Topical Review

The Cerebellum: It's About Time! But Timiug Is Not EverythiugNew Iusights Iuto the Role of the Cerebellum iu Timiug Motor aud Coguitive Tasks
Michael S. Salman, MSc, MRCP

ABSTRACT Converging evidence from different research studies supports a role for the cerebellum in timing neural processes. The cerebellum is part of a distributed system for motor control. The timing hypothesis provides a specific functional role for the unique contribution of the cerebellum. The timing capabilities of the cerebellum appeal- to extend beyond motor conirol into tasks focusing on perceptual processing tliat require the precise representation of temporal information and sensorimotor learning. Behavioral and modeling studies suggest that the cerebellar timing system is best characterized as providing a near-infmite set of interval-type timers rather than as a single clock with pacemaker or oscillatory properties, but this is controvei-sial. In addition to learning precisely thiied motor responses, the cerebellum is involved in on-line processing using feed-forward systems for which sensory input is used prior to movement execution to improve movement accuracy. This would be a mechamsm for triggering accurate "time." The cerebellum continues to fascinate scientists, and although sur\ival is possible without the cerebellum, tl^e resultant quality of lile is significantly compromised with clum.siness, ataxia, hypotonia, dysarthria, slowing of various cognitive perceptual processes, and impaired fine motor and ocular-motor coordination. The last tliree decades have seen the development of research that has focused on how the (erebellum functions. Further neurophysiologic research in cerebellaj- cortical neuroLransniission is likely to further our understanding of the cerebellar contribution to timing sensorimotor processes. (J Child Neural 2002; 17:1-9).

The foiuth dimension of our world, time, has tended to be forgotten in many theories of perception and motor contrcjl. Yet actions and events take place overtime. It is not mandatory to postulate that temporal information is represented explicitly. Variation in the speed and durat ion of a reaching movement might be an emergent property of the rate at which miLscle units are recruited. Therefore, temporal regularities in sequential actions may not reflect direct control processes. Nonetheless, many phenomena suggest the existence of an internal timing system in which temporal information is explicitly represented. For example, humans caii

Received June 15, 2001. Accepted for publication Juty 17,2001, From tlie Division of Neurology, Hospital for Sick Cliildien, Toronto, ON, 'Hie author is finimc'ially supported by a research traiiung fellowship awaixied by the Hospital for Sick t'hildren's Research Institute. Addiess correspondence to Dr Michael S, Salman, Divisions of Neurology ;md Psychology, Hospital forSick Children, 55ii University Avenue, Toronto, ON M5(J 1X8. Tel: 416-813-6918; fax: 416-813-6334; e-mail: Micliae!,Salmani&isickkids.on.ca

easily discriminate among intervals between event-s separated by 400 or 425 msec or show high sensitivity to perturbations in a stream of rhythmic events.' It has been hypothesized that the cerebellum operates as a specialized module for timing. in this review, the evidence is presented for and against this hypothesis. The evidence is based on a sensitive but not specific literature search strategy. This approacth was taken in order not to miss important and relevant studies in this area. The initial MEDLINE search identified over 500 studies; 52 were initially selected based on the relevance of the title and abstract. Twenty-two studies were then retrieved in full aiid reviewed in detail. A fuither 12 relevant references that were mentioned in important studies were also retrieved to verify some of the quoted "facts." Animal and human studies in health and disease were included to ensure adequate coverage of the topic. Different lines of evidence were reviewed (eg, theoretical, physiologic, neuropsychologic, and neuroradiologic) to enrich the discussion. The cerebellum's role in oculomotor control is not discussed In this review. i 1

Journal of Child Neurology I Volume 17, Number 1, January 2002

NEUROBIOLOGY OF TEMPORAL PROCESSING Tlie passage of time can be estimated by means of (I) stimulus duration by verbal estimation, for which subjects are r equired to evaluate verbally the dui-adon of either filled (eg, a tone is heard) or empty intervals; (2) temporal discrimination tests, for which subjects are presented with two stimulus durations and are asked to determine whether the second one is shorter or longer than the first stimulus; (3) temporal production, for wliich the subject is asked to produce a certain interval (eg, by pressing a button); and (4) temporal reproduction, for which the subject is presented with specific stimulus duration and is asked to produce that duration. In temporal orientation, the subject is asked 10 identify the hour of the day, the day of the week, and the month of Uxe year. This task is different and not correlated to time estimation, although both require tlie optimal use of memoiy It is speculated that bniin mechanisms involved in tlie c:onect estimation of slioit durations (msec) are different firom tluise involved in long durations (sec); in the fonner, the subject is unable to luse cues such as counting.- The cerebellum and the basal ganglia are considered to be candidate regions within the brain for intenial clock representation. Other aj eas include the prefrontal cortex and temporal and parietal lobes, but the evidence for this is scanty and not convincing.' The circadian rhythm is another example of a biologic clock (~ 24 hours). It is located in the hypothalamlc suprachiasmatic nucleus. There are many factors that are thought to influence time estimation in noniial subjects. Tliese factors have been discussed in detail- and include (a) isolation from obvious temporal markers (eg, day/night, cycle), (b) attention, (c) short term memory (hypothesized but not proven), (d) aiousal, (e) development (adult levels are reached around 11 years of age), (f) type of stimulus (auditory stimuli are judged to be longer than visual stimuh of equivalent durations), and (g) cognitive strategies (eg, counting silently).

