Introduction
The term ‘paranoia’ refers to thoughts of excessive mistrust and suspicion toward other people. These thought impairments lie on the continuum that extends from mild manifestations, frequently observed in the general population sample, to paranoid delusions associated with a high level of conviction in thoughts, distress, and reduced social functioning (Bebbington et al., Reference Bebbington, McBride, Steel, Kuipers, Radovanovic, Brugha and Freeman2013; Freeman et al., Reference Freeman, Garety, Bebbington, Smith, Rollinson, Fowler and Dunn2005). It has been shown that the latent structure of paranoia is rather dimensional than categorical across clinical and nonclinical samples (Edens, Marcus, & Morey, Reference Edens, Marcus and Morey2009). Therefore, dimensional approaches are needed to better understand the mechanisms underlying the occurrence and progression of paranoid ideation.
Contextual processes are not only the content of paranoid thoughts but also play an important role in their development (Jester et al., Reference Jester, Thomas, Sturm, Harvey, Keshavan, Davis and Jeste2023). Over recent years, some studies have focused on the role of social disconnection as both the cause and consequence of paranoia (Fulford & Holt, Reference Fulford and Holt2023). Social disconnection is usually understood in light of two constructs, that is, social isolation and loneliness. The first one serves as an objective phenomenon referring to a low number of social contacts. The second one is subjective and occurs when individual social bonds are self-perceived as insufficient. A recent meta-analysis of cross-sectional studies demonstrated that psychotic experiences, especially paranoia, are associated with loneliness (Chau, Zhu, & So, Reference Chau, Zhu and So2019). However, little is known about longitudinal associations that have been studied for psychotic experiences in general, but not paranoia specifically (Endo et al., Reference Endo, Yamasaki, Nakanishi, DeVylder, Usami, Morimoto and Nishida2022; Qiao et al., Reference Qiao, Lafit, Lecei, Achterhof, Kirtley, Hiekkaranta and van Winkel2024). Experimental recollection of loneliness has been found to increase the level of paranoia, supporting causal effects (Gollwitzer, Wilczynska, & Jaya Reference Gollwitzer, Wilczynska and Jaya2018; Lamster, Nittel, Rief, Mehl, Lincoln Reference Lamster, Nittel, Rief, Mehl and Lincoln2017). Some insights have also been provided by the experience sampling method studies, showing that individuals with psychosis tend to spend time alone and experience higher levels of paranoia while being alone (Fett et al., Reference Fett, Hanssen, Eemers, Peters and Shergill2022). It has also been observed that momentary isolation from others and loneliness might predict the emergence of psychotic experiences in daily life (Bell et al., Reference Bell, Velthorst, Almansa, Myin-Germeys, Shergill and Fett2023; Chau, So, & Barkus, Reference Chau, So and Barkus2023; Misiak et al., Reference Misiak, Kowalski, Bogudzinska, Piotrowski, Gelner, Gaweda and Samochowiec2024).
Evolutionary frameworks posit that social interactions are necessary to increase the chances of reproduction and survival by providing a greater defensive capacity (Cacioppo, Cacioppo, & Boomsma, Reference Cacioppo, Cacioppo and Boomsma2014; Cacioppo & Hawkley, Reference Cacioppo and Hawkley2009). Taking into consideration these benefits, humans show an inherent ability to form social connections. In this context, the experience of loneliness is aversive but might be adaptive while signaling the need to activate cognitive processes and behaviors that aim to reconnect with others (Cacioppo & Hawkley, Reference Cacioppo and Hawkley2009). However, some individuals who experience loneliness do not activate processes and behaviors that aim to restore social connections. They tend to neglect opportunities to reconnect and approach a defensive cognitive style by focusing on social cues that inform them about potential threats from others. This strategy is supposed to be defensive as it aims to protect against threats associated with being alone. These individuals are likely to show a number of cognitive biases, including hypervigilance to social threats, attributional biases, expectations of rejection from others, negative perceptions of self and others, the focus on prevention-oriented goals, and low self-efficacy (Spithoven, Bijttebier, & Goossens, Reference Spithoven, Bijttebier and Goossens2017). Eventually, these cognitive biases further contribute to social disconnection through a self-defeating loop.
