Mnemonic Coding of Visual Space in the ... - Stanford University

JOURNALOFNEUROPHYSIOLOGY Vol. 6 1, No. 2. February 1989. Printed

in U&4.

Mnemonic Coding of Visual Space in the Monkey’s Dorsolateral Prefrontal Cortex SHINTARO

FUNAHASHI,

Section of Neuroanatomy, SUMMARY

AND

CHARLES

J. BRUCE,

AND

Yale University School of Medicine,

CONCLUSIONS

1. An oculomotor delayed-response task was used to examine the spatial memory functions of neurons in primate prefrontal cortex. Monkeys were trained to fixate a central spot during a brief presentation (0.5 s) of a peripheral cue and throughout a subsequent delay period (l-6 s), and then, upon the extinction of the fixation target, to make a saccadic eye movement to where the cue had been presented. Cues were usually presented in one of eight different locations separated by 45O. This task thus requires monkeys to direct their gaze to the location of a remembered visual cue, controls the retinal coordinates of the visual cues, controls the monkey’s oculomotor behavior during the delay period, and also allows precise measurement of the timing and direction of the relevant behavioral responses. 2. Recordings were obtained from 288 neurons in the prefrontal cortex within and surrounding the principal sulcus (PS) while monkeys performed this task. An additional 31 neurons in the frontal eye fields (FEF) region within and near the anterior bank of the arcuate sulcus were also studied. 3. Of the 288 PS neurons, 170 exhibited task-related activity during at least one phase of this task and, of these, 87 showed significant excitation or inhibition of activity during the delay period relative to activity during the inter-trial interval. 4. Delay period activity was classified as div~tional for 79% of these 87 neurons in that significant responses only occurred following cues located over a certain range of visual field directions and were weak or absent for other cue directions. The remaining 2 1% were omnidirectional, i.e., showed comparable delay period activity for all visual field locations tested. Directional preferences, or lack thereof, were maintained across different delay intervals (l-6 s). 5. For 50 of the 87 PS neurons, activity during the delay period was significantly elevated above the neuron’s spontaneous rate for at least one cue location; for the remaining 37 neurons only inhibitory delay period activity was seen. Nearly all (92%) neurons with excitatory delay period activity were directional and few (8%) were omnidirectional. Most (62%) neurons with purely inhibitory delay period activity directional, but a substantial minority . . were . (38%) was ommdn-ectronal. 6. Fifteen of the neurons with excitatory directional delay period activity also had significant inhibitory delay period activity for other cue directions. These inhibitory responses were usually strongest for, or centered about, cue directions roughly opposite those optimal for excitatory responses. 7. The distribution of preferred cue locations was examined across the population of PS neurons having directional delay period activity. All possible cue locations were represented (left, right, up, down, and obliques); however, preferred cue locations in the contralateral hemifield predominated. 8. Tuning curves were calculated by the Gaussian formula to determine the directional specificity of delay period activity. The mean tuning indices (7’J of PS neurons were 26.8” for excitatory

0022-3077/89

$1 SO Copyright

PATRICIA

S. GOLDMAN-RAKIC

New Haven, Connecticut 06510 delay period activity and 43S” for inhibitory delay period activity. 9. Delay period activity was examined on trials in which the monkey made large saccadic errors for cues in a neuron’s preferred direction. Delay period activity was either truncated or absent altogether on such trials. IO. Of the 3 1 FEF neurons examined, 22 exhibited task-related activity: 17 had delay period activity and 10 showed directional delay period activity. The mean tuning index (Td) of excitatory directional delay period activity in the FEF was 27.4’. I I. These results indicate that prefrontal neurons (both PS and FEF) possess information concerning the location of visual cues during the delay period of the oculomotor delayed-response task. This information appears to be in a labeled line code: different neurons code different cue locations and the same neuron repeatedly codes the same location. This mnemonic activity occurs during the I- to 6-s delay interval-in the absence of any overt stimuli or movements-and it ceases upon the execution of the behavioral response. These results strengthen the evidence that the dorsolateral prefrontal cortex participates in the process of working or transient memory and further indicate that this area of the cortex contains a complete “memory” map of visual space.

INTRODUCTION

The role of the prefrontal cortex, particularly the principal sulcal (PS) region, in spatial memory has been strongly supported by lesion and developmental studies over the past two decades (17, 28, 57). Studies of single neuron activity in monkeys during performance of delayed-response tasks have also provided strong evidence of a prefrontal contribution to memory (4, 5, 15, 16,20,22, 37,4 1, 42, 46, 50-53). Although many types of neural activity have been found in the prefrontal cortex, neurons that show sustained activity during the delay period are particularly relevant to the issueof mnemonic processing. Usually these neurons increase their discharge rates following the brief cue presentation and continue firing tonically during the delay period until the response is executed (5, 15, 20, 22, 52, 53). Such delay activity is often “directional” (44, 50-52) i.e., dependent on the left-right location of the cue or response, with some neurons responding only in conjunction with “left” trials and others only with “right” trials. Because experimental lesions of the dorsolateral prefrontal cortex profoundly affect the ability of monkeys to perform spatial delayed-response tasks correctly ( 17, 28) this delay activity has been assumed to be the cellular expression of a mnemonic code for the left-right direction of the cue or, equivalently, of the impending response (17, 18, 28).

0 1989 The American

Physiological

Society

331

332

FUNAHASHI,

BRUCE, AND

In the present study we have explored further the properties of neural activity in the prefrontal cortex of monkeys performing delayed-response tasks. We sought to determine if prefrontal neurons were capable of accessing and maintaining activity reflecting the location of a target stimulus or motor response in any part of the visual field or if they were particularly tuned to left-right categorization. Further, we were interested in finding out if prefrontal neurons code stimuli in both the contralateral and ipsilatera1 visual fields equally well. To answer these questions, it was necessary to precisely control the retinotopic location of the cues, a requirement that has not been met in most previous studies. Accordingly, in the present study, we used an oculomotor analogue of the classical, manual delayedresponse task, in which the monkeys were trained to maintain fixation during the cue presentation, and the visual cue could be presented at multiple locations around the visual field. This method allowed us to record the monkey’s direction of gaze throughout the task, and consequently to know the precise retinotopic location of the peripheral visual cue. In addition, the fixation requirement during the delay period insured comparable behavior in this period on every trial and also made it difficult for the monkey to adopt postural rather than mnemonic strategies to perform the task. Similar oculomotor delay tasks have previously been used for neurophysiological experiments in the frontal eye fields (6), the posterior parietal cortex (30), the caudate nucleus (32), and the substantia nigra (33) as well as in the prefrontal cortex (37, 40). The present study, however, is the first to explore the full perimetry of visual space with this paradigm. Preliminary results from this study have previously been presented in abstract form (15). METHODS

Subjects Three adult rhesus monkeys (Maeaca mulatta, 3.2-5.3 kg, one male and two females) served as subjects. Prior to surgery and training they were adapted to a primate chair for 5- 10 days.

Surgical procedures

GOLDMAN-RAKIC

Behavioral techniques The monkey sat in the primate chair during the experiment and his head was fixed to the chair by the restraining receptacle. Recording and training sessionsusually lasted 2-3 h. A program on the PDP- 1l/23 computer presented the visual stimuli, sampled the monkey’s gaze coordinates as well as neuron activity, and delivered rewards. Visual stimuli were presented on a monochrome CRT (1g-in., RCA TC 1119) subtending 48 X 38” in visual angle using a GRAPH- 11 graphics card (Pacific Binary Systems). The fixation target was a small white spot (0.1’ diam.), usually presented at the center of the CRT. The peripheral visual cues were filled white squares (0.7 X 0.7”). Gaze coordinates were obtained with Robinson’s search coil method (56). The field coil and associated electronic equipment were made by C-N-C Engineering, Seattle, WA. Details concerning the sampling and storage of the eye and neural signals are given below. The monkeys were rewarded with drops (0.2 ml) of lightly sweetened water via an electronic metering pump (A741, Waltham). Water was not available in the home cage; instead the monkeys worked to satiety (150-250 ml) each day in the laboratory. They were given monkey chow immediately upon return to their home cages and their intake of chow and body weight were closely monitored.

Oculomotor delayed-response task Figure

1 shows schematic drawings

of the oculomotor

delayed-

response task. Figure 1A depicts the timing of presentation of the fixation target, peripheral cues, and reward. Figure 1B illustrates records of horizontal and vertical eye movements and single neuron activity on a typical trial of the oculomotor delayed-response task. Following a 5-s inter-trial interval (ITI), the fixation target appeared at the center of the CRT. The monkey looked at the fixation target and maintained fixation for 0.75 s (the fixation period), whereupon the visual cue was presented for 0.5 s (the cue period) at one of four or eight peripheral locations (13” eccentricA

Oculomotor

delayed-response n-l

task F I

fixation

C

D

R

I I

II

target

peripheral cue

Monkeys were prepared for chronic single-neuron recording using aseptic surgical technique and barbiturate anesthesia (pen-

tobarbital sodium, intravenous to effect). For the purpose of recording eye movements (see below), a search coil was placed under the conjunctiva in one eye using the technique of Judge et al. (38). To secure the implant, stainless steel’bolts with flattened heads were run along slots in the skull with the bolt head under the skull. The bolts, a connector for the search coil, and a stainless steel receptacle for attaching the monkey’s head to the primate chair were bound together with dental acrylic. Because dura exposed during the lengthy training period (usually 3-4 mo) would thicken and impede the passageof electrodes, a second surgery to install the recording cylinder was performed only after the monkey became proficient on the oculomotor delayed-response task. Trephine holes (20 mm diam.) were made over the prefrontal region and a stainless steel recording cylinder placed over the hole. Additional acrylic was applied to bond the recording cylinder to the existing implant. Monkeys were given systemic antibiotics, fruits, and ad lib water and chow for 5-8 days following each surgery.

reward

B An example

of an oculomotor ITI

horizontal gaze coordinate

delayed-response , F ,C, III

trial D

, R, I I I I

vertical gaze coordinate neuron activity 10” 1 set

FIG. 1. A: temporal sequence of events in the oculomotor delayed-response task. B: eye movements and single neuron activity during an oculomotor delayed-response trial. ITI, intertrial interval; F, fixation period (0.75 s); C, cue period (0.5 s); D, delay period (3 s); R, response period (0.5 9.

MEMORY

CODING

IN PREFRONTAL

ity). Stimulus location was randomized over trials so that the monkey could not predict where the cue would appear on any given trial. A crucial feature of the task was that the monkey had to maintain fixation throughout the cue period and also throughout the subsequent delay period. At the end of the delay period, the fixation target was extinguished; this was the “go” signal to make a saccade. If the monkeys made a saccadic eye movement within the next 0.5 s (the response period) to the location where the cue had been presented, they were rewarded with a drop of liquid. A correct response was defined as an eye movement that fell within a window (6” diam.) around the cue location. We usually used one fixed delay interval (3 s) during recording, but longer delays (6 s) or combinations of two (3 and 6 s) or three (1.5, 3, and 6 s) different delays were sometimes used to examine activity across different delay intervals. Many of these neurons were also studied with other tasks (visual probe task, visually guided saccade task, etc.): these results will be presented in a future paper. After analyzing the response properties of neurons, we sometimes employed microstimulation through the recording microelectrode to determine whether the recording site was in the frontal eye fields, based on the criteria of Bruce et al. (7).

