INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Three possible oculomotor roles have been suggested for the central mesencephalic reticular formation (cMRF) (Waitzman et al. 1996). Subthreshold low-frequency electrical microstimulation and single neuron recording of a low-frequency, long-latency (15-100 ms) discharge before saccades support the idea that the cMRF participates in saccade triggering (Cohen et al. 1985, 1986; Handel and Glimcher 1997; Waitzman 1982, 1992; Waitzman et al. 1996). Second, existence of cMRF neurons with contralateral movement fields that increase their discharge with the horizontal but not the vertical component of movement suggests that these cells could serve as a spatial filter extracting the horizontal component of movements from the superior colliculus (SC) output (Sparks 1986; Sparks and Mays 1990; Waitzman et al. 1996). Cells in the rostral portion of the MRF (see accompanying paper) may participate in the generation of the vertical component of saccadic eye movement (Handel and Glimcher 1997). Third, by virtue of a burst of activity that peaks just before and during saccades and dynamics of the neural discharge that correlate closely with either eye velocity and/or displacement, we have hypothesized that cMRF neurons could participate in the feedback control of saccades (Waitzman et al. 1996).

The impact of each of these hypotheses on saccade generation is illustrated with the help of two feedback models of oculomotor control (Fig. 1). The eye position model shown in Fig. 1A was based on the original, local-feedback model of Robinson (1975). Current eye position in space (Eye) was subtracted from target position in space (Targ) by the retina, to produce a retinal error signal (R_err). Robinson's major contribution was to suggest that retinal error was added to an internal copy of eye position (i.e., efference copy, or corollary discharge, E') in craniotopic coordinates to create an estimate of target position with respect to the head not the retina (Tar_est). In a subsequent step, efference copy (E') was subtracted from a delayed copy of target position to generate a motor error signal (e_m) used to drive the burst neurons in the pontine reticular formation (B). Integration of the velocity output of the burst neurons (V_c) by the neural integrator (NI) produced an eye position signal used to drive the ocular motoneurons. Two unique properties emerged from this model. First, by virtue of local feedback, burst output continued for as long as necessary to get the eyes onto the target and explained many aspects of the relationship between saccade amplitude and duration. Second, the input to the oculomotor system was a target position with respect to the head signal. This property in particular made it easy to incorporate vestibular inputs (Robinson 1975). However, since its proposal, a number of objections have been raised to this model and question its applicability to the oculomotor system. One primary concern has been that few regions of the brain contain eye position activity [i.e., nucleus prepositus hypoglossi (NPH) and the interstitial nucleus of Cajal (INC)]. More importantly these regions do not project back to areas such as the SC, which should receive feedback of this efference copy of eye position.

View larger version (25K): [in this window] [in a new window]   Fig. 1. Two local feedback models for generation of saccadic eye movements. A: an eye position model modified from Robinson (1975). Target position in space is compared with eye position in space by the retina generating retinal error (Rerr) at summing junction 1 (SJ1). Retinal error is then added to an efference copy of current eye position (E') at SJ2 to produce an estimate of target position with respect to the head (Tarest). After a delay of ~250 ms (i.e., central processing of visual information including target selection), current eye position (E'), is now subtracted from desired eye position (ed) at SJ3 to generate a motor error (em). Motor error, which represents how far the eyes must move to point toward the target, drives the burst neurons in the pontine reticular formation (PPRF) through a switch (trigger) thought to represent the omnipause neurons. The "velocity command" [Vc; output of the medium lead burst neurons (MLBs), B] is applied to the neural integrator (NI) to generate current eye position (an internal representation of eye position) and to the oculomotor plant to produce the actual eye movement. Reductions in desired eye position will lead to hypometric saccades, whereas interruption in the generation of the current eye position (i.e., loss of feedback) will generate hypermetric saccades. B: the local feedback portion of the above model (dashed rectangle) is recast into "change in eye position" coordinates (Jurgens and Becker 1981). The input to the local feedback loop is now a desired eye displacement (E) that is compared with current eye displacement to generate motor error (em). Again em drives the MLBs through a switch. However, now the output of the MLBs, Vc, is fed back through a resettable integrator (RI) to generate current eye displacement (E'). Vc is also applied to the NI to hold the eyes steady at a new location following each saccade, eye position (E'). This diagram illustrates that an increase in the input (E) to the feedback loop can lead to saccade hypermetria (Hyper #1). Damage to the RI or feedback path itself could also lead to saccade hypermetria (Hyper #2). Reduction in the trigger threshold will initiate saccades earlier (i.e., reduced saccade latency). Increased delays in the 2nd feedback pathway (Hyper #3) will lead to macrosaccadic square-wave jerks and in some instances, hypermetria (see DISCUSSION).

The eye displacement model shown in Fig. 1B addressed these issues by placing a resettable integrator (RI) into the local feedback pathway. This modification transformed the inputs of the model into retinotopic coordinates (Jurgens et al. 1981; see Waitzman et al. 1991, 1996 for further discussion of this model). Briefly, the input to the model, desired eye displacement (E), was thought to arise from the frontal eye fields (FEF) and dorsomedial frontal cortex (DMFC). The desired displacement was compared with current eye displacement (E') to produce a motor error (e_m) that was thought to reside in the long lead burst neurons (LLBNs) of the paramedian zone of the pontine reticular formation (PPRF). This motor error signal was then relayed through a switch (controlled by a trigger signal) to the medium-lead burst neurons (B) in the PPRF. The output of the burst neurons was a velocity command (V_c) that was directed to both the NI and a RI (E' or the efference copy) whose output was reset to zero at the end of each saccade. The purpose of the NI was to hold the eyes steady following the occurrence of each saccade while the output of the RI was used to update higher structures of the current displacement of the eyes. The NI for the horizontal saccade component is generated in the NPH (Cannon and Robinson 1987), and the NI for the vertical component of saccades is thought to originate from the INC (Crawford et al. 1991). The source of the trigger signal used to initiate saccades is thought to be the omnipause neurons located in the nucleus raphe interpositus (RIP).

Predictions about the specific oculomotor deficits, which may occur after inactivation of brain stem structures, are easier to understand by reference to these models. Shifts in the input to either model, that is a more distant orbital position (EP model), or larger eye displacement (ED model), would result in saccades that overshoot the goal (Fig. 1, A and B, Hyper #1). A shift in input could occur if cMRF neurons performed a spatial filter role for the SC and FEF output (Sparks 1986; Sparks and Mays 1990; Waitzman et al. 1996). Simulations of these various aspects of the models are presented in the DISCUSSION.

