Over the past three decades, the cerebellum has been recognized as playing a role in a variety of cognitive and social functions as well as affective regulation (Tedesco et al., 2011; Buckner, 2013; Mariën et al., 2014; Sokolov et al., 2017; Leggio and Olivito, 2018; Stoodley and Schmahmann, 2018; Schmahmann, 2019). However, how the cerebellum peculiarly operates in such multifarious domains is still being discussed. According to the “sequence detection theory” postulated by Leggio and colleagues (Leggio et al., 2011) the cerebellum is engaged for detecting and simulating repetitive patterns of temporally and spatially organized events. Once the sequence of events has been identified, the cerebellum contributes to the implementation of internal models that allow one to make predictions about the consequences of events and to detect any discrepancies between expected and occurring events, thus making online corrections when errors occur (Ito, 2008; Leggio et al., 2011; Leggio and Molinari, 2015). Specifically, the cerebellum compares actual input and preceding stimuli, makes associations among them in space and time, and then tests potential discordances. When an incoming stimulus corresponds to the predicted one, cerebellar output is marginal. On the contrary, when a discordance or an error is revealed, the activity of the cerebellum grows, alerting specific areas of the cerebral cortex based on the type of stimuli (Ito, 2008). Accordingly, sequencing is not acknowledged as a discrete cognitive function, instead is defined as a supramodal function that subserves several functional domains ranging from sensorimotor to advanced decision-making domains (Molinari and Masciullo, 2019), supported by different brain networks. Through the cerebello-cerebral collaboration, sequences of events that have been met earlier are employed to create internal models, in turn used to make predictions. To identify sequential relationships between events, the cerebellum must understand the links between them in the space-time domain. Each event is retained and then compared with a subsequent event, matching their characteristics so that the cerebellum can place them correctly in space and time. Subsequently, due to the presence of the already established internal models, the cerebellum makes it possible to predict the succession of future events by setting the correct excitatory levels in brain areas responsible for responding adaptively to a given pattern of stimuli if it is recognized as a previously encountered model (Molinari and Masciullo, 2019). In this way, cerebellar detection and prediction of sequential events contributes to feedforward control based on anticipation, ultimately reducing the prediction error. Thus, the cerebellum represents a chief structure to be included in the “predictive brain” (Ito, 2008; Molinari and Masciullo, 2019).

Cerebellar sequential functioning has been demonstrated to contribute to information processing in the sensorimotor system (Tesche and Karhu, 2000; Doyon et al., 2002; Restuccia et al., 2007; Morgan et al., 2021) and for cognitive functions, such as visuospatial functions (Molinari et al., 1997; Leggio et al., 1999), language (Fabbro et al., 2000; Leggio et al., 2000, 2011; Tedesco et al., 2011; Lupo et al., 2021a), verbal working memory (Silveri et al., 1998; Chiricozzi et al., 2008; Leggio et al., 2011; Sheu and Desmond, 2022) and script sequences (Leggio et al., 2008; Molinari et al., 2008; Morgan et al., 2021).

More recently, the role of the cerebellum as a sequential and prediction processor has also been investigated in the social domain, with a focus on Theory of Mind (ToM) skills. In this domain, the construction of internal models reflects the correct implementation of the sequence of social actions, thus allowing people to predict one’s own and others’ behaviors and reactions, and to adjust unexpected events when violations from predicted scenarios are met to finally adjust social interaction accordingly by modulating cerebral cortex activity (Clausi et al., 2019; Heleven et al., 2019; van Overwalle et al., 2019a,b, 2022). The contribution of lobules Crus I and II has been evidenced for those functions, which is consistent with the widespread connections the posterior cerebellum has with areas in the cerebral cortex involved in high-order functions (van Overwalle and Mariën, 2016; van Overwalle et al., 2022).

