Delay-period Activity in the Prefrontal CortexOne Function Is Sensory Gating

Bradley R Postle

Abstract

amp The prefrontal cortex (PFC) contributes to workingmemory functions via executive control processes that donot entail the storage per se of mnemonic representationsOne of these control processes may be a sensory gatingmechanism that facilitates retention of representations inworking memory by down-regulating the gain of the sensoryprocessing of intervening irrelevant stimuli This idea wastested by scanning healthy young adults with functionalmagnetic resonance imaging while they performed a delayedface-recognition task The 2 2 factorial design varied thefactors of Memory (present absent) and Distraction (presentabsent) During memory-present trials target and probestimuli were individual gray-scale male faces Memory-absenttrials were identical except that they employed the same

recurring female faces (denoting a lsquolsquono memoryrsquorsquo trial)Distraction-present trials featured rapid serial visual presenta-tion of bespectacled male faces during the two middle secondsof the delay The first step of the analyses identifieddorsolateral PFC (dlPFC) and inferior occipitotemporal cortex(IOTC) voxels exhibiting delay-period activity in memory-presentdistraction-absent trials that is the lsquolsquounfilledrsquorsquo delayWithin these voxels distraction-evoked activity in the dlPFCwas markedly higher during trials that required the concurrentshort-term retention of information than on those that did notwhereas the opposite effect was seen in the IOTC Theseresults are consistent with the view that processes related tosensory gating account for a portion of the delay-period activitythat is routinely observed in the dlPFC amp

INTRODUCTION

The prefrontal cortex (PFC) makes important contribu-tions to the control of cognitive processes includingworking memory (Stuss amp Knight 2002 Monsell ampDriver 2000) But what is the function of the delay-period activity that is observed almost universally inthe PFC of monkeys and humans performing simpletasks of short-term memory that make no overt de-mands on executive control By one influential viewit represents the active short-term retention (STR) ofinformation (Courtney 2004 Goldman-Rakic amp Leung2002) This view in effect equates different regionsof the PFC with the storage buffers of Baddeley andHitchrsquos (1974) model of working memory for a moredetailed review see Postle Druzgal and DrsquoEsposito(2003) An alternative approach to this question incontrast comes from a framework that holds thatworking memory functions do not arise from the oper-ation of one or more specialized systems but rather thatthey arise from the recruitment via attention of systemsthat have evolved to accomplish sensory representa-tional andor motoric goals From this perspective thePFC is not the neural substrate for working memorystorage buffers because the STR of information is

supported in a domain-specific manner in areas re-sponsible for perception knowledge representationlanguage and motor control (eg Jonides Lacey ampNee 2005 Pasternak amp Greenlee 2005 Postle AwhJonides Smith amp DrsquoEsposito 2004 Postle et al 2003Ruchkin Grafman Cameron amp Berndt 2003 PostleBerger amp DrsquoEsposito 1999 Postle amp DrsquoEsposito 19992003) This view derives support from a considerablebody of neuropsychological data indicating that the STRof information (as measured by span and delay tasks)does not depend on the integrity of the PFC whenmonkeys (eg Petrides 2000 Malmo 1942) and hu-mans (eg DrsquoEsposito amp Postle 1999 Chao amp Knight1998 Chao amp Knight 1995 Milner 1964) are testedunder optimal conditions

If it is not storage what might be the function ofdelay-period activity in PFC There are several possibil-ities including attentional selection (Rowe StephanFriston Frackowiak amp Passingham 2005 LebedevMessinger Kralik amp Wise 2004 Kane amp Engle 2003Passingham amp Rowe 2002 Rowe Toni Josephs Fracko-wiak amp Passingham 2000) top-down control (vari-ously described eg as lsquolsquoguided activationrsquorsquo [Miller ampCohen 2001]lsquolsquoadaptive codingrsquorsquo [Duncan amp Miller2002] and lsquolsquoproactive controlrsquorsquo [Braver Gray amp Burgessin press]) stimulus transformation and response prep-aration (eg Curtis Rao amp DrsquoEsposito 2004 FukushimaUniversity of WisconsinndashMadison

D 2005 Massachusetts Institute of Technology Journal of Cognitive Neuroscience 1711 pp 1679ndash1690

Hasegawa amp Miyashita 2004 Bor Duncan Wisemanamp Owen 2003 Takeda amp Funahashi 2002 2004 Pochonet al 2001 DrsquoEsposito Ballard Zarahn amp Aguirre 2000Fuster 1995) and motivation andor reward expectancy(eg Watanabe 1996 2002) For extensive reviews ofthis question see Curtis and DrsquoEsposito (2003) andPassingham and Sakai (2004) The present study wasdesigned to test yet another candidate function of PFCdelay-period activitymdashthe mediation of interference

One indication that the PFC plays an important role ininterference mediation comes from the fact that impair-ments begin to emerge in the working memory per-formance of PFC-lesioned subjects when distraction orinterference is present in the environment (DrsquoEspositoamp Postle 1999 Chao amp Knight 1998 Chao amp Knight1995 Malmo 1942) or in the cognitive system (as is thecase with proactive interference [PI] Thompson-Schillet al 2002 Milner 1964) Further evidence comes fromelectrophysiological and neuroimaging studies Withelectrophysiology Chao and Knight (1998) providedevidence that lateral PFC contributes to lsquolsquoattentionalfilteringrsquorsquo or lsquolsquosensory gatingrsquorsquo by showing that theevoked response to distracting auditory stimuli wasgreater for patients with PFC lesions than for controlsubjects With PET and fMRI several groups have inves-tigated the selective sensitivity to PI of the portion of theleft inferior frontal gyrus corresponding to Brodmannrsquosarea (BA 45) (eg Postle amp Brush 2004 Postle Brush ampNick 2004 Postle Berger Goldstein Curtis amp DrsquoEspo-sito 2001 Jonides Marshuetz Smith Reuter-Lorenz ampKoeppe 2000 DrsquoEsposito Postle Jonides amp Smith1999 Jonides Smith Marshuetz Koeppe amp Reuter-Lorenz 1998) Two other neuroimaging studies havegone beyond demonstrating the PFCrsquos sensitivity tointerference by showing that the PFC plays an active rolein its mediation In one Sakai Rowe and Passingham(2002) found that PFC activity during an unfilled delayperiod predicted task accuracy on trials when this un-filled period was followed by a distractor task Theyattributed this to a PFC-controlled lsquolsquoactive maintenancersquorsquoprocess that strengthened mnemonic representations viathe strengthening of the coupling of activity between thesuperior frontal cortex (BA 8) and intraparietal sulcus1 Inanother study Gray Chabris and Braver (2003) foundthat PFC activity correlated with performance on trials ofan n-back working memory task that featured a high levelof PI but only in subjects who had been independentlydetermined to be of high general fluid intelligence Likethese two the present study was also designed toevaluate evidence for a PFC-based control mechanisminvolved in the mediation of interference Its designwould permit a two-stage process of (1) the identificationof delay-period activity from unfilled delay periods fol-lowed by (2) an assessment of whether the voxelsidentified in Step 1 might contribute to a sensory gatingfunction Thus it would explicitly evaluate one candidateexplanation of the function of PFC delay-period activity

Although there are many ways to implement a gatingmechanism the present study was designed to detectone that operates by regulating the gain of sensoryprocessing such that the processing of potentially dis-rupting afferent sensory information is down-regulatedwhen this information might interfere with the STR ofother behaviorally prioritized information There isconsiderable evidence for such a mechanism from stud-ies measuring event-related potentials In these studiesan increase in alpha-band power over posterior areasduring short-term memory task performance is inter-preted as evidence for lsquolsquoinhibition of task irrelevantprocessesrsquorsquo (Klimesch Doppelmayr Schwaiger Auingeramp Winker 1999 p 403 see also Jensen GelfandKounios amp Lisman 2002 Chao amp Knight 1998) Thetask featured in this report was delayed face recognitionimplemented in a 2 2 design that varied the factors ofmemory (present absent) and distraction (presentabsent Figure 1) The logic of the experiment dependedon the assumption that the PFC delay-period activitythat was expected on memory-presentdistraction-absent trials would not reflect the STR of face informa-tion This assumption was supported by two previousstudies of delayed face recognition indicating that PFCdelay-period activity was logically incompatible with aSTR function One of these studies employed a multi-step ABBA-like design intended to winnow out delay-period activity that may be correlated with but notnecessary for the STR of face information Althoughthe results from each subject revealed Delay 1-specificactivity in many brain areas including the PFC posteriorfusiform gyrus and posterior parietal cortex only asubset of these voxels retained the Delay 1 signal duringDelay 2 and the posterior fusiform gyrus was the onlyregion in which in each subject voxels retained theDelay 1 signal during Delay 3 (Postle et al 2003) Thesecond study demonstrated that varying instructionsmdashlsquolsquoremember face(s)rsquorsquo versus lsquolsquoremember scene(s)rsquorsquomdashmodulated the delay-period activity of the fusiform facearea or the parahippocampal place area in a category-specific manner but had no effect on PFC (ie PFCactivity did not distinguish between STR of faces vs STRof houses Ranganath DeGutis amp DrsquoEsposito 2004)Instead of STR the sensory gating hypothesis posits thatsome portion of PFC delay-period activity on memory-presentdistraction-absent trials reflects a tonically activemechanism that monitors the environment for poten-tially distracting stimuli or events and that regulates thegain of sensory processing when needed

And so the logic of the experimental approach was (1)to identify face memory-related activity in the dorsolat-eral PFC (dlPFC) and inferior occipitotemporal cortex(IOTC) by identifying voxels in these two regions dem-onstrating significant delay-period activity in memory-presentdistraction-absent trials (ie in lsquolsquostandardrsquorsquodelayed-recognition conditions) and (2) to assess thedistraction-evoked activity in these voxels of interest

1680 Journal of Cognitive Neuroscience Volume 17 Number 11

(VOIs) by measuring delay-period activity from memory-absentdistraction-present and memory-presentdistrac-tion-present trials We assumed that a dlPFC-basedsensory gating mechanism would be engaged to a largerextent by the conjunction of the need to retain amemory with the presence of distracting noise in theenvironment than by the presence of distraction aloneThus we predicted greater delay-period activity in dlPFCVOIsmdashthe hypothesized source of controlmdashin memory-presentdistraction-present than in memory-absentdistraction-present trials In contrast we predicted thatthe gain-modulating effects of this gating process wouldresult in decreased activity at the site of its action duringthe conjunction of memory and distraction that is torelatively decreased delay-period activity in IOTC VOIs inmemory-presentdistraction-present as compared withmemory-absentdistraction-present trials Restated interms of our experimental design we predicted aninteraction such that the manipulation of the factor ofmemory on distraction trials would have a greater effecton dlPFC VOIs than on IOTC VOIs We did not knowa priori whether the outcome would be a crossoverinteraction (ie whether in IOTC VOIs the delay-periodeffect on memory-presentdistraction-present trialswould in fact be quantitatively lower than on memory-absentdistraction-present trials) or whether the differ-ence would be in the same direction as in dlPFC VOIsbut markedly smaller

RESULTS

Behavior

Behavioral results presented in Table 1 indicated thatperformance was high in all conditions (Due to techni-cal problems accuracy data were not collected from two

subjects and reaction time [RT] data were not collectedfrom four subjects) An analysis of variance (ANOVA)assessing accuracy as a function of memory (presentabsent) and distraction (present absent) revealed amain effect of memory that approached significanceF(113) = 36 p = 08 a main effect of distractionF(113) = 52 p lt 05 and an interaction F(113) =181 p lt 001 The analogous analysis of RT datarevealed main effects of memory F(111) = 512p lt 0001 and distraction F(111) = 228 p lt 001and an interaction F(111) = 215 p lt 001 The maineffect of distraction reflects the fact that RTs were fasterfor trials with distraction than without it Although thiseffect was not expected it may be that the rapid serialvisual presentation (RSVP) distraction elicited a generalincrease of arousal or a greater level of motor prepara-tion (eg Connolly Goodale Goltz amp Munoz 2005)Together the accuracy and RT results confirmed theintuition that of the two critical conditions in this

Figure 1 Illustration of a

memory-presentdistraction-

absent trial and a memory-

absentdistraction-present trial

Table 1 Behavioral Performance Accuracy (Mean ProportionCorrect) and RT (in Milliseconds)

Memory

Distraction Present Absent

Present

Accuracy (SE) 0874 (0020) 0931 (0023)

RT (SE) 7661 (401) 4744 (368)

Absent

Accuracy (SE) 0945 (0021) 0934 (0024)

RT (SE) 8036 (381) 6339 (478)

Postle 1681

study memory-absentdistraction-present and memory-presentdistraction-present the latter would be themore difficult

fMRI

Delay-period activity was observed on memory-presentdistraction-absent trials in the dlPFC and the IOTC ineach of the 16 subjects (Figure 2) Within the function-ally defined VOIs distractor-evoked activity was mark-edly higher for memory-present than for memory-absenttrials in the dlPFC whereas the IOTC showed a trend inthe opposite direction (Figures 3 and 4) ANOVA assess-ing raw distraction-evoked activity in VOIs in the tworegions (dlPFC IOTC) and the two critical trial types(memory-absentdistraction-present memory-presentdistraction-present trials) revealed a main effect of re-gion F(115) = 1197 p lt 005 an effect of trial typethat approached significance F(115) = 396 p = 07and an interaction F(115) = 1523 p = 001 Theseresults were broadly consistent with our predictionsHowever because the main effect of region indicatedthat the overall evoked response was greater in theIOTC than in the dlPFC it was possible that the differ-

ential patterns of activity observed in these two VOIswere attributable at least in part to a difference in theinherent level of activity in the two regions in which theywere located To rule out this possibility the effectsfrom each subjectrsquos VOIs were normalized by dividingwithin VOI the distractor-evoked effects from both trialtypes of interest by the memory-absentdistraction-present effect This had the effect of transforming thememory-absentdistraction-present effect in both re-gions from each subject to the arbitrarily selected valueof 10 and transforming the memory-presentdistraction-present effect into a proportion of the memory-absentdistraction-present effect After this transformation wasperformed we excluded the data from one subject(12) from further analyses because the normalizedmemory-presentdistraction-present effect from thedlPFC VOIs of this subject was in excess of 3 standarddeviations greater than the mean2 For the remaining 15subjects the normalized data indicated that the meandelay-evoked response on memory-presentdistraction-present trials in dlPFC VOIs was 139 greater than onmemory-absentdistraction-present trials [a significantpairwise effect at t(14) = 21 p = 05] whereas in IOTCVOIs the mean delay-evoked response on memory-

Figure 2 Two representative slices from each subject illustrating dlPFC and IOTC VOIs Memory-presentdistraction-absent activity wasobserved in the dlPFC bilaterally in 13 subjects and unilaterally in the right hemisphere in 2 subjects and in the left hemisphere in 1 subject

in the IOTC it was observed bilaterally in 4 subjects unilaterally in the right hemisphere in 9 subjects and unilaterally in the left hemisphere

in 3 subjects

1682 Journal of Cognitive Neuroscience Volume 17 Number 11

presentdistraction-present trials was 75 smaller thanit was on memory-absentdistraction-present trials [apairwise effect that was not significant t(14) = 12ns] Thus the pattern of the normalized effects retained

the features of the raw data (Figure 4B) as confirmed byANOVA assessing the normalized results from the re-maining 15 subjects which produced a main effect ofregion F(114) = 49 p lt 05 a borderline effect of trial

Figure 3 An example of

results from a single subject

(21 in Figure 2) The two

top plots illustrate the trial-averaged time series data from

dlPFC and IOTC VOIs from

memory-presentdistraction-absent trials (fMRI MPDA)

and the corresponding delay-

period covariates scaled by

their parameter estimates(delay effect) The gray

bars along the horizontal axes

indicate the duration of the

delay period The two bottomplots illustrate the delay effects

from these same VOIs from

memory-presentdistraction-present (MPDP) and

memory-absentdistraction-

present (MADP) trials

Note that the signal intensityscales differ on plots for the

two regions

Figure 4 Group averageeffects (A) Memory-present

distraction-absent delay-period

VOIs with unscaled data (B)memory-presentdistraction-

absent delay-period VOIs with

normalized data (C) target-

evoked VOIs with unscaleddata (D) target-evoked VOIs

with normalized data MPDA =

memory-presentdistraction-

absent MPDP = memory-presentdistraction-present

MADP = memory-absent

distractionpresent Errorbars represent SEM

Postle 1683

type F(114) = 38 p = 07 and an interaction F(114) =49 p lt 05

One alternative to the sensory gating interpretation ofthese results might be that PFC delay-period activity onmemory-presentdistraction-present trials reflected dualtaskingmdashthe STR of target face information plus theincidental encoding (or some other type of processing)of the distracting faces This seems unlikely howeverbecause PFC delay-period activity was numerically loweron memory-presentdistraction-present trials than onmemory-presentdistraction-absent trials [although thiseffect did not achieve significance t(15) = 13 nsFigure 4A]

How specific was this pattern of results To assess thisquestion we applied the same contrasts to voxels in thedlPFC and IOTC that were selected for a differentpropertymdashresponsivity to visually presented faces Spe-cifically we identified voxels in the dlPFC and IOTC thatresponded to the Target epoch of the task (and re-moved from these VOIs any voxels that were also activeduring the delay period of memory-presentdistraction-absent trials) In this way we identified voxels from thesame regions in which we tested the sensory gatinghypothesis but whose function was different (presum-ably related to visual perception andor memory encod-ing but not to retention or another delay-periodprocess) For the raw data from these target-evokedVOIs (Figure 4C) as compared to the memory-evokedVOIs there was again a strong main effect of regionF(114) = 334 p lt 0001 a weaker borderline effect oftrial type F(114) = 33 p = 09 and now only a trendtoward an interaction F(114) = 40 p 07 When datafrom the target-evoked VOIs were normalized in thesame manner as had been the data from the delay-evoked VOIs (Figure 4D) their analysis with ANOVArevealed only a borderline effect of VOI F(114) = 40p = 06 for the first time a reliable effect of trial typeF(114) = 51 p lt 05 and again a trend toward aninteraction F(114) = 40 p = 06

The weakening of the Region Trial type interactionin target-evoked VOIs as compared with delay-evokedVOIs could be consistent with the claim that this inter-action predicted to be the fMRI signature of a sensorygating mechanism was selectively seen only in the delay-evoked VOIs It could also however simply reflect asituation in which the interaction terms for the twotypes of VOI fell on either side of the 05 thresholdbut did not differ from each other To distinguishbetween these two possibilities we compared the ef-fects of the experimental manipulation on delay-periodactivity across the two types of VOIs with an omnibusANOVA that directly compared the effect of memory onnormalized distraction-evoked activity across target-evoked and delay-evoked VOIs The factors for theomnibus ANOVA were region (dlPFC IOTC) trial type(memory present memory absent) and epoch definingthe VOIs (target delay) A three-way interaction would

be necessary to support the claim that the differentialeffect of memory demands on distraction-evoked activityin the dlPFC and IOTC was greater for delay-evoked thanfor target-evoked VOIs The ANOVA revealed a maineffect of region F(114) = 49 p lt 05 a borderlineeffect of trial type F(114) = 43 p = 06 and no effect ofepoch F(114) = 28 p gt 1 no interaction of Epoch Trial type F(114) = 28 p gt 1 but reliable inter-actions of Epoch Region F(114) = 47 p lt 05and of Region Trial type (F(114) = 49 p lt 05 and areliable interaction of Epoch Region Trial typeF(114) = 47 p lt 05 Thus it confirmed the specificityof the sensory gating effect to delay-evoked VOIs

DISCUSSION

The results demonstrate that delay-sensitive voxels intwo brain regions the dlPFC and the IOTC responddifferentially to the RSVP of faces conditioned on thebehavioral context in which these faces are presentedThe dlPFC response was 139 greater during concur-rent face memory whereas the IOTC response was 75lower This pattern was also seen to be selective becausea different group of dlPFC voxels (those that are targetsensitive) increased its activity by only 38 and theanalogously different group of IOTC voxels increased itsactivity by 3 These results are therefore consistentwith a model in which one function of dlPFC voxels thatare active during unfilled delay periods of delayed-recognition tasks is to perform a sensory gating functionBy this account the detection of potentially interferingstimuli in the environment triggers activity in networksof dlPFC neurons3 that in turn has the effect of sup-pressing perhaps via thalamic relays the afferent flow ofsensory information (Knight Staines Swick amp Chao1999) This would have beneficial effects for the signal-to-noise ratio in the networks of posterior neurons thatare responsible for the STR of information Thus a PFC-based control process may serve to protect the contentsof working and short-term memory from the deleteriouseffects external interference although it does not sup-port STR operations themselves

This account makes two broader claims about theworking memory literature First it offers an explanationas to why primates with PFC lesions can perform nor-mally on tasks of working and short-term memory whenpotential sources of distraction and inference are mini-mal but are impaired on the same tasks when distractionandor interference is high (Thompson-Schill et al2002 DrsquoEsposito amp Postle 1999 Chao amp Knight 1998Chao amp Knight 1995 Milner 1964 Malmo 1942)Second it proposes that sensory gating may accountfor a portion of the unfilled delay-period activity that isroutinely observed in PFC across a variety of neuro-imaging studies including those that do not explicitlyinclude distraction In relation to this second claim how-

1684 Journal of Cognitive Neuroscience Volume 17 Number 11

ever we cannot rule out the possibility that the dlPFCactivity observed during memory-presentdistraction-absent trials was due to a specific feature of our designBecause subjects could not predict which memorytrials would also contain distraction an effective strat-egy may have been to implement a sensory gatingstrategy on every memory trial To rule this out wouldrequire a two stage process of first scanning lsquolsquonaıversquorsquosubjects performing several blocks of only memory-presentdistraction-absent trials then introducing thefactor of delay-period distraction for a subsequent setof scans

The present study fits within a growing body of resultsthat document control rather than STR functions forthe PFC These include the monitoring and transforma-tion of information held in working memory (Bor et al2003 DrsquoEsposito Postle Ballard amp Lease 1999 Pet-rides 1995) and the attentional selection of information(Lebedev et al 2004 Sakai et al 2002 de Fockert ReesFrith amp Lavie 2001) These approaches which proposewell-defined testable mechanisms represent an advancebeyond the vague notion of lsquolsquomaintenancersquorsquo that is ofteninvoked in interpretations of PFC activity The presentstudy suggests one such alternative to lsquolsquomaintenancersquorsquo toaccount for sustained PFC activity observed during thesimplest tests of STR of information

METHODS

Subjects

Sixteen healthy young adults (mean age = 224 13 men)participated in this study All provided informed consentand were paid in compensation for participating in thestudy

Behavioral Procedure

The 2 2 factorial design featured the factors ofMemory (present absent) and Distraction (presentabsent) Each trial had the same structure Target (1 sec)followed by Delay (7 sec) followed by Probe (1 sec)followed by ITI (13 sec) During memory-present trialstarget and probe stimuli were individual gray-scale malefaces (none wearing glasses) appearing on white back-grounds and subjects indicated with a button presswhether the latter matched the former ( p = 5) Memory-absent trials were procedurally identical except thatthey featured the same recurring female facesmdashfacelsquolsquoArsquorsquo always the target which denoted a lsquolsquono-memoryrsquorsquotrial face lsquolsquoBrsquorsquo always the probe On memory-absenttrials subjects were cued by the probe to indicate witha button press whether or not distracting faces appearedduring the delay period ( p = 5) thereby matching themotor response demands of memory-present trialsDistraction-present trials featured RSVP (3 Hz) of thefaces of males wearing glasses (to differentiate distrac-

tion items from memoranda) during the two middleseconds of the delay (Figure 1) Subjects were instructedto look at the distracting faces during all trials on whichthey appeared and to fixate centrally on distraction-absent trials Stimuli were presented via eyepiecesmounted on the head coil (Avotec Visible Eye StuartFL) and fitted with an infrared-based eye tracking system(SMI Inc Teltow Germany) Fixation was monitoredduring scanning and analyzed off-line with iLab software(Gitelman 2002) Trials during which subjects failedto fixate centrally from time 3 sec to time 5 sec werediscarded from subsequent analyses (Figure 5) Thiscriterion ensured that any evidence for differencesin delay-period activity between conditions could notbe attributed to theoretically uninteresting differencesin for example the amount of visual stimulation pro-duced by the distracting stimuli or the amount ofoculomotor activity Trials of each of the four types wereequiprobable and presented in a randomly determinedorder

fMRI Data Acquisition

Whole-brain images were acquired with a 3T scanner(GE Signa VHI Waukesha WI) High-resolution T1-weighted images (32 axial slices 09375 09375 4 mm)were obtained in every participant and a gradient-echoecho-planar sequence (TR = 2000 msec TE = 50 msec)was used to acquire data sensitive to the blood oxygenlevel dependent (BOLD) signal (Kwong et al 1992Ogawa et al 1992) within a 64 64 matrix (32 axialslices coplanar with T1 375 375 4 mm) Scans ofthe delayed-recognition task were preceded by a scan inwhich we derived an estimate of the hemodynamicresponse function (HRF) for each participant Duringthis scan each participant performed a simple RT taskthat required a bimanual button press once every 20 secin response to a brief change in shape of the fixationstimulus A partial F test associated with a Fourier basiscovariate set ( Josephs Turner amp Friston 1997) wasused to evaluate the significance of task-correlatedactivity in each voxel of primary somatosensory andmotor cortical regions of interest (Aguirre Zarahn ampDrsquoEsposito 1998) An HRF estimate was extracted fromthe suprathreshold voxels of these ROIs by spatiallyaveraging their time series filtering the resultant aver-aged fMRI time series to remove high (gt0244 Hz)and low (lt005 Hz) frequencies adjusting it to removethe effects of nuisance covariates (Friston HolmesPoline Heather amp Frackowiak 1995) and trial aver-aging The HRF characterizes the fMRI response result-ing from a brief impulse of neural activity (BoyntonEngel Glover amp Heeger 1996) and can vary markedlyacross participants (Aguirre et al 1998) The subject-specific HRFs were used to convolve independent vari-ables entered into the modified general linear model(GLM Worsley amp Friston 1995) that we used to analyze

