Exploring the Working Mechanisms between Anxiety, Mental Imagery, and Auditory Verbal Hallucinations: A Longitudinal Study

Abstract

Background

Auditory verbal hallucinations (AVHs) are the most common hallucinations within the psychosis spectrum and are often accompanied by anxiety. Integrating mental imagery into current psychological theories may increase the understanding of working mechanisms and effectiveness of cognitive behavioral therapy for AVHs. Therefore, the current study examined associations over time between levels of anxiety, mental imagery, and AVHs in individuals with psychosis spectrum disorders.

Study Design

Baseline data were analyzed from a replicated single-case series study using a within person mediation model approach and time-lagged multilevel analyses to examine associations over time between levels of anxiety, mental imagery, and AVHs. The baseline comprised a 2-week period before starting with an imagery intervention in 32 individuals diagnosed with psychosis spectrum disorders.

Study Results

There was a positive but non-significant association between anxiety and subsequent mental imagery (P = .061,  = 0.044). There was a significant positive association between mental imagery and AVHs (P < .001,  = 0.042), and between anxiety and AVHs (P < .001,  = 0.157). Additionally, anxiety was found to mediate the association between mental imagery and AVHs (P < .001  = 0.034).

Conclusion

The present findings suggest that mental imagery may play a role in the development and the maintenance of AVHs by intensifying anxiety, which then precedes an increase in AVHs. Capturing mental imagery alongside verbal interpretations of AVHs may help to understand the dynamic interplay between mental imagery, anxiety, and AVHs and to refine cognitive frameworks for more effective psychological treatments.

Introduction

Auditory verbal hallucinations (AVHs) are the most common hallucinations in psychosis spectrum disorders.^1,2 Cognitive behavioral therapy (CBT), the current standard psychological treatment for AVHs, is grounded in cognitive and learning theories,^3–5 and complemented by biological and neurocognitive models, offering a comprehensive basis for understanding and treating AVHs. While CBT for AVHs is valuable, it has shown only moderate effects,⁶ often with a high risk of relapse.⁷ These limitations may stem from the focus of CBT on patients’ verbal appraisals, potentially overlooking the role of mental imagery in AVHs. Mental imagery is defined as “perceptual information accessed from memory, giving rise to the experience of ‘seeing with the mind’s eye’ or ‘hearing with the mind’s ear’, without the presence of external stimuli.”,⁸ and can be either emotional or non-emotional.^9,10 Emotional mental imagery, in particular, plays a central role in psychopathology,^11,12 with 3 key aspects identified as relevant in regard to psychology: frequency, quality, and imagery appraisals.^10,13–16 Notably, emotional mental imagery can induce a wide range of affective experiences, including both positively and negatively valenced emotions. In the present study, when we refer to “mental imagery,” we specifically refer to emotional mental imagery and incorporate its frequency, quality, and appraisals. Mental imagery impacts emotion, motivation, and behavior across various mental disorders and holds promise as an additional avenue for psychological treatment.^12,17 Mental imagery-based interventions have already shown good treatment effects in treating AVHs^18–23, yet the relationship between mental imagery and AVHs is currently insufficiently clear.

Mental imagery and AVHs share defining features as perceptual experiences without external stimuli, leaving it open to debate whether voice-hearing reflects a continuum of misinterpreted thoughts or distinct processes that interact. To date, a few studies have investigated the relationship between mental imagery and psychotic symptoms, leading to hypotheses on the underlying mechanisms. These studies suggest that mental imagery is associated with symptoms of psychosis (ie, delusions or hallucinations), but also with the experience of anxiety.^24,25 Morrison²⁶ suggests that symptoms of psychosis arise from misinterpretations of intrusive, both verbal and image-based, thoughts (ie, continuum hypothesis). Morrison’s hypothesis aligns with other theoretical perspectives, suggesting that symptoms of anxiety and associated cognitions may be misinterpreted, ultimately leading to the experience of one’s own thoughts as alien and perceived instead as voices.²⁷ In recent years, growing academic attention has been paid to a cognitive model of mental imagery specifically applied to bipolar disorders by Holmes, Geddes, Colom, and Goodwin,¹⁴ often referred to as the emotional amplifier model. Holmes, Geddes, Colom, and Goodwin¹⁴ posit that mental imagery amplifies core symptoms of bipolar disorder by impacting not only mood but also anxiety levels and associated threat appraisals. Viewed transdiagnostically, mental imagery may increase emotions such as anxiety, and these emotions may act as a trigger that intensifies mental imagery, creating a feedback loop impacting other symptom formation. This proposition is of particular relevance as studies have highlighted the prominent role of anxiety in the recurrent activation of AVHs.^28,29 Based on these studies, it can be hypothesized that anxiety levels and AVHs represent distinct constructs, interacting dynamically, with each symptom potentially amplifying or modulating the other over time.^30–32 Symptoms of depression also seem to contribute to the development of psychosis³³, but the present study focused on the role of anxiety in relation to mental imagery in psychosis.

