Endogenously Generated and Visually Guided after Lesions of the Human Frontal Eye Fields

Avishai Hen& Ben Gurion University of the Negev, Israel

Robert Rafhl Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/6/4/400/1755171/jocn.1994.6.4.400.pdf by guest on 18 May 2021 University of California, Davis Dell Rhodes Reed College

Abstract Nine patients with chronic, unilateral lesions of the dorso- toward the ipsilesional $eld had abnormally prolonged laten- lateral including the frontal eye fields (FEF) cies; they were comparable to the latencies observed for vol- made saccades toward contralesional and ipsilesional fields. untary SdCcddeS. The effect of FEF lesions on saccacles The saccades were either voluntarily directed in response to contrasted with those observed in a second experiment re- arrows in the center of a visual display, or were reflexively quiring a key press response: FEF lesion patients were slower summoned by a peripheral visual signal. latencies were in making key press responses to signals detected in the con- compared to those made by seven neurologic control patients tralesional field. To assess covert and preparatory set with chronic, unilateral lesions of dorsolateral prefrontal cortex the effects of precues providing advance information were sparing the FEF, and by 13 normal control subjects. In both the measured in both saccade and key press experiments. Neiher normal and neurologic cohl[rolsubjects, reflexive saccades had patient group showed any deficiency in using precues to shili shorter Latencies than voluntary sdccades . In the FEF lesion attention or to prepare saccades. patients, voluntary saccades had longer latencies toward the The FEF facilitates the generation of voluntary saccatles and contralesional field than toward the ipsilesional field. The op- also inhibits reflexive saccades to exogenous signals. FEF Ic- posite pattern was found for reflexive saccades: latencies of sions may disinhibit the ipsilesional midbrain which in turn saccades to targets in the contralesional field were shorter than may inhibit the opposite colliculus to slow reflexive s:iccades saccades summoned to ipsilesional targets. Reflexive saccades toward the ipsilesional field.

INTRODUCTION saccades. Monkeys with lesions show an increase in saccade latency (Goldberg & Wurtz, 1972; The frontal lobes are the latest structures to develop both Schiller, Sandell, & Maunsell, 1987). In humans the SLI- phylogenetically and ontogenetically: Their development perior colliculus is involved not only in triggering re- is most advanced in the human and their myelin- flexive saccades (Rafal, Smith, Krantz, Cohen, & Brennan, ization is not complete until long after birth. They are 1990) but also in moving visual attention to exogenous involved in complicated activities that require planning. signals (Rafal, Henik, & Smith, 1991; Rafal, Posner, Fried- This regulation of voluntary, goal-directed behavior may man, Inhoff, & Bernstein, 1988). require modulating or adapting subcortical reflexes. It is Cortical mechanisms, on the other hand, are necclcd likely, for example, that the frontal eye fields (FEF) affect for generating voluntary saccadic eye movements under reflexive saccades generated by subcortical systems strategic guidance (Bruce & Goldberg, 1985; Deng, Gold- (Goldberg & Segraves, 1989). berg, Segraves, Ungerleider, & Mishkin, 1986; Goldberg Saccades can be reflexively summoned, as when turn- & Segraves, 1989). In early work, Holmes (1938) ob- ing toward a movement seen out of the corner of the served that patients with frontal cortex lesions have dif- eye; or they can be deployed endogenously, as when ficulty executing saccades in response to verbal looking both ways before crossing the street. Brainstem command. Luria, Karpov, and Yarbuss (1966) suggested programs involving primarily the superior colliculus are that ocular scanning patterns of these patients are con- critical in triggering reflexive, peripherally summoned trolled entirely by external stimuli rather than by instruc-

0 1994 Masacbusens Institute of Technology Journal of Cognitiue Neuroscience 64,pp. 400-41 I

