Obsessive-compulsive disorder (OCD) is a highly debilitating condition affecting between 1.1 and 1.8% of the population worldwide (1). It is characterized by unpleasant, unwanted obsessive thoughts, and repetitive compulsive rituals to try and neutralize these thoughts, and as such, can severely impair all aspects of quality of life (2). Elucidating the cognitive underpinnings of OCD is of great importance in improving our understanding of this condition to better help patients. In order to do this, we must appreciate the spectral nature of psychiatric conditions and the variability in their manifestations between individuals, and thus taking a dimensional approach is valuable. A clear evidence-based endophenotype may allow identification of individuals with a predisposition for OCD, and the development of more targeted screening measures and interventions. A number of theories have been proposed to explain the cognitive processes underlying OCD symptoms, and there have been several attempts to develop a unified account of the origins of OCD [e.g., (3, 4)], but as yet there are still substantial ambiguities and empirical inconsistencies.

Over the years, patterns of executive dysfunction in OCD have been extensively studied. These have suggested that deficits in certain domains of executive function, such as set shifting and inhibition, are relatively specific to OCD, while other domains such as planning, verbal fluency and working memory, are less impaired (5, 6). Furthermore, these deficits in set shifting and inhibition are trait-like, “stable characteristics that are consistently deviant from normal even during remission” [(5), p. 1032], meaning that they are independent of symptomatology, and thus may play an etiological role in the development or maintenance of OCD (5). These findings are supported by neuroimaging studies demonstrating under-activation in dorsal frontal-striatal regions in OCD patients during task-switching, while healthy controls showed significant activation in these areas (7).

These findings have led to an interest in the study of excessive cognitive rigidity in OCD [e.g., (8, 9)]. Experimentally, cognitive flexibility has widely been shown to be impaired in OCD [e.g., (9)], as well as in obsessive-compulsive personality disorder (OCPD) (10, 11), the latter being characterized by extreme perfectionism, order and neatness, often associated with anxiety, with a prevalence of 3–8% in adult populations (12). However, impaired cognitive flexibility has not been demonstrated to such an extent in other obsessive-compulsive spectrum disorders (OCSDs) such as trichotillomania (13), or other psychiatric disorders such as generalized anxiety disorder (14, 15). This suggests that low cognitive flexibility may act as a specific predictor of OCD, providing support for cognitive flexibility accounts of OCD. Impaired cognitive flexibility has also been demonstrated in unaffected first-degree relatives of both adult (8) and pediatric (6) OCD patients, suggesting that it may be an endophenotype (a heritable, intermediate marker of risk between phenotype and genetics) of OCD. Furthermore, this may be specific to OCD, as cognitive flexibility does not seem to be an endophenotype for other psychiatric disorders such as schizophrenia (16). The combination of clinically observed rigidity, and trait-like and even endophenotypic cognitive inflexibility in OCD patients, accounts for the focus on cognitive inflexibility as a potential causal factor, and highlights the importance of further study in this area.

Nonetheless, there has been some dispute over the association between cognitive inflexibility and OCD, with some studies finding broad impairments in executive function in patients with OCD (17, 18), rather than specific deficits in cognitive flexibility. A meta-analysis conducted by Fradkin et al. (19) found no evidence for impaired cognitive flexibility in OCD. However, this only included studies using measures of flexibility that are externally-cued (or reactive), such as set-shifting and task-switching, in which the participant must switch by reacting to cues, but not measures of self-directed (or generative) flexibility, in which the participant must internally generate the switching. Dissociations between performance on externally-cued and self-directed measures of cognitive flexibility have previously been demonstrated in patients with Parkinson's disease (20), and furthermore, the two types of cognitive flexibility have distinct neural correlates (21). Therefore, it may be that different paradigms in fact measure different aspects of the cognitive flexibility construct, and that only certain aspects of cognitive flexibility are impaired in OCD.