Single neuron activity was recorded with Parylene-coated tungsten microelectrodes (2-5 Mit at 1 kHz, Micro Probe) that were advanced through the intact dura with a hydraulic microdrive (MO-95B, Narishige). Single neuron activity was amplified 1O,OOO-fold by a differential amplifier (MDA-4, BAK). This signal passed through a low-pass filter to eliminate artifact from the search coil driver and then to a window discriminator (DIS- 1, BAK) to isolate single neuron activity for the computer. We monitored both the raw activity signal and the window discriminator output simultaneously by an oscilloscope (5 1 10, Tektronix) and an audio monitor. Usually one electrode track was explored each recording session. Electrode tracks were separated on a 1-mm grid. For the microstimulation, trains of pulses were generated by a biphasic pulse generator (BPG- 1, BAK). Each pulse was 0.2 ms in duration, stimulation frequency was 330 Hz, and the train duration was 70 ms. The pulse generator output was connected to a constant-current stimulus isolation unit (BSI- 1, BAK). Current was monitored by the voltage drop across a I-kft resistor in series between the output of the isolation unit and the microelectrode. Current tested ranged up to 200 PA. A

Visually

guided

saccade

B

3.0

CORTEX

333

Data ucquisit ion The on-line computer system, in addition to carrying out the behavioral paradigms, sampled neural and ocular signals and stored these data in relation to task events on magnetic media. Voltages from the phase-sensitive detector representing the horizontal and vertical eye coordinates were digitized every 2 ms (500 Hz). These signals were also electronically differentiated to yield horizontal and vertical eye velocity. The velocity signals were also digitized and were used by the on-line computer program to identify each saccade that the monkey made via the algorithm of van Gisbergen et al. (25). Two types of data storage files were stored. Event bz&Gr files contained the time of every event that the computer had access to, including the time of each discriminated action potential, the time, duration, and amplitude of each saccade, and the time of events such as the appearance and disappearance of visual cues. Individual event buffer files usually contained 50- 100 trials. AnaI~~~u~ bu&r files contained multiple records ( l-2-s epoches) of all of the analogue signals being sampled, together with the discriminated action potentials and a code representing progress through the task paradigm (see Fig. 1B). Two types of computer-generated displays were monitored during the experiments. In one display, horizontal gaze, vertical gaze, and overall eye velocity were displayed as scrolling traces on a digital oscilloscope (1345A, Hewlett Packard). Discriminated neuron activity was scrolled below these traces and tics representing on-line saccade recognition were placed on the velocity trace. Superimposed on the scrolling display were two-dimensional representations of gaze, visual stimuli, and the fixation window. In the other display, rasters and histograms were viewed on the same oscilloscope to observe the overall neural data collection.

Datu anaivsis Using the stored event buffer files, we examined rasters and histograms of neuron activity for each cue location. These rasters and histograms were made with different alignment points including I) the onset of cue, 2) the start of the delay period, 3) the end of the delay period, 4) the initiation of the saccadic eye movement during the response period, and 5) the appearance of the fixation target. Rasters and histograms were examined on the digital oscilloscope and copies were made using a digital plotter (7470A, Hewlett-Packard). Neural activity during the delay period was also analyzed quantitatively for all neurons recorded. The average discharge rate C

s delay

6.0

s delay

20.

20

4

: --I-

*

20

----

** I” - *

20

FIG. 2. The distribution of eye positions following the saccade in the visually guided saccade task (A), in the oculomotor delayed-response task with 3-s delay (B) and with 6-s delay (0 These plots contain both correct and incorrect trials (if any). All cue eccentricities were 13”.

I

FUNAHASHI,

334

BRUCE,

AND

during the delay period was calculated for each trial, and then overall mean discharge rates and standard deviations for each cue location were computed. We tested for significant delay period activity by comparing the mean discharge rate during the delay period for all trials having a given cue direction versus the mean discharge rate during the intertrial interval over all trials, using a two-tailed unpaired Student’s t statistic and an alpha level of 0.05. Differences in delay period activity across different cue locations were evaluated using an analysis of variance (ANOVA). 1350

c

II

I

Ill

D

I

IUI I II I

GOLDMAN-RAKIC

Histological

After 2-8 mo of nearly daily recording sessions the monkeys were killed with an overdose of pentobarbital sodium and perfused with saline followed by buffered Formalin. The brains were photographed. Frozen coronal sections were taken and stained with thionin. Individual recording sites that had been marked with electrolytic lesions (20 PA, lo- 15 s, tip negative) were identified. How450

C I II Illlll II III II I Ill I II I Ill1 I IF I Ill I II II II II III u I I Ill I I I

R

111111

analysis

IIII

I

I

Ill III I I IIIIllI lllll II I ll’ II 1 I I ‘111 Ill1I I I I II IIIY II I Ill II ll Ill II IIII II I I I I Ill1 III I IIII I III llIIIIII I I I Ill1 I I II lull II II I I II II Ill I II 111 I II I Ill II I 11111111 Ill1 I Ill Ill II I Ill I II IlllId I I l ll II’ll I I I I Ill1 II IIIII I I ll lull I

IIIa II II lllll I I III I I

D I

II II

I II Ill IIll I

III

I

Ill1 I

I ~1111 II

I II

I

II

II

I

I11 Ill

3 llllll Ill I I II11 II I Ill Ill I Ill1 I; ; I iiiiimu III

I Ill I I ‘I ‘I I

II

I II III Ill1 I I I

I

I’1 II III1

1’1 II

5211-O-3

1800 ICI

D

I R

lllll I I I Ill II II!1 I I ‘I m I 11~11~11111II I I A

I “:ll1,l I’ ’ I III1111111I Ill I’

II Ill

II

Ill

I

I

I

I!’ A II I Yl

I II

lllll I

I II II “~‘1’:

1111

II

II

I I

III

I

I Ill I Illllll

I

I I

III I

I ’ I

I I I

I I

II allllII I I III I llllllll I I 1ly yml

II I

I Ill

I

Ill I Illll

II

11111111 II I 1 Ill II I I

Ill I

cl O

225

I I I II II Ill I I I Ill1 I I I ‘lmllll I Ill m II II III I I III I I II I IIll1lIII I I II1 Ill I Ill 111 I I I IIII

II II i III1 II II 111 II I I II Ill II I I II 11111111 I Ill I I II I II II I Ill11 I

521

1 .O-5

II llllll II

IIII I I I Ill I II II II I II I Ill II III I

Ii11 I I III‘I I’ I I llllI I II

D R IC I I I 11111~11~111 II II 1111111111111 I 11111 II1 lllll Illlln~mmImII Ill I I Il I I II I IIII I I ItI I I IHI II IIIuIIIIII I I I11II I III I ii1111 IIIII n~~~wml~iilllilmiimi~~~~~~i i iii Fl

II ““Ill”::’ Ill1

I

~~IrzKl~:~l~ I

I I Ill

IIamImlIll

I I III11 111 Ii I

I

I I /l

315O

c

I’ll I :‘I1 Ill 1111111111 I11 11111 Ill II I I I I YI I II I III I

a I I

1111 II

III II

I III1111 I l!l III I ilmil ii I II I

.O-6

I II

II II

I

III1 I I I I I II II I II I II I III I

IIII 1 I

II

Ill

I II

‘I I IIll l l II I ll I I

I I I II I II I I Ill I I I lllllI I II I

IIIII I I IlIlY I

I II lllllllll Ill I I I I

II II I

1 .0-O

II

5211

I I

II I

3150

270"

R

D

II11

I

270' 521

c

Ill I I II Ill I

El

5211.0-4

225O

I I I I

II mi Ill1 II I I IIII IIII

111111Ill1 II III III

IIA

IR

D

I

D

R

II II I Ill I I I II II IllI\ Ill/ II II !

II

I

II I III I 1 I

Illlll I I I I Ill I II I II II ll II I ‘;I1 I

5211.0-7

FIG. 3. Directional delay period activity of a principal sulcus neuron during the oculomotor delayed-response task. This neuron (52 1 1, left hemisphere) had strongly directional delay period activity (fi’ = 48.35; df = 7, 68; P < 0.00 l), responding only when the cue had been presented at the bottom (270”) location. It was suppressed during the delay when the cue was presented in the upper visual field, and in all 3 cases delay period activity was significantly below the IT1 rate (45”, l = 2.350, df = 84, P < 0.025; 90”, t = 3.45 1, df = 85, P < 0.001; 135”, t = 2.607, df = 84, I-’ < 0.025). Visual cues were randomly presented at 1 of the 8 locations indicated in the center diagram. All cue eccentricities were 13” and all delay periods were 3 s.

I III I Ill II IIIII I I I Ill1 I I Ill II II I11 II1 WI II llnll

IIII

Ill

I

I

I

Ill II I

II

MEMORY

CODING

IN

PREFRONTAL

335

CORTEX

ever, the long duration of recording and large number of electrode penetrations precluded identification of most penetrations, and their locations in the brain were estimated from their microdrive coordinates.

the distribution of saccade end points of one monkey in a visually guided saccade task (Fig. 24, and in the 3-s (Fig. 2B) and 6-s delay conditions (Fig. 2C) of the oculomotor delayed-response task. Although the distribution of saccade end points is wider in the delay task than in the visually RESULTS guided saccade task, nearly all saccades end near the appropriate location, even with 6-s delays. Task yerfimnance . The mean latencies for saccadic eye movements in the responseperiod were 2 16.4 t 33.0 (SD) ms in the visually All monkeys performed the oculomotor delayed-response task proficiently and the percentage correct was guided saccadetask, 298.3 t 34.4 ms in the 3-s delay task, usually 90% or better for all cue locations. Figure 2 shows and 293.0 t 37.4 ms in the 6-s delay task. The difference in 1350

I

c

I

1800

I I

D

I I I 111111I II L I Ill llllllll1lI#l 11111111 II II lllllll Ianllll II II I mmllIIIIIImul Illll I WI I I HII Ill1 llll1lNmII Ml liii I I I II I III III ilii~i~l~m~tw l lll~~~~~n~il I I ii ml~rn~~irn~~iin Ill Ill I II II II nl I llHlI HllMIImIllml lllll IIN Ill Ii I iiim II~rnmiw I III I 111 I I I II IH II rrllll I II Mlllllll III Ill1 11111 111111mIII ! I I

