Acta Psychologica 152 (2014) 10–18

Contents lists available at ScienceDirect

Acta Psychologica

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

Post-error slowing is influenced by cognitive control demand☆

Shirley Regev ⁎, Nachshon MeiranDepartment of Psychology, Ben-Gurion University of the Negev, IsraelZlotowski Center for Neuroscience, Ben-Gurion University of the Negev, Israel

☆ We wish to thank Shelly Gil, Joseph Sarfaty, Liat Shprdata collection. This research was supported by researchScience Foundation to the second author.⁎ Corresponding author at: Department of Psychology

Negev, Beer-Sheva 84105, Israel.E-mail address: dorchins@post.bgu.ac.il (S. Regev).

http://dx.doi.org/10.1016/j.actpsy.2014.07.0060001-6918/© 2014 Elsevier B.V. All rights reserved.

a b s t r a c t

a r t i c l e i n f o

Article history:Received 26 January 2014Received in revised form 10 July 2014Accepted 12 July 2014Available online xxxx

PsycINFO classification:2340 Cognitive Processes

Task-switchingStroopPost-error slowingCognitive controlReaction-time

Post-error slowing (PES) has been shown to reflect a control failure due to automatic attentional capture by theerror. Here we aimed to assess whether PES also involves an increase in cognitive control. Using a cued-task-switching paradigm (Experiment 1) and a Stroop task (Experiment 2), the demand for top down control wasma-nipulated. In Experiment 1, one group received dimension cues indicating the relevant stimulus dimension(e.g., “number”) without specifying the response-category-to-key mapping, hence requiring considerable topdown control. Another group was shown mapping cues providing information regarding both the relevanttask identity and its category-to-key mapping (e.g., “one three”), requiring less top down control, and the lastgroup received both types of cues, intermixed. In Experiment 2, one group performed a pure incongruent Stroopcondition (name ink color of incongruent color names, high control demand), and another group received a pureneutral Stroop condition (name color patches, low control demand). In Experiment 2a, participants received thetwo conditions, intermixed. A larger PESwas observedwith dimension cues as comparedwithmapping cues, andwith incongruent Stroop stimuli as compared to neutral stimuli, but not when the conditions were intermixed.Thesefindings reveal that PES is influencedby the control demands that characterize the given block-wide exper-imental context and show that proactive cognitive control is involved in PES.

ung and Sharon Dvir for help ingrant no. 1939/12 of the Israel

, Ben-Gurion University of the

© 2014 Elsevier B.V. All rights reserved.

1. Introduction

Interestingly, after committing an error, normal individuals tend toslow down their performance on the next trial. This phenomenon iscalled “post-error slowing” (PES; Laming, 1979; Rabbitt, 1966; Smith& Brewer, 1995; see Danielmeier & Ullsperger, 2011, for a review) andis in the focus of behavioral studies on error monitoring and processing.

PES was first described by Rabbitt (1966) as reflecting either disrup-tion of regularity or precaution. To date, there is still an ongoing debateregarding the mechanism underlying this phenomenon. Some theoriesview transient control failure as the cause for both the occurrence of anerror and the slowing observed afterward. For example, Cheyne,Carriere, Solman, and Smilek (2011) have recently suggested that tran-sient failures of sustained attention impair task performance. The error,in turn, being a significant outcome causes an additional failure insustained attention that is being reflected in subsequent responseslowing. Another theory (Notebaert et al., 2009) that interprets PES asa result of an attentional lapse is the orienting account. This theoryviews PES as an outcome of an involuntary attentional shifting towards

a rare event (error). Such attention reorientation results in slowing. Insupport, the authors showed PES in a typical condition with infrequenterrors, but also showed response slowing that came after correct butinfrequent trials. In a similar vein, rare erroneous responses and rarecorrect responses led to an increase in P3 amplitude (Núñez-Castellar,Kühn, Fias, & Notebaert, 2010), an event-related brain potential (ERP)component associated with involuntary attentional capture by novelevents (P3a) and limited-capacity memory updating (P3b, see Polich,2007, for a review). Moreover, in Núñez-Castellar et al.'s (2010) study,PES size was positively correlated with P3 amplitude, but not with twoother ERP components: error-related negativity and feedback-relatednegativity, which arguably reflects error detection and evaluation of neg-ative feedback regarding outcomes, respectively (Falkenstein, Hoormann,Christ, & Hohnsbein, 2000; Hajcak, Moser, Holroyd, & Simons, 2006).Although some studies did find a positive correlation between PES andthe amplitude of error-related negativity (e.g.: Debener et al., 2005), thisevidence has been inconsistent (see, Van Veen & Carter, 2006, for areview). In sum, Núñez-Castellar et al.'s neurophysiological findingssupport the idea that PES is not associated with error monitoring butrather with attentional processing of unexpected novel events. This con-clusion was further supported by a functional MRI study demonstratingthat both error and novelty processingwere associatedwith brain activityin common cortical and subcortical regions (Wessel, Danielmeier,Morton, & Ullsperger, 2012).

The aforementioned theories seem to share the assumptionconcerning automatic/involuntary attention orientation as the main

11S. Regev, N. Meiran / Acta Psychologica 152 (2014) 10–18

cause. Supportive findings for the notion that automatic processing un-derlies PES show that shortening the response-to-stimulus interval ledto PES increase (Danielmeier & Ullsperger, 2011; Dudschig & Jentzsch,2009). These results may be interpreted as showing that PES does notdepend on capacity-limited adjustments that require sufficient time totake place, but is rather triggered automatically by the error and isthus shown to be larger with short intertrial intervals when the errorsignal is still strong and had not yet decayed. Nonetheless, these resultsare inconclusive since one could argue, for example, that the detectionof the error occupies the central processing bottleneck (Pashler &Johnston, 1989). This in turn leads to a postponement in subsequentresponse selection that must wait for the bottleneck to be freed(e.g., Houtman & Notebaert, 2013; Jentzsch & Dudschig, 2009).

The fact that PES may be influenced by involuntary attention orien-tation does not rule out the possibility that, in addition, it also reflectscontrolled processing. Specifically, cognitive control can be conceptual-ized as involving both reactive and proactive components (according toDual Mechanisms of Control framework, DMC; Braver, 2012; Braver,Gray, & Burgess, 2007). Reactive control operates in response to animperative event (such as an error) immediately after its occurrence.In contrast, proactive control is engaged in advance, based on goal-relevant information maintained active over a period of time. It has re-cently been proposed that these modes can interact (Ridderinkhof,Forstmann,Wylie, Burle, & van denWildenberg, 2010) such that anticipa-tory top-down control can proactively amplify reactive online control,contingent to performance difficulty, in order to prevent further errors.Accordingly, reactive control manages the recruitment of cognitivecontrol which depends on error detection. On the other hand, proactivecontrol mechanisms adjust control involvement in a proactive mannerand may thus serve to amplify post-error control adjustments. Belowwe detail the putative involvement of both reactive and proactive controlin the PES phenomenon.

An account that stresses the role of increased reactive control inpost-error processing is the conflict monitoring theory (Botvinick,Braver, Barch, Carter, & Cohen, 2001; Yeung, Botvinick, & Cohen,2004). This theory holds that slower post-error performance reflectsreactive implementation of cognitive control, elicited by the detectionof conflict. Specifically, error trials entail response conflict between co-activated representations of the correct and erroneous responses. Asystem responsible for detecting conflicts in information processingthen leads to a relatively more conservative and controlled behavioron subsequent trials.

There is also evidence suggesting that proactive control is also in-volved in PES. The evidence comes from studies showing that widecontext-level manipulations of control involvement or demand influ-ence PES size. For example, PES was shown to increase when accuracywas emphasized over speed and when punishment followed errors(Jentzsch & Leuthold, 2006; Riesel, Weinberg, Endrass, Kathmann, &Hajcak, 2012; Ullsperger & Szymanowski, 2004). Additionally, PES wasabolished with time on task (Boksem, Meijman, & Lorist, 2006; Lorist,Boksem, & Ridderinkhof, 2005), following sleep deprivation (Murphy,Richard, Masaki, & Segalowitz, 2006) and when participants believedthat their errors were caused by an external source and not by them-selves (Steinhauser & Kiesel, 2011), conditions believed to compromisecognitive control. Moreover, incentive given after the fatigue inductionled to PES reappearance (Boksem et al., 2006), presumably showing arecovery of control resources.

