INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The supplementary eye field (SEF), an area located on the dorsomedial shoulder of the frontal lobe in macaque monkeys, has been thought since its discovery 10 years ago to serve oculomotor functions (Schlag and Schlag-Rey 1985, 1987). This view has been supported by studies demonstrating that electrical stimulation of the SEF at reasonably low currents (<50 µA) evokes saccadic eye movements (Chen and Wise 1995b; Fujii et al. 1995; Lee and Tehovnik 1995; Mann et al. 1988; Mitz and Goldschalk 1989; Russo and Bruce 1993; Tehovnik and Lee 1993; Tehovnik and Sommer 1997; Tehovnik et al. 1994; Tian and Lynch 1995) and that neurons in the SEF fire during the preparation and execution of saccades, exhibiting selectivity for particular saccade directions (Bon and Lucchetti 1992; Chen and Wise 1995a,b, 1996, 1997; Hanes et al. 1995; Mann et al. 1988; Mushiake et al. 1996; Russo and Bruce 1996; Schall 1991a,b; Schlag and Schlag-Rey 1985, 1987; Schlag-Rey et al. 1997). However, the contributions of the SEF to oculomotor control probably are not as straightforward as those of the other major frontal oculomotor area, the frontal eye field (FEF). In the SEF, more frequently than in the FEF, neuronal activity varies across the course of learning as monkeys acquire arbitrary associations between visual patterns and eye-movement directions (Chen and Wise 1995b). Further, around half of SEF neurons, unlike neurons in the FEF, fire differentially during combined movements of the arm and eye as compared with eye movements alone (Mushiake et al. 1996). Finally, higher levels of electrical current must be delivered to the SEF than to the FEF to elicit saccades (Russo and Bruce 1993; Tehovnik and Sommer 1997). These observations suggest that the SEF is removed farther than the FEF from processes occurring at the oculomotor periphery and that its functions, while encompassing oculomotor control, may not be restricted to it.

A potentially valuable approach to understanding the functions of the SEF is to characterize the spatial reference frames with respect to which it operates. This requires answering the question: insofar as specific sites or neurons in the SEF represent particular eye-movement directions, with respect to what reference frame are these directions specified? Studies carried out to date have yielded evidence for three forms of spatial sensitivity in the SEF: eye-centered, head-centered, and object-centered. 1) Evidence that the SEF encodes directions relative to an oculocentric reference frame arose from studies based on both electrical-stimulation and single-neuron recording. Electrical-stimulation studies demonstrated that fixed vector saccades (saccades having a particular size and direction regardless of the eyes' starting point in the orbit) could be elicited from certain sites in the SEF (Bon and Lucchetti 1992; Mitz and Godschalk 1989; Russo and Bruce 1993; Schlag and Schlag-Rey 1987). Likewise, some SEF neurons were shown to fire in conjunction with saccades in preferred directions regardless of the eyes' starting point (Mitz and Godschalk 1989; Russo and Bruce 1996; Schlag and Schlag-Rey 1987). 2) There are also signs that the SEF encodes directions relative to a craniocentric frame. The fact that electrical stimulation at some sites in the SEF seems to elicit goal-directed saccades, driving the eyes to a certain angle in the orbit regardless of initial direction, has been taken by some as evidence for craniocentric encoding (Tehovnik 1995; Tehovnik and Lee 1993; Tehovnik et al. 1994), although others have interpreted this phenomenon as arising from failure of SEF stimulation to engage cerebellar mechanisms that correct for variations in ocular mechanics across orbital position (Russo and Bruce 1993). At the level of single-neuron recording, some SEF neurons have been shown to possess craniocentric gaze fields, firing as a function of the angle of the eyes in the head during motivated fixation of external targets (Bon and Lucchetti 1990, 1992; Lee and Tehovnik 1995; Schlag et al. 1992). 3) Finally, studies carried out in our laboratory during the last several years have indicated that some SEF neurons are sensitive to the allocentric directions of eye movementsdirections as defined with respect to objects in the external world. In monkeys planning and executing eye movements to the left or right end of a horizontal bar, around half of SEF neurons fire differentially on bar-left and bar-right trials even when the location of the bar on the screen is manipulated so as to keep the location of the target on the screen the same (Olson and Gettner 1995, 1999). The object-centered spatial selectivity of these neurons suggests that they are involved in eye-movement control at the level of target specification rather than of motor programming.

The aim of the experiment described here was to extend our understanding of object-centered direction selectivity in the SEF by answering the question does this phenomenon depend on the nature of the reference image and, in particular, on its physical continuity? In previous studies, monkeys were required to make eye movements to the left or right end of only a single image, a physically continuous horizontal bar. Here we trained monkeys to perform a task in which, on interleaved trials, they had to make eye movements to the right or left end of a bar, as in the previous experiments, or, alternatively, to the right or left element in an array consisting of two horizontally separated dots. We recorded from SEF neurons during performance of this task to determine whether firing was different under bar and dot conditions. We found only subtle differences in neuronal activity across the two conditions. This result suggests that SEF neurons carry comparatively pure object-centered spatial signalssignals that reflect the location of the target with respect to the selected reference image but are not influenced to a major degree by the reference image's intrinsic properties.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects

Two adult male rhesus monkeys were used (Macaca mulatta; laboratory designations Ju and Po). Experimental procedures were approved by the Carnegie Mellon University Animal Care and Use Committee and were in compliance with the guidelines set forth in the United States Public Health Service Guide for the Care and Use of Laboratory Animals.

Preparatory surgery

At the outset of the training period, each monkey underwent sterile surgery under general anesthesia maintained with isofluorane inhalation. The top of the skull was exposed, bone screws were inserted around the perimeter of the exposed area, a continuous cap of rapidly hardening acrylic was laid down so as to cover the skull and embed the heads of the screws, a head-restraint bar was embedded in the cap, and scleral search coils were implanted on the eyes with the leads directed subcutaneously to plugs on the acrylic cap (Remmel 1984; Robinson 1963). After initial training, a 2-cm-diam disk of acrylic and skull, centered on the midline of the brain approximately at anterior 23 mm (Horsley-Clarke coordinates), was removed, and a cylindrical recording chamber was cemented into the hole with its base just above the exposed dural membrane.

Single-neuron recording

At the beginning of each day's session, a varnish-coated tungsten microelectrode with an initial impedance of several megohms at 1 kHz (Frederick Haer and Company, Bowdoinham, ME) was advanced vertically through the dura into the immediately underlying cortex. The electrode could be placed reproducibly at points forming a square grid with 1 mm spacing (Crist et al. 1988). The action potentials of a single neuron were isolated from the multineuronal trace by means of an on-line spike-sorting system using a template matching algorithm (Signal Processing Systems, Prospect, Australia). The spike-sorting system, on detection of an action potential, generated a pulse the time of which was stored with 1-ms resolution.

