Processing of visual signals related to self-motion in the cerebellum of pigeons

In this paper I describe the key features of optic flow processing in pigeons. Optic flow is the visual motion that occurs across the entire retina as a result of self-motion and is processed by subcortical visual pathways that project to the cerebellum. These pathways originate in two retinal-recipient nuclei, the nucleus of the basal optic root (nBOR) and the nucleus lentiformis mesencephali, which project to the vestibulocerebellum (VbC) (folia IXcd and X), directly as mossy fibers, and indirectly as climbing fibers from the inferior olive. Optic flow information is integrated with vestibular input in the VbC. There is a clear separation of function in the VbC: Purkinje cells in the flocculus process optic flow resulting from self-rotation, whereas Purkinje cells in the uvula/nodulus process optic flow resulting from self-translation. Furthermore, Purkinje cells with particular optic flow preferences are organized topographically into parasagittal “zones.” These zones are correlated with expression of the isoenzyme aldolase C, also known as zebrin II (ZII). ZII expression is heterogeneous such that there are parasagittal stripes of Purkinje cells that have high expression (ZII+) alternating with stripes of Purkinje cells with low expression (ZII−). A functional zone spans a ZII± stripe pair. That is, each zone that contains Purkinje cells responsive to a particular pattern of optic flow is subdivided into a strip containing ZII+ Purkinje cells and a strip containing ZII− Purkinje cells. Additionally, there is optic flow input to folia VI–VIII of the cerebellum from lentiformis mesencephali. These folia also receive visual input from the tectofugal system via pontine nuclei. As the tectofugal system is involved in the analysis of local motion, there is integration of optic flow and local motion information in VI–VIII. This part of the cerebellum may be important for moving through a cluttered environment.


INTRODUCTION
As an observer moves through an environment consisting of numerous objects and surfaces, visual motion occurs across the entire retina. This is known as "optic flow" (Gibson, 1954). The processing of optic flow is important for numerous behaviors and processes including perception of self-motion, the control of posture and locomotion, and navigation (Waespe and Henn, 1987;Srinivasan et al., 1996;Warren et al., 2001;Kearns et al., 2002). Although research has shown that the cortical area MST is important for the analysis of optic flow in primates (e.g., Duffy and Wurtz, 1991; for review see Duffy, 2004), there is a much older literature showing that the terminal nuclei of the accessory optic system (AOS) and the nucleus of the optic tract in the pretectum process optic flow (Simpson and Alley, 1974;Collewijn, 1975; for reviews see Simpson, 1984;Gamlin, 2005;Giolli et al., 2005). The AOS and pretectum are found in all vertebrate classes (Fite, 1985;McKenna and Wallman, 1985;Weber, 1985) and are highly conserved with respect to physiological response properties and neuroanatomical connections (Ibbotson and Price, 2001;Voogd and Wylie, 2004).
In birds, optic flow analysis begins with two retinal recipient nuclei: the nucleus of the basal optic root (nBOR; homologous to the terminal nuclei in mammals) of the AOS, and the pretectal nucleus lentiformis mesencephali (LM; homologous to the nucleus of the optic tract). Retinal input to nBOR arises from a distinct subset of ganglion cells; "displaced" ganglion cells, so called because they are found in the inner plexiform layer rather than the ganglion cell layer Reiner et al., 1979;Fite et al., 1981). The connections of LM and nBOR are extensive and include structures involved in axial motor control, oculomotor control, and nuclei in other visual pathways (Clarke, 1977;Brecha et al., 1980;Gamlin and Cohen, 1988;Wild, 1989;Wylie et al., 1997Wylie et al., , 1998b. This review will focus on my work describing how optic flow information is processed en route to, and within, the cerebellum of pigeons although I note similarities and differences with other species. There are several reasons why pigeons are the subjects of this research. In addition to practical considerations such as expense, availability, and manageability, pigeons are especially useful for studying optic flow processing in the cerebellum for several reasons. The avian and mammalian visual pathways are very similar with respect to anatomical and functional organization (Karten and Shimizu, 1991;Nguyen et al., 2004). This is particularly true for the visual-cerebellar pathways involved in processing optic (Voogd and Wylie, 2004; see below). Furthermore, birds in general have a highly developed cerebellum (Larsell, 1967) which is easily accessible for electrophysiological and anatomical study. Finally, pigeons are a diurnal species, and as creatures of flight, the analysis of optic flow is critical to their survival.

