Exploring the domain of the cerebellar timing system

TL;DR: The hypothesis that the cerebellum can be conceptualized as a relatively task-independent timing mechanism that is capable of representing temporal information ranging from a few milliseconds to an upper bound of a few seconds is reviewed.

Abstract: The ability of an animal to process temporal information has adaptive significance across different temporal ranges. The ability to encode and utilize temporal information allows an animal to predict and anticipate events. However, the time scales vary widely. The predictable event might be based on information that changes over relatively long periods such as a year or a day, or over periods comprising much shorter durations, events that change within a few minutes or milliseconds. Are there a single set of neural mechanisms that are essential for representing temporal information over these different scales? Despite the fact that numerous neural structures have been linked to successful performance on a variety of timing tasks, this question has received relatively little attention. In this chapter, we will focus on the role of the cerebellum in a variety of timing tasks. We will review the hypothesis that the cerebellum can be conceptualized as a relatively task-independent timing mechanism. An important feature of this hypothesis is that the range of the cerebellar timing system is assumed to be relatively restricted. Specifically, we assume that the cerebellum is capable of representing temporal information ranging from a few milliseconds to an upper bound of a few seconds. What remains unclear is whether the cerebellum is involved on tasks spanning longer durations. Cognitive processes such as attention and memory become clearly important here, and indeed, may dominate performance for longer intervals. The animal literature points to non-cerebellar structures as playing a critical role in these tasks and we will provide a brief review of this work. Finally, we will present the preliminary results from two experiments designed to directly test the hypothesis that the cerebellum's temporal capabilities are limited to relatively short durations.

Summary

1. Studies with patients with cerebellar lesions

• These results required a reconccptualization of the domain of cerebellar function.
• This structure has generally been linked to motor functions, or sensorimotor learning.
• The fact that the patients were impaired on a purely perceptual task suggested that its domain should be specified in terms of a particular mental operation, namely the representation of the temporal relationships between events.
• The authors have hypothesized that this computational capability is invoked across a wide range of tasks that require this form of representation.

2. The cerebellum and sensorimotor learning

• It remains difficult to specif3.' the learning domain of the cerebellum (see Ivry, 1993) .
• This hypothesis emphasizes the task domain of the cerebellum and focuses on the fact that the climbing fiber pathway provides a salient error signal for shaping appropriate skeletal responses.
• By this way of thinking, the cerebellum is associated with NMR conditioning because this type of learning is only adaptive if it is appropriately timed (Kecle and lvr3'.
• That is, learning an association and forming the temporal representation of that association can not be thought of as distinct.
• Due to negative feedback loops and physiological processes such as slow IPSPs, other consequences of the mossy fiber activity may not be evident for hundreds of milliseconds.

3. Potential limitations of cerebellar timing

• Second, behavioral studies in humans suggest that there may be a qualitative change around 2-4 scc in their capacity to represent temporal infommtion.
• Below this duration, successive events are seen as belonging to a conunon temporally-defined group, regardless of whether this group has a rh}ethmic structure or lacks such organization.
• Ps.vchophysical studies have also indicated an increase in the Weber fraction on duration discrimination tasks for intervals longer than 2 sec (Getty.
• On the motor side, Mates et al. (1994) have shox~aa that when tapping with a periodic pacing signal, people shift from a predictive to an reactive mode as the target interval becomes longer.
• For intervals less than 2-3 sec. the subjects' responses tended to anticipate the tones.

4. Animal models of temporal discrimination

• To date, this animal research has ignored the cerebellum, perhaps because this structure has been assumed to be limited to the motor domain while these tasks focus on perceptual and memor), processes.
• Moreover, while some studies have used stimuli that are less than 1 sec (Allan and Gibbon, 1991) , the maiority of this work has involved stimuli that are considerably longer, frequently ranging tip to 40 sec.
• As noted above, the authors have hypothesized that the ccrcbcllar timing system is limited to relatively short durations.
• The authors working model is that this timing process is relatively immune to cognitive influences.
• The onset of a stimulus may automatically activate different sets of neurons, and mcmor), demands are minimal.

4.1. EXPERIMENT I

• To obtain psychometric fi~nctions, the test phase included both the training durations and probe durations.
• For the SR task, there were nine probe durations ranging from 200 to 850 ms, with seven of these durations falling between the endpoint values.
• On 50% of the trials, one of the two training durations was selected and correct responses were reinforced.
• On the other 50% of the trials, one of the nine probe durations was selected and no reward was possible.
• The rats completed 36 sessions of the test phase prior to surgery.

