JOURNALOFNEUROPHYSIOLOGY Vol. 58, No. 6, December 1987. Printed in U.S.A

Functional Properties of Corticotectal Neurons in the Monkey’s Frontal Eye Field

MARK A. SEGRAVES AND MICHAEL E. GOLDBERG Laboratory of SensorimotorResearch, National Eye Institute, National Institutes of Health, Bethesda,Maryland 20892

SUMMARY AND CONCLUSIONS 5. The remaining 20% of the corticotectal neurons were a heterogeneous group whose 1. We examined the activity of identified activity could not be classified as movement, corticotectal neurons in the frontal eye field visuomovement or foveal. Their responses of awake behaving rhesus monkeys (Macaca included postsaccadic, anticipatory, and re- mulatta). Corticotectal ne urons were anti- ward-related activity, as well as activity dromical y excited using biphasic cu rrent modulated during certain directions of pulses passedthrough monopolar microelec- smooth-pursuit eye movements. One neuron trodes within the superior colliculus. The ac- was unresponsive during all of the behavioral tivity of single corticotectal neurons was tasks used. There were no corticotectal studied while the m .onkey performed behav- neurons that could be classified as primarily ioral tasks designed to test the relation of the responsive to peripheral visual stimuli. neuron’s discharge to visual and oculomotor 6. Histological reconstructions of elec- events. trode penetrations localized corticotectal 2. Fifty-one frontal eye field corticotectal neurons to layer V of the frontal eye field. neurons were examined in two monkeys. For 22 corticotectal neurons tested, each had Current thresholds for antidromic excitation its minimum threshold for antidromic exci- ranged from 6 to 1,200 PA, with a mean of tation within the superior colliculus, as 330 PA. Antidromic latencies ranged from judged by either histological confirmation, 1.2 to 6.0 ms, with a mean of 2.25 ms. or surrounding neuronal responsesrecorded 3. Fifty-three percent of the identified through the stimulation microelectrode. The corticotectal neurons were classified as hav- majority of these neurons had minimum ing movement-related activity. They had lit- threshold sites within the intermediate tle or no response to visual stimuli, but very layers; a few minimum threshold sites were strong activity before both visually guided located within the superficial or deep collic- and memory-guided saccades.An additional ular layers. 6% of corticotectal n.eurons had v ,isuomove- 7. The lowest thresholds for antidromic ment activity, combining both a visual- and excitation were obtained when the optimal a saccade-related response. In each case, vi- saccade vectors associated with the frontal suomovement neurons antidromically ex- eye field recording and collicular stimulation cited from the superior colliculus had move- sites were closely matched. There was a ment-related activity, which was much strong correlation between a measure of the stronger than the visual component of their difference between saccadesassociated with response. recording and stimulation sitesand the log of 4. Twenty-two percent of the corticotectal threshold for antidromic excitation. This re- neurons were primarily responsive to visual lationship was such that small increases in stimulation of the fovea. These included the vector difference between frontal eye both neurons responding to the onset and field and collicular saccadeswere accompa- neurons responding to the disappearance of a nied by large increases in threshold. light flashed on the fovea. 8. In comparison to the entire population

1387 1388 M. A. SEGRAVES AND M. E. GOLDBERG of frontal eye field neurons examined by field neurons were not related to spontane- Bruce and Goldberg (7), we conclude that ous saccades made in the dark. The only there is a selective enrichment within the neurons related to saccades were a small per- population of corticotectal projection centage (30/700) that became active only neurons for neurons with eye movement-re- after the initiation of a saccadic eye move- lated activity and neurons with fovea1 visual ment. More recent single-neuron studies activity, and a paucity of neurons with pe- have indicated a role for the frontal eye field ripheral visual receptive fields and postsac- in the initiation of saccades. Mohler, Gold- cadic activity. berg, and Wurtz (39) demonstrated that cells 9. These data suggest that the visual activ- in the frontal eye field discharge in response ity prevalent within the frontal eye field is to visual stimuli. This visual activity is en- likely to help generate the activity of move- hanced when the monkey makes a saccade to ment-related neurons, but it is not a source the stimulus in the receptive field (62), and for visual activity within the superior collic- the enhancement only occurs before sac- ulus. cades and not before arm movements or 10. The frontal eye field’s projection to other activity in which saccades do not occur the superior colliculus provides several mes- (9). Bruce and Goldberg (7) recently reported sages relevant for oculomotor performance, that only - 19% of frontal eye field neurons including information pertinent to the main- have purely postsaccadic activity. The ma- tenance and release of fixation, and targeting jority of neurons recorded from by Bruce information regarding an intended saccade. and Goldberg in awake behaving monkeys had some form of activity that preceded visually guided saccadic eye movements, INTRODUCTION including a range of neuron activity from Since Ferrier’s original demonstration ( 12) purely visual to purely movement related. that electrical stimulation of the monkey’s It is clear there are signals in the frontal eye prearcuate frontal cortex produced conju- field that could drive saccadic eye move- gate eye movements, it has been postulated ments and neural pathways by which these that this region participates in the voluntary signals could reach brain stem oculomotor control of gaze. However, a number of ex- centers. However, to understand how the ce- periments have raised questions about the rebral cortex controls a specific behavior, it is exact role of the frontal eye field in eye- not sufficient to know the types of activity in movement control. This uncertainty results, the cortex and the anatomical projections of in part, from the existence of parallel path- that region. One must also know what infor- ways that are likely to be involved in the mation is carried by the corticofugal signals. input of eye movement control signals to the These output signals are likely to be pro- brain stem ocuomotor centers. The frontal duced bv a subset of the activity types within eye field can affect the oculomotor system the entire neuronal population, and the out- via three pathways (26, 33-35, 55). The put population will be enriched for some sig- first is by a direct projection to perioculomo- nal types and lacking others. To begin to an- tor regions in the midbrain and pons (33). swer this question, we identified corticotectal The second is by a projection to the caudate projection neurons by antidromic excitation nucleus, which might then affect the inhibi- from the superior colliculus and then charac- tory pathway from the substantia nigra pars terized these neurons in awake behaving ani- reticulata to the superior colliculus (11, 24). mals according to the scheme described by The final one is a direct projection to the Bruce and Goldberg (7). A preliminary re- intermediate layers of the superior colliculus port of these experiments has been published (1, 30). The relative contribution that each elsewhere (5 3). pathway makes to eye-movement control has not been determined. Additional prob- METHODS lems arise when one considers the number of different neuron activity types that have Preoperative training been described in the frontal eye field. Bizzi Two adult rhesus monkeys (Macaca mulatta) (4) originally found that most frontal eye were trained preoperatively to do a simple visual FRONTAL EYE FIELD CORTICOTECTAL NEURONS fixation task using established techniques (60). criteria for eye position relative to target position The monkey was seated in a primate chair and and for saccade amplitude and direction, in order began a trial by pressing a metal bar in front of to receive a liquid reward. Eye position was mea- him. This resulted in the onset of a target light on sured by the magnetic search-coil system using the a screen in front of the monkey. The monkey was C.N.C. Engineering phase-sensitive detector. Eye required to detect the dimming of the light and position measurement was accurate to 15 min of signal this detection by releasing the metal bar to arc within a range of 20” from the center of gaze receive a liquid reward. and was not corrected for cosine error. Each monkey was trained to do several behav- Surgery ioral tasks (Fig. 1). Each task began with the ap- Surgery was performed under aseptic condi- pearance of a central stationary light on the screen tions. The monkey was anesthetized with ket- (Fig. 1, FP). The computer monitored the mon- amine hydrochloride (10 mg/kg im) followed by key’s eye position, and when the monkey had pentobarbital sodium through an intravenous achieved fixation for 100 ms the computer then catheter as needed. During the surgical procedure, began the various timing intervals of each task: a subconjunctival wire coil for the measurement 1) FIXATION TASK (FIG. 1A). The monkey was of eye position with the magnetic search-coil tech- required to hold fixation throughout the trial. nique was implanted in one eye (27,44). Trephine During some trials the fixation light was turned holes were made through the skull over the supe- off at an unpredictable moment for a brief period rior colliculi and over both left and right frontal of time. The interval during which the light was eye fields. Stainless steel bolts to strengthen the turned off was the same in any given block of bond of dental acrylic to the skull were fastened in trials, but was varied from 200 to 500 ms between slots cut through the skull and extending away blocks of trials. The monkey was rewarded for from the edges of the trephine holes. Three re- maintaining eye position within a window sur- cording cylinders, a steel receptacle to fix the rounding the target such that the sum of the hori- monkey’s head during recording sessions, and the zontal and vertical differences between eye and connector for the eye coil were fixed in place and target position was ~5” (54). This was a criterion bonded to the skull with dental acrylic. Immedi- that the monkeys met easily, and usually sur- ately after surgery, the monkey was given genta- passed. With practice each monkey was able to micin sulfate (5 mg/kg im) as a prophylactic mea- maintain fixation of the target’s initial position sure against infection. The monkey received daily throughout the trial, including periods when the dosages of gentamicin sulfate for 1 wk postopera- light was turned off. This paradigm enabled us to tively. Postoperative training was begun between test the effects of visual stimulation of the fovea 1.5 and 2 wk after surgery. on neuronal activity. Postoperative training 2) VISUAL NO-SACCADE TASK (FIG. 1B). The monkey fixated the stationary light spot in the After surgery, each monkey received additional center of the tangent screen for the duration of the training in several visual and saccade tasks. The trial. During the trial, at an unpredictable time, a computer hardware and software used to monitor second light spot (Fig. lB, S) was flashed for a and control the monkey’s behavior has recently variable period of time. The position of the sec- been described in detail elsewhere (17). Briefly, ond light was controlled by mirror galvanometers the monkey was seated in a primate chair with its and could be placed anywhere within the mon- head fixed and centered within horizontal and key’s visual field. The monkey was rewarded for vertical magnetic field coils. Visual stimuli pre- maintaining fixation of the center light through- sented to the monkey included stationary and out the trial. The peripheral stimulus had no be- movable light stimuli originating from yellow havioral significance, but its position was used to light-emitting diodes rear-projected onto a tan- map the visual receptive fields of frontal eye field gent screen 57 cm in front of the monkey. Each neurons. projected stimulus was -0.25’ in diameter with a brightness of 0.4 log units above a background of 3) VISUALLY GUIDED SACCADE TASK (FIG. 1 cd/m*. The movable stimulus was positioned by 1 C). After a variable period of fixation, the a pair of servo-controlled mirror galvanometers center light was extinguished at the same moment (General Scanner, Watertown, MA) driven by an- that a peripheral target light (Fig. 1, C and D, T) alog signals synthesized by 12-bit digital-to-analog was turned on. When the peripheral target ap- converters under the control of the computer. peared, the monkey was required to make a sac- Since, after surgery, it was possible to accurately cade to it within 500 ms. After the saccade, the measure eye position, the monkey was no longer monkey was rewarded either for maintaining eye required to detect the dimming of the stimulus. position within a window with -3-6’ radius sur- Instead, the monkey had to meet predetermined rounding the new target position or for having 1390 M. A. SEGRAVES AND M. E. GOLDBERG