Postle 1685

the data from the scans of the delayed-recognition task4

The eight scans of the delayed-recognition task eachlasted 6 min 12 sec (5 min 52 sec of task preceded by20 sec of dummy pulses to achieve a steady state of tis-sue magnetization)

fMRI Data Processing

The fMRI time series data were filtered and adjusted asdescribed previously (Postle Zarahn amp DrsquoEsposito2000) The principle of the fMRI time series analysiswas to model the fMRI signal changes evoked byeach stimulus presentation epoch with covariates com-posed of BOLD impulse response functions shiftedalong the time line of the task to represent various trialepochs (Figure 1 Postle et al 2000 Zarahn Aguirre

amp DrsquoEsposito 1997) The least-squares solution of theGLM of the fMRI time series data yielded parameterestimates that were associated with each covariateof interest The smoothness of the fMRI response toneural activity allows fMRI-evoked responses that arisefrom temporally dependent events to be resolved onthe order of 4 sec (That is if the covariates modelingthese events are spaced at 4-sec intervals the covariatepositioned at time 0 sec will account for virtually all ofthe variance attributable to an event starting at time 0 secand of duration of 1 sec or less thereby leaving thecovariate positioned at time 4 sec relatively uncontam-inated by variance attributable to the time 0 sec eventsee Figure 6 of Zarahn et al 1997) Differences in fMRIsignal (either between conditions or vs baseline) weretested by computing t statistics from contrasts between

Figure 5 Examples of eye-position data from four representative trials For all four examples trial epoch is color coded red = target stimulus

green = early delay blue = distraction portion of the delay period fuchsia = late delay black = probe stimulus plus first second of ITI (A) A

memory-presentdistraction-present trial with acceptable fixation during distraction The top panel is a 2-D representation of eye position on the

display lsquolsquoscreenrsquorsquo over the course of the trial axis labels correspond to degrees of visual angle Bottom panels display horizontal (x) and vertical ( y)eye position as a function of time with time in seconds on the horizontal axes and position in pixels on the vertical axes (each pixel measured

058 mm2) (B) A memory-absentdistraction-present trial with acceptable fixation during distraction All conventions same as in (A) Interrupted

portion of late delay corresponds to a blink when eyes were closed (C) A memory-presentdistraction-present trial that was discarded due

to unacceptable fixation during distraction All conventions same as (A) (D) A memory-absentdistraction-present trial that was discarded dueto unacceptable fixation during distraction All conventions same as in (A)

1686 Journal of Cognitive Neuroscience Volume 17 Number 11

parameter estimates associated with the covariates inquestion

Our analysis strategy was to identify delay-specificactivity in two regions the dlPFC and the IOTC5 frommemory-presentdistraction-absent trials This wouldidentify delay-period activity comparable to that ob-served in virtually all whole-brain neuroimaging studiesof similar tasks (eg with face stimuli Postle et al 2003Ranganath amp DrsquoEsposito 2001 Druzgal amp DrsquoEsposito2001 Courtney Ungerleider Keil amp Haxby 1997 Court-ney Ungerleider Keil amp Haxby 1996) Masks cor-responding to these two regions were drawn on eachsubjectrsquos anatomical scans Voxels located within thesemasks which became the VOIs were identified on tmaps of delay-period activity from memory-presentdis-traction-absent trials thresholded to a regionwise a of05 (one-tailed) with Bonferroni correction for the num-ber of voxels in the mask Note that with this approachwith spatially unsmoothed data individual voxels whoset value exceeds threshold can be interpreted as lsquolsquomean-ingfulrsquorsquo data (Postle et al 2000) Thus this procedurediffered somewhat from other applications of lsquolsquoregion-

of-interest (ROI) analysesrsquorsquo in that no selection criteriawere imposed with respect to cluster size or proximity ofcritical voxels This was done among other reasonsbecause we did not know a priori in which subregionof PFC to expect to find sensory-gating-related activityWithin the VOIs we evaluated the responses evokedby RSVP of faces during memory-presentdistraction-present trials versus memory-absentdistraction-presenttrials (Memory-absentdistraction-absent trials do notfigure directly into the test of the sensory gating hy-pothesis but were necessary to prevent anticipation ofdistraction on memory-absent trials)

Acknowledgments

The author thanks Olufunsho Faseyitan and Craig Rypstat forprogramming assistance and OF and Christopher Jordan forhelp with data collection and analysis This research was sup-ported by the National Institutes of Health RO1 MH064498

Reprint requests should be sent to Bradley R Postle Depart-mentofPsychologyUniversityof WisconsinndashMadison1202WestJohnson St Madison WI 53706 or via e-mail postlewiscedu

Figure 5 (continued)

Postle 1687

The data reported in this experiment have been depositedwith the fMRI Data Center archive (wwwfmridcorg) Theaccession number is 2-2005-119B9

Notes

1 This may be an empirical example of the lsquolsquoactive main-tenancersquorsquo proposed by Miller and Cohen (2001)2 The raw value of this subjectrsquos memory-absentdistraction-present effect in the dlPFC was 51 and so dividing thememory-presentdistraction-present by this value inflated it bya factor of almost 2 the opposite direction than that effectedby the normalization procedure in all other subjects exceptone whose raw memory-absentdistraction-present effect inthe dlPFC was 983 The fact that delay-period activity in the dlPFC VOIs wasnumerically lower during memory-presentdistraction-presentthan during memory-presentdistraction-absent trials is con-sistent with the idea that the sensory gating mechanism oncetriggered by neurons that are tonically active during unfilleddelay periods is implemented by a broader PFC network thatincludes regions that are not typically active during unfilleddelay periods Although the data are not shown here it iscertainly the case that in every subject considerably more dlPFCvoxels were active during the memory-presentdistraction-present delay than during the memory-presentdistraction-absent delay4 The HRF also varies in shape across brain regions with-in an individual participant (eg Miezen Maccotta OllingerPetersen amp Buckner 2000) However recent work indi-cates that the magnitude of variability in the shape of theHRF is greater across individuals within a region than it isacross regions within an individual (Handwerker Ollinger ampDrsquoEsposito 2004) Additionally it indicates that employinga participant-specific HRF can improve the sensitivity andmagnitude estimates of the least-squares solution of the GLMover comparable analyses that employ a generic HRF model(Handwerker et al 2004)5 The broader search space of the IOTC was used instead ofa more circumscribed area such as the fusiform face areabecause as can be seen in Figure 2 there was a considerableamount of topological variability of memory-presentdistraction-absent delay-period activity in the posterior ventral stream andno one area was activated in all subjects

REFERENCES

Aguirre G K Zarahn E amp DrsquoEsposito M (1998) Thevariability of human BOLD hemodynamic responsesNeuroimage 8 360ndash369

Baddeley A D amp Hitch G J (1974) Working memoryIn G H Bower (Ed) The psychology of learning andmotivation (Vol 8 pp 47ndash89) New York Academic Press

Bor D Duncan J Wiseman R J amp Owen A M (2003)Encoding strategies dissociate prefrontal activity fromworking memory demand Neuron 37 361ndash367

Boynton G M Engel S A Glover G H amp HeegerD J (1996) Linear systems analysis of functionalmagnetic resonance imaging in human V1 The Journalof Neuroscience 16 4207ndash4221

Braver T S Gray J R amp Burgess G C (in press)Explaining the many varieties of working memoryvariation Dual mechanisms of cognitive control In AConway C Jarrold M Kane A Miyake amp J Towse(Eds) Variation in working memory Oxford OxfordUniversity Press

Chao L amp Knight R (1998) Contribution of humanprefrontal cortex to delay performance Journal ofCognitive Neuroscience 10 167ndash177

Chao L L amp Knight R T (1995) Human prefrontal lesionsincrease distractibility to irrelevant sensory inputsNeuroReport 6 1605ndash1610

Connolly J D Goodale M A Goltz H C amp Munoz D P(2005) fMRI activation in the human frontal eye field iscorrelated with saccadic reaction time Journal ofNeurophysiology 94 605ndash611

Courtney S M (2004) Attention and cognitive control asemergent properties of information representation inworking memory Cognitive Affective amp BehavioralNeuroscience 4 501ndash516

Courtney S M Ungerleider L G Keil K amp Haxby J(1996) Object and spatial visual working memory activateseparate neural systems in human cortex Cerebral Cortex6 39ndash49

Courtney S M Ungerleider L G Keil K amp Haxby J V(1997) Transient and sustained activity in a distributedneural system for human working memory Nature 386608ndash611

Curtis C E amp DrsquoEsposito M (2003) Persistent activity inthe prefrontal cortex during working memory Trends inCognitive Sciences 7 415ndash423

Curtis C E Rao V Y amp DrsquoEsposito M (2004) Maintenanceof spatial and motor codes during oculomotor delayedresponse tasks The Journal of Neuroscience 243944ndash3952

DrsquoEsposito M Ballard D Zarahn E amp Aguirre G K(2000) The role of prefrontal cortex in sensory memoryand motor preparation An event-related fMRI studyNeuroimage 11 400ndash408

DrsquoEsposito M amp Postle B R (1999) The dependence of spanand delayed-response performance on prefrontal cortexNeuropsychologia 37 1303ndash1315

DrsquoEsposito M Postle B R Ballard D amp Lease J (1999)Maintenance versus manipulation of information held inworking memory An event-related fMRI study Brain ampCognition 41 66ndash86

DrsquoEsposito M Postle B R Jonides J amp Smith E E(1999) The neural substrate and temporal dynamicsof interference effects in working memory as revealedby event-related functional MRI Proceedings of theNational Academy of Sciences USA 96 7514ndash7519

de Fockert J W Rees G Frith C D amp Lavie N (2001) Therole of working memory in visual attention Science 2911803ndash1806

Druzgal T J amp DrsquoEsposito M (2001) A neural networkref lecting decisions about human faces Neuron 32947ndash955

Duncan J amp Miller E K (2002) Cognitive focus throughadaptive neural coding in the primate prefrontal cortexIn D Stuss amp R Knight (Eds) Principles of frontal lobefunction (pp 278ndash291) Oxford Oxford University Press

Friston K J Holmes A P Poline J-B Heather J D ampFrackowiak R S J (1995) Analysis of fMRI time-seriesrevisited Neuroimage 2 45ndash53

Fukushima T Hasegawa I amp Miyashita Y (2004)Prefrontal neuronal activity encodes spatial targetrepresentations sequentially updated after nonspatialtarget-shift cues Journal of Neurophysiology 911367ndash1380

Fuster J M (1995) Memory in the cerebral cortexCambridge MIT Press

Gitelman D R (2002) ILAB A program for post experimentaleye movement analysis Behavioral Research MethodsInstruments and Computers 34 605ndash612

1688 Journal of Cognitive Neuroscience Volume 17 Number 11

Hasegawa amp Miyashita 2004 Bor Duncan Wisemanamp Owen 2003 Takeda amp Funahashi 2002 2004 Pochonet al 2001 DrsquoEsposito Ballard Zarahn amp Aguirre 2000Fuster 1995) and motivation andor reward expectancy(eg Watanabe 1996 2002) For extensive reviews ofthis question see Curtis and DrsquoEsposito (2003) andPassingham and Sakai (2004) The present study wasdesigned to test yet another candidate function of PFCdelay-period activitymdashthe mediation of interference

One indication that the PFC plays an important role ininterference mediation comes from the fact that impair-ments begin to emerge in the working memory per-formance of PFC-lesioned subjects when distraction orinterference is present in the environment (DrsquoEspositoamp Postle 1999 Chao amp Knight 1998 Chao amp Knight1995 Malmo 1942) or in the cognitive system (as is thecase with proactive interference [PI] Thompson-Schillet al 2002 Milner 1964) Further evidence comes fromelectrophysiological and neuroimaging studies Withelectrophysiology Chao and Knight (1998) providedevidence that lateral PFC contributes to lsquolsquoattentionalfilteringrsquorsquo or lsquolsquosensory gatingrsquorsquo by showing that theevoked response to distracting auditory stimuli wasgreater for patients with PFC lesions than for controlsubjects With PET and fMRI several groups have inves-tigated the selective sensitivity to PI of the portion of theleft inferior frontal gyrus corresponding to Brodmannrsquosarea (BA 45) (eg Postle amp Brush 2004 Postle Brush ampNick 2004 Postle Berger Goldstein Curtis amp DrsquoEspo-sito 2001 Jonides Marshuetz Smith Reuter-Lorenz ampKoeppe 2000 DrsquoEsposito Postle Jonides amp Smith1999 Jonides Smith Marshuetz Koeppe amp Reuter-Lorenz 1998) Two other neuroimaging studies havegone beyond demonstrating the PFCrsquos sensitivity tointerference by showing that the PFC plays an active rolein its mediation In one Sakai Rowe and Passingham(2002) found that PFC activity during an unfilled delayperiod predicted task accuracy on trials when this un-filled period was followed by a distractor task Theyattributed this to a PFC-controlled lsquolsquoactive maintenancersquorsquoprocess that strengthened mnemonic representations viathe strengthening of the coupling of activity between thesuperior frontal cortex (BA 8) and intraparietal sulcus1 Inanother study Gray Chabris and Braver (2003) foundthat PFC activity correlated with performance on trials ofan n-back working memory task that featured a high levelof PI but only in subjects who had been independentlydetermined to be of high general fluid intelligence Likethese two the present study was also designed toevaluate evidence for a PFC-based control mechanisminvolved in the mediation of interference Its designwould permit a two-stage process of (1) the identificationof delay-period activity from unfilled delay periods fol-lowed by (2) an assessment of whether the voxelsidentified in Step 1 might contribute to a sensory gatingfunction Thus it would explicitly evaluate one candidateexplanation of the function of PFC delay-period activity

Although there are many ways to implement a gatingmechanism the present study was designed to detectone that operates by regulating the gain of sensoryprocessing such that the processing of potentially dis-rupting afferent sensory information is down-regulatedwhen this information might interfere with the STR ofother behaviorally prioritized information There isconsiderable evidence for such a mechanism from stud-ies measuring event-related potentials In these studiesan increase in alpha-band power over posterior areasduring short-term memory task performance is inter-preted as evidence for lsquolsquoinhibition of task irrelevantprocessesrsquorsquo (Klimesch Doppelmayr Schwaiger Auingeramp Winker 1999 p 403 see also Jensen GelfandKounios amp Lisman 2002 Chao amp Knight 1998) Thetask featured in this report was delayed face recognitionimplemented in a 2 2 design that varied the factors ofmemory (present absent) and distraction (presentabsent Figure 1) The logic of the experiment dependedon the assumption that the PFC delay-period activitythat was expected on memory-presentdistraction-absent trials would not reflect the STR of face informa-tion This assumption was supported by two previousstudies of delayed face recognition indicating that PFCdelay-period activity was logically incompatible with aSTR function One of these studies employed a multi-step ABBA-like design intended to winnow out delay-period activity that may be correlated with but notnecessary for the STR of face information Althoughthe results from each subject revealed Delay 1-specificactivity in many brain areas including the PFC posteriorfusiform gyrus and posterior parietal cortex only asubset of these voxels retained the Delay 1 signal duringDelay 2 and the posterior fusiform gyrus was the onlyregion in which in each subject voxels retained theDelay 1 signal during Delay 3 (Postle et al 2003) Thesecond study demonstrated that varying instructionsmdashlsquolsquoremember face(s)rsquorsquo versus lsquolsquoremember scene(s)rsquorsquomdashmodulated the delay-period activity of the fusiform facearea or the parahippocampal place area in a category-specific manner but had no effect on PFC (ie PFCactivity did not distinguish between STR of faces vs STRof houses Ranganath DeGutis amp DrsquoEsposito 2004)Instead of STR the sensory gating hypothesis posits thatsome portion of PFC delay-period activity on memory-presentdistraction-absent trials reflects a tonically activemechanism that monitors the environment for poten-tially distracting stimuli or events and that regulates thegain of sensory processing when needed

And so the logic of the experimental approach was (1)to identify face memory-related activity in the dorsolat-eral PFC (dlPFC) and inferior occipitotemporal cortex(IOTC) by identifying voxels in these two regions dem-onstrating significant delay-period activity in memory-presentdistraction-absent trials (ie in lsquolsquostandardrsquorsquodelayed-recognition conditions) and (2) to assess thedistraction-evoked activity in these voxels of interest

1680 Journal of Cognitive Neuroscience Volume 17 Number 11

(VOIs) by measuring delay-period activity from memory-absentdistraction-present and memory-presentdistrac-tion-present trials We assumed that a dlPFC-basedsensory gating mechanism would be engaged to a largerextent by the conjunction of the need to retain amemory with the presence of distracting noise in theenvironment than by the presence of distraction aloneThus we predicted greater delay-period activity in dlPFCVOIsmdashthe hypothesized source of controlmdashin memory-presentdistraction-present than in memory-absentdistraction-present trials In contrast we predicted thatthe gain-modulating effects of this gating process wouldresult in decreased activity at the site of its action duringthe conjunction of memory and distraction that is torelatively decreased delay-period activity in IOTC VOIs inmemory-presentdistraction-present as compared withmemory-absentdistraction-present trials Restated interms of our experimental design we predicted aninteraction such that the manipulation of the factor ofmemory on distraction trials would have a greater effecton dlPFC VOIs than on IOTC VOIs We did not knowa priori whether the outcome would be a crossoverinteraction (ie whether in IOTC VOIs the delay-periodeffect on memory-presentdistraction-present trialswould in fact be quantitatively lower than on memory-absentdistraction-present trials) or whether the differ-ence would be in the same direction as in dlPFC VOIsbut markedly smaller

RESULTS

Behavior

Behavioral results presented in Table 1 indicated thatperformance was high in all conditions (Due to techni-cal problems accuracy data were not collected from two

subjects and reaction time [RT] data were not collectedfrom four subjects) An analysis of variance (ANOVA)assessing accuracy as a function of memory (presentabsent) and distraction (present absent) revealed amain effect of memory that approached significanceF(113) = 36 p = 08 a main effect of distractionF(113) = 52 p lt 05 and an interaction F(113) =181 p lt 001 The analogous analysis of RT datarevealed main effects of memory F(111) = 512p lt 0001 and distraction F(111) = 228 p lt 001and an interaction F(111) = 215 p lt 001 The maineffect of distraction reflects the fact that RTs were fasterfor trials with distraction than without it Although thiseffect was not expected it may be that the rapid serialvisual presentation (RSVP) distraction elicited a generalincrease of arousal or a greater level of motor prepara-tion (eg Connolly Goodale Goltz amp Munoz 2005)Together the accuracy and RT results confirmed theintuition that of the two critical conditions in this

Figure 1 Illustration of a

memory-presentdistraction-

absent trial and a memory-

absentdistraction-present trial

Table 1 Behavioral Performance Accuracy (Mean ProportionCorrect) and RT (in Milliseconds)

Memory

Distraction Present Absent

Present

Accuracy (SE) 0874 (0020) 0931 (0023)

RT (SE) 7661 (401) 4744 (368)

Absent

Accuracy (SE) 0945 (0021) 0934 (0024)

RT (SE) 8036 (381) 6339 (478)

Postle 1681

study memory-absentdistraction-present and memory-presentdistraction-present the latter would be themore difficult

fMRI

Delay-period activity was observed on memory-presentdistraction-absent trials in the dlPFC and the IOTC ineach of the 16 subjects (Figure 2) Within the function-ally defined VOIs distractor-evoked activity was mark-edly higher for memory-present than for memory-absenttrials in the dlPFC whereas the IOTC showed a trend inthe opposite direction (Figures 3 and 4) ANOVA assess-ing raw distraction-evoked activity in VOIs in the tworegions (dlPFC IOTC) and the two critical trial types(memory-absentdistraction-present memory-presentdistraction-present trials) revealed a main effect of re-gion F(115) = 1197 p lt 005 an effect of trial typethat approached significance F(115) = 396 p = 07and an interaction F(115) = 1523 p = 001 Theseresults were broadly consistent with our predictionsHowever because the main effect of region indicatedthat the overall evoked response was greater in theIOTC than in the dlPFC it was possible that the differ-

ential patterns of activity observed in these two VOIswere attributable at least in part to a difference in theinherent level of activity in the two regions in which theywere located To rule out this possibility the effectsfrom each subjectrsquos VOIs were normalized by dividingwithin VOI the distractor-evoked effects from both trialtypes of interest by the memory-absentdistraction-present effect This had the effect of transforming thememory-absentdistraction-present effect in both re-gions from each subject to the arbitrarily selected valueof 10 and transforming the memory-presentdistraction-present effect into a proportion of the memory-absentdistraction-present effect After this transformation wasperformed we excluded the data from one subject(12) from further analyses because the normalizedmemory-presentdistraction-present effect from thedlPFC VOIs of this subject was in excess of 3 standarddeviations greater than the mean2 For the remaining 15subjects the normalized data indicated that the meandelay-evoked response on memory-presentdistraction-present trials in dlPFC VOIs was 139 greater than onmemory-absentdistraction-present trials [a significantpairwise effect at t(14) = 21 p = 05] whereas in IOTCVOIs the mean delay-evoked response on memory-

Figure 2 Two representative slices from each subject illustrating dlPFC and IOTC VOIs Memory-presentdistraction-absent activity wasobserved in the dlPFC bilaterally in 13 subjects and unilaterally in the right hemisphere in 2 subjects and in the left hemisphere in 1 subject

in the IOTC it was observed bilaterally in 4 subjects unilaterally in the right hemisphere in 9 subjects and unilaterally in the left hemisphere

in 3 subjects

1682 Journal of Cognitive Neuroscience Volume 17 Number 11

presentdistraction-present trials was 75 smaller thanit was on memory-absentdistraction-present trials [apairwise effect that was not significant t(14) = 12ns] Thus the pattern of the normalized effects retained

the features of the raw data (Figure 4B) as confirmed byANOVA assessing the normalized results from the re-maining 15 subjects which produced a main effect ofregion F(114) = 49 p lt 05 a borderline effect of trial

Figure 3 An example of

results from a single subject

(21 in Figure 2) The two

top plots illustrate the trial-averaged time series data from

dlPFC and IOTC VOIs from

memory-presentdistraction-absent trials (fMRI MPDA)

and the corresponding delay-

period covariates scaled by

their parameter estimates(delay effect) The gray

bars along the horizontal axes

indicate the duration of the

delay period The two bottomplots illustrate the delay effects

from these same VOIs from

memory-presentdistraction-present (MPDP) and

memory-absentdistraction-

present (MADP) trials

Note that the signal intensityscales differ on plots for the

two regions

Figure 4 Group averageeffects (A) Memory-present

distraction-absent delay-period

VOIs with unscaled data (B)memory-presentdistraction-

absent delay-period VOIs with

normalized data (C) target-

evoked VOIs with unscaleddata (D) target-evoked VOIs

with normalized data MPDA =

memory-presentdistraction-

absent MPDP = memory-presentdistraction-present

MADP = memory-absent

distractionpresent Errorbars represent SEM

Postle 1683

type F(114) = 38 p = 07 and an interaction F(114) =49 p lt 05

One alternative to the sensory gating interpretation ofthese results might be that PFC delay-period activity onmemory-presentdistraction-present trials reflected dualtaskingmdashthe STR of target face information plus theincidental encoding (or some other type of processing)of the distracting faces This seems unlikely howeverbecause PFC delay-period activity was numerically loweron memory-presentdistraction-present trials than onmemory-presentdistraction-absent trials [although thiseffect did not achieve significance t(15) = 13 nsFigure 4A]

How specific was this pattern of results To assess thisquestion we applied the same contrasts to voxels in thedlPFC and IOTC that were selected for a differentpropertymdashresponsivity to visually presented faces Spe-cifically we identified voxels in the dlPFC and IOTC thatresponded to the Target epoch of the task (and re-moved from these VOIs any voxels that were also activeduring the delay period of memory-presentdistraction-absent trials) In this way we identified voxels from thesame regions in which we tested the sensory gatinghypothesis but whose function was different (presum-ably related to visual perception andor memory encod-ing but not to retention or another delay-periodprocess) For the raw data from these target-evokedVOIs (Figure 4C) as compared to the memory-evokedVOIs there was again a strong main effect of regionF(114) = 334 p lt 0001 a weaker borderline effect oftrial type F(114) = 33 p = 09 and now only a trendtoward an interaction F(114) = 40 p 07 When datafrom the target-evoked VOIs were normalized in thesame manner as had been the data from the delay-evoked VOIs (Figure 4D) their analysis with ANOVArevealed only a borderline effect of VOI F(114) = 40p = 06 for the first time a reliable effect of trial typeF(114) = 51 p lt 05 and again a trend toward aninteraction F(114) = 40 p = 06

The weakening of the Region Trial type interactionin target-evoked VOIs as compared with delay-evokedVOIs could be consistent with the claim that this inter-action predicted to be the fMRI signature of a sensorygating mechanism was selectively seen only in the delay-evoked VOIs It could also however simply reflect asituation in which the interaction terms for the twotypes of VOI fell on either side of the 05 thresholdbut did not differ from each other To distinguishbetween these two possibilities we compared the ef-fects of the experimental manipulation on delay-periodactivity across the two types of VOIs with an omnibusANOVA that directly compared the effect of memory onnormalized distraction-evoked activity across target-evoked and delay-evoked VOIs The factors for theomnibus ANOVA were region (dlPFC IOTC) trial type(memory present memory absent) and epoch definingthe VOIs (target delay) A three-way interaction would

be necessary to support the claim that the differentialeffect of memory demands on distraction-evoked activityin the dlPFC and IOTC was greater for delay-evoked thanfor target-evoked VOIs The ANOVA revealed a maineffect of region F(114) = 49 p lt 05 a borderlineeffect of trial type F(114) = 43 p = 06 and no effect ofepoch F(114) = 28 p gt 1 no interaction of Epoch Trial type F(114) = 28 p gt 1 but reliable inter-actions of Epoch Region F(114) = 47 p lt 05and of Region Trial type (F(114) = 49 p lt 05 and areliable interaction of Epoch Region Trial typeF(114) = 47 p lt 05 Thus it confirmed the specificityof the sensory gating effect to delay-evoked VOIs

DISCUSSION

The results demonstrate that delay-sensitive voxels intwo brain regions the dlPFC and the IOTC responddifferentially to the RSVP of faces conditioned on thebehavioral context in which these faces are presentedThe dlPFC response was 139 greater during concur-rent face memory whereas the IOTC response was 75lower This pattern was also seen to be selective becausea different group of dlPFC voxels (those that are targetsensitive) increased its activity by only 38 and theanalogously different group of IOTC voxels increased itsactivity by 3 These results are therefore consistentwith a model in which one function of dlPFC voxels thatare active during unfilled delay periods of delayed-recognition tasks is to perform a sensory gating functionBy this account the detection of potentially interferingstimuli in the environment triggers activity in networksof dlPFC neurons3 that in turn has the effect of sup-pressing perhaps via thalamic relays the afferent flow ofsensory information (Knight Staines Swick amp Chao1999) This would have beneficial effects for the signal-to-noise ratio in the networks of posterior neurons thatare responsible for the STR of information Thus a PFC-based control process may serve to protect the contentsof working and short-term memory from the deleteriouseffects external interference although it does not sup-port STR operations themselves

