The Journal of Neuroscience, May 12, 2004 • 24(19):4491– 4497 • 4491

Behavioral/Systems/Cognitive

Multiple Reference Frames for Motion in the

Primate Cerebellum
Aasef G. Shaikh, Hui Meng, and Dora E. Angelaki
Department of Neurobiology, Washington University School of Medicine, St. Louis, Missouri 63110

Knowledge of body motion through space is necessary for spatial orientation, self-motion perception, and postural control. Yet, sensory
afferent signals may not directly provide such information to the brain. Because motion detected by the vestibular end organs is encoded
in a head-fixed frame of reference, a coordinate transformation is thus required to encode body motion. In this study, we investigated
whether cerebellar motion-sensitive neurons encode the translation of the body through space. We systematically changed both the
direction of motion relative to the body and the static orientation of the head relative to the trunk. The activities of motion-sensitive
neurons in the most medial of the deep cerebellar nuclei, the rostral fastigial nucleus, were compared with those in the brainstem
vestibular nuclei. We found a distributed representation of reference frames for motion in the rostral fastigial nucleus, in contrast to cells
in the vestibular nuclei, which primarily encoded motion in a head-fixed reference frame. This differential representation of motion-
related information implies potential differences in the functional roles of these areas.
Key words: self-motion; cerebellum; frame of reference; sensorimotor; multisensory; gain field; coordinate transformation

Introduction Extensive convergence of vestibular and somatosensory signals

The vestibular system forms an essential component in the sens- has been reported in motion-sensitive areas of the brainstem ves-
ing and generation of movement. Our inertial motion sensors, tibular nuclei (VN), cerebellar cortex (e.g., anterior and posterior
the otolith organs, located in the inner ear cavity, are highly con- cerebellar vermis), and deep cerebellar nuclei (Boyle and Pom-
served throughout evolution and function to provide an accurate peiano, 1981; Anastasopoulos and Mergner, 1982; Wilson et al.,
moment-to-moment estimate of our motion through space. Be- 1990; Manzoni et al., 1998, 1999; Gdowski and McCrea, 1999,
ing fixed in the head, the corresponding signals carried to the 2000; McCrea et al., 1999). At present, it is not known whether
brain by primary vestibular afferents detect self-motion in a this convergence reflects an underlying reference frame transfor-
head-fixed coordinate system. However, a reference frame trans- mation to compute body motion in space. Here, we specifically
formation might take place centrally because day-to-day func- address one of the possible functions of these multisensory inter-
tions often require knowledge of body position, orientation, and actions. Results provide novel evidence that a population of neu-
movement. For example, systematic alterations in vestibulo- rons in the most medial of the deep cerebellar nuclei, the rostral
spinal reflex properties after altered static orientations of head- fastigial nucleus (FN), encode the translation of the body through
on-trunk have been reported (Nashner and Wolfson, 1974; space. In contrast, motion-sensitive cells in the VN seem to main-
Kennedy and Inglis, 2002). In addition, information on body tain a head-fixed reference frame for encoding self-motion.
orientation and movement is also important for the perception of
self-motion and location of objects in extra-personal space Materials and Methods
(Lackner and Graybiel, 1978; Mergner et al., 1991, 1992). Single-unit activities in the rostral FN and rostral medial and lateral VN
were recorded in two juvenile monkeys (Maccaca Fascicularis), which
How does the brain acquire access to an internal estimate of
were chronically implanted with an eye coil and a circular delrin ring for
body motion if this information is not provided by primary ves- head stabilization. A guide tube delrin platform with a staggered array of
tibular afferents? By combining otolith signals, which encode holes (Angelaki and Dickman, 2000; Dickman and Angelaki, 2002) was
translation in head coordinates, with neck proprioceptive infor- stereotaxically secured to the skull inside the head ring. To provide better
mation that signals the position of the head relative to the body, a access to the medial cerebellar nuclei, the platform was slanted in the
coordinate transformation could take place, whereby the brain horizontal plane 10° from anterior to posterior and 10° from left to right.
constructs an internal estimate of body motion through space. All surgical procedures were performed under sterile conditions in ac-
cordance with institutional and National Institutes of Health guidelines.
During experiments, the monkeys were seated in a primate chair with
Received Jan. 11, 2004; revised March 8, 2004; accepted March 30, 2004. their bodies secured with multiple shoulder and lap belts while their
This work was supported by grants from the National Institutes of Health (DC04260, F32 DC006540) and NASA
extremities were tied to the chair. Their head ring was fixed to the primate
(NAG2–1493).
Correspondence should be addressed to Dr. Dora Angelaki, Department of Anatomy and Neurobiology, P.O. Box
chair through a 2.5 inch center bore-bearing piece that could be locked in
8108, Washington University School of Medicine, 660 South Euclid Avenue, St. Louis, MO 63110. E-mail: place at different head-on-trunk orientations. The primate chair was
angelaki@pcg.wustl.edu. then secured inside the inner frame of a vestibular turntable consisting of
DOI:10.1523/JNEUROSCI.0109-04.2004 a three-dimensional rotator on top of a 2 m linear sled (Acutronics,
Copyright © 2004 Society for Neuroscience 0270-6474/04/244491-07$15.00/0 Pittsburgh, PA). The linear acceleration was measured with a three-axis
4492 • J. Neurosci., May 12, 2004 • 24(19):4491– 4497 Shaikh et al. • Reference Frames for Motion in the Cerebellum