Behavioral apparatus

All aspects of the behavioral experiment, including presentation of stimuli, monitoring of eye movements, monitoring of neuronal activity, and delivery of reward, were under the control of a 486- or pentium-based computer running Cortex software provided by R. Desimone, Laboratory of Neuropsychology, National Institute of Mental Health. Eye position was monitored by means of a scleral search coil system (Remmel Labs, Ashland, MA, or Riverbend Instruments, Birmingham, AL) and the x and y coordinates of eye position were stored with 10-ms resolution. Stimuli generated by an active matrix LCD projector (Sharp, XG H4OU) were rear-projected on a frontoparallel screen 25 cm from the monkey's eyes. Reward in the form of ~0.1 ml of water or juice was delivered through a spigot under control of a solenoid valve on successful completion of each trial.

ANOVA and t-test analysis of data from individual neurons

Details of statistical analysis are provided in the text. The general approach was to analyze results obtained with a given behavioral paradigm by applying a set of identical procedures to data collected from each neuron. The trial epoch under consideration was defined as the period between two identifiable events. The mean firing rate during the epoch was computed for each trial completed successfully during recording from the neuron. Then an ANOVA or t-test was carried out to determine whether firing rate varied significantly across the trials as a function of the conditions by which trials differed from each other.

² analysis of population data

A population of neurons might exhibit trait a or b in one context and trait x or y in another context. For example, among neurons significantly selective for horizontal direction in two tasks, each neuron might prefer right (a) or left (b) in the first task and right (x) or left (y) in the second task. In such cases, to test whether the distribution of neurons with respect to a and b was correlated with the distribution with respect to x and y, we employed the following procedure. We took as observed values the four counts Oax, Oay, Obx, and Oby, where Oax was the number of neurons observed to express trait a in the first context and trait x in the second context, and so on for Oay, Obx, and Oby. We then computed the sum of the counts, S = Oax + Oay + Obx + Oby, and the four frequencies, Fa = (Oax + Oay)/S, Fb = (Obx + Oby)/S, Fx = (Oax + Obx)/S, and Fy = (Oay + Oby)/S. Then on the assumption that the distribution of neurons with respect to a and b was uncorrelated with the distribution with respect to x and y, we computed the four expected counts: Eax = Fa*Fx*S, Eay = Fa*Fy*S, Ebx = Fb*Fx*S, and Eby = Fb*Fy*S. Finally, we used a ² test with 1 df to determine the level of significance of the deviation of the observed values (Oax, Oay, Obx, and Oby) from the expected values (Eax, Eay, Ebx, and Eby).

Localization of recording sites

In each monkey, recording was carried out in a pair of regions, each a few mm in extent, disposed approximately symmetrically across the interhemispheric midline. One of the monkeys (Po) is still under study in behavioral experiments. In the other monkey (Ju) the brain was photographed after it was killed with an overdose of pentobarbital sodium and transcardiac perfusion with 10% formalin. Marks indicating the location of the recording chamber were compared with gross anatomic landmarks including the hemispheric midline and the arcuate and principal sulci. On the basis of the grid coordinates at which the electrode had been placed, recording sites then were projected onto the image of the cortical surface.

Bar-dot task

Both monkeys were trained to perform a task requiring them to make eye movements to one end or the other of a reference image which could be physically continuous or discontinuous. Essential features of the task are summarized in Fig. 1, A and B. At the beginning of each trial, while the monkey was fixating a central spot, a sample was presented, either in the form of a solid horizontal bar (Fig. 1A2) or in the form of a pair of dots corresponding to the ends of a virtual horizontal bar (Fig. 1B2). Then one end of the sample was cued (3). After a delay, a target appeared, identical to the sample in form but not necessarily in location (5). After a second delay, extinction of the central fixation spot (7) signaled the monkey to make an eye movement (8). If the monkey made a saccade directly to the end of the target corresponding to the cued end of the sample, then 100 ms after target-attainment, a white spot came on at the now-fixated target location, thus providing positive feedback. However, the monkey was required to maintain fixation on the target for an additional variable period (300-450 ms). Only at the end of this period was the display extinguished and reward delivered. These postattainment steps were introduced to prevent the monkey from following up the first saccade with a second one to some other part of the display. Any such behavior would have led to interpretational difficulties inasmuch as activity around the time of the first saccade might have been related to programming of the second one. To perform this task successfully, monkeys had to perceive and remember the location of the cue relative to a reference frame centered on the sample and had to do so regardless of whether the sample was a continuous horizontal bar or a pair of dots forming a horizontal array.

View larger version (44K): [in this window] [in a new window]   Fig. 1. Bar-dot task. A, 1-8: screen in front of the monkey during successive epochs of a single representative trial in which the reference image was a continuous bar. Center of each hatched circle indicates the monkey's direction of gaze during the corresponding trial epoch and the arrow indicates the direction of the eye movement. All other items are patterns visible to the monkey. 1: a white fixation spot appeared at the center of the screen and the monkey achieved foveal fixation. 2: a horizontal sample bar appeared in the superior visual field. 3: a white cue flashed on 1 end of the sample bar. 4: during an ensuing delay period of variable length, the monkey maintained central fixation. 5: the target bar appeared. 6: during a 2nd delay period, the monkey continued to maintain central fixation. 7: offset of the fixation spot signaled the monkey to initiate an eye movement. 8: monkey was required to respond by making an eye movement directly to the end of the target bar corresponding to the cued end of the sample bar. After foveation of the target, the following events occurred (not shown): 100 ms elapsed; then a 0.8 × 0.8° white feedback spot came on at the target location, to provide positive feedback; then a random interval in the range 300-450 ms elapsed; then the entire display was extinguished and reward was delivered. B: equivalent display for a trial in which the reference image was a pair of dots marking the ends of a virtual bar. C: factors varying across bar trials included the location of the 3.1 × 0.2° blue sample bar (a or b), the location of the 0.8 × 0.8° white cue (c, d, or e), the location of the 3.1 × 0.2° blue target bar (f, g, or h), and the direction of the required eye movement (1, 2, 3, or 4). Dot trials were identical with the exception that 2-dot arrays were presented at the locations where bars would have been. Each array consisted of two 0.4 × 0.4° white dots separated horizontally by a 3.1° center-to-center offset. All stimuli were presented at the same elevation (5.8°) relative to the fixation spot (F). Vertical offset of a-e from f-h in this figure is simply a convention adopted for clarity. D: these tables summarize the features defining 12 bar conditions and 12 corresponding dot conditions.

