The role of the cerebellum in schizophrenia: an update of clinical, cognitive, and functional evidences.

The role of the cerebellum in schizophrenia has been highlighted by Andreasen's hypothesis of "cognitive dysmetria," which suggests a general dyscoordination of sensorimotor and mental processes. Studies in schizophrenic patients have brought observations supporting a cerebellar impairment: high prevalence of neurological soft signs, dyscoordination, abnormal posture and propioception, impaired eyeblink conditioning, impaired adaptation of the vestibular-ocular reflex or procedural learning tests, and lastly functional neuroimaging studies correlating poor cognitive performances with abnormal cerebellar activations. Despite those compelling evidences, there has been, to our knowledge, no recent review on the clinical, cognitive, and functional literature supporting the role of the cerebellum in schizophrenia. We conducted a Medline research focusing on cerebellar dysfunctions in schizophrenia. Emphasis was given to recent literature (after 1998). The picture arising from this review is heterogeneous. While in some domains, the role of the cerebellum seems clearly defined (ie, neurological soft signs, posture, or equilibrium), in other domains, the cerebellar contribution to schizophrenia seems limited or indirect (ie, cognition) if present at all (ie, affectivity). Functional models of the cerebellum are proposed as a background for interpreting these results.

Introduction

The cerebellum is traditionally regarded as an organ that subserves coordination, balance, gait, and fine motor control. Nevertheless, its involvement in cognition has also been suggested1 and a cerebellar dysfunction could underlay psychiatric disorders such as schizophrenia2 or autism.3 The potential role of the cerebellum in schizophrenia has been highlighted by Andreasen's hypothesis of “cognitive dysmetria,” which suggests an impaired sequencing and coordination of sensorimotor and mental processes,2 resulting in dyscoordinated language and disordered thought, the core feature of the cognitive dysmetria hypothesis. Functional brain imaging studies have suggested the involvement of cortico-cerebellar networks in cognition.4

Converging evidence suggests that schizophrenia may be associated with cerebellar anomalies. Neuropathological studies have reported a reduction in size and density of Purkinje cells,5 and magnetic resonance spectroscopy studies have shown an altered expression of synaptic proteins in the cerebellum of schizophrenic patients.6 In addition, an atrophy of the cerebellar vermis has been described, though not always found.7 There is also available evidence that a cerebellar dysfunction could underlay some of the clinical psychiatric and neurological symptoms as well as cognitive dysfunctions observed in schizophrenia. Recent reviews have already focused on the implication of the cerebellum in psychiatric illnesses in general,8 in emotion9 or behavior and cognition.10 Here, we propose an updated and synthetic review of clinical, cognitive, behavioral, and functional neuroimaging studies supporting the involvement of the cerebellum in schizophrenia. We aimed to highlight the domains where convincing evidence is already available. Some functional models of the cerebellum will be briefly discussed, yet the review of all hypothesized functions of the cerebellum itself, is beyond the scope of this review.

METHOD

For every specific domain, Medline searches, limited to studies in humans and in English language, were performed in order to be as exhaustive as possible. Preference was given to studies later than 1998, except when the search retrieved mainly older studies. All relevant works, either supporting or challenging a cerebellar dysfunction in schizophrenia, were included. In general, the searches included the keywords “cerebellum” + “schizophrenia” and 1 or 2 terms used in the text below as section titles, eg, “affectivity” or “saccades”. Variants or related terms were also used, such as “affect,” “affective,” or even “emotion.” In the case of functional imaging studies, special attention was paid to the inclusion of the cerebellum in the analysis because it is often excluded due to technical constraints. When cited studies did not include the cerebellum in the analysis, this is clearly stated.

Cerebellum and Psychiatric Symptoms

Hallucinations

A few brain imaging studies in schizophrenic patients have provided evidence of the involvement of the cerebellum in hallucinations reported (table 1). Significant reductions in gray matter volume in the cerebellum, bilaterally, in the left superior temporal gyrus and the left thalamus were reported correlated with higher scores in the hallucinations item of the Brief Psychiatric Rating Scale.11 Two studies functional magnetic resonance imaging (fMRI) studies by Shergill et al12,13 investigated the neural basis of inner speech and auditory verbal imagery in schizophrenic patients with prominent auditory hallucinations. Compared to controls, patients showed an attenuated activation in a cortical-subcortical network involving the posterior cerebellar cortex while imagining external speech. The authors concluded that hallucinated patients may present a dysfunction in areas implicated in verbal self-monitoring. Yet, 2 other studies using semi-automated algorithms to analyze structural MRIs of schizophrenic patients found correlations of structural changes with the severity of hallucinations.14,15 Shin et al14 reported greater gray matter in the frontal and temporal cortices of hallucinated patients vs non-hallucinated patients, but not in the cerebellum, while Gaser et al15 did not examine the cerebellum.

Table 1

Involvement of the cerebellum in schizophrenia – Non-cognitive aspects.

Psychiatric symptoms
Hallucinations	Shergill et al, 2003 13	+++
	Shergill et al, 2000 12	+++
	Neckelman et al, 2006 11	++
	Gaser et al 2004 14	NIA
	Shin et al 2005 15	−
Formal Thought Disorder	Kircher et al, 2001 20	+++
	Kircher et al, 200118	+++
	Levitt et al, 1999 21	++
	Weinstein et al, 2006 17	−
	Kircher et al, 2002 19	−
Affect	Stip et al, 2005 25	+++
	Paradiso et al, 2003 26	+++
	Abel et al, 2003 27	+++
	Takahashi et al, 2004 24	−
Neurological symptoms
Neurological Soft Signs	Bottmet et al, 2005 7	+++
	Mouchet-Mages et al, 2007 43	+++
	Keshavan et al, 2003 44	−
	Dazzan et al, 2004 45	NIA
	Schroder et al, 1995 46	NIA
Cerebellar symptoms	Ho et al, 2004 33	++
	Varambally et al, 2006 47	++
Motricity & propioception	Daskalakis et al, 2005 50	+++
	Muller et al, 2002 53	+++
	Stephan et al, 2001 52	+++
	Ridler et al 2006 48	++
	Bays et al, 2006 55	+
	Sukhwinder et al, 2005 54	+
	Muller et al, 2002 51	−
Posture	Marvel et al, 2004 56	++
	Sullivan et al, 2004 58	++
	Deshmukh et al, 2002 59	++
Oculomotricity
Saccadic movements	Keedy et al, 2006 68	−
	Raemaekers et al, 2002 69	−
	Schultze et al 2006 72	−
	Tu et al 2006 70	−
	McDowell et al 2002 71	−
Smooth pursuit movements	Pivik, 1991 74	+
	Avila et al, 2002 75	+
Nondeclarative learning
Eyeblink	Sears et al, 2000 77	+
	Brown et al, 2005 78	+
	Stevens et al, 2002 79	−
	Hofer et al, 2001 80	−
Vestibulo-Ocular-Reflex	Jones et al, 1983 82	+
	Pivik et al, 1987 83	+
	Warren et al, 1998 84	−
Procedural Learning	Kodama et al, 2001 94	+
	Bigelow et al, 2006 95	+
	Ninomiya et al, 1998 96	−

No cerebellar involvement reported.

