DOI: 10.1212/01.wnl.0000261482.99569.fb 2007;68;1515-1523 Neurology

M. P. Alexander, D. T. Stuss, T. Picton, T. Shallice and S. Gillingham Regional frontal injuries cause distinct impairments in cognitive control

This information is current as of June 20, 2007

http://www.neurology.org/cgi/content/full/68/18/1515located on the World Wide Web at:

The online version of this article, along with updated information and services, is

Print ISSN: 0028-3878. Online ISSN: 1526-632X. published continuously since 1951. Copyright © 2007 by AAN Enterprises, Inc. All rights reserved. Neurology is the official journal of AAN Enterprises, Inc. A bi-monthly publication, it has been

at SWETS SUB SERVICE on June 20, 2007 www.neurology.orgDownloaded from

Regional frontal injuries cause distinctimpairments in cognitive control

M.P. Alexander, MDD.T. Stuss, PhDT. Picton, MD, PhDT. Shallice, PhDS. Gillingham

ABSTRACT Background: Lesions of the frontal lobes may impair the capacity of patients to controlotherwise intact cognitive operations in the face of ambiguous sensory input or conflicting possibleresponses. Objective: To address the question of whether focal lesions in different regions of thefrontal lobes produced specific impairments in cognitive control. Methods: We evaluated 42 pa-tients with chronic frontal lesions and 38 control subjects on a modified Stroop test that allowedmeasurement of reaction times and errors. Planned, stratified analyses permitted identification ofdiscrete frontal lesions that are critical for impaired performance. Results: Lesions of the left ventro-lateral region produced an increased number of incorrect responses to distractors. Lesions of a largeportion of the right superior medial region, including anterior cingulate, supplementary motor area(SMA), pre-SMA, and dorsolateral areas, caused a slow reaction time and a decreased number ofcorrect responses to targets. Conclusion: Lesions in two distinct frontal regions impair cognitive con-trol for a Stroop task, and the mechanisms of impairment are specific to the region of injury. This issupport for a general proposal that the supervisory system is constructed of distinct subsystems.NEUROLOGY 2007;68:1515–1523

Patients with damage to the frontal lobes often have difficulty controlling their behavior.This may take many forms: impulsiveness, poor appreciation of risk, blunted motivation,intrusiveness, inadequate planning, and others. This loss of control may be a result of im-paired emotional computations or an inability to determine complex reward consequences ofbehaviors,1 but one straightforward difficulty common after frontal lesions is defective con-trol of behavior in the face of choice, complexity, or ambiguity.2

Cognitive control can be characterized as a capacity-limited system that “guides voluntary,complex actions,”3 and it has been most productively studied with a variety of experimentalstrategies using imaging, predominantly fMRI, in normal subjects. Cognitive control is re-quired when “task demands change unpredictably,”4 when a stimulus is ambiguous and po-tentially conflicting responses might be generated, or when a high probability response (onethat has been used frequently before) must be blocked.2 Cognitive control comprises severalprocesses that are required for rapid response to a single stimulus, to repeating stimuli, tochanging stimuli, to simultaneous stimuli, to stimuli that might occur in a quiet backgroundor a complex, distracting one, to a unique stimulus, or to a stimulus that shares propertieswith irrelevant stimuli. These processes include preparation to respond, sustaining responsecapability while awaiting a stimulus (vigilance) or during a task (concentration), setting spe-cific stimulus–response (S-R) guides, whether the stimulus is internal (drives) or external,monitoring the occurrence of the stimulus, the timing of expected responses and the proper-ties and accuracy of the response, suppressing responses to competing stimuli, inhibitingpracticed or primed potent responses, and switching to entirely new S-R challenges whencontext or stimuli demand. These processes are partly the fundamental phenomena of con-trolled attention but largely the mechanisms of controlling attention to allow control ofresponses, and they are essential to accomplish goal-directed, context-dependent behaviors.5

From Behavioral Neurology (M.P.A.), Beth Israel Deaconess Medical Center, and Harvard Medical School (M.P.A.), Boston, MA; RotmanResearch Institute of Baycrest (M.P.A., D.T.S., T.P., T.S., S.G.) and University of Toronto (D.T.S., T.P.), Ontario, Canada; SISSA (T.S.), Trieste,Italy; and Institute of Cognitive Neuroscience (T.S.), London, UK.

Disclosure: The authors report no conflicts of interest.

