executive control deficit in depression: event-related potentials in a go/nogo task

Psychiatry Research: Neuroimaging 122(2003) 169–184

0925-4927/03/$ - see front matter� 2003 Elsevier Science Ireland Ltd. All rights reserved.doi:10.1016/S0925-4927(03)00004-0

Executive control deficit in depression: event-related potentials ina GoyNogo task

Stefan Kaiser , Joerg Unger , Markus Kiefer , Jaana Markela , Christoph Mundt ,a b b a a

Matthias Weisbrod *a,

Department of Psychiatry, University of Heidelberg, Heidelberg, Germanya

Department of Psychiatry III, University of Ulm, Ulm, Germanyb

Received 14 May 2001; received in revised form 1 May 2002; accepted 21 May 2002

Abstract

Growing evidence suggests an impairment of executive control functions in depression. The aim of this study wasto investigate whether depressive patients show a specific impairment of executive control in a response inhibitiontask and to investigate its neurophysiological correlates using event-related potentials. We analyzed data from 16patients with unipolar depression and 16 healthy controls using an auditory GoyNogo task. High resolution event-related potentials(ERPs) were recorded. Depressive patients performed similar to controls in the Go task, but worsein the Nogo task, which required response inhibition. ERPs revealed the neurophysiological correlate of this deficit.Both groups showed the same voltage pattern in the Go task. However, in the Nogo task depressive patients showeda reduction of an early fronto-temporal positivity in the N2 time window, which was associated with responseinhibition in healthy subjects. This effect could not be explained by increased task difficulty in the Nogo task. Therewas no difference between groups in later stages of processing as indexed by the P3 complex. Therefore, the findingssuggest a specific deficit in response inhibition, which requires executive control. This deficit is thought to reflectdysfunctional activation of the network subserving executive control during an early stage of cortical processing.� 2003 Elsevier Science Ireland Ltd. All rights reserved.

Keywords: Response inhibition; Depressive disorder; N2

1. Introduction

In addition to its effects on mood, major depres-sion is associated with cognitive and particularlyattentional deficits(Austin et al., 2001). It hasbeen a matter of debate whether these impairments

*Corresponding author. Psychiatrische Universitatsklinik,¨Voss-Strasse 4, 69115 Heidelberg, Germany; Tel.:q49-6221-562745; fax:q49-6221-565477.

E-mail address: [email protected](M. Weisbrod).

reflect a general slowing of information processingor dysfunction of specific cognitive systems(Mialet et al., 1996). Growing evidence suggestsan impairment at the level of executive control,which might account in part for patients’ difficul-ties in cognitive performance, especially whenflexible responses are required(Channon andGreen, 1999).Theories of attention have proposed an execu-

tive control system that regulates information proc-essing and response selection in situations, where

170 S. Kaiser et al. / Psychiatry Research: Neuroimaging 122 (2003) 169–184

routine mechanisms are unavailable or inadequatefor task performance(Norman and Shallice, 1986;Posner and Dehaene, 1994). According to Normanand Shallice, these situations include planning,error correction, novel or difficult situations andresponse inhibition. The neural basis of this exec-utive system is thought to be a distributed networkinvolving the anterior cingulate gyrus and prefron-tal areas(Posner and DiGirolamo, 1998; Smithand Jonides, 1999).Depressive patients have been reported to show

impaired performance on tests of executive func-tion, such as the Wisconsin Card Sorting Test(Merriam et al., 1999) and the Stroop Color WordTest(Trichard et al., 1995). The interpretation thatthese performance deficits genuinely reflect exec-utive dysfunction, as opposed to general cognitiveslowing, has been challenged by some authors(Degl’Innocenti et al., 1998). However, in neu-roimaging studies, hypofrontality has been dem-onstrated in depressive patients, which suggests adysfunction in areas critical for executive functions(Dolan et al., 1994). There is also some evidencefor prefrontal and anterior cingulate dysfunction inthose patients when functional activation duringexecutive task performance is studied(Elliott etal., 1997; George et al., 1997).While the functional neuroimaging methods

used in these studies provide good spatial resolu-tion, electrophysiological methods yield preciseinformation concerning the time course of brainactivation. In this study, we used event-relatedpotentials (ERPs), which are electroencephalo-graphic (EEG) changes time-locked to sensory,motor, or cognitive events. To our knowledge,ERPs have not yet been applied to address exec-utive function in depression.Response inhibition is one of the situations that

require activation of the executive control system.The inhibition of a prepared response can bestudied with GoyNogo paradigms. In such tasks,subjects usually have to respond to a given targetstimulus in the Go task, while they have towithhold the response to the target stimulus in theNogo task. While Go and Nogo stimuli should besimilar in stimulus processing, the Nogo taskadditionally demands inhibition of a response tothe target. Several versions of the GoyNogo para-

digm have been applied in electrophysiologicalstudies to investigate the effects of response inhi-bition. Two major ERP components related toresponse inhibition have been described. First, theN2, a negative deflection over fronto-central elec-trodes at approximately 200–300 ms, is found tobe increased in the Nogo compared to the Go task(Jodo and Kayama, 1992; Kopp et al., 1996). Aconcurrent positive deflection has been describedover fronto-temporal regions and interpreted as apolarity-inversed N2 component(Kiefer et al.,1998). Second, the P3, a positive deflection atapproximately 300–600 ms, shows a more frontaldistribution in the Nogo task(Pfefferbaum et al.,1985; Strik et al., 1998). This is in contrast to theparietal P3 observed in the Go task. In addition, aleft lateralization of Nogo-P3 over fronto-centralelectrodes has been reported(Kiefer et al., 1998;Roberts et al., 1994). These effects are thought tobe electrophysiological correlates of response inhi-bition and to reflect activity in the cortical networksubserving executive control.The aim of our study was to investigate whether

depressive patients show a specific impairment ofexecutive control in a response inhibition task andto investigate the neurophysiological correlatesusing ERPs. For this purpose we examined depres-sive patients and healthy controls in an uncuedauditory GoyNogo task during which high-resolu-tion ERPs were recorded. A methodological issuewas a possible confound between deficits in audi-tory stimulus perception as opposed to genuinedeficits in response inhibition, i.e. depressivepatients might perform worse because they wereunable to correctly identify the stimuli. In order toaddress this problem, we adjusted pitch disparityto the subject’s individual discrimination ability. Afurther objective was to separate the effects ofinhibition requirement and task difficulty. There-fore, subjects performed Go and Nogo tasks in asimple and a difficult condition, which differed inpitch disparity between rare and frequent tones.This experimental design of an uncued auditoryGoyNogo task has already been validated in twoprevious studies with healthy subjects(Kiefer etal., 1998) and schizophrenic patients(Weisbrod etal., 2000).