Individual difference studies also suggest that PES is influenced bycognitive control. Specifically, PES is large among individuals who arerelatively more accurate (Steinborn, Flehmig, Bratzke, & Schröter,2012), have higher academic achievements (Hirsh & Inzlicht, 2010),have higher cardiorespiratory fitness (Themanson & Hillman, 2006)and are more physically active (Themanson, Hillman, & Curtin, 2006).Nonetheless, these individual and group-difference studies are notcompletely conclusive since PES was not correlated with workingmemory capacity (Unsworth, Redick, Spillers, & Brewer, 2012) and

was shown to be numerically larger among old adults than amongyoung adults despite the known deterioration in executive functioningin aging (e.g., Band & Kok, 2000; Smith & Brewer, 1995).

Further support for the increased control position comes from stud-ies showing greater reduction in compatibility-related interference(indicating better resolution of interference) on trials following errorscompared to those following correct responses (De Bruijn, Miedl, &Bekkering, 2011; King, Korb, von Cramon, & Ullsperger, 2010). Howev-er, post-error reduction of interference was not found consistentlyacross experiments (Bombeke, Schouppe, Duthoo, & Notebaert, 2013;Carp & Compton, 2009), and even if reliable, this effect was found toact independently from PES effect (e.g., Bombeke et al., 2013).

Most relevant in the present context are the few studies whichmanipulated control demands. One study (Hogan, Vargha‐Khadem,Kirkham, & Baldeweg, 2005) reported larger PES and a correspondingdecrease in self-corrected errors for incompatible stimuli during afour-choice response task, than for compatible stimuli in a two-choiceresponse task. Another study manipulated cognitive demands in aflanker task by reversing stimulus–response mappings between blocks.This study found enhanced PES and reduced post-error accuracy in themore demanding switch blocks (Schroder, Moran, Infantolino, &Moser, 2013). This evidence further shows that PES increases withincreased task complexity.

Unfortunately, many of thefindingswhich presumably indicate con-trolled processing in PES are equally well explained by the involuntaryattention account. According to the increased control position,more de-manding conditions, incentives, higher ability, and conditions in whichcontrol is not compromised are associated with increased control ingeneral, resulting in more robust behavioral adjustments followingerrors. According to the reduced control position, when the task iscomplex, it requires more control resources than when it is lesscomplex. Errors, as unexpected events, grab the necessary resourcesneeded to execute the required task, resulting in poorer performance.An analogous point can be made with respect to incentives, individualdifferences, and conditions involving compromised control such asmental fatigue.

To conclude, while there is strong support for the involvement oflow-level attention-grabbing processes in PES, clear cut evidence thatPES also reflects top-down controlled processing is still lacking.

Thus, the aim of the present experimentswas to test the influence ofcontrol demands manipulated globally and locally on post-error pro-cessing. This was done by examining the influence of variables knownto involve strategic top-down control. In Experiment 1, we used acued task switching paradigm (see Kiesel et al., 2010; Meiran, 2010;Monsell, 2003; Vandierendonck, Liefooghe, & Verbruggen, 2010, for areview) in which participants are required to constantly switchbetween a number of simple tasks, with or without task-repetition.Experiment 2 employed the Stroop task (Stroop, 1935; see MacLeod,1991, for a review).

The task switching paradigm was chosen based on theoretical andmethodological considerations despite the fact that it is not commonlyused in PES research (but see e.g., Gupta, Kar, & Srinivasan, 2009;Jentzsch & Leuthold, 2006; Nee, Kastner, & Brown, 2011; Themansonet al., 2006). Specifically, since we were interested in the involvementof control processes, we needed a paradigm in which the degree ofcontrol demand can be easily manipulated. Indeed, Notebaert andcolleagues (Núñez-Castellar et al., 2010) had noted that errors made ina paradigm with stimuli affording only one response such as those usedin their studies (oddball task in Notebaert et al., 2009; four-choice color-discrimination task in Núñez-Castellar et al., 2010), are qualitativelydifferent from errors made in the tasks typically employed in PES studiesin which the stimuli afford several (sometimes competing) responses,such as the Stroop task and the flanker task (Eriksen & Eriksen, 1974).

Our manipulation of cognitive control demand was based on Mayrand Kliegl (2000) who employed in one of their experiments a cuedtask switching paradigm in which the tasks changed randomly and

12 S. Regev, N. Meiran / Acta Psychologica 152 (2014) 10–18

each trial began with the presentation of a task cue indicating whichtask to execute. In that experiment, they manipulated the type of taskcue which was used. Accordingly, in the present Experiment 1, thetask cues either indicated the relevant stimulus dimensionwithout indi-cating the response-category-to-key mapping (e.g., “Fill Fill”; hence-forth: “dimension cues”) or indicated the response-category-to-keymapping (e.g., “Empty Full”; henceforth “mapping cues”), hence imply-ing the relevant dimension as well (see Fig. 1). Mayr and Kliegl haveshown evidence for a lesser involvement of cognitive control withmapping cues than with dimension cues, seen for example in largertask-switching costs, especially given a short task preparation time.These authors argued that dimension cues requiremore control becausethey require the retrieval of the category-to-keymapping fromworkingmemory.

In Experiment 1, participants switched between four speeded choicetasks, and the relevant taskwas indicated at the beginning of the trial bymeans of a task cue. Some participants received just dimension cues forall four tasks. Others received just mapping cues. We used the cue-typemanipulation based on a series of unpublished experiments showinglarger PES with dimension cues than with mapping cues. We added agroup in which the two types of cues were intermixed such that 2taskswere consistently cuedwith dimension cues and 2 taskswere con-sistently cued with mapping cues. As explained below, this conditionpermitted us to characterize the type of control being involved in PES.The employment of group design was chosen over a blocked design inorder to avoid the need to deal with carryover effects.

We reasoned that if PES reflects bottom-up processing as suggestedby the decreased control theories, cue type is not expected to influencethe saliency of the errors, at least as long as error rates are comparableacross cue type conditions. Alternatively, if PES reflects top-down pro-cessing as suggested by the increased control position, then greaterPES is expected with dimension cues than with mapping cues, becausethe former requires more controlled processing to be performed. Fur-thermore, if PES is influenced by reactive control as suggested by theconflict monitoring theory, difficulty-related increase in PES due tocue type should be similar regardless of whether the cue type waschanged on a trial-by-trial basis (in the mixed-cue condition) or be-tween participants (in the conditions in which a given participant wasexposed to one cue type only). This is because reactive cognitive controlexpected to be elicited in response to errors, should differ depending onthe task demands in the erroneous trial (i.e., trial N − 1). A trial withmore challenging dimension cues (compared to easier mapping cues)is expected to increase post-error cognitive processing. On the otherhand, if PES involves proactive control in the sense that it is employedstrategically based on expected task difficulty, it should depend on theglobal (experiment-wide or block-wide) context but not on local (triallevel) difficulty (see also Los, 1996). Specifically, PES was not predicted

Fig. 1. The experimental paradigm. The first screen presents the task cues and the secondscreen adds the target stimulus. We used Hebrew words as task cues.

to differ across cue-types when they are intermixed and change unpre-dictably on a trial-by-trial basis. Rather, PES was predicted to differ onlyacross cue-types when they are not intermixed.

2. Experiment 1

2.1. Method

2.1.1. ParticipantsSeventy-eight students from Ben-Gurion University (69 females, 65

right-handed, mean age = 22.8 years, SD = 1.05, range = 20–26)participated in the experiment for partial course credit. All participantsreported having normal or corrected-to-normal vision with no historyof attention disorder or learning disabilities. Participants were assignedto six conditions (n = 13 per condition) of a 3 (dimension cues, map-ping cues, mixed cues) × 2 (with or without task repetition trials)between-subjects design according to the order of entering the study.The results of the mixed-cue condition are reported as Experiment 1a,because we decided to increase the sample size in this condition, as ex-plained below.