accelerometer mounted on the primate chair. Thus, as their trunk and Phase was the difference (in degrees) between peak neural activity and
extremities were softly restrained to the chair, the motion measured by peak linear acceleration.
the accelerometer reflected the motion of the body through space. We Neural response gains and phases were subsequently plotted as a func-
point out that in this presentation, body refers to the single block of torso tion of the direction of translation and fitted (Levenberg–Marquardt
that always faced forward relative to the primate chair. optimization method) with a two-dimensional spatiotemporal model
For each recording session, the eye coil signals and the linear acceler- that represents a more generalized model than cosine tuning (Angelaki
ometer output were low-pass filtered (200 Hz; 6-pole bessel), digitized at and Dickman, 2000). This fitting procedure allowed the estimation of
a rate of 833.33 Hz (model 1401, 16-bit resolution; Cambridge Electron- four parameters: direction, gain, and phase of the maximum response as
ics Design, Cambridge, UK), and stored for off-line analysis. Extracellu- well as the minimum response gain (Angelaki, 1991; Angelaki and Dick-
lar recordings from single FN and VN neurons were obtained with man, 2000). The goodness of fit was based on the variance-accounted-for
epoxy-coated tungsten microelectrodes inserted into 26-gauge guide (VAF), computed as VAF ⫽ {1 ⫺ [var (model ⫺ data)/var (data)]}. The
tubes. Neural activity was amplified, filtered (300 Hz to 6 kHz), and fitting procedure was first applied to each head-on-trunk position sepa-
passed through a BAK Instruments (Germantown, MD) dual-time am- rately (analysis step I). This allowed estimation of all four response pa-
plitude window discriminator. Single-unit spikes triggered acceptance rameters, separately for each of the three head-on-trunk positions. This
pulses (BAK window discriminator) that were stored on a computer first step in analysis was necessary to investigate which, if any, spatial
using the event channel of a Cambridge Electronics Design (model 1401) tuning properties changed as a function of head-on-body position. Once
data acquisition system. In addition, eye movement signals, stimuli, and it was established that the only systematic difference was a spatial shift of
neural activity were stored on a digital audio tape recorder for off-line the tuning curves, data in all three head-on-trunk positions were also
spike discrimination. Stimulus protocols and data acquisition were com- fitted simultaneously by assuming that the maximum response orienta-
puter controlled using scripts written for the Spike2 (Cambridge Elec- tion for each cell was shifted through an angle, ⫾␭ 30°, for the rotated
tronics Design) software environment. head-in-trunk positions (analysis steps II and III). Thus, in a second stage
During initial experiments in each animal, the abducens nuclei were of analysis (step II), all data for each cell were fitted simultaneously with
first identified bilaterally. The recording sites in the FN and VN were then each of two four-parameter models corresponding to the body-fixed
identified on the basis of their stereotaxic location relative to the abdu- reference frame model (␭ ⫽ 0; i.e., no spatial shift of maximum response
cens nuclei and the fourth ventricle. All neural responses were obtained direction for different head-on-trunk positions) or the head-fixed refer-
from vestibular-only (VO) neurons. These neurons were recorded from ence frame model (␭ ⫽ 1; i.e., a ⫾30° systematic shift in the maximum
rostral portions of the medial and lateral VN within 1 mm of areas where response direction for different head-on-trunk positions). Finally, be-
eye movement-sensitive cells were recorded (Angelaki et al., 2001; Dick- cause it is possible that cells exhibit an in-between behavior, a third fitting
man and Angelaki, 2002). In the FN, neurons were recorded in the rostral procedure was used, where ␭ was allowed to vary and was one of the
portions of the nucleus, anterior to all pursuit-related and saccade- parameters fitted (analysis step III). This latter ␭-variable model had a
related cells that we consistently encountered in the caudal FN (Gardner total of five parameters. A sequential F test (Draper and Smith, 1998) was
and Fuchs, 1975; Buttner et al., 1991). To characterize cells as VO or eye used to ascertain the statistical significance of increasing the model pa-
movement-sensitive, a standard protocol was used, including rotations, rameters by comparing the errors associated with the ␭ variable and each
fixations, and smooth pursuit eye movements (Angelaki et al., 2001). of the two ␭-fixed models while allowing for differences in their degrees
VN and FN cells that were sensitive to linear motion also were tested of freedom. The sequential F test was the statistical test used to quantify,
using different directions of translation (0, 30, 60, 90, 120, 150, and 180°; on a cell-by-cell basis, whether the tuning of each neuron was best de-
defined relative to the body) (see Fig. 1), with the head secured at three scribed by the head-fixed/body-fixed or an intermediate reference frame
different positions relative to the trunk (see Fig. 1, monkey schematic model. For both the ␭-fixed and ␭-variable models (analysis steps II and
drawing). Initially, the animal’s head was straight-ahead relative to the III), all other parameters of cell tuning (e.g., maximum and minimum
body (h ⫽ 0°) (see Fig. 1, blue drawing). Subsequently, using the specially response gain and phase) were independent of head-on-trunk position.
constructed head holder, the head was secured 30° to the left relative to
the animal’s body (h ⫽ 30°) (see Fig. 1, green drawing), and the six-
direction translation protocol was repeated. Finally, the motion protocol Results
was also delivered with the animal’s head secured 30° to the right (h ⫽ The head and body coordinate systems were dissociated by sys-
⫺30°) (see Fig. 1, red drawing). The order in which these three head-on- tematically varying both the direction of motion and the static
trunk positions were tested was pseudo-randomized for different neu- orientation of the head relative to the body (Fig. 1). Neural activ-
rons. Such reorientations of the head in relation to the body were in- ities were recorded as animals were translated along different
tended to dissociate between body and head coordinates. Data were directions (0, 30, 60, 90, 120, 150, and 180°; defined relative to
collected during sinusoidal translational motion stimulation either at 0.5 their body) in the horizontal plane with the head fixed to the
or 2 Hz (⫾0.2 g, with g ⫽ 9.81 m/sec2), whichever frequency provided the trunk in one of three different positions: straight-ahead (Fig. 1,
largest cell modulation. Results were similar at the two frequencies and
blue traces) (h ⫽ 0°) as well as rotated 30° to the left (h ⫽ 30°) or
thus are presented together.
Data analyses were performed off-line using custom-written scripts in
right (h ⫽ ⫺30°) (Fig. 1, green and red traces, respectively). Ac-
Matlab (MathWorks, Natick, MA). For each action potential, an occur- cordingly, if a neuron encodes the motion of the body, its spatial
rence time was logged. Instantaneous frequency (IFR) was then calcu- tuning should be independent of the change in head position. In
lated as 1/(interspike interval) and assigned to the middle of the interval. contrast, if a cell exclusively detects the motion of the head inde-
For each stimulus cycle (e.g., n th cycle), an integral (n ⫺ 1) times the pendently of body motion, its preferred movement direction
period was subtracted from the timing for all of the instantaneous fre- should systematically shift to the left or right to reflect the shifted
quency values for that specific cycle. The result was to “fold” all instan- direction of motion in head coordinates (Fig. 1, monkey head
taneous frequency values into a single stimulus cycle. This procedure orientation drawing). Neurons with properties reflecting either
provides no averaging, because all spike occurrences are represented in the body or the head coordinate systems were encountered in the
time. Neural response amplitude and phase were then determined by rostral fastigial nucleus, as illustrated for the two cells in Figure
fitting a sine function to both response and stimulus using a nonlinear
1 A (body coordinates) and Figure 1 B (head coordinates). For the
least squares (Levenberg–Marquardt) minimization algorithm. Portions
of the response cycle in which no spikes occur (silencing or rectification) neuron illustrated in Figure 1 A, the directions of maximum and
does not pose a problem to the fitting procedure, because only positive minimum neural response modulation (0 and 90° motion direc-
IFR values exist in the response cycle and are weighed in the least-squares tions, respectively) were the same for all three head-on-body po-
estimations. Examples of such fits have been illustrated for two cells in sitions. In contrast, for the neuron in Figure 1 B, the directions of
Figure 1. Neural response gains were expressed as spikes/second/gravity. maximum and minimum response modulation shifted for the
Shaikh et al. • Reference Frames for Motion in the Cerebellum J. Neurosci., May 12, 2004 • 24(19):4491– 4497 • 4493

