NeuroImage 64 (2013) 197–208

Contents lists available at SciVerse ScienceDirect

NeuroImage

j ourna l homepage: www.e lsev ie r .com/ locate /yn img

Neural correlates of control operations in inverse priming with relevant andirrelevant masks

Daniel Krüger a, Susan Klapötke a, Stefan Bode b, Uwe Mattler a,⁎a Georg-Elias-Müller Institute for Psychology, Georg-August University Göttingen, Germanyb Psychological Sciences, The University of Melbourne, Australia

⁎ Corresponding author at: Georg-Elias-Müller InstituteUniversity Göttingen, Gosslerstr, 14, D-37073 Göttingen,

E-mail address: uwe.mattler@psych.uni-goettingen.d

1053-8119/$ – see front matter © 2012 Elsevier Inc. Allhttp://dx.doi.org/10.1016/j.neuroimage.2012.09.018

a b s t r a c t

a r t i c l e i n f o

Article history:Accepted 5 September 2012Available online 16 September 2012

Keywords:PrimingNegative compatibility effectConsciousnessCognitive controlInhibition

The inverse priming paradigm can be considered one example which demonstrates the operation of control pro-cesses in the absence of conscious experience of the inducing stimuli. Inverse priming is generated by a primethat is followed by a mask and a subsequent imperative target stimulus. With “relevant”masks that are composedof the superposition of both prime alternatives, the inverse priming effect is typically larger than with “irrelevant”masks that are free of task-relevant features. We used functional magnetic resonance imaging (fMRI) to examinethe neural substrates that are involved in the generation of inverse priming effects with relevant and irrelevantmasks. We found a network of brain areas that is accessible to unconscious primes, including supplementarymotor area (SMA), anterior insula, middle cingulate cortex, and supramarginal gyrus. Activation of these brainareas were involved in inverse priming when relevant masks were used. With irrelevant masks, however, onlySMA activation was involved in inverse priming effects. Activation in SMA correlated with inverse priming effectsof individual participants on reaction time, indicating that this brain area reflects the size of inverse priming effectson the behavioral level. Findings aremost consistentwith the view that a basic inhibitorymechanism contributes toinverse primingwith either type ofmask and additional processes contribute to the effect with relevantmasks. Thisstudy provides new evidence showing that cognitive control operations in the human cortex take account of taskrelevant stimulus information even if this information is not consciously perceived.

© 2012 Elsevier Inc. All rights reserved.

Introduction

Cognitive control has been considered one instance which requiresconscious processing (e.g., Norman and Shallice, 1986). However, severalstudies have shown that the processing of a target stimulus can beinfluenced by a preceding unconscious prime stimulus (e.g., Neumannand Klotz, 1994; Vorberg et al., 2003). Typically, performance benefitsare observed when prime and target belong to the same category(congruent condition) as opposed to alternative categories (incongruentcondition). On this background, it is a matter of debate whetherunconscious stimuli can also modulate cognitive control operations(e.g., Dehaene et al., 2003; Rees et al., 2002). One instance in whichcognitive control is required is the inhibition of automatic responses(e.g., Norman and Shallice, 1986). The inhibition of automatic responseactivation has been assumed to account for the inverse priming effect(e.g., Eimer and Schlaghecken, 1998). Inverse priming is characterizedby performance deficits on congruent trials which can result when aseparatemasking stimulus is presented between prime and target stim-uli (e.g., Eimer and Schlaghecken, 1998). The size of inverse primingdepends on the stimulus onset asynchrony (SOA) between mask and

for Psychology, Georg-AugustGermany.e (U. Mattler).

rights reserved.

target stimulus (e.g., Mattler, 2007; Schlaghecken and Eimer, 2000)and the structure of the mask (e.g. Lleras and Enns, 2004). With shortmask-target SOAs priming effects tend to be positive, with increasingmask-target SOA inverse priming effects occur. As outlined below, it isassumed that this reversal of priming effects with long SOA is linkedto increased automatic control processing on congruent trials. Withmasks that consist of task-relevant features inverse priming effects aretypically larger than with masks that consist of irrelevant features.Here we examined the inverse priming paradigm in a functional mag-netic resonance imaging (fMRI) study to determine the automatic con-trol operations in the human cortex which are involved in inversepriming with relevant and irrelevant masks.

To account for themodulation of inverse priming effects by the struc-ture of the mask the literature provides at least three differentapproaches (Krüger et al., 2011). According to the Co-activeMechanismsapproach, an inhibitory mechanism is effective with both types of masksand an additional mechanism contributes to the effect with relevantmasks (Lleras and Enns, 2006). According to the Single-Mechanismapproach, the inhibitory mechanism generates inverse priming effectswith both kinds of masks and it is more productive with relevantmasks (Jaśkowski and Verleger, 2007). According to the Separate Mech-anisms approach, one mechanism accounts for the entire effect withrelevant masks, and a different inhibitory mechanism is only operatingwith irrelevant masks (Klapp, 2005). Therefore, clear evidence for the

198 D. Krüger et al. / NeuroImage 64 (2013) 197–208

Separate Mechanisms approach consists in an activation of non-overlapping brain regions with relevant and irrelevant masks. TheSingle-Mechanism approach predicts that the same brain regions areinvolved in inverse priming with relevant and irrelevant masks. Sharedand specific brain regions in the case of relevant and irrelevant masks,however, would be most consistent with the Co-active Mechanismsapproach.

Three previous fMRI studies have addressed the neuronal sourcesof inverse priming effects (Aron et al., 2003; Boy et al., 2010a, 2010b).Aron et al. (2003) employed arrow stimuli and irrelevant masks andvaried the mask-target SOA between 0 and 150 ms. Analyses of prim-ing effects in pre-specified regions-of-interest revealed that inversedpriming effects were associated to the activation in subcortical regionsof the basal ganglia (thalamus and caudate). Boy et al. (2010b) employedthe same paradigm but did not vary the prime target SOA. A specialregion-of-interest analysis that was based on a selection of the 20%most activated voxels yielded increased activity on congruent trials inregions of the supplementary motor area (SMA). An involvement ofSMA in inverse primingwas suggested by the absence of inverse primingeffects in a patient with a micro-lesion in SMA (Sumner et al., 2007) andby the correlation of the regional concentration of the neurotransmitterGABA in SMA with the magnitude of inverse priming effects as measurein the same paradigm (Boy et al., 2010a). We advance beyond thesestudies by comparing the effect of mask-target SOAs with relevant andirrelevant masks. This design enabled a model based approach wherewe compared the effect of SOA on congruent trials to the SOA effect onincongruent trials with either type of mask to examine whether theeffect results from common or separate mechanisms in the conditionwith relevant and irrelevant masks. We used whole-brain analyses toexceed the narrow focus of pre-specified regions-of-interest analyses.

Methods

Participants

28 healthy students from the University of Göttingenwere recruitedto participate in the four sessions of the experiment. Two of them vol-untarily quitted during the first two practice sessions, two subjectsdid not perform the task correctly, one showed anatomical irregulari-ties, and another subject's data were lost due to technical failure. Theremaining 22 subjects constituted the final sample (7 male, mean age22.6 years, ranging from 19 to 27 years). All had normal or corrected-to-normal vision andwere right-handed according to self-report. Beforeparticipation, subjects gave written informed consent and completed aquestionnaire to ensure MRT safety requirements and to rule out ahistory of neurological or psychiatric illness. Participants received anallowance of 51 € for their participation. The study was approved bythe Ethics Committee of the Medical Faculty of the University ofGöttingen.

Stimuli

The trial structure and timing parameters were the samethroughout all four sessions of the experiment (see Fig. 1). Primestimuli were left- or rightward pointing double arrows (≪ or ≫)presented at the center of the screen. Each target stimulus consistedof a pair of identical double arrows, pointing towards the left or tothe right, presented below and above fixation. Targets pointed tothe same side as the primes on half of the trials (congruent condition)and to opposite sides otherwise (incongruent condition). Primes werefollowed by a masking stimulus presented at the center of the screen.A relevant mask was presented on half of the trials and an irrelevantmask otherwise. The relevant mask consisted of a superposition ofboth prime alternatives. The irrelevant mask consisted of 130 linesof different length and width each approximately centered on the13×10 intersections of a virtual grid with a small random spatial jitter.