contribution to regulating motor behavior relates to realtime on-line processing it performs to optimize the out(nit of the motor system ciuring the performance of coordinated ajid novel movements." Over tlie last decade, functional ncuroimaging studies have implicated the cerebellum in diverse higher cognitive and behavioral functions including exe( utive function (jilaiming. abstj-act reasoning, working memoiy), spatial cognition (visuospatial organization, memory), language (fluency, prosody), and emotional regulation.'" Tliis is not surprising given the substantial input to the cerebellum from almost all levels of tlie centi'al nei-vous system.'* In general terms, these inputs are via vestibulocerebellar (governing body equilibrium and eye movements), spinocerebellar (controlling execution of limb movements), and cerebrocerebellar (through the pontine nuclei; implicated in the initiation, planning, and timing of movements) pathways. The cerebellum plays a cnicial piu1 in the visual giiidiuice of movement. The midline cerebellum is concerned with calibrating reflex movements of tlie eyes, whereas the cerebellar hemispheres are concerned with the voluntaiy control of eye movements." The main contribution of the cerebellum to tlie control of these systems is prediction, the ability to anticipate events and thus adapt movements fluenlly to their required goal.'" This may be regarded as one aspect of how the cerebellum controls the quality of movement. More recently, molecular and genetic stutlies have identified many genes and mutations in cerebellar ion channels (eg, the Ca-* channel in episodic ataxia tyjie 2) and defective proteins in diseases causing spinocerebellar degeneration (eg, mitochondrial iron transport protein in Friedreich's ataxia). These advances have added another angle t(j t-erebellar reseai'ch tliat will help us understand and advance the basic mechanisms underlying cerebellai- function and, in pjuticular, its contribution to timing. ANATOMY AND PHYSIOLOGY OF THE CEREBELLUM The internal structure of the cerebellum seems simple and surprisingly consistent throughout the different cerebellar regions."* It is also highly conserved across species. I'his suggests that the cerebellmn performs the same general computation for many different tasks. However, important regional heterogeneities within the cerebellum in citlior chemistry or physiology exist For example, it was found tliat the cerebellar polypeptides, zebiiiis I and II, are Purkiiye c ell specific within the cerebellum.'' The m^jor anatomic compartments of the cerebellum are the cortex and the deep nuclei. Purkir\ie cells, the only outputs from the cerehellar cortex, project through inhibitory connections to the deep cerebeUar nuclei, which provide the output to other brain regi(3ns. Inputs are traiismitted to tlie cerebellum over climbing fibers and mossy fibere, two pathways with fundamentally different physiology and anatomy (Figure 1). The inferior olives give rise to climbing fibers. Climbing fibers project to Purkinje cells. Each Purkinjc cell receives monosynaptic inputs from just one climbing fiber.

FUNCTIONAL ROLE OF THE CEREBELLUM

The cerebellum constitutes only 10% of the brain by weight ;md volume' but contains ~ 5 X 10'" neiu-ons, which is in the same range as tbe number of neurons in the whole cerebral cortex,'' estimated to be 22 X 10'". Most of the brain's inhibitoi-y neurotransmitter, 7-aminobutyric acid (GABA), is located in the cerebellum. Tiaditionally, the cerebellum was thought to be devoted entirely to the quality of movement, particularly the coordination of skilled volimtary movement including eye movements and control of motor tone, posture, and gait (ie, the cerebellum is concerned not with movement per se but the quality of movement). This popularized \iew states that the principal role of the cerebellum is tlie acquisition and .storage of the plastic changes underlying learned and conditioned motor behaviors. An alternative view proposes that, although the cerebellum is likely involved in skill acquisition, its most critical

The Cerebellmn / Salvian

Cerebellar cortex 25-msec cycle Purtunje cli = 300 msec

Climbing fiber

= 400 msec

Counter Excitation Inhibition

I Oeep cerebeilar nucleus

Interior olive Figure 1. Schematic of the basic cerebellar circuit, which is iterated throughout the structure. A mossy fiber - granule ceil * parallel fiber input. An inferior olive -- climbing fiber input and some, but not all, of the intricate connections of the inhibitory interneurons are shown. Arrows indicate direction of impulse transmission. (Reprinted with permission from Raymond JL, Lisberger SG, Mauk MD:The cerebellum: A neural learning machine? Science 1996;272:n26-1131. Copyright 1996 American Association for the Advancement of Science.)

300 msec

400 msec

500 msec
O 1B96 Current Opinion In Neurobmlogv

wliich makes it emit complex spikes, occiuiing at 1 or a few Ilz. Complex spikes occur infrequently and hence are not thought to code for infoniiation transfer. This led to the sug.t*estion that climbing fibers-Purkinje cells are involved in keeping time for coordination. The mossy fiber inputs aiise from a variety of pontine nuclei and spinal cord. They influence Purkinje cell firing through their comiection with granule cells. Granule cells form parallel fibers thai make excitatory connections with numerous Purkirge cells and inhibitory connections on interneurons. Inhibitory intenieurons synapse on Purkii\je cells aiifl also provide inhibitory feedback to the granule cells. This airangement offers many opportmiities for spatial and temporal integration. Mossy fibers, via the granule celLs, cause Purkinje cells to emit simple spikes at rates of up to 100 Hz. TMs is probably compatible with a frequency code for information transfer. There are also direct axonal collaterals between mossy and climbing fibers and the deep cerebellar nuclei.'Climbing fiber input is capable of producing a longterm depression of Purkii\ie cell responsiveness to parallel fiber inputs.'' Long term depression had been observed inde|)endently in different laboratories using the conjunctive stimulation paradigms in which long-term depression is produced by pairing the stimuli tliat activate the parallel and climbing fiber inputs to the same population of Purkin.je neurons. However, the sustained high rates of climbing fiber activation that are needed to produce long-term depression liave not been observed under more behavioral conditions. Natural stimuli have not been shown directly to produce long-