MOTION PROCESSING IN THE NUCLEUS OF THE BASAL OPTIC ROOT AND LENTIFORMIS MESENCEPHALI
As self-motion causes visual motion across the entire retina, one would expect a system that analyzes this optic flow would respond to motion over large parts of the retina. Indeed, LM (Figure 2A) and nBOR ( Figure 2B) neurons have large contralateral receptive fields ( Figure 2C) averaging 60 • in diameter with the largest encompassing the entire monocular visual field. These neurons are directionally selective in response to large stimuli, such as random dot patterns, checkerboards, and gratings ( Figure 2D) (Burns and Wallman, 1981;Morgan and Frost, 1981;Gioanni et al., 1984). A tuning curve for a nBOR neuron is shown in Figure 2E (Wylie and Frost, 1990a). Although broadly tuned, the neuron shows a maximal response to upward motion (preferred direction) and is inhibited by downward motion (anti-preferred direction). Neurons in nBOR and LM show a complementary pattern of direction selectivity. In LM, most (>50%) neurons prefer forward (i.e., temporal-to-nasal) motion ( Figure 2F) (Winterson and Brauth, 1985;Wylie and Frost, 1996;Wylie and Crowder, 2000). In contrast, neurons preferring upward, downward and backward (i.e., nasal-to-temporal) motion are about equally abundant in nBOR, but fewer (5-10%) prefer forward motion ( Figure 2G) (Gioanni et al., 1984;Wylie and Frost, 1990a;Crowder et al., 2003).

DISTINGUISHING SELF-TRANSLATION AND SELF-ROTATION IN THE VESTIBULOCEREBELLUM (VBC)
The motion of any object through 3-dimensional space can be described with reference to its translation between two points, and its rotation about an intrinsic axis. This can also be applied to self-motion of an organism, and vertebrates do have mechanisms to detect both self-translation and self-rotation. The vestibular system consists of the semicircular canals, which detect head rotation, and the otolith organs, which detect head acceleration resulting from gravity and self-translation (Wilson and Melvill Jones, 1979). A neural system involved in analyzing optic flow can also encode self-translation and self-rotation. The patterns of optic flow resulting from self-translation and self-rotation are quite different. Figures 3A and B show, respectively, the patterns of optic flow resulting from translation along, and rotation about, the z-axis. These are shown as projected onto imaginary spheres surrounding the animal, where the arrows indicate local motion within the flowfield (Gibson, 1954). Assuming no eye movements, during self-translation there is a focus of expansion in the direction of self-motion, and backward motion along the equator of this sphere in both visual fields ( Figure 3A). Not visible in the figure, there would also be a focus of contraction behind the animal's head. For self-rotation about the z-axis, there is circular motion about the axis of rotation, but along the equator of FIGURE 1 | A reduced schematic, showing the visual pathways (A) to the cerebellum (B) in birds. The cerebellum is divided into folia, numbered I-X from anterior to posterior (Larsell, 1967). Folia IXcd and X comprise the vestibulocerebellum, which receives optic flow input from the nucleus of the basal optic root (nBOR) and lentiformis mesencephali (LM) via a climbing fiber pathway through the medial column of the inferior olive (mcIO) (blue pathway) (Arends and Voogd, 1989;Lau et al., 1998). The LM and nBOR also project directly to folium IXcd as mossy fibers (green pathways) (Brauth and Karten, 1977;Clarke, 1977;Brecha et al., 1980;Wylie and Linkenhoker, 1996). The LM also projects to folia VI-VIII (red pathway), which are part of the oculomotor cerebellum (Voogd and Barmack, 2006). These folia also receive visual motion signals from a tecto-pontine system (red pathway) (Reiner and Karten, 1982). See text for additional details. LMm, i, l: the medial, intercalatus, and lateral divisions of LM. nBORd, p: the dorsalis and proper divisions of nBOR. VTA: ventral tegmental area. Indicates that most nBOR and LM neurons have large receptive fields in the contralateral visual field and exhibit directional tuning in response to largefield motion (e.g., Burns and Wallman, 1981;Winterson and Brauth, 1985). (D) Shows the response of a nBOR neuron to upward (excitation) and downward (inhibition) motion of a large drifting sine-wave grating (from Crowder and Wylie, 2001). (E) Shows a directional tuning curve of a typical nBOR neuron. Firing rate (spikes/s) is plotted as a function of the direction of motion in polar coordinates, and the gray circle represents the neuron's spontaneous firing rate. The directions are indicated as follows: U, upward; D, downward; F, forward or temporal-to-nasal (T-N), and B, backward or nasal-to-temporal (N-T). (F) Shows a distribution of the direction preferences of LM neurons in pigeons: most prefer forward motion (adapted from Wylie and Crowder, 2000). (G) Shows a distribution of the direction preferences of nBOR neurons in pigeons: most prefer upward, downward, or backward motion (Crowder et al., 2003).