45

C

clllllllll II I ill Ill1 II I I II II Ill II Ill Ill1 WI I I IIN IIIIIIII N~IIIII I HI IIIHIII~ I ll I Il 1111111 Ill mu 111 II I IIIIIIIHII Irn I Ill IlmlI II11 Dal I llHIIIIII

i II IIDIII III III 1 l rnwiiiu~ m ii1111 aiiii I II I II II I I I I I I lllll I II I lllltl ll I II II Ill 1111111 I 1111111111111111 II llllllI II III ‘r:~lllll II I Ill II 111111111111 11111111 I 111111111111 I I II II I II1 lmlmlBNlllrl Ill I Ill11 lll1ml I I111111111111111 11111111111##11) 111111 I ll III III I lllll II I I I I I II 11111111111111 HI I I I I I I UII I 4

O

Ill IIII I I I I I I lllll I II III II I III Ill II II II1 I II II II I III II I II III II I I IIIII I Ill1 I I I Ill

00

ICI

II

C D IR II I II II I I I II ‘Ill II I II1 I I lIlti lllllll I IIII I I I I II I lull 11111111111 II II III I ! I I 1111111 IIIII I ml I I ll IIll IllIll I1 11111111111 II I I IIII III I I III Ill II I I I111111 Ill11 II lI III II I

I111 II II 1111111lll llllll ;ly’Il/; 111111 III 1111111111111111I II i ii 1111i iimll iiiiiiiiwiinmi II HII 1111II II II Ill1 II IliIuMln 11111111

I III I III I II II 1111111111I111111 I II II Ill1 I Ill1

II11 IN lllll I IIIBII III111 II11 I I II 1111 Ill1 II

2250

C II Ill1 I II I II I I ll I IIIHIIIII I I II IIIII Illlllll I u I llllI I Ill I I I II I III IlllllllI I I Ill 1111111111111 Ill II I llllll 11111 Ill1 Ill

2700 IIII I II11 I II llllll II I DIlamlmll II I 11111II111 lllmmI I IllllIlII Ill1 Ill Ill 111111 Ill II

C R I I II I 11; Ill1 II II I III II I # I\1 I lllll 11111111111 Ill II Ill ill III I II I II I I II I II 111II II Ill1 I I I Ill Ill lllll I11 II I IllllIlllII Ill1 Ml Ill1 I I Ill II II II I I I 11111111 I I II I I 11111I Ill 11111 111 I I I IIH 11111HI III Ill1 IIIIII II Ill1 Illllllll Pill Ill I

315O Ill1 I I I II II I II I II 11111 I II I II II II III I

I lllll IIIII Ill1 II II Ill ml ll Ill III I ! ill1 I III II Ill WIIIII IIII I I I III I II I I llllllI I I11111 Ill II lull I 111111 I

R

D

‘II 1111 llllll I11 IIIIIllIIII I I Ill1 II I II I ID II I II II 11111 I III I II III1 II II I I 111 llllll II I IUI I INIIIIllII Ill I I 1111 I 111111I II I I I III I II I I III I Ill I Ill1 I Ill II I II I II II I Ill I II I llllllll 1:’ I llllllll II I I I I II II Ill II Ill I IIIl I II II III IIII I llllllI II II I IIII II 1111111111 1111 I II Ill I I llll11ll

I

50s

5111.1-6

1 s

FIG. 4. Directional delay period activity of a principal sulcus neuron during the oculomotor delayed-response task. This neuron (5 11 1, right hemisphere) exhibited directional delay period activity (F = 25.23; df = 7, 56; P < 0.00 l), responding only when the cue had been presented at locations in the upper quadrant of the contralateral visual field (90, 135, and 1800). The strongest delay period activity occurred at the 135” location. Visual cues were randomly presented at 1 of the 8 locations indicated in the center diagram. All cue eccentricities were 13” and all delay periods were 3 s.

IS

336

FUNAHASHI,

BRUCE,

AND

&w-al

mean latency between the visually guided saccade task and each of the delay tasks was statistically significant (3-s delay: t = 25.93, df = 454, P < 0.00 1; 6-s delay: t = 24.08, df = 493, P < O.OOl), whereas that between the 3-s and the 6-s delay tasks was not significant (t = 1.60, df = 477, P > 0.05). The mean duration of saccadic eye movements was 46.2 t 8.5 (SD) ms in the visually guided saccade task, 44.9 t 7.6 ms in the 3-s delay task, and 46.2 t 8.2 ms in the 6-s delay task. There were no significant differences in mean saccade duration among the three tasks (ANOVA, F = 1.982; df = 2, 706; P > 0.1). 900

IC I

lllnl III I I I I 111 Ill II III I Ill I I II I 1111 I I I llllI I I Ill II I I II I nim III II n I I i

GOLDMAN-RAKIC

data base

We recorded from 3 19 neurons in the prefrontal cortex of three rhesus monkeys while they performed the oculomotor delayed-response task. As discussed in detail below, 288 of these neurons were from cortex within and surrounding the caudal half of the principal sulcus. We term these PS (principal sulcus) neurons and most of this report concerns their activity. The other 3 1 neurons were located within and near the anterior bank of the arcuate sulcus and were classified as frontal eye field (FEF) neurons. At some sites this FEF classification was confirmed by low microstimulation thresholds (~50 pA) for the elicitation of sacD

II I11 Ill lllll m Ill I Ill I I I I I l!~llIII I Ill1 I Ill1 II I II Ill1 Illll I II 111 I I I III I I pll 1::’ IIll” II I I II II I lIlanIl II iiiui III I l 1

180'

270"

C

FIG. 5. Directional delay period activity of a principal sulcus neuron. This neuron (5050, right hemisphere) showed tonic inhibitory delay period activity only when the cue was presented at 90” (F = 4.870; df = 3, 39; P < 0.025). Visual cues were randomly presented at 1 of the 4 locations indicated in the center diagram. All cue eccentricities were 13” and all delay periods were 3 s.

MEMORY

CODING

IN PREFRONTAL

cadic eye movements. Data concerning these 3 1 FEF neurons is separately presented at the end of the RESULTS. Of the 288 PS neurons thoroughly analyzed, most (170 neurons, 59% of total sample) had task-related activity in that their average discharge rate during at least one phaseof the delayed-response task differed significantly (P < 0.05, t test) from their IT1 (InterTrial Interval) rate. Of these 170 neurons, 87 neurons (30% of total sample, 5 1% of total task-related neurons) had significant de/a-v period activity. For 33 (38%) of these neurons the delay period activity was the only significant task-related activity, but most (54 neurons, 62%) also had significant activity in the task segments immediately preceding or following the delay period

337

CORTEX

(delay and cue periods: 11 neurons, 13%; delay and response periods: 32 neurons, 36%; delay, cue, and response periods: 11 neurons, 13%). The remaining 83 task-related neurons exhibited significant activity only in the cue period (n = 12), only in the response period (n = 55) in both cue and response periods (n = lo), or at reward presentation (n = 6).

Directional

specifrcitv . period activity . ” of’delav .

Most PS neurons with delay period activity showed a significant increase or decrease in this activity relative to their base-line discharge rate only when the cue had been 45O

1111 I Ill Ill1 I I 111 II II II I I I II II I I I II I Ill Ill Ill Ill1 I II III I Ill IFII11 I Ill1 I Ill1 II llllll Ill Ill I WI 1I i ii l l l ii llla II I liliil l i I IIII IIII la1 I I Ill ia l ii ii1111III limaii iiiai II II

1t i

I III1 I IllI I I I Ill I III I ia I I I IIIll1 I I

I I a I II II I I I I UII III II I I I II ~II~III~III I I I I II I I II llllll III lllllll II llllll Ill1 ll I I II Ill1 I I I 11III II II I Ill1II II I I I I I II Ill11 II11 I11 III II II Ill I II II I II II I I I I I ;llul “;I I I Ill I I I ll II II IIII I I ll 1111 I III II Ill I I II 1111 III Ill I I11111111111111111 II III I I

I II

I II

I

II

II

II

II

I

II

I

I II

II

i

D

C

R

iii nu ii i aili i ii liiilliiil1llll illii lllill II III I I I II I I IIIII IIIIIIIIII i II a I Ill1 I I I II II I Ill II III llllI IIU II llllllrlll IIII III IIII IIIIIII~ lliilliiama I I iiiil II I UII II II I II IIIU I I I aim I ;l: II 1;: a Ill I Ill II I I

I

I11

Ill

llllll I I I Ill I I lllll

II I

I II Ill I

111 IIII I11 III ;;i

Ill1 I II I Ill II II I II Ill I II I I II II I I II

III I II II II III I I II I I I

III i a imii ai

II

I III1

I

i

Ill IIIII

5011.0-2

1800

ICI

II I I I I I I Ill 11111 I I Illlll I I I II I III1I II I Ill I II lllll I I I II II I lllll 111 II l!~.l

2250 I II

C

lllllll

I Ill1 II II WI II II ii i iii I I I I iiii

D

IR

I

I II I II 1111111 III IIII II II I Illll I II II I IIIIIIIII iiuiaiiiii II III I I Ill1 I 1lllI I I II II II I I I I II I Ill II I Ill II I I I Ill I I I I I I I II II I I II 111:~1,::1~:1 l:I l:)lll

D

,;:1

gooLo

3150

R

I 11111111111 Ill1 I Ill1 Ill1 I 111 III I Ill1 I I Ill Ill I I II I I II III II II I iiii ii iial I I I al iii i i i l I IIIIIDI II laiiullla II 111II I iii ii l 111

C

R

D

I I II II I I I III iii111 I I ii11 ii u l ;l~l~ I : I IIII; l;l I

I

II II II I u I u II I I III I Ill Ill1 - I I I II III II1111 I II I I I ii11111 I II I IIIIIIII IIIII IIIII ii iimll I IIII I I llila I iii IIIII I allliilllw I II III IUI IIII III I I wa I imi iy I I II Ill lllllll Ill Ill I It

501

1 .O-6

FIG. 6. Omnidirectional delay period activity of a principal sulcus neuron. This neuron (50 1 1, right hemisphere) showed tonic excitatory delay period activity independent of cue location (F = 0.63; df = 7,46; P > 0.1). Visual cues were randomly presented at 1 of the 8 locations indicated in the center diagram. All cue eccentricities were 13” and all delay periods were 3 s.

I

l.

.....

a....

5211.0

5041.0

C

D

. ....

l’

51111

E

. ....

. . ..a

.**

l’

i 708/s

.*.

5066.0

F

l.

l .. . .. ..

... .