Taking this all together, the experience of anxiety, mental imagery, and AVHs might interact dynamically over time, though the underlying mechanisms remain unclear. The emotional amplifier model¹⁴ may also apply to psychosis, suggesting that increased anxiety and mental imagery could elevate AVHs. Evidence linking verbal cognitions to psychotic symptoms demonstrates that appraisals determine whether an experience is perceived as threatening, which triggers an emotional response, such as anxiety, that can further contribute to both the development and maintenance of psychosis.^34–37 Accordingly, we propose that the content or meaning attributed to a mental image (ie, imagery appraisal) may increase anxiety levels, which in turn reinforces the intensity and persistence of AVHs, a perspective we illustrated with a case example in Box 1. However, current literature does not yet clarify whether AVHs and mental imagery constitute distinct or partially overlapping processes, nor the role that anxiety may play in mediating the potential relationship between AVHs and mental imagery. We aim to take an initial step toward examining the associations between levels of anxiety, mental imagery, and AVHs, thereby providing a basis for generating hypotheses about the underlying mechanisms that may contribute to a refinement of cognitive models and interventions for AVHs.

Box 1.

Case Example

Participant

Male, 30 years, pseudonym: John.

Clinical context

John meets the criteria for schizoaffective disorder, depressive type. He is employed part-time, working 2 days per week in a supermarket. Clinically, he presents with persistent negative self-beliefs and auditory verbal hallucinations. For several years, he has reported hearing voices that speak negatively about him.

Context anxiety, mental imagery and auditory verbal hallucinations

When John is required to explain something to a customer at his supermarket job, he becomes anxious (trigger) and interprets the situation as a risk of being negatively judged or rejected (core belief indicating threat). Becoming increasingly upset, he experiences a highly vivid and detailed mental image (mental imagery quality) of groups of people laughing at him and saying he is incapable. This mental image increases his anxiety level. John appraises the image as evidence that the anticipated event will occur and that he will indeed fail (imagery appraisal). Concurrently, John reports hearing voices saying that he is a loser and will fail.

The current study examined associations over time between levels of anxiety, mental imagery, and AVHs in psychosis spectrum disorders using a within-person mediation model approach and time-lagged multilevel analyses. In order to test whether the emotional amplifier model of Holmes, Geddes, Colom, and Goodwin¹⁴ applies to AVH development in individuals with psychosis sensitivity, we propose a working mechanism comprising 2 interconnected segments of associations between levels of anxiety, mental imagery, and AVH as presented in Figure 1, hypothesizing that associations between levels of anxiety at successive time points are mediated by mental imagery levels, and associations between levels of mental imagery and AVHs are mediated by anxiety levels. More specifically, we tested the 3 hypotheses as illustrated in Figure 1.

Figure 1.

Hypothesized Working Mechanism between Levels of Anxiety, Mental Imagery, and AVHs Over Time: H1: Increased Anxiety Levels at Time t − 1 Predict an Increase in Mental Imagery Levels at Time t; H2: Both Increased Levels of Anxiety at Time t − 1 and Mental Imagery at Time t are Associated with an Increase in Anxiety Levels at Time t, with Mental Imagery Levels at Time t Mediating the Relationship between Anxiety Levels at Time t − 1 and Anxiety Levels at Time t; H3: Increased Levels of AVHs at Time t are Associated with Both Increased Levels of Mental Imagery and Anxiety at Time t, with Anxiety Levels at Time t Mediating the Relationship between Levels of Mental Imagery at Time t and AVHs at Time t. Abbreviations: AVHs, Auditory verbal hallucinations; H1, Hypothesis 1; H2, Hypothesis 2; H3, Hypothesis 3. Dotted circle, First hypothesized segment/Hypothesis 2. Filled circle: Second hypothesized segment/Hypothesis 3.

