Three planar views of functionalmagnetic resonance image activations for the β images of NoGo (left)and Go (middle) activations for healthy subjects (upper row) and patientswith schizophrenia (lower row). The sections were chosen to illustrate maximalactivations on the axial, coronal, and sagittal cuts. The event-related brainpotentials (ERPs) from the Fz, Cz, and Pz electrode sites are overlaid forthe appropriate β images. The contrast images for NoGo-Go activationsare shown on the right. Positivity in the ERPs is plotted up. Images on rightof the figure are from the right side of the brain.
The NoGo functional magnetic resonanceimaging (fMRI) β image (left) (approximately 12-16 seconds) and NoGoevent-related brain potentials (ERPs) (middle) (P300 lasts approximately 200milliseconds) for healthy control subjects, as shown in Figure 1. The NoGoP300 is circled. Correlations between fMRI activations and P300 amplitudes(recorded at the Fz [red blobs], Cz [green blobs], and Pz [yellow blobs] electrodesites) appear as activations on the right (P<.01 uncorrected;contiguous voxels = 6). Scatter plots and correlation coefficients appearon the far right, and fMRI effects represent the average value for that region.Centroids are given in Talairach coordinates (millimeters).
The NoGo functional magnetic resonanceimaging (fMRI) β image (left) (approximately 12-16 seconds) and NoGoevent-related brain potentials (ERPs) (top right) (P300 lasts approximately200 milliseconds) for patients with schizophrenia, as shown in Figure 2. TheNoGo P300 is circled. Correlations between fMRI activations and P300 amplitudes(recorded at the Fz [red blobs], and Pz [yellow blobs] electrode sites) appearas activations on the middle right (P<.01 uncorrected; contiguousvoxels = 6). A scatter plot and correlation coefficient appear on the lowerright, and fMRI effects represent the average value for that region. Centroidsare given in Talairach coordinates (millimeters).
Judith M. Ford, Max Gray, Susan L. Whitfield, And U. Turken, Gary Glover, William O. Faustman, Daniel H. Mathalon. Acquiring and Inhibiting Prepotent Responses in SchizophreniaEvent-Related Brain Potentials and Functional Magnetic ResonanceImaging. Arch Gen Psychiatry. 2004;61(2):119–129. doi:10.1001/archpsyc.61.2.119
Schizophrenia is associated with deficits in using context to establish
prepotent responses in complex paradigms and failures to inhibit prepotent
responses once established.
To assess prepotent response establishment and inhibition in patients
with schizophrenia using event-related brain potential (ERP) and functional
magnetic resonance imaging (fMRI) in a simple NoGo task. To combine fMRI and
ERP data to focus on fMRI activations associated with the brief (approximately
200 ms) moment of context updating reflected in the NoGo P300 ERP component.
Design and Setting
We collected ERP and fMRI data while subjects performed a NoGo task
requiring a speedy button press to X stimuli (P =
.88) but not to K stimuli (P = .12). The ERPs were
collected at the Veterans Affairs Palo Alto Health Care System, Palo Alto,
Calif; fMRIs were collected at Stanford University, Stanford, Calif.
We recruited patients with DSM-IV schizophrenia
(n = 11) from the community and the VA hospital and sex- and age-matched healthy
control subjects (n = 11) from the community.
Main Outcome Measures
Behavioral accuracy, P300 amplitudes and latencies, and fMRI activations
suggested that patients with schizophrenia did not establish as strong a prepotent
tendency to respond to the Go stimulus as healthy subjects. In healthy subjects,
NoGo P300 was related to activations in the anterior cingulate cortex, dorsal
lateral prefrontal cortex, and right inferior parietal lobule and caudate
nucleus, perhaps reflecting conflict experienced when withholding a response,
control needed to inhibit a response, and stopping a response in action, respectively.
In patients with schizophrenia, NoGo P300 was modestly related to activations
in the anterior cingulate cortex, which is consistent with experiencing conflict.