Another possible explanation for these conflicting findings may be that patients with clinical OCD often show more general cognitive impairment due to comorbidity with other forms of psychiatric disorder (22). It is thus of great value to study cognitive flexibility in subclinical populations to eliminate this confounding factor. In addition, as clinical and subclinical obsessive-compulsive symptoms share similar features and are continuous with one another (22), cognitive inflexibility in subclinical OCD may be a predisposing factor for the development of OCD. Impaired cognitive flexibility has been demonstrated in subclinical OCD by Sternheim et al. (23), using a set-shifting task and a self-report measure of cognitive flexibility. Although this study provides strong support for the cognitive inflexibility specificity hypothesis of OCD, the participants were all female students within a narrow age range. Student populations can be expected to show relatively high levels of cognitive ability, and so the sample may not be representative of the general population. Additionally, this study used an externally-cued, rather than self-directed, paradigm to measure behavioral flexibility. The present study aims to demonstrate the relationship between cognitive flexibility and subclinical OCD symptoms in more representative samples, using a self-directed, generative measure of cognitive flexibility, the Alternative Uses Task, AUT (24–26, 57). As well as flexibility, the AUT is also scored on elaboration, fluency and originality, which can act as controls for other cognitive processes, ensuring that inferences are based on flexibility specifically and not general cognitive ability or fluency.

Another longstanding theory of the cognitive origins of OCD, proposed by Graybiel and Rauch (27), is that OCD may be a clinical manifestation of an over-reliance on habits, involving maladaptive habit learning. Studies in both animals and humans have found evidence for a shift from goal-directed to habitual control in OCD, using behavioral measures such as outcome devaluation (28) and contingency degradation (29), as well as self-report measures of habits (60). However, very few studies have explored how individual differences in habitual tendencies may be associated with subclinical OCD traits in the general population [see (30)], and this analysis is crucial in order to augment our understanding of preliminary or predisposing OCD traits.

An apparent interplay between the different neural and cognitive mechanisms underlying OCD has led to suggestions that an interaction of several factors underlies OCD and its subtypes (4). This dimensional model suggests that the interaction of three factors (anxiety or emotional vulnerability, cognitive inflexibility, and an imbalance between goal-directed and habitual control over behavior) results in the development of OCD (4). To the best of our knowledge, previous studies have not explored whether this interaction is present, and if so, whether it may have a role in predicting or screening for clinical OCD. Therefore, the present study aims to determine whether individual differences in habitual tendencies and cognitive flexibility may interact to predict subclinical OCD symptomatology.

Consequently, the objectives of the present study were two-fold: firstly, to examine the extent to which behaviourally-assessed cognitive flexibility is negatively related to subclinical OCD symptomatology; and secondly, to explore the nature of any interactions between habitual tendencies and cognitive flexibility to study whether these, in conjunction, may predict OCD traits.

Study 1 sought to investigate associations between cognitive flexibility, subclinical OCD symptomatology, and the Habitual Tendencies Questionnaire (HTQ), in order to improve our understanding of the cognitive processes underlying OCD traits in the general population.

We addressed two main hypotheses as follows:

H1—Cognitive flexibility is negatively correlated with levels of OCD traits [as cognitive flexibility has been shown to be impaired in subclinical OCD, e.g., (23)] or OCPD [e.g., (10)].

H2—Habitual tendencies measured by the HTQ interact with cognitive flexibility in predicting OCD traits [due to OCD involving an over-reliance on habits, e.g., (27)].

Study 1 demonstrated that cognitive inflexibility and habitual compulsivity independently predict subclinical symptomatology, and that habitual compulsivity moderates the negative association between cognitive flexibility and OCD traits, such that it is significant at high but not low levels of habitual compulsivity. Study 2 was conducted in order to reproduce the findings of Study 1 in a larger sample. The aims of Study 2 were as follows: (1) to replicate the relationships of cognitive flexibility and compulsive habitual tendencies with OCD traits (as measured by the OCI); (2) to replicate the interaction effect between cognitive flexibility and compulsive habitual tendencies in predicting OCD traits.