. 5011.0

i 258/s

5008.0

i 258/s

FIG. 7. Directional tuning of 6 principal sulcus neurons with delay period activity. In each radial plot, the neuron’s average discharge rate during the delay period for each cue location is depicted by the radial eccentricity of the plot in that direction, the standard deviation of this rate is depicted by the radial line, and the neuron’s average IT1 discharge rate is the radius of the circle. Asterisks (*) indicate cue directions with statistically significant (t test, P < 0.05) differences between delay period activity and the IT1 rate. A and B: 2 principal sulcus neurons (A: 504 1, right hemisphere; B: 52 11, left hemisphere) with very specific directional delay period activity (A, F = 10.12; df = 7,54; P < 0.00 1; B, F = 48.35; df = 7,68; P < 0.00 1). They have a significant elevation of delay period activity for only 1 direction. Notice that there is significant inhibition of delay period activity for cue directions roughly opposite the excitatory directions. C and D: 2 principal sulcus neurons (C: 5 1 1 1, right hemisphere; D: 5066, right hemisphere) with directional, but broadly tuned, delay period activity (C, F = 25.23; df = 7,56; P < 0.00 1; D, F = 3.80; df = 7,7 1; P < 0.005). Notice that for Cthe responses are excitatory relative to the IT1 rate whereas for D the only significant responses are inhibitory. E and F: 2 principal sulcus neurons (E: 5011, right hemisphere; F: 5008, right hemisphere) with omnidirectional delay period activity (E, F = 0.63; df = 7, 46; P > 0.1; F, F = 1.26; df = 7, 65: P > 0.1). One of these is excitatory (E) and the other inhibitory (F).

338

MEMORY

CODING

IN

presented at one or a few among the 4 or 8 locations tested. Of the 87 PS neurons with delay period activity, 69 (79%) were directional, i.e., a statistically significant difference (P < 0.05) in delay period discharge rates across different cue locations. Figure 3 shows directional activity in a PS neuron tested with eight cue locations. Strong tonic excitatory activation appeared throughout the delay period only when the cue was presented at the downward (270”) location. This response began - 100 ms after the cue presentation and ceased 130 ms after the initiation of the saccadic eye movement. Moreover, activity during the delay period was suppressed relative to the IT1 discharge rate for the three cue locations in the upper (opposite) visual field (45”, t = 2.350, df = 84, P < 0.025; 90”, t = 3.451, df = 85, P < 0.00 1; 135O, t = 2.607, df = 84, P < 0.025). This phenomenon of an excitation of delay period activity for one set of cue locations and an inhibition of delay period activity for other directions is discussed further below. Figure 4 shows another PS neuron with directional activity that exhibited broader tuning than the neuron shown in Fig. 3. The delay period activity was significantly elevated for the three locations that span the upper quadrant of the contralateral visual field (90, 135, and 180”) with the highest rate on 135’ trials. The activity ceased abruptly 170 ms after the initiation of the saccadic eye movement. For the three cues spanning the opposite quadrant (lower ipsilateral, 270, 3 15, and 0”) this neuron’s activity was depressed below its ITI rate (see also Fig. 7C), although this depression was not statistically significant for any individual direction. Neurons with only inhibitory activity during the delay period could also be directional. The neuron shown in Fig. 5 was suppressed during the delay period only when the cue had been presented at the upper location (90”, t = 6.036, df = 53, P < 0.001); there was no significant change in activity during the delay period for the three other cue locations tested. The remaining 18 neurons (2 1%) with significant delay period activity were classified as omnidirc~ctional because they responded similarly during the delay period regardless of cue locations and the ANOVA criterion for directional selectivity was not met. Figure 6 shows an example of a PS neuron with omnidirectional delay period activity. The tonic excitatory delay period responses for all locations was significantly above the IT1 rate by individual t test and the slight differences between locations were not statistically significant. To more concisely show directional specificity of directional delay period activity, we constructed polar plots that 1. Response tIvpe of direct ion al und omnidirectional delay period activit wv of. principal sz~lcus neurons

TABLE

Response

type

Directional

Omnidirectional

Total

Excitation Inhibition

46 23

4 14

50 37

Total

69

18

87

PREFRONTAL

CORTEX Distribution for excitatory

of best delay

cue directions activity (n=46)

45" (1)

180" (11)

315" (4

225" (2) 270" (8)

Dis tribution for inhibitory

of best delay

cue directions acti vity (n=23)

90” (5)

180" 0"

(9)

(5) 315"

225" (1)

(0) 270" (3)

FIG. 8. Distribution of the preferred directions for delay period activity of principal sulcus neurons tested with the oculomotor delayed-response task. Length of radial lines is proportional to the number of cells for which that cue direction gave the best delay period response: the cell counts are also given beneath the directions. Preferred directions were normalized by a right-hemisphere convention: the preferred direction of each left hemisphere neuron was mirror reversed across the vertical meridian. Top: this plot summarizes the preferred directions for the 46 neurons having excitatory directional delay period activity. Bottom: this plot summarizes the preferred directions for the 23 neurons having inhibitory directional delay period activity. Note that 15 of the 46 neurons in the top plot also had significant inhibitory delay period activity, usually for cue directions roughly opposite their best excitatory direction. The best inhibitory direction for these neurons are not plotted in the hottom part, and hence the neuron populations comprising the top and bottom plots are mutually exclusive. The overall preference in both plots for directions along the horizontal and vertical meridians, as opposed to oblique directions, is an artifact because many neurons were only tested for four cue directions (left, right, up, and down) and not for oblique directions.

depict the delay period discharge rate of a neuron as a function of polar direction by plotting discharge rate as a function of polar eccentricity. The base-line (ITI) rate of the neuron is shown on the plots by the radius of a circle, the standard deviations of the delay period discharge for each direction are shown by radial bars, and statistically significant departures from the IT1 rate are indicated by asterisks. Figure 7, B, C, and E, show polar plots for the narrowly tuned, broadly tuned, and omnidirectional

FUNAHASHI,

BRUCE,

AND

GOLDMAN-RAKIC

FIG. 9. Directional tuning of delay period activity for 2 principal sulcus neurons. Plots show discharge rate during the delay period for the 8 different cue directions, with a Gaussian function fit to the data. A (521 1, left hemisphere) shows narrowly tuned directional activity ( Td = 19.7O, preferred cue direction = 270”) and B (5077, right hemisphere) shows broadly tuned directional activity (T(, = 48.4”, preferred cue direction = 135”).

20

10

1

I

180"

I

270" cue

[5211.0]

0"

I

0

0”

90"

locat Ion

Excitatory e and inhihitorv * dduv- puiod

uctivit w v

As evident from the polar plots in Fig. 7, we observed instances of both elevation and depression of delay period activity relative to a neuron’s base-line rate. To further B

A D

180"

270"

po77.01

neurons whose histograms were illustrated in the previous three figures (Fig. 3, Fig. 4, and Fig. 6, respectively), whereas Fig. 7, A, D, and c(‘, show three additional examples: another PS neuron with narrowly tuned directional delay period activity is shown in Fig. 7A. For this PS neuron, the preferred cue direction was leftward (1 SOO), whereas for most other cue directions, the discharge rate during the delay period was suppressed below base line. A neuron with directional delay period activity that was suppressed below base line when the to-be-remembered cues were in the upper-left quadrant of the visual field is shown in Fig. 70. Finally, Fig. 7F shows a neuron with omnidirectional inhibition of delay period activity: comparable inhibition occurred for all eight cue locations.

C

90”

understand the relation between the “sign” of this activity and directional specificity, we classified each neuron as having either excitatory or inhibitory delay period activity by the statistical criteria previously described. A neuron was classified as excitatory (or inhibitory) if delay period rate for any cue direction was significantly elevated (or on/y reduced) relative to the neuron’s ITI. Table 1 shows the incidence of neurons for the different combinations of delay period response sign and directionality. As is evident from examining the Table, neurons having excitatory delay period activity were more likely to be directional (92%) than neurons with solely inhibitory responses (62%). Furthermore, we observed that directional tuning for inhibitory delay period activity was usually broader than that for excitatory directional activity. The phenomenon of a single neuron having both an elevation of delay period activity for one set of directions and depression of delay period activity for another set was frequently encountered. Usually the best cue directions for saccade

R

2os/s

0.5

C

FIG. 10. The time course of excitatory and inhibitory delay period activity. These histograms sum neural activity at the preferred cue direction for all 46 principal sulcus neurons with excitatory directional delay period activity (A, B) and all 23 principal sulcus neurons with inhibitory directional delay period activity (C’, D). A and C’ were aligned at the cue presentation; B and D were aligned at the initiation of the saccadic eye movements. All delay periods were 3 s.

s

saccade D

R

2osi’s

MEMORY

CODING

IN

eliciting inhibitory responses were opposite to the best directions for eliciting excitatory responses. For 15 neurons the criteria of having a statistically significant elevation of delay period activity for at least one direction and a statistically significant inhibition of activity for at least one other direction was met. However, several other neurons, such as the ones illustrated in Figs. 4 and 7C, had delay period discharge rates below base line for one or more directions that did not meet our statistical criteria yet seemed genuine because the inhibition was present for two or more adjacent directions opposite to the excitatory direction.

Distribution

of prefh-red direct ions .

To determine the distribution of directional preferences of delay period activity, we used the largest t statistic to classify the preferred direction of each neuron as the cue location where it showed the most prominent excitation or inhibition. For analytical and graphic purposes, the preferred directions of left-hemisphere neurons were transformed into mirror-image directions, as if all neurons were recorded from the right hemisphere. Among the 69 PS neurons with directional delay period activity, 30 were studied in the 4-cue condition and 39 were studied in the 8-cue condition. As shown in Fig. 8, the majority of neurons with excitatory directional delay period activity had preferred directions on the side contralateral to the hemisphere recorded from. This contralateral bias was statistically significant by a chi-square test (x2 = 8.758, df = 1, P < 0.0 1). There was no statistical significance in the directional preference of inhibitory directional delay period activity (X2 = 1.667, df = 1, P > 0. l), although most inhibitory directions were also contralateral.

A

PREFRONTAL

Tuning ofL directional

where j‘(d) is discharge frequency, d is cue direction, and the remaining terms are constants: B is the discharge rate during the ITI, D is the best direction, R indexes delay period response strength, and Td indexes tuning with respect to cue direction. The function was implemented by fixing B at the average background discharge rate, and then iteratively converging to the best least-squares solution for R, D, and T’. This Gaussian approach has previously been used to describe visual receptive fields and presaccadic movement fields in the FEF (6). We applied this Gaussian formula to all 69 PS neurons that exhibited directional delay period activity. However, we omitted 42 PS neurons because 30 PS neurons were only studied with 4 cue directions and the Gaussian fits to 8 data points for 12 PS cells were poor. The remaining 27 PS neurons had good fits based on 8 cue locations. Figure 9 shows the tuning curves obtained for two of these PS neurons, one with narrowly tuned excitatory delay period activity (Td = 19.7’, Fig. 9A) and another with broadly tuned excitatory delay period activity ( Td = 48.4”, Fig. 9B). Excitatory delay period activity tended to have narrower tuning than inhibitory delay period activity. The mean tuning index (Td) was 26.8 t 14.2” (SD) (n = 22) for neurons with significant excitatory delay period activity versus 48.2 t 43.5O (n = 5) for neurons with only inhibitory delay period activity. These means are not significantly different (t = 1.985; df = 25; 0.05 < P < 0.1) proba-

D

IR

270 O Il I l’

I ‘7 I Il II I I

delay period activit *v

To determine the directional specificity of delay period activity, tuning curves were estimated using the Gaussian function .f(d) = B + R * e-l/2”dy~)/~d)2

1800 I I I III Ilulalm l wll I I IntIll lmmBllmlI IIIll l I 111 Nrn I II I II I Il1 II 1 1111 UBII I I HI l II I III II II IIIIW lml I I l Il1 II III II I I IlNIHlYIII 111 I l l1lI I I I II Il I InIYIIIII I Il I I I Il I I Il 111111 II

341

CORTEX

D

II

I II 1111

I I I

I I

O0 I I III

c Il

I”

II’ III Il

111’ I11111

Q

R

I MI I

I II 1111111

I II II Il

I I

II “II 01

I

5018.0-2

6

R

II I

II11111 I I1111111 II WIIlIIII IU 1111111111 II I I II I I I IlIN II I I I I nllHlnl II II II l mlIl!JnwIII II I l’nl Il lulI I 111111 Il I II I Il IBml ml11 II Il 1111111111111 II 1111 I

FIG. 1 1. Directional delay period activity of a principal sulcus neuron (50 18, right hemisphere) for 2 different delay durations (3-s delay in A and 6-s delay in B), examined sequentially (3 s first, then 6 s). For each set of histograms the visual cues were presented randomly at 1 of 4 locations indicated (0” = right, 90” = up, 180” = left, and 270” = down). All cue eccentricities were 13 O.