This account makes two broader claims about theworking memory literature First it offers an explanationas to why primates with PFC lesions can perform nor-mally on tasks of working and short-term memory whenpotential sources of distraction and inference are mini-mal but are impaired on the same tasks when distractionandor interference is high (Thompson-Schill et al2002 DrsquoEsposito amp Postle 1999 Chao amp Knight 1998Chao amp Knight 1995 Milner 1964 Malmo 1942)Second it proposes that sensory gating may accountfor a portion of the unfilled delay-period activity that isroutinely observed in PFC across a variety of neuro-imaging studies including those that do not explicitlyinclude distraction In relation to this second claim how-

1684 Journal of Cognitive Neuroscience Volume 17 Number 11

ever we cannot rule out the possibility that the dlPFCactivity observed during memory-presentdistraction-absent trials was due to a specific feature of our designBecause subjects could not predict which memorytrials would also contain distraction an effective strat-egy may have been to implement a sensory gatingstrategy on every memory trial To rule this out wouldrequire a two stage process of first scanning lsquolsquonaıversquorsquosubjects performing several blocks of only memory-presentdistraction-absent trials then introducing thefactor of delay-period distraction for a subsequent setof scans

The present study fits within a growing body of resultsthat document control rather than STR functions forthe PFC These include the monitoring and transforma-tion of information held in working memory (Bor et al2003 DrsquoEsposito Postle Ballard amp Lease 1999 Pet-rides 1995) and the attentional selection of information(Lebedev et al 2004 Sakai et al 2002 de Fockert ReesFrith amp Lavie 2001) These approaches which proposewell-defined testable mechanisms represent an advancebeyond the vague notion of lsquolsquomaintenancersquorsquo that is ofteninvoked in interpretations of PFC activity The presentstudy suggests one such alternative to lsquolsquomaintenancersquorsquo toaccount for sustained PFC activity observed during thesimplest tests of STR of information

METHODS

Subjects

Sixteen healthy young adults (mean age = 224 13 men)participated in this study All provided informed consentand were paid in compensation for participating in thestudy

Behavioral Procedure

The 2 2 factorial design featured the factors ofMemory (present absent) and Distraction (presentabsent) Each trial had the same structure Target (1 sec)followed by Delay (7 sec) followed by Probe (1 sec)followed by ITI (13 sec) During memory-present trialstarget and probe stimuli were individual gray-scale malefaces (none wearing glasses) appearing on white back-grounds and subjects indicated with a button presswhether the latter matched the former ( p = 5) Memory-absent trials were procedurally identical except thatthey featured the same recurring female facesmdashfacelsquolsquoArsquorsquo always the target which denoted a lsquolsquono-memoryrsquorsquotrial face lsquolsquoBrsquorsquo always the probe On memory-absenttrials subjects were cued by the probe to indicate witha button press whether or not distracting faces appearedduring the delay period ( p = 5) thereby matching themotor response demands of memory-present trialsDistraction-present trials featured RSVP (3 Hz) of thefaces of males wearing glasses (to differentiate distrac-

tion items from memoranda) during the two middleseconds of the delay (Figure 1) Subjects were instructedto look at the distracting faces during all trials on whichthey appeared and to fixate centrally on distraction-absent trials Stimuli were presented via eyepiecesmounted on the head coil (Avotec Visible Eye StuartFL) and fitted with an infrared-based eye tracking system(SMI Inc Teltow Germany) Fixation was monitoredduring scanning and analyzed off-line with iLab software(Gitelman 2002) Trials during which subjects failedto fixate centrally from time 3 sec to time 5 sec werediscarded from subsequent analyses (Figure 5) Thiscriterion ensured that any evidence for differencesin delay-period activity between conditions could notbe attributed to theoretically uninteresting differencesin for example the amount of visual stimulation pro-duced by the distracting stimuli or the amount ofoculomotor activity Trials of each of the four types wereequiprobable and presented in a randomly determinedorder

fMRI Data Acquisition

Whole-brain images were acquired with a 3T scanner(GE Signa VHI Waukesha WI) High-resolution T1-weighted images (32 axial slices 09375 09375 4 mm)were obtained in every participant and a gradient-echoecho-planar sequence (TR = 2000 msec TE = 50 msec)was used to acquire data sensitive to the blood oxygenlevel dependent (BOLD) signal (Kwong et al 1992Ogawa et al 1992) within a 64 64 matrix (32 axialslices coplanar with T1 375 375 4 mm) Scans ofthe delayed-recognition task were preceded by a scan inwhich we derived an estimate of the hemodynamicresponse function (HRF) for each participant Duringthis scan each participant performed a simple RT taskthat required a bimanual button press once every 20 secin response to a brief change in shape of the fixationstimulus A partial F test associated with a Fourier basiscovariate set ( Josephs Turner amp Friston 1997) wasused to evaluate the significance of task-correlatedactivity in each voxel of primary somatosensory andmotor cortical regions of interest (Aguirre Zarahn ampDrsquoEsposito 1998) An HRF estimate was extracted fromthe suprathreshold voxels of these ROIs by spatiallyaveraging their time series filtering the resultant aver-aged fMRI time series to remove high (gt0244 Hz)and low (lt005 Hz) frequencies adjusting it to removethe effects of nuisance covariates (Friston HolmesPoline Heather amp Frackowiak 1995) and trial aver-aging The HRF characterizes the fMRI response result-ing from a brief impulse of neural activity (BoyntonEngel Glover amp Heeger 1996) and can vary markedlyacross participants (Aguirre et al 1998) The subject-specific HRFs were used to convolve independent vari-ables entered into the modified general linear model(GLM Worsley amp Friston 1995) that we used to analyze

Postle 1685

the data from the scans of the delayed-recognition task4

The eight scans of the delayed-recognition task eachlasted 6 min 12 sec (5 min 52 sec of task preceded by20 sec of dummy pulses to achieve a steady state of tis-sue magnetization)

fMRI Data Processing

The fMRI time series data were filtered and adjusted asdescribed previously (Postle Zarahn amp DrsquoEsposito2000) The principle of the fMRI time series analysiswas to model the fMRI signal changes evoked byeach stimulus presentation epoch with covariates com-posed of BOLD impulse response functions shiftedalong the time line of the task to represent various trialepochs (Figure 1 Postle et al 2000 Zarahn Aguirre

amp DrsquoEsposito 1997) The least-squares solution of theGLM of the fMRI time series data yielded parameterestimates that were associated with each covariateof interest The smoothness of the fMRI response toneural activity allows fMRI-evoked responses that arisefrom temporally dependent events to be resolved onthe order of 4 sec (That is if the covariates modelingthese events are spaced at 4-sec intervals the covariatepositioned at time 0 sec will account for virtually all ofthe variance attributable to an event starting at time 0 secand of duration of 1 sec or less thereby leaving thecovariate positioned at time 4 sec relatively uncontam-inated by variance attributable to the time 0 sec eventsee Figure 6 of Zarahn et al 1997) Differences in fMRIsignal (either between conditions or vs baseline) weretested by computing t statistics from contrasts between

Figure 5 Examples of eye-position data from four representative trials For all four examples trial epoch is color coded red = target stimulus

green = early delay blue = distraction portion of the delay period fuchsia = late delay black = probe stimulus plus first second of ITI (A) A

memory-presentdistraction-present trial with acceptable fixation during distraction The top panel is a 2-D representation of eye position on the

display lsquolsquoscreenrsquorsquo over the course of the trial axis labels correspond to degrees of visual angle Bottom panels display horizontal (x) and vertical ( y)eye position as a function of time with time in seconds on the horizontal axes and position in pixels on the vertical axes (each pixel measured

058 mm2) (B) A memory-absentdistraction-present trial with acceptable fixation during distraction All conventions same as in (A) Interrupted

portion of late delay corresponds to a blink when eyes were closed (C) A memory-presentdistraction-present trial that was discarded due

to unacceptable fixation during distraction All conventions same as (A) (D) A memory-absentdistraction-present trial that was discarded dueto unacceptable fixation during distraction All conventions same as in (A)

1686 Journal of Cognitive Neuroscience Volume 17 Number 11

parameter estimates associated with the covariates inquestion

Our analysis strategy was to identify delay-specificactivity in two regions the dlPFC and the IOTC5 frommemory-presentdistraction-absent trials This wouldidentify delay-period activity comparable to that ob-served in virtually all whole-brain neuroimaging studiesof similar tasks (eg with face stimuli Postle et al 2003Ranganath amp DrsquoEsposito 2001 Druzgal amp DrsquoEsposito2001 Courtney Ungerleider Keil amp Haxby 1997 Court-ney Ungerleider Keil amp Haxby 1996) Masks cor-responding to these two regions were drawn on eachsubjectrsquos anatomical scans Voxels located within thesemasks which became the VOIs were identified on tmaps of delay-period activity from memory-presentdis-traction-absent trials thresholded to a regionwise a of05 (one-tailed) with Bonferroni correction for the num-ber of voxels in the mask Note that with this approachwith spatially unsmoothed data individual voxels whoset value exceeds threshold can be interpreted as lsquolsquomean-ingfulrsquorsquo data (Postle et al 2000) Thus this procedurediffered somewhat from other applications of lsquolsquoregion-

of-interest (ROI) analysesrsquorsquo in that no selection criteriawere imposed with respect to cluster size or proximity ofcritical voxels This was done among other reasonsbecause we did not know a priori in which subregionof PFC to expect to find sensory-gating-related activityWithin the VOIs we evaluated the responses evokedby RSVP of faces during memory-presentdistraction-present trials versus memory-absentdistraction-presenttrials (Memory-absentdistraction-absent trials do notfigure directly into the test of the sensory gating hy-pothesis but were necessary to prevent anticipation ofdistraction on memory-absent trials)

Acknowledgments

The author thanks Olufunsho Faseyitan and Craig Rypstat forprogramming assistance and OF and Christopher Jordan forhelp with data collection and analysis This research was sup-ported by the National Institutes of Health RO1 MH064498

Reprint requests should be sent to Bradley R Postle Depart-mentofPsychologyUniversityof WisconsinndashMadison1202WestJohnson St Madison WI 53706 or via e-mail postlewiscedu

Figure 5 (continued)

Postle 1687

The data reported in this experiment have been depositedwith the fMRI Data Center archive (wwwfmridcorg) Theaccession number is 2-2005-119B9

Notes

1 This may be an empirical example of the lsquolsquoactive main-tenancersquorsquo proposed by Miller and Cohen (2001)2 The raw value of this subjectrsquos memory-absentdistraction-present effect in the dlPFC was 51 and so dividing thememory-presentdistraction-present by this value inflated it bya factor of almost 2 the opposite direction than that effectedby the normalization procedure in all other subjects exceptone whose raw memory-absentdistraction-present effect inthe dlPFC was 983 The fact that delay-period activity in the dlPFC VOIs wasnumerically lower during memory-presentdistraction-presentthan during memory-presentdistraction-absent trials is con-sistent with the idea that the sensory gating mechanism oncetriggered by neurons that are tonically active during unfilleddelay periods is implemented by a broader PFC network thatincludes regions that are not typically active during unfilleddelay periods Although the data are not shown here it iscertainly the case that in every subject considerably more dlPFCvoxels were active during the memory-presentdistraction-present delay than during the memory-presentdistraction-absent delay4 The HRF also varies in shape across brain regions with-in an individual participant (eg Miezen Maccotta OllingerPetersen amp Buckner 2000) However recent work indi-cates that the magnitude of variability in the shape of theHRF is greater across individuals within a region than it isacross regions within an individual (Handwerker Ollinger ampDrsquoEsposito 2004) Additionally it indicates that employinga participant-specific HRF can improve the sensitivity andmagnitude estimates of the least-squares solution of the GLMover comparable analyses that employ a generic HRF model(Handwerker et al 2004)5 The broader search space of the IOTC was used instead ofa more circumscribed area such as the fusiform face areabecause as can be seen in Figure 2 there was a considerableamount of topological variability of memory-presentdistraction-absent delay-period activity in the posterior ventral stream andno one area was activated in all subjects

REFERENCES

Aguirre G K Zarahn E amp DrsquoEsposito M (1998) Thevariability of human BOLD hemodynamic responsesNeuroimage 8 360ndash369

Baddeley A D amp Hitch G J (1974) Working memoryIn G H Bower (Ed) The psychology of learning andmotivation (Vol 8 pp 47ndash89) New York Academic Press

Bor D Duncan J Wiseman R J amp Owen A M (2003)Encoding strategies dissociate prefrontal activity fromworking memory demand Neuron 37 361ndash367

Boynton G M Engel S A Glover G H amp HeegerD J (1996) Linear systems analysis of functionalmagnetic resonance imaging in human V1 The Journalof Neuroscience 16 4207ndash4221

Braver T S Gray J R amp Burgess G C (in press)Explaining the many varieties of working memoryvariation Dual mechanisms of cognitive control In AConway C Jarrold M Kane A Miyake amp J Towse(Eds) Variation in working memory Oxford OxfordUniversity Press

Chao L amp Knight R (1998) Contribution of humanprefrontal cortex to delay performance Journal ofCognitive Neuroscience 10 167ndash177

Chao L L amp Knight R T (1995) Human prefrontal lesionsincrease distractibility to irrelevant sensory inputsNeuroReport 6 1605ndash1610

Connolly J D Goodale M A Goltz H C amp Munoz D P(2005) fMRI activation in the human frontal eye field iscorrelated with saccadic reaction time Journal ofNeurophysiology 94 605ndash611

Courtney S M (2004) Attention and cognitive control asemergent properties of information representation inworking memory Cognitive Affective amp BehavioralNeuroscience 4 501ndash516

Courtney S M Ungerleider L G Keil K amp Haxby J(1996) Object and spatial visual working memory activateseparate neural systems in human cortex Cerebral Cortex6 39ndash49

Courtney S M Ungerleider L G Keil K amp Haxby J V(1997) Transient and sustained activity in a distributedneural system for human working memory Nature 386608ndash611

Curtis C E amp DrsquoEsposito M (2003) Persistent activity inthe prefrontal cortex during working memory Trends inCognitive Sciences 7 415ndash423

Curtis C E Rao V Y amp DrsquoEsposito M (2004) Maintenanceof spatial and motor codes during oculomotor delayedresponse tasks The Journal of Neuroscience 243944ndash3952

DrsquoEsposito M Ballard D Zarahn E amp Aguirre G K(2000) The role of prefrontal cortex in sensory memoryand motor preparation An event-related fMRI studyNeuroimage 11 400ndash408

DrsquoEsposito M amp Postle B R (1999) The dependence of spanand delayed-response performance on prefrontal cortexNeuropsychologia 37 1303ndash1315

DrsquoEsposito M Postle B R Ballard D amp Lease J (1999)Maintenance versus manipulation of information held inworking memory An event-related fMRI study Brain ampCognition 41 66ndash86

DrsquoEsposito M Postle B R Jonides J amp Smith E E(1999) The neural substrate and temporal dynamicsof interference effects in working memory as revealedby event-related functional MRI Proceedings of theNational Academy of Sciences USA 96 7514ndash7519

de Fockert J W Rees G Frith C D amp Lavie N (2001) Therole of working memory in visual attention Science 2911803ndash1806

Druzgal T J amp DrsquoEsposito M (2001) A neural networkref lecting decisions about human faces Neuron 32947ndash955

Duncan J amp Miller E K (2002) Cognitive focus throughadaptive neural coding in the primate prefrontal cortexIn D Stuss amp R Knight (Eds) Principles of frontal lobefunction (pp 278ndash291) Oxford Oxford University Press

Friston K J Holmes A P Poline J-B Heather J D ampFrackowiak R S J (1995) Analysis of fMRI time-seriesrevisited Neuroimage 2 45ndash53

Fukushima T Hasegawa I amp Miyashita Y (2004)Prefrontal neuronal activity encodes spatial targetrepresentations sequentially updated after nonspatialtarget-shift cues Journal of Neurophysiology 911367ndash1380

Fuster J M (1995) Memory in the cerebral cortexCambridge MIT Press

Gitelman D R (2002) ILAB A program for post experimentaleye movement analysis Behavioral Research MethodsInstruments and Computers 34 605ndash612

1688 Journal of Cognitive Neuroscience Volume 17 Number 11

(VOIs) by measuring delay-period activity from memory-absentdistraction-present and memory-presentdistrac-tion-present trials We assumed that a dlPFC-basedsensory gating mechanism would be engaged to a largerextent by the conjunction of the need to retain amemory with the presence of distracting noise in theenvironment than by the presence of distraction aloneThus we predicted greater delay-period activity in dlPFCVOIsmdashthe hypothesized source of controlmdashin memory-presentdistraction-present than in memory-absentdistraction-present trials In contrast we predicted thatthe gain-modulating effects of this gating process wouldresult in decreased activity at the site of its action duringthe conjunction of memory and distraction that is torelatively decreased delay-period activity in IOTC VOIs inmemory-presentdistraction-present as compared withmemory-absentdistraction-present trials Restated interms of our experimental design we predicted aninteraction such that the manipulation of the factor ofmemory on distraction trials would have a greater effecton dlPFC VOIs than on IOTC VOIs We did not knowa priori whether the outcome would be a crossoverinteraction (ie whether in IOTC VOIs the delay-periodeffect on memory-presentdistraction-present trialswould in fact be quantitatively lower than on memory-absentdistraction-present trials) or whether the differ-ence would be in the same direction as in dlPFC VOIsbut markedly smaller

RESULTS

Behavior

Behavioral results presented in Table 1 indicated thatperformance was high in all conditions (Due to techni-cal problems accuracy data were not collected from two

subjects and reaction time [RT] data were not collectedfrom four subjects) An analysis of variance (ANOVA)assessing accuracy as a function of memory (presentabsent) and distraction (present absent) revealed amain effect of memory that approached significanceF(113) = 36 p = 08 a main effect of distractionF(113) = 52 p lt 05 and an interaction F(113) =181 p lt 001 The analogous analysis of RT datarevealed main effects of memory F(111) = 512p lt 0001 and distraction F(111) = 228 p lt 001and an interaction F(111) = 215 p lt 001 The maineffect of distraction reflects the fact that RTs were fasterfor trials with distraction than without it Although thiseffect was not expected it may be that the rapid serialvisual presentation (RSVP) distraction elicited a generalincrease of arousal or a greater level of motor prepara-tion (eg Connolly Goodale Goltz amp Munoz 2005)Together the accuracy and RT results confirmed theintuition that of the two critical conditions in this

Figure 1 Illustration of a

memory-presentdistraction-

absent trial and a memory-

absentdistraction-present trial

Table 1 Behavioral Performance Accuracy (Mean ProportionCorrect) and RT (in Milliseconds)

Memory

Distraction Present Absent

Present

Accuracy (SE) 0874 (0020) 0931 (0023)

RT (SE) 7661 (401) 4744 (368)

Absent

Accuracy (SE) 0945 (0021) 0934 (0024)

RT (SE) 8036 (381) 6339 (478)

Postle 1681

study memory-absentdistraction-present and memory-presentdistraction-present the latter would be themore difficult

fMRI

Delay-period activity was observed on memory-presentdistraction-absent trials in the dlPFC and the IOTC ineach of the 16 subjects (Figure 2) Within the function-ally defined VOIs distractor-evoked activity was mark-edly higher for memory-present than for memory-absenttrials in the dlPFC whereas the IOTC showed a trend inthe opposite direction (Figures 3 and 4) ANOVA assess-ing raw distraction-evoked activity in VOIs in the tworegions (dlPFC IOTC) and the two critical trial types(memory-absentdistraction-present memory-presentdistraction-present trials) revealed a main effect of re-gion F(115) = 1197 p lt 005 an effect of trial typethat approached significance F(115) = 396 p = 07and an interaction F(115) = 1523 p = 001 Theseresults were broadly consistent with our predictionsHowever because the main effect of region indicatedthat the overall evoked response was greater in theIOTC than in the dlPFC it was possible that the differ-

ential patterns of activity observed in these two VOIswere attributable at least in part to a difference in theinherent level of activity in the two regions in which theywere located To rule out this possibility the effectsfrom each subjectrsquos VOIs were normalized by dividingwithin VOI the distractor-evoked effects from both trialtypes of interest by the memory-absentdistraction-present effect This had the effect of transforming thememory-absentdistraction-present effect in both re-gions from each subject to the arbitrarily selected valueof 10 and transforming the memory-presentdistraction-present effect into a proportion of the memory-absentdistraction-present effect After this transformation wasperformed we excluded the data from one subject(12) from further analyses because the normalizedmemory-presentdistraction-present effect from thedlPFC VOIs of this subject was in excess of 3 standarddeviations greater than the mean2 For the remaining 15subjects the normalized data indicated that the meandelay-evoked response on memory-presentdistraction-present trials in dlPFC VOIs was 139 greater than onmemory-absentdistraction-present trials [a significantpairwise effect at t(14) = 21 p = 05] whereas in IOTCVOIs the mean delay-evoked response on memory-

Figure 2 Two representative slices from each subject illustrating dlPFC and IOTC VOIs Memory-presentdistraction-absent activity wasobserved in the dlPFC bilaterally in 13 subjects and unilaterally in the right hemisphere in 2 subjects and in the left hemisphere in 1 subject

in the IOTC it was observed bilaterally in 4 subjects unilaterally in the right hemisphere in 9 subjects and unilaterally in the left hemisphere

in 3 subjects

1682 Journal of Cognitive Neuroscience Volume 17 Number 11

presentdistraction-present trials was 75 smaller thanit was on memory-absentdistraction-present trials [apairwise effect that was not significant t(14) = 12ns] Thus the pattern of the normalized effects retained

the features of the raw data (Figure 4B) as confirmed byANOVA assessing the normalized results from the re-maining 15 subjects which produced a main effect ofregion F(114) = 49 p lt 05 a borderline effect of trial

Figure 3 An example of

results from a single subject

(21 in Figure 2) The two

top plots illustrate the trial-averaged time series data from

dlPFC and IOTC VOIs from

memory-presentdistraction-absent trials (fMRI MPDA)

and the corresponding delay-

period covariates scaled by

their parameter estimates(delay effect) The gray

bars along the horizontal axes

indicate the duration of the

delay period The two bottomplots illustrate the delay effects

from these same VOIs from

memory-presentdistraction-present (MPDP) and

memory-absentdistraction-

present (MADP) trials

Note that the signal intensityscales differ on plots for the

two regions

Figure 4 Group averageeffects (A) Memory-present

distraction-absent delay-period

VOIs with unscaled data (B)memory-presentdistraction-

absent delay-period VOIs with

normalized data (C) target-

evoked VOIs with unscaleddata (D) target-evoked VOIs

with normalized data MPDA =

memory-presentdistraction-

absent MPDP = memory-presentdistraction-present

MADP = memory-absent

distractionpresent Errorbars represent SEM

Postle 1683

type F(114) = 38 p = 07 and an interaction F(114) =49 p lt 05

One alternative to the sensory gating interpretation ofthese results might be that PFC delay-period activity onmemory-presentdistraction-present trials reflected dualtaskingmdashthe STR of target face information plus theincidental encoding (or some other type of processing)of the distracting faces This seems unlikely howeverbecause PFC delay-period activity was numerically loweron memory-presentdistraction-present trials than onmemory-presentdistraction-absent trials [although thiseffect did not achieve significance t(15) = 13 nsFigure 4A]

How specific was this pattern of results To assess thisquestion we applied the same contrasts to voxels in thedlPFC and IOTC that were selected for a differentpropertymdashresponsivity to visually presented faces Spe-cifically we identified voxels in the dlPFC and IOTC thatresponded to the Target epoch of the task (and re-moved from these VOIs any voxels that were also activeduring the delay period of memory-presentdistraction-absent trials) In this way we identified voxels from thesame regions in which we tested the sensory gatinghypothesis but whose function was different (presum-ably related to visual perception andor memory encod-ing but not to retention or another delay-periodprocess) For the raw data from these target-evokedVOIs (Figure 4C) as compared to the memory-evokedVOIs there was again a strong main effect of regionF(114) = 334 p lt 0001 a weaker borderline effect oftrial type F(114) = 33 p = 09 and now only a trendtoward an interaction F(114) = 40 p 07 When datafrom the target-evoked VOIs were normalized in thesame manner as had been the data from the delay-evoked VOIs (Figure 4D) their analysis with ANOVArevealed only a borderline effect of VOI F(114) = 40p = 06 for the first time a reliable effect of trial typeF(114) = 51 p lt 05 and again a trend toward aninteraction F(114) = 40 p = 06

The weakening of the Region Trial type interactionin target-evoked VOIs as compared with delay-evokedVOIs could be consistent with the claim that this inter-action predicted to be the fMRI signature of a sensorygating mechanism was selectively seen only in the delay-evoked VOIs It could also however simply reflect asituation in which the interaction terms for the twotypes of VOI fell on either side of the 05 thresholdbut did not differ from each other To distinguishbetween these two possibilities we compared the ef-fects of the experimental manipulation on delay-periodactivity across the two types of VOIs with an omnibusANOVA that directly compared the effect of memory onnormalized distraction-evoked activity across target-evoked and delay-evoked VOIs The factors for theomnibus ANOVA were region (dlPFC IOTC) trial type(memory present memory absent) and epoch definingthe VOIs (target delay) A three-way interaction would

be necessary to support the claim that the differentialeffect of memory demands on distraction-evoked activityin the dlPFC and IOTC was greater for delay-evoked thanfor target-evoked VOIs The ANOVA revealed a maineffect of region F(114) = 49 p lt 05 a borderlineeffect of trial type F(114) = 43 p = 06 and no effect ofepoch F(114) = 28 p gt 1 no interaction of Epoch Trial type F(114) = 28 p gt 1 but reliable inter-actions of Epoch Region F(114) = 47 p lt 05and of Region Trial type (F(114) = 49 p lt 05 and areliable interaction of Epoch Region Trial typeF(114) = 47 p lt 05 Thus it confirmed the specificityof the sensory gating effect to delay-evoked VOIs