The difference in ERP and fMRI responses to Go and NoGo stimuli suggested
that inhibiting a response was easier for patients with schizophrenia than
for healthy subjects. Correlations of P300 and fMRI data suggested that patients
with schizophrenia and healthy subjects used different neural structures to
inhibit responses, with healthy subjects using a more complex system.
Schizophrenia has been associated with failures of inhibitory controlin some but not all experimental tasks. Tasks in which prepotent responsesare reflexive, overlearned, or automatic consistently reveal response inhibitiondeficits in patients with schizophrenia. For example, when the reflex to movethe eyes toward a light (prosaccade) must be inhibited and the eyes must bewillfully moved in the opposite direction (antisaccade), patients with schizophreniahave inordinate difficulty suppressing prosaccades.1,2 Whenpresented with the word red written in green inkduring the Stroop color test, patients with schizophrenia have difficultyinhibiting the overlearned tendency to read the word when their task is toname the ink color.3,4
When prepotent response biases are newly learned during an experiment,healthy subjects have more difficulty inhibiting these responses than patientswith schizophrenia, perhaps because patients with schizophrenia do not usecontext to acquire prepotent responses as readily, as is the case with theAX version of the continuous performance task.5- 8 Thefailure to establish new prepotent responses in patients with schizophreniais not absolute; they have a high rate of perseverative errors on the WisconsinCard Sorting Task.9,10 Both tasksinvolve fairly complex yet very different stimulus-response mapping rulesrequiring varying degrees of working memory, learning, set-shifting, and responseinhibition. Whether patients with schizophrenia show deficits in developingprepotent response tendencies or in inhibiting them when cognitive demandsare minimized can be addressed more simply in a NoGo task where subjects pressa button to one stimulus (Go) but not to the other (NoGo). Even in this simpletask, patients with schizophrenia sometimes make more false-alarm errors toNoGo stimuli than healthy subjects11,12 butnot always.13,14 Nevertheless,when combined with brain imaging tools, this task can assess whether patientswith schizophrenia establish prepotent response biases and which neural structuresare recruited to inhibit these responses.
Two in vivo, noninvasive brain-imaging methods can be used to understandneural responses to NoGo stimuli, electrophysiological and hemodynamic. Electrophysiologicalmethods, using electroencephalography, allow real-time measures of neuronalactivity with millisecond temporal resolution. Individual electroencephalogramsare averaged to produce an event-related potential (ERP) whose componentsdevelop and resolve within tens or hundreds of milliseconds. A less directmeasure of neural activity is hemodynamic brain imaging, the most common ofwhich is functional magnetic resonance imaging (fMRI). This operates on amuch more delayed time scale; depending on age, task, and brain region, itmay take up to 16 seconds to develop and resolve.15 However,it has superior spatial resolution, allowing a more precise delineation ofbrain structures and circuits activated during specific tasks.
The ERP has been used for many years to study inhibition of prepotentresponses.16 At approximately 400 ms followinga NoGo stimuli, a positive ERP component lasting about 200 ms peaks; it iscalled the NoGo P300. The Go and NoGo P300s both may reflect context updating17 necessary for successful ongoing execution and inhibitionof prepotent responses. When context has been established to respond to almostevery trial (Go), the context does not need to be updated until an exceptionto the context is presented (NoGo), resulting in a relatively small Go P300and a large NoGo P300. However, if context to respond or "go" has not beenestablished, relatively large Go P300s and relatively small NoGo P300s willbe elicited.