ahead relative to the body (h ⫽ 0°). There

was no systematic dependence of any of
these parameters on static neck inputs, as
illustrated by linear regression (Fig. 3A,
solid lines; dashed lines illustrate 95% con-
fidence intervals). The main effect of
changing the head-on-body position was a
systematic shift in the spatial tuning
curves, including the maximum and min-
imum response directions. Figure 3B illus-
trates the mean spatial tuning curve shift,
plotted separately for the cerebellar and
brainstem populations (gray squares vs
open circles, respectively). Only the VN
data were consistent with the spatial shift
expected from a head-fixed reference
frame coding of motion. The difference in
the spatial shift of the tuning curves for FN
and VN neurons was statistically signifi-
cant (F(1,31) ⫽ 11.4; p ⬍ 0.01).
Because other parameters of cell firing
rate properties remained independent of
changes in the head-on-trunk position
(including the maximum gain of the cells),
the spatiotemporal model was next used to
fit the neural gain and phase dependence
on motion direction for all three head-on-
trunk positions simultaneously. Accord-
ing to the population behaviors in Figure
Figure 1. Instantaneous firing rates of two rostral FN neurons during passive whole-body translation at 0.5 Hz (⫾0.2 g) 3B, the maximum sensitivity direction of
along different directions in the horizontal plane. Data are color coded for the three head-on-body positions; blue, red, and green each cell was assumed to shift through an
colors are used for straight-ahead (h ⫽ 0°) and 30° to the right (h ⫽ ⫺30°) or left (h ⫽ 30°) head-on-trunk positions. angle ⫾␭ 30° for the rotated head-on-
Superimposed solid lines represent best-fit sine functions. A, FN neuron encoding motion in a body reference frame, where the trunk positions. Thus, in a second analysis
firing rate of the cell is independent of head-on-trunk position. B, FN neuron encoding motion in a head reference frame. The firing procedure (step II; see Materials and
rate of the cell changes for the different head-on-trunk positions. In both panels, the minimum responses are marked with Methods), the value of ␭ was fixed to either
asterisks. The black traces represent the linear acceleration stimulus. Motion stimuli are defined relative to the body and, thus, ␭ ⫽ 0 (body-fixed reference frame model;
change direction relative to the head (monkey drawing).
e.g., no changes in the direction of maxi-
mum response gain as a function of
three different head-on-body positions, such that they would re- head-on trunk orientation) or ␭ ⫽ 1 (head-fixed reference frame
main fixed relative to the head. This shift is better seen for the model; i.e., allowing for a ⫾30° shift in the maximum response
minimum response direction, which shifts from 90° (h ⫽ 0°) (Fig. direction as a function of head-on-trunk position). The VAF
1, blue traces) to 120° (h ⫽ 30°) (Fig. 1, green traces) and 60° (h ⫽ values for each of the head-fixed and body-fixed reference frame
⫺30°) (Fig. 1, red traces), shown with asterisks in Figure 1 B. models were compared for each cell in Figure 4 A. In the figure,
To describe these differences, neural response gains and each symbol corresponds to one cell in either the VN (open cir-
phases were plotted as a function of motion direction separately cles) or the FN (gray squares), and the dotted line illustrates the
for each of the three head-on-body positions (Fig. 2). The spatial unity–slope line. For all but one of the VN neurons, the head-
tuning curves for the three head-on-body positions superimpose fixed reference frame model gave as good or higher VAF values
for the neuron shown in Figure 2 A, suggesting that this cell en- than the corresponding body-fixed reference frame model (Fig.
codes the translational motion of the body (i.e., encoding linear 4 A; all but one of the open circles fall in the top left half of the
acceleration in body coordinates). In contrast, the spatial tuning diagram). In contrast, only approximately one-third of the FN
for the neuron in Figure 2 B shifts for the three head-on-body neurons was better fitted with the head-fixed reference frame
positions, indicating that this cell encodes the translational mo- model. The majority of the FN cells better complied with the
tion of the head. To quantify these observations, a spatiotemporal body-fixed reference frame model. However, many neurons were
tuning model was used to fit neural gain and phase as a function equally well fitted by either model, illustrating the need for the
of motion direction (Angelaki, 1991; Angelaki and Dickman, third analysis procedure.
2000). Three different analysis steps were performed. First, the Thus, to further investigate the reference frame used by FN
spatiotemporal model was fitted separately to data for each head- and VN neurons to encode motion, an additional analysis was
on-trunk position (analysis step I). This allowed an independent performed where the value of ␭ was allowed to vary and was
estimation of the gain and phase of neural firing in response to actually one of the parameters fitted for data of each cell (analysis
motion along the maximum response direction of the neuron for step III; see Materials and Methods). This resulted in a model
each head-on-body position. Figure 3A illustrates these values for with one additional parameter. To quantify whether this higher
each of the two rotated head-on-body positions (h ⫽ ⫾30°) plot- parameter model fitted the data better compared with the lower
ted versus the corresponding values when the head was straight- parameter head-fixed or body-fixed model (step II), a sequential
4494 • J. Neurosci., May 12, 2004 • 24(19):4491– 4497 Shaikh et al. • Reference Frames for Motion in the Cerebellum

df ⫽ 14; p ⫽ 0.05), having a median value of 0.4 and 0.9,

respectively.

Discussion
These results demonstrate for the first time that subcortical
motion-sensitive neurons exist that encode translation of the
body, unlike primary vestibular afferents, which detect motion in
a head-fixed reference frame (Cullen and Minor, 2002). Interest-
ingly, the body reference frame was primarily observed in the
deep cerebellar nuclei, where approximately half of the FN cells
encoded motion of the body through space. A similar result was
not seen in brainstem neurons, raising the possibility that the
coordinate transformation might take place in the cerebellar cor-
tex. The rostral fastigial nucleus represents the main output of the
medial zone of the anterior vermis (Voogd, 1989), the role of
which in vestibular–somatosensory interactions has received
strong experimental support (Manzoni et al., 1998, 1999). In fact,
Manzoni et al. (1999) have previously reported that static neck
input modulates the responses of anterior vermis Purkinje cells
during complex vestibular stimulation in decerebrate cats. The
present results show that these vestibular–somatosensory inter-
actions implement a coordinate transformation to estimate mo-
tion of the body through space. The outcome of the reference
frame change is observed in the firing rates of a subpopulation of
FN neurons, an area that represents a main cerebellar outflow to
the spinal cord, premotor brainstem centers, and the thalamo-
cortical system.

