| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
1Graduate Program in Neurobiology and Behavior, 2Department of Physiology and Biophysics, and 3Washington National Regional Primate Research Center, University of Washington, Seattle, Washington
Submitted 9 March 2006; accepted in final form 21 October 2006
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
; Scudder et al. 2002; Sparks et al. 2001). These structures include the supplementary eye field (SEF), the frontal eye field (FEF), the superior colliculus (SC), and the pontine brain stem burst generator for horizontal saccades (Scudder et al. 2002). With the head free to turn, stimulation of either the SEF or SC produces combined eye and head movements whose metrics match those of natural, visually guided gaze shifts (Freedman et al. 1996; Martinez-Trujillo et al. 2003). Moreover, single neurons in the caudal SC discharge a burst of spikes whose characteristics are strongly correlated with the overall gaze shift (Freedman and Sparks 1997b). Because the FEF receives projections from the SEF and in turn projects to the caudal SC (Schall 1997; Stanton et al. 1988b), and FEF neurons have been reported to discharge a burst of spikes for contralateral head movement (Bizzi and Schiller 1970), our initial hypothesis was that the FEF is involved not only with the generation of eye saccades but also with the generation of combined eye–head gaze shifts when the head is free to turn.
Currently, there is little evidence to either support or refute this hypothesis. The three investigations of the effect of electrical stimulation of the FEF on head-free gaze shifts have reached contradictory conclusions. The first study evoked eye movements that were sometimes coupled with head movements that contributed to the gaze shift—i.e., the head moved during the gaze movement. However, this head contribution to gaze was small except when high-intensity current (200 µA) was used to evoke large-amplitude (
70°) gaze movements (Tu and Keating 2000). The two more recent studies evoked gaze movements composed of eye movements alone, with any head movement occurring after the end of the gaze movement (Chen 2006; Sparks et al. 2001). Thus the former study concluded that the FEF participates in gaze shifts with the head unrestrained, whereas the latter two concluded it does not.
To resolve whether the FEF is involved with combined eye–head gaze shifts, we stimulated the FEF under conditions chosen to facilitate the occurrence of a head movement. These included: concentrating on the dorsomedial portion of the FEF where stimulation elicited large gaze shifts, which under normal conditions would be accompanied by a head component (Freedman and Sparks 1997b; Phillips et al. 1995; Tomlinson and Bahra 1986a), using naturally occurring initial eye and head positions, which were more likely to produce a head movement in a normal gaze shift, and recording unit activity before stimulation to determine the neuronal substrate associated with the evoked gaze shifts. Moreover, we used the same low-intensity stimulus currents that Bruce and colleagues (1985) used to elicit eye saccades in head-restrained monkeys, thereby defining the low-threshold FEF.
Here we demonstrate that stimulation of the dorsomedial FEF produces gaze shifts with eye and head movement components that are characteristic of normal, visually guided gaze shifts. In a companion paper, we will show that many single neurons recorded at these same sites in the FEF discharge a burst of spikes related to the amplitude of the head movement component of large gaze shifts. These data suggest that under natural conditions the FEF is involved not only with the generation of saccadic eye movements but also with the generation of head movements required for large eye–head gaze shifts. Some of these results were previously presented in abstract form (Knight and Fuchs 2001).
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
These experiments were performed with two male rhesus monkeys, Macaca mulatta (weight 4–9 kg). During asceptic surgery, a scleral search coil for measuring gaze position in space was implanted on each monkey’s left eye (Fuchs and Robinson 1966) and a similar coil for measuring head position in space was mounted on the head in the coronal plane, just anterior to the axis of horizontal head rotation. We implanted a magnetic resonance imaging (MRI)–compatible recording cylinder (Crist Instruments, Damascus, MD) over each FEF in the first animal and over the right FEF in the second animal. During experimental sessions, each animal sat in a primate chair, which restricted its upper body rotations, and made gaze shifts from one target position to another with the head completely unrestrained. The chair was placed in a dark sound-deadened booth and the animal’s condition was monitored with an infrared video system.
Experimental procedures
Each animal was trained to make normal, visually guided gaze shifts to a step change in target position between two sequentially (within 5 ms) illuminated light-emitting diodes (LEDs) on an array in front of the animal (LED spacing every 1° of visual angle, ±80° horizontally, ±30° vertically and obliquely). If the gaze shift landed within a reward window surrounding the illuminated (target) LED, the animal received an applesauce reward. The size of the window was roughly ±10% of the size of the target step. The LED target steps varied in amplitude from 1 to 100°. During training, the animals were also rewarded for maintaining gaze position if the fixated initial target was extinguished and was not followed by a subsequently illuminated LED. This reduced the likelihood of nontargeting, spontaneous gaze shifts (see "catch trials" below). The locations of target LEDs and the intertrial intervals (500–1,500 ms) were varied pseudorandomly to reduce the predictability of the size, direction, and timing of a target step.
Once the animal was trained, we introduced a tungsten recording microelectrode (impedance 
1.0 M
) through the dura by a cannula and advanced it into the cortex with a hydraulic microdrive (Trent Wells). After the recording of single-unit or multiunit activity related to gaze shifts, we applied microstimulation at that same site by the same electrode. Stimulation trials were intermixed randomly with normal, visually guided gaze shifts at a ratio of about 1:3. Each stimulation trial began at a fixated LED position that was varied across trials to elicit different initial eye and head positions. After eye and head positions were stable at an initial gaze position for about 200 ms, we extinguished the fixation target and started electrical microstimulation at pseudorandom delays ranging from 50 to 300 ms after fixation target offset. No visual stimulus was present during microstimulation (typically 200 ms). Four hundred to 1,000 ms after the end of microstimulation and after any microstimulation-evoked movement had ended, the target reappeared at a random position and the animal was rewarded if it made a gaze shift to the new target position. Therefore during a stimulation trial there was never a different target presented before the onset of microstimulation and indeed not until 
650 ms after offset of the fixation target. "Catch trials" (roughly one trial in 10 stimulation trials), where the target was extinguished for >500 ms and no microstimulation was applied, were used to confirm that the animals maintained initial fixation until after a visual target appeared. These catch trials minimized the possibility that spontaneous (such as centripetal movements from peripheral positions) movements interacted with microstimulation-evoked movements.