III 111

l

II I II I I I II I I II I II I I IIIM III IIll I I Il1 Il1 HI I I IIIU I III llI I I II II I Il1 III1111 II BIII ID1 I I Ilrll I I I

342

FUNAHASHI,

BRUCE,

bly because of the small number of inhibitory neurons that could be fit and the large SD of inhibitory tunings.

Time course of dela wk)period act ivit w c’ To further characterize the overall activity of PS neurons during the oculomotor delayed-response task, we made histograms that summed activity across our sample of neurons, selecting for each neuron the trials with cues in that neuron’s preferred direction. Figure 10 shows these composite histograms. Neurons with excitatory (n = 46) and inhibitory (n = 23) directional delay period activity were summed separately to prevent these activities from canceling each other. The left pair of histograms (Fig. 10, A and C) depict excitatory and inhibitory composite activities during the 3-s delay period, as well as during the fixation, cue, and response periods, and part of the ITI. For the excitatory composite histogram the overall delay period activity was 13.9 spikes/s, in comparison to 7.6 s/s during the first 1.5 s of the histogram (the IT1 and fixation interval). For the inhibitory composite histogram the overall delay period discharge rate was 4.5 s/s, in comparison to 8.1 s/s during the ITI and fixation interval. Thus PS neurons with typical excitatory directional activity discharge at approximately twice their base-line rate during the delay period, whereas PS neurons with typical inhibitory directional activity discharge at approximately onehalf their base-line rate. The composite histograms also indicate the time course of PS neuron activity during the delay period. Although individual neurons have a variety of temporal changes during the delay period, these composite histograms indicate that overall PS neuron activity is well-maintained during a 3-s delay. In fact, there is an increase in discharge rate over the 3-s delay period in the excitatory composite histogram ( 12.4, 14.9, and 14.5 s/s in the first, second, and third second, respectively) and a decrease in the inhibitory composite histogram (4.8, 4.7, and 3.9 s/s). A phasic response is evident during the 0.5-s cue period in the excitatory composite histogram (Fig. 1OA). The earliest indication of this response begins -70 ms following cue onset (all latencies were estimated using a cumulative histogram of the same data, which is not shown); but individual neurons had a wide variety of latencies, and many did not respond at all during the cue period. It should be noted that many PS neurons having strong visual responses, but lacking delay period activity, were not included in these composite histograms. Finally, the composite histograms depict a phasic increase in activity during the response period for both the excitatory and inhibitory activity. The latency of this activity is 180 ms following the disappearance of the fixation target for the excitatory activity and 300 ms for the inhibitory activity. Since the typical saccade latency during the oculomotor delayed-response task is -290 ms, it would appear that the excitatory burst leads the saccade in the excitatory composite histogram but follows the saccade in the inhibitory composite histogram. Histograms aligned at the saccade initiation (Fig. 10, B and n) support this interpretation: the additional burst of activity in the excitatory composite begins -30 ms prior to the saccade initiation

AND

GOLDMAN-RAKIC

and ends -40 ms after the saccade initiation (Fig. 1OB). In contrast, the increased discharge in the inhibitory composite begins -40 ms following the saccade initiation (which is approximately at the saccade termination) and continues for -0.5 s (Fig. lo@.

Eighteen PS neurons that exhibited directional delay period activity were tested either with two (3 and 6 s) or three (1.5, 3, and 6 s) different delay durations. The monkeys had little difficulty with any of these delay times and maintained a high level (>80%) of correct performance at all delays tested. Varying the delay length over this range of values had little effect on delay period activity. In particular, the directional preference was unchanged and the duration of excitation or inhibition expanded with longer delays. For example, Fig. 11A shows a neuron with tonic excitation during a 3-s delay period only when the cue was presented at the 180’ location. Figure 11B shows the same neuron’s A __

I CDI I

Ill

11111 ll I II I 111111~1 1111111111 I I 111111111111 llllll I II II IIIII UII I11111I I I

I R I I II lllll II II I I llrnl1 I I II I I I II IllllIll Ill II I II I II IIII I I I I II1 I III1 I Illll II 111111111 Ill 1111 Ill1 llill I II I I i I I

Ill11

ll1Dll II11 II I n ll I I ll lllll I Ill II1111111 Ill1 llllll I Ill1 I I Ill1 II I II I I I I1111111 II III II I I I I I 111111111I 11111111III Ill1 I II11 II Ill II11 lllllI I Ill Ill I II I1111111 II11111111 llllll ll I

I1111 MM lllll Ill Ill

5070.3-2

B

5070.3-6

C

I CI

D

I R

y II11 I II I Ill I I Ill I II Al

I

FIG. 12. Directional delay period activity of a principal sulcus neuron (5070, right hemisphere) for three different delay durations (1 S-s delay in A, 3-s delay in B, and 6-s delay in CT). The data are from an experiment with 12 different types of trials: 1.5, 3-, and 6-s delay durations crossed with 4 visual cue locations (0” = right, 90” = up, 180” = left, and 270” = down). All cue eccentricities were 13”. Only the 3 histograms from the preferred direction (180” location) are shown.

MEMORY

Neuron

CODING

IN

PREFRONTAL

343

CORTEX

5018

correct

trials

3

error

u

!!!&

trial

’

e-k-!

Neuron

!

! ! It-u!!

!! !

5111

correct

trials C 1

*j+*

2

UJ!f

D

R fgj

H!

\

u

FIG. 13. Discharges of 2 principal sulcus neurons with directional delay period activity during correct performance of the oculomotor delayed-response task and during trials with errors. For each trial depicted, discharge traces are shown on the left and cue locations and saccade directions are shown on the right. Top: this neuron’s (5018, right hemisphere) strongest delay period activity was associated with leftward cue presentations. On the error trial, its discharge ceased about midway through the delay period. Bottom: this neuron’s (5 111, right hemisphere) strongest delay period activity was associated with cue presentations in the upper-left visual field. On the error trial, its discharge in the delay period was absent.

3-###+#* m error

+

trial

t

:

! ! !!+S+t!

II IIaI

1111 Illl

I##+#+# I 7-T

activity during the delayed-response task with a 6-s delay. The tonic excitation still appeared only when the cue was presented at 180’ and was maintained during the entire 6-s delay period. With both delays, the tonic excitation ceased at the initiation of the response. Eight other PS neurons with tonic delay period activity were similarly tested with different delay durations; all showed similar results, that is, the response e panded to fill the entire delay. Figure 12 sx ows another PS neuron tested across different delay durations. In this case, four cue locations (0, 90, 180, and 270”) at three different delay durations (1.5, 3, and 6 s) were randomly intermixed. The rasters and histograms in Fig. 12 show neural activity during three different delay durations for cues presented at 180”, the preferred direction for this neuron. Notice that the activity in the first 1.5 s of the delay is very similar in all three conditions, that is little or no change in activity relative to the preceding fixation interval. Furthermore, the activity in the next 1.5 s of the delay period is the same in the 3-s and 6-s delay conditions, both showing a gradual rise in activity. This result was obtained in six other PS neurons that increased discharge rate during the delay. Likewise, two PS neurons with gradually decreasing discharge rates during the delay showed analogous patterning of activity when the delay duration was changed.

Directional

delay period activity in error trials

Analysis of activity in error trials often aids the interpretation of delay period activity (5, 53, 54, 58,64-66). Therefore, we examined the activity when the monkeys made

errors on trials with the preferred cue direction for the neuron being studied. For this analysis, failure to saccade and misdirected saccades(i.e., ending outside the window) were counted as errors; however, nearly all errors were misdirected saccades.Premature saccadesmade during the delay period did not count as errors. Of 69 PS neurons A

Correct

trials

4Qsls

1s

B

Error

trials C

D

R

FIG. 14. Comparison of delay period activity for correct trials with activity for error trials. The first correct trial and the first error trial were taken from each of the 9 principal sulcus neurons that had excitatory directional delay period activity and for which the monkey made at least 1 error on a trial in the neuron’s preferred direction. The tophistogram sums the 9 correct trials and the bottom histogram sums the 9 error trials.

344

FUNAHASHI,

BRUCE,

AND

exhibited directional delay period activity, only 13 PS neurons (9 excitatory and 4 inhibitory) had one or more such error trials. Nevertheless, these data indicate that the responses of PS neurons during the delay period were significantly weaker on trials where the monkey made an error during the response period. Figure 13 shows examples of neural activity in both correct and error trials for two neurons with directional delay period activity. The neuron shown at the top of the figure had its strongest delay period activity with leftward (1 80°) cue presentations. For trials on which the monkey’s eventual saccade was to this location the neuron responded tonically throughout the delay period. However, on the one trial in which an error was made (saccade to upper right) the neuron’s discharge ceased about midway through the delay period. The neuron at the bottom of Fig. 13 showed a similar pattern except that its delay period activity was absent altogether on the one trial in which an error was made. To examine the error data as a whole we summed the first error trial (usually the only one) across the nine excitatory neurons with an error trial in the neuron’s preferred direction. This histogram was compared to a histogram composed of the first correct trial (from the same event buffer) of these nine neurons at their preferred direction. Figure 14 shows these two histograms. The delay period activity on the correct trials was significantly greater than activity on the error trials by a paired t test (t = 2.76, df = 8, P < 0.025).