Although associated with sensorimotor inhibition,18 NoGoP300 cannot be a direct reflection of motor inhibition; it is elicited byNoCount stimuli that are equiprobable with Count stimuli,16 andit is unaffected by motor response priming.19
Although a NoGo P300 is elicited when Go and NoGo stimuli are equiprobable,20 probability affects NoGo P300 in the same way itaffects the Go P300. The more improbable the NoGo stimulus, the larger theNoGo P300 amplitude,21 reflecting the prepotencyof the Go response and the difficulty inhibiting it. An improbable Go (target)stimulus elicits a large P300 with a parietal maximum, whereas an equallyimprobable NoGo stimulus elicits a P300 with a central frontal scalp distribution.22
Because of its association with frontal lobe function and because patientswith schizophrenia are known to have wide-reaching functional and structuralfrontal lobe deficits,23- 25 NoGoP300s should be affected by schizophrenia. Surprisingly, few NoGo ERP studieshave been done on patients with schizophrenia. These have had varying degreesof success, perhaps because of the peculiarities of the population studied,12 the stimulation and acquisition parameters,11 and demands on working memory.26
While there are many fMRI studies of inhibitory control broadly defined,there are relatively few studies using a simple NoGo paradigm. However, amongthe studies reported, there is remarkable concordance in confirming the ERPliterature that preceded them. NoGo stimuli activate frontal lobe structuresincluding the anterior cingulate cortex (ACC),20,27- 29 premotorcortex,20,27,28 dorsolateralprefrontal cortex (DLPFC),4,20,27- 30 andposterior right frontal inferior cortex.31 Righthemisphere structures are more activated than left, particularly in the middleand inferior frontal gyri, frontal limbic area, anterior insula, and inferiorparietal lobule (IPL).32 In addition, caudatenucleus activation28 and temporal,4,20,30 parietal,4,20,27,28 andvisual4,27,30 corticalactivations have been noted. Some of these fMRI findings may be affected bythe inclusion of errors when block designs were used for analysis27,28 and by the inclusion of low-probabilityevents.4,20 Although eliminatingerror trials is important, balancing probabilities between Go and NoGo stimuliworks against the processes we most want to understand: the establishmentof prepotent responses and their successful inhibition.
Few NoGo studies using fMRI have been done in patients with schizophrenia.Although patients performed normally in NoGo and Stop signal tasks, they showedreduced activation in the left ACC during both tasks and reduced left rostralDLPFC activation during the Stop task.13 However,because error trials were included in the analysis and because patients withschizophrenia have different neural responses to errors,5,33,34 thesefindings may not be specific to schizophrenia-related differences in responseinhibition.
To maximally engage executive control, we attempted to establish a strongprepotent bias to respond to Go stimuli. To build up expectancy for Go stimuli,we skewed stimulus probabilities (Go stimuli = 88%; NoGo stimuli = 12%), amanipulation that has proved successful in activating frontal lobe structuresassociated with executive control in other response inhibition tasks.35 In addition, we pretrained subjects to respond tothe stimulus that subsequently became the NoGo stimulus, and we emphasizedspeed rather than accuracy.
Hemodynamic activity associated with NoGo stimuli is likely to reflectmany processes including sensation, perception, attention, response selection,response inhibition, response monitoring, self-evaluation, planning for thenext trial, and any number of other processes happening in the 4 to 8 secondsit takes for the hemodynamic response to peak and the subsequent 6 to 8 secondsit takes to return to its baseline state. To focus on fMRI activations inthose brain regions associated with context updating, we combined NoGo P300data with NoGo fMRI data recorded from the same subjects in the same paradigm.
We predicted that patients with schizophrenia would have a smaller differencebetween NoGo and Go P300s than healthy subjects and a smaller difference betweenNoGo and Go fMRI activations.12,13 Whetherthis was related to reduced neural activity subserving response inhibitionto NoGo stimuli or excessive neural activity subserving response executionto Go stimuli was assessed by separate analyses of Go and NoGo responses.Compared with healthy control subjects, we predicted that Go stimuli wouldbe associated with greater effort than NoGo stimuli in patients with schizophreniaowing to their deficient use of context to establish prepotent response tendencies.Thus, we expected patients with schizophrenia would exhibit relatively moreomission than false-alarm errors, relatively larger Go than NoGo P300s, andgreater fMRI activation to Go than NoGo stimuli. Finally, by correlating NoGoP300 and NoGo fMRI data, we focused on fMRI activations associated with thebrief moment (approximately 200 milliseconds) associated with context updating.