A new irrelevant mask was generated for each trial. All stimuli werepresented in black on a white background. Stimulus presentation wasrealized with a CRT-monitor during the first two practice sessions inthe lab, and with MRT-compatible LCD-glasses during the last twosessions in the scanner, both running at 60 Hz. All double arrowssubtended 1.5°×1° of visual angle. The two double arrowswhich consti-tuted the target stimuli were presented 2.3° above and below fixation,respectively. The virtual grid underlying the structure of the irrelevantmasks subtended 1.7°×1.1° of visual angle. The color of the fixationpoint was changed for 1000 ms to give feedback at the end of each trial(green and red following correct and incorrect responses, respectively).

Tasks

(a) Choice-reaction time task: During the initial two practice sessionsand the third session (in the scanner), subjects were required to judgethe orientation of the target stimuli as fast as possible avoiding errors.When the target arrows pointed to the left a response with the leftindex finger was required and otherwise a response with the rightindex finger. (b) Prime recognition task: To estimate the visibility ofprimes on the LCD-glasses of the scanner, we conducted a final primerecognition session with the same setup as in the previous session inthe scanner. In this session, subjects were informed about the presenceof the prime and they had to report its orientation in the same way asduring the first three sessions but without speed stress. Again, a left-ward pointing prime required a response with the left index fingerand a rightward pointing prime with the right index finger. Trial wiseerror feedback was given on each task.

Procedure

Prior to the MRT-session participants performed two identicalpractice sessions outside the scanner because pilot testing suggestedthat inverse priming effects increased with practice. The temporalstructure of the trials was the same in all four sessions of the exper-iment (Fig. 1). The beginning of each trial was signaled by a large fix-ation cross presented for 500 ms. The prime followed 300 ms afterthe offset of the fixation cross for 17 ms. The mask followed 17 msafter prime offset for 100 ms. The mask-target SOA varied randomlybetween trials with values of 33 or 150 ms. Targets were presentedfor 100 ms. After target onset, the computer waited 1433 ms or1317 ms for a response on trials with short or long SOA, respectively.Within this sequence of stimuli a fixation dot was presentedwhen noother stimulus was presented. The intertrial interval was 1033 ms.Thus, in total the duration of each trial was 4333 ms. In the MRT-session, volume acquisitionwas triggeredwith a temporal jitter relativeto the trial onset because the repetition time of the scanner (TR) was2 s. Therefore, the jitter varied between 0, 333, 667, 1000, 1333, and1667 ms. Each experimental condition was realized equally often witheach of these jitter conditions.

Design

In each session subjects performed eight runs with 96 trials each.Only the data of the two final sessions in the scanner were analyzed.We used a 2×2×2×2 repeated measures design with factors PrimeOrientation (left vs. right), Target Orientation (left vs. right), SOA (33vs. 150 ms), and Mask Structure (relevant vs. irrelevant). Each ofthese 16 conditions occurred six times per run and was combinedonce per run with each possible jitter between trial onset and theonset of volume acquisition. Thus, we acquired 48 trials in eachexperimental condition. Apart from the instructed task, choice-RTand prime recognition sessions were identical. Response time (RT),error rate, and the BOLD signal served as dependent measures inthe choice-RT session. The analysis of prime recognition perfor-mance focused on the effects of independent variables SOA and

+Fixation500 ms

300 ms

Prime17 ms

ISI17 ms

+

time

Fixation500 ms

300 ms

Prime17 ms

ISI17 ms

Mask alone33 ms

Mask + Target67 ms

1333 ms

Feedback1000 ms

1033 ms

Mask-Target SOA 33 ms

Target alone33 ms

Mask100 ms

Mask-Target SOA 150 ms

Target100 ms

ISI50 ms

1217 ms

Feedback1000 ms

1033 ms

Fig. 1. Stimulus sequence used in the experiment. Primes were followed by either a relevant or an irrelevant mask which in turn was followed by the target stimulus. Primes andtargets were either congruent or incongruent. Four incongruent trials are depicted which result from the factorial combination of short vs. long SOA, and relevant vs. irrelevantmask. Note that the durations of mask and target stimuli were constant 100 ms irrespective of SOA. Feedback was given by a red fixation point on error trials and a green pointon correct trials. Trial duration was constant 4333 ms. With a mask-target SOA of 33 ms positive priming was expected while with 150 ms inverse priming should occur. SOAand type of mask varied randomly between trials.

199D. Krüger et al. / NeuroImage 64 (2013) 197–208

Mask Structure on the dependent measure percentage of correctlyreported prime stimuli.

Data acquisition

Participants were scanned with a 3.0 T MRI System (SiemensMagnetom Trio). An echo-planar imaging (EPI) sequence wasimplemented. During the choice-RT task in the scanner we acquired 33slices aligned to the AC–PC plane in ascending order (TR=2000 ms,TE=30 ms, flip angle=70°, matrix 64×64, spatial resolution: 3×3×3.6 mm with 0.6 mm gap). For each run, 215 EPI T2*-weighted wholebrain images were acquired. The first two volumes of each run werediscarded to allow for T1 saturation effects. Preceding functional data ac-quisition, a high-resolution T1-weighted anatomical image was acquired

using an MPRage sequence (spatial resolution 1×1×1 mmwith no gap,TE=3.26 ms, TR=2250 ms, flip angle=9°).

Data analyses

MRT-data were analyzed using SPM8 (http://www.fil.ion.ucl.ac.uk/spm; Friston et al., 1995). First, all functional and the anatomicalimages were reoriented such that the commissura anterior becamethe origin of the image. Second, ArtRepair 4 (Mazaika et al., 2009)was used to remove bad slices and to eliminate extracerebral noise.Bad slices were replaced by an interpolation of the preceding andfollowing scan. On average 1.0% of all slices were corrected by thisprocedure (ranging from 0.0 to 4.1% between subjects). Next, a slicetime correction compensated for varying slice acquisition times. All

200 D. Krüger et al. / NeuroImage 64 (2013) 197–208

functional images were realigned to the first image of the first run tocorrect for head movement. At the same time, a mean image of allfunctional images was generated. Then, the anatomical image wasco-registered to this mean image and segmented into gray andwhite matter and cortico-spinal fluid proportions. Parameters forthe normalization to the template image provided by the MontrealNeurological Institute (MNI template) were obtained from the seg-mented anatomical image and applied to all functional and the ana-tomical images. The normalized functional images were smoothedusing a Gaussian Kernel (full width at half maximum: 6 mm). Lowfrequency artifacts were removed by applying a high-pass filter tothe time-series data at each voxel with a cutoff at 128 s.

Single subject analyses were performed by fitting a general linearmodel (GLM) to the time-series data. The 16 experimental conditionswere used as explanatory variables. Separately for every run, eachcondition was modeled as a time-series of events based on prime on-sets, excluding error trials. The resulting stick function was convolvedwith the canonical hemodynamic response function as implementedin SPM 8. Additionally, seven covariates-of-no-interest were added tothe GLM: one regressor modeled error trial and six regressors modeledhead movements (3 translation, 3 rotation parameters given by therealignment algorithm). Effects of our experimental manipulation onthe BOLD signal were assessed by linear combinations of the GLMparameter estimates (beta values). At the level of group statistics, thereliability of effects was tested by one-sample t-tests.

Statistical analysis of fMRI data

Data analyses are based on a response activation model of inversepriming which is described below. The model predicts different effectsof SOA on congruent and incongruent trials in distinct brain areas. Totest the assumptions of the model the main effect of SOA on congruenttrials is compared to the main effect of SOA on incongruent trials foreach mask separately. Additionally, the model enables the analysis ofbrain areas that are related to response conflict and unconscious controloperations. To distinguish control processes that are related to relevantmasks from those related to irrelevant masks and control processeswhich are involved with either type of mask, we determined brainareas which show specific patterns of activation indicative for controlprocesses (i) exclusively on trials with relevant masks, (ii) exclusivelyon trials with irrelevant masks, and (iii) for both masks.