Results attributed, on theoretical grounds, to the cerebellum.

Behavioural or structural imaging studies (correlation studies) supporting a cerebellar dysfunction.

Functional studies reporting a cerebellar involvement (fMRI, PET, TMS).

(Cerebellum) Not Included in Analysis

Altogether, despite several reports of cerebellar structural abnormalities in schizophrenia,7 the evidence supporting a cerebellar involvement in hallucinations, formal thought disorder (FTD), and blunted affect seems, to date, limited and not conclusive. The most direct support comes from fMRI studies12,13 reporting decreased activation of a networks including the cerebellar cortex in inner speech control, while structural MRI lead to discrepant results,11,14 possibly because of differences in techniques as well as methods to induce inner speech control. The interpretation of cerebellar hypoactivations or decreased gray matter volume in the pathophysiology of hallucinations has thus to be considered cautiously considering the extension of other cortical areas also concerned. Thus, caution is warranted interpreting the role of the cerebellum.

Formal Thought Disorder

Thought disorder is one of the fundamental features of schizophrenia.16 It encompasses disorders in the content (delusions and hallucinations) and disorders in the formal aspects of thought (formal thought disorder, FTD). FTD includes a variety of symptoms such as poverty of content of speech, incoherence, or neologisms. FTD has been associated with dysfunctions in the left superior temporal gyrus,17 but the cerebellum could also be involved. Three studies by Kircher et al18–20 examined FTD with fMRI (table 1). Though interpretation of the results highlighted fronto-temporal dysfunctions in patients, cerebellar hypoactivations were also found in two of these works. In one of them,18 FTD was yielded through a 3-minute speech about Rorschach inkblots. The severity of incoherence and neologisms was positively correlated with the activity in the cerebellar vermis, right caudate nucleus and the precentral gyrus and it was negatively correlated with changes in the Wernicke area. Another fMRI controlled study by the same group20 reported the cerebellum of patients with FTD significantly less activated compared with healthy controls. Non-FTD patients also presented decreased cerebellar activation but difference was less marked. In this latter study contrasts were obtained between 3 conditions: sentence completion, word-choice and a reading condition. Finaly, Levitt et al21 found greater vermis white matter volume associated positively with the severity of thought disorder, positive symptoms and impairments in verbal logical memory.

In conclusion, though 2 fMRI studies17,19 associated FTD with dysfunctions in left temporal regions, another 2 fMRI studies18,20 found correlations between measures of FTD and reduced activity in the cerebellum, among other regions. A structural MRI work21 also reported correlations between cerebellar white matter volume and the severity of thought disorder, positive symptoms, and verbal memory impairments. These somewhat contradictory results could be explained by the heterogeneity of the patients groups concerning age, medication, and duration of illness, as well as different measures of FTD and different paradigms of speech production used. More homogeneous groups and replication studies are needed before any definite conclusion.

Affectivity

Some studies have reported the presence of psychotic and affective features in patients with lesions of the cerebellum. A “cognitive affective cerebellar syndrome,” initially described in patients after surgery for posterior fossa tumors,22 has also been reported in patients with cerebellar degenerative diseases.23 Many structural and functional neuroimaging studies in affective disorders have reported cerebellar abnormalities.8,9 The induction of sadness has been correlated with increased cerebellar regional blood flow (rCBF) in the vermis,8 a region particularly sensitive to stress due to its high concentration of glucocorticoids receptors.8

In schizophrenia, functional studies have used the presentation of emotionally charged images to elicit an affective response. Emotional recognition tasks assess the capacity to perceive one aspect of visual stimuli (ie, faces) as much as emotional prosody can be perceived in spoken language. In schizophrenic patients under neuroleptic treatment, functional imaging of blunted affect has reported 2 types of pattern of activation: either decreased activation of widespread cortical and subcortical areas24 or the implication of areas not normally concerned in controls.25 The cerebellum has been described as both hypoactivated24 and hyperactivated.25 Neuroleptic free patients, though correctly recognizing the emotional stimulus, underactivate the networks normally involved including the amygdala, cerebellum, thalamus, or prefrontal cortex (PFC).26 It has been suggested that emotional blunting is associated with a shift in the relative contribution of brain regions subserving cognitive and emotional processing. The noncompetitive antagonist of the glutamate receptor ketamine (an anesthetic agent) produces emotional blunting in healthy subjects. Ketamine-induced blunted affect has been correlated with a reduced rCBF in cerebellum and cingulate and visual cortex in fMRI, a pattern similar to that reported in schizophrenic patients.27

Functional imaging studies in patients with blunted affect reported either increased24 or decreased25,26 activation of the cerebellum during emotional recognition tasks (table 1). Medication status does not seem to account for those differences. Indeed, the fMRI study in medicated patients reporting decreased cerebellar activation25 is in line with the one in medication-free patients.26 Both these investigations did not distinguish patients presenting, or not, blunted affect. On the contrary, Stip et al25 divided the patient group in those presenting (BA+) or not presenting (BA-) blunted affect, the latter being those who presented, among other regions, increased cerebellar activation.

Cerebellum and Neurological Dysfunction in Schizophrenia

Neurological disturbances are commonly found in schizophrenia, including both “hard signs” and “soft signs”28. However, “hard signs,” such as abnormal developmental reflexes, are rather seldom, and their prevalences have not always been found significantly different between patients and controls.29 In most cases, the neurological dysfunctions are subtle. In line with these arguments, a set of motor coordination, equilibrium, and sensory integration items (eg, graphesthesia or stereognosia) have been grouped under the term Neurological Soft Signs (NSS). In schizophrenia, different NSS scales have been used (eg, Neurological Evaluation Scale [NES],30 NSS scale31). It is usually assumed that NSS in schizophrenic patients reflect minimal brain dysfunction, resulting from abnormal neurodevelopment. Despite the common view that these signs are nonlocalizing, there is growing interest in clarifying the precise structural and psychiatric correlates of NSS taking into account the heterogeneity of schizophrenic symptoms.32 In particular, some neurological soft signs could reflect mild cerebellar dysfunction such as motor dyscoordination or adiadochokinesia.28 Moreover, some studies have focussed on specific neurological dysfunctions including cerebellar signs, posture, and gait.28,33

Neurological Soft Signs

Compared to healthy controls, never-treated, first-episode schizophrenic patients have higher NSS scores,28,34 while the nonaffected siblings show intermediate NSS scores.32,35 Furthermore, several studies have significantly correlated NSS with severity of illness,36 lower social functioning,33,36 and negative symptoms.32,35–37 By comparison with affective psychiatric patients, first-episode schizophrenic patients also present higher NSS scores.38 Interestingly, NSS are prevalent in 2 pathological states also related with cerebellar pathology: chronic alcoholism39,40 and autism.41,42

A few recent studies have investigated the neural substrate of NSS. Bottmer et al7 reported diminished volume of both cerebellar hemispheres in a group of 37 first-episode schizophrenic patients matched to controls. Furthermore, a significant inverse correlation between NSS scores and the volume of the right cerebellar hemisphere was found. A study by our own group43 in first-episode schizophrenia patients with no history of substance abuse showed significant white matter reductions in the left posterior cerebellum and the right insula in patients with high NSS scores compared with patients with low scores. The subscore of motor integration was negatively associated with the gray matter volume of the cerebellum, the right inferofrontal, right occipital, and the left postcentral gyri. Noteworthy, the score of sensory integration was negatively correlated with bilateral gray matter volume of the cerebellum. In the same line, Keshavan et al44 comparing 17 neuroleptic naive schizophrenic patients and 18 controls reported correlations of the cognitive/perceptual factor of the NES scale with smaller volumes of the left heteronodal association cortices. Two other studies45,46 found correlations between NSS and basal ganglia, but those studies excluded the cerebellum from their analysis.