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/jocn.1994.6.4.400 by guest on 27 September 2021 tions. Recent work suggests that interactions between the the saccade is being withheld (Bruce & Goldberg, 1985). frontal cortex and the superior colliculus may arbitrate To study this aspect of preparatory set, saccade targets between external and internal demands on the oculo- were preceded by a preparatory cue (see Fig. 2). The motor apparatus (Bruce & Goldberg, 1985; Schiller et al., cue could provide correct information about the spatial 1987; Schiller, True, & Conway, 1980). Surgical lesions direction of the forthcoming required response (infor- of frontal lobes in humans produce a deficit in inhibiting mative cue) or supply no directional information (neutral “reflexive glances” (Guitton, Buchtel, & Douglas, 1985) cue). The advance information provided by a informative and unilateral lesions of the FEF may shorten latencies cue should allow the subject to prepare to respond to of reflex saccades to targets in the contralesional field the target, shortening the latency to respond compared (Pierrot-Deseillgny, Rivaud, Penet, & Rigolet, 1987). to neutral cue trials. It was predicted that FEF lesions Single cell recordings provide evidence for a specific might cause a loss of facilitation from advance informa- role of the frontal eye fields in endogenous saccade tion, at least for endogenously generated saccades. Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/6/4/400/1755171/jocn.1994.6.4.400.pdf by guest on 18 May 2021 generation (Bruce & Goldberg, 1985; Goldberg & Se- graves, 1989). These show that the FEF contains cells that respond in temporal correlation with purposive saccades RESULTS even in the absence of an exogenous visual stimulus. In Experiment 1: Saccade Task addition, some cells show anticipatory activity preceding Normal Control Subjects a cue fir saccade execution if the monkey could predict the dimensions of the saccade. Bruce and Goldberg con- Saccade latencies for the control subjects were subjected cluded that “The frontal eye fields may be the means by to a three-way repeated measures analysis of variance. which cortical decisions can access the brainstem’s sac- The three within-subject factors were target type (central cade generator. In contrast, the superior colliculus may arrow vs. peripheral target), cue validity (informative VS. provide a direct and rapid mechanism whereby visual neutral cue), and saccade direction (right vs. left). Mean stimuli can trigger saccades” (pp. 632-633) (Bruce & reaction times for target type and cue validity, averaged Goldberg, 1985). over saccade direction, are presented in Table 2. Thus, the FEF may be necessary both for endogenously Saccade latencies were shorter to a peripheral target generating voluntary saccades, and for inhibiting more than from a central arrow [F(1,12)= 7.00,p < 0.051, and automatic brainstem mechanisms that generate reflexive were shorter following informative cues [F(1,12) = saccades summoned by exogenous signals. The present 118.53, p < O.OOl]. A target type X validity interaction work measured the latencies of these two types of sac- [F(1,12) = 7.27, p < 0.051 obtained because saccade cades in patients with chronic unilateral lesions of the latencies were shorter to peripheral than to central t;ir- dorsolateral prefrontal cortex. Patients whose lesions in- gets following neutral cues [F(1,12) = 9.72,p < 0.011, volved the superior dorsolateral prefrontal cortex, in- but not following informative cues. Saccade latencies cluding areas thought to include FEF, were compared to were shorter when saccades were made to the right than two groups of control subjects: age-matched normal sub- when they were made to the left [right = 334 msec, jects, and other neurologic patients with lesions of dor- left = 349 msec; F(1,12) = 11.82,p < 0.0051. Saccade solateral prefrontal cortex sparing the FEF (see Table 1 direction did not interact with effects of target type or and Fig. 1). Visually guided saccades were made to a validity. bright signal appearing 10” eccentric to fixation. Endog- enously generated saccades were directed by a central Patients symbol (an arrow head pointing to left or to right) that appeared at fixation (see Fig. 2). It was predicted that Saccade latencies of the patients were subjected to a four- FEF lesions would increase the latencies of endogenously way analysis of variance. The three within-subject factors generated sdccades to the contralesional field, and might were target type (central arrow vs. peripheral target ), decrease the latencies of reflexive saccades to the con- cue validity (informative vs. neutral cue), and field to tralesional field. To determine whether any obtained ef- which saccades were directed (ipsilesional vs. contrale- fects are specific to oculomotor responses, a second sional field). The between groups factor was subject experiment required the same subjects to make a choice group (FEF vs. NFEF). Mean saccade latencies for the reaction time (RT) key press response to the same stim- various conditions are presented in Table 3. uli. The two patient groups did not differ in mean saccade In addition to studying the role of the FEF in saccade latencies [F(1,14) = 1.54, p > 0.21. As in the control execution, a second goal of the study was to investigate subjects, peripheral targets produced saccades with the hypothesis that the FEF might have a role in prepar- shorter latencies than did central arrows [F(1~4) = ing eye movements, i.e.,in oculomotor set. This objective 20.94,~< 0.0011, and informative cues expedited saccade was based on the observation that some single units in latencies relative to neutral cues [F(1,14) = 13.83,p < FEF discharge when a monkey is preparing to make a 0.0051. There were no significant interactions involving saccade to a remembered target during a period in which cue validity for the patients: Both patient groups bene-

Heiik el al. 401

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/jocn.1994.6.4.400 by guest on 27 September 2021 Table 1. Clinical Information for Each Patienta Lesion Vintage volume Clinical Patient Agelsex Lesion Hemisphere (years) (cc) stgm FEF Lesion Group Rh4 65 M Stroke R 4 29 FR 69 M Stroke L 5 31 Aphasia CI 67 F Stroke L 7 70 Aphasia, hemiparesis HS 66 M Stroke I. 7 81 Aphasia, hemiparesis Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/6/4/400/1755171/jocn.1994.6.4.400.pdf by guest on 18 May 2021 LS 60 F Meningioma L 10 28 Al. 61 F Stroke L 10 51 Aphasia FS 75 M Stroke L 10 49 Aphasia, hemiparesis RK 60 M Stroke R 2 106 Aphasia, hemiparesis .Ic 63 M Stroke L 3 103 Aphasia, hemiparesis No FEF Lesion Group TI 64 M Stroke L 3 47 Aphasia MM 65 M Stroke R 3 52 RT 71 M Stroke L 6 39 Aphasia MH 67 M Stroke R 6 49 JD 62 M Stroke L 11 31 Aphasia MG 24 M AVM R 4 25 JH 61 M Stroke L 8 77 Aphasia, hemiparesis

“Vintage refers to the number of years since the stroke, or since surgical resection. AVM, aneriovenous malformation