Vestibular–somatosensory convergence
The concept of vestibular-proprioceptive interactions was origi-
nally introduced by von Holst and Mittelstaedt (1950). Several
lines of evidence have supported this idea. First and foremost,
control of muscular coordination in relationship to movement is
crucial for maintaining posture. Vestibular-proprioceptive inter-
actions have been reported for limb control (Manzoni et al., 1983;
Ezure and Wilson, 1984; Wilson et al., 1986) and neck reflexes
Figure 2. Neural response gain and phase plotted as a function of the direction of translation
for each of the three head-on-body positions. A, FN neuron encoding motion in a body reference
(Peterson et al., 1985). In addition, sensations of tilting and fall-
frame (same data as in Fig. 1 A). The firing rate of the cell is independent of head-on-trunk ing, as well as altered perception of spatial orientation can be
position. B, FN neuron encoding motion in a head reference frame (same data as in Fig. 1 B). The induced by electrical stimulation or vibration of the neck (Kar-
firing rate of the cell changes for the different head-on-trunk positions. Different symbols are nath et al., 1994; Ivanenko et al., 1999). Additional psychophys-
used for data obtained for the three head-on-trunk positions. The corresponding lines illustrate ical experiments using trunk rotation relative to the head have
the fit of the ␭-variable spatiotemporal tuning model. sp/sec/g, Spikes/second/gravity. demonstrated that normal human subjects derive trunk motion
perception from a combination of vestibular and neck cues
F test was used (Draper and Smith, 1998). The fitted ␭ values are (Mergner et al., 1983, 1991). In addition, vibration of muscles or
summarized in Figure 4 B. For neurons encoding motion in a muscle tendons causes proprioceptive “misinformation” pro-
body reference frame, the spatial shift parameter, ␭, was close to ducing an illusory sensation of movement (Goodwin et al., 1972).
zero. This was the case for many FN but not VN neurons. In 43% In normal subjects, vibration applied to the neck muscles causes a
(15 of 35) of the FN cells, but only one VN cell, the ␭-variable displacement of body orientation, a shift of subjective midline,
model fit was better ( p ⬍ 0.05) than the head-fixed model fit and a consistent error in pointing and body sway (Biguer et al.,
(␭ ⫽ 1) and statistically indistinguishable ( p ⬎ 0.05) from the 1988; Pyykko et al., 1989; Karnath et al., 1994).
body-fixed model fit (␭ ⫽ 0). Therefore, this subpopulation of The interaction between vestibular and joint afferents already
FN cells encodes motion in a body-fixed reference frame. The occurs at the level of the vestibular nuclei, the first central synapse
opposite result, consistent with the head-fixed reference frame of primary vestibular afferents (Boyle and Pompeiano, 1981;
predictions, was encountered in 34% (12 of 35) of the FN cells Anastasopoulos and Mergner, 1982; Wilson et al., 1990). Many
and 80% (14 of 20) of the VN cells. Thus, the majority of VN VN neurons respond to both vestibular and proprioceptive stim-
neurons and a subpopulation of FN cells encode motion in a ulation (Brink et al., 1980; Kasper et al., 1988). Given such exten-
head-fixed reference frame. The remaining neurons (eight FN sive vestibular-proprioceptive convergence in the VN and the
and five VN cells) did not favor either one of the ␭-fixed models. fact that the paramedian zone of the anterior vermis projects
These neurons had 0.4 ⬍ ␭ ⬍ 0.7 and complied with neither the directly to the lateral vestibular nucleus (Voogd, 1989), the ob-
head-fixed nor the body-fixed coordinate system predictions, servation that the VN cell population described here used a mo-
suggesting an intermediate reference frame. The difference in the tion reference frame that was not body-fixed might appear sur-
distributions for the spatial shift parameter, ␭, between the FN prising. Several studies have demonstrated recently that there are
and VN cell populations was statistically significant (␹ 2 ⫽ 23.5; differences in the VN neuron responses during rotations in which
Shaikh et al. • Reference Frames for Motion in the Cerebellum J. Neurosci., May 12, 2004 • 24(19):4491– 4497 • 4495