A drawback of the NSS literature, in general, is the lack of a consensual tool. Though the NES is often used, only a selection of items is generally used44 varying from one study to another and other studies used different scales,43 sometimes not standardized,33 limiting the comparability of the results. Two studies support the involvement of the cerebellum in NSS,7,43 but the main limitation is the small size of the samples. Unfortunately, the 2 studies associating NSS in patients to basal ganglia dysfunctions45,46 did not include the cerebellum in their analysis, so to date, it is not possible to weigh the relative importance of these 2 structures.

Cerebellar Symptoms in Schizophrenia

Ho et al33 in a study considering specifically cerebellar signs, assessed equilibrium, intentional tremor, dysdiadochokinesis, and hypotonia in 155 schizophrenic patients and controls. Twenty-one percent of the patients versus 4.5% of controls presented at least 1 cerebellar sign. The Romberg test and tandem gait test were the most common signs. Patients with cerebellar signs (n = 32) had more severe negative symptoms, poorer premorbid social adjustment, and smaller total cerebellar volume than patients with no cerebellar signs. These latter showed equivalent cerebellar volumes to controls. These findings were independent of medication status. Another recent study by Varambally et al,47 compared 32 schizophrenic patients to matched controls by using 4 scales: the NES (for the NSS), the International Cooperative Ataxia Rating Scale, ICARS (for cerebellar signs), and the SANS and simplified acute physiology score (for psychopathology). Discriminant analysis revealed 2 ICARS subscores, kinetic abnormalities and dysarthria, (but none of the NES scores) to be significant (P < .0001) accounting for 78% of discrimination between patients and controls. Furthermore, ICARS total score, posture subscore, and oculomotor subscore had significant positive correlation with negative syndrome score. Altogether, these works support an intrinsic cerebellar dysfunction in schizophrenia.

Hence, cerebellar signs per se may represent an important part of NSS as a whole, as suggested by 2 studies reporting (1) their relative importance in schizophrenic patients33 and (2) their discriminative power between patients and controls47 (this latter, altogether with other 2 cerebellar subscores).

Motor Tasks and Propioception

In addition to its well-known role in motor coordination, the cerebellum takes an important part in the processing of sensory inputs. Cerebellar lesions are indeed frequently associated with ataxic movements and sensory dysfunctions.23

Frontal cortico-cerebellar systems implicated in adult executive functions are anatomically related to systems undergoing maturation during normal childhood motor development. Disruption of this anatomical system may underlie both the early developmental and adult cognitive abnormalities in schizophrenia, as supported by the Finnish prospective study,48 correlating motor function at 1 year of age with structural MRI measures and executive functions at age 33–35 years (table 1).

Connections between the cerebellum and the motor cortex have been explored using transcranial magnetic stimulation.49,50 Schizophrenic patients, whether medicated or not, showed an increased excitability of the motor cortex.49 A deficit in corticocortical inhibition has been incriminated, but a decreased cerebellar inhibition to the motor cortex could possibly contribute to this higher excitability.50 Altered connectivity has also been reported in fMRI studies, both in the basal ganglia51 and cerebellum.52 Stephan et al,52 using simple unilateral self-paced finger-tapping tasks, reported, after 3 weeks of treatment with the antipsychotic olanzapine, a change in the pattern of cerebellar connectivity, thus suggesting that medication is indeed an important confounding variable in studies assessing motor functions. Furthermore, in this study, olanzapine seemed to affect differently the function of the right and the left cerebellar hemispheres. Similar conclusions were reported by another fMRI study in patients and healthy controls performing a finger-tapping task.53

Sensory propioceptive consequences of self-generated movements have been reported to be altered in schizophrenia.54 Propioceptive inputs are normally attenuated by a sensory prediction mechanism differentiating self-generated from externally generated sensory inputs.55 Sukhwinder et al54 observed that patients demonstrated less sensory attenuation than controls. Using a force-matching task that dissociates sensory inputs and self-generated movements, they concluded that patients have a dysfunctional predictive mechanism. Sensory prediction has been proposed as one of the basic cerebellar functions (see below).

Behavioral studies usually analyze the motor responses of subject. Yet, as the motor responses are usually interpreted in the context of other sensory or cognitive functions, studies specifically focusing on motor performance are surprisingly sparse in schizophrenia. Based on works with 3 different techniques (TMS,49,50 fMRI,51–53 sensory decoupling 54,55), arguments exist to hypothesize dysfunctions in the motor system of schizophrenia patients before frontal executive control regions intervene. But, to date, evidence is limited and diverse.

Posture

Posture and gait specifically involve the cerebellum. First reports of altered postural reflexes in schizophrenia patients date from the 1970s.56 Measures of body postural sway are a way to assess postural control. In schizophrenia, 1 study (Marvel et al56) explored the postural sway of patients and controls with a pressure-sensitive platform. Patients demonstrated more postural sway than controls, independently of medication. Differences remained significant, even after removing from the analysis the patients with tardive dyskinesia. Bloem et al57 reported altered long-latency postural reflexes in a heterogeneous group of medicated psychotic patients compared with different types of parkinsonism and controls. Stimuli consisted of predictable or unpredictable sequences of movements of a platform where the subjects were standing. Psychiatric patients showed adapted reflex amplitudes to predictable variations of the stimuli. Under unpredictable displacements, psychiatric and early-onset parkinsonian patients failed to present the stereotypic anticipated response, which was present in controls. Impairments in postural control in patients have also been found by Varambally et al47 (see “Cerebellar Symptoms in Schizophrenia”) who reported that the posture subscore of the ICARS scale for cerebellar evaluation was significantly higher compared with controls.

Alcohol abuse is highly prevalent among schizophrenia patients (50%–70%).58 Alcohol can alter cerebellar structure and function. However, its physiopathology on the cerebellum and any possible interaction are unknown. Sullivan et al58 measured posture and gait in 4 groups: schizophrenic patients, alcoholic patients, comorbid schizophrenic-alcoholic patients, and controls. All 3 clinical groups were impaired compared with controls, but the comorbid group was significantly more impaired than both the alcoholic and schizophrenic groups, especially when tested with open eyes. This suggested 2 distinct and possibly additive physiopathological processes involving the cerebellum in patients comorbid for schizophrenia and alcohol abuse. This argument was also suggested by Deshmuck et al59 observing a higher prevalence of dysdiadochokinesia in the schizophrenic group after a comparison of schizophrenics, alcoholics, and controls. Reinforcing the argument of an intrinsic cerebellar dysfunction in schizophrenia, the subgroup of schizophrenic patients showed a lesser prevalence of alcohol abuse, thus suggesting an intrinsic cerebellar dysfunction in schizophrenia.