Figure 2. Two mechanisms for representing temporal information. A, Clock-counter models postulate a pacemaker that produces output to a counter. Longer intervals are represented by increases in the number of pacemaker outputs that accumulate in the counter B, Intervalbased models assume that different intervals are represented by distinct elements, each corresponding to a specific duration. (Reprinted from Current Opinion in Neurobiology, Vol. 6, Ivry RB.The representation of temporal information in perception and motor control, 851-857, Copyright 1996, with permission from Elsevier Science.)

term depression-related phenomenon. In addition, longterm depression is usually not a reversible experimental phenomenon.''" This aigues against the cerebellum as a storage site.
GAIN CHANGE AND DYNAMIC SELECTION HYPOTHESES

Interestingly, it has been shown that short-term enhancement can occur in Purkirye cells in response to lnossy-parallel fiber input. In the gain change hyjiothesis, this occurs when climbing fiber activation takes place within a critical period prior to the arrival of other synaptic inputs to Purkii\je cells. These observations have been incorporated into the dynamic selection hypothesis, which proposes a mechanism by which the climbing fiber inputs can act to specify specific spatial distributions of Purkinje cells that will be most responsive and hence most highly modulated by combinations of mossy fiber inputs originating from many sources. The well-characterized oscillatory activity of the olivary neurons (6-10 Hz) is postulated to provide a pacemaking signal and lo restiict the control process to particular moments in time (Figure 2). This view requires a task-specific, behaviorally dependent activation of specific populations of olivary neurons reflecting various pattems

Journal of Child Neurology I Volume 17, Number 1, Janiiary 2002

of converging inputs (ie, muscle synergies) to the inferior olive."'' The theory emphasizes the Involvement of the olivocerebellar system in on-line, real-time processing, allowing movements to be executed rapidly in a feed-forward manner with minimum sensory guidance and feedback rather than in processes involved in establishing memoi-y in the cerebellum." This process is not easily observed by recording the activity of individual Purkin.je cells, one at a time, as this may not appear to be significantly related to particular instances of movement. However, tlie activity of groups of olivary neurons, producing distributed but synchronous activity within sets of Purkirye cells, is clearly related to movement.'^ There does not seem to be any shortage of theories explaining the function of the cerebellum. In one theory,"' it was proposed that the different speeds of conduction through parallel fibers, together with the simultaneous input from climbing fibers, cause increased synaptic discharge in a previously facilitated Purkinje cell synapse hence, Purkinje cell output codes for specific temporal information. In a more recent theory,'^ it was proposed tliat Golgi c ells fulfill a gating task by forming short, and welldefmed time windows within which granule cells can reach firing threshold, thus organizing neural activity in discrete "time slices." In this theory, the climbing fiber's spikes act as synchronization and as a teaching signal, not as an enor signal. ATAXIA TELANGIECTASIA AND TIME PERCEPTION Ataxia telangiectasia is an autosomal-recessive progressive disorder, with an early childliood onset of around age 2 to 3 yeai\s. It presents with ataxia. extrapyramidal movement disorder, and c utaneous telangiectasia. Tliere is an increased susceptibility to chromosomal breakage with failure in its repair mechanisms. Ataxia telangiectasia causes acquired cereboliai' cortical degeneration affecting Purkirye and g!"anular cell layers. In an experiment using a test of judgment of duration,'" 17patients with this disease were compared with 21 control subjects. The age range was 7 to 21 yeare in both groups, with significantly more girls in the control group. The Wechsler Verbal IQ was significantly higher in the controls (107 versus 74). Fi'equency perception of sound was used as a control task. For judgment of duration (perception task), subjects compared successive time intervals generated by two pairs of (73-dB) tones that were 50 msec in duration and 1 kHz in frequency. The first pair of tones was separated by 550 msec; the second pair, presented 1 sec later, had variable intervals that were either shorter or longer than 550 msec; 10 practice trials for each task were offered before the start of the experiment. After statistically controlling for Verbal IQ, children and adolescents with this disease performed significantly worse than controls on judgment of duration (P = .01) but not of pitch, suggesting that the cerebelliun may be critical for judging explicit time intervals.

In this study, only seven participants in the ataxia telangiectasia group had evidence of cerebellar atrophy on brain imaging (magnetic resonance imaging [MRII/coniput ed tomography [CT]). Imaging was noniial in four on CT and not available for the rest. It is not cleai- when the scans were done. This matters becaiLse the disease is progressive, and the results may be biased as other abnormalities in this disease are reported in the basal ganglia, albeit uncommonly. Also, ataxia telangiectasia causes increased susceptibility to infections and brain lymphoma later in cliiidhood and early adulthood that may be silent initially; this can further bias the results. Although age was matched, the age range chosen was wide, and, as mentioned earlier, there is developmental influence on time estimation that reaches adult values around the age of 11 years, this could have confounded the results, albeit in both groups. In addition, the study was not blinded and had a small niuuber of subjects. CEREBELLAR TUMORS AND TIME ESTIMATION CerebeUar lesions in childhood are not rare. The posterior fossa is the site of half of all < hildhood tumors, of which medulloblastoma and astrocytomas are the most common types. They displace the cerebellum but rarely infiltrate it. In a study done on long-tenn survivors with these tumors,''' few timing functions were studied, including short-tiuration perception and long-duration estimation. The paiiicipants included 40 controls, 20 subjects with cerebellar astrocytoma, and 20 subjects with medullobiastoma. The mean age of diagnosis was 8 years. The mean age of testing was done 14 years after diagnosis. Wechsler Verbal or Performance IQ was greater than 70 in all participants. In the first experiment, short-duration and frequency perception were measured using a two-alternative forcedchoice procedure. The ability to discriminate among intervals in the 400-msec range was tested. The participants made their choices by pressing on a left or right button according to their ability to discriminate between two empty intervals marked by a 1-kHz tone of 50-nisec duration. Participants were instructed to press the left button if they thought that the first duration was longer and the right button if they thought that the second duration was longer. The tumor group was less able to discriminate among durations in the 400-msec range (P = .006). The two groups did not differ in their ability to discriminate frequency differences In the 3-kHz range. In the second experiment, a retrospective estimation of 60-minute duration was assessed by asking the subjects to estimate the duration of time passed in the waiting room without the aid of external cues. Prospective time estimation was assessed by asking the subjects to tell the examiner when they believed that 30 minutes had elapsed. The two groups did not differ significantly in their ability to estimate long durations, althougli timior-relatf^d prospective memory deficits interfered with the ability to produce longduration prospective estimates.