DISCUSSION

The results demonstrate that delay-sensitive voxels intwo brain regions the dlPFC and the IOTC responddifferentially to the RSVP of faces conditioned on thebehavioral context in which these faces are presentedThe dlPFC response was 139 greater during concur-rent face memory whereas the IOTC response was 75lower This pattern was also seen to be selective becausea different group of dlPFC voxels (those that are targetsensitive) increased its activity by only 38 and theanalogously different group of IOTC voxels increased itsactivity by 3 These results are therefore consistentwith a model in which one function of dlPFC voxels thatare active during unfilled delay periods of delayed-recognition tasks is to perform a sensory gating functionBy this account the detection of potentially interferingstimuli in the environment triggers activity in networksof dlPFC neurons3 that in turn has the effect of sup-pressing perhaps via thalamic relays the afferent flow ofsensory information (Knight Staines Swick amp Chao1999) This would have beneficial effects for the signal-to-noise ratio in the networks of posterior neurons thatare responsible for the STR of information Thus a PFC-based control process may serve to protect the contentsof working and short-term memory from the deleteriouseffects external interference although it does not sup-port STR operations themselves

This account makes two broader claims about theworking memory literature First it offers an explanationas to why primates with PFC lesions can perform nor-mally on tasks of working and short-term memory whenpotential sources of distraction and inference are mini-mal but are impaired on the same tasks when distractionandor interference is high (Thompson-Schill et al2002 DrsquoEsposito amp Postle 1999 Chao amp Knight 1998Chao amp Knight 1995 Milner 1964 Malmo 1942)Second it proposes that sensory gating may accountfor a portion of the unfilled delay-period activity that isroutinely observed in PFC across a variety of neuro-imaging studies including those that do not explicitlyinclude distraction In relation to this second claim how-

1684 Journal of Cognitive Neuroscience Volume 17 Number 11

ever we cannot rule out the possibility that the dlPFCactivity observed during memory-presentdistraction-absent trials was due to a specific feature of our designBecause subjects could not predict which memorytrials would also contain distraction an effective strat-egy may have been to implement a sensory gatingstrategy on every memory trial To rule this out wouldrequire a two stage process of first scanning lsquolsquonaıversquorsquosubjects performing several blocks of only memory-presentdistraction-absent trials then introducing thefactor of delay-period distraction for a subsequent setof scans

The present study fits within a growing body of resultsthat document control rather than STR functions forthe PFC These include the monitoring and transforma-tion of information held in working memory (Bor et al2003 DrsquoEsposito Postle Ballard amp Lease 1999 Pet-rides 1995) and the attentional selection of information(Lebedev et al 2004 Sakai et al 2002 de Fockert ReesFrith amp Lavie 2001) These approaches which proposewell-defined testable mechanisms represent an advancebeyond the vague notion of lsquolsquomaintenancersquorsquo that is ofteninvoked in interpretations of PFC activity The presentstudy suggests one such alternative to lsquolsquomaintenancersquorsquo toaccount for sustained PFC activity observed during thesimplest tests of STR of information

METHODS

Subjects

Sixteen healthy young adults (mean age = 224 13 men)participated in this study All provided informed consentand were paid in compensation for participating in thestudy

Behavioral Procedure

The 2 2 factorial design featured the factors ofMemory (present absent) and Distraction (presentabsent) Each trial had the same structure Target (1 sec)followed by Delay (7 sec) followed by Probe (1 sec)followed by ITI (13 sec) During memory-present trialstarget and probe stimuli were individual gray-scale malefaces (none wearing glasses) appearing on white back-grounds and subjects indicated with a button presswhether the latter matched the former ( p = 5) Memory-absent trials were procedurally identical except thatthey featured the same recurring female facesmdashfacelsquolsquoArsquorsquo always the target which denoted a lsquolsquono-memoryrsquorsquotrial face lsquolsquoBrsquorsquo always the probe On memory-absenttrials subjects were cued by the probe to indicate witha button press whether or not distracting faces appearedduring the delay period ( p = 5) thereby matching themotor response demands of memory-present trialsDistraction-present trials featured RSVP (3 Hz) of thefaces of males wearing glasses (to differentiate distrac-

tion items from memoranda) during the two middleseconds of the delay (Figure 1) Subjects were instructedto look at the distracting faces during all trials on whichthey appeared and to fixate centrally on distraction-absent trials Stimuli were presented via eyepiecesmounted on the head coil (Avotec Visible Eye StuartFL) and fitted with an infrared-based eye tracking system(SMI Inc Teltow Germany) Fixation was monitoredduring scanning and analyzed off-line with iLab software(Gitelman 2002) Trials during which subjects failedto fixate centrally from time 3 sec to time 5 sec werediscarded from subsequent analyses (Figure 5) Thiscriterion ensured that any evidence for differencesin delay-period activity between conditions could notbe attributed to theoretically uninteresting differencesin for example the amount of visual stimulation pro-duced by the distracting stimuli or the amount ofoculomotor activity Trials of each of the four types wereequiprobable and presented in a randomly determinedorder

fMRI Data Acquisition

Whole-brain images were acquired with a 3T scanner(GE Signa VHI Waukesha WI) High-resolution T1-weighted images (32 axial slices 09375 09375 4 mm)were obtained in every participant and a gradient-echoecho-planar sequence (TR = 2000 msec TE = 50 msec)was used to acquire data sensitive to the blood oxygenlevel dependent (BOLD) signal (Kwong et al 1992Ogawa et al 1992) within a 64 64 matrix (32 axialslices coplanar with T1 375 375 4 mm) Scans ofthe delayed-recognition task were preceded by a scan inwhich we derived an estimate of the hemodynamicresponse function (HRF) for each participant Duringthis scan each participant performed a simple RT taskthat required a bimanual button press once every 20 secin response to a brief change in shape of the fixationstimulus A partial F test associated with a Fourier basiscovariate set ( Josephs Turner amp Friston 1997) wasused to evaluate the significance of task-correlatedactivity in each voxel of primary somatosensory andmotor cortical regions of interest (Aguirre Zarahn ampDrsquoEsposito 1998) An HRF estimate was extracted fromthe suprathreshold voxels of these ROIs by spatiallyaveraging their time series filtering the resultant aver-aged fMRI time series to remove high (gt0244 Hz)and low (lt005 Hz) frequencies adjusting it to removethe effects of nuisance covariates (Friston HolmesPoline Heather amp Frackowiak 1995) and trial aver-aging The HRF characterizes the fMRI response result-ing from a brief impulse of neural activity (BoyntonEngel Glover amp Heeger 1996) and can vary markedlyacross participants (Aguirre et al 1998) The subject-specific HRFs were used to convolve independent vari-ables entered into the modified general linear model(GLM Worsley amp Friston 1995) that we used to analyze

Postle 1685

the data from the scans of the delayed-recognition task4

The eight scans of the delayed-recognition task eachlasted 6 min 12 sec (5 min 52 sec of task preceded by20 sec of dummy pulses to achieve a steady state of tis-sue magnetization)

fMRI Data Processing

The fMRI time series data were filtered and adjusted asdescribed previously (Postle Zarahn amp DrsquoEsposito2000) The principle of the fMRI time series analysiswas to model the fMRI signal changes evoked byeach stimulus presentation epoch with covariates com-posed of BOLD impulse response functions shiftedalong the time line of the task to represent various trialepochs (Figure 1 Postle et al 2000 Zarahn Aguirre

amp DrsquoEsposito 1997) The least-squares solution of theGLM of the fMRI time series data yielded parameterestimates that were associated with each covariateof interest The smoothness of the fMRI response toneural activity allows fMRI-evoked responses that arisefrom temporally dependent events to be resolved onthe order of 4 sec (That is if the covariates modelingthese events are spaced at 4-sec intervals the covariatepositioned at time 0 sec will account for virtually all ofthe variance attributable to an event starting at time 0 secand of duration of 1 sec or less thereby leaving thecovariate positioned at time 4 sec relatively uncontam-inated by variance attributable to the time 0 sec eventsee Figure 6 of Zarahn et al 1997) Differences in fMRIsignal (either between conditions or vs baseline) weretested by computing t statistics from contrasts between

Figure 5 Examples of eye-position data from four representative trials For all four examples trial epoch is color coded red = target stimulus

green = early delay blue = distraction portion of the delay period fuchsia = late delay black = probe stimulus plus first second of ITI (A) A

memory-presentdistraction-present trial with acceptable fixation during distraction The top panel is a 2-D representation of eye position on the

display lsquolsquoscreenrsquorsquo over the course of the trial axis labels correspond to degrees of visual angle Bottom panels display horizontal (x) and vertical ( y)eye position as a function of time with time in seconds on the horizontal axes and position in pixels on the vertical axes (each pixel measured

058 mm2) (B) A memory-absentdistraction-present trial with acceptable fixation during distraction All conventions same as in (A) Interrupted

portion of late delay corresponds to a blink when eyes were closed (C) A memory-presentdistraction-present trial that was discarded due

to unacceptable fixation during distraction All conventions same as (A) (D) A memory-absentdistraction-present trial that was discarded dueto unacceptable fixation during distraction All conventions same as in (A)

1686 Journal of Cognitive Neuroscience Volume 17 Number 11

parameter estimates associated with the covariates inquestion

Our analysis strategy was to identify delay-specificactivity in two regions the dlPFC and the IOTC5 frommemory-presentdistraction-absent trials This wouldidentify delay-period activity comparable to that ob-served in virtually all whole-brain neuroimaging studiesof similar tasks (eg with face stimuli Postle et al 2003Ranganath amp DrsquoEsposito 2001 Druzgal amp DrsquoEsposito2001 Courtney Ungerleider Keil amp Haxby 1997 Court-ney Ungerleider Keil amp Haxby 1996) Masks cor-responding to these two regions were drawn on eachsubjectrsquos anatomical scans Voxels located within thesemasks which became the VOIs were identified on tmaps of delay-period activity from memory-presentdis-traction-absent trials thresholded to a regionwise a of05 (one-tailed) with Bonferroni correction for the num-ber of voxels in the mask Note that with this approachwith spatially unsmoothed data individual voxels whoset value exceeds threshold can be interpreted as lsquolsquomean-ingfulrsquorsquo data (Postle et al 2000) Thus this procedurediffered somewhat from other applications of lsquolsquoregion-

of-interest (ROI) analysesrsquorsquo in that no selection criteriawere imposed with respect to cluster size or proximity ofcritical voxels This was done among other reasonsbecause we did not know a priori in which subregionof PFC to expect to find sensory-gating-related activityWithin the VOIs we evaluated the responses evokedby RSVP of faces during memory-presentdistraction-present trials versus memory-absentdistraction-presenttrials (Memory-absentdistraction-absent trials do notfigure directly into the test of the sensory gating hy-pothesis but were necessary to prevent anticipation ofdistraction on memory-absent trials)

Acknowledgments

The author thanks Olufunsho Faseyitan and Craig Rypstat forprogramming assistance and OF and Christopher Jordan forhelp with data collection and analysis This research was sup-ported by the National Institutes of Health RO1 MH064498

Reprint requests should be sent to Bradley R Postle Depart-mentofPsychologyUniversityof WisconsinndashMadison1202WestJohnson St Madison WI 53706 or via e-mail postlewiscedu

Figure 5 (continued)

Postle 1687

The data reported in this experiment have been depositedwith the fMRI Data Center archive (wwwfmridcorg) Theaccession number is 2-2005-119B9

Notes

1 This may be an empirical example of the lsquolsquoactive main-tenancersquorsquo proposed by Miller and Cohen (2001)2 The raw value of this subjectrsquos memory-absentdistraction-present effect in the dlPFC was 51 and so dividing thememory-presentdistraction-present by this value inflated it bya factor of almost 2 the opposite direction than that effectedby the normalization procedure in all other subjects exceptone whose raw memory-absentdistraction-present effect inthe dlPFC was 983 The fact that delay-period activity in the dlPFC VOIs wasnumerically lower during memory-presentdistraction-presentthan during memory-presentdistraction-absent trials is con-sistent with the idea that the sensory gating mechanism oncetriggered by neurons that are tonically active during unfilleddelay periods is implemented by a broader PFC network thatincludes regions that are not typically active during unfilleddelay periods Although the data are not shown here it iscertainly the case that in every subject considerably more dlPFCvoxels were active during the memory-presentdistraction-present delay than during the memory-presentdistraction-absent delay4 The HRF also varies in shape across brain regions with-in an individual participant (eg Miezen Maccotta OllingerPetersen amp Buckner 2000) However recent work indi-cates that the magnitude of variability in the shape of theHRF is greater across individuals within a region than it isacross regions within an individual (Handwerker Ollinger ampDrsquoEsposito 2004) Additionally it indicates that employinga participant-specific HRF can improve the sensitivity andmagnitude estimates of the least-squares solution of the GLMover comparable analyses that employ a generic HRF model(Handwerker et al 2004)5 The broader search space of the IOTC was used instead ofa more circumscribed area such as the fusiform face areabecause as can be seen in Figure 2 there was a considerableamount of topological variability of memory-presentdistraction-absent delay-period activity in the posterior ventral stream andno one area was activated in all subjects

REFERENCES

Aguirre G K Zarahn E amp DrsquoEsposito M (1998) Thevariability of human BOLD hemodynamic responsesNeuroimage 8 360ndash369

Baddeley A D amp Hitch G J (1974) Working memoryIn G H Bower (Ed) The psychology of learning andmotivation (Vol 8 pp 47ndash89) New York Academic Press

Bor D Duncan J Wiseman R J amp Owen A M (2003)Encoding strategies dissociate prefrontal activity fromworking memory demand Neuron 37 361ndash367

Boynton G M Engel S A Glover G H amp HeegerD J (1996) Linear systems analysis of functionalmagnetic resonance imaging in human V1 The Journalof Neuroscience 16 4207ndash4221

Braver T S Gray J R amp Burgess G C (in press)Explaining the many varieties of working memoryvariation Dual mechanisms of cognitive control In AConway C Jarrold M Kane A Miyake amp J Towse(Eds) Variation in working memory Oxford OxfordUniversity Press

Chao L amp Knight R (1998) Contribution of humanprefrontal cortex to delay performance Journal ofCognitive Neuroscience 10 167ndash177

Chao L L amp Knight R T (1995) Human prefrontal lesionsincrease distractibility to irrelevant sensory inputsNeuroReport 6 1605ndash1610

Connolly J D Goodale M A Goltz H C amp Munoz D P(2005) fMRI activation in the human frontal eye field iscorrelated with saccadic reaction time Journal ofNeurophysiology 94 605ndash611

Courtney S M (2004) Attention and cognitive control asemergent properties of information representation inworking memory Cognitive Affective amp BehavioralNeuroscience 4 501ndash516

Courtney S M Ungerleider L G Keil K amp Haxby J(1996) Object and spatial visual working memory activateseparate neural systems in human cortex Cerebral Cortex6 39ndash49

Courtney S M Ungerleider L G Keil K amp Haxby J V(1997) Transient and sustained activity in a distributedneural system for human working memory Nature 386608ndash611

Curtis C E amp DrsquoEsposito M (2003) Persistent activity inthe prefrontal cortex during working memory Trends inCognitive Sciences 7 415ndash423

Curtis C E Rao V Y amp DrsquoEsposito M (2004) Maintenanceof spatial and motor codes during oculomotor delayedresponse tasks The Journal of Neuroscience 243944ndash3952

DrsquoEsposito M Ballard D Zarahn E amp Aguirre G K(2000) The role of prefrontal cortex in sensory memoryand motor preparation An event-related fMRI studyNeuroimage 11 400ndash408

DrsquoEsposito M amp Postle B R (1999) The dependence of spanand delayed-response performance on prefrontal cortexNeuropsychologia 37 1303ndash1315

DrsquoEsposito M Postle B R Ballard D amp Lease J (1999)Maintenance versus manipulation of information held inworking memory An event-related fMRI study Brain ampCognition 41 66ndash86

DrsquoEsposito M Postle B R Jonides J amp Smith E E(1999) The neural substrate and temporal dynamicsof interference effects in working memory as revealedby event-related functional MRI Proceedings of theNational Academy of Sciences USA 96 7514ndash7519

de Fockert J W Rees G Frith C D amp Lavie N (2001) Therole of working memory in visual attention Science 2911803ndash1806

Druzgal T J amp DrsquoEsposito M (2001) A neural networkref lecting decisions about human faces Neuron 32947ndash955

Duncan J amp Miller E K (2002) Cognitive focus throughadaptive neural coding in the primate prefrontal cortexIn D Stuss amp R Knight (Eds) Principles of frontal lobefunction (pp 278ndash291) Oxford Oxford University Press

Friston K J Holmes A P Poline J-B Heather J D ampFrackowiak R S J (1995) Analysis of fMRI time-seriesrevisited Neuroimage 2 45ndash53

Fukushima T Hasegawa I amp Miyashita Y (2004)Prefrontal neuronal activity encodes spatial targetrepresentations sequentially updated after nonspatialtarget-shift cues Journal of Neurophysiology 911367ndash1380

Fuster J M (1995) Memory in the cerebral cortexCambridge MIT Press

Gitelman D R (2002) ILAB A program for post experimentaleye movement analysis Behavioral Research MethodsInstruments and Computers 34 605ndash612

1688 Journal of Cognitive Neuroscience Volume 17 Number 11

study memory-absentdistraction-present and memory-presentdistraction-present the latter would be themore difficult

fMRI

Delay-period activity was observed on memory-presentdistraction-absent trials in the dlPFC and the IOTC ineach of the 16 subjects (Figure 2) Within the function-ally defined VOIs distractor-evoked activity was mark-edly higher for memory-present than for memory-absenttrials in the dlPFC whereas the IOTC showed a trend inthe opposite direction (Figures 3 and 4) ANOVA assess-ing raw distraction-evoked activity in VOIs in the tworegions (dlPFC IOTC) and the two critical trial types(memory-absentdistraction-present memory-presentdistraction-present trials) revealed a main effect of re-gion F(115) = 1197 p lt 005 an effect of trial typethat approached significance F(115) = 396 p = 07and an interaction F(115) = 1523 p = 001 Theseresults were broadly consistent with our predictionsHowever because the main effect of region indicatedthat the overall evoked response was greater in theIOTC than in the dlPFC it was possible that the differ-

ential patterns of activity observed in these two VOIswere attributable at least in part to a difference in theinherent level of activity in the two regions in which theywere located To rule out this possibility the effectsfrom each subjectrsquos VOIs were normalized by dividingwithin VOI the distractor-evoked effects from both trialtypes of interest by the memory-absentdistraction-present effect This had the effect of transforming thememory-absentdistraction-present effect in both re-gions from each subject to the arbitrarily selected valueof 10 and transforming the memory-presentdistraction-present effect into a proportion of the memory-absentdistraction-present effect After this transformation wasperformed we excluded the data from one subject(12) from further analyses because the normalizedmemory-presentdistraction-present effect from thedlPFC VOIs of this subject was in excess of 3 standarddeviations greater than the mean2 For the remaining 15subjects the normalized data indicated that the meandelay-evoked response on memory-presentdistraction-present trials in dlPFC VOIs was 139 greater than onmemory-absentdistraction-present trials [a significantpairwise effect at t(14) = 21 p = 05] whereas in IOTCVOIs the mean delay-evoked response on memory-

Figure 2 Two representative slices from each subject illustrating dlPFC and IOTC VOIs Memory-presentdistraction-absent activity wasobserved in the dlPFC bilaterally in 13 subjects and unilaterally in the right hemisphere in 2 subjects and in the left hemisphere in 1 subject

in the IOTC it was observed bilaterally in 4 subjects unilaterally in the right hemisphere in 9 subjects and unilaterally in the left hemisphere

in 3 subjects

1682 Journal of Cognitive Neuroscience Volume 17 Number 11

presentdistraction-present trials was 75 smaller thanit was on memory-absentdistraction-present trials [apairwise effect that was not significant t(14) = 12ns] Thus the pattern of the normalized effects retained

the features of the raw data (Figure 4B) as confirmed byANOVA assessing the normalized results from the re-maining 15 subjects which produced a main effect ofregion F(114) = 49 p lt 05 a borderline effect of trial

Figure 3 An example of

results from a single subject

(21 in Figure 2) The two

top plots illustrate the trial-averaged time series data from

dlPFC and IOTC VOIs from

memory-presentdistraction-absent trials (fMRI MPDA)

and the corresponding delay-

period covariates scaled by

their parameter estimates(delay effect) The gray

bars along the horizontal axes

indicate the duration of the

delay period The two bottomplots illustrate the delay effects

from these same VOIs from

memory-presentdistraction-present (MPDP) and

memory-absentdistraction-

present (MADP) trials

Note that the signal intensityscales differ on plots for the

two regions

Figure 4 Group averageeffects (A) Memory-present

distraction-absent delay-period

VOIs with unscaled data (B)memory-presentdistraction-

absent delay-period VOIs with

normalized data (C) target-

evoked VOIs with unscaleddata (D) target-evoked VOIs

with normalized data MPDA =

memory-presentdistraction-

absent MPDP = memory-presentdistraction-present

MADP = memory-absent

distractionpresent Errorbars represent SEM

Postle 1683

type F(114) = 38 p = 07 and an interaction F(114) =49 p lt 05

One alternative to the sensory gating interpretation ofthese results might be that PFC delay-period activity onmemory-presentdistraction-present trials reflected dualtaskingmdashthe STR of target face information plus theincidental encoding (or some other type of processing)of the distracting faces This seems unlikely howeverbecause PFC delay-period activity was numerically loweron memory-presentdistraction-present trials than onmemory-presentdistraction-absent trials [although thiseffect did not achieve significance t(15) = 13 nsFigure 4A]

How specific was this pattern of results To assess thisquestion we applied the same contrasts to voxels in thedlPFC and IOTC that were selected for a differentpropertymdashresponsivity to visually presented faces Spe-cifically we identified voxels in the dlPFC and IOTC thatresponded to the Target epoch of the task (and re-moved from these VOIs any voxels that were also activeduring the delay period of memory-presentdistraction-absent trials) In this way we identified voxels from thesame regions in which we tested the sensory gatinghypothesis but whose function was different (presum-ably related to visual perception andor memory encod-ing but not to retention or another delay-periodprocess) For the raw data from these target-evokedVOIs (Figure 4C) as compared to the memory-evokedVOIs there was again a strong main effect of regionF(114) = 334 p lt 0001 a weaker borderline effect oftrial type F(114) = 33 p = 09 and now only a trendtoward an interaction F(114) = 40 p 07 When datafrom the target-evoked VOIs were normalized in thesame manner as had been the data from the delay-evoked VOIs (Figure 4D) their analysis with ANOVArevealed only a borderline effect of VOI F(114) = 40p = 06 for the first time a reliable effect of trial typeF(114) = 51 p lt 05 and again a trend toward aninteraction F(114) = 40 p = 06

The weakening of the Region Trial type interactionin target-evoked VOIs as compared with delay-evokedVOIs could be consistent with the claim that this inter-action predicted to be the fMRI signature of a sensorygating mechanism was selectively seen only in the delay-evoked VOIs It could also however simply reflect asituation in which the interaction terms for the twotypes of VOI fell on either side of the 05 thresholdbut did not differ from each other To distinguishbetween these two possibilities we compared the ef-fects of the experimental manipulation on delay-periodactivity across the two types of VOIs with an omnibusANOVA that directly compared the effect of memory onnormalized distraction-evoked activity across target-evoked and delay-evoked VOIs The factors for theomnibus ANOVA were region (dlPFC IOTC) trial type(memory present memory absent) and epoch definingthe VOIs (target delay) A three-way interaction would

be necessary to support the claim that the differentialeffect of memory demands on distraction-evoked activityin the dlPFC and IOTC was greater for delay-evoked thanfor target-evoked VOIs The ANOVA revealed a maineffect of region F(114) = 49 p lt 05 a borderlineeffect of trial type F(114) = 43 p = 06 and no effect ofepoch F(114) = 28 p gt 1 no interaction of Epoch Trial type F(114) = 28 p gt 1 but reliable inter-actions of Epoch Region F(114) = 47 p lt 05and of Region Trial type (F(114) = 49 p lt 05 and areliable interaction of Epoch Region Trial typeF(114) = 47 p lt 05 Thus it confirmed the specificityof the sensory gating effect to delay-evoked VOIs

DISCUSSION

The results demonstrate that delay-sensitive voxels intwo brain regions the dlPFC and the IOTC responddifferentially to the RSVP of faces conditioned on thebehavioral context in which these faces are presentedThe dlPFC response was 139 greater during concur-rent face memory whereas the IOTC response was 75lower This pattern was also seen to be selective becausea different group of dlPFC voxels (those that are targetsensitive) increased its activity by only 38 and theanalogously different group of IOTC voxels increased itsactivity by 3 These results are therefore consistentwith a model in which one function of dlPFC voxels thatare active during unfilled delay periods of delayed-recognition tasks is to perform a sensory gating functionBy this account the detection of potentially interferingstimuli in the environment triggers activity in networksof dlPFC neurons3 that in turn has the effect of sup-pressing perhaps via thalamic relays the afferent flow ofsensory information (Knight Staines Swick amp Chao1999) This would have beneficial effects for the signal-to-noise ratio in the networks of posterior neurons thatare responsible for the STR of information Thus a PFC-based control process may serve to protect the contentsof working and short-term memory from the deleteriouseffects external interference although it does not sup-port STR operations themselves

This account makes two broader claims about theworking memory literature First it offers an explanationas to why primates with PFC lesions can perform nor-mally on tasks of working and short-term memory whenpotential sources of distraction and inference are mini-mal but are impaired on the same tasks when distractionandor interference is high (Thompson-Schill et al2002 DrsquoEsposito amp Postle 1999 Chao amp Knight 1998Chao amp Knight 1995 Milner 1964 Malmo 1942)Second it proposes that sensory gating may accountfor a portion of the unfilled delay-period activity that isroutinely observed in PFC across a variety of neuro-imaging studies including those that do not explicitlyinclude distraction In relation to this second claim how-

1684 Journal of Cognitive Neuroscience Volume 17 Number 11

ever we cannot rule out the possibility that the dlPFCactivity observed during memory-presentdistraction-absent trials was due to a specific feature of our designBecause subjects could not predict which memorytrials would also contain distraction an effective strat-egy may have been to implement a sensory gatingstrategy on every memory trial To rule this out wouldrequire a two stage process of first scanning lsquolsquonaıversquorsquosubjects performing several blocks of only memory-presentdistraction-absent trials then introducing thefactor of delay-period distraction for a subsequent setof scans

The present study fits within a growing body of resultsthat document control rather than STR functions forthe PFC These include the monitoring and transforma-tion of information held in working memory (Bor et al2003 DrsquoEsposito Postle Ballard amp Lease 1999 Pet-rides 1995) and the attentional selection of information(Lebedev et al 2004 Sakai et al 2002 de Fockert ReesFrith amp Lavie 2001) These approaches which proposewell-defined testable mechanisms represent an advancebeyond the vague notion of lsquolsquomaintenancersquorsquo that is ofteninvoked in interpretations of PFC activity The presentstudy suggests one such alternative to lsquolsquomaintenancersquorsquo toaccount for sustained PFC activity observed during thesimplest tests of STR of information