Our analysis of unconscious control processes that are involved ininverse priming is derived from a simple response activation model ofinverse priming. Consistent with electrophysiological measures ofresponse activation in the inverse priming paradigm, the modelassumes that the processing of prime, mask, and target induces asequence of response activations and control processes which serveto compensate premature response tendencies. The sequence of pro-cesses depends on Congruency and SOA (see Fig. 2A). According tothemodel, increased regulatory demands are predicted only on congru-ent trials with long rather than short SOA (Con: 150 ms SOA>33 msSOA; see Fig. 2A upper left versus upper right panel). Note that this ap-proachdoes not need to determine the specific functional characteristicsof the control operations involved in inverse priming with long SOA. Itsimply assumes that the activation of behaviorally relevant control oper-ations on congruent trials is larger with long rather than short SOA.However, this contrast is confounded with the effects of the differencesin the physical stimulus sequences with long and short SOA. To controlfor these differences in visual processing, the effect of long SOAwas com-pared to short SOA on incongruent trials (Inc: 150 ms SOA>33 ms SOA.This analysis is based on the assumption that regulatory demands donot varywith SOAon incongruent trials where themask-induced rever-sal of response activation drives accumulation towards the correctresponse and target induced activation simply continues this tendency(see Fig. 2A; lower left versus lower right panel).

Three predictions can be derived from this model. First, the maineffect of SOA on incongruent trials should be restricted to visualbrain areas. Second, on congruent trials all these areas should exhibita comparable main effect of SOA. Third, brain areas which are respon-sive to SOA variation on congruent but not on incongruent trials canbe considered neural correlates of cognitive control operations.

To test these predictions, we examined the SOA effect on congruentand incongruent trials in the entire brain and compared the main SOAeffect on congruent trials to that on incongruent trials. We aimed to dis-tinguish brain regions according to five specific patterns of activation(Fig. 2B). First, to identify brain areas which are exclusively responsiveto physical differences in the stimulus sequence but not to varying con-trol demands we searched for voxels which respond to SOA on congru-ent and incongruent trials (see Fig. 2B, left panel). For this purpose, weused a conjunction analysis across the SOA effect on congruent trialsand the SOA effect on incongruent trials. From these voxelswe excludedvoxels which showed a larger SOA effect on congruent trials comparedto the SOA effect on incongruent trials. Note that the conjunction anal-ysis that we used tests the logical AND, i.e. the null hypothesis that anyof the two contrasts (e.g., congruent or incongruent) does not show aneffect (Nichols et al., 2005). Therefore, a significant effect in our con-junction analyses results only if the effect in each and every contrastis significant. In other words, conjunction analyses determine brainregions that are responsive to two (or more) experimental manipula-tions at the same time. Moreover, in our analyses the contrasts thatwere entered into any reported conjunction analysis were orthogonalof one another.

Second, to identify brain areas which are related to both visual pro-cessing and control processes we searched for voxels which respondwith a larger SOA effect on congruent than on incongruent trials(Fig. 2B, middle panel). This was accomplished by a conjunction analy-sis of SOA effects on congruent trials, SOA effects on incongruent trials,and the interaction of SOA and Congruency specified by larger SOAeffects on congruent trials.

Third, to identify brain areaswhich are specifically related to controloperations but not to perceptual operations we searched for voxelswhich respond to SOA on congruent trials but not on incongruent trials(Fig. 2B, right panel). This was accomplished by a conjunction analysisof SOA effects on congruent trials and the interaction of SOA andCongruency, and we excluded voxels that showed an SOA effect onincongruent trials. This conjunction analysis was tailored in threeways: To identify brain areas which are exclusively related to controloperations on trials with relevant (irrelevant) masks, we excludedvoxels which showed an SOA effect on congruent trials and the interac-tion SOA×Congruency with irrelevant (relevant) masks. To identifybrain areas which are related to control operations with either mask,the conjunction analysis had to be significant with either mask. Notethat these analyses increase the sensitivity for control operations ofour study as compared to that of a previous one (Boy et al., 2010b)which simply contrasted incongruent and congruent trials with longSOA. The model in Fig. 2 shows that contrasting congruent trials withlong and short SOAs enables a more exhaustive identification of controlrelated operations than the simple contrast of incongruent and congru-ent trials with long SOA.

Finally, we examined the relation between the activation in the SMAwhich is related to control operations and the inverse priming effect onperformance measures of RT. This analysis was stimulated by recentfindings which suggest a relation between the SMA and the size of thebehavioral inverse priming effect (Boy et al., 2010a, 2010b; Sumner etal., 2007). To this end, we correlated individuals' inverse priming effecton RTs – determined as the difference of RT on incongruentminus RT oncongruent trials –with individuals' increase in SMA activity on congru-ent trials from short to long SOA. This analysis was performed for eachmask separately, first, with the SMA voxel that showed the largestSOA effect on congruent trials, and second, with all SMA voxelswhich responded to SOA on congruent trials. Group-based statistical

A

B

congruent incongruentco

ng

ruen

t

SOA 33 ms

Correctresponse

Incorrectresponse

Correctresponse

Incorrectresponse

SOA 150 ms

inco

ng

ruen

t

Perception Perception & Control Control

33 150

Mask-Target SOA [ms]33 150

Mask-Target SOA [ms]33 150

Mask-Target SOA [ms]

time

Fig. 2. A) Response activationmodel for inverse priming. Themodel is based on the accumulatormodel of Vorberg et al. (2003) and research demonstrating the time course of lateralizedreadiness potentials in inverse priming (Eimer and Schlaghecken, 1998; Praamstra and Seiss, 2005). The figure describes the time course of the activation difference of left and right handresponses. Three consecutive phases can be distinguished: first, the prime activates the associated response. Second, the presentation of the mask induces a response activation whichreverses the prime-related response activation. The mask induced activation drives the differential response activation away from the prime associated response until the target ispresented. Third, targets induce response activation.When prime and target stimuli are congruent, the differential response activation is reversed again. Prime-incongruent targets, how-ever, simply continue themask induced activation. The model assumes that an overt response is triggered when the differential response activation reaches a threshold. We assume thatreversals of response activation induce response conflicts which require regulatory demands. Response conflict increases with longmask-target SOA on congruent trials leading to higherregulatory demands than on incongruent trials. Therefore, neural activation which increases with SOA on congruent trials but not on incongruent trials reflects the increase of regulatorydemands in inverse priming. B) Predicted effects of SOA and regulatory demands on brain activation. Left panel: brain activation in a brain region which is modulated by SOA in the sameway on congruent and incongruent trials reflects differences in perceptual processing which result from physical differences in the stimulus sequence with long and short SOA. Middlepanel: brain activation in regions which respond differentially to SOA on congruent and incongruent trials with a larger SOA effect on congruent than on incongruent trials. Activationin these regions is related to both perceptual and control operations. Right panel: brain activation in a region which responds to SOA only on congruent but not on incongruent trials.Activation in these regions reflects the increased processing of control operations on congruent trials with long SOA. Note, the model predicts a very specific interaction of SOA and Con-gruency as a sign of brain areas which are involved in control operations.

201D. Krüger et al. / NeuroImage 64 (2013) 197–208

inferences were made with an uncorrected pb0.001 and a clusterextent of more than 10 contiguous voxels.

Results

Behavioral results

RT dataPriming effects depended on SOA and Mask Structure as indexed

by the three-way interaction of Congruency, SOA, and Mask Structure

in a repeated measures ANOVA, F(1, 21)=57.7, pb .001 (see Fig. 3).With short SOAs, significant positive priming effects were observed(t(21)=12.4, and 13.2, pb .001 for relevant and irrelevant masks,respectively) with faster responses on congruent trials (relevantmask: 322 ms, irrelevant mask: 320 ms) as compared to incongruenttrials (365 ms and 363 ms, respectively). These effects were indepen-dent of mask structure because the interaction of Congruency andMaskStructure was not significant (F(1, 21)b1). With long SOAs, however,inverse priming effects were obtained: responses were slower on con-gruent trials (350 ms and 332 ms) when compared to incongruent

202 D. Krüger et al. / NeuroImage 64 (2013) 197–208

trials (322 and 330 ms). The size of inverse priming effects differedbetween the two masks as reflected in the significant interaction ofCongruency andMask Structure, F(1, 21)=60.1, pb .001. With the rele-vant mask, a substantial inverse priming effect of −28 ms was foundwhich significantly differed fromzero, t(21)=−6.8, pb .001 (one-tailed).In contrast, with the irrelevantmask, however, the inverse priming effectdid not reach significance (−2 ms, t(21)=− .5, p=.33, one-tailed).

