Common Neural Substrates for Response Selection across ...

Common Neural Substrates for Response Selection acrossModalities and Mapping Paradigms

Yuhong Jiang1 and Nancy Kanwisher12

Abstract

amp In many situations people can only compute one stimulus-to-response mapping at a time suggesting that responseselection constitutes a lsquolsquocentral processing bottleneckrsquorsquo inhuman information processing Using fMRI we tested whethercommon or distinct brain regions were involved in responseselection across visual and auditory inputs and across spatialand nonspatial mapping rules We isolated brain regionsinvolved in response selection by comparing two conditionsthat were identical in perceptual input and motor output butdiffered in the complexity of the mapping rule In the visual ndashmanual task of Experiment 1 four vertical lines were positionedfrom left to right and subjects pressed one of four keys to reportwhich line was unique in length In the auditoryndashmanual task ofExperiment 2 four tones were presented in succession and

subjects pressed one of four keys to report which tone wasunique in duration For both visual and auditory tasks themapping between target position and key position was eitherspatially compatible or incompatible In the verbal task ofExperiment 3 subjects used nonspatial mappings that wereeither compatible (lsquolsquosamersquorsquo if colors matched lsquolsquodifferentrsquorsquo if theymismatched) or incompatible (the opposite) Extensive activa-tion overlap was observed across all three experiments forincompatible versus compatible mapping in bilateral parietaland frontal regions Our results indicate that common neuralsubstrates are involved in response selection across inputmodalities and across spatial and nonspatial domains ofstimulus-to-response mapping consistent with behavioralevidence that response selection is a central process amp

INTRODUCTION

Major research efforts in cognitive neuroscience haveaddressed the neural mechanisms underlying both theperceptual analysis of stimuli and the preparation ofmotor responses However successful interaction withthe world requires that perception and action be linkedThe present investigation focuses on the process ofmapping stimuli onto responses (called lsquolsquoresponse selec-tionrsquorsquo [RS])

A substantial literature in cognitive psychology hassuggested that RS represents a fundamental bottleneckin human information processing (for a review seePashler 1994 1998) Dual-task interference studies onchoice RT tasks indicate that several perceptual process-es can be conducted in parallel without interference(Treisman amp Davies 1973) and several motor responsescan be produced simultaneously without interference(Pashler 1990) but under most conditions only a singlestimulus-to-response (SndashR) mapping can be computedat a time (Pashler 1984) Importantly this lsquolsquocentralprocessing bottleneckrsquorsquo is general across different mo-dalities of input and output Thus while we can see ashape and hear a tone at the same time and we can

press a key and say a word at the same time we cannotsimultaneously determine which key to press on thebasis of the shape and which word to say on the basis ofthe tone Exceptions to this rule occur under somecircumstances when the SndashR mapping is highly com-patible (Pashler Carrier amp Hoffman 1993) after sub-stantial practice (Schumacher et al 1999 SchumacherSeymour Glass Kieras amp Meyer 2001) or when sub-jects are encouraged to adopt a lsquolsquodaringrsquorsquo responsestrategy (Schumacher et al 2001) While the implica-tions of these exceptions on the nature of the bottle-neck is subject to debate (Ruthruff Pashler amp Klaassen2001 Schumacher et al 1999 2001 Meyer et al 1995Meyer amp Kieras 1997 Pashler 1994) the persistence ofdual-task interference is a robust finding

Despite considerable behavioral evidence that RS mayform a central processing bottleneck this property ofhuman cognition has been largely ignored by research-ers in cognitive neuroscience Indeed when consideredfrom a neuroanatomical perspective the very idea of thecentral bottleneck represents a puzzle It seems reason-able to assume that two processes can interfere witheach other only if they engage common neural mecha-nisms (or different but interacting mechanisms) Butwhat neural mechanisms are shared (or interact) be-tween the two tasks of pressing a key based on a visualshape and uttering a word based on a tone

1 Massachusetts Institute of Technology 2 Athinoula A MartinosCenter for Biomedical Imaging Charlestown

copy 2003 Massachusetts Institute of Technology Journal of Cognitive Neuroscience 158 pp 1080ndash1094

If RS is a central cognitive process then RS involvingdifferent input and output modalities should sharecommon neural substrates This hypothesis is testedhere by asking whether there are brain regions thatare activated in common by different kinds of RSindependent of input and output modalities A compan-ion article (Jiang amp Kanwisher 2003) addresses a secondprediction based on the central bottleneck view (Pashler1989) that some regions involved in RS should not alsobe involved in perceptual processing

This study tests whether common regions are in-volved in various kinds of RS by asking three questionsFirst what brain regions are engaged in RS in a visualndashmanual task Second are these regions specific to thevisual modality or are they also involved in RS forauditory stimuli Third are any common brain regionsthat are recruited in both visual and auditory RS en-gaged only when the two tasks share manual output orspatial mapping requirements Although a few priorinvestigations of response interference in Stroop andother tasks have addressed the first question (Barch etal 2001 Dassonville et al 2001 Connolly Goodale

Desouza Menon amp Vilis 2000 Hazeltine Poldrackamp Gabrieli 2000 Botvinick Nystrom Fissell Carter ampCohen 1999 Carter et al 1998 Carter Botvinick ampCohen 1999 Iacoboni Woods amp Mazziotta 1996Pardo Pardo Janer amp Raichle 1990) few if any priorstudies have investigated the second and third ques-tions (Schumacher amp DrsquoEsposito 2002) Based on thebehavioral literature we predicted that RS based ondifferent input modalities mapping paradigms (whetherspatial or not) and output modalities should recruitcommon brain regions

RESULTS

Behavioral Performance

Figure 1 illustrates the SndashR mapping rules used inExperiments 1 and 2 Responses were either spatiallycompatible or incompatible with the stimuli Table 1shows mean RT and accuracy collected during scan-ning in the three experiments In all experiments in-compatible mapping was substantially slower thancompatible mapping In the auditoryndashmanual task itwas also less accurate Thus we were successful inmanipulating the difficulty of SndashR mappings (ie RS)in all experiments

fMRI Data from the Visual ndashManual Task

We split data from the visual RS task into two indepen-dent halves and carried out a random effects analysis onone half to generate group regions of interest (ROI)involved in visualndashmanual RS The percent signal change(PSC) within these ROIs was then measured in the otherhalf of the visual RS data in the auditoryndashmanual RSdata and in the visual-verbal RS data

Random Effects Analysis

A random effects analysis was conducted on one half ofeach subjectrsquos data (N = 17) searching for any brainregions that had higher BOLD for the spatially incom-patible mapping than compatible mapping ( p lt 001uncorrected) This analysis revealed (see Figure 2)bilateral parietal activation that followed the intraparietalsulcus (IPS) and extended to the superior parietal lobule

Figure 1 Sample display and the SndashR mapping rules for thevisualndashmanual task The task was to pick out the line with uniquelength and report its spatial position among the four lines Targetlocation was mapped to response location according to either naturalmapping (left) or unnatural mapping (right) The same mapping rules(natural and unnatural mapping) were used in the auditory ndashmanualtask in which subjects reported the temporal position of a unique toneamong four tones

Table 1 Behavioral Performance during Scanning

Accuracy () RT (msec)

Experiment Unnatural Natural SE p Unnatural Natural SE p

1 Visual RS 94 95 2 ns 561 383 22 0001

2 Auditory RS 76 82 2 01 717 544 22 0001

3 Verbal naming NA 511 371 24 0001

RS = response selection SE = standard error ns = nonsignificant NA = not applicable

Jiang and Kanwisher 1081

bilaterally and to the precuneus in the right hemispherebilateral frontal eye fields (FEFs) bilateral inferiormiddlefrontal gyrus surrounding area 46 (GFiGFm) bilateralopercularinsula regions surrounding area 4447 (frontaloperculuminsula) and right cerebellum Because acti-vation in the IPS covered a large anatomical region whichmay not be functionally homogeneous we subdivided itinto an anterior and a posterior volume in the ROIanalysis discussed below Group ROIs were definedfunctionally as regions that were activated by the incom-patible gt compatible mapping contrast in the randomeffects analysis ( p lt 001 uncorrected for multiplecomparison see Table 2)

Validation of the VisualndashManual RS Activation in anROI Analysis

To confirm that the activations we saw in the randomeffects analysis were replicable we calculated within thegroup ROIs the PSC relative to fixation in the other halfof the data These PSC results are statistically indepen-dent from the data used to define the ROIs Table 3shows the PSC for each of the ROIs