Cognitive biases and impairments of social cognition have been widely documented among individuals with psychosis and those at clinical high risk of psychosis (Gaweda et al., Reference Gaweda, Kowalski, Aleksandrowicz, Bagrowska, Dabkowska and Pionke-Ubych2024; van Donkersgoed et al., Reference van Donkersgoed, Wunderink, Nieboer, Aleman and Pijnenborg2015; Weinreb, Li, & Kurtz, Reference Weinreb, Li and Kurtz2022). Also, there is evidence that cognitive biases are associated with psychotic experiences in nonclinical samples (Livet, Navarri, Potvin, & Conrod, Reference Livet, Navarri, Potvin and Conrod2020). From a clinical perspective, impairments across cognitive processes are recognized as potential targets for therapeutic interventions (Nijman, Veling, van der Stouwe, & Pijnenborg, Reference Nijman, Veling, van der Stouwe and Pijnenborg2020; Sauve et al., Reference Sauve, Lavigne, Pochiet, Brodeur and Lepage2020). Although the evolutionary theory of loneliness has largely improved our understanding of cognitive processes involved in the development and progression of loneliness, it remains unknown whether it might be relevant for understanding the occurrence of paranoid ideation in the general population. It is also unknown whether loneliness and social isolation differ in terms of their associations with psychosis. Complex processes related to psychopathology are increasingly being addressed using network models that recognize their dynamic associations over time and provide insights into the relative importance of concurrent phenomena by estimating centrality metrics (Borsboom et al., Reference Borsboom, Deserno, Rhemtulla, Epskamp, Fried, McNally and Waldorp2021). Recent advancements in network modeling have provided a novel approach, known as the cross-lagged panel network (CLPN) analysis (Wysocki, van Bork, Cramer, & Rhemtulla, Reference Wysocki, van Bork, Cramer and Rhemtulla2022). This approach provides opportunities to model longitudinal associations while controlling for autoregressive processes without the use of a predefined causal conceptualization. Taking into consideration existing research gaps and recent advancements in the development of network models, the present study had two aims. First, this study aimed to investigate whether cognitive processes mediate the longitudinal association of social disconnection with paranoid ideations in the general population. Second, it aimed to disentangle whether social isolation and loneliness differ in terms of their associations with paranoia in the general population sample.
Methods
Participants
Recruitment procedures were implemented by means of the quota sampling to ensure sample representativeness with respect to sociodemographic characteristics (age, gender, the level of education, employment status, and place of residence). Baseline assessments were performed using the internet-based survey between July and August 2024. Some results from the baseline assessment were published previously (Misiak, Reference Misiak2024). The follow-up assessments were carried out in February 2025. Recruitment was performed by a research company using its own online access panel of registered participants. Members of this panel are enrolled by means of regular campaigns. For underrepresented groups (e.g. old-age individuals and ethnic minorities), additional recruitment campaigns are regularly initiated. All panel members aged at least 18 years were considered eligible. The participants were informed about the confidentiality of all assessments. They provided informed consent to participate in this study. The study received approval from the Bioethics Committee at Wroclaw Medical University, Wroclaw, Poland (Approval number: 553/2024).
