Abstract

Previous studies have shown the involvement of the ventrolateral prefrontal cortex (PFC) and the caudate nucleus when performing a set shift. However, the effect of set shifting on the frontostriatal activity observed during the later trials within a series of same-set classifications has yet to be determined. Here, young healthy adults underwent the functional magnetic resonance imaging while performing a card-sorting task in which the classification rule was provided prior to each trial. We observed a significant activation in the dorsolateral PFC, regardless of whether a set shift occurred or not. By contrast, the ventrolateral PFC and caudate nucleus showed an increased activity in both the shifting trials versus the control and in trials where the same rule was applied for a few trials before a set shift occurred, unlike trials where the same rule was applied for a longer period. Finally, decreased activity in the caudate nucleus correlated with an increasing trial position in trials where no set shift occurred, suggesting that the more a rule is executed, the better it is established. We argue that a new rule needs to be performed multiple times until the brain areas usually associated with the set shifting are no longer significantly required anymore.

Introduction

A key contribution of the prefrontal cortex (PFC) appears to be the ability to execute a set shift necessary for the cognitive flexibility, as has been shown in the lesion (Milner 1963) and functional neuroimaging studies (Berman et al. 1995; Rogers et al. 2000; Monchi et al. 2001; Nagahama et al. 2001; Monchi et al. 2006). Set shifting refers to the ability to change our attention from one response set to another according to the changing goals of a task.

Specific areas of the PFC and the caudate nucleus are known to be involved in set shifting. Using a computerized version of the Wisconsin Card Sorting Task (WCST), a neuropsychological test widely used to assess impairments in set shifting in humans (Milner 1963), Monchi et al. (2001) conducted an event-related functional magnetic resonance imaging (fMRI) study that demonstrated the involvement of area 47/12 of the ventrolateral PFC and the caudate nucleus during the period of planning a set-shifting response, while the junction of areas 6, 8, and 44 in the posterior PFC and the putamen were observed during the actual execution of the set shift. More recently, Monchi et al. (2006) have shown that the caudate nucleus is specifically involved when a new rule needs to be established without a specific indication from the environment.

Wylie and Allport (2000) reported behavioral results showing the effect of set shifting within a task. They observed that the reaction time increased in the trial following a change of rule and remained significantly higher for several trials after the set shift had been performed. To the best of our knowledge, no neuroimaging study has investigated set shifting as a gradual process in time, and by extension, the activity patterns of brain regions involved in set shifting in an increasing number of trials using the same rule of classification following a set shift. Previous studies have already shown a relationship between an increase in frontostriatal activity and cognitive load demands (Baker et al. 1996; Owen et al. 1996; van den Heuvel et al. 2003). Indeed, as the cognitive loads increase in the context of manipulation of the information within a working memory task, significant activity was observed in the dorsolateral PFC as well as in the caudate nucleus. However, cognitive load has never been studied before in the context of the residual effects of set shifting on the subsequent trials.

In the present study, using fMRI, we focussed our attention on 2 aspects of the performance of a WCST-like task, where the rule for classification is given to the participant prior to each trial. First, we wanted to examine the evolution of brain activity across several trials following a set shift. Second, we examined the transient brain activity in conditions in which participants needed either to execute a set shift on each trial for a long period, to apply the same rule for a long period, or to sporadically execute a set shift. We hypothesized that when a small number of trials implementing the same rule must be performed following a set shift, the pattern of brain activity will be similar to the pattern observed during the first trial following a set shift, which consists of a greater activity in the ventrolateral PFC and caudate nucleus. Moreover, we expected to see augmenting activation in these same 2 regions when the number of set shifts that are performed consecutively increases, as task complexity with respect to the cognitive resources being solicited should augment in this context. Furthermore, in a condition in which no set shift occurs for a long period, we predicted a decrease in activity across the number of trials in the brain regions usually associated with those observed during a set shift, perhaps related to a habituation effect.

