Prefrontal cortex activation in task switching: an event-related fMRI study

Cognitive Brain Research 9 Ž2000. 103–109 www.elsevier.comrlocaterbres Short communication Prefrontal cortex activation in task switching: an event-related fMRI study Anja Dove ) , Stefan Pollmann, Torsten Schubert, Christopher J. Wiggins, D. Yves von Cramon Max Planck Institute of CognitiÕe Neuroscience, Stephanstr. 1a, D-04103 Leipzig, Germany Accepted 15 June 1999 Abstract When a switch between two tasks has to be carried out, performance is slower than in trials where the same task is performed repeatedly. This finding has been attributed to time-consuming control processes required for task switching. Previous results of other paradigms investigating cognitive control processes suggested that prefrontal cortex is involved in executive control. We used event-related fMRI to investigate prefrontal cortex involvement in task switching. Regions in the lateral prefrontal and premotor cortex bilaterally, the anterior insula bilaterally, the left intraparietal sulcus, the SMArpre-SMA region and the cuneusrprecuneus were activated by the task repetition condition and showed additional activation in the task switch condition. This confirmed the hypothesis that lateral prefrontal cortex is involved in task switching. However, the results also showed that this region is neither the only region involved in task switching nor a region specifically involved in task switching. q 2000 Elsevier Science B.V. All rights reserved. Keywords: Task-switching; Prefrontal cortex; Control processes; Event-related fMRI The task switching paradigm is an experimental approach to examine executive processes w1,7,17,19x. In this paradigm subjects are required to alternate back and forth between a task A and a task B. A well known result are switch costs w1,7,17,19x that is an increase of reaction time when a task has to be performed as a task switch trial Žtask A then task B. compared to executing the same task as a task repetition trial Žtask A then task A.. These costs are attributed to time-consuming control processes. Behavioral research has shown that switch costs may vanish, if stimuli are presented which unambiguously cue only one task or the other and different responses are required in both tasks Že.g. task 1: name the opposite of ‘COLD’, task 2: subtract ‘3’ from number ‘47’. w25x. On the other hand, switch costs increase when ambiguity of the stimuli is heightened by the presentation of stimuli which contain stimulus dimensions relevant for both tasks w19x and when responses are ambiguous because the same responses are used for both tasks Že.g. the same fingers. w15x. Thus, one can conclude that task switch costs do not simply reflect ) Corresponding dove@cns.mpg.de author. Fax: q49-341-9940-221; E-mail: processes of re-configuration for a new task per se w1,17,19x. Configuration means selecting, linking and setting parameters of processing modules required for the upcoming task w17,19x. Instead, switch costs presumably reflect the heightened demand on executive control due to the requirement to configure the system for one task in the context of massive interference elicited by the second task w1,15,21x. These characteristic features of the paradigm make it valuable for an investigation of neural structures involved in certain aspects of executive control. Task switching especially involves ‘task management’ processes, i.e. scheduling processes which require switching of focused attention between tasks, and ‘attention and inhibition’ processes, i.e., focusing attention on relevant information and processes and inhibiting irrelevant ones w24x. The first goal of our study was to identify brain regions involved in task switching. Since prefrontal cortex is often thought to be involved in executive control processes w20,23,24x, we were especially interested to specify prefrontal cortex involvement in task switching. To our knowledge no fMRI study of task switching has yet been published which explicitly examined task switching Žbut cf. discussion.. However, data from patients with partly extended left and right frontal damage indicated that the 0926-6410r00r$ - see front matter q 2000 Elsevier Science B.V. All rights reserved. PII: S 0 9 2 6 - 6 4 1 0 Ž 9 9 . 