All parietal ROIs the FEF the left GFm GFi opercu-lum and the right cerebellum showed a significantlyhigher PSC in the incompatible than in the compatiblemapping conditions in this analysis of the second half ofthe data thus independently validating the significanceof this effect in each of these ROIs in a fashion that avoidsthe statistical problems with multiple voxelwise compar-isons Activity in the right GFm and right operculumfailed to reach significance on this stringent test suggest-ing that their involvement in the RS task was less robustHowever the failure to reach significance in these ROIsin the second analysis is subject to type II error In sumthe split-half ROI analysis showed that the majority ofthe ROIs showed significant activation for RS and thata fronto-FEF-parietal network is engaged during thevisualndashmanual mapping task

fMRI Data on the AuditoryndashManual Task

One primary piece of evidence that the processingbottleneck resides at a central cognitive level comesfrom dual-task studies involving inputs from differentmodalities (Pashler 1990) Although visual and auditory

Figure 2 Regions of the brain that showedhigher BOLD during unnatural than naturalmapping between a visual target and a manualresponse (n = 17 p lt 001 random effectsuncorrected all the significant voxels are shownin red) The green arrows point to voxelscontaining most significant voxel of the ROIThese ROIs include 1 Left anterior IPS (LaIPS)2 Left posterior IPS (LpIPS) 3 Right anterior IPS(RaIPS) 4 Right posterior IPS (RpIPS) 5 Rightprecuneus 6 Left frontal eye field (LFEF)7 Right frontal eye field (RFEF) 8 Left inferiorfrontal gyrus (LGFi) 9 Left middle frontal gyrus(LGFm) 10 Right middle frontal gyrus (RGFm)11 Left operculum 12 Right operculum 13Right cerebellum Left hemisphere is shown onthe left side of the figure

1082 Journal of Cognitive Neuroscience Volume 15 Number 8

perception can proceed in parallel (Duncan Martens ampWard 1996 Treisman amp Davies 1973) strong interfer-ence occurs when two responses are selected onebased on visual and the other based on auditory input(Borger 1963 Pashler 1994) Thus RS for stimulipresented in different input modalities relies on thesame cognitive mechanism A natural prediction fromthese behavioral findings is that common brain regionsshould be activated whether RS is based on visual orauditory input Fourteen subjects who were tested inthe visual ndash manual mapping task were also tested in anauditory ndash manual mapping task The same mappingrulesmdashcompatible and incompatiblemdashwere used inboth visual and auditory tasks

Random Effects Analysis on the Auditory ndash ManualMapping Experiment

We analyzed each subjectrsquos data individually and thenperformed a random effects analysis across subjects( p lt 005 uncorrected N = 14) The SPM map ofauditory RS was qualitatively similar to the map of visualRS Figure 3 shows the activation for visual and auditorymodalities within the same 14 subjects revealing thatthe two tasks produced largely overlapping activation(shown in green)

The overlap in the parietal and FEF ROIs was strikingVisual and auditory RS appear to rely on the same regionsin these ROIs In addition to the similarities there werealso apparent differences For example the inferiormiddle frontal gyri were significant for the visual butnot the auditory RS task while the thalamusbasal ganglia

and the cingulate cortex showed the reversed patternSuch differences may be attributed partly to thresholdsetting such that one contrast passed the thresholdwhile the other just missed it Consistent with thishypothesis an interaction test between modality andmapping rule failed to reveal any region that was activat-ed more by one modality than the other In addition ROIanalysis on regions that did not show overlap suggeststhat the trend of an incompatibility effect was the samefor the visual and the auditory RS Thus the brain regionsinvolved in RS appeared to be largely independent ofinput modality at least for visual and auditory stimuli

ROI Analysis Are Visual RS ROIs Also Involved inAuditory RS

To characterize the functional properties of the ROIsdefined by visual RS in Experiment 1 we measured thePSC within these group ROIs in the auditory RS task andin the second half of the visual RS task Thus both datasets were independent of the visual RS data used todefine these ROIs Table 4 shows the mean PSCs in theauditory ndash manual and visual ndash manual mapping tasks

The regions that were strongly involved in visual RS mdashthe parietal ROIs and the FEFs mdashwere also stronglyinvolved in auditory RS The regions whose activationin visual RS was weaker or failed to reach significance mdashthe GFm and the operculummdashalso showed a weakeractivation pattern in auditory RS Thus brain activationsfor RS were qualitatively similar across input modalities

An ANOVA on modality (visual vs auditory) andmapping rule (compatible vs incompatible) within each

Table 2 Regions Showing Significant BOLD Activation for Unnatural gt Natural Mapping in the Visual ndashManual Mapping Task(n = 17 random effect p lt 001 uncorrected)

Voxels [x y z] mm in MNI Hemisphere Location t ROI Radius

900 iexcl36 iexcl54 43 L anterior IPS (area 40) 959 6

iexcl30 iexcl69 45 L posterior IPS (area 407) 714 6

775 42 iexcl45 45 R anterior IPS (area 40) 733 12

30 iexcl66 48 R posterior IPS (area 7) 678 12

18 iexcl66 60 R precuneus (area 7) 711 6

222 iexcl27 3 48 L FEF (area 6) 536 12

92 33 3 48 R FEF (area 6) 539 12

182 iexcl48 9 21 L GFi (area 4446) 502 12

iexcl45 33 18 L GFm (area 46) 558 9

90 36 24 27 R GFm (area 46) 500 6

46 iexcl33 24 0 L operculum (area 4447) 534 12

40 48 21 iexcl12 R operculum (area 4447) 522 12

19 33 iexcl72 iexcl33 R cerebellum 428 12

MNI = Montreal Neurological Institute L = left R = right


ROI showed that the PSCs were generally higher duringthe auditory task than the visual task for both compat-ible and incompatible mappings This main effect ofmodality was significant in all ROIs except in the bilateralposterior IPS and right GFm This modality effect may bedue to the fact that it was harder to map an auditorysequence onto the keys than to map a visual array ontothe keys reflected by subjectsrsquo self-report of effort Themain effect of mapping rule was significant in all theROIs except the bilateral GFm and the right operculumFor the majority of the ROIs the interaction betweenmodality and mapping rule was not significant suggest-ing that visual and auditory RS produced similar activa-tion The only ROI that showed a significant interactionwas the left anterior IPS p lt 01 where higher RS-related activity was found for the auditory than thevisual modality Even in this ROI however both visualand auditory RS were significant suggesting that the leftaIPS was not restricted to processing RS from onemodality only

The similarity in activations between visual and audi-tory RS cannot be attributed to the selection of groupROIs that averaged across many different brains In afurther analysis we selected ROIs based on each indi-vidual subjectrsquos visual RS and replicated the resultsreported above Thus both the random effects SPManalysis and the ROI analysis showed that RS based onvisual and on auditory inputs involved similar regions ofthe brain Further when ROIs were defined based ontheir activity in the auditory RS task those whichshowed a significant effect for the auditory task werealso significant for the visual task ( ps lt 05) Thesefindings support the hypothesis that RS constitutes acentral process occurring after perceptual processing

Comparing the left and right ROIs for the visual andthe auditory RS task we found a significant differencebetween modalities in the laterality of parietal activationp lt 05 Whereas the RS effect was larger in the left thanin the right IPS for the auditory modality it was larger inthe right than in the left for the visual modality How-ever note that the RS effect was significant on both leftand right IPS in both modalities Thus although RSactivities were not identical across visual and auditorymodalities they were largely modality independent

fMRI Data from the Verbal Naming Task

In the verbal naming task we tested whether the regionsinvolved in a spatial visualndashmanual RS task were alsoinvolved in a nonspatial verbal RS task The purpose ofthis experiment was to test the generality of the regionsfor RS found earlier in the manual tasks Specifically wesought to test whether the activity we saw in the visualand auditory RS tasks might be due to the fact that bothtasks required spatial mapping as well as manual re-sponses In this task subjects decided whether twopatches of colormdashpresented successively at the centerof the screen mdash were the same or different In thecompatible mapping condition they responded lsquolsquosamersquorsquoverbally if the colors matched and lsquolsquodifferentrsquorsquo if thecolors did not In the incompatible mapping conditionthey made the opposite response The response map-ping here was thus completely nonspatial and themotor output was verbal If the common activationbetween visual and auditory RS was specific to spatialmapping between stimulus space and motor space orspecific to making manual responses they should not beinvolved in the verbal naming task

Table 3 Percent Signal Change above Fixation for Each ROI in the Natural and Unnatural Mapping Conditions in the Second DataSet of the VisualndashManual Task and p Level of the t Test for Unnatural Versus Natural Mapping across Subjects