171S. Kaiser et al. / Psychiatry Research: Neuroimaging 122 (2003) 169–184

Table 1Demographic data and results of pitch discrimination test

Sex Age Edu BDI HRSD Pitch simple Pitch difficult(myf) (S.D.) (S.D.) (range) (range) (S.D.) (S.D.)

Controls 4y12 38.3(7.7) 11.5 (1.4) 3.4 (0–10) – 1065.0(56.5) 1048.4(42.6)Patients 4y12 40.0(8.7) 11.5 (1.7) 29.8 (15–43) 22.7 (17–30) 1119.4(73.2) 1089.1(55.3)

Mean age and years of school education(with standard deviation). Mean scores(with range) for BDI (sBeck DepressionInventory) and HRSD(sHamilton Rating Scale for Depression). Mean frequencies in Hz(with standard deviation) used for thedeviant stimulus in the simple and difficult condition of the main experiment.

Regarding the comparison of depressive patientswith healthy controls, our hypotheses were:(1)depressive patients would perform similar to con-trols in the Go task. In contrast, they wouldperform worse in the Nogo task, which requiresresponse inhibition.(2) In the ERP data depressivepatients would show a modification of the inhibi-tion effects observed in healthy subjects.

2. Methods

2.1. Subjects

Twenty-three patients with unipolar majordepression(DSM-IV classification 296.x) and 19healthy control subjects participated in the study.Six patients and three controls had to be excludedbecause an insufficient number of correct artifact-free trials were available. In one patient recording,there was an amplifier problem. Thus, 16 patientsand 16 control subjects were included in the finalanalysis. Groups were matched for gender, age andyears of education(see Table 1 for demographicdata). Exclusion criteria were a history of neuro-logical illness, major medical disorders, orimpaired hearing. Handedness was assessed by aGerman version of the Edinburgh Inventory(Old-field, 1971) and only right-handed subjects wereincluded in the study.Patients were recruited from the wards of the

University of Heidelberg Psychiatric Hospital.Diagnosis was confirmed by Structured ClinicalInterviews for DSM-IV (Wittchen et al., 1997).All patients with a history of an Axis-I disorderother than unipolar depression were excluded fromthe study. Severity of depression was assessedusing the Hamilton Rating Scale for Depression(Hamilton, 1967) and the Beck Depression Inven-

tory (Beck et al., 1961) (see Table 1). At the timeof the experiment, all patients were treated withantidepressant medication. Four patients were tak-ing SSRIs, six patients tricyclics and six patientsa combination of SSRI and tricyclic. Six patientswere also receiving benzodiazepines and four wereadditionally treated with neuroleptic medication.Control subjects were recruited from the hospitalstaff. None of the controls had a personal or familyhistory of psychiatric disorders or were taking anymedication that might potentially affect cognition.The Beck Depression Inventory was administeredto screen for depressive symptomatology(seeTable 1).The study protocol was approved by the local

ethics committee and all subjects gave writteninformed consent after the experiment had beenfully explained.

2.2. Task and procedure

The uncued GoyNogo task we used(Kiefer etal., 1998) was a modification of the auditoryoddball paradigm. Subjects heard frequent low-pitched and rare high-pitched tones. In the Gotask, a response to rare tones was required using amouse-button. In the Nogo task, subjects respond-ed to the frequent tones and had to withhold theresponse to the rare tones. Response side wasbalanced across subjects to control for lateralizedmotor activity. Both tasks were performed in asimple and a difficult condition, which differed inpitch disparity. Thus, the experiment consisted offour blocks(Go-simple, Go-difficult, Nogo-simple,Nogo-difficult), which were counterbalancedacross subjects. All experiments were conductedbetween 9.00 h and 12.00 h. Subjects were seated


upright in a dimly lit room. They were instructedto fixate a red dot on a gray screen and to avoidblinking and movement of any sort. Between thefour experimental blocks of approximately 5 min,subjects were allowed to rest. The whole experi-ment including pitch discrimination test, electrodeplacement, main experiment and breaks tookapproximately 2 h to complete.Auditory stimuli were presented via headphones

using the Stim software package(Neuroscan Inc.).Each block consisted of 200 stimuli with aninterstimulus interval varying between 1.3 and 1.7s. All tones had a duration of 40 ms. The soundpressure level was adjusted to 97 dB on the Stimsystem, which corresponds to approximately 70dB hearing level. Subjects reported the tones to beof comfortable intensity. In all conditions, thefrequent tones(1000 Hz) occurred with 80%probability. The rare tones occurred with 20%probability. Their pitch was adjusted according tothe individual subject’s discrimination ability. Forthis purpose, subjects performed a pitch discrimi-nation test before the main experiment. In this taskthey heard tone pairs, where the first tone alwayshad a pitch of 1000 Hz, while the second variedbetween 1000 and 1200 Hz(1000, 1010, 1015,1020, 1025, 1030, 1040, 1060, 1100, 1150 and1200 Hz). Subjects had to detect tone pairs differ-ing in pitch. The difficulty of the GoyNogo taskwas adjusted to the individual discrimination abil-ity using the following procedure. For the simplecondition of the GoyNogo task, we selected thedeviant stimulus with the smallest pitch disparity,which had been identified as different in 100% ofpairs. For the difficult condition, the deviant stim-ulus with 80% recognition rate was chosen.Patients performed more poorly on the preliminarytest than controls(see Table 1), which was con-firmed by t-tests on the tone frequencies for thesimple(ts2.32,P-0.05) and difficult conditions(ts2.35,P-0.05).

