0.046). However, there was no significant interaction effect between Malevolence and NAt (β = 0.010, s.e. = 0.007, p = 0.168), and Benevolence and NAt (β = −0.008, s.e. = 0.006, p = 0.187). This indicated that the belief about voices as omnipotent exacerbated an increase in AVH at the subsequent time point, following NA.
Results remained unchanged after controlling for baseline CDSS score, BAI score, age, and gender. The key findings are summarized in Fig. 1 in a schematic manner.
Fig. 1 Schematic diagram illustrating relationships between variables within and across time points. All relationships shown are positive.
Discussion
The present study employed ESM to reveal the moment-to-moment associations between AVH and NA, and the moderating effect of beliefs about voices on their temporal dynamics, among a group of patients with frequent AVH. Momentary level of AVH predicted an increase in NA at the subsequent moment. Crucially, the momentary level of NA also predicted an increase in AVH at the subsequent moment. These findings remained robust after controlling for baseline demographic characteristics and mood symptoms. This demonstrated, to our knowledge the first time, a bi-directional relationship between AVH and NA, whereby patients following a hallucinatory episode experienced more NA that further triggered subsequent hallucinations, leading to a perpetual negative feedback loop.
This study added to the current ESM literature on AVH and emotions in several ways. Firstly, we replicated Peters et al.'s (Reference Peters, Lataster, Greenwood, Kuipers, Scott, Williams and Myin-Germeys2012a) finding of a positive association between AVH and negative emotions within the same moment, and further revealed that such positive association was also true across moments. Secondly, rather than setting an arbitrary threshold of a ‘hallucinatory episode’ like Delespaul et al. (Reference Delespaul, Devries and van Os2002) and Oorschot et al. (Reference Oorschot, Lataster, Thewissen, Bentall, Delespaul and Myin-Germeys2012), we treated the severity of AVH as a continuous variable which allowed for greater statistical power and sensitivity to examine variations across time points. Thirdly, we included multiple NA items and adopted an aggregate score rather than restricted to one single emotion (e.g. anxiety, Delespaul et al., Reference Delespaul, Devries and van Os2002). This is supported by previous ESM studies (Kimhy et al., Reference Kimhy, Delespaul, Corcoran, Ahn, Yale and Malaspina2006; Myin-Germeys, Delespaul, & van Os, Reference Myin-Germeys, Delespaul and van Os2005; So et al., Reference So, Peters, Swendsen, Garety and Kapur2013) where a high correlation between the NA items was consistently reported.
Mapping out the vicious cycle between AVH and negative emotions carries treatment implications. In cognitive behavioral therapy for psychosis (CBTp), individuals are typically asked to identify triggers and situations that are associated with higher or lower severity of AVH, a procedure called functional analysis (van der Gaag, Valmaggia, & Smit, Reference van der Gaag, Valmaggia and Smit2014). Individuals are also encouraged to identify behavioral and emotional responses to AVH, and how they in turn lead to more AVH (Tarrier et al., Reference Tarrier, Beckett, Harwood, Baker, Yusupoff and Ugarteburu1993; Thomas et al., Reference Thomas, Hayward, Peters, van der Gaag, Bentall, Jenner and McCarthy-Jones2014). In addition to recent ESM studies (Chan et al., Reference Chan, So, Mak, Siah, Chan and Wu2019; Kimhy et al., Reference Kimhy, Wall, Hansen, Vakhrusheva, Choi, Delespaul and Malaspina2017; Leung et al., Reference Leung, So, Kwok, Ng, Chan, Lo and Lee2019; So et al., Reference So, Peters, Swendsen, Garety and Kapur2013, Reference So, Peters, Swendsen, Garety and Kapur2014), we have demonstrated the feasibility and reliability of mapping out moment-to-moment dynamics between symptoms and subjective experiences. This indicates that ESM, with assistance from a therapist, can be used more routinely in clinical practice to assist patients’ understanding of their illness and to facilitate illness management (Bell, Lim, Rossell, & Thomas, Reference Bell, Lim, Rossell and Thomas2017; Bell et al., Reference Bell, Fielding-Smith, Hayward, Rossell, Lim, Farhall and Thomas2018a, Reference Bell, Fielding-Smith, Hayward, Rossell, Lim, Farhall and Thomas2018b; Myin-Germeys et al., Reference Myin-Germeys, Birchwood and Kwapil2011; Simons et al., Reference Simons, Drukker, Evers, van Mastrigt, Höhn, Kramer and Wichers2017; van Os et al., Reference van Os, Verhagen, Marsman, Peeters, Bak, Marcelis and Delespaul2017). Our results further suggest that activities to lift affect may be effective in reducing the frequency and severity of AVH. Indeed, singing and physical activity have frequently been described as effective coping strategies (Carter, Mackinnon, & Copolov, Reference Carter, Mackinnon and Copolov1996).