Figure 3. Parameters of a spatiotemporal tuning model fitted to gain and phase as a function of motion direction separately for each head-on-trunk position (analysis step I). A, The gain and
phase of the cell during motion along the maximum response direction for each of the two rotated head-on-body positions (h ⫽ ⫾30°) plotted versus the corresponding values when the head was
straight ahead relative to the body (h ⫽ 0°). Each pair of open and filled symbols corresponding to the same abscissa illustrates data from a cell for which the spatial tuning was tested in both left
and right head-on-trunk positions. A few cells were only tested either with the left (h ⫽ 30°) or the right head-on-trunk (h ⫽ ⫺30°) positions. Regression equations: y ⫽ 0.2 ⫹ 0.92 x, r 2 ⫽ 0.85
(gain); y ⫽ 0.7 ⫹ 1.01 x, r 2 ⫽ 0.96 (phase). B, The average spatial shift in the tuning curves for FN and VN neurons was plotted as a function of head-on-body position. Data were normalized before
the calculation of the averages by subtracting from the spatial angle of the maximum response direction for h ⫽ ⫾30° the respective angle with the head centered (h ⫽ 0°).

erence frame for translational motion in a

subpopulation of FN neurons. Extensive
vestibular-proprioceptive interactions that
might underlie reference frame transforma-
tions also have been reported in rostral fasti-
gial neurons during rotational motion
(Buttner et al., 2003). The rostral fastigial nu-
cleus, among all deep cerebellar nuclei, is the
one most important in the central processing
of motion-related vestibular information.
Anatomical and physiological studies have
separated the fastigial nucleus into rostral
and caudal subdivisions (Noda et al., 1990;
Buttner et al., 1991). Neurons of the caudal
fastigial nucleus are modulated during
saccadic or smooth pursuit eye move-
ments (Gardner and Fuchs, 1975; Buttner
et al., 1991; Fuchs et al., 1993). In contrast,
the rostral division is populated by neu-
Figure 4. Coordinate frames for FN and VN neurons. A, Comparison between the goodness-of-fit (VAF) for two ␭-fixed spatial rons that are modulated by vestibular
shift models (analysis step II): a body-fixed reference frame model with ␭ ⫽ 0 (abscissa) and a head-fixed reference frame model stimulation but do not show sensitivity to
with ␭ ⫽ 1 (ordinate). Open circles, VN neurons; gray squares, FN neurons. The dotted line illustrates the unity–slope line. B,
eye movements (Gardner and Fuchs, 1975;
Variable spatial shift model (analysis step III): distributions of ␭ values for the VN and FN cell populations (black and gray,
respectively). Vertical gray and black lines illustrate the medians of the two populations, ␭ (median for FN neurons) ⫽ 0.4 and ␭ Buttner et al., 1991; Siebold et al., 1997).
(median for VN neurons) ⫽ 0.9. Previous studies established that rostral
FN neurons respond to sinusoidal rotation
in horizontal and vertical planes (Gardner
the head was fixed relative to the body versus rotations where the and Fuchs, 1975; Siebold et al., 1997, 1999)
head was allowed to rotate freely (Gdowski and McCrea, 1999; as well as to translational movements (Zhou et al., 2001). Given
McCrea et al., 1999; Roy and Cullen, 2001). Yet, the majority of the afferent and efferent projections of rostral FN neurons, it is
VN neurons were closer to head than body coordinates. It is believed that they are involved in vestibulospinal control, includ-
possible that if a clear body reference frame exists in the VN for ing regulation of gait and postural mechanisms (Gardner and
rotational or translational movements, the location of such neu- Fuchs, 1975; Buttner et al., 1991; Siebold et al., 1997).
rons might be in more posterior areas than those explored here. Knowledge of the orientation and motion of the body in space
is important for the control of body, limb, and neck posture as
Role of fastigial neurons in motion detection well as motion perception and spatial localization of objects in
The present experiments, reporting a specific and systematic spa- extra-personal space (Mergner et al., 1991, 1992; Horak et al.,
tial shift in the tuning curves (without gain or phase changes), 1994). All of these functions could be served by projections of FN
represent the first evidence regarding the existence of a body ref- neurons. For example, the main fastigial efferent projections are
4496 • J. Neurosci., May 12, 2004 • 24(19):4491– 4497 Shaikh et al. • Reference Frames for Motion in the Cerebellum