Methods

Participants

The current study used data from a replicated single-case series study using a within-person mediation model approach to explore associations over time between momentary daily measured levels of mental imagery, AVHs, and anxiety measured during a 2-week baseline period before starting with several imagery interventions.³⁸

In total, 41 patients aged 16 to 65 years were recruited from a specialized mental health care center for people with voice hearing and psychotic experiences at a large psychiatric hospital in the Netherlands. Thirty-nine patients showed interest in participating, from which 32 patients were included in the study. Inclusion criteria for the study were as follows:

• (1) Hearing voices as indicated by an intensity score of 4 or more on subscale 1.3 (“perceptual abnormalities”) of the Comprehensive Assessment of At-Risk Mental States (CAARMS),³⁹ or as indicated by a score of 3 or more on item P3 (“hallucinatory behavior”) of the Positive and Negative Syndrome Scale.⁴⁰

• (2) A DSM-5⁴¹ diagnosis in the psychosis spectrum (codes: 297.1; 298.8; 295.40; 295.90; 295.70; 298.8; 298.9) as confirmed by the administration of the mini-SCAN,⁴²

• (3) or a condition defined as Ultra High Risk/At Risk Mental State according to the CAARMS estimated by a clinician.

Full details of the exclusion criteria and the informed consent procedure are described by Janssen, Van den Berg, Keijsers, Paulik, Newman-Taylor, Taylor, Steel, and Marcelis.³⁸

The diagnosis of a participant was confirmed by the participants’ psychologist or psychiatrist, along with a case note review and the administration of the mini-SCAN.⁴² The mini-SCAN was also employed to screen participants for comorbid diagnoses.

Procedure

After inclusion, participants started a 2-week baseline period without intervention during which they completed thrice-daily online self-reports assessing mental imagery, AVHs, and anxiety, yielding 42 assessments. Participants completed the daily online self-report measurements in their natural environments. Daily assessments were prompted at equidistant time points with a 4-hour interval in between. Participants received a text message on their mobile phone with a link to the self-report measurement. They were asked to complete the measurement immediately after the alert or, if impossible, the same day. We established a minimum data threshold of over 33%, based on previous studies employing the Experience Sampling Method, to determine eligibility for this measure.⁴³ Participants who failed to meet this threshold were excluded from further analysis. Data were collected in a secure web-based system, Research Manager. This trial was registered at Clinicaltrials.gov (identifier NCT05603260). Ethical approval was given by METC azM/UM (NL79610.068.21/METC21-077).

Measures

Visual analogue scales of imagery characteristics

Based on Holmes, Hales, Young, and Di Simplicio¹⁵ and van den Berg, Voncken, Hendrickson, Houterman, and Keijsers,¹⁶ the following 7 questions were used to assess imagery characteristics and imagery severity scores: (1) “How vivid and clear were your images?” (2) “How compelling were your images?” (3) “Did you have the feeling that this image encouraged you to make plans to do something?” (4) “To what extent did this image tell you something about yourself?” (5) “To what extent did you believe this might happen now you thought about it?” (6) “To what extent did this image feel real, like something that is happening right now?” And lastly, (7) “From what perspective did you view this image? Through your own eyes or observing yourself from an external point of view?” Questions 1 to 6 were rated on 11-point visual analogue scale (VAS), with scores ranging from 0 (“not at all”) to 10 (“all the time or very much/intense”). The seventh item was rated on a VAS scale, ranging from 0 (“observing myself from an external view”), 5 (“mixed”), and 10 (“completely through my own eyes”). By summing the scores of all items at each daily measurement point, we computed a total severity score of momentary mental imagery, which ranges from 0 to 70. Summing the scores aligns with prior studies^43–45 and better reflects mental imagery severity than individual items.^16,47 Previous studies using similar scales report alpha values of 0.80 to 0.90, indicating strong reliability^11,16.

Visual analogue scales of AVHs

Five AVH questions were administered thrice daily using online self-report methods. The questions were: “How badly or how much did you suffer from the voices?,” “How often were you bothered by the voices?,” “How much sense of control did you experience over the voices?,” and “How loud were the voices the last time you had them?” The items were rated on 11-point VAS scales, ranging from 0 (“not at all”) to 10 (“all the time or very much/intense”). We computed a total severity score of momentary AVHs at each daily measurement point, ranging from 0 to 50. Previous research using comparable series of questions to assess the severity of auditory vocal hallucinations typically reported alpha values of 0.70 to 0.90, suggesting acceptable to excellent internal consistency.⁴⁸

Visual analogue scale of anxiety

The level of anxiety was measured using one 11-point VAS scale, ranging from 0 (“not anxious at all”) to 10 (“very anxious”).