Materials and Methods

Subjects

Fifteen subjects (7 males and 8 females), 20–29 years old (mean age: 23.9 years; standard deviation [SD]: 2.8 years), participated in this study. They were all right handed, as assessed with the Edinburgh Handedness Inventory (Oldfield 1971) and had no history of psychiatric or neurological disorders. Informed written consent was obtained from all participants according to the requirements of the research ethics committee of the Regroupement Neuroimagerie Québec (CMER-RNQ). This committee follows the guidelines of the Tri-Council Policy Statement of Canada, the civil code of Quebec, the Declaration of Helsinki, and the code of Nuremberg.

Cognitive Task

The present fMRI study used a mixed-design paradigm in order to look at the specific change in the blood oxygen level–dependent (BOLD) signal during the selection process. This new cognitive protocol was developed using the stimulus presentation software “Media Control Function” (Digivox, Montreal, Canada). The stimuli were presented via an LCD projector onto a mirror placed in front of the participant in the scanner. As in the classical WCST, the same 4 reference cards were displayed throughout the experiment. The 4 reference cards showing 1 red triangle, 2 green stars, 3 yellow crosses, and 4 blue circles, were displayed horizontally at the top of the screen (Fig. 1). On each trial, a different test card was presented in the middle of the screen, and the participants were required to pair the test card with one of the reference cards based on a shared attribute (color or shape or number) that was specified before each trial, unlike in the classical WCST in which the subject has to discover, by a trial and error, the classification rule. Here, the rule of classification in each trial was provided to the subject before the trial started, in order to examine the patterns of brain activity linked to the implementing of a rule rather than in searching for it. A letter cue (“N” = number, “S” = shape, “C” = color, or “I” = identical) was presented at the beginning of each trial to give the rule for the classification of that trial. The letter cue “I” (identical) indicated that the test card would be the same as one of the reference cards, and the subject would have to classify it on the basis of identity. The letter cue “I” was the instruction in the CONTROL condition. The subjects indicated their selection by using a magnetic-resonance–compatible button box: the left button moved a cursor under the reference cards from left to right, while the right button was used to validate the subject's selection. The button box was placed in the right hand of the participant who used the index finger to press the left button and the middle finger to press the right button. After each selection, the cursor reappeared at the position where it was last seen, so that no specific motor component could be associated with one particular reference card.

Figure 1.

Description of the cognitive task. The top panel shows an example of 2 typical trials of the cognitive task. First, the rule of classification is presented for a fixed period of 2000 ms. The participant is then asked to select one of the reference cards according to the previous rule displayed on the screen. Finally, the screen freezes for 1900 ms before the beginning of a new trial. In the lower section of the figure, the 4 conditions are presented each with an example of a sequence of the 4 conditions presented for 12 trials. In the CONTINUOUS SHIFT condition, participants never execute the same rule twice in row. In the SAME RULE condition, the same rule is executed continuously for the 12 trials. In the ALTERNATING condition, a change of rule is executed after 2, 3, or 4 consecutive trials of applying the same rule of classification. An overall total of either 5 or 6 set shifts are executed, however, the first trial was removed from our analysis leaving us always with a total of 5 set shifts. In this condition, we specified 2 distinct conditions: one in which we consider the sporadic set shifts and one in which we consider the execution of the same rule for a short amount of trials. In the CONTROL condition, participants need to match the test card with its twin.

Figure 1.

Description of the cognitive task. The top panel shows an example of 2 typical trials of the cognitive task. First, the rule of classification is presented for a fixed period of 2000 ms. The participant is then asked to select one of the reference cards according to the previous rule displayed on the screen. Finally, the screen freezes for 1900 ms before the beginning of a new trial. In the lower section of the figure, the 4 conditions are presented each with an example of a sequence of the 4 conditions presented for 12 trials. In the CONTINUOUS SHIFT condition, participants never execute the same rule twice in row. In the SAME RULE condition, the same rule is executed continuously for the 12 trials. In the ALTERNATING condition, a change of rule is executed after 2, 3, or 4 consecutive trials of applying the same rule of classification. An overall total of either 5 or 6 set shifts are executed, however, the first trial was removed from our analysis leaving us always with a total of 5 set shifts. In this condition, we specified 2 distinct conditions: one in which we consider the sporadic set shifts and one in which we consider the execution of the same rule for a short amount of trials. In the CONTROL condition, participants need to match the test card with its twin.