fitted equally from cues that instructed them to prepare contralesional field of patients with FEF lesions [F( 1,s)= saccades, and this benefit of preparation was not differ- 11.53,~< 0.011, is not present in the “good” field ipsi- entially impaired for saccades to the contralesional field lateral to the FEF lesion. in either group of patients. The major results of this experiment are evident in interactions between contralesional and ipsilesional sac- Individual Patients cade latencies in the FEF lesion group. These are shown The major result of this study is that FEF lesions had in Figure 3, which reveals a significant group X field X opposite effects on voluntary and reflexive saccades. Fig- target type interaction [F(1,14) = 13.33,p < 0.0051. For ure 4 depicts the data for individual patients in the two patients with lesions involving FEF, there was a target groups to demonstrate the consistency of this dissocia- gpe x field interaction [F(1,8)= 19.88,p < 0.0051: tion. Figure 4A shows that centrally controlled saccades voluntan saccades had longer latencies toward the con- had longer latencies toward the contralesional field in tralesional than ipsilesional field [F(1,8)= 11.16,p < seven of nine of the patients in the FEF group, and that 0.051; whereas reflexive saccades had longer latencies seven of nine showed shorter latencies for reflexive sac- toward the ipsilesional than contralesional field [F( 1,8) = cades to the contralesional field. Only one patient in this 9.82,~< 0.05I. Patients with lesions of dorsolateral pre- group showed no effect of the FEF lesion on either type frontal cortex that spared the FEF showed no such target of saccade. It is noteworthy that this patient was the only type x field interaction [F(1,6) < 11 nor did they show patient in the FEF lesion group who had not had a stroke. ;I main effect of field [F( 1,6) < 11. She had had a meningioma resected; and this lesion, 1:igurc 3A shows another interesting finding in the growing slowly over a long time, may have resulted in patients with lesions of the FEF. In the ipsilesional field an atypical functional architecture of the dorsolateral there is no difference in saccade latency between reflex- prefrontal cortex. Figure 4B shows the results for indi- ive and endogenous saccades [F( 1,s) = 0.001.That is, the vidual patients in the NFEF group. No consistent field- usual aclvantage for reflexive saccades, which is present related effects on saccade latency are evident for either in the normal and non-FEF control subjects and in the type of saccade in these patients.

402 Joicrnul of‘ Cognitit v Neuroscience

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/jocn.1994.6.4.400 by guest on 27 September 2021 FRONTAL EYE FIELD LESION 1 2 3 Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/6/4/400/1755171/jocn.1994.6.4.400.pdf by guest on 18 May 2021

4 7

NO FRONTAL EYE FIELD LESION

1 2 3

4 5 1: 6 7

Figure 1. Averaged lesion extent in patients with lesions extending into the frontal eye fields (top) and those in whom the lesions spared the frontal rye fields (bottom). The scale refers to the percent of patients in a group with lesions in that area. The lines on the lateral reconstruc- tion indicate corresponding axial cuts. All lesions are reflected onto the left side.

Other Neuroanutomic Anu(vses For these reasons we conducted several other analyses. These analyses used, as a variable, the field difference The emphasis on the consistency of effects across indi- scores, for reflexive and voluntary saccades, and related vidual subjects is important because the patient them to appropriate anatomic parameters. (1) Correla- subgroups were not entirely homogeneous. For exam- tion of lesion volume across the 16 patients; (2) t tests ple, although the two groups did not differ signzjkantly comparing grouping of the 16 patients according to in lesion volume, there was a tendency for lesions in the whether the lesion affected the left or right hemisphere; FEF group to be larger; and to extend a bit more deeply and (3) t tests comparing grouping of the 16 patients into subcortical regions (Fig. 1).

Heniketal. 403

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/jocn.1994.6.4.400 by guest on 27 September 2021 Figure 2. The design of the saccdde experiment (Experi- Informative Neutral ment 1 ). In the centrally con- trolled sdccade task (right hottom), the subject made an directed by a large arrow in the center of the diaplay. In the peripherally 300ms summoned saccade task (leti bottom), the subject made an 4 eye movement to a large mer- isk presented 10" to left or right. Each trial began with a precue that might or might nut Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/6/4/400/1755171/jocn.1994.6.4.400.pdf by guest on 18 May 2021 provide advance information to i enable the subject to prepro- Precue-TargetInterval gram the direction of the forthcoming saccade. The pre- cue could he informative (top 500 rnr left) or it could be neutral (top A right). The efficacy of saccade v preparation was meawred as a 1250 mr facilitation in saccade latency in the informative precue con- 4 dition compared to the neutral precue condition. In Experi- ment 2 the visual display was identical to that shown above i for Experiment 1. The subjects' task was different: they main- TARGET Until tained fixation and made a 0 El 0'0 Response choice RT key press response to the target.

- Peripheral Central

Table 2. Response Latency in msec (SD in Parentheses) for Normal Control Subjects in the Sdccade Task (Experiment 1) :ind in the Key Press Task (Experiment 2)

~~ ~~ Informative Neutrul Infomtizle precue pecue precue Endogenously Generated Saccade Visually Guided Saccade Saccade 33 1 406 304 346 Tack (40) (53) (49) (59)

Central Target Peripheral Target Key Press 342 362 438 Task (66) (78) (74)

according to whether the lesion extended subcortically Experiment 2: Key Press Response into the striatum or internal capsule (4 out of 16). The purpose of Experiment 2 was to determine whether Lesion volume did not correlate significantly with field the effects of FEF lesions seen in Experiment 1 were asymmetries for either voluntary or reflexive saccades. specific to saccades. The subject5 responded to the same Hemisphere of lesion did not interact with hemifield targets but made manual key press responses rather than asymmetries (t = 0.56), nor did subcortical extension saccades. interact with hemifield asymmetries.

404 Journal of Cognitive Neuroscience Volume 6, Number 4

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/jocn.1994.6.4.400 by guest on 27 September 2021 Table 3. Mean Patient Saccade Latencies in msec toward the Contralesional and the lpsilesional Field (SD in Parentheses).