4.1.2. Results And l)iscnssion.

• The effect of task and the interaction were not significant.
• In contrast, the consistency measure revealed a dissociation between the two tasks.
• This result is in accord with the hypothesis that the temporal range of the cerebellum is limited to short intervals.
• For the first two postsurgery sessions on the SR task, the consistency scores were -5.25 and -4.88.
• Second, their animals received many more trials post-surger).,' than are commonly used in lcsion studies.

4.2. EXPERIMENT 2

• Third, all sessions were four hours in duration and alternated by session between the two tasks.
• Wc did not include any mixed sessions since an auditory stimulus was used for both tasks.
• On each day, one group was tested on the SR task and the other group was tested on the ID task.
• All animals received extensive training on the two tasks prior to surgeD'.
• After a one-week recover3.' period, they were tested for an additional 26 sessions post-surgery.

4.2.2. Results And Discttssion.

• Lesions (Ivry and Keele, 1989) . and included tests of both time production and time perception.
• Rather than focus on particular tasks (e.g., reflex conditioning, motor control, perceptual processing), it can be useful to consider the operation performed by a structure and then explore how widely that operation is employed.
• In the case of the cerebellum, it appears that its timing capability can help account for why this structure is essential for coordinated movement, speech production, sensorimotor learning, as well as certain perceptual functions (see Ivry, 1993) .
• A timing system that is used for coordinating the actions of multiple joints or perceiving the velocity of a moving stimulus is unlikely to be useful for tasks that span many seconds or minutes.
• Researchers utilizing tasks such as the peak procedure have suggested vet3' different evolutionary pressures for why rats might benefit from being able to determine when 40 see has elapsed.

Full text

Time, Internal Clocks and Movement
M.A. Pastor and J. Artieda (Editors)
9 1996 Elsevier Science B.V. All fights reserved.
257
EXPLORING THE DOMAIN OF THE CEREBELLAR TIMING
SYSTEM
SEAN CLARKE, RICHARD IVRY, JACK GRINBAND, SETH
ROBERTS and NAOMI SHIMIZU
Department of P.wcholo~v. llmvervity qf Cal!/brnia, Berkeley, CA 94720
ABSTRACT. Tile ability of all animal to process temporal information has
adaptive significance across different temporal ranges. The ability to encode and
utilize temporal information allows an animal to predict and anticipate events.
However, the lime scales var3.' widely. The predictable event might be based on
information that changes over relatively long periods such as a year or a day. or
over periods comprising much shorter durations, events that change within a few
minutes or milliseconds. Are there a single set of neural mechanisms thai are
essential for representing temporal information over these different scales?
Despite the facl thai numerous neural sln~clures have been linked to successfid
performance on a varieD' of timing tasks, this question has received relalively
little allen!ion. In this chapler, we will focus on the role of the cerebellum in a
varie .ty of timing tasks. We will review the hypothesis that the cerebellum can be
conceptualized as a relatively task-independent tinting mechanism. An important
feature of this hypothesis is that the range of the cerebellar tinting s.vslem is
assumed to be relatively reslriclcd. Specifically, we assume thai the cerebellum is
capable of representing temporal information ranging from a few milliseconds to
an upper bound of a few seconds. What remains unclear is whether the
cerebellum is involved on tasks spanning longer durations. Cognitive processes
such as attention and memor3.' become clearly imporlant here. and indeed, may
dominate performance for longer inlervals. The animal literature points !o non-
cerebellar stn~ctures as playing a crilical role in these tasks and we will provide a
brief review of this work. Finally. we will present the preliminary resulls from
two experimenls designed to directly test the hypothesis that the cerebeilum's
temporal capabilities are limilcd Io rclalivcly shorl durations.
1. Studies with patients with cerebellar lesions
Ivry and Keele (1989) assessed the performance of a variety of neurological
patients and age-matched control subjects on two tasks that were designed to