METHODS

Subjects

Sixteen healthy young adults (mean age = 224 13 men)participated in this study All provided informed consentand were paid in compensation for participating in thestudy

Behavioral Procedure

The 2 2 factorial design featured the factors ofMemory (present absent) and Distraction (presentabsent) Each trial had the same structure Target (1 sec)followed by Delay (7 sec) followed by Probe (1 sec)followed by ITI (13 sec) During memory-present trialstarget and probe stimuli were individual gray-scale malefaces (none wearing glasses) appearing on white back-grounds and subjects indicated with a button presswhether the latter matched the former ( p = 5) Memory-absent trials were procedurally identical except thatthey featured the same recurring female facesmdashfacelsquolsquoArsquorsquo always the target which denoted a lsquolsquono-memoryrsquorsquotrial face lsquolsquoBrsquorsquo always the probe On memory-absenttrials subjects were cued by the probe to indicate witha button press whether or not distracting faces appearedduring the delay period ( p = 5) thereby matching themotor response demands of memory-present trialsDistraction-present trials featured RSVP (3 Hz) of thefaces of males wearing glasses (to differentiate distrac-

tion items from memoranda) during the two middleseconds of the delay (Figure 1) Subjects were instructedto look at the distracting faces during all trials on whichthey appeared and to fixate centrally on distraction-absent trials Stimuli were presented via eyepiecesmounted on the head coil (Avotec Visible Eye StuartFL) and fitted with an infrared-based eye tracking system(SMI Inc Teltow Germany) Fixation was monitoredduring scanning and analyzed off-line with iLab software(Gitelman 2002) Trials during which subjects failedto fixate centrally from time 3 sec to time 5 sec werediscarded from subsequent analyses (Figure 5) Thiscriterion ensured that any evidence for differencesin delay-period activity between conditions could notbe attributed to theoretically uninteresting differencesin for example the amount of visual stimulation pro-duced by the distracting stimuli or the amount ofoculomotor activity Trials of each of the four types wereequiprobable and presented in a randomly determinedorder

fMRI Data Acquisition

Whole-brain images were acquired with a 3T scanner(GE Signa VHI Waukesha WI) High-resolution T1-weighted images (32 axial slices 09375 09375 4 mm)were obtained in every participant and a gradient-echoecho-planar sequence (TR = 2000 msec TE = 50 msec)was used to acquire data sensitive to the blood oxygenlevel dependent (BOLD) signal (Kwong et al 1992Ogawa et al 1992) within a 64 64 matrix (32 axialslices coplanar with T1 375 375 4 mm) Scans ofthe delayed-recognition task were preceded by a scan inwhich we derived an estimate of the hemodynamicresponse function (HRF) for each participant Duringthis scan each participant performed a simple RT taskthat required a bimanual button press once every 20 secin response to a brief change in shape of the fixationstimulus A partial F test associated with a Fourier basiscovariate set ( Josephs Turner amp Friston 1997) wasused to evaluate the significance of task-correlatedactivity in each voxel of primary somatosensory andmotor cortical regions of interest (Aguirre Zarahn ampDrsquoEsposito 1998) An HRF estimate was extracted fromthe suprathreshold voxels of these ROIs by spatiallyaveraging their time series filtering the resultant aver-aged fMRI time series to remove high (gt0244 Hz)and low (lt005 Hz) frequencies adjusting it to removethe effects of nuisance covariates (Friston HolmesPoline Heather amp Frackowiak 1995) and trial aver-aging The HRF characterizes the fMRI response result-ing from a brief impulse of neural activity (BoyntonEngel Glover amp Heeger 1996) and can vary markedlyacross participants (Aguirre et al 1998) The subject-specific HRFs were used to convolve independent vari-ables entered into the modified general linear model(GLM Worsley amp Friston 1995) that we used to analyze

Postle 1685

the data from the scans of the delayed-recognition task4

The eight scans of the delayed-recognition task eachlasted 6 min 12 sec (5 min 52 sec of task preceded by20 sec of dummy pulses to achieve a steady state of tis-sue magnetization)

fMRI Data Processing

The fMRI time series data were filtered and adjusted asdescribed previously (Postle Zarahn amp DrsquoEsposito2000) The principle of the fMRI time series analysiswas to model the fMRI signal changes evoked byeach stimulus presentation epoch with covariates com-posed of BOLD impulse response functions shiftedalong the time line of the task to represent various trialepochs (Figure 1 Postle et al 2000 Zarahn Aguirre

amp DrsquoEsposito 1997) The least-squares solution of theGLM of the fMRI time series data yielded parameterestimates that were associated with each covariateof interest The smoothness of the fMRI response toneural activity allows fMRI-evoked responses that arisefrom temporally dependent events to be resolved onthe order of 4 sec (That is if the covariates modelingthese events are spaced at 4-sec intervals the covariatepositioned at time 0 sec will account for virtually all ofthe variance attributable to an event starting at time 0 secand of duration of 1 sec or less thereby leaving thecovariate positioned at time 4 sec relatively uncontam-inated by variance attributable to the time 0 sec eventsee Figure 6 of Zarahn et al 1997) Differences in fMRIsignal (either between conditions or vs baseline) weretested by computing t statistics from contrasts between

Figure 5 Examples of eye-position data from four representative trials For all four examples trial epoch is color coded red = target stimulus

green = early delay blue = distraction portion of the delay period fuchsia = late delay black = probe stimulus plus first second of ITI (A) A

memory-presentdistraction-present trial with acceptable fixation during distraction The top panel is a 2-D representation of eye position on the

display lsquolsquoscreenrsquorsquo over the course of the trial axis labels correspond to degrees of visual angle Bottom panels display horizontal (x) and vertical ( y)eye position as a function of time with time in seconds on the horizontal axes and position in pixels on the vertical axes (each pixel measured

058 mm2) (B) A memory-absentdistraction-present trial with acceptable fixation during distraction All conventions same as in (A) Interrupted

portion of late delay corresponds to a blink when eyes were closed (C) A memory-presentdistraction-present trial that was discarded due

to unacceptable fixation during distraction All conventions same as (A) (D) A memory-absentdistraction-present trial that was discarded dueto unacceptable fixation during distraction All conventions same as in (A)

1686 Journal of Cognitive Neuroscience Volume 17 Number 11

parameter estimates associated with the covariates inquestion

Our analysis strategy was to identify delay-specificactivity in two regions the dlPFC and the IOTC5 frommemory-presentdistraction-absent trials This wouldidentify delay-period activity comparable to that ob-served in virtually all whole-brain neuroimaging studiesof similar tasks (eg with face stimuli Postle et al 2003Ranganath amp DrsquoEsposito 2001 Druzgal amp DrsquoEsposito2001 Courtney Ungerleider Keil amp Haxby 1997 Court-ney Ungerleider Keil amp Haxby 1996) Masks cor-responding to these two regions were drawn on eachsubjectrsquos anatomical scans Voxels located within thesemasks which became the VOIs were identified on tmaps of delay-period activity from memory-presentdis-traction-absent trials thresholded to a regionwise a of05 (one-tailed) with Bonferroni correction for the num-ber of voxels in the mask Note that with this approachwith spatially unsmoothed data individual voxels whoset value exceeds threshold can be interpreted as lsquolsquomean-ingfulrsquorsquo data (Postle et al 2000) Thus this procedurediffered somewhat from other applications of lsquolsquoregion-

of-interest (ROI) analysesrsquorsquo in that no selection criteriawere imposed with respect to cluster size or proximity ofcritical voxels This was done among other reasonsbecause we did not know a priori in which subregionof PFC to expect to find sensory-gating-related activityWithin the VOIs we evaluated the responses evokedby RSVP of faces during memory-presentdistraction-present trials versus memory-absentdistraction-presenttrials (Memory-absentdistraction-absent trials do notfigure directly into the test of the sensory gating hy-pothesis but were necessary to prevent anticipation ofdistraction on memory-absent trials)

Acknowledgments

The author thanks Olufunsho Faseyitan and Craig Rypstat forprogramming assistance and OF and Christopher Jordan forhelp with data collection and analysis This research was sup-ported by the National Institutes of Health RO1 MH064498

Reprint requests should be sent to Bradley R Postle Depart-mentofPsychologyUniversityof WisconsinndashMadison1202WestJohnson St Madison WI 53706 or via e-mail postlewiscedu

Figure 5 (continued)

Postle 1687

The data reported in this experiment have been depositedwith the fMRI Data Center archive (wwwfmridcorg) Theaccession number is 2-2005-119B9

Notes

1 This may be an empirical example of the lsquolsquoactive main-tenancersquorsquo proposed by Miller and Cohen (2001)2 The raw value of this subjectrsquos memory-absentdistraction-present effect in the dlPFC was 51 and so dividing thememory-presentdistraction-present by this value inflated it bya factor of almost 2 the opposite direction than that effectedby the normalization procedure in all other subjects exceptone whose raw memory-absentdistraction-present effect inthe dlPFC was 983 The fact that delay-period activity in the dlPFC VOIs wasnumerically lower during memory-presentdistraction-presentthan during memory-presentdistraction-absent trials is con-sistent with the idea that the sensory gating mechanism oncetriggered by neurons that are tonically active during unfilleddelay periods is implemented by a broader PFC network thatincludes regions that are not typically active during unfilleddelay periods Although the data are not shown here it iscertainly the case that in every subject considerably more dlPFCvoxels were active during the memory-presentdistraction-present delay than during the memory-presentdistraction-absent delay4 The HRF also varies in shape across brain regions with-in an individual participant (eg Miezen Maccotta OllingerPetersen amp Buckner 2000) However recent work indi-cates that the magnitude of variability in the shape of theHRF is greater across individuals within a region than it isacross regions within an individual (Handwerker Ollinger ampDrsquoEsposito 2004) Additionally it indicates that employinga participant-specific HRF can improve the sensitivity andmagnitude estimates of the least-squares solution of the GLMover comparable analyses that employ a generic HRF model(Handwerker et al 2004)5 The broader search space of the IOTC was used instead ofa more circumscribed area such as the fusiform face areabecause as can be seen in Figure 2 there was a considerableamount of topological variability of memory-presentdistraction-absent delay-period activity in the posterior ventral stream andno one area was activated in all subjects

REFERENCES

Aguirre G K Zarahn E amp DrsquoEsposito M (1998) Thevariability of human BOLD hemodynamic responsesNeuroimage 8 360ndash369

Baddeley A D amp Hitch G J (1974) Working memoryIn G H Bower (Ed) The psychology of learning andmotivation (Vol 8 pp 47ndash89) New York Academic Press

Bor D Duncan J Wiseman R J amp Owen A M (2003)Encoding strategies dissociate prefrontal activity fromworking memory demand Neuron 37 361ndash367

Boynton G M Engel S A Glover G H amp HeegerD J (1996) Linear systems analysis of functionalmagnetic resonance imaging in human V1 The Journalof Neuroscience 16 4207ndash4221

Braver T S Gray J R amp Burgess G C (in press)Explaining the many varieties of working memoryvariation Dual mechanisms of cognitive control In AConway C Jarrold M Kane A Miyake amp J Towse(Eds) Variation in working memory Oxford OxfordUniversity Press

Chao L amp Knight R (1998) Contribution of humanprefrontal cortex to delay performance Journal ofCognitive Neuroscience 10 167ndash177

Chao L L amp Knight R T (1995) Human prefrontal lesionsincrease distractibility to irrelevant sensory inputsNeuroReport 6 1605ndash1610

Connolly J D Goodale M A Goltz H C amp Munoz D P(2005) fMRI activation in the human frontal eye field iscorrelated with saccadic reaction time Journal ofNeurophysiology 94 605ndash611

Courtney S M (2004) Attention and cognitive control asemergent properties of information representation inworking memory Cognitive Affective amp BehavioralNeuroscience 4 501ndash516

Courtney S M Ungerleider L G Keil K amp Haxby J(1996) Object and spatial visual working memory activateseparate neural systems in human cortex Cerebral Cortex6 39ndash49

Courtney S M Ungerleider L G Keil K amp Haxby J V(1997) Transient and sustained activity in a distributedneural system for human working memory Nature 386608ndash611

Curtis C E amp DrsquoEsposito M (2003) Persistent activity inthe prefrontal cortex during working memory Trends inCognitive Sciences 7 415ndash423

Curtis C E Rao V Y amp DrsquoEsposito M (2004) Maintenanceof spatial and motor codes during oculomotor delayedresponse tasks The Journal of Neuroscience 243944ndash3952

DrsquoEsposito M Ballard D Zarahn E amp Aguirre G K(2000) The role of prefrontal cortex in sensory memoryand motor preparation An event-related fMRI studyNeuroimage 11 400ndash408

DrsquoEsposito M amp Postle B R (1999) The dependence of spanand delayed-response performance on prefrontal cortexNeuropsychologia 37 1303ndash1315

DrsquoEsposito M Postle B R Ballard D amp Lease J (1999)Maintenance versus manipulation of information held inworking memory An event-related fMRI study Brain ampCognition 41 66ndash86

DrsquoEsposito M Postle B R Jonides J amp Smith E E(1999) The neural substrate and temporal dynamicsof interference effects in working memory as revealedby event-related functional MRI Proceedings of theNational Academy of Sciences USA 96 7514ndash7519

de Fockert J W Rees G Frith C D amp Lavie N (2001) Therole of working memory in visual attention Science 2911803ndash1806

Druzgal T J amp DrsquoEsposito M (2001) A neural networkref lecting decisions about human faces Neuron 32947ndash955

Duncan J amp Miller E K (2002) Cognitive focus throughadaptive neural coding in the primate prefrontal cortexIn D Stuss amp R Knight (Eds) Principles of frontal lobefunction (pp 278ndash291) Oxford Oxford University Press

Friston K J Holmes A P Poline J-B Heather J D ampFrackowiak R S J (1995) Analysis of fMRI time-seriesrevisited Neuroimage 2 45ndash53

Fukushima T Hasegawa I amp Miyashita Y (2004)Prefrontal neuronal activity encodes spatial targetrepresentations sequentially updated after nonspatialtarget-shift cues Journal of Neurophysiology 911367ndash1380

Fuster J M (1995) Memory in the cerebral cortexCambridge MIT Press

Gitelman D R (2002) ILAB A program for post experimentaleye movement analysis Behavioral Research MethodsInstruments and Computers 34 605ndash612

1688 Journal of Cognitive Neuroscience Volume 17 Number 11

presentdistraction-present trials was 75 smaller thanit was on memory-absentdistraction-present trials [apairwise effect that was not significant t(14) = 12ns] Thus the pattern of the normalized effects retained

the features of the raw data (Figure 4B) as confirmed byANOVA assessing the normalized results from the re-maining 15 subjects which produced a main effect ofregion F(114) = 49 p lt 05 a borderline effect of trial

Figure 3 An example of

results from a single subject

(21 in Figure 2) The two

top plots illustrate the trial-averaged time series data from

dlPFC and IOTC VOIs from

memory-presentdistraction-absent trials (fMRI MPDA)

and the corresponding delay-

period covariates scaled by

their parameter estimates(delay effect) The gray

bars along the horizontal axes

indicate the duration of the

delay period The two bottomplots illustrate the delay effects

from these same VOIs from

memory-presentdistraction-present (MPDP) and

memory-absentdistraction-

present (MADP) trials

Note that the signal intensityscales differ on plots for the

two regions

Figure 4 Group averageeffects (A) Memory-present

distraction-absent delay-period

VOIs with unscaled data (B)memory-presentdistraction-

absent delay-period VOIs with

normalized data (C) target-

evoked VOIs with unscaleddata (D) target-evoked VOIs

with normalized data MPDA =

memory-presentdistraction-

absent MPDP = memory-presentdistraction-present

MADP = memory-absent

distractionpresent Errorbars represent SEM

Postle 1683

type F(114) = 38 p = 07 and an interaction F(114) =49 p lt 05

One alternative to the sensory gating interpretation ofthese results might be that PFC delay-period activity onmemory-presentdistraction-present trials reflected dualtaskingmdashthe STR of target face information plus theincidental encoding (or some other type of processing)of the distracting faces This seems unlikely howeverbecause PFC delay-period activity was numerically loweron memory-presentdistraction-present trials than onmemory-presentdistraction-absent trials [although thiseffect did not achieve significance t(15) = 13 nsFigure 4A]

How specific was this pattern of results To assess thisquestion we applied the same contrasts to voxels in thedlPFC and IOTC that were selected for a differentpropertymdashresponsivity to visually presented faces Spe-cifically we identified voxels in the dlPFC and IOTC thatresponded to the Target epoch of the task (and re-moved from these VOIs any voxels that were also activeduring the delay period of memory-presentdistraction-absent trials) In this way we identified voxels from thesame regions in which we tested the sensory gatinghypothesis but whose function was different (presum-ably related to visual perception andor memory encod-ing but not to retention or another delay-periodprocess) For the raw data from these target-evokedVOIs (Figure 4C) as compared to the memory-evokedVOIs there was again a strong main effect of regionF(114) = 334 p lt 0001 a weaker borderline effect oftrial type F(114) = 33 p = 09 and now only a trendtoward an interaction F(114) = 40 p 07 When datafrom the target-evoked VOIs were normalized in thesame manner as had been the data from the delay-evoked VOIs (Figure 4D) their analysis with ANOVArevealed only a borderline effect of VOI F(114) = 40p = 06 for the first time a reliable effect of trial typeF(114) = 51 p lt 05 and again a trend toward aninteraction F(114) = 40 p = 06

The weakening of the Region Trial type interactionin target-evoked VOIs as compared with delay-evokedVOIs could be consistent with the claim that this inter-action predicted to be the fMRI signature of a sensorygating mechanism was selectively seen only in the delay-evoked VOIs It could also however simply reflect asituation in which the interaction terms for the twotypes of VOI fell on either side of the 05 thresholdbut did not differ from each other To distinguishbetween these two possibilities we compared the ef-fects of the experimental manipulation on delay-periodactivity across the two types of VOIs with an omnibusANOVA that directly compared the effect of memory onnormalized distraction-evoked activity across target-evoked and delay-evoked VOIs The factors for theomnibus ANOVA were region (dlPFC IOTC) trial type(memory present memory absent) and epoch definingthe VOIs (target delay) A three-way interaction would

be necessary to support the claim that the differentialeffect of memory demands on distraction-evoked activityin the dlPFC and IOTC was greater for delay-evoked thanfor target-evoked VOIs The ANOVA revealed a maineffect of region F(114) = 49 p lt 05 a borderlineeffect of trial type F(114) = 43 p = 06 and no effect ofepoch F(114) = 28 p gt 1 no interaction of Epoch Trial type F(114) = 28 p gt 1 but reliable inter-actions of Epoch Region F(114) = 47 p lt 05and of Region Trial type (F(114) = 49 p lt 05 and areliable interaction of Epoch Region Trial typeF(114) = 47 p lt 05 Thus it confirmed the specificityof the sensory gating effect to delay-evoked VOIs

DISCUSSION

The results demonstrate that delay-sensitive voxels intwo brain regions the dlPFC and the IOTC responddifferentially to the RSVP of faces conditioned on thebehavioral context in which these faces are presentedThe dlPFC response was 139 greater during concur-rent face memory whereas the IOTC response was 75lower This pattern was also seen to be selective becausea different group of dlPFC voxels (those that are targetsensitive) increased its activity by only 38 and theanalogously different group of IOTC voxels increased itsactivity by 3 These results are therefore consistentwith a model in which one function of dlPFC voxels thatare active during unfilled delay periods of delayed-recognition tasks is to perform a sensory gating functionBy this account the detection of potentially interferingstimuli in the environment triggers activity in networksof dlPFC neurons3 that in turn has the effect of sup-pressing perhaps via thalamic relays the afferent flow ofsensory information (Knight Staines Swick amp Chao1999) This would have beneficial effects for the signal-to-noise ratio in the networks of posterior neurons thatare responsible for the STR of information Thus a PFC-based control process may serve to protect the contentsof working and short-term memory from the deleteriouseffects external interference although it does not sup-port STR operations themselves

This account makes two broader claims about theworking memory literature First it offers an explanationas to why primates with PFC lesions can perform nor-mally on tasks of working and short-term memory whenpotential sources of distraction and inference are mini-mal but are impaired on the same tasks when distractionandor interference is high (Thompson-Schill et al2002 DrsquoEsposito amp Postle 1999 Chao amp Knight 1998Chao amp Knight 1995 Milner 1964 Malmo 1942)Second it proposes that sensory gating may accountfor a portion of the unfilled delay-period activity that isroutinely observed in PFC across a variety of neuro-imaging studies including those that do not explicitlyinclude distraction In relation to this second claim how-

1684 Journal of Cognitive Neuroscience Volume 17 Number 11

ever we cannot rule out the possibility that the dlPFCactivity observed during memory-presentdistraction-absent trials was due to a specific feature of our designBecause subjects could not predict which memorytrials would also contain distraction an effective strat-egy may have been to implement a sensory gatingstrategy on every memory trial To rule this out wouldrequire a two stage process of first scanning lsquolsquonaıversquorsquosubjects performing several blocks of only memory-presentdistraction-absent trials then introducing thefactor of delay-period distraction for a subsequent setof scans

The present study fits within a growing body of resultsthat document control rather than STR functions forthe PFC These include the monitoring and transforma-tion of information held in working memory (Bor et al2003 DrsquoEsposito Postle Ballard amp Lease 1999 Pet-rides 1995) and the attentional selection of information(Lebedev et al 2004 Sakai et al 2002 de Fockert ReesFrith amp Lavie 2001) These approaches which proposewell-defined testable mechanisms represent an advancebeyond the vague notion of lsquolsquomaintenancersquorsquo that is ofteninvoked in interpretations of PFC activity The presentstudy suggests one such alternative to lsquolsquomaintenancersquorsquo toaccount for sustained PFC activity observed during thesimplest tests of STR of information

METHODS

Subjects

Sixteen healthy young adults (mean age = 224 13 men)participated in this study All provided informed consentand were paid in compensation for participating in thestudy

Behavioral Procedure

The 2 2 factorial design featured the factors ofMemory (present absent) and Distraction (presentabsent) Each trial had the same structure Target (1 sec)followed by Delay (7 sec) followed by Probe (1 sec)followed by ITI (13 sec) During memory-present trialstarget and probe stimuli were individual gray-scale malefaces (none wearing glasses) appearing on white back-grounds and subjects indicated with a button presswhether the latter matched the former ( p = 5) Memory-absent trials were procedurally identical except thatthey featured the same recurring female facesmdashfacelsquolsquoArsquorsquo always the target which denoted a lsquolsquono-memoryrsquorsquotrial face lsquolsquoBrsquorsquo always the probe On memory-absenttrials subjects were cued by the probe to indicate witha button press whether or not distracting faces appearedduring the delay period ( p = 5) thereby matching themotor response demands of memory-present trialsDistraction-present trials featured RSVP (3 Hz) of thefaces of males wearing glasses (to differentiate distrac-

tion items from memoranda) during the two middleseconds of the delay (Figure 1) Subjects were instructedto look at the distracting faces during all trials on whichthey appeared and to fixate centrally on distraction-absent trials Stimuli were presented via eyepiecesmounted on the head coil (Avotec Visible Eye StuartFL) and fitted with an infrared-based eye tracking system(SMI Inc Teltow Germany) Fixation was monitoredduring scanning and analyzed off-line with iLab software(Gitelman 2002) Trials during which subjects failedto fixate centrally from time 3 sec to time 5 sec werediscarded from subsequent analyses (Figure 5) Thiscriterion ensured that any evidence for differencesin delay-period activity between conditions could notbe attributed to theoretically uninteresting differencesin for example the amount of visual stimulation pro-duced by the distracting stimuli or the amount ofoculomotor activity Trials of each of the four types wereequiprobable and presented in a randomly determinedorder

fMRI Data Acquisition

Whole-brain images were acquired with a 3T scanner(GE Signa VHI Waukesha WI) High-resolution T1-weighted images (32 axial slices 09375 09375 4 mm)were obtained in every participant and a gradient-echoecho-planar sequence (TR = 2000 msec TE = 50 msec)was used to acquire data sensitive to the blood oxygenlevel dependent (BOLD) signal (Kwong et al 1992Ogawa et al 1992) within a 64 64 matrix (32 axialslices coplanar with T1 375 375 4 mm) Scans ofthe delayed-recognition task were preceded by a scan inwhich we derived an estimate of the hemodynamicresponse function (HRF) for each participant Duringthis scan each participant performed a simple RT taskthat required a bimanual button press once every 20 secin response to a brief change in shape of the fixationstimulus A partial F test associated with a Fourier basiscovariate set ( Josephs Turner amp Friston 1997) wasused to evaluate the significance of task-correlatedactivity in each voxel of primary somatosensory andmotor cortical regions of interest (Aguirre Zarahn ampDrsquoEsposito 1998) An HRF estimate was extracted fromthe suprathreshold voxels of these ROIs by spatiallyaveraging their time series filtering the resultant aver-aged fMRI time series to remove high (gt0244 Hz)and low (lt005 Hz) frequencies adjusting it to removethe effects of nuisance covariates (Friston HolmesPoline Heather amp Frackowiak 1995) and trial aver-aging The HRF characterizes the fMRI response result-ing from a brief impulse of neural activity (BoyntonEngel Glover amp Heeger 1996) and can vary markedlyacross participants (Aguirre et al 1998) The subject-specific HRFs were used to convolve independent vari-ables entered into the modified general linear model(GLM Worsley amp Friston 1995) that we used to analyze

Postle 1685

the data from the scans of the delayed-recognition task4

The eight scans of the delayed-recognition task eachlasted 6 min 12 sec (5 min 52 sec of task preceded by20 sec of dummy pulses to achieve a steady state of tis-sue magnetization)

fMRI Data Processing

The fMRI time series data were filtered and adjusted asdescribed previously (Postle Zarahn amp DrsquoEsposito2000) The principle of the fMRI time series analysiswas to model the fMRI signal changes evoked byeach stimulus presentation epoch with covariates com-posed of BOLD impulse response functions shiftedalong the time line of the task to represent various trialepochs (Figure 1 Postle et al 2000 Zarahn Aguirre

amp DrsquoEsposito 1997) The least-squares solution of theGLM of the fMRI time series data yielded parameterestimates that were associated with each covariateof interest The smoothness of the fMRI response toneural activity allows fMRI-evoked responses that arisefrom temporally dependent events to be resolved onthe order of 4 sec (That is if the covariates modelingthese events are spaced at 4-sec intervals the covariatepositioned at time 0 sec will account for virtually all ofthe variance attributable to an event starting at time 0 secand of duration of 1 sec or less thereby leaving thecovariate positioned at time 4 sec relatively uncontam-inated by variance attributable to the time 0 sec eventsee Figure 6 of Zarahn et al 1997) Differences in fMRIsignal (either between conditions or vs baseline) weretested by computing t statistics from contrasts between