In line with sequence detection theory, the cerebellum is recognized as a prediction machine that harmonizes whole-brain function by updating certain neural networks about the required responses to a given stimulus in each context (Ito, 2008; Leggio and Molinari, 2015), providing internal coherence between externally and internally produced signals (D’Angelo and Casali, 2012), all functions needed to properly interpret events and to adapt to changes. Impaired prediction abilities have been proposed as the basis of the behavioral deficits in individuals with disorders of different etiologies, such as patients with primary cerebellar damages (Rustemeier et al., 2016; Clausi et al., 2019), individuals with neurodevelopmental conditions, such as autism (Sinha et al., 2014; Siciliano and Clausi, 2020), and psychiatric disorders, such as schizophrenia (D’Angelo and Casali, 2012; Abram et al., 2022), all conditions in which cerebellar alterations have been documented (D’Mello et al., 2015; Mothersill et al., 2016; Clausi et al., 2021b).

The acknowledged role of the cerebellum in advanced and supramodal functions involving affection and cognition has widely broadened the spectrum of pathologies in which a role of the cerebellum has been hypothesized and investigated (Baldaçara et al., 2008; D’Mello and Stoodley, 2015; Lupo et al., 2019; Moreno-Rius, 2019; Siciliano et al., 2022). It is assumed that, despite the specific symptomatology, cerebellum-related cognitive and affective signs might reflect either exaggerated-hypermetric (overshoot) or diminished-hypometric (undershoot) reactions to externally or internally triggered stimuli, likely to be busted by poor cerebellar modulatory function to widespread brain networks (Schmahmann et al., 2007). In line with this and with the above-mentioned sequence detection theory, the cerebellum guarantees the proper equilibrium between processing of endogenous and exogenous stimuli and, thus, allows to choose the best corresponding responses in the environment, by modulating the activity in specific brain networks (e.g., sensory, motor, memory, attention, language, affective and social) (Schmahmann et al., 2007; D’Angelo and Casali, 2012; Leggio and Molinari, 2015).

Recently, a particular focus has been placed on cerebellar engagement in bipolar disorders (BD). BD is a debilitating and chronic psychopathology characterized by phases of depression and mania or hypomania, with varying inter-episodes of symptoms’ remission. Specific neural mechanisms involving cortical-striatal-limbic circuits are supposed to trigger the diverse mood states that characterized the two peculiar phases of BD (Spielberg et al., 2016). While diverse meta-analytic studies have currently proven the presence of structural changes in certain subcortical areas in BD, such as hippocampus, thalamus, and amygdala (Rossi et al., 2013; Hibar et al., 2016; Ding and Hu, 2021), considerable inconsistency between studies still exists about cerebellar structural changes in BD (Ding and Hu, 2021). However, the involvement of the cerebellum in this disorder is currently being investigated on the basis of clinical and scientific evidence suggesting its implication in the pathology (Mills et al., 2005; Argyropoulos et al., 2021; Lupo et al., 2021b; Olivito et al., 2022a,b). For example, the overshoot and undershoot that characterize responses to external or internal stimuli following cerebellar alterations have been associated with the states of mania and depression typical of bipolar disorder (Schmahmann et al., 2007). An imbalanced “mind-world synchronization” driven by cerebellar dysfunctional prediction ability has been hypothesized to contribute to the onset of psychotic symptoms and might be also associated with the poor mood homeostasis that leads to depressive and manic phases typical of BD (Ito, 2008; Moberget and Ivry, 2019). In support of the hypothesis of a cerebellar role in affective regulation, the occurring of both depressive and manic symptoms has recently been reported in patients with cerebellar disease of different etiology, creating a bridge between the two disorders (Clausi et al., 2019; Lupo et al., 2019). Though, the interest in deepening the contribution of the cerebellum in BD brain networks has increased lately (Argyropoulos et al., 2021; Olivito et al., 2022a), promoted by the plethora of studies that are nowadays available about the anatomical and functional connections that link the cerebellum and subcortical limbic and cortical associative areas involved in emotional and affective regulation (Stoodley and Schmahmann, 2009; Adamaszek et al., 2017; Baumann and Mattingley, 2022; Schutter, 2022). In BD, the persistence of functional connectivity changes during the euthymic phases has been reported. It has been hypothesized that they represent a vulnerability resulting from emotional brain networks (i.e., anterior limbic network) that become hypersensitive due to acute episode, in turn representing a risk for persistent affective and cognitive disturbances in BD in the euthymic phase (Strakowski et al., 2005; Syan et al., 2018; Olivito et al., 2022a). In this regard, the interest in examining the neural and behavioral mechanisms underlying the phases of symptom remission in BD has grown. In BD patients in the euthymic phase, cognitive deficits, such as visuospatial and verbal difficulties, and affective and ToM impairments have been described, strongly impacting patients’ quality of life (Martinez-Aran et al., 2007; Bora et al., 2009; Gogos et al., 2010; Osher et al., 2011; Samamé et al., 2015; de Siqueira Rotenberg et al., 2020; Olivito et al., 2022b). Furthermore, decreased volumes in the gray matter (GM) of cerebellar vermian and posterior areas have been documented (Mills et al., 2005; Lupo et al., 2021b) as well as disrupted functional and structural connectivity, which likely represents chronic impairments that persist during periods of symptom remission (Argyropoulos et al., 2021; Olivito et al., 2022a). In addition, a recent study revealed specific patterns of GM decrease in cerebellar vermis and Crus I and II both in patients with bipolar disorder type 1 and 2 in the euthymic phase, likely to explain the occurrence of ToM deficits (Olivito et al., 2022b). It must be noted that sequential abilities and prediction abilities related to social and non-social functions have never been investigated in BD.