Statistical analysis

R version 4.3.3 ⁴⁹ was used for the statistical analyses. Package lme4⁵⁰ was used to estimate the within person mediation model. The R script that was used is available in the supplementary material. In order to evaluate the fit, effect sizes (partial s) were calculated for all parameters following the guidelines of Cohen.⁵¹

In order to test the 3 hypotheses outlined in the introduction, a time-lagged mixed regression model analysis was performed⁵² to evaluate whether the within-person relationships between anxiety levels at successive time points were mediated by mental imagery levels and whether levels of mental imagery and AVHs within a person were mediated by anxiety levels. The first level was formed by the 42 baseline measurements of anxiety, mental imagery, and AVHs of each participant measured 3 times a day, and the second level was formed by the 32 participants.

Regression equations

The multilevel model consisted of 3 regression equations (Figure 2A). In the first equation at group level, mental imagery levels at time t were predicted by the level of anxiety at time t − 1, representing the hypothesis that increased anxiety levels lead to an increase of imagery levels (Hypothesis 1, Path a  Figure 2A).

Figure 2.

Predictive within-Person Mediation Models for Levels of Anxiety, Mental Imagery, and Auditory Verbal Hallucinations. (A) Within Person Mediation Model in Which Anxiety Levels at Time t − 1 Predicts Mental Imagery Levels at Time t, Which in Turn Predicts Anxiety Levels at Time t, Which Increases AVHs at Time t. (B) Within Person Mediation Model in Which Anxiety Levels at Time t − 1 Predicts Mental Imagery Levels at Time t, Which Directly Elevates AVHs at Time t. Abbreviations: AVHs, auditory verbal hallucinations; T, time.

In the second equation at group level, both anxiety levels at time t − 1 and mental imagery levels at time t served as predictors of anxiety levels at time t (Hypothesis 2, Path b  Figure 2A), representing the autocorrelation of anxiety over time and the hypothesis that levels of mental imagery and anxiety reinforce each other over time, and that mental imagery levels at time t mediate the relationship between levels of anxiety at time t − 1 and anxiety at time t (Path d  Figure 2A).

In the third equation at group level, AVHs at time t were expected to be predicted by levels of anxiety and mental imagery at the same moment t (Path c  Figure 2A), and anxiety levels at time t were expected to mediate the relationship between levels of mental imagery and AVHs at the same moment t (Hypotheses 3, Path e  Figure 2A). Note that theoretically it is assumed that within a very short time frame levels of anxiety induces the experience of mental imagery, which in turn induces AVH levels.

Covariance of the random effects

Since a multi-level was used, the random effects were estimated for each parameter to model the individual variation. The degree of individual variation can be used to evaluate whether the estimated value of the fixed effects is representative at the individual level.

Figure 2A shows the variables in the mediation paths. The multilevel structure is visible in Figure 2A by the underscores j, which indicate that the parameters , , and are estimated for each of the j = 1, … 32, participants by using random factors. Parameter is the relationship between levels of anxiety at t − 1 and mental imagery at t; is the relationship between levels of mental imagery and anxiety at t, is the direct relationship between levels of anxiety at t − 1 and anxiety at t, and is the direct relationship between levels of mental imagery and AVHs. is the indirect relationship between levels of anxiety at t − 1 and anxiety at t mediated by mental imagery levels. is the indirect relationship between levels of anxiety at t and AVHs at t mediated by mental imagery levels at t. is the covariance of the random variables and . A positive value indicates that participants that have a positive relationship between levels of anxiety at t − 1 and mental imagery, also show a positive relationship between levels of mental imagery and anxiety at t.   is the covariance of the random variables and . A positive value would indicate that participants that have a positive relationship between levels of mental imagery and anxiety at t, also show a positive relationship between levels of anxiety and AVHs at t. The total effect , is the sum of , , , , and .