Stimulation was applied with a constant current stimulator (Nuclear Chicago). At all sites described herein, we first verified that saccades were evoked with low-threshold stimulation current (threshold <50 µA, symmetric 0.25-ms biphasic pulses at 350 Hz with 70-ms train duration; Bruce et al. 1985). Threshold was taken as the current that evoked saccades on 50% of the trials under these stimulation conditions. We then increased current to twice threshold and stimulated at 350 or 200 Hz for durations of 200 ms, except where otherwise noted (e.g., Figs. 10 and 11).
|
|
|
|
|
|
|
|
|
|
|
The positions of the gaze (G) and head (H) in space were measured by the magnetic search coil technique (Fuchs and Robinson 1966; Robinson 1963). Analog signals of target, gaze, and head position were low-pass filtered, recorded to VCR tape (Vetter model 4000A PCM recorder, A. R. Vetter), and digitized at 1 kHz with a Macintosh computer (National Instruments A/D conversion board, PCIMO6). Stimulation pulses were digitized with 10-µs resolution. Off-line data analysis was performed with laboratory-designed software and eye position in the head (E) was calculated off-line as G – H. Digitized G, H, and E position signals were digitally filtered and differentiated to produce the respective velocity signals. Behavioral and stimulation data from both visually guided and stimulus-evoked gaze shifts were analyzed on a Macintosh G4 computer with an interactive program that displayed the positions and velocities together with the stimulation pulses. The program detected and marked features of each gaze shift and its eye and head components (e.g., start and end times when movement velocities exceeded and then fell to <10°/s, start and end positions, peak velocities, etc.). The program also detected and marked the onset and offset of the stimulus train and the start and end times and positions of the target movement (Phillips et al. 1995). The authors inspected the automatic markings and corrected the occasional errors. Because FEF neurons may have disynaptic connections to extraocular motoneurons (Moschovakis et al. 2004), no trials where gaze movements started at <5 ms after stimulation onset were included in the analysis (only 1% of all trials had gaze latencies between 5 and 10 ms). Those evoked movements with latencies >150 ms were also not considered. The program used these markings to determine movement metrics and stimulation event times and these data were then exported to Matlab (The MathWorks) and Statview (SAS Institute) for further analyses.
Because the head typically continues to move after the gaze movement has ended [and gaze is stabilized in space because of the vestibuloocular reflex (VOR)], we considered two aspects of the head movement separately. Total head movement amplitude (head) was the change in vectorial head position from head movement start time to head movement end time. The head contribution to the gaze shift amplitude (head contribution) was the change in head position from gaze movement start time to gaze movement end time (see Freedman and Sparks 1997a).
Calculation of initial position effects
We quantified the effect of initial eye and head positions on stimulus-evoked movement amplitude with a multiple linear regression (MLR) analysis, which determined the effects of horizontal and vertical initial eye and head positions (IEPh, IEPv, IHPh, IHPv, respectively) on each evoked movement component’s amplitude with the model
 | (1) |
to analyze initial position effects on saccades evoked by head-restrained FEF stimulation. The range of these indices was typically between 0.0 (which characterized the evoked movements from a stimulation site as independent of initial position, i.e., "fixed vector" movements) and –1.0 (which characterized movements as ending at the same position regardless of initial position, i.e., "goal-directed" movements).
Our MLR analysis proceeded as follows. First, at each site we analyzed only those trials that evoked a movement. To confirm the linearity of the remaining data for each site, we inspected simple linear regression (SLR) fits and residual plots of each movement component’s amplitude regressed on each initial position (IP) variable (e.g., Gh on IEPh). To protect against misinterpretations of MLR results arising from intercorrelations among the IP variables, we confirmed that variables were relatively uncorrelated (i.e., the R2 of any independent variable predicted by all of the remaining independent variables was <0.90, variable inflation test; Kleinbaum et al. 1988). We used the equal correlations test (Kleinbaum et al. 1988) to determine which component amplitude was best correlated with which IP variable, and then we added other IP variables to obtain the best-fitting MLR model for that component (extra sum of squares/partial F test; Draper and Smith 1981; Kleinbaum et al. 1988). We confirmed this result with forward and backward stepwise MLR. These methods accounted for correlation among the independent variables and resulted in models that included only those independent variables that had statistically significant effects on the dependent variable (movement component amplitude). Standardized regression coefficients provided a comparison of the relative importance of the different variables included in the model of best fit.
MLR analyses also yielded the characteristic vectors (CVs) for gaze, eye, head, and head contribution (Klier et al. 2001; Martinez-Trujillo et al. 2003; Russo and Bruce 1993) for each stimulation site. The CV approximates the typical two-dimensional (2D) movement elicited at a particular site as if the movement had been evoked from the straight-ahead position (i.e., the initial positions of the eyes and head were zero). The 2D CV for each movement type (gaze, eye, head, and head contribution) was derived from all movements evoked at that site. The 2D CV for each movement type was calculated (Pythagorean theorem) from the horizontal and vertical component CVs of each movement. The horizontal and vertical component CVs were the y-intercepts (b0 in Eq. 1) of the models of best fit for each component across all trials where a movement was elicited. Because CVs reflect the effect of initial position on movement amplitude, they are more representative of the evoked movements for a given site than are average amplitudes and thus provide comparable descriptions of the evoked movements at each stimulation site. We used similar methods to calculate the characteristic latencies, characteristic durations (CDs), and characteristic peak velocities (CPVs) of the evoked movements.