The Cerebeilimi / Salman

The results of the experiments were the same regardless of the pathology or treatments used. There was no evidence of functional recovery of short-duration perception. However, this study is not longitudinal, and the ability to estimate short durations close to the time of diagnosis is unknown. The two tumor groups were affected similarly despite the more extensive lesions associated with the medulloblastoma group. Possible bias may have arisen from the wide age range at the time of diagnosis (1 '/i2-15^/i2 years). The wide age range assumes that the effect of having tumor and treatment, including surgery and radiation, in early childhood is similar to having it in late childhood. Many studies have shown this not to be the case; for example, radiation effects on the developing brain are particularly harmful below the age of 3 to 4 years. Overall, tl\is study was well designed and showed that childhood lesions of the cerebellum produced enduring deficits in shortduration perception. i CEREBELLAR MOTOR CONTROL AND TIMING Motor control can be defined as the process of restricting the output of the motor nervous system so tliat meaningful and coordinated beha\ior ensues. Timing is a critical aspect of movement. This is not surprising as controlling movements inlierently involves activating motor units at the correct times. Accurate anii movement requires correct time courses in the command for force generation in each muscle, as well as suitable amplitude of the force created by each contraction. There are over G O skeletal muscles in the O human body, each having many motor units. Complex movement sequences often involve many muscles, resulting in an astronomical number of simultaneous and sequential muscle contractions. Continuous control through time has extremely high computational overload for which not enough neurons are present. The stereotypical and timelocked performance of multiple and clearly independent muscles suggests that they are activated by a single command and controlled as a functional group (ie, muscle synergies). It may be necessary to use motor learning to regulate the timing of motor commands to produce coordinated smooth movements'^-*^ that meet motor demands under various circumstances, but real-time, on-line processing has also been suggested." '^ Defective motor coordination is a principal feature of diseases of tlie cerebellum, manifested by a lack of smoothness dining movement execution. One possible explanation revolves aroimd the disruption in the timing of the normal patterning of agonist and antagonist muscle activity in the course of the movement^'-' In support of this idea, people with cerebellar disease have excessive amounts of agonist-antagonist co-contraction at movement onset, improper timing of phasic bursts of activity in agonist and antagonist muscle pairs causing a delay in movement initiation. This is caused by a delay in tlie onset of phasic motor cortex neural discharge owing to decreased input from the cerebellar hemispheres. Dysmetria

results from abnonnal timing and the intensity of the antagonist burst necessary to break the movement. In addition, the time delay in tracking movements is increased, possibly because of an increase in reaction time for movement initiation.-^ Dysdiadochokinesia is slowing and abnormal performance of rapid, alternating movements. It may be explained by slowness at the turning points caused by delays in movement initiation and/or dysmetria at the end of the movement,-'' compounded by delay or absence of agonist pause. This results in abnormalities of movement velocity, acceleration, and deceleration.''-'' The cerebellum is involved in sequencing agonist and antagonist activity; hence, timing may be faulty in cerebellar disease. In addition to timing, the cerebellum has been shown to be involved in other aspects of movement control including modulating reflex gain (eg, long latency reflexes),^''' thereby maintaining effective joint compliance; compensating for inlierent mechanical instability, controlling movements requiring multiple joints,^" and updating motor acts to correct for any mismatch between the current and the desired limb position. In an interesting si udy on the effect of weight on cerebellar hypennetria,-' it was found that adding weight to a rapidly n\oving limb increased the overshoot in patients with cerebellar disease and reduced it in nonnal controls. This contrasts with previous studies showing a reduction in kinetic tremor with added weiglit to the hmb. It was postulated that kinetic tremor and hypermetila in cerebellar disease seem to have different pathophysiologic mechanisms. The authors suggested that the lateral cerebellar cortex is involved not only in programming the time onset for the antagonist muscle contraction but also the amplitude and intensity of the conti action based on Ihe initial position of the limb, the position of the target, and the inertia to be overcome. Hence, the cerebellum is iTsponsible for providing more than just timing function. Physiologic, anatomic, and clinical studies-^ have also associated the lateral cerebellum with movement planning and programming, whereas the intermediate and medial regions were associated with movement execution. Based on these fmdings and the Wing and Kristofferson model-'" of the two processes involved in periodic behavior, namely, timekeeper and implementation systems (Figiu'e 3),^ an experiment was designed to measure movement timing and movement execution using rhythmic tiipping in patients with localized cerebellar damage. Seven patients aged 19 to 66 years were tested. Four of them had predominantly lateral cerebellar lesions, whereas in three the lesions were centered in the medial zone of the cerebellum. This was only an ^proximation, however, based on clinical, surgical, and CT scan reports. The lesions were either caused by vascular strokes including hemorrhage or excised tumors. Tiie authors did not give details on the etiology of the hemonhage (eg, trauma). This is relevant as the effects of concussion may not show up on clinical or radiologic examination but may, nevertheless, ii\fiuence the subject's performance. Some of the patients were evaluated in

