Intolerance of Uncertainty (IU) can be defined as the “individual’s dispositional incapacity to endure the aversive response triggered by the perceived absence of salient, key, or sufficient information, and sustained by the associated perception of uncertainty” (Carleton, 2016, p. 31); individuals high in IU find situations that are uncertain threatening, upsetting, and undesirable, regardless of the actual probability of a negative event to occur (Dugas et al., 1998). IU as a construct was originally developed by the Laval team working on models and treatments of generalized anxiety disorder (GAD) in the early nineties (Freeston et al., 1994; Dugas et al., 1998). Among the systematically validated explanatory models of GAD (for a complete review, refer to Behar et al., 2009), the Intolerance of Uncertainty Model of GAD (IUM) proposed by Dugas et al. (1998) originally asserted that four factors contribute to the development and the maintenance of worry, the core feature of GAD (American Psychiatric Association [APA], 2013): IU, Positive Beliefs about Worry (PBW), Negative Problem Orientation (NPO), and Cognitive Avoidance (CA).

A robust body of research bolstered the evidence of IU, the model’s main feature, as a cognitive vulnerability factor for worry, as well as a maintaining factor for GAD (e.g., Ladouceur et al., 2000; Sexton et al., 2003; Koerner and Dugas, 2008). Negative beliefs about uncertainty usually interfere with the individual’s ability to effectively deal with these situations and promote the use of worry as a dysfunctional strategy to cope with or prevent feared outcomes (Dugas et al., 1998; Behar et al., 2009). PBW are distorted beliefs about the usefulness of worry; false contingencies usually act as both positive (e.g., worrying can sometimes produce effective solutions) and negative (the non-occurrence of a feared event) reinforcements maintaining the use of worry (Freeston et al., 1994; Dugas et al., 1998). Indeed, individuals endorsing PBW believe that worry is a positive personality feature (i.e., “being a worrier means being thoughtful”), and that worrying is an effective problem solving strategy, capable of preventing negative situations to occur, avoiding unpleasant emotions associated with negative events, and motivating to act in life (Freeston et al., 1994). NPO refers to a negative attitude toward problems and consists of a set of negative cognitive and emotional reactions that are activated when a problem situation occurs. It is associated with low levels of confidence about the ability to solve problems successfully, low personal control over the problem solving process, and pessimism about problem-solving outcomes (D’Zurilla and Nezu, 1999, 2006). When facing a problem situation, individuals with GAD usually do not focus on a problem solving strategy and do not consider themselves as effective problem solvers although they have an adequate knowledge of problem solving skills (Dugas et al., 1995, 1997; Robichaud and Dugas, 2005b; Koerner and Dugas, 2006). Lastly, CA is a cognitive process endorsed to avoid and/or suppress unwanted mental content, especially fearful mental images inducing peripheral physiological activation (Borkovec et al., 1993, 2004). CA includes a number of strategies (both automatic and controlled) such as substituting threatening thoughts with neutral or positive ones, transforming mental images into verbal-linguistic thoughts, etc., and it is particularly relevant to GAD (Dugas et al., 1998; Dugas and Koerner, 2005). Notably, since avoidance impairs the emotional processing of fear (Foa and Kozak, 1986), CA usually leads to the maintenance of high levels of worry and anxiety (Dugas et al., 1998; Borkovec et al., 2004). Importantly, the clinical efficacy of the IUM-based treatments for GAD relative to wait-list control conditions (Dugas and Ladouceur, 2000; Ladouceur et al., 2000; Dugas et al., 2003, 2010; Dugas and Robichaud, 2007) has been supported in multiple studies.