Contralesional Ipsilesional Informatiue Neutral Informutitie Neutral precue precue precue prectie Endogenously Generated Saccade FEF 446 558 389 46.'; Patien 1,s (129) (121) (88) (109) Non-FEF 411 488 414 Patient5 (51) (85) (90)

Visually Guided Saccade Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/6/4/400/1755171/jocn.1994.6.4.400.pdf by guest on 18 May 2021 FEF 292 338 403 Patienis (65) (149) (97) Non-FEF 308 281 Patients (89) (58)

- -< 8-- Central Target Peripheral Target T - T 550 b FEF LesimMimh

250 J 250 I Contralesional lpsilesional Contralesional lpsilesional Saw& Saccade saccade Sam&

Figure 3. Mean of the median saccade latencies in msec (averaged across the two precue conditions) for endogenously generated (central target) and visually guided (peripheral target) saccades toward the contralesional and ipsilesional tields. (A) FEF lesion group. (B) No FEF lesion group

(A) Central Target Peripheral Targel Central Target Peripheral Targd 300 1 (B) 300 200 7 200 , Contralesional 100 Contralesionai 100 I I* minus I 'T' minus 7 losilesional 0 1~1 .Ill( lpsibsronal 0 I

Saccade ~ , Latency I (msec) -200 -300 I I I -400 ' -400-300 ! FM R a HS LS AL R w x ffi !A4 TJ AT JD JH hM

I I

Figure 4. Individual patient differences (shown by a vertical line for each patient) in median saccade latencies toward the ipsileaional and contralesional fields for endogenously generated (central target) and visually guided (peripheral target) saccades. (A) FEF lesion group. (B) No FEF lesion group.

Henik et al. 405

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/jocn.1994.6.4.400 by guest on 27 September 2021 Normal Control Subjects press experiment, the latencies to contralesional periph- eral targets tended to be longer than to ipsilesional tar- Key press latencies for the control subjects were sub- gets. jected to a three-way (target type X cue validity X side of key press) repeated measures analysis of variance identical to that performed on saccade latencies in Ex- DISCUSSION periment 1. Mean reaction times for target type and cue validity, averaged over side of key press, are presented The major findings in patients with lesions of the frontal in Table 2. eye fields were as follows: (1) Longer latencies for cen- Mean key press latencies did not differ significantly trully-directed saccades toward the contralesional coni- between the central arrow and peripheral target condi- pared to the ipsilesional field. (2) Shorter latencies for

tions. Targets following informative cues were re- vkual!y guided saccades toward the contralesional coni- Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/6/4/400/1755171/jocn.1994.6.4.400.pdf by guest on 18 May 2021 sponded to faster than targets following neutral cues pared to the ipsilesional field. This speeding of responses [F(1,12) = 76.34, p < 0.0011. As in the saccade task to contralesional compared to ipsilesional signals was experiment, informative precues produced greater facil- specific to eye movements. There were no comparable itation for central targets [F(1,12) = 17.31,p < 0.0051. reductions in key press latencies to targets in the con- There were no significant effects involving side of key tralesional field; in fact key press latencies to contrale- press. sional peripheral targets tended to be longer than to ipsilesional targets. (3) Abnormally prolonged latencies for visually guided saccades toward the ipsilesionul field. Purients These saccades had longer latencies than did the reflex- ive saccades toward the contralesional field in these pa- Key press latencies of the patients were subjected to a tients. Furthermore, peripherally summoned saccades to four-way (group X target type X cue validity x field) the ipsilesional field did not appear to be reflexively analysis of variance identical to that performed on sac- triggered: their latencies were not different from volun- cade latencies in Experiment 1. Mean RTs are presented tary saccades directed towards the ipsilesional field. (4) in Table 4. No impairment in the ability to use advance information As was the case for saccades, key press RTs were from precues to prepare saccades toward the contrale- shorter following informative cues [F( 1,14) = 103.40,p < sional field for either endogenous or reflexive saccades 0.0011, and cue validity did not interact with group or In the patients with lesions of the dorsoiateral prefrontal field, suggesting unimpaired ability to utilize advance cortex that spared the FEF, no field asymmetry for either information in the patients. Figure 5 shows that the target voluntary or for reflexive saccades was evident. These ’)pe X field interaction for the FEF patients in the key results lead to several conclusions about the role of the press experiment [F(l$) = 5.26,~= 0.051 was opposite frontal eye fields in generating endogenous saccades, and of that seen for saccades [experiment X target type X about frontocollicular interactions for modulating reflex- field interaction: F( 1,s) = 20.93,p < 0.0051. In the sac- ive, peripherally summoned saccades. cade experiment, response latencies for saccades to pe- ripheral targets in the contralesional field were shorter 1. The FEF plays an important role in generating vol- than toward the ipsilesional field, whereas in the key untary saccades toward the contralesional field. The tind-

Table 4. Patient Key Press Reaction Time in insec for Contralesional and Ipsilesional Targets (SD in Parentheses) Contrulesionul tpsilesional

f njormutii je Neutrul Informutiiie Neirtrul precire precue precue precirr Central Target FE F 449 565 451 576 Patients (91 1 (85) (1011 (86) 389 494 38 1 4-77 (101) (122) (122) (160)

Peripheral Target 52 1 612 485 578 (105) (158) (81) (115)

J06 &~rrnzal of Cognrtii ie Neuroscierice Volrrnie 6, Nrrniber 4

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/jocn.1994.6.4.400 by guest on 27 September 2021 FEF Leslon Patlents (B) No FEF Lesion Patients --< ,-- Central Target Peripheral Target T - -. 8-- Central Target -Peripheral Target

I 400’ 400 I Contralateral Contralateral lpsilateral lpsilateral Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/6/4/400/1755171/jocn.1994.6.4.400.pdf by guest on 18 May 2021 Response Response Response Response