Left Hemisphere ROI Right Hemisphere ROI

Natural Unnatural p Natural Unnatural p

aIPS 017 029 008 025

pIPS 010 027 010 026

FEF 022 037 016 027

GFm iexcl002 004 + iexcl005 000 ns

Operculum 000 006 + iexcl003 007 ns

Precuneus NA 007 029

GFi 009 022 NA

Cerebellum NA 019 027 +

ns = nonsignificant NA = not applicable PSCs were calculated from the raw data after preprocessing (motion correct normalizationand smoothing)+p lt 10

p lt 01

p lt 001


Head Motion

Although most fMRI studies have avoided the use ofovert speech during scanning because of concerns abouthead motion recent studies suggest that meaningfuldata can be obtained while subjects speak aloud (BarchBraver Sabb amp Noll 2000 Huang et al 2002 Palmer etal 2001 Rosen Ojemann Ollinger amp Petersen 2000Barch et al 1999) In our verbal naming task subjectswere instructed to make an overt verbal response whileminimizing lip and mouth movements For the two scansinvolving overt naming the overall amount of motioncalculated by SPM motion correction was below 05 mmin each of the three translations and below 0018 in eachof the three rotations The amount of motion during thetwo scans involving overt verbal response was not signif-icantly larger than that during the two preceding scanswithout overt speech (eg during auditory RS)

Random Effects Analysis on Verbal Naming

A random effects analysis was performed on the groupdata for the verbal naming task (N = 12 p lt 005

uncorrected unnatural gt natural naming) Figure 4shows the voxels that reached significance with thevisual RS activation from a different group of 12 sub-jects overlaid

The activation for the naming task in parietal regionswas largely overlapping with the visual RS task al-though it did not extend as far posteriorly Overlappingactivation was also observed in the FEF the GFiGFmand the operculuminsula In addition resolving inter-ference in verbal naming activated the paracingulatesulcus region ( pre-SMA [9 21 48] Rushworth HadlandPaus amp Sipila 2001) Activation unique to the namingtask can also be seen near the fusiform gyrus which hasbeen reported to be involved in color processing (egTootell amp Hadjikhani 2001 Beauchamp Haxby Jen-nings amp De Yoe 1999) These differences were testedin a mapwise interaction between experiment andcompatibility The visualndashmanual task more significantlyactivated left anterior IPS [iexcl39 iexcl54 45] while theverbalndashcolor task more significantly activated middletemporal gyrus [iexcl57 0 iexcl30 iexcl45 33 iexcl15] the fusiformgyrus [iexcl39 iexcl51 iexcl21 39 iexcl51 iexcl15] and the anteriormedial frontal regions [9 42 36 9 51 6] An ROI analysis

Figure 3 Overlapping activation (green)across visual (red) and auditory (blue)response selection in a random effects groupanalysis (n = 14 p lt 005 uncorrected)


based on these regions showed that left anterior IPSwas involved in both visual and verbal RS while theother ROIs failed to reach significance for the incom-patibility effect in either visual or verbal RS Thus it isunlikely that the regions that showed interaction in themapwise comparison were involved in RS for only onemodality

ROI Analysis Are VisualndashManual ROIs Limited toSpatial SndashR Mapping or Manual Responses

To test the functional properties of the group ROIsdefined in the visualndashmanual task here we measuredthe PSC in the natural and reversed verbal naming taskswithin the group ROIs defined in Experiment 1 We alsomeasured PSC within the group ROIs from a subset of12 subjects from Experiment 1 The same number ofsubjects (N = 12) were selected in the two tasks sothat they had comparable statistical power The ROIswere the ones defined in Experiment 1 Table 5 showsthe results

Twelve of the 13 ROIs defined by their visualndashmanualRS activity were also significantly involved in the verbalnaming task the only exception was the right cerebel-lum which may not be covered to the same extentacross subjects As in the random effects analysis theROI analysis revealed common activation between thevisual RS and the verbal naming task Some differencesbetween the two tasks were also noted For example inthe verbal naming task the posterior IPS and the right

precuneus showed signal lower than the fixation base-line This may be related to the finding that the midlineposterior brain including the posterior cingulate andthe precuneus typically shows negative signals duringdemanding cognitive tasks (Gusnard amp Raichle 2001Raichle et al 2001 Shulman et al 1997) That thesignal was above fixation during the visualndashmanualtask suggests that in addition to their possible involve-ment in RS these regions are also sensitive to spatialprocessing (Berman et al 1999 Labar Gitelman Par-rish amp Mesulam 1999 Culham et al 1998) When theexperiment (verbal naming vs visualndashmanual mapping)was entered in an ANOVA as a between-subject factorwe failed to find a significant interaction between theexperiment and mapping condition in any of the ROIssuggesting that activity in these ROIs was not specificto spatial or nonspatial RS and was not specific tomanual responses

Activation in the Anterior Cingulate Cortex acrossExperiments

The anterior cingulate cortex (ACC) has been implicatedin monitoring response conflict (Barch et al 2001 VanVeen Cohen Botvinick Stenger amp Carter 2001 LeungSkudlarski Gatenby Peterson amp Gore 2000 Carteret al 1998) and a priori it appears to be a likelycandidate for the central processing bottleneck Howev-er in this study the ACC failed to pass the threshold in

Table 4 Percent Signal Change above Fixation within the Visual RS ROIs in the AuditoryndashManual and the VisualndashManualMapping Tasks


AuditoryndashManual VisualndashManual Auditory ndashManual VisualndashManual

Natural Unnatural Natural Unnatural Natural Unnatural Natural Unnatural

aIPS 029 048 012 018+ 026 037 004 019

pIPS 009 030 007 019 008 021 005 020

FEF 042 060 024 038 025 040 012 024

GFm 025 032 ns iexcl008 000 ns 002 004 ns iexcl006 iexcl007 ns

Operculum 034 039 ns 001 006 ns 026 032 ns iexcl009 iexcl003 ns

Precuneus NA 025 049 006 028

GFi 051 061 007 019 NA

Cerebellum NA 049 061 024 034

ns = nonsignificant NA = nonapplicable+p lt 10

p lt 05

p lt 01

p lt 001


Experiment 1 and was not further tested To examineactivity in the ACC here we defined an anatomical ROIcentered on the ACC ([0 33 30] Van Veen et al 2001)It included a spherical volume of 33 voxels with a radiusof 6 mm

The ACC failed to be activated by the visualndashmanualmapping task (iexcl007 signal change for natural map-ping iexcl004 signal change for unnatural mappingp gt 25) or by the verbal color matching task (iexcl010signal change for natural report 005 signal change forreversed report p gt 10) Interestingly the ACC activa-tion in the auditory RS task approached significance(007 for natural mapping 017 for unnatural map-ping p lt 053) Past studies have argued that the ACCplays an important role in monitoring potential conflictwhich correlates with error monitoring but is notidentical to error detection (Botvinick et al 1999 Carteret al 1998 1999) In all three tasks conflict monitoringis necessary and may be greatest in the auditory RS taskreflected by a significant accuracy effect This hypothesisabout conflict monitoring explains activation in theauditory RS task but does not account for the lack ofactivation in the other two experiments Such lack ofconsistent ACC involvement in different kinds of RS

argues against the ACC as the cognitive processingbottleneck

Negative Activation

It is not uncommon to observe decreased BOLD signalsassociated with a more effortful task in certain brainareas (Raichle et al 2001) Comparing the compatibleand incompatible visualndashmanual mapping in a randomeffects analysis a decreased BOLD signal for the incom-patible mapping condition was seen in several regionstypically associated with decreased activation the cu-neusprecuneus [9 iexcl90 27] the posterior cingulatecortex [6 iexcl51 24] the superior frontal gyrus [24 39 51]the medial frontal cortex [iexcl12 66 19] and the superiormiddle temporal gyrus [57 iexcl6 iexcl18] In the auditoryndashmanual mapping task the same subtraction led tonegative activation in the superior temporal gyrus[iexcl39 iexcl12 9] In the verbal naming task negative activa-tion was seen in the precuneus [iexcl6 iexcl51 33] theposterior cingulate cortex [iexcl3 iexcl30 48] the superiorfrontal gyrus [iexcl21 42 42] the medial frontal cortex[iexcl6 57 iexcl3] and the middle temporal gyrus [30 0iexcl33] In all these lsquolsquodeactivatedrsquorsquo regions the PSCs for