Consistent with Peters et al. (Reference Peters, Williams, Cooke and Kuipers2012b) and others, we found that, at baseline, beliefs about voices as malevolent and omnipotent were associated with resistance of voices, whereas the belief about voices as benevolent was associated with engagement with voices. However, we found omnipotence to also be significantly associated with engagement with voices. Although beliefs about voices were not significantly associated with momentary levels of NA or AVH, omnipotence was associated with subsequent increase in hallucinatory experiences following NA. Our result is in line with van der Gaag et al.'s (Reference van der Gaag, Hageman and Birchwood2003) argument of differential effects of the two types of beliefs about voices (malevolence and omnipotence). Believing one's voices as powerful (omnipotent) may or may not lead to resistance against the voices, but what matters to a resistant response is when the power is interpreted as carrying a bad intention. Considering the potential collinearity between malevolence and omnipotence on NA, we conducted a post-hoc test of Malevolence × Omnipotence interaction. We found no significant interaction effect. Our results raised the possibility that, for some patients who believe their voices to be powerful, they may engage with voices more in times when they are in distress. Indeed, Hacker, Birchwood, Tudway, Meaden, and Amphlett (Reference Hacker, Birchwood, Tudway, Meaden and Amphlett2008) found that individuals who believed their voices as omnipotent engaged more frequently in safety-seeking behaviors, which included conversing with their voices. These suggest that the relationship between beliefs about voices and engagement/resistance may be less straightforward than previously expected, which warrants further research.
There are several limitations to this study. Firstly, beliefs about voices were only assessed once at baseline. Although appraisals about hallucinations were assumed to be habitual (Csipke & Kinderman, Reference Csipke and Kinderman2006), Peters et al. (Reference Peters, Lataster, Greenwood, Kuipers, Scott, Williams and Myin-Germeys2012a) reported that beliefs about power and control of voices may fluctuate across moments. This may explain the absence of associations between beliefs about voices and momentary AVH and NA. The one-time assessment of beliefs about voices in the present study also did not allow us to test for mediation effects. Secondly, although the focus of this study was voices, other co-occurring psychotic symptoms (such as visual hallucinations, delusions, and negative symptoms) may have convoluted the relationships between affect and AVH, and factors such as life events and emotional processing may also affect hallucinatory experiences in ways that we did not measure (Larøi et al., Reference Larøi, Thomas, Aleman, Fernyhough, Wilkinson, Deamer and McCarthy-Jones2019). Thirdly, the random-signal contingency led to varying duration between beeps. We are unsure whether the results would be different if the duration was consistent, or that any between-moment relationships may have been lost due to the duration being too long or too short.
This study demonstrated a vicious cycle between hallucinatory experiences and emotional experiences, which occur in real time in the flow of daily life among patients with drug-resistant hallucinations. How an individual interprets the power and intent of their voices has an impact on their hallucinatory experiences and emotions. The interaction between hallucinations, emotions, and beliefs about voices reported in this study is consistent with cognitive models of psychosis and may inform CBTp.
Acknowledgements
We would like to thank the participants for giving up their time to take part in this project.
Financial support
This work was supported by the Health and Medical Research Fund, the Food and Health Bureau, The Government of the Hong Kong Special Administrative Region (grant number 13140131).
Conflict of interest
None.
Ethical standards
The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008.