258
S. Clarke et aL
require the explicit reprcscntation of temporal information. For the time
production task, the participants produced a series of simple keypresses,
attempting to produce isochronous intervals between each pair of
keypresses. For the time perception task, the participants judged whether a
comparison interval was shorter or longer than a standard interval. There
were three primary groups of patients" those with cerebellar lesions, those
with Parkinson's disease which would indicate basal ganglia pathology, and
those with cortical lesions cncompassing premotor regions. Anatomical
models as well as consideration of the symptoms associated with cercbcllar
lesions prompted the inclusion of the first group. The basal ganglia and
cortical groups were included both for comparison purposes and because of
earlier neuropsychological research implicating basal ganglia (Wing et al.,
1984) or frontal/temporal regions (Milner, 1971) in time production or
perception. In temas of variability on the repetitive tapping task, the patients
with Parkinson's Disease perfonned comparably to age-matched control
subjects. Surprisingly, these null results were obtained under both the on
and off medication state. In contrast, patients with either cortical or
cerebellar lesions were found to have increased variability on the repetitive
tapping task. The total variability was decomposed into two components,
that associated with central control processes and that associated with motor
implementation (Wing and Kristofferson. 1973). From this analysis, the
patients' deficits were attributcd to both sources. However, a second study
focused on patients with unilateral cerebellar lesions, either in medial or
lateral regions. Here, a double dissociation was obtained. Whereas medial
lesions led to increased implementation variability, lateral lesions led to an
increase in central variability (Ivr)., et al., 1988). This dissociation is in
accord with neuroanatomical models which emphasize ascending projections
from the lateral cerebellum and descending projections from the medial
cerebellum. From the tapping results, it is not easy to determine whether the
cerebellum is critical for regulating timing, or some other aspect of motor
performance.
While Wing and Kristoffcrson (1973) labeled the central component.
"clock variability", this componcnt actually includes all sources of
variability not included in the estimate of motor implementation variability
(Ivry and Hazeltine, 1995). For this rcason, the perception task provides an
opportunity to determine whether a particular structure was essential for
internal timing. Correlational studies have suggested that a common
mechanism is invoked in both motor and perceptual timing (Keele et al.,

Exploring the Domain of the Cerebellar
259
1985). From this, we might expect to find that lesions of a particular brain
region will impair performance on both time production and perception
tasks. Only the patients with cerebellar lesions showed this dual-deficit.
They were significantly impaired on the time perception task, requiring a
larger difference between the comparison and standard intervals in order to
achieve a criterion level of pcrfommnce. The perceptual deficit was specific
for time discrimination in that the cerebellar patients were unimpaired on an
intensity discrimination task. Importantly,, the cortical group was normal on
the time perception task, but impaired on the intensity task. Thus, the
perception task provided a second double dissociation suggesting a special
role for the cerebellum in both motor and perceptual tasks that require
precise timing. As with the tapping results, the Parkinson patients
perfomaed within nonnal bounds on the time perception task.
These results required a reconccptualization of the domain of cerebellar
function. This structure has generally been linked to motor functions, or
sensorimotor learning. The fact that the patients were impaired on a purely
perceptual task suggested that its domain should be specified in terms of a
particular mental operation, namely the representation of the temporal
relationships between events. We have hypothesized that this computational
capability is invoked across a wide range of tasks that require this form of
representation. For example. Ivr3' and Diener ( 1991) reported that cerebellar
patients were impaired on a velocity perception task and that this perceptual
problem could not be attributed to a problem in occulomotor control.
Indeed, they proposed that some of the eye movement problems observed
following cerebeilar lesions may reflect an inability to represent the metrical
properties of a moving stimulus.
2. The cerebellum and sensorimotor learning
Impressive progress has been made over the past few decades towards
identi~,ing the neural structures involved with different forms of learning
and memoD'. Given the obvious advantages imposed by learning, it is not
surprising that a large number of neural structures have been implicated in
these processes. An important endeavor has been to specify the domain of
these structures and develop computational models to explain their
contributions.