2.1.2. Instrument and stimuliThe experiment was programmed in E-Prime 1.0 (Schneider,

Eschman, & Zuccolotto, 2002), and presented on Pentium 4 computerswith 17-in.monitors. During each trial, participants categorized a targetstimulus according to one of four dimensions. The relevant dimensionwas determined by a task cue that appeared before (and during) thepresentation of the target stimulus. The task cue consisted of a pair ofHebrew words presented on the right and on the left of the upper sideof the screen in Arial font, each 0.8 × 1.5–3.5 cm. In the mapping-cuecondition, the two words were the response categories presented in aposition compatible with the corresponding response (e.g., in a taskthat required a right key press for a full object and a left key press foran empty object, task cues were “Full” on the right side of the screenand “Empty” on its left side). The dimension cues consisted of two iden-tical Hebrewwords (e.g., “Fill” on the left and right sides of the screen).The cues were presented inside two black 4 cm (height) × 9 cm (width)rectangles, on a white background. The blue target stimulus waspresented in the center of the display. It was composed of 1 or 3 circles(or cloud-shapes) that were empty (or full) andwere crossed by a 7 cmvertical (or horizontal) line.

2.1.3. ProcedureParticipants provided informed consent and were seated at approx-

imately 60 cm from the computer screen. At the beginning of the exper-iment, participants were given on-screen instructions explaining thefour tasks and the meaning of their cues. Participants were told that atthe beginning of each trial, the initial display would consist of a cuepresented on both sides of the screen to indicate the relevant task.This cue is then followed by a target stimulus for which the relevanttask rules need to be applied. The first task was Quantity, indicated bythe dimension cue “Quantity Quantity” or by the mapping cue “OneThree”. In this task, participants were required to determine whetherthe target stimulus is made of one or three objects and respond accord-ingly. The second task was Fill, with the dimension cue “Fill Fill” or themapping cue “Empty Full” to indicate to participants that they need todetermine if the target stimulus is empty or full. The third task wasFrame and had the dimension cue “Frame Frame” or the mapping cue“Circle Empty”, for which participants were to determine if the shapeof the object was circular or cloud-like. Finally, the fourth task wasLine Orientation, and had the dimension cue “Line Orientation LineOrientation” or the mapping cue “Horizontal Vertical”. In this task,participants determined if the line crossing the object was horizontalor vertical. Hence, in each trial, participants performed one out of foursimple tasks, as indicated by a task cue that specified the relevantdimension in the target stimulus.

13S. Regev, N. Meiran / Acta Psychologica 152 (2014) 10–18

Participants were instructed to respond to the target stimulus withtheir two index fingers. The QWERTY keyboard keys P (right) and E(left) were used for responses. In each cue type condition, half of theparticipants had both switch and repeat trials across task blocks, whilethe other half had only switch trials. This was conducted in order toevaluate whether the presence of repetition trials would modulateany of the effects. Repeat trials refer to trials for which the cue task ofthe previous trial (trial N − 1) is presented again on the current trial(trialN), requiring participants to perform the same task on two consec-utive trials. Switch trials refer to trials for which the cue task has alter-nated from one trial to the next. Each participant was required toexecute onepractice block consistingof 20 trials and three experimentalblocks; each consisted of 240 trials. Thus, excluding the practice, therewere 720 trials for each participant.

In an effort to keep participants alert, a mandatory 5-s break wasgiven between experimental blocks, after which it persisted until theparticipant pressed a key on the keyboard. The importance of bothspeed and accuracy was equally emphasized.1

Participants in the dimension cues condition received instructionsemphasizing the need to memorize category–response mapping of thetasks rules in order to perform correctly. Participants in the mappingcues condition were explicitly told that there is no need for them tomemorize the task rules since these are provided by the task cues.Participants in the mixed-cue condition received both types ofinstructions, which varied as a function of the cue assigned to eachtask. In this condition, two tasks had dimension cues and the othertwo had mapping cues. The assignment of tasks to cue types wascounterbalanced across participants in that group.

A trial consisted of a task cue presented for 800 ms, followed byadding the target stimulus until a response was given. Errors werefollowed by a 400-ms beep tone from the computer monitor.

2.2. Results

2.2.1. Data exclusionData from one participant were excluded from the analysis due to

equipment failure. Moreover, the analysis was conducted only on par-ticipants with at least four trials in each cell, leaving 25 participants(out of 26) in the dimension cues condition and 24 participants in themapping cues condition. As outliers, we excluded RTs two standard de-viations from the mean for each participant, separately for post-errorand post-correct trials (a total of 4.7% of experimental trials). For RTanalyses, errors were also excluded. Analyses performed on conditionswith and without repetition trials were based only on switch trials inorder to equate these conditions. We used α = .05 in all statisticaltests reported in this paper.2

2.2.2. RTRT was analyzed using a three-way mixed Analysis of Variance

(ANOVA) with Cue Type (dimension, mapping) and Repetition (condi-tion including both switch and repeat trials, condition including onlyswitch trials) as between-subjects independent variable and PreviousAccuracy (correct, error) as a within-subjects independent variable.

There was a significant main effect of Cue Type, F(1,45) = 7.90,MSE = 257,393.33, p = .007, η2

p= .15, and a significant main effect

of Previous Accuracy, F(1, 45) = 73.68, MSE = 101,396.28, p b .0001,η2

p= .62, indicating a substantial PES effect due to slower performance

after errors (M = 1334 ms, SD = 80 ms) compared to correct trials(M = 780 ms, SD = 32 ms). There was also a significant interaction

1 We chose not to stress the importance of speed over accuracy (such as byimplementing response deadlines), even though that could have served as amethod of in-creasing error rate. We were concerned that it would lead to a reduction in PES effect(e.g., Ganushchak & Schiller, 2006; Ullsperger & Szymanowski, 2004), possibly becausecorrecting errors become less important.

2 An analysis based on how Dutilh et al. (2012) redefined PES indicated similar resultsas those reported below.

between Cue Type and Previous Accuracy, F(1, 45) = 4.44, MSE =101,396.28, p = .041, η2

p= .09. Follow-up planned contrasts indicated

that the simple main effect of PES was significant in both groups(p b .0001). Importantly, PES was larger with dimension cues (M =689 ms, SD = 66 ms; F(1, 45) = 58.50) than with mapping cues(M = 417 ms, SD = 68 ms; F(1, 45) = 20.50) (see Fig. 2). All othereffects were non-significant.

2.2.3. Proportion of errors (PE)An analogous ANOVA revealed a significant main effect of Repeti-

tion, F(1, 45)= 5.09,MSE= 0.002, p= .029,η2p= .10, and a significant

interaction between Previous Accuracy and Repetition, F(1, 45) = 4.71,MSE= 0.002, p= .035, η2

p= .09. Follow-up planned contrasts demon-

strated a significant post-error accuracy decrease (from .062 to . 035,F(1, 45) = 5.12, p = .029) for the condition that included both switchand repeat trials, but a non-significant effect when there were onlyswitch trials (from .023 to .032, F b 1.00). All other effects were non-significant.

3. Experiment 1a

Preliminary analyses of the mixed-cue condition revealed no signif-icant influence of current cue type or previous cue type on PES size.However, endorsing the null hypothesis may be associated with aType II error, and as such may be premature given the poor statisticalpower (equaling 1-Type II error) associated with a small sample size.Thus we decided to increase the statistical power of the within-subjects ANOVAs in the mixed-cue condition by nearly doubling thesample size. Specifically, we added twenty-four similar participants tothe existing twenty-six participants, resulting in N = 50 (34 females,43 right-handed, mean age = 24.1 years, SD= 1.58, range = 21–28).

3.1. Results

Thenumber of errors committed by each participantwas insufficientto examine the joint influence of Previous Cue Type and Current CueType in a single ANOVA, and hence we conducted 2 separate ANOVAs.

3.1.1. Data exclusionData from one participant were excluded due to error rate greater

than 3 SDs above the general mean (26%). In addition, the analyseswere conducted only on participants with at least four trials in eachcell, leaving 40 participants in the analysis with Previous Cue Type,and 44 participants in the analysis with Current Cue Type. As outliers,we excluded RTs two standard deviations from themean in each condi-tion for each participant (an averaged total of 4.8% of trials). For RT anal-yses, errors were also excluded.