In separate sessions, we recorded ERP and fMRI data while 11 patientswith DSM-IV36 schizophreniaand 11 healthy comparison subjects performed a NoGo task. All gave writteninformed consent after the procedures had been fully described. Demographicand clinical data are included in Table1.37,38
Patients with schizophrenia were recruited from community mental healthcenters, as well as from inpatient and outpatient services of the Palo AltoVeterans Affairs Healthcare System, Palo Alto, Calif. All patients with schizophrenia,who were taking stable doses of antipsychotic medications, met DSM-IV criteria for schizophrenia based either on the diagnosis froma Structured Clinical Interview for DSM-IV conductedby a psychiatrist or psychologist or by consensus of a Structured ClinicalInterview for DSM-IV conducted by a trained researchassistant and a clinical interview by a psychiatrist or psychologist. In 1case, a psychiatrist made the diagnosis by reviewing the patient's medicalrecord. Prospective patient and control participants were excluded if theymet DSM-IV criteria for alcohol or drug abuse within30 days prior to the study. In addition, patient and control participantswere excluded for significant head injury and neurological or other medicalillnesses compromising the central nervous system. Patient symptoms were assessedusing the 18-item Brief Psychiatric Rating Scale.39,40
Comparison subjects were recruited by newspaper advertisements and wordof mouth, screened by telephone using questions from the Structured ClinicalInterview for DSM-IV,41 andexcluded for any significant history of Axis I psychiatric illness.
Subjects viewed an irregular sequence of K (12%) and X (88%) stimuli,presented for 100 milliseconds each. The stimulus onset asynchrony was 1,2, or 3 seconds, with each occurring with equal probability. The intervalbetween 2 K stimuli varied between 7 and 24 seconds.30
Participants lifted a lever attached to the index finger of their responsehand each time an X stimulus was presented and withheld responding to anyK stimuli. There were 42 K stimuli and 288 X stimuli. (Because of a PsyScope[software developed by Cohen et al42] buffererror, the last third of the trials in the fMRI environment were presentedat a constant 2-second interstimulus interval. These trials were omitted fromthe analysis of the fMRI data. The same error was not present in the ERP environmentwhere the stimulus presentation was controlled by STIM software in Neuroscan[Neuroscan, El Paso, Tex.]) To increase the prepotent tendency to respondto K stimuli, we pretrained subjects to respond to K stimuli and not X stimuliin an oddball target detection task. Subjects were told to go as fast as possibleand if they made errors, to keep going and not slow down. All but 1 subjectmade right-handed responses. (All subjects responded with their right handsexcept for the left-handed patient. The results of the analysis were not changedwhen she was eliminated from the analysis. Because of considerations of power,her data have been included in the analysis presented herein.)
A pressure-sensitive piezoelectric transducer that produced a continuousmeasure of response activity and was sensitive to vigor or acceleration ofthe response was used to record motor responses. Thus, a slow and weak buterroneous response to a K stimulus might register as a response. These trialswere eliminated by setting a very low criterion for a motor response (>15%of the rolling average amplitude of 20 surrounding trials). A very brisk butpartial response might also register as a response, as could small fingertwitches and any pressure changes against the device.
Participants were seated in a sound-attenuating, electrically shieldedbooth. Electroencephalography data recorded from the F3, Fz, F4, C3, Cz, C4,P3, Pz, and P4 electrode sites are reported herein. Vertical electrooculogramdata were recorded from electrodes placed above and below the right eye, andhorizontal electrooculogram data were recorded from electrodes placed at theouter canthus of each eye. Data were sampled at 500 Hz and bandpass filteredat 0.05 to 30 Hz.