0 0 0 2 9 - 4 104 A. DoÕe et al.r CognitiÕe Brain Research 9 (2000) 103–109 left frontal lobe seems to be involved in task switching w20x. A PET study which compared activation in a task switch block with activation elicited in single task blocks suggested that left dorsolateral prefrontal cortex ŽBA 9, 44, 45, 46., the anterior cingulate gyrus ŽBA 32., the premotor cortex ŽBA 6., the posterior parietal lobe ŽBA 7, 39, 40. and right cerebellum showed task switch specific activation w16x. An interesting observation in this PET study led us to the second goal of our study. Activations in the single task blocks were compared to a baseline condition w16x. It was observed that close to some of the regions involved in task switching activations occurred in the single task blocks, whereas other regions involved in task switching appeared to be additional areas of activation which were not involved in the single task blocks. For instance, a region within the posterior parietal lobe was activated both in the task repetition and in the task switch condition, whereas left dorsolateral prefrontal cortex only showed activation in the task switch condition w16x. This result suggests that it is possible to distinguish between areas involved both in task switch and task repetition and ‘executive areas’ exclusively involved in task switching. However, the correct test to show whether such a distinction can be made would have been to analyze whether exactly the same areas are involved both in the task switch and the task repetition condition or only in the task switch condition. If it could be demonstrated that the same region is involved in task switching, but not in task repetition, this would indicate that the region is specifically involved in executive processes. To conduct such an analysis, we defined regions of interest ŽROI. in areas activated in task switching and investigated whether the same ROIs showed task switch related and task repetition related activation or exclusively task switch related activation. Sixteen subjects Žnine female, age 21–29. took part in the fMRI experiment. Informed written consent was obtained from all subjects in accordance with the Declaration of Helsinki. The activation paradigm consisted of two tasks, a basic task and a switch task, in which an unpredictable switch had to be carried out. In basic task trials, a green ‘q’ or ‘y’ sign was presented at the center of the screen. Participants were instructed to press the left key of a response board when a ‘q’, and the right key when a ‘y’ was presented. They used the index and middle fingers of the right hand for responding. In the switch task the stimulus color changed to red and participants had to reverse the response mapping Ž‘q’ right key, ‘y’ left key.. A simple reversal of responses was used for several reasons. First, it ensured that working memory load was low in switch trials. Second, interference between tasks was high because the same stimulus dimensions were relevant for both tasks. Third, the same basic task processes Že.g., classification of simple objects, responses. were necessary in both tasks. This ensured that task switch trials would not elicit different activation than task repetition trials simply due to the fact that the tasks per se would require different processes which therefore would activate different brain areas. After presentation of one switch trial, a minimum number of four basic task trials was presented. Pilot studies showed that switch costs are higher when the amount of interspersed task repetition trials is increased. Switch trials were presented in exponentially declining frequency over the subsequent trials to ensure that no valid expectation of the occurrence of a switch could be generated by the participants. This warranted that no preparation for the task switch was possible which would reduce switch costs w14,15,19x. In the fMRI experiment, stimuli were presented for 1 s with an interstimulus interval ŽISI. of 15 s during which a fixation square was presented. One hundred twenty-five basic task trials and 20 unpredictable switch trials were presented. Data were collected at 3T using a 30r100 Medspec scanner ŽBruker, Ettlingen, Germany.. A two-shot EPI sequence was used in which each slice was measured twice with an interval of 1 s between the two measurements of the same slice ŽTR s 2000 ms, TE s 40 ms, RF flip angle of 408, FOV s 25 cm, matrix size s 128 = 64.. Seven horizontal slices were acquired parallel to the AC– PC-plane. Slice thickness was 10 mm, interslice distance 2 mm Žmost ventral slice: 22 mm below AC–PC.. Functional data were processed using the in-house software BRIAN w12x. The first goal of our experiment was to identify regions involved in task switching. To achieve this goal we averaged the data of the 16 participants in order to identify those regions which were commonly activated in task switching across subjects. After preprocessing Žmotion correction w4x, gaussian filter with a FWHMs 3.30 mm, baseline correction, low