3.1.2. Previous cue typeA two-way ANOVA was performed with Previous Cue Type

(dimension, mapping) and Previous Accuracy (correct, error) aswithin-subjects independent variables. For RTs, the analysis revealed amain effect of Previous Accuracy, F(1, 39) = 62.08,MSE = 118,261.17,p b .001, η2

p= .61. Performance was slower after errors (M= 1187 ms,

SD= 67 ms) than after correct responses (M = 758 ms, SD = 31 ms),indicating a significant PES effect. The main effect of Previous Cue Typewas nearly significant, F(1, 39) = 3.33, MSE = 29,649.51, p = .08,η2

p= .08, with longer RTs when previous trials had mapping cues

(M = 997 ms, SD = 50 ms) than when they had dimension cues(M= 948 ms, SD= 43ms). Importantly, the interaction between Pre-vious Accuracy and Previous Cue Typewas not significant (F b 1.00) andits trend was actually opposite to that seen in Experiment 1. Mean PESwas 403 ms (SD = 30 ms) when previous trials had dimension cuesand 453 ms (SD = 47 ms) when previous trials had mapping cues. Asimilar analysis of PE as a dependent variable revealed that the only sig-nificant effect was for Previous Accuracy, with a higher PE after correct

Fig. 2. Mean reaction times according to current cue type and previous accuracy — Experiment 1 (left) and Experiment 1a (right). Error bars represent the 95% repeated-measureconfidence intervals (Hollands & Jarmasz, 2010).

14 S. Regev, N. Meiran / Acta Psychologica 152 (2014) 10–18

trials (M= .057) than after errors (M= .045), demonstrating that thePES effect revealed in this analysis may possibly reflect a speed–accura-cy tradeoff. All other effects were non-significant (all ps N .10).

3.1.3. Current cue typeA two-way ANOVA was performed with Current Cue Type

(dimension, mapping) and Previous Accuracy (correct, error) as within-subjects independent variables. The main effect of Previous Accuracywas significant, F(1, 43) = 71.49, MSE = 116,842.25, p b .001, η2

p=

.62, with slower RTs after errors (M= 1206 ms, SD= 67 ms) than aftercorrect trials (M = 770 ms, SD = 29 ms). The main effect of CurrentCue Type was also significant, F(1, 43) = 10.10, MSE= 21,997.69, p =.003, η2

p= .19. This effect indicated that trials with dimension cues had

a longer response time (M=1024ms, SD=48ms) than trialswithmap-ping cues (M=952ms, SD=44ms). In otherwords, the level of difficul-ty was not the same between the two types of cues when they changedon a trial-by-trial basis. Nevertheless, PES effect was not significantly in-fluenced by this differential task difficulty as evident from the absenceof a significant interaction between Current Cue Type and Previous Accu-racy (p N .20). Mean PES was 457 ms (SD = 39 ms) when current trialshad dimension cues and 413 ms (SD = 41 ms) when current trials hadmapping cues. Note that the direction of this effect is similar to that in Ex-periment 1, but that it is also considerably attenuated: PES difference of44ms in Experiment 1a as comparedwith 272ms in Experiment 1. A sim-ilar analysis of PE as a dependent variable revealed a main effect of CueType, F(1, 43) = 12.15, MSE = 0.002, p = .001, η2

p= .22, with a higher

PE for trialswith dimension cues (M= .063) than for trialswithmappingcues (M= .038). This also indicates that trials with dimension cues wereharder to perform than trials with mapping cues. Despite that, Cue Typedid not interact significantly with Previous Accuracy (F b 1.00), as allother effects were non-significant (all ps N .1; See Fig. 2).

4. Discussion

We demonstrated in Experiments 1 and 1a that cognitive demandswere higher for trials with dimension cues than for trials with mappingcues, and this was true whether cue types differed between contexts(i.e., between subjects) or on a trial-by-trial basis. However, cognitivedemand related-increase in PES was significant and quite large onlywhen demands altered between different experiment-wide contexts.When the conditions changed unpredictably from one trial to another,the effect was either reversed (previous cue-type) or was considerablysmaller and non-significant (current cue-type).3

3 Of note, PES effect (averaged across cue-types) in Experiment 1a (M=436, SD=38)was numerically smaller than in Experiment 1 (M=554, SD=48). The conditions in thetwo experiments were not strictly comparable for at least two reasons. One is perhaps theoverall difficulty in the block changes if there are easy trials included in it. Moreover, in Ex-periment 1 there were only dimension cues followed by dimension cues and mappingcues followed by mapping cues, and it did not include conditions in which cues were dif-ferent in the current and previous trial.

Therefore, we conclude that the block/experiment level cognitivedemands increase readiness to engage in post-error cognitive process-ing and only (or predominantly) in a proactive manner, when thesetask demands are determined in advance with regard to the context inwhich one operates.

5. Experiment 2

Previous PES-related research has mainly employed RT tasks thatwere considerably simpler than the task switching paradigm that weemployed in Experiment 1. Hence, it is possible that the cognitivecontrol-related differences in PES effect found in that experiment can-not inform previous research. Moreover, in Experiment 1, errors werefollowed by an auditory feedback, while correct responses were not.This may have led to an additional processing after errors that couldhave had contributed to the relatively large PES effect that was found(and the group differences therein). Experiment 2 addressed both ofthese shortcomings. First, we employed a key-press version of theStroop task that has been used quite extensively in previous researchon PES (Carp & Compton, 2009; Gehring & Fencsik, 2001; Hajcak &Simons, 2002, 2008; Hirsh & Inzlicht, 2010; Kerns et al., 2004; Larson,Fair, Good, & Baldwin, 2010; Unsworth et al., 2012). Moreover, errorswere not followed by any feedback. In order to manipulate theexperiment-wide control demand, we compared two conditions(between groups). In one condition, all the trials were incongruent(the relevant ink color was different from the color name stimulus),thus requiring a relatively high degree of cognitive control. In theother condition, all the trials were neutral trials (requiring a responseto the ink color inwhich alphanumeric strings such as “#$%” appeared),requiring much less cognitive control.

5.1. Methods

5.1.1. ParticipantsTwenty-four students from Ben-Gurion University (15 females; 20

right-handed; mean age = 24.0 years, SD= 1.6, range = 21–28) par-ticipated in the experiment for payment (20 NIS, approximately $6).All participants were native Hebrew speakers, reported having normalor corrected-to-normal vision with no history of attention disorder orlearning disabilities. Participants were randomly assigned to either apure incongruent or a pure neutral condition (n = 12 per condition).

5.1.2. Instrument and stimuliThe experiment was programmed in E-Prime 2.0 (Schneider et al.,

2002), and presented on Pentium 4 computers with 17-in. monitors.In the pure incongruent condition, stimuli were four Hebrew colorwords (equivalent to RED, GREEN, BLUE, YELLOW), each presented inthe three colors that did not match its meaning (e.g.: the word BLUEwas printed in red color). The words were shown in bold 40-pointCourier New font. They were 7 cm (width) × 2 cm (height) presented

15S. Regev, N. Meiran / Acta Psychologica 152 (2014) 10–18

in the middle of the screen, on a white background. In the neutral con-dition, stimuli were short strings of alphanumeric characters colored inred, green, blue or yellow (e.g.: “#$%” printed in red). The neutral stim-ulus was 5 cm (width) × 2 cm (height) presented in the middle of thescreen on a white background.

5.1.3. ProcedureA similar procedure was used as in Experiment 1. However, no dem-

onstration of the task was given and there was no practice block due tothe simplicity of the task. Participants were instructed to respond to thetarget stimulus with the middle and index fingers of their right and lefthands, pressing the keys X, C, N and M, on the QWERTY keyboard indi-cating green, blue, yellow and red, respectively. The experiment wascomposed of seven experimental blocks; each consisted of 96 trials,with a short break after every block. Thus, there were 672 trials intotal for each participant.