Figure 4 Overlapping activation (in green)between the visual RS task of Experiment 1 (in red)and the verbal naming task (in blue)


both natural and unnatural mapping were below fixa-tion with the unnatural mapping more negative De-creased BOLD in these regions may be related toreduced self-referencing mental activity during demand-ing cognitive tasks (Raichle et al 2001)

DISCUSSION

We used fMRI to identify brain regions engaged duringRS a process thought to represent a central bottleneckin human information processing (Pashler 1984 19941998 see also Schumacher et al 1999 2001 Meyeret al 1995 Meyer amp Kieras 1997) The regions thatproduced significantly higher BOLD signal duringincompatible versus compatible SndashR mappings acrosstasks include the bilateral IPS extending into the SPLand the right precuneus the bilateral FEFs the rightcerebellum and less robustly into the lateral frontalregions

These regions support two properties of the hypoth-esized central processing bottleneck (Pashler 1994) itsindependence from perceptual input and from motoreffector First we found that largely overlapping re-gions of the brain were recruited when subjects select-ed responses based on visual and auditory inputsSecond we observed common activation of a set ofregions for RS across tasks involving manual outputand overt speech Finally the similarity in activationbetween the visualndashmanual task and the verbal re-sponse task also indicates that this activation was notrestricted to spatial mapping These findings are con-

sistent with the hypothesis that RS is a central cognitiveprocess

Relationship to Prior Studies of RS

Broadly defined RS includes any process involved inmapping a stimulus onto a response A number ofpossible subprocesses satisfy this criterion In the visu-al ndashmanual mapping task for example increasing thecomplexity of the mapping rule leads to increaseddemand on any of several related subprocesses (1)SndashR mapping (2) inhibition of prepotent responsesbased on the natural mapping rule and (3) spatialremapping between the target space and the responsespace Increasing the complexity of the mapping rulealso increases the difficulty of the task The selection ofthis task was guided by the behavioral literature whichhas shown that a more complex mapping rule occupiesthe central bottleneck for a longer time (Pashler 1994McCann amp Johnston 1992)

Another study (Jiang amp Kanwisher in preparation)tested whether the activation we saw was related to SndashR mapping That study tested 12 subjects in a furtherexperiment that did not involve inhibition of prepotentresponses or spatial remapping Subjects were presentedwith geometric shapesmdasha square or a circlemdashand madeeither consistent responses to these shapes (eg one keyfor square and another for circle) or noncontingentresponses (repeatedly keypresses independent of theshape) Thus contingent minus noncontingent re-sponses tested the presence of RS uncontaminated by

Table 5 Percent Signal Change above Fixation in the Natural and Reversed Naming and in the Natural and Unnatural Visual RS inthe Group ROIs Defined by the Visual ndashManual Task


Naming Task VisualndashManual Naming Task VisualndashManual

Natural Reversed Natural Unnatural Natural Reversed Natural Unnatural

aIPS iexcl003 010 012 027 iexcl003 015 007 023

pIPS iexcl022 iexcl009+ 008 026 iexcl021 iexcl009 010 027

FEF 012 021 018 033 009 021 014 025

GFm iexcl009 002 iexcl007 iexcl001+ iexcl005 002+ iexcl009 iexcl004 ns

Operculum 007 016 iexcl002 003 ns 021 038 iexcl001 005 ns

Precuneus NA iexcl035 iexcl025+ 003 026

GFi 012 025 007 018 NA

Cerebellum NA 034 032 ns 018 029


p lt 05

p lt 01

p lt 001


inhibition of prepotent responses or spatial remappingThis subtraction led to highly significant activation inthe bilateral IPS and FEFs and weaker but significantactivation in the lateral frontal regions the sameregions activated in the experiments reported hereThus we believe that the regions seen here in ourstudy are the ones that play important roles in RSwhether inhibition of prepotent response or spatialremapping is involved or not An unanswered questionis whether the RS regions are also activated by processesrelated to RS such as inhibition of prepotent responsesand increased working memory load Further experi-ments that parametrically manipulate levels of RS difficul-ty (Schumacher Elston amp DrsquoEsposito 2003) and thatisolate each subprocess experimentally are likely to pro-vide more conclusive answers

Although it is hard to rule out the hypothesis that ouractivations reflect general difficulty previous studiessuggest that at least some difficult tasks may not alwaysactivate the frontal cortex (Schumacher amp DrsquoEsposito2002 Owen 2000 Barch et al 1997 Demb et al 1995)and the parietal regions particularly on the right (Mar-ois Chun amp Gore 2000 Wojciulik amp Kanwisher 1999)The issue of task difficulty will be directly tested in thecompanion article (Jiang amp Kanwisher 2003) There weused two word tasks differing in task difficulty and failedto find increased right parietal activation for the moredifficult task The pattern of activation across the wholebrain also differed depending on which type of difficultywas involved Although task difficulty may be hard todisentangle from the increased demand on specificcognitive processes it is unlikely that general difficultycan fully account for all common activations seen in ourstudy

A number of prior imaging studies have tested RS byvarying response competition These include studiesusing the Stroop colorndashword naming task (MacLeod1991 Stroop 1935) the Eriksen flanker task (Hazeltineet al 2000 Botvinick et al 1999) the anti-saccade andanti-pointing task (Connelly et al 2000) the gono-gotask (Liddle Kiehl amp Smith 2001 Konishi et al 1999)and spatial incompatibility tasks (Schumacher amp DrsquoEs-posito 2002 Dassonville et al 2001 Iacoboni et al1996) The areas we saw activated in the current studynamely the fronto-FEF-parietal network are also typi-cally seen in many previous studies using other para-digms For example in the Stroop task the brain regionsthat are more active during incompatible than compat-ible trials typically include the inferior frontal cortex theposterior parietal cortex and the ACC (Zysset MullerLohmann von Cramon 2001 Banich et al 2000 Leunget al 2000 Bovinick et al 1999 Carter et al 1999 Bushet al 1998 Bench et al 1993 Pardo et al 1990) In theflanker task the lateral prefrontal cortex the supple-mentary motor areas and the parietal cortex showhigher BOLD signal on incongruent trials than on con-gruent trials (Hazeltine et al 2000 Bunge Hazeltine

Scanlon Rosen amp Gabrieli in press) The fronto-FEF-parietal network has been implicated in the anti-saccadetask (Connolly et al 2000) the gono-go task (Liddleet al 2001 Konishi et al 1999) and spatial incompat-ibility tasks (eg Schumacher amp DrsquoEsposito 2002) Atthe behavioral level both response competition (withor without inhibition of prepotent responses) and in-creased mapping complexity are assumed to occupy thecentral bottleneck (Pashler 1994) Whether the exactsame network is used for different subprocesses of RSremains to be sorted out in future studies that test thesame subjects in different paradigms

Practice Effects

The persistence of a dual-task interference when twoRSs are made is strong evidence that RS is a centralcognitive process (Pashler 1994) Although dual-taskinterference between two concurrent or adjacent RStasks is the general rule this interference can some-times be reduced or even eliminated after practice(Schumacher et al 2001 Van Selst Ruthruff amp John-ston 1999) However even thousands of trials ofpractice are not always sufficient to eliminate theinterference (Ruthruff et al 2001 Schumacher et al2001) Other important conditions for reduced inter-ference are equal emphasis on the two tasks subjectrsquosadoption of a lsquolsquodaringrsquorsquo versus a lsquolsquoconservativersquorsquo strategyand highly compatibleSndashR mapping in each task (Schu-macher et al 2001) Whether the practice effect issufficient to reject the existence of a central bottleneckis currently under debate (Levy amp Pashler 2001 Ruth-ruff et al 2001 Van Selst et al 1999 Meyer amp Kieras1997 Pashler 1994) Future neuroimaging studies mayshed light on these questions For example one unan-swered question concerns the change in neural activa-tion after practice When dual-task interference from RSis diminished is the same fronto-FEF-parietal networkreported here still recruited or does extensive trainingwith these tasks lead to the construction of alternateneural pathways dedicated to that task (or combinationof tasks)

Differences among the ROIs

The co-activation of the frontal FEF and parietal ROIs inour tasks and in many prior studies (Cabeza amp Nyberg2000) poses a real challenge for any effort to distinguishbetween the functions of these regions However thisco-activation does not mean that these regions carry outidentical functions (Hazeltine et al 2000 Bunge et al2002) and may instead reflect the strong anatomicalconnections between frontal and parietal cortices (Pet-rides amp Pandya 1984)