Reduction or damage to the pathways within the feedback loop would eventually produce a reduction in either the current eye position (EP model) or eye displacement (ED model) feedback signals. This reduction would increase the duration of the motor error signal (e_m), and the eyes would continue to move beyond their goal, albeit at a slower velocity (Fig. 1, A and B, Hyper #2). Thus, in the ED model if the reticulotectal, long lead burst neurons (RTLLBNs) of the cMRF provide a conduit for a velocity signal from the PPRF to the SC, or participate in the process of integrating eye velocity (i.e., the RI), loss of these cells should produce saccade hypermetria. This result would correlate well with the feedback hypothesis (Waitzman et al. 1996). However, damage to the feedback mechanism of the ED model could not produce a change in initial eye position or generate a saccade goal.

Reduction of feedback or damage to the neural integrator itself in the EP model would also produce hypermetric, slow saccades (Fig. 1A, Hyper #2). However, in this instance, shifts in initial position and generation of a saccade goal relative to the head could result. Moreover, damage in the second portion of the feedback pathway of the EP model (Fig. 1A, Hyper #3) could increase delays in the generation of the Tar_est and cause repeated saccades to a virtual target that continues to reappear (see DISCUSSION). Finally, making the saccade trigger easier to flip from opened to closed and vice versa could make saccade latency shorter. This might occur if excitatory activity from cMRF neurons important for maintaining the tonic firing of omnipause neurons was removed (i.e., the triggering hypothesis) (Cohen et al. 1985; Hepp and Henn 1982, 1983; Waitzman et al. 1996). Providing clear neurophysiological evidence to support each of these hypotheses of MRF participation in oculomotor control has proven difficult. The midbrain tegmentum contains both cells and fibers in passage from the superior colliculus and other structures. As a result, the destruction or activation of the collicular output may have biased previous electrolytic lesion and electrical microstimulation experiments (Cohen et al. 1982, 1985, 1986; Komatsuzaki et al. 1972).

The current group of experiments has been designed to circumvent some of these difficulties. Following electrical microstimulation and single and multiunit identification of the MRF, we have made microinjections of muscimol, a GABA_A agonist. We demonstrate that the MRF can be divided into two separate regions. Inactivation of a ventral-caudal region, which corresponds to the nucleus subcuneiformis (the cMRF), leads to oblique (contraversive and up) saccade hypermetria, higher saccade velocity, reduced saccade duration, and marked instability in fixation with the development of macrosaccadic square-wave jerks to a specific goal in the orbit (the current paper). Inactivation of the rostral portion of the MRF results in severe hypometria primarily of vertical, but not horizontal saccades (see accompanying paper, Waitzman et al. 2000). The implications of these findings are discussed with reference to the two models and three possible hypotheses for cMRF function just presented. Abstracts of these findings have appeared previously (Silakov and Waitzman 1996; Waitzman and Silakov 1994; Waitzman et al. 1997).

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The methods for recording eye movements and single neurons, electrical microstimulation, and data analysis in awake behaving primates in these experiments are essentially the same as those described in detail elsewhere (Waitzman et al. 1991, 1996). All procedures were approved by the University Animal Care and Use Committee.

Injection and recording procedures

In brief, four male rhesus monkeys (G, C, K, and T) were surgically prepared under isoflurane inhalational anesthesia with two eye coils (Judge et al. 1980), a head restraining device, and two stainless steel chambers to allow separate access to the MRF and the SC. The MRF cylinder was positioned over the posterior portion of the cerebral cortex tilted 15° off the sagittal plane (Waitzman et al. 1996). The MRF, located just lateral to the oculomotor nuclei, was identified by the characteristic features of single neurons that discharge with contraversive saccadic eye movements and electrical microstimulation that elicited contraversive, conjugate saccades at short latency (Silakov et al. 1995; Waitzman et al. 1996). Eye movements were recorded using the magnetic search coil technique and were accurate to 0.1° (Judge et al. 1980). In two monkeys, a series of guide tubes were placed parallel to each other and sampled the rostral, mid, and caudal portions of the MRF. The tubes were semipermanently positioned using a grid (spacing of 1 mm) fixed within the stainless steel recording chamber. In the third and fourth monkeys, only the caudal portion of the MRF was sampled. The arrangement of a rostral-caudal orientation of the guide tubes allowed for repeated testing and subsequent permanent identification of the sites of muscimol injections. A customized microinjection/recording needle (Crist et al. 1988) attached to a Hamilton syringe allowed for physiological confirmation of neuronal activity related to saccades before an injection and monitoring of neuronal activity after the injection. In monkey T, a picospritzer apparatus was substituted for the Hamilton syringe (Dias and Segraves 1997).

Behavioral paradigms

Figure 2A shows the fixation paradigm, and Fig. 2, B-D, illustrates the visually guided saccade (VGS) paradigm. Visual targets were positioned at eight different directions [0° (position 0), 45° (1), 90° (2), 135° (3), 180° (4), 225° (5), 270° (6), and 315° and/or 45° (7)] and five amplitudes (5, 10, 15, 20, and 25°) along each of these directions for a total of 40 different target locations. The monkey was trained to fixate a central light-emitting diode. After a variable interval of 200-400 ms, the light was extinguished, and a new target light appeared that was the cue for the monkey to shift his eyes and fixate the new visual target (15° saccades are shown in Fig. 2B and 20° in Fig. 2C). The monkey was rewarded for moving the eyes to within ±2° window of the visual target. After the injections this window was relaxed to ±7° and in some cases ±12° so that all attempted saccades to the visual target would be collected. The trajectories in each direction of Fig. 2C show five repetitions. Note the regularity, accuracy, and straightness of the trajectories. Following a control injection of saline in another monkey, the trajectories of the saccades were unchanged from baseline (compare Fig. 2D, saline, to Fig. 4A, same monkey 25° saccades, no injection). Filled circles show the average of all endpoints of control saccades to the same visual target. Saccades of five different amplitudes (5-25°; 8 randomized directions × 5 repetitions of each saccade + errors) were collected into separate files for each injection. Each file took ~6-15 min for the monkey to complete. A "complete" set of data covering all 40 positions was comprised of 5 files (1 for each amplitude, total collection time of 30-50 min). The amplitudes sampled during the first two or three files were repeated at the end of the sequence to document changes that occurred while the drug had diffused.