Individual trials differed not only with respect to the nature of the reference image (bar or dots), but also with respect to the location of the sample (Fig. 1C: a or b; each bar indicates the location at which a bar or array could appear), the cued end of the sample (right or left), and the location of the target (Fig. 1C: f, g, or h; each bar indicates the location at which a bar or array could appear). Systematic variation in these factors gave rise to 24 conditions summarized in Fig. 1D. Trials corresponding to these 24 conditions were interleaved pseudorandomly according to the rule that one trial of each type had to be completed successfully before initiation of the next block. An essential feature of this design was the dissociation of relative location (the right or left end of the bar or array) from certain other factors that might influence neuronal activity in the SEF, notably the location of the cue on the screen (and thus the location of its image on the retina) and the screen location of the target. A cue at one screen location (Fig. 1C: d) could mark either the right end of a left-displaced sample (Fig. 1C: b) or the left end of a right-displaced sample (Fig. 1C: a). Similarly a target at one screen location (Fig. 1C: 3) might be either the right side of a left-displaced target (Fig. 1C: h) or the left side of a right-displaced target (Fig. 1C: g).

It may be noted that the location of the sample image was different in this task than in the task employed in previous studies (Olson and Gettner 1995, 1999). In those studies, it was placed to one side of fixation. Here, in contrast, it was presented in the upper visual field where it appeared to the left or right of the midline on interleaved trials. This change was instituted so as to eliminate the asymmetry with respect to the visual field midline inherent in the earlier design. Results obtained with the design used here can be interpreted identically regardless of the recording hemisphere because the task is perfectly symmetric with respect to the visual field and thus with respect to hemispheric representation of visual space.

Memory-guided saccade task

We also recorded neuronal activity during performance of a standard oculomotor test requiring the monkey to make eye movements to targets in the form of small dots located at 9.6° eccentricity above, below, to the right of, and to the left of the fixation point. The main stages of a single representative trial lasting ~1.5 s are summarized in Fig. 13, A-F. The staggered panels in this figure represent the display on the screen in front of the monkey during successive stages of the trial. In each panel, a circle indicates the monkey's direction of gaze. While the monkey maintained fixation on a central spot (A), four potential targets were presented (B) and one of the targets was cued (C). The monkey then was required to maintain central fixation during a delay period (D) at the end of which the fixation spot was extinguished (E), whereupon the monkey had to make an eye movement rapidly and directly to the previously cued target (F). If the monkey made a saccade directly to the target, then 100 ms after target-attainment, the now fixated target increased in size, thus providing positive feedback. However, the monkey was required to maintain fixation on the target for an additional variable period (300-450 ms) before reward was delivered. Trials were imposed in pseudorandom sequence according to the rule that the monkey had to complete successfully one trial in each direction before moving on to the next block. Data collection continued until ~16 successful trials conforming to each of the four conditions had been completed.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Task performance

Both monkeys learned to perform the bar-dot task at a level well above chance, and each experienced moderately more difficulty on dot than on bar trials. Monkey Ju scored 99.9% on bar trials as compared with 99.6% on dot trials (averages computed across all neuronal data collection sessions; consideration restricted to trials in which the monkey made an eye movement to one end or the other of the target). The difference between the two percent-correct scores, although only a fraction of a percent, was significant (2-tailed paired t-test, P = 0.03). Monkey Po scored 95.9 and 89.4% on bar and dot trials, respectively; these values differed at a high level of significance (P < .0001).

The behavioral reaction time (the interval between offset of the fixation spot and initiation of the saccadic eye movement) also was measured as a function of cue condition in each monkey. In monkey Ju, there was a minor but significant (2-tailed paired t-test, P = 0.004) tendency for reaction times to be longer on bar than on dot trials (150 vs. 148 ms). The same tendency was present and significant in monkey Po (164 vs. 160 ms; P = 0.008). Decision time was not a factor in this effect because a long delay intervened between the instructional cue and the imperative signal. Perhaps it was related to subtle differences in the eye movements executed on bar and dot trials, as described in the following text.

Recording sites

Our approach in selecting recording sites was to record from neurons at the rough location of the SEF, as estimated on the basis of stereotaxic coordinates and, having identified sites at which there was robust eye-movement-related activity, to record from these sites and then move out from them to adjacent sites over successive recording sessions. At each site, we recorded from neurons located in the superficial cortex, remaining within the initially encountered gray matter and never passing through white matter into buried cortex. The mean recording depth (as measured relative to the level at which neural activity first was detected) was 875 ± 448 µm (mean ± SD; minimum = 178 µm, maximum = 1,988 µm) in monkey Ju and 781 ± 681 µm (minimum = 0 µm; maximum = 2,724 µm) in monkey Po. In the context of the bar-dot task, we characterized a total of 77 neurons from monkey Ju (17 and 60 in the left and right hemispheres, respectively) and 66 neurons from monkey Po (29 and 37 in the left and right hemispheres, respectively). The tangential distribution of bar-dot recording sites in monkey Ju is shown in Fig. 2A, where each dot represents one site and the size of the dot indicates how many neurons at that site contributed data to the present paper. Monkey Po is still under behavioral study, therefore it is not possible to describe precisely the relation of the recording sites to gross morphological landmarks.

View larger version (16K): [in this window] [in a new window]   Fig. 2. A: recording sites superimposed on a dorsal view of the cerebral hemispheres of monkey Ju. Each dot indicates a recording site. Area of each dot is proportional to the number of neurons at that site contributing data to the present paper. Largest dot, in the right hemisphere, represents 14 neurons, whereas the smallest dots represent one neuron each. B: extent of the supplementary eye field (SEF) as defined by mapping with intracortical microstimulation in studies from 10 laboratories listed in Table 1 of Tehovnik (1995). Tehovnik brought results from different laboratories into register by use of two landmarks: for mediolateral register, the hemispheric midline and, for anteroposterior register, the genu of the arcuate sulcus, as marked by the horizontal line spanning A and B in this figure. For each study listed by Tehovnik, we generated a rectangle encompassing, to the nearest mm, the anterior, posterior, lateral and medial limits of the region in which electrical stimulation elicited eye movements. Then, on each point in a 1 × 1 mm grid spanning the cortex, we superimposed a dot the area of which was proportional to the number of rectangles including that point. Number of times a site was counted could range from 0 (not marked on this figure: sites not implicated by any study) through 1 (smallest dots visible in this figure: sites implicated by just 1 study) to 10 (largest dots visible in this figure: sites implicated by all 10 studies). Note that sites in which recording was carried out in this study (dots in A) overlap the relatively anterior region in which electrical stimulation in previous studies most consistently elicited eye movements (largest dots in B). as, arcuate sulcus; as, genu, genu of the arcuate sulcus; cs, central sulcus; ps, principal sulcus.