Anatomical confirmation of stimulation sites
Postoperative magnetic resonance imaging (MRI) was used to identify the anatomical locations of the electrode paths in the FEFs of both monkeys (Kim and Shadlen 1999). In addition, electrolytic lesions were made in the right and left FEFs of the first animal (T) by passing 30 µA of positive DC current (Nuclear Chicago) for 30 s at two depths along a medial and lateral recording track. After 10 days, the monkey was anesthetized and killed, after which the brain was removed and processed for histology (Robinson et al. 1994).
The Animal Care and Use Committee at the University of Washington approved all the surgical, training, and experimental procedures. The veterinary staff of the Washington National Regional Primate Research Center cared for the animals that were housed under conditions that comply with National Institutes of Health standards, as stated in the Guidelines for the Care and Use of Mammals in Neuroscience and Behavioral Research (compiled by the National Research Council, Washington, DC, National Academy Press, 2003), and recommended by the Institution of Laboratory Resources and the American Association for Accreditation of Laboratory Care International.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
We evoked eye and head movements with low-intensity microstimulation at 60 sites in two monkeys. Thirty-four of these sites (23 in monkey T; 11 in monkey P) provided large data sets with varying initial positions, which we subjected to the detailed analyses reported here. Stimulus trains with durations of 70 ms often evoked head movements in association with an eye saccade. However, head movements were reliably evoked (28 of 34 sites) only with longer stimulus durations of 200 ms. Figure 1 indicates the range of stimulus-evoked, head-unrestrained gaze shifts by illustrating horizontal gaze, eye, and head movements evoked with 200-ms stimuli from three representative sites. The evoked gaze shifts began after a short latency from stimulus onset and could end before or after the end of stimulation. Gaze shifts evoked at most sites were primarily horizontal with head movements varying from large (left column) to moderate (middle column) to little, if any (right column). The middle vertical line in each panel indicates the time of the end of the gaze movement and thereby identifies the end of the associated head contribution to the gaze movement. Whenever a head movement occurred (Fig. 1, A and B), the head contribution was always less than the overall head movement (which usually continued past the end of the gaze movement). As it does in natural gaze shifts, the head movement usually continued beyond the end of the gaze shift and the VOR served to keep gaze stable in space. As we demonstrate later (Figs. 10 and 11), longer-stimulus trains at some sites elicited head movements as large as 115°, which ended only when the stimulation did. Stimulus durations of 200 ms did not evoke multistep eye/gaze movements; all analyses (except those illustrated in Figs. 10 and 11) included only movements evoked with a stimulus duration of 200 ms.
Movement latency from stimulation onset
We calculated mean gaze and head movement latencies from stimulation onset for evoked movements of 1° in amplitude. Twenty-three of 34 sites were stimulated at 350 Hz and 200-ms duration. The gaze latency (mean ± SD: 31 ± 12 ms) was shorter than the head latency (60 ± 16) for all but three of these sites (paired t-test, P < 0.001). Twenty of the 34 sites were stimulated at 200 Hz and 200 ms. For these stimulation conditions, the mean latencies for both gaze (50 ± 17 ms) and head movement (74 ± 21) were also significantly different from one another (paired t-test, P < 0.001) and longer than the latencies resulting from 350-Hz stimulation (Wilcoxon signed-rank test, P < 0.05, 10 sites stimulated at both frequencies). At neither stimulus frequency were gaze latencies more tightly related to the stimulus onset than were head latencies (variance ratio F-test, P > 0.10; Commenges and Seal 1985; Hanes et al. 1995; Zar 1999). As with normal visually guided movements, gaze and eye start times were equivalent (Freedman and Sparks 1997b; Martinez-Trujillo et al. 2003).
These mean latencies were calculated without regard to initial positions. To test for position effects on movement latencies, we calculated "characteristic latencies" for gaze and head movements for each stimulation site. The characteristic latencies for almost all of the 34 sites were no different from the mean latencies reported earlier, and position effects were minimal except at 14 sites, which showed strong (R2 > 0.25, P < 0.05) effects of initial eye position on gaze latency. At these sites, gaze latencies were shorter when the eyes were initially positioned in the direction opposite that of the ensuing evoked gaze movement (i.e., in the "off-direction").
Two-dimensional trajectories of evoked movements
To compare the movements elicited at different loci in the FEF, we constructed 2D plots of the evoked movements. Figure 2A shows the trajectories of all gaze, eye, and head movements evoked with 350 Hz and 200-ms stimulus trains from different initial positions at a representative site. For clarity, Fig. 2B illustrates the same trajectories as vectors connecting the start and end (open symbol) positions of each movement. The open arrows indicate the site-typical evoked movement’s direction and amplitude as a characteristic vector (CV). Stimulation at this site in the right FEF elicited contraversive and predominantly horizontal gaze movements with head movements up to about 40° in amplitude if gaze was initiated from the off-direction. Although the evoked eye movements often had a prominent vertical component, the head movements evoked from all initial positions were primarily horizontal; this was true for nearly all sites at which a head movement was evoked. Normal gaze shifts to oblique target steps also have little if any vertical head movement (Freedman and Sparks 1997b).