A

GOLDMAN-RAKIC

Directional and omnidirectional the frontal eye .fields

delay period activity e in

Of the 3 19 examined neurons, 3 1 were recorded from the frontal eye fields (FEF), where low- (~50 PA) and intermediate-threshold (between 50 and 100 PA) microstimulation generated saccadic eye movements (7). Of these 3 1 neurons, 22 showed task-related activity during the oculomotor delayed-response task, with 17 (54.8%) having significant responses during the delay period ( 10 excitatory and 7 inhibitory). Of these 17 FEF neurons, 10 exhibited directional delay period activity. Figure 15 provides an example of a neuron with a preferred direction at 90”; this FEF neuron showed tonic excitation during the delay period (see Fig. 15B) and also exhibited presaccadic excitation at the time of response when the monkey made saccadic eye movements to the cue location which was the neuron’s preferred direction for delay period activity (see Fig. 15C). Like some PS neurons, the delay period activity was suppressed when the cue was presented at roughly opposite directions to its preferred direction (135, 180, and 225”). Of the 10 FEF neurons with directional delay period activity, 5 had preferred directions into the contralateral hemifield, 4 preferred upward directions, and 1 preferred ipsilaterally directed stimuli. The mean tuning index (Td) was 27.4 t 20.0° (n = 7) for FEF neurons with excitatory directional delay period activity. The three FEF neurons with inhibitory directional delay period activity were very broadly tuned.

D **So

B 90”

D

Ic 1 I ‘I

III11 I

I .I

IR

I I I I I 1 111111111111 II il 1. limlmmll I III II l II I111111111 I I m,l~mlll II I I lanll II I t IIIBYIII I1111 I ihIl I il I l I IIIumuIIII I II I Il t I

I11111 YIIIII,

I Il111 I

E

I II

I I I I II II I11111 II I I I l I I I I I5 I

a

uo503s.

000

.

4ows

50sls I

5035.0-2 200

ms

FIG. 15. Directional delay period activity (F = 29.582; df = 7, 33; P < 0.00 1) of a frontal eye fields neuron. This neuron (5035, right hemisphere) had significant tonic excitatory delay period activity only when the cue was presented at the 45 and 90” locations and was suppressed during the delay when the cue was presented at roughly opposite locations (135”, t = 2.647, df = 43, P < 0.025; 180”. t = 3.158, df = 42, P < 0.005; 225”, t = 3.759, df = 46, P < 0.001). Visual cues were presented randomly at 1 of the 8 locations. All cue eccentricities were 13” and all delay durations were 3 s. A: radial plot showing the directional tuning of delay period activity. B and D: neuron’s delay period activity on the 90” trial and the 225” trial, respectively. C and E: activity aligned at the start of saccadic eye movements at the response period on the 90” trial and the 225” trial, respectively. Note that this neuron had presaccadic activity for 90” saccades and postsaccadic activity for 225” saccades.

200

ms

MEMORY

CODING

IN

*I3 -I--

n

Directional

0

Omnidirectional

A

Both

Activity __I--

Activity ,

FIG. 16. Tracings of the dorsolateral prefrontal cortex showing the location of electrode penetrations for the 3 monkeys (monkq~.s 2,3, and 5) studied. The perspective of the tracings is approximately normal to the dorsolateral prefrontal cortical surface, roughly midway between standard lateral and dorsal views. Locations of neurons with directional delay period activity (a), omnidirectional delay period activity (m), and locations where both types of activity were found (A). Small &ts indicate surface locations of electrode penetrations where neurons with delay period activity were not found. PS, principal sulcus; AS, arcuate sulcus.

Cortical distribution of neurons with directional omnidirectional dela*v*period activity in the prefrontal cortex

and

Recording sites in each hemisphere were located on reconstructions of the dorsolateral prefrontal region. Most electrode penetrations were in the dorsal and ventral banks of the medial and posterior part of the principal sulcus and also in the surrounding regions of the principal sulcus [Walker’s area 46 (63)]; however, in one hemisphere nine penetrations were made in the prearcuate region bordering the anterior bank of the arcuate sulcus where the frontal eye fields are located. Figure 16 illustrates the location of all penetrations and also shows the location of neurons with directional and omnidirectional delay period activity in each of five hemispheres studied. We were unable to discern any clear localization or clustering of neurons with directional delay period activity and neurons with omnidirectional delay period activity. DISCUSSION

Memoryfields

of prefivntal .

neurons

Prefrontal neurons with directionally selective delay period activity were first described in conjunction with the classical manual version of the delaved-response task (50)

PREFRONTAL

CORTEX

345

and a related task, spatial delayed alternation (5 1, 52). Prefrontal neurons in previous studies were characterized as directionally selective because their delay period activity was specific for the left- or right-of-center locations of the cues or the left or right direction of the manual response (50-52). Similar selective delay period activity has been observed in prefrontal neurons during performance of delayed discrimination (22, 47, 58) and conditional response (64) tasks. Although the overall percentage of neurons with directionally selective delay period activity in previous studies of the prefrontal cortex was small [i.e., 6% of total “task-related” neurons in Niki (50); 13% in Niki (5 1); 19% in Niki and Watanabe (53); 10% in Fuster et al. (22)], these neurons are nevertheless reflective of central mnemonic processes guiding the correct choice at the end of delay ( 18, 28, 50). The present study differs from previous recording efforts in that we employed fixation and perimetry in order to extend the analysis of delay-related activity beyond the traditional two-choice (left and right) paradigms (5, 16,20,22, 37, 40-42, 45, 50, 53, 54, 58, 67). We found that the delay period activity of prefrontal neurons codes the spatial coordinates of visual cues during a delayed-response task and thus provides a mnemonic code for direction over the full perimetry of visual space, not for left-right direction alone. Each neuron with directional delay period activity had a “mnemonic” receptive field: only when the cue was presented in that field did the neuron show excitation or inhibition during the subsequent delay period. Moreover, directional delay period activity expanded when the delay was lengthened and faltered on occasional trials when errors were made. Therefore, we propose that this area of the visual field be termed the memory field of the neuron analogous to the receptive fields of visual neurons or the movement fields of oculomotor neurons. Memory fields may be the cellular expression of a working memory process (3) that allows mnemonic information to guide behavior. The mechanisms by which a memory field is constructed are not known at present. It is highly relevant that the caudal prefrontal areas around the principal sulcus and prearcuate cortex are heavily interconnected with the posterior parietal cortex (2, 9, 1 1, 35, 36, 55, 59) which is a major center for processing of spatial vision (1, 49). Our recent anatomic studies of the posterior parietal cortex ( 10) indicate that visuotopic information may be transmitted to posterior parietal areas from a recently described visual area on the medial surface of the occipital lobe (area PO) which receives a direct input from the striate cortex (12) and where neurons are responsive to peripheral stimuli ( 13). The posterior parietal cortex also receives afferents from the superior temporal sulcus, from area MT, from the inferotemporal cortex, and from other prestriate areas ( 10). Thus we speculate that prefrontal neurons may access visuotopic information via the parietoprefrontal pathway, although that information may be also available from other cortical and subcortical sources. Delay period activity has been reported for neurons in several cortical or subcortical structures during performance of delayed-response tasks, as well as in prefrontal cortical areas. These cortical and subcortical structures in-

346

FUNAHASHI,

BRUCE,

elude the posterior parietal cortex (5, 30, 39) the inferotemporal cortex (23, 24), the cingulate cortex (54) the basal ganglia (32, 33) the mediodorsal nucleus of the thalamus (2 l), and the hippocampus (67). Interestingly, almost all of these structures are reciprocally connected with the prefrontal cortex (17, 29, 62). These data are compatible with the idea that the mnemonic information in the prefrontal cortex reflected by neurons with directional delay period activity may be a result of the neural interaction among these cortical and subcortical structures (28, 29) and that many of these structures may be part of a distributed network of cortical and subcortical structures dedicated to spatially guided behavior (29, 6 1).

Later-alit I17of. memorl?- .fields Previous studies of prefrontal cortex in delay tasks have not emphasized the laterality (left or right, contralateral versus ipsilateral) of neuronal activity. In contrast, our study found a predominance of neurons preferring contralateral cues for their delay period activation. Among the 69 PS neurons with directional delay period activity, one half (49.3%) had memory fields centered in the visual field contralateral to the hemisphere where the neurons were located, only 17.4% had ipsilateral memory fields, and 17.4 and 15.9% had memory fields in upward and downward directions, respectively. This result indicates that, although the population of neurons in each hemisphere may be capable of coding cues over the entire visual field, the prefrontal neurons within each hemisphere code mainly contralateral locations. Accordingly, these results provide evidence that memory for spatial location is lateralized, with memories for locations in the right visual field coded mainly in the left hemisphere and memories for left field locations coded in the right hemisphere. This physiological evidence of lateralization of spatial memory is supported by results from lesion studies. Fuster and Alexander (19) observed a higher proportion of errors to the contralateral side in monkeys with unilateral cryogenic depression of the dorsolateral prefrontal cortex. In a study using the oculomotor delayed-response task (15) we found that monkeys with a unilateral surgical ablation in the principal sulcus had marked difficulty making correct saccadic eye movements to remembered visual targets which had been presented in the visual field contralateral to the lesioned hemisphere and only mild or no difficulty making saccades to remembered targets which had been presented in the ipsilateral visual field. A similar result was obtained by Deng et al. for a unilateral lesion of the frontal eye fields on a task requiring memory over a IOO-ms time interval (14). Because the monkeys with unilateral prefrontal lesions made normal saccadic eye movements to targets in both fields when their eye movements were visually guided, the difficulty of making saccades to remembered targets in the oculomotor delayed-response task seems to reflect a mnemonic deficit, which we have termed a “mnemonic hemianopia” ( 15) or “mnemonic scotoma.”

Incidence and specificit . ev of. dela -v period activit _v More research is still needed to know if the principal sulcus region has one population of neurons with delay

AND

GOLDMAN-RAKIC

period activity specific for the oculomotor delayed-response task and a separate population specific for manual types of delay tasks. We saw no indication of an anatomic subregion of the prefrontal cortex particularly dense in neurons with directional delay period activity for the oculomotor task, and our sampling area largely overlapped with the regions where neurons with directional activity on manual types of delay tasks have been found (16, 22, 50, 5 1, 53). Thus it is possible that the same region of the prefrontal cortex codes mnemonic representations across different types of delay tasks, and may even generalize across both sensory (visual versus auditory) and response (hand versus eye) modality. Some evidence favors such a type of generalized coding by individual prefrontal neurons, primarily the relatively high incidence of neurons with directional delay period activity during the oculomotor task (40.5% of total “task-related” neurons) obtained in the present study. If there were separate populations of prefrontal neurons for mnemonic representations in conjunction with eye movements and hand movements, then both populations would be active during the conventional, manual types of delayed-response tasks because primates invariably first direct their gaze with a saccade to stimuli that are to be touched or grasped with the hand. Therefore, previous studies should have found an even higher proportion of prefrontal neurons with delay period activity than we found with the oculomotor delay task. In fact, the reverse is the case, with 6-19% being the incidence of directional delay neurons from previous studies using the manual task, as reviewed earlier. Thus it is possible that the same prefrontal neurons code delay period representations for both oculomotor and manual response systems. On the other hand, there is no anatomic evidence in primates that would indicate that single neurons in the prefrontal cortex have axon collaterals to both eye- and hand-movement centers. Therefore, the question of “supramodal” neurons in the prefrontal cortex must remain open for future evaluation. The higher proportion of prefrontal neurons with directional delay period activity may be explained by two other differences between our study and most previous ones. First, our use of four or eight cue locations increases the number of responsive neurons relative to studies employing only two (left and right) cue locations; many of our neurons with directional delay period activity responded best to cues above or below the fixation point (see Figs. 3 and 4) and would have been classified as nondirectional or nonresponsive if only the histograms for the left and right cue locations are considered. Another important reason is that other observers seem reluctant to attend to inhibitory responses; however, neurons with on/y inhibitory responses account for nearly one-half of our sample of neurons with significant delay period activity and slightly over one-third of the population of directional delay period neurons.