this sphere there is upward and downward motion in the right and left visual fields respectively. Although the neurons in LM and nBOR have large receptive fields for analyzing optic flow, they cannot distinguish optic flow patterns resulting from selfrotation and self-translation. For example, a neuron preferring upward motion, such as that depicted in Figures 2C-E, would respond equally well to downward-translation and a rightward roll of the head. For a predominantly lateral-eyed animal such as a pigeon, a simple solution is to integrate information from the ipsi-and contralateral visual fields. This is what occurs in the olivo-vestibulocerebellar pathway shown in blue in Figure 1.
In Figures 3C and D, examples are shown from the VbC on the left side of the brain, where directional tuning to largefield moving stimuli was measured for both the ipsilateral and contralateral eyes. The neuron in Figure 3C responded best to backward (nasal-to-temporal) motion in both eyes, which would result from forward self-translation. The neuron in Figure 3D responded best to upward motion in the ipsilateral eye, and downward motion in the contralateral eye, which would result from a rightward rotation about the z-axis (roll). Although there are a few neurons in nBOR, LM, and the ventral tegmental area (VTA) that have such binocular receptive fields that respond to particular patterns of optic flow resulting from self-translation and self-rotation Frost, 1990b, 1999b;Wylie, 2000), almost all neurons in mcIO and the VbC have panoramic receptive fields (Wylie and Frost, 1991, 1999aWinship and Wylie, 2001). Moreover there is a clear topographic organization of neurons responsive to translational and rotational optic flow (Winship and Wylie, 2001;Pakan et al., 2005;Graham and Wylie, 2012). The pathway from the nBOR and LM to the VbC is as follows. The mcIO receives a projection from the ipsilateral LM (Clarke, 1977). This projection is mainly directed to the caudal mcIO (Wylie, 2001;Pakan et al., 2010) and arises from a distinct group of medium-sized fusiform cells found in a thin strip along the border of the medial and lateral subnuclei of LM (LMm, LMl) (Gamlin and Cohen, 1988;Wylie, 2001;. We have referred to this region as the intercalated nucleus of LM (LMi). The mcIO also receives a bilateral input from nBOR. It is directed mainly to the rostral mcIO (Wylie, 2001;Pakan et al., 2010) and arises from small neurons in nBOR dorsalis (nBORd) and the adjacent VTA (Brecha et al., 1980;Wylie, 2001;

C) and (D)
Show the directional tuning curves in response to large-field stimulation of the ipsi-and contralateral eyes for Purkinje cells in the vestibulocerebellum. The arrows represent the peak best fit sine wave to the tuning curve, and serves as a proxy for the preferred direction.
The cell in (C) preferred backward (b) [i.e., nasal-to-temporal (n-t)] motion in both eyes, which would result from forward self-translation (adapted from Graham and Wylie, 2012). The cell in (D) preferred upward (u) motion in the ipsilateral eye, and downward motion in the contralateral eye, which results from rotation of the head about the z-axis (roll) (adapted from Wylie and Frost, 1991). The gray circles represent the spontaneous firing rates of these neurons, which is typically about 1 spikes/s. d: downward motion; f: forward motion [i.e., temporal-to-nasal (t-n)]. . The mcIO projects to the VbC (folia IXcd and X) as climbing fibers. This projection is topographic such that the medial mcIO projects to the lateral VbC and the lateral mcIO projects to the medial VbC (Arends and Voogd, 1989;Lau et al., 1998;Wylie et al., 1999;Crowder et al., 2000). The lateral VbC is known as the flocculus, whereas the medial VbC is the uvula/nodulus.

ENCODING OF ROTATIONAL OPTIC FLOW
Generally, one describes the rotation of an object in space with reference to its component rotations about three orthogonal axes: roll (z), pitch (x) and yaw (y). As outlined in this section, a three-axis reference frame underlies the analysis of rotational optic flow. These axes are orthogonal, but they are not roll, pitch, and yaw.