In conclusion, abnormal postural control in schizophrenia is expected to reflect cerebellar dysfunction. However, only one controlled study assessed posture in patients, reporting abnormal swaying. More extended studies are needed to confirm this result, excluding comorbid alcohol abuse.

To summarize, a consistent and growing set of results suggest that the cerebellum is implicated in a significant part of NSS,7,43 some of the items of NSS scales being well-known procedure classically used to assess the cerebellum.33,47 Postural impairments are among these signs though evidence is still limited to confirm its presence and physiopathological origin.56,58,59

Oculomotricity, Cerebellum, and Schizophrenia

Saccadic Ocular Movements

The oculomotor vermis (lobuli VI–VII) is part of the control network of ocular movements.60 The activation of this region is correlated to the amplitude of saccades,60 its adaptation to constant shifts in amplitude,61 and the velocity of smooth eye-tracking movements.62 Schizophrenic patients present abnormalities in a variety of saccadic movement paradigms. While reactive or simple saccades (to look at a target in a reflexive way) are virtually not impaired,63 performances in other paradigms have been reported abnormal: predictive saccades64 (a rapid repetitive sequence of targets elicits anticipated saccadic movements), remembered or memory saccades65 (a saccade produced after a delay toward one or more targets briefly shown), and antisaccades66 (inhibition of a reactive saccade, with a saccadic movement to the opposite direction). So far, several dysfunctions have been proposed in networks concerned with oculomotor control in schizophrenic patients.64–67 However, fMRI studies on saccadic movements (reactive saccades and antisaccades) and smooth eye tracking comparing patients and controls have reported decreased activations of the frontal eye field, parietal eye field, supplementary eye field, the cingulate cortex, and the dorsolateral PFC (DLPFC)68–71 without revealing abnormal activations of the cerebellum. Furthermore, no structural MRI abnormality has been correlated with performance in antisaccades or smooth pursuit in a large population of patients, their relatives, and healthy controls.72

The adaptation of the amplitude of visually guided saccades to unaware constant shifts in the target's position is a function usually ascribed to the cerebellum.73 This can be supported by studies in monkeys60 and humans.61 Despite this relative specificity, no studies have been published using adaptative saccade tasks in schizophrenia.

Smooth Pursuit Eye Movements

Abnormalities in smooth pursuit and in fixation paradigms have been consistently reported in schizophrenic patients.67 Pivik74 demonstrated that eye-tracking disruptions in schizophrenic patients were “normalized” in dark-adapting conditions. This effect was attributed to the diminished cerebellar influence, inactivated in the dark because optimal visual fixation is precluded.74 Avila et al75 reported that ketamine-induced eye-tracking abnormalities were similar to those observed in relatives of schizophrenic patients, involving a network linking the frontal eye field to the cerebellum (affected by ketamine). However, fMRI (Keedy et al68 cited above) did not report cerebellar activations during smooth eye tracking.

Despite the fact that the cerebellum is included in networks of brain areas controlling oculomotricity60–62 and that numerous ocular abnormalities have repeatedly been reported in schizophrenia,63–66 to date, no functional evidence directly supports that those anomalies are related to cerebellar dysfunctions in schizophrenic patients. Saccadic adaptation maybe the most specific ocular movement paradigm involving the cerebellum61 but has not been yet studied in schizophrenic patients (table 1).68,69

Nondeclarative Learning

Learning and memory systems can be classified as declarative (explicit, verbally mediated) and nondeclarative (implicit). Declarative learning is evaluated with verbal tasks (see “Language” and “Memory”). Nondeclarative, implicit learning comprehends different paradigms: (1) conditioning paradigms usually involving the eyeblink reflex or the vestibular-ocular reflex (VOR) and (2) procedural learning paradigms.

Eyeblink-Conditioned Stimulus

Classical eyeblink conditioning is known as a cerebellar associative motor learning task.76 The role of cerebellum is to adjust in time the blinking response triggered by the conditioning stimuli (usually a puff of air on the eye). So far, results in schizophrenia are conflicting: facilitation (ie, faster) in some studies,77 impairment (ie, slower),78 or no patients-controls differences79 have been reported. Hofer et al,80 combining an eyeblink conditioning paradigm with reinforcement learning, reported that schizophrenic patients poorly detected reinforced stimuli; in other terms, they failed to increase eyeblink responses on reinforced trials. The authors attributed these results to temporal (septal-hippocampal) dysfunctions. Thus, even if a cerebellar dysfunction is suggested,77,78 cortical dysfunctions cannot be excluded as an alternative hypothesis.80

Vestibular-Ocular-Reflex

VOR is the ocular response that corrects gaze position in accordance with head movements. It includes a long smooth eye movement phase followed by a fast saccadic corrective phase. VOR can be elicited experimentally through caloric stimulation of the tympanic membrane. Because of its relative simplicity, and its well-documented structure and physiology, VOR has been proposed as a model of cerebellar motor learning.81 The cerebellar flocculus was found to be responsible for VOR adaptation.81 In schizophrenia, Jones and Pivik82 reported dysrhythmic saccades and a slower saccadic component of the VOR. Schizophrenic patients failed to suppress VOR with visual fixation, particularly in patients presenting an active symptomatology. Pivik et al83 confirmed those former results and correlated this impairment in VOR suppression with dysfunctions in smooth eye tracking. Interestingly, patients' tracking performance was decreased solely when the cerebellar contribution is supposed to be stronger, ie, in light-adapting conditions. Abnormal VOR suppression by gaze fixation has also been recently reported with a moving circling chair.84 Though a cortical inhibitory deficit has been proposed at the origin of this impaired VOR inhibition,84 other studies82–85 have proposed complex and possibly abnormal interactions between the VOR circuitry and the smooth eye-tracking circuitry to account for abnormal VOR adaptation in schizophrenia. Supporting this view, a study in neurological patients has correlated intentional tremor (considered as being of cerebellar origin), impairments in VOR inhibition and smooth eye movement.85

Although only 3 studies evaluated VOR suppression in schizophrenia,82–84 contrasting with the literature concerning eyeblink conditioning, consistent defective inhibition of the VOR was found in patients that was attributed to well-documented cortically based inhibition abnormalities81 present in schizophrenia. A refutation of this hypothesis in favor of a cerebellar dysfunction remains to be done.