View larger version (28K): [in this window] [in a new window]   Fig. 2. Behavioral paradigms and control experiment. A: fixation paradigm. With fixation point onset the monkey makes a saccade to orient the fovea toward a small (<1°) visual stimulus. Note maintenance of fixation for 400 ms. Traces are horizontal (H) and vertical (V) eye position and fixation point appearance (FP). Saccades to the right are indicated as upward deflections. B: visually guided saccade (VGS) paradigm. Trial begins with fixation point onset toward which the monkey directs his eyes (similar to A). At a random interval later the FP disappears and a peripheral target appears (Tar). After a variable delay period of 200-400 ms, the monkey makes a visually guided saccade to look at the peripheral target. Calibration in A applies to B. C: control trajectories of visually guided saccades from primary position (straight ahead) to 8 different target positions (0-7) before injection (monkey K). Zero degrees was defined as saccades to the right and 180° as saccades to the left. Saccades down and to the right (i.e., 315°) were relabeled for ease of presentation as 45° (Fig. 11). D: 20° visually guided saccades to the 8 peripheral targets after the injection of 1 µl of saline in the rostral MRF of monkey G (for location see Fig. 1 of Waitzman et al. 2000). , average final eye position for ten 20° control saccades to each of these targets (g0217b.1).

Data analysis

Following each experiment raw eye movement records were processed by software that identified the beginning and end of each eye movement using a template matching algorithm (Waitzman et al. 1991). Each trial was visually inspected, and marks indicating the beginning and end of horizontal and vertical components of each saccade were corrected as needed. Corrective saccades following the primary movement were specifically excluded from the current analysis. Variations in saccade amplitude and direction following the injections were evaluated in a number of ways. One analytic technique calculated the fractional change in saccade amplitude and direction following the injection (Fig. 4D). A "difference coefficient" for each of these metrics was calculated by taking the difference between post- and pre-parameters and then expressing this as a fraction of the prevalue. This technique effectively normalized the data so that changes in eye movements of different amplitudes or directions could be compared. A negative value for the difference coefficient for amplitude (Diff. Amp.) indicated saccade hypometria, and a positive value reflected saccade hypermetria. Difference coefficients were plotted against target direction. In the analysis of changes following an injection, we also tested the "null" hypothesis that saccades in a particular direction were not deviated from their normal trajectory (Fig. 4E). If direction was not modified, then the direction at saccade end should be no different from target direction. The absolute difference between the angle at saccade end and target direction was the saccade deviation that was plotted as a function of target direction. If the deviation was positive (i.e., the postinjection angle was larger than target direction), then by definition this was plotted as a counterclockwise deviation (CCW), and if the deviation was negative (i.e., postinjection angle smaller than target direction), then this was scored as a clockwise (CW) deviation.

Two midbrain structures could be influenced by inactivation of the MRF by muscimol: 1) the nuclei of the optic tract (NOT) and 2) the INC. Contraversive slow phases of nystagmus develop after inactivation of the NOT, and position-dependent vertical postsaccadic drift occurs after inactivation of the INC (Cohen et al. 1992; Crawford and Vilis 1993; Crawford et al. 1991). To calculate the slow-phase eye velocity, instantaneous eye velocity was averaged from the end of the current saccade to just before the beginning of the subsequent saccade. This was done for the horizontal component of all spontaneous saccades (in total darkness) that occurred just before the paradigm began (including control injections of saline). The horizontal slow-phase eye velocity plotted for a single time point represented the average of all intersaccadic intervals for a particular file (~70-100 movements per file spanning 5-10 min). Time points were collected starting just before the injection and for each subsequent file following each injection until recording ended.

Drift amplitude [the amplitude of slow movement from saccade offset to the end of the drift as per Crawford and Vilis (1993)] was measured for at least 10 spontaneous eye movements occurring in each file. A running average (Student's t-test) was used to decide when significant vertical drift had occurred.

Duration is directly proportional to vectorial amplitude for pure horizontal saccades (Fuchs 1967). However, for oblique saccades component stretching occurs to produce saccade trajectories that are straight. As a result, component (horizontal or vertical) duration is proportional to vectorial amplitude (King et al. 1986) and is used to display the duration data here. Comparison of the slopes of saccade duration versus vectorial amplitude was made by t-test to determine whether a change in component duration had occurred after muscimol injections. In a similar fashion, the log relationship between vectorial amplitude and velocity (Fuchs 1967; King et al. 1986) was compared before and after muscimol to decide whether saccades had been displaced off this main sequence.

Histology

Once all data were collected and the most productive eye movement regions identified, a pressure injection of 1-2 µl of fluorescent labeled microspheres (green and red, LumaFluor, ~0.05 µm diam; blue, Polyscience, BB19773, 0.05 µm diam) was made to positively localize the sites of microinjection in three monkeys. The location of the electrode tracks in one monkey was identified by placement of a small electrolytic lesion. At the conclusion of the experiments, monkeys were deeply anesthetized with pentobarbital sodium and perfused. The brains were removed, and 50 µm vibratome sections were made through the brain stem. Unstained sections (with fluorescent beads) were mounted wet and photographed under both white and fluorescent light. Alternating sections were stained with thionin and drawn onto paper using an inverted microscope. Drawings were then scanned into the computer and traced to produce the final anatomic representation of injection sites.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Neuronal effects of muscimol: areas of inactivation

Eight injections of the GABA_A agonist muscimol were made in four monkeys at sites in the MRF where eye movement-related cells were recorded (Table 1). Of the eight injections, seven were made in head-fixed animals, and these injections were used to summarize the effects of muscimol inactivation. The eighth injection was done in the head-free animal to demonstrate the interaction between head and eye initial position shift. Besides these eight injections, two injections of inactive muscimol (determined empirically) produced no changes in eye movements and were used as controls. Changes in eye movements were noted as early as 5 min after a 1.0 µg injection of muscimol (Sigma, 0.5 µg/µl in sterile NaCl) into the MRF and could last for up to 7 h. Typically, electrical silence was noted at 20-30 min, and thus early time points were repeated after this initial inactivation period. Data collection began with the start of the injection and continued for as long as the monkey could perform the behavioral tasks.