To provide an effective stimulus for neurons responsive to rotational optic flow in the flocculus of rabbits, Jerry Simpson and Werner Graf designed a planetarium projector, which projected a flowfield onto the floor, walls and ceiling of the room (Simpson et al., 1981. The projector was suspended in gimbals, such that axis of rotation could be aligned to any orientation within 3-dimensional space. We used a similar device, depicted in Figure 4A, to stimulate the complex spike activity of Purkinje cells in the pigeon flocculus. Our findings  were essentially identical to those of Graf et al. (1988). In the flocculus, there are two types of neurons: one prefers rotational optic flow about the vertical (y) axis (VA neurons) and the other prefers rotational optic flow about an horizontal axis oriented 45 • to the midline (HA neurons). Figure 4B shows the responses of a VA neuron in the left flocculus to rotational optic flow about four axes in the sagittal (yz) plane. Each peri-stimulus time histogram (PSTH) is summed from 10 sweeps, where each sweep consisted of 5 s of rotation in one direction followed by 5 s of rotation in the opposite direction. An elevation tuning curve is also shown, where the firing rate (solid red line) is plotted as a function of the axis of rotation. The broken circle represents the spontaneous rate, and the broken red line indicates the preferred axis as calculated from the best fit sine wave. The direction of each curved arrow represents the direction of head motion that would cause the presented flowfield. Thus, the cell responds best to leftward rotation of the head about the vertical (y) axis. The flowfield that maximally stimulates VA neurons in shown in Figure 4C, and the best axes of several VA neurons are shown in Figure 4D. The largest arrow with the broken shaft represents the mean of the distribution. Figures 4E-G shows axis tuning for an HA neuron. Figure 4F shows the azimuth tuning curve plotting the responses to rotation about axes in the horizontal (xz) plane, whereas Figure 4G shows the azimuth tuning curve in a vertical plane that intersects the horizontal plane through 45 • contralateral azimuth (45 • c) for the same neurons. This vertical plane is depicted in Figure 4E, and the axes numbered 1-4 in Figure 4E correspond to those in Figure 4G. This neuron responded best to Considering both sides of the brain, it consists of three orthogonal axes: the vertical (y) axis and two horizontal axes oriented 45 • to the midline. All responses in this and subsequent figures refer to recording from neurons in the VbC on the left side of the brain. These data are from . See text for additional details.
rotation about an horizontal axis oriented at 45 • c/135 • i azimuth. The flowfield that maximally stimulates HA neurons in shown in Figure 4H, and the best axes of several HA neurons are shown in Figure 4I. Figure 4J shows the reference frame of rotational optic flow neurons considering neurons on both sides of the brain. It consists of three orthogonal axis: the vertical axis, and two horizontal axes oriented 45 • to the midline. Simpson, Graf and colleagues noted that this is the same reference frame as the vestibular semicircular canals, and the eye muscles (Simpson et al., 1979(Simpson et al., , 1981Ezure and Graf, 1984;Simpson and Graf, 1985;Graf et al., 1988; see also Wylie and Frost, 1996). The horizontal canals are maximally responsive to rotation about the vertical axis, whereas one anterior canal (and the contralateral coplanar posterior canal) responds best to rotation about a horizontal axis oriented 45 • to the midline. With respect to the eye muscles, the horizontal recti rotate the eyes about the vertical axis, whereas the vertical recti and oblique muscles rotate the eyes about an horizontal axis oriented at 45 • to the midline. Thus, the sensory systems involved in analysis of self-rotation (vestibular and optic flow) and the output of this system (i.e., the eye muscles which generate compensatory rotary eye movements) all share the same spatial reference frame.