Figure 5 Examples of eye-position data from four representative trials For all four examples trial epoch is color coded red = target stimulus

green = early delay blue = distraction portion of the delay period fuchsia = late delay black = probe stimulus plus first second of ITI (A) A

memory-presentdistraction-present trial with acceptable fixation during distraction The top panel is a 2-D representation of eye position on the

display lsquolsquoscreenrsquorsquo over the course of the trial axis labels correspond to degrees of visual angle Bottom panels display horizontal (x) and vertical ( y)eye position as a function of time with time in seconds on the horizontal axes and position in pixels on the vertical axes (each pixel measured

058 mm2) (B) A memory-absentdistraction-present trial with acceptable fixation during distraction All conventions same as in (A) Interrupted

portion of late delay corresponds to a blink when eyes were closed (C) A memory-presentdistraction-present trial that was discarded due

to unacceptable fixation during distraction All conventions same as (A) (D) A memory-absentdistraction-present trial that was discarded dueto unacceptable fixation during distraction All conventions same as in (A)

1686 Journal of Cognitive Neuroscience Volume 17 Number 11

parameter estimates associated with the covariates inquestion

Our analysis strategy was to identify delay-specificactivity in two regions the dlPFC and the IOTC5 frommemory-presentdistraction-absent trials This wouldidentify delay-period activity comparable to that ob-served in virtually all whole-brain neuroimaging studiesof similar tasks (eg with face stimuli Postle et al 2003Ranganath amp DrsquoEsposito 2001 Druzgal amp DrsquoEsposito2001 Courtney Ungerleider Keil amp Haxby 1997 Court-ney Ungerleider Keil amp Haxby 1996) Masks cor-responding to these two regions were drawn on eachsubjectrsquos anatomical scans Voxels located within thesemasks which became the VOIs were identified on tmaps of delay-period activity from memory-presentdis-traction-absent trials thresholded to a regionwise a of05 (one-tailed) with Bonferroni correction for the num-ber of voxels in the mask Note that with this approachwith spatially unsmoothed data individual voxels whoset value exceeds threshold can be interpreted as lsquolsquomean-ingfulrsquorsquo data (Postle et al 2000) Thus this procedurediffered somewhat from other applications of lsquolsquoregion-

of-interest (ROI) analysesrsquorsquo in that no selection criteriawere imposed with respect to cluster size or proximity ofcritical voxels This was done among other reasonsbecause we did not know a priori in which subregionof PFC to expect to find sensory-gating-related activityWithin the VOIs we evaluated the responses evokedby RSVP of faces during memory-presentdistraction-present trials versus memory-absentdistraction-presenttrials (Memory-absentdistraction-absent trials do notfigure directly into the test of the sensory gating hy-pothesis but were necessary to prevent anticipation ofdistraction on memory-absent trials)

Acknowledgments

The author thanks Olufunsho Faseyitan and Craig Rypstat forprogramming assistance and OF and Christopher Jordan forhelp with data collection and analysis This research was sup-ported by the National Institutes of Health RO1 MH064498

Reprint requests should be sent to Bradley R Postle Depart-mentofPsychologyUniversityof WisconsinndashMadison1202WestJohnson St Madison WI 53706 or via e-mail postlewiscedu

Figure 5 (continued)

Postle 1687

The data reported in this experiment have been depositedwith the fMRI Data Center archive (wwwfmridcorg) Theaccession number is 2-2005-119B9

Notes

1 This may be an empirical example of the lsquolsquoactive main-tenancersquorsquo proposed by Miller and Cohen (2001)2 The raw value of this subjectrsquos memory-absentdistraction-present effect in the dlPFC was 51 and so dividing thememory-presentdistraction-present by this value inflated it bya factor of almost 2 the opposite direction than that effectedby the normalization procedure in all other subjects exceptone whose raw memory-absentdistraction-present effect inthe dlPFC was 983 The fact that delay-period activity in the dlPFC VOIs wasnumerically lower during memory-presentdistraction-presentthan during memory-presentdistraction-absent trials is con-sistent with the idea that the sensory gating mechanism oncetriggered by neurons that are tonically active during unfilleddelay periods is implemented by a broader PFC network thatincludes regions that are not typically active during unfilleddelay periods Although the data are not shown here it iscertainly the case that in every subject considerably more dlPFCvoxels were active during the memory-presentdistraction-present delay than during the memory-presentdistraction-absent delay4 The HRF also varies in shape across brain regions with-in an individual participant (eg Miezen Maccotta OllingerPetersen amp Buckner 2000) However recent work indi-cates that the magnitude of variability in the shape of theHRF is greater across individuals within a region than it isacross regions within an individual (Handwerker Ollinger ampDrsquoEsposito 2004) Additionally it indicates that employinga participant-specific HRF can improve the sensitivity andmagnitude estimates of the least-squares solution of the GLMover comparable analyses that employ a generic HRF model(Handwerker et al 2004)5 The broader search space of the IOTC was used instead ofa more circumscribed area such as the fusiform face areabecause as can be seen in Figure 2 there was a considerableamount of topological variability of memory-presentdistraction-absent delay-period activity in the posterior ventral stream andno one area was activated in all subjects

REFERENCES

Aguirre G K Zarahn E amp DrsquoEsposito M (1998) Thevariability of human BOLD hemodynamic responsesNeuroimage 8 360ndash369

Baddeley A D amp Hitch G J (1974) Working memoryIn G H Bower (Ed) The psychology of learning andmotivation (Vol 8 pp 47ndash89) New York Academic Press

Bor D Duncan J Wiseman R J amp Owen A M (2003)Encoding strategies dissociate prefrontal activity fromworking memory demand Neuron 37 361ndash367

Boynton G M Engel S A Glover G H amp HeegerD J (1996) Linear systems analysis of functionalmagnetic resonance imaging in human V1 The Journalof Neuroscience 16 4207ndash4221

Braver T S Gray J R amp Burgess G C (in press)Explaining the many varieties of working memoryvariation Dual mechanisms of cognitive control In AConway C Jarrold M Kane A Miyake amp J Towse(Eds) Variation in working memory Oxford OxfordUniversity Press

Chao L amp Knight R (1998) Contribution of humanprefrontal cortex to delay performance Journal ofCognitive Neuroscience 10 167ndash177

Chao L L amp Knight R T (1995) Human prefrontal lesionsincrease distractibility to irrelevant sensory inputsNeuroReport 6 1605ndash1610

Connolly J D Goodale M A Goltz H C amp Munoz D P(2005) fMRI activation in the human frontal eye field iscorrelated with saccadic reaction time Journal ofNeurophysiology 94 605ndash611

Courtney S M (2004) Attention and cognitive control asemergent properties of information representation inworking memory Cognitive Affective amp BehavioralNeuroscience 4 501ndash516

Courtney S M Ungerleider L G Keil K amp Haxby J(1996) Object and spatial visual working memory activateseparate neural systems in human cortex Cerebral Cortex6 39ndash49

Courtney S M Ungerleider L G Keil K amp Haxby J V(1997) Transient and sustained activity in a distributedneural system for human working memory Nature 386608ndash611

Curtis C E amp DrsquoEsposito M (2003) Persistent activity inthe prefrontal cortex during working memory Trends inCognitive Sciences 7 415ndash423

Curtis C E Rao V Y amp DrsquoEsposito M (2004) Maintenanceof spatial and motor codes during oculomotor delayedresponse tasks The Journal of Neuroscience 243944ndash3952

DrsquoEsposito M Ballard D Zarahn E amp Aguirre G K(2000) The role of prefrontal cortex in sensory memoryand motor preparation An event-related fMRI studyNeuroimage 11 400ndash408

DrsquoEsposito M amp Postle B R (1999) The dependence of spanand delayed-response performance on prefrontal cortexNeuropsychologia 37 1303ndash1315

DrsquoEsposito M Postle B R Ballard D amp Lease J (1999)Maintenance versus manipulation of information held inworking memory An event-related fMRI study Brain ampCognition 41 66ndash86

DrsquoEsposito M Postle B R Jonides J amp Smith E E(1999) The neural substrate and temporal dynamicsof interference effects in working memory as revealedby event-related functional MRI Proceedings of theNational Academy of Sciences USA 96 7514ndash7519

de Fockert J W Rees G Frith C D amp Lavie N (2001) Therole of working memory in visual attention Science 2911803ndash1806

Druzgal T J amp DrsquoEsposito M (2001) A neural networkref lecting decisions about human faces Neuron 32947ndash955

Duncan J amp Miller E K (2002) Cognitive focus throughadaptive neural coding in the primate prefrontal cortexIn D Stuss amp R Knight (Eds) Principles of frontal lobefunction (pp 278ndash291) Oxford Oxford University Press

Friston K J Holmes A P Poline J-B Heather J D ampFrackowiak R S J (1995) Analysis of fMRI time-seriesrevisited Neuroimage 2 45ndash53

Fukushima T Hasegawa I amp Miyashita Y (2004)Prefrontal neuronal activity encodes spatial targetrepresentations sequentially updated after nonspatialtarget-shift cues Journal of Neurophysiology 911367ndash1380

Fuster J M (1995) Memory in the cerebral cortexCambridge MIT Press

Gitelman D R (2002) ILAB A program for post experimentaleye movement analysis Behavioral Research MethodsInstruments and Computers 34 605ndash612

1688 Journal of Cognitive Neuroscience Volume 17 Number 11

type F(114) = 38 p = 07 and an interaction F(114) =49 p lt 05

One alternative to the sensory gating interpretation ofthese results might be that PFC delay-period activity onmemory-presentdistraction-present trials reflected dualtaskingmdashthe STR of target face information plus theincidental encoding (or some other type of processing)of the distracting faces This seems unlikely howeverbecause PFC delay-period activity was numerically loweron memory-presentdistraction-present trials than onmemory-presentdistraction-absent trials [although thiseffect did not achieve significance t(15) = 13 nsFigure 4A]

How specific was this pattern of results To assess thisquestion we applied the same contrasts to voxels in thedlPFC and IOTC that were selected for a differentpropertymdashresponsivity to visually presented faces Spe-cifically we identified voxels in the dlPFC and IOTC thatresponded to the Target epoch of the task (and re-moved from these VOIs any voxels that were also activeduring the delay period of memory-presentdistraction-absent trials) In this way we identified voxels from thesame regions in which we tested the sensory gatinghypothesis but whose function was different (presum-ably related to visual perception andor memory encod-ing but not to retention or another delay-periodprocess) For the raw data from these target-evokedVOIs (Figure 4C) as compared to the memory-evokedVOIs there was again a strong main effect of regionF(114) = 334 p lt 0001 a weaker borderline effect oftrial type F(114) = 33 p = 09 and now only a trendtoward an interaction F(114) = 40 p 07 When datafrom the target-evoked VOIs were normalized in thesame manner as had been the data from the delay-evoked VOIs (Figure 4D) their analysis with ANOVArevealed only a borderline effect of VOI F(114) = 40p = 06 for the first time a reliable effect of trial typeF(114) = 51 p lt 05 and again a trend toward aninteraction F(114) = 40 p = 06

The weakening of the Region Trial type interactionin target-evoked VOIs as compared with delay-evokedVOIs could be consistent with the claim that this inter-action predicted to be the fMRI signature of a sensorygating mechanism was selectively seen only in the delay-evoked VOIs It could also however simply reflect asituation in which the interaction terms for the twotypes of VOI fell on either side of the 05 thresholdbut did not differ from each other To distinguishbetween these two possibilities we compared the ef-fects of the experimental manipulation on delay-periodactivity across the two types of VOIs with an omnibusANOVA that directly compared the effect of memory onnormalized distraction-evoked activity across target-evoked and delay-evoked VOIs The factors for theomnibus ANOVA were region (dlPFC IOTC) trial type(memory present memory absent) and epoch definingthe VOIs (target delay) A three-way interaction would

be necessary to support the claim that the differentialeffect of memory demands on distraction-evoked activityin the dlPFC and IOTC was greater for delay-evoked thanfor target-evoked VOIs The ANOVA revealed a maineffect of region F(114) = 49 p lt 05 a borderlineeffect of trial type F(114) = 43 p = 06 and no effect ofepoch F(114) = 28 p gt 1 no interaction of Epoch Trial type F(114) = 28 p gt 1 but reliable inter-actions of Epoch Region F(114) = 47 p lt 05and of Region Trial type (F(114) = 49 p lt 05 and areliable interaction of Epoch Region Trial typeF(114) = 47 p lt 05 Thus it confirmed the specificityof the sensory gating effect to delay-evoked VOIs

DISCUSSION

The results demonstrate that delay-sensitive voxels intwo brain regions the dlPFC and the IOTC responddifferentially to the RSVP of faces conditioned on thebehavioral context in which these faces are presentedThe dlPFC response was 139 greater during concur-rent face memory whereas the IOTC response was 75lower This pattern was also seen to be selective becausea different group of dlPFC voxels (those that are targetsensitive) increased its activity by only 38 and theanalogously different group of IOTC voxels increased itsactivity by 3 These results are therefore consistentwith a model in which one function of dlPFC voxels thatare active during unfilled delay periods of delayed-recognition tasks is to perform a sensory gating functionBy this account the detection of potentially interferingstimuli in the environment triggers activity in networksof dlPFC neurons3 that in turn has the effect of sup-pressing perhaps via thalamic relays the afferent flow ofsensory information (Knight Staines Swick amp Chao1999) This would have beneficial effects for the signal-to-noise ratio in the networks of posterior neurons thatare responsible for the STR of information Thus a PFC-based control process may serve to protect the contentsof working and short-term memory from the deleteriouseffects external interference although it does not sup-port STR operations themselves

This account makes two broader claims about theworking memory literature First it offers an explanationas to why primates with PFC lesions can perform nor-mally on tasks of working and short-term memory whenpotential sources of distraction and inference are mini-mal but are impaired on the same tasks when distractionandor interference is high (Thompson-Schill et al2002 DrsquoEsposito amp Postle 1999 Chao amp Knight 1998Chao amp Knight 1995 Milner 1964 Malmo 1942)Second it proposes that sensory gating may accountfor a portion of the unfilled delay-period activity that isroutinely observed in PFC across a variety of neuro-imaging studies including those that do not explicitlyinclude distraction In relation to this second claim how-

1684 Journal of Cognitive Neuroscience Volume 17 Number 11

ever we cannot rule out the possibility that the dlPFCactivity observed during memory-presentdistraction-absent trials was due to a specific feature of our designBecause subjects could not predict which memorytrials would also contain distraction an effective strat-egy may have been to implement a sensory gatingstrategy on every memory trial To rule this out wouldrequire a two stage process of first scanning lsquolsquonaıversquorsquosubjects performing several blocks of only memory-presentdistraction-absent trials then introducing thefactor of delay-period distraction for a subsequent setof scans

The present study fits within a growing body of resultsthat document control rather than STR functions forthe PFC These include the monitoring and transforma-tion of information held in working memory (Bor et al2003 DrsquoEsposito Postle Ballard amp Lease 1999 Pet-rides 1995) and the attentional selection of information(Lebedev et al 2004 Sakai et al 2002 de Fockert ReesFrith amp Lavie 2001) These approaches which proposewell-defined testable mechanisms represent an advancebeyond the vague notion of lsquolsquomaintenancersquorsquo that is ofteninvoked in interpretations of PFC activity The presentstudy suggests one such alternative to lsquolsquomaintenancersquorsquo toaccount for sustained PFC activity observed during thesimplest tests of STR of information

METHODS

Subjects

Sixteen healthy young adults (mean age = 224 13 men)participated in this study All provided informed consentand were paid in compensation for participating in thestudy

Behavioral Procedure

The 2 2 factorial design featured the factors ofMemory (present absent) and Distraction (presentabsent) Each trial had the same structure Target (1 sec)followed by Delay (7 sec) followed by Probe (1 sec)followed by ITI (13 sec) During memory-present trialstarget and probe stimuli were individual gray-scale malefaces (none wearing glasses) appearing on white back-grounds and subjects indicated with a button presswhether the latter matched the former ( p = 5) Memory-absent trials were procedurally identical except thatthey featured the same recurring female facesmdashfacelsquolsquoArsquorsquo always the target which denoted a lsquolsquono-memoryrsquorsquotrial face lsquolsquoBrsquorsquo always the probe On memory-absenttrials subjects were cued by the probe to indicate witha button press whether or not distracting faces appearedduring the delay period ( p = 5) thereby matching themotor response demands of memory-present trialsDistraction-present trials featured RSVP (3 Hz) of thefaces of males wearing glasses (to differentiate distrac-

tion items from memoranda) during the two middleseconds of the delay (Figure 1) Subjects were instructedto look at the distracting faces during all trials on whichthey appeared and to fixate centrally on distraction-absent trials Stimuli were presented via eyepiecesmounted on the head coil (Avotec Visible Eye StuartFL) and fitted with an infrared-based eye tracking system(SMI Inc Teltow Germany) Fixation was monitoredduring scanning and analyzed off-line with iLab software(Gitelman 2002) Trials during which subjects failedto fixate centrally from time 3 sec to time 5 sec werediscarded from subsequent analyses (Figure 5) Thiscriterion ensured that any evidence for differencesin delay-period activity between conditions could notbe attributed to theoretically uninteresting differencesin for example the amount of visual stimulation pro-duced by the distracting stimuli or the amount ofoculomotor activity Trials of each of the four types wereequiprobable and presented in a randomly determinedorder

fMRI Data Acquisition

Whole-brain images were acquired with a 3T scanner(GE Signa VHI Waukesha WI) High-resolution T1-weighted images (32 axial slices 09375 09375 4 mm)were obtained in every participant and a gradient-echoecho-planar sequence (TR = 2000 msec TE = 50 msec)was used to acquire data sensitive to the blood oxygenlevel dependent (BOLD) signal (Kwong et al 1992Ogawa et al 1992) within a 64 64 matrix (32 axialslices coplanar with T1 375 375 4 mm) Scans ofthe delayed-recognition task were preceded by a scan inwhich we derived an estimate of the hemodynamicresponse function (HRF) for each participant Duringthis scan each participant performed a simple RT taskthat required a bimanual button press once every 20 secin response to a brief change in shape of the fixationstimulus A partial F test associated with a Fourier basiscovariate set ( Josephs Turner amp Friston 1997) wasused to evaluate the significance of task-correlatedactivity in each voxel of primary somatosensory andmotor cortical regions of interest (Aguirre Zarahn ampDrsquoEsposito 1998) An HRF estimate was extracted fromthe suprathreshold voxels of these ROIs by spatiallyaveraging their time series filtering the resultant aver-aged fMRI time series to remove high (gt0244 Hz)and low (lt005 Hz) frequencies adjusting it to removethe effects of nuisance covariates (Friston HolmesPoline Heather amp Frackowiak 1995) and trial aver-aging The HRF characterizes the fMRI response result-ing from a brief impulse of neural activity (BoyntonEngel Glover amp Heeger 1996) and can vary markedlyacross participants (Aguirre et al 1998) The subject-specific HRFs were used to convolve independent vari-ables entered into the modified general linear model(GLM Worsley amp Friston 1995) that we used to analyze

Postle 1685

the data from the scans of the delayed-recognition task4

The eight scans of the delayed-recognition task eachlasted 6 min 12 sec (5 min 52 sec of task preceded by20 sec of dummy pulses to achieve a steady state of tis-sue magnetization)

fMRI Data Processing

The fMRI time series data were filtered and adjusted asdescribed previously (Postle Zarahn amp DrsquoEsposito2000) The principle of the fMRI time series analysiswas to model the fMRI signal changes evoked byeach stimulus presentation epoch with covariates com-posed of BOLD impulse response functions shiftedalong the time line of the task to represent various trialepochs (Figure 1 Postle et al 2000 Zarahn Aguirre

amp DrsquoEsposito 1997) The least-squares solution of theGLM of the fMRI time series data yielded parameterestimates that were associated with each covariateof interest The smoothness of the fMRI response toneural activity allows fMRI-evoked responses that arisefrom temporally dependent events to be resolved onthe order of 4 sec (That is if the covariates modelingthese events are spaced at 4-sec intervals the covariatepositioned at time 0 sec will account for virtually all ofthe variance attributable to an event starting at time 0 secand of duration of 1 sec or less thereby leaving thecovariate positioned at time 4 sec relatively uncontam-inated by variance attributable to the time 0 sec eventsee Figure 6 of Zarahn et al 1997) Differences in fMRIsignal (either between conditions or vs baseline) weretested by computing t statistics from contrasts between

Figure 5 Examples of eye-position data from four representative trials For all four examples trial epoch is color coded red = target stimulus

green = early delay blue = distraction portion of the delay period fuchsia = late delay black = probe stimulus plus first second of ITI (A) A

memory-presentdistraction-present trial with acceptable fixation during distraction The top panel is a 2-D representation of eye position on the

display lsquolsquoscreenrsquorsquo over the course of the trial axis labels correspond to degrees of visual angle Bottom panels display horizontal (x) and vertical ( y)eye position as a function of time with time in seconds on the horizontal axes and position in pixels on the vertical axes (each pixel measured

058 mm2) (B) A memory-absentdistraction-present trial with acceptable fixation during distraction All conventions same as in (A) Interrupted

portion of late delay corresponds to a blink when eyes were closed (C) A memory-presentdistraction-present trial that was discarded due

to unacceptable fixation during distraction All conventions same as (A) (D) A memory-absentdistraction-present trial that was discarded dueto unacceptable fixation during distraction All conventions same as in (A)

1686 Journal of Cognitive Neuroscience Volume 17 Number 11

parameter estimates associated with the covariates inquestion

Our analysis strategy was to identify delay-specificactivity in two regions the dlPFC and the IOTC5 frommemory-presentdistraction-absent trials This wouldidentify delay-period activity comparable to that ob-served in virtually all whole-brain neuroimaging studiesof similar tasks (eg with face stimuli Postle et al 2003Ranganath amp DrsquoEsposito 2001 Druzgal amp DrsquoEsposito2001 Courtney Ungerleider Keil amp Haxby 1997 Court-ney Ungerleider Keil amp Haxby 1996) Masks cor-responding to these two regions were drawn on eachsubjectrsquos anatomical scans Voxels located within thesemasks which became the VOIs were identified on tmaps of delay-period activity from memory-presentdis-traction-absent trials thresholded to a regionwise a of05 (one-tailed) with Bonferroni correction for the num-ber of voxels in the mask Note that with this approachwith spatially unsmoothed data individual voxels whoset value exceeds threshold can be interpreted as lsquolsquomean-ingfulrsquorsquo data (Postle et al 2000) Thus this procedurediffered somewhat from other applications of lsquolsquoregion-

of-interest (ROI) analysesrsquorsquo in that no selection criteriawere imposed with respect to cluster size or proximity ofcritical voxels This was done among other reasonsbecause we did not know a priori in which subregionof PFC to expect to find sensory-gating-related activityWithin the VOIs we evaluated the responses evokedby RSVP of faces during memory-presentdistraction-present trials versus memory-absentdistraction-presenttrials (Memory-absentdistraction-absent trials do notfigure directly into the test of the sensory gating hy-pothesis but were necessary to prevent anticipation ofdistraction on memory-absent trials)

Acknowledgments

The author thanks Olufunsho Faseyitan and Craig Rypstat forprogramming assistance and OF and Christopher Jordan forhelp with data collection and analysis This research was sup-ported by the National Institutes of Health RO1 MH064498

Reprint requests should be sent to Bradley R Postle Depart-mentofPsychologyUniversityof WisconsinndashMadison1202WestJohnson St Madison WI 53706 or via e-mail postlewiscedu

Figure 5 (continued)

Postle 1687

The data reported in this experiment have been depositedwith the fMRI Data Center archive (wwwfmridcorg) Theaccession number is 2-2005-119B9

Notes

1 This may be an empirical example of the lsquolsquoactive main-tenancersquorsquo proposed by Miller and Cohen (2001)2 The raw value of this subjectrsquos memory-absentdistraction-present effect in the dlPFC was 51 and so dividing thememory-presentdistraction-present by this value inflated it bya factor of almost 2 the opposite direction than that effectedby the normalization procedure in all other subjects exceptone whose raw memory-absentdistraction-present effect inthe dlPFC was 983 The fact that delay-period activity in the dlPFC VOIs wasnumerically lower during memory-presentdistraction-presentthan during memory-presentdistraction-absent trials is con-sistent with the idea that the sensory gating mechanism oncetriggered by neurons that are tonically active during unfilleddelay periods is implemented by a broader PFC network thatincludes regions that are not typically active during unfilleddelay periods Although the data are not shown here it iscertainly the case that in every subject considerably more dlPFCvoxels were active during the memory-presentdistraction-present delay than during the memory-presentdistraction-absent delay4 The HRF also varies in shape across brain regions with-in an individual participant (eg Miezen Maccotta OllingerPetersen amp Buckner 2000) However recent work indi-cates that the magnitude of variability in the shape of theHRF is greater across individuals within a region than it isacross regions within an individual (Handwerker Ollinger ampDrsquoEsposito 2004) Additionally it indicates that employinga participant-specific HRF can improve the sensitivity andmagnitude estimates of the least-squares solution of the GLMover comparable analyses that employ a generic HRF model(Handwerker et al 2004)5 The broader search space of the IOTC was used instead ofa more circumscribed area such as the fusiform face areabecause as can be seen in Figure 2 there was a considerableamount of topological variability of memory-presentdistraction-absent delay-period activity in the posterior ventral stream andno one area was activated in all subjects

REFERENCES

Aguirre G K Zarahn E amp DrsquoEsposito M (1998) Thevariability of human BOLD hemodynamic responsesNeuroimage 8 360ndash369

Baddeley A D amp Hitch G J (1974) Working memoryIn G H Bower (Ed) The psychology of learning andmotivation (Vol 8 pp 47ndash89) New York Academic Press

Bor D Duncan J Wiseman R J amp Owen A M (2003)Encoding strategies dissociate prefrontal activity fromworking memory demand Neuron 37 361ndash367

Boynton G M Engel S A Glover G H amp HeegerD J (1996) Linear systems analysis of functionalmagnetic resonance imaging in human V1 The Journalof Neuroscience 16 4207ndash4221