Journal of Child Neurology I Volume 17, Number 1, January 2002

Central
signal Peripherol fesponse

CEREBELLUM AND EYELID CONDITIONING

Central signat
Peripherot response

B
Figure 3. Hypothetical series of inter-response intervals resulting from the two-process model of Wing and Kristofferson.^" A, Series of intervals in which the only variability is introduced on the third response resulting from added implementation (motor delay) time. S, Series of intervals in which the only variability is introduced by the long clock time on the third interval. C = clock; I = interval; D - motor implementation (delay). (Reprinted with permission from Experimental Brain Research, Dissociation of the lateral and medial cerebellum in movement timing and movement execution, Ivry RB, Keete SW, Diener HC, Vol. 73, 167-180, Figures IA and IB, 1988, Copyright notice of Springer-Verlag.)

Germany, others in the United States. Hence, the evaluation was not identical, and. in some cases, the examiner was not blinded Io the outcome of the tapping performance. Tlie subjects were instructed to tap a microswitch first with and then without (65-dB) tones of 50-msec duration presented at sufficiently slow regular intewals of 550 msec. The index fmger of either the normal or impaired hand was chosen randomly and put on a mit roswitch. mounted on a wooden block, and linked to a computer. The computer recorded the subjects' response to the nearest millisecond after a few practice trials. Control subjects were not included in the experimental design as the authors chiimed that in a pre\'iously published study on nomial subjects, no difference in the same task proficiency was found based on hand dominance. All of the patients were found to have increased variability in performing rhythmic tapping with the impaired hand (ipsilateral to the lesion) compared with the ctjntrol hand. Further analysis, biised on the Wing and Kristofferson model, which assumes that the timing and motor implementation processes operate independently, each witli its normal variance (ie, open-loop mode only), revealed that the poor performance of patients with the lateral lesion could be attributed to a deficit in the central timing process. Patients with medial lesions were able to accurately determine when to make a response but were unable to implement the response at the desired time. After laking the problems (higliliglited above) in tJie design of tJiis study into account, the conclusions of the study are interesting and a{id furtlier weight to the involvement of the cerebellum in temporal processing. Many studies, in addition to this study, have suggested that the cerebellum can operate in an open-loop (feedback free) system but that timing is unlikely to be the sole function of the lateral cerebellar cortex, as discussed earlier.

The temporal specificity of cerebellar motor learning has been revealed most clearly from analysis of eyelid conditioning. Eyelid conditioning provides a relatively direct look at the temporal specificity of cerebellar learning. In short, training involves a relatively netitral st imuliLs such as an autiitory tone followed by a reinforcing stimulus such as stimulation aroimd the eye (eg, air puff). The air puff acts as an unconditioned stimulus, eliciting a reflex eyelid closure (retraction of the eyeball aiid closure of the nictitating menibr<me and the eyelids) with 25- to 40-msec latency. Repeated pairing of the tone and the air puff promotes the acquisition of a conditioned response: the eyelids close in response to the tone alone. Many studies have shown that the sound is conveyed to the cerebellum via mossy fibers from the auditoiy poiiion of the doi'solateral pontine nucleus. The air puff is conveyed to the anterior lobe of the cerebellar cortex via climbing fiber inputs (and possibly mossy fibers)'- and their collaterals to the deep cerebellar nuclei. The output of the cerebellum from the anterior ii\terpositus nucleus to the red nucleus and oUier brainsteni nuclei is responsible foi- pxi)ression of tJie conditioned responses.-" Ablation of the anterior interpositus nucleus abolishes the learned eyelid responses to the tone without preventing the reflex response to the puff. FurtheiTuore, learning still occurred when electrical stimulation of the mossy and the cUmbing fibers replaced the tone and the air puff. Reversible ablation of the red nucleus during trainhig abolished the expression of the eyelid response temporarily, but learning still occurred, as indicated by the expression of the leamecl response fiftor the inactivation was removed. These experiments demonstrate that the cerebellum must be the site for learning.'^ As the basic properties of the cerebellar inputs are known, the temporal properties of cerebellai- learning can therefore be inferred from the temporal properties of eyelid conditioning. Tliere is a dependence on the delay between the onset of the tone and the air puff. No learning occurs witli delays < 100 msec. Learning occurs with delays between 150 and 500 msec, which gradually declines as the delay increases beyond 500 msec. When learning occurs, the responses are precisely timed to peak at the time of the air puff This does not reflect the fact that strong responses are fast and weaker responses are slower. The cerebellar cortex appears to play a special role in regulating the timing of learned movement. After a period of conditioning in which t he tone and air puff ai e paired with a short interval, it was found that when delivered alone, the tone elicited a brief, short-latency conditioned respoase of the eyelid. Lesioas that included the anterior lobe of the cerebellar cortex caused little change in the timing of the coiv ditioned reflex. When the two stimuli were condiiioned with a longer intewaJ (ie, the aii' puff occiured at a long interval after the onset of the tone), it was found that the tone elicited a prolonged, longer-latency eyelid response. I^osions that include tlie anterior lobe of the cerebellar cortex triuisformed the prolonged, long-latency conditioned response