We made parallel tracks 1 and 2 mm away from a 1.0 µg/µl muscimol injection in one monkey. Data from these tracks showed that the blocked region (electrical silence) extended no greater than 1.5 mm laterally from the site of injection. Cell activity 2 mm lateral to the injection was normal. Monitoring of activity above and below this site demonstrated a vertically blocked region of 2.5 mm above the site of the injection. No eye movements could be elicited from within the blocked region using electrical microstimulation at 3 times threshold, but eye movements could be elicited below the blocked region. Twenty-four hours later, neuronal activity in the blocked area had recovered, electrical microstimulation could elicit saccades, and eye movements had returned to normal. These experiments suggested that an injection of 1.0 µg/µl of muscimol inactivated an ellipsoid portion of the brain stem 2.2 mm in diameter and 1.5-2.5 mm in length. After a control injection of saline, neuronal activity was suppressed for ~3 min (Fig. 2D, monkey G), but returned to normal levels within 5-10 min. Following this control injection, saccades to the eight different target positions located 20° from primary position were straight, accurate, and thus unaffected by the injection (Fig. 2D).

Our initial hypothesis was that the MRF [corresponding to nucleus cuneiformis and nucleus subcuneiformis of Olszewski and Baxter (1954)] was physiologically a homogeneous region. As our experiments progressed, it was clear that some division of the "MRF" was necessary, because the effects on eye movements were quite different if injections were made rostral or caudal to the posterior commissure. Specifically, an analysis of caudal injection sites showed that oblique, upward, contraversive saccades became hypermetric (Fig. 3, all caudal injections), whereas vertical saccades became hypometric after rostral injections (Fig. 4A of accompanying paper, Waitzman et al. 2000).

View larger version (36K): [in this window] [in a new window]   Fig. 3. One representative section of the caudal brain stem showing the locations of all 8 muscimol injections made in 4 monkeys (). All of these injections produced saccade hypermetria, and their results are presented in the current paper. Abbreviations for all anatomic sections: III, 3rd nerve nucleus; IV, trochlear nucleus; aq, aqueduct of sylvius; BC, brachium conjunctivum; BCdec, decussation of the brachium conjunctivum; BIC, brachium of the inferior colliculus; BSC, brachium of the superior colliculus; CG, central gray; HBL, medial habenular nucleus; inc, interstitial nucleus of Cajal; LL, lateral lemniscus; ML, medial lemniscus; MLF, medial longitudinal fasciculus; NOT, N. of the optic tract; NPC, N. of posterior commissure; PB, N. parabigeminus; PC, posterior commissure; riMLF, rostral interstitial nucleus of the MLF; RPC or NRPC, N. reticularis ponto caudalis; RTP or NRTP, N. reticularis tegmenti pontis.

Inactivation of the ventral-caudal MRF (the cMRF): synopsis

Muscimol inactivation of the ventral-caudal MRF produced seven primary oculomotor effects in the monkey: 1) hypermetria of contraversive oblique saccades; 2) reduced saccade latency; 3) a moderate reduction in saccade duration, with an increase in saccade velocity following many injections; 4) repetitive macrosaccadic square-wave jerks to a specific goal in the orbit; 5) spontaneous saccades in the dark directed toward the same specific goal in space (relative to the head); 6) straight trajectories of saccades directed toward the orbital goal and curved trajectories of saccades directed toward adjacent locations to the goal; and 7) contraversive head-tilt following five of eight ventral-caudal injections.

These effects were consistent across monkeys and did not occur after inactivation of adjacent locations in the brain stem. The effects on saccade metrics (duration, velocity, and latency) will be illustrated for seven injections. The eighth injection was made in a monkey free to move its head, and thus the data for saccade metrics were not comparable to the head-fixed case. Careful examination of saccade duration will point to which portions of the oculomotor models could account for the observed changes. No change in saccade duration would suggest the input to the local feedback loop had shifted, whereas increased duration would suggest loss of feedback. Shorter duration and higher saccade velocity suggest a combination of effects on model parameters.

Analysis of square-wave jerks and the goal-directed nature of postinjection saccades are presented for all injections made in head-fixed animals. Each of these injections had a different goal to which spontaneous saccades were directed repeatedly. Data from two injections will be presented in detail, one in the left and the other in the right MRF. The rest of the data are presented in summary format to illustrate the range of effects observed.

Inactivation of the cMRF: changes in saccade metrics

Seven injections (c0416, c0419, c0521, g0217, k0329, k0331, and k0403) placed into the caudal MRF of three head-fixed monkeys produced hypermetric saccades. The results of one muscimol injection (1 µg/2 µl) placed at the site of cMRF long-lead burst neurons that discharged before contraversive (rightward) saccades is shown in Fig. 4. Multiunit contraversive eye movement related activity was registered through the recording syringe at a similar depth at which the single cells of Fig. 4F (movement fields) had previously been recorded. Electrical microstimulation was not performed at this site. Within 5 min after the end of the injection (duration of 20 min), 25° saccades up and to the contraversive side became hypermetric (Fig. 4B, positions 1 and 2; , averaged endpoints of control saccades). During the next hour of observation (5-30 min shown) all visually guided saccades up and to the right became hypermetric (Fig. 4, B-D). This hypermetria affected the vertical more than the horizontal component of movement (Fig. 4D). For a 15° oblique saccade the horizontal component of movement was increased by 20%, whereas the increment in the vertical component approached 50% (Fig. 4D, compare  with ).

View larger version (37K): [in this window] [in a new window]   Fig. 4. Hypermetric saccades developed after a muscimol injection in the caudal MRF of monkey G (g0217). A: control trajectories before injection. B: VGS trajectories 5 min following the injection of 1 µg (2 µl of 0.5 µg/µl) of muscimol into the left MRF. Note contraversive (rightward) saccade hypermetria to target positions 1 and 2. C: large almost 25° hypermetric saccades generated in direction 1 to a visual target that was located 10° up and to the right. Times indicate the onset of eye movement recording after muscimol injection. , averages of all preinjection control eye positions endpoints. D: vectorial difference coefficients for 4 different amplitude VGS in each of 8 target directions (, , and ). Horizontal and vertical difference coefficients for 15° VGS in each of 8 directions ( and ). Note that both the horizontal and vertical components are modified by the injection. Different symbols indicate which amplitude was affected by the injection. E: deviation of the endpoints of saccades occurred after the muscimol injection. Key in D applies to E. CW, clockwise rotation of saccade endpoints are shown along the ordinate (negative); CCW, counterclockwise rotation of saccade endpoints are shown along the ordinate (positive). F: 2-dimensional cartesian plots of the contraversive (right) movement fields of 3 neurons recorded from the left MRF at the site of this muscimol injection. G: site of this injection (g0217) and a control saline injection (g0223). Abbreviations as in Fig. 3.