pass filter to reduce physiological and system noise w13x. data were averaged over participants. A voxelwise t-test was conducted to compare activation in the switch condition against the task repetition condition Žimplemented as described in Ref. w18x.. A boxcar reference waveform was designed to represent the task switch and task repetition conditions. Each switch trial and each repetition trial was represented by only one ‘on’ image. By choosing only one timestep per condition, most adjacent observations within one condition were separated by seven timesteps, and should thus be only minimally autocorrelated. The ‘on’-period covered the time window from 4 s–6 s. At this timestep the peak of the BOLD response is typically reached w6x and thus differences between the experimental conditions should be maximal. The obtained t-values were converted into z-values via the corresponding p-value. The z-map was administered with a threshold z s 3.09 Žcorresponding to a p s 0.001.. Regions with z ) 3.09 were then assessed for their significance using the Friston–Worsley Formula w5x with a p s 0.05. We focused on activations and did not analyze deactivations in this analysis. A. DoÕe et al.r CognitiÕe Brain Research 9 (2000) 103–109 The second goal of our study was to further analyze regions involved in task switching. We investigated 105 whether the same ROI showed task switch related and task repetition related activation or exclusively task switch Fig. 1. Analysis of brain regions involved in task switching. Ža. Areas activated in the switch condition. Activations are overlaid on the MR image of a single subject. The left hemisphere is on the left. Numbers correspond to the location of activation foci identified in Table 1. Žb. Reaction times in the switch and task repetition condition. The error bars indicate the standard error of means. Žc–i. Averaged event-related BOLD signal changes for the switch and the task repetition condition. The error bars indicate the standard error of means. A. DoÕe et al.r CognitiÕe Brain Research 9 (2000) 103–109 106 related activation. The steps of the ROI analyses were as follows. In order to define ROIs, the former group analysis was used to select anatomical structures involved in task switching. The borders of these structures were then defined anatomically in each individual brain. We always used the same slice to define anatomical structures in order to avoid slice acquisition order effects between subjects. Within the anatomical borders a cluster of four voxels showing maximal signal intensity in task switch trials were selected as ROI. To identify these voxels, a t-test was conducted which compared activation elicited by task switch trials to fixation baseline. The BOLD signal was pre-processed as described before but without low pass filter. The low pass filtering was omitted because we wanted to analyze and present eventrelated timecourses which are comparable to others reported in the literature Žespecially Refs. w10,11x.. After pre-processing the mean signal intensity of each timestep was determined for each ROI. To convert data into percent change from baseline, we first computed baseline activation for each trial. Average signal intensity at timestep 0–2 s Žat stimulus presentation, cf. Fig. 1c–j. and the two last timesteps of the prior trial Žtimesteps presented during the intervals 10–12 s and 12–14 s, cf. Fig. 1c–j. was defined to represent the baseline for the trial. Percent change from baseline was calculated within each trial. Afterwards, data were averaged for the switch condition, respectively the task repetition condition for each participant. Two statistical analyses were calculated on the ROI data. First, it was analyzed whether there was a significant difference between the BOLD-responses elicited in the task switch and task repetition condition. This was necessary to show that the individually defined ROI showed significant switch related activation. To confirm these differences, the BOLD responses measured at the peak of the BOLD-response Žin the time window of 4–6 s after stimulus presentation. w6x in both conditions were compared using a t-test for dependent measures. To investigate whether the activation in the task repetition condition was significantly increased relative to baseline or not, activation at the peak of the response was compared to the measurement acquired one timestep prior to the presentation of the task repetition stimulus. This timestep is most suited to represent the baseline because it is least affected by activation elicited by task performance in the previous trial. In all analyses we only selected one