The Cerebellum / Salman

into a brief, short-latency response. Thus, the tone after tlie lesion had the same short latency independent of its timing before the lesion. In both conditioning experiments, deep cerebellar nuclear lesions abolished tliis learned response, presumably because these lesions destroy all of the output from the cerebellum."' The interpretation of tliis experiment was that the cerebellar cortex is required for the expression of lemTied and timed responses that are delayed relative to the stimuli tliat elicit them. The cerebellar cortex may therefore be the site in which the memory of learned timing is stored. The cerebellum not only leanis but also learns to change its output at the coiTect time, hi eyelid conditioning, when a mossy fiber input repeatedly predicts a climbing fiber input, there will be an increase in cerebellar outi^ut timed to peak Just prior to the arrival time of the climbing fiber input. The learning capacity that eyelid conditioning reveals is consistent with a classic feed-foi'ward use of sensory input to improve movement accuracy. Feedback of sensory infomiation is used during the execution of movement to produce accurate movement, but it is inherently hmited hy its sluggishness and tendency to oscillate when forced to operate quickly. Feed-forward control obviates this problem by using sensory information available prior to movement execution to make decisions about ensuing motor commands. FUNCTIONAL NEUROIMAGING AND TIMING Relatively little timing research has been conducted using functional imaging techniques. An advantage of this method is that the importance of a neural site relative to others can be examined more readily than in lesion or phannacologic st udy. Activation of a particular system can also provide insight into the cognitive processes underlying behavior, w hen there is a general agreement about the system's ftuiction. In a controlled study using positron emission tomography (PET) to localize cerebeUar timing function,^' six healthy adults aged 23 to 41 years compared a test interval (either 200- or 400-msec duration), defined by two tones of 50-msec duration and 1 kHz, with a standard 300-msec interval. The subjects used their right index finger if they thought that the test interval was shorter or the right middle finger if they thought that the test interval was longer to make their choices. In the control task, the two intervals had identical duration, and subjects were instructed to make alternative choices with their right index and middle fingers. Regional cerebnil blood flow using the C'''O, inhalation technique was examined. For anatomic reference, T,-weighted MRI was obtained. A significant increase in blood flow was found in the inferior parts of the ipsilateral cerebeUar hemisphere reflecting finger movements when the control task was compared with rest conditions. Bilateral activation of the temporal lobes, donsolatenil prefrontal cortex, anterior cingulate cortex, and right caudate nucleus was seen during the control task. Additional activation occurred separately during the

thning task in the cerebeUar vemiis and hemisphere bilaterally. On average, subjects correctly identified 97% of test intervals presented in the timing condition. The authors concluded that the cerebellum is involved in time perception. However, an alternative interpretation of the results has been suggested,'- whereby activation of the cerebellar hemispheres may have been the result of their involvement in sequential stimulus discrimination, a possible nonmotor cerebeUar function. In another study using PET,-^' both auditoiy (li-kllz) tones and visual stimuli (white squares appearing sequentially on a computer screen) were presented to 12 normal subjects aged 20 to 30 years in a set of experiments tiiat had a similar design to the experiments mentioned above. The subjects, however, had to reproduce the sequences presented during a pause by tapping on a computer keyboard using their right index finger. The sequences presented were (1) isosynchronous short (250-msec) or long (750msec) duration or (2) novel (ie, a mixture of 250- or 750msec) duration. Regional cerebral blood flow using the C'-labeled H,,O bolus technique was examined. In addition to the basal ganglia and modality-specific activation in the sensoty association and frontal cortices, the results revealed a supramodal contribution of the lateral cerebellar cortex and cerebellai' vennis to the production of a timed motor response, paiticiUarly when it is complex and/or novel. Tlie authors suggested that the cerebellum may contribute in two ways: (1) in computing the temporal parameters of incoming sensory stimuli and outgoing movements and (2) in learning novel, temporally precise motor responses. The results also gave suppoil to the involvement of the basal gangUa structures in motor timing that may be more directly related to implementation of the motor response than to timing per se. The role of the cerebellum in timing is conceptualized not as a clock or counter but simply as the structure that provides the necessary circuitry for the sensory system to extract temporal information and for the motor system to learn to produce a precisely timed response. The impUcit assumption in tliis study that the timing requirements increased across tasks, with the fixed interval sequence being the easiest and tlie novel sequences with mixed intervals being the most difficult to time, was not verified by the behavioral data. Experimental manipulations of a cognitive function may alter brain activity in many sites; hence, it is difficult to directly associate brain activity in a specific- aiea witli one but not another behavioral measure. Additionally, functional imaging cunvntly relies heavily on the subtractive logic, wherein the subtraction of two tasks is assumed to reflect functional activity associated with a mental operation in one but not the other task. This may not be tenable or difficult to test These limitations can seriously obscure interpretations of causal linkages between brain structure and function.' The results of functional neuroimaging from studies of motor timing are discrepant, particularly with regard to the role of the cerebellum and basal ganglia. This may be attributable to different experimental designs.