Figure 5. Key press RTs in nisec for central and peripheral targets in either the contralesional or ipsilesional field. (A) FEF lesion group. (8) No FEI: lesion group

ings indicate an important FEF role in generating contralateral to the cortical lesion: “This initial hemi- endogenously generated saccades in humans, and con- anopia is apparently due to depression of function of the verge with observations of oculomotor behavior after colliculus ipsilateral to the cortical lesion, a depression FEF lesions in monkeys (Deng et al., 1986). maintained by influx of inhibition from the crossed col- 2. FEF lesions may result in disinhibition of the ipsi- liculus. Thus, removal of the contralateral tectum, or lateral midbrain structures, and this ipsilesional disinhi- splitting of the collicular commisure, abolishes this in- bition may in turn result in inhibition of the opposite hibition and allows the return of function in the ipsilat- colliculus with the result that reflexive saccades are de- era1 colliculus, and with it the recovery from layed toward the ipsilesional field and must be supported hemianopia” (Sprague, 1966, p. 1544). More recent stud- endogenously. ies suggest that this “Sprague effect” is mediated by 3. The normal effects of preparatory cues suggest that GABAegic projections from the substantia nigra pars dorsolateral prefrontal cortex may not be essential either reticulata (Wallace, Rosenquist, & Sprague, 1989, 1990). for orienting covert visual attention or for preprogram- The current results suggest that lesions of the frontal eye ming oculomotor set preceding saccades. Our experi- fields may have the opposite effect on the extrageniculate ment did not reveal any deficit in using advance from occipital lesions, i.e., disinhibition of information in preparing eye movements. However, only the ipsilesional substantia nigrapars reticulata produces saccade latency was measured, not saccade accuracy or suppression of collicular function contralateral to the other saccade parameters. Moreover, our study did not cortical lesion. incorporate invalid cues, it., a condition in which an eye The findings of the current study are consistent with movement is prepared in one direction, but which must previous observations that patients with le- subsequently be executed to a target appearing at an sions have trouble making “antisaccades” away from a unexpected location. Nagel-Leiby, Buchtel, and Welch peripheral visual signal (Guitton et al., 1985). Further, (1990) showed that frontal cortex lesions did affect the our findings help to clarify the reasons for this difficulty, amplitude of saccades after invalid cues. and why it is present for signals in both visual fields. Consider just what the requires. First the The observation of slowed visually guided saccades to subject must inhibit a “reflexive glance” toward the pe- the ipsilesional field in patients with FEF lesions is a new ripheral signal. In addition, the peripheral signal also finding. We propose that frontal eye field lesions elimi- tells the subject to saccade toward the opposite hemi- nate inhibitory influences of ipsilesional midbrain path- field, a saccade that must be generated endogenously ways, possibly to the substantia nigra (pars reticulata). based on the symbolic meaning (“go to the other field”) The disinhibited substantia nigra may then inhibit the of the cue. From our findings it seems clear that patients contraleral colliculus. This inhibition abolishes the re- with FEF lesions must be confounded by the antisaccade flexive component of visually guided saccades toward task for two reasons: (1) When the signal appears in the ipsilesional signals resulting in longer latencies typical contralesional field, they have trouble inhibiting a re- of endogenously generated saccades. flexive saccade toward it; and (2) when it appears in the This account is derived from observations made by ipsilesional field, they are impaired in generating an Sprague (1966) and by Sherman (1974). Sprague re- endogenous saccade toward the contralesional field. ported that occipitotemporal cortical ablation in cats pro- The current study examined only patients with chronic duced stable hemianopia. However, visually guided lesions of the dorsolateral prefrontal cortex, more than behavior was restored by ablation of the dorsal midbrain 2 years after the ictus. This differs from most previous