Address correspondence andreprint requests to Dr. M.P.Alexander, Beth Israel DeaconessMedical Center, BNU, KS 253,330 Brookline Ave., Boston, MA02215malexand@bidmc.harvard.edu

See also page 1450

Copyright © 2007 by AAN Enterprises, Inc. 1515 at SWETS SUB SERVICE on June 20, 2007 www.neurology.orgDownloaded from

Neuroimaging experiments have been em-ployed in normal subjects to investigate theroles of various frontal regions, particularlythe anterior cingulate and dorsolateral (DL)regions, in performance of conflict-loadedtasks. Anterior cingulate gyrus (ACG) acti-vation rises in response to the cognitive effortof more controlled performances,3,6,7 and ac-tivation is greater when the conflict is in re-sponse control than when conflict isenmeshed in the stimulus ambiguity.8,9 Inresponse to ACG, activation rises in DL re-gions associated with setting S-R contingen-cies for a given task,3,10 particularly for a lessautomatic response.3 This DL activation isprimarily left sided, areas 93 and 44.10 Sometasks with the higher working memory de-mands activate right or bilateral areas 93,8 or46/9.7 These studies suggest a template forexpectations about the effects of brain le-sions, but the translation from fMRI findingsto clinical findings is not always transparent.Confirmatory convergence with findings inlesion studies should clarify the significanceof the model proposed through imaging. Inaddition, it is the only avenue for under-standing the relevant consequences for a clin-ical population.

In this report we describe the effects of fo-cal prefrontal cortex lesions on a task withcomplex S-R rules, requiring intense mainte-nance of attention across time and a chal-lenging response suppression. Based on ourprevious studies11-13 and on the models of in-teractions between distinct frontal regionssuggested by fMRI studies, we had two pre-dictions: 1) Setting the S-R rules should besensitive to left ventrolateral (VL) lesions.Action schema theories suggest that sometasks require two S-R contingencies within asingle action schema: one to set the more fre-quent prepotent responses and another to setthe suppression of that response. If that is themechanism of managing suppression, sup-pressing prepotent responses should also besensitive to left VL lesions. 2) Initiating andsustaining plans for and execution of re-sponses should be sensitive to superior me-dial lesions. “Activation” is a term that hascome to have multiple meanings—physio-

logic, hemodynamic, psychological, and be-havioral—but we use it here only to specifythe drive to prepare, initiate, and maintainattention to task and response.

METHODS Subjects. We tested 42 patients with frontal le-sions and 38 nonpatient control subjects (CTLs), matched asclosely as possible to the patients for sex, age, and education.All patients were at least 2 months (all but one past 3.6 months)post onset (mean � 22 months; range � 2 to109 months). Otherinclusion criteria were absence of aphasia, visual neglect, andany other significant neurologic or psychiatric disorders; IQwithin the normal range (patients [mean � 108, SD � 8.9];CTLs [mean � 112, SD � 6.7]; all scores �90). All patients andCTLs gave informed consent in accordance with the Institu-tional Review Board requirements of the University of Torontoand Baycrest Centre for Geriatric Care.

All subjects responded with their dominant hand (40/42 pa-tients used the right). None of the patients had weakness orimpaired motor control. All subjects had normal color vision.Measures of neglect (line bisection and double simultaneousstimulation) were normal in the patients. To provide a measureof general intellectual ability, the National Adult ReadingTest–Revised was administered. Other neuropsychological testmeasures included Digit Span forward and backward, TokenTest of language comprehension,14 Boston Naming Test,15

Judgment of Line Orientation,16 and Beck DepressionInventory.17

The etiology of the lesions was an acquired acute disorder,including infarction, hemorrhage (including ruptured aneu-rysms with secondary infarction or intracerebral hemorrhage),trauma, and tumors. Patients with trauma had well-defined fo-cal contusions and no or only brief loss of consciousness. Alltrauma patients were more than 5 months post onset (mostmore than 24 months) allowing for complete recovery fromacute-phase factors such as edema or hemorrhage. Patientswith tumors all had resection of meningiomas or low-grade gli-omas, were tested more than 6 months post surgery, and had noevidence of recurrence. None had brain radiation. Whereas itwould be preferable to create groups differing only in lesion siteand not in etiology, it is impossible to gather patients with le-sions in all frontal regions because different etiologies have pre-dispositions for different regions5 and the location of lesions ismore critical than the etiology in determining cognitive defi-cits.18 Etiology has had no effect on results in any of our previ-ous reports with the same patient population.11,12,19

The frontal patients were divided into the following ana-tomic classifications based on our previous research12: left lat-eral frontal (LL, n � 11); right lateral frontal (RL, n � 6);inferior medial (IM, n � 15); superior medial (subjects in thisgroup may have had extension into inferior medial; SM, n �

10). Patients in the SM and IM groups could have lesions ineither hemisphere or both. The lateral groups could include DLsubcortical lesions involving deep frontal white matter and dor-sal caudate. There were six patients (four RL, two SM) inwhom pathology extended to nonfrontal structures, the non-frontal extension being less than 10% of the entire lesion (range3.3 to 8.1%, mean � 6.1%. In one patient in the IM group, thenonfrontal extension was 35%). All lesions were localized witha standard template20,21 using postacute CT in 61% and lateMRI, relying predominantly on T1-weighted images, in 39%.Lesion size was quantified by superimposing the lesion contourfor each axial slice on a constant pixel diagram and counting

1516 Neurology 68 May 1, 2007 at SWETS SUB SERVICE on June 20, 2007 www.neurology.orgDownloaded from

the number of pixels within the lesion area. The percentage oftotal brain area damaged was obtained by dividing the lesioncount by the total pixel count for all axial slices.