Measures
Loneliness: To measure loneliness, the 11-item version of De Jong Gierveld Loneliness Scale (DJGLS) was administered (de Jong-Gierveld & Kamphuls, Reference de Jong-Gierveld and Kamphuls1985; Grygiel et al., Reference Grygiel, Humenny, Rebisz, Świtaj and Sikorska2013). The items include five possible responses (“yes!,” “yes,” “more or less,” “no,” and “no!”). There are two subscales of the DJGLS. The first one refers to emotional aspects of loneliness (six items). In turn, the second one measures social aspects of loneliness (five items). The total loneliness score is calculated by summing positive and neutral responses (“yes!,” “yes,” and “more or less”) across items measuring emotional loneliness together with negative and neutral responses (“no!,” “no,” and “more or less”) to those referring to social loneliness. The total loneliness score ranges between 0 and 11. Higher scores reflect a greater level of loneliness. In this study, the Cronbach’s alphas for emotional and social loneliness subscales were 0.874 and 0.810, respectively.
Social isolation: To measure the level of social isolation, the six-item version of the Lubben Social Network Scale (LSNS-6) was used (Lubben et al., Reference Lubben, Blozik, Gillmann, Iliffe, von Renteln Kruse, Beck and Stuck2006). Specific items record the number of family members and friends who are seen or heard at least once a month, with whom the respondent can talk about private matters, and who can be called on for help. Each item is based on a 6-point scale. The total score ranges between 0 and 20. Higher scores reflect a greater network of social contacts. To ease the interpretation of findings and show the level of social isolation, the total LSNS-6 was reversed. The Cronbach’s alpha of the LSNS-6 was 0.876 in the present study.
Cognitive processes: In the present study, the 18-item version (Gaweda et al., Reference Gaweda, Prochwicz, Krezolek, Klosowska, Staszkiewicz and Moritz2018) of the Davos Assessment of Cognitive Biases (DACOBS-18) (van der Gaag et al., Reference van der Gaag, Schutz, Ten Napel, Landa, Delespaul, Bak and de Hert2013) was administered. Each item is based on a 7-point scale (1 – “strongly disagree” and 7 – “strongly agree”). The items measure two cognitive biases (attributional biases and safety behaviors), social cognition problems, and subjective cognitive problems. In study, the Cronbach’s alphas were as follows: 0.801 for attributional biases, 0.745 for safety behaviors, 0.733 for social cognition problems, and 0.809 for subjective cognitive problems.
In turn, to assess rejection sensitivity, the Adult Rejection Sensitivity Questionnaire was administered (Berenson et al., Reference Berenson, Gyurak, Ayduk, Downey, Garner, Mogg and Pine2009). It is based on nine vignettes describing various social contexts. The respondent is asked to imagine the involvement in each situation and the rejection concern (1 – “very unconcerned” and 6 – “very concerned”) and rejection expectancy (1 – “very unlikely” and 6 – “very likely”). For each vignette, the rejection sensitivity score is estimated by multiplying the rejection concern score by the rejection expectancy score. Next, the mean across all vignettes is estimated, representing the total rejection sensitivity score. In this study, the Cronbach’s alphas were 0.868 for the rejection concern subscale and 0.826 for the rejection expectancy subscale.
Paranoid ideation: To assess paranoid ideation, the Revised Green et al. Paranoid Thoughts Scale (R-GPTS) was used (Freeman et al., Reference Freeman, Loe, Kingdon, Startup, Molodynski, Rosebrock and Bird2021). The R-GPTS items represent two subscales recording ideas of reference and ideas of persecution in the preceding month. Each item includes a 5-point scale (0 – “not at all” and 4 – “totally”). The items cover the experiences that did not appear as the effects of substance use. The optimal threshold scores for discriminating clinically relevant symptoms are ≥16 for ideas of reference and ≥11 for ideas of persecution (Freeman et al., Reference Freeman, Loe, Kingdon, Startup, Molodynski, Rosebrock and Bird2021). In this study, the Cronbach’s alphas were 0.930 for ideas of reference and 0.964 for ideas of persecution.