Our data reveal some differences in the pattern ofactivation across ROIs between the visualndashmanual taskand the verbal naming task Consider two ROIs the right


posterior IPS and the right operculum In the visualndashmanual mapping task the signal increase for incompat-ible versus compatible RS was high in the right pIPS butlow in the right operculum In contrast in the verbalnaming task the RS activation was weak in the right pIPSand strong in the right operculum This three-wayinteraction between task (visualndashmanual vs naming)mapping rule and region was significant ( p lt 03) Inhindsight this difference makes sense because the pIPS(extending into the SPL and the precuneus) is typicallyengaged during spatial processing (Berman et al 1999Labar et al 1999 Culham et al 1998) while the frontalopercularinsular region falls in the vicinity of Brocarsquosarea which has been implicated in linguistic processing(Friederici Meyer amp von Cramon 2000 Price amp Friston1997) and tongue movement (Corfield et al 1999)Thus RS based on different tasks may not induceidentical activation in all ROIs However this conclusionis based only on a post hoc analysis that used a lenientstatistical criterion ( p lt 05 uncorrected) Convergingempirical evidence especially from techniques withhigher temporal resolution is needed to determinethe differences in the contributions made by each ofthese regions to RS

To summarize in three fMRI experiments we variedthe difficulty of mapping a percept onto a responsewhile holding constant the percept and the motorresponse Compared with compatible mapping incom-patible SndashR mapping produced higher BOLD signals inthe parietal cortex the FEF the cerebellum and thelateral prefrontal cortex This activation was seenwhen the RS was based on visual or auditory inputswhen the mapping was spatial or nonspatial and whenthe motor output was manual or verbal The generalityof activation for RS across input and output modalitiesand across mapping paradigms supports the hypothesisthat it constitutes a central cognitive process Howeverto qualify as the neural counterpart of the cognitivebottleneck characterized by the behavioral literature(Levy amp Pashler 1995 Pashler 1989 1992) one woulddemonstrate that these regions are NOT involved inperceptual attention Thus it remains to be seen wheth-er any of these regions are selective for RS per se orwhether they have more general functions This issuewill be pursued in the companion article (Jiang ampKanwisher 2003)

METHODS

Subjects

Twenty-seven subjects between the ages of 18 and 42participated in these studies (12 women and 15 men)They all had normal or corrected-to-normal vision andnormal color vision Seventeen subjects were tested inExperiment 1 14 in Experiment 2 and 12 in Experiment3 Some subjects were scanned in multiple experiments

Testing Procedure

Subjects received a 5-min practice in each task on thesame day or the day before the scan They were scannedon a Siemens 30 T head-only scanner at the Athinoula AMartinos Center for Biomedical Imaging in CharlestownMA We first acquired 3-D structural images based on128 sagittal slices ( TR = 11 msec TE = 4 mm field ofview [FoV] = 256 pound 256) and acquired high resolutionT1 images (default TR = 700 msec FoV = 200 pound200 mm resolution = 078 pound 078 mm) Twenty obliqueaxial slices 6 mm thick with 0 mm distance betweenslices were scanned The slices extended from the top ofthe brain to most or all of the cerebellum We used aT2-weighted EPI sequence (TR = 2000 msec TE =30 msec flip angle = 908 resolution = 313 pound 313 pound600 mm) for the functional scans Each scan lasted 6 min4 sec the first 8 sec were discarded During scanningthe visual display was back projected onto a glass screenwhose image was reflected by a 458 mirror to thesubjects Auditory stimuli were transmitted by plastictubes and magnetic-compatible head phones to subjectsbinaurally In the visual scans (Experiments 1 and 3) thecollection of images was evenly distributed in the 2-secTR In the auditory scans (Experiment 2) the collectionwas bunched to the first 1312 sec of the 2-sec TRleaving a 688-msec silent period during which time theauditory stimuli were presented

Scan Composition

Each functional scan was a blocked design of threeconditions fixation (F) Task A and Task B Each taskblock lasted 64 sec and each fixation block lasted 20 secThe scan was composed of a series of blocks in whichtasks alternated and fixation blocks intervened betweeneach task block An example of one scan sequence islsquolsquoFAFBFAFBFrsquorsquo In all experiments the two tasks werematched in perceptual input and in motor outputDifferences between tasks were introduced by instruc-tions The first four fixation blocks were each composedof 15 sec fixation followed by 5 sec instruction Theorder between Tasks A and B was counterbalancedwithin subjects as ABAB or BABA

Materials and Tasks

Stimuli were presented using the Psychtoolbox imple-mented in MATLAB (Brainard 1997) Presentation of thestimuli was synchronized with scanning using a triggersignal sent by the scanner to the computer

Experiment 1 VisualndashManual Mapping

Each trial (2 sec) of the visualndashmanual task started witha visual display of 100 msec followed by a 100-msecmask and then a 1800-msec fixation display Each


stimulus contained four vertical lines three of which hadidentical length whereas one was either shorter orlonger than the others The longer line of each displaywas either 318 or 288 whereas the shorter line waseither 108 or 088 The stimuli were presented against amid-gray background The lines were evenly spaced on a6258 by 6258 display (Figure 1) The mask was made of18 vertical and 18 horizontal lines (length = 6258) semi-irregularly displaced

The task on each trial was to identify the line withunique length and report its spatial position among thefour lines by pressing one of four keys Subjects restedtheir index middle ring and little fingers of the righthand on keys 1 2 3 and 4 The target position wasmapped onto the keys according to a natural mappingrule or an unnatural mapping rule The instructionspreceding the natural mapping blocks schematicallyillustrated a straightforward mapping (Figure 1 left)The instructions preceding the unnatural mapping in-formed subjects to press key 3 4 1 and 2 for target atpositions 1 2 3 and 4 respectively Each subjectcompleted four scans

Experiment 2 AuditoryndashManual Mapping

The auditory task was designed to be as similar aspossible to the visual task On each trial subjects werepresented with a sequence of four tones lasting a totalof 480 msec Three of the tones were identical and theother one was unique in duration (or frequency forsome subjects) In the frequency task each tone lasted45 msec with a blank interval of 100 msec betweenadjacent tones The frequency of the higher tone ofeach trial was either 2000 or 1800 Hz and the frequencyof the lower tone was either 500 or 450 Hz Each tonewas a sine wave with a rise and fall period of 5 msecSubjects were instructed to identify the tone with theunique frequency and report its temporal positionwithin the series by pressing one of four keys In theduration task the frequency of all the tones was fixed at1000 Hz Three of the tones had short duration (20msec) while the unique tone was long (120 msec) aninterval of 100 msec separated consecutive tones Sub-jects were to report the temporal position of the longertone Of the 14 subjects the first 6 were tested in thefrequency task and the last 8 were tested in the durationtask The dimension of judgment was changed becausethe frequency task was reported to be too difficult Datafrom the duration and frequency tasks were pooledtogether because analyses of each task alone showedthat the two tasks produced very similar mapwiseactivation Furthermore in the ROI analysis when sub-ject group (frequency vs duration group) was enteredas a between-subject factor it failed to show any signif-icant main effect or interaction with other factors inany of the ROIs Only pooled data are reported in theresults here

Subjects were scanned in a blocked design identical tothat of Experiment 1 and performed either compatibleor incompatible mapping Each subject received eitherfour or two functional scans (in ABAB and BABA order)The 480-sec sequence was presented in the middle ofthe 688-msec silent period in the bunched scanningsequence Subjects wore earplugs to reduce the loud-ness of the scanner noise The loudness of the tones wasindividually adjusted The typical loudness was 100 dBbefore the dampening (of about 30 dB) by the ear plugsand the foam padding

Experiment 3 Verbal Color-Matching Task

On each trial two color patches (diameter = 093 cm)each presented for 100 msec and separated by a100-msec interval were presented at fixation Subjectswere asked to judge whether the colors were identicalor different The colors were chosen from two shades ofgreen (RGB values [0 255 0] and [0 255 30]) and twoshades of blue (RGB values [0 0 255] and [0 30 255])Half of the trials were match trials that is the two colorswere identical The other half were nonmatch trials onewas blue and the other was green