The patients and CTLs reported in this study are identicalto the groups in an earlier article except for omission of onepatient (no. 2,151) and one CTL for whom data on this testwere incomplete. Detailed summaries of demographic data andneuropsychological test results and the lesion maps of patientscan be found in the earlier article.19

Procedure. The test is a variant of the Stroop interferenceparadigm that uses very simple stimuli (designed by Jeff Toth,PhD) A series of letters was presented in the center of a 14- or15-in color monitor, controlled by a personal computer (486 orPentium). The subject sat approximately 16 in from the moni-tor. Programming was done onMEL2. The letters occurred at arate of once every 3 to 4 seconds. The letters are either red orblue. The subject was asked to press Button 1 when either a redX or a blue O—the targets—appeared and press Button 2 forany other stimulus. Twenty-five percent of the stimuli were oneof the targets; 25% were distractors (either a red O or blue X),and 50% were others (letters other than X and O in blue or redand also excluding the potentially visually confusing letters C,D, G, Q, K, and Y). There was an initial practice session withitem-by-item feedback about accuracy and then another briefpractice without feedback at the presentation rate of the actualtest. There were then two blocks of recordings, each containing102 stimuli; the first two are discarded.

Statistical analyses. The major dependent variables were re-action time (RT) and errors. A sequence of planned, repeated-measure analyses of variance (ANOVAs) evaluated theexperimental effects.12 Four separate ANOVAs compared eachof the patient groups with CTL to determine if any patientgroup was behaving abnormally. For any group with signifi-cant difference from CTLs, a subsequent ANOVA comparedthat patient group with all the other patients combined to de-termine if the abnormality was specific to that group. All anal-

yses of the reaction times were performed with and without acovariate of the simple RT measured in the same testing ses-sion. The covariate did not affect the results, and we report thestatistics for the analyses with the covariate. We used a criterionfor significance of p� 0.05. Errors were characterized as false pos-itives, that is, responding to either other or distractor stimuli astargets with other and distractor analyzed separately, or as falsenegatives, that is, responding to targets as nontargets.

If the planned ANOVAs indicated impairment in one ormore of the groups, a detailed lesion assessment was per-formed. This methodology is described in detail in earlier pub-lications.11,12 Working from standard templates20,21 for frontallocalizations in each patient, every Petrides–Pandya architec-tonic area21 was coded as damaged (greater than 25% of itsarea) or not. Only areas that were in a group region identifiedas impaired compared with the CTL group and that were in-volved in three or more subjects were included. For each areathus identified, we compared the group of patients who hadlesions in that area with all patients who had no damage to thatarea using t tests for both RT and errors. This approach mapsthe full range of the behavioral variable (the latency of the RTor the number of errors) on to different areas of the brain andidentifies “hot spots,” architectonic areas most related to theabnormal findings. It is susceptible to type 1 error because ofthe large number of comparisons, so the results thereforeshould be taken as suggesting a more precise localization ratherthan indicating it definitively. Claims for specific areas of local-ization will demand demonstration of consistency of localiza-tions across studies and methods.

RESULTS Demographic variables. There were nomajor demographic differences between groups andno significant correlations between any neuropsy-chological measure and the dependent variables inthe experimental task.

RT. There was a main effect of stimulus types in thedirect comparisons of RT in all subjects (F[2, 138] �30.31,p � 0.000); post hoc analysis revealed that RTfor distractors was significantly longer than for“others.” Every group showed expected the profileof RT interference: distractors greater than targetsand both significantly greater than others. Our vari-ation of the Stroop test produced the expected dis-tribution of performance and can be consideredgenerally informative about the Stroop effect. Inpairwise comparisons with CTLs, only the SMgroup was slower than CTLs (F[2, 148] � 8.839,p � 0.000). A separate analysis of SM against theother frontal groups combined also demonstratedslowing (F[2, 80] � 5.374,p � 0.006). Post hoc anal-ysis demonstrated significant interference effects forthe SM group compared with CTLs for all stimuliand compared with the other patient groups com-bined for the distractor stimuli. The results are sum-marized in figure 1.

The hotspot lesion analysis identified a largearea of the right superior and medial frontal lobethat was disproportionately injured in patients withprolonged RT (all three stimulus types, correct re-

Figure 1 Mean reaction times

Mean reaction times withstandard deviations for thethree stimuli—targets,distractors, and others—forthe five experimental groups.Analysis of differences is inthe text.

Neurology 68 May 1, 2007 1517 at SWETS SUB SERVICE on June 20, 2007 www.neurology.orgDownloaded from

sponses only). This area included ACG (areas 24and 32) and areas 9/46, 9, and 6A as well as the leftmedial area 9 (figure 2).