Depressive symptoms: To measure depressive symptoms, the Patient Health Questionnaire-9 (PHQ-9) was used (Kroenke, Spitzer, & Williams, Reference Kroenke, Spitzer and Williams2001). The PHQ-9 measures depressive symptoms in the preceding 2 weeks. The items are based on a 4-point scale (0 – “not at all” and 3 – “nearly every day”). The total score is 0–27 (higher scores refer to a greater level of depressive symptoms). The optimal threshold score for discriminating clinically relevant depressive symptoms has been estimated at ≥10 (Negeri et al., Reference Negeri, Levis, Sun, He, Krishnan and Wu2021). In this study, the Cronbach’s alpha was 0.897.
Anxiety symptoms: Anxiety symptoms were assessed using the Generalized Anxiety Disorder-7 (GAD-7) (Spitzer, Kroenke, Williams, & Lowe, Reference Spitzer, Kroenke, Williams and Lowe2006). The GAD-7 is based on seven items (rated on 4-point scale: 0 – “not at all” and 3 – “nearly every day”) that refer to the presence of generalized anxiety symptoms in the preceding 2 weeks. Items are rated on a 4-point scale (0 – “not at all” and 3 – “nearly every day”). The total score is 0–21 (higher scores indicate a greater level of anxiety symptoms). The optimal threshold score for discriminating clinically relevant generalized anxiety has been estimated at ≥10 (Spitzer et al., Reference Spitzer, Kroenke, Williams and Lowe2006). In this study, the Cronbach’s alpha was 0.944.
Data analysis
Individuals who completed assessments at both time points (further referred to as completers) and those who did not complete the follow-up assessment (further referred to as non-completers) were compared with respect to sociodemographic characteristics, symptom profiles, cognitive biases, social cognition, and subjective cognitive problems using the χ 2-test (categorical variables) and t-tests (continuous variables). Differences between both groups were considered significant if the p-value < 0.05. These analyses were performed in the JASP software.
Next, a two-step approach to analyze the data was followed. In the first step, the data were analyzed using CLPN models. Next, specific mediation models were analyzed. This mixed-model design was used as a formal analysis of mediation cannot be performed using CLPN. However, CLPN analysis does not impose a strict causal structure (Borsboom et al., Reference Borsboom, Deserno, Rhemtulla, Epskamp, Fried, McNally and Waldorp2021), and thus it might inform more focused models generating specific hypotheses about mediation. The CLPN and mediation models were analyzed among the study completers. No missing data were present in this group.
The following variables were analyzed with respect to cross-lagged and autoregressive effects using CLPN models: (1) loneliness; (2) social isolation; (3) attributional biases; (4) safety behaviors; (5) rejection sensitivity; (6) social cognition problems; (7) subjective cognitive problems; (8) ideas of reference; and (9) ideas of persecution. Age, gender, the level of education, employment status, monthly income, a lifetime history of psychiatric treatment, substance use in the preceding month (except for nicotine and alcohol), and baseline depressive and anxiety symptoms were used as covariates. The least absolute shrinkage and selection operator regressions were used to estimate autoregressive and cross-lagged effects to avoid indicating weak associations (Tibshirani, Reference Tibshirani1996). The CLPN was modeled in the R Studio. Regressions were analyzed in the glmnet package (Friedman, Hastie, & Tibshirani, Reference Friedman, Hastie and Tibshirani2010). In turn, the qgraph package was used to visualize the results (Epskamp et al., Reference Epskamp, Cramer, Waldorp, Schmittmann and Borsboom2012).
To assess the importance of specific nodes in the network, strength centrality was estimated (Jones, Mair, & McNally, Reference Jones, Mair and McNally2018). It shows the sum of edge weights between a specific node and other nodes in the network. This centrality metric is most commonly used in network models of psychotic experiences in community samples (Misiak, Pytel, & Stanczykiewicz, Reference Misiak, Pytel and Stanczykiewicz2025). In the case of CLPN, two categories of centrality are analyzed, that is, the output and input centrality. Out-strength shows the sum of edge weights from a specific node to all other nodes in the network. In turn, in-strength refers to the sum of edge weights to a specific node from all other nodes in the network.