Accuracy dataErrors occurred on 2.9% of all trials. Accuracy data were arc-sine

transformed and subjected to a repeated measures ANOVA. Primingeffects were significantly modulated by SOA and Mask Structure,F(1, 21)=8.2, pb .01 (see Fig. 3). With short SOA, significant positivepriming effects occurred with fewer errors on congruent (relevantmask: M=0.4%, irrelevant mask: 0.5%) as compared to incongruenttrials (2.7% and 3.4%, respectively; t(21)=4.5 and 4.0, pb .01, for rele-vant and irrelevant masks, respectively. The main effect of Congruencywas not significant, F(1, 21)b1. With long SOAs, inverse priming effectsoccurred with more errors on congruent (5.7% and 5.7% for relevant andirrelevant masks, respectively) as compared to incongruent trials (1.0%and 3.9%; t(21)=−5.3, pb .01, and−1.5, p=.07, one-tailed, for relevantand irrelevant masks, respectively). Inverse priming effects were largerwith relevant masks (−4.7%) than with irrelevant masks (−1.8%) asreflected in the significant interaction of Congruency andMask Structure(F(1, 21)=5.6, pb .03), resembling the effects on RT.

Prime recognition performanceOverall, participants correctly reported the orientation of the prime

stimulus on 50.7% of the trials. The effect of Mask Structure was signif-icant (F(1, 21)=9.8, pb .01) because the frequency of correct responseswas smaller with relevant masks (49.0%) than with irrelevant masks(52.3%). However, prime recognition performance did not differ signif-icantly from chance level neither with relevant (t(21)=−1.1, p>.28)nor with irrelevant masks (t(21)=1.5, p>.14). No other effect reachedsignificance.

fMRI results

Perceptual effects of SOAFor each type of mask we examined main effects of SOA separately

on congruent and incongruent trials. On incongruent trials, increasedbrain activity with long as compared to short SOA was restricted to

33 150

-40

-20

0

20

40

Mask-Target SOA [ms]

RT

inco

ng

ruen

t - R

Tco

ng

ruen

t [m

s]

relevant

irrelevant

-5

0

5

Err

or

Rat

e in

con

gru

ent -

Err

or

Rat

e co

ng

ruen

t [%

]

Fig. 3. Behavioral results. Net priming effect for RTs (lines) and error rates (bars) atshort and long SOA separate for the relevant (red) and irrelevant mask (blue). Primingeffects are determined as the difference in performance measures on incongruent trialsminus congruent trials. While short mask-target SOAs led to positive priming effects ofcomparable magnitude for both masks, long SOAs produced inverse priming whichwas larger and significantly below zero for relevant masks.

the visual system (Fig. 4A). For both types of masks we found symmet-rical clusters of activation at the transition from occipital to temporallobe with the maxima located on the left middle occipital gyrus andright inferior temporal gyrus. With irrelevant masks, SOA modulatedthe activity of larger brain regions than with relevant masks (see sup-plementary Table 1). This finding is consistent with our assumptionthat cognitive control demands do not vary with SOA on incongruenttrials (see Fig. 2A).

On congruent trials, the effect of SOA involved the same brainregions as on incongruent trials, but a large number of additionalregions were also significantly more activated with long rather thanshort SOA (Fig. 4B). These additional brain areas included the SMA,precentral gyrus and the postcentral gyrus (for a complete summarysee supplementary Table 2). This finding is consistentwith our assump-tion that the control demands increase on congruent trials when SOAincreases (Fig. 2A).

To separate purely perceptual effects from control related SOA ef-fects we determined voxels with comparable SOA effects on congru-ent and incongruent trials, and excluded voxels which showed aninteraction between SOA and Congruency (left panel of Fig. 2B).Purely perceptual effects of SOA were found bilaterally in virtuallythe same higher visual areas with relevant and irrelevant masks inthe anterior occipital lobe at the transition to the temporal lobe(Fig. 4C). The focus of perceptual SOA effects for each mask was lo-cated within a cytoarchitectonically defined region called hoc5 cor-responding to the area V5/MT+ (Malikovic et al., 2007). Tworegions with purely perceptual effects for both masks were found,in the left inferior occipital gyrus (MNI coordinates −42, −76, −4;z=3.73, p=0.000096, uncorrected) and in the right middle tempo-ral gyrus (MNI coordinates 46, −68, −2; z=3.76, p=0.000085,uncorrected; see the purple regions in Fig. 4C).

Perceptual and control related effects of SOATo identify brain regions which are involved in both perceptual

and control related operations we determined those voxels in whichthe SOA effect on congruent trials exceeded the SOA effect on incon-gruent trials (middle panel in Fig. 2B). However, with the preset sta-tistical threshold we did not find any brain regions with this patternof activity with neither type of mask.

Control related effects of SOAFirst, we identified brain regions that are involved in control oper-

ations with either mask. As outlined above, this analysis is based on aspecific pattern of the interaction between SOA and Congruency. Thegeneral interaction is reported in supplementary Table 3 for relevant,irrelevant, and both masks. We identified brain regions that are in-volved exclusively with relevant or irrelevant masks.

Then we specified the interaction by searching for increased activ-ity with long SOA as compared to short SOA only on congruent trialsand no SOA effect on incongruent trials (Fig. 2B, right panel). This pat-tern of results was found for both types of masks only in one clusterwithin the SMA (supplementary Table 4) where the most significantinteraction between SOA and Congruency over both masks was iden-tified (MNI coordinates 2,−8, 52; z=3.34, p=0.00042, uncorrected;see Fig. 6A). According to our model, the activation of this brain areais related to control processes which are relevant for inverse primingeffects with either type of mask. Therefore, this brain area could rep-resent a basic mechanism which is operating with both relevant andirrelevant masks.

Next, we tried to identify control related activation which is exclu-sively found with one type of mask. With this tailored analysis we didnot find any brain region which was activated only on trials with irrel-evant masks. In contrast, nine clusters were exclusively found withrelevant masks (Fig. 5; supplementary Table 4). On the one hand,mask specific control related activation was found in an additionalpart of the SMA. On the other hand, additional clusters were situated

A

B Con: SOA = 150 > SOA = 33

C [Con: SOA = 150 > SOA = 33] & [Inc: SOA = 150 > SOA = 33]

Relevant mask

Irrelevant mask

Both masks

0 9

T-values

Inc: SOA = 150 > SOA = 33

Fig. 4. Main effect of SOA on brain activation. Activations on trials with relevant masks (red), irrelevant masks (blue), and overlapping activation for both masks (purple). Resultsunderscore the validity of the response activation model. A) Incongruent trials: SOA affected only visual areas. B) Congruent trials: SOA affected the same areas as on incongruenttrials plus a number of additional areas within the parietal and frontal cortex. C) Common main effect of SOA on both congruent and incongruent trials: Visual effects peaked withinV5/MT+. All statistical results were thresholded at pb .001 (uncorrected) and only clusters of k>10 adjacent voxels are shown. We excluded voxels which showed stronger SOAeffects on congruent as compared to incongruent trials thresholded by the default SPM exclusion threshold of pb .05 (uncorrected).

203D. Krüger et al. / NeuroImage 64 (2013) 197–208

ventral to the SMA within the posterior part of the anterior cingulatecortex, the rostral cingulate zone (RCZ, Picard, and Strick, 1996). Thesame specific pattern of activation was also found within insulaand the surrounding rolandic operculum bilateral, and also withinthe left postcentral gyrus and the left supramarginal gyrus withpeak activations in the cytoarchitectonically defined subregionsPFcm and PFop (Caspers et al., 2006). Beyond this, the same patternof activation was also found in regions of the right cerebellum.