In a previous study, we reported ToM and cerebellar alterations in BD in the euthymic phase. In the present study, we aimed to further explore these features in this population by also focusing on sequential and predictive processing, which represent the key modes of cerebellar processing. To this end, we used implicit and explicit sequential tests with or without ToM engagement, both requiring prediction processing. Specifically, we employed the Faux Pas test (Stone et al., 1998; Liverta Sempio et al., 2005), that explicitly requires to process and predict advanced social cues and implicitly needs to order social events sequentially to be well performed, and the Sequence test (Leggio et al., 2008), that allows to evaluate sequencing and predicting abilities demanding the processing of stimuli that entail behavioral, verbal, or spatial/abstract elements. Furthermore, we aimed to verify the presence of cerebellar alterations by conducting a voxel-based morphometry (VBM) analysis on MRI data of BD patients.

The key role of the cerebellum in adaptive control and predictive coding by means of a sequential operational mode has been documented by many studies in diverse domains ranging from sensorimotor to advanced cognitive and ToM functions (Ito, 2008; Molinari et al., 2008; Leggio et al., 2011; D’Angelo and Casali, 2012; Sokolov et al., 2017; Clausi et al., 2019; Heleven et al., 2019; van Overwalle et al., 2019a,b). Nevertheless, thus far, most of those studies have focused on investigating the cerebellar operational mode in healthy individuals and in patients with pure cerebellar lesions or degenerative disorders. Cerebellar alterations and ToM deficits have been described in patients with remitted BD (Shamay-Tsoory et al., 2009; Bora et al., 2016; Argyropoulos et al., 2021; Olivito et al., 2022b). However, no study has previously focused on the ability to sequence and predict events in this clinical population. To address this gap in the current study, we investigated the performance of individuals with bipolar disorder in the euthymic phase in non-explicitly sequential ToM tasks and sequential non-ToM tasks, both requiring prediction capacities. In addition, VBM analysis was used to explore the patterns of cerebellar GM and assess whether our sample of BD patients showed the concomitant presence of cerebellar structural alterations and sequential and predicting deficits.