Finally, nonparametric and case-drop bootstrapping was performed to analyze the network accuracy and stability in the bootnet package (Epskamp, Borsboom, & Fried, Reference Epskamp, Borsboom and Fried2018). These analyses provide the correlation stability coefficient (CS-C). It shows the maximal proportion of data that can be dropped to retain the correlation coefficient of >0.70 between data from the original sample and the data retained. The network is considered stable if the CS-C is higher than 0.25 (Epskamp & Fried, Reference Epskamp and Fried2018). Bootstrapping was also used to estimate the significance of differences between centrality metrics and edge weights.
Direct and indirect paths (with one intermediary node) between the nodes representing social disconnection (either loneliness or social isolation) and paranoid ideation (ideas of reference and ideas of persecution) were analyzed. Additional analyses of mediation were performed using the PROCESS macro (Model 4) (Hayes, Reference Hayes2018). Specific models were tested for the paths where one intermediary node representing cognitive processes connected either loneliness or social isolation with ideas of reference or ideas of persecution in the CLPN model. Results of these analyses were interpreted as significant if the 95% confidence interval did not include the zero value.
Results
Descriptive characteristics of the sample
The flow diagram of participants is shown in Supplementary Figure S1. Altogether, 5,099 individuals completed the baseline assessment (aged 44.9 ± 15.4, 47.7% men). In turn, a total of 3,275 participants completed the follow-up assessment, resulting in the study retention rate of 64.2%. Follow-up completers and non-completers did not differ significantly across sociodemographic characteristics and all other measures used in the present study (Table 1).
Table 1. Descriptive characteristics of the sample
Note: Data are reported as mean ± SD or n (%).
Abbreviations: ARSQ, Adult Rejection Sensitivity Questionnaire; DACOBS, Davos Assessment of Cognitive Biases Scale; DJGLS, De Jong Gierveld Loneliness Scale; GAD-7, Generalized Anxiety Disorder-7; LSNS-6, Lubben Social Network Scale-6; PHQ-9, Patients Health Questionnaire-9; R-GPTS, the Revised Green et al. Paranoid Thoughts Scale.
a Except for nicotine and alcohol.
b Optimal thresholds for discriminating clinically relevant symptoms (see the main text).
c Reversed coding was used to show the level of social isolation.
Network analysis
Overview of the network: The network estimated in the present study is visualized in Figure 1. There were 38 nonzero cross-lagged edges (46.9%). Negative edges (i.e. those showing negative correlations) were not found. Autoregressive effects are shown in Figure 2. The highest autoregressive effect was observed for loneliness, while the lowest one was found for ideas of reference. The strongest cross-lagged effect led from loneliness to rejection sensitivity, while the weakest one was from ideas of reference to rejection sensitivity (Supplementary Figure S2). The connection from loneliness to rejection sensitivity was significantly stronger compared to almost all cross-lagged associations in the network (except for the effect of loneliness on social isolation, as well as cross-lagged effects between ideas of persecution and ideas of reference).
Figure 1. The network estimated in the present study. Arrows show cross-lagged associations. Thicker and more saturated arrows refer to stronger cross-lagged effects. To ease visual interpretation, autoregressive effects were removed.
Figure 2. Autoregressive effects across variables measured in the present study.
Paths from social disconnection to paranoia: Loneliness directly predicted ideas of reference, but not ideas of persecution. Other paths from loneliness to ideas of persecution, with one intermediary node, led through ideas of reference, rejection sensitivity, and attributional biases. Also, other paths (with one intermediary node representing cognitive processes) from loneliness to ideas of reference led through rejection sensitivity, subjective and social cognition problems, and attributional biases.
Social isolation directly predicted loneliness; however, it did not directly predict ideas of reference and ideas of persecution. The path from social isolation to ideas of reference led through the effects of loneliness. There were no paths leading from social isolation to ideas of persecution through one intermediary node.