A trial began with a 350-ms presentation of a fixation cross(0.8 × 0.8 cm) at the center of the screen, after which the target stimu-lus appeared on the screen at the same location until a response wasgiven. A blank screen served as a response–stimulus interval (RSI) andwas presented for 500-ms, between each key press response to theonset of the fixation cross for the next trial.

5.2. Results

5.2.1. Data exclusionData from one participant in the pure-incongruent condition were

excluded from the analysis due to error rate greater than 3 SDs abovethe general mean (22%). All remaining participants had at least four tri-als in each cell. As outliers, we excluded RTs two standard deviationsfrom the mean for each participant, separately for post-error and post-correct trials (a total of 4.5% of trials). For RT analyses, errors were alsoexcluded.

5.2.2. RTRT was analyzed using a two-way mixed ANOVA with Condition

(pure-neutral, pure-incongruent) as a between-subjects independentvariable and Previous Accuracy (correct, error) as a within-subjectsindependent variable.

No significant main effect for Condition was seen (F b 1.00). None-theless, mean RT was 718 ms (SD= 75 ms) and 825 ms (SD = 75 ms)in the groups receiving neutral and incongruent Stroop stimuli, respec-tively. As expected, a significant main effect was found for PreviousAccuracy, F(1, 21) = 12.11,MSE = 16,493.01, p = .002, η2

p= .37, indi-

cating a substantial PES effect, due to slower performance after errors(M = 832 ms, SD = 71 ms) compared to correct trials (M = 700 ms,SD = 43 ms). Critically, there was a significant interaction betweenGroup and Previous Accuracy, F(1, 21) = 6.00, MSE = 16,493.01, p =.02, η2

p= .22. Follow-up planned contrasts indicated that the simple

main effect of PES was significant in the pure-incongruent group(M = 225 ms, SD = 40 ms), F(1, 21) = 16.85, MSE = 16,493.01, p =

Fig. 3.Mean reaction times according to condition and previous accuracy — Experiment 2 (lefintervals (Hollands & Jarmasz, 2010).

.0005, but was not significant in the pure-neutral group (M = 39 ms,SD= 39 ms; F b 1.00; see Fig. 3).

One possible difference between the groups refers to the type of er-rors that theymade. Specifically, the RT distribution of wrong responsesto incongruent stimulimay reflect amixture of two populations of trials,those associated with a correct word response (i.e., errors due tostimulus-based triggering of the competing response) and those associ-ated with erroneous color responses. However, the RT distribution ofwrong responses to neutral stimuli cannot result from the formersource of errors. In order to ensure that the results do not reflect theinfluence of error-type, we conducted an additional analysis in whichwe excluded errors in the pure-incongruent condition for which theerroneous response was the required response according to theconflicting word. This exclusion left 56% of the error trials in the pure-incongruent condition. Importantly, the numerical PES effect in thepure-incongruent group was not reduced. If anything, it was slightlyincreased to 229 ms (SD = 65 ms). This shows that the enlarged PESin this condition was not biased upward by a different type of errors.

5.2.3. Proportion of errors (PE)An analogous ANOVA revealed only a significant main effect for

Previous Accuracy, F(1, 21) = 5.32, MSE = 0.014, p = .031, η2p= .20,

demonstrating a significant post-error accuracy decrease (of .03, from.95 to .92). The main effect of Condition was not significant, (F b 1.00),with PE of .059 and .095 in the groups receivingneutral and incongruentStroop stimuli, respectively. Likewise, no significant interaction wasfound between Previous Accuracy and Condition in PE (F = 1.51, n.s.).

6. Experiment 2a

In this experiment, we randomly intermixed the two Stroop stimu-lus types in a block. This design allowed us to examine a change inPES as a function of Stroop condition when it varies randomly. Theaim was to see if the differences in PES size between stimuli reflectstimulus differences and/or trial level fluctuations as opposed toblock/experiment context differences. Specifically, we used two differ-ent conditions: incongruent and neutral. The two conditions were ran-domly intermixed in the same experimental session. The incongruentcondition included the twelve possible word–color pairs taken fromthe pure-incongruent condition of Experiment 2. The neutral conditionincluded the four alphanumeric stringswhichwere pairedwith the fourcolors on random basis.

6.1. Method

6.1.1. ParticipantsTwenty-two students from Ben-Gurion University (14 females; 21

right-handed; mean age = 23.8 years, SD= 1.0, range = 22–26) par-ticipated in the experiment for payment (20 NIS, approximately $6).All participants were native Hebrew speakers who reported having

t) and Experiment 2a (right). Error bars represent the 95% repeated-measure confidence

4 In contrast to Experiments 1 and 1a in which PES size differed, PES effects in Experi-ments 2 (M = 132 ms, SD = 28 ms) and 2a (M = 138 ms, SD = 26 ms) were similar.As noted in Experiment 1, the conditions are not strictly comparable, and perhaps mixingeasy trials in a block has a different influence for cue-mixing and for the mixing of Stroopconditions.

16 S. Regev, N. Meiran / Acta Psychologica 152 (2014) 10–18

normal or corrected-to-normal visionwith nohistory of attention disor-der or learning disabilities.

6.1.2. ProcedureThe stimuli from the pure-neutral condition were three times more

likely to appear than each of the incongruent stimuli. Hence, each of theseven experimental blocks consisted of 96 trials, half of which were in-congruent stimuli and half of which were neutral stimuli. Thematchingin frequency of appearance between neutral and incongruent stimuliwas done to rule out different probabilities of Stroop categories as anartifact.

6.2. Results

The number of errors was insufficient to examine the joint influenceof Previous Stroop Condition and Current Stroop Condition on PES effectin a single ANOVA, and hence we conducted 2 separate ANOVAs.

6.2.1. Data exclusionThe two following analyses were conducted only on participants

with at least four trials in each cell, which led to the exclusion of two dif-ferent participants (out of 22) in each analysis. As outliers, we excludedRTs two standard deviations above the mean in each condition for eachparticipant (an averaged total of 4.6% of trials). For RT analyses, errorswere also excluded.

6.2.2. Previous Stroop conditionA two-way ANOVA was performed with Previous Stroop Condition

(incongruent, neutral) and Previous Accuracy (correct, error) aswithin-subjects independent variables. For RTs, the analysis revealeda main effect of the Previous Accuracy, F(1, 19) = 12.7, MSE =13,475.35, p = .002, η2

p= .40. RTs were slower after errors (M =

812 ms, SD = 52 ms) than after correct responses (M = 720 ms,SD = 35 ms). However, the main effect of Previous Stroop Conditionwas not significant nor did this variable interact with PES effect(F b 1.00). Mean PES was 70 ms (SD = 20 ms) and 115 ms (SD =27 ms) in trials following neutral and incongruent Stroop stimuli,respectively. Note that the direction of this difference is the same as inExperiment 2, but that it is considerably attenuated (45ms as comparedwith 186 ms in Experiment 2). A similar analysis of PE as a dependentvariable revealed neither main effects nor interactions (all ps N .1).

6.2.3. Current Stroop conditionA two-way ANOVA was performed with Current Stroop Condition

(incongruent, neutral) and Previous Accuracy (correct, error) aswithin-subjects independent variables. The main effect of PreviousAccuracy was significant, F(1, 19) = 11.7, MSE = 32,910.35, p = .003,η2

p= .38, with slower RTs after errors (M = 853 ms, SD = 69 ms)

compared to correct trials (M = 715 ms, SD = 34 ms). The maineffect of Current Stroop Condition was also significant, F(1, 19) = 7.11,MSE = 27,446.43, p = .015, η2

p= .27. RTs were 834 ms (SD = 65 ms)

and 735 ms (SD = 39 ms) for incongruent and neutral stimuli, respec-tively, indicating a Stroop incongruency effect. Importantly, CurrentStroop Condition × Previous Accuracy interaction was non-significant(F b 1.00). Mean PES was 122 ms (SD = 25 ms) and 155 ms (SD =53 ms) in neutral and incongruent trials, respectively (see Fig. 3).Again, the trend is similar to that in Experiment 2, but considerably at-tenuated (33 ms as compared with 186 ms in Experiment 2). A similaranalysis of PE as a dependent variable revealed neither main effects norinteractions (all ps N .10).