Study 2 was preregistered on the Open Science Framework at the following link as “Hypothesis #5”: https://osf.io/3ag79/?view_only=b9fd67d034c7425dbf559c05d2421cb3. After the preregistration, some changes were made as follows: the data for binge eating, alcohol addiction, smoking habits and apathy were not analyzed in relation to habitual tendencies, as this was beyond the scope of the present paper.

The present study has demonstrated that behaviourally-assessed cognitive inflexibility predicts high levels of OCD traits in the general population, and that cognitive inflexibility interacts with habitual compulsivity to predict the highest severity of OCD symptoms. Notably, both rigidity and habitual compulsivity are necessary, and neither is sufficient alone for high OCD symptoms. Thus, in line with our first hypothesis, H1, individuals with lower cognitive flexibility showed increased subclinical OCD symptomatology. In line with our second hypothesis, H2, cognitive inflexibility interacted with habitual compulsivity to account for 49.4% of the variance in subclinical OCD symptomatology in Study 1, and 37.3% in Study 2, demonstrating reproducibility across two independent samples. Simple slope analyses demonstrated that cognitive inflexibility differentiated between high and low levels of OCD traits when habitual compulsivity was high, but not when it was low, suggesting that habitual compulsivity moderates the negative relationship between cognitive flexibility and subclinical OCD symptomatology. Thus, in a group of individuals with low habitual compulsivity, individuals with high cognitive flexibility will have low levels of OCD symptoms, but this effect will not be visible in individuals with high habitual compulsivity. Furthermore, the highest levels of OCD traits were seen in participants with high HTQ Compulsivity scores and low AUT Flexibility scores, suggesting a compensatory effect (see Figures 2C, 4C).

The present study demonstrated cognitive rigidity to be a significant independent predictor of OCD traits. This provides support for cognitive flexibility accounts of OCD, and suggests that impaired cognitive flexibility may be one of the core cognitive underpinnings of OCD. Across the literature, there is marked variation in methods of measuring cognitive flexibility, and this may help to explain the heterogeneity of the findings concerning impaired cognitive flexibility in OCD. Cognitive flexibility is a complex multidimensional construct that has been subdivided in a number of different ways. For example, it may be broken down according to the various tasks used to measure it, including set shifting, reversal learning, cued task switching, and cognitive or motor inhibition tasks (39). Kanen et al. (40) further subdivided one of these tasks, and demonstrated that a subcomponent of probabilistic reversal learning was increased in individuals with substance use disorder, but decreased in individuals with OCD, relative to controls. Similarly, a meta-analysis by Chamberlain et al. (41) found that extra-dimensional, but not intra-dimensional, set-shifting is consistently impaired in OCD. These findings suggest that not all aspects of cognitive flexibility may be impaired in OCD. The difficulty with which rule changes may be detected, and thus the levels of feedback processing required, varies across the wide range of tasks used to measure cognitive flexibility (19). Therefore, deficits in feedback processing may, in part, account for the cognitive inflexibility found in OCD (19). The use of self-directed tasks to measure cognitive flexibility, such as the AUT used in the present study, may eliminate the confounding effects of feedback processing, as the shifting is internally generated. Therefore, these tasks may be able to specifically target the subdimensions of cognitive flexibility impaired in OCD.

The subdivision of cognitive flexibility into its constituent parts may be an important future direction in gaining more specific insight into the cognitive processes underlying OCD, as well as the heterogeneity of OCD symptoms (42). Indeed, another explanation for the highly mixed nature of the literature regarding the cognitive deficits associated with OCD may be the wide range of its clinical presentations. In a meta-analysis by Bragdon et al. (42), increased symmetry and ordering symptoms were associated with reduced cognitive flexibility (measured using a mixture of externally-cued and self-directed tasks across the included studies) and verbal working memory, while increased obsessing and checking symptoms were associated with poorer memory and verbal memory. Therefore, in addition to the subdivision of constructs such as cognitive flexibility, it may also be important to separate OCD into its different symptom groups, in order to gain more specific insight into its cognitive underpinnings.