Alternative model comparison

To further validate Hypothesis 3 that levels of mental imagery indirectly lead to AVH levels, mediated by anxiety levels at the same moment t, an additional check was conducted by testing the same model again, but this time without levels of anxiety at time t (Figure 2B). Note that the only difference with Figure 2A is that path and the covariance were removed. Should parameter not increase in this more parsimonious model, and the absolute and relative fit of the model not be reduced compared to the first model (Figure 2A), this would indicate that the relationship between levels of mental imagery and AVHs is not mediated by anxiety levels.

To compare models A and B in Figure 2, the effect size of parameter was compared, as well as the overall fit. Information Criteria BIC,⁵³ and AIC⁵⁴ were used to evaluate the relative fit⁵⁵ and a log-likelihood ratio test⁵⁶ was used to evaluate the absolute fit of the 2 nested models.

Test statistics and effect size

For each relationship (parameter a through e of the total effect f) a raw regression weight (ie, estimated value) was estimated and the SE of the estimate. Test statistics were calculated by dividing the raw regression weight by the SE. Note that these test statistics follow a Student t-distribution. Using this t-distribution, we calculated the probability that the raw regression weights are zero in the population (ie, P-value) of the dependent variable by the parameter, when corrected for all other parameters in the model. Note that a partial of 0.01 indicates a small effect, 0.06 a medium effect and 0.14 a large effect.⁵¹

Results

Descriptive statistics

Table 1 comprehensively summarizes participant demographics and characteristics, alongside disorder-specific information. Adherence to the measurement protocol was notably high, with no dropouts during the baseline period. All participants completed a minimum of 2 weeks of thrice-daily assessments, resulting in minimal missing data (5%). Here, missing data indicates the absence of 1 of the 3 assessments a day.

Table 1.

Characteristics of the Study Cohort Including Demographics, Diagnosis, Comorbidity, and Medication

	N = 32
Demographic information
Age years, mean (SD)	31 (12.33)
Gender, n (%)
Female	14 (43.8%)
Male	18 (56.2%)
Level of education, n (%)
Primary education	4 (12.5%)
Pre-vocational secondary education/secondary vocational education	16 (50%)
Senior general secondary/pre-university/higher professional	12 (37.5%)
work status, n (%)
Employed (incl. volunteers, side job and sickness benefit)	20 (62.5%)
Unemployed	12 (37.5%)
Diagnosis schizophrenia spectrum and UHR/ARMS, n (%)
Schizophrenia spectrum and other psychotic disorders	17 (53.1%)
UHR/ARMS	15 (46.9%)
Comorbidity, n (%)
Mood disorders	10 (31.3%)
Anxiety disorders	2 (6.3%)
Trauma-related disorders	5 (15.6%)
Somatic symptom disorders	2 (6.3%)
Alcohol/use disorders (light)	4 (12.5%)
Dissociative disorders	2 (6.3%)
Personality disorders	10 (31.3%)
Autism spectrum disorder	9 (28.1%)
ADHD	1 (3.2%)
Clinical course in years, n (%)
0-1 years	10 (31.3%)
2-5 years	15 (46.8%)
6-10 years	4 (12.5%)
>10 years	3 (9.4%)
Medication at screening, n (%)
Antipsychotic	13 (40.6%)
Antidepressant	8 (25%)
Anxiolytic	6 (18.6%)

Abbreviations: ARMS = At risk mental state; UHR = Ultra-high risk.

Regression equations

Hypothesis 1:

Increased anxiety levels at time t − 1 predict an increase in mental imagery levels at time t.

The multilevel analysis results in Table 2 showed that the unique association between anxiety at t − 1 and mental imagery at t (path a) was non-significant positive with a small to medium effect size.

Hypothesis 2:

Both increased levels of anxiety at time t − 1 and mental imagery at time t are associated with an increase in anxiety levels at time t, with mental imagery levels at time t mediating the association between levels of anxiety at time t − 1 and anxiety at time t.

Table 2.

Parameter Estimates of the Direct and Indirect Effects and Covariance

Parameter	Estimated value	SE	P-value	Partial
Model 1
a	0.30	0.16	.061	.044
b	0.29	0.05	<.001	.042
c	1.06	0.24	<.001	.157
d	0.14	0.13	.269	.021
e	0.23	0.04	<.001	.034
Covariance ab	0.04
Covariance bc	0.04
f	0.84
Model 2
a	0.31	0.16	.051	.047
b	0.31	0.05	<.001	.046
d	0.15	0.13	.230	.022
e	0.31	0.05	<.001	.047
Covariance ab	0.01

Abbreviations: Estimated Value = Raw regression weight; SE = standard error for the raw regression weight.