Because we almost immediately located sites of oculomotor activity in each hemisphere of each monkey, we had no occasion to carry out extensive mapping, defining the borders of the SEF or identifying adjacent regions such as the supplementary motor area. Accordingly, it is reasonable to ask whether the sites from which we recorded were indeed in the SEF. To answer this question, we compared our recording sites to maps of the SEF generated in previous studies as summarized by Tehovnik (1995). Table 1 of Tehovnik's review summarizes the results of 10 studies in which electrical stimulation was used to map out the SEF, indicating, for each study, the area's mediolateral extent (ML, defined relative to the interhemispheric midline) and anterior-posterior extent (AP, defined relative to the genu of the arcuate sulcus). These results are translated, in Fig. 2B, into a graph in which the area of each dot corresponds to the fraction of the 10 studies in which electrical stimulation at the dot's location elicited eye movements (the dots in Fig. 2B range in area from 1only one study reported elicitation of eye movements by stimulation at that location to 10all 10 studies reported a positive result). Loci at which electrical stimulation elicited eye movements in a large number of studies are marked by a cluster of large dots extending 3-7 mm anterior to the level of the genu of the arcuate sulcus. We may now compare the recording sites in monkey Ju to sites of electrical stimulation in these studies. Recording sites in monkey Ju extended 4-9 mm anterior to the genu of the arcuate sulcus (Fig. 2A: as, genu), with an average of ~6 mm. We conclude that recording sites in monkey Ju were toward the front of the cortical territory in which electrical stimulation has been reported to elicit eye movements and that they overlapped the part of this territory in which electrically induced eye movements have been obtained with greatest frequency. Recording sites in monkey Ju also overlapped the SEF as identified by electrical stimulation in later studies not considered by Tehovnik. Chen and Wise (1995b, Fig. 8A) show sites positive for elicitation of eye movements as extending 2-6 mm anterior to the genu, whereas Fujii et al. (1995, Fig. 1) show such sites at levels 1-8 mm anterior to the genu. Finally, it should be noted that recording sites in monkey Ju do not overlap the zone rostral to the SEF in which Bon and Lucchetti (1994) have described electrical stimulation as eliciting ear movements. This zone extends ~10-14 mm anterior to the genu (Bon and Lucchetti 1994, Fig. 2A). This set of comparisons, although not as conclusive as electrical stimulation mapping carried out in the same monkey and although limited by the accuracy with which gross morphological landmarks can be identified in published figures, nevertheless suggests strongly that recording sites in this study were confined to the SEF.

Object-centered direction selectivity

We will refer to a neuron as exhibiting object-centered direction selectivity if it fired at different rates on trials requiring an eye movement to the left versus the right end of a reference image even when the retinal location of the cue and the location of the target on the screen were held constant across trials. An example of a neuron exhibiting strong object-centered direction selectivity under both bar and dot conditions is shown in Fig. 3. During delay 1, the period between presentation of the cue and onset of the target bar, this neuron's rate of firing was markedly higher on trials in which the right side of the image had been cued (Fig. 3, C and D) than on those in which the left side had been cued (Fig. 3, A and B) regardless of whether the image was a bar (Fig. 3, B and D) or a pair of dots (Fig. 3, A and C). This difference in level of activity cannot have resulted from any difference in the retinal location of the cue because, under all four illustrated conditions, the cue was at the same location, directly above fixation.

View larger version (28K): [in this window] [in a new window]   Fig. 3. Data from a neuron selective, during delay 1, for those conditions in which the right end of the sample image had been cued. Each histogram, with accompanying raster display, represents neuronal activity under a single set of conditions defined with respect to type of sample image (bar or dot), location of sample image (right or left on screen), and location of cue (right or left on sample). Juxtaposed to each histogram is a panel depicting the sample image and cue in relation to the fixation spot (black dot). Numbers in each panel indicate the trial conditions under which the stimuli were in this configuration (cf. Fig. 1D). Data from successive successfully completed trials were aligned on target-onset. Times of cue onset and trigger onset are indicated as a range because delay 1 and delay 2 were variable. Ticks on the horizontal axis mark 250-ms intervals. A: dot trials with cue presented on the left end of the right-displaced reference image. B: dot trials with cue presented on the right end of the left-displaced reference image. C: bar trials with cue presented on the left end of the right-displaced reference image. D: bar trials with cue presented on the right end of the left-displaced reference image. Note that the location of the cue on the screen is the same across all sets of conditions.

To obtain an objective estimate of the frequency with which neurons exhibited object-centered direction selectivity, we carried out analyses of variance on data collected from each neuron during three trial epochs: delay 1 (from cue onset until target onset), delay 2 (from target-onset until fix-spot-offset) and the movement period (from the initiation of the saccade until 100 ms after its completion). There was a solid rationale for using these epochs, insofar as object-centered signals, if they waxed and waned during a trial, generally did so in the vicinity of the epoch boundaries. Nevertheless the divisions should be viewed as essentially heuristic with full appreciation of the fact that continuous activity might be parsed into multiple epochs (e.g., in the case shown in Fig. 3, where object-centered signals carried over from delay 1 to delay 2). In each analysis, there was one dependent variable (firing rate) and there were two factors: object-centered direction (right or left) and image type (bar or dot). Consideration during delay 1 was restricted to a subset of conditions in which the screen location of the cue was balanced across the two factors (conditions 2, 4, 6, 7, 9, 11, 14, 16, 18, 19, 21, and 23 in Fig. 1D). Consideration during delay 2 and the movement period was restricted to a subset of conditions in which the screen location of the target was balanced across the two factors (conditions 2, 3, 4, 5, 8, 9, 10, 11, 14, 15, 16, 17, 20, 21, 22, and 23 in Fig. 1D). A significance criterion of P < 0.05 was employed. The results, summarized in Table 1, indicate that around half of the tested neurons showed a main effect of object-centered direction during each epoch (65/143 during delay 1, 89/143 during delay 2, and 57/143 during the movement period).

To determine whether neurons exhibiting object-centered direction selectivity were arranged within the recording zone according to any clear global pattern, we computed for each recording site the frequency with which neurons at that site yielded a significant main effect for object-centered direction. Three tests had been carried out on each neuron, assessing activity during delay 1, delay 2, and the movement period. Thus at a cortical site where n neurons had been studied, 3*n tests were carried out. The results of these tests are summarized for each recording site in Fig. 4, A and B. In this figure, the size of each circle indicates the percentage of tests revealing significant selectivity for object-centered direction. Although there was some variation from site to site in the proportion of tests yielding a significant result, there was no clear mediolateral or anterior-posterior trend in the arrangement of sites with a high yield.