Subjects were required to make an overt verbal reportby saying lsquolsquosamersquorsquo or lsquolsquodifferentrsquorsquo while minimizing lip andmouth movement The instructions were lsquolsquosay SAMEwhen the colors are the same say DIFFERENT whenthe colors are differentrsquorsquo during the compatible blocksand lsquolsquosay DIFFERENT when the colors are the same saySAME when the colors are differentrsquorsquo during the incom-patible blocks Verbal responses were not recorded inthe scanner To measure the speed of subjectsrsquo re-sponse subjects were instructed to press a key withtheir right index finger simultaneously with their verbalresponse regardless of which verbal response theymade Behavioral studies have shown that when tworesponses are made to the same feature they do notinterfere with each other (Fagot amp Pashler 1992) Thusadding the constant task of pressing an index fingershould not interfere with verbal responses In additionbecause the same finger responses were made in allconditions the effect of pushing a key could be sub-tracted out Each subject performed two scans in thistask the blocked design and the scanning parameterswere identical to those of Experiment 1

fMRI Data Analysis

Data were analyzed using SPM99 (httpwwwfilionuclacukspmspm99html) Each subjectrsquos data weremotion corrected and then normalized onto a commonbrain space (the MNI template) Data were thensmoothed with a gaussian filter (full width half maxi-mum = 8 mm) and high-pass filtered during analysisWe also performed the SPM analysis on unsmootheddata to make sure that the common activation we found


across experiments was not due to spatial smoothingResults based on unsmoothed data were qualitativelysimilar to those on smoothed data reported here

We carried out a random effects analysis ( p lt 001uncorrected for Experiment 1 based on one half of eachsubjectrsquos data and p lt 005 uncorrected for Experi-ments 2 and 3) The threshold was set to be relativelylenient to reduce Type II error Possible Type I error wasfurther controlled by the ROIs analysis on independentdata sets as follows

A test of the functional properties of the activeregions found in Experiment 1 was performed usingan ROI analysis In Experiment 1 the four scans ofeach subject were divided into two sets of two scanseach set was counterbalanced in presentation order(ABAB or BABA) One set of the data were entered inthe random effects analysis and defined group ROIsfunctionally Each group ROI was centered on thelocal maximum of the random effect analysis with asphere including active voxels Regions containingfewer than five significant voxels were not consideredThe distance between the center of any two ROIs wasat least 12 mm apart Each group ROI was within aspherical volume containing the significant voxels theradius of the ROIs was between 6 and 12 mm withthe constraint that different ROIs did not overlapBecause these group ROIs were generated based ona random effects analysis they permitted generaliza-tion to the population at large Results focused on theROIs defined by the visualndashmanual task because ittested a large number of subjects on multiple scansWe also conducted ROI analysis on ROIs defined bythe auditoryndashmanual or the colorndashverbal task andfound similar activation across experiments

PSC relative to fixation baseline [PSC = 100 rawBOLD magnitude for (the task iexcl fixation)raw BOLDmagnitude for fixation] was calculated for both eachmapping condition within each ROI for each subjectThese values were then entered in an ANOVA thesignificance of the ANOVA was reported in each tableBecause the data defining the ROIs were independentfrom data used to calculate PSCs Type I errors weredrastically reduced

Acknowledgments

This work was supported by a Human Frontiersrsquo grant to NKYJ was supported by a research fellowship from the HelenHay Whitney Foundation We thank Miles Shuman fortechnical assistance and Kyungmouk Lee for data analysisand Silvia Bunge Mark DrsquoEsposito John Duncan Molly PotterRebecca Saxe and Eric Schumacher for helpful comments onthe manuscript

Reprint requests should be sent to Yuhong Jiang currentlyat the Department of Psychology Harvard University33 Kirkland St Room 820 Cambridge MA 02138 USA orvia e-mail yuhongwjhharvardedu

The data reported in this experiment have been deposited inthe fMRI Data Center (httpwwwfmridcorg) The accessionnumber is 2-2003-113T5

REFERENCES

Banich M T Milham M P Atchley R A Cohen N J WebbA Wszalek T Kramer A F Liang Z-P Wright AShenker J Magin R Barad V Gullett D Shah C ampBrown C (2000) fMRI studies of Stroop tasks reveal uniqueroles of anterior and posterior brain systems inattentional selection Journal of CognitiveNeuroscience 12 988 ndash1000

Barch D M Braver T S Akbudak E Conturo T OllingerJ amp Snyder A (2001) Anterior cingulate cortex andresponse conflict Effects of response modality andprocessing domain Cerebral Cortex 11 837ndash848

Barch D M Braver T S Nystrom L E Forman S D NollD C amp Cohen J D (1997) Dissociating working memoryfrom task difficulty in human prefrontal cortexNeuropsychologia 35 1373 ndash1380

Barch D M Braver T S Sabb F W amp Noll D C (2000)Anterior cingulate and the monitoring of response conflictEvidence from an fMRI study of overt verb generationJournal of Cognitive Neuroscience 12 298 ndash309

Barch D M Sabb F W Carter C S Braver T S Noll D Camp Cohen J D (1999) Overt verbal responding during fMRIscanning Empirical investigations of problems and potentialsolutions Neuroimage 10 642 ndash657

Beauchamp M S Haxby J V Jennings J E amp DeYoe E A(1999) An fMRI version of the Farnsworth ndashMunsell 100-Huetest reveals multiple color-selective areas in human ventraloccipitotemporal cortex Cerebral Cortex 9 257 ndash263

Bench C J Frith C D Grasby P M Friston K J PaulesuE Frackowiak R S amp Dolan R J (1993) Investigations ofthe functional anatomy of attention using the Stroop testNeuropsychologia 31 907 ndash922

Berman R A Colby C L Genovese C R Voyvodic J TLuna B Thulborn K R amp Sweeney J A (1999) Corticalnetworks subserving pursuit and saccadic eye movements inhumans An fMRI study Human Brain Mapping 8209 ndash225

Borger R (1963) The refractory period and serial choicereactions Quarterly Journal of Experimental Psychology15 1ndash12

Botvinick M Nystrom L E Fissell K Carter C S amp CohenJ D (1999) Conflict monitoring versus selection-for-actionin anterior cingulate cortex Nature 402 179 ndash181

Brainard D H (1997) The psychophysics toolbox SpatialVision 10 433 ndash436

Bunge S A Hazeltine E Scanlon M D Rosen A C ampGabrieli J D E (2002) Dissociable contributions of pre-frontal and parietal cortices to response selectionNeuroimage 17 1562ndash1571

Bush G Whalen P J Rosen B R Jenike M A McInerneyS C amp Rauch S L (1998) The counting Stroop Aninterference task specialized for functionalneuroimagingmdashvalidation study with functional MRIHuman Brain Mapping 6 270 ndash282

Cabeza R amp Nyberg L (2000) Imaging cognition II Anempirical review of 275 PET and fMRI studies Journal ofCognitive Neuroscience 12 1ndash47

Carter C S Botvinick M M amp Cohen J D (1999) Thecontribution of the anterior cingulate cortex to executiveprocesses in cognition Reviews in the Neurosciences 1049ndash57

Carter C S Braver T S Barch D M Botvinick M M Noll


D amp Cohen J D (1998) Anterior cingulate cortex errordetection and the online monitoring of performanceScience 280 747ndash749

Connolly J D Goodale M A Desouza J F Menon R S ampVilis T (2000) A comparison of frontoparietal fMRIactivation during anti-saccades and anti-pointingJournal of Neurophysiology 84 1645 ndash1655

Corfield D R Murphy K et al (1999) Cortical andsubcortical control of tongue movement in humans Afunctional neuroimaging study using fMRI Journal ofApplied Physiology 86 1468 ndash1477

Culham J C Brandt S A Cavanagh P Kanwisher N GDale A M amp Tootell R B (1998) Cortical fMRI activationproduced by attentive tracking of moving targets Journal ofNeurophysiology 80 2657 ndash2670

Dassonville P Lewis S M Zhu X H Ugurbil K Kim S GAshe J (2001) The effect of stimulusndashresponsecompatibility on cortical motor activation Neuroimage13 1ndash14

Demb J B Desmond J E Wagner A D Vaidya C JGlover G H amp Gabrieli J D (1995) Semantic encodingand retrieval in the left inferior prefrontal cortex Afunctional MRI study of task difficulty and process specificityJournal of Neuroscience 15 5870ndash5878

Duncan J Martens S amp Ward R (1996) Restrictedattentional capacity within but not between sensorymodalities Nature 379 808 ndash810

Fagot C amp Pashler H (1992) Making two responses to asingle object Implications for the central attentionalbottleneck Journal of Experimental Psychology HumanPerception and Performance 18 1058ndash1079

Friederici A D Meyer M amp von Cramon D Y (2000)Auditory language comprehension An event-related fMRIstudy on the processing of syntactic and lexical informationBrain and Language 75 289ndash300

Gusnard D A amp Raichle M E (2001) Searching for abaseline Functional imaging and the resting human brainNature Reviews Neuroscience 2 685 ndash694

Hazeltine E Poldrack R amp Gabrieli J D (2000) Neuralactivation during response competition Journal ofCognitive Neuroscience 12 118 ndash129