The unique association between mental imagery and anxiety at t (Path b) was a significant positive, small- to medium-size effect. The unique direct effect, d, from anxiety at t − 1 and anxiety at t corrected for mental imagery at t was small and not significant (Path d).

Hypothesis 3:

Increased levels of AVHs at time t are associated with both increased levels of mental imagery and anxiety at time t, with anxiety levels at time t mediating the association between levels of mental imagery at time t and AVHs at time.

There was a strong, unique, significant association between anxiety at t and AVHs at t (path c). The unique direct effect, e, from mental imagery at t to AVHs at t corrected for anxiety at t was small to medium and significant (Path e).

Covariance of the random effects

The covariance of the random effects of paths a and b was positive, indicating that, on average, a positive association between anxiety at t − 1 and mental imagery at t is associated with a positive association between mental imagery and anxiety at t.

The covariance of the random effects of paths b and c was positive, indicating that, on average, a positive association between mental imagery and anxiety at time t is associated with a positive association between anxiety and AVHs at time t.

The total effect, f, is 0.840, so the variance explained by the indirect—mediation—effect is , the indirect—mediation—effect is . The variance explained by the direct effect, d, is 17%. The variance explained by the direct effect, e, is 27%.

Finally, the percentage of the indirect mediation effect that is explained by the covariance between a and b is 31% confirming that there is quite some individual variation in the size of the mediation effect of mental imagery between anxiety at t − 1 and anxiety at t. The percentage of the indirect mediation effect that is explained by the covariance between b and c is 12% indicating that the individual variation in the size of the mediation effect of anxiety at t between mental imagery and AVHs at t is much smaller. The results indicate that the second part of the hypothesized working mechanism (ie, the associations from anxiety at t to mental imagery at t and subsequently to AVHs at t) was stronger than the pathway from anxiety at t − 1 to mental imagery at t and to AVHs at t. This suggests that the second part of the hypothesized working mechanism is likely more consistent across individuals, as evidenced by reduced variability in the data.

Alternative model comparison: Validation of hypothesis 3 that levels of mental imagery indirectly lead to AVHs, mediated by levels of anxiety at the same moment t

Parameter estimates of Model B in Figure 2, in which Path c is removed, were similar to those of Model A, except for the estimate of Path e (Table 2). The direct effect of mental imagery on AVHs is much larger in the second model, indicating that anxiety at t explains quite some variance between mental imagery and AVHs.

Model A (df = 29) demonstrated a better fit with lower AIC (24544) and BIC (24725) compared to Model B (df = 12, AIC = 25333, BIC = 25408). The log-likelihood tests to test the difference in model fit between the 2 models using a chi-square distribution is significant (L. Ratio = 823, P < .001), indicating that the Model A showed a better fit. These findings are consistent with the hypothesis that the association between mental imagery and AVHs operates through the experience of anxiety.

Discussion

The present study investigated the associations between levels of anxiety, mental imagery, and changes in AVHs during an average period of 2 weeks, using online daily self-report measures in individuals diagnosed with psychosis spectrum disorders. The aim was to gain a better understanding of the working mechanisms underlying the frequency and severity of AVHs and the potential role of the experience of anxiety and mental imagery in eliciting, preceding, or elevating these facets of AVHs. Therefore, we hypothesized a working mechanism outlining the potential pathways between levels of anxiety, mental imagery, and AVHs consisting of 2 interconnected segments of associations as presented in Figure 1. Contrary to our first hypothesis, an increase in anxiety levels at the preceding time point (t − 1) failed to predict an increase in mental imagery levels at the subsequent time point (t). The subsequent pathways of the hypothesized working mechanisms were confirmed: an increase in mental imagery levels predicted an increase in anxiety levels, and an increase in anxiety levels predicted an increase in AVH levels. While mental imagery did not mediate the association between levels of anxiety at t − 1 and anxiety at t, levels of anxiety at t did function as a mediator between levels of mental imagery and AVHs at t. Notably, the second segment of the hypothesized working mechanism (mental imagery → anxiety → AVHs) showed stronger and more consistent effects, offering a more robust explanation than the first segment (anxiety → mental imagery → anxiety).