to spinal-related brainstem structures as well as direct projections cleus in saccade generation. I. Neuronal discharge pattern. J Neurophysiol
to the spinal cord, where they directly synapse on spinal mo- 70:1723–1740.
Fukushima K, Peterson BW, Uchino Y, Coulter JD, Wilson VJ (1977) Direct
toneurons (Batton et al., 1977; Fukushima et al., 1977; Wilson et
fastigiospinal fibers in the cat. Brain Res 126:538 –542.
al., 1977). The activities of limb musculature for postural re- Gardner EP, Fuchs AF (1975) Single-unit responses to natural vestibular
sponses during movement have been shown to correlate with stimuli and eye movements in deep cerebellar nuclei of the alert rhesus
body, but not head, displacement as required for postural stabil- monkey. J Neurophysiol 38:627– 649.
ity, suggesting that vestibulospinal reflexes might be organized in Gdowski GT, McCrea RA (1999) Integration of vestibular and head move-
body coordinates (Manzoni et al., 1998). The same study also ment signals in the vestibular nuclei during whole-body rotation. J Neu-
rophysiol 82:436 – 449.
showed that intact activity in the anterior vermis is necessary for
Gdowski GT, McCrea RA (2000) Neck proprioceptive inputs to primate
this function. The spatially shifted reference frames characteriz- vestibular nucleus neurons. Exp Brain Res 135:511–526.
ing the tuning of FN neurons might represent a neural substrate Goodwin GM, McCloskey DI, Matthews PB (1972) A systematic distortion
for this processing. In addition to spinal projections, the FN nu- of position sense produced by muscle fibration. J Physiol (Lond)
cleus, similar to all other cerebellar nuclei, projects to the thala- 221:8P–9P.
mus (Batton et al., 1977; Nakano et al., 1980; Asanuma et al., Horak FB, Shupert CL, Dietz V, Horstmann G (1994) Vestibular and so-
matosensory contributions to responses to head and body displacements
1983). Although the cortical targets of the FN remain unknown,
in stance. Exp Brain Res 100:93–106.
some studies have suggested FN projections to parietal cortex Ivanenko YP, Grasso R, Lacquaniti F (1999) Effect of gaze on postural re-
(Sasaki et al., 1976; Amino et al., 2001) as well as to frontal cortex sponses to neck proprioceptive and vestibular stimulation in humans.
regions and multimodal visual areas (Kyuhou and Kawaguchi, J Physiol (Lond) 519:301–314.
1987). Fastigial-thalamocortical projections could thus mediate Karnath HO, Sievering D, Fetter M (1994) The interactive contribution of
spatial perception functions in which knowledge of body motion neck muscle proprioception and vestibular stimulation to subjective
straight-ahead orientation in man. Exp Brain Res 101:140 –146.
in space is essential.
Kasper J, Schor RH, Wilson VJ (1988) Response of vestibular neurons to
head rotations in vertical planes. II. Response to neck stimulation and
References vestibular-neck interaction J Neurophysiol 60:1765–1778.
Kennedy PM, Inglis JT (2002) Interaction effects of galvanic vestibular stim-
Amino Y, Kyuhou S, Matsuzaki R, Gemba H (2001) Cerebello-thalamo-
ulation and head position on the soleus H reflex in humans. Clin Neuro-
cortical projections to the posterior parietal cortex in the macaque mon-
physiol 113:1709 –1714.
key. Neurosci Lett 309:29 –32.
Kyuhou S, Kawaguchi S (1987) Cerebellocerebral projection from the fasti-
Anastasopoulos D, Mergner T (1982) Canal-neck interaction in vestibular
gial nucleus onto the frontal eye field and anterior ectosylvian visual area
nuclear neurons of the cat. Exp Brain Res 46:269 –280.
in the cat. J Comp Neurol 259:571–590.
Angelaki DE (1991) Dynamic polarization vector of spatially tuned neu-
Lackner JR, Graybiel A (1978) Some influences of touch and pressure cues
rons. IEEE Trans Biomed Eng 38:1053–1060.
on human spatial orientation. Aviat Space Environ Med 49:798 – 804.
Angelaki DE, Dickman JD (2000) Spatiotemporal processing of linear accel-
Manzoni D, Pompeiano O, Srivastava UC, Stampacchia G (1983) Re-
eration: primary afferent and central vestibular neuron responses. J Neu-
sponses of forelimb extensors to sinusoidal stimulation of macular laby-
rophysiol 84:2113–2132.