Journal of Child Ntjuroloyy I Vuhuiie 17, Number 1, Januaiy 2002

In recent years, functional MRI has emerged as a promising and noninvasive technique that may help in the functional localization of various brain regions. It does not involve the use of radioactive isotopes that aie used in PET studies. The assumption underlying the use of functional MRI is that during a specific task, blood tlow to the processing region increases with consequent increase in oxygen delivery, extraction, and use. Hence, by subtracting the resting image from the image acquired during the task, the result should be a functional, task-specific map of the brain. In a study on memory-timed finger movement t;isk.'^ functional MRI was used to measure regional cerebral hemodynamic responses. The subjects were eight healthy male volunteers aged 19 to 27 years. All subjects perfonned the following four tasks: (1) memory-timed finger movement (using the right index finger to click on a mouse every 1500 msec without any preceding cue), (2) visually cued finger movement (by clicking on a mouse as above when the brightness of light at a fixation point changes randomly between 500 and 2500 msec, (3) silent articulation (saying "pi" silently every 1500 msec without any cue), and (4) resting baseline. The results showed that niomory-timed fmger niovenieuts significantly activated the anterior lobe of the cerebellum bilaterally (lobules IV and V), the contralateral primary motor area, the dorsal premotor area bUateraUy, tlie supplementary motor area, the inferior frontal cortex bUaterally, the left intraparietal cortex, and the right inferior paiietal lobe conipai ed with the control resting conditions. The same areas in the ipsilateral right anterior cerebellum, contralateral primary motor area, bilateral dorsal premotor area, left intraparietal cortex, and right inferior parietal cortex were also significantly activated during the visuaUy cued movement task. The supplementary motor area was activated in a more posterior location compared with the memory task. Additional areas were also activated, including the ipsilateral ventral premotor area, right middle frontal coriex, left inferior parietal cortex, right superior temporal cortex, left insula, and right thalamus. In the silent articulation task, the supplementary motor area and the left inferior frontiil cortex were activated compared with the control resting conditions and in the same areas as the first task. The anterior lobe of the cerebellum was only minimally activated in the silent articulation task. It is known that ipsilateral activation of the anterior cerebellum is associated with the index finger movement. However, the authors postulate that bilateral activation of the anterior lobes of the cerebellum, supplementary motor area, and left prefrontal cortex were probably involved in the generation of accurate timing, functioning as a clock in the central nervous system. The left prefrontal cortex is associated with language processing in terms of perception, production, and memory; hence, its activation may be involved in the subvocalization associated with chronometric counting. Similar reasoning was postulated by the authors to explain supplementary motor area activation in this study, for which the authors quoted few studies

Memory

DedsloQ

Figure 4. Diagram of the information-processing mode! of interval timing derived from scalar theory. The clock stage component consists of a pacemaker that discharges pulses, which are gated by a switch and then passed into an accumulator to be counted. Accumulated pulses are encoded into working memory on each trial and stored in reference memory over trials.The decision process compares pulse counts from the accumulator with those in memory to determine how or when to respond. Attention can influence the clock process by delaying (or hastening) the onset of the switch or changing the pulse count in the accumulator. Attention and strategic processes influence the decision stage by biasing response thresholds. (Reprinted from Harrington DL. Haaiand KY: Neural underpinnings of temporal processing: A review of focal lesions, pharmacological, and functional imaging research. flev/Veu/-osc/1999;10:91-n6, by permission of Freund Publishing House Ltd.)

supporting the role of the supplementary motor area in sequence generation from memory that fit into a precise timing plan. The absence of basal ganglia activation in this study was blamed on susceptibility artifacts, hi a functional MRI study investigating explicit time estimation,''"' participants were given time intervals (12-24 sec) and were asked to indicate when it elapsed without external clock cues. Compared to control tasks (counting backwjird or forward), lateral cerebellar and inferior lobe activation was seen. This study implicated the involvement of the cerebellum in estimating longer time intervals (in the order of seconds) compared with other studies that implicated the cerebellum in timing tasks in the order of milliseconds. The above studies show that multiple neural systems support temporal processing, consistent with the scalar timing theory, which assumes tJiat the scalar property of timing is attributableto multiplicative variance mechjinisms that can be influenced by the clock, memory, and decision processes (Figure 4). Isolating these processes and, in particular, expUcit timing from those involved in nontemporal processes in functional neuroimaging studies is complicated, with many limiting factors as described earlier.
Acknouiledgmenl
I w<Hild like Ia tliaiik Dr Maiirci'ii ii.s fur her (ioiicral sii|)p<iii iuid cTiiliiisi-

iLsm iilxniE liie cerelx'llum and Ore William Mackay and Jaines A. Sliarpe for their helpful .suggestions and comments on the manuscript

The Cerebellum ! Salimin

R^erences
1. Ivry RB: The representation of temporal information in perception aiid motor control. Ctirr Opin NeuroMol 1996;6:851-857. :;. I^londe R, Hiuineqiiin I): The neiirobiologiral basis of time estimation aiui temponil order. Rev Nrum.sci lf)99;10:l.'Jl-17;i. '!. HaiTiiigton DL, Haaiand KY: Neural underpinnings of temporal processing: A review of focal lesions, phamiacologic:al, and fimctional imaging research. Rev Ncumsci 1999; 10:91-116. 1. Dolari RI: A cognitive affective role for the cerebellum. Brain Ii)fi8; 121:545-546. 5. Robinson FR: Role of Ihe cerebellum in movement control and adaptation. Curr Opin Neurobiot 1995;5:755-762. G. Bloedel JR. Bracha V, I^^rson PS: Real time operations of tlie cerebeUar cortex. Can J Neurol Sci 1993;20(StippI 3):S7-Si8. 7. Schmahinitnn JD, Sherma JC: The cerebellar cognitive affective syndrome. Brain 1998;121:561-57f). 8. Honi AKE, Buttner U, Buttner-Ennever JA: Brainstem and cerebeUar stnictures for eye movement generation. Adv OtorhinoInryngol im?l;55:l-2B. 9. I^eigli R(j, Zee DS: Tlie proiwrtJes and netiral sutratrate of eye niovemenls, in Leigh RG, Zee DS (eds): The Neurology of Eye Movements, 3rd ed. New York, Oxford University Press, 1999, :J-15. 10. Stein JF, Giickstein MG: Role of cerebeUum in visual guidance of movement. Physiol Rev 1992; 72:968-1017. 11. Hawkes R, Blytli S, Chockkan V. e( al: Structural and molecular compartmentation in the cerebellum. Can J Neurol Set 12. Raymond JL, Lisberger SG, Mauk MD: The cerebeUum: A neural leamuig machine? Sfr;ni.rPl996;272:1126-1131. 13. Ito M: Synaptic plasticity in the cerebeUar cortex and its role in motor learning. Can J Neurol Sci 1993;20(Suppl 3):S70-S74. 14. Llinas R, Welsh JP: On Ihe cerebelltmi and motor learning. Curr Opin Biol 1993;3:958-965. 15. WeLsh JP, Llinas R: Some organizing principles for the control of movement based on olivocerebellar physiology. Pivg Brain Res 1997;1I4:449-461. Slrehler BL: A new theoiy of cerebellar function: Movement control through phase-dependent recognition of identities between time-based neural inrormational symbols. Synap.'ie 19{)0;5:l-32. 17, Kisder WM, Henimen JL: Delayed reverberation tlirough time windows as a key to cerebellar function. Biol Cybern 1999:81:373-380. 18, Mostofsky SH, Kunze JC, Cutting LE: Judgement of duration in individuals with ataxia telangiectasia. Dev Neuropsychol 2000;! 7(1 ):63-74.