For this site and most others, the vectors of the evoked gaze, eye, and head movements varied with initial position. Generally, when initial positions were in directions opposite to the evoked movements, the elicited movements were larger. Furthermore, as seen for this representative site, individual trajectories of gaze and eye movements tended to converge toward a location in space; i.e., these evoked movements exhibited some degree of goal-directed behavior. If initial position had had no effect on these evoked movements, each trajectory would have been identical—i.e., they would have fixed vectors.
At some of the same sites (13 of 34) where stimulation elicited gaze movements like those illustrated in Fig. 2, stimulation from some initial positions occasionally elicited a head movement without an associated rapid eye movement; however, the eye underwent counterrotation similar to that observed with the VOR. Figure 2B identifies the 26 "head-only" trials for this stimulation site in the "Gaze" (solid black squares; note lack of movement vector) and "Head" movement plots (solid black circles with short, left, and down movement vectors). Head-only movements occurred in only a small number of the stimulation trials at these 13 sites (two to 54 trials per site). These ranged from 2 to 13° in amplitude, usually moved centrifugally in the on-direction for that site, and had peak velocities and latencies (9 to 138 ms) similar to those of head movements of the same amplitude evoked in association with saccadic eye movements. These head-only movements were evoked mostly when initial eye position was deviated in the direction of the evoked gaze movement. Although they were only a small percentage of stimulation-evoked trials, these head-only movements represent an extreme example of the effect of initial position on stimulus-evoked movements. These trials were excluded from calculation of the CVs in Fig. 2B and all further analyses. Stimulation never evoked a head movement where the eyes remained immobile in the head—either a saccadic eye movement was also evoked or an eye counterrotation occurred.
Characteristic vectors
To provide a concise description of the typical movements evoked from each stimulation site, we calculated the characteristic vectors (CVs; see METHODS) for gaze, eye, head, and head contribution and plotted them as arrows originating at the origin (see Fig. 2B). CVs could be readily calculated for most stimulation sites (24 of 34) where movement amplitudes were a linear function of initial eye or head position. For the remaining 10 sites, the data were nonlinear: movements either were evoked from most initial positions, but not at all from others, or they were oppositely directed from opposite initial positions. For these sites, we separated the data based on initial position into subsets within which there was linearity. We then calculated a CV for each subset and averaged those CVs to get the site-specific CV for each movement.
SPATIAL DISTRIBUTION OF CVS. Figure 3 (left column) shows the distribution of CVs for gaze (A), eye (C), head (E), and head contribution (G) movements for all 34 sites. The endpoint of each CV is indicated by a square (gaze and eye) or circle (head and head contribution), and all CVs start at the origin. Of 34 sites 29 had contraversive gaze and eye CVs (open squares). At the five sites with ipsiversive CVs (solid black squares), stimulation elicited small gaze and eye movements with little or no head movement. Most gaze and eye CVs had substantial horizontal components with varying amounts of upward deviation; the gaze and eye CVs of 13 (including three ipsiversive) sites were predominantly vertical. Head and head contribution CVs were predominantly horizontal and they were contraversive at all but one site (Fig. 3, E and G, solid black circle).
The histograms in Fig. 3 (right column) compare the distributions of CV amplitudes (in white) for gaze (B), eye (D), head (F), and head contribution (H) to those of the maximum amplitudes elicited at the same sites (in black; overlap in gray) for stimulus durations of 200 ms. The mean CV amplitude for each movement type and all sites is indicated by a black vertical line with the crossing horizontal line at the top indicating the 95% confidence interval. The mean CVs for gaze, eye, head, and head contribution across the 34 sites were all >0. Gaze movement CVs ranged from 1 to 41° (mean ± SD: 17 ± 11; number of trials 16 to 210, mean 86) and maximum evoked gaze movement amplitudes ranged from 4 to 86° (Fig. 3B). Eye movement CVs ranged from 1 to 36° (15 ± 9) and maximum evoked movement amplitudes ranged from 3 to 56° (Fig. 3D). Head CVs ranged from 0 to 17° (4 ± 4) and maximum movement amplitudes ranged from 0 to 60° (Fig. 3F). Most head contribution amplitudes were small, with CVs ranging from 0 to 11° (2 ± 3) and maximum evoked movements from 0 to 34° (Fig. 3H). Although the head contribution CVs at 22 sites evoked with 200-ms stimulation were <1°, individual head movements with amplitudes >2° could be evoked at most of these sites.
EFFECTS OF INITIAL EYE AND HEAD POSITION. As we saw in Fig. 2, stimulation at certain sites in the FEF evoked gaze shifts that clearly depended on the initial position of the eye and/or head. To quantify this position dependency we calculated the index of initial position effect for each movement component from the best-fitting MLR model (also used to calculate CVs) for each of the 24 sites where the data exhibited linear relations between the components of the evoked movement and initial movement positions.
The amplitude of the horizontal component of gaze (Gh) and eye (Eh) movements was best predicted by horizontal initial eye position (IEPh) at most sites (18 and 19 of 24, respectively). That is, the simple regression of Gh versus IEPh had the statistically best fit (equal correlations test, P < 0.05), and IEPh was the primary variable (its coefficient was statistically significant and its standardized regression coefficient was the greatest) in the MLR model. Similarly, the vertical component of gaze (Gv) and eye (Ev) movements was best related to IEPv (16 of 24 for both). Finally, the horizontal component of head (Hh) movement was best related to IHPh at 14 of 24 sites, although the vertical component was best related to IHPv at only four sites. In nearly all cases, the MLR model of best fit consisted of either one or two initial position variables. The primary variable was nearly always related to the movement component along the same direction, i.e., IEPh with the horizontal component of eye movement. Because the identity of the secondary variable was seldom the same across sites, we considered only the relation between the movement component and the primary initial position variable.