Errors. Both the left DL group (F[2, 148] � 3.449,p � 0.034) and the SM group (F[2, 80] � 6.194, p �0.003) had more errors than the CTL group. Boththe left DL (F(2,80) � 4.655, p � 0.012) and SM(F[2, 80] � 3.061, p � 0.052) groups also made moreerrors when compared with the other frontal pa-tients. Post hoc analyses demonstrated that the leftDL group had significantly more errors than CTLsonly on the distractor items and that the SM grouphad significantly more errors than CTLs on the tar-get items (figure 3). One-way ANOVAs were per-formed comparing each group with CTLs and withthe combined other groups on each stimulus type.Only left DL had more errors on distractors (falsepositives) than CTLs (F[1, 74] � 9.896, p � 0.002)and than other patients (F[1, 40] � 7.146, p �0.011). Only SM had more errors on target (falsenegatives) than CTLs (F[1, 74] � 10.416, p � 0.002)and than other patients (F[1, 40] � 4.04, p � 0.051).

The absolute number of errors was not high, butCTLs and the RL and IM patients made so few er-rors that even a low rate of error in the left DL andSM groups was significant. False-positive errors oc-curred on a mean of 5.4% of the 50 distractor stim-

uli in the left DL group. This is four times the rate incontrols and three times the rate in the next highestpatient group (SM, 1.8%). False-negative errors ontargets occurred on 9% of the 50 targets in the SMgroup; this is three times the control rate (3%) andalmost twice as frequent as in the next highest pa-tient group (IM, 5.5%). The only difference be-tween the two blocks in frequency of errors was anincrease in false-positive endorsements of otherstimuli as targets by the SM group in the secondblock (p � 0.008 vs CTLs and p � 0.001 vs otherpatients). The total number of these errors wassmall, but there were essentially no errors of thistype made by any other group in either block, andthe SM group’s error rate increased fourfold fromthe first to the second block.

The parametric hotspot lesion analysis identifiedseveral contiguous regions of lesion in the left lateralfrontal lobe that correlated with false-positive er-rors on distractor items: Petrides and Pandya desig-nations of Brodmann areas 9/46v, 45, 47/12, and 6A(figure 4).

Lesions in one confluent region of right medialfrontal lesions correlated with errors on targets.This region included anterior and dorsal cingulate(areas 32 and 24) and pre-supplementary motor area(SMA) (area 9).

Figure 2 Cortical regions associated with prolonged reaction timesDark shading at a level of p �

0.05; light shading at a levelof p � 0.05 to 069.

1518 Neurology 68 May 1, 2007 at SWETS SUB SERVICE on June 20, 2007 www.neurology.orgDownloaded from

The effect of responses prior to an error and theinteraction of errors and RT can both potentiallyinform about response monitoring, but with only200 stimuli and relatively low error rates, the vari-ance was too great for analysis.

DISCUSSION This paradigm contains the same es-sential demands as the classic Stroop task with theadded information of measured RT and the varia-tion that task demands can change from item toitem unlike the classic Stroop, which is presented ina blocked manner. This test revealed the expectedoverall pattern of RTs: RTs for the “other” stimuliwere significantly faster than for both the distractorand the target stimuli that require active discrimina-tion, and rejecting distractors requires the longestconsideration. That the stimuli are presented un-blocked requires that the subject establish distinctanticipatory contingent response criteria; otherwiseeach stimulus requires analysis of features. Thepresence of distractors that share features with thetargets requires suppression of one response and im-plementation of the correct response.

The SM group was significantly slower thanCTLs on all stimuli and slower than other patientson distractors, even when controlled for underlyingresponse speed on a simple RT. This demandingversion of the Stroop test was affected by lesions ina very large region of right medial frontal lobe in-cluding ACG, SMA, and pre-SMA and of right DLfrontal lobe. SM patients also made a large numberof errors on this unblocked task. This error pattern

and the prolonged RT suggest slowing or completefailure of the process of initiating behavior.

The left DL group made significantly more er-rors on the distractor items. We have demonstratedthis difficulty with response bias and excessive false-positive responses in other, much different,tasks.11,13 The error rate on the current task was nothigh, but in any unfolding task in real time, evenoccasional failures to control S-R parameters couldripple through a long complicated activity: drivingin traffic, telling a story, planning a trip, performinga card sorting task, etc.

There have been relatively few prior studies ofthe effects of frontal lesions on cognitive control onStroop or Stroop-like tasks. Some of these studieshad limitations induced by the population selectedor available for study: patients with poorly localizedtumors22 or patients who had undergone partialfrontal lobectomies for seizure control in whom thepre-SMA and superior medial regions are uniformlyinvolved but the VL regions almost uniformly notinvolved.23 This limited literature supports our find-ings and hypotheses: left lateral lesions affect settingS-R contingencies and SM lesions affect initiation ofresponse. Left frontal lesions affect “the efficiencyof S-R associations under circumstances in whichresponse specification is not trivial and subject tocontrol errors.”23

Our group evaluated 37 patients with chronicfrontal lesions (mostly infarcts with no overlap withthe patient group in the current report) grouped asin the current study on the standard Stroop test.24

Patients with lesions in the left VL region (areas 44and 45) uniquely made errors on color naming (6%errors, largely commission). The SM group had sig-nificantly more errors (20%, both omission andcommission) and overall slower RT on the incon-gruent task, and there was a strong association oferrors with either left- or right-sided lesions inpre-SMA (area 6) and perhaps in dorsal ACG (24and 32).