timestep per trial for each condition to ensure that temporal autocorrelation did not affect our analyses. Fig. 1b shows the behavioral results of the fMRI experiment. As expected, there were significant differences in reaction time ŽRT. between switch and task repetition trials Ž t Ž15. s 14.87; p - 0.001.. Participants were also significantly less accurate in the task switch condition Ž8.44% errors. than in the task repetition condition Ž1.44% errors; t Ž15. s 3.93; p - 0.001.. Image analysis revealed significant activation in the switch condition compared to the task repetition condition in several regions specified in Fig. 1a and Table 1. Fig. 1c–i shows the results of the ROI analyses which were conducted to analyze whether regions involved in task switching show activation in the task repetition condition or not. All cortical areas were analyzed except posterior cingulate, because in this region only few voxels showed switch-related activation. There was significantly Žtwo-sided Bonferroni corrected t-test with aŽ0.05. s 0.007. more activation at the peak of the BOLD-response in the task switch condition than in the task repetition condition in all analyzed regions except Table 1 Activation foci Žtask switch–task repetition condition.. Regions of significantly increased MR signal in the switch condition compared to the task repetition condition. Only regions with more than 10 active voxels were reported. LOC no. refers to activated regions shown in Fig. 1a. Brodmann’s areas ŽBA. are derived from the stereotactic atlas of Talairach and Tournoux w26x. Coordinates Žin mm, expressed as distances from the anterior commissure. refer to the location of maximal activation in standard stereotactic space w26x LOC no. Activated areas BA b 1 2 3 4 5 6 7 left precentral and inferior frontal sulcus right precentral and inferior frontal sulcus left anterior insulaa right anterior insula SMA and pre-SMA left intraparietal sulcus cuneus and precuneus 9r6r44 9r6 44r45 44r45 6r32 40 7 8 posterior cingulate gyrus 23r31 9 10 left thalamus right thalamus a b Max. Z score 5.45 6.16 7.31 7.26 7.56 6.12 4.72 3.88 4.36 3.91 5.08 4.73 5.31 Coordinates x y z y44 40 y36 28 y8 y32 0 y8 y4 y4 y8 4 20 5 8 20 23 11 y50 y75 y68 y16 y36 y13 y22 y27 37 36 13 8 47 45 42 57 33 34 8 6 0 The activation reported here extended from anterior insula to prefrontal cortex. To obtain the coordinates of the prefrontal activation the threshold was increased Ž z s 4.5. to separate the activation from the left insula activation. A. DoÕe et al.r CognitiÕe Brain Research 9 (2000) 103–109 cuneusrprecuneus which just missed significance Žleft prefrontal: t Ž15. s 4.08, p s 0.001; right prefrontal: t Ž15. s 5.53, p - 0.001; left anterior insula: t Ž15. s 7.22, p 0.001; right anterior insula: t Ž15. s 8.89, p - 0.001; SMArpre-SMA: t Ž15. s 6.65, p - 0.001; left IPS: t Ž15. s 7.97, p - 0.001; cuneusrprecuneus: t Ž15. s 2.84, p s 0.013.. This analysis confirmed that we selected subregions within the anatomically defined regions which showed task switch related activation. Peak activation in the task repetition condition was significantly different from fixation in all analyzed areas Žaccording to a two-sided t-test for repeated measurements with a Bonferroni corrected aŽ0.05. s 0.007; left prefrontal: t Ž15. s 4.75; p - 0.001, right prefrontal: t Ž15. s 4.77; p - 0.001, left anterior insula: t Ž15. s 6.46; p - 0.001, right anterior insula: t Ž15. s 11.76; p - 0.001, SMArpre-SMA: t Ž15. s 7.49; p - 0.001, left IPS: t Ž15. s 9.28; p - 0.001, cuneusrprecuneus: t Ž15. s 4.63; p - 0.001.. These results showed that all analyzed ROI not only showed task switch related activation but also activation in the task repetition condition. Thus, we did not find areas exclusively involved in task switch trials which were not activated in task repetition as well. To conclude, we identified regions in the lateral prefrontal and lateral premotor cortex bilaterally, the anterior insula bilaterally, the left intraparietal sulcus, the SMArpre-SMA region, the cuneusrprecuneus, the posterior cingulate and the thalamus bilaterally as network involved in task switching. Our left prefrontal and left intraparietal sulcus activation were comparable to regions identified by Meyer et al. w16x in their switching study. The goal of our study was to investigate prefrontal cortex involvement in task switching and, more generally, in executive control. Therefore, we will first show how the prefrontal activations observed in our study are related to results of other task switching studies and to results of other studies which used different paradigms to examine executive