Paths from paranoia to social disconnection: Ideas of reference, but not ideas of persecution, directly predicted loneliness. This effect was not significantly stronger compared to the reversed one (i.e. from loneliness to ideas of reference; Supplementary Figure S2). The paths from ideas of persecution to loneliness with one intermediary node led through ideas of reference and attributional biases. In turn, the paths from ideas of reference to loneliness with one intermediary node led through attributional biases, but not through other cognitive processes.
Social isolation was directly predicted by loneliness but not by paranoid thoughts (either ideas of reference or ideas of persecution). The path from ideas of reference to social isolation with one intermediary node led through loneliness and attributional biases. In turn, the path from ideas of persecution to social ideation with one intermediary node led through attributional biases.
Strength centrality: The highest out-strength centrality was found for loneliness (Figure 3). It was significantly higher compared to all other nodes in the network (Supplementary Figure S3). The highest in-strength centrality was identified for ideas of reference. It was significantly higher than the in-strength centrality of loneliness, safety behaviors, attributional biases, and subjective cognitive problems (Supplementary Figure S4).
Figure 3. The strength centrality metrics.
Network stability and accuracy: The CS-C values were as follows: 0.672 for edges, 0.595 for out-strength centrality, and 0.439 for in-strength centrality. Therefore, the network might be interpreted as stable while retaining various proportions of data. Results of stability and accuracy analyses are visualized in Supplementary Figures S5 and S6.
Mediation analyses
There were significant effects of baseline loneliness on follow-up ideas of reference (Supplementary Table S2). Vice versa, baseline levels of ideas of reference directly predicted follow-up levels of loneliness. Indirect effects of cognitive processes (i.e. attributional biases, subjective cognitive problems, social cognition problems, and rejection sensitivity) in these associations were also significant, indicating the presence of partial mediation. Direct associations of loneliness with ideas of persecution with loneliness were not significant. However, significant indirect effects of cognitive processes in the association of loneliness with ideas of persecution were found, indicating the presence of full mediation.
Discussion
Findings from the present study imply that loneliness might be bidirectionally associated with the occurrence of paranoid thoughts, especially ideas of reference. However, loneliness might be more closely related to paranoid thoughts. It also had the greatest importance in terms of predicting other variables in the network. The observations extend the results of a recent meta-analysis showing that psychotic experiences are cross-sectionally associated with loneliness across the psychosis continuum, with stronger and more robust correlations observed for paranoia (Chau et al., Reference Chau, Zhu and So2019). Previous studies have also shown that loneliness and social isolation are modestly correlated, suggesting different underlying mechanisms (Cacioppo & Hawkley, Reference Cacioppo and Hawkley2009; Cacioppo, Hawkley, & Bernston, Reference Cacioppo, Hawkley and Bernston2003). It is also needed to note that while loneliness is always a distressing experience, social isolation might be congruent with individual social needs. A recent study, based on the experience sampling method, revealed that momentary social isolation does not exert negative effects on emotional responses when it is autonomous (choiceful) and does not accumulate over days (Weinstein, Vuorre, Adams, & Nguyen, Reference Weinstein, Vuorre, Adams and Nguyen2023). Other studies have shown that the preference for social isolation might be a strategy to regulate high-arousal and negative emotions (Nguyen, Konu, & Forbes, Reference Nguyen, Konu and Forbes2025; Nguyen, Ryan, & Deci, Reference Nguyen, Ryan and Deci2018). Finally, it has been suggested that social isolation might provide opportunities for the development of self-knowledge and self-connection that are difficult to achieve while spending time with others (Long & Averill, Reference Long and Averill2003).