Before baseline correction, single-trials were corrected for eye blinkand eye movement artifacts based on correlations between electroencephalogramsrecorded at each electrode site and vertical and horizontal electrooculograms43. Trials exceeding +100 µV were then rejected.The linear component of each averaged ERP from –100 to 1000 millisecondswas removed before peaks were identified and measured. Only trials with correctresponses to X stimuli or successful inhibition of responses to K stimuliwere included in the ERPs.
The P300 peak was identified as the maximum positive voltage between280 and 600 milliseconds. Its amplitude was quantified as the average voltagearound the peak (+50 milliseconds) relative to a 100 milliseconds prestimulusbaseline. Before P300 was measured, data were low-pass filtered at 12 Hz.
Univariate repeated-measures analyses of variance were performed forP300 amplitude and latency for the following factors: group (comparison subjects,patients with schizophrenia), stimulus (K stimulus vs X stimulus), anteriorposterior scalp site (frontal, central, parietal), and lateral scalp site(left, middle, right).
Images were acquired on a GE 3 Tesla magnetic resonance imaging scanner(General Electric, Milwaukee, Wis) using a custom-made head coil with a spiralgradient echo sequence.44 Subjects were stabilizedwith a bite bar made from their dental impression.
Image processing was performed with statistical parametric mapping (SPM99;Wellcome Department of Cognitive Neurology, London, England).
Rests (the first and last 14 trials) and trials associated with a PsyScopebuffer error were not included in this analysis. X stimuli hits, X stimuliomissions, K stimuli successful inhibitions, and K stimuli false alarms weremodeled for all but the 5 subjects who had no X stimuli omissions. Analyseswere done in 2 stages. First, per subject per voxel β estimates werecomputed, producing brain maps of parameters for all of the explanatory variables.Second, a random-effects model was applied to individual subject images derivedduring first-level analyses separately for healthy subjects and patients withschizophrenia. Specific responses to Go and NoGo stimuli were assessed byseparate examination of Go and NoGo β images (reflecting the peak amplitudeof the fitted hemodynamic response), and group contrasts for Go and NoGo βimages were each performed separately. In addition, NoGo-Go and Go-NoGo contrastswere estimated separately for patients with schizophrenia and healthy subjects,using 2-sample t tests. Group contrasts of thesecontrasts were also estimated.
We correlated NoGo fMRI values at each voxel with the NoGo P300 amplitudeat Fz, Cz, and Pz electrode sites entered as the covariate in the simple regressionoption in SPM99. This was done separately for both groups.
As presented in Table 2,patients with schizophrenia had a higher overall error rate than healthy subjects.Across ERP and fMRI, patients with schizophrenia made a lower percentage offalse-alarm errors to NoGo stimuli than healthy subjects (50.4% vs 68.2%)and a higher percentage of omission errors to Go stimuli, which is consistentwith a failure to form a strong prepotent bias to respond to Go stimuli. Poorerperformance during fMRI than ERP could be due to fMRI testing preceding ERPtesting for most of the subjects or to the awkward posture, restriction ofmovement, ambient noise, and anxiety associated with fMRI.
The ERPs to Go and NoGo stimuli are shown in Figure 1. (Apparent group X stimulus differences in the N1 and N2ERP components were not significant, which is consistent with other comparisonsof patients with schizophrenia and control subjects in NoGo tasks [Kiehl etal12].) P300 areas were compared in a 4-wayanalysis of variance for stimulus, anterior posterior scalp site, laterality,and group (Table 3). There wasa significant main effect of stimulus (F1,20 = 30.84; P<.001) with NoGo stimuli eliciting larger P300s across all scalpsites than Go stimuli, and there was an interaction of group X stimulus (F1,20 = 5.14; P<.04) reflecting a smallerstimulus effect in patients with schizophrenia (2.4 µV) than in healthysubjects (5.7 µV). To determine whether the reduced NoGo-Go effect inpatients with schizophrenia was due to larger P300s to Go or smaller P300sto NoGo, the group effect was assessed for Go and NoGo P300 separately. WhileGo P300 was larger in patients with schizophrenia (4.8 µV) than in healthysubjects (3.4 µV) (F1,20 = 1.47; P =.24) and NoGo P300 was smaller in patients with schizophrenia (7.2 µV)than in healthy subjects (9.1 µV) (F1,20 = 1.03; P = .32), neither was significant.