BD patients performed significantly worse in the Faux Pas Test than HS. Specifically, the ability to identify the occurrence of the faux pas and to justify others’ behavior when the faux pas was correctly identified were impaired in BD patients, whereas a trend toward significantly worse performance was detected for the affective component. Our results on preserved affective component of ToM are in line with previous studies on BD (Montag et al., 2010; Mercer and Becerra, 2013). While most studies have regularly reported deficits in the cognitive component of ToM in BD, inconsistencies have often occurred as concerned with the affective component (Shamay-Tsoory et al., 2009; Ioannidi et al., 2011). In the present study, BD revealed to be able to “feel” the emotional impact that an inconvenient situation could have on the person who might be hurt by it, but they were unable to use this information to provide appropriate justifications on the protagonist’s behavior, demonstrating cognitive ToM difficulties. In previous studies it has been speculated that BD patients typically generate extreme emotional responses due to an over activation of limbic regions during the appraisal of emotional material (Shamay-Tsoory et al., 2009). This hyperactivation has also been hypothesized to account for “hyper empathy” that BD show and to be related to affective ToM inconsistency found among studies, in which BD often reveal to process affective ToM information as healthy subjects do (Shamay-Tsoory et al., 2009). Additionally, affective ToM impairments are reported to be higher during the acute phases of BD, and to normalize during the euthymic phases (Bora et al., 2016). This effect might be related to affective regulation and stabilization due to medications, that might impact more on affective processing of stimuli rather than on cognitive appraisal of those. The ability to adequately respond to the questions posed by the examiner partly entails sequential processing; however, it is not as explicitly required as noted in the Sequence Test. Indeed, to identify the faux pas and to understand that a person has said something awkward with respect to expected behavior, it is necessary for the individual to implicitly understand the correct sequence of social actions occurring in the story and to detect an error compared to a properly predicted behavior (Clausi et al., 2019). In stories with no faux pas, BD patients exhibited performance equal to HS. Indeed, in these stories, it is easier to predict others’ social behavior given that the social situations present no errors to be detected. Thus, when no conflict occurs between expected and current conditions, individuals can rely on internal models of previous experiences with no need for correction, whereas a higher predictive ability is crucial to properly identify the faux pas occurrence. In the latter, questions are asked about why the person committed a faux pas and why he or she should not have done so. Thus, this test investigates a complex level of ToM that entails the ability to predict and verbalize inferences about specific consequences of a series of unexpected and inconvenient events, requiring online comparisons between external stimuli and internalized models of social expectations (de Siqueira Rotenberg et al., 2020).

An explicit demand in terms of sequential processing is associated with the Sequence Test, whereas no ToM component is involved. In this test, BD patients exhibited poorer performance than HS in all three conditions examined, although for the verbal sequences the difference only denotes a trend toward a significance. Our results revealed that sequential difficulties are experienced by these patients despite the content of the stimuli. In the Sequence Test, the high sequential demand is associated with the need to detect the errors associated with the non-logical sequence as presented to the subjects, to identify the proper associations between events, to determine precedence and to finally place events in chronological order to logically reconstruct the sequence (Leggio et al., 2008). The requisite to assemble cards in the proper order implies making predictive connections among the cards. Consequently, the sequence misinterpretation might be due to the inability to detect the errors in the non-logical sequences presented and thus to properly predict the way each card is temporarily and spatially linked to the next. Previous studies found that patients with cerebellar disease failed to accomplish this test regardless of the specific content (Leggio et al., 2008; Tedesco et al., 2011).