Correlation of SMA activity and behavioral inverse priming effectsWe performed correlation analysis in order to investigate whether

control related SOA effects in the SMAwere related to behavioral primingeffects. First, we determined the voxel, which was most strongly affectedby SOA on congruent trials (identified by the largest t-value) for the

conjunction of both masks in the entire group of subjects (0, −4, 54MNI, located in the SMA). Then we calculated the correlation betweenthe neural control related effect (βcongruent SOA=150−βcongruent SOA=33)in this voxel and the behavioral inverse priming effect (RTincongruent−RT-congruent) on trials with long SOA. Significant negative correlations wereobserved for both the relevantmask (r=− .43, pb .05) and the irrelevantmask (r=− .5, pb .02; Fig. 6A). These correlations indicate that subjectswith a larger increase of SMA activation on congruent trials with increas-ing SOA showed larger behavioral inverse priming effects. Second, to ruleout that our findingwas restricted to one voxel only, we determined cor-relations for all SMA voxels as indicated by the SMA-ROI from theMARSBAR toolbox for SPM (http://marsbar.sourceforge.net/, Tzourio-Mazoyer et al., 2002). For this, we selected all voxels within the SMAwhose activity was larger at long as compared to short SOA in the

T-values

Relevant Masks only

2 5

incongruent

incongruent - congruent

33 ms 150 ms0

1

2

3

Mea

n be

ta

Relevant Mask

33 ms 150 ms

Irrelevant Mask

CON: 150-33 INC: 150-33

0

1

Diff

eren

ce

CON: 150-33 INC: 150-33

Rolandic Operculum (48 - 24 22)

33 ms 150 ms0

1

2

3

Mea

n be

ta

Relevant Mask

33 ms 150 ms

Irrelevant Mask

CON: 150-33 INC: 150-33

0

1

Diff

eren

ce

CON: 150-33 INC: 150-33

SMA (12 -8 64)

33 ms 150 ms0

1

2

3

Mea

n be

ta

Relevant Mask

33 ms 150 ms

Irrelevant Mask

CON: 150-33 INC: 150-33

0

1

Diff

eren

ce

CON: 150-33 INC: 150-33

Insula lobe (38 2 10)

33 ms 150 ms0

1

2

3

Mea

n be

ta

Relevant Mask

33 ms 150 ms

Irrelevant Mask

CON: 150-33 INC: 150-33

0

1

Diff

eren

ce

CON: 150-33 INC: 150-33

Insula lobe (30 24 8)

33 ms 150 ms0

1

2

3

Mea

n be

ta

Relevant Mask

33 ms 150 ms

Irrelevant Mask

CON: 150-33 INC: 150-33

0

1

Diff

eren

ce

CON: 150-33 INC: 150-33

Postcentral Gyrus (-60 -18 18)

33 ms 150 ms0

1

2

3

Mea

n be

ta

Relevant Mask

33 ms 150 ms

Irrelevant Mask

CON: 150-33 INC: 150-33

0

1

Diff

eren

ce

CON: 150-33 INC: 150-33

Rolandic Operculum (-44 -4 8)

33 ms 150 ms0

1

2

3

Mea

n be

ta

Relevant Mask

33 ms 150 ms

Irrelevant Mask

CON: 150-33 INC: 150-33

0

1

Diff

eren

ce

CON: 150-33 INC: 150-33

Supramarginal Gyrus (- 50 -32 30)

33 ms 150 ms0

1

2

3

Mea

n be

ta

Relevant Mask

33 ms 150 ms

Irrelevant Mask

CON: 150-33 INC: 150-33

0

1

Diff

eren

ce

CON: 150-33 INC: 150-33

MiddleCingulateCortex (10 10 38) congruent

Fig. 5. Brain areas which are exclusively involved in control operations with relevant masks. Activity in these brain areas increased with SOA on congruent trials but not on incon-gruent trials. These areas were determined by a significant SOA effect on congruent trials and a significant interaction of Congruency and SOA excluding brain areas with a signif-icant SOA effect on incongruent trials. To find brain regions which are exclusively involved in mask specific control processes, we also excluded brain areas which showed either aneffect of SOA on congruent trials or the interaction Congruency×SOA on trials with the opposite type of mask. Eight brain areas showed the specific pattern of activity which sig-nifies control operations exclusively with relevant masks. Bar plots depict mean betas per condition averaged across subjects. Error bars indicate standard error across subjects.Thresholds are the same as in Fig. 4. Coordinates refer to MNI-space.

204 D. Krüger et al. / NeuroImage 64 (2013) 197–208

group analysis at pb .001 (yielding 305 voxels). An analysis of the corre-lations revealed that 159 SMA-voxels showed a significant (pb .05) neg-ative correlation with relevant masks, and 40 voxels with irrelevantmasks. The entire distributions of correlation coefficients for all selectedSMA voxels are given in Fig. 6B.

Discussion

The present study was designed to examine the neural correlatesof unconscious control operations in inverse priming. We used rele-vant and irrelevant masks to render primes invisible. Rigorous

-2 -1 0 1 2-70

-60

-50

-40

-30

-20

-10

0

10

20

30

40

βcon SOA 150 - βcon SOA 33

RT

inco

ngru

ent -

RT

cong

ruen

t [m

s]

A

-0.6 -0.4 -0.2 00

5

10

15

20

25

r(βcon SOA 150 - βcon SOA 33,RTinc - RTcon)

num

ber

of v

oxel

s

B irrelevant mask

relevant mask

33 ms 150 ms 0

1

2

3

Mea

n be

ta

33 ms 150 ms

con: 150-33 inc: 150-33

0

1

Diff

eren

ce

con: 150-33 inc: 150-33

Relevant Mask Irrelevant Mask

SMA (2 -8 52)

congruent incongruent difference

A

B C

Fig. 6. Brain areaswhich are involved in control operationswith both relevant and irrelevantmasks. A) Only SMA activity showed the expected pattern of activationwith eithermask. Barplots depict mean betas per condition averaged across subjects. Error bars indicate standard error across subjects as in Fig. 5. Thresholds are the same as in Fig. 4. Coordinates refer toMNI-space. B) Scatterplot illustrating the relation between individuals' SOA effect in the SMA on congruent trials and their behavioral inverse priming effect separated for the relevant(red) and irrelevantmasks (blue). Activity was determined in the SMA voxel which wasmaximally affected by SOA variation on congruent trials. We obtained a significant negative cor-relation for both, relevant (r=− .43, pb .05) and irrelevant masks (r=− .50, pb .05). C) Distribution of correlation coefficients in those SMA voxels which respond to SOA on congruenttrials, separated for trials with relevant (red) and irrelevant masks (blue). Negative correlations smaller than the value indicated by the vertical dashed line are significant at pb .05.

205D. Krüger et al. / NeuroImage 64 (2013) 197–208

psychophysical testing revealed that prime recognition performancewas at chance level with either type of mask. This finding suggeststhat participants have not consciously perceived the effective stimuliin our study. Nonetheless, the activation of several brain areas whichare related to high level control processes has been modulated by un-conscious primes. Findings suggest that the SMA is the common neuralsite of control processes in inverse priming on trials with relevant andirrelevant masks. Exclusively on trials with relevant masks, the activa-tion of several additional brain areas in the frontal and parietal cortexwas also modulated by unconscious primes. These findings supportthe view that inverse priming is governed by a common mechanismwhich operates with either type of mask, and an additional mechanismwhich operates only on trials with relevant masks (Co-active Mecha-nisms approach; Krüger et al., 2011). Beyond this, findings suggestthat various control operations in frontal and parietal cortex are suscep-tible to the effects of unconscious stimuli.

The present approach examined control processes in inverse prim-ing by comparing brain activity with long and short mask-targetSOAs on congruent trials because regulatory demands increase with

increasing SOA on such trials as suggested by our response activationmodel. Based on model assumptions it seems reasonable to assumethat this comparison comprises all relevant types of control operationswhich contribute to the behaviorally defined inverse priming effect.However, the specific characteristics of these control operations arenot entirely clear. The subliminal prime induces an initial responseactivation which is followed by a counteracting response activationtriggered by the mask and final counteracting response activationby the conscious target stimulus. On the one hand, the reversal of un-conscious response activation may need unconscious conflict processingand/or unconscious inhibitory control (e.g., Boy et al., 2010a, 2010b;van Gaal et al., 2010b). On the other hand, the reversal of mask triggeredresponse activation by the conscious target may require a different kindof conflict processing and voluntary motor control. Our analysis assumesthat the activity of control related brain areas increases with increasingamounts of misleading response activation. However, irrespective ofwhat specific kind of control operations involved, our analysis is basedon the simple assumption that there is a larger amount of control opera-tions involved on congruent trials with long SOA as compared to short

206 D. Krüger et al. / NeuroImage 64 (2013) 197–208

SOA. Similarly, on incongruent trials we assume that the amount of con-trol operations does not differ substantially with long and short SOA. Thefunctional interpretation of the resulting brain areas provides support forthis analysis.