There was a counterclockwise, upward deviation of the endpoints of contraversive, horizontal (position 0), and oblique, upward saccades (position 1, 45°) after this injection. The endpoints of pure upward movements (position 2, 90°) were deviated downward (i.e., negative direction, clockwise in Fig. 4E). A similar, albeit smaller reversal of saccade endpoint deviation occurred between positions 5 (225°) and 6 (270°; Fig. 4E). These reversals of saccade deviation correspond with zero crossings from counterclockwise to clockwise (Fig. 4E, arrows). This defined a plane tilted ~25° from the vertical toward which saccade endpoints were deviated (Fig. 4B). This plane also influenced the trajectory of the saccade. Saccade trajectories close to the plane remained almost straight, whereas saccade trajectories in other directions became curved. For example, the trajectories of upward vertical saccades to position 2 were bowed away from this plane (but their endpoints were closer to the plane), whereas the trajectories of oblique saccades to position 1 and those of horizontal saccades were bowed upward toward this plane. Similarly, downward saccades to position 6 were bowed away from the tilted vertical plane (Fig. 4B). Such curvature suggests discoordination in the generation of the horizontal or vertical components of the saccade such that the vertical component reached peak velocity before the horizontal component.

Details of the upward saccade hypermetria following this injection (g0217) are shown in Fig. 5. Hypermetric oblique saccades (position 1) had an increase in peak velocity compared with preinjection movements (Fig. 5A, horizontal; Fig. 5B, vertical). Saccades in the opposite direction (down and to the left, position 5) were only slightly hypermetric (Fig. 5D). In both cases (position 1 and 5), the amplitude and velocity of the vertical component was affected more than the horizontal component. In fact, horizontal component amplitude for ipsiversive saccades (position 5) was slightly hypometric (Fig. 5C, solid line). The overall increases in vectorial peak velocity for both directions (positions 1 and 5) were matched by a commensurate increase in saccade amplitude and duration. As a result, these postinjection saccades remained on the amplitude versus peak velocity main sequence (Fig. 5E, Table 2, P > 0.05).

View larger version (24K): [in this window] [in a new window]   Fig. 5. Changes in saccade dynamics and metrics that followed the left caudal MRF injection (g0217) of Fig. 4. A and B: eye position and velocity trajectories of 10° VGS directed up and to the contraversive (right) side (A and B: position 1, same movements as Fig. 4C) and down and to the ipsiversive side (C and D: position 5). Traces from top to bottom: horizontal eye position and eye velocity (A and C); vertical eye position and velocity (B and D; right and up are positive). Control saccades are indicated by dotted lines, and solid lines show postinjection saccades. Note that the vertical component of postinjection saccades toward the 10° target at position 1 had significant variability in saccade amplitude and velocities of up to 750°/s (B). Postinjection saccades to the ipsilateral side had slightly lower horizontal velocity (C), but higher vertical velocity compared with control (D). E: complementary changes in saccade velocity and amplitude produced saccades that fell on the main sequence relating vectorial amplitude and velocity. , preinjection saccades; , postinjection movements; · · ·, control regression; , postinjection regression.

The increased amplitude of postinjection saccades was matched by a commensurate increase in horizontal saccade duration. This maintained the same linear amplitude-duration relationship as before injection [Fig. 6A, slopes (m) not different]. However, duration of the vertical saccade component was longer than the associated vectorial amplitude would have required (Fig. 6B), while the slope of postinjection vertical duration versus amplitude relationship rose. The difference in slope did not reach statistical significance (see Fig. 10C). On the other hand, the latency to onset for saccades of all amplitudes was significantly reduced following this injection. Contraversive, upward saccades were initiated the fastest and some latencies (150 ms) approached that of express saccades (Fig. 6C).

View larger version (33K): [in this window] [in a new window]   Fig. 6. Changes in saccade duration, latency, and nystagmus following the muscimol injection of Fig. 4. A: horizontal component saccade duration (ordinate) was not different from control following this injection. Vectorial amplitude is shown on the abscissa. B: vertical component duration (ordinate) was longer after the injection, although the slopes were not statistically different. Key is the same as in A. C: postinjection latency (, 30 min postinjection) was shorter than control () for saccades of all amplitudes. D: horizontal slow-phase eye velocity following this same muscimol injection was not different (P > 0.5) from control (only contraversive slow phases are shown). Each symbol represents a different time point collected up to 60 min after the injection. Note that ipsiversive slow phases were also analyzed and were also not different from control (P > 0.5).

To determine whether muscimol had spread to include the NOT, dorsal to the MRF, horizontal slow-phase eye velocity (slow movements between saccades) was calculated after the injection (see METHODS) (Cohen et al. 1992). Control preinjection files demonstrated <3°/s of contraversive, slow-phase eye velocity (Fig. 6D, ). Following this muscimol injection, no contraversive horizontal nystagmus was found (Fig. 6D, ). These results suggested that the changes in saccade amplitude, velocity, and latency could not be accounted for by spread of the muscimol to involve the nucleus of the optic tract.

Inactivation of the cMRF: square-wave jerks and changes in initial position

At the end of 25 min after the injection, the monkey developed pronounced contraversive, upward macrosaccadic square-wave jerks (Fig. 7). The requirement of this particular paradigm was for the monkey to maintain stable fixation (Fig. 7, control, dotted line: see also Fig. 2A). After the injection the monkey made repeated saccades that were in the same direction as the previously described hypermetria. Each eye movement was separated by a minimal intersaccadic time interval of 150-200 ms. These movements were very stereotypic and brought the eye to a specific location in the orbit (Fig. 8A).

View larger version (25K): [in this window] [in a new window]   Fig. 7. Macrosaccadic square-wave jerks developing 20 min after a muscimol injection in the left ventral-caudal MRF (g0217). A and B: horizontal and vertical eye position, respectively, following the injection. ···, control fixation; , postinjection fixation. Note the >25° repetitive up and to the right (contraversive) saccades generated with barely one (200 ms) intersaccadic delay.

View larger version (41K): [in this window] [in a new window]   Fig. 8. Trajectories of macrosaccadic square-wave jerks and spontaneous saccades. A: Cartesian x-y plot of macrosaccadic square-wave jerks that interrupted fixation (see Fig. 7) were directed up and to the contraversive side. The middle of the box here and in B represents the average of the endpoints of all saccades, the edges of the box are ±1 SD around the mean. B: vectorial representation of amplitude and direction of spontaneous saccades made in total darkness 60 min after a left-sided muscimol injection (g0217). C and D: the slopes of the component horizontal and vertical amplitude vs. initial position plots formed the perturbation indexes for the spontaneous saccades of B. Note the effect of initial position on saccade amplitude.