ENCODING OF TRANSLATIONAL OPTIC FLOW
To simulate translational optic flow we designed the device shown in Figure 5A, which projected panoramic translational optic flow on to the walls, ceiling, and floor of the room, and we recorded from Purkinje cells in the uvula/nodulus in pigeons (Wylie et al., 1998a;Wylie and Frost, 1999a). There are four types of optic flow neurons in the uvula/nodulus: Contraction, Expansion, Ascent, and Descent. Figure 5B shows the responses of a Contraction neuron in the left uvula/nodulus in response to translational optic flow along several axes. PSTHs show the responses to translational optic flow along 4 axes in the horizontal (xz) plane. Each PSTH is summed from 20 sweeps, where each sweep consisted of 5 s of translation in one direction followed by a 5 s pause, then 5 s of motion in the opposite direction followed by a 5 s pause. An azimuth tuning curve (xz plane) is shown as well as an elevation tuning curve in a vertical plane that intersects the horizontal place at 45 • c azimuth. The direction of each arrow represents the direction of head motion that would cause the presented flowfield. This cell responds best to backward translation along an horizontal axis oriented at 45 • c/135 • i azimuth. This results in a flowfield with a focus of contraction at 45 • c azimuth. Figures 5C-E shows tuning curves for the other three types of translation neurons in the VbC: Descent, Ascent and Expansion. The flowfields that maximally stimulate each of the four types are shown in Figure 5F,  Wylie and Frost (1999b). See text for additional details.
Frontiers in Behavioral Neuroscience www.frontiersin.org February 2013 | Volume 7 | Article 4 | 6 and the best axes of translation for the four groups are shown in Figure 5G. Figure 5H shows the common reference frame for translational and rotational optic flow responses in the VbC. Considering both sides of the brain, the reference frame consists of three orthogonal axes: the vertical (y) axis and two horizontal axes oriented 45 • to the midline. Note that this is the same reference frame as that of the rotational optic flow system in the flocculus. We have previously argued how this reference frame is optimal on several accounts (see Frost and Wylie, 2000). Although the processing of rotational optic flow in the flocculus is essentially identical in pigeons and rabbits (see previous section), the same cannot be said for the uvula/nodulus. In the uvula/nodulus of rabbits, VA and HA neurons are found Wylie et al., 1994) in addition to some Purkinje cells where the complex spike activity is modulated by vestibular stimulation originating in the otolith organs and vestibular canals (Barmack and Shojaku, 1992). Purkinje cell complex spike activity responsive to translational optic flow has yet to be observed in any species but the pigeon. However, Yakusheva et al. (2008) showed that simple spike activity of Purkinje cells in the uvula/nodulus in monkeys responds to self-translation. Thus is seems that the uvula/nodulus in mammals may be involved in processing both self-translation and self-rotation. Figure 6A depicts the flowfield that would result from a rightward rotation about the roll axis. To construct a receptive field sensitive to this flowfield, one could pool information from local motion detectors with predictably varying direction preferences: leftward/downward at S 1 , downward at S 2 , upward at S 3 , etc. This is not the case for the optic flow cells in the VbC. Rather they have a receptive field structure that provides a crude approximation to the preferred optic flow pattern by pooling information from two motion detectors with opposing direction preferences as illustrated in Figure 6B. Such a "bipartite" receptive field was suggested by Simpson et al. (1979Simpson et al. ( , 1981Simpson et al. ( , 1988 for the HA neurons in the rabbit flocculus.  showed that the bipartitie receptive field type of arrangement underlies the receptive field structure for neurons in the pigeon flocculus and uvula/nodulus. Figure 6C shows some of the critical data for an HA neuron in the pigeon flocculus. The cell was stimulated with the two composite stimuli depicted. We predicted that if the receptive field was precisely tuned to rotation (as in Figure 6A), the cell would modulate equally to the "horizontal shear" and "vertical shear" conditions as an equal number of motion detectors would be excited by both stimulus configurations. However, the cell showed maximal modulation to the vertical shear configuration and no modulation to the horizontal shear condition, indicating the underlying receptive field is bipartite as indicated in Figure 6B. Data for all (n = 22) flocculus HA neurons are shown in Figure 6D. Here the normalized depth of modulation is shown in response to the vertical and horizontal shear stimuli, as well as true rotation. Note that the cells showed little modulation to the horizontal shear, and more to the vertical shear compared to rotation. Again, these data support the idea of a bipartite receptive field organization.