There have been even fewer studies of patientswith lesions restricted to ACG as it is an uncommonisolated lesion site. In one report two patients withchronic lesions largely restricted to ACG were givena modified Stroop test.25 The patient with a middleright lesion (area 24) had a reduction of the normalinterference effect due to reduced preparedness torespond. The patient with the rostral left lesion (ar-eas 24/32) had increased errors on the incongruenttrials. Both patients had increased RTs and exces-sive errors on the trials with low frequency of incon-gruent stimuli, the condition that requires thegreater executive control of responses. In anotherreport, three patients with relatively isolated left-

Figure 3 Mean number of errors with standard deviations

Target errors are responsesto a target as though it is adistractor or other; therewere 50 targets and 50distractors. Distractor errorsare responses to a distractoras though it is a target (falsepositive). The right lateral andinferomedial groups did notdiffer from the controls.LDL � left dorsolateral;SM � superior medial; CTL �

control subjects.

Neurology 68 May 1, 2007 1519 at SWETS SUB SERVICE on June 20, 2007 www.neurology.orgDownloaded from

sided chronic ACG lesions26 were given an un-blocked Stroop test that had either high (80%) orlow (20%) frequencies of incongruent stimuli. Thepatients had slower RTs but did not differ fromCTLs on errors, interference effect, or sensitivity tocongruence frequency. The essentially negative re-sult from left-sided lesions suggests that the relativelateralization of deficits to right ACG lesions repre-sents a real asymmetry in ACG function.

Functional neuroimaging studies of cognitive con-trol. There have been many functional neuroimag-ing studies of Stroop-like tasks requiring responsesuppression. The studies vary substantially in ex-perimental parameters from the standard Strooptest to more complicated procedures. Some utilizelong S-R intervals or long interstimulus intervals toallow separate serial measures of activation afterthe stimulus or after the response. There are alsoinvestigations of tasks that make heavy demands ontask setting without being specifically Stroop-like.27-29 These studies are broadly consistent withour findings in a patient population and suggest thatthe left VL region and the SM region, includingACG, must play different roles in these responsesuppression tasks.

First, consider the left VL (areas 44 and 45) re-

gion: DL activation (following ACG activation)rises in association with setting S-R contingenciesfor a given task,3,10 particularly for less automaticresponses.3 This DL activation is primarily leftsided, areas 93 and 44.10 Serial scans after onset ofincongruent stimuli have shown the earliest re-sponse in left VL followed by symmetric increases inDL and caudate.30 The left VL region appears to setselective S-R criteria. Various complex manipula-tions of cuing to task switch and to criterion settingwithin task have also demonstrated that setting S-Rcriteria within a task activates left lateral re-gions,31,32 and the activation is greater for the lessautomatic (more demanding) task.

Translation between languages has Stroop inter-ference qualities that have been explored with PETscans on bilingual subjects as they either read ortranslated words in both blocked and unblocked(switching) trials.29 The switching vs blocked com-parison revealed activation in left VL frontal (area44), perhaps representing the necessity of continu-ous setting of S-R criteria. Apparently simple wordretrieval can be manipulated by requiring either anautomatic response to a stimulus or a more effortfulresponse to stimulus that requires suppression ofthe automatic response.27 Three separate tasks allproduced activation in left inferior frontal gyrus

Figure 4 Cortical regions associated with false-positive errors on distractorsDark shading at a level of p �

0.05; light shading at a levelof p � 0.05 to 0.063.

1520 Neurology 68 May 1, 2007 at SWETS SUB SERVICE on June 20, 2007 www.neurology.orgDownloaded from

(IFG) for the more effortful task. In a recent experi-ment designed to define the areas involved inswitching between competing outputs, activationspecifically related to switching was found in leftIFG but not in ACG. Although the task was createdto assess switching effects,28 the authors describe therole of the left IFG as “critical for our ability to biasactivation patterns in response to conflict that oc-curs when a strongly activated representation mustbe suppressed.” During switch responses, left tem-poral regions show reduced activation, perhaps in-dicating active suppression by left IFG. (In aseparate study of verb generation task with similarmanipulations, patients with left VL lesions mademore word retrieval errors when the less automaticresponse was required and more task errors com-pared with patients with left frontal lesions not in-volving the VL region.33 There was no overlapbetween the left VL group and the other two lesiongroups, and 90% of the variance in accuracy wasaccounted for by lesion volume in area 44.)

In a comprehensive review of all neuroimagingstudies on task switching and on the Stroop testpublished from 2000 through 2004 with sufficientdata for a meta-analysis, the merged localizationmaximas for both switching and Stroop were com-puted.4 For the Stroop studies, the major localiza-tions were in the left IFG (areas 44 and 6),apparently the critical region for updating task rep-resentations. The reinforcement of the lower proba-bility S-R contingency against a dominant onewithin a task (the Stroop paradigm) requires thisregion’s integrity.