Henikeral. 407

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/jocn.1994.6.4.400 by guest on 27 September 2021 studies in both animals and humans which chietly ex- The patients were subdivided into two subgroups amined the effects of acute lesions were examined (Guit- based upon whether the lesion involved regions of thc ton et al., 1985; Pierrot-Deseillgny et ai., 1987; Pierrot- superior dorsolateral prefrontal cortex thought t() in- Deseillgny, Rivaud, Gaymard, & Agid, 1991; Schiller et clude the frontal eye fields (FEF). In nine patients the al., 1980, 1987). For some time after frontal cortex lesions FEF was affected (FEF group); in seven patients the FEF there are diaschesis effects producing dysfunction in sub- was spared (NFEF group). The two subgroups of pxieiits cortical nucleii including thalamus, basal ganglia, and did not differ significantly in age (t = 1.075,p = 0.3) : superior colliculus (Deuel & Collins, 1984). It is not Mean age of the FEF patients was 65 years (SD = 4.9); surprising, therefore, that some of the observations in mean age of the NFEF patients was 59 years (SD = 15.81. the current study differ from previous findings. For ex- The average lesion volume was 63 cni' for the FEI; group ample, none of the previous studies has reported a con- and 46 cm' for the NFEF group. The average lesion sistently shorter latency for reflexive saccades toward volume did not differ significantly between the two pa- Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/6/4/400/1755171/jocn.1994.6.4.400.pdf by guest on 18 May 2021 contralesional targets. tient subgroups [F(1,14) = 1.66,p > 0.21. Contemporary methods of neuroimaging now make it All the patients were selected on the basis of having ;I possible to study, in humans, the role of quite specific single lesion restricted to the dorsolateral prefrontal cor- cortical regions. Our approach has been to show that a tex (Table 1). All the lesions were caused by strokes lesion of the superior dorsolateral prefrontal cortex pre- except in two patients (one in each of the subgroups). sumed to include the FEF is necessary to produce the In all, the lesion was chronic (at least 2 years after onset kinds of effects described. This region includes areas 6 of illness). All patients were functioning independently and 8 (Damasio & Damasio, 1989), and encompasses the and were experienced participants in behavioral studies. region that has been linked to saccades on the basis of None had any history of drug or alcohol dependency or electric stimulation studies (Penfield & Ja5per, 1954; Pen- of mental illness. None had any clinical sign of visual field & Rasmussen, 1950); activation studies with positron field defect, extinction, or neglect. In all, routine bedside emission tomography (Fox, Fox, Raichle, & Burde, 1985), examination revealed no obvious defect of saccade spced and other patient lesion studies (Pierrot-Deseillgny et al., or amplitude, and movements and opti- 1991). Since the control patients with lesions of dorso- cokinetic nystagnius appeared to be normal. lateral prefrontal cortex did not show these effects, it is possible to be reasonably confident that the FEF lesion Lesion Location is the necessary condition for producing these effects. Since the patients with the FEF lesion did, however, also Lesion location was verified by CT scan or MKI in dl have involvement of other parts of the dorsolateral pre- patients. Lesion area were reconstructed onto axial tem- frontal cortex, it cannot be concluded that the FEF lesion plates drawn from an atlas (De Armond, Fusco, & Dewey, is also sufficient to cause the oculomotor behavior that 1976). Individual reconstructions were then computed was observed. from the axial sections (Frey, Woods, Knight, & Scabini, In summary, the superior dorsolateral prefrontal cor- 1987). Figure 1 shows composite diagrams of the patients tex was shown to play an important role in regulating separated into two groups: those in whom the lesion eye movements, and that this function is not directly involved the superior dorsolateral prefrontal cortex, and related to covert visual attention. Lesions of the FEF those in whom the superior part of the dorsolateral produce a deficit in generating endogenous saccades. prefrontal cortex was spared. The division of patients They also appear to result in disinhibition of ipsilateral into these two groups was done with the intent of coni- midbrain visuomotor pathways. This disinhibiton results paring those in whom the lesion likely involved the in a facilitation of reflexive saccades to the contralesional frontal eye fields (FEF group) and those in whom the. field and, presumably due to inhibition of the opposite superior dorsolateral prefrontal cortex was not involved colliculus, also to a loss of reflexive saccades to signals and the frontal eye fields were believed likely to he in the ipsilesional field. spared (NFEF Group). The division of patients into the two groups was based on whether the lesion in prefrontal cortex extended SUBJECTS AND METHODS above the highest cut in which the lateral ventricles were visible. Since patients with ventricular enlargement were Subjects not included, this division was reasonably uniform across Thirteen neurologically normal control subjects and 16 patients in terms of cortical areas affected. This criteria patients with lesions of dorsolateral prefrontal cortex for group assignment was selected to provide consistent volunteered to participate and were tested after obtaining anatomic landmarks to identify regions known to gen- informed consent. One patient was 26 years old, one erate eye movements when electrically stimulated (Pen- normal control 23. Otherwise, patients ranged in age field & Jasper, 1954; Penfield & Rasmussen, 1950), from 61 to 74 (mean = 63) and normal controls ranged Brodman's area 8 (Damasio & Damasio, 1989), the region in age from 45 to 83 (mean = 65). showing increased metabolism during saccades with pos-

408 Journul of Cognitive Neuroscience

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/jocn.1994.6.4.400 by guest on 27 September 2021 itron emission tomography (Fox et al., 1985), and the screen at eccentricities of 10”. These provided the sub- region identified as FEF in the human lesion study of jects with target locations for their saccades. Pierrot-Deseillgny et al. (1991). In the study of Pierrot- In addition to examining the effect of prefrontal cortex Deseillgny et al. (1991) the center of the lesion in their lesions on saccade execution, another purpose of the FEF patient group was on the lateral frontal convexity 24 study was to determine whether these lesions affected mm below the upper limit of the cerebral hemispheres, the ability to preprogram eye movements. This was ac- which was anterior to the precentral at the level complished by using precues that could provide advance of the second frontal . These sulcal landmarks were information about the direction of the forthcoming sac- identified in each of our patients, and this area was cade. Before the target “go” signal on each trial, one of involved in all the patients included in the FEF lesion two types of precues was presented in the center of the group. display (Fig. 2): an informative precue (a small 1” arrow- head pointing to left or to right) that instructed the Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/6/4/400/1755171/jocn.1994.6.4.400.pdf by guest on 18 May 2021 subject to get ready to make a saccade in that direction, Procedure or a neutral precue (a double-headed arrow diamond Subjects sat in a quiet, dimly lit room facing a video shape) that required the subject to wait for the target to display on which visual precues (“get ready”) and targets learn which way the eyes had to move. Half of the pre- (“go”)were presented. There were two types of tasks. In cues were informative and half were neutral. Preparation one (saccade experiments) subjects made saccades to a effects were inferred from the decreased response laten- peripheral marker in response to the target signals that cies following informative cues in comparison to laten- instructed them to move their eyes either left or right. cies following neutral cues. The decrease in saccade The other was a detection task (key press experiments) latency in the informative precue condition compared to in which the target signal instructed them which of two the neutral cue condition provided an index of how keys to press. The main dependent variable was reaction effective the subject was in preprogramming the saccade. time (KT), the time from the appearance of the target to The intertrial interval was 1500 msec during which the the subject’s response. For each type of task, two different subject maintained fixation at the center of the screen. types of target go signals were tested in separate trial Each trial began with the appearance of the precue for blocks: either a peripheral signal, an asterisk appearing 300 msec. After a variable interval the target “go” signal 10” from fixation; or a large arrowhead in the center of appeared and remained on the screen until the subject the display pointing to left or to right (Fig. 2). Thus, four responded. The intervals between the onset of the cue conditions were generated by crossing response mode and the onset of the target were 500,750,1000, and 1250 (saccade or key press) and target type (peripheral or msec. central 1. For both peripheral and central target tasks, the cue In the eye movement experiments a peripheral target, in the center of the screen was extinguished before the presented 10” to right or left, was used to elicit reflexive, target go signal, and an interval ranging from 200 to 950 exogenously triggered saccades. This will be referred to msec passed while the subject waited for the go signal. as the peripherally summoned saccade condition. To Since there was no fixation point during this variahle measure voluntary, endogenously generated saccades, an (and on some trials quite long) foreperiod it was im- arrowhead in the center of the display, pointing either portant to ensure that subjects did not move their eyes right or left, was presented to instruct the subject which during this foreperiod, and to exclude from analysis any way to move the eyes. This will be referred to as the trial on which fixation was broken before the go signal. centrally controlled saccade condition. The key press Subjects were therefore supplied with feedback for any experiments were done after the saccade experiments. eye movement that occurred during this foreperiod and The visual display was identical to that used in the sac- the trial was aborted: if any eye movement was made cade experiments, only the required response was dif- between cue onset and 100 msec after the target’s onset ferent. The purpose of the key press control experiment the message “TOO SOON” appeared on the screen and was to determine whether any effects of FEF lesions stayed there for 2 sec. Since it can be quite difficult to found in the saccade experiments were specific for oc- maintain such strict fixation over this range of time with- ulomotor responses, and to determine whether the le- out blinking or drifting, this strict criteria resulted in sions had separate effects on covert visual attention. exclusion of 18% of trials. Since subjects were given a The peripheral asterisk target subtended 1.8” visual “TOO SOON” error signal on these trials, this procedure angle, was 16 cam2 in luminance, and filled one of two also tended to make the subjects cautious in their re- empty boxes that were displayed continuously on the sponses. As a result, the saccade latencies in this study left and right. The central arrowhead subtended 2.5” of tended to be rather longer than in other saccade studies. visual angle. In the voluntary saccade condition of Ex- However, given the goal of manipulating saccade prep- periment 1, small paper squares (the same size as the aration as a variable in the current study, this design was empty squares used in the peripherally summoned sac- adopted intentionally, in order to provide information cade task) were taped on the left and right sides of the on saccade preparation.