Huang J Carr T H amp Cao Y (2002) Comparing corticalactivations for silent and overt speech using event-relatedfMRI Human Brain Mapping 15 39ndash53

Iacoboni M Woods RP amp Mazziotta JC (1996) Brainndashbehavior relationships Evidence from practice effects inspatial stimulusndashresponse compatibility Journal ofNeurophysiology 76 321 ndash331

Jiang Y amp Kanwisher N (2003) Common neural mechanismfor response selection and perceptual processing Journalof Cognitive Neuroscience 15 1095 ndash1110

Jiang Y amp Kanwisher N (in preparation) Involvement of theparietal-frontal regions in nonspatial response selection

Konishi S Nakajima K Uchida I Kikyo H Kameyama Mamp Miyashita Y (1999) Common inhibitory mechanism inhuman inferior prefrontal cortex revealed by event-relatedfunctional MRI Brain 122 981ndash991

LaBar K S Gitelman D R Parrish T B Mesulam M (1999)Neuroanatomic overlap of working memory and spatialattention networks A functional MRI comparison withinsubjects Neuroimage 10 695 ndash704

Leung H C Skudlarski P Gatenby J C Peterson B S ampGore J C (2000) An event-related functional MRI study ofthe Stroop color word interference task Cerebral Cortex10 552 ndash560

Levy J amp Pashler H (1995) Does perceptual analysiscontinue during selection and production of a speededresponse Acta Psychologica 90 245 ndash260

Levy J amp Pashler H (2001) Is dual-task slowing instructiondependent Journal of Experimental Psychology HumanPerception and Performance 27 862 ndash869

Liddle P F Kiehl K A amp Smith A M (2001) Event-relatedfMRI study of response inhibition Human Brain Mapping12 100 ndash109

MacLeod C M (1991) Half a century of research on theStroop effect An integrative review Psychological Bulletin109 163 ndash203

Marois R Chun M M amp Gore J C (2000) Neural correlatesof the attentional blink Neuron 28 299 ndash308

McCann R S amp Johnston J C (1992) Locus of thesingle-channel bottleneck in dual-task interference Journalof Experimental Psychology Human Perception andPerformance 18 471ndash484

Meyer D E amp Kieras D E (1997) A computational theoryof executive cognitive processes and multiple-taskperformance Part 2 Accounts of psychologicalrefractory-period phenomena Psychological Review 104749 ndash791

Meyer D E Kieras D E Lauber E Schumacher E HGlass J Zurbriggen E Gmeindl L amp Apfelblat D (1995)Adaptive executive control Flexible multiple-taskperformance without pervasive immutableresponse-selection bottlenecks Acta Psychologica 90163 ndash190

Owen A M (2000) The role of the lateral prefrontal cortex inmnemonic processing The contribution of functionalneuroimaging Experimental Brain Research 133 33ndash43

Palmer E D Rosen H J Ojemann J G Buckner R LKelley W M Petersen S E (2001) An event-related fMRIstudy of overt and covert word stem completionNeuroimage 14 182 ndash193

Pardo J V Pardo P J Janer K W amp Raichle M E (1990)The anterior cingulate cortex mediates processing selectionin the Stroop attentional conflict paradigm Proceedingsof the National Academy of Sciences USA 87 256 ndash259

Pashler H (1984) Processing stages in overlapping tasksEvidence for a central bottleneck Journal of ExperimentalPsychology Human Perception and Performance 10358 ndash377

Pashler H (1989) Dissociations and dependencies betweenspeed and accuracy Evidence for a two-component theoryof divided attention in simple tasks Cognitive Psychology21 469 ndash514

Pashler H (1990) Do response modality effects supportmultiprocessor models of divided attention Journal ofExperimental Psychology Human Perception andPerformance 16 826ndash840

Pashler H (1992) Shifting visual attention and selectingmotor responses Distinct attentional mechanisms Journalof Experimental Psychology Human Perception andPerformance 17 1023 ndash1040

Pashler H (1994) Dual-task interference in simple tasks Dataand theory Psychological Bulletin 116 220 ndash244

Pashler H (1998) The Psychology of Attention CambridgeMIT Press

Pashler H Carrier M amp Hoffman J (1993) Saccadic eyemovements and dual-task interference Quarterly Journalof Experimental Psychology 46A 51ndash82

Petrides M amp Pandya D N (1984) Projections to thefrontal cortex from the posterior parietal region in therhesus monkey Journal of Comparative Neurology 228105 ndash116

Price C J amp Friston K J (1997) The temporal dynamics ofreading A PET study Proceedings of the Royal Society ofLondon B Biological Sciences 264 1785 ndash1791

Raichle M E MacLeod A M Snyder A Z Powers W J


Gusnard D A amp Shulman G L (2001) A default mode ofbrain function Proceedings of the National Academy ofScience USA 98 676 ndash682

Rosen H J Ojemann J G Ollinger J M amp Petersen S E(2000) Comparison of brain activation during word retrievaldone silently and aloud using fMRI Brain and Cognition42 201 ndash217

Rushworth M F S Hadland K A Paus T amp Sipila P K(2001) Role of the human medial frontal cortex in taskswitching A combined fMRI and TMS study Journal ofNeurophysiology 87 2577 ndash2592

Ruthruff E Pashler H amp Klaassen A (2001) Processingbottlenecks in dual-task performance Structural limitationor strategic postponement Psychonomic Bulletin andReview 8 73ndash80

Schumacher E H amp DrsquoEsposito M (2002) Neuralimplementation of response selection in humans as revealedby localized effects of stimulusndashresponse compatibility onbrain activation Human Brain Mapping 17 193 ndash201

Schumacher E H Elston P H amp DrsquoEsposito M (2003)Neural evidence for representation-specific responseselection Journal of Cognitive Neuroscience 15 1111 ndash1121

Schumacher E H Lauber E J Glass J M Zurbriggen E LGmeindl L Kieras D E amp Meyer D E (1999) Concurrentresponse-selection processes in dual-task performanceEvidence for adaptive executive control of task schedulingJournal of Experimental Psychology Human Perceptionand Performance 25 791 ndash814

Schumacher E H Seymour T L Glass J M Kieras D Eamp Meyer D E (2001) Virtually perfect time sharing in

dual-task performance Uncorking the central attentionalbottleneck Psychological Science 12 101 ndash108

Shulman G L Fiez J A Corbetta M Buckner R L MiezinF M Raichle M E amp Petersen S E (1997) Commonblood flow changes across visual tasks II Decreases incerebral cortex Journal of Cognitive Neuroscience 9648 ndash663

Stroop J R (1935) Studies of interference in serial verbalreactions Journal of Experimental Psychology 18643 ndash662

Tootell R B amp Hadjikhani N (2001) Where is lsquolsquodorsal V4rsquorsquo inhuman visual cortex Retinotopic topographic andfunctional evidence Cerebral Cortex 11 298 ndash311

Treisman A amp Davies A (1973) Dividing attention to ear andeye In S Kornblum (Ed) Attention and performance IV(pp 101ndash107) New York Academic Press

Van Selst M Ruthruff E amp Johnston J C (1999) Canpractice eliminate the psychological refractory period effectJournal of Experimental Psychology Human Perceptionand Performance 25 1268 ndash1283

Van Veen V Cohen J D Botvinick M M Stenger V A ampCarter C S (2001) Anterior cingulate cortex conflictmonitoring and levels of processing Neuroimage 141302 ndash1308

Wojciulik E amp Kanwisher N (1999) The generality ofparietal involvement in visual attention Neuron 23747 ndash764

Zysset S Muller K Lohmann G amp von Cramon D Y(2001) Colorndashword matching stroop task Separatinginterference and response conflict Neuroimage 13 29ndash36


If RS is a central cognitive process then RS involvingdifferent input and output modalities should sharecommon neural substrates This hypothesis is testedhere by asking whether there are brain regions thatare activated in common by different kinds of RSindependent of input and output modalities A compan-ion article (Jiang amp Kanwisher 2003) addresses a secondprediction based on the central bottleneck view (Pashler1989) that some regions involved in RS should not alsobe involved in perceptual processing

This study tests whether common regions are in-volved in various kinds of RS by asking three questionsFirst what brain regions are engaged in RS in a visualndashmanual task Second are these regions specific to thevisual modality or are they also involved in RS forauditory stimuli Third are any common brain regionsthat are recruited in both visual and auditory RS en-gaged only when the two tasks share manual output orspatial mapping requirements Although a few priorinvestigations of response interference in Stroop andother tasks have addressed the first question (Barch etal 2001 Dassonville et al 2001 Connolly Goodale