A

Fixation VH m-/ L I

B Visual VH No-Saccade FP-I 1 s I 1

C

Visually VH Guided \ / FPJ I Saccade T IL

D Memory VH \‘I 10" Guided / FP-J I Saccade T I L r 1 250 ms

FIG. 1. Behavioral tasks. H, horizontal eye position; V, vertical eye position. An upward deflection signifies movement to the right for H and up for V. FP, fixation point; S, peripheral stimulus light; T, peripheral target light. For FP, S, and T, an upward deflection signifies light on. Each trial began when the monkey fixated a light in the center of the tangent screen and maintained fixation for at least 100 ms. A: fixation task. The sketch to the right of the eye position traces shows that only the fixation point appears in this task. During the trial, the fixation point was turned off for 200 ms. B: visual no-saccade task. Sketch at right indicates that both fixation point and stimulus flash occur in this task, but no saccade is made. C: visually guided saccade task. Sketch at right indicates that a saccade is made to the position of the target light after the fixation point is turned off. D: memory-guided saccade task. Sketch indicates that saccade is made to the remembered position of the target light after the disappearance of the fixation point. See text for an additional description of each task.

made a saccade whose end point was within the cade to a remembered target location. This task window. The movable light spot was again used enabled us to distinguish between visual and for the peripheral target and could be positioned movement associated neuronal activity during the anywhere on the tangent screen. In this task the trial. simultaneous disappearance of the center fixation In this study, the location of visual stimuli and light and onset of peripheral target light were the targets are expressed in polar coordinates, radius cue for the monkey to make a saccade. and angle, where radius is equal to the distance, in 4) MEMORY-GUIDED SACCADE TASK (FIG. degrees of arc, of the target or stimulus from the 10). In a second more difficult form of saccade central fixation point. An angle of 0” describes a task, the center light came on to start the trial, as rightward horizontal direction, and a 90” angle before, and the monkey was required to begin and describes an upward vertical direction. maintain fixation of the center light until it was turned off. During the central fixation period, a peripheral target was turned on for 50-300 ms. Electrophysiology When the center light was turned off the monkey A diagram of the stimulation and recording was required to make a saccade to the position of setup is provided in Fig. 2, illustrating the anti- the flashed peripheral target light. The duration dromic excitation of a single neuron in the frontal and time of occurrence of the target flash were eye field by a monopolar stimulating electrode in adjusted so that the target light was turned off the superior colliculus. In these experiments, we during the fixation period, requiring the monkey defined the frontal eye field as the area of cortex, to maintain fixation for up to 250 ms after the located primarily on the rostra1 bank of the ar- target light’s disappearance and then make a sac- cuate sulcus, where eye movements can be evoked FRONTAL EYE FIELD CORTICOTECTAL NEURONS 1391

Stimulation

Superior Colliculus Frontal Eye Fields

FIG. 2. Stim ulation and neuron recording configuration . Single neurons isolated by a icroelectrode in the frontal eye field were antidromically excited by a monopolar stimulating microelectrode in t superior collicul us. with thresholds of 50 PA or less (7, 8). Initial repeated penetrations through areas where anti- neuron recording in both the superior colliculus dromically excitable neurons were found pre- and the frontal eye field was done with glass- viously and improved the recording stability. coated platinum-iridium electrodes (59) intro- Twenty-seven (5 3%) of the antidromically excited duced through the intact dura. For the superior and physiologically characterized neurons in- colliculi, penetrations through the dura were used cluded in this report were recorded from elec- to define the collicular topography relative to the trodes passed through implanted guide tubes. recording cylinder. Once this topography was Stimulation through the superior colliculus known, an M-gauge guide tube was introduced electrode to excite frontal eye field neurons anti- through the dura with the monkey under ket- dromically was done with biphasic negative first amine anesthesia (5- 10 mg/kg). The portion of pulses, with 0.2-ms pulse width. The output of the the guide tube exposed above the dura was ce- stimulus generator was connected to the electrode mented to the wall of the recording cylinder with through constant-current optical isolators. The dental acrylic. The tip of the guide tube was posi- amount of stimulus current was determined from tioned a few millimeters above the surface of the the voltage measured across a l-k0 resistor in superior colliculus. This made it possible to make series with the electrode. Searching for antidromi- repeated penetrations with fine tungsten micro- tally excitable neurons was routinely done with electrodes through the same region of the superior stimulus currents of 1 .O mA. Once an antidromi- colliculus and reduced the amount of daily prepa- tally excitable neuron was isolated, its excitation ration time before the actual search for antidrom- threshold was determined and the stimulus cur- ically excitable neurons could begin. When the rent was then set to an amount -50% greater guide tube was not in use, a wire matched to the than threshold. We defined threshold as the cur- inside diameter of the 1&gauge tubing and coated rent intensity at which antidromic spikes were with antibiotic (3% tetracycline HCl ointment) obtained in response to -50% of the stimulus was inserted into it. In several instances, a single pulses. Single-shock stimulation never appeared guide tube in the superior colliculus remained to have any effect on the monkey; no eye or skele- implanted for 1 mo without adverse effect. The tal movement were ever evoked, nor did the tangential area of the superior colliculus sampled monkey ever display a behavior that could be in- through the guide tube could be increased by terpreted as resulting from an aversive stimulus. slightly bending the electrode - 1 cm from its tip, In fact, the monkey frequently fell asleep during resulting in the area “seen” by penetrations the interval between behavioral trials when the through the guide tube to be conical in shape. superior colliculus stimulus was administered. Neuronal recording in the frontal eye field was Antidromically excited neuronal responses were done both with platinum-iridium electrodes intro- identified by their fixed latency and by our ability duced through the dura and with tungsten elec- to collide the neuronal spike with spontaneous trodes passed through guide tubes. The use of spikes originating in the frontal eye field (3, 16). guide tubes in the cortex made it easier to make The amplified electrode signal was sampled at 12 1 1392 M. A. SEGRAVES AND M. E. GOLDBERG

c kHz by an analog-to-digital converter and saved A in computer memory. One sampling of the signal from the electrode lasted for 8.5 ms and was trig- gered by either the onset of the collicular stimulus or the beginning of a spontaneous neuronal spike in the frontal eye field. The digitized electrode signal trace was displayed on an oscilloscope and could be averaged with successive traces following B the same trigger. Single and averaged traces were saved by the computer for future analysis and dis- play. Figure 3 shows examples of antidromic ex- citation (Fig. 3, A and C) and collision (Fig. 3, B and 0). Two smaller neuronal spikes, driven by the collicular stimulus but not collidable by the larger neuron spike, are indicated by arrows in C Fig. 3, C and D. Trains of pulses were occasionally applied at a frequency of 330 Hz through the recording micro- electrode, to establish that the electrode was lo- cated in the low-threshold frontal eye field or the intermediate layer of the superior colliculus. Trains of stimulation in the deeper colliculus, close to the central gray occasionally evoked ste- reotyped skeletal movements, or, more rarely, aversive reactions. In the latter cases train stimu- lation was stopped and the electrode elevated to a 1 ms nonaversive site. Train durations of 70 ms were * U A * A A used to evoke saccades, and durations >300 ms were used to evoke smooth pursuit (8). FIG. 3. Example of antidromic excitation, collision. We searched for frontal eye field corticotectal Each trace begins with the occurrence of a spontaneous neurons in two different ways. We obtained the spike generated by a neuron isolated in the frontal eye highest yield of antidromically excitable neurons field. Traces are aligned upon the superior colliculus stimulus artifact. A and B: single traces 8.5 ms in dura- by stimulating at 50- to loo-pm intervals along a tion digitized by a high-speed analog-to-digital converter penetration through cortex, testing the activity of (see METHODS); C and D: averages of 10 consecutive a neuron during behavioral tasks only when it had traces. In A, 1 marks the spontaneous neuronal spike; 2, been identified as a corticotectal neuron. Alterna- the superior colliculus stimulus artifact; and 3, the anti- tively, we made penetrations through the cortex dromically excited neuronal spike. When the spontane- and tested the activity of isolated neurons during ous spike and superior colliculus stimulation were sepa- the various behavioral tasks and then stimulated rated by 2.5 ms (A and C), the isolated neuron was the colliculus to see if the neuron could be excited excited antidromically with a latency of 1.5 ms. How- antidromically. The latter method was used to ever, when the separation between spontaneous spike and stimulus was reduced to 0.6 ms, the antidromically make sure that our electrode penetrations sam- evoked spike collided with the spontaneous spike and pled all of the neuron types previously identified was not seen at the frontal eye field electrode (B and D). in the frontal eye field (7). Note that two smaller neuron spikes evoked at constant Behavioral and neuronal activity were sampled latencies of 2.0 and 4.6 ms (arrow in C and D) were not at 1 kHz. Rasters and histograms were con- affected. Time between ticks in lowest trace was 1 ms. structed with bin widths of 4 ms (250 Hz) on-line Stimulus intensity, 50 PA. from this I-kHz sampled data and were then stored on magnetic disk for off-line analysis. Be- cause the original I-kHz data was resampled at Furthermore, two spikes falling in one bin were 250 Hz, minor quantization differences occurred treated as one spike, and thus a pair of spikes with between pairs of rasters and histograms con- frequency ~250 Hz could fall in one bin and structed from the same original data set but with a count as a single spike in one raster, or fall in fixed relative offset, for example beginning and adjacent bins and count as two spikes in another end of a target flash. Since the precise positioning raster. This effect is evident in the histograms of of sampling bins would not necessarily be the Figs. 8, B and C’, 10, -4 and B, and 11, B and C. same for two different pairs of rasters and histo- These minor differences had absolutely no influ- grams, a given spike could fall in one bin on one ence upon the classification of the firing activity of raster and in an adjacent bin in the other raster. these units. FRONTAL EYE FIELD CORTICOTECTAL NEURONS 1393

Data analysis barbital sodium and were perfused transcardially NEURON CLASSIFICATION. The behavioral with saline followed by 10% formalin. The brains tasks described above were used to classify were then removed, photographed, and frozen neurons according to the schema proposed by sectioned at 48 pm. Two planes of section were Bruce and Goldberg (7). The largest class of used. Sections through the superior colliculus neurons in the frontal eye field are presaccadic were made in the coronal/frontal plane. Sections neurons, which discharge before a visually guided through the arcuate sulcus were made in a plane eye movement. Presaccadic neurons are classified running rostrocaudally, parallel to the principal as visual, movement, or visuomovement. VC2& sulcus (see Fig. 5). Every 10th section was Y~~URVZSrespond to peripheral visual stimuli mounted and stained with cresyl violet. Addi- whether or not the monkey actually uses the stim- tional sections were stained as needed. ulus as the target for a saccade. The response of a visual neuron may be enhanced when the monkey RESULTS does make a saccade to the visual target, but the neuron does not fire when the monkey makes the Physiological classzj?cation of. identical eye movement in a learned saccade task antidromically excited neurons in which the target light is not presented. Move- We physiologically characterized 5 1 anti- ment neuronshave little or no response to visual dromically excited neurons. Twenty-four stimuli, but very strong activity before both vi- neurons were recorded from the left frontal sually guided and memory-guided saccades. eye field of monkey M30, and 22 neurons Movement neurons have definable movement from the right and 5 neurons from the left fields, but give little or no discharge before sponta- neous saccades, of the appropriate direction, frontal eye field of monkey IM50. Table 1 made in the dark. They do not discharge in re- shows the number and percentage of corti- sponse to visual stimulation of the fovea. Visuo- cotectal neurons in each category of neuro- movementneurons show both visual and move- nal activity. ment-related activity, but their optimal discharge LMOVEMENT NEURONS. Fifty-three per- occurs before visually guided saccades, which dis- cent of the corticotectal neurons (27 tinguishes them from movement neurons whose neurons) identified and characterized in this activity is similar before both visually and mem- study belonged to the movement category. ory-guided saccades. The visuomovement neurons occupy a continuum from strongly visual Figure 4 shows the activity, recorded during neurons with weak movement discharge to the memory-guided saccade task (Fig. 1D), strongly movement neurons with weak visual ac- from a movement neuron antidromically tivity. All presaccadic neurons can have anticipa- excited from the superior colliculus. The tory activity, beginning their discharge before the neuron did not discharge in response to the signal to make a saccade in paradigms where the appearance of the target flash (Fig. 4A) but monkey expects to make a saccade of a certain discharged briskly before the saccade (Fig. amplitude and direction. There are a number of 4B), with a burst starting - 160 ms before, neurons that cannot be classified as presaccadic. Two types are of interest to this study: fovea1 neurons and postsaccadic neurons. Fovea/ neuronsrespond to visual stimulation of the fovea TABLE 1. Distribution ofactivity typesjtir and include neurons with on or off responses to neurons antidromically excited from fovea1 stimulation. We found uncomplicated fo- veal neurons that could only be driven by fovea1 the superior colliculus stimuli and complex fovea1 neurons whose fovea1 responses were combined with visual, visuomove- Monkey Monkey Activity Type M30 hf.50 Total Percent ment, or movement anticipatory activity. Post- saccadicneurons begin to discharge during and Visual 0 0.0 after saccades, but not before. Visuomovement 3 3 5.9 Movement 1.5 12 27 52.9 Histology Postsaccadic 2.0 Marking lesions were made at the site of an Fovea1 2 9 11 21.6 antidromically excited neuron in each hemi- Others 4 8 15.7 sphere and at the lowest threshold collicular sites Nonresponsive 2.0 from which these neurons could be antidromi- Totals 24 27 51 100.1 tally excited. At the conclusion of the experiment, the monkeys were given a lethal dose of pento- See METHODS for description of neuron activity categories. 1394 M. A. SEGRAVES AND M. E. GOLDBERG