Each trial consisted of 2 distinct events. First, the rule of classification was displayed for 2000 ms (Fig. 1). The second event was characterized by the matching period. The 4 reference cards as well as the test card were presented until a selection was made by the participant. The length of the matching periods depended on the subject's response time and provided the desynchronization necessary between the stimuli and the frame acquisition, in order to perform an event-related analysis. After a selection was made, the screen froze for 1900 ms before the beginning of another trial.

The 4 conditions used during scanning were unknown to the participant (Fig. 1). In the “CONTINUOUS SHIFT” condition, a different rule of classification was provided on each one of the 12 consecutive trials. In the “SAME RULE” condition, participants matched the test card with one of the reference cards according to the same rule for the 12 consecutive trials. The rule was selected randomly by the computer program, and all rules were used equally often during the experiment. In the “ALTERNATING” condition, a different rule of classification was provided to the participants every 3, 4, or 5 consecutive trials, in a random fashion. Throughout a block of this condition, 5 or 6 set shifts were performed depending on whether or not the rule required for the first trial was the same as the one required for the last trial of the previous condition. For consistency purposes, we removed the first trial of this particular condition from the analysis so that 5 set shifts were always taken into account. Within the alternating condition, we could dissociate 2 different events. First, we could observe a shifting period, which occurred when the rule suddenly changed after a few consecutive trials. We defined that event as being the “RANDOM SHIFT” period. Then, we defined the event in which the same rule needed to be applied for 2, 3, or 4 trials as being the “RANDOM SAME” period (Fig. 1). The number of trials performed within this condition could vary from 12 to 24 depending on the number of consecutive trials executing the same rule before each set shift. The number of consecutive trials executing the same rule was randomly chosen by the computer each time. A post hoc analysis revealed that each set shift came about after an average of 2.6 trials were completed (SD = 0.4), which gives an average of 15.6 trials for the overall condition. In the “CONTROL” condition, the test card was the same as one of the reference cards, and the subjects were required to match it with its twin.

All participants needed to complete 3 blocks of conditions, where a block is defined as the 4 conditions randomly presented. Each condition was presented in a randomized fashion inside each block.

Before moving to the scanner, each participant was properly trained by completing the equivalent of one run, which consisted of 3 blocks of the 4 conditions randomly presented. Each participant was considered ready for scanning when his/her performance reached at least 95% correct responses, which occurred for each participant.

MRI Acquisitions

Scanning was performed in the 3T Siemens Trio Magnetom MRI Scanner at the Functional Neuroimaging Unit of the Centre de Recherche de l’Institut universitaire de Gériatrie de Montréal. The session started with a scout for positioning the subject. This was followed by a T1-weighted acquisition for the anatomical localization. Then, it was followed by 4 runs of T2*-weighted functional acquisitions. Each functional run lasted 10.5 min and consisted of 252 frames, each acquired at every 2.5 s. Each frame contained 36 slices placed along the anterior and posterior commissures with a matrix of 64 × 64 pixels, an isotropic voxel size of 3.4 × 3.4 × 3.4 mm3. The flip angle was 90° and the time echo was 30 ms.