Procedural Learning

Procedural learning is a rule-based learning in which performance facilitation occurs with practice without conscious awareness. Motor skill learning (eg, with the Serial Reaction-Time Task) has been associated with activation in motor cortical areas, cerebellum, basal ganglia, and thalamus.86 Cognitive habit learning, assessed with executive functions tests involving the frontal lobes (eg, Tower of London test), has been associated with the DLPFC and association cortices.87 Cortical and subcortical activations seem to evolve inversely as performance improves. Thus, an fMRI study reported that while cortical activations decreased, activity in the cerebellar dentate nucleus, thalamus, and putamen progressively increased.88 Furthermore, the activation of the cerebellar cortex seems to follow a lateral-to-medial course during procedural learning tasks. A positron emission tomography (PET) study by Matsumura et al89 in healthy subjects reported that the lateral cerebellum was predominantly activated during the early phases of learning while the parasagittal cerebellum diminished its activity with training, correlating inversely with task performance.

Though not always, schizophrenic patients usually present deficits in cognitive and motor procedural learning (eg, rotary pursuit task), independently from medication or abnormal motricity.90 Early reports91 suggested that problem-solving difficulties (rather than a true learning defect) were at the origin of procedural learning defects in schizophrenia. However, the neural substrate of both aspects and their mutual relation are still unclear.92–94 Procedural learning impairments have been related to the severity of the disorganization syndrome during the acute phase of the illness, “normalized” procedural learning performances appearing after stabilization.92 In addition to the cerebellum, the premotor area, the presupplementary motor area,93 or the basal ganglia have been related to motor procedural learning impairments in schizophrenia.94 The relative weight of problem-solving (attributed to the frontal lobes) and of other learning deficits (possibly related to subcortical structures) is not clear. The phase of the illness could be an important confounding factor, as suggested by a study reporting “normalized” procedural learning performances after stabilization.92

A simple pointing task performed while wearing special distorting goggles was used by Bigelow et al95 to test procedural learning in schizophrenic patients and controls. Patients were significantly more impaired than controls in the pointing task suggesting an impairment to adapt their motor performance to the visual distortion. After goggles removal, readaptation was similar in both groups, but patients had significantly greater difficulties in reorientation. The authors attributed this impairment to dysfunctions of a network including the frontostriatal cortices, basal ganglia, and the cerebellum. A similar study by Ninomiya et al96 reported that schizophrenic patients were significantly worse than control in the pointing motor responses. In spite of the convergent findings of those 2 latter studies, it is worth noting that different dependent variables were used: accuracy of responses in the one,95 velocity of the responses in the other.96 The former attributed the results, on a theoretical basis, to dysfunctions in a fronto-striatal-cerebellar network, which is rather unspecific concerning the motor system.

In summary, though the cerebellar involvement in procedural learning paradigms is established in healthy humans,76,81,86,88,89 there is only indirect evidence suggesting a specific cerebellar dysfunction in procedural learning impairments in schizophrenia (table 1).

Cognition and the Cerebellum in Schizophrenia

Global Cognitive Function and the Cerebellum in Schizophrenia

Five studies attempted to correlate cerebellar volumes to measures of global cognitive functioning in schizophrenia: Toulopoulou et al97 and 4 other studies reviewed by Antonova et al98 (table 2). Toulopoulou et al97 reported that though several cerebral measures correlated with IQ, the cerebellar volume correlated with delayed verbal memory but not IQ. Szeszko et al99 computing the mean of 6 cognitive domains (executive, motor, language, visuospatial, memory, and attention functions) found that the cerebellar volume correlated significantly with better global functioning in healthy subjects but not among patients. Antonova et al98 reviewed 4 studies correlating cerebellar structure and cognition in schizophrenia and reported no significant changes in total cerebellar volume in patients. Additionally, cerebellar volume was found positively related to global cognitive functioning in healthy people and affected females but not in affected men. Nevertheless, 2 studies cited by Antonova et al98 report results that have not been reproduced: (a) the correlation of diminished anterior vermis volume in patients, with diminished total and verbal IQ and (b) the increased white matter proportion in vermis of patients linked to decreased logic memory.

Table 2

Involvement of the cerebellum in schizophrenia – Studies correlating cerebellar volume and global measures of cognitive function.

Study, by author (All MRI studies)	Groups			Correlations reported with cerebellum
	SZ	HC
Flaum et al 1994 98	50 M 22 F		32 M 27 F	Left CB Vol > Righ CB Vol direct ↔ global IQin HC and affected women.Lack of this relation in affected men.
Nopoulos et al 1999 98	65 M		65 M	Total CB Vol preserved.↓ anterior vermis in SZ direct ↔↓ global IQ and ↓verbal IQ
Levitt et al 1999 21	15 M		15 M	Total CB Vol preserved. Greater vermis in SZGreater left-than-right CB asymmetry in grey matter in SZ.↑ vermal white matter in SZ direct ↔↓ logical memory; ↑positive symp and ↑ FTD
Szesko et al 2003 99	48 M 33 F		14 M9 F	Total CB Vol direct ↔ global cog. score(constructed with 6 sub-scores: attention, memory, executive, visuospatial , motor, language)This relation, found in HC was absent in SZTotal CB Vol ↔ memory, executive, visuospatial scores in HC, not in SZ
Toulopoulou et al 2004 97	56 SZ	90 relat.	55 HC	Total CB Vol direct ↔ delayed memoryGlobal IQ and visual IQ direct ↔ with toal brain vol. and right hippocampus.Right hippocampus direct ↔ performance IQLeft hippocampus direct ↔ verbal IQ

Note: CB : cerebellum; ↔ : correlation; SZ : schizophrenic patients; HC: healthy controls.

Hence, evidence exist to suggest a role of the cerebellum in global aspects of cognition, such as measured by the IQ, in healthy people and affected females yet not in affected men. So far, patient groups studied have been heterogeneous regarding duration of illness and medication status. Future studies should take into account these issues as well as gender differences in order to obtain more reproducible results.

Attention

fMRI studies in healthy subjects have implicated the cerebellum in attentional tasks.100 Behavioral studies in cerebellar patients101 have raised the question of a role of the cerebellum in attention. It has been proposed that the cerebellum could participate in preparing and reorienting attention.102 Studies in autistic patients support this view.103

Attention impairments are well documented in schizophrenia.104–108 Abnormalities in processing salient novel stimuli in oddball tasks have been consistently replicated.104 Schizophrenic patients display poorer discrimination and slower reaction times than controls. Using fMRI in patients and controls while identifying a letter among similar-looking letters, performance of the 2 groups was comparable, but patients showed a diminished activation in the inferior frontal cortex and an abnormally enhanced activation in right postcentral gyrus, right medial temporal lobe, and left cerebellum, possibly reflecting an increased effort.105 During auditory oddball tasks in patients, decreased rCBF was seen in a large network including cerebellum and thalamus.106 Honey et al107 assessed in schizophrenic patients the Continuous Performance Test in degraded and nondegraded conditions, with fMRI focusing on basal ganglia, amygdala, frontal, temporal, and parietal regions. The degraded condition elicited decrements in sensitivity thought to reflect increased demands on the limited capacity of visual attention. In this case, an attenuated activation of the anterior cingulate and cerebellum was found only in patients. Furthermore, patients presented functional disruptions in 2 networks: cerebellum-medial superior frontal gyrus and cerebellum -anterior cingulate.