Figure 4 plots the index of initial position effect of the relation between component amplitude and the primary initial position variable for the vertical and horizontal components of gaze, eye, and head movement for these 24 sites. For the horizontal and vertical gaze movement components (Fig. 4A), the mean (± SD) index of initial position effect was –0.31 ± 0.23 and –0.35 ± 0.25, respectively (ranges: –0.70 to 0 and –0.78 to 0). For four of the 24 sites, IHP, not IEP, was the primary variable influencing evoked gaze movements (median index: –0.19). For the horizontal and vertical eye movement components (Fig. 4B), the mean index of initial position effect was –0.35 ± 0.23 and –0.35 ± 0.23, respectively (ranges: –0.79 to 0 and –0.76 to 0). Again, at four stimulation sites, IHP was the primary variable influencing evoked eye movements. Head movements were much less dependent on initial position (Fig. 4C). Fifteen of 24 MLR models that fit horizontal head movement amplitude to an initial position variable included statistically significant values for IHPh (Fig. 4C), but the mean index of initial position effect was not statistically different from zero (–0.10 ± 0.16; range: –0.63 to 0). Only five models fitting vertical head movement amplitude were statistically significant and the mean index of initial position effect was not different from zero (–0.04 ± 0.10; range: –0.37 to 0).
The movements illustrated in Fig. 2 exhibited indices of initial position effect that were above average. At this site, the horizontal and vertical gaze and eye movements were characterized by coefficients of –0.70, –0.69, –0.78, and –0.66, respectively, demonstrating that the gaze and eye movements were more strongly influenced by initial eye position than not. Horizontal head movement was also more strongly influenced by initial position, but vertical head movement position dependency was similar to the mean (indices of –0.41 and –0.07, respectively). Thus an index of initial position effect of –0.35 indicates an influence on evoked gaze and eye movements somewhere between the effect seen in the more goal-directed site of Fig. 2 and a hypothetical fixed-vector site.
The remaining 10 "nonlinear" sites exhibited perhaps the strongest initial position effects because the evoked movements tended to converge more dramatically toward a goal. Figure 5 shows the 2D trajectories (redrawn as vectors) of movements elicited from two such sites. For the site in Fig. 5A, stimulation evoked rightward horizontal gaze and eye movements (endpoints indicated by squares) if the movements started from the left, and leftward movements if they started from the right. Movements evoked from upward initial positions pointed to a goal to the left and down. Data from six other sites showed similar goal-directed evoked movements. For the site characterized in B, stimulation evoked upward gaze and eye movements that appeared to converge at an eccentric location to the right and up but never reversed direction (as in Fig. 5A). A similar convergence of evoked movements to an eccentric goal was seen at two other sites. Evoked head movements at the 10 "nonlinear" sites usually were small, if present at all, but at two of these sites, the evoked head movements also exhibited reversals in direction.
TOPOGRAPHIC ORGANIZATION. Larger gaze shifts with head movement components most often were elicited from stimulation of dorsomedial rather than ventrolateral sites in the FEF. Locations of the 34 stimulation sites were reconstructed with the aid of magnetic resonance (MR) images of each FEF and its recording cylinder. In monkey T, sites within the left FEF were also verified by histological examination of electrolytic lesions. Stimulation sites were deep within the cortex, but are represented as locations on the surface of the FEF in Fig. 6 ("+," as viewed normal to the cortical surface). Gaze, eye, and head movement CVs start at the primary direction of gaze and end at the symbols as in Fig. 3. Fifteen sites lay in the left (A) and eight in the right (B) FEF of monkey T and 11 in the right FEF of monkey P (C).
Stimulation of dorsomedial sites (those medial to a line drawn from the caudal tip of the principal sulcus to the convergence of the dorsal and ventral sulci) elicited gaze shifts that ranged from primarily vertical to primarily horizontal. For sites where evoked gaze shift CVs were large (>25°) and tilted toward horizontal, there usually was a clear, often prominent, head movement CV that was always horizontal (three, one, and six sites in Fig. 6, A, B, and C, respectively). For ventrolateral sites where evoked gaze shifts were predominantly vertical and/or of small amplitude, there was little if any head movement component (nine, three, and two in Fig. 6, A, B, and C, respectively).
In summary, stimulus durations of 200 ms elicited head movements, but did so reliably only under certain conditions. The major factor appeared to be the size of the evoked gaze shift. If a normal, visually elicited gaze shift of the same amplitude usually included a head movement, the stimulus-evoked gaze shift also did. Initial positions of the eye and the head appeared less important, but did influence whether a head movement was evoked and what amplitude it attained. If the eyes were deviated in the off-direction, a head movement was less likely and usually smaller in amplitude. If the eyes were aimed in the on-direction, however, a head movement was more likely to occur and occasionally to do so in isolation. If the head was deviated away from the direction of the evoked gaze shift before stimulation onset, a head component was more likely to be evoked. As we demonstrate later, the head movement component also depends strongly on stimulus duration.
Comparison of stimulation-evoked and visually guided gaze shifts
Saccades evoked by FEF stimulation with the head restrained have the metrics (velocity and duration) of normal visually guided targeting saccades, suggesting that stimulation appears to engage a physiologically relevant descending command (Robinson and Fuchs 1969). Therefore we asked whether stimulus-evoked gaze shifts also exhibited normal metrics with the head unrestrained. We first compared all movements evoked by 350 Hz and 200-ms stimulus trains at 18 stimulation sites [six sites each from the left and right FEFs of monkey T (about 900 trials) and six sites from monkey P (about 450 trials)] with accurate (gain 0.8 to 1.2) visually guided gaze shifts with similar distributions of directions and initial eye and head positions. Because we observed significant effects of initial positions on evoked movements (e.g., Fig. 4), we also compared stimulus-evoked (all 34 sites, stimulation at 350 and 200 Hz, 200-ms duration) and visually guided gaze shifts initiated with the eyes centered in the head and with the head centered in space.