Data Analysis

The methods for data analysis were the same as those in Monchi et al. (2001, 2006) using the fmristat analysis software developed by Worsley et al. (2002), http://www.math.mcgill.ca/keith/fmristat/, combined with the minc tools. The first 2 frames in each run were discarded. Images from each run were first realigned to the third frame for motion correction and were smoothed using a 6-mm full-width at half-maximum isotropic Gaussian kernel. The statistical analysis of the fMRI data was based on a linear model with correlated errors. The design matrix of the linear model was first convolved with a difference of 2-gamma hemodynamic response functions timed to coincide with the acquisition of each slice. The correlation structure was modeled as an autoregressive process. At each voxel, the autocorrelation parameter was estimated from the least squares residuals, after a bias correction for correlation induced by the linear model. The autocorrelation parameter was first regularized by spatial smoothing and was then used to “whiten” the data and the design matrix. The linear model was reestimated using least squares on the whitened data to produce estimates of effects and their standard errors. The resulting effects and the standard effect files were then spatially normalized by nonlinear transformation into the standard proportional stereotaxic space of Talairach and Tournoux (1988) using the algorithm of Collins et al. and the ICBM152 atlas as an approximation (Collins et al. 1994). Anatomical images were also normalized to the same space using the same transformation. In a second step, runs, sessions, and subjects were combined using a mixed effects linear model. A random effects analysis was performed by first estimating the ratio of the random effects variance to the fixed effects variance, then regularizing this ratio by spatial smoothing with a Gaussian filter. The amount of smoothing was chosen to achieve 100 effective degrees of freedom (Worsley 2005). Statistical maps were thresholded at P < 0.05 correcting for multiple comparisons using the minimum between a Bonferroni and random field correction. This corresponds to a t-statistics equal to or above 4.7 or a cluster size larger than 550 mm3, and only those peaks are reported here.

We performed an event-related analysis considering only the selection period, which corresponded to the time period that started with the presentation of the stimuli and ended with the selection of one reference card, indicating the choice of the participant. Six contrasts were generated from the statistical analysis for the comparison between trials, while 2 statistical maps were produced for the correlation analysis. We examined the following contrasts: 1) CONTINUOUS SHIFT minus the CONTROL condition; 2) SAME RULE minus CONTROL condition; 3) RANDOM SHIFT minus CONTROL condition; 4) RANDOM SAME minus CONTROL condition; 5) CONTINUOUS SHIFT minus RANDOM SAME; and 6) RANDOM SHIFT minus RANDOM SAME.

In addition to the contrast analysis, we performed a correlation analysis to assess the effect of trial position on the BOLD signal with the trial position within blocks of the CONTINUOUS SHIFT and SAME RULE conditions (not contrasted with the control condition), separately. In this analysis, a covariate was added assigning different weights according to the trial position within a given block. Only positive peaks are reported for the contrast analysis, while both positive and negative peaks are reported for the correlation analyses.

Results

Behavioral Data

All 15 participants completed the task. Behavioral data were analyzed using SPSS 15.0 for Windows. The mean percentage of errors for all conditions together was 1.6% (range from 0.8% to 2.6%). Specifically, we observed 3.4% of errors in the CONTINUOUS SHIFT condition, 0.8% in the SAME RULE condition, 2.6% in the RANDOM SHIFT trials, 1.5% in the RANDOM SAME trials, and 1.1% of errors in the CONTROL condition. We found a significant difference in the number of errors in the CONTINUOUS SHIFT versus the SAME RULE condition (t4213 = 2.19, P < 0.05) and in the RANDOM SHIFT compared with the SAME RULE condition (t3080 = 2.83, P < 0.05). Also, there was a significant difference when comparing the RANDOM SHIFT with the CONTROL condition (t3054 = 2.00, P < 0.05).

The average reaction time for the CONTINUOUS SHIFT condition was 1515 ms (SD = 645 ms), 1425 ms (SD = 633 ms) for the SAME RULE condition, 1481 ms (SD = 583 ms) for the alternating condition, and 1325 ms (SD = 593 ms) for the CONTROL condition. When the events in the alternating condition were separated, the average reaction time for the RANDOM SHIFT was 1505 ms (SD = 586 ms) and 1472 ms (SD = 582 ms) for the RANDOM SAME trials.