2.3. Data acquisition and analysis

Subjects’ responses were recorded with the stim-ulation computer using the Stim software package.In order to obtain a behavioral measure suitablefor direct comparison between Go and Nogo task

performance, the sensitivity indexd9 according toGreen and Swets(1966) was calculated asd9sz(hit rate frequent tone)yz (error rate rare tone).For statistical analysis, ANOVA withd9 as thedependent measure was performed with group asthe between-subject factor, and task and difficultyas the within-subject factors. Newman–Keuls testswere used for post hoc comparisons. Statistica 5.1for Windows (StatSoft Inc.) was used for allstatistical computations. The significance level wasset toP-0.05, statistical trends ofP-0.1 are alsoreported.Scalp voltages were recorded using a 61-channel

Easy Cap(Falk Minow Systems) with sinteredAgyAgCl-electrodes(see Fig. 1). The referenceelectrode was placed on the vertex, while groundwas linked to a frontal midline electrode. Addi-tionally, eye movements were monitored withsupra- and infraorbital electrodes and with elec-trodes on the external canthi. Electrode impedancewas maintained below 10 kV for all recordings.Electrical signals were recorded continuously withSynamps amplifiers(bandwidth DC-70 Hz, 50-Hznotch filter) and digitized(sampling rate 250 Hz).EEG data were processed offline using the

software BrainVision Analyzer 1.1(Brain ProductsGmbH). The continuous EEG was digitally low-pass filtered at 16 Hz. The EEG was then seg-mented into epochs 100 ms pre-stimulus to 1000ms post-stimulus and baseline corrected to themean amplitude 100 ms before stimulus. First, allsegments with a difference between minimum andmaximum value exceeding 200mV in any EEGor EOG channel were automatically excluded fromfurther processing. Second, eyeblinks and bulbousmovements were corrected separately according tothe method of Gratton et al.(1983). After thiscorrection procedure, all segments were screenedvisually and those with remaining artifacts wereremoved. The average-reference transform wasapplied. All segments with correct responses torare tones were averaged and at least 20 trialswere available for each subject and condition.Grand averages were computed to identify com-ponents and time windows for statistical analysis.Three scalp regions of interest, which have consis-tently shown effects of response inhibition in ourprevious studies(Kiefer et al., 1998), were select-


Fig. 1. Electrode locations on the Easy Cap, view from top(figure courtesy of Falk Minow Systems). Electrodes analyzed statisticallyare depicted in black. G and R denote ground and recording reference, respectively. Three bilateral electrode pairs were selectedfrom each of the following scalp regions(corresponding 10y20 system positions in parentheses): fronto-temporal: 49y37(Af79yAf89),61y51 (F9yF10), 60y52 (Ft99yFt109); fronto-central: 19y9 (F19yF29), 18y10 (Fc39yFc49), 7y3 (Fc19,Fc29); centro-parietal 15y13(P19yP29), 29y26 (P39yP49), 28y27 (Po19yPo29).

ed: (inferior) fronto-temporal; fronto-central; andcentro-parietal. Each region was represented bythree electrodes over each hemisphere(see Fig. 1and Fig. 2), which were pooled(i.e. averaged) forthe represented waveforms(Figs. 3–5).The components of interest were P3 at fronto-

central and parietal electrode sites, as well as thefronto-central N2 and its accompanying fronto-temporal positivity. Peak latencies for these com-ponents did not differ significantly between groupsand conditions. Therefore, mean scalp voltageswere analyzed statistically in only one time win-dow for each component. Time windows werecentered on the peak latencies of the respectivecomponents in the Nogo task and their duration

was chosen according to the previous validationstudy (Kiefer et al., 1998). Thus, the followingtime windows were selected for statistical analysis:N2 (225–285 ms) at fronto-temporal and fronto-central electrode sites; frontal-P3(318–398 ms)at fronto-central; and parietal-P3(422–542 ms) atcentro-parietal sites. For each time window andscalp region, separate ANOVAs were calculatedwith group (depressive patientsycontrol subjects)as the between-subject factor and task(GoyNogo),difficulty (simpleydifficult), hemisphere (leftyright) and electrode site as the within-subjectfactors. Greenhouse–Geisser corrections wereapplied where appropriate. Homogeneity of vari-ances between groups was confirmed by comput-


Fig. 2. Accuracy of signal detection(d9 index according toGreen and Swets, 1966) for controls and patients. Both tasks(GoyNogo) are shown with difficulty(simpleydifficult) aver-aged. Error bars represent standard error of mean. * and nsdenote significant and non-significant interactions,respectively.

ing HartleyF-max statistics. Newman–Keuls testswere used for post hoc comparisons.

3. Results

The focus of this article is the comparisonbetween depressive patients and healthy subjects.Therefore, we only report main effects and thoseinteractions that involve group as a factor.

3.1. Behavioral data

Reaction times did not differ significantlybetween groups. The ANOVA with the sensitivityindex d9 as the dependent variable showed maineffects of task (GoyNogo) (F s54.01, P-1,30

0.001) and difficulty (simpleydifficult) (F s1,30

31.64,P-0.001). Subjects performed the Go taskbetter than the Nogo task and the simple betterthan the difficult condition. Most important werethe effects involving groups. There was a maineffect of group (F s7.26, P-0.05) showing1,30

that over all conditions controls performed betterthan patients. This effect was modified by agroup=task interaction(F s6.96,P-0.05) as1,30

shown in Fig. 2. Post hoc tests revealed that groupsdid not differ significantly in the Go task, whilepatients performed worse than controls in the Nogo

task (P-0.05). There was neither a significantgroup=difficulty nor a group=task=difficultyinteraction.Thus, depressive patients showed a performance

deficit specific to the Nogo task, which did notinteract with task difficulty. This result raised thequestion whether it reflects a higher rate in errorsof omission (failure to respond to the frequenttone) or errors of commission(erroneous responseto the rare tone). For further analysis of the errorsin the Nogo task, an ANOVA was calculated withgroup as the between-subject factor and error type(omissionycommission) and difficulty (simpleydifficult) as the within-subject factors. This anal-ysis yielded a main effect of group(F , P-1,30

0.001), which showed that in sum patientscommitted more errors than controls. This effectwas modified by a group=error type interaction(F s13.24,P-0.005). Both groups showed a1,30

similar rate of omission errors. In contrast, patientsshowed a higher commission error rate than con-trols (P-0.01). In the Go condition groups didnot differ significantly in either commission oromission errors.