FIGURE 6 | (A) and (B),
Respectively, show receptive fields either "precisely" tuned for rotational optic flow by pooling many local motion detectors with different direction preferences, or "approximately" tuned with a bipartite receptive field structure. (C) Shows the responses of an HA neuron to composite stimuli. The neuron responded much better to vertical shear as opposed to horizontal shear, indicating that it has a bipartite receptive field structure shown in (B) . (D) Shows the normalized depth of modulation [(CCW−CW)/(CCW+CW)] for all rotation units (n = 22) in response to the three stimulus configurations illustrated (mean ± s.e.m). Note that the cells responded better to the vertical shear as opposed to the true rotation . (E) Shows the responses of a rotation sensitive neuron in monkey area MST to similar stimuli. It responded equally well to vertical and horizontal shear, indicating that it has a precisely tuned receptive field (from . (F) Shows the responses of an optic flow neuron in the lobula plate in blowflies to local stimulation. These neurons have an underlying receptive field with precise tuning (from Krapp et al., 1998). Optic flow neurons sensitive to translational and rotational patterns are also found in the primate cortical area MST (Duffy and Wurtz, 1991) and in the lobula plate in blowflies (Krapp and Hengstenberg, 1996), and these neurons have an underlying receptive field with precise tuning. Figure 6E shows data from a MST neuron that preferred clockwise (CW) optic flow. Note that it responded equally well to vertical and horizontal shear .  have used other composite stimuli to stimulate MST neurons and shown that the closer the stimulus matches the preferred flowfield, the greater the response of the neuron. Figure 6F shows data from a neuron in the blowfly visual system that preferred rotational optic flow (Krapp et al., 1998). The local direction preferences within the flowfield are shown by the arrows. Thus, optic flow neurons in the MST and blowfly visual system, unlike VbC neurons, do pool information from several local motion detectors with predictable differences in direction preference to create a receptive field with precise tuning.

VESTIBULAR INPUT TO THE VESTIBULOCEREBELLUM
Given that the flocculus and uvula/nodulus are involved in the processing of optic flow resulting from self-rotation and selftranslation, respectively, one might expect that the flocculus would be associated with vestibular input from the semi-circular canals, whereas the uvula/nodulus would be associated with input from the otolith organs. By examining the input from the vestibular nuclei to the VbC, we have shown that this is generally the case . Shown in Figure 7, we injected retrograde tracers in the flocculus and uvula/nodulus and analyzed the distribution of retrogradely labeled cells in the vestibular nuclei ( Figure 7A) and compared this to descriptions of the primary vestibular afferents from the canals and otolith organs to the vestibular nuclei ( Figure 7B) (Schwarz and Schwarz, 1983;Dickman and Fang, 1996). Although hardly absolute, the regions that project to the flocculus tend to receive input from the semicircular canals, whereas the regions that project to the uvula/nodulus receive input from the otolith organs. For example, in both the descending vestibular nucleus (VeD) and superior vestibular nucleus (VeS), the lateral portion receives input primarily from the otolith organs (blue in Figure 7B) and projects primarily to the uvula-nodulus (blue in Figure 7A), whereas the medial portions of VeD and VeS receive input primarily from the vestibular canals (yellow in Figure 7B) and project primarily to the flocculus (yellow in Figure 7A).

ZONAL ORGANIZATION OF THE VESTIBULOCEREBELLUM
The functional units of the cerebellum are series of "zones" that lie in the sagittal plane, perpendicular to the axes of the folia. This organization is revealed in several aspects: afferents to the cerebellar cortex terminate in parasagittal bands (Voogd and Bigare, 1980;Wu et al., 1999;Ruigrok, 2003), and Purkinje cells within a sagittal band show similar response properties (Andersson and Oscarsson, 1978). As outlined by Simpson (2011), the flocculus is no exception in this regard, and this has been extensively studied in rabbits. Based on converging evidence examining the inferior olivary inputs to the flocculus, the projections of flocculus to the vestibular nuclei, eye movements elicited by stimulation of the FIGURE 7 | (A) Shows the areas of the vestibular nuclei that project to the flocculus (yellow) and uvula-nodulus (blue). Green represents areas of overlap, which contains cells that project to the flocculus and the uvula/nodulus (based on . (B) Shows the areas of the vestibular nuclei that receive input from the otolith organs (blue) and semicircular canals (yellow). The areas in green receive input from both the semicircular canals and otolith organs (based on Schwarz andSchwarz, 1983 andDickman andFang, 1996). Abbreviations: VeMc, r, pc, mc: the caudal, rostral, parvocellular, and magnocellular divisions of the medial vestibular nucleus; VeD: descending vestibular nucleus; FLM: medial longitudinal fasciculus; ph: prepositus hypoglossi; CbM: medial cerebellar nucleus; CbL: lateral cerebellar nucleus; INF: infracerebellar nucleus; VDL: dorsolateral vestibular nucleus; VeLv: lateral vestibular nucleus, ventral division; Ta: tangential nucleus; pcv: cerebellovestibular process. Scale bar = 1 mm.