control processes. Then we will discuss the results obtained in our ROI analyses. The left prefrontal activation is in line with the data of Rogers et al. w20x, who showed that patients with left-sided frontal damage were impaired in task switching. Left prefrontal activation was also shown in the PET study of Meyer et al. w16x. In a further event-related switch study conducted by us in which a different switch task was used, we again obtained bilateral prefrontal activation at very similar locations as in this study w3x. Koechlin et al. w9x observed activation in posterior dorsolateral prefrontal cortex bilaterally in their ‘dual task’ conditions compared to conditions where only one task had to be performed. The dual task conditions involved performing two tasks successively using the same stimuli and responses for both tasks, and thus should better be labeled as ‘task switching’. Other neuroimaging studies which used different paradigms to investigate related aspects of executive control Ž‘attention and inhibition’ andror ‘task management’. 107 also observed prefrontal activation. Konishi et al. w10x examined brain activation during performance of a computerized version of the Wisconsin Card Sorting Test ŽWCST., which required shifting of cognitive set to identify and attend to new perceptual dimensions. The authors detected transient activation at the banks of the inferior frontal sulcus bilaterally whenever a cognitive shift was necessary. Bilateral dorsolateral prefrontal cortex activation was observed in a dual task condition where two tasks were performed simultaneously and thus had to be coordinated w2x. Left inferior frontal gyrus was identified as a region involved in Stroop-task performance w27x, a task which also requires cognitive control due to an induced conflict between an automatic irrelevant response and a required response w17x. In a random generation task which required self-determined generation of random finger sequences activation of left inferior frontal sulcus was observed w22x. In a verbal item-recognition task, the stimulus material included familiar distracter probes which presumably led to interference when the decision ‘target vs. non-target’ had to be made w8x. Again, activation of left lateral prefrontal cortex was elicited, albeit more ventral than our left inferior frontal sulcus activation. Konishi et al. w11x observed activation at the banks of the right inferior frontal sulcus in no-go compared to go-trials, which confirmed the suggestion that prefrontal cortex is involved in response inhibition. Inhibition of the response of the task presented prior to the switch is likely to be required in task switching. Thus, one can conclude that the prefrontal activations obtained in our study are comparable to prefrontal activations observed in imaging studies which used other related paradigms to investigate executive control. Since we used event-related fMRI, we could avoid the shortcomings of the block design used in several reported studies of executive control. Note that in dual task or task switching blocks working memory load is higher than in single task blocks w19x. For instance, in the study of Meyer et al. w16x in the task switch blocks rules for two tasks had to be kept in mind, whereas in the single task blocks only rules for one task had to be available. To identify regions involved in task switching, activation in single task blocks was subtracted from activation in the task switch block. Thus, the identified regions could either be involved in task switching or be activated due to the higher working memory load in the task switch blocks. Since the prefrontal cortex is known to be involved in working memory w24x, it is not clear whether the prefrontal activation is due to task switching or to the higher working memory load in the task switching block. The same confound was present in the studies of Koechlin et al. w9x and D’Esposito et al. w2x, because their dual task conditions were compared to conditions which only involved one task. This confound was not present in our experiment, since we used event-related fMRI which allowed us to present task switch and task repetition trials within one block. 