A B

Frontal Eye Field Activity

H- / H vv vv r loo FP L p1 FP ‘-1 T f 1 T1

...... --...... I ...... w-..- m ...... ” .w ...... w .. . ..q ...... s ...... -.

...... -...... ” ......

. ..“-. .. .w .. . ..“......

.- ...... B-.-w ...... ” ...... -.a- ...... e- ... WV ...... , ... "S. . W” ...... “_ ...... -.s...... -w......

. .- ...... s- ...... e . . . ” ...

s -...... -.- ...... w -...... _.” ......

. . ..-.s ...... - ... .

. . “M...... I I......

. . . .m ......

...... -...... ” ..

. . .-- ..“...... *- ...... , . . . . .

..... -.- ..- . ..” ...... -. “W ... I......

; ’ .:

. 200 ms/div

C D

Superior Colliculus Activity

H- vv v\ loo FP 1 FP ~-1 T I 1 T1

1 I1 11 I I I I I I I I I I I I I I I 200 ms/div

FIG. 4. Activity of a frontal eye field movement neuron antidromically excited from the superior colliculus (A, B) and multineuron activity recorded at the collicular stimulation site with lowest threshold for antidromic excitation of the frontal eye field neuron (C, D). Activity was obtained during performance of the memory-guided saccade task (Fig. lD), and each pair of illustrations compares activity synchronized on the appearance of the target (A, C) with activity synchronized on the beginning of the saccade (B, D). Each portion of this figure includes a raster, histogram, and sample eye movement traces for the paradigm employed. Each raster dot represents one neuron spike (sampled at 250 Hz), every raster line includes the neuron activity from a 2-s interval of a single trial. Each histogram is the summation of the raster illustrated above it with 4-ms bin width. The calibration mark to the left of each histogram FRONTAL EYE FIELD CORTICOTECTAL NEURONS 1395 and peaking at the beginning of the saccade. separated in time by ~750 ms. The activity The center of the movement field for this of this neuron in the visually guided saccade neuron was 5’ from the fixation point in the task is shown in Fig. 6, C and D. Note the lower right quadrant (5” radius, 308’ angle brisker response for the visually guided sac- in polar coordinates). The latency for anti- cade, in which both the movement and the dromic excitation from the superior collicu- visual components occur simultaneously. lus was 1.4 ms. An antidromic excitation The latency for antidromic excitation of this threshold of 110 PA was first recorded for neuron was 1.2 ms. The minimum current this neuron while stimulating from a depth threshold for excitation was 160 PA. Figure 7 within the superior colliculus where purely shows photomicrographs of the recording visual activity was recorded prior to stimula- and stimulation sites for this neuron. As was tion. However, the minimum threshold for the case for the movement neuron described this neuron was obtained when the stimulat- above, the recording site was localized to ing electrode was lowered 1 mm deeper into cortical layer V, and the stimulation site with the superior colliculus. At this depth, the an- lowest current threshold was within the in- tidromic excitation threshold was 70 PA. termediate gray layer of the superior collicu- The threshold for excitation increased as the lus. Neurons with visuomovement activity electrode was advanced further, reaching a occurred infrequently in our sample of fron- level of 115 PA at a depth of 1.5 mm below tal eye field corticotectal neurons (3 neurons, the initial stimulation site. Multiunit activity 5.9%). Each of the other identified visuo- recorded at the superior colliculus stimula- movement neurons had activity that resem- tion site with lowest threshold showed a pre- bled that shown in Fig. 6 with a weak visual ponderance of activity surrounding the be- response overshadowed by strong movement ginning of the eye movement, with only a activity. slight visual response as is shown in Fig. 4, C The activity preceding saccades for both and D. The center of the movement field for visuomovement and movement corticotectal this multiunit activity was of greater ampli- neurons occurred before both visually guided tude, but in roughly the same direction into and memory-guided saccades. the lower right quadrant ( 11 O radius, 328’ 3. FOVEAL NEURONS. Eleven neurons angle). The lesions marking the recording (21.6%) responded to visual stimulation of site for this neuron in the frontal eye field, as the retinal fovea. Fovea1 neurons included well as the stimulation site with lowest six neurons that exhibited a relatively un- threshold in the superior colliculus are illus- complicated response to either the onset or trated in Fig. 5. The lesion marking the loca- disappearance of a fovea1 stimulus. The re- tion of the corticotectal neuron was within maining five neurons exhibited fovea1 re- cortical layer V. The site of lowest threshold sponses in combination with activity that for antidromic excitation was within the in- was anticipatory, evoked by peripheral visual termediate gray layer of the superior collic- stimuli, or related to some other behavioral ulus. or environmental parameter. The response 2. VISUOMOVEMENT NEURONS. we anti- of a foveal-on neuron is illustrated in Fig. 8. dromically excited three visuomovement This neuron increased its firing rate immedi- neurons (5.9%), one of which is shown in Fig. ately after the appearance of the central fixa- 6. In the memory-guided saccade task, this tion light at the start of the trial (Fig. 8A) in neuron had a weak visual response (Fig. 6A) the fixation task. Its activity level decreased followed by much stronger activity asso- during the time when the fixation light was ciated with the eye movement (Fig. 6B). turned off during the trial (Fig. 8B) and was Note that for this task, the beginning of the reactivated following the reappearance of the visual response and movement activity are light (Fig. 8C). Figure 9 shows the activity of

FIG. 4, cont. represents a firing rate of 100 spikes/s. The vertical line passing through both raster and histogram is the point of alignment for the activity. The abbreviations and conventions for the eye movement and stimulus traces at the top of each raster are identical to those of Fig. 1. 1396 M. A. SEGRAVES AND M. E. GOLDBERG

anterior

ran

B . ai-3 cm ‘i

FIG. 5. Recording and stimulation sites for the same corticotectal neuron illustrated in Fig. 4. A: recording site for antidromically excited neuron marked by a lesion in cortical layer V in the fundus of the arcuate sulcus--black- ened in drawing of section through arcuate sulcus, and indicated by white arrowhead in photomicrograph of same section. Cortical sections were made in a plane that was roughly parallel to that of the principal sulcus. The neuron was located in the left frontal eye field, and the cortical sections are oriented so that anterior is to the left, posterior to the right. In the section drawing, the shaded area on the anterior bank ofthe arcuate sulcus marks a region of necrotic tissue, probably resulting from the effects of previous penetrations. P.S.,principal sulcus; AS, arcuate sulcus. B: lesion marking stimulation site in the intermediate gray layer of the left superior colliculus. Coronal section. SGS, stratum griseum superticiale; SGI, stratum griseum intermediale; SGP, stratum griseum profundum. FRONTAL EYE FIELD CORTICOTECTAL NEURONS 1397

FP L FP 1 T I L T

. . . ,* . ..s. . s-.--w ...... w w ...... * _..L ...... , ...... IS .e......

. . m--m . . . . .

...... MLI.... . e

...... m--w.” .

...... “.-... .

..-m. . -...... w-.-.m. .

. .M ...... -we-. . . .

...... , . . . . . “.--...... e . ..w.-.-.- . . *

200 ms/div

C i f l loo FP [ FP -

200 ms/div

FIG. 6. Corticotectal visuomovement neuron activity. A, B: activity during memory-guided saccade task. C, D: visually guided saccade task. Rasters and histograms aligned on the onset of the peripheral target in A and C and on the beginning of the saccadic eye movement in I? and D. Note that the visuomovement neuron has both visual activity, seen in the first burst in A and B, and movement activity, seen in the second burst in A and B. In Cand D the visual activity burst and the movement activity burst are superimposed, and the activity is greater than either of the bursts in A and II. The visual nature of the onset of the burst is illustrated by its excellent synchrony with the appearance of the target in C. The burst continues through the movement in each trial. 1398 M. A. SEGRAVES AND M. E. GOLDBERG

anterior Asposterior

- $jj3? 1

FIG. 7. Recording and stimulation sites for the corticotectal neuron illustrated in Fig. 6. Same conventions as used in Fig. 5. A: the lesion marking the recording site for this neuron was in layer V in the posterior bank of the arcuate sulcus near the fundus of the sulcus. B: the stimulation site was in the intermediate gray layer of the superior colliculus. a neuron that at first appeared to be move- in the memory-guided saccade task and was, ment related (Fig. 94. However, this activity in fact, best interpreted as a visual response occurred before saccades in every direction to the disappearance of the fixation light FRONTAL EYE FIELD CORTICOTECTAL NEURONS 1399

A B C H H H V V V loo [ FP f FP ,-I I FP 1 I

I II 1 III I III I I I I I I I I I I I I I I I I I I I I I I 200 ms/div

IG. 8. Corticotectal neuron with foveal-on response. Fixation task. Rasters and histograms aligned. A: onset of the fixation light at the beginning of the trial. B: disappearance of fixation light while the monkey maintained fixation. C: reappearance of the fixation light 500-ms later.