View larger version (35K): [in this window] [in a new window]   Fig. 4. Cortical distribution of neurons exhibiting significant object-centered direction selectivity in the bar-dot task for monkey Ju (A) and monkey Po (B). Coordinates are with respect to the center of the recording grid (0,0). Each site at which recording was carried out during performance of the bar-dot task is marked by a circle. The area of each dark circle is proportional to the percentage of tests revealing significant object-centered direction selectivity at the corresponding site. Where n neurons were studied, 3*n tests were carried out (direction selectivity was assessed independently during delay 1, delay 2, and the movement period for each neuron). The largest circles represent cases in which 100% of tests yielded a significant result. Sites at which 0% of tests yielded a significant result are indicated by small open circles. Note that sites marked in A correspond one-to-one to recording sites projected onto the dorsal view of the frontal lobe in Fig. 2A.

It was obvious on casual inspection of the data that neurons exhibiting object-centered direction selectivity under bar conditions also did so under dot conditions (Fig. 3). To assess this effect systematically, we carried out an additional step of analysis. For each recorded neuron during each trial epoch, we computed the directional signal (firing rate under left-side-cued trials minus firing rate on right-side-cued trials) independently for bar and dot conditions, restricting consideration to conditions in which the retinal location of the cue and the screen-location of the target were balanced across object-centered direction. The results are summarized in the graphs of Fig. 5, which plot the directional signal for dot trials, on the vertical axis, against the directional signal for bar trials, on the horizontal axis, with each neuron represented as a single point. The clear positive correlation between directional signals recorded during dot and bar trials (significant at P < .0001 for each monkey during each epoch) indicates that neurons firing more strongly during left-side-cued (or right-side-cued) trials under dot conditions tended to display the same pattern under bar conditions. In monkey Ju, the R² values for delay 1, delay 2, and the movement period were 0.574, 0.584, and 0.405, respectively. In monkey Po, the corresponding values were 0.810, 0.449, and 0.324. 

View larger version (27K): [in this window] [in a new window]   Fig. 5. Scatter plots of object-centered directional activity on dot trials (vertical axis) vs. bar trials (horizontal axis) for all neurons in each monkey. Each point represents 1 neuron. Directional signal was computed by subtracting the mean firing rate on trials when the right end of the reference image had been cued from the mean firing rate on trials when the left end had been cued, with consideration restricted to successfully completed trials within a subset of conditions selected so that the retinal location of the cue and the screen-location of the target were fully balanced across object-centered direction.

A few neurons, although exhibiting object-centered direction selectivity under both bar and dot conditions, nevertheless appeared to fire at different rates under the two conditions or appeared to carry object-centered signals of different strength. In the neuron of Fig. 6, firing during delay 1 was stronger on trials in which the right end of the image had been cued (Fig. 6: C and D vs. A and B). In addition, activity was stronger under bar conditions than under corresponding dot conditions (Fig. 6: B and D vs. A and C). The converse was true of the neuron shown in Fig. 7. During delay 2, after onset of the target and before the signal to respond, this neuron fired more strongly on trials when the right end of the reference image was the target (Fig. 7: E-H vs. A-D). However, its activity differed across dot and bar trials. On dot trials, it fired more strongly and showed an enhancement of the object-centered directional signal (the difference in firing rate between conditions in which the left or right end of the reference image had been cued). The strength of the directional signal can be estimated in Fig. 7 by comparing horizontally juxtaposed histograms (A vs. E; B vs. F; C vs. G; D vs. H). In each pair, the left histogram represents activity on trials in which the left end of the reference image had been cued and the right histogram represents activity on trials in which the right end had been cued, with other factorsthe retinal location of the cue and the location of the target on the screenheld constant.

View larger version (39K): [in this window] [in a new window]   Fig. 6. Data from a neuron influenced, during delay 1, both by the cued end of the reference image (stronger firing for right) and by the type of reference image (stronger firing for bar). All conventions as for Fig. 3.

View larger version (27K): [in this window] [in a new window]   Fig. 7. Data from a neuron selective, during delay 2, both for the selected end of the target image (stronger firing for right) and for the type of target image (stronger firing for dot). Each histogram, with accompanying raster display, represents neuronal activity under a condition defined with respect to type of target image (bar or dot), location of target image (right, middle or left), and location of selected spot on target image (right or left). Juxtaposed to each histogram is a panel depicting the target image and the direction of the impending eye movement. The number in each panel indicates the corresponding trial condition (cf. Fig. 1D). A and E: conditions differing with respect to the selected end of the target array (A = left; E = right) but identical with respect to the location of the cue and the target on the screen. Three other pairs (B and F; C and G; D and H) stand in the same relation to each other. Other conventions as for Fig. 3.

The frequency with which neurons differentiated between bar and dot conditions is indicated by results summarized in Table 1. On the basis of the frequency with which main effects and interaction effects involving image type occurred, we draw the following conclusions. 1) Around a quarter of tested neurons showed a main effect of image type (bar vs. dot) during each epoch (31/143 during delay 1, 44/143 during delay 2, and 36/143 during the movement period). Each of these proportions is greater than expected by chance (P < .0001, ² test). 2) Among neurons in which there was a main effect of image type, those firing more strongly under dot conditions were markedly preponderant during later epochs (34/44 during delay 2 and 30/36 during the movement period). Each of these proportions is greater than expected by chance (P < 0.001, ² test). 3) In a few neurons, there was an interaction between object-centered direction and image type (17/143 during delay 1, 39/143 during delay 2, and 22/143 during the movement period). Each of these proportions is greater than expected by chance (P < 0.001, ² test). 4) Among neurons exhibiting a significant interaction between object-centered direction and image type, the preponderant pattern during later epochs was for the directional signal (the difference in firing rate between left-side-cued and right-side-cued conditions) to be stronger under the dot condition (26/39 during delay 2 and 16/22 during the movement period). Each of these proportions is greater than expected by chance (P < 0.05, ² test). The general conclusion arising from these observations is that neuronal activity (both net activity and differential activity dependent on object-centered direction) tended to be stronger under the dot condition but that the effect was weak.