Drug-free patients also present differences in neural activation. A PET study compared drug-naive patients and controls in 3 conditions: counting regular auditory clicks, counting in silence, and rest.108 Cerebellar activation was present in both groups in all conditions. Yet, differences were observed only in the frontal and inferior parietal gyrus when mentally counting without auditory stimulation. This led the authors to conclude that the role of cerebellum was predominantly sensory. Interestingly, 1 fMRI study investigating the effects of rivastigmine (a central nervous system-selective cholinesterase inhibitor) on attention, given in cotherapy to schizophrenic patients, reported that the improvement of behavioral measures with rivastigmine correlated with increased rCBF only in the cerebellum.109

In conclusion, fMRI studies in patients have revealed the involvement of the cerebellum in tasks such as visual oddball detection105 (enhanced cerebellar activation) or auditory oddball detection106 (diminished cerebellar activation), albeit the cerebellum was included in large and distributed cortico-cerebellar networks (table 3). Increased visual attentional demands yielded attenuated activation and functional disruption of a cingulate-cerebellum network only in the patient group.107 The involvement of cerebellum in attentional tasks supports the “attentional-shift” hypothesis of cerebellar function (see below). The decreased activation of this structure in schizophrenia patients106,107 could thus be interpreted as a result of increased demands over a dysfunctional attentional network. Yet, this does not explain the work reporting hyperactivation of the cerebellum in patients during a visual oddball detection tasks.105 A differential involvement of the cerebellum depending on the sensorial modality (as suggested by Pivik74 in the context of oculomotor studies) could explain these contradictory results, although it does not account for the cerebellar activations in PET scan found during an auditory driven counting condition, which disappeared during a mentally counting condition, both in patients and controls.108 Differences between groups were reported in frontal and parietal regions when mentally counting without auditory stimulation. The authors interpreted the role of the cerebellum as predominantly sensory (rather than related to attention). This interpretation is in-line with studies in healthy humans proposing the cerebellum as a sensory prediction organ (see below). Yet the absence of differences between patients and controls in the cerebellum challenged the existence of a cerebellar dysfunction in attentional tasks, at least in the auditory modality. Paucity of results and diversity of methods impede any definitive conclusion.

Table 3

Involvement of the cerebellum in schizophrenia – Cognitive aspects.

Cognition
Attention
	Eyler et al, 2004 105	+++
	Kiehl et al, 2005 106	+++
	Honey et al, 2005 107	+++
	Aasen et al, 2005 109	+++
	Ojeda et al, 2002 108	-
Language
	Shergill et al, 2003 118	+++
	Boksman et al 2005 130	+++
	Kircher et al 2005 131	+++
	Marvel et al, 2004 132	+
	McGuire et al, 1996 129	-
	Hofer et al, 2003 119	NIA
	Ragland et al, 2004 120	NIA
	Weiss et al, 2006 121	NIA
	Sommer et al, 2001122	NIA
	Bonner-Jackson et al 2005 123	NIA
	Jessen et al 2003 124	NIA
	Koeda et al 2006 125	NIA
	Mitchell et al 2004 126	NIA
	Kubici et al 2003 127	NIA
	Sommer et al 2003 128	NIA
Memory (all types)
	Mendrek et al, 2005 139	+++
	Mendrek et al, 2004 142	+++
	Meyer-Lindenberg et al 2001 144	+++
	Crespo-Facorro et al, 2001 148	+++
	Kindermann et al 2004 140	+++
	Schlosser et al 2003 143	+++
	Whyte et al 2006 147	+++
	Toulopoulou et al, 2004 97	++
	Antonova et al, 2004 98	++
	Szeszko et al, 2003 99	++
Timing
	Volz et al, 2001 155	-
	Ortuno et al, 2005 156	-
Planninga

Note: The key is the same as for table 1.

No studies in SZ patients reporting or suggesting a cerebellar involvement.

Language

There are some evidences suggestive of the role the cerebellum in speech perception,110 word production,111 and syntactic112 and semantic aspects of oral and written language.113,114 Nevertheless, until now, no consensus was found. The right posterolateral cerebellum seems particularly involved in speech perception,110 lexical semantic retrieval deficits,115 or agrammatism.116 Cerebellar language functions could be lateralized as much as it is for cerebral cortex areas.117

Studies on language in schizophrenia focussed preferentially on: (a) correlations of language measures with clinical or cognitive aspects (ie, with hallucinations,118 episodic memory deficits119); (b) studies on episodic memory (word encoding and recognition,120 see below); (c) studies on language lateralization121,122; and (d) studies on affective prosody.126 Unfortunately, most functional neuroimaging studies investigating the neural correlates of language in schizophrenia did not take into account the cerebellum in their analysis119–128 and, in particular, those evaluating language lateralization121,122,128 and affective prosody.126 This fact sensibly decreases the weight of results that do take the cerebellum into account and which report a cerebellar involvement.

A predisposition to verbal hallucinations was associated with a failure of inner speech monitoring. In a task of inner speech production, schizophrenic patients showed an attenuated response in the right temporal, parietal, parahippocampal, and cerebellar cortex compared with controls (Shergill et al118). Inner monitoring abnormalities were also reported by McGuire et al129 though cerebellar activations were not seen.

Regarding word production deficits, a 4-T fMRI investigation on word fluency by Boksman et al130 reported decreased cerebellar activation during speech production in patients compared with healthy controls. These differences remained significant when controlling for medication effects. An original work by Kircher et al131 compared the structures associated to word production depending on the syntax of sentences. Patients spoke freely about 7 Rorschach inkblots during an fMRI scan. Patients showed activation of the cerebellum during the production of syntactically simple sentences (though not in the “complex sentences” condition). The contrast between patients and healthy controls reinforced this finding.

It has been proposed that whilst the cerebellum could be involved in word search, the PFC could be preferentially involved in word selection. Marvel et al132 reported that schizophrenic patients showed impairments in both search and selection conditions. The authors concluded that a search deficit underlies word production problems in schizophrenia and that this may involve fronto-cerebellar circuits.

In summary, inner speech monitoring deficits have been associated in patients with attenuated activation of the cerebellar cortex among other regions,118 but this finding was not reproduced by another study.129 In addition, decreased cerebellar activation during speech production was found in patients depending on the syntactic complexity of the produced sentences: only the simple sentences condition activated the cerebellum.131 A word search defect in schizophrenia, related to fronto-cerebellar networks, has been proposed and need confirmation.132 However, word tasks used in the studies reviewed do not permit to conclusively exclude a pure motor cerebellar function, even in the study that assessed more abstract levels of speech production (simple vs complex syntactic structures). The same concern is raised from inner speech control studies,118,129 where brain regions implicated are supposed not to produce speech but rather to control its production. Altogether, any conclusion about the role of the cerebellum in language-related functions would be premature (table 3).