Henik et al. 409

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/jocn.1994.6.4.400 by guest on 27 September 2021 Eye position was monitored throughout on a slave Apparatus scope by the experimenter. Blinks (in which the globes An Apple 2e microcomputer controlled the display and rotate upward) could be easily distinguished by the ex- recorded response directions and RTs. In the keypress perimenter from left or right saccades. However, because task, the subject placed the index finger and middle of crosstalk generated by vertical globe deviations,.the finger of one hand on adjacent plastic keys mounted on computer could not distinguish these reliably. During a keyboard. Light pressure on either key activated a mi- practice the experimenter gave subjects feedback on er- croswitch that recorded RT. In the saccade task, eye po- rors-that is saccades made to left or right before the sition was recorded with an Eye-Trac 210 infrared scleral target appeared or saccades made in the opposite direc- reflectance recording device mounted on spectacle tion of that indicated by the target go signal. Data were frames. The experimenter monitored an eye position not collected until the task was done correctly. During cursor on a separate slave scope, which also replicated Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/6/4/400/1755171/jocn.1994.6.4.400.pdf by guest on 18 May 2021 experimental sessions errors were very rare (mean < the display viewed by the subject. At the beginning of 2%) and did not exceed 4% in any subject. each session the Eye-Trac was calibrated by centering For each of the four tasks, subjects were given a prac- the cursor on a central fixation and the gain of the tice block of 48 trials; then trials were run in blocks of "+" device was then adjusted such that when the subject 80 trials until approximately 240 trials (in the saccade made a saccade to a target on left or right, the cursor experiment, uncontaminated by eye blinks) were col- moved to the position of the target on the slave scope. lected. For normal control subjects, all data were ob- The Eye-Trac was interfaced with a microcomputer tained in a single session. For the patients, the saccade through a device that produced a velocity (approximately and key press experiments were run in separate sessions. 65 degreeshecond) (first derivative) transformation of The saccade experiment was always run first, and the the Eye-Trac signal. When a threshold velocity was ex- central arrow target task was run first for both saccade ceeded, a response signal was sent to the computer. and key press experiments. Stimuli were white on a black background and were displayed on a Magnavox RGB 40 monitor.