Despite the role of all components in worry and GAD has been considerably demonstrated (e.g., Freeston et al., 1994; Borkovec and Roemer, 1995; Dugas et al., 1997, 2005; Ladouceur et al., 1998, 1999; Robichaud et al., 2003; Buhr and Dugas, 2006; Koerner and Dugas, 2008), only a few studies were conducted to shed light on the mechanisms through which the IUM components act in determining worry and somatic anxiety levels, i.e., the core features of GAD. For example, Dugas et al. (2007) examined the associations between the IUM components and different indicators of severity (diagnostic severity, worry severity, and somatic symptom severity) in patients with a primary diagnosis of GAD. Results from correlations showed that IU and NPO were correlated with all the three indicators of severity, whereas CA was related with diagnostic and somatic symptom severity and PBW were only associated with diagnostic severity. Furthermore, when age, gender, and depressive symptoms were controlled, IU and NPO maintained their positive correlations with worry, and IU showed stronger correlations with worry severity rather than with somatic symptom severity. Lastly, the conduction of linear regressions highlighted that, partialling out the contribution of IU, NPO was no longer a significant predictor of diagnostic severity, worry severity, and somatic symptom severity, thus suggesting that the role of IU in determining worry levels is more relevant than the role played by the other variables (Dugas et al., 2007). Nonetheless, the correlational nature of this study does not allow understanding how the IUM components interact in producing worry and somatic anxiety. A subsequent attempt to examine a mediational path from IU to worry is represented by a study by Koerner and Dugas (2008), who tested the mediating role of appraisals of ambiguous situations in the relationship between IU and worry through a vignette task. University students high and low in IU were compared on their appraisals of ambiguous, negative, and positive situations; the former appraised all the situations they were presented as more distressing than the latter (and this was particularly true for ambiguous ones). Furthermore, the authors observed that appraisals (cognitive bias) partially mediated the path from IU to worry, and that worry partially mediated the path from IU to appraisals. Despite results provided support to the hypothesis of reciprocal relationships between worry and cognitive processes, the other IUM components were not taken into account in this study (Koerner and Dugas, 2008).

The IUM for GAD identified four main components associated with the development and maintenance of worry: IU, PBW, NPO, and CA (Dugas et al., 1998). Among them, IU was conceptualized as a background factor and the literature of the last 20 years supports the notion of IU as a vulnerability factor for worry (Dugas et al., 1998; Ladouceur et al., 2000; Sexton et al., 2003; Koerner and Dugas, 2008). Despite studies showing the contributions of PBW, NPO, and CA and occasionally in a mediational path from IU to worry (e.g., Dugas et al., 2007; Koerner and Dugas, 2008), whether and how these components can turn IU into worry and somatic anxiety still remains to be clarified. Therefore, the current studies aimed at gathering a better understanding of the original Laval model by testing two mediation models using bootstrapping approaches in two university samples made up of UK and Italian undergraduate students.

Our first hypothesis stated that PBW would have mediated the relationship between IU and worry but not the relationship between IU and somatic anxiety. As far as concern the path to worry, such an expectation was only partially sustained: Indeed, PBW emerged as a mediator only in the Italian sample. The different mediational patterns observed across samples is no better explained by a different degree of reliance in such beliefs across samples, since groups scored similarly on the WW-III (F_1,657 = 0.85, p = 0.36). Nonetheless, it is noteworthy that PBW have been proposed to initially drive the worry process, thus fostering the initial use of worry as a coping strategy and, once established across time, their role is variable (Dugas et al., 2007). Consequently, in light of the higher levels of both IU (F_1,657 = 70.61; p < 0.001; ηp2 = 0.10) and worry (F_1,657 = 62.20, p =< 0.001; ηp2 = 0.09) demonstrated by the UK students when compared with the Italian ones, one could hypothesize that a greater establishment of worry – altogether with a minor endorsement of PBW – might have already occurred in the former and, therefore, PBW could have already played their function. As regards the path from IU to somatic anxiety, consistently with our hypotheses, PBW emerged not to play a mediational role in both cultures; this confirms that, given their content specificity, PBW are logically associated with worry (Freeston et al., 1994).

The second hypothesis of the current study was that NPO and CA would have been mediators in the paths from IU to both worry and somatic anxiety. As far as concern NPO, it emerged to play a mediational role in the path from IU to worry both in Studies 1 and 2 thus demonstrating its involvement across cultures. On the other hand, NPO did not emerge to play any mediational role in the path from IU to somatic anxiety, which is in contrast with previous evidence highlighting this cognitive variable to be relevant also to somatic anxiety (e.g., Dugas et al., 2007; Fergus and Wu, 2011). Nonetheless, as outlined in other studies, NPO (as well as PBW and CA) are theorized to contribute directly to worry, which is then posited to contribute to somatic anxiety symptoms (e.g., Dugas and Robichaud, 2007; Dugas et al., 2007). In light of this, our result is barely surprising. An opposite pattern of results emerged when analyzing the role of CA. Indeed, findings revealed that it mediated only the path from IU to somatic anxiety and only in the Italian sample. Its involvement in the association between IU and somatic anxiety provides further support to the evidence stating that CA strategies are relevant in affecting peripheral activation and somatic symptoms (e.g., Borkovec et al., 2004), even though understanding why this emerged only in the Italian student sample is struggling. Notably, also in this case it is to note that UK undergraduates showed higher levels of CA than Italian ones (F_1,657 = 62.29; p < 0.001; ηp2 = 0.09), therefore hypothesizing such a finding as a function of a differential endorsement of CA strategies across samples is not explanatory. Furthermore, since the two samples were administered different measures of anxiety symptoms (i.e., the DASS-21 Anxiety in the UK sample and the BAI in the Italian one), speculating about differences between samples in relation to somatic anxiety levels is not possible. As already mentioned, CA emerged not to mediate the path from IU to worry in both samples. This is quite an unexpected finding, considering that CA is considered an important maintenance factor of worry (Borkovec et al., 1993, 2004; Dugas et al., 1998), and raises questions about the involvement of this component into the model.