Braver T S Gray J R amp Burgess G C (in press)Explaining the many varieties of working memoryvariation Dual mechanisms of cognitive control In AConway C Jarrold M Kane A Miyake amp J Towse(Eds) Variation in working memory Oxford OxfordUniversity Press

Chao L amp Knight R (1998) Contribution of humanprefrontal cortex to delay performance Journal ofCognitive Neuroscience 10 167ndash177

Chao L L amp Knight R T (1995) Human prefrontal lesionsincrease distractibility to irrelevant sensory inputsNeuroReport 6 1605ndash1610

Connolly J D Goodale M A Goltz H C amp Munoz D P(2005) fMRI activation in the human frontal eye field iscorrelated with saccadic reaction time Journal ofNeurophysiology 94 605ndash611

Courtney S M (2004) Attention and cognitive control asemergent properties of information representation inworking memory Cognitive Affective amp BehavioralNeuroscience 4 501ndash516

Courtney S M Ungerleider L G Keil K amp Haxby J(1996) Object and spatial visual working memory activateseparate neural systems in human cortex Cerebral Cortex6 39ndash49

Courtney S M Ungerleider L G Keil K amp Haxby J V(1997) Transient and sustained activity in a distributedneural system for human working memory Nature 386608ndash611

Curtis C E amp DrsquoEsposito M (2003) Persistent activity inthe prefrontal cortex during working memory Trends inCognitive Sciences 7 415ndash423

Curtis C E Rao V Y amp DrsquoEsposito M (2004) Maintenanceof spatial and motor codes during oculomotor delayedresponse tasks The Journal of Neuroscience 243944ndash3952

DrsquoEsposito M Ballard D Zarahn E amp Aguirre G K(2000) The role of prefrontal cortex in sensory memoryand motor preparation An event-related fMRI studyNeuroimage 11 400ndash408

DrsquoEsposito M amp Postle B R (1999) The dependence of spanand delayed-response performance on prefrontal cortexNeuropsychologia 37 1303ndash1315

DrsquoEsposito M Postle B R Ballard D amp Lease J (1999)Maintenance versus manipulation of information held inworking memory An event-related fMRI study Brain ampCognition 41 66ndash86

DrsquoEsposito M Postle B R Jonides J amp Smith E E(1999) The neural substrate and temporal dynamicsof interference effects in working memory as revealedby event-related functional MRI Proceedings of theNational Academy of Sciences USA 96 7514ndash7519

de Fockert J W Rees G Frith C D amp Lavie N (2001) Therole of working memory in visual attention Science 2911803ndash1806

Druzgal T J amp DrsquoEsposito M (2001) A neural networkref lecting decisions about human faces Neuron 32947ndash955

Duncan J amp Miller E K (2002) Cognitive focus throughadaptive neural coding in the primate prefrontal cortexIn D Stuss amp R Knight (Eds) Principles of frontal lobefunction (pp 278ndash291) Oxford Oxford University Press

Friston K J Holmes A P Poline J-B Heather J D ampFrackowiak R S J (1995) Analysis of fMRI time-seriesrevisited Neuroimage 2 45ndash53

Fukushima T Hasegawa I amp Miyashita Y (2004)Prefrontal neuronal activity encodes spatial targetrepresentations sequentially updated after nonspatialtarget-shift cues Journal of Neurophysiology 911367ndash1380

Fuster J M (1995) Memory in the cerebral cortexCambridge MIT Press

Gitelman D R (2002) ILAB A program for post experimentaleye movement analysis Behavioral Research MethodsInstruments and Computers 34 605ndash612

1688 Journal of Cognitive Neuroscience Volume 17 Number 11

ever we cannot rule out the possibility that the dlPFCactivity observed during memory-presentdistraction-absent trials was due to a specific feature of our designBecause subjects could not predict which memorytrials would also contain distraction an effective strat-egy may have been to implement a sensory gatingstrategy on every memory trial To rule this out wouldrequire a two stage process of first scanning lsquolsquonaıversquorsquosubjects performing several blocks of only memory-presentdistraction-absent trials then introducing thefactor of delay-period distraction for a subsequent setof scans

The present study fits within a growing body of resultsthat document control rather than STR functions forthe PFC These include the monitoring and transforma-tion of information held in working memory (Bor et al2003 DrsquoEsposito Postle Ballard amp Lease 1999 Pet-rides 1995) and the attentional selection of information(Lebedev et al 2004 Sakai et al 2002 de Fockert ReesFrith amp Lavie 2001) These approaches which proposewell-defined testable mechanisms represent an advancebeyond the vague notion of lsquolsquomaintenancersquorsquo that is ofteninvoked in interpretations of PFC activity The presentstudy suggests one such alternative to lsquolsquomaintenancersquorsquo toaccount for sustained PFC activity observed during thesimplest tests of STR of information

METHODS

Subjects

Sixteen healthy young adults (mean age = 224 13 men)participated in this study All provided informed consentand were paid in compensation for participating in thestudy

Behavioral Procedure

The 2 2 factorial design featured the factors ofMemory (present absent) and Distraction (presentabsent) Each trial had the same structure Target (1 sec)followed by Delay (7 sec) followed by Probe (1 sec)followed by ITI (13 sec) During memory-present trialstarget and probe stimuli were individual gray-scale malefaces (none wearing glasses) appearing on white back-grounds and subjects indicated with a button presswhether the latter matched the former ( p = 5) Memory-absent trials were procedurally identical except thatthey featured the same recurring female facesmdashfacelsquolsquoArsquorsquo always the target which denoted a lsquolsquono-memoryrsquorsquotrial face lsquolsquoBrsquorsquo always the probe On memory-absenttrials subjects were cued by the probe to indicate witha button press whether or not distracting faces appearedduring the delay period ( p = 5) thereby matching themotor response demands of memory-present trialsDistraction-present trials featured RSVP (3 Hz) of thefaces of males wearing glasses (to differentiate distrac-

tion items from memoranda) during the two middleseconds of the delay (Figure 1) Subjects were instructedto look at the distracting faces during all trials on whichthey appeared and to fixate centrally on distraction-absent trials Stimuli were presented via eyepiecesmounted on the head coil (Avotec Visible Eye StuartFL) and fitted with an infrared-based eye tracking system(SMI Inc Teltow Germany) Fixation was monitoredduring scanning and analyzed off-line with iLab software(Gitelman 2002) Trials during which subjects failedto fixate centrally from time 3 sec to time 5 sec werediscarded from subsequent analyses (Figure 5) Thiscriterion ensured that any evidence for differencesin delay-period activity between conditions could notbe attributed to theoretically uninteresting differencesin for example the amount of visual stimulation pro-duced by the distracting stimuli or the amount ofoculomotor activity Trials of each of the four types wereequiprobable and presented in a randomly determinedorder

fMRI Data Acquisition

Whole-brain images were acquired with a 3T scanner(GE Signa VHI Waukesha WI) High-resolution T1-weighted images (32 axial slices 09375 09375 4 mm)were obtained in every participant and a gradient-echoecho-planar sequence (TR = 2000 msec TE = 50 msec)was used to acquire data sensitive to the blood oxygenlevel dependent (BOLD) signal (Kwong et al 1992Ogawa et al 1992) within a 64 64 matrix (32 axialslices coplanar with T1 375 375 4 mm) Scans ofthe delayed-recognition task were preceded by a scan inwhich we derived an estimate of the hemodynamicresponse function (HRF) for each participant Duringthis scan each participant performed a simple RT taskthat required a bimanual button press once every 20 secin response to a brief change in shape of the fixationstimulus A partial F test associated with a Fourier basiscovariate set ( Josephs Turner amp Friston 1997) wasused to evaluate the significance of task-correlatedactivity in each voxel of primary somatosensory andmotor cortical regions of interest (Aguirre Zarahn ampDrsquoEsposito 1998) An HRF estimate was extracted fromthe suprathreshold voxels of these ROIs by spatiallyaveraging their time series filtering the resultant aver-aged fMRI time series to remove high (gt0244 Hz)and low (lt005 Hz) frequencies adjusting it to removethe effects of nuisance covariates (Friston HolmesPoline Heather amp Frackowiak 1995) and trial aver-aging The HRF characterizes the fMRI response result-ing from a brief impulse of neural activity (BoyntonEngel Glover amp Heeger 1996) and can vary markedlyacross participants (Aguirre et al 1998) The subject-specific HRFs were used to convolve independent vari-ables entered into the modified general linear model(GLM Worsley amp Friston 1995) that we used to analyze

Postle 1685

the data from the scans of the delayed-recognition task4

The eight scans of the delayed-recognition task eachlasted 6 min 12 sec (5 min 52 sec of task preceded by20 sec of dummy pulses to achieve a steady state of tis-sue magnetization)

fMRI Data Processing

The fMRI time series data were filtered and adjusted asdescribed previously (Postle Zarahn amp DrsquoEsposito2000) The principle of the fMRI time series analysiswas to model the fMRI signal changes evoked byeach stimulus presentation epoch with covariates com-posed of BOLD impulse response functions shiftedalong the time line of the task to represent various trialepochs (Figure 1 Postle et al 2000 Zarahn Aguirre

amp DrsquoEsposito 1997) The least-squares solution of theGLM of the fMRI time series data yielded parameterestimates that were associated with each covariateof interest The smoothness of the fMRI response toneural activity allows fMRI-evoked responses that arisefrom temporally dependent events to be resolved onthe order of 4 sec (That is if the covariates modelingthese events are spaced at 4-sec intervals the covariatepositioned at time 0 sec will account for virtually all ofthe variance attributable to an event starting at time 0 secand of duration of 1 sec or less thereby leaving thecovariate positioned at time 4 sec relatively uncontam-inated by variance attributable to the time 0 sec eventsee Figure 6 of Zarahn et al 1997) Differences in fMRIsignal (either between conditions or vs baseline) weretested by computing t statistics from contrasts between

Figure 5 Examples of eye-position data from four representative trials For all four examples trial epoch is color coded red = target stimulus

green = early delay blue = distraction portion of the delay period fuchsia = late delay black = probe stimulus plus first second of ITI (A) A

memory-presentdistraction-present trial with acceptable fixation during distraction The top panel is a 2-D representation of eye position on the

display lsquolsquoscreenrsquorsquo over the course of the trial axis labels correspond to degrees of visual angle Bottom panels display horizontal (x) and vertical ( y)eye position as a function of time with time in seconds on the horizontal axes and position in pixels on the vertical axes (each pixel measured

058 mm2) (B) A memory-absentdistraction-present trial with acceptable fixation during distraction All conventions same as in (A) Interrupted

portion of late delay corresponds to a blink when eyes were closed (C) A memory-presentdistraction-present trial that was discarded due

to unacceptable fixation during distraction All conventions same as (A) (D) A memory-absentdistraction-present trial that was discarded dueto unacceptable fixation during distraction All conventions same as in (A)

1686 Journal of Cognitive Neuroscience Volume 17 Number 11

parameter estimates associated with the covariates inquestion

Our analysis strategy was to identify delay-specificactivity in two regions the dlPFC and the IOTC5 frommemory-presentdistraction-absent trials This wouldidentify delay-period activity comparable to that ob-served in virtually all whole-brain neuroimaging studiesof similar tasks (eg with face stimuli Postle et al 2003Ranganath amp DrsquoEsposito 2001 Druzgal amp DrsquoEsposito2001 Courtney Ungerleider Keil amp Haxby 1997 Court-ney Ungerleider Keil amp Haxby 1996) Masks cor-responding to these two regions were drawn on eachsubjectrsquos anatomical scans Voxels located within thesemasks which became the VOIs were identified on tmaps of delay-period activity from memory-presentdis-traction-absent trials thresholded to a regionwise a of05 (one-tailed) with Bonferroni correction for the num-ber of voxels in the mask Note that with this approachwith spatially unsmoothed data individual voxels whoset value exceeds threshold can be interpreted as lsquolsquomean-ingfulrsquorsquo data (Postle et al 2000) Thus this procedurediffered somewhat from other applications of lsquolsquoregion-

of-interest (ROI) analysesrsquorsquo in that no selection criteriawere imposed with respect to cluster size or proximity ofcritical voxels This was done among other reasonsbecause we did not know a priori in which subregionof PFC to expect to find sensory-gating-related activityWithin the VOIs we evaluated the responses evokedby RSVP of faces during memory-presentdistraction-present trials versus memory-absentdistraction-presenttrials (Memory-absentdistraction-absent trials do notfigure directly into the test of the sensory gating hy-pothesis but were necessary to prevent anticipation ofdistraction on memory-absent trials)

Acknowledgments

The author thanks Olufunsho Faseyitan and Craig Rypstat forprogramming assistance and OF and Christopher Jordan forhelp with data collection and analysis This research was sup-ported by the National Institutes of Health RO1 MH064498

Reprint requests should be sent to Bradley R Postle Depart-mentofPsychologyUniversityof WisconsinndashMadison1202WestJohnson St Madison WI 53706 or via e-mail postlewiscedu

Figure 5 (continued)

Postle 1687

The data reported in this experiment have been depositedwith the fMRI Data Center archive (wwwfmridcorg) Theaccession number is 2-2005-119B9

Notes

1 This may be an empirical example of the lsquolsquoactive main-tenancersquorsquo proposed by Miller and Cohen (2001)2 The raw value of this subjectrsquos memory-absentdistraction-present effect in the dlPFC was 51 and so dividing thememory-presentdistraction-present by this value inflated it bya factor of almost 2 the opposite direction than that effectedby the normalization procedure in all other subjects exceptone whose raw memory-absentdistraction-present effect inthe dlPFC was 983 The fact that delay-period activity in the dlPFC VOIs wasnumerically lower during memory-presentdistraction-presentthan during memory-presentdistraction-absent trials is con-sistent with the idea that the sensory gating mechanism oncetriggered by neurons that are tonically active during unfilleddelay periods is implemented by a broader PFC network thatincludes regions that are not typically active during unfilleddelay periods Although the data are not shown here it iscertainly the case that in every subject considerably more dlPFCvoxels were active during the memory-presentdistraction-present delay than during the memory-presentdistraction-absent delay4 The HRF also varies in shape across brain regions with-in an individual participant (eg Miezen Maccotta OllingerPetersen amp Buckner 2000) However recent work indi-cates that the magnitude of variability in the shape of theHRF is greater across individuals within a region than it isacross regions within an individual (Handwerker Ollinger ampDrsquoEsposito 2004) Additionally it indicates that employinga participant-specific HRF can improve the sensitivity andmagnitude estimates of the least-squares solution of the GLMover comparable analyses that employ a generic HRF model(Handwerker et al 2004)5 The broader search space of the IOTC was used instead ofa more circumscribed area such as the fusiform face areabecause as can be seen in Figure 2 there was a considerableamount of topological variability of memory-presentdistraction-absent delay-period activity in the posterior ventral stream andno one area was activated in all subjects

REFERENCES

Aguirre G K Zarahn E amp DrsquoEsposito M (1998) Thevariability of human BOLD hemodynamic responsesNeuroimage 8 360ndash369

Baddeley A D amp Hitch G J (1974) Working memoryIn G H Bower (Ed) The psychology of learning andmotivation (Vol 8 pp 47ndash89) New York Academic Press

Bor D Duncan J Wiseman R J amp Owen A M (2003)Encoding strategies dissociate prefrontal activity fromworking memory demand Neuron 37 361ndash367

Boynton G M Engel S A Glover G H amp HeegerD J (1996) Linear systems analysis of functionalmagnetic resonance imaging in human V1 The Journalof Neuroscience 16 4207ndash4221

Braver T S Gray J R amp Burgess G C (in press)Explaining the many varieties of working memoryvariation Dual mechanisms of cognitive control In AConway C Jarrold M Kane A Miyake amp J Towse(Eds) Variation in working memory Oxford OxfordUniversity Press

Chao L amp Knight R (1998) Contribution of humanprefrontal cortex to delay performance Journal ofCognitive Neuroscience 10 167ndash177

Chao L L amp Knight R T (1995) Human prefrontal lesionsincrease distractibility to irrelevant sensory inputsNeuroReport 6 1605ndash1610

Connolly J D Goodale M A Goltz H C amp Munoz D P(2005) fMRI activation in the human frontal eye field iscorrelated with saccadic reaction time Journal ofNeurophysiology 94 605ndash611

Courtney S M (2004) Attention and cognitive control asemergent properties of information representation inworking memory Cognitive Affective amp BehavioralNeuroscience 4 501ndash516

Courtney S M Ungerleider L G Keil K amp Haxby J(1996) Object and spatial visual working memory activateseparate neural systems in human cortex Cerebral Cortex6 39ndash49

Courtney S M Ungerleider L G Keil K amp Haxby J V(1997) Transient and sustained activity in a distributedneural system for human working memory Nature 386608ndash611

Curtis C E amp DrsquoEsposito M (2003) Persistent activity inthe prefrontal cortex during working memory Trends inCognitive Sciences 7 415ndash423

Curtis C E Rao V Y amp DrsquoEsposito M (2004) Maintenanceof spatial and motor codes during oculomotor delayedresponse tasks The Journal of Neuroscience 243944ndash3952

DrsquoEsposito M Ballard D Zarahn E amp Aguirre G K(2000) The role of prefrontal cortex in sensory memoryand motor preparation An event-related fMRI studyNeuroimage 11 400ndash408

DrsquoEsposito M amp Postle B R (1999) The dependence of spanand delayed-response performance on prefrontal cortexNeuropsychologia 37 1303ndash1315

DrsquoEsposito M Postle B R Ballard D amp Lease J (1999)Maintenance versus manipulation of information held inworking memory An event-related fMRI study Brain ampCognition 41 66ndash86

DrsquoEsposito M Postle B R Jonides J amp Smith E E(1999) The neural substrate and temporal dynamicsof interference effects in working memory as revealedby event-related functional MRI Proceedings of theNational Academy of Sciences USA 96 7514ndash7519

de Fockert J W Rees G Frith C D amp Lavie N (2001) Therole of working memory in visual attention Science 2911803ndash1806

Druzgal T J amp DrsquoEsposito M (2001) A neural networkref lecting decisions about human faces Neuron 32947ndash955

Duncan J amp Miller E K (2002) Cognitive focus throughadaptive neural coding in the primate prefrontal cortexIn D Stuss amp R Knight (Eds) Principles of frontal lobefunction (pp 278ndash291) Oxford Oxford University Press

Friston K J Holmes A P Poline J-B Heather J D ampFrackowiak R S J (1995) Analysis of fMRI time-seriesrevisited Neuroimage 2 45ndash53

Fukushima T Hasegawa I amp Miyashita Y (2004)Prefrontal neuronal activity encodes spatial targetrepresentations sequentially updated after nonspatialtarget-shift cues Journal of Neurophysiology 911367ndash1380

Fuster J M (1995) Memory in the cerebral cortexCambridge MIT Press

Gitelman D R (2002) ILAB A program for post experimentaleye movement analysis Behavioral Research MethodsInstruments and Computers 34 605ndash612

1688 Journal of Cognitive Neuroscience Volume 17 Number 11

the data from the scans of the delayed-recognition task4

The eight scans of the delayed-recognition task eachlasted 6 min 12 sec (5 min 52 sec of task preceded by20 sec of dummy pulses to achieve a steady state of tis-sue magnetization)

fMRI Data Processing

The fMRI time series data were filtered and adjusted asdescribed previously (Postle Zarahn amp DrsquoEsposito2000) The principle of the fMRI time series analysiswas to model the fMRI signal changes evoked byeach stimulus presentation epoch with covariates com-posed of BOLD impulse response functions shiftedalong the time line of the task to represent various trialepochs (Figure 1 Postle et al 2000 Zarahn Aguirre

amp DrsquoEsposito 1997) The least-squares solution of theGLM of the fMRI time series data yielded parameterestimates that were associated with each covariateof interest The smoothness of the fMRI response toneural activity allows fMRI-evoked responses that arisefrom temporally dependent events to be resolved onthe order of 4 sec (That is if the covariates modelingthese events are spaced at 4-sec intervals the covariatepositioned at time 0 sec will account for virtually all ofthe variance attributable to an event starting at time 0 secand of duration of 1 sec or less thereby leaving thecovariate positioned at time 4 sec relatively uncontam-inated by variance attributable to the time 0 sec eventsee Figure 6 of Zarahn et al 1997) Differences in fMRIsignal (either between conditions or vs baseline) weretested by computing t statistics from contrasts between

Figure 5 Examples of eye-position data from four representative trials For all four examples trial epoch is color coded red = target stimulus

green = early delay blue = distraction portion of the delay period fuchsia = late delay black = probe stimulus plus first second of ITI (A) A

memory-presentdistraction-present trial with acceptable fixation during distraction The top panel is a 2-D representation of eye position on the

display lsquolsquoscreenrsquorsquo over the course of the trial axis labels correspond to degrees of visual angle Bottom panels display horizontal (x) and vertical ( y)eye position as a function of time with time in seconds on the horizontal axes and position in pixels on the vertical axes (each pixel measured

058 mm2) (B) A memory-absentdistraction-present trial with acceptable fixation during distraction All conventions same as in (A) Interrupted

portion of late delay corresponds to a blink when eyes were closed (C) A memory-presentdistraction-present trial that was discarded due

to unacceptable fixation during distraction All conventions same as (A) (D) A memory-absentdistraction-present trial that was discarded dueto unacceptable fixation during distraction All conventions same as in (A)

1686 Journal of Cognitive Neuroscience Volume 17 Number 11

parameter estimates associated with the covariates inquestion

Our analysis strategy was to identify delay-specificactivity in two regions the dlPFC and the IOTC5 frommemory-presentdistraction-absent trials This wouldidentify delay-period activity comparable to that ob-served in virtually all whole-brain neuroimaging studiesof similar tasks (eg with face stimuli Postle et al 2003Ranganath amp DrsquoEsposito 2001 Druzgal amp DrsquoEsposito2001 Courtney Ungerleider Keil amp Haxby 1997 Court-ney Ungerleider Keil amp Haxby 1996) Masks cor-responding to these two regions were drawn on eachsubjectrsquos anatomical scans Voxels located within thesemasks which became the VOIs were identified on tmaps of delay-period activity from memory-presentdis-traction-absent trials thresholded to a regionwise a of05 (one-tailed) with Bonferroni correction for the num-ber of voxels in the mask Note that with this approachwith spatially unsmoothed data individual voxels whoset value exceeds threshold can be interpreted as lsquolsquomean-ingfulrsquorsquo data (Postle et al 2000) Thus this procedurediffered somewhat from other applications of lsquolsquoregion-

of-interest (ROI) analysesrsquorsquo in that no selection criteriawere imposed with respect to cluster size or proximity ofcritical voxels This was done among other reasonsbecause we did not know a priori in which subregionof PFC to expect to find sensory-gating-related activityWithin the VOIs we evaluated the responses evokedby RSVP of faces during memory-presentdistraction-present trials versus memory-absentdistraction-presenttrials (Memory-absentdistraction-absent trials do notfigure directly into the test of the sensory gating hy-pothesis but were necessary to prevent anticipation ofdistraction on memory-absent trials)

Acknowledgments

The author thanks Olufunsho Faseyitan and Craig Rypstat forprogramming assistance and OF and Christopher Jordan forhelp with data collection and analysis This research was sup-ported by the National Institutes of Health RO1 MH064498

Reprint requests should be sent to Bradley R Postle Depart-mentofPsychologyUniversityof WisconsinndashMadison1202WestJohnson St Madison WI 53706 or via e-mail postlewiscedu

Figure 5 (continued)

Postle 1687

The data reported in this experiment have been depositedwith the fMRI Data Center archive (wwwfmridcorg) Theaccession number is 2-2005-119B9

Notes

1 This may be an empirical example of the lsquolsquoactive main-tenancersquorsquo proposed by Miller and Cohen (2001)2 The raw value of this subjectrsquos memory-absentdistraction-present effect in the dlPFC was 51 and so dividing thememory-presentdistraction-present by this value inflated it bya factor of almost 2 the opposite direction than that effectedby the normalization procedure in all other subjects exceptone whose raw memory-absentdistraction-present effect inthe dlPFC was 983 The fact that delay-period activity in the dlPFC VOIs wasnumerically lower during memory-presentdistraction-presentthan during memory-presentdistraction-absent trials is con-sistent with the idea that the sensory gating mechanism oncetriggered by neurons that are tonically active during unfilleddelay periods is implemented by a broader PFC network thatincludes regions that are not typically active during unfilleddelay periods Although the data are not shown here it iscertainly the case that in every subject considerably more dlPFCvoxels were active during the memory-presentdistraction-present delay than during the memory-presentdistraction-absent delay4 The HRF also varies in shape across brain regions with-in an individual participant (eg Miezen Maccotta OllingerPetersen amp Buckner 2000) However recent work indi-cates that the magnitude of variability in the shape of theHRF is greater across individuals within a region than it isacross regions within an individual (Handwerker Ollinger ampDrsquoEsposito 2004) Additionally it indicates that employinga participant-specific HRF can improve the sensitivity andmagnitude estimates of the least-squares solution of the GLMover comparable analyses that employ a generic HRF model(Handwerker et al 2004)5 The broader search space of the IOTC was used instead ofa more circumscribed area such as the fusiform face areabecause as can be seen in Figure 2 there was a considerableamount of topological variability of memory-presentdistraction-absent delay-period activity in the posterior ventral stream andno one area was activated in all subjects

REFERENCES

Aguirre G K Zarahn E amp DrsquoEsposito M (1998) Thevariability of human BOLD hemodynamic responsesNeuroimage 8 360ndash369

Baddeley A D amp Hitch G J (1974) Working memoryIn G H Bower (Ed) The psychology of learning andmotivation (Vol 8 pp 47ndash89) New York Academic Press

Bor D Duncan J Wiseman R J amp Owen A M (2003)Encoding strategies dissociate prefrontal activity fromworking memory demand Neuron 37 361ndash367

Boynton G M Engel S A Glover G H amp HeegerD J (1996) Linear systems analysis of functionalmagnetic resonance imaging in human V1 The Journalof Neuroscience 16 4207ndash4221

Braver T S Gray J R amp Burgess G C (in press)Explaining the many varieties of working memoryvariation Dual mechanisms of cognitive control In AConway C Jarrold M Kane A Miyake amp J Towse(Eds) Variation in working memory Oxford OxfordUniversity Press

Chao L amp Knight R (1998) Contribution of humanprefrontal cortex to delay performance Journal ofCognitive Neuroscience 10 167ndash177

Chao L L amp Knight R T (1995) Human prefrontal lesionsincrease distractibility to irrelevant sensory inputsNeuroReport 6 1605ndash1610