Frontiers in Behavioral Neuroscience
www.frontiersin.org February 2013 | Volume 7 | Article 4 | 8 flocculus, and the responses of Purkinje cells to rotational optic flow, it has been determined that there are four optic flow zones in the rabbit flocculus: two VA zones interdigitated with two HA zones DeZeeuw et al., 1994;Van der Steen et al., 1994;Tan et al., 1995). A strikingly similar organization has been found in pigeons: there are 2 VA zones interdigitated with 2 HA zones ( Figure 8A). In other species it has also been shown that there is an interdigitation of HA and VA zones. In rats, there are 2 HA zones and 2 or 3 VA zones (Sugihara et al., 2004;Schonewille et al., 2006) whereas in macaques there appear to be two VA zone but only one HA zone (Voogd et al., 2012). Thus, the organization of optic rotational optic flow zones is highly conserved across birds and mammals (Voogd and Wylie, 2004). This evolutionary conservation does not extend to the uvula/nodulus. The topographic organization of the zones in the uvula/nodulus of pigeons gleaned from several of our studies up until 2003 (Winship and Wylie, 2003;Wylie et al., 2003a,b) is shown in Figure 8A. In pigeons we showed that the Contraction, Expansion and Descent neurons were organized in three adjacent zones, from medial to lateral. We were uncertain as to the location of the Ascent neurons. In the mammalian uvula/nodulus, the VA, HA, and vestibular-responsive neurons are organized into parasaggital zones Barmack and Shojaku, 1992).

THE RELATIONSHIP BETWEEN THE OPTIC FLOW ZONES AND ZEBRIN STRIPES
A parasagittal organization is also seen in the cerebellum with respect to the expression of numerous molecular markers (Herrup and Kuemerle, 1997). The most thoroughly studied of these is zebrin II (ZII; the metabolic isoenzyme aldolase C), which is expressed almost exclusively by Purkinje cells (Brochu et al., 1990;Ahn et al., 1994;Hawkes and Herrup, 1995). ZII immunopositive (ZII+ve) Purkinje cells are distributed as a parasagittal array of stripes, separated by intervening ZII immunonegative (ZII−ve) stripes (Sillitoe et al., 2005; Frost, 1993, 1999b;Wylie et al., , 2003aWinship and Wylie, 2003;Pakan et al., 2005). The location of the and Hawkes, 2006) (see Figure 8B). ZII stripes have been shown in several mammalian and avian species, with a strikingly similar pattern (Iwaniuk et al., 2009;Marzban and Hawkes, 2011). Thus, the pattern of ZII stripes is highly conserved, and likely critical for normal cerebellar function. The ZII stripes are apparent in folium IXcd of the pigeon VbC ; Figures 8B-D). However, in folium X there are no ZII stripes, as all the Purkinje cells are ZII+ve (Figures 8C  and D). There are seven stripe pairs in IXcd numbered, from the midline, P1± to P7±. Importantly, the P1± stripe is divided into medial and lateral halves (P1− med, P1− lat) by a thin ZII+ stripe that is only 1-3 Purkinje cells wide (see "?" in Figures 8C  and D). Also, the P2+ stripe is divided into medial and lateral halves (P2+ med, P2+ lat) by a thin ZII−ve notch (see inverted triangle in Figure 8C). In a recent series of studies Pakan et al., 2011;Graham and Wylie, 2012), we have attempted to determine if the ZII stripes are correlated with the optic flow zones in the VbC. We found a clear relationship: each optic flow zone spans a ZII+ve/−ve stripe pair. Data regarding the floccular zones is shown in Figures 9A-H, from our most comprehensive case (Pakan et al., 2011). The procedure was to record from identified HA and VA neurons, mark the recording sites with an injection of red or green fluorescent tracer (biotinylated dextran amine; BDA), then subsequently process the tissue for ZII to determine the location of the recordings. Figure 9A shows the view of the flocculus through the surgical microscope with six injection pipettes superimposed. Those marked C, E, and G contained green BDA, whereas the others contained red BDA. When the perfused brain was dissected, the six injections could be clearly seen under a dissecting microscope (Figure 9B). At sites C, F, and G, VA neurons were recorded, whereas HA neurons were recorded at sites D, E, and H. As shown in the corresponding panels with the ZII expression pattern visualized in coronal sections, VA injections were localized to stripes P4+ (Figure 9C), P6+ (Figure 9F), and P6− (Figure 9G), whereas HA injections were found in stripes P5+ (Figure 9D), P5− (Figure 9E), and P7− (Figure 7H). Supplemented with data from other cases, we determined that the medial and lateral VA zones spanned the P4± and P6± stripe pairs, respectively, and the medial and lateral HA zones spanned the P5± and P7± stripe pairs, respectively (see Figure 9J). A similar story emerged for the translation optic flow zones in the uvula. Figure 9I shows the locations of identified neurons superimposed on the ZII stripes in IXcd from several cases . The Contraction cells were localized to the P1+ and P1−med stripes, and the Descent cells were localized to the P2+lat and P2− stripes. Ascent and Expansion cells were found intermingled in the P2+med and P1−lat stripes. We did localize some cells to the P3+ stripe, but these were not modulated by the optic flow stimuli. The relationship between the ZII stripes and the optic flow zones in the VbC is summarized in Figure 9J. Each ZII+ve/−ve stripe pair spans an optic flow zone. Each one of these optic flow zones contains neurons with the same optic flow preference, with the exception of the one zone that contains both Ascent and Expansion neurons. Why this zone is peculiar in this regard is unknown, as is the function of the P3± stripe pair.