Finally, we assessed whether the tendency of neurons to fire differently on bar and dot trials was related to their cortical location. For each neuron during each of three trial epochsdelay 1, delay 2, and the movement periodan ANOVA had been carried out indicating whether or not firing rate was significantly (P < 0.05) dependent on two factors (object-centered direction and image type) or their interaction. Thus at a location in the cortex where n neurons had been recorded, there were 3*n tests that might reveal a main effect of image type and 3*n tests that might reveal an interaction effect involving image type. For each cortical location, we counted the number of significant outcomes in each of four categories: main effect (firing greater under dot conditions), main effect (firing greater under bar conditions), interaction effect (difference in firing rate between left-on-image and right-on-image trials greater under dot conditions), and interaction effect (difference in firing rate between left-on-image and right-on-image trials greater under bar conditions). The results are shown in Fig. 8, A-H, where the size of each circle indicates the number of tests on data from that site yielding the indicated outcome. The figure reveals no clear trend toward segregation of sites exhibiting different patterns of dependence on image type.

View larger version (49K): [in this window] [in a new window]   Fig. 8. Cortical distribution of neurons exhibiting significant dependence on image type (bar vs. dot) in the bar-dot task. Left: monkey Ju. Right: monkey Po. Coordinates are with respect to the recording grid: (0,0) was at the center of the chamber. In each box, the area of each dark circle is proportional to the number of tests carried out on neurons at that site which yielded a particular form of significant dependence on image type. At a site where n neurons were studied, 3*n tests were carried out (direction selectivity was assessed independently during delay 1, delay 2, and the movement period for each neuron). The largest circle (in A) corresponds to a count of 14. A and B: main effect of image type, with the mean firing rate greater in dot trials. C and D: interaction effect between object-centered direction and image type, with the directional signal greater on dot trials. E and F: main effect of image type, with the mean firing rate greater in bar trials. G and H: interaction effect between object-centered direction and image type, with the directional signal greater in bar trials. Directional signal = the absolute value of difference in firing rate between trials on which the left vs. the right end of the image was cued.

In summary, our main finding is that most SEF neurons exhibiting object-centered direction selectivity under the standard condition used in our previous experiments (horizontal bar as reference image) also did so under a new condition (a pair of dots in a horizontal array as reference image). During each trial epoch, the firing of around a quarter of the neuronal sample was significantly affected by the type of image (bar or dot) either in the form of a main effect or in the form of an interaction with object-centered direction. Even in these cases, however, the preferred object-centered direction was the same under both conditions.

Possible influence of variations in ocular landing position

It is important to ask whether the signals interpreted in the preceding section as being object-centered possibly could have arisen from minor variations in the physical trajectory of the eyes. Accordingly, we analyzed saccades executed on bar and dot trials. We found that the trajectory of the eyes did vary slightly as a function of whether the target was the left or right end of a reference image. Especially in the case of a bar, the eyes did not land precisely on the end of the image but rather deviated inward toward its center. This is illustrated in Fig. 9, which shows eye-movement data from a single data-collection session. The symbols represent eye position over a period extending from 100 ms before to 100 ms after the instant of peak eye velocity for ~12 eye movements under each of 12 conditions. The reference image could be at any of three locations (left = L, middle = M, and right = R), and the target could be either the right end (r) or the left end (l) of the image. Thus there were six conditions in which the reference image was a bar and six in which it was a pair of dots. Among the six bar conditions (Fig. 9A), there were two pairs in which the targets were at the same location on the screen but at opposite ends of a bar (Lr vs. Ml and Mr vs. Rl). It is clear that the terminal direction of gaze was offset to the left by around half a degree on trials when the target was the right end of a bar (Lr and Mr) as compared with corresponding trials when target was the left end of a bar (Ml and Rl, respectively). In contrast, under conditions in which the target was an array of dots (Fig. 9B), this tendency was vanishingly small.

View larger version (39K): [in this window] [in a new window]   Fig. 9. Superimposed eye movement traces for 12 conditions (12 trials each) during a single data collection session. Each bundle of traces is identified by an uppercase letter indicating the location of the reference image on the screen (L, left; M, middle; and R, right) and the locus of the target on the reference image (l = left end, r = right end). Traces collected during eye movements to the left versus right end of a reference image are represented by red crosses versus green circles. The consistent difference between endpoints of corresponding saccades executed under bar conditions (A) and dot conditions (B) indicates that, in bar conditions, the saccade was directed not to the extreme end of the bar but, rather, to a point a few tenths of a degree in from the end. Eye position was stored every 10 ms. On each trial, the instant of maximal eye velocity was identified. Trace for that trial consisted of 10 readings before and 10 readings after this instant, spanning a total period of 190 ms. All targets were at an elevation of 5.8°. The outermost targets (Ll and Rr) were 4.65° to the right and left of the center of the screen, respectively. Conditions shown here correspond to conditions 2, 4, 6, 7, 9, 11, 14, 16, 18, 19, 21, and 23 of Fig. 1D.

To determine how consistent this pattern was, we computed the mean landing point of the eyes (the location to which gaze was directed 70-100 ms after the instant of peak velocity) under each of four spatial conditions (Lr, Ml, Mr, and Rl) for both bar and dot reference images. The results are summarized in Fig. 10, which shows the mean, across all data collection sessions, of the ocular landing position associated with each condition (all SDs were between 0.05 and 0.25°). In each of eight comparisons (2 screen locations × 2 image types × 2 monkeys), the eyes landed at significantly different loci when the targets were at the same location on the screen but at opposite ends of their respective reference images (paired t-test, P < 0.05). In seven of eight comparisons, the eyes deviated toward the center of the reference image so as to land farther to the left when the target was the right end of a reference image and farther the right when it was the left end. The sole exception arose in monkey Ju on comparison of eye movements to the right end of the middle dot array (Mr) and the left end of the right dot array (Rl). Across both monkeys and all four target locations, the mean horizontal displacement of the landing position on image-right as compared with image-left trials was 0.85° under the bar condition and 0.10° under the dot condition. This pattern of deviation is similar to the one observed by Edelman and Keller (1998) in monkeys trained to make eye movements to single target spots and exposed to occasional trials in which two spots came on simultaneously at radial directions 45° apart. On those trials, the eyes tended to land between the two targets; indeed when the saccades were of express latency (<90 ms), they landed close to the middle of the array. The effect was mild in our study because, unlike Edelman and Keller, we trained monkeys to select one end or the other of the distributed pattern, withholding reward if they landed outside a target window centered on the correct end.

View larger version (25K): [in this window] [in a new window]   Fig. 10. Mean landing positions of the eyes under 8 critical trial conditions (conditions used for statistical analysis of object-centered direction selectivity) in each of 2 monkeys. Positions were measured during the period from 70 to 100 ms after attainment of maximal saccadic velocity. Each mean was computed by finding the average horizontal and vertical positions of the eyes for each data collection session and then averaging the averages across all sessions. Filled and open symbols represent landing positions on bar and dot trials, respectively. Lr: target = right end of reference image at left location (conditions 11 and 23 in Fig. 1D). Ml: target = left end of reference image at middle location (conditions 4 and 16 in Fig. 1D). Mr: target = right end of reference image at middle location (conditions 9 and 21 in Fig. 1D). Rl: target = left end of right reference image (conditions 2 and 14 in Fig. 1D). In cases Lr and Ml, the target was located at superior 5.8°, left 4.65°, relative to presaccadic fixation. In cases Mr and Rl, the target was located at superior 5.8°, right 4.65°. Degrees on the graphs are positive upward and to the right.