3.2. ERP data

3.2.1. Grand averages and topographic mapsN2 and P3 were elicited reliably for all subjects

in all conditions. We computed averages usingonly trials with a correct response to the rare tone,i.e. Go trials with successful response to targetstimuli or Nogo trials with successful inhibition.Inspection of the grand averages(Figs. 3–5) andtopographic maps(Fig. 6) showed the followingspatio-temporal distribution for these components.N2 emerged as a fronto-centrally distributed neg-ativity peaking at approximately 255 ms(Fig. 3).This negative peak was accompanied by a concur-rent inferior frontotemporal positivity in the N2time window (Fig. 4). Both the fronto-central N2and the concurrent fronto-temporal positivity areclearly visible in the topographic maps, with thelatter being virtually absent in the patient’s Nogodata(Fig. 6). N2 was followed by two overlappingpositive components in all conditions. Fronto-central electrodes showed an early componentpeaking at approximately 360 ms, which we refer


Fig. 3. Grand averages for pooled(i.e. averaged) fronto-central electrode sites(see also Fig. 1) over both hemispheres(leftyright)for both tasks(goynogo). N2 is visible as a negative peak at approximately 260 ms, while P3 is visible as a positive peak atapproximately 360 ms. N2 and P3 did not differ significantly between groups. In addition, N1 is visible as a negative peak atapproximately 100 ms.

to as frontal-P3(Fig. 3). Over parietal electrodes,a broad positivity peaking at approximately 480ms was found, for which we use the term parietal-P3 (Fig. 5). Over fronto-central electrodes anoverlapping negative shift between 250 and 700ms of the ERP waveform was seen particularly inthe Go task(Fig. 3), which was thought to reflectmotor-related activity.

3.2.2. N2 time windowAnalysis of the inferior fronto-temporal positiv-

ity yielded a main effect of group(F s4.55,1,30

P-0.05). Mean amplitude was more positive forcontrols than depressive patients. This effect wasmodified by a group=task interaction(F s1,30

8.69,P-0.01) as shown in Fig. 7. Post hoc testsrevealed that groups did not differ significantly inthe Go task(P)0.1). In contrast, the Nogo task

showed a more positive voltage for controls thanfor patients (P-0.001). Post hoc comparisonswithin groups revealed that in controls voltagebecame more positive from Go to Nogo task(P-0.05), while it became more negative in patients(P-0.05). Thus, controls and patients showed anopposite modulation of this N2 effect from Go toNogo task. The group=task interaction was notmodified by any higher order interaction. Further-more, a group=electrode interaction was found(F s4.72, P-0.05). Healthy subjects had sig-1,30

nificantly more positive voltages at both anteriorelectrode sites(49y37,61y51 see Fig. 1), but notat the most posterior electrode site(60y52).At fronto-central electrodes no main effects were

found. A group=task=electrode interaction(F s3.91, P-0.05) was the only significant2,60

effect involving groups. Post hoc tests revealed


Fig. 4. Grand averages from pooled inferior fronto-temporal electrodes over both hemispheres for both tasks. At approximately 260ms, a positive peak is visible, which most likely corresponds to a polarity-inverted N2. Patients show less positive N2 time windowamplitude in the Nogo task. In addition, a polarity-inverted N1 component and a sustained negativity(SN) peaking at approximately400 ms are visible.

that healthy controls had more negative voltagesin the Nogo task only at the most posteriorelectrode sites(7y3).

3.2.3. Frontal-P3 time windowAt fronto-central electrodes main effects of task

(F s28.13, P-0.001) and difficulty (F s1,30 1,30

11.37, P-0.001) were found for P3. Amplitudewas larger in the Nogo than the Go task and inthe simple than the difficult condition. Further-more, a main effect of electrode site(F s12.12,1,30

P-0.001) showed that the largest frontal-P3amplitudes were elicited at the most posteriorelectrode pair(7y3). Inspection of the ERP datasuggested a modification of Nogo-P3 by groupand difficulty, but there was no significant effectinvolving groups in this time window(Table 2).

3.2.4. Parietal-P3 time windowAt centro-parietal electrodes, main effects for

task (F s10.91, P-0.005) and difficulty1,30

(F s8.99,P-0.01) were found. Amplitude was1,30

larger in the Go than the Nogo task and in thesimple than the difficult condition. Main effectsconcerning the spatial distribution of parietal-P3were also found for hemisphere(F s7.0, P-1,30

0.05) and electrode site(F s3.9, P-0.05).1,30

Amplitude was largest over the left hemisphereand electrode sites closest to the midline{ 15y13,28y27} . There was no significant effect involvinggroups.The topographic maps at 255 ms suggest that

the control group’s parietal P3 develops later inthe Nogo task condition(Fig. 6). Therefore, we


Fig. 5. Grand averages for pooled centro-parietal electrode sites over both hemispheres(leftyright) for both tasks(GoyNogo). P3is visible as a broad positivity at approximately 480 ms, which did not differ significantly between groups.

performed an exploratory ANOVA for the timewindow from 200 to 400 ms that showed a trendtowards a group=task interaction(F s3.98,1,30

P-0.1).

3.2.5. Other ERP componentsERP graphs of fronto-central(Fig. 3) and fron-

to-temporal(Fig. 4) electrodes suggest differencesbetween groups in the N1 time window. This effectdid not reach significance(P)0.1). Over fronto-temporal regions the ERP waveforms(Fig. 4)show a sustained negativity lasting up to 600 ms,which in the Nogo condition is more pronouncedin the patient group. Part of the N2 effect couldbe due to this overlapping negativity. However, inthe time window from 300 to 600 ms, an ANOVAshowed only a trend(F s3.83,P-0.1) towards1,30

a group=condition interaction.