Interpretation of findings

Contrary to the emotional amplifier model of Holmes, Geddes, Colom, and Goodwin,¹⁴ we found no significant association between levels of anxiety at t − 1 and mental imagery at t, suggesting that the experience of anxiety does not trigger mental imagery in psychosis spectrum disorders. As the first study to investigate the segmented relationships between levels of anxiety, mental imagery, and AVHs over time, this finding, although unexpected, is interesting as it may imply that the symptom sequence does not necessarily begin with the experience of anxiety. Our results suggest that the symptom sequence rather starts with the experience of mental imagery, which subsequently leads to anxiety levels and then to an elevation of AVHs. Alternative explanations for the non-significant association between levels of anxiety at t − 1 and mental imagery at t are possible. Another factor may be involved in the working mechanism between the levels of anxiety, mental imagery, and AVHs, such as imagery appraisals or threat beliefs as also proposed by Holmes, Geddes, Colom, and Goodwin.¹⁴ It is also possible that the factors in our hypothesized working mechanism occur in rapid succession, or nearly simultaneously, making it challenging to detect effects between distinct measurement points. In this sense, levels of anxiety, mental imagery, and AVHs may be conceptually and phenomenologically intertwined instead of just co-occurring. Possibly, the experience of anxiety, mental imagery, and AVHs may not be distinct experiences unfolding sequentially but overlapping phenomena that shape and intensify each other. Indeed, AVHs often involve imagery-like qualities and are experienced as affectively charged, making it difficult to separate the contribution of the experience of mental imagery from that of anxiety. This view aligns with earlier work of Ratcliffe and Wilkinson²⁷ and Morrison²⁶, suggesting that anxiety-related cognitions may be misattributed, resulting in one’s own thoughts being experienced as voices. In our predictive working mechanism model, variables following the experience of anxiety at the first measurement (t − 1), that is, levels of mental imagery, anxiety, and AVHs, were assessed at the same time points (t), and significant associations were observed among them. This finding suggests that the temporal proximity of the factors in the first segment of the hypothesized working mechanism may play a role in their interaction, warranting further investigation with increased temporal frequency to accurately capture these dynamic processes.

The finding that heightened anxiety levels were associated with increases in AVHs is consistent with previous network approach studies demonstrating the crucial role of anxiety in the dynamic interplay of psychotic symptoms.^30–32 This finding is also partially in line with theoretical models proposing an affective pathway to psychosis³³; however, these models primarily attribute this pathway to depression-related mechanisms. It is therefore crucial that future research examine not only the role of anxiety, but also that of depression when investigating mechanisms linking mental imagery to AVHs. It is also important to note that the association between heightened anxiety levels and an increase in AVHs was observed at a single point in time, precluding conclusions about causality and leaving open the possibility that the sequence of symptoms may be reversed. Nevertheless, future clinical frameworks for AVHs could improve treatment efficacy by considering the experience of anxiety as both a response to, and a driver factor of AVHs. Additionally, the association between levels of mental imagery at t and AVHs at t indicates that elevated levels of mental imagery might predict an increase in AVHs severity, highlighting the potential role of mental imagery in the emergence of AVHs. This finding aligns with previous cross-sectional studies^24,57,58 and meaningfully advances the literature by being among the first to utilize longitudinal data. Importantly, this association supports the clinical notion that both verbal and mental imagery-based thoughts contribute to AVHs.

As hypothesized and in line with the emotional amplifier model,¹⁴ we found that the association between levels of mental imagery and AVHs was mediated by the experience of anxiety. This finding also supports the hypothesis of Holmes, Geddes, Colom, and Goodwin¹⁴ that the experience of anxiety reinforces the cycle between levels of mental imagery and other psychological symptoms, including AVHs. Our findings align with studies showing that the experience of anxiety bridges psychotic and other imagery-based symptoms, such as trauma intrusions,^30,59 and with research linking the experience of anxiety and mental imagery in psychosis.^24,25

Although the associations between the variables in the second segment of our hypothesized working mechanism (mental imagery → anxiety → AVHs) were measured concurrently (at time t), our findings do provide hypotheses about a potential reinforcing process among these factors. In line with previous findings in mood disorders,^14,60 our findings might suggest that the experience of mental imagery precedes an increase in anxiety levels, and anxiety levels subsequently predict an increase in other psychological symptoms, including AVHs in psychosis. Our findings might also align with the continuum hypothesis based on the work of Morrison²⁶, suggesting that the experience of anxiety may exacerbate the misinterpretation of mental imagery (ie, misinterpretation of intrusive thoughts), which in turn leads to AVHs. Future research should aim to explore the directionality of the specific working mechanisms through which mental imagery influences anxiety and AVHs over time. Our findings reinforce the notion that addressing the experience of anxiety and mental imagery in therapeutic settings may be crucial for alleviating the distress associated with AVHs and may highlight the need for integrated treatment approaches that target both constructs. Naturally, such approaches take place within a holistic perspective, in which the personal contextual origins and resiliency factors are central to the understanding of the individual’s mental (dis)balance.