Although we have shown that each optic flow zone can be subdivided into a strip containing ZII+ve Purkinje cells and a strip containing ZII−ve Purkinje cells, the functional consequence of this remains unknown, as the function of ZII is not known. However, there are a few clues. First, shown in Figure 10, Pakan et al. (2010) found that most of the mossy fiber inputs from LM and nBOR (the green pathway shown in Figure 1) project adjacent to the ZII+ve stripes in IXcd. Thus, although both ZII+ve and ZII−ve neurons within a given optic flow zone receive visual input via climbing fibers from the mcIO, the ZII+ve cells seem to be getting more visual input via the mossy fiber pathways. Whether there are vestibular or somatomotor mossy afferents that project preferentially to ZII−ve stripes remains to be seen, but one could speculate that the ZII+ve and ZII−ve cells are processing different sensory information. Second, it has been shown that Purkinje cells in the ZII+ve and ZI−ve stripes within an optic flow zone likely project to different areas in the vestibular and cerebellar nuclei (Sugihara, 2011;Wylie et al., 2012). Finally, some studies have suggested that ZII+ve and ZII−ve cells may have different roles in plasticity (Nagao et al., 1997; Wadiche   Jahr, 2005;Ebner et al., 2012). For example, Paukert et al. (2010) showed that climbing fibers contacting ZII+ve Purkinje cells release more glutamate per action potential than those contacting ZII−ve Purkinje cells. They proposed that the ZII+ve Purkinje cells undergo more activity-dependent synaptic plasticity as a result. Thus, within an optic flow zone, there could be one system originating in ZII+ve stripes running in parallel with another system originating in ZII−ve stripes that differ with respect to: (1) mossy fiber inputs, (2) outputs to the vestibular and cerebellar nuclei, and (3) plasticity.

INTEGRATION OF LOCAL MOTION AND OPTIC FLOW IN FOLIA VI-VIII OF THE CEREBELLUM
In addition to the projection to IXcd, the LM also projects heavily to folia VI-VIII (Clarke, 1977), which is known as the oculomotor cerebellum (Voogd and Barmack, 2006).  investigated this projection using retrograde techniques. After injections of tracer into folia VI-VIII, most retrogradely labeled cells were found in LMm, whereas injections into IXcd labeled more cells in LMl (Figure 11). Injections into VI-VIII also labeled cells in the medial and lateral pontine nuclei. Previous reports have shown that the optic tectum projects to the pontine nuclei (Reiner and Karten, 1982). Thus, it appears that local motion from the tectum, and optic flow from LM may be integrated in the posterior cerebellum. What could be the function of this visualvisual integration? A few studies have shown that there is integration of local and optic flow information in primate visual cortex, and it has been suggested that this is important for "steering" to avoid obstacles during locomotion through cluttered environments (Page and Duffy, 2008;Elder et al., 2009). Hellmann et al. (2004) have suggested that the tecto-pontine pathway in birds is involved in avoidance behavior. Thus, perhaps the integration of optic flow and local motion signals in the posterior cerebellum of birds is important for obstacle avoidance as they fly through cluttered environments.