Post hoc moderated mediation analyses were performed, when appropriate, to test whether IU levels moderated the mediational effect of the other components in the paths to worry and somatic anxiety. Findings revealed that IU emerged as a moderator only in the path to worry through NPO, and only in Italian students, thus providing only partial support to the idea of IU as a moderator of this relationship. Such a difference could be cautionary explained in terms of cross-cultural issues. Indeed, whilst UK has been reported as a culture whose members are more prone to take and manage unknown risks, Italy is included among the “strong uncertainty avoidance cultures” (De Mooij, 2000, 2003; Stremersch and Tellis, 2004; Wennekers et al., 2007). Notably, economic and financial crises have troubled Italy in the last decade; this might have somewhat impacted on a population (and on its young adult portion in particular) supposed to be resistant to change and particularly conservative. Notably, this does not necessarily mean that Italian university students are more intolerant of uncertainty than their UK counterparts; rather, they are likely to attribute a different (and more negative) meaning to uncertainty and/or to be less used to it (as demonstrated by scores obtained on the IUS-12, where scores shown by the Italian undergraduates were even significantly lower than those obtained by UK undergraduates). Nonetheless we acknowledge that, overall, our findings failed to demonstrate that the contribution of PBW, NPO, and CA depends on the level of IU, thus not providing additional help in understanding why higher IU leads to higher levels of worry (e.g., Ladouceur et al., 2000; Koerner and Dugas, 2008).

To summarize, the current re-examination of the IUM for GAD showed that the original conceptualization of the model, claiming that IU contributes to worry and anxiety through PBW, NPO, and CA is substantially supported but only in the Italian undergraduate sample, whereas findings emerged from the UK one are controversial. The discrepancies characterizing main results obtained in the two samples further strengthen the recommendation of performing cross-cultural studies in the field of anxiety, since it looks like that people from different cultures might actually interpret and express their symptoms in a different way, and cross-cultural research should help in understanding why and how (Malgady, 1996; Kirmayer, 2001; Norton, 2005). For example, related to this and with specific regard to IU, Norton (2005) explored the psychometric properties and the structure factor of the IUS in four different racial groups (African, Caucasian, Hispanic, and Southeast Asian) and findings showed that, despite substantially similar reliability and validity values across the four racial groups emerged, differences in the factor structures across groups were found; the extent to which cultural or racial variables had influenced results was not clearly definable but could not be excluded. A number of issues might have contributed in determining different patterns across samples and should be acknowledged as weaknesses characterizing the present research. First, the UK undergraduate sample was mainly made up of female participants (86.5%), whereas the Italian one was only slightly unbalanced in regard to gender (61.7% females). Such a disproportion in terms of gender composition may have affected results, since evidence highlighting that women are more likely to refer fear and higher levels of physiological hyperarousal, catastrophic cognitions, and anxiety sensitivity than men have been reported in literature (e.g., Armstrong and Khawaja, 2002; McLean and Anderson, 2011). On the other hand, in the present study female gender emerged to be (weakly) associated with higher scores on the PSWQ in both samples (and with scores obtained on the BAI in the Italian sample), therefore the two samples appear to be similarly characterized despite the number of females included. Possibly related to this, UK students showed higher levels of IU, NPO, CA, and worry than their Italian counterparts. Such differences could be reasonably attributed to a gender effect (but interactions Group × Gender for all measures resulted p > 0.05); to a “subject effect,” since the Italian sample was made up not only of Psychology students like the UK sample, but also of Engineering students who can show both a different approach and sensitivity to “psychology stuff”; to a culturally different attitude to metacognition; to different mechanisms intervening in different cultures. Indeed, emerged findings could also suggest that UK students are more prone to worry than Italian ones, but the processes through which they worry are different. Future studies should further investigate these topics. Another already mentioned difference between samples is in terms of measures employed to assess somatic anxiety and depression levels; this limits the possibility of tracing reliable comparisons between the results emerged in the two samples. Nonetheless, excellent convergent validity values between the DASS-21, the BAI, and the BDI-II have been extensively reported across cultures (Antony et al., 1998; Daza et al., 2002; Bados et al., 2005; Bottesi et al., 2015a), thus suggesting that such instruments substantially measure the same constructs. Furthermore, it is to note that scores ranges observed in the two samples on these measures were in line with previous literature on large non-clinical samples (e.g., Henry and Crawford, 2005; Sica and Ghisi, 2007).