Methodological considerations

The findings of the present study should be interpreted with caution due to several limitations. Although, the sample size was relatively modest for group-level multilevel analyses, the large within-person effects observed in our treatment study data³⁸ increased confidence that mediation effects could also be detected. Accordingly, the multilevel mediation analyses were intended to explore the consistency of these within-person associations across individuals, rather than to replicate treatment efficacy findings. In addition, we deviated from our pre-registered plan by including only anxiety levels in the analyses. This decision was based on the complexity of simultaneously modelling both anxiety and depression, and the fact that existing imagery research has primarily focused on anxiety.^14,24,26 Examining the role of depression levels in the relation between mental imagery and AVHs represents an important step for future research. Another deviation from the pre-registered plan was that our daily questionnaires contained too few items to reliably distinguish between the constructs of quality and appraisal in emotional mental imagery. Therefore, we measured emotional mental imagery at a general level, integrating both quality and appraisal aspects, which is also consistent with current practices in the literature.^15,16 A third limitation is that, although measuring factors 3 times daily, we related certain factors in the predictive model (ie, the second segment of the hypothesized working mechanism in Figure 1) at the same time points (time t). Our findings should therefore be interpreted with caution, as no conclusions about causality can be drawn, and we can only propose hypotheses regarding the underlying mechanisms of the observed associations. Nevertheless, the observed associations appear robust, as the model was tested longitudinally within participants across all 42 measured time points. Another limitation is that we assumed a conceptual distinction between AVHs and mental imagery. Although this distinction is theoretically supported, partial conceptual overlap between these concepts or underlying processes cannot be ruled out. Future research should therefore further examine the extent to which AVHs and mental imagery represent distinct vs overlapping concepts and investigate the cognitive processes that underlie their potential interrelations. Lastly, we did not measure the factor of threat beliefs, which is integrated in the emotional amplifier theory. Future research should aim to address these distinctions to enhance our understanding of the role of mental imagery in AVHs. Additionally, incorporating biological markers of anxiety, such as cortisol measurements, could allow for more precise and temporally detailed associations between anxiety and other symptoms.

Clinical implications

Mental imagery can and is even likely to be active alongside verbal interpretations of AVHs, probably influencing each other and particularly operating predominantly through the experience of anxiety, a relationship that warrants further investigation in future research. Further research is also essential to clarify specific aspects of mental imagery (eg, quality, appraisals) that influence AVHs-anxiety dynamics, offering insights for refined interventions.

Therefore, this line of work may hold several implications for clinical practice. While existing and standard treatment with CBT for AVHs is highly valuable, CBT addresses verbal and semantical representations of patients’ experiences. The effects of CBT may be further strengthened by incorporating attention to mental imagery in relation to symptoms of anxiety, also. Our results provide insights into potential precursors of AVHs and automatic thoughts, as outlined in cognitive models of current standard CBT for AVHs. We propose that a focus on the origins of AVHs, with a possible role for the experience of mental imagery and anxiety, may enhance therapeutic effectiveness.

Conclusion

In summary, the findings of the present study suggest that the experience of mental imagery has a potential role in eliciting, maintaining, and elevating AVHs. Mental imagery appears to precede increases in anxiety levels, which in turn precede increases in AVHs. These results are largely in line with the emotional amplifier model,¹⁴ suggesting its relevance to psychosis. However, our findings need to be replicated in the future. Future research could focus on measuring additional characteristics of mental imagery alongside verbal thought interpretations of AVHs to clarify their interplay and to improve cognitive theories, potentially enhancing the effectiveness of psychosocial interventions for psychosis.

Funding

This research was financially supported by Mental Health Institute Eindhoven (GGzE) and Maastricht University.

Conflicts of Interest

None declared.