108 A. DoÕe et al.r CognitiÕe Brain Research 9 (2000) 103–109 So far we focused on prefrontal cortex activation. On the other hand, other brain regions showed switch related activation, too. In several of the reported neuroimaging studies of executive control, other areas besides the prefrontal cortex showed activation in conditions requiring executive control w2,3,9,10,16,27x. We asked whether it is possible to distinguish between prefrontal activation and activations at other sites by investigating activation in the task repetition trials. If it could be shown that only some areas including prefrontal cortex showed switch specific activation, this would lend support to the view that these switch specific areas can be regarded as ‘executive areas’ instructing other areas to perform the task switch correctly. However, the results indicated that there was no such distinction between ‘executive areas’ and other areas. All analyzed areas showed similar activation patterns: A relatively small but significant BOLD-response elicited in task repetition trials and a greater BOLD-response elicited in task switch trials. Before we consider what this result might reveal about the functions of the prefrontal cortex, we will first discuss alternative explanations of our result. One objection to our interpretation would be that we observed global effects of effort in regions which were identified as task switch related. However, this explanation is not likely since behavioral results of other studies indicated that in the task switch paradigm basic task processes and task switch processes are independent of each other. For instance, Rubinstein et al. w21x varied the difficulty of task execution processes and the difficulty of the task switch operation. The results showed that the manipulation of basic task difficulty only led to higher reaction times in task repetition trials but did not affect task switch costs and that the manipulation of task switching difficulty only led to higher switch costs, but did not affect reaction times of task repetition trials. This showed that the task switch process is a separate process and not a more difficult basic task process. In light of this result, it is unlikely that our task switch condition only required the same processes as the task repetition condition. Furthermore, one could argue that in addition to effects of task switching other components of the task may have contributed to the task switching costs. Since switch trials were presented infrequently relative to task repetition trials, an ‘oddball’ response might have led to activation of prefrontal cortex. While we cannot completely rule out this explanation, it seems unlikely. Both in the study of Meyer et al. w16x and in another event-related fMRI switch study by us w3x, switch trials were not infrequent relative to repetition trials. Nonetheless prefrontal cortex activation was observed. Finally, it could be argued that in fact brain regions exist which show switch specific activation but that we failed to activate these regions. However, the strong behavioral switch costs indicated that our switch trials required control operations. Besides we did observe switch related activation in several brain regions, including the prefrontal cortex. Thus, in our view we succeeded to identify regions involved in task switching per se. We additionally found that these regions were activated in the task repetition condition and showed additional activation in the task switch condition. Our results therefore indicate that the task switch operation is performed by a network of areas which are involved in the task repetition condition already, not by one or several areas exclusively involved in task switching. More generally, this suggests that executive control is not exerted by brain areas exclusively involved in executive control, but within some of the areas involved in basic task processes. To resume, we identified several regions involved in task switching, including lateral prefrontal and premotor cortex bilaterally, anterior insula bilaterally, left intraparietal sulcus, SMArpre-SMA, cuneusrprecuneus, posterior cingulate and thalamus. Our prefrontal activation was comparable to activations reported in several other neuroimaging studies of executive control w2,3,9,10,16,22,27x. Therefore, our results further confirmed the hypothesis that the lateral prefrontal cortex is involved in executive control. On the other hand, our results also showed switch-related activations in other areas besides the prefrontal cortex. Thus, prefrontal cortex is not the only area involved in executive control. Furthermore, the prefrontal cortex, as anterior insula, intraparietal sulcus and SMArpre-SMA, did not only show switch specific activation but also activation in the task repetition condition. This result suggests that executive control in task switching is not exerted by specific ‘executive’ brain areas. Acknowledgements The authors thank the anonymous reviewer for helpful comments on this research. References w1x D.A. Allport, E.A. Styles, S. Hsieh, Shifting intentional set: exploring the dynamic control of tasks, in: C. Umilta, M. Moscovitch ŽEds.., Attention and Performance XV, Erlbaum, Hillsdale, 1994, pp. 421–452. w2x M. D’Esposito, J.A. Detre, D.C. Alsop, R.K. Shin, A. Scott, M. Grossman, The neural basis of the central executive system of working memory, Nature 378 Ž1995. 