Connolly J D Goodale M A Goltz H C amp Munoz D P(2005) fMRI activation in the human frontal eye field iscorrelated with saccadic reaction time Journal ofNeurophysiology 94 605ndash611

Courtney S M (2004) Attention and cognitive control asemergent properties of information representation inworking memory Cognitive Affective amp BehavioralNeuroscience 4 501ndash516

Courtney S M Ungerleider L G Keil K amp Haxby J(1996) Object and spatial visual working memory activateseparate neural systems in human cortex Cerebral Cortex6 39ndash49

Courtney S M Ungerleider L G Keil K amp Haxby J V(1997) Transient and sustained activity in a distributedneural system for human working memory Nature 386608ndash611

Curtis C E amp DrsquoEsposito M (2003) Persistent activity inthe prefrontal cortex during working memory Trends inCognitive Sciences 7 415ndash423

Curtis C E Rao V Y amp DrsquoEsposito M (2004) Maintenanceof spatial and motor codes during oculomotor delayedresponse tasks The Journal of Neuroscience 243944ndash3952

DrsquoEsposito M Ballard D Zarahn E amp Aguirre G K(2000) The role of prefrontal cortex in sensory memoryand motor preparation An event-related fMRI studyNeuroimage 11 400ndash408

DrsquoEsposito M amp Postle B R (1999) The dependence of spanand delayed-response performance on prefrontal cortexNeuropsychologia 37 1303ndash1315

DrsquoEsposito M Postle B R Ballard D amp Lease J (1999)Maintenance versus manipulation of information held inworking memory An event-related fMRI study Brain ampCognition 41 66ndash86

DrsquoEsposito M Postle B R Jonides J amp Smith E E(1999) The neural substrate and temporal dynamicsof interference effects in working memory as revealedby event-related functional MRI Proceedings of theNational Academy of Sciences USA 96 7514ndash7519

de Fockert J W Rees G Frith C D amp Lavie N (2001) Therole of working memory in visual attention Science 2911803ndash1806

Druzgal T J amp DrsquoEsposito M (2001) A neural networkref lecting decisions about human faces Neuron 32947ndash955

Duncan J amp Miller E K (2002) Cognitive focus throughadaptive neural coding in the primate prefrontal cortexIn D Stuss amp R Knight (Eds) Principles of frontal lobefunction (pp 278ndash291) Oxford Oxford University Press

Friston K J Holmes A P Poline J-B Heather J D ampFrackowiak R S J (1995) Analysis of fMRI time-seriesrevisited Neuroimage 2 45ndash53

Fukushima T Hasegawa I amp Miyashita Y (2004)Prefrontal neuronal activity encodes spatial targetrepresentations sequentially updated after nonspatialtarget-shift cues Journal of Neurophysiology 911367ndash1380

Fuster J M (1995) Memory in the cerebral cortexCambridge MIT Press

Gitelman D R (2002) ILAB A program for post experimentaleye movement analysis Behavioral Research MethodsInstruments and Computers 34 605ndash612

1688 Journal of Cognitive Neuroscience Volume 17 Number 11

parameter estimates associated with the covariates inquestion

Our analysis strategy was to identify delay-specificactivity in two regions the dlPFC and the IOTC5 frommemory-presentdistraction-absent trials This wouldidentify delay-period activity comparable to that ob-served in virtually all whole-brain neuroimaging studiesof similar tasks (eg with face stimuli Postle et al 2003Ranganath amp DrsquoEsposito 2001 Druzgal amp DrsquoEsposito2001 Courtney Ungerleider Keil amp Haxby 1997 Court-ney Ungerleider Keil amp Haxby 1996) Masks cor-responding to these two regions were drawn on eachsubjectrsquos anatomical scans Voxels located within thesemasks which became the VOIs were identified on tmaps of delay-period activity from memory-presentdis-traction-absent trials thresholded to a regionwise a of05 (one-tailed) with Bonferroni correction for the num-ber of voxels in the mask Note that with this approachwith spatially unsmoothed data individual voxels whoset value exceeds threshold can be interpreted as lsquolsquomean-ingfulrsquorsquo data (Postle et al 2000) Thus this procedurediffered somewhat from other applications of lsquolsquoregion-

of-interest (ROI) analysesrsquorsquo in that no selection criteriawere imposed with respect to cluster size or proximity ofcritical voxels This was done among other reasonsbecause we did not know a priori in which subregionof PFC to expect to find sensory-gating-related activityWithin the VOIs we evaluated the responses evokedby RSVP of faces during memory-presentdistraction-present trials versus memory-absentdistraction-presenttrials (Memory-absentdistraction-absent trials do notfigure directly into the test of the sensory gating hy-pothesis but were necessary to prevent anticipation ofdistraction on memory-absent trials)

Acknowledgments

The author thanks Olufunsho Faseyitan and Craig Rypstat forprogramming assistance and OF and Christopher Jordan forhelp with data collection and analysis This research was sup-ported by the National Institutes of Health RO1 MH064498

Reprint requests should be sent to Bradley R Postle Depart-mentofPsychologyUniversityof WisconsinndashMadison1202WestJohnson St Madison WI 53706 or via e-mail postlewiscedu

Figure 5 (continued)

Postle 1687

The data reported in this experiment have been depositedwith the fMRI Data Center archive (wwwfmridcorg) Theaccession number is 2-2005-119B9

Notes

1 This may be an empirical example of the lsquolsquoactive main-tenancersquorsquo proposed by Miller and Cohen (2001)2 The raw value of this subjectrsquos memory-absentdistraction-present effect in the dlPFC was 51 and so dividing thememory-presentdistraction-present by this value inflated it bya factor of almost 2 the opposite direction than that effectedby the normalization procedure in all other subjects exceptone whose raw memory-absentdistraction-present effect inthe dlPFC was 983 The fact that delay-period activity in the dlPFC VOIs wasnumerically lower during memory-presentdistraction-presentthan during memory-presentdistraction-absent trials is con-sistent with the idea that the sensory gating mechanism oncetriggered by neurons that are tonically active during unfilleddelay periods is implemented by a broader PFC network thatincludes regions that are not typically active during unfilleddelay periods Although the data are not shown here it iscertainly the case that in every subject considerably more dlPFCvoxels were active during the memory-presentdistraction-present delay than during the memory-presentdistraction-absent delay4 The HRF also varies in shape across brain regions with-in an individual participant (eg Miezen Maccotta OllingerPetersen amp Buckner 2000) However recent work indi-cates that the magnitude of variability in the shape of theHRF is greater across individuals within a region than it isacross regions within an individual (Handwerker Ollinger ampDrsquoEsposito 2004) Additionally it indicates that employinga participant-specific HRF can improve the sensitivity andmagnitude estimates of the least-squares solution of the GLMover comparable analyses that employ a generic HRF model(Handwerker et al 2004)5 The broader search space of the IOTC was used instead ofa more circumscribed area such as the fusiform face areabecause as can be seen in Figure 2 there was a considerableamount of topological variability of memory-presentdistraction-absent delay-period activity in the posterior ventral stream andno one area was activated in all subjects

REFERENCES

Aguirre G K Zarahn E amp DrsquoEsposito M (1998) Thevariability of human BOLD hemodynamic responsesNeuroimage 8 360ndash369

Baddeley A D amp Hitch G J (1974) Working memoryIn G H Bower (Ed) The psychology of learning andmotivation (Vol 8 pp 47ndash89) New York Academic Press

Bor D Duncan J Wiseman R J amp Owen A M (2003)Encoding strategies dissociate prefrontal activity fromworking memory demand Neuron 37 361ndash367

Boynton G M Engel S A Glover G H amp HeegerD J (1996) Linear systems analysis of functionalmagnetic resonance imaging in human V1 The Journalof Neuroscience 16 4207ndash4221

Braver T S Gray J R amp Burgess G C (in press)Explaining the many varieties of working memoryvariation Dual mechanisms of cognitive control In AConway C Jarrold M Kane A Miyake amp J Towse(Eds) Variation in working memory Oxford OxfordUniversity Press

Chao L amp Knight R (1998) Contribution of humanprefrontal cortex to delay performance Journal ofCognitive Neuroscience 10 167ndash177

Chao L L amp Knight R T (1995) Human prefrontal lesionsincrease distractibility to irrelevant sensory inputsNeuroReport 6 1605ndash1610

Connolly J D Goodale M A Goltz H C amp Munoz D P(2005) fMRI activation in the human frontal eye field iscorrelated with saccadic reaction time Journal ofNeurophysiology 94 605ndash611

Courtney S M (2004) Attention and cognitive control asemergent properties of information representation inworking memory Cognitive Affective amp BehavioralNeuroscience 4 501ndash516

Courtney S M Ungerleider L G Keil K amp Haxby J(1996) Object and spatial visual working memory activateseparate neural systems in human cortex Cerebral Cortex6 39ndash49

Courtney S M Ungerleider L G Keil K amp Haxby J V(1997) Transient and sustained activity in a distributedneural system for human working memory Nature 386608ndash611

Curtis C E amp DrsquoEsposito M (2003) Persistent activity inthe prefrontal cortex during working memory Trends inCognitive Sciences 7 415ndash423

Curtis C E Rao V Y amp DrsquoEsposito M (2004) Maintenanceof spatial and motor codes during oculomotor delayedresponse tasks The Journal of Neuroscience 243944ndash3952

DrsquoEsposito M Ballard D Zarahn E amp Aguirre G K(2000) The role of prefrontal cortex in sensory memoryand motor preparation An event-related fMRI studyNeuroimage 11 400ndash408

DrsquoEsposito M amp Postle B R (1999) The dependence of spanand delayed-response performance on prefrontal cortexNeuropsychologia 37 1303ndash1315

DrsquoEsposito M Postle B R Ballard D amp Lease J (1999)Maintenance versus manipulation of information held inworking memory An event-related fMRI study Brain ampCognition 41 66ndash86

DrsquoEsposito M Postle B R Jonides J amp Smith E E(1999) The neural substrate and temporal dynamicsof interference effects in working memory as revealedby event-related functional MRI Proceedings of theNational Academy of Sciences USA 96 7514ndash7519

de Fockert J W Rees G Frith C D amp Lavie N (2001) Therole of working memory in visual attention Science 2911803ndash1806

Druzgal T J amp DrsquoEsposito M (2001) A neural networkref lecting decisions about human faces Neuron 32947ndash955

Duncan J amp Miller E K (2002) Cognitive focus throughadaptive neural coding in the primate prefrontal cortexIn D Stuss amp R Knight (Eds) Principles of frontal lobefunction (pp 278ndash291) Oxford Oxford University Press

Friston K J Holmes A P Poline J-B Heather J D ampFrackowiak R S J (1995) Analysis of fMRI time-seriesrevisited Neuroimage 2 45ndash53

Fukushima T Hasegawa I amp Miyashita Y (2004)Prefrontal neuronal activity encodes spatial targetrepresentations sequentially updated after nonspatialtarget-shift cues Journal of Neurophysiology 911367ndash1380

Fuster J M (1995) Memory in the cerebral cortexCambridge MIT Press

Gitelman D R (2002) ILAB A program for post experimentaleye movement analysis Behavioral Research MethodsInstruments and Computers 34 605ndash612

1688 Journal of Cognitive Neuroscience Volume 17 Number 11

The data reported in this experiment have been depositedwith the fMRI Data Center archive (wwwfmridcorg) Theaccession number is 2-2005-119B9

Notes

1 This may be an empirical example of the lsquolsquoactive main-tenancersquorsquo proposed by Miller and Cohen (2001)2 The raw value of this subjectrsquos memory-absentdistraction-present effect in the dlPFC was 51 and so dividing thememory-presentdistraction-present by this value inflated it bya factor of almost 2 the opposite direction than that effectedby the normalization procedure in all other subjects exceptone whose raw memory-absentdistraction-present effect inthe dlPFC was 983 The fact that delay-period activity in the dlPFC VOIs wasnumerically lower during memory-presentdistraction-presentthan during memory-presentdistraction-absent trials is con-sistent with the idea that the sensory gating mechanism oncetriggered by neurons that are tonically active during unfilleddelay periods is implemented by a broader PFC network thatincludes regions that are not typically active during unfilleddelay periods Although the data are not shown here it iscertainly the case that in every subject considerably more dlPFCvoxels were active during the memory-presentdistraction-present delay than during the memory-presentdistraction-absent delay4 The HRF also varies in shape across brain regions with-in an individual participant (eg Miezen Maccotta OllingerPetersen amp Buckner 2000) However recent work indi-cates that the magnitude of variability in the shape of theHRF is greater across individuals within a region than it isacross regions within an individual (Handwerker Ollinger ampDrsquoEsposito 2004) Additionally it indicates that employinga participant-specific HRF can improve the sensitivity andmagnitude estimates of the least-squares solution of the GLMover comparable analyses that employ a generic HRF model(Handwerker et al 2004)5 The broader search space of the IOTC was used instead ofa more circumscribed area such as the fusiform face areabecause as can be seen in Figure 2 there was a considerableamount of topological variability of memory-presentdistraction-absent delay-period activity in the posterior ventral stream andno one area was activated in all subjects

REFERENCES

Aguirre G K Zarahn E amp DrsquoEsposito M (1998) Thevariability of human BOLD hemodynamic responsesNeuroimage 8 360ndash369

Baddeley A D amp Hitch G J (1974) Working memoryIn G H Bower (Ed) The psychology of learning andmotivation (Vol 8 pp 47ndash89) New York Academic Press

Bor D Duncan J Wiseman R J amp Owen A M (2003)Encoding strategies dissociate prefrontal activity fromworking memory demand Neuron 37 361ndash367

Boynton G M Engel S A Glover G H amp HeegerD J (1996) Linear systems analysis of functionalmagnetic resonance imaging in human V1 The Journalof Neuroscience 16 4207ndash4221

Braver T S Gray J R amp Burgess G C (in press)Explaining the many varieties of working memoryvariation Dual mechanisms of cognitive control In AConway C Jarrold M Kane A Miyake amp J Towse(Eds) Variation in working memory Oxford OxfordUniversity Press

Chao L amp Knight R (1998) Contribution of humanprefrontal cortex to delay performance Journal ofCognitive Neuroscience 10 167ndash177

Chao L L amp Knight R T (1995) Human prefrontal lesionsincrease distractibility to irrelevant sensory inputsNeuroReport 6 1605ndash1610

Connolly J D Goodale M A Goltz H C amp Munoz D P(2005) fMRI activation in the human frontal eye field iscorrelated with saccadic reaction time Journal ofNeurophysiology 94 605ndash611

Courtney S M (2004) Attention and cognitive control asemergent properties of information representation inworking memory Cognitive Affective amp BehavioralNeuroscience 4 501ndash516

Courtney S M Ungerleider L G Keil K amp Haxby J(1996) Object and spatial visual working memory activateseparate neural systems in human cortex Cerebral Cortex6 39ndash49

Courtney S M Ungerleider L G Keil K amp Haxby J V(1997) Transient and sustained activity in a distributedneural system for human working memory Nature 386608ndash611

Curtis C E amp DrsquoEsposito M (2003) Persistent activity inthe prefrontal cortex during working memory Trends inCognitive Sciences 7 415ndash423

Curtis C E Rao V Y amp DrsquoEsposito M (2004) Maintenanceof spatial and motor codes during oculomotor delayedresponse tasks The Journal of Neuroscience 243944ndash3952

DrsquoEsposito M Ballard D Zarahn E amp Aguirre G K(2000) The role of prefrontal cortex in sensory memoryand motor preparation An event-related fMRI studyNeuroimage 11 400ndash408

DrsquoEsposito M amp Postle B R (1999) The dependence of spanand delayed-response performance on prefrontal cortexNeuropsychologia 37 1303ndash1315

DrsquoEsposito M Postle B R Ballard D amp Lease J (1999)Maintenance versus manipulation of information held inworking memory An event-related fMRI study Brain ampCognition 41 66ndash86

DrsquoEsposito M Postle B R Jonides J amp Smith E E(1999) The neural substrate and temporal dynamicsof interference effects in working memory as revealedby event-related functional MRI Proceedings of theNational Academy of Sciences USA 96 7514ndash7519

de Fockert J W Rees G Frith C D amp Lavie N (2001) Therole of working memory in visual attention Science 2911803ndash1806

Druzgal T J amp DrsquoEsposito M (2001) A neural networkref lecting decisions about human faces Neuron 32947ndash955

Duncan J amp Miller E K (2002) Cognitive focus throughadaptive neural coding in the primate prefrontal cortexIn D Stuss amp R Knight (Eds) Principles of frontal lobefunction (pp 278ndash291) Oxford Oxford University Press

Friston K J Holmes A P Poline J-B Heather J D ampFrackowiak R S J (1995) Analysis of fMRI time-seriesrevisited Neuroimage 2 45ndash53

Fukushima T Hasegawa I amp Miyashita Y (2004)Prefrontal neuronal activity encodes spatial targetrepresentations sequentially updated after nonspatialtarget-shift cues Journal of Neurophysiology 911367ndash1380

Fuster J M (1995) Memory in the cerebral cortexCambridge MIT Press

Gitelman D R (2002) ILAB A program for post experimentaleye movement analysis Behavioral Research MethodsInstruments and Computers 34 605ndash612

1688 Journal of Cognitive Neuroscience Volume 17 Number 11

Goldman-Rakic P S amp Leung H-C (2002) Functionalarchitecture of the dorsolateral prefrontal cortex inmonkeys and humans In D T Stuss amp R T Knight (Eds)Principles of frontal lobe function (pp 85ndash95) OxfordUK Oxford University Press

Gray J R Chabris C F amp Braver T S (2003) Neuralmechanisms of general f luid intelligence NatureNeuroscience 6 316ndash322

Handwerker D A Ollinger J M amp DrsquoEsposito M (2004)Variation of BOLD hemodynamic responses acrosssubjects and brain regions and their effects on statisticalanalyses Neuroimage 21 1639ndash1651

Jensen O Gelfand J Kounios J amp Lisman J E (2002)Oscillations in the alpha band (9ndash12 Hz) increase withmemory load during retention in a short-term memorytask Cerebral Cortex 12 877ndash882

Jonides J Lacey S C amp Nee D E (2005) Processes ofworking memory in mind and brain Current Directionsin Psychological Science 14 2ndash5

Jonides J Marshuetz C Smith E E Reuter-Lorenz P Aamp Koeppe R A (2000) Age differences in behavior and PETactivation reveal differences in interference resolution inverbal working memory Journal of Cognitive Neuroscience12 188ndash196

Jonides J Smith E E Marshuetz C Koeppe R A ampReuter-Lorenz P A (1998) Inhibition of verbal workingmemory revealed by brain activation Proceedings of theNational Academy of Sciences USA 95 8410ndash8413

Josephs O Turner R amp Friston K (1997) Event-relatedfMRI Human Brain Mapping 5 243ndash248

Kane M J amp Engle R W (2003) The role of prefrontalcortex in working-memory capacity executive attentionand general f luid intelligence An individual-differencesperspective Psychonomic Bulletin amp Review 9 637ndash671

Klimesch W Doppelmayr M Schwaiger J Auinger P ampWinker T (1999) lsquolsquoParadoxicalrsquorsquo alpha synchronizationin a memory task Cognitive Brain Research 7 493ndash501

Knight R T Staines W R Swick D amp Chao L L (1999)Prefrontal cortex regulates inhibition and excitation indistributed neural networks Acta Psychologica 101159ndash178

Kwong K K Belliveau J W Chesler D A Goldberg I EWeisskoff R M Poncelet B P Kennedy D N HoppeltB E Cohen M S Turner R Cheng H-M Brady T Jamp Rosen B R (1992) Dynamic magnetic resonance imagingof human brain activity during primary sensory stimulationProceedings of the National Academy of Sciences USA89 5675ndash5679

Lebedev M A Messinger A Kralik J D amp Wise S P (2004)Representation of attended versus remembered locationsin prefrontal cortex PLoS Biology 2 1919ndash1935

Malmo R B (1942) Interference factors in delayed responsein monkey after removal of the frontal lobes Journal ofNeurophysiology 5 295ndash308

Miezen F M Maccotta L Ollinger J M Petersen S E ampBuckner R L (2000) Characterizing the hemodynamicresponse Effects of presentation rate sampling procedureand the possibility of ordering brain activity based on relativetiming Neuroimage 11 735ndash759

Miller E K amp Cohen J (2001) An integrative theory ofprefrontal cortex function Annual Review of Neuroscience24 167ndash202

Milner B (1964) Some effects of frontal lobectomy in manIn J M Warren amp K Akert (Eds) The frontal granularcortex and behavior (pp 313ndash334) New York McGraw-Hill

Monsell S amp Driver J (Eds) (2000) Control of cognitiveprocesses Attention and performance XVIII CambridgeMIT Press

Ogawa S Tank D W Menon R Ellermann J M Kim S GMerkle H amp Ugurbil K (1992) Intrinsic signal changesaccompanying sensory stimulation Functional brainmapping using MRI Proceedings of the National Academyof Sciences USA 89 5951ndash5955

Passingham D amp Sakai K (2004) The prefrontal cortexand working memory Physiology and brain imagingCurrent Opinion in Neurobiology 14 163ndash168

Passingham R E amp Rowe J B (2002) Dorsal prefrontalcortex Maintenance in memory or attentional selection InD T Stuss amp R T Knight (Eds) Principles of frontal lobefunction (pp 221ndash232) Oxford Oxford University Press

Pasternak T amp Greenlee M W (2005) Working memory inprimate sensory systems Nature Reviews Neuroscience 697ndash107

Petrides M (1995) Functional organization of humanfrontal cortex for mnemonic processing Evidence fromneuroimaging studies In Structure and functions ofthe human prefrontal cortex (Vol 769 pp 85ndash96)New York Annals of the New York Academy of Sciences

Petrides M (2000) Dissociable roles of mid-dorsolateralprefrontal and anterior inferotemporal cortex invisual working memory Journal of Neuroscience 207496ndash7503

Pochon J-B Levy R Poline J-B Crozier S Lehericy SPillon B Deweer B Le Bihan D amp Dubois B (2001) Therole of dorsolateral prefrontal cortex in the preparation offorthcoming actions An fMRI study Cerebral Cortex 11260ndash266

Postle B R Awh E Jonides J Smith E E amp DrsquoEsposito M(2004) The where and how of attention-based rehearsalin spatial working memory Cognitive Brain Research 20194ndash205

Postle B R Berger J S amp DrsquoEsposito M (1999) Functionalneuroanatomical double dissociation of mnemonic andexecutive control processes contributing to workingmemory performance Proceedings of the NationalAcademy of Sciences USA 96 12959ndash12964

Postle B R Berger J S Goldstein J H Curtis C E ampDrsquoEsposito M (2001) Behavioral and neurophysiologicalcorrelates of episodic coding proactive interference andlist length effects in a running span verbal workingmemory task Cognitive Affective and BehavioralNeuroscience 1 10ndash21

Postle B R amp Brush L B (2004) The neural bases of theeffects of item-nonspecific proactive interference in workingmemory Cognitive Affective amp Behavioral Neuroscience4 379ndash392

Postle B R Brush L B amp Nick A M (2004) Prefrontalcortex and the mediation of proactive interference inworking memory Cognitive Affective amp BehavioralNeuroscience 4 600ndash608

Postle B R amp DrsquoEsposito M (1999) lsquolsquoWhatrsquorsquomdashthenmdashlsquolsquowherersquorsquo in visual working memory An event-related fMRIstudy Journal of Cognitive Neuroscience 11 585ndash597

Postle B R amp DrsquoEsposito M (2003) Spatial working memoryactivity of the caudate nucleus is sensitive to frame ofreference Cognitive Affective and BehavioralNeuroscience 3 133ndash144

Postle B R Druzgal T J amp DrsquoEsposito M (2003) Seekingthe neural substrates of working memory storage Cortex39 927ndash946

Postle B R Zarahn E amp DrsquoEsposito M (2000) Usingevent-related fMRI to assess delay-period activity duringperformance of spatial and nonspatial working memorytasks Brain Research Protocols 5 57ndash66

Ranganath C DeGutis J amp DrsquoEsposito M (2004)Category-specific modulation of inferior temporal activity

Postle 1689

during working memory encoding and maintenanceCognitive Brain Research 20 37ndash45

Ranganath C amp DrsquoEsposito M (2001) Medial temporallobe activity associated with active maintenance of novelinformation Neuron 31 865ndash873

Rowe J B Stephan K E Friston K J Frackowiak R S Jamp Passingham R E (2005) The prefrontal cortex showscontext-specific changes in effective connectivity to motoror visual cortex during the selection of action or colourCerebral Cortex 15 85ndash95

Rowe J B Toni I Josephs O Frackowiak R S J ampPassingham R E (2000) The prefrontal cortex Responseselection or maintenance within working memory Science288 1656ndash1660

Ruchkin D S Grafman J Cameron K amp Berndt R S(2003) Working memory retention systems A state ofactivated long-term memory Behavioral and BrainSciences 26 709ndash777

Sakai K Rowe J B amp Passingham R E (2002) Activemaintenance in prefrontal area 46 createsdistractor-resistant memory Nature Neuroscience 5479ndash484

Stuss D T amp Knight R T (Eds) (2002) Principles offrontal lobe function New York Oxford University Press

Takeda K amp Funahashi S (2002) Prefrontal task-relatedactivity representing visual cue location or saccadedirection in spatial working memory tasks Journal ofNeurophysiology 87 567ndash588

Takeda K amp Funahashi S (2004) Population vectoranalysis of primate prefrontal activity during spatialworking memory Cerebral Cortex 14 1328ndash1339

Thompson-Schill S Jonides J Marshuetz C Smith E EDrsquoEsposito M Kan I P Knight R T amp Swick D (2002)Effects of frontal lobe damage on interference effects inworking memory Cognitive Affective and BehavioralNeuroscience 2 109ndash120

Watanabe M (1996) Reward expectancy in primate prefrontalneurons Nature 382 629ndash632

Watanabe M (2002) Integration across multiple cognitiveand motivational domains in monkey prefrontalcortex In D T Stuss amp R Knight (Eds) Principles offrontal lobe function (pp 326ndash337) Oxford OxfordUniversity Press

Worsley K J amp Friston K J (1995) Analysis of fMRItime-series revisitedmdashagain Neuroimage 2 173ndash182

Zarahn E Aguirre G K amp DrsquoEsposito M (1997) Atrial-based experimental design for fMRI Neuroimage6 122ndash138

1690 Journal of Cognitive Neuroscience Volume 17 Number 11