As argued above, the contrast between brain activation on congruenttrials with short and long SOAs is confounded by the effect of the differ-ences in the temporal stimulus sequence. To control for these perceptualsequence effects, we analyzed SOA effects on incongruent trials. Ourassumptions were nicely confirmed by our findings. Irrespective ofMask structure and Congruency, the different stimulus sequencesmodulated activity within area V5/MT+ (Fig. 5C; Malikovic et al.,2007) which has been associated with motion perception (Wilmset al., 2005). Goebel et al. (1998) reported that apparent motionevoked larger activity in this brain region than a flicker control con-dition. We suggest that the sequence of central prime and mask presen-tation followed by the peripheral presentation of target stimuli hasinduced motion signals in our participants. Perception of apparent mo-tion depends on the spatial separation and the temporal delay betweenstimuli (Shepard and Zare, 1983). Therefore, motion signals which weregenerated by our stimulus sequence might have been stronger withlong rather than short SOA.

To localize control related brain areas we contrasted brain activa-tion on congruent trials with long and short SOA given that no suchSOA effect occurred on incongruent trials. SMA has been related toinverse priming with irrelevant masks in previous studies (Boy etal., 2010a, 2010b; Sumner et al., 2007). Our data extend these findingsby demonstrating a general role of the SMA in inverse priming withboth relevant and irrelevant masks. For both types of masks, activityin the SMA seems to mirror the need for control processes to inhibitoutdated response tendencies. A comparable relation between SMAactivity and response conflict has been foundwith the Eriksen Flankertask (Hazeltine et al., 2000). Findings are consistent with the viewthat SMA activation is related to motor control processes, which areinvolved in the suppression of planned but currently inappropriateresponse tendencies (Swick et al., 2011), the sudden adaptations ofplanned movements (Matsuzaka, and Tanji, 1996), response inhibi-tion in reaction to sudden task changes (Chen et al., 2010), and theinstantiation of currently relevant movements (see Nachev et al.,2008). The absence of any effect in pre-SMA supports the view thatthere are specific inhibitory control processes located in regions ofSMA proper which are absent in pre-SMA (Boy et al., 2010b; Jaffardet al., 2008; Sumner et al., 2007). It has been suggested to distinguishbetween externally triggered, transient “reactive” inhibition and endog-enously guided sustained “proactive” inhibition (Aron, 2011). In thisperspective, the present findings are evidence for a reactive inhibitionprocess located in regions of SMA as part of a motor control systemwhich is susceptible to unconscious stimuli. Results are consistentwith the view that SMA is part of an automatic control system whichcan be triggered in specific task contexts byunconscious stimuli, where-as pre-SMA is related to complex higher level control processes that areexperienced as voluntary (Lau et al., 2004; van Gaal et al., 2010b). Insum, inverse priming might be a helpful paradigm to distinguish thecomplex function of pre-SMA and SMA because it seems to activateonly SMA but not pre-SMA.

With relevant masks we found additional brain areas where themetabolic activation increased when the unconscious need for controlprocesses which was triggered by unconscious stimuli increased. Theseadditional operations might be responsible for increased behavioralinverse priming effects with relevant masks. One cluster which wasspecifically involvedwith relevantmasks is situated in themiddle frontalcortex, in the RCZ which has previously been associated with responsemonitoring, response conflict and decision uncertainty (Ridderinkhofet al., 2004a, 2004b; Roger et al., 2010). Activation in bilateral clusterswithin the insula lobe was also correlated with regulatory demands.More specifically, the specific control related pattern of activation wasfound in the anterior part of the insula and a cluster in the posterior

portion of the right-hemisphere, and in a single cluster of the posteriorregion of the left insula. This finding accords well with the meta-analysis of Swick et al. (2011) which identified the anterior insula (pre-dominantly in the right hemisphere) as a commonmodule for successful-ly withholding a prepared response. According to meta-analyses ofDosenbach et al. (2006, 2007) MCG and insula are central componentsof the core task-set system which controls goal-directed behaviorthrough the stable maintenance of task sets. Thus, our data suggestthat unconscious stimuli have access to this cingulo-opercular task-setnetwork. This conclusion is in line with recent findings of Van Gaal etal. (2010a) who report an effect of unconscious no-go stimuli onpre-SMA and bilateral inferior frontal regions.

Activity in the left post-central gyrus and in the left supramarginalgyrus (SMG) was also modulated by unconscious primes. Rushworthet al. (2001) proposed a manual motor attention system in the leftsupramarginal gyrus and intraparietal sulcus, a view which is con-firmed by a recent MRI study (Cotti et al., 2011). Our data suggestthat this system is also susceptible to unconscious stimuli: The pro-cessing of increased unconsciously activated response conflict seemsto modulate motor attention for the upcoming movement. Together,these areas constitute an additional network which contributes to in-verse priming in the case of relevant masks.

In sum, this pattern of results is most consistent with the Co-activeMechanisms approach which assumes that common processes areinvolved with both types of masks and that additional processes areresponsible for the increased size of inverse priming effects in thecase of relevant masks. However, we cannot rule out that our findingsresult from the reduced inverse priming effects on trials with irrele-vant masks which might have led to the reduced activation of brainareas. In this perspective, the present findings are also compatiblewith the Single-Mechanism approach, which assumes that inversepriming with relevant and irrelevant masks results from the sameoperations. The Separate Mechanisms approach, however, seems toconflict with the present results, unless one argues that different pro-cesses are responsible for SMA activation with relevant and irrelevantmasks.

In addition, our findings contribute to the discussion regarding thesource of inverse priming effects. The anatomical location of the controlareas that are modulated by the congruency between unconsciousprimes and visible target stimuli used in this study seems to be consis-tent with the view that inverse priming is generated in the responsesystem (Boy and Sumner, 2010; Jaśkowski and Verleger, 2007; Klappand Hinkley, 2002; Lleras, and Enns, 2006; Schlaghecken and Eimer,2002) rather than at perceptual levels of processing (Huber, 2008;Krüger and Mattler, 2012; Krüger et al., 2011; Mattler, 2006, 2007;Sohrabi and West, 2009). Clearly, this interpretation depends on thecurrent functional specification of the responsive brain areas. Nonethe-less, a comparison of the literature suggests that different amounts ofresponse compatibility of the primes could explain why some studiesreport evidence for a perceptual origin of inverse priming and other stud-ies evidence in favor of a post-perceptual mechanism (cf. Jaśkowski, andŚlósarek, 2007). Evidence for a post-perceptual origin is provided mostlyby studies which employed arrow stimuli. Arrows are characterized by ahigh degree of dimensional overlap to the response set (Kornblum et al.,1990) which could result in automatic activation of the correspondingresponse (D'Ostilio, and Garraux, 2011; Eimer, 1995; Georgopoulos etal., 1989; Kornblumet al., 1990; Proctor et al., 1995). In contrast, evidencefor a perceptual origin of inverse priming was found with stimuli whichdo not overlap with response dimensions (Krüger et al., 2011; Mattler,2006, 2007). These stimuli might be processed on a different, less auto-matic route which involves controlled response selection processes(Kornblum et al., 1990) and avoids automatic response activation.Further research is needed to examine this hypothesis.

Cognitive control has traditionally been linked to conscious influ-ences on cognitive processing and behavior, which has been contrastedto automatic processing. In this view cognitive control is related to

207D. Krüger et al. / NeuroImage 64 (2013) 197–208

conscious control processes that are voluntarily operated and cannot beinfluenced by unconscious information. Both of these aspects have beenquestioned recently. For one thing, everyday experience already castsdoubt on the view that we have to be aware of every stimulus whichhas to be taken into account to behave successfully for instance whenwe drive a car in a difficult situation. In the same line, experimental evi-dence demonstrated that participants' voluntary free choice of a motorresponse can be influenced by unconscious information (e.g., Mattlerand Palmer, 2012). Beyond this, the inverse priming paradigm demon-strates that control processes do not have to be voluntarily operated bythe participants. Instead, control processes can be triggered unconscious-ly by a specific stimulus sequence in the given task context.