3.3. Effects of medication and psychopathology

In order to further elucidate a possible relationwith clinical variables, we conducted the followinganalyses using differences in fronto-temporal N2time window amplitude andd9 index betweenNogo and Go tasks as dependent variables.For each medication group(tricyclics, SSRIs,

neuroleptics and bezodiazepines) we computedt-tests comparing patients taking each medicationwith those not on the medication. In this explora-tory analysis, we found no significant effect ofmedication on either N2 time window differencesin amplitude ord9 index. Furthermore, we com-puted correlations between severity of depression(as measured by Beck and Hamilton scores) withbehavioral and electrophysiological data. No sig-nificant linear correlation was found.


Fig. 6. Spherical spline map at 255 ms(mean of N2 time window) for controls and patients in the difficult Go and Nogo tasks. Inthe Go task, both groups show the following ERP components:(1) a bilateral negativity over fronto-central electrodes;(2) aninferior fronto-temporal positivity; and(3) a parietal positivity. In the Nogo task, healthy controls show an increased frontal positivity,while fronto-central negativity is slightly decreased and parietal positivity not visible. In contrast, depressive patients show virtuallyno fronto-temporal positivity in the Nogo task, while fronto-central negativity and parietal positivity resemble the Go task.


Fig. 7. Mean voltages in the N2 time window(225–285 ms)over inferior fronto-temporal electrode sites for controls andpatients. Both tasks(GoyNogo) are shown with difficulty lev-els (simpleydifficult) averaged. Error bars represent standarderror of mean. * and ns denote significant and non-significantinteractions, respectively.

Table 2Mean amplitudes and statistical results for ERP components

Control Patient Group Group=task Group=difficulty

Go Nogo Go Nogo F P F P F P

Frontal positivity 0.42"0.75 1.09"0.61 0.24"0.49 y0.45"0.74 4.55 0.041* 8.69 0.006** 2.39 0.13in N2 window

N2 fronto-central y0.89"0.77 y0.60"0.48 y0.54"0.93 y0.39"0.85 0.25 0.62 0.16 0.69 0.92 0.34P3 fronto-central 0.26"1.13 1.65"0.97 0.23"0.99 1.54"0.78 0.012 0.91 0.026 0.87 0.45 0.51P3 parietal 4.44"1.83 2.84"1.17 4.15"1.86 3.56"1.32 0.23 0.63 2.36 0.14 1.69 0.21

Mean amplitudes for each component inmV with standard deviations for both groups and task conditions. An ANOVA wasperformed with mean amplitude as dependent variable, and results for the group main effect and the critical group=task andgroup=difficulty interactions are shown(see results).

Significant results(P-0.05).*

Highly significant results(P-0.01).**

4. Discussion

We assessed executive control in depressivepatients and healthy controls in an uncued auditoryGoyNogo tasks that requires response inhibition.Task difficulty was adjusted on an individual basisby adapting pitch difference to the individualdiscrimination ability. The poorer performance ofdepressive patients in the pitch discrimination test

demonstrates that the individual calibration of taskdifficulty was important for the present study.The main focus of the present study was the

comparison between groups with respect to the Goand Nogo tasks. Stimuli and mode of responsewere identical for both task conditions. Hence, thecritical difference between the two tasks was theinstruction, which required inhibition of a motorresponse in the Nogo task. There was no signifi-cant performance difference between groups in theGo task, but depressive patients performed worsein the Nogo task.Since the Nogo task is more difficult to perform

than the Go task as reflected by the behavioraldata, it may be objected that non-specific effectsof motivation or cognitive slowing might have ledto these results. To address this problem, weanalyzed two important within-subject compari-sons:(1) Go vs. Nogo task reflecting the effectsof the inhibition requirement; and(2) simple vs.difficult condition reflecting difficulty of pitchdiscrimination as a cognitive process not involvingresponse inhibition. Non-specific cognitive deficitsin depressive patients should affect both of thesecomparisons in a comparable fashion, while inhib-itory deficits should only affect the GoyNogocomparison. The group=task interaction showsthat performance is more affected in the Nogocondition in patients than controls. The lack of asignificant group=difficulty or group=task=difficulty interaction indicates that increasing dif-ficulty of pitch discrimination affected both groups


and did not differentiate between patients andcontrols. Therefore, the patients’ poorer perform-ance in the Nogo task is probably not reducible toa generalized impairment due to non-specific fac-tors. The more likely explanation is that thepatient’s poorer performance depended on theinhibition requirement. However, we did notmanipulate the difficulty of response inhibitionand, thus, cannot draw conclusions about theeffects of challenging executive functions at dif-ferent levels of difficulty.The interpretation that the poor performance of

the depressive patients on the Nogo task wasmainly due to impaired response inhibition wasconfirmed by further analysis of the errors made.In the Nogo task, patients showed a higher rate ofcommission errors, but a similar rate of omissionerrors. This reflects a failure to inhibit the responseto the rare tones. In sum, these behavioral findingspoint towards a specific impairment of responseinhibition in depressive patients.The neurophysiological basis of the observed

inhibition deficits can be explored by analyzingthe ERP data. Both groups showed similar activa-tion in the Go task. In the Nogo task, there wasno difference between groups in later stages ofprocessing as indexed by the P3 complex. How-ever, voltage of the inferior fronto-temporal posi-tivity in the N2 time window differed betweencontrols and patients in the Nogo task. Controlshad a more positive voltage in the Nogo taskcompared with the Go task. In contrast, patientsshowed a decreased voltage in the Nogo task. Bothgroups showed similar effects of task difficulty onERP components. Hence, the difference betweengroups observed in the Nogo task most likelyreflects the inhibition requirement and can proba-bly not be explained as a result of increasingdifficulty. The present results for the N2 timewindow point towards dysfunctional activation ofthe generators underlying the modulation of thisearly inferior fronto-temporal positivity in theNogo task.These results could be confounded by medica-

tion effects in the patient group. We did not findany differential effects of medication on eitherbehavioral or electrophysiological data in our anal-ysis, which is certainly limited by sample size.