We examined the difference in the reaction times for the 4 conditions. First, there were significant reaction time differences among all conditions ([F3 = 41.60; P < 0.001]; CONTINUOUS SHIFT versus SAME RULE, t4281 = 4.63, P < 0.001; CONTINUOUS SHIFT versus ALTERNATING, t5412 = 1.96, P < 0.05; CONTINUOUS SHIFT versus CONTROL, t4261 = 10.03, P < 0.001; SAME RULE versus ALTERNATING, t5421 = 3.42, P < 0.001; SAME RULE versus CONTROL, t4270 = 5.32, P < 0.001; ALTERNATING versus CONTROL, t5401 = 9.62, P < 0.001). Second, we compared the reaction time differences by separating the events in the alternating condition into its 2 subconditions (i.e., RANDOM SHIFT and RANDOM SAME). Even when we considered the 5 types of events, we still observed marked difference among conditions (CONTINUOUS SHIFT vs. RANDOM SAME, t4433 = 2.33, P < 0.02; SAME RULE vs. RANDOM SHIFT, t3123 = 3.35, P < 0.001; SAME RULE vs. RANDOM SAME, t4442 = 2.61, P < 0.009; RANDOM SHIFT vs. CONTROL, t3103 = 7.88, P < 0.001), except in the CONTINUOUS SHIFT versus RANDOM SHIFT (t4281 = 0.43, P = 0.66), and in the RANDOM SHIFT versus the RANDOM SAME comparison (t3275 = 1.46, P = 0.15).

fMRI Data

A whole-brain analysis was performed for the contrasts of interest. For the fMRI analysis, trials with errors were removed. In conditions where a set shift was needed to be performed, we observed significant increase of the activation in the ventrolateral area of the PFC as well as in the striatum, both regions usually associated with set shifting. As predicted, we also observed significantly increased activity in the same brain areas in the condition when the same rule needed to be performed for a small number of trials, that is, in the RANDOM SAME trials. In this case, a set shift was required sporadically after 3, 4, or 5 trials. Finally, a significant decrease of activity in the dorsolateral PFC and the caudate nucleus was observed across time in the SAME RULE condition.

For ease of description, only positive frontal peaks (including the motor cortex) and subcortical ones are reported in this manuscript (both texts and tables). Tables are presented for all the active conditions versus control while only a description of the results in the text is given for the 2 comparisons between 2 active conditions. For the correlation analysis, both the positive (i.e., correlating with increasing trial position) and the negative (i.e., correlating with decreasing trial position) peaks are reported in the present manuscript.

It should be noted that we have added to all of our frontal peaks, the corresponding Brodmann areas according to the general anatomical landmarks. However, it should be noted that they are reported to improve the description of localization, but that this study using fMRI averaged over a group of individuals does not permit us to determine with certainty the exact Brodmann area(s) at the origin of a given peak of activation. Indeed, it has been suggested that caution should be used while using this classification because of intersubject variability (Amunts et al. 1999; Devlin and Poldrack 2007).

CONTINUOUS SHIFT versus CONTROL Condition

We found significantly increased activity in the left dorsolateral PFC (area 9, 9/46), the ventrolateral PFC (area 47/12), the Supplementary Motor Area (SMA) bilaterally, and the left posterior PFC (Fig. 2 and Table 1). Subcortically, we observed significant activation in the caudate nucleus and the ventrolateral nucleus of the thalamus bilaterally, and the left posterolateral nucleus of the thalamus.