279–281. w3x A. Dove, T. Schubert, S. Pollmann, D. Norris, D.Y. von Cramon, Event-related fMRI of task switching, Neuroimage 9 Ž1999. S332. w4x K.J. Friston, S. Williams, R. Howard, R.S.J. Frackowiak, R. Turner, Movement-related effects in fMRI time-series, Magn. Reson. Med. 35 Ž1996. 346–355. w5x K.J. Friston, K.J. Worsley, R.S.J. Frackowiak, J.C. Maziotta, A.C. Evans, Assessing the significance of focal activations using their spatial extent, Hum. Brain Mapp. 1 Ž1994. 210–220. w6x X. Hu, T.H. Le, K. Ugurbil, Evaluation of the early response in A. DoÕe et al.r CognitiÕe Brain Research 9 (2000) 103–109 w7x w8x w9x w10x w11x w12x w13x w14x w15x w16x fMRI in individual subjects using short stimulus duration, Magn. Reson. Med. 37 Ž1997. 877–884. A. Jersild, Mental set and shift, Arch. Psychol., Whole No. 89 Ž1927. . J. Jonides, E.E. Smith, C. Marshuetz, R.A. Koeppe, P.A. ReuterLorenz, Inhibition in verbal working memory revealed by brain activation, Proc. Natl. Acad. Sci. U. S. A. 95 Ž1998. 8410–8413. E. Koechlin, G. Basso, P. Pietrini, S. Panzer, J. Grafman, The role of anterior prefrontal cortex in human cognition, Nature 399 Ž1999. 148–151. S. Konishi, K. Nakajima, I. Uchida, M. Kameyama, K. Nakahara, K. Sekihara, Y. Miyashita, Transient activation of inferior prefrontal cortex during cognitive set shifting, Nat. Neurosci. 1 Ž1998. 80–84. S. Konishi, K. Nakajima, I. Uchida, K. Sekihara, Y. Miyashita, No-go dominant brain activity in human inferior prefrontal cortex revealed by functional magnetic resonance imaging, Eur. J. Neurosci. 10 Ž1998. 1209–1213. F. Kruggel, G. Lohmann, BRIAN ŽBrain Image Analysis. — a toolkit for the analysis of multimodal brain data sets, in: H.U. Lemke, M.W. Vannier, K. Inamura, A.G. Farman ŽEds.., Computer Assisted Radiology, Proceedings of the International Symposium on Computer and Communication Systems for Image Guided Diagnosis and Therapy, Paris, June 1996, Elsevier, Amsterdam, 1996, pp. 323–328. F. Kruggel, D.Y.F. von Cramon, X. Descombes, Comparison of filtering methods for fMR datasets, submitted. N. Meiran, Reconfiguration of processing mode prior to task performance, J. Exp. Psych. Learn. Mem. Cogn. 22 Ž1996. 1423–1442. N. Meiran, The reconfiguration of the stimulus task-set and the response task-set during task-switching: putting the model to test, Paper presented at the meeting Attention and Performance XVIII: Control of Cognitive Processes, Windsor, England, July 13–17, 1998. D.E. Meyer, J.E. Evans, E.J. Lauber, J. Rubinstein, L. Gmeindl, L. w17x w18x w19x w20x w21x w22x w23x w24x w25x w26x w27x 109 Junck, R.A. Koeppe, Activation of brain mechanisms for executive mental processes in cognitive task switching, Poster presented at the meeting of the Cognitive Neuroscience Society, Boston, MA, March 23–25, 1997. S. Monsell, Control of mental processes, in: V. Bruce ŽEd.., Mysteries of the Mind: Tutorial Essays on Cognition, Erlbaum, Hove, 1996, pp. 93–148. W.H. Press, S.A. Teukolsky, W.T. Vetterling, B.P. Flannery, Numerical Recipes in C, Cambridge Univ. Press, 2nd edn., Cambridge, 1992, p. 616. R.D. Rogers, S. Monsell, Costs of a predictable switch between simple cognitive tasks, J. Exp. Psych. Gen. 124 Ž1995. 207–231. R.D. Rogers, B.J. Sahakian, R.J. Hodges, C.E. Polkey, C. Kennard, T.W. Robbins, Dissociating executive mechanisms of task control following frontal lobe damage and Parkinson’s disease, Brain 121 Ž1998. 815–842. J. Rubinstein, D.E. Meyer, J.E. Evans, Executive control of cognitive processes in task switching, submitted. T. Schubert, D.Y. von Cramon, Common activation of the dorsolateral prefrontal cortex in self-determined action and working memory, submitted. T. Shallice, From Neuropsychology to Mental Structure, Cambridge Univ. Press, Cambridge, 1996. E.E. Smith, J. Jonides, Storage and executive processes in the frontal lobes, Science 283 Ž1999. 1657–1661. A. Spector, I. Biederman, Mental set and mental shift revisited, Am. J. Psychol. 89 Ž1976. 669–679. J. Talairach, P. Tournoux, Co-Planar Stereotaxic Atlas of the Human Brain. 3-Dimensional Proportional System: An Approach to Cerebral Imaging, Georg Thieme Verlag, Stuttgart, 1988. S.F. Taylor, S. Kornblum, E.J. Lauber, S. Minoshima, R.A. Koeppe, Isolation of specific interference processing in the Stroop task: PET activation studies, Neuroimage 6 Ž1997. 81–92.