It is also important to note that ideas of reference were more closely related to loneliness than ideas of persecution. Indeed, the latter dimension of paranoid thoughts was not found to be directly associated with loneliness. Cognitive processes, especially attributional biases, appeared to be more closely tied to the association between loneliness and ideas of persecution (full mediation) than the one between loneliness and ideas of reference (partial mediation). In addition to attributional biases, other cognitive biases may play an important role in explaining the associations between social disconnection and paranoid ideation. The present study included rejection sensitivity, a cognitive-affective bias that reflects heightened expectations of social rejection and exaggerated concern about its occurrence (Berenson et al., Reference Berenson, Gyurak, Ayduk, Downey, Garner, Mogg and Pine2009). Rejection sensitivity has been previously implicated in the emergence and maintenance of paranoia, particularly among individuals with psychosis and those at risk (Kesting & Lincoln, Reference Kesting and Lincoln2013; Lincoln, Johnson, Winters, & Laquidara, Reference Lincoln, Johnson, Winters and Laquidara2021). In the present study, rejection sensitivity was found to mediate the association of loneliness with ideas of reference. This finding aligns with theoretical models suggesting that loneliness may activate maladaptive cognitive-affective processes, such as heightened rejection sensitivity, which in turn amplify mistrust and threat-related interpretations of social interactions (Spithoven et al., Reference Spithoven, Bijttebier and Goossens2017).
From a clinical perspective, these results underscore the potential value of addressing cognitive biases, including rejection sensitivity and attributional biases, in psychological interventions aimed at reducing paranoid ideation and loneliness. Previous meta-analyses have shown that psychological interventions might be effective in reducing the level of loneliness, albeit with low-to-medium effect size estimates (Hickin et al., Reference Hickin, Kall, Shafran, Sutcliffe, Manzotti and Langan2021; Masi, Chen, Hawkley, & Cacioppo, Reference Masi, Chen, Hawkley and Cacioppo2011). The meta-analysis by Masi et al., (Reference Masi, Chen, Hawkley and Cacioppo2011) concluded that social cognition interventions have the greatest effectiveness in reducing loneliness. In turn, in the meta-analysis by Hickin et al. (Reference Hickin, Kall, Shafran, Sutcliffe, Manzotti and Langan2021), cognitive-behavioral therapy (CBT), targeting perceptual and cognitive biases that result in hypervigilance to negative social information, was the most common intervention for loneliness. Potential effectiveness of CBT might be related to the fact that this intervention encourages individuals to seek out disconfirming evidence to challenge their perceptions of loneliness and enhance self-efficacy, with the goal of changing behaviors, building social connections, and ultimately reducing loneliness (Hickin et al., Reference Hickin, Kall, Shafran, Sutcliffe, Manzotti and Langan2021; Kall et al., Reference Kall, Jagholm, Hesser, Andersson, Mathaldi, Norkvist and Andersson2020).
Furthermore, it is important to note that CBT might be effective for reducing general and positive symptoms of psychosis with small-to-medium effect size estimates (Berendsen et al., Reference Berendsen, Berendse, van der Torren, Vermeulen and de Haan2024). However, evidence with respect to the efficacy of CBT in reducing the level of loneliness among individuals with psychosis and its subclinical presentations is scarce. To date, only one study investigated whether the state-of-the-art CBT for psychosis might be beneficial in terms of reducing the levels of loneliness (Winkler et al., Reference Winkler, Lincoln, Wiesjahn, Jung and Schlier2024). However, caution is warranted when considering the extent to which the findings obtained from individuals with psychotic disorders (Berendsen et al., Reference Berendsen, Berendse, van der Torren, Vermeulen and de Haan2024; Winkler et al., Reference Winkler, Lincoln, Wiesjahn, Jung and Schlier2024) correspond to those from the present study, conducted in a nonclinical cohort. The authors found significant symptom reductions during this treatment; however, changes in the levels of loneliness were not significant. Nevertheless, it is important to note that the CBT protocol assessed in this study did not include a specific module targeting loneliness.