Given that the orbital directions of the eye movements varied subtly but systematically between image-right and image-left trials, we considered the possibility that neurons exhibiting apparent object-centered direction selectivity were simply selective for the orbital directions of eye movements. To assess this possibility, we carried out a test summarized in Fig. 11. The test was applied to each neuron for which the ANOVA had revealed a significant main effect of object-centered direction. For each such neuron, the test was applied independently to each epoch in which a significant effect had been present. For each such epoch, it was applied to each image type. Each test focused on those four trial conditions in which the target was the right end of an image at the left location (Lr), the left end of an image at the middle location (Ml), the right end of an image at the middle location (Mr), or the left end of an image at the right location (Rl). For each of these conditions, we computed the mean horizontal coordinate of the eyes' landing position: X(Lr), X(Ml), X(Mr), and X(Rl). We also computed the mean observed firing rate: O(Lr), O(Ml), O(Mr), and O(Rl). We next fitted a line to the four points representing O as a function of X. Then for each condition, we computed the firing rates predicted on the assumption that firing rate was a linear function of X: P(Lr), P(Ml), P(Mr), and P(Rl). Finally, we computed two object-centered directional signals: the one actually observed0.5 * [O(Mr)  O(Rl) + O(Lr)  O(Ml)]and the one predicted from the linear function0.5 * [P(Mr)  P(Rl) + P(Lr)  P(Ml)]. Figure 11 shows the results of applying this procedure to a single caseneuron ju152a41, delay 2, dot conditionsfor which eye-position data are shown in Fig. 9B and firing rate data in Fig. 7, A, C, E, and G. The observed object-centered directional signal was 9.6 spikes/s, in marked contrast to the object-centered directional signal predicted on the basis of linear dependence on horizontal landing position (0.23 spikes/s). Given the fact that the predicted signal was 42 times smaller in amplitude than the observed signal, not to mention opposite in sign, we conclude that orbital direction selectivity cannot explain this neuron's object-centered direction selectivity.

View larger version (23K): [in this window] [in a new window]   Fig. 11. Method for assessing whether measured values of object-centered direction selectivity could arise spuriously from systematic variation of ocular landing position across task conditions. Method is illustrated with data collected from neuron ju152a41 during the 2nd delay period under dot conditions. Mean neuronal activity under conditions Rl, Ml, Mr, and Lr can be judged by inspection of the histograms of Fig. 7, A, C, E, and G, respectively. Mean ocular landing position under the same conditions can be judged by inspection of Fig. 9B. Four points in this graph are at locations corresponding to the combinations of observed firing rate (vertical axis: O) and horizontal ocular landing position (horizontal axis: x) observed under the 4 conditions. Line (y = 15.57 + 1.27x) is the linear function relating firing rate (spikes/s) to the horizontal coordinate of the ocular landing position (deg) that provides the best fit to these points. Positive slope reflects an overall tendency for firing to be stronger on trials in which the eye movement had a rightward component. Observed index of object-centered direction selectivityas based on observed firing rate values and computed according to the formula 0.5 * [O(Lr)  O(Ml) + O(Mr)  O(Rl)]was positive and large (9.6 spikes/s), reflecting the tendency of the neuron to fire more strongly on trials when the right end of the reference image was the target. To determine whether the observed pattern of object-centered direction selectivity could arise spuriously through a mechanism based solely on neuronal sensitivity to ocular landing position, we noted the firing rates at which the best-fit line (representing neuronal sensitivity to ocular landing position) intersected the observed landing positions. These firing rates [P(Lr), P(Ml), P(Mr), and P(Rl)] were the ones predicted on the basis of neuronal sensitivity to ocular landing position alone. Predicted index of object-centered direction selectivityas based on these predicted firing rates and computed according to the formula 0.5 * [P(Lr)  P(Ml) + P(Mr)  P(Rl)]was negative and small (0.23 spikes/s). From the fact that the predicted index was forty times smaller than the observed index, and, incidentally, of opposite sign, we conclude that the appearance of object-centered direction selectivity in this neuron could not have arisen simply as a secondary manifestation of sensitivity to the ocular landing position. Method illustrated here was used to generate the measures of observed and predicted object-centered direction selectivity which are plotted against each other in Fig. 12.

The results for all neurons and epochs are summarized in Fig. 12, where, in each panel, the observed object-centered signal is plotted on the horizontal axis and the object-centered signal predicted on the basis of the landing-position hypothesis is plotted on the vertical axis. The range of predicted object-centered signals is obviously miniscule as compared with the range of observed object-centered signals. In monkey Ju, the standard deviation of the observed values was greater than the standard deviation of the predicted values by factors of 9.7 and 38.1 under bar and dot conditions, respectively. In monkey Po, the corresponding values were 5.3 and 56.1. In summary, if we assume that neuronal activity is related only to the eyes' landing position, form the best estimate of the linear function relating firing rate to landing position, take into account the differences in landing position across different conditions, and compute the spurious "object-centered" directional signal predicted on the basis of the differences in landing position, then we find that the predicted spurious signals are extremely small as compared with the signals actually observed in the experiment. We conclude that object-centered direction selectivity is not an artifact arising from subtle variations of the eyes' landing position across conditions.

View larger version (31K): [in this window] [in a new window]   Fig. 12. Assessment of whether measured values of object-centered direction selectivity could arise simply from systematic variation of ocular landing position across task conditions. Each point represents data obtained from 1 neuron during a single epoch (delay 1, delay 2, or movement period) during use of a single type of reference image (bar or dots). Consideration was restricted to epochs in which statistical analysis had revealed significant object-centered direction selectivity. For each epoch, 2 indices were computed: the observed index of object-centered direction selectivityas computed without taking into account the systematic variation of ocular landing position across conditionsand a predicted indexas computed by taking systematic variations into account and assuming that neuronal activity was solely a linear function of the horizontal landing position of the eyes. See legend to Fig. 11 for further details. Large measured values of object-centered direction selectivity (horizontal axis) are incommensurate with the small values predicted on the basis of sensitivity to ocular landing position alone (vertical axis). Interpretation that object-centered direction selectivity is an artifact of sensitivity to ocular landing position therefore is rejected.