Although changes of visually guided saccades with shifts in initial position were not specifically studied in this monkey, spontaneous saccades made in total darkness were collected just after this fixation file. The trajectories of the spontaneous saccades (whose vectors are shown in Fig. 8B) demonstrate that the eyes were directed to a specific goal in the orbit located 13° to the right and 19° up (error bars are ±1 SD). We compared an average of the endpoints of the macrosaccadic square-wave jerks from the fixation paradigm with the location of the endpoints of all of the spontaneous saccades and found an extensive overlap (compare rectangular boxes in Fig. 8, A and B). The dependence of postinjection spontaneous saccades on initial eye position was assessed by calculation of an "orbital perturbation index." This reflects the slope of the regression line relating component saccade amplitude with initial position (Russo and Bruce 1993). The indexes for horizontal and vertical saccade components were markedly elevated, supporting a strong effect of initial eye position on saccade amplitude. In summary, this injection demonstrated that inactivation of the cMRF was critical for the generation of saccade hypermetria. Within 1 h of this injection the monkey generated repetitive macrosaccadic square-wave jerks that brought the eyes to a specific goal in the orbit. Two hours after the injection, the monkey's head was released and a contraversive head tilt was noted.

Six other cMRF sites in two additional monkeys produced essentially the same results but to varying degrees. The results of an injection on the opposite side of the brain stem of another monkey are shown in Fig. 9. The first visually guided saccades were collected 88 min after the injection (Fig. 9B). The monkey could still make 25° saccades in all directions; however, the initial position of the eyes was displaced up and slightly to the contraversive side. Downward saccades were hypometric missing the target, even when the shift in initial eye position was taken into account. Contraversive (left) upward saccades were displaced clockwise toward the earth vertical. Thirty minutes later (127 min postinjection), saccades intended for the 5° contraversive target position (position 3) were markedly hypermetric and drawn toward a specific position in the orbit (Fig. 9C). Other contraversive movements were either very hypermetric or could not be generated. Ipsiversive 5° saccades were normal. Fixation was persistently interrupted by macrosaccadic square-wave jerks toward a contraversive upward goal (x = 11.4°; y = 19°, Fig. 9D). One hundred forty-two minutes after the end of the injection, spontaneous saccades were directed toward this same location, 11.8° left and 19.4° up (Fig. 9E). Slopes of the regression of initial eye position on component saccade amplitude were markedly elevated. This confirmed a strong effect of initial position on the amplitude and direction of spontaneous saccades (K_h = 0.61; K_v = 1.6). Last, postinjection saccades (particularly those to position 3) were displaced above the main sequence relating vectorial amplitude and velocity. However, the slope of the log regression for all postinjection saccades while higher did not reach statistical significance compared with preinjection control (Table 2). This result would not be expected from displacement of the saccade input (Hyper #1, Fig. 1), or interruption of the feedback loop (Hyper #2, Fig. 1), which would have left saccade velocity normal or lower, respectively.

View larger version (36K): [in this window] [in a new window]   Fig. 9. Generation of repetitive macrosaccadic square-wave jerks following a 2nd muscimol injection (1.0 µl of 1.0 µg/µl) in the right caudal MRF (c0521). A: control saccade trajectories. B: trajectories of saccades made to eight 25° targets 88 min after a muscimol injection in the right, caudal MRF. Note moderate saccade deviation and hypermetria of saccades to position 3. , average location of control saccades. C: a similar file of 5° visually guided saccades collected 127 min after the muscimol injection. Note the marked hypermetria of contraversive saccades made to a 5° target at position 3. D: macrosaccadic square-wave jerks that interrupted steady fixation at a target located at the center of the tangent screen. The center of the rectangular box corresponds to the average, and its edges are ±1 SD of final eye position for the entire group of square-wave jerks. E: spontaneous saccades made in total darkness 142 min after this muscimol injection. Note that the saccades were directed toward a goal whose center was located 11.8° to the left and 18.4° up. The rectangular box around these endpoints is determined as in D. Note extensive overlap with endpoints of saccades in D.

Summary of results: caudal injections

All but one muscimol injection showed a significant reduction in saccade latency (Fig. 10). The higher control latency for monkey C was most likely the result of difficulty with down gaze in the right eye (the recorded eye) following surgical procedures to correct tethering of the eye-coil in this eye. However, the reduction in latency was significant, and control latency returned back to 600 ms the day following the c0416, c0419, and c0521 injections. Taken together, these latency results suggest a role for the ventral-caudal MRF in saccade initiation and maintenance of fixation.

View larger version (51K): [in this window] [in a new window]   Fig. 10. Summary of changes in saccade latency and duration following muscimol inactivation of the caudal MRF. Data from 7 muscimol injections are included of which 4 were performed on the left side and 3 were placed in the right MRF (c0419, c0521, k0331). A: note the reduction in latency of all saccades with the exception of one experiment (k0331). The high preinjection latencies for monkey C were noted after an eye coil repair that slowed downward saccades. The postinjection reduction in saccade latency for the 3 injections in this monkey was still significant when downward saccades were excluded from the analysis. B: slopes of the component duration vs. vectorial amplitude graphs as shown in Fig. 6B (horizontal, B, and vertical, C) before and after injection. Horizontal saccade duration was reduced for all injections with exception of k0403, but reached significance for c0416 and c0419. C: a similar reduction in vertical component duration for all injections with the exception of c0521 and g0217, both of which showed moderate increases. Only the reduced duration of postinjection saccades of c0416 reached statistical significance.

Because of the differential effect on saccade velocity following the two injections shown in detail, we also examined the duration of the saccades for each of the seven injections. Note that in six of seven injections, horizontal saccade duration was reduced and in five of the seven injections, vertical saccade duration was reduced compared with control. These trends reached significance in only two of the horizontal and one of the vertical measurements (Fig. 10, B and C). This corresponded closely to an increase in saccade velocity that was higher than expected for amplitude and positioned these movements above the main sequence. These results suggest that inactivation of the ventral-caudal MRF could influence portions of the saccade burst generator.

Effects of muscimol inactivation of the caudal MRF on the amplitude of postinjection saccades are shown in Fig. 11. To compare the degree of hypermetria of one injection to that of another, the difference coefficient data were plotted so that the direction of hypermetria was up and to the right (i.e., 45°). Thus all injections were displayed as if they had occurred on the left side of the brain stem. Difference coefficients for the time point for which the monkey demonstrated the greatest hypermetria but was still capable of making saccades in all other directions are illustrated. All MRF injections caudal to the posterior commissure produced contraversive saccade hypermetria, albeit to a small degree in two injections (k0329 and k0403). The direction of saccade hypermetria was primarily up and to the contraversive side. In one case, horizontal saccades were hypermetric (c0419, ), but most often oblique upward saccades were hypermetric. This family of curves illustrates two points. First, the saccade hypermetria following ventral-caudal MRF inactivation ranged from ~5% of the control value to >50%. Second, there was a small secondary peak in saccade hypermetria 180° in the opposite direction of the primary hypermetria. This occurred for all injections except k0403, which had a small degree of hypometria in the opposite direction.