7. Discussion

The findings from Experiments 2 and 2a demonstrate a relativelylarge PES effect in the pure-incongruent condition that present conflict-ing color words, while a much smaller and statistically non-significant

PES effect was seen for the pure-neutral condition that used non-conflicting strings of alphanumeric characters. Of importance, PESdifference between incongruent and neutral trials was considerablysmaller and non-significant when stimulus type varied between trials(Experiment 2a).4

A somewhat peculiar finding in Experiment 2 was the lack of signif-icant difference between conditions involving incongruent and neutraltrials. However, the effect was numerically larger in that experimentas compared with Experiment 2a in which it was significant. We attri-bute the lack of significance in Experiment 2 to the poor statisticalpower to detect this particular effect as a result of the between-subjects design that was employed in Experiment 2 as compared tothe within-subject design in Experiment 2a. Note, however, that this isless of an issue concerning the predicted interaction between Conditionand Previous Accuracy because this interaction employs the within-subjects error term in both cases. According to an alternative accountsuggested by an anonymous reviewer, in Experiment 1 (between-subject design), the response threshold was higher with dimensioncues than with mapping cues. This account is consistent with thecomparable error rates combined with a difference in RT between theconditions. Moreover, a high response threshold may indicate that er-rors become more important, a fact that could explain the elevatedPES with dimension cues. In contrast, Experiment 1a (within-subjectdesign) varied Cue-Type within blocks and had thus eliminated theaforementioned response threshold differences and accordingly, alsothe PES differences. A similar pattern was found in Experiments 2 and2a. Namely, the difference between incongruent and neutral stimuli inthe between-subjects design was numerically larger than the equiva-lent difference in the within-subject design. Hence, the conditionrelated-increase in PES can be attributed to differences in the general re-sponse threshold. Nonetheless, the alternative account is probablywrong as can be seen when comparing the neutral condition in Experi-ment 2 (PE= .05 and .07 after correct and error responses, respectively)to the current-neutral condition in Experiment 2b (PE = .05 and .10,respectively). As can be seen in Fig. 3, RT was comparable in the twoconditions. Importantly, the error rates suggest either a comparableresponse criterion in the two conditions or a higher response criterionin the all-neutral condition (Experiment 2). If the alternative accountwere correct, PES should have been larger (or equal) in the neutralcondition of Experiment 2 than in Experiment 2a. The results are in anopposite-to-predicted direction, given the much smaller PES effect inthe neutral condition of Experiment 2 than in the current-neutral condi-tion of Experiment 2a (see Fig. 3).

In summary, Experiment 2 shows that the trend found in Experi-ment 1 is not restricted to the (less frequently used in that context)task-switching paradigm. Moreover, the fact that the trend was obtain-ed without error feedback rules out the alternative explanation that theinvolvement of top down control in PES depends on error-feedbackprocessing.

8. General discussion

Previous work has indicated the involvement of bottom-up atten-tional orientation (e.g.: Notebaert et al., 2009) and top-down control(e.g.: Botvinick et al., 2001) in the PES phenomenon. Whereas the ex-tant evidence for bottom-up factors is relatively compelling, evidencefor top-down involvement is more equivocal. The aim of the presentstudy was to examine the involvement of top-down control processingin PES and to begin clarifying its nature. Specifically, we asked whetherPES reflects bottom-up processing of the error that are automatic in

17S. Regev, N. Meiran / Acta Psychologica 152 (2014) 10–18

nature leading to a failure in cognitive control or whether PES also re-flects top-down controlled processing evoked in response to the error.Moreover, we wanted to determine if top-down control influences arereactive or proactive in nature.

To this end, we examined the influence of varying task demands onperformance after errors. In Experiment 1, we used a task switchingparadigm and compared conditions differing in the type of task-cuesbeing used: dimension cues (more demanding) vs. mapping cues (lessdemanding). The main result shows a larger PES with dimension cuesthan with mapping cues. Importantly, this difference was statisticallyabolished when the two different cues were intermixed. This resultindicates that PES is influenced by the control demands of the experi-mental context rather than by the local demands of a given trial. Thus,the result indicated that proactive cognitive control is involved in PES.

Of note, PES differences were obtained despite similar accuracylevels seen in the two groups. This fact rules out differential novelty ofthe errors as the alternative explanation. Specifically, involuntarycapture of attention due to error saliency should depend on errorfrequency (Notebaert et al., 2009). In addition, the finding that PESwas not significantly influenced by current cue type in themixed condi-tion rules out the notion that, once the core task is more demanding, itbecomes more sensitive to attention grabbing by the error. Namely,when we compared cue types between groups, PES was measuredwith different cue types, and the fact that one cue type was moredifficult could have resulted in greater susceptibility to interference.Nonetheless, PES was only slightly and non-significantly influenced bycurrent cue type when cue types were intermixed rules. This resultrules out the aforementioned possibility.

Experiment 2 had similar results, this time in a key-press version ofthe Stroop task, which has been extensively used in previous studies onthe PES phenomenon (Carp & Compton, 2009; Gehring & Fencsik, 2001;Hajcak & Simons, 2002, 2008; Hirsh & Inzlicht, 2010; Kerns et al., 2004;Larson et al., 2010; Unsworth et al., 2012). Namely, PESwas larger in themore demanding incongruent condition than in the less demandingneutral condition. Moreover, PES was statistically comparable in incon-gruent and neutral trials when theywere intermixed. Thus, we can con-cludewith reasonable confidence that our conclusions are not restrictedto the task switching paradigm.

The DMC framework (Braver, 2012; Braver et al., 2007) differenti-ates between two types of top-down cognitive control processes, reac-tive and proactive. We used this theoretical framework to understandthe involvement of cognitive control in PES. From this perspective, thecognitive system shifts bias between reactive and proactive controlmodes, based on characteristics of the situation (and the person), suchas task demands for WM (Braver et al., 2007). Moreover, reactivecontrol may interact with proactive control, such that reactive controlis adjusted by proactive control when situation demands require so(Ridderinkhof et al., 2010).

We suggest that PES obviously reflects a reaction to a special event(error). Our results show that this reaction depends onproactive controlsince PES depended on the experiment-wide cue conditions rather thanthe current cue or the cue in the trial inwhich the error occurred. In thissense, we interpret our findings as being in line with the duality sug-gested by Ridderinkhof et al. (2010) between online and anticipatorycontrol processes. In this view, anticipatory control mechanisms areable to proactively modulate online adjustments activated in responseto an error.

However, while we showed that PES involves top-down control, wedid not rule out the possibility that it additionally involves post-noveltyprocesses. It is possible that errors elicit several different processes inthe brain. For example, PES can reflect both controlled processing andattentional capture by the error. Support for this position comes fromseveral studies. Specifically, Jentzsch and Dudschig (2009) claimedthat the error first activates a detection system that occupies limitedcentral resources creating a bottleneck that interferes with subsequentprocessing. With sufficient time, this process is then followed by

another process of criterion adjustment resulting in a more cautiousbehavior. Hochman andMeiran (2005) suggested the existence of auto-matically generated error-related processing, being triggered in spite ofattempts to suppress error correction. They further reported the addi-tional involvement of limited capacity error processing that could bepostponed due to ongoing central executive processes.

The precise nature of the cognitive control processes that are activat-ed by an error is still unclear. Our results, together with those of previ-ous studies mentioned above, may pose a challenge to the popularnotion that PES reflects response criterion adjustment (Botvinick et al.,2001). This is especially true regarding Experiment 2 in which therewas both a significant PES and post-error accuracy decrease (ratherthan an increase, that would have been predicted based on responsecriterion adjustment). A more promising direction may be based onBrown, Reynolds, and Braver (2007), who suggested that control pro-cessing involves shifts in bias between exploitation and exploration.Specifically, these authors suggested that task switching results in ashift towards exploration, which is characterized by slower and error-prone processing. A similar shift in bias might happen in response toerrors.

In sum,we show that PES found in task-switching paradigms as wellas in the Stroop task (and quite likely, inmany other tasks) reflects stra-tegic controlled processing that is both reactive and proactive in nature.Specifically, post-error processing is reactive in the sense that it occurslocally in response to the error and proactive in the sense that it is setin a top-downmanner as a function of the control demands that charac-terize the wide (block/experiment-wide) context.