Limitations of the study by Winkler et al. (Reference Winkler, Lincoln, Wiesjahn, Jung and Schlier2024) might also be related to using only one item that directly recorded the feelings of loneliness, pointing to the discussion on how loneliness should be measured. Although single-item measures are commonly used to record loneliness, they tend to provide lower prevalence rates compared to those obtained by studies based on indirect indicators, for example, DJGLS (Stegen et al., Reference Stegen, Duppen, Savieri, Stas, Pan, Aartsen and De Donder2024). This discrepancy might be related to the observation that individuals with loneliness tend to underreport this experience (van den Broek, Lam, & Potente, Reference van den Broek, Lam and Potente2024). This phenomenon might be related to the stigma of loneliness (Fan et al., Reference Fan, Shi, Leng, Cui and Li2025; Ko, Wei, Rivas, & Tucker, Reference Ko, Wei, Rivas and Tucker2022; Shiovitz-Ezra & Ayalon, Reference Shiovitz-Ezra and Ayalon2012).
The observation that cognitive mechanisms are not critical in the association between loneliness and paranoia indicates that other psychological processes might also be relevant. For instance, there is evidence that loneliness might be perceived as a stressful experience associated with a greater likelihood to experience negative emotions and physiological responses (Cacioppo et al., Reference Cacioppo, Hawkley and Bernston2003; Luo & Shao, Reference Luo and Shao2023; Steptoe, Owen, Kunz-Ebrecht, & Brydon, Reference Steptoe, Owen, Kunz-Ebrecht and Brydon2004). Previous studies have also shown that individuals with psychosis and those at risk of psychosis show increased stress sensitivity, manifesting in elevated emotional responses to minor daily stressors (Myin-Germeys & van Os, Reference Myin-Germeys and van Os2007). Impaired emotion regulation has also been observed among individuals experiencing loneliness (Patrichi, Rimbu, Miu, & Szentagotai-Tatar, Reference Patrichi, Rimbu, Miu and Szentagotai-Tatar2025). In this regard, the models considering the relevance of both cognitive and emotional processes might provide further insights into the paths between social isolation, loneliness, and paranoia.
The present study is characterized by some limitations. The sample should be interpreted as nonclinical; therefore, translation of findings into clinical contexts needs to be approached with caution. At this point, it is also needed to note that the level of paranoid ideation was not validated using in-person clinical assessments. Moreover, the study did not record the full spectrum of cognitive biases related to the psychosis spectrum disorders, for example, jumping to conclusions and belief inflexibility. It is also needed to note that impairments across cognitive processes were not assessed using objective measures. This might be of importance for two domains of DACOBS, that is, subjective cognitive problems and social cognition problems. Another limitation is that a potential selection bias related to sampling among internet users cannot be excluded. Also, the cohort retention rate of 64.2% was relatively low. Furthermore, it should be noted that the data collection was based on two time points and, thus, insights into the temporal ordering of mediating effects of cognitive processes might be limited. Indeed, in all mediation models, either predictors and mediators or mediators and outcome variables were measured at the same time point. The limitations related to network modeling, where the results might depend on the inclusion of specific nodes, should also be considered. Finally, the study was observational and, thus, it informs about a temporal ordering of variables rather than causal effects.
In conclusion, the findings indicate that social disconnection, especially loneliness, is bidirectionally associated with paranoid thoughts in the general population. Altered information processing might link loneliness and social isolation with paranoia in some individuals. Future studies need to further investigate the mechanisms explaining the association of loneliness and social isolation with paranoia, beyond those related to cognitive processes. It is also needed to replicate the findings in clinical and general population cohorts with longer observation periods, as well as clinical trials of interventions targeting cognitive mechanisms associated with psychosis.
Supplementary material
The supplementary material for this article can be found at http://doi.org/10.1017/S003329172510130X.
Funding statement
The present study was funded by the resources of the Development Strategy of Wroclaw Medical University, Poland, under the project “Wroclaw Medical University in the Light of Scientific Excellence 2024– 2026” (task number: IDUB.C230.24.001).
Competing interests
The authors declare none.
Open access