rinth and neck receptors. Archs Ital Biol 121:205–214.
Angelaki DE, Green AM, Dickman JD (2001) Differential sensorimotor
Manzoni D, Pompeiano O, Andre P (1998) Convergence of directional ves-
processing of vestibulo-ocular signals during rotation and translation.
tibular and neck signals on cerebellar Purkinje cells. Pflügers Arch
J Neurosci 21:3968 –3985.
435:617– 630.
Asanuma C, Thach WT, Jones EG (1983) Distribution of cerebellar termi- Manzoni D, Pompeiano O, Bruschini L, Andre P (1999) Neck input modi-
nations and their relation to other afferent terminations in the ventral fies the reference frame for coding labyrinthine signals in the cerebellar
lateral thalamic region of the monkey. Brain Res 286:237–265. vermis: a cellular analysis. Neuroscience 93:1095–1107.
Batton III RR, Jayaraman A, Ruggiero D, Carpenter MB (1977) Fastigial McCrea RA, Gdowski GT, Boyle R, Belton T (1999) Firing behavior of ves-
efferent projections in the monkey: an autoradiographic study. J Comp tibular neurons during active and passive head movements: vestibule-
Neurol 174:281–305. spinal and other non-eye-movement related neurons. J Neurophysiol
Biguer B, Donaldson IM, Hein A, Jeannerod M (1988) Neck muscle vibra- 82:416 – 428.
tion modifies the representation of visual motion and direction in man. Mergner T, Nardi GL, Becker W, Deecke L (1983) The role of canal-neck
Brain 111:1405–1424. interaction for the perception of horizontal trunk and head rotation. Exp
Boyle R, Pompeiano O (1981) Responses of vestibulospinal neurons to neck Brain Res 49:198 –208.
and macular vestibular inputs in the presence or absence of the paleocer- Mergner T, Siebold C, Schweigart G, Becker W (1991) Human perception
ebellum. Ann NY Acad Sci 374:373–394. of horizontal head and trunk rotation in space during vestibular and neck
Brink EE, Hirai N, Wilson VJ (1980) Influence of neck afferents on vestib- stimulatin. Exp Brain Res 85:389 – 404.
ulospinal neurons. Exp Brain Res 38:285–292. Mergner T, Rottler G, Kimmig H, Becker W (1992) Role of vestibular and
Buttner U, Fuchs AF, Markert-Schwab G, Buckmaster P (1991) Fastigial neck inputs for the perception of subject motion in space. Exp Brain Res
nucleus activity in the alert monkey during slow eye and head movements. 89:655– 668.
J Neurophysiol 65:1360 –1371. Nakano K, Takimoto T, Kayahara T, Takeuchi Y, Kobayashi Y (1980) Dis-
Buttner U, Glasauer S, Glonti L, Guan Y, Kipiani E, Kleine J, Siebold C, tribution of cerebellothalamic neurons projecting to the ventral nuclei of
Tchelidze T, Wilden A (2003) Multimodal signal integration in vestib- the thalamus: an HRP study in the cat. J Comp Neurol 194:427– 439.
ular neurons of the primate fastigial nucleus. Ann NY Acad Sci Nashner LM, Wolfson P (1974) Influence of head position and propriocep-
1004:241–251. tive cues on short latency postural reflexes evoked by galvanic stimulation
Cullen KE, Minor LB (2002) Semicircular canal afferents similarly encode of the human labyrinth. Brain Res 67:255–268.
active and passive head-on-body rotations: implications for the role of Noda H, Sugita S, Ikeda Y (1990) Afferent and efferent connections of the
vestibular efference. J Neurosci 22:RC226. oculomotor region of the fastigial nucleus in the macaque monkey.
Dickman JD, Angelaki DE (2002) Vestibular convergence patterns in ves- J Comp Neurol 302:330 –348.
tibular nuclei neurons of alert primates. J Neurophysiol 88:3518 –3533. Peterson BW, Goldberg J, Bilotto G, Fuller JH (1985) Cervicocollic reflex: its
Draper NR, Smith H (1998) Applied regression analysis. New York: Wiley. dynamic properties and interaction with vestibular reflexes. J Neuro-
Ezure K, Wilson VJ (1984) Interaction of tonic neck and vestibular reflexes physiol 54:90 –109.
in the forelimb of the decerebrate cat. Exp Brain Res 54:289 –292. Pyykko I, Aalto H, Seidel H, Starck J (1989) Hierarchy of different muscles
Fuchs AF, Robinson FR, Straube A (1993) Role of the caudal fastigial nu- in postural control. Acta Otolaryngol [Suppl] 468:175–180.
Shaikh et al. • Reference Frames for Motion in the Cerebellum J. Neurosci., May 12, 2004 • 24(19):4491– 4497 • 4497