References

Band, G. P., & Kok, A. (2000). Age effects on response monitoring in a mental-rotationtask. Biological Psychology, 51, 201–221, http://dx.doi.org/10.1016/S0301-0511(99)00038-1.

Boksem,M.A., Meijman, T. F., & Lorist, M. M. (2006). Mental fatigue, motivation and actionmonitoring. Biological Psychology, 72, 123–132, http://dx.doi.org/10.1016/j.biopsycho.2005.08.007.

Bombeke, K., Schouppe, N., Duthoo, W., & Notebaert, W. (2013). The effect of alcohol andplacebo on post-error adjustments. Frontiers in Human Neuroscience, 7, http://dx.doi.org/10.3389/fnhum.2013.00003.

Botvinick, M. M., Braver, T. S., Barch, D.M., Carter, C. S., & Cohen, J.D. (2001). Conflict mon-itoring and cognitive control. Psychological Review, 108, 624–652, http://dx.doi.org/10.1037/0033-295X.108.3.624.

Braver, T. S. (2012). The variable nature of cognitive control: A dual-mechanisms frame-work. Trends in Cognitive Sciences, 16, 106–113, http://dx.doi.org/10.1016/j.tics.2011.12.010.

Braver, T. S., Gray, J. R., & Burgess, G. C. (2007). Explaining the many varieties of workingmemory variation: Dual mechanisms of cognitive control. In A.R. A. Conway, C.Jarrold, M. J. Kane, A. Miyake, & J. N. Towse (Eds.), Variation in working memory(pp. 76–106). New York, NY: Oxford University Press.

Brown, J. W., Reynolds, J. R., & Braver, T. S. (2007). A computational model of fractionatedconflict-control mechanisms in task-switching. Cognitive Psychology, 55, 37–85,http://dx.doi.org/10.1016/j.cogpsych.2006.09.005.

Carp, J., & Compton, R. J. (2009). Alpha power is influenced by performance errors.Psychophysiology, 46, 336–343, http://dx.doi.org/10.1111/j.1469-8986.2008.00773.x.

Cheyne, J. A., Carriere, J. S., Solman, G. J., & Smilek, D. (2011). Challenge and error: Criticalevents and attention-related errors. Cognition, 121, 437–446, http://dx.doi.org/10.1016/j.cognition.2011.07.010.

Danielmeier, C., & Ullsperger, M. (2011). Post-error adjustments. Frontiers in Psychology, 2,233, http://dx.doi.org/10.3389/fpsyg.2011.00233.

De Bruijn, E. R., Miedl, S. F., & Bekkering, H. (2011). How a co-actor's task affects monitor-ing of own errors: Evidence from a social event-related potential study. ExperimentalBrain Research, 211, 397–404, http://dx.doi.org/10.1007/s00221-011-2615-1.

Debener, S., Ullsperger, M., Siegel, M., Fiehler, K., Von Cramon, D. Y., & Engel, A. K. (2005).Trial-by-trial coupling of concurrent electroencephalogram and functional magneticresonance imaging identifies the dynamics of performance monitoring. The Journalof Neuroscience, 25, 11730–11737, http://dx.doi.org/10.1523/JNEUROSCI.3286-05.2005.

Dudschig, C., & Jentzsch, I. (2009). Speeding before and slowing after errors: Is it all juststrategy? Brain Research, 1296, 56–62, http://dx.doi.org/10.1016/j.brainres.2009.08.009.

Dutilh, G., van Ravenzwaaij, D., Nieuwenhuis, S., van der Maas, H. L., Forstmann, B. U., &Wagenmakers, E. J. (2012). How to measure post-error slowing: A confound and asimple solution. Journal of Mathematical Psychology, 56, 208–216, http://dx.doi.org/10.1016/j.jmp.2012.04.001.

Eriksen, B.A., & Eriksen, C. W. (1974). Effects of noise letters upon the identification of atarget letter in a nonsearch task. Perception & Psychophysics, 16(1), 143–149, http://dx.doi.org/10.3758/BF03203267.

18 S. Regev, N. Meiran / Acta Psychologica 152 (2014) 10–18

Falkenstein, M., Hoormann, J., Christ, S., & Hohnsbein, J. (2000). ERP components onreaction errors and their functional significance: A tutorial. Biological Psychology, 51,87–107, http://dx.doi.org/10.1016/S0301-0511(99).

Ganushchak, L. Y., & Schiller, N. O. (2006). Effects of time pressure on verbal self-monitoring: An ERP study. Brain Research, 1125, 104–115, http://dx.doi.org/10.1016/j.brainres.2006.09.096.

Gehring, W. J., & Fencsik, D. E. (2001). Functions of the medial frontal cortex in theprocessing of conflict and errors. The Journal of Neuroscience, 21, 9430–9437.

Gupta, R., Kar, B. R., & Srinivasan, N. (2009). Development of task switching and post-error-slowing in children. Behavioral and Brain Functions, 5, 38, http://dx.doi.org/10.1186/1744-9081-5-38.

Hajcak, G., Moser, J. S., Holroyd, C. B., & Simons, R. F. (2006). The feedback-related nega-tivity reflects the binary evaluation of good versus bad outcomes. BiologicalPsychology, 71, 148–154, http://dx.doi.org/10.1016/j.biopsycho.2005.04.001.

Hajcak, G., & Simons, R. F. (2002). Error-related brain activity in obsessive–compulsiveundergraduates. Psychiatry Research, 110, 63–72, http://dx.doi.org/10.1016/S0165-1781(02)00034-3.

Hajcak, G., & Simons, R. F. (2008). Oops!…I did it again: An ERP and behavioral study ofdouble-errors. Brain and Cognition, 68, 15–21, http://dx.doi.org/10.1016/j.bandc.2008.02.118.

Hirsh, J. B., & Inzlicht, M. (2010). Error‐related negativity predicts academic performance.Psychophysiology, 47, 192–196, http://dx.doi.org/10.1111/j.1469-8986.2009.00877.x.

Hochman, E. Y., & Meiran, N. (2005). Central interference in error processing. Memory &Cognition, 33, 635–643, http://dx.doi.org/10.3758/BF03195330.

Hogan, A.M., Vargha‐Khadem, F., Kirkham, F. J., & Baldeweg, T. (2005). Maturation ofaction monitoring from adolescence to adulthood: An ERP study. DevelopmentalScience, 8, 525–534, http://dx.doi.org/10.1111/j.1467-7687.2005.00444.x.

Hollands, J. G., & Jarmasz, J. (2010). Revisiting confidence intervals for repeated measuresdesigns. Psychonomic Bulletin & Review, 17, 135–138, http://dx.doi.org/10.3758/PBR.17.1.135.

Houtman, F., & Notebaert, W. (2013). Blinded by an error. Cognition, 128, 228–236, http://dx.doi.org/10.1016/j.cognition.2013.04.003.

Jentzsch, I., & Dudschig, C. (2009).Why dowe slow down after an error? Mechanisms un-derlying the effects of posterror slowing. The Quarterly Journal of ExperimentalPsychology, 62, 209–218, http://dx.doi.org/10.1080/17470210802240655.

Jentzsch, I., & Leuthold, H. (2006). Control over speeded actions: A common processinglocus for micro- and macro-trade-offs? The Quarterly Journal of ExperimentalPsychology, 59, 1329–1337, http://dx.doi.org/10.1080/17470210600674394.

Kerns, J. G., Cohen, J.D., MacDonald, A. W., Cho, R. Y., Stenger, V. A., & Carter, C. S. (2004).Anterior cingulate conflict monitoring and adjustments in control. Science, 303,1023–1026, http://dx.doi.org/10.1126/science.1089910.

Kiesel, A., Steinhauser, M., Wendt, M., Falkenstein, M., Jost, K., Philipp, A.M., et al. (2010).Control and interference in task switching—A review. Psychological Bulletin, 136, 849,http://dx.doi.org/10.1037/a0019842.

King, J. A., Korb, F. M., von Cramon, D. Y., & Ullsperger, M. (2010). Post-error behavioraladjustments are facilitated by activation and suppression of task-relevant and task-irrelevant information processing. The Journal of Neuroscience, 30, 12759–12769,http://dx.doi.org/10.1523/JNEUROSCI.3274-10.2010.