EYE AND HEAD CONTRIBUTIONS TO GAZE.
For normal visually guided gaze shifts, the relative contribution of eye and head movements to the overall gaze amplitude is rather consistent (Freedman and Sparks 1997b). Similarly consistent relations also characterized the visually guided gaze shifts in our study. Figure 7 plots the mean vectorial eye, head, and head contribution movement amplitudes (±SD; calculated for each 10° gaze amplitude bin) versus the gaze movement amplitude for visually guided movements (open symbols). The eye and head movement contributions to stimulus-evoked gaze movements (filled symbols) also were very reproducible across gaze movement amplitudes. They did not differ statistically from the movement components of visually evoked gaze shifts at any gaze movement amplitude for either monkey when initial gaze positions were allowed to vary (Fig. 7, A–C and G–I). For both visually and stimulus-evoked gaze shifts, head movements contributed to gaze movement at gaze amplitudes 
20°. When movements elicited from only central initial eye and head positions (IEPh and IEPv 10°, and IHPh and IHPv 5° of center) were compared, the mean amplitudes of stimulus-evoked movements of the first animal were similar, but statistically smaller than those of visually guided movements (Fig. 7, D–F; "*," Wilcoxon signed-rank test, P < 0.05); those of the second animal were mostly equivalent (Fig. 7, J–L).
VELOCITY–AMPLITUDE RELATION. The peak polar (angular) velocity of stimulus-evoked gaze, eye, and head movements varied with movement amplitude for both monkey T (Fig. 8, A–C) and monkey P (Fig. 8, G–I). For each monkey, the mean peak polar gaze and eye velocity of stimulus-evoked movements (solid squares) increased sharply for small-movement amplitudes and attained a soft saturation between 30 and 40°. When movements elicited from all initial positions were considered, mean peak polar head velocity increased linearly with head movement amplitude. For monkey T (Fig. 8, A–C), the shapes of the peak velocity versus amplitude relations for visually evoked movements (open squares) resembled those for stimulus-evoked movements; however, the peak velocities of the visually evoked gaze and eye movements were statistically lower across nearly all amplitudes ("*," paired t-test, P < 0.05). For monkey P (Fig. 8, G–I), the mean peak velocities of visually guided gaze and eye movements were statistically higher at small-movement amplitudes and lower at large amplitudes. For both animals, the peak head velocity of visually guided saccades increased linearly with head movement amplitude as it did for stimulus-evoked movements, but attained a lower peak velocity at most amplitudes. For a given gaze amplitude, movements of monkey T elicited from central initial positions had lower mean polar peak velocities than movements elicited from all initial positions, although these means were still higher than those of visually guided movements (Fig. 8, D–F, "*," Wilcoxon signed-rank test, P < 0.05). The mean peak velocities of stimulus-evoked movements of monkey P evoked from central initial positions were not different from those of visually guided movements across most gaze movement amplitudes (Fig. 8, J–L).
DURATION–AMPLITUDE RELATION. Movement duration increased linearly with gaze, eye, and head duration for both stimulus-evoked (solid symbols) and visually evoked (open symbols) gaze shifts in both animals, T and P (Fig. 9). For monkey T, the gaze and eye durations of evoked saccades were not statistically different from those of comparable visually evoked gaze shifts except at one amplitude, but the durations of stimulus-evoked head movements were statistically shorter at all amplitudes (Fig. 9, A–C, "*," paired t-test, P < 0.05). For monkey P, gaze and eye movement durations for stimulus-evoked and visually evoked movements were statistically equivalent until movement amplitudes exceeded 50 and 30°, respectively (Fig. 9, G and H). Durations of stimulus- and visually evoked head movements were mostly equivalent, except for gaze amplitudes between 30 and 50° (Fig. 9I). For movements elicited from only central positions across all 34 sites, durations of stimulus-evoked movements were lower for animal T (Fig. 9, D–F, "*," Wilcoxon signed-rank test, P < 0.05), but mostly equivalent for animal P (Fig. 9, J–L).
Effect of increased stimulus duration
Because many head movements for large natural gaze shifts have durations of 300 ms, we examined whether increasing stimulus duration would increase the size of gaze shifts still farther and therefore make a substantial head movement more likely. Therefore at 26 sites, where we had stimulated with durations of 200 ms, we also varied stimulus duration to a maximum of 800 ms. Figure 10 shows examples of the gaze shifts evoked by stimulus durations of 200 (left column) and 800 ms (right column) at one of these sites. The longer stimulus train elicited a "staircase" of gaze shifts and eye movement components with a total change in gaze movement during stimulation of almost 200°. The first "step" of the evoked gaze shift (and the eye movement) increased in amplitude and duration with the increased stimulus duration (cf. Fig. 10, A and B). For example, the amplitude of gaze shifts evoked at this site by stimulus durations of 200 ms and from similar initial positions ranged from 35 (Fig. 10A) to 86°, whereas the first gaze shift elicited by the 800-ms train was about 90° (Fig. 10B). The effect of increased stimulus duration was even greater on the head movement component, however. The 200-ms train elicited horizontal head movement components of no more than 40° (about 20° in Fig. 10A), whereas a stimulus duration of 800 ms elicited a head movement of 115° (Fig. 10B). Although several gaze shifts occurred during the 800-ms train, the head component appeared to be a single large movement because there was no sign of multiple movements in the smooth head velocity trace (Fig. 10D). Similar patterns of gaze movement staircases and large, smooth head movements were elicited at 19 of 26 sites where we used long-duration-stimulus trains.