Roy JE, Cullen KE (2001) Selective processing of vestibular reafference dur- Voogd J (1989) Parasagittal zones and compartments of the anterior vermis
ing self-generated head motion. J Neurosci 21:2131–2142. of the cat cerebellum. Exp Brain Res 17:3–19.
Sasaki K, Kawaguchi S, Oka H, Sakai M, Mizuno N (1976) Electrophysio- Wilson VJ, Uchino Y, Susswein A, Fukushima K (1977) Properties of direct
logical studies on the cerebellocerebral projections in monkeys. Exp Brain fastigiospinal fibers in the cat. Brain Res 126:543–546.
Res 24:495–507. Wilson VJ, Schor RH, Suzuki I, Park BR (1986) Spatial organization of neck
Siebold C, Glonti L, Glasauer S, Buttner U (1997) Rostral fastigial nucleus and vestibular reflexes acting on the forelimbs of the decerebrate cat.
activity in the alert monkey during three dimensional passive head move- J Neurophysiol 55:514 –526.
ments. J Neurophysiol 77:1432–1446. Wilson VJ, Yamagata Y, Yates BJ, Schor RH, Nonaka S (1990) Response of
Siebold C, Kleine JF, Glonti L, Tchelidze T, Buttner U (1999) Fastigial nu- vestibular neurons to head rotations in vertical planes. III. Response of
cleus activity during different frequencies and orientations of vertical vestibulocollic neurons to vestibular and neck stimulation. J Neuro-
vestibular stimulation in the monkey. J Neurophysiol 82:34 – 41. physiol 64:1695–1703.
von Holst E, Mittelstaedt H (1950) Das reafferenzprinzip. Naturwissen- Zhou W, Tang BF, King WM (2001) Responses of rostral fastigial neurons
schaften 37:464 – 476. to linear acceleration in an alert monkey. Exp Brain Res 139:111–115.