To determine whether reduced NoGo-Go P300 amplitude differences seenin patients with schizophrenia could be attributed to symptom severity, theywere regressed against total Brief Psychiatric Rating Scale score. Fewer symptomaticpatients with schizophrenia had larger NoGo-Go P300 differences (r = −0.62; P<.04), suggesting thatclinical severity contributed to attenuated differences in NoGo vs Go stimulusprocessing.
Analysis of variance of P300 latencies revealed a significant groupeffect (F1,20 = 6.23; P<.03) with P300being earlier in healthy subjects than in patients with schizophrenia anda significant stimulus effect (F1,20 = 19.38; P<.001) with P300 being later to NoGo (461 ms) than to Go (411 ms)stimuli (Table 3). Importantly,there was a trend toward a significant group X stimulus interaction (F1,20 = 4.11; P<.06) due to the larger differencebetween Go and NoGo responses in healthy subjects (73 ms) than in patientswith schizophrenia (27 ms), again suggesting that patients with schizophreniaresponded more similarly to Go and NoGo stimuli.
No activated voxels reached a corrected significance level of P<.05 when adjusted for the entire volume for the NoGo-Gocontrast. Instead, we used a height threshold of P<.01(uncorrected) and an extent threshold of 6. Significant gray matter voxel–levelactivations are reported. (To determine if patients with schizophrenia movedmore than comparison subjects, as has sometimes been reported,45 movementparameters derived during the realignment step of the analysis were comparedin two 2-way analyses of variance for group and dimension [one analysis X,Y, Z and one for pitch, roll, yaw]. Movements in the Z plane were significantlygreater than in the X and Y planes [P<.001], butnone of the movements was affected by group.)
Many brain regions were more activated to NoGo than Go stimuli in healthycontrol subjects, reflecting a combination of response conflict, responseinhibition, stimulus improbability, and task relevance (eg, DLPFC, ACC, IPL,basal ganglia), particularly in the right hemisphere. These can be seen in Figure 1 and are listed in Table 4.46 To understand which brainregions were more active to Go than NoGo stimuli, the Go-NoGo contrast wasestimated. No voxels survived our threshold, indicating healthy control subjectsexpended very little neural energy making the automatic prepotent responseto Go stimuli relative to inhibiting responses to NoGo stimuli. When we droppedthe threshold to P<.05, activations were seenin the left somatosensory cortex, confirming motor response involvement.
That there was greater activation to NoGo than to Go stimuli is alsoreflected in the β images (Figure 1)(Table 4), which indicate activatedvoxels with β values significantly different from zero. NoGo and Go stimuliwere associated with activation of 1098 and 100 gray matter voxels, respectively.The majority (66%) of activated voxels to the Go stimulus were in the leftsomatosensory or motor cortex, reflecting the motor response.
The NoGo-Go contrast revealed modest activations (106 gray matter voxels),mostly in the right frontal and parietal cortices (Figure 1) (Table 4).With the reverse contrast (Go-NoGo), 61 gray matter voxels were activatedin the patients with schizophrenia, indicating relatively greater recruitmentof effortful processes compared with healthy control subjects.
That there was greater activation to Go (222 gray matter voxels) thanNoGo (69 gray matter voxels) stimuli is also reflected in the β imagesshown in Figure 1. About 33% ofvoxel activations to the Go stimulus were in the left somatosensory or motorcortex, reflecting the motor response; however, about 60% were in regionsassociated with target detection (frontal, temporal, and parietal lobes, andinsula and thalamus).