Furthermore, the moderation of cognitive flexibility by habitual tendencies adds nuance to the habit hypothesis of OCD proposed by Graybiel and Rauch (27), suggesting that cognitive flexibility must be studied along with habitual tendencies, thus building on previous contributions to this debate. For example, Gillan et al. (43) found that excessive habits in OCD patients were associated with hyperactivation in the caudate nucleus, an important region for goal-directed behavior that is implicated in the pathophysiology of OCD. Therefore, habit-forming biases in OCD may be a result of impairments in this neural circuitry. Another study by Gillan et al. (28) used outcome devaluation paradigms to demonstrate increased habitual tendencies in OCD patients. As the tasks used in Gillan et al. (28) study did not involve obsessions, it has been suggested that the tendency toward developing compulsive-like habits in OCD is independent of obsessions, and thus compulsions are a core component of OCD (44). Gillan et al. (45) further found that individuals with OCD showed increased habitual responding in a shock-avoidance paradigm, and when trying to explain their behavior, erroneously deduced that if they felt compelled to perform the habitual response, then there must be something to fear. In contrast with the majority of the preceding literature on OCD, this reverse inference suggests that obsessions may in fact arise as a result of compulsions. This provides further evidence for the habit hypothesis of OCD, and emphasizes the importance of studying compulsivity in relation to habits and OCD. The HTQ, and in particular its Compulsivity subscale, may be valuable as a predictive or screening measure for OCD traits (30). The present study extends the potential applications of the HTQ, such that it may be used in conjunction with measures of cognitive flexibility as an even stronger predictor of subclinical OCD symptomatology.

The moderation of cognitive flexibility by habitual tendencies suggests that neither construct alone is sufficient to explain the roots of OCD, but that the interactions between the two must be studied in order to develop more comprehensive theories. A recent review conducted by Fullana et al. (46) was unable to identify any biomarkers with diagnostic specificity for OCD. However, as biomarkers are only one small aspect of the RDoC framework, perhaps more integrative RDoC approaches using combinations of biomarkers and their interactions with other constructs may provide a solution. In addition, future use of interactionist, individual differences approaches may help to explain past empirical inconsistencies regarding excessive cognitive inflexibility and habitual tendencies in OCD [e.g., (17, 18, 47)].

The importance of studying individual differences in cognitive flexibility, habitual tendencies, and the interactions between them in order to elucidate the cognitive processes underlying OCD is emphasized by neurobiological findings. Vaghi et al. (48) discovered that reduced functional connectivity between the caudate and the ventrolateral prefrontal cortex was selectively associated with reduced cognitive flexibility, while reduced functional connectivity between the putamen and the dorsolateral prefrontal cortex was selectively associated with goal-directed performance, as well as OCD symptom severity. Another double dissociation was found by Tyagi et al. (49), who showed that anteromedial subthalamic nucleus deep brain stimulation in patients with treatment refractory OCD selectively improved cognitive flexibility, while ventral capsule/ventral striatal deep brain stimulation selectively improved mood. Deep brain stimulation at each site independently reduced OCD symptoms. The above neurobiological findings are reflected in the integrated model of OCD proposed by Robbins et al. (4), and corroborate the importance of the three main factors in OCD of varying severity, namely cognitive inflexibility; habitual rather than goal-directed behavior; and mood. Furthermore, these neurobiological findings have potential therapeutic applications for those suffering from severe, treatment-refractory OCD.