The data from the site illustrated in Fig. 10 suggest that the amplitude and duration of stimulus-evoked gaze, eye, and head movements depend on stimulus duration and that this effect is independent of, and in addition to, initial position effects. We next tested the effect of stimulus duration on the key metrics of gaze shifts across several stimulation sites by calculating the stimulus-evoked gaze, eye, and head movement CVs, characteristic durations (CDs), and characteristic peak velocities (CPVs) because these measures account for varying initial position. We examined data from 10 dorsomedial FEF sites for which there was a head contribution CV >1° and for which two to five different stimulus durations (70 to 800 ms) had been applied, yielding a data set of 31 points (i.e., each site contributed a datum for each stimulus duration applied at that site). Only the first gaze and eye movement and the single head movement were considered, and the CDs and CPVs were calculated in the same manner as were CVs (see METHODS). Figure 11, A–C demonstrates that the CVs of gaze, eye, and head movements increased linearly with increasing stimulus duration. However, the relation with the head movement was more robust (r2 = 0.55, slope of 3°/100 ms; P < 0.0005) than with either gaze or eye CVs (r2 = 0.30 and 0.19, respectively; equal correlations test, P < 0.05). As with amplitude, CDs of evoked gaze, eye, and head movements increased linearly with stimulus duration (Fig. 11, D–F). Again, the CD of the elicited head movement was much more strongly related to stimulus duration (r2 = 0.81, slope of 59 ms/100 ms; P < 0.0005) than were either gaze or eye CD (r2 = 0.43 and 0.45, respectively; equal correlations test, P < 0.05). Note too, that for stimulus durations >200 ms, head CDs exceeded those of gaze. Finally, head CPV (Fig. 11, G–I) increased significantly with stimulus duration (r2 = 0.39, slope of (9° s–1)/100 ms, P < 0.0005) but gaze and eye CPVs did not.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
; Stryker and Schiller 1975; Tomlinson and Bahra 1986a). Taken together, these data demonstrate that electrical microstimulation of the dorsomedial FEF can produce a combined eye and head gaze command. Moreover, this command drives eye and head movements according to initial eye positions. Comparison with other studies
Three previous studies examined the effects of FEF stimulation in monkeys that were free to turn their heads (Chen 2006; Sparks et al. 2001; Tu and Keating 2000). Our results are in general agreement with those of Tu and Keating (2000) who also evoked head movements in conjunction with eye saccades at most dorsomedial FEF sites. Moreover, the latencies and amplitudes of their evoked eye–head gaze shifts, the strong effect of initial eye position on eye movement, and the presence of evoked head-only movements are in agreement with our results. However, it is unclear whether the head movements these authors evoked were from sites within the low-threshold FEF because they required stimulus currents of 200 µA to evoke large gaze shifts with head movement (and did not identify which sites required the higher stimulus intensity). We evoked gaze shifts with head movement—and a substantial head contribution to gaze—with currents <100 (2x TH of 50) µA, the same maximum intensity used to define the low-threshold FEF (Bruce et al. 1985). In contrast, when Sparks et al. (2001) and Chen (2006) stimulated the dorsomedial FEF, they evoked little, if any, head movement that, if present, began largely after the gaze shift; peak head velocity occurred on average 100 ms after the end of the gaze shift (Chen 2006). These investigators therefore concluded that the FEF does not encode a combined eye–head gaze shift, but only eye saccades.
It is unclear why these latter investigators observed only small, late head movements and no head contribution to gaze. The absence of an evoked head movement was not the result of a failure to explore different initial positions because they deliberately had trained the monkeys to align their heads and eyes separately to a range of different target positions before stimulation—a method also used by Tu and Keating (2000). Chen (2006) used stimulus durations (300–500 ms) and frequencies (100–200 Hz) similar to those used here and by Tu and Keating. Despite these similarities, Chen reports mean latencies to gaze movement onset and the variability around those means (81 ± 48 or 91 ± 50 ms for different behavioral tasks), which are greater than those of head-restrained reports using 350 Hz (30 to 45 ms; Bruce et al. 1985; Russo and Bruce 1993), those of Tu and Keating (47 ± 2.4-ms latency at 250 Hz; Tu and Keating 2000), and those presented here (31 ± 12 ms at 350 Hz; 50 ± 17 ms at 200 Hz).
Perhaps we had more success at eliciting head movements with low-intensity FEF stimulation because we tried to optimize the occurrence of a head movement in several ways. First, we explored the entire FEF region, especially those regions where stimulation with the head restrained evoked the largest gaze shifts (Bruce et al. 1985; Robinson and Fuchs 1969), because only large (20°), mostly horizontal, naturally occurring gaze shifts have a head movement component (Freedman and Sparks 1997b; Phillips et al. 1995; Tomlinson and Bahra 1986a). Second, we stimulated from a variety of naturally occurring initial eye and head positions that were attained as the animal fixated a single target LED at various starting locations. Training animals to separately align eye and head positions, as was done in these previous studies, may have modified or suppressed the generation of more natural head movements. Third, unlike the previous studies, we stimulated only at those sites that appeared to have single-unit (22 of 34 sites) or multiunit activity (12 of 34) related to head-unrestrained gaze shifts.