However, the lack of medication effects is consis-tent with behavioral data using the Stroop test(Killian et al., 1984) and electrophysiological find-ings concerning P300 amplitude in depressivepatients(Sara et al., 1994).The modulation of the fronto-temporal positivity

in the N2 time window has been previously report-ed for response inhibition in a GoyNogo task(Kiefer et al., 1998). Figs. 3 and 4 show that it issimilar in shape and time course to the fronto-central N2 component observed in this study,which suggests that it reflects activity of the samedipole. The interpretation of these effects is com-plicated by the fact that activity in this timewindow contains at least two major components.First, the ‘classic’ modality-dependent N2 com-ponent is generally observed in oddball tasks(Simson et al., 1977). Second, a modification ofthis component is thought to reflect an effect ofresponse inhibition(Jodo and Kayama, 1992).In the present study the Go task condition shows

activity in the N2 time window without overlap-ping effects of inhibition. This activity did notdiffer significantly between controls and patients.Auditory N2 in oddball tasks has been reported tobe reduced in depressive patients(el Massioui etal., 1996), but results generally have been incon-sistent(Ogura et al., 1993; Sara et al., 1994). Thepresent data do not support the notion of a gener-ally reduced auditory N2. Therefore, we focus onthe inhibition-related modulation of activity in theN2 time window.Increased N2 amplitude is a well-known corre-

late of response inhibition and this modulation hasbeen termed the Nogo-N2(Jodo and Kayama,1992; Kopp et al., 1996). This effect has beendescribed mainly at frontal and central electrodesites at or near the midline. Our main findings inthe N2 time window concern an increased positiv-ity at inferior fronto-temporal sites, while no majoreffect involving task was found at fronto-centralelectrodes.Three considerations are important regarding the

interpretation of the present findings. First, overfronto-central sites we observed a negative shift asearly as 250 ms, which was most pronounced inGo trials. This most likely reflects motor-relatedpotentials. The overlap with the fronto-central N2


renders findings at superior sites close to the motorcortex difficult to interpret. Second, most studiesreporting a large fronto-central Nogo-N2 have usedlinked mastoids or earlobes as reference, while wehave used an average reference. Third, in auditorytasks, the Nogo-N2 at frontal sites close to themidline is often found to be small compared tothe Nogo-N2 elicited by visual stimulation(Fal-kenstein et al., 1999). To explain these findings,Falkenstein et al.(1999) have suggested modality-specific Nogo-N2 generators in the prefrontal cor-tex, which is in line with findings in animal studies(Gemba and Sasaki, 1990). According to ourresults from this and previous studies(Kiefer etal., 1998), measurements at inferior frontal elec-trodes might reflect the auditory Nogo-N2 betterthan superior sites. The scarcity of reports con-cerning this effect might be due to the fact that inprevious GoyNogo studies using auditory stimuli,electrodes have rarely been placed in the respectiveregion (Falkenstein et al., 1999). In addition tothe Nogo-N2 in controls, we have observed majordifferences between groups at inferior fronto-tem-poral sites. Following the previous considerations,we interpret these findings as a modulation ofNogo-N2 in depressive patients as compared tohealthy controls.The generators of the visual Nogo-N2 have been

localized to right inferio-lateral prefrontal cortexusing source analysis(Jackson et al., 1999). In anauditory task, Kiefer et al.(1998) localized theinhibitory effects in the N2 time window to bilat-eral inferio-lateral prefrontal areas. Differencesbetween the studies could be due in part to thedifferent modalities used. However, if one transfersanimal findings to humans, one would expect theseareas to be adjacent in the prefrontal cortex witha distance below the resolution of current electro-physiological methods(Gemba and Sasaki, 1990).Therefore, other parameters of the experiment suchas the complexity of the stimuli or the use of awarning cue might be more important factors. Ingeneral, electrophysiological findings are in linewith neuroimaging studies stressing the role ofinferio-lateral prefrontal cortex in response inhibi-tion (Garavan et al., 1999; Konishi et al., 1999).In sum, the result of a reduced Nogo-N2 indepressive patients points towards reduced activa-

tion in a prefrontal area critical for responseinhibition during an early stage of processing.The more anterior distribution of P3 in Nogo

trials has also been commonly described as anelectrophysiological marker of response inhibition(Pfefferbaum et al., 1985; Strik et al., 1998). ThisNogo-P3 is thought to be generated by a distrib-uted network, which probably involves prefrontal,premotor and anterior cingulate areas(Kiefer etal., 1998). In the present study, P3 was found tobe frontally distributed in the Nogo task for thecontrol as well as the patient group. Thus, we wereable to replicate the previous findings for healthysubjects. Depressive patients showed a similarpattern of Nogo-P3 distribution. The absence ofsignificant differences between groups could bedue to a mixture of elevated and diminished Nogo-P3 in depressive patients. Inspection of the indi-vidual averages did not reveal this pattern andvariance was not increased in the patient group. Inaddition, the topographic maps suggest that parietalP3 in the Nogo task condition develops later incontrols. This could reflect either a genuine effectof P3 suppression(Jackson et al., 1999) or anoverlap with the inhibitory components mainlyobserved at inferior frontal sites. However, in thepresent analysis, this effect did not reachsignificance.Further information on the spatio-temporal pat-

tern of response inhibition and its deficits can beobtained by comparing the results for depressiveand schizophrenic patients. Our group has previ-ously reported data for schizophrenic patients per-forming the same response inhibition task(Weisbrod et al., 2000). Both patient groupsshowed impaired performance in the Nogo task,which requires response inhibition. This behavioraldeficit was reflected differently in the ERP datafor both groups. Schizophrenic patients showednormal activation in the N2 time window, but alack of P3 lateralization in the Nogo task. Incontrast, depressive patients showed dysfunctionalactivation in the N2 time window, while P3 seemedto be intact. The two components Nogo-N2 andNogo-P3 probably reflect different stages and gen-erators in the processing involved in responseinhibition.