Next, consider the superior medial/ACG regions.Three early PET studies of the standard Stroop taskall demonstrated activation in the ACG during theincongruent block.34-36 One study involved serialscans across the entire 15 minutes of testing anddemonstrated steadily increasing ACG activation.36

Another early PET study included an “anticipatory”scan performed after instruction for a task but be-fore the task was actually presented.37 ACG activa-tion was seen during the anticipatory period; that is,simple preparation for the cognitive task may besufficient to activate the ACG.

All fMRI versions of the Stroop task demon-strate increased activation in bilateral ACG com-pared with baselines,3,4,30 but many demonstratespecific parameters that influence activation. PeakACG activation may occur when a cue directsswitching to a task with increased S-R complexi-ty10,38 or when low rates of incongruence increaseinterference effect.32 In the work of Cohen et al.,ACG activation rises whether subsequent responsesto incongruent stimuli are correct or not.3,6,7 All of

these studies suggest that the ACG activation re-flects the effort required for the more controlledperformance. ACG activation is greater when theconflict is in response control than when conflict isenmeshed in the stimulus ambiguity.8,9

Translation produces greater activation in bilat-eral ACG and SMA (and striatum and cerebellum)than reading, perhaps representing the extra re-sources required for translation.29 In tasks requiringeither response generation or suppression with alow rate (17%) or a high rate (83%) of conflictedS-R associations, bilateral ACG activation occursduring the low frequency blocks.38 This result wasinterpreted as suppression of a more automatic re-sponse, but it could just as well be increased activa-tion for the more demanding task.

In the meta-analysis of neuroimaging studies ontask switching and on the Stroop, the second largestactivations (after left VL, summarized above) werein bilateral superior medial cortex (areas 32/6 and32/9).4 The authors did not emphasize the bilateralACG/pre-SMA region, but it is plausible that itserves as a general activator for any task that is de-manding, even for preparation for the task.

Regarding our predictions: 1) Lesions of the leftVL region (areas 44 and 45 and perhaps area 6)damage the capacity to establish the contingent rela-tionship of stimuli to responses, particularly underconditions that require continuous refreshing andsuppression of more salient responses. Setting andsuppressing responses in these Stroop paradigmsmay be two sides of the same action schema func-tion. This may be related to a more complex claimthat left lateral frontal structures are critical for de-fining a set of responses for all nonroutinized tasks,including setting the response criteria for not re-sponding—suppressing—a response.39 We havedemonstrated these same linked deficits—settingand suppressing—in the same population as the cur-rent study in a different task with similar demandsfor rapidly establishing S-R contingencies.11

These results in a population with focal lesionsconverge with the evidence from neuroimagingstudies but demonstrate, as only lesion studies can,that the left VL region is critical for these opera-tions, not just involved. Why the left and why ven-tral and not dorsal? That is unclear. Perhapshumans always draft a rough verbal code for behav-iors until they are practiced sufficiently to form atop-down action code, even when there is no ex-plicit verbalization of S-R contingencies. On theother hand, the task goes too quickly to rely on anexplicit verbal model, and left VL may be an impor-tant region for implementing all top-down actionschemas.39 Perhaps left VL connectivity through the

Neurology 68 May 1, 2007 1521 at SWETS SUB SERVICE on June 20, 2007 www.neurology.orgDownloaded from

“ventral stream” to temporal association cortex iscritical for establishing object or word discrimina-tive targets.

2) Lesions of the superior medial frontal lobesimpair initiating and sustaining the response state.The extent of lesion required depends on the de-mands of the task, and it is possible that the rightmedial frontal region is more critical. This can beviewed as anticipatory preparation for a demandingtask or as recruitment of greater attentional capac-ity to meet ambiguous or conflicted stimuli or tomanage conflicting possible responses. We havedemonstrated impairment in initiating and sustain-ing the response state in several other experimentsin this same group of patients.11,12,19

This conclusion is compatible with the knownconnectivity of the both the ACG40,41 and the SMA.42

Both receive substantial sensory input from all pos-terior association centers. Both have bilateral pro-jections to extensive prefrontal regions andstriatum. Both receive major dopaminergic inputsfrom the ventral tegmental area that appear criticalfor general properties of cognitive and motor initia-tion.43 The ACG responds in a relatively uniformphysiologic manner to all forms of errors, effort,orienting, and response demands, and its activationis followed, in turn, by phase-locked activity in awide variety of cortical sites.44 When fatigue devel-ops during a demanding task, there is a reduction ofactivity in ACG.45 The ACG has functional subdivi-sions with specific patterns of connectivity fromrostral to caudal41,46—affective, cognitive, motor,spatial, and mnestic—but much of the ACG roleappears to be a general impetus to act—think, plan,or move—and to maintain that arousal for the du-ration of a task.

We propose that these findings are robust evi-dence for fractionation of capacities in the frontallobes. The left VL region establishes S-R contingen-cies. The SM regions respond to internal and exter-nal drives to engage response systems. Thesupervisory system is the context-driven simulta-neous operation of these localized subsystems.

Received May 9, 2006. Accepted in final form February 7, 2007.

REFERENCES1. Bechara A, Tranel D, Damasio H. Characterization of the

decision-making deficit of patients with ventromedial pre-frontal cortex lesions. Brain 2000;123:2189–2202.