To account for priming effects of unconscious stimuli, a notion of“conditional automaticity” has been suggested recently which assumesthat a specific task set may prepare the processing system in such awaythat task relevant stimuli can be processed irrespective of consciousawareness (e.g., Kiefer and Martens, 2010; Sumner and Husain, 2008).Experimental evidence for the influence of unconscious stimuli on cog-nitive control processes has been provided only recently. Behavioralpriming studies suggested that unconscious primes can induce changesof task sets (Mattler, 2003, 2005, 2007), a finding which has been sub-stantiated by fMRI findings (Lau and Passingham, 2007). Behavioralstudies provided evidence for an effect of unconscious stimuli on con-trol processes in go–no-go and stop signal tasks and physiological stud-ies located this effect in the frontal cortex (Van Gaal et al., 2008, 2009,2010a, 2010b, 2010c). Unconsciously modulated inhibitory controloperations have been suggested by behavioral studies on the inversepriming effect and initial physiological studies have substantiated thisview (Boy et al., 2010a, 2010b; Sumner et al., 2007). The present fMRIstudy complements this evidence using an elaborated inverse primingparadigm which shows that small changes of mask structure canchange the control operations which respond to unconscious stimuli.In sum, the findings suggest that a broad variety of control operationscan integrate stimulus information which is not consciously perceived.

Acknowledgments

We thank Peter Dechent and the group MR-Research in Neurologyand Psychiatry at the University of Göttingen for allowance and technicalsupport in using the scanner for this study, and Torsten Wüstenberg forhis grateful help in data analyses. This research was funded by theDeutsche Forschungsgemeinschaft Grant MA 2276/3-2 awarded to UweMattler.

Appendix A. Supplementary data

Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.neuroimage.2012.09.018.

References

Aron, A.R., 2011. From reactive to proactive and selective control: developing a richermodel for stopping inappropriate responses. Biol. Psychiatry 69, e55–e68.

Aron, A.R., Schlaghecken, F., Fletcher, P.C., Bullmore, E.T., Eimer, M., Barker, R., Sahakian,B.J., Robbins, T.W., 2003. Inhibition of subliminally primed responses is mediatedby the caudate and thalamus: evidence from functional MRI and Huntington's dis-ease. Brain 126, 713–723.

Boy, F., Sumner, P., 2010. Tight coupling between positive and reversed priming in themasked priming paradigm. J. Exp. Psychol. Hum. Percept. Perform. 36 (4), 892–905.

Boy, F., Evans, C.J., Edden, R.A.E., Singh, K.D., Husain, M., Sumner, P., 2010a. Individualdifferences in subconscious motor control predicted by GABA concentration inSMA. Curr. Biol. 20, 1779–1785.

Boy, F., Husain, M., Singh, K.D., Sumner, P., 2010b. Supplementary motor area activationsin unconscious inhibition of voluntary action. Exp. Brain Res. 206, 441–448.

Caspers, S., Geyer, S., Schleicher, A., Mohlberg, H., Amunts, K., Zilles, K., 2006. Thehuman inferior parietal cortex: cytoarchitectonic parcellation and interindividualvariability. Neuroimage 33, 430–448.

Chen, X., Scangos, K.W., Stuphorn, V., 2010. Supplementary motor area exerts proactiveand reactive control of arm movements. J. Neurosci. 30 (44), 14657–14675.

Cotti, J., Rohenkohl, G., Stokes, M., Nobre, A.C., Coull, J.T., 2011. Functionally dissociatingtemporal and motor components of response preparation in left intraparietal sulcus.Neuroimage 54, 1221–1230.

D'Ostilio, K., Garraux, G., 2011. Automatic stimulus-induced medial premotor cortexactivation without perception or action. PLoS One 6 (2), e16613.

Dehaene, S., Artiges, E., Naccache, L., Martelli, C., Viard, A., Schurhoff, F., et al., 2003. Consciousand subliminal conflicts in normal subjects and patients with schizophrenia: the role ofthe anterior cingulate. Proc. Natl. Acad. Sci. U. S. A. 100, 13722–13727.

Dosenbach, N.U.F., Visscher, K.M., Palmer, E.D., Miezin, F.M., Wenger, K.K., Kang, H.C., etal., 2006. A core system for the implementation of task sets. Neuron 50, 799–812.

Dosenbach, N.U.F., Damien, A.F., Miezin, F.M., Cohen, A.L., Wenger, K.K., Dosenbach,R.A.T., et al., 2007. Distinct brain networks for adaptive and stable task control inhumans. Proc. Natl. Acad. Sci. U. S. A. 104, 11073–11078.

Eimer, M., 1995. Stimulus–response compatibility and automatic response activation:evidence from psychophysiological studies. J. Exp. Psychol. Hum. Percept. Perform.21, 837–854.

Eimer, M., Schlaghecken, F., 1998. Effects of masked stimuli on motor activation: behav-ioral and electrophysiological evidence. J. Exp. Psychol. Hum. Percept. Perform. 24,1737–1747.

Friston, K.J., Holmes, A.P., Worsley, K.J., Poline, J.P., Frith, C.D., Frackowiak, R.S.J., 1995.Statistical parametric maps in functional imaging: a general approach. Hum.Brain Mapp. 2, 189–210.

Georgopoulos, A.P., Lupito, J.T., Petrides, M., Schwartz, A.B., Massey, J.T., 1989. Mentalrotation of the neuronal population vector. Science 243, 234–236.

Goebel, R., Khorram-Sefat, D., Muckli, L., Hacker, H., Singer, W., 1998. The constructivenature of vision: direct evidence from functional magnetic resonance imagingstudies of apparent motion and motion imagery. Eur. J. Neurosci. 10, 1563–1573.

Hazeltine, E., Poldrack, R., Gabrieli, J.D.E., 2000. Neural activation during response conflict.J. Cogn. Neurosci. 12, 118–129.

Huber, D.E., 2008. Immediate priming and cognitive aftereffects. J. Exp. Psychol. Gen.137, 324–347.

Jaffard, M., Longcamp, M., Velay, J.L., Anton, J.L., Roth, M., Nazarian, B., Boulinguez, P.,2008. Proactive inhibitory control of movement assessed by event-related fMRI.Neuroimage 42, 1196–1206.

Jaśkowski, P., Ślósarek, M., 2007. How important is the prime's gestalt for subliminalpriming? Conscious. Cogn. 16, 485–497.

Jaśkowski, P., Verleger, R., 2007. What determines the direction of subliminal priming.Adv. Cogn. Psychol. 3 (1–2), 181–192.

Kiefer, M., Martens, U., 2010. Attentional sensitization of unconscious cognition: task setsmodulate subsequent masked semantic priming. J. Exp. Psychol. Gen. 139, 464–489.

Klapp, S.T., 2005. Two versions of the negative compatibility effect: comment on Llerasand Enns (2004). J. Exp. Psychol. Gen. 134, 431–435.

Klapp, S.T., Hinkley, L.B., 2002. The negative compatibility effect: unconscious inhibitioninfluences reaction time and response selection. J. Exp. Psychol. Gen. 131, 255–269.

Kornblum, S., Hasbroucq, T., Osman, A., 1990. Dimensional overlap: cognitive basis forstimulus–response compatibility — a model and taxonomy. Psychol. Rev. 97, 253–270.

Krüger, D., Mattler, U., 2012. Inverse cue priming is not limited to masks with relevantfeatures. Conscious. Cogn. 21 http://dx.doi.org/10.1016/j.concog.2012.04.003.

Krüger, D., Klapötke, S., Mattler, U., 2011. PRP-paradigm provides evidence for a perceptualorigin of the negative compatibility effect. Conscious. Cogn. 20, 866–881.

Lau, H.C., Passingham, R.E., 2007. Unconscious activation of the cognitive control system inthe human prefrontal cortex. J. Neurosci. 27, 5805–5811.