The pattern of GM reduction found in BD patients includes both anterior and posterior cerebellar lobules, which is consistent with the results of previous studies we performed using BD patients (Lupo et al., 2021b; Olivito et al., 2022b). The reduced GM in the anterior cerebellar areas might be related to psychomotor agitation that typically characterizes affective episodes (Kraepein, 1921; Sani et al., 2016) and that could lead to persisting structural changes in motor and sensorimotor regions of the cerebellum, such as lobules I-IV and V (Lupo et al., 2021b). On the other hand, the GM reduction in the posterior cerebellum is consistent with the behavioral profile of BD patients and with the well-known functional topography of the cerebellum (Sokolov, 2018; Stoodley and Schmahmann, 2018). Indeed, a larger cluster of significant GM reduction was evidenced in the right Crus II, a region known to be involved in mentalizing processes, primarily when a high level of abstraction is in demand (van Overwalle and Mariën, 2016; van Overwalle et al., 2020). Several fMRI studies on healthy subjects have acknowledged the activation of the posterior cerebellum, specifically the Crus I and II, together with cerebral areas known to be part of the social brain during social mentalizing both when the tasks required and did not require sequential processing (van Overwalle and Mariën, 2016; van Overwalle et al., 2022). Together with functional activation, cerebellar Crus II and mentalizing brain regions have been found to be effectively connected by closed loops, further demonstrating the cerebellar modulation of advanced social processing (van Overwalle et al., 2020). In addition, dysfunctional advanced ToM skills, which were investigated by means of the FP test, have been previously related to decreased GM in cerebellar Crus II in patients with pure cerebellar disease (Clausi et al., 2019, 2021a,b) and in individuals with autism spectrum disorders (Clausi et al., 2021b) and bipolar disorder types 1 and 2 (Olivito et al., 2022b). Coherently, altered functional connectivity between the Crus II and social mentalizing areas has been reported in all the above-mentioned clinical conditions (Olivito et al., 2018, 2022a; Clausi et al., 2019). The most sustained hypothesis holds that structural alterations in the cerebellar posterior areas might impede the proper functional modulation that the cerebellum exerts on regions of the neocortex, thus affecting appropriate high-order functions (Clausi et al., 2019). Additionally, the interplay among the posterior cerebellum, the prefrontal cortex and temporoparietal cortex is associated with the ability of the human brain to construct internal models of complex stimuli occurring in the external world and to control inner manipulation of related processing and thought (Ito, 2008). Consistently, a complex human process such as prediction error, recruits the most lateral part of the cerebellum (Ploghaus et al., 2000). Additionally, activity in the right posterior cerebellum is triggered when people are asked to predict an event in the future, whereas no cerebellar activation was detected for past episode recollection. These findings likely indicate that the posterior cerebellum might be pivotal for the creation or reshaping of novel internal models based on experiential models (Szpunar et al., 2006; Ito, 2008).

The prediction capacity of the cerebellum is currently considered a supramodal function that modulates whole-brain function alerting specialized brain circuits based on the type of incoming input and of the response required in the specific context (Tesche and Karhu, 2000; Restuccia et al., 2007; Moberget et al., 2014; Molinari and Masciullo, 2019). The participation of the cerebellum in the predictive brain is ascribed to its role in detecting sequences and violations from correct sequences and creating internal models of those sequences used to make proper predictions, thus allowing feedforward control based on anticipation (Molinari and Masciullo, 2019). Sequential processing represents the basic functional mechanism of the cerebellum in several functional domains, including motor and cognitive control, language, spatial processing, decision-making, and theory of mind (Braitenberg et al., 1997; Leggio et al., 2008; Molinari et al., 2008; Tedesco et al., 2011; Clausi et al., 2019; van Overwalle et al., 2019b). As previously stated, poor performance in the Sequence Test has been reported in patients with cerebellar diseases (Leggio et al., 2008; Tedesco et al., 2011). Specifically, patients with lesions occurring in the posterior cerebellum were associated with worse performances (Tedesco et al., 2011). Interestingly, our sample of BD patients showed altered performances in the Sequence Test in presence of a large cluster of decreased cerebellar GM in posterior regions, such as the right lobule Crus II with extensions in the right lobules VI and Crus I and in the left Crus I-II.