2. Burgess PW, Shallice T. Response suppression, initiationand strategy use following frontal lobe lesions. Neuropsy-chologia 1996;34:263–272.

3. MacDonald AW 3rd, Cohen JD, Stenger VA, Carter CS.Dissociating the role of the dorsolateral prefrontal andanterior cingulate cortex in cognitive control. Science2000;288:1835–1838.

4. Derrfuss J, Brass M, Neumann J, von Cramon DY. In-volvement of the inferior frontal junction in cognitivecontrol: meta-analyses of switching and Stroop studies.Hum Brain Map 2005;25:22–34.

5. Stuss DT, Shallice T, Alexander MP, Picton TW. A multi-disciplinary approach to anterior attentional functions.Ann NY Acad Sci 1995;769:191–211.

6. Botvinick M, Nystrom LE, Fissell K, Carter CS, CohenJD. Conflict monitoring versus selection-for-action in an-terior cingulate cortex. Nature 1999;402:179–181.

7. Carter CS, Braver TS, Barch DM, Botvinick MM, Noll D,Cohen JD. Anterior cingulate cortex, error detection, andthe online monitoring of performance. Science 1998;280:747–749.

8. Holroyd CB, Nieuwenhuis S, Yeung N, et al. Dorsal ante-rior cingulate cortex shows fMRI response to internal andexternal error signals. Nat Neurosci 2004;7:497–498.

9. van Veen V, Holroyd CB, Cohen JD, Stenger VA, CarterCS. Errors without conflict: implications for performancemonitoring theories of anterior cingulate cortex. BrainCogn 2004;56:267–276.

10. van Veen V, Cohen JD, Botvinick MM, Stenger VA,Carter CS. Anterior cingulate cortex, conflict monitoring,and levels of processing. Neuroimage 2001;14:1302–1308.

11. Alexander MP, Stuss DT, Shallice T, Picton TW, Gilling-ham S. Impaired concentration due to frontal lobe dam-age from two distinct lesion sites. Neurology 2005;65:572–579.

12. Stuss DT, Alexander MP, Shallice T, et al. Multiple fron-tal systems controlling response speed. Neuropsychologia2005;43:396–417.

13. Stuss DT, Binns MA, Murphy KJ, Alexander MP. Disso-ciations within the anterior attentional system: effects oftask complexity and irrelevant information on reactiontime speed and accuracy. Neuropsychology 2002;16:500–513.

14. De Renzi E, Vignolo LA. The token test: a sensitive test todetect receptive disturbances in aphasics. Brain 1962;85:665–678.

15. Kaplan EF, Goodglass H, Weintraub S. The Boston Nam-ing Test. 2nd ed. Philadelphia: Lea & Febiger, 1983.

16. Benton AL, Hamsher K, Varney NR, Spreen O. Contribu-tions to neuropsychological assessment. New York: Ox-ford University Press, 1983.

17. Beck AT, Steer RA. Beck Depression Inventory. San Anto-nio: Psychological Corp., 1993.

18. Elsass P, Hartelius H. Reaction time and brain disease:relations to location, etiology and progression of cerebraldysfunction. Acta Neurol Scand 1985;71:11–19.

19. Picton TW, Stuss DT, Shallice T, Alexander MP, Gilling-ham S. Keeping time: effects of focal frontal lesions. Neu-ropsychologia 2006;44:1195–1209.

20. Damasio H, Damasio AR, eds. Lesion analysis in neuro-psychology. New York: Oxford University Press, 1989.

21. Petrides M, Pandya DN. Comparative architectonic anal-ysis of the human and macaque frontal cortex. In: BollerF, Grafman J, eds. Handbook of neuropsychology. Am-sterdam: Elsevier, 1994:17–57.

22. Perret E. The left frontal lobe of man and the suppressionof habitual responses in verbal categorical behaviour.Neuropsychologia 1974;12:323–330.

23. Richer F, Decary A, Lapierre MF, Rouleau I, Bouvier G,Saint-Hilaire JM. Target detection deficits in frontal lo-bectomy. Brain Cogn 1993;21:203–211.

1522 Neurology 68 May 1, 2007 at SWETS SUB SERVICE on June 20, 2007 www.neurology.orgDownloaded from

24. Stuss DT, Floden D, Alexander MP, Levine B, Katz D.Stroop performance in focal lesion patients: dissociationof processes and frontal lobe lesion location. Neuropsy-chologia 2001;39:771–786.

25. Swick D, Jovanovic J. Anterior cingulate cortex and theStroop task: neuropsychological evidence for topographicspecificity. Neuropsychologia 2002;40:1240–1253.

26. Fellows LK, Farah MJ. Is anterior cingulate cortex neces-sary for cognitive control? Brain 2005;128:788–796.

27. Thompson-Schill SL, D’Esposito M, Aguirre GK, FarahMJ. Role of left inferior prefrontal cortex in retrieval ofsemantic knowledge: a reevaluation. Proc Natl Acad SciUSA 1997;94:14792–14797.