The data from the two studies with schizo-phrenic and depressive patients suggest that thesestages can be selectively impaired. Further char-acterization of these processes would benefit fromthe combination of functional imaging and electro-physiological methods. In addition, it is an inter-esting perspective that the spatio-temporal patternof brain activation in certain cognitive tasks mighthelp to differentiate between psychiatric disorders.Further research is needed to verify the specificityof these findings by directly comparing depressivepatients with other patient groups. In this context,it is also an important issue whether the observeddeficits can be linked to a disease like majordepression or rather to a symptom like psycho-motor retardation(Dolan et al., 1993).In the neuropsychological literature, response

inhibition is thought to require activity of theexecutive control system(Norman and Shallice,1986). However, it is still an open issue whetherthere is a unitary executive system or there areseparate subprocesses(Stuss et al., 1995). Theneuroimaging literature suggests an intermediatesolution with overlapping activity in dorsolateralprefrontal and anterior cingulate areas, but alsotask and modality-specific prefrontal activation(Smith and Jonides, 1999). The areas specific forinhibitory control seem to localize mainly to theinferior part of the large dorsolateral prefrontalcortex(Garavan et al., 1999; Konishi et al., 1999).But it is an interesting observation that these areasare activated not only in tasks of pure motorinhibition, but also in more complicated tests ofexecutive functions such as the Wisconsin CardSorting Test(Konishi et al., 1999). In this context,it seems reasonable to view the inhibitory deficitswe observed in the present study in conjunctionwith evidence for prefrontal cortex dysfunction indepression from functional neuroimaging studies(Drevets, 1998; Videbech, 2000). Most PET stud-ies have scanned subjects in a resting condition,and evidence for dysfunctional activation of pre-frontal cortex in tasks challenging executive con-trol is sparse. One study examined depressivesubjects using the Tower of London planning taskand found dysfunction in prefrontal, cingulate andstriate areas(Elliott et al., 1997). George et al.(1997) specifically analyzed anterior cingulate

activation in the Stroop attentional conflict taskand found a lack of activation in depressivepatients. To integrate neuroimaging findings, May-berg (1997) has proposed a model of limbic-cortical dysregulation in depression. According tothis model, executive and inhibitory deficits couldbe explained by a failure of coordinated interac-tions between limbic and neocortical areas, whichinvolve a down-regulation of dorsolateral prefron-tal activity. It remains an issue of debate whetherthe executive systems of the prefrontal cortex areaffected as a whole or if certain subsystems canbe delineated. Furthermore, it is not yet entirelyclear to what extent the observed dysfunctionsrepresent a state or a trait marker, even thoughresolution of abnormalities has been reported afterrecovery from depression(Mayberg et al., 1999).From a clinical perspective, cognitive impair-

ment is an important feature of major depression.On a behavioral level, executive deficits and par-ticularly inhibitory deficits could explain difficul-ties with flexible courses of action, which arecommonly observed in depressive patients(Chan-non and Green, 1999). Perhaps even more impor-tantly, further analysis of executive dysfunctioncould define specific patterns of brain dysfunctionand help to pave the way for a cognitive neuro-psychiatric model for depression.

Acknowledgments

Stefan Kaiser and Jaana Markela were supportedby the Deutsche Forschungsgemeinschaft throughthe University of Heidelberg Graduate Program inClinical Emotion Research. We thank SabineMeidner for her competent help during EEGacquisition.

References

Austin, M.P., Mitchell, P., Goodwin, G.M., 2001. Cognitivedeficits in depression: possible implications for functionalneuropathology. British Journal of Psychiatry 178, 200–206.

Beck, A.T., Ward, C.H., Mendelson, M., Mock, J.E., Erbaugh,J.K., 1961. An inventory for measuring depression. Archivesof General Psychiatry 4, 561–571.

Channon, S., Green, P.S., 1999. Executive function in depres-sion: the role of performance strategies in aiding depressedand non-depressed participants. Journal of Neurology, Neu-rosurgery and Psychiatry 66, 162–171.


Degl’Innocenti, A., Agren, H., Backman, L., 1998. Executive˚deficits in major depression. Acta Psychiatrica Scandinavica97, 182–188.

Dolan, R.J., Bench, C.J., Brown, R.G., Scott, L.C., Frackowiak,R.S., 1994. Neuropsychological dysfunction in depression:the relationship to regional cerebral blood flow. Psycholog-ical Medicine 24, 849–857.

Dolan, R.J., Bench, C.J., Liddle, P.F., Friston, K.J., Frith, C.D.,Grasby, P.M., Frackowiak, R.S., 1993. Dorsolateral prefron-tal cortex dysfunction in the major psychoses; symptom ordisease specificity? Journal of Neurology, Neurosurgery andPsychiatry 56, 1290–1294.

Drevets, W.C., 1998. Functional neuroimaging studies ofdepression: the anatomy of melancholia. Annual Review ofMedicine 49, 341–361.

el Massioui, F., Everett, J., Martin, M.T., Jouvent, R., Wid-locher, D., 1996. Attention deficits in depression: an electro-¨physiological marker. Neuroreport 7, 2483–2486.

Elliott, R., Baker, S.C., Rogers, R.D., O’Leary, D.A., Paykel,E.S., Frith, C.D., Dolan, R.J., Sahakian, B.J., 1997. Prefron-tal dysfunction in depressed patients performing a complexplanning task: a study using positron emission tomography.Psychological Medicine 27, 931–942.

Falkenstein, M., Hoormann, J., Hohnsbein, J., 1999. ERPcomponents in GoyNogo tasks and their relation to inhibi-tion. Acta Psychologica(Amsterdam) 101, 267–291.

Garavan, H., Ross, T.J., Stein, E.A., 1999. Right hemisphericdominance of inhibitory control: an event-related functionalMRI study. Proceedings of the National Academy of Sci-ences of the Unites States of America 96, 8301–8306.

Gemba, H., Sasaki, K., 1990. Potential related to no-go reactionin goyno-go hand movement with discrimination betweentone stimuli of different frequencies in the monkey. BrainResearch 537, 340–344.

George, M.S., Ketter, T.A., Parekh, P.I., Rosinsky, N., Ring,H.A., Pazzaglia, P.J., Marangell, L.B., Callahan, A.M., Post,R.M., 1997. Blunted left cingulate activation in mooddisorder subjects during a response interference task(theStroop). Journal of Neuropsychiatry and Clinical Neurosci-ences 9, 55–63.

Gratton, G., Coles, M.G., Donchin, E., 1983. A new methodfor off-line removal of ocular artifact. Electroencephalog-raphy and Clinical Neurophysiology 55, 468–484.