It is worth noting that no influences of intellectual level and executive functions were found on patients’ sequences and ToM performances, as revealed by the absence of correlations among those (see Supplementary Table 3 in Supplementary materials). The only correlation we found occurred between the behavioral condition of the Sequence Test and the Corsi forward test. This result is consistent with the knowledge that to recognize the proper order of specific stimuli, the information extrapolated from one stimulus and the comparisons and the predictive link between stimuli must be kept active in a working memory system (Leggio et al., 2008, 2011).

Overall, the results of the present study indicate BD patients’ impairment in ToM and sequential abilities that require prediction processing and error detection and correction to accomplish the requested tasks. Consistent with widespread literature, we suggest that the above-mentioned compromised performances might be related to reduced GM occurring in the posterior cerebellar areas. It is likely that the structural alterations in the posterolateral cerebellum might affect the cerebellar modulation on cortical projection areas, which is necessary to inform about the proper sequential responses to be made when unexpected events are met. As a supramodal function of the cerebellum, prediction failure might influence cognitive and social domains in various components, thus impeding the optimization of a range of functions and behaviors (Clausi et al., 2019). Accordingly, damage to the cerebellum has been described as a shared risk factor among different psychopathologies, emphasizing the usefulness of focusing on transdiagnostic dimensional features between categorical mental disorders for both empirical research and applied practice (Romer et al., 2018; Siciliano et al., 2022). In this regard, we suggest that the impact of prediction deficits should be considered when assessing a patient’s psychosocial functioning and his or her quality of life. In the context of our study, we propose that the imbalanced mood homeostasis typical of BD might be accompanied by an impaired capacity to properly process temporal and spatial information between stimuli, occurring in both the external and internal world. Under normal conditions, the cerebellum is needed to create appropriate internal models of actual events for use in the future for proper interpretations and prediction of events and, in general, to adapt behaviors to a given context and situation. Thus, our preliminary results provide the breeding ground for future hypothesis-testing studies assessing the causal link between cerebellar alterations and sequential and predictive deficits in BD.

Some limitations of the present study must be mentioned. Our sample of BD patients included both patients with BD types 1 and 2 given that the total number of patients recruited did not allow us to differentiate and characterize the two different subtypes. Thus, future research is needed to further characterize predictive ToM and sequential processing and the direct link with subtending cerebellar structural alterations in the two BD subtypes.

Finally, as reported in Supplementary Table 2, the BD patients enrolled were medicated, many of whom were receiving treatment involving polypharmacotherapy. Indeed, the enrollment of drug-free patients was not possible due to ethical and clinical concerns and given that pharmacotherapy is essential for patients to achieve a clinical stability pivotal for remission of symptoms and to maintain a euthymic state. In addition, it has to be considered that, although it is possible that research with participants receiving psychotropic medications might reveal confounding imaging results, most of the evidence from morphometric imaging studies shows that the effect of medication is unlikely to explain the neuroimaging parameters’ variances occurring between bipolar individuals and healthy controls (Hafeman et al., 2012).

Additionally, the present study does not provide evidence of a causal link between the observed changes in GM in the cerebellum and the reported performances on behavioral tasks. In particular, since we did not detect a monotonic relationship between each pair of variables, namely between values of GM reduction in each lobule and the performances at each SC tests, we could not conduct correlation analysis.

Despite these limitations and the preliminary nature of the research, the present study represents the first investigation proving the presence of sequencing and predicting deficits in BD patients, and the first step to provide a theoretical link between these deficits and cerebellar alterations.

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 Ethics Committee of Fondazione Santa Lucia IRCCS (Prot. CE/PROG.579 and 22 November 2016). The patients/participants provided their written informed consent to participate in this study.

LS, GO, MLu, and MLe: conceptualization. LS, GO, and MLu: methodology, writing—original draft preparation, and writing—review and editing. LS, GO, MLu, and NU: formal analysis and data curation. LS, GO, MLu, NU, AG, and MS: investigation. RD and MLe: supervision. All authors have read and agreed to the published version of the manuscript.