28. Hirshorn EA, Thompson-Schill SL. Role of the left infe-rior frontal gyrus in covert word retrieval: neural corre-lates of switching during verbal fluency.Neuropsychologia 2006;44:2547–2557.

29. Price CJ, Green DW, von Studnitz R. A functional imag-ing study of translation and language switching. Brain1999;122:2221–2235.

30. Leung HC, Skudlarski P, Gatenby JC, Peterson BS, GoreJC. An event-related functional MRI study of the stroopcolor word interference task. Cereb Cortex 2000;10:552–560.

31. Brass M, von Cramon DY. Selection for cognitive control:a functional magnetic resonance imaging study on the se-lection of task-relevant information. J Neurosci 2004;24:8847–8852.

32. Weissman DH, Gopalakrishnan A, Hazlett CJ, WoldorffMG. Dorsal anterior cingulate cortex resolves conflictfrom distracting stimuli by boosting attention toward rel-evant events. Cereb Cortex 2005;15:229–237.

33. Thompson-Schill SL, Swick D, Farah MJ, D’Esposito M,Kan IP, Knight RT. Verb generation in patients with focalfrontal lesions: a neuropsychological test of neuroimagingfindings. Proc Natl Acad Sci USA 1998;95:15855–15860.

34. Bench CJ, Frith CD, Grasby PM, et al. Investigations ofthe functional anatomy of attention using the Stroop test.Neuropsychologia 1993;31:907–922.

35. Pardo JV, Pardo PJ, Janer KW, Raichle ME. The anteriorcingulate cortex mediates processing selection in the

Stroop attentional conflict paradigm. Proc Natl Acad SciUSA 1990;87:256–259.

36. Ravnkilde B, Videbech P, Rosenberg R, Gjedde A, GadeA. Putative tests of frontal lobe function: a PET-study ofbrain activation during Stroop’s Test and verbal fluency. JClin Exp Neuropsychol 2002;24:534–547.

37. Murtha S, Chertkow H, Beauregard M, Dixon R, EvansA. Anticipation causes increased blood flow to the ante-rior cingulate cortex. In: Human brain mapping. Wiley-Liss, 1996:103–112.

38. Braver TS, Barch DM, Gray JR, Molfese DL, Snyder A.Anterior cingulate cortex and response conflict: effects offrequency, inhibition and errors. Cereb Cortex 2001;11:825–836.

39. Frith CD, Shallice T, Dolan RJ. “Sculpting the responsespace”—an account of left prefrontal activation at encod-ing. Neuroimage 2000;12:404–417.

40. Baleydier C, Mauguiere F. The duality of the cingulategyrus in monkey. Neuroanatomical study and functionalhypothesis. Brain 1980;103:525–554.

41. Vogt BA, Nimchinsky EA, Vogt LJ, Hof PR. Human cin-gulate cortex: surface features, flat maps, and cytoarchi-tecture. J Comp Neurol 1995;359:490–506.

42. Jurgens U. The efferent and afferent connections of thesupplementary motor area. Brain Res 1984;300:63–81.

43. Lindvall O, Bjorklund A, Moore RY, Stenevi U. Mesence-phalic dopamine neurons projecting to neocortex. BrainRes 1974;81:325–331.

44. Wang C, Ulbert I, Schomer DL,Marinkovic K, Halgren E.Responses of human anterior cingulate cortex microdo-mains to error detection, conflict monitoring, stimulus–response mapping, familiarity, and orienting. J Neurosci2005;25:604–613.

45. Lorist MM, Boksem MA, Ridderinkhof KR. Impairedcognitive control and reduced cingulate activity duringmental fatigue. Brain Res Cogn Brain Res 2005;24:199–205.

46. Devinsky O, Morrell MJ, Vogt BA. Contributions of theanterior cingulate cortex to behavior. Brain 1995;118:279–306.

Neurology 68 May 1, 2007 1523 at SWETS SUB SERVICE on June 20, 2007 www.neurology.orgDownloaded from

DOI: 10.1212/01.wnl.0000261482.99569.fb 2007;68;1515-1523 Neurology

M. P. Alexander, D. T. Stuss, T. Picton, T. Shallice and S. Gillingham Regional frontal injuries cause distinct impairments in cognitive control

This information is current as of June 20, 2007

& ServicesUpdated Information

http://www.neurology.org/cgi/content/full/68/18/1515including high-resolution figures, can be found at:

Related Articles http://www.neurology.org/cgi/content/full/68/18/1450

A related article has been published:

Subspecialty Collections

http://www.neurology.org/cgi/collection/attention Attention

http://www.neurology.org/cgi/collection/executive_function Executive function avior

http://www.neurology.org/cgi/collection/all_neuropsychology_beh All Neuropsychology/Behavior

following collection(s): This article, along with others on similar topics, appears in the

Permissions & Licensing

http://www.neurology.org/misc/Permissions.shtmlor in its entirety can be found online at: Information about reproducing this article in parts (figures, tables)

Reprints http://www.neurology.org/misc/reprints.shtml

Information about ordering reprints can be found online:

at SWETS SUB SERVICE on June 20, 2007 www.neurology.orgDownloaded from