Desouza Menon amp Vilis 2000 Hazeltine Poldrackamp Gabrieli 2000 Botvinick Nystrom Fissell Carter ampCohen 1999 Carter et al 1998 Carter Botvinick ampCohen 1999 Iacoboni Woods amp Mazziotta 1996Pardo Pardo Janer amp Raichle 1990) few if any priorstudies have investigated the second and third ques-tions (Schumacher amp DrsquoEsposito 2002) Based on thebehavioral literature we predicted that RS based ondifferent input modalities mapping paradigms (whetherspatial or not) and output modalities should recruitcommon brain regions

RESULTS

Behavioral Performance

Figure 1 illustrates the SndashR mapping rules used inExperiments 1 and 2 Responses were either spatiallycompatible or incompatible with the stimuli Table 1shows mean RT and accuracy collected during scan-ning in the three experiments In all experiments in-compatible mapping was substantially slower thancompatible mapping In the auditoryndashmanual task itwas also less accurate Thus we were successful inmanipulating the difficulty of SndashR mappings (ie RS)in all experiments

fMRI Data from the Visual ndashManual Task

We split data from the visual RS task into two indepen-dent halves and carried out a random effects analysis onone half to generate group regions of interest (ROI)involved in visualndashmanual RS The percent signal change(PSC) within these ROIs was then measured in the otherhalf of the visual RS data in the auditoryndashmanual RSdata and in the visual-verbal RS data

Random Effects Analysis

A random effects analysis was conducted on one half ofeach subjectrsquos data (N = 17) searching for any brainregions that had higher BOLD for the spatially incom-patible mapping than compatible mapping ( p lt 001uncorrected) This analysis revealed (see Figure 2)bilateral parietal activation that followed the intraparietalsulcus (IPS) and extended to the superior parietal lobule

Figure 1 Sample display and the SndashR mapping rules for thevisualndashmanual task The task was to pick out the line with uniquelength and report its spatial position among the four lines Targetlocation was mapped to response location according to either naturalmapping (left) or unnatural mapping (right) The same mapping rules(natural and unnatural mapping) were used in the auditory ndashmanualtask in which subjects reported the temporal position of a unique toneamong four tones

Table 1 Behavioral Performance during Scanning

Accuracy () RT (msec)

Experiment Unnatural Natural SE p Unnatural Natural SE p

1 Visual RS 94 95 2 ns 561 383 22 0001

2 Auditory RS 76 82 2 01 717 544 22 0001

3 Verbal naming NA 511 371 24 0001

RS = response selection SE = standard error ns = nonsignificant NA = not applicable


























222 iexcl27 3 48 L FEF (area 6) 536 12

92 33 3 48 R FEF (area 6) 539 12

182 iexcl48 9 21 L GFi (area 4446) 502 12

iexcl45 33 18 L GFm (area 46) 558 9

90 36 24 27 R GFm (area 46) 500 6














aIPS 017 029 008 025

pIPS 010 027 010 026

FEF 022 037 016 027




GFi 009 022 NA



p lt 01

p lt 001


Head Motion



















aIPS 029 048 012 018+ 026 037 004 019

pIPS 009 030 007 019 008 021 005 020

FEF 042 060 024 038 025 040 012 024



Precuneus NA 025 049 006 028

GFi 051 061 007 019 NA



p lt 05

p lt 01

p lt 001





Negative Activation





DISCUSSION











aIPS iexcl003 010 012 027 iexcl003 015 007 023


FEF 012 021 018 033 009 021 014 025




GFi 012 025 007 018 NA



p lt 05

p lt 01

p lt 001






Practice Effects








METHODS

Subjects


Testing Procedure


Scan Composition


Materials and Tasks













fMRI Data Analysis







Acknowledgments




REFERENCES




































































































222 iexcl27 3 48 L FEF (area 6) 536 12

92 33 3 48 R FEF (area 6) 539 12

182 iexcl48 9 21 L GFi (area 4446) 502 12

iexcl45 33 18 L GFm (area 46) 558 9

90 36 24 27 R GFm (area 46) 500 6














aIPS 017 029 008 025

pIPS 010 027 010 026

FEF 022 037 016 027




GFi 009 022 NA



p lt 01

p lt 001


Head Motion



















aIPS 029 048 012 018+ 026 037 004 019

pIPS 009 030 007 019 008 021 005 020

FEF 042 060 024 038 025 040 012 024



Precuneus NA 025 049 006 028

GFi 051 061 007 019 NA



p lt 05

p lt 01

p lt 001





Negative Activation





DISCUSSION











aIPS iexcl003 010 012 027 iexcl003 015 007 023


FEF 012 021 018 033 009 021 014 025




GFi 012 025 007 018 NA



p lt 05

p lt 01

p lt 001






Practice Effects








METHODS

Subjects


Testing Procedure


Scan Composition


Materials and Tasks













fMRI Data Analysis







Acknowledgments




REFERENCES




















































































aIPS 017 029 008 025

pIPS 010 027 010 026

FEF 022 037 016 027




GFi 009 022 NA



p lt 01

p lt 001


Head Motion



















aIPS 029 048 012 018+ 026 037 004 019

pIPS 009 030 007 019 008 021 005 020

FEF 042 060 024 038 025 040 012 024



Precuneus NA 025 049 006 028

GFi 051 061 007 019 NA



p lt 05

p lt 01

p lt 001





Negative Activation





DISCUSSION











aIPS iexcl003 010 012 027 iexcl003 015 007 023


FEF 012 021 018 033 009 021 014 025




GFi 012 025 007 018 NA



p lt 05

p lt 01

p lt 001






Practice Effects








METHODS

Subjects


Testing Procedure


Scan Composition


Materials and Tasks













fMRI Data Analysis







Acknowledgments




REFERENCES












































































Head Motion



















aIPS 029 048 012 018+ 026 037 004 019

pIPS 009 030 007 019 008 021 005 020

FEF 042 060 024 038 025 040 012 024



Precuneus NA 025 049 006 028

GFi 051 061 007 019 NA



p lt 05

p lt 01

p lt 001





Negative Activation





DISCUSSION











aIPS iexcl003 010 012 027 iexcl003 015 007 023


FEF 012 021 018 033 009 021 014 025




GFi 012 025 007 018 NA



p lt 05

p lt 01

p lt 001






Practice Effects








METHODS

Subjects


Testing Procedure


Scan Composition


Materials and Tasks













fMRI Data Analysis







Acknowledgments




REFERENCES























































































aIPS 029 048 012 018+ 026 037 004 019

pIPS 009 030 007 019 008 021 005 020

FEF 042 060 024 038 025 040 012 024



Precuneus NA 025 049 006 028

GFi 051 061 007 019 NA



p lt 05

p lt 01

p lt 001





Negative Activation





DISCUSSION











aIPS iexcl003 010 012 027 iexcl003 015 007 023


FEF 012 021 018 033 009 021 014 025




GFi 012 025 007 018 NA



p lt 05

p lt 01

p lt 001






Practice Effects








METHODS

Subjects


Testing Procedure


Scan Composition


Materials and Tasks













fMRI Data Analysis







Acknowledgments




REFERENCES















































































Negative Activation





DISCUSSION











aIPS iexcl003 010 012 027 iexcl003 015 007 023


FEF 012 021 018 033 009 021 014 025




GFi 012 025 007 018 NA



p lt 05

p lt 01

p lt 001






Practice Effects








METHODS

Subjects


Testing Procedure


Scan Composition


Materials and Tasks













fMRI Data Analysis







Acknowledgments




REFERENCES













































































DISCUSSION











aIPS iexcl003 010 012 027 iexcl003 015 007 023


FEF 012 021 018 033 009 021 014 025




GFi 012 025 007 018 NA



p lt 05

p lt 01

p lt 001






Practice Effects








METHODS

Subjects


Testing Procedure


Scan Composition


Materials and Tasks













fMRI Data Analysis







Acknowledgments




REFERENCES
















































































Practice Effects








METHODS

Subjects


Testing Procedure


Scan Composition


Materials and Tasks













fMRI Data Analysis







Acknowledgments




REFERENCES














































































METHODS

Subjects


Testing Procedure


Scan Composition


Materials and Tasks













fMRI Data Analysis







Acknowledgments




REFERENCES




















































































fMRI Data Analysis







Acknowledgments




REFERENCES
















































































Acknowledgments




REFERENCES












































































Common Neural Substrates for Response Selection across ...

Documents

Transcript of Common Neural Substrates for Response Selection across ...