(Fig. 9B). In the fixation task, this neuron visually guided saccade task where the pe- became quiet when the fixation light was ripheral target light was continually present. turned on at the beginning of the trial (Fig. The omnipresence of the peripheral target 9C). It fired during the period when the fixa- prevented it from evoking a phasic response tion point was off in the middle of the trial from the neuron. The neuron was active (Fig. 90) and ceasedfiring -75 ms after the while the monkey fixated the central fixation reappearance of the fixation point (Fig. 9E). point, but it showed a suppression of activity It became active once more, 100 ms after the when the fixation point disappeared. This disappearance of the fixation point at the end suppression was followed by a reactivation of of the trial (Fig. 9F). It should be noted that the neuron’s activity coincident with the the activity of this neuron in the saccadetask onset of the saccade that would attain the is not fully explained by a foveal-off re- target, that is the future fixation point, in the sponse, since the increase in firing rate actu- periphery. Note that this increase in firing ally occurs before the disappearance of the occurs before the target light is foveated. fixation light (Fig. 9B). This may represent There did not appear to be a preferred target activity that anticipates the impending eye location associated with this activity. The movement. In the fixation task, where sac- neuron’s activity was linked to the fixation of cades are suppressed,the neuron’s firing rate a peripheral target light, rather than the ret- did not increase until after the disappearance inotopic location of that target. of the fixation light (Fig. 90). Corticotectal fovea1 neurons did not ap- As mentioned above, about one-half pear to be localized to a particular portion of (5/ 11) of the fovea1 corticotectal neurons ex- the topographic map of the frontal eye field. hibited more than a simple on or off re- They were found in the vicinity of move- sponse. The activity of one of these complex ment neurons with preferred saccade ampli- fovea1 neurons is illustrated in Fig. 10. In the tudes ranging from 6 to 30’. Likewise, the fixation task, this neuron stopped firing fovea1 neurons were antidromically excited when the fixation point was turned off dur- from a wide range of points within the topo- ing the trial (Fig. 1OA) and resumed firing in graphic representation in the superior collic- response to the reappearance of the fixation ulus. light (Fig. 1OB). However, its firing rate was affected by more than fovea1 visual stimula- 4. POSTMOVEMENT NEURONS. There was tion alone. Figure 10, C and D, shows this one neuron (2.0%) in our sample of cortico- neuron’s activity during a variation of the tectal neurons classified as postmovement. 1400 M. A. SEGRAVES AND M. E. GOLDBERG

FP; FP 1 FP I T1 T1

. .-. - ...... “. . . . ._...“...... s . -.-.. . .-, ...... -...... w ...... , ...... - ...... -...... - ...... s...... m .. . .-. .... * . . . . . - . . . . . -. - ...... - ...... - ...... -...... -...... w ...... - ...... -...... -...... “._...... - ... .-...... w C.. ... w ...... -. . ..m -...... ” ...... - ...... -. ... w.- - ...... -...- . . . . . *.*I . . SW.“..“...... -. ‘I”. . -----**-. --... . .* ......

*

‘ I II I I 1 I 1 200 ms/div

D E F H H H V V V loo [ FP I 1 FP t I FP 1

. . . .

. 1. . . . - . . . .

.-. .

. . ..^.. . ..“..

FIG. 9. Neuron responsive to disappearance of fixation light. A, B: memory-guided saccade task. Data in A aligned to beginning of eye movement, B aligned to disappearance of fixation point-the signal to make a saccade. C-l? fixation task. Data in C aligned to when the fixation light appeared at the beginning of the trial. D, disappearance of fixation light during the trial; E, reappearance of the fixation light; and F, disappearance of the fixation light at the end of the trial.

Its activity was related to saccades, but began the visually guided saccade task and antici- after the beginning of the eye movement. pated the occurrence of the reward (1 neuron), 3) a burst of activity before an eye 5. OTHER NEURON TYPES. There were eight movement followed by a high activity rate up antidromically excited neurons ( 15.7%), four to the start of the next trial (1 neuron), 4) a in each monkey, that could not be classified sustained response to visual stimuli in the as visual, visuomovement, movement, post- ipsilateral hemifield (1 neuron), 5) activity saccadic, or foveal. They were a heteroge- that combined an enhanced visual response, neous group, and included neurons with I) anticipatory, and postmovement activity (1 high spontaneous activity that was sup- neuron), or anticipatory and postmovement pressed 100 ms before and during saccades to activity (1 neuron), and 6) neurons that dis- the contralateral hemifield (1 neuron), 2) ac- charged briskly during smooth pursuit in tivity that began after the eye movement in certain directions (2 neurons). The activity of FRONTAL EYE FIELD CORTICOTECTAL NEURONS 1401

...... - m......

.” . . .” . . . .

.... ” - ...... -a...... “...... m”...... " ...... w ...... “...““. . . . u...... “......

. em. .

.” ......

......

...... ” , ...... “..

...... ” . .” ......

......

. . . . .” . . . . . 8 ......

. . . . a...... “.

...... ” .W..” , ...... ”

. m ...... “m . .”

200 ms, div

FP FP T

. . .““. “. .. - .. ” ...... -. ... “......

-...-. ” ...... ““. .“-m.o.-...... _.““.

.“...... “. .“. ..-...... ” ...... “. .”

. ... _....“.“...... m S..“.“......

...... “.” .. ” . . . . . *.. . . .L .. ..” .... .

...... w . ..w ...... ” ......

” .... .” ... .“...... - .... .“...... - “...... ” ...

* . . “...... - ...... “..“...... - .. ...

...... - ...... I ... .-.-.e..- ...... ““.” ...... -..B” ...... “” ..- ...... ” .. .” m- ..” . . .w . . ..““.o. ..” ... me .rn..” ...... “.N.W” ...

...... “_. . . “. . . ..““. . .

...... “...... w m ... “I ...... - ..” ” .. .“.

...... w.m.- ...... -- ... .

.

200 ms/div

FIG. 10. Response of a complex fovea1 neuron in the fixation (A-B) and saccade (C-D) tasks. A: aligned on disappearance of fixation light during fixation task trial. B: reappearance of fixation light. C-D: activity in a variation of the visually guided saccade task where the peripheral target is always on. C: aligned on disappearance of fixation light. D: aligned on the beginning of the saccade. Note that in C and D, the target light status line begins and ends in the “on” condition, since the target light was never turned off in this task. The monkey was trained to always fixate the central light whenever it was present and make a saccade to and fixate the peripheral target light only when the central light was turned off. two of these neurons is described in more the initial large burst of firing at the begin- detail below. ning of fixation, the neuron’s activity re- Figure 11 illustrates the activity of a mained at a high rate throughout the trial neuron whose greatest activity occurred after and ended with a burst of activity that began the monkey attained fixation at the very be- after the start of the eye movement (Fig. ginning of the trial in the memory-guided 11 D). This postmovement component of the saccade task (Fig. 1 IA). This neuron did not activity was strongest after saccades to targets respond to either the onset or disappearance located on the horizontal meridian in the of the stimulus light (Fig. 11, B and C). After contralateral hemifield at - 12’ away from 1402 M. A. SEGRAVES AND M. E. GOLDBERG

*+==== B!y=E

FP J FP T I 1 T J lp -. . -. . T”“” ..-.w...... am.. . .---w- ..“..S...... Wm..

200 ms,

7 FP L T- T 1

.I”...... em .. . “I...... m.. .. .

-M.-w. ... S”. m..-m-...... , .-S.W”. .... “.w”...-- - ... ., ......

. . ... -w..#...... W”. . s ... ..-...... - ... ..-...... -...... * ...... -...w ... . .

S.-s...... -.. . - . . .. 6 . . . ..- .. ..I ...... M .- . . . .

...... a.. .. w...... “.I, . . .

.-- ...... w- .... M”..“...... w ...... w.- ...... w ...... a-- . -” ...... a.....-- .. .WW” .. . SW. .. -...... -e- I. - . . -. , . . . . . - ... -. .. ..-...a.-_. I ...... , ...... --.m ...... -..-...... “...“...... - ... . . “.....- ...... B-w ...... “...... s-s ..M”...... -...-. ..- . .... -w- s-.- ...... - ...... - . ..I . ...

-*- ...... -- .... - ...... --. ...

.s- .- ...... m... .-. .... -..- ... . -. .- .” .... “-w-w-...... -..-. .

/ 1 / *-

200 ms/div

FIG. 11. Activity of a corticotectal neuron with combined anticipatory and postmovement activity, during a memory-guided saccade task. A: activity aligned on the achievement of fixation, at the beginning of the trial. B: activity aligned on the appearance of the target. C: activity aligned on the disappearance of the target. D: activity aligned on the beginning of the saccade.

the center of gaze. It also occurred after sac- the monkey is required to make eye move- cades made in the dark. Although this ments to the same target location in consecu- neuron responded weakly to the reappear- tive trials. Under these conditions, the events ance of the fixation light in the fixation task, of the trial become predictable, and the the magnitude of the activity occurring be- monkey anticipates their occurrence (7). For fore the eye movement in the task illustrated the neuron shown in Fig. I 1, the early activ- is best described as a form of anticipatory ity was not a response to a visual stimulus. activity. In general, anticipatory neuron ac- The activity preceded the beginning of the tivity in the frontal eye field is strongest when eye movement by - 1,200 ms, so it could not FRONTAL EYE FIELD CORTICOTECTAL NEURONS 1403 be considered strictly movement related. In- perior colliculus, could not be activated dur- stead, it appears to have anticipated events ing any of the behavioral tasks. that would occur near the end of the trial, in particular, the saccade and its associated re- Neurons adjacent to ant idrom ically ward. excitable neurons Another neuron showed a modulation of Most antidromically excitable neurons response during different directions of were isolated during electrode penetrations smooth-pursuit eye movements. It did not where only responses to electrical stimula- respond to either peripheral or fovea1 visual tion of the superior colliculus were routinely stimuli, and it did not respond in saccade examined, and neuron activity during visual tasks. However, it was active while the mon- and eye movement tasks was studied only key tracked~ slowly moving targets with after an antidromically excitable neuron had smooth-pursuit eye movements. The neuron been isolated. However, we frequently re- preferred tracking movements directed up to corded the activity characteristics of all the left, a direction contralateral to the neurons encountered in a single penetration neuron’s location in cortex. It was also ac- to ascertain the overall distribution of tive, to a lesser extent, during smooth-pursuit neuron types that were being sampled by the eye movements down to the left, but was electrode. As an example, Fig. 12 illustrates nearly silent during tracking movements di- the activity of neurons encountered during a rected ipsilaterally. The neuron was best ac- penetration through 6 mm of cortex made tivated by slow stimulus velocity in the pre- with a platinum-iridium electrode. Because ferred direction. Using a target with sinus- of the length of this penetration, it is likely oidally varied velocity, and a peak-to-peak that the electrode traveled obliquely through excursion of 20”, the preferred frequency the cortical layers of the frontal eye field, be- was 0.1 Hz, producing a mean eye speed of ginning at layer I and ending in white matter 4”/s. This neuron had a high rate of activity after passing through layer VI. The thickness during the inter-trial interval, and its best ac- of the frontal eye field cortex from surface to tivity during smooth pursuit tasks did not white matter is normally ~2 mm. Two anti- exceed its activity during the intertrial inter- dromically excitable neurons were isolated val. Electrical stimulation (0.2-ms pulse during this penetration. Neuronal activity width, 330 Hz, 400-ms train, 75PA pulses) at recorded at the beginning of the penetration the recording site of this neuron in the fron- contained only visual responses, including tal eye field produced upward slow eye tonic multineuron visual responses recorded movements that began -45 ms after stimu- at a depth of 1,502 pm (Fig. 12A). Move- lus onset, and lasted until the stimulus was ment-related activity was first encountered at turned off. 3,300 pm, and at this depth the movement 6. NONRESPONSIVE NEURONS. one neuron component of the multineuron activity was (2.0%), antidromically excited from the su- stronger than the visual component. The first

FIG. 12. Neurons encountered during penetration through 6 mm of frontal eye field cortex. A-G shows neuronal activity. In each truce H and V are horizontal and vertical eye position, respectively, with 10” calibration mark shown for each pair. For each example of neural activity (A-G) two rasters and histograms are shown. In the left of each pair activity is synchronized on the appearance of the target. In the right of each pair activity is synchronized on the beginning of the saccade. A: multineuron tonic visual activity recorded at 1,502 pm. Memory-guided saccade task. Note that activity starts with stimulus and ends at movement. B: antidromically excited neuron with movement- related presaccadic activity, 3,75 1 pm. Memory-guided saccade task. Note absence of response to target onset (left). C and D: neurons with postmovement activity recorded at 3,742 and 3,841 pm. Memory-guided saccade task with zero delay between target light and fixation point disappearance. Note lack of response to target onset (left) and response beginning with beginning of saccade (right).E: second antidromically excited neuron isolated, premove- ment activity combined with strong anticipatory activity. Memory-guided saccade task with zero delay. Note activity preceding target onset (left) and burst occurring before movement (right).F and G: postmovement neurons. Note no activity synchronized on target onset. (left) and activity beginning after the beginning of the saccade (right).H indicates the locations of the recording sites for the neurons illustrated in A-G. The vertical rule shows distance traversed by electrode. The asterisks mark the sites at which antidromically excited neurons were found. The activity of these antidromically excited neurons is shown in B and E. See text for additional details. 1loo FP FP 7 FP FP 1 T f I T1 T I 1 T-

200 ms/div

H - v r 110° FP 1 FP T I ‘I T- ...... _---...... -...... - _. -.--.-._-_...... --...... -. t_...... -...... -__. -...... -._..--_.-_...... _. . _ ...... _____.__._...... _..._”...... _____ ...... _. I.... I,__--_-.-...... -...... - ......