260
S. Clarke et al.
One approach for understanding the computational requirements of
different learning situations is to consider the temporal properties imposed
by different tasks. For example, the learning process in classical
conditioning is constrained by the temporal relationship between the CS and
US (see Flaherty, 1985; Jenkins, 1984). A prerequisite for learning across a
wide range of paradigms is that the onset of the CS precede the US. The
most effective inter-stimulus interval (ISI), however, varies depending on the
type of learning. Three general categories can be described. 1) Conditioning
of simple skeletal reflexes such as the eyeblink reflex is limited to short ISls.
being strongest when this interval is less than 1 see. 2) Conditioning of
autonomic responses such as heart-rate conditioning can occur with these
short ISis, but can also be robust when the ISI is extended to the minutes
range. 3) Conditioning of avoidance behavior such as in food aversion
experiments can be found when the CS and US are separated by durations
up to many hours. Moreover. whereas the pairing of a CS and US may lead
to multiple CRs, the timing of these learned responses can be quite different.
Conditioning of the rabbit nictitating membrane response (NMR) has
become a model paradigm for investigating the neural substrates of basic
associative learning and memor3., processes associated with simple skeletal
reflex responses. In NMR conditioning, a neutral CS such as a tone or light
is paired with an aversive US (e.g.. an airpuff directed near the eye). After
relatively few presentations, the animal begins to extend the membrane in
response to the CS alone. The rate of NMR learning is highly dependent on
the ISI. Smith (1968) reported that an ISI of 200 ms produced the highest
percentage of CRs in comparison to ISis of 100, 400, and 800 ms (see also,
Steinmetz, 1990). Few CRs were observed with ISis of 50 ms. Conditioned
NMRs can be found with longer intervals, although the rate and efficacy of
learning are reduced. In addition, the topography of the CR is highly
constrained by the ISI. The maximum extension of the nictitating membrane
occurs just prior to the presentation of the US. Indeed, it is this feature that
makes this CR highly adaptive. It pem~its the organism to attenuate the
aversive effects of the US. The importance of timing in NMR conditioning
was made clear by the work of Kehoe et al. (1989). When rabbits were
conditioned simultaneously with two ISls, they produced two CRs, each one
timed to be maximal just prior to the onset of the upcoming US.
Learning related changes during NMR conditioning have been observed
in neurons in several brain regions, including the hippocampus (Berger and
Thompson, 1978) and the cerebellum (McCormick and Thompson, 1984).

Exploring the Domain of the Cerebellar
261
However, lesion studies have provided compelling evidence that the
cerebellum is essential for NMR conditioning. The exact site of plasticity
within the cerebellum has been a source of controversy. Thompson (1986)
has argued that the critical locus is the interpositus nucleus, whereas other
studies have focused on the importance of the cerebellar cortex (e.g., Yeo et
al., 1984). Given that the principal cerebeilar inputs, the mossy and climbing
fibers, im~ervate both sites, it is reasonable to assume that learning-related
changes may occur in both sites (Perret et al., 1993). If this is so, then we
want to consider potential computational differences between nuclear and
cortical learning. One possibility is that nuclear mechanisms might support a
basic associative process for the fonnation of a CR, while changes in the
cerebellar cortex are essential for shaping the topography of the CR. That is,
the precise timing of the CR may result from changes in the cerebcllar
cortex. This hypothesis is supported by the findings that lesions of the
cerebellar cortex disnlpt the timing of the CR (Perrett et ai., 1993).
It remains difficult to specif3.' the learning domain of the cerebellum (see
Ivry, 1993). One possibility is that this structure is essential for forming
sensorimotor associations that result in skeletal responses to avoid aversive
stimuli (Thompson, 1990). This hypothesis emphasizes the task domain of
the cerebellum and focuses on the fact that the climbing fiber pathway
provides a salient error signal for shaping appropriate skeletal responses.
An alternative h)~othesis is that the domain of cerebellar learning extends to
those situations in which the animal must precisely represent temporal
information. By this way of thinking, the cerebellum is associated with NMR
conditioning because this type of learning is only adaptive if it is
appropriately timed (Kecle and lvr3'. 1991). That is, learning an association
and forming the temporal representation of that association can not be
thought of as distinct. Of course, other types of associations may not have
the same temporal requirements and. as such, would not be expected to be
dependent on the cerebellum. For example, the timing of conditioned
autonomic responses in the NMR paradigm seems to be relatively
independent of the ISI and this form of learning is unaffected by cerebcllar
lesions, even when the NM response itself is abolished (Lavond et al., 1984).
Buonomano and Mauk (1994) have presented a computational model of
the cerebellum that produces the associations seen in NMR conditioning as
well as the precise topography of the CR (see also, Bullock et al., 1904).
This model does not depend on delay lines or arrays of oscillators, but rather
emphasizes known anatomical and physiological properties of this neural