View larger version (32K): [in this window] [in a new window]   Fig. 11. Summary of difference coefficients for the 7 muscimol injections performed in the caudal portion of the MRF of 3 monkeys. The last difference coefficient curve for which the monkey could make saccades in all directions is displayed. , , , and , left-sided injections; , , , right-sided injections. The 3 caudal MRF injections (g0217, c0416, and c0419) associated with prominent contraversive hypermetria (25-50%) were also associated with a smaller but significant degree of ipsiversive hypermetria, directed 180° in the opposite direction of the largest contraversive point. Control injection of saline is shown by the bold solid line with opened circles. The data are plotted so that direction of largest hypermetria was located at 45°. This demonstrates 2 points. First, there was moderate hypermetria in the opposite direction for 3 injections. For example g0217 showed contraversive hypermetria at 45° (up and to the right) and ipsiversive hypermetria at 225° (down and to the left). Second, the degree of hypermetria ranged from 5% (k0403) to >50% (g0217).

A final aspect of the hypermetria was that it directed saccades toward a specific region in the orbit regardless of initial eye position. This was demonstrated by analyzing the direction and final endpoints of files of spontaneous saccades recorded toward the end of the monkey's ability to generate visually guided saccades. The different regions to which spontaneous saccades were directed are summarized in Fig. 12 for seven injections and two control days. Regions determined by right-sided MRF injections are shown by open symbols, and those following left-sided injections are filled. Note that all regions were located in the top half plane of movement and most (6 of 8) were contraversive. This trend of spontaneous saccades toward a specific goal in the orbit was confirmed by calculating an orbital perturbation index for both the horizontal and vertical components of spontaneous saccades. A clear effect (P < 0.05) of initial eye position on the vertical component of spontaneous saccades was found in three of seven injections (data not shown). The horizontal orbital perturbation indexes were increased over control (5 of 7) but did not reach statistical significance.

View larger version (22K): [in this window] [in a new window]   Fig. 12. Evidence for goal-directed saccades following muscimol injections. Plot showing the locations of the saccade goals of the spontaneous saccades for each of 7 muscimol injections made in the caudal MRF. All goals were located in the top half of the contraversive field of movement with the exception of 2 (k0329 and k0403). These goals corresponded closely with the location of the endpoints of square-wave jerks during fixation. Open symbols are right-sided injections, and filled symbols are left-sided injections.

Head tilt and shift of initial eye position

Three of eight injections in the ventral-caudal MRF were associated with a shift in initial eye position (Fig. 13). The shift in initial eye position increased over time, and typically the monkey made no attempt to compensate for the shift. The shift was contraversive in two injections (c0419 and c0521) and ipsiversive and up after one injection (c0416).

View larger version (24K): [in this window] [in a new window]   Fig. 13. Changes in initial position following all injections in 3 monkeys. The initial position was measured for successive time points and plotted in x-y coordinates. Note a small but consistent shift in the initial position following 3 injections (c0521, c0419, and c0416). , , , and , injections made in the left side; , , and , injections made on the right side. Two injections (c0419 and c0521) produced shifts to the contraversive side ( and ). One injection produced a shift up and to the ipsiversive side ().

One key to a better understanding of the changes noted in initial eye position came from studying the contralateral head tilt generated after the ventral-caudal muscimol injections. One possibility was that the shifts in initial position noted with the head fixed were compensatory for an attempted head movement in the opposite direction. Another possibility was that the MRF contributed to maintenance of initial eye position via its connections to the omnipause region of the PPRF. Loss of a fixation signal would produce destabilization of fixation similar to the macrosaccadic square-wave jerks shown above (Fig. 7). To examine whether the shift in initial eye position was compensatory, we measured the head tilt in one monkey free to move its head following a muscimol injection in the right ventral-caudal MRF.

With the use of an additional coil fixed to the head in the coronal plane, horizontal and vertical, but not torsional displacement of the head could be recorded (only 2 coils, 1 eye, and 1 head could be monitored). An almost immediate contralateral head roll of ~30° was confirmed visually and via photographs. The coronal coil demonstrated a contraversive and downward head displacement (Fig. 14, D, E, and G). This was associated with a compensatory shift of the initial position of the eyes up and to the ipsilateral (right) side (Fig. 14D, ; F, horizontal, ; H, vertical, ). This combination of head tilt and shift in eye position resulted in gaze (combination of head and eyes) being directed toward the center of the screen (Fig. 14B). This injection was performed using <0.5 µl of muscimol from a picospritzer apparatus, limiting spread of muscimol to adjacent structures. This suggests that the shift of initial eye position seen after the muscimol injections could have been compensatory for an intended head tilt.

View larger version (25K): [in this window] [in a new window]   Fig. 14. Head tilt and initial position changes 21 min after a muscimol injection in the right MRF of a head-free monkey (t0917). A and C: preinjection initial gaze, eye, and head positions. Gaze is centered. Head and eye positions are distributed around initial eye position. D: 30 min after muscimol injection there was a downward and small contraversive shift in initial head position (open circles) measured by a head coil in the coronal plane. Initial eye position (gray circles) was shifted upward and to the ipsiversive side. Gaze (filled circles, B) showed no change. E-H: histograms of initial head (E, horizontal, mean 1.86 ± 3.56; G, vertical, mean 4.50 ± 4.47) and eye position (F, horizontal, mean 2.46 ± 2.68; H, vertical, mean 2.82 ± 4.31) before (open bars) and initial head (E, horizontal, mean 0.47 ± 4.34; G, vertical, mean 8.75 ± 3.062) and eye position (F, horizontal, mean 1.75 ± 3.04; H, vertical, mean 8.15 ± 3.14) after (filled bars) injection. These histograms show the significant difference in initial head and eye positions that exactly compensated for each other. The postinjection means of both initial eye and head positions were significantly different from control at the P < 0.05 level.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

To better characterize the oculomotor function of neurons in the MRF, injections of the GABA_A agonist, muscimol, were placed at the sites of midbrain neurons that discharged with, or where electrical microstimulation induced, contraversive, conjugate saccades (Cohen et al. 1985; Handel and Glimcher 1997; Waitzman et al. 1996). Two previous, careful studies of single-neuron activity in behaving monkeys have revealed only a gross topographic arrangement of the oculomotor functions in this region. In particular, cMRF neurons, adjacent to the oculomotor nuclei, began to discharge 150 ms and peaked 8-10 ms before saccades with a contraversive horizontal or downward oblique component of movement (Waitzman et al. 1996). More rostrally located MRF neurons, adjacent to the INC also had long-lead activity, but were most sensitive to contraversive oblique and vertical saccades (Handel and Glimcher 1997). The movement fields for both groups of neurons were large and could extend for up to 40°.