Relation to selectivity for saccade direction in the memory-guided saccade task

Even if our recording sites were within the SEF as defined on morphological grounds, which we believe to have been the case, nevertheless neurons exhibiting object-centered direction selectivity in the bar-dot task might constitute a population distinct from intermingled neurons exhibiting selectivity for saccade direction in standard oculomotor tasks as described by previous authors. To cast light on this issue, we compared results obtained in the bar-dot task (Fig. 1) with those obtained in a memory-guided saccade task (Fig. 13). The latter task required monkeys to make eye movements to four targets at rightward, upward, leftward, and downward locations relative to fixation. The use of four targets at directions 90° apart, common in studies of the SEF (Chen and Wise 1995a,b, 1996, 1997; Schall 1991a,b), is warranted because SEF neurons are very broadly tuned for saccade direction and amplitude (Russo and Bruce 1996). In the context of the memory-guided saccade task, we studied a total of 125 neurons from monkey Ju (56 and 69 in the left and right hemispheres, respectively) and 156 neurons from monkey Po (79 and 77 in the left and right hemispheres, respectively). Many but not all of these neurons also were studied in the bar-dot task (62 in monkey Ju and 52 in monkey Po).

View larger version (15K): [in this window] [in a new window]   Fig. 13. Memory-guided saccade task. A-F: screen in front of the monkey during successive epochs of a single representative trial. Center of each circle indicates the monkey's direction of gaze during the corresponding trial epoch and the arrow indicates the direction of the eye movement. All other items are patterns visible to the monkey. A: white fixation spot appeared at the center of the screen and the monkey achieved foveal fixation. B: 4 potential targets (white dots 0.4° in diameter) appeared at locations 9.6° rightward, leftward, upward, and downward relative to fixation. C: white cue (1.2° in diameter) flashed on 1 of the targets. D: during an ensuing delay period, the monkey maintained central fixation. E: extinction of the central fixation spot signaled the monkey to initiate an eye movement. F: monkey made a saccade directly to the previously cued target.

First we asked whether there was any systematic pattern to the topographic distribution of neurons exhibiting saccade-direction selectivity. We based this analysis on all neurons studied in the memory-guided saccade test regardless of whether they were studied in the bar-dot task. The significance (P < 0.05) of each neuron's selectivity for eye-movement direction was assessed by means of an ANOVA with direction (right, up, left, or down) as the single factor and with firing rate as the dependent variable. This was done independently for the delay period (from onset of the cue to offset of the fixation spot) and the movement period (from offset of the fixation spot to 100 ms after completion of the saccade). Thus at a location in the cortex where n neurons had been recorded, 2*n tests of significance were carried out. For each cortical location, we computed the percentage of tests that yielded a significant outcome. The results are shown in Fig. 14, A and, B, where the size of each circle indicates the percentage of tests, on data from that site, indicating significant selectivity for eye-movement direction. Inspection of this figure reveals that orbital direction selectivity was comparatively widespread across the recording sites sampled in each monkey. Further, the patterns of regional arrangement showed no clear trends of a form consistent across hemispheres or monkeys. Finally, comparison of sites yielding significant selectivity for eye-movement direction (Fig. 14, A and B) to sites yielding significant selectivity for object-centered direction (Fig. 4, A and B) reveals that the two were largely overlapping.

View larger version (38K): [in this window] [in a new window]   Fig. 14. Cortical distribution of neurons exhibiting significant eye-movement direction selectivity in the memory-guided saccade task for monkey Ju (A) and monkey Po (B). Coordinates are with respect to the center of the recording grid (0,0). Each site at which recording was carried out during performance of this task is marked by a circle. Area of each dark circle is proportional to the percentage of tests revealing significant eye-movement direction selectivity at the corresponding site. Where n neurons were studied, 2*n tests were carried out (direction selectivity was assessed independently during the delay period and the movement period for each neuron). Largest circles represent cases in which 100% of tests yielded a significant result. Sites at which 0% of tests yielded a significant result are indicated by small open circles.

Next we carried out a set of comparisons intended to reveal whether the presence or sign of direction selectivity in the memory-guided saccade task was correlated with the presence or sign of object-centered direction selectivity in the bar-dot task. This analysis was restricted to 114 neurons studied in both tasks. Data collected from each neuron during the memory-guided saccade task were assessed to determine whether the firing rate was significantly affected by vertical direction (upward vs. downward trials) or horizontal direction (rightward vs. leftward trials). Each comparison was carried out on data from the delay period (cue onset to fix-spot offset) and the movement period (fix-spot offset to 100 ms after completion of the saccade). Four t-tests indicated whether the neuron was significantly (P < 0.05) selective for direction as defined with respect to the horizontal and vertical axes during the delay and movement epochs. Neurons were tested for object-centered direction selectivity by means of an ANOVA as described in an earlier section.

We first asked whether selectivity for vertical direction, as observed in the memory-guided saccade task, was related to object-centered direction selectivity, as observed in the bar-dot task. It was reasonable to pose this question because all eye movements required in the bar-dot task were in an upward direction. We calculated the numbers of neurons exhibiting various combinations of selectivity during three pairs of epochs: delay 1 in the bar-dot task versus delay in the memory-guided saccade task; delay 2 in the bar-dot task versus delay in the memory-guided saccade task; and movement period in the bar-dot task versus movement period in the memory-guided saccade task. The results are presented in Fig. 15, A and B. In each panel, the rows contain counts of cells exhibiting (S) or not exhibiting (N) significant image-centered direction selectivity in the bar-dot task. Likewise, the columns contain counts of cells not exhibiting vertical direction selectivity (N) or significantly favoring downward (D) or upward (U) movements in the memory-guided saccade task. We carried out two ² tests on these counts. The first test, applied to all neurons, assessed whether the presence of selectivity for eye-movement direction as defined with respect to the vertical axis (in the memory-guided saccade task) was correlated with the presence of selectivity for object-centered direction (in the bar-dot task). Overall, across all pairs of epochs in both monkeys, there was a slight trend for object-centered direction selectivity to be more common among neurons selective for vertical direction than among those not so selective (56 vs. 46%). However, this tendency did not achieve significance (P < 0.05) for any pair of epochs in either monkey. The second test, applied only to neurons exhibiting selectivity for direction as defined with respect to the vertical axis, assessed whether the preferred vertical eye-movement direction (upward or downward) was correlated with the presence of selectivity for object-centered direction. Overall, across all pairs of epochs in both monkeys, there was a slight trend for object-centered direction selectivity to be more common in neurons selective for upward than in those selective for downward movement (58 vs. 41%). However, this tendency did not achieve significance (P < 0.05) for any pair of epochs in either monkey.

View larger version (40K): [in this window] [in a new window]   Fig. 15. Counts of neurons exhibiting v