200 ms/div FRONTAL EYE FIELD CORTICOTECTAL NEURONS 1405

*- . . . . *... . : . . . . . * . , . . . . * .: *. * .* :: . *-. . . *...... ‘1.: . *. :‘. .- * . . ’ ...... ‘i* a...... :;,* ;;: .:I:: i .i : . . . * : ,l;‘fl;!.~~;“i Iiq!WI$/i !ii~!!!!li:l.:/

$iI;;]]i[ji;j .:I : -- i’- , : ;;!ii;i!i;. I;‘!: a* -.,I’ Ii’t ;..,. . :. *I. a..- ;:i: I r-: :-!.. 0 :. :.. . . . : 0 . . : . .:i ;’ ; . - . . . LJ ..* . . ; . . : 7. . . . . 1406 M. A. SEGRAVES AND M. E. GOLDBERG antidromically excitable neuron was isolated lated at 4,050 pm (Fig. 12E) was the sup- at 3,75 1 pm (Fig. 12B). This neuron was pression of its activity surrounding eye classified as a movement neuron. It in- movements in the off direction. When the creased its firing rate - 100 ms before the monkey was required to make a saccade di- start of the eye movement. However, the rected away from the optimal direction for bulk of its activity occurred after the begin- the movement activity of the neuron, there ning of the eye movement. A neuron isolated was a strong suppression of the neuron’s ac- adjacent to the corticotectal neuron at a tivity beginning 100 ms after the onset of the depth of 3,742 pm (Fig. 12C) had entirely target light in the visually guided saccade postmovement activity; it did not begin to task, and continuing through the eye move- fire until after the beginning of the eye move- ment (Fig. 13). The optimal presaccadic ac- ment. A second postmovement neuron was tivity for this neuron was obtained when the encountered at 3,841 pm (Fig. 120). At target light was presented on the horizontal 4,050 pm, the second antidromically excit- meridian, 10’ to the right. The suppression able neuron was encountered (Fig. 12E). of activity surrounding the eye movement This was also a movement neuron, however, was optimal when the target light was pre- its activity was very different from that of the sented at 10’ to the left of the center of gaze, first corticotectal neuron. It exhibited strong near the horizontal meridian. It is interesting anticipatory activity, beginning immediately to note that the optimal target location for after the target flash onset in the memory- the postmovement neurons illustrated in Fig. guided saccade task, and included a pre- 12 was in the lower left (Fig. 12, C, D, and F) movement burst of activity that peaked be- and lower right (Fig. 12G) quadrants. The fore the start of the eye movement. Two ad- optimal target location for the visual neurons ditional postmovement neurons were recorded at the beginning of this penetration isolated at 4,176 and 4,70 1 pm (Fig. 12, F were in the upper right quadrant. Together and G). The penetration was ended at 6,000 these results suggest that frontal eye field vi- pm when the electrode entered white matter. sual neurons have an excitatory effect upon One of the most impressive components of movement neurons, and postmovement the activity of the corticotectal neuron iso- neurons have an inhibitory effect. However,

A B

H V loo FP 1 FP 1 T t I T t

...... s.

......

. . .

200 ms/div

FIG. 13. Suppression surrounding eye movements in the off direction for antidromically excited neuron isolated at 4,050 pm in the penetration illustrated in Fig. 12 (this neuron’s activity is shown in Fig. 12~9. Suppression begins 100 ms after the onset of target light ( 12,176”) in visually guided saccade task (A) and continues until after the end of the eye movement (B). FRONTAL EYE FIELD CORTICOTECTAL NEURONS 1407

n

n = 45 Mean = 330 pamps SD = 324 pamps

~0O- :B $1 2 o= n = 24

SD = 265 pamps n = 26 Mean = 1.97 ms SD = 0.89 ms

0 1000 1500

0 5 10 0 0 n 0 d- 0 m 0 cl 0 -

0 1000 1500 Threshold (pamps)

5 10 FIG. 15. Percent distributions of thresholds for anti- Latency (ms) dromic excitation. Bin width is 50 PA.

FIG. 14. Percent distribution of latencies for antidro- mic excitation. A: distribution for all corticotectal not movement activity. We isolated more neurons isolated in this study. B: distribution for neurons classified as movement. C: neurons classified than 100 of these neurons during the course as foveal. Bin width is 0.5 ms. of our exploration of the frontal eye field, and we were never able to excite one anti- dromically. We made a number of penetra- these effects do not fully account for the ac- tions specifically looking for visual neurons, tivity of the corticotectal neuron illustrated both through guide tubes from which we in Figs. 12E and 13, since the suppression of successfully isolated corticotectal neurons, its activity begins 100 ms before the start of and from simple searching penetrations the eye movement, and thus precedes the ac- made after passing through the dura with tivity in the postmovement neurons. glass-coated platinum-iridium electrodes. One of the striking results of this study was Because visual neurons were common, and our failure to excite antidromically neurons antidromic excitation technically feasible in that had nerinheral visual recebtive fields but our hands. we feel it is unlikelv that the fron- M. A. SEGRAVES AND M. E. GOLDBERG

TABLE 2. Mean latencies jbr antidromic excitation

Mean SD SE n P

A. AlI latenciesjor ail neurons All neurons 2.25 1.12 0.16 50 Movement neurons 1.97 0.89 0.17 26 NS Fovea1 neurons 3.19 1.41 0.42 11

Entries include all corticotectal neurons isolated in the frontal eye field, neurons whose activity was classified as movement, and those classified as foveal. Significance levels are for comparison of movement or fovea1 to all neurons, and comparison of movement to foveal. A, includes all latencies for all neurons in each category; B, latencies >5.0 ms have been excluded. tal eye field sends a significant visual signal a standard deviation equal to 324 PA, and to the superior colliculus. ranged from 6 to 1,200 ,uA (Fig. 15). To ensure that high stimulation currents Antidromic Latencies were not exciting extracollicular frontal The latencies for antidromic excitation fibers, we studied the threshold for antidro- were unimodally distributed (Fig. 14). The mic stimulation as a function of location of mean latency for all corticotectal neurons the collicular electrode for 22 of the 5 1 iso- was 2.25 ms (range 1.2-6.0 ms). Movement lated antidromically excitable neurons. To neurons had a mean latency of 1.97 ms that do this we moved the stimulating electrode was slightly shorter than that for all cortico- through the superior colliculus and recorded tectal neurons, but this difference was not the threshold current for antidromic excita- significant. Fovea1 neurons had a mean la- tion at 250-pm intervals in depth. In a few tency of 3.19 ms, which was 0.9 ms longer cases, an electrolytic lesion was made at the than the mean for all corticotectal neurons, minimum threshold site so that its position and 1.2 ms longer than the mean for move- could be determined histologically. Two ment neurons. This difference was signifi- such lesions have been reconstructed, and cant in both instances (Table 2A). Three each was located within the intermediate neurons had latencies >5 ms, and appear gray layers of the superior colliculus (see separate from the main distribution in each Figs. 5 and 7). However, it was difficult to graph. However, when these longer latencies recover electrolytic lesionsmade more than a were excluded from the calculation and com- few weeks before the time at which the mon- parison of means, the same relationships de- key was perfused, and the number of lesions scribed above were maintained (Table 2B). made in each monkey was limited so that The estimated conduction velocity of single lesion sites could be identified with a frontal eye field corticotectal neurons based high degree of certainty. As an alternative upon their mean antidromic excitation la- means for identifying the anatomical sites tency is 18 m/s (range 7-34 m/s). This as- with lowest threshold, the locations in depth sumes an estimated distance from fundus of where minimum thresholds were obtained the arcuate sulcus (AP+22.0), through inter- were correlated with the neuronal activity nal capsule and thalamus, to midcolliculus properties recorded at those depths. For 14 of (AP- 1.5) of 40.5 mm. the 22 antidromic neurons examined, such comparisons could be made directly with Antidromic thresholds collicular neuronal activity recorded during The current thresholds for antidromic ex- the threshold determination experiment. For citation had a mean of 330 PA (n = 45), with the other eight antidromic neurons neuronal FRONTAL EYE FIELD CORTICOTECTAL NEURONS

Multiunit Responses: 0 Surface Visual Visual Visual 1 Visual > Saccade Saccade > Visual

Saccade > Visual 2 Saccade > Visual Deep layer

t”“l”“l”“l”“l”“I”“; 0 100 200 300 400 500 600

Threshold for Antidromic Activation (ramps)

FIG. 16. Threshold for antidromic excitation compared with depth from collicular surface. Based upon multi- neuron activity indicated on the right side of the figure, the transition between superficial and intermediate layers was at a depth of 1 mm from the collicular surface. The border between intermediate and deep layers was at 2.5 mm. The minimum threshold for antidromic excitation occurred at the transition point between intermediate and deep layers. Antidromic excitation thresholds increased both above and below this depth. activity could not be recorded by the stimu- tivity was diminished in strength relative to lation electrode, but the stimulating elec- that recorded at 1.750 mm. At 2.500 mm, trode depths could be determined by com- the phasic movement burst associated with parison with other penetrations. By these cri- intermediate layer neuron activity was re- teria each of the 22 neurons tested had their placed by activity that increased before the minimum thresholds within the superior col- monkey achieved fixation and continued liculus. The minimum thresholds for these until after the end of the eye movement in neurons ranged from 7 to 500 PA. the memory-guided saccade task. Of the 13 For all but one neuron, the minimum neurons where accurate correlations could threshold site in the colliculus was a reversal be made between threshold reversals and point, with increasing thresholds above and collicular neuronal activity, the reversal below the minimum threshold depth. For points of three neurons were located at one neuron, thresholds were only tested depths where superficial layerlike properties above the minimum threshold site, so we were present, including a prominent visual cannot rule out the possibility that its thresh- response, little or no movement activity, and old would have decreased at lower depth. a relatively high threshold (generally >50 Figure 16 shows a comparison of antidromic PA) for electrically evoked eye movements. threshold and collicular neuronal activity for Nine neurons had reversal points at depths a single neuron stimulated at various sites on where movement activity was recorded and a colliculus penetration. Note that the thresholds for electrically evoked eye move- threshold reached a minimum at a depth in ments were low ~50 PA). These include the the colliculus of 2.5 mm. This depth corre- following. I) Two neurons with reversal sponded to the transition between interme- points at the transition between superficial diate and deep layer activity. Just above the and intermediate layerlike activity-identi- reversal point, at 2.250 mm, the threshold fied by the presenceof both visual and move- for electrically evoked eye movements began ment activity components, with a promi- to increase with increasing depth, and the nence of visual over movement activity. 2) movement component of the multiunit ac- Five neurons with reversal points located at B

ii ~~8 200 ms/dlv 200 ms/dlv

FIG. 17. Comparison of frontal eye field and superior colliculus premovement activity for low-threshold corticotectal neuron. A: activity in visually guided saccade task for a frontal eye field neuron excited from the superior colliculus with a threshold of 7 PA. B: single-neuron activity recorded at low-threshold site within the superior colliculus. Rasters are aligned upon target onset. In both A and B the amplitude of the target’s displacement from the fixation point was 10”. FRONTAL EYE FIELD CORTICOTECTAL NEURONS 1411

