The role of the cerebellum in timing

the cerebellum’s involvement in timing


Sensory prediction, timing and the cerebellum
Two prominent proposed functions of the cerebellum are that it generates a forward model for checking sensory predictions against reality [1] and that it allows for action and perception requiring delicate timing [2].The benefit of generating a forward model is that it allows for quick adaptation if predictions are not met.For example, if you mistakenly take an extra step beyond the top of a flight of stairs, your cerebellum's (failed) sensory prediction combined with a suitable alarm mechanism could prevent you from stumbling.We here propose that the timing capabilities of the cerebellum are integral to building accurate sensory predictions and to acting upon them.
The first studies to reveal the cerebellum's involvement in timing were lesion studies [3] showing that patients with cerebellar lesions performed worse than controls on several tasks: for example, tapping rhythmically [4], perceiving the duration of time segments [4] and judging the velocity of movement [5].The first magnetoencephalography (MEG) study consistent with cerebellar involvement in timing rhythmic somatosensory stimuli [6] showed that hypothesised cerebellar source activations were compatible with field patterns evoked by sequences of rhythmically presented somatosensory stimuli.Very interestingly, oscillatory cerebellar activity that appeared to anticipate subsequent somatosensory stimuli was reported as well.More recently, a study [7] using a similar paradigm with rhythmically presented somatosensory stimuli, but using modern beamforming source reconstruction methods, localised beta band (14-30 Hz) activity related to differences in stimuli expectations to the cerebellum.This study also indicated that the cerebellum responds to unexpected omissions of stimuli.In a subsequent study [8], we found that the strength of cerebellar beta band responses to omissions was dependent on the temporal regularity between somatosensory stimulations preceding the omission.Our latest work [9] elaborates that this cerebellar beta band activity supports timing processes that are informative to action -in that we found a correlation between cerebellar activation and performance in a detection task.Cerebellar beta band responses have also been found in relation to the timing of predicted auditory responses [10], indicating that cerebellar beta band responses might have a general, modality-independent role in building temporal expectations.Supporting this modality-independent role of the cerebellum, in a study [11] using transcranial magnetic stimulation (TMS), the authors found that inhibitory cerebellar stimulation caused the sensory attenuation of self-generated auditory stimuli to diminish.Behavioural and functional magnetic resonance imaging (fMRI)

]] ]] ]]]]]]
studies have found that a crucial factor in whether sensory processing of self-generated actions is attenuated or not is the time interval between the action and the subsequent sensory feedback [12][13][14].Common to all the neuroimaging studies reported here is that they report effects in cerebellar lobule VI and cerebellar crus I, to which MEG has good sensitivity [15] (Figure 1).Interestingly, cerebellar lobule VI is also found to show differential functional connectivity to primary somatosensory cortex depending on whether a touch is self-generated or externally generated [15].Finally, a meta-analysis [16] of nine timing studies using fMRI [17][18][19][20][21][22][23][24][25] also finds cerebellar lobule VI to be a common factor in timing.