As presented in Table 4,healthy control subjects had greater activations than patients with schizophreniafor the NoGo-Go contrast, predominantly in brain regions activated in healthysubjects. No voxels were more activated in patients with schizophrenia thanhealthy subjects for this contrast. With the reverse contrast (Go-NoGo), patientswith schizophrenia had greater activations than healthy control subjects,with 177 gray matter voxels being activated in the somatosensory and motorcortex, ACC, DLPFC, striatum, and insula. No voxels were more activated incontrol subjects than patients with schizophrenia for this contrast.
β images to NoGo stimuli were contrasted for patients with schizophreniaand healthy subjects (Table 4).Healthy subjects had significantly more activation than patients with schizophrenia,but no voxels were more activated in patients with schizophrenia than healthysubjects for this contrast.
β images to Go stimuli were contrasted for patients with schizophreniaand healthy subjects (Table 4).Patients with schizophrenia had significantly more activation than healthysubjects in the thalamus, IPL, and caudate nucleus. Only 2 voxels (in Brodmannarea 8) were more activated in healthy subjects than patients with schizophreniafor this contrast.
In healthy subjects, larger NoGo P300s were associated with greaterNoGo fMRI activations centered in the ACC, DLPFC, IPL, and caudate nucleus,depending on the ERP electrode site (Figure2). In patients with schizophrenia, the same analysis revealed modestcorrelations in the ACC for NoGo P300 recorded from the Pz electrode site(Figure 3).
Our intention was to establish a strong prepotent tendency to respondon every trial, thereby increasing the difficulty of response inhibition toinfrequently occurring NoGo stimuli in an otherwise cognitively simple NoGotask. Based on the behavioral data, we achieved this goal in healthy subjectswho rarely failed to respond to Go stimuli but who often failed to inhibitresponses to NoGo stimuli. Patients with schizophrenia did not establish asstrong an automatic or prepotent response bias and instead may have fullyevaluated each stimulus on each trial, making a deliberate choice to respondor not to respond. This was evident in their greater percentage of omissionerrors than false-alarm errors, consistent with previous findings by Carteret al,5 Barch et al,6 Heniket al,7 and Servan-Schreiber et al8 demonstrating that patients with schizophrenia aredeficient in the use of context to establish prepotent response biases.
As expected, compared with Go stimuli, NoGo stimuli activated more right(957 voxels) than left (474 voxels) hemisphere structures,31,47 aswell as many voxels in the frontal lobe including the ACC,20,27- 29 DLPFC,4,20,27- 30 andinferior frontal cortex.31 In addition, weconfirmed greater temporal lobe,4,20,30 parietallobe,4,20,27,28 andcaudate nucleus28 activations to NoGo comparedwith Go stimuli. Although ACC activation has also been associated with thecommission of errors,30 because of our extrememeasures to eliminate errors and partial errors from the analysis, it is unlikelythat ACC involvement in this case reflects the contribution of errors. Insteadit may reflect monitoring for conflict and the detection of potential forerror48 or simply the infrequency of the NoGostimulus.4 Recent data from Milham et al49 suggest that the ACC monitors for the presence ofcompeting or conflicting actions in an effort to prevent execution of erroneousmotor actions. On detecting conflict, a signal may be sent to the frontallobe structures to implement top-down executive control to stop the initiationof an inappropriate motor response.35 In thistask, the final effort to stop responses in progress, as observed in Stopsignal tasks, may be responsible for both caudate nucleus27 andIPL activations.50
As predicted, patients with schizophrenia had a smaller difference infMRI activation to NoGo compared with Go stimuli than healthy subjects, althoughthe same brain regions (right frontal and parietal) were activated in bothgroups. This reduced difference in patients with schizophrenia can be attributedboth to greater activation to Go stimuli and less activation to NoGo stimuli,consistent with behavioral data suggesting Go responses were more effortfuland deliberate for patients with schizophrenia than healthy subjects and thatNoGo responses were not as difficult for patients with schizophrenia to inhibit.Importantly, most of the activations associated with Go stimuli in patientswith schizophrenia were in brain regions typically associated with effortfultarget detection and implicated in generating visual target P300s.51- 53
As expected,21 infrequent NoGo stimulielicited larger and later P300s than frequent Go stimuli in healthy subjects.Patients with schizophrenia showed less P300 amplitude and latency distinctionbetween NoGo and Go stimuli, suggesting that they processed Go and NoGo stimulisimilarly. This represents a relatively new finding. Few NoGo studies havebeen done in patients with schizophrenia, and those that have been done aredifficult to interpret owing to the peculiarities of populations studied12 and differences in recording parameters.11 Importantly, P300 data are consistent with fMRI datain suggesting that reduced differences in patients with schizophrenia weredue to both larger Go P300s and smaller NoGo P300s, although neither effectwas significant.