-2 zq 001

4 f4 .\ / )

\ / 4 1412 M. A. SEGRAVES AND M. E. GOLDBERG sites where activity characteristic of the in- We have con rpared antidromic excitation termediate layers was recorded. These char- thresholds with a value expressing the vector acteristics include strong movement activity, difference between frontal eye field and col- weak or no visual response, and the lowest licular saccade vectors for data obtained for threshold for electrically evoked eye move- 25 frontal eye field visuomovement and ments. 3) Two neurons with reversal points movement neurons (Fig. 19). The “index of at the transition between intermediate and vector difference” was equal to the ampli- deep collicular layers. At this depth, there tude of the vector difference between opti- was a decrease in the strength of the move- mal frontal eye field and collicular saccade ment activity, a return of visual activity, and vectors divided by the sum of the amplitudes an increased threshold for electrically evoked of the two saccade vectors. This method was eye movements. One neuron’s reversal point chosen to enable us to compare data from was localized to a depth with deep layerlike both large and small eye movement sites. activity where, in addition to increased visual Our hypothesis was that stimulation and and decreased movement activity, the elec- recording sites with similar optimal saccade trical stimulus train used to evoke eye move- vectors would have low thresholds. Since ments appeared to be mildly aversive to the equal saccade vectors result in an index of monkey at currents of 75 PA, suggesting that vector difference of 0, and opposite saccade the electrode was in the vicinity of the central vectors result in an index value of 1, it was gray matter. expected that low thresholds would be asso- In general, the lowest thresholds occurred ciated with index values near 0, and high when the movement fields of the collicular thresholds with index values approaching 1. multiunit activity and the frontal eye field In fact, there was a significant linear correla- neuron were alike. An example of this rela- tion (Y = 0.62) between the regression of tionship is provided in Figs. 17 and 18. The index of vector difference upon log of thresh- corticotectal neuron illustrated in Fig. 17A old for antidromic excitation (P < 0.001). had movement activity that was strongest These data suggest that small increases in the preceding saccades to target positions on the vector difference between the stimulation vertical meridian, 10’ below the horizontal meridian. This neuron was antidromically excited from the superior colliculus with a n = 25 0 a minimum threshold of 7 PA. Premovement r = 0.62

p < 0.001 l l single neuron activity recorded at the collic- 8 l l ular stimulation site (Fig. 17B) had a pre- l ferred target location that was similar to the best target location for the frontal eye field neuron. In contrast, the optimal target loca- tions for the corticotectal neuron illustrated in Fig. 18A and multineuron activity re- corded at its collicular stimulation site (Fig. 18B) differed in both amplitude and direc- tion from the center of gaze. The center of 1 o-o,,, I I rI11.., I I I .““7 1 10 100 1000 the movement field of the frontal eye field Threshold (ramps) neuron was at 15” radius and 45” angle, and at 10’ radius and 270” angle for the collicu- FIG. 19. Effect of topography of recording and stimu- lar activity. The minimum threshold for an- lation sites upon threshold for antidromic excitation. Index of vector difference is equal to the amplitude of tidromic excitation of this visuomovement the vector difference between optimal saccade vectors at neuron was 160 PA. The ideal situation ex- frontal eye field and collicular sites divided by the sum of emplified by the electrode pair illustrated in the amplitudes of the two vectors. The index of vector Fig. 17 was rarely realized, however, because difference would equal 0 for identical saccade vectors and 1 for saccade vectors that are in opposite directions. of the different topographical organizations The data were taken from 25 antidromically excited vi- of the superior colliculus and frontal eye suomovement and movement neurons. Threshold is ‘. field. nlotted on a log scale. FRONTAL EYE FIELD CORTICOTECTAL NEURONS 1413 and recording sites result in large increases in frontal eye field and superior colliculus, as threshold. For example, a 23% increase in well as several technical aspects concerning difference between saccade vector angles, these experiments. that is for saccades of equal loo amplitude Nonhomogeneous projection of frontal eye but increase in angular difference from 44 to jield activity to the colliculus 54”, would be accompanied by a doubling of threshold for antidromic excitation from 50 Figure 20 compares the distributions of to 100 ,uA. our population of identified corticotectal neurons (solid bars) with the population of all frontal eye field neurons described by DISCUSSION Bruce and Goldberg (hatched bars) (7). In this study, we have demonstrated that Fifty-three percent of the corticotectal the projection from the frontal eye field to neurons were classified as having movement the superior colliculus, first described using activity, compared with 11% in the general anatomical techniques, can also be shown sample. The second largest single group of using electrophysiological techniques in corticotectal neurons were those with fovea1 awake monkeys. The most striking result is responses. They included 20% of the cortico- that there seems to be a selective enrichment tectal neurons, but only 7% of all frontal eye for two neuron types, those in which move- field neurons. There were no neurons anti- ment activity predominates, and those with dromically excited from the superior collic- fovea1 visual activity. There is also a selection ulus that could be classified as preferring against two types of neurons, those with pe- purely peripheral visual stimuli. Neurons ripheral visual receptive fields and those whose predominant activity occurs during whose discharge only occurs during and after and after the saccade only rarely project to saccades. We will discuss the implications of the superior colliculus. these results for our understanding of the Given such a striking difference between generation of saccadic eye movements by the all the neurons in the area and the collicular

All frontal eye field neurons (Bruce & Goldberg, 1985), n = 752 0 q GO q Neurons activated antidromically from superior colliculus, n = 51