Effect of stimulus parameters: stimulus-evoked versus normal visually evoked gaze shifts
We report stimulus-evoked gaze, eye, and head movements with higher peak velocities (Fig. 8) and shorter durations (Fig. 9) than those of natural gaze shifts when movements evoked with 350-Hz stimulation from varied initial positions were considered. Because we had observed strong initial eye position effects on movement amplitude, we also compared movements elicited from only central initial eye and head positions (with both 350- and 200-Hz stimulation). The differences in kinematic relations between the stimulus-evoked and visually guided movements remained for the first animal but were reduced for the second animal. These differences were not attributable to an underlying difference in our animals’ behavior because the velocity–amplitude relations for visually guided gaze shifts (with central or varied initial positions) were not different from those of previous studies (Freedman and Sparks 1997b; Phillips et al. 1995; Tomlinson and Bahra 1986a).
Instead, these differences in kinematics may be explained by the frequency of our microstimulation. To meet the criterion of stimulating within the low-threshold FEF, and to allow comparison with previous studies with the head restrained (Bruce et al. 1985; Russo and Bruce 1993), we started the investigation of each site with stimulation at 350 Hz for 70 ms. More than half of the sites here were tested at 200-ms-stimulus duration and 350 Hz, and the comparisons of kinematics across all initial positions examined only movements evoked with 350-Hz stimulation. Thus this relatively high frequency of stimulation may have elevated movement velocities. A comparison of the effects of different stimulus frequencies at the SC shows that higher frequency stimulation (500 Hz) results in evoked gaze peak velocities (similar to those of monkey T) greater than those evoked with 250-Hz stimulus and those of visually guided movements (Freedman et al. 1996). Unfortunately, we did not systematically apply 200-ms stimulus trains at 200 and 350 Hz at enough sites to allow comparison of peak velocity–amplitude relations across different stimulus frequencies. Interestingly, the peak velocities of our stimulus-evoked gaze and eye movements were not different from those reported by Chen (2006).
Although our electrical microstimulation produced movements that were somewhat faster than normal, these movements were within a range observed by other investigations of head-unrestrained gaze shifts. Moreover, these movements were correctly coordinated to produce the same relative contributions of eye and head component amplitudes seen in normal, visually guided gaze shifts. Thus our results of FEF microstimulation may indeed reflect the underlying physiology and reveal a previously undisclosed role for the FEF: the coordination of the eye and head contributions of normal, large head-free gaze shifts.
FEF contribution to gaze programming is modulated by initial position
Stimulation of the FEF with the head unrestrained demonstrated a major dependency of the evoked movements on the initial position of the eye in the head. This dependency was present for at least half the sites where evoked gaze and eye movements were between goal-directed and fixed-vector types (Figs. 2, 4, and 5). We found no clear dependency of gaze, eye, or head movements on initial head position, but head-only movements were occasionally evoked from eccentric "on-direction" eye positions at several sites. We therefore suggest that the motor output of the FEF is a gaze command that may generate head movement coordinated with eye movement. By "coordinated," we do not mean that each eye movement is always coupled with a head movement, as it is for SEF stimulation (Martinez-Trujillo et al. 2003). Instead, depending on eye-in-head position, "coordinated" recruitment of eye and head movement may call for eye-only gaze movements, head-only gaze stabilization with eye counterrotation, or combined eye–head gaze movements.
In contrast to our data, previous studies report that initial eye/gaze position has little effect on stimulus-evoked eye movements with the head restrained. In those studies, evoked saccades were described as "fixed vector" with roughly the same amplitude and direction irrespective of initial eye position (Robinson and Fuchs 1969; Russo and Bruce 1993). However, our results are consistent with results of studies of head-unrestrained stimulation of the FEF, SC, and SEF. Those studies demonstrated that under more natural conditions (when the head is free to turn), evoked eye movements are indeed strongly dependent on initial eye position (Freedman et al. 1996; Martinez-Trujillo et al. 2003; Tu and Keating 2000). Although our mean index of initial eye position effect on eye movement (–0.35) is comparable to the –0.46 reported by Tu and Keating (2000), they found an effect of initial head position on head movements and little or no effect of initial gaze position on gaze movements. However, those authors neither used MLR to protect against effects of intercorrelations with initial eye position nor reported statistical significance. We therefore feel that our results, which demonstrate no significant effect of initial head position on movement amplitude, more accurately represent the effect of FEF stimulation. Taken together, our data are consistent with the conclusions of others that the SC, SEF, and FEF issue a gaze command, although we add that the gaze command from the FEF is affected by initial eye position.
The coordination of an eye saccade with a head movement to accurately shift gaze to a target at a large angular distance requires sophisticated neuronal processing. An early model proposed by Bizzi and colleagues suggested that a gaze shift is modified by head movement signals transmitted through the VOR (Bizzi et al. 1971; Morasso et al. 1973). However, recent behavioral evidence suggests that the VOR is suppressed during active gaze shifts (Laurutis and Robinson 1986; Tomlinson and Bahra 1986b). Moreover, the interneuron of the VOR is turned off during head-fixed and head-free gaze shifts (Fuchs et al. 2005; McCrea and Gdowski 2003; Scudder et al. 2002). A more recent view suggests that the eye movement component of a gaze shift is influenced by an active head movement signal (such as an efferent copy of the command for head movement). However, there is disagreement as to whether gaze is controlled directly or eye and head movements are controlled independently. Our results suggest that the FEF could provide a unified eye–head gaze command (Scudder et al. 2002).
Observations of head-unrestrained gaze shifts made by animals
, gaze; 
, head). Note the "reversal" of gaze and eye movements from opposite IPs. B: a similar plot for a site representative of 3 nonlinear sites that did not show "reversal." Gaze and eye movements from left (negative) IPs have a large horizontal component, whereas those from right IPs have little or none. Abscissae and ordinates as in 
), eye (