Data Analyses Acknowledgment For every subject in each target type condition, median RTs were calculated separately for the correct left and This work wa.. supported by US. PHS Grant MH 41544. right responses following informative and neutral cues. Reprint requests should be sent to Dr. Avishai Henik, Depan- Responses with a latency of less than 100 msec or more ment of Behavioral Sciences, Ben Gurion University of the than 1500 nisec were excluded from analysis. Median Negev, Beer Sheva, Israel. response latencies for correct responses were entered into analyses of variance. The main purpose of the study was to determine whether unilateral lesions of the su- REFERENCES perior dorsolateral prefrontal cortex (including FEF) had Bruce, C.J., Goldberg, M. E. (1985). Primate frontal eye fields. ;I lateralized effect on saccade latencies, and to determine I. Single neurons discharging before saccades. Jounzal of whether any lateralized effects were specific to saccades Neuropbysiology, 53, 603-635. Damasio, H., & Damasio, A. R. (1989). Lesion ana/ysis in neu- and not to manual detection responses; and if so, to ropsycbolo~~New York: Oxford University Press. determine whether these lesions had differential effects De Armond, S. J., Fusco, M. M., & Dewey, M. M. (1976). Struc- of reflexive and voluntary saccades. Therefore, the pri- ture of the burrun brain: A photographic atlac New York: inar): analysis was a repeated measures ANOVA with &e Oxford University Press. t>emeen-group FactcIr (FEF lesion versus no FEF lesion), Dew S-Y., Goldberg, M. E., Segraves, M. A., Ungerleider, L. G., & Mishkin, M. (1986). The effect of unilateral ablation and three within-subject factors: Task (central or periph- of the frontal eye fieldson saccadic performance in the era1 go signal), Cue Validity (informative or neutral), and monkey, In E, L. Keller & D, s, zee (Eds,),Adaptjliepro. Field (contralesional or ipsilesional). This ANOVA was ceses in visudl and oculomotor .system (pp. 201-208). Ox- done selm-ately for both the saccade and key press ex- ford: ['ergamon Press. per~ments,~1,~~~ [he m0vArevealed a statistically reli- Deuel, R. K., & Collins, R. C. (1984). The functional anatomy of frontal lobe neglect in monkeys: Behavioral and 2-deox- able interaction, appropriate post-hoc tests were done to yglucose studies, of NeuroloM! Is, 521-529, cletermine the source of the interaction. In addition, to F~~,p, T,, F()~, J, M,, Raichle, M, E,, & R, M, ( 1985). determine whether other anatomic variance between The role of the in the generation of volun- groups may \lave contributed to the effects found (e,g,, tary saccades: a positron emission tomographic study. Jour- leti versus right hemisphere lesion, or involvement of nal of Neurophysiology, 54, 348-369. Frey, R., Woods, D. L., Knight, R. T., C(r Scabini, D. (1987). subcortical structures such as striatum or internal cap- Definingv functional cortical area.^ with averaged CT sca,ls, sule ), other ANOVA'i were conducted to explore these socieg, of Neuroscience AbsWacfi, 1.3, 1266.71 variables as possibly relevant between group factors. Goldberg, M. E., & Segraves, M. A. (1989). The visual and

41 0 Journul of Cognitiie Neuroscience Volume 6, Niinihc.r 4

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/jocn.1994.6.4.400 by guest on 27 September 2021 frontal cortices. In R. H. Wurtz & M. E. Goldberg (Eds.), butions to reflexive visual orienting in normal humans: A The neurobiology of saccadic eye rnoi1ement.s (pp. 283- temporal hemifield advantage. Journal of Cognititle Neuro- 313 ). Amsterdam: Elsevier Science Publishers BV. science, 3, 323-329. Goldberg, M. E., & Wurtz, R. H. (1972). Activity of superior Rafal, R., Smith, J., Krantz, J., Cohen, A,, & Brennan, C. (1990). colliculus neurons in behaving monkeys. Journal of Neu- Extrageniculate vision in hemianopic humans: Saccade in- ropkysiology, 35, 42-91. hibition by signals in the blind field. Science, 250, 118-121. Guitton. D., Buchtel, H. A., & Douglas, R. M. (1985). Frontal Rafal, R. D., Posner, M. I., Friedman, J. H., Inhoff, A W., & lobe lesions in man cause difficulties in suppressing reflex- Bernstein, E. (1988). Orienting of visual attention in pro- ive glances and in generating goal directed saccadrs. Ex- gressive supranuclear palsy. Brain, I1I, 267-280. peri?~zcrztalBrain Research, 58, 4 5 5-472. Schiller, P. H., Sandell, J. H., & Maunsell, J. H. (1987). The Holmes, G. (1938). The cerebral integration of the ocular effect of frontal eye field and superior colliculus lesions on movements. British Medical Journal, 2, 107-1 12. saccadic latencies in the rhesus monkey. Journal of Neuro- Luria, A. R., Karpov, B., & Yarbuss, A L. (1966). Disturbances pbysiolog); 57, 1033-1049. of active visual perception with lesions of frontal lobes. Schiller, P. H., True, S. D., & Conway, J. L. (1980). Deficits in Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/6/4/400/1755171/jocn.1994.6.4.400.pdf by guest on 18 May 2021 Corttx, 2, 202-2 12. eye movements following frontal eye-field and superior Nagel, L. S., Buchtel, H. A, & Welch, K. M. (1990). Cerebral colliculus ablations. Journal of Neuropbysioloa, 44, 1175- control of directed visual attention and orienting saccddes. 1189. Brain, 113, 237-286. Sherman, S. M. (1974). Visual fields of cats with cortical and Penfield. W., &Jasper, H. H. (1954). Epilepg and theJunc- tectal lesions. Science, 185, 355-357. tiornil anatomy of the . Boston: Little, Brown. Sprague, J. M. (1966). Interaction of cortex and superior colli- Penfield, W., & Rasmussen, T. (1950). The cerebral cortex of culus in mediation of peripherally summoned behavior in man. A clinical study of localization of function. New the cat. Science, 153, 1544-1547. York: Macmillan. Wallace, S. F., Rosenquist, A. C., & Sprague, J. M. (1989). Re- Pierrot-Deseillgny,C., Rivaud, S., Gaymard, B., & Agid, Y. covery from cortical blindness mediated by destruction of (1991 ). Cortical control of reflexive visually-guided sac- nontectotectal fibers in the commisure of the superior col- cades. Brain, 114, 1473-1485. liculus in the cat. Journal of Cornparatitle Neurology, 284, Pierrot-Deseillgny, C., Rivaud, S.,Penet, C., & Rigolet, M-H. 429-450. ( 1987 ). Latencies of visually guided sdccades in unilateral Wallace, S. F., Rosenquist, A. C., & Sprague, J. M. (1990). Ibo- hemispheric cerebral lesions. Annals of Neurology, 21, tenic acid lesions of the lateral substantia nigra restore vis- 138-148. ual orientation behavior in the hemianopic cat. Journal of Rafal, R., Henik, A, & Smith, J. (1991). Extrageniculate contri- Cornparatitle Neurology, 296, 222-252.

Hen& etal. 411

Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/jocn.1994.6.4.400 by guest on 27 September 2021