Movement

Visual movement

FIG. 20. Comparison of distribution among activity categories of frontal eye field neurons excited antidromically from the superior colliculus (solid bars) to the distribution of all frontal eye field neurons reported by Bruce and Goldberg (7) (hatched bars). M. A. SEGRAVES AND M. E. GOLDBERG projection neurons, we must consider current spread are only rough approxima- whether or not this difference could be an tions. experimental artifact. The first possibility is Frontal eye field axons projecting to the always sampling error. However, in order to midbrain and pons follow both transtha- gather our modest population of antidromic lamic and pedunculopontine pathways (26, neurons we had to record from a much larger 32) (Stanton, personal communication). population of frontal eye field neurons. We Transthalamic fibers travel within and adja- made a special effort to find neurons of the cent to the internal medullary lamina. At the types that we could not drive antidromically. level of the posterior thalamus, the transtha- Therefore probabilistic sampling error is un- lamic pathway bifurcates, with corticotectal likely. The second possibility is that there is a axons passing dorsally through the pretec- physical difficulty that makes it possible for turn to terminate in the superior colliculus, antidromic spikes to invade the somata of and the remaining axons passing ventrolat- some corticotectal neurons, but not others. erally toward the mesencephalic reticular This is unlikely because anatomical studies formation and pontine tegmentum. Frontal showing the somata of frontotectal projec- eye field axons within the pedunculopontine tion neurons fail to distinguish multiple pathway travel within the cerebral peduncle morphologies for such neurons. They are all and project to the pontine and reticularis large layer V pyramidal neurons (15, 34). tegmentis pontis nuclei. Recent evidence Therefore it is likely that the population of suggests that this pathway does not include antidromically excited neurons was not terri- corticotectal axons (Stanton, personal com- bly biased by a systematic error. munication). The next problem is whether or not an Although the potential range of our stimu- electrode in the superior colliculus excites lating currents were often large, there was no frontal eye field fibers other than frontotectal indication that neurons antidromically ex- ones. This might seem especially likely in cited in the frontal eye field were being ex- view of the large currents needed to excite cited by current spread outside of the supe- some neurons. Threshold currents for anti- rior colliculus. In every case tested (4 1% of dromic excitation ranged from 6 to 1,200 the antidromically excited neurons), the re- PA, with a mean of 330 PA and standard versal point in depth where the minimum deviation of 324 PA. The administration of threshold for antidromic excitation was ob- these currents within the superior colliculus tained was located within the superior collic- undoubtedly resulted in both passive and ulus. Thus it is unlikely that we were stimu- transsynaptic spread of electrical activity for lating frontal axons projecting to the mesen- some distance away from the tip of the stim- cephalic reticular formation or the region of ulating electrode (2, 38, 42). An estimate the oculomotor nuclei instead of frontotectal based upon Ranck’s review of several studies axons. of mammalian CNS stimulation suggests It is most likely that high thresholds for that the spread of passive current from a antidromic excitation were the result of a 330-PA source could excite axons l,OOO- lack of overlap between the movement fields 1,500 pm from the electrode tip. A ~-PA at the site of the collicular electrode and the source (one standard deviation below the movement field of the antidromically excited mean threshold in this study) could excite neuron. The frontal eye field-collicular path- axons from 70 to 140 pm away, and a 654-PA way connects equivalent points within each stimulus has a potential range of 1,300- topographic representation (29, 55), and our 1,800 pm. Our stimulation sites were pre- efforts to stimulate and record from sites sumably in the vicinity of unmyelinated ax- with similar movement fields were meant to onal endings and terminals of frontal eye take advantage of this relationship. However, field neurons. The experiments reviewed by it was often difficult to position our elec- Ranck involved stimulation of myelinated trodes within overlapping movement fields axons only. Unfortunately there have been because of the different topological represen- no reports of the relative sensitivities of axon tations of the intermediate layers of the supe- terminals versus axons en passage to applied rior colliculus and the frontal eye field. The extracellular current. Thus our estimates of colliculus contains a simple visuotopic map FRONTAL EYE FIELD CORTICOTECTAL NEURONS 1415 of desired movement (45), but the frontal eye effect upon saccade dynamics, including sac- field combines a dorsoventral representation cadic velocity ( 10, 25). of saccade amplitude superimposed upon re- petitive representations of saccade direction, FOVEAL NEURONS. An unexpected finding such that a single saccade vector is repre- in these experiments is the strength within sented at multiple nonadjacent locations (8, the frontotectal projection of foveally re- 46). This produced frequent changes in sac- sponsive neurons. Fovea1 neurons comprised cade direction during a single frontal eye a small fraction of the total population of field penetration and resultant mismatches frontal eye field neurons examined by Bruce with the fixed collicular saccade vector. In and Goldberg (7), but formed the second the cases where the saccade vectors over- largest population of frontal eye field corti- lapped, the threshold was usually quite low; cotectal neurons. Twenty-two percent of an- when they did not overlap, the threshold was tidromically excited neurons were classified higher. as fovea1 neurons versus 7% in the total pop- ulation of frontal eye field neurons. These Functional aspects offrontal eye field neurons were typically suppressed or excited neuron types by fovea1 light stimuli and discharged most MOVEMENT NEURONS. We have demon- strongly in our task at the signal to make a strated that the principal component of the saccade or during attentive fixation, respec- frontal eye field’s input to the superior collic- tively. There were equal numbers of foveal- ulus consists of increased neuronal activity on and foveal-off neurons. The observation preceding saccadic eye movements, both vi- that frontal eye field neurons discharge dur- sually and memory guided. This input origi- ing fixation was first made by Bizzi (4). The nates from neurons with presaccadic activity neurons were studied most extensively by in cortical layer V that terminate in the supe- Suzuki and his co-workers (56, 57), who rior colliculus in a topographic manner. Our showed that the fovea1 responsiveness of results complement the findings of recent these neurons explained much but not all of stimulation and neuron recording studies (7, their activity. 8) showing that minimum threshold stimula- Fovea1 neurons may play a role in the act tion sites (as low as 10 PA) within the frontal of attentive fixation, which is known to be eye field are likely to be located near neurons physiologically different from merely main- with movement activity. Low-threshold sites taining the eye in a given orbital position (14, were generally located within cortical layers 19, 40). They send a message complemen- V and VI, and 63% of the movement and tary to that of the movement neurons: a sig- visuomovement neurons were located within nal either to suppress or initiate a saccade. It this region. In addition, Bruce and colleagues is possible that the artificiality of our task, in found a close correspondence between the which a visual stimulus bears the signal to optimal direction within the movement suppress or initiate a saccade, may be respon- fields of adjacent neurons and the saccade sible for the fovea1 responsiveness of these vector generated by electrical stimulation. neurons, which are otherwise more closely These present results suggest, therefore, that associated with a fixation or end-of-fixation one pathway by which electrical stimulation role in the untrained monkey. Presumably or natural neuronal activity in the frontal eye neurons active during attentive fixation field can initiate saccades is through this di- would provide a suppressive effect upon su- rect projection to the intermediate layers of perior colliculus movement neurons and the superior colliculus. This contribution thereby assist in the maintenance of fixation. will be better understood when we are able to This effect may be direct or indirect. Layer V determine the information content of the pyramidal neurons in the frontal eye field do signal carried by the axons of corticotectal not appear to contain the inhibitory trans- movement neurons. It is possible that the mitter y-aminobutyric acid (52), although frontal eye field’s movement output is specif- they may use another inhibitory neurotrans- ically related to motor parameters for the im- mitter. It is more likely that they act indi- pending saccade, since lesions of the frontal rectly via collicular interneurons. This po- eye field and superior colliculus both have an tential inhibitory effect of frontal eye field 1416 M. A. SEGRAVES AND M. E. GQLDBERG neurons is similar to that demonstrated for saccades. Fifty percent of frontal eye field vi- substantia nigra neurons that project to the sual neurons are enhanced before saccades to colliculus (24). The frontotectal suppressive the stimulus in their receptive field (62), and, effect would be stronger specifically during unlike other types of attentive movements, attentive fixation, unlike the nigra signal, this enhancement is spatially selective and which is always maximal, except around a specific to saccades (18). Our results suggest saccade. that these neurons participate in the intrinsic generation of a movement signal within the OTHER CORTICOTECTAL NEURON TYPES. frontal eye field, rather than project directly Two neurons in our sample of corticotectal to the oculomotor system. Of the visuo- neurons seemed to have activity related to movement neurons, some with strong move- the performance of smooth-pursuit eye ment activity project to the tectum, but those movements. This finding is of interest in with visual predominance do not. This result light of recent reports of deficits in smooth- reaffirms the likelihood that visuomovement pursuit eye movements in monkeys with neurons do not comprise a unique class of frontal eye field lesions (28, 36). Neurons ac- neurons by themselves, but rather belong to tive during smooth pursuit have been de- a continuum extending from visual to move- scribed previously in the frontal eye field (5), ment-related activity. where they were found at the same sites as- sociated with the evocation of smooth pur- Comparison to other corticotectal neurons suit by electrical stimulation (8). These Corticotectal neurons have not been exhaustively charac- STRIATE CORTICOTECTAL. neurons within striate cortex have been ex- terized to determine if they could be move- amined in both cats (41) and monkeys (13). ment neurons for pursuit or if they are in- In both species these neurons belong to a stead discharging in response to target move- subset of complex visual receptive-field ment or the retinal slip induced by the types. In cats, removal of visual cortical stationary environment on the moving ret- input to the colliculus results in a loss of di- ina. The superior colliculus has not been rection selectivity as well as input from the shown to have a role in smooth pursuit, and ipsilateral eye in superficial layer neurons it is possible that these neurons are, in fact, and also affects receptive-field properties of suppressing small saccades, similar to the deep layer neurons (47, 58). However, in the role postulated for the fovea1 neurons. monkey, ablation or cooling of visual cortex A few other neurons discharged in relation only affects the visual responses of neurons to anticipation of the reward or the next trial, within the intermediate and deep layers of were suppressed around saccades, or com- the superior colliculus (50). The average bined anticipatory, postmovement activity, conduction latencies for corticotectal and/or visual activity. These neurons were neurons in monkey striate cortex were longer unusual in these experiments and unusual in (4.6 ms) (13) than those in our sample of previous ones, and it is difficult at this point frontal eye field neurons (2.25 ms). The con- to understand their function or significance. duction velocity reported for striate cortico- NEURONS WITH PERJPHER +t VISUAL tectal neurons was an average of 8 m/s (range FIELDS. None of the corGcotecta1 neurons 3- 19 m/s) versus an estimated 18 m/s for in our sample could be classified as periph- frontal eye field corticotectal neurons. eral visual neurons. Thus one can assume Although prestriate and parietal neurons that the visual activity present in the superior have been shown by anatomical methods to colliculus does not originate from the frontal project to the intermediate layers of the supe- eye field. Only 10% of the visual neurons rior colliculus (3 1, 37) antidromic studies characterized in the work of Bruce and col- have not been done to characterize them leagues (8) were located within a region physiologically. Other prefrontal neurons, where thresholds for evoked eye movements especially those more anterior to the frontal were ~50 PA. Although they do not project eye field, have also been shown by anatomi- to the superior colliculus, visual neurons are cal methods to project to the colliculus (20, still likelv to be involved in the generation of 34), but the nature- of this projection is also FRONTAL EYE FIELD CORTICOTECTAL NEURONS 1417 not known. We did not attempt antidromic discharge before all saccades (49, 6 1) and the excitation in frontal areas that we had not region has a monosynaptic projection to the characterized as being in the low-threshold long-lead bursters in the brain stem reticular arcuate frontal eve4 field. formation (43). The arcuate frontal eye field can affect this presaccadic final common The role of the frontal eyefield in the path in three ways. The first is through the generation of eye movements direct targeting signal to the colliculus that These results contribute to the growing ev- we have demonstrated here. The second is idence that the arcuate frontal eye field is through the fovea1 signals, which could di- important in the initiation of purposive sac- rectly trigger or suppress a collicular saccadic cadic eye movements. We have now estab- signal. The third is by affecting the substantia lished that the oculomotor region of the su- nigra. The frontal eye field projects to the perior colliculus, the intermediate layers, re- head of the caudate, which contains a pre- ceives a distinct oculomotor message from saccadic signal similar in quality but reversed the frontal eye field. The message is twofold: in sign from that of the substantia nigra. This it tells about the state of fixation and fovea1 frontal-caudate signal could inhibit the sub- stimulation and sends a command to make a tantia nigra and result in a presaccadic re- movement of certain dimensions. lease of the nigral suppression of the superior The superior colliculus and the frontal eye colliculus. Thus the frontal command to the field can function independently in the gen- colliculus would be exceedingly powerful be- eration of saccades. Lesions of the superior cause it combines an excitatory signal with colliculus do not affect the generation of sac- the release of a suppressive one. When a cades by electrical stimulation of the frontal monkey is actively fixating, thresholds for eye field (48), and monkeys can make vi- evoking saccades from the superior colliculus sually guided saccades in the absence of the (22) are elevated. Presumably this is true be- frontal eye field or the superior colliculus, cause saccades that occur without changes in although they cannot make visually guided frontal and nigral activitv must overcome the saccades at all when both are ablated (5 1). suppression which inhibits them. Since Recent results suggest that some types of sac- monkeys and humans can make normal sac- cades do, in fact, require the frontal eye field. cades after lesions of the frontal eye field, the We have shown that monkeys with unilateral colliculus must be able to function on its frontal eye field lesions have difficulty learn- own, and under certain circumstances, for ing to make saccades to remembered targets, example spontaneous saccades made in total and when they do learn, the motor perfor- darkness, there is little or no frontal signal mance of memory-guided but not visually (4, 7). We submit, however, that when the guided saccades is impaired ( 10). Bruce has normal monkey makes purposive saccades, shown that although normal monkeys can these saccades are organizaed by the fronto- make predictive saccades, monkeys with tectal pathways discussed here. frontal lesions cannot (6). Guitton and col- leagues have shown that humans with frontal ACKNOWLEDGMENTS lesions have difficulty making saccades away The authors are grateful to the technical staff of the from a visual target, and instead make inap- Laboratory of Sensorimotor Research for their invalu- propriate saccades toward the target (2 1). able help: A. Ziminsky for electronic support, C. Crist Thus there is a repertory of saccades, charac- and T. Ruffner for machining, G. Creswell and L. Coo- terized by behavioral complexity, that re- per for histology, G. Snodgrass and J. Pellegrini for ani- quire the presence of the frontal eye field. mal care and surgical assistance, A. Hayes for computer hardware support, J. Steinberg for manuscript prepara- Our results suggest that in normal mon- tion, and N. Hight for facilitating everything. Dr. Lance keys the signal for these as well as visually M. Optican wrote the graphics package with which the guided saccades progress from the frontal eye illustrations were produced. We thank the photographic field to the intermediate layers of the supe- staff of the National Eye Institute for preparation of rior colliculus, which in normal monkeys figures. then serve as a final common path for sac- Received 29 March 1987; accepted in final form 30 cades (49). Neurons in the superior colliculus July 1987. 1418 M. A. SEGRAVES AND M. E. GOLDBERG