The behavioral, fMRI, and ERP data all suggest that the NoGo task engageddifferent response inhibition strategies in patients with schizophrenia andhealthy subjects. This conclusion is underscored by the correlation analysisof ERP and fMRI data, which allowed us to focus narrowly on those brain activationsassociated with the brief (approximately 200 milliseconds) moment of contextupdating following a NoGo stimulus. Specifically, correlations between ERPand fMRI data were found in the ACC, DLPFC, caudate nucleus, and right IPL.Healthy subjects with larger NoGo P300 amplitudes had greater NoGo fMRI activationsin these brain regions associated with executive control. The sequencing ofthese structures in the service of response inhibition awaits refinementsin methods to detect small temporal differences in fMRI activations. The correlationsin patients with schizophrenia were different. Patients with larger NoGo P300shad larger NoGo fMRI activations only in the ACC, suggesting the experienceof conflict is associated with the elicitation of the NoGo P300. That patientswith schizophrenia who had a more normal pattern of P300s did not have a morenormal pattern of fMRI response may reflect their inability to recruit DLPFC,IPL, and striatum, even when a normal strategy is attempted.
While large regions of brain were activated by NoGo stimuli, especiallyin healthy subjects, very few of these regions were correlated with NoGo P300.Instead, these must subserve processes other than those related to contextupdating reflected in the NoGo P300 component following response inhibition.This correlational technique does not provide source localization for ERPcomponents but rather capitalizes on individual differences in neural (ERP)and hemodynamic (fMRI) responses, identifying areas of the brain that havegreater hemodynamic responses in people who have larger ERPs. Our resultssuggest that healthy subjects set up prepotent response biases and when responseshave to be inhibited, effort must be expended. This ongoing effort, reflectedin NoGo P300 amplitude, is associated with the engagement of neural structuresassociated with executive control. Patients with schizophrenia, however, didnot set up strong prepotent response biases, and their NoGo P300s insteadreflect the simple experience of conflict associated with ACC activation.
This analysis is an example of how we might start to combine ERP andfMRI data to generate hypotheses about different task strategies used by neuropsychiatricpopulations and the neural structures recruited to implement them. However,our conclusions are limited by the small sample sizes studied, the low probabilitythreshold used, the fact that patients with schizophrenia were all medicated,and the fact that subjects used their preferred hand for responding. Also,because we did not correct for multiple comparisons, these findings need tobe replicated in a new sample.
Corresponding author and reprints: Judith M. Ford, PhD, Departmentof Psychiatry and Behavioral Sciences, Stanford University School of Medicine,Stanford, CA 94305-5550 (e-mail: email@example.com).
Submitted for publication January 6, 2003; final revision received June4, 2003; accepted June 12, 2003.
This study was supported by grants MH40052 and MH 58262 from the NationalInstitute of Mental Health, Bethesda, Md (Dr Ford), and grants from the Departmentof Veterans Affairs, Palo Alto, Calif.
We acknowledge Mark Rothrock, MD, for recruiting patient volunteersto our studies and Margaret J. Rosenbloom, MA, for help preparing the manuscript.