19. Hetherington R, Dennis M, Spiegler B: Perc eption and estimation of time in long-term survivors of childhood posterior fossa tumoi-s. ./ hU Ncn rapnycbol Sor 200();6:682-()92. 20. MaukMD, Medina JF. Noies WL,OhyainaT: CerebeUar fimcUon: Coordination, learning or timing? Cmr Biol 2000; 10:R522-R525. 21. Dichgans J: Clinical symptoms of cerebellar dysfunction and their topodiagnostical significance. Hum Neurobiol 1984;2:269-279. 22. Diener HC, Dichgans J: Pathophysiology of cerebellar ataxia Mov Disord 1992;2:95-109, 23. Gilman S: Cerebellar control of movement, Ann Neurol 1994;;i5(l):3-4. 24. Hallet M, Massaquoi SG: Physiologic studies of dysmelria in patients with cerebellar deficit.s. Can J Neuwl Sci 1993;20(Suppl 3):S83-S92. 25. Diener HC, Hore J, Ivry R, Dichgans J: Cerebellar dysftmction of movement and perception. Can J Neurol Sci 1993;20(SuppI 3):S62-Sfi9. 26. Becker W,I, Morrice BL, Clark AW, Lee RG: Multi-joint rea* hiiig movements and eye-himd tracking in cerebeUar incoordination: Investigation of a patient wilh complete loss of Purkirije cells. Can J Neurol Sci 1991;18:476-487. 27. MacKi^ WA, Muri)hy JT CerebcUar modulation of refiex gain. Prog Neurobiol 1979; 13:361^17. 28. Ivry RB, Keele SW, Diener HC: Dissociation of Ihe lateral and medial cej-ebellum in movement timing and movement execution. E3i:p Brain Res 1988; 73:167-180. 29. Wing A, Kristofferson A: Response delays and the timing of discrete motor responses. Percept Psycliopfty.s 1973; 14:5-12. 30. IvryRB: Cerebeilai timing systems, inSchm;dimaimJD(ed): The CerebeUum and Cognition. Inlernationat Revien' of Neurobiology, vol. 41. San Diego, Academy Press, 1997;41:565-573. 31. Jueptner M, Rijnt,jes M, Weiller C, et al: Localization of a cerebeUar liming process u.sing PET. Neurology 1995:45:1540-1545. 32. Seitz RJ: CerebeUar timing process. Neurology 1996;47:306-a07. 33. Penhtine VB, Zatorre RJ, Evans AC: Cerebellar contributions to motor timing: A PET stndy of audit ory and visual i hyl hni reproduction. / Cogn NeunKsci I998;10:752-7f>5. 34. Kawashinia R, Okuda J, Umetsu A, et al: Human cerebellum plays an important role in memory-timed finger movement: An fMRl
35.

Tracy JI, Faro SH, Mohamed FB, et al: Functional localization of a "time keeper" fimction separate from attentJonal resources and task strategy. Neuroimagc 2000;ll:22S-242.

Announcement
Bernard L. Maria, MD, MBA, Chair, Department of Child Health, and Pediatrician-in Chief
Congratulations to Bernard L. Maria, MD, MBA, a long-term member of the Editorial Board of the Journal of Child Neurology, for having joined a select group of child neurologists who have ascended to the position of chairman of a department of pediatrics. On September 1, 2001, Dr Maria was named Chair of the Department of Child Health at, the University of Missouri-Columbia School of Medicine and Pediatrician-in-Chief at Children's Hospital. Dr Maiia received his medical degree in 1981 from the Universite de Sherbrooke (Sherbrooke, Quebec, Canada). He then completed a residency in pediatrics at McGill University (Montreal, Quebec, Caixada) in 1983, a residency in pediatric neurology at Johns Hopkins Hospital (Baltimore, Maryland) in 1986, and a fellowship in neiiro-oncology at the M. D. Anderson Cancer Center (Houston, Texas) in 1988. He returned to Canada as the country's first pedialric neurooncologist. Later he Joined the University of Florida College of Medicine (Gainesville, Florida), where he founded the neuro-oncology program and was professor and chief of pediatric netirology. He also received a master's degree in business administration with honors from the University of Florida in 1995. The Editorial Staff and Publisher of the Journal of Child Neuwlogy compliment Dr Maria on this outstanding achievement.