Laming, D. (1979). Choice reaction performance following an error. Acta Psychologica, 43,199–224, http://dx.doi.org/10.1016/0001-6918(79)90026-X.

Larson, M. J., Fair, J. E., Good, D. A., & Baldwin, S. A. (2010). Empathy and error processing.Psychophysiology, 47, 415–424, http://dx.doi.org/10.1111/j.1469-8986.2009.00949.x.

Lorist, M. M., Boksem, M.A., & Ridderinkhof, K. R. (2005). Impaired cognitive control andreduced cingulate activity during mental fatigue. Cognitive Brain Research, 24,199–205, http://dx.doi.org/10.1016/j.cogbrainres.2005.01.018.

Los, S. A. (1996). On the origin of mixing costs: Exploring information processing in pureand mixed blocks of trials. Acta Psychologica, 94, 145–188, http://dx.doi.org/10.1016/0001-6918(95)00050-X.

MacLeod, C. M. (1991). Half a century of research on the Stroop effect: An integrative re-view. Psychological Bulletin, 109, 163–203, http://dx.doi.org/10.1037/0033-2909.109.2.163.

Mayr, U., & Kliegl, R. (2000). Task-set switching and long-term memory retrieval. Journalof Experimental Psychology: Learning, Memory, and Cognition, 26, 1124–1140, http://dx.doi.org/10.1037/0278-7393.26.5.1124.

Meiran, N. (2010). Task switching: Mechanisms underlying rigid vs. flexible self control.In R. Hassin, K. Ochsner, & Y. Trope (Eds.), Self control in society, mind and brain(pp. 202–220). New York, NY: Oxford University Press, http://dx.doi.org/10.1093/acprof:oso/9780195391381.003.0011.

Monsell, S. (2003). Task switching. Trends in Cognitive Sciences, 7, 134–140, http://dx.doi.org/10.1016/S1364-6613(03)00028-7.

Murphy, T. I., Richard, M., Masaki, H., & Segalowitz, S. J. (2006). The effect of sleepiness onperformance monitoring: I know what I am doing, but do I care? Journal of SleepResearch, 15, 15–21, http://dx.doi.org/10.1111/j.1365-2869.2006.00503.x.

Nee, D. E., Kastner, S., & Brown, J. W. (2011). Functional heterogeneity of conflict, error,task-switching, and unexpectedness effects within medial prefrontal cortex.NeuroImage, 54, 528–540, http://dx.doi.org/10.1016/j.neuroimage.2010.08.027.

Notebaert, W., Houtman, F., Opstal, F. V., Gevers, W., Fias, W., & Verguts, T. (2009). Post-error slowing: An orienting account. Cognition, 111, 275–279, http://dx.doi.org/10.1016/j. cognition .2009.02.002.

Núñez-Castellar, E., Kühn, S., Fias, W., & Notebaert, W. (2010). Outcome expectancy andnot accuracy determines posterror slowing: ERP support. Cognitive, Affective, &Behavioral Neuroscience, 10, 270–278, http://dx.doi.org/10.3758/CABN.10.2.270.

Pashler, H., & Johnston, J. C. (1989). Chronometric evidence for central postponement intemporally overlapping tasks. The Quarterly Journal of Experimental Psychology, 41,19–45, http://dx.doi.org/10.1080/14640748908402351.

Polich, J. (2007). Updating P300: An integrative theory of P3a and P3b. ClinicalNeurophysiology: Official Journal of the International Federation of ClinicalNeurophysiology, 118, 2128–2148, http://dx.doi.org/10.1016/j.clinph.2007.04.019.

Rabbitt, P.M. (1966). Errors and error correction in choice-response tasks. Journal ofExperimental Psychology, 71, 264–272, http://dx.doi.org/10.1037/h0022853 (Journalof Experimental Psychology).

Ridderinkhof, K. R., Forstmann, B. U., Wylie, S. A., Burle, B., & van den Wildenberg, W. P.(2010). Neurocognitive mechanisms of action control: Resisting the call of the sirens.Wiley Interdisciplinary Reviews: Cognitive Science, 2, 174–192, http://dx.doi.org/10.1002/wcs.99.

Riesel, A., Weinberg, A., Endrass, T., Kathmann, N., & Hajcak, G. (2012). Punishment has alasting impact on error‐related brain activity. Psychophysiology, 49, 239–247, http://dx.doi.org/10.1111/j.1469-8986.2011.01298.x.

Schneider, W., Eschman, A., & Zuccolotto, A. (2002). E-prime. Pittsburgh, PA: PsychologySoftware Tools.

Schroder, H. S., Moran, T. P., Infantolino, Z. P., & Moser, J. S. (2013). The relationshipbetween depressive symptoms and error monitoring during response switching.Cognitive, Affective, & Behavioral Neuroscience, 13, 790–802, http://dx.doi.org/10.3758/s13415-013-0184-4.

Smith, G. A., & Brewer, N. (1995). Slowness and age: Speed–accuracy mechanisms.Psychology and Aging, 10, 238, http://dx.doi.org/10.1037/0882-7974.10.2.238(Psychology and Aging).

Steinborn, M. B., Flehmig, H. C., Bratzke, D., & Schröter, H. (2012). Error reactivity in self-paced performance: Highly-accurate individuals exhibit largest post-error slowing.The Quarterly Journal of Experimental Psychology, 65, 624–631, http://dx.doi.org/10.1080/17470218.2012.660962.

Steinhauser, M., & Kiesel, A. (2011). Performance monitoring and the causal attribution oferrors. Cognitive, Affective, & Behavioral Neuroscience, 11, 309–320, http://dx.doi.org/10.3758/s13415-011-0033-2.

Stroop, J. R. (1935). Studies of interference in serial verbal reactions. Journal ofExperimental Psychology, 18, 643–662, http://dx.doi.org/10.1037/h0054651.

Themanson, J. R., & Hillman, C. H. (2006). Cardiorespiratory fitness and acute aerobicexercise effects on neuroelectric and behavioral measures of action monitoring.Neuroscience, 141, 757–767, http://dx.doi.org/10.1016/j.neuroscience.2006.04.004.

Themanson, J. R., Hillman, C. H., & Curtin, J. J. (2006). Age and physical activity influenceson action monitoring during task switching. Neurobiology of Aging, 27, 1335–1345,http://dx.doi.org/10.1016/j.neurobiolaging.2005.07.002.

Ullsperger, M., & Szymanowski, F. (2004). ERP correlates of error relevance. In M.Ullsperger, & M. Falkenstein (Eds.), Errors, conflicts, and the brain. Current opinionson performance monitoring (pp. 171–177). Leipzig: MPI for Human Cognitive andBrain Sciences.

Unsworth, N., Redick, T. S., Spillers, G. J., & Brewer, G. A. (2012). Variation in workingmemory capacity and cognitive control: Goal maintenance and microadjustmentsof control. The Quarterly Journal of Experimental Psychology, 65, 326–355, http://dx.doi.org/10.1080/17470218.2011.597865.

Van Veen, V., & Carter, C. S. (2006). Error detection, correction, and prevention in thebrain: A brief review of data and theories. Clinical EEG and Neuroscience, 37,330–335, http://dx.doi.org/10.1177/155005940603700411.

Vandierendonck, A., Liefooghe, B., & Verbruggen, F. (2010). Task switching: Interplay ofreconfiguration and interference control. Psychological Bulletin, 136, 601–626,http://dx.doi.org/10.1037/a0019791.

Wessel, J. R., Danielmeier, C., Morton, J. B., & Ullsperger, M. (2012). Surprise and error:Common neuronal architecture for the processing of errors and novelty. The Journalof Neuroscience, 32, 7528–7537, http://dx.doi.org/10.1523/JNEUROSCI.6352-11.2012.

Yeung, N., Botvinick, M. M., & Cohen, J.D. (2004). The neural basis of error detection:Conflict monitoring and the error-related negativity. Psychological Review, 111, 931,http://dx.doi.org/10.1037/0033-295X.111.4.931.