The primary findings of this study were that inactivation of the ventral-caudal MRF 1) generated conjugate, contraversive, upward saccade hypermetria; 2) reduced saccade latency; and 3) produced a moderate increase in saccade velocity accompanied by a moderate reduction in saccade duration. Many ventral-caudal injections also produced macrosaccadic square-wave jerks that repetitively brought the eyes to a fixed place in the orbit. Similarly, spontaneous saccades executed in total darkness were directed toward the same specific orbital position. The distribution of the orbital goals was across the contralateral upper field of movement. Interestingly, three of the muscimol injections induced a displacement of the initial position of the eyes. These findings suggested a number of possible roles for cells in the cMRF: 1) participation in feedback of an eye position or displacement signal, 2) stabilization/maintenance of fixation, and 3) activation of the saccade burst generator. These results are discussed in light of the various anatomic connections of the MRF and how these cells could participate in the circuitry needed for the control of saccades and stabilization of fixation. Simulations of the two models presented in the INTRODUCTION are used to demonstrate that an eye position and not eye displacement model can best explain the current observations.

Localization of muscimol inactivation

GABA containing interneurons have been localized to the MRF (Nagai et al. 1983). Our basic assumption was that muscimol activation of GABAergic synapses on saccade related long-lead burst neurons in the MRF produced the observed oculomotor effects and left fibers in passage (i.e., the central tegmental tract) unaffected (Andrews and Johnston 1979; Krogsgaard-Larsen et al. 1979; Ritchie 1979). Single-unit recordings through and around the region blocked by muscimol demonstrated no neural activity in a sphere of radius 1.1 mm centered on the site of injection and support this view. Saline injections produced no oculomotor effects, thus eliminating a mechanical pressure gradient as the source of our findings (Fig. 2). Inadvertent inactivation of a number of oculomotor structures adjacent to the MRF including the nucleus reticularis tegmenti pontis (NRTP), the SC, and the NOT, could color the above interpretation of our results. However, muscimol inactivation of these regions has produced little or no saccade hypermetria (Aizawa and Wurtz 1998; Hikosaka and Wurtz 1985a; Lee et al. 1988; Munoz and Wurtz 1993; Quaia et al. 1998; van Opstal et al. 1996). Destabilization of fixation has occurred following rostral SC inactivation, but macrosaccadic square-wave jerks directed toward a specific location in the orbit were not generated (Munoz and Wurtz 1993). Possible inactivation of the NOT could produce horizontal, contraversive nystagmus (Cohen et al. 1992). In five of seven injections, no nystagmus was found (Table 1). In one of the other two remaining injections, onset of hypermetria occurred before contraversive nystagmus developed. Taken together this constellation of findings suggests that inactivation of the MRF and not adjacent structures produced saccade hypermetria.

Oculomotor functions of cMRF: anatomic connections

Intracellular filling of MRF neurons has shown them to be of at least three types (Scudder et al. 1996). RTLLBNs located within the central MRF (i.e., the nucleus subcuneiformis) direct their axons toward the ipsilateral SC where they arborize within the intermediate and deep layers. These cells provide a collateral branch that crosses the intracollicular commissure to innervate the contralateral SC (Moschovakis et al. 1988). This group of neurons could influence the generation of saccades in the SC. A second group of MRF LLBNs [probably reticulospinal LLBNs (RSLLBNs)] is located lateral to the INC, rostral to the RTLLBNs just described (Scudder et al. 1996). The RTLLBNs have contralateral movement fields, whereas the RSLLBNs have vertical movement fields (Handel and Glimcher 1997; Scudder et al. 1996; Waitzman et al. 2000). These cells have axons that descend toward the pons to innervate the raphe nuclei (raphe pontis, nucleus RIP, raphe obscuris) and the medullary reticular formation (primarily the inhibitory burst neuron region caudal to the abducens nucleus) (Scudder et al. 1996). The RSLLBNs could interact with both saccade and head generation networks in the pons and spinal cord (see accompanying paper, Waitzman et al. 2000). Scudder and colleagues (1996) have also described a third group of saccade-related neurons whose cell bodies are probably located within the caudal MRF (i.e., again nucleus subcuneiformis) and whose discharge is similar to the RTLLBNs. However, the axons of these cells (cRSLLBNs) were directed caudally within the predorsal bundle and innervated the NRTP, RIP, the nucleus reticularis pontis oralis and caudalis (NRPo-NRPc; including the excitatory and inhibitory burst neurons) and sent a descending axon to cervical levels of the spinal cord. An ipsilateral projection from the cMRF and cuneate reticular nucleus to the ventral horn of the cervical spinal cord, separate from that of the INC, has been confirmed repeatedly in both cat (Castiglioni et al. 1978) and monkey (Crawford and Villis 1993; Fukushima 1987; Fukushima et al. 1987; Kokkoroyannis et al. 1996; Robinson et al. 1994; Scudder et al. 1996). Moreover, the descending MRF projections were more numerous than the projections from the INC to the cervical spinal cord (Robinson et al. 1994; Scudder et al. 1996). Furthermore, the MRF also receives head-related proprioception via direct afferents from the cervical spinal cord and dorsal column nuclei (Bjorkland and Boivie 1984; Pechura and Liu 1986). In sum, this pattern of connectivity suggests that caudal MRF neurons could participate in the generation of combined head and eye movements. The output of cRSLLBNs to the omnipause neurons in the RIP and the precerebellar saccade generating machinery in the NRTP and NRPo-NRPc could account for the marked reduction in saccade latency found after cMRF inactivation (J. Buttner, personal communication; Buttner-Ennever and Buttner 1988; Horn et al. 1994; Scudder et al. 1996). In the normal state, MRF activity could enhance the tonic firing rate of omnipause neurons thereby suppressing unwanted saccades as has been suggested by the results of electrical microstimul