Memory

Short-term,133 long-term,133 and episodic memory134 deficits have been reported in schizophrenia with no apparent link to the severity of psychopathology or the duration of illness.133 In addition to its role in nondeclarative implicit memory systems (see above), some studies also suggested the implication of the cerebellum in declarative systems, ie, verbal working memory,135 fact retrieval,136 and autobiographical memory137 (for a review, see Weiss and Heckers138)

Verbal and spatial working memories are impaired in schizophrenia and in a lesser extent in their first-degree relatives.139 Regarding spatial working memory, Kindermann et al140 reported patterns of activation significantly different between patients and healthy controls. Increased activation of the anterior cerebellum was present only in the patients group. Verbal working memory has been more frequently evaluated in schizophrenia than spatial working memory. In-line with fMRI studies in healthy subjects,141 verbal working memory tasks in schizophrenic patients activate a fronto-thalamo-cerebellar network of regions.139,142 Using an n-back working memory task during fMRI in schizophrenic patients, before and after partial remission, Mendrek et al142 showed that partial remission was associated with a normalized activity in a right network including the right DLPFC-right thalamus-left cerebellum and cingulate gyrus, although the contralateral network remained disturbed. In a similar controlled study with stable patients, these authors reported relative underactivation in the left DLPFC and the right cerebellum, contrasting with an overactivation in the left cerebellum,139 the latter possibly reflecting a compensatory mechanism or an increased effort. An fMRI study by Schlosser et al143 also reported altered fronto-cerebellar connectivity in schizophrenia. These authors found that while the fronto-cerebellar and the cerebello-thalamic loops were hypoactivated, the thalamo-frontal loop revealed hyperactivated. This was interpreted as compensatory increments in the presence of decreased cerebellar inputs. Numerous PET studies have linked verbal working memory with corticocerebellar activations in medication-free schizophrenic patients: Meyer-Lindenberg et al144 using an n-back verbal working memory task explained more than half the variance of activations in patients, by changes in the inferotemporal, parahippocampal, and cerebellar regions.

Some researches used paradigms of word encoding (reading and memorizing) and word recognition (as “seen” or “not seen”) to study the neural substrate of episodic verbal memory.145–148 Two studies by Ragland et al145,146 used this kind of tasks with PET145 and 3-T fMRI.148 In both, cerebellar activations were reported in the patients group during the encoding condition. However, when comparing with the encoding patterns in healthy controls, cerebellar activations were not significant.145,146 Noteworthy, medication effects may play a major role. When only medication-naive patients were considered, the cerebellum and the primary motor cortex showed greater activation.145 Greater cerebellar activation was also reported in high-risk groups (ie, patients' relatives) during the recognition phase.147 These results are in-line with a study by Crespo-Facorro et al148 also reporting abnormal cerebro-cerebellar activations in patients during a memory task (recall of lists of words, either recently learned, or well known), compared with controls. This hyperactivation possibly reflects the effect of increased effort over a dysfunctional circuit.

In summary, an important body of literature supports the implication of the cerebellum in mnemonic processes in schizophrenia (table 3). One spatial working memory140 and 3 verbal working memory (n-back) fMRI studies139,142,143 report cerebellar underactivations in the context of cortico-cerebellar circuits. Though medicated effects cannot be excluded, all 4 studies had control groups and results were consistent between studies. In episodic verbal memory, 2 studies report greater cerebellar activation in patients during word encoding,145,146 while other fMRI studies reported greater cerebellar involvement in the recognition (recalling) phase in patients148 and in high-risk relatives.147 Different word recognition tasks could partly explain this apparent contradiction.

Timing

Timing has been proposed as a basic function of the cerebellar cortex-climbing fiber system.149 Transient inhibition of the cerebellar vermis with transcranial magnetic stimulation has yielded more variable finger-tapping responses in healthy subjects, ie, altered timed movements.150 Studies on duration perception usually report the activation of a network including the cerebellum, frontal and parietal cortices, supplementary motor area, basal ganglia, and thalamus.151,152 Schizophrenic patients are less accurate in temporal generalization tasks (recognition of a standard duration)153 and temporal bisection tasks (categorization of durations as short or long),153 both in auditory and visual modalities.154 Timing and working memory could be correlated because recognition of an interval implies storing it (in working memory) for deferred comparison.152 Investigating if the poorer performances of patients in timing tasks were due to a more global impairment of working memory, Elvevag et al153 reported no correlations between working memory scores and timing performance, suggesting 2 independent processes. Volz et al155 explored the neural basis of time estimation in schizophrenia comparing patients and controls in an fMRI study in 3 conditions: an auditory time estimation task, a pitch discrimination task, and a rest condition. Compared with controls, patients presented patterns of hypoactivity in a fronto-thalamo-striatal network in the timing vs rest contrast and the pitch vs rest contrast.155 No differences in the cerebellum were reported in this work. A PET study by Ortuno et al156 contrasted 2 timing task conditions (counting 1-Hz auditory clicks at 1 Hz and mentally counting at the same frequency of the heard clicks) to a passively hearing condition and a rest condition. They also reported alterations in fronto-striatal networks in patients (greater frontal activation: area BA 6) and no cerebellar involvement.

Altogether, evidence suggests that a timing estimation deficit exist in schizophrenia153,154 independent of working memory impairments.153 Whether the cerebellum is involved or not in time estimation impairments in patients is still subject to debate. Two neuroimaging studies155,156 suggest that the cerebellum is not concerned in timing in schizophrenia (table 3). However, these results challenge studies in healthy subjects showing the opposite152 and are based on studies in rather small samples (9 and 11 patients). Neuroimaging studies in greater groups are warranted in this topic.

Planning

Planning performances in schizophrenia have been reported altered both in the cognitive domain157 and in the motor domain,158 thelatter being correlated to the severity of the disorganized syndrome.158 Patients may less anticipate their actions, thus impairing their strategy in complex or unfamiliar situations159 as shown in a study comparing the performance of patients and controls in 3 tasks: a simple line-copying task, a more complex figure-copying task, and a standard psychomotor test, the Digit Symbol Test. Patients appeared about one-third slower in their total performance time in all 3 tasks. Increased figure complexity or decreased familiarity significantly prolonged the initiation time, particularly in patients with higher scores on negative symptoms. A recent behavioral study using the Cambridge Neuropsychological Test Automated Battery160 reported poor spatial working memory as a significant predictor of planning impairments. However, in the cognitive domain, even if patients take significantly more moves to solve a series of problems in the Tower of London Test,157 when latencies of movements are adjusted to consider the slower responses overall, the patients planning times are not significantly increased, thus challenging the existence of a true cognitive planning deficit in schizophrenia.

Although the cerebellum has been involved in movement planning both in cerebellar patients161,162 and controls,163 studies in healthy subjects dissociating planning phases and control of execution have linked the cerebellum preferentially to control of execution.164,165 Functional imaging studies on planning functions in schizophrenia are needed to clarify this point.

Hence, though some studies suggest a true planning deficit in schizophrenia,157,158 when latencies of movements are adjusted to consider the slower responses, the patients' planning times are not significantly increased, thus challenging the existence of a true planning deficit in schizophrenia. So far, functional neuroimaging studies lack in this domain. Hypothesis about a role for the cerebellum in (at least) motor planning still need direct evidence.

Functional Models of the Cerebellum

Functional models permit to interpret and integrate important amounts of data, often issued from diverse methodologies. Some of the concepts reviewed below, particularly the concepts of error detection and learning, have permitted the development of computational models and hence the simulation of the network functions with variable success.166 We will briefly summarize the main models proposed and the specific function attributed to the cerebellum.