Inhihitorv e delab7 w period activitl?. Inhibitory delay period activity was much more prevalent than previous studies have indicated. Nearly one-half of the neurons with significant delay period activity were solely inhibitory. Inhibitory responses were less likely to be

MEMORY

CODING

IN PREFRONTAL

directional than excitatory ones; even so, nearly one-third of the directional responses were classified as directional because of selective inhibitory responses in the delay period. In addition, some neurons with excitatory delay period activity for particular cue locations had inhibitory activity for other cue locations, usually such inhibition being strongest for, or centered about, cue directions opposite to those optimal for excitatory responses. These inhibitory responses may have several functions. First, neurons with inhibitory omnidirectional delay period activity comprised 16% of all the PS cells with significant delay period activity. We suggest that a tonic reduction in activity over this prefrontal pathway during the delay period may help the monkey suppress all saccadic eye movements for the duration of the delay period. There is an intense projection from the prefrontal cortex to the intermediate and deeper layers of the superior colliculus, the collicular zones responsible for saccadic eye movements (27, 48). A tonic reduction in activity over this pathway may provide a mechanism whereby the prefrontal cortex can prevent inappropriate saccades; patients with frontal lobe lesions have difficulty withholding responses to salient sensory stimuli (3 1). Another role for the inhibitory delay period activity is suggested by the neurons with inhibitory activity for directions opposite the excitatory cues. For these neurons inhibitory activity may be sharpening the spatial tuning of excitatory delay period activity. Opposing patterns of activity are reminiscent of the complex response patterns of neurons in sensory and motor centers that respond with excitation to one parameter of stimulation but are inhibited by the opposite or complementary stimulation, e.g., the opponent vector organization of parieta1 neurons, the color opponency of visual cortical neurons, and the reciprocal activation of motor cortex neurons by extension and flection. Our findings indicate that mnemonic coding is organized on a common physiological principle.

We recorded neurons with memory fields from both the principal sulcal (PS) region [Walker’s area 46 (63)] and from the frontal eye fields (FEF) region in and near the anterior bank of the arcuate sulcus. Although the number of FEF neurons sampled was small, several had directional delay period activity similar to that of PS neurons and the tuning of directional delay period activity was comparable (mean Td = 26.8’ in PS and 27.4” in FEF). Furthermore, the directional tuning of delay period activity in PS is comparable to the directional tuning of presaccadic movement fields in FEF (mean Td of visuomovement neurons = 3 1.3’ I, Ref. 6). Bruce and Goldberg (6) reported that many FEF neurons, particularly those having either tonic visual activity or visuomovement activity, responded throughout a l-s delay period preceding saccades into their visual or move’ All Td estimates of Bruce and Goldberg (6) are inflated by a factor of 1.4 142 because of a program error discovered while completing the present study. Thus their average Tci of 44.2” for visuomovement neurons should be 3 1.3”.

CORTEX

347

ment field. Our FEF results are in general agree with that report and our PS results indicate that many prefrontal neurons anterior to the FEF behave similarly in the context of the delayed-saccade task. The primary pathway of mnemonic information pertinent to saccadic eye movements in the frontal lobe may be from the PS to the FEF. We suggested in the previous section that PS neurons could code mnemonic representations of previous stimuli and impending responses across different delayed-response tasks. The FEF may construct appropriate saccadic commands based on these inputs from the PS region. Other cortex, such as the supplementary motor area, may fabricate hand movements based on similar PS inputs, with the context of the task potentiating one or both of these motor-specific cortical zones. Indeed, Goldberg and Bushnell (26) found that FEF visual activity was only enhanced in the context of eye movements, but Bushnell et al. (8) found that the visual activity on the surface of the inferior parietal lobule, which projects to the PS region, was enhanced in the context of both eye and hand movements. On the other hand, the PS and FEF regions also could function independently with respect to the delayed-saccade task. Both areas receive visual inputs from the temporal and parietal cortices (17, 62) and both have direct projections to the superior colliculus (27, 34, 43, 48) as well as indirect projections there via the corticostriatal system (60, 6 1). The actual circuit for memory guided behavior is probably complex, and at present, we can only postulate that the PS region handles delay information for both saccades and for arm movements whereas the FEF are specific for eye movements.

Oculomotor

dcla evcd-rcsponsc task

We used an oculomotor delayed-response task to examine the functions of the primate prefrontal cortex, whereas most information about previous studies of spatial delayed-response have been conducted in a Wisconsin General Test Apparatus (WGTA), in which freely moving monkeys reach through the bars of a small cage to retrieve a food reward manually after an imposed delay. Testing in a WGTA precludes precise experimental control over the animal’s regard of visual stimuli as well as its behavior during the delay period. In neurophysiological experiments modified versions of the using behaving monkeys, WGTA-based task, such as two-choice (left and right) paradigms, have been used. Although the monkey’s head is fixed and stimulus presentations and response movements are more controlled in this situation, the monkey’s behavior during the delay period, especially its eye movements, have still not been controlled. The question could be raised as to whether, under these testing conditions, a monkey really needs memory to perform correctly, because he could maintain an ocular or postural orientation to the appropriate response key throughout the delay period. Obviously, if a monkey looked at the prospective response window continuously during this period, he would not need to remember the correct cue location. Although investigators have sometimes recorded electrooculographic data during task performance (21, 22, 37, 40, 65, 66) they

34%

FUNAHASHI,

BRUCE,

have not usuallv continuouslv monitored or controlled eve movements during the collection of single neuron activity. In this respect, the oculomotor delayed-response paradigm that we used has several advantages. First, the monkey’s behavior can be controlled more precisely during the performance of the delayed-response task. Because the monkey must maintain fixation of the center spot of light during the delay period, and as there is no clue available during this period to indicate the direction of the correct response, the monkey is required to use working memory to achieve a high level performance. Second, this paradigm allows the use of multiple cue locations in the monkey’s visual field in order to more rigorously test whether directionally selective activity of prefrontal neurons reflects coding of specific spatial information rather than just a general preparatorv set. Finally, recording eye position during the delay as-well as the latency, direction, and amplitude of the eye movement at the time of response allows a more accurate analysis of correlations between prefrontal single neuron activity and behavior. Thus the oculomotor delayed-response task may be a powerful tool for analyzing the mnemonic functions of the prefrontal cortex in both neurophysiological and ablation experiments. The authors thank L. Ladewig and S. Morgenstern for surgical assistance and animal preparation, J. Coburn and M. Pappy for histological assistance, and J. Musco for photography. The authors are also grateful to D. Burman, M. Chaffee, T. Preuss, T. Sawaguchi, and F. Wilson for reading an earlier draft and providing constructive comments. This work was supported by National Institute of Mental Health Grants MH-38546 and MH-00298 to P. S. Goldman-Rakic, National Eye Institute Grant EY-04740 to C. J. Bruce, and by Jacob Javits Center for Excellence in Neuroscience Grant NS-22807. Received

25 April

1988; accepted

in final

form

07 October

1988.

REFERENCES 1. ANDERSEN, R. A. Inferior parietal lobule function in spatial perception and visuomotor integration. In: Handbook of Physiology. The Nervous System. Bethesda, MD: Am. Physiol. Sot., 1987, vol. V, p. 483-518. 2. ANDERSEN, R. A., ASANUMA, C., AND COWAN, W. M. Callosal and prefrontal associational projecting cell populations in area 7A of the macaque monkey: a study using retrogradely transported fluorescent dyes. J. Comp. Neural. 232: 443-455, 1985. 3. BADDELEY, A. D. Working memory. Philos. Trans. R. Sot. Land. B Biol. Sci. 302: 3 1 l-324, 1983. 4. BATUEV, A. S., ORLOV, A. A., AND PIROGOV, A. A. Short-term spatiotemporal memory and cortical unit reactions in the monkey. Acta Physiol. Acad. Sci. Hung. 58: 207-2 16, 198 1. 5. BATUEV, A. S., SHAEFER, V. I., AND ORLOV, A. A. Comparative characteristics of unit activity in the prefrontal and parietal areas during delayed performance in monkeys. Behav. Brain Res. 16: 57-70, 1985. 6. BRUCE, C. J. AND GOLDBERG, M. E. Primate frontal eye fields. I. Single neurons discharging before saccades. J. Neurophysiol. 53: 606-635, 1985. 7. BRUCE, C. J., GOLDBERG, M. E., BUSHNELL, M. C., AND STANTON, G. B. Primate frontal eye fields. II. Physiological and anatomical correlates of electrically evoked eye movements. J. Neurophysiol. 54: 7 14-734, 1985. 8. BUSHNELL, M. C., GOLDBERG, M. E., AND ROBINSON, D. L. Behavioral enhancement of visual responses in monkey cerebral cortex. I. Modulation in posterior parietal cortex related to selective visual attention. J. Neurophysiol. 46: 755-772, 198 1. 9. CAVADA, C. AND GOLDMAN-RAKIC, P. S. Parieto-prefrontal connections in the monkey: topographic distribution within the prefrontal

AND

10.

11.

12.

13.

14.

15.

16.

17. 18.

19. 20. 2 1.

22.