What about the basal ganglia?
An unresolved debate however when it comes to cerebellar timing systems is what role the basal ganglia play in timing, if any.The early lesion evidence did not result in an unequivocal conclusion -one of the control groups in the aforementioned cerebellar lesion studies [3] comprised patients with Parkinson's disease, in whom no evidence was found for timing deficits.However, around the same time, evidence for timing deficits in duration perception and finger tapping was reported for patients with Parkinson's disease [26][27][28].Due to the dysfunction of the basal ganglia in Parkinson's disease, it has been proposed that the basal ganglia also play a role in timing.A theory explaining this apparent redundancy states that the brain has several timing systems and that the cerebellum is involved in subsecond timing and the basal ganglia in suprasecond timing [29].This view seems to be at odds with some of our recent findings reported with MEG [8,9] in which we find evidence that the cerebellum engages in tracking intervals of 1.5 sand also an older MEG finding [6] in which ongoing cerebellar activity is found up to 4 s after an expected but omitted stimulus, with a sudden increase 50-100 ms before the next expected stimulus.These findings [6,8,9] seem to suggest that cerebellum is also involved in suprasecond timing.However, MEG's sensitivity to the basal ganglia is relatively low, and few studies report basal ganglia activity, making it hard to say anything conclusive about the basal ganglia's role in these studies [6,8,9].Neuronal recordings from monkeys shed some more light on the potential different roles of the basal ganglia and the cerebellum in timing.A study [30] found that cerebellar preparatory activity for making a selftimed saccade following a delay interval was expressed the same whether the delay interval was 400, 600, 1200 or 2400 ms, that is, it began maximally 700 ms before the monkeys were cued to begin the self-timing.Interestingly, a subsequent study [31] replicated the cerebellar findings while also recording from the striatum of the basal ganglia.In contrast to the cerebellar activity, the striatal activity began immediately after the cue and persisted until a saccade was made, likely reflecting timing of the interval.A possible interpretation of these results is that the basal ganglia are responsible for timing, sub-and suprasecond and that the cerebellum is responsible for translating the basal ganglia's timing signals into informed action.This interpretation is seemingly challenged by TMS studies [32,33] that find that repetitive, inhibitory stimulation of the cerebellum only affects time perception in the subsecond range and not in the suprasecond range.Even more so, it has also been found [34] that inhibiting the dorsolateral prefrontal cortex affects time perception in the suprasecond range while not affecting time perception in the subsecond range.We argue that the challenge from these TMS studies is mitigated by considering the cerebellar preparatory activity reported earlier [30,31].In the subsecond case, there is not enough time for the cerebellar activity to build up, that is, in those cases [30,31], it was found to build up immediately, thus a change in performance would be expected (Figure 2a).In the suprasecond cases, where activity in the normal case started 700 ms before the saccade was to be made, there would be longer time for this preparatory activity to build up, thus there does not have to be a change in performance (Figure 2b).

The cerebellum as an enabler of proactive action
In tasks like this requiring motor action, it is likely that the cerebellar information is relayed to the primary motor cortex through the ventral intermediate nucleus of the thalamus [35].It should be mentioned that the basal ganglia also project to the thalamus [36], thus the timing information may partly reach the motor cortex from here.But as shown in Ref. [31], striatal responses peak at the Expected responses for timed saccades with and without TMS following our line of argumentation that (a) without TMS; for both short (green) and long (orange) time segments, cerebellar preparatory builds up unimpeded (see Figure 3d in [31] for the real data this is based on).(b) with TMS; for short (green) time segments, the preparatory response cannot reach its full height due to the inhibitory nature of the TMS, but for long (orange) time segments, the preparatory response could reach its full height, albeit with a slower rise.
The role of the cerebellum in timing Andersen and Dalal 3 moment of the saccade both for the sub-and suprasecond responses (just as the cerebellar responses do).Thus, the striatal responses are not sufficient for enabling precise timing behaviour -otherwise, TMS [32][33][34] of the cerebellum should not have an effect for either subsecond or suprasecond timing.However, the prefrontal cortex also seems to play a role in suprasecond timing as TMS-induced inhibition of the dorsolateral prefrontal cortex has been reported to induce underestimation of suprasecond intervals [37,38] but also overestimation [34].The role of the prefrontal cortex in timing should be investigated further.
One might rightfully ask why the basal ganglia's timing signals would have to be relayed to the cerebellum before being made relevant to action.Our proposal is that the cerebellum is a hub for integrating sensory expectations, be they temporal [6][7][8][9][10], spatial [15], or spatiotemporal [12][13][14] -and cerebellar lobule VI and cerebellar crus I may be especially relevant to this [16], as these are the common areas found across the literature [7][8][9][10][11][12][13][14][15][16].In this sense, the cerebellum is responsible for timing in a restricted manner: that is, it integrates timing information from the basal ganglia with spatial information, creating a forward model [1] that predicts both sensory and temporal aspects of upcoming stimuli.Given accurate predictions, this makes both temporally informed action [9] and delicately timed action possible [30,31].
This information is likely relayed to the thalamus [9,35] where it can be incorporated to establish an action plan to be carried out by the primary motor cortex.In terms of the example from the introduction where an extra step is taken at the top of a flight of stairs -the missing confirmation of the cerebellar prediction of an extra step could be relayed to the thalamus such that appropriate, proactive action can be taken, hopefully making sure that one does not stumble.The action is proactive in the sense that you do not have to finish the motor programme before you can adapt and change the action.Our prediction would be that with either a dysfunctional cerebellum or thalamus, one would be more likely to stumble because of not being able to build sensory expectations or converting them into action, respectively.The model that we propose (Figure 3) is anatomically feasible as tracing studies of the monkey [39][40][41] have shown that the basal ganglia and cerebellum communicate.Output of the subthalamic nucleus reaches the cerebellar cortex through the pontine nuclei.It has also been shown that cerebellar output is passed on to the motor output regions of the thalamus and further to primary motor cortex from there [42].The primary motor cortex projects to the cerebellum through the pontine nuclei [43].Finally, the basal ganglia have connections directly to the thalamus [44].Evidence of functional connectivity between the cerebellar crus I and the thalamus has been reported when expectations are relevant to action [9] as well as functional connectivity between cerebellar lobule VI and the primary and secondary somatosensory cortices when predicted touches are attenuated [15].Cerebellar lobule VI, specifically, has been found in a tracing study to project to the motor cortex through the dentate nucleus and the thalamus [45,46].