Lau, H.C., Rogers, R.D., Haggard, P., Passingham, R.E., 2004. Attention to intention. Science303, 1208–1210.

Lleras, A., Enns, J.T., 2004. Negative compatibility or object updating? A cautionary taleof mask-dependent priming. J. Exp. Psychol. Gen. 133, 475–493.

Lleras, A., Enns, J.T., 2006. How much like a target can a mask be? Geometric, spatial, andtemporal similarity in priming: a reply to Schlaghecken and Eimer (2006). J. Exp.Psychol. Gen. 135, 495–500.

Malikovic, A., Amunts, K., Schleicher, A., Mohlberg, H., Eickhoff, S.B., Wilms, M., Palomero-Gallagher, N., Armstrong, E., Zilles, K., 2007. Cytoarchitectonic analysis of the humanextrastriate cortex in the region of V5/MT1: a probabilistic, stereotaxic map of areahOc5. Cereb. Cortex 17, 562–574.

Matsuzaka, Y., Tanji, J., 1996. Changing directions of forthcoming arm movements:neuronal activity in the presupplementary and supplementary motor area of themonkey cerebral cortex. J. Neurophysiol. 76 (4), 2327–2342.

Mattler, U., 2003. Priming of mental operations by masked stimuli. Percept. Psychophys.65, 167–187.

Mattler, U., 2005. Inhibition and decay of motor and non-motor priming. Percept.Psychophys. 67, 285–300.

Mattler, U., 2006. On the locus of priming and inverse priming effects. Percept. Psychophys.68, 975–991.

Mattler, U., 2007. Inverse target- and cue-priming effects of masked stimuli. J. Exp.Psychol. Hum. Percept. Perform. 33, 83–102.

Mattler, U., Palmer, S., 2012. Time course of free-choice priming effects explained by asimple accumulator model. Cognition 123, 347–360.

Mazaika, P.K., Hoeft, F., Glover, G.H., Reiss, A.L., 2009. Methods and software for fMRIanalysis of clinical subjects. NeuroImage 47 (Supplement 1), S39–S41.

Nachev, P., Kennard, C., Husain, M., 2008. Functional role of the supplementary andpre-supplementary motor areas. Nat. Rev. Neurosci. 9, 856–869.

Neumann, O., Klotz, W., 1994. Motor responses to nonreportable, masked stimuli:where is the limit of direct parameter specification. In: Umiltà, C., Moscovitch, M.(Eds.), Attention and Performance XV: Conscious and Nonconscious InformationProcessing. MIT Press, Cambridge, MA, pp. 123–150.

Nichols, T., Brett, M., Andersson, J., Wager, T., Poline, J.B., 2005. Valid conjunction inferencewith the minimum statistic. Neuroimage 25, 653–660.

208 D. Krüger et al. / NeuroImage 64 (2013) 197–208

Norman, D.A., Shallice, T., 1986. Attention to action: willed and automatic control ofbehavior. In: Davidson, R.J., Schwartz, G.E., Shapiro, D. (Eds.), Consciousness andSelf-regulation, Vol. 4. Plenum Press, New York, pp. 1–18.

Picard, N., Strick, P.L., 1996. Motor areas of the medial wall: a review of their locationand function. Cereb. Cortex 6, 342–353.

Praamstra, P., Seiss, E., 2005. The neurophysiology of response competition: motor cortexactivation and inhibition following subliminal response priming. J. Cogn. Neurosci. 17(3), 483–493.

Proctor, R.W., Lu, C.-H.,Wang, H., Dutta, A., 1995. Activation of response codes by relevantand irrelevant stimulus information. Acta Psychol. 90, 275–286.

Rees, G., Kreiman, G., Koch, C., 2002. Neural correlates of consciousness in humans. Nat.Rev. Neurosci. 3, 261–270.

Ridderinkhof, K.R., van Wildenberg, W.P.M., Segalowitz, S.J., Carter, C.S., 2004a.Neurocognitive mechanisms of cognitive control: the role of prefrontal cortex inaction selection, response inhibition, performance monitoring, and reward-basedlearning. Brain Cogn. 56, 129–140.

Ridderinkhof, K.R., Ullsperger, M., Crone, E.A., Nieuwenhuis, S., 2004b. The role of medialfrontal cortex in cognitive conrol. Science 306, 443–447.

Roger, C., Bénar, C.G., Vidal, F., Hasbroucq, T., Burle, B., 2010. Rostral cingulate zone andcorrect response monitoring: ICA and source localization evidences for the unicityof correct- and error-negativities. Neuroimage 51, 391–403.

Rushworth, M.F.S., Krams, M., Passingham, R.E., 2001. The attentional role of the leftparietal cortex: the distinct lateralization and localization of motor attention inthe human brain. J. Cogn. Neurosci. 13 (5), 698–710.

Schlaghecken, F., Eimer, M., 2000. A central-peripheral asymmetry in masked priming.Percept. Psychophys. 62, 1367–1382.

Schlaghecken, F., Eimer, M., 2002. Motor activation with and without inhibition: evi-dence for a treshold mechanism in motor control. Percept. Psychophys. 64,148–162.

Shepard, R.N., Zare, S.L., 1983. Path-guided apparent motion. Science 220, 632–634.Sohrabi, A., West, R.L., 2009. Positive and negative congruency effects in masked priming:

a neuro-computational model based on representation, attention and conflict. BrainRes. 1289, 124–132.

Sumner, P., Husain, M., 2008. At the edge of consciousness: automatic motor activationand voluntary control. Neuroscientist 14 (5), 474–486.

Sumner, P., Nachev, P., Morris, P., Peters, A.M., Jackson, S.R., Kennard, C., Husain, M., 2007.Human medial frontal cortex mediates unconscious inhibition of voluntary action.Neuron 54 (5), 697–711.

Swick, D., Ashley, V., Turken, U., 2011. Are the neural correlates of stopping and not goingidentical? Quantitative meta-analysis of two response inhibition tasks. Neuroimage56, 1655–1665.

Tzourio-Mazoyer, N., Landeau, B., Papathanassiou, D., Crivello, F., Etard, O., Delcroix, N.,Mazoyer, B., Joliot, M., 2002. Automated anatomical labeling of activations in SPMusing a macroscopic anatomic parcellation of the MNI MRI single-subject brain.Neuroimage 15, 273–289.

Van Gaal, S., Ridderinkhof, K.R., Fahrenfort, J.J., Scholte, S., Lamme, V.A.F., 2008. Fron-tal cortex mediates unconsciously triggered inhibitory control. J. Neurosci. 28,8053–8062.

Van Gaal, S., Ridderinkhof, K.R., Van denWildenberg, W.P.M., Lamme, V.A.F., 2009. Dissoci-ating consciousness from inhibitory control: evidence for unconsciously triggeredresponse inhibition in the stop signal task. J. Exp. Psychol. Hum. Percept. Perform. 35,1129–1139.

Van Gaal, S., Ridderinkhof, K.R., Scholte, S., Lamme, V.A.F., 2010a. Unconscious activationof the prefrontal no-go network. J. Neurosci. 30, 4143–4150.

Van Gaal, S., Scholte, S., Lamme, V.A.F., Fahrenfort, J.J., Ridderinkhof, K.R., 2010b. Pre-SMAgray-matter density predicts individual differences in action selection in the face ofconscious and unconscious response conflict. J. Cogn. Neurosci. 23, 382–390.

Van Gaal, S., Lamme, V.A.F., Fahrenfort, J.J., Ridderinkhof, K.R., 2010c. Dissociable brainmechanisms underlying the conscious and unconscious control of behavior. J. Cogn.Neurosci. 23, 91–105.

Vorberg, D., Mattler, U., Heinecke, A., Schmidt, T., Schwarzbach, J., 2003. Different timecourses for visual perception and action priming. Proc. Natl. Acad. Sci. U. S. A. 100,6275–6280.

Wilms, M., Eickhoff, S.B., Specht, K., Amunts, K., Shah, N.J., Malikovic, A., Fink, G.R., 2005.Human V5/MT+: comparison of functional and cytoarchitectonic data. Anat.Embryol. 210, 485–495.