GOLDMAN-RAKIC cortex of sectors connected with the lateral and medial posterior parieta1 cortex. Sot. Neurosci. Abstr. 11: 323, 1985. CAVADA, C. AND GOLDMAN-RAKIC, P. S. Subdivisions of area 7 in the rhesus monkey exhibit selective patterns of connectivity with limbit, visual and somatosensory cortical areas. Sot. Neurosci. Abstr. 12: 262, 1986. CHAVIS, D. A. AND PANDYA, D. N. Further observations on corticofrontal connections in the rhesus monkey. Brain Res. 117: 369-386, 1 Q7C; I/I”. COLBY, C. L., GATTASS, R., OLSON, C. R., AND GROSS, C. G. Topographic organization of cortical afferents to extrastriate visual area PO in the macaque: a dual tracer study. J. Comp. Neural. 269: 392-413, 1988. COVEY, E., GATTASS, R., AND GROSS, C. G. A new visual area in the parieto-occipital sulcus of the macaque. SX. Neurosci. Abstr. 8: 68 1, 1982. DENG, S.-Y., GOLDBERG, M. E., SEGRAVES, M. A., UNGERLEIDER, L. G., AND MISHKIN, M. The effect of unilateral ablation of the frontal eye fields on saccadic performance in the monkey. In: Adaptive Processes in Visual and Oculomotor Systems, edited by E. L. Keller and D. S. Zee. New York: Pergamon, 1987, p. 20 l-208. FUNAHASHI, S., BRUCE, C. J., AND GOLDMAN-RAKIC, P. S. Perimetry of spatial memory representation in primate prefrontal cortex: evidence for mnemonic hemianopia. Sot. Neurosci. Abstr. 12: 554, 1986. FUSTER, J. M. Unit activity in prefrontal cortex during delayed-response performance: neuronal correlates of transient memory. J. Neurophysiol. 36: 6 l-78, 1973. FUSTER, J. M. The Prefrontal Cortex. New York: Raven, 1980. FUSTER, J. M. The prefrontal cortex and temporal integration. In: Cerebral Cortex, edited by A. Peters and E. G. Jones. New York: Plenum, 1985, vol. 4, p. 15 l- 177. FUSTER, J. M. AND ALEXANDER, G. E. Delayed response deficit by cryogenic depression of frontal cortex. Brain Res. 20: 85-90, 1970. FUSTER, J. M. AND ALEXANDER, G. E. Neuron activity related to short-term memory. Science Wash. DC 173: 652-654, 197 1. FUSTER, J. M. AND ALEXANDER, G. E. Firing changes in cells of the nucleus medialis dorsalis associated with delayed response behavior. Brain Res. 61: 79-91, 1973. FUSTER, J. M., BAUER, R. H., AND JERVEY, J. P. Cellular discharge in the dorsolateral prefrontal cortex of the monkey in cognitive tasks.

Exp. Neural. 77: 679-694, 1982. 23. FUSTER, J. M., BAUER, R. H., AND JERVEY, J. P. Functional interactions between inferotemporal and prefrontal cortex in a cognitive task. Brain Res. 330: 299-307, 1985. 24. FUSTER, J. M. AND JERVEY, J. P. Neuronal firing in the inferotemporal cortex of the monkey in a visual memory task. J. Neurosci. 2: 361-375, 1982. 25. VAN GISBERGEN, J. A. M., ROBINSON, D. A., AND GIELEN, S. A quantitative analysis of generation of saccadic eye movements by burst neurons. J. Neurophysiol. 45: 4 17-442, 198 1. 26. GOLDBERG, M. E. AND BUSHNELL, M. C. Behavioral enhancement of visual responses in monkey cerebral cortex. II. Modulation in frontal eye fields specifically related to saccades. J. Neurophysiol. 46: 773-787, 1981. 27. GOLDMAN, P. S. AND NAUTA, W. J. H. Autoradiographic demonstration of a projection from prefrontal association cortex to the superior colliculus in the rhesus monkey. Brain Res. 116: 145- 149, 1976. 28. GOLDMAN-RAKIC, P. S. Circuitry of primate prefrontal cortex and regulation of behavior by representational memory. In: Handbook of Physiology. The Nervous System. Bethesda, MD: Am. Physiol. Sot., 1987, vol. V, p. 373-417. 29. GOLDMAN-RAKIC, P. S. Topography of cognition: paralleled distributed networks in primate association cortex. Ann. Rev. Neurosci. 1 I : 137-156, 1988. 30. GNADT, J. W. AND ANDERSEN, R. A. Memory related motor planning activity in posterior parietal cortex of macaque. Exp. Brain Res. 70: 2 16-220, 1988. 31. GUITTON, D., BUCHTEL, H. A., AND DOUGLAS, R. M. Frontal lobe lesions in man cause difficulties in suppressing reflexive glances and in generating goal-directed saccades. Exp. Brain Res. 58: 455-472, 1985. 32. HIKOSAKA, 0. AND SAKAMOTO, M. Cell activity in monkey caudate nucleus preceding saccadic eye movements. Exp. Brain Res. 63: 659-662, 1986. 33. HIKOSAKA, 0. AND WURTZ, R. H. Visual and oculomotor functions

MEMORY

34.

35.

36.

37.

38.

39.

40. 41.

42.

43.

44.

45.

46.

47.

48.

49.

CODING

IN

of monkey substantia nigra pars reticulata. III. Memory-contingent visual and saccade responses. J. Neurophysiol. 49: 1268- 1284, 1983. HUERTA, M. F., KRUBITZER, L. A., AND KAAS, J. H. Frontal eye field as defined by intracortical microstimulation in squirrel monkeys, owl monkeys, and macaque monkeys. I. Subcortical connections. J. Clomp. Neurd. 253: 4 15-439, 1986. JACOBSON, S. AND TROJANOWSKI, J. Q. Prefrontal granular cortex of the rhesus monkey. I. Intrahemispheric cortical afferent. Brain Res. 132:209-233,1977. JONES, E. G. AND POWELL, T. P. S. An anatomical study of converging sensory pathways within the cerebral cortex of the monkey. Brain 93:793-820, 1970. JOSEPH, J. P. AND BARONE, P. Prefrontal unit activity during a delayed oculomotor task in the monkey. Exp. Brain Rc~s.67: 460-468, 1987. JUDGE, S. J., RICHMOND, B. J., AND CHU, F. C. Implantation of magnetic search coils for measurement of eye position: an improved method. Vision Rcs. 20: 535-538, 1980. KOCH, K. W. AND FUSTER, J. M. Single unit activity in the parietal cortex of the primate during a haptic delayed matching-to-sample task. Sot. Ncurosci. Ahstr. 1 1: 1275, 1985. KOJIMA, S. Prefrontal unit activity in the monkey: relation to visual stimuli and movements. Exp. Neurd. 69: 1 lo- 123, 1980. KOJIMA, S. AND GOLDMAN-RAKIC, P. S. Delay-related activity of prefrontal neurons in rhesus monkeys performing delayed response. Brain Res. 248: 43-49, 1982. KOJIMA, S. AND GOLDMAN-RAKIC, P. S. Functional analysis of spatially discriminated neurons in prefrontal cortex of rhesus monkey. Brain Rev. 29 1: 229-240, 1984. KOMATSU, H. AND SUZUKI, H. Projections from the functional subdivisions of the frontal eye field to the superior colliculus in the monkey. Brain Rcs. 327: 324-327, 1985. KUBOTA, K. AND FUNAHASHI, S. Direction-specific activities of dorsolateral prefrontal and motor cortex pyramidal tract neurons during visual tracking. J. Ncurophysiol. 47: 362-376, 1982. KUBOTA, K., IWAMOTO, T., AND SUZUKI, K. Visuokinetic activities of primate prefrontal neurons during delayed-response performance. J. N~~urophysiol. 37: 1 197- 12 12, 1974. KUBOTA, K. AND NIKI, H. Prefrontal cortical unit activity and delayed alternation performance in monkeys. J. Ncuruph~vsid. 34: 337-347,1971. KUBOTA, K., TONOIKE, M., AND MIKAMI, A. Neuronal activity in the monkey dorsolateral prefrontal cortex during a discrimination task with delay. Brain Res. 183: 29-42, 1980. LEICHNETZ, G. R., SPENCER, R. F., HARDY, S. G. P., AND ASTRUC, J. The prefrontal corticotectal projection in the monkey: an anterograde and retrograde horseradish peroxidase study. Neurosckvzce 6: 1023-1041, 1981. MOUNTCASTI,E, V. B. The world around us: neural command functions for selective attention. Ncurosci. Rex Program Bull. 14, Supp!.. l-47,1976.

PREFRONTAL

CORTEX

349

50. NIKI, H. Differential activity of prefrontal units during right and left delayed response trials. Brain Rcs. 70: 346-349, 1974. 51. NIKI, H. Prefrontal unit activity during delayed alternation in the monkey. I. Relation to direction of response. Brain Res. 68: 185- 196, 1974. 52. NIKI, H. Prefrontal unit activity during delayed alternation in the monkey. II. Relation to absolute versus relative direction of response. Brain Rcs. 68: 197-204, 1974. M. Prefrontal unit activity and delayed 53. NIKI, H. AND WATANABE, response: relation to cue location versus direction of response. Brain Rc.s. 105: 79-88, 1976. M. Cingulate unit activity and delayed 54. NIKI, H. AND WATANABE, response. Brain Rcs. 1 10: 38 l-386, 1976. 55. PETRIDES, M. AND PANDYA, D. N. Projections to the frontal cortex from the posterior parietal region in the rhesus monkey. J. Camp. N~urol. 228: iO5- 1 16, 1984. 56. ROBINSON, D. A. A method of measuring eye movement using a scleral search coil in a magnetic field. IEEE Trans. Biomd Eng. 10: 137-145, 1963. C. E. Functional heterogeneity of the prefrontal cortex 57. ROSENKILDE, in the monkey: a review. Bchav. Nc~urd. Bid. 25: 30 l-345, 1979. C. E., BAUER, R. H., AND FUSTER, J. M. Single cell 58. ROSENKILDE, activity in ventral prefrontal cortex of behaving monkeys. Brain Res. 209:375-394, 1981. M. L. AND GOLDMAN-RAKIC, P. S. Callosal and intra59. SCHWARTZ, hemispheric connectivity of the prefrontal association cortex in rhesus monkey: relation between intraparietal and principal sulcal cortex. J. Camp. Ncurd. 226: 403-420, 1984. L. D. AND GOLDMAN-RAKIC, P. S. Longitudinal topogra60. SELEMON, phy and interdigitation of corticostriatal projections in the rhesus monkey. J. Ncurosci. 5: 776-794, 1985. L. D. AND GOLDMAN-RAKIC, P. S. Common cortical and 61. SELEMON, subcortical target areas of the dorsolateral prefrontal and posterior parietal cortices in the rhesus monkey: evidence for a distributed neural network subserving spatially guided behavior. J. Ncurosci. 8: 4049-4068, 1988. D. F. The Frontal Lobes. New York: 62. STUSS, D. T. AND BENSON, Raven, 1986. A. E. A cytoarchitectural study of the prefrontal area of the 63. WALKER, macaque monkey. J. Camp. Ncurd. 73: 59-86, 1940. M. Prefrontal unit activity during delayed conditional 64. WATANABE, discriminations in the monkey. Brain Res. 225: 5 l-65, I98 1. M. Prefrontal unit activity during delayed conditional 65. WATANABE, Go/No-Go discrimination in the monkey. I. Relation to the stimulus. Brain Res. 382: I- 14, 1986. M. Prefrontal unit activity during delayed conditional 66. WATANABE, Go/No-Go discrimination in the monkey. II. Relation to Go and No-Go responses. Brain Ros. 382: 15-27, 1986. T. AND NIKI, H. Hippocampal unit activity and delayed 67. WATANABE, response in the monkey. Brain Res. 325: 24 l-254, 1985.