Attention-Shift

Courchesne and Allen167 highlighted the role of the cerebellum in orienting attentional resources. According to these authors, the cerebellum rapidly primes task-relevant systems in order to improve neural responsiveness. This anticipatory effect is supposed to affect sensory, motor, cognitive, affective, and autonomic systems.166,167 Doing so, the cerebellum ensure that attention resources be implemented in a fast and coordinated manner. The cerebellar cortex would detect patterns of inputs, which should shift attention toward the systems implicated in the ongoing action. Afterward, the cerebellar nuclei, through cortico-cerebellar loops, would control performance.168

Error Detection and Learning

Reports of modified cerebellar output and increased activation after unexpected sensory perturbations169 or during the initial phases of skill learning89 suggest that detection and correction of errors could be elicited as the basic cerebellar function.170 Formerly proposed for the motor domain,171 the hypothesis has been generalized to the cognitive domain on the basis of verbal tasks in cerebellar patients.171

Prediction and Timing

Anticipatory adjustments have been hypothesized as a cerebellar function.172 Through experience (ie, learning), a particular predicted sensory context or cognitive state could become associated with a particular motor or cognitive response. Prediction is a recurrent concept in cerebellar theories.167,170,172 A variant of the prediction hypothesis is that the cerebellum operates as an internal timing system,173 providing the precise temporal representation across a range of tasks. The timing hypothesis is coherent with proposals of the cerebellum as a structure devoted to sensory prediction.174 However, recent studies relate timing abilities to distributed networks.175

Discussion

The role of the cerebellum in schizophrenia has been highlighted by the hypothesis of the “cognitive dysmetria,” which assumes that dysfunctions in the cortico-cerebellar-thalamo-cortical circuits could result in schizophrenic symptoms, via impaired coordination of mental processes.2 After reviewing recent clinical, cognitive, and functional imaging studies involving the cerebellum in schizophrenia, a heterogeneous picture emerges. While in some domains the cerebellum seems to be incriminated (ie, NSS, posture, equilibrium), in other domains its contribution seems limited or indirect (ie, cognition) if present at all (ie, affectivity). As a whole, the review is congruent with the classical view of the cerebellum as specifically implicated in sensorimotor control. However, some results point to the involvement of the cerebellum in cognitive functions as well, in particular fMRI studies on disordered thought18,19 or verbal working memory139,142,143 (independent from motor performance) or structural imaging studies linking cerebellar volume with cognitive parameters.97–99

A significant gap persists however between what we know about the cerebellum in motor function and the evidence for the role of a putative cerebellar dysfunction in schizophrenia. Functional models, mostly tested with motor tasks, have been difficult to generalize to cognitive tasks, even more to schizophrenia, where distributed cerebral abnormalities preclude any simple, localizing interpretation. The computational perspective could theoretically bridge the gap between sensorimotor and cognitive functions.176 This perspective links the cerebellar structural homogeneity and functional unity, it does not restrict a priori the domains of application and it enables modularity, in the sense that the distinct channels could all implement the same unique cerebellar computation166,172 (hypothesis usually assumed).

Several complementary arguments may explain the heterogeneity of the reviewed results:

Methodological issues. The diversity of domains explains the variety of techniques employed. In addition, in the context of each considered area of research, methodological diversity continues to be the rule, thus making comparisons difficult and generalizations risky. Differences between studies concern the following 2 factors. (1) The characteristics of the patients groups. Most studies, maybe due to their exploratory character, assess medicated patients with varied age and duration of illness. Notwithstanding, it has been reported that the age,177 duration of illness,178 and medication51,52 are important (usually confounding) variables to consider when interpreting data in schizophrenia. (2) Diversity of behavioral paradigms employed to test the same function. While this diversity could be useful because it permits complementarity of results, the limited number of works in most domains and the absence of replication studies render it a drawback rather than an advantage.

Not all schizophrenic patients present a cerebellar dysfunction. How could we clinically characterize those patients in whom a cerebellar involvement would be highly probable? Could this be related to a specific etiology? This raises the question of defining more homogeneous endophenotypes in schizophrenia research.2 Schizophrenic patients with cerebellar anatomical or functional abnormalities have been related to negative symptoms,32 more impaired cognitive profile,32,98 poorer outcome,179 and higher NSS scores.7

Cerebellar dysfunction in schizophrenia could be restricted to specific parts of the cerebellum. Thus, impairments in some tasks could coexist with a normal cerebellar function assessed through other paradigms. Considering the fact that atrophy of the vermis is the most cited structural cerebellar abnormality in schizophrenia,7 it could be proposed that only vermis-based cerebellar functions (eg, ocular motricity and posture58–62) should be dysfunctional, whereas the hemisphere-based cerebellar functions (those specially related to the frontal lobes and cognition100,101,110,111,113) should be preserved.

The possibility to estimate the influence of the cerebellum depends on the complexity of the considered neural network. The more a neural network is large and complex, the more it can be difficult to disentangle the precise function of one of its components. Cerebellar functions may be more accessible experimentally in the context of relatively small networks such as those responsible of eyeblink conditioning, VOR, or postural reflexive control, than in the context of larger networks such as those involved in cognition. In these latter, putative cerebellar dysfunctions should be interpreted in the context of networks that have not completely been described (eg, executive functions, language). Moreover, cortical dysfunctions present in schizophrenia and the existence of compensation effects warrant caution in interpreting the data. Investigating hypothesized cerebellar dysfunctions in such a context would be not far from speculation. Functional imaging, though not completely solving these difficulties, could be the best method, to date, to associate behavioral and neural networks level.

The cerebellum could indirectly influence cognition through abnormal sensory integration. If the role of the cerebellum in the cognitive aspects of schizophrenia seems limited, how can we interpret reports linking cerebellar volumes or activations with cognitive tasks?33,97–99 Sensory inputs, presumed to be abnormally integrated by the cerebellum, could result secondarily in abnormal cognitive processes. Sensory integration impairments, associated with dysfunctions in the cerebellum28 and heteromodal cortices,43,44 could in turn facilitate the emergence of abnormal cortically based cognitive processes as it seems to be the case for hallucinations118 and arguably for FTD18,19 and verbal working memory.139,142,143

In conclusion, different lines of research converge to suggest that a cerebellar dysfunction could exist, at least in some patients with schizophrenia, and/or could account for some of the psychiatric, neurological, or cognitive symptoms present in this disease although negative reports are also found. Several explanations could explain divergent findings found in the literature including the clinical heterogeneity, the heterogeneity of the cerebellar structure and functions, and the complexity of the involved networks. The relative prominence of the cerebral cortex in cognition as well as a postulated indirect influence of the cerebellum via impaired sensory or sensorimotor integration could hamper the delimitation of a specific role of the cerebellum in schizophrenia. Computational models appear promising to synthesize the role of the cerebellum in motricity and its elusive role in cognition. Further studies in well-characterized and ideally more homogeneous groups of patients are warranted to fully understand the place of the cerebellum in the symptoms and deficits associated with schizophrenia.