Perspectives
To understand the relations between the areas proposed here (Figure 3), it is a challenge to find the optimal neuroimaging modality.While it has been shown that MEG can register activity from the cerebellum [47,48], the thalamus and basal ganglia are very challenging targets for noninvasive electrophysiology [49,50].Insights from a new generation of deep brain stimulation (DBS) equipment [51,52] may shed some light on this -instead of just stimulation, new equipment can also record activity from these regions.Frequent targets of DBS are basal ganglia (Parkinson's disease) and the thalamus (essential tremor) [53], but cerebellum has also been targeted in movement disorders such as dystonia, ataxia, and tremor [54] with at least two case studies showing clinical improvements of tremor following DBS of the dentate nucleus [55,56].Invasive recording of the cerebellum in humans may thus be feasible in the near future.
In terms of understanding the cerebellum's interplay with subcortical regions, fMRI offers the possibility to also image subcortical regions -however, the temporal resolution of fMRI is on the order of seconds, so while it Proposed model for the cerebellum's involvement in timing action: the basal ganglia pass timing information on to the cerebellum.Sensory expectations in the cerebellum then inform the thalamus, which in turn informs primary motor cortex.At the same time, sensory predictions and feedback are sent back and forth between the cerebellum and primary motor cortex.The basal ganglia can also send information directly to the thalamus.may reveal differences in cerebellar, basal ganglia, or thalamic timing-related activity between carefully controlled conditions, it does not offer the temporal precision of electrophysiological methods such as MEG.This temporal precision is necessary to differentiate when, for example, cerebellar and basal ganglia responses begin [31].In terms of increasing the sensitivity to cerebellar activity through better sensor coverage, newly available optically pumped magnetometers (OPM) for MEG [48,57] allow free placement of sensors, while groups working with electroencephalography (EEG) have started extending EEG caps [58].However, the gain in sensitivity of these improvements, if any, to thalamus and basal ganglia is unclear.
In conclusion, our proposal that the cerebellum is a hub for integrating information, temporal as well as spatial, necessary for building sensory expectations and for utilising those very expectations in informed action can be evaluated using a combination of old, fMRI, DBS without measurement, TMS, EEG and MEG, and new techniques, OPM-MEG, DBS with measurement and EEG with cerebellar coverage [48].

Figure 1 Current
Figure 1

Figure 2 Current
Figure 2

Figure 3 Current
Figure 3