Green, D.M., Swets, J.A., 1966. Signal Detection Theory andPsychophysics. Wiley, New York.

Hamilton, M., 1967. Development of a rating scale for primarydepressive illness. British Journal of Social and ClinicalPsychology 6, 278–296.

Jackson, S.R., Jackson, G.M., Roberts, M., 1999. The selectionand suppression of action: ERP correlates of executivecontrol in humans. Neuroreport 10, 861–865.

Jodo, E., Kayama, Y., 1992. Relation of a negative ERPcomponent to response inhibition in a GoyNo-go task.Electroencephalography and Clinical Neurophysiology 82,477–482.

Kiefer, M., Marzinzik, F., Weisbrod, M., Scherg, M., Spitzer,M., 1998. The time course of brain activations during

response inhibition: evidence from event-related potentialsin a goyno go task. Neuroreport 9, 765–770.

Killian, G.A., Holzman, P.S., Davis, J.M., Gibbons, R., 1984.Effects of psychotropic medication on selected cognitiveand perceptual measures. Journal of Abnormal Psychology93, 58–70.

Konishi, S., Nakajima, K., Uchida, I., Kikyo, H., Kameyama,M., Miyashita, Y., 1999. Common inhibitory mechanism inhuman inferior prefrontal cortex revealed by event-relatedfunctional MRI. Brain 122, 981–991.

Kopp, B., Mattler, U., Goertz, R., Rist, F., 1996. N2, P3 andthe lateralized readiness potential in a nogo task involvingselective response priming. Electroencephalography andClinical Neurophysiology 99, 19–27.

Mayberg, H.S., 1997. Limbic-cortical dysregulation: a pro-posed model of depression. Journal of Neuropsychiatry andClinical Neurosciences 9, 471–481.

Mayberg, H.S., Liotti, M., Brannan, S.K., McGinnis, S.,Mahurin, R.K., Jerabek, P.A., Silva, J.A., Tekell, J.L.,Martin, C.C., Lancaster, J.L., Fox, P.T., 1999. Reciprocallimbic-cortical function and negative mood: converging PETfindings in depression and normal sadness. American Jour-nal of Psychiatry 156, 675–682.

Merriam, E.P., Thase, M.E., Haas, G.L., Keshavan, M.S.,Sweeney, J.A., 1999. Prefrontal cortical dysfunction indepression determined by Wisconsin Card Sorting Testperformance. American Journal of Psychiatry 156, 780–782.

Mialet, J.P., Pope, H.G., Yurgelun-Todd, D., 1996. Impairedattention in depressive states: a non-specific deficit? Psy-chological Medicine 26, 1009–1020.

Norman, D.A., Shallice, T., 1986. Attention to action: willedand automatic control of behavior. In: Davidson, R.,Schwartz, G., Shapiro, D.(Eds.), Consciousness and Self-Regulation. Plenum Press, New York, pp. 1–18.

Ogura, C., Nageishi, Y., Omura, F., Fukao, K., Ohta, H.,Kishimoto, A., Matsubayashi, M., 1993. N200 componentof event-related potentials in depression. Biological Psychi-atry 33, 720–726.

Oldfield, R.C., 1971. The assessment and analysis of handed-ness: the Edinburgh Inventory. Neuropsychologia 9, 97–113.

Pfefferbaum, A., Ford, J.M., Weller, B.J., Kopell, B.S., 1985.ERPs to response production and inhibition. Electroen-cephalography and Clinical Neurophysiology 60, 423–434.

Posner, M.I., Dehaene, S., 1994. Attentional networks. Trendsin Neurosciences 17, 75–79.

Posner, M.I., DiGirolamo, G., 1998. Executive attention: con-flict, target detection and cognitive control. In: Parasuraman,R. (Ed.), The Attentive Brain. MIT Press, Cambridge, pp.401–423.

Roberts, L.E., Rau, H., Lutzenberger, W., Birbaumer, N., 1994.Mapping P300 waves onto inhibition: GoyNo-Go discrimi-nation. Electroencephalography and Clinical Neurophysiol-ogy 92, 44–55.

Sara, G., Gordon, E., Kraiuhin, C., Coyle, S., Howson, A.,Meares, R., 1994. The P300 ERP component: an index ofcognitive dysfunction in depression? Journal of AffectiveDisorders 31, 29–38.


Simson, R., Vaughan Jr., H.G., Ritter, W., 1977. The scalptopography of potentials in auditory and visual GoyNoGotasks. Electroencephalography and Clinical Neurophysiology43, 864–875.

Smith, E.E., Jonides, J., 1999. Storage and executive processesin the frontal lobes. Science 283, 1657–1661.

Strik, W.K., Fallgatter, A.J., Brandeis, D., Pascual-Marqui,R.D., 1998. Three-dimensional tomography of event-relatedpotentials during response inhibition: evidence for phasicfrontal lobe activation. Electroencephalography and ClinicalNeurophysiology 108, 406–413.

Stuss, D.T., Shallice, T., Alexander, M.P., Picton, T.W., 1995.A multidisciplinary approach to anterior attentional func-tions. Annals of the New York Academy of Sciences 769,191–211.

Trichard, C., Martinot, J.L., Alagille, M., Masure, M.C., Hardy,P., Ginestet, D., Feline, A., 1995. Time course of prefrontallobe dysfunction in severely depressed in-patients: a longi-tudinal neuropsychological study. Psychological Medicine25, 79–85.

Videbech, P., 2000. PET measurements of brain glucosemetabolism and blood flow in major depressive disorder: acritical review. Acta Psychiatrica Scandinavica 101, 11–20.

Weisbrod, M., Kiefer, M., Marzinzik, F., Spitzer, M., 2000.Executive control is disturbed in schizophrenia: evidencefrom event-related potentials in a GoyNoGo task. BiologicalPsychiatry 47, 51–60.

Wittchen, H.U., Wunderlich, U., Gruschwitz, S., Zaudig, M.,1997. Strukturiertes Klinisches Interview fur DSM-IV.¨Hogrefe, Gottingen.¨

executive control deficit in depression: event-related potentials in a go/nogo task

Documents