In order to ensure that the present study was well-powered to detect the present effects, we conducted a conservative post-hoc power analysis (α = 0.001, n = 130 for Study 1; n = 259 for Study 2), using the pwr package in the statistical software R Studio. For the effect sizes, we used the smallest relevant Pearson's correlations: 0.390 for Study 1 and 0.362 for Study 2, which reflected the correlations between AUT flexibility and the OCI (see Tables 1, 4). This revealed power of 0.912 for Study 1 and 0.997 for Study 2, indicating that the sample sizes were sufficient. Future studies using even larger sample sizes would help to replicate and extend the present findings. The present investigation furthers our understanding of the cognitive underpinnings of obsessive-compulsive symptoms, illustrating their interactive underpinnings in rigidity and habitual tendencies. Replication of these findings in children and adolescents would be useful in considering developmental trajectories and in exploring whether early compulsivity or cognitive rigidity can help to predict the development of OCD in later life. This could have important implications in the development of early interventions for OCD. Notably, adolescents with OCD do not show deficits in performance on the Wisconsin Card Sort Test, an externally-cued measure of cognitive flexibility (50), and as such, an important future direction would be to administer a self-directed measure of cognitive flexibility, such as the AUT, in this age group. In order to explore whether the present findings are consistent across cultural contexts, cross-cultural replication is imperative. Interestingly, cognitive rigidity may also account for the finding of heightened moral rigidity in OCD (51). Furthermore, impaired cognitive flexibility is not a characteristic limited to OCD, but has been demonstrated in other disorders, such as binge eating disorder (52) and addictions (53). Therefore, it is important to explore whether habitual tendencies, especially compulsivity, interact with cognitive inflexibility to predict symptom severity in these other conditions, or whether this interaction effect is specific to OCD.

The present study is not without its limitations. Firstly, the use of an online convenience sample may not be representative of the general population, and so future research may administer the study in a laboratory setting or a population sampled specifically for representativeness. Secondly, although the present findings are in a healthy population without any formal clinical diagnoses of OCD, a number of participants did show high levels of OCD symptoms, with mean OCI-R scores of 16 in both samples. Although this is just above the cut-off score of 15 for mild vs. moderate severity in Abramovitch et al. (54) paper, it is well below the cut-off score of 21 for clinical OCD suggested by the developers of the OCI-R (32). Nevertheless, replication in samples with lower mean OCI scores, as well as in patients with clinically diagnosed OCD and OCPD would be valuable. A limitation of the OCI-R used in the present study is its inclusion of hoarding, which has recently been suggested to be a separate disorder. A fine-grained approach to OCI symptom specificity is thus encouraged in future studies. Thirdly, anxiety and depression may also be associated with lower cognitive flexibility, so future studies may address this by controlling for these factors. Finally, the present study focuses on one aspect of the cognitive flexibility construct, by using the AUT. In order to fully characterize the cognitive flexibility construct, future research must deconstruct it, by including additional measures of cognitive flexibility, such as the Wisconsin Card Sort Test (55), or the Intra-Dimensional Extra-Dimensional set-shift CANTAB task (56).

To conclude, the present study demonstrated that both cognitive flexibility and habitual compulsivity acted as independent significant predictors of subclinical OCD symptomatology, but also exhibited an important interaction. The interaction between cognitive flexibility and habitual compulsivity accounted for almost half of the variance in subclinical OCD symptomatology. Furthermore, both cognitive inflexibility and habitual compulsivity are necessary for high levels of OCD symptoms, while neither is sufficient alone. These findings may prove useful for future research into both subclinical and clinical OCD traits, as well as other disorders involving cognitive inflexibility. The results may also help in the development of interventions targeting the impaired cognitive flexibility and maladaptive habits proposed to underlie OCD, thus helping to alleviate the debilitating symptoms that cause so many disruptions and difficulties in patients' daily lives.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by Cambridge University Psychology Department Research Ethics committee. The patients/participants provided their written informed consent to participate in this study.

SR, TR, and LZ: study conceptualization and write-up. SR and LZ: data collection and analysis. All authors contributed to the article and approved the submitted version.

This work was funded by Wellcome Trust, Grant/Award Number: 104631/z/14/z to TR and from Gates Cambridge Scholarship to LZ.

The 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.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.