REFERENCES 1. ASTRUC, J. Corticofugal connections of area 8 antidromic excitation by the collision test: Problems (frontal eye field) in Macaca mulattta. Brain Rex of interpretation. Brain Res. 112; 283-298, 1976. 33: 241-256, 1971. 17. GOLDBERG, M. E. Studying the neurophysiology of 2. BAGSHAW, E. V. AND EVANS, M. H. Measurement behavior: Methods for recording single neurons in of current spread from microelectrodes when stimu- awake behaving monkeys. In: Methods in Cellular lating within the central nervous system. Exp. Brain Neurobiology, edited by J. L. Barker and J. F. Res. 25: 39 l-400, 1976. McKelvy. New York: Wiley, 1983, vol. 3, chapt. 8, 3. BISHOP, P. O., BURKE, W., AND DAVIS, R. Single- p. 225-248. unit recording from antidromically activated optic 18 GOLDBERG, M. E. AND BUSHNELL, M. C. Behav- radiation neurones. J. Physiol. Lond. 162: 432-450, ioral enhancement of visual responses in monkey 1962. cerebral cortex. II. Modulation in frontal eye fields 4. BIZZI, E. Discharge of frontal eye field neurons dur- specifically related to saccades. J. Neurophysiol. 46: ing saccadic and following eye movements in unan- 773-787, 1981. esthetized monkeys. Exp. Brain Rex 6: 69-80, 19. GOLDBERG, M.E., BUSHNELL, M.C., ANDBRUCE, 1968. C. J. The effect of attentive fixation on eye move- 5. BIZZI, E. AND SCHILLER, P. H. Single unit activity in ments evoked by electrical stimulation of the frontal the frontal eye fields of unanesthetized monkeys eye fields. Exp. Brain Res. 6 1: 579-584, 1986. during head and eye movement. Exp. Brain Res. 10: 20. GOLDMAN, P. S. AND NAUTA, W.J. H.Autoradio- 151-158,197O. graphic demonstration of a projection from pre- 6. BRUCE, C. J. AND BORDEN, J. A. The primate fron- frontal association cortex to the superior colliculus tal eye fields are necessary for predictive saccadic in the rhesus monkey. Brain Res. 116: 145- 149, tracking. Sot. Neurosci. Abstr. 12: 1086, 1986. 1976. 7. BRUCE,~. J. AND GOLDBERG, M.E.Primate fron- 21 GUITTON, D., BUCHTEL, H. A., AND DOUGLAS, tal eye fields. I. Single neurons discharging before R. M. Disturbances of voluntary saccadic eye move- saccades. J. Neurophysiol. 53: 603-635, 1985. ment mechanisms following discrete unilateral 8. BRUCE,~. J., GOLDBERG, M. E., STANTON,G. B., frontal lobe removals. In: Functional Basis of Ocu- AND BUSHNELL, M. C. Primate frontal eye fields. II. lar Motility Disorders, edited by G. Lennerstrand, Physiological and anatomical correlates of electri- D. S. Zee, and E. L. Keller. Oxford, UK: Pergamon cally evoked eye movements. J. Neurophysiol. 54: 1982, p. 497-500. 714-734, 1985. 22 GUTHRIE, B.L., PORTER, J.D., ANDSPARKS, D.L. 9. BUSHNELL, M.C., GOLDBERG, M.E., ANDROBIN- Corollary discharge provides accurate eye position SON, D. L. Behavioral enhancement of visual re- information to the oculomotor system. Science sponses in monkey cerebral cortex. I. Modulation in Wash. DC221: 1193-l 195, 1983. posterior partietal cortex related to selective visual 23. HIKOSAKA,~. AND SAKAMOTO, M.Cellactivityin attention. J. Neurophysiol. 46: 755-772, 198 1. monkey caudate nucleus preceding saccadic eye movements. Exp. Brain Res. 63: 659-662, 1986. 10. DENG,S.-Y.,GOLDBERG, M.E., SEGRAVES, M.A., 24. HIKOSAKA,O.ANDWURTZ, R.H.Visualandocu- UNGERLEIDER, L.G., ANDMISHKIN, M.The effect lomotor functions of monkey substantia nigra pars of unilateral ablation of the frontal eye fields on reticulata. IV. Relation of substantia nigra to supe- saccadic performance in the monkey. In: Adaptive rior colliculus. J. Neurophysiol. 49: 1285- 130 1, Processes in the Visual and Oculomotor Systems, 1983. edited by E. Keller and D. Zee. New York: Elsevier hd. HIKOSAKA, 0. AND WURTZ, R. H. Saccadic eye North-Holland, 1986, p. 20 l-208. movements following injection of lidocaine into the 11. EVARTS, E., KIMURA, M., WURTZ, R. H., ANDHI- superior colliculus. Exp. Brain Res. 6 1: 53 l-539, KOSAKA, 0. Behavioral correlates of activity in basal 1986. ganglia neurons. Trends Neurosci. 7: 447-453, 26. HUERTA, M. F., KRUBITZER, L. A., AND KAAS, 1984. J. H. Frontal eye field as defined by intracortical 12. FERRIER, D. The localization of function in the microstimulation in squirrel monkeys, owl mon- brain. Proc. R. Sot. Lond. B Biol. Sci. 22: 229-232, keys, and macaque monkeys: I. Subcortical connec- 1874. tions. J. Comp. Neurol. 253:4 15-439, 1986. 13. FINLAY, B.L., SCHILLER, P.H., ANDVOLMAN,S. F. 27. JUDGE, S. J., RICHMOND, B. J., AND CHU, F. C. Quantitative studies of single-cell properties in Implantation of magnetic search coils for measure- monkey striate cortex. IV. Corticotectal cells. J. ment of eye position: an improved method. Vision Neurophysiol. 39: 1352- 136 1, 1976. Res. 20: 535-538, 1980. 14. FISCHER, B. AND BOCH, R. Peripheral attention 28. KEATING, E. G., GOOLEY, S. G., AND KENNEY, versus central fixation: modulation of the visual ac- D. V. Impaired tracking and loss of predictive eye tivity of prelunate cortical neurons of the rhesus movements after removal of the frontal eye fields. monkey. Brain Res. 345: 111-123, 1985. Sot. Neurosci. Abstr. 11: 472, 1985. 15. FRIES,W. Cortical projections to the superior collic- 29. KOMATSU, H. AND SUZUKI, H. Projections from ulus in the macaque monkey: A retrograde study the functional subdivisions of the frontal eye field to using horseradish peroxidase. J. Comp. Neurol. 230: the superior colliculus in the monkey. Brain Res. 55-76, 1984. 327: 324-327, 1985. 16. FULLER.- J. H. AND SCHLAG. J. D. Determination of 30. KUNZLE. H. AND AKERT. K. Efferent connections FRONTAL EYE FIELD CORTICOTECTAL NEURONS 1419

of cortical area 8 (frontal eye field) in Macaca fascic- ments evoked by stimulation of frontal eye fields. J. ularis. A reinvestigation using the autoradiographic Neurophysiol. 32: 637-648, 1969. technique. J. Comp. Neural. 173: 147-164, 1977. 47. ROSENQUIST, A. C. AND PALMER, L. A. Visual re- 31. KUYPERS, H. G. J. M. AND LAWRENCE, D. G. Cor- ceptive field properties of cells of the superior collic- tical projections to the red nucleus and the brain ulus after cortical lesions in the cat. Exp. Neural. 33: stem in the rhesus monkey. Brain Res. 4: 15 1- 188, 629-652, 1971. 1967. 48. SCHILLER, P. H. The effect of superior colliculus 32. LEICHNETZ, G. R. The prefrontal cortico-oculomo- ablation on saccades elicted by cortical stimulation. tor trajectories in the monkey. J. Neuro. Sci. 49: Brain Res. 122: 154- 156, 1977. 387-396, 1981. 49. SCHILLER, P. H. AND KOERNER, F. Discharge char- 33. LEICHNETZ, G., SMITH, D. J., AND SPENCER, R. F. acteristics of single units in superior colliculus of the Cortical projections to the paramedian tegmental alert rhesus monkey. J. Neurophysiol. 34: 920-936, and basilar pons in the monkey. J. Comp. Neural. 1971. 228: 388-408, 1984. 50. SCHILLER, P. H., STRYKER, M. P., CYNADER, M., 34. LEICHNETZ, G. R., SPENCER, R. F., HARDY, AND BERMAN, N. Response characteristics of single S. G. P., AND ASTRUC, J. The prefrontal corticotec- cells in the monkey superior colliculus following tal projection in the monkey: An anterograde and ablation or cooling of visual cortex. J. Neurophysiol. retrograde horseradish peroxidase study. Neuro- 37: 181-194, 1974. science 6: 1023-1041, 1981. 51. SCHILLER, P. H., TRUE, S. D., AND CONWAY, J. L. 35. LEICHNETZ, G. R., SPENCER, R. F., AND SMITH, Deficits in eye movements following frontal eye D. J. Cortical projections to nuclei adjacent to the field and superior colliculus ablations. J. Neuro- oculomotor complex in the medial dien-mesence- physiol. 44: 1175- 1189, 1980. phalic tegmentum in the monkey. J. Comp. Neural. 52. SCHWARTZ, M. L., ZHENG, D. S., AND GOLDMAN- 228: 359-387, 1984. RAKIC, P. S. Laminar and tangential variation in the 36 LYNCH, J. C. AND ALLISON, J. C. A quantitative morphology and distribution of GABA-containing study of visual pursuit deficits following lesions of neurons in rhesus monkey prefrontal cortex. Sot. the frontal eye fields in Rhesus monkeys. Sot. Neurosci. Abstr. 11: 503, 1985. Neurosci. Abstr. 1 1: 473, 1985. 53. SEGRAVES, M. A. AND GOLDBERG, M. E. Func- 37. LYNCH, J. C., GRAYBIEL, A. M., AND LOBECK, L. J. tional properties of tectal projection neurons in the The differential projection of two cytoarchitectonic monkey frontal eye field. Sot. Neurosci. Abstr. 11: subregions of the inferior parietal lobule of macaque 472, 1985. upon the deep layers of the superior colliculus. J. 54. SPARKS, D. L. AND HOLLAND, R. Computer con- Comp. Neural. 235: 241-254, 1985. trol of eye position and velocity. Behav. Res. 38. MCILWAIN, J. T. Lateral spread of neural excitation Methods Instrum. 7: 115- 119, 1975. during microstimulation in intermediate gray layer 55. STANTON, G. B., BRUCE, C. J., AND GOLDBERG, of cat’s superior colliculus. J. Neurophysiol. 47: M. E. Organization of subcortical projections from 167-178, 1982. saccadic eye movement sites in the macaque’s fron- 39. MOHLER, C. W., GOLDBERG, M. E., AND WURTZ, tal eye fields. Sot. Neurosci. Abstr. 8: 293, 1982. R. H. Visual receptive fields of frontal eye neurons. 56. SUZUKI, H. AND AZUMA, M. Prefrontal neuronal Brain Res. 61: 385-389, 1973. activity during gazing at a light spot in the monkey. 40. MOUNTCASTLE, V. B., ANDERSON, R. A., AND Brain Res. 126: 497-508, 1977. MOTTER, B. C. The influence of attentive fixation 57. SUZUKI, H., AZUMA, M., AND YUMIYA, H. Stimu- upon the excitability of the light-sensitive neurons lus and behavioral factors contributing to the acti- of the posterior parietal cortex. J. Neurosci. 1: vation of monkey prefrontal neurons during gazing. 1218-1235, 1981. Jpn. J. Physiol. 29: 47 l-489, 1979. PALMER, L. A. AND ROSENQUIST, A. C. Visual re- 41. 58. WICKELGREN, B. G. AND STERLING, P. Influence of ceptive fields of single striate cortical units project- visual cortex on receptive fields in the superior col- ing to the superior colliculus in the cat. Brain Res. liculus of the cat. J. Neurophysiol. 32: 16-23, 1969. 67: 27-42, 1974. 59. WOLBARSHT, M. L., MACNICHOL, E. F., JR, AND 42. RANCK, J. B., JR. Which elements are excited in WAGNER, H. G. Glass insulated platinum micro- electrical stimulation of mammalian central ner- vous system: A review. Brain Res. 98: 417-440, electrode. Science Wash. DC 132: 1309- 13 10, 1960. 1975. 60. WURTZ, R. H. Visual receptive fields of striate cor- 43. RAYBOURN, M. S. AND KELLER, E. L. Colliculore- tex neurons in awake monkeys. J. Neurophysiol. 32: titular organization in primate oculomotor system. 727-742, 1969. J. Neurophysiol. 40: 86 l-878, 1977. 61. WURTZ, R. H. AND GOLDBERG, M. E. Activity of 44. ROBINSON, D. A. A method of measuring eye move- superior colliculus in behaving monkey. III. Cells ment using a scleral search coil in a magnetic field. discharging before eye movements. J. Neurophysiol. IEEE Trans. Biomed. Eng. 10: 137- 145, 1963. 35: 575-586, 1972. 45. ROBINSON, D. A. Eye movements evoked by collic- 62. WURTZ, R. H. AND MOHLER, C. W. Enhancement ular stimulation in the alert monkey. Vision Res. 12: of visual response in monkey striate cortex and 1795-1808, 1972. frontal eye fields. J. Neurophysiol. 39: 766-772, 46. ROBINSON, D. A. AND FUCHS, A. F. Eye move- 1976.