Several further limitations characterizing the present study should be taken into account. The sample sizes employed, despite not extremely small, do not allow to generalize emerged results. A further obstacle to the generalization of results is represented by the cross-sectional nature of the current data. Indeed, we tested a conceptually (and partially empirically demonstrated) plausible model, but many cross-sectional models have equally plausible alternatives; furthermore, a number of biases characterizing the tests of mediation when using cross-sectional data have been highlighted (e.g., Maxwell and Cole, 2007). Therefore, longitudinal research in this field is highly recommended. Moreover, given the nature of samples, studies testing the same models also in community as well as in clinical populations are advocated. Indeed, whilst the adequacy and the advantages of employing analog samples when conducting research on clinical phenomena has been claimed (e.g., Ruscio et al., 2001; Olatunji et al., 2010; Abramowitz et al., 2014), current findings are not necessarily generalizable to other populations and, therefore, need to be intended as a starting point for additional research focusing on differently characterized samples. Another shortcoming is represented by the fact that some of the measures administered in the UK sample were not-validated refinements of previous versions (i.e., the WW-III and the R-CAQ), and some questionnaires used in the Italian sample were translations – and not proper validations – of the correspondent English ones (e.g., the IUS-12, the WW-III, the NPOQ, and the R-CAQ); nevertheless, internal consistency values were adequate both in the present study and in previous researches (Bottesi et al., 2014, 2015b). An additional issue deserving attention is the decision we made about partialling out depression in the models we tested. The rationale for doing so was eliminating a third variable scenario (depression) and increasing the specificity of findings to worry and somatic anxiety; however, in the current study measures of depression correlated with measures of worry and somatic anxiety in the 0.48 to 0.59 range and partialling out depression certainly took out a lot of variance. Therefore, such an issue raises questions about what is the portion of worry/anxiety that does not overlap with depression conceptually: Future developments should consider controlling for anhedonia rather than for shared features. Related to this, it is worthy to further highlight that increasing evidence supporting the notion of IU as a transdiagnostic factor involved in several psychopathologies beyond GAD, such as obsessive compulsive disorder (e.g., Tolin et al., 2003; Holaway et al., 2006), depression (e.g., Buhr and Dugas, 2002; Dugas et al., 2004; van der Heiden et al., 2010), post-traumatic stress disorder (e.g., Banducci et al., 2016; Oglesby et al., 2016), panic disorder and agoraphobia (e.g., McEvoy and Mahoney, 2012; Carleton et al., 2013), and social anxiety (e.g., Riskind et al., 2007; Boelen and Reijntjes, 2009; Carleton et al., 2010), has been documented. Future studies aiming to achieve a more in-depth knowledge of the commonalities that may exist between disorders are recommended; moreover, investigating the functioning of high-level factors, such as IU, in different phenomenologies might increase our understanding of comorbidities (Gentes and Ruscio, 2011).

Keeping in mind the above-mentioned limits, overall present results provide support to the current refinement of the original Laval model, where the IUM components are better conceptually integrated. Differential cross-cultural IU mechanisms, as well as potentially different worry domains across cultures, might have contributed to observed findings and deserve further investigation. Clinically, the findings tentatively suggest that understanding how people of different cultures understand and express uncertainty and IU may be central in our ability to target it effectively in therapy.

GB: Performed literature review, statistical analyses and wrote the manuscript. MG: Performed literature review and contributed in writing up the manuscript. EC, NB, and RP: Performed data collection, data entry, and support to data analyses. MF: Designed and supervised the entire project.

MF declares he has received training honoraria and book royalties on closely related topics. All the other authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.