Table 1

Continuous shift versus control condition

Anatomical areas Stereotaxic coordinates 
 x y z t values Cluster size 
Positive peaks 
Dorsolateral PFC (area 9, 46/9) 
    Left −48 30 24 9.47 34 872 
Ventrolateral PFC (area 47/12) 
    Left −28 22 5.91 34 872 
    Right 32 28 4.67 1544 
Paracingulate/SMA 
    Left −4 16 50 4.85 3864 
    Right 10 16 48 4.76 3864 
Posterior PFC (area 6, 8) 
    Left −42 34 7.47 34 872 
    Left −50 16 32 6.56 34 872 
Caudate nucleus (body) 
    Left −16 −10 26 4.93 34 872 
    Right 18 −16 26 5.18 6912 
Thalamus (ventrolateral) 
    Left −14 −10 10 4.64 34 872 
    Right 20 −14 18 4.88 6912 
Thalamus (posterolateral) 
    Left −20 −22 16 5.25 34 872 
Caudate nucleus 
    Right 26 −30 12 4.73 6912 
Anatomical areas Stereotaxic coordinates 
 x y z t values Cluster size 
Positive peaks 
Dorsolateral PFC (area 9, 46/9) 
    Left −48 30 24 9.47 34 872 
Ventrolateral PFC (area 47/12) 
    Left −28 22 5.91 34 872 
    Right 32 28 4.67 1544 
Paracingulate/SMA 
    Left −4 16 50 4.85 3864 
    Right 10 16 48 4.76 3864 
Posterior PFC (area 6, 8) 
    Left −42 34 7.47 34 872 
    Left −50 16 32 6.56 34 872 
Caudate nucleus (body) 
    Left −16 −10 26 4.93 34 872 
    Right 18 −16 26 5.18 6912 
Thalamus (ventrolateral) 
    Left −14 −10 10 4.64 34 872 
    Right 20 −14 18 4.88 6912 
Thalamus (posterolateral) 
    Left −20 −22 16 5.25 34 872 
Caudate nucleus 
    Right 26 −30 12 4.73 6912 

SAME RULE versus CONTROL Condition

We found significant activation in the left dorsolateral PFC (area 9/46) and the left lateral premotor cortex (area 6) (Table 2).

Table 2

Same rule versus control condition

Anatomical areas Stereotaxic coordinates 
 x y z t values Cluster size 
Dorsolateral PFC (area 46/9) 
    Left −48 30 28 7.57 >10 000 
Lateral premotor cortex (area 6) 
    Left −40 34 6.88 >10 000 
Anatomical areas Stereotaxic coordinates 
 x y z t values Cluster size 
Dorsolateral PFC (area 46/9) 
    Left −48 30 28 7.57 >10 000 
Lateral premotor cortex (area 6) 
    Left −40 34 6.88 >10 000 

RANDOM SHIFT versus CONTROL Condition

Significant activation was found in the left dorsolateral PFC (area 9/46), left ventrolateral PFC (area 47/12), left lateral posterior PFC (area 8), left paracingulate/SMA, and left precentral cortex (area 6) (Fig. 2 and Table 3). In subcortical regions, we found significant activation in the caudate nucleus bilaterally.

Table 3

Random shift versus control condition

Anatomical areas Stereotaxic coordinates 
 x y z t values Cluster size 
Positive peaks 
    Mid-dorsolateral PFC (area 9/46) 
        Left −50 30 28 8.86 >10 000 
    Ventrolateral PFC (area 47/12) 
        Left −30 22 10 5.44 1720 
    Lateral posterior PFC (area 8) 
        Left −52 20 32 6.48 >10 000 
    Paracingulate/SMA 
        Left −6 18 48 4.53 1712 
    Precentral gyrus (area 6) 
        Left −42 34 7.34 >10 000 
        Left −28 58 4.84 >10 000 
    Caudate nucleus (body) 
        Left −18 −2 20 4.71 2168 
        Right 18 −16 24 4.61 1720 
Anatomical areas Stereotaxic coordinates 
 x y z t values Cluster size 
Positive peaks 
    Mid-dorsolateral PFC (area 9/46) 
        Left −50 30 28 8.86 >10 000 
    Ventrolateral PFC (area 47/12) 
        Left −30 22 10 5.44 1720 
    Lateral posterior PFC (area 8) 
        Left −52 20 32 6.48 >10 000 
    Paracingulate/SMA 
        Left −6 18 48 4.53 1712 
    Precentral gyrus (area 6) 
        Left −42 34 7.34 >10 000 
        Left −28 58 4.84 >10 000 
    Caudate nucleus (body) 
        Left −18 −2 20 4.71 2168 
        Right 18 −16 24 4.61 1720 

Note: sc, same cluster

RANDOM SAME versus CONTROL Condition

We found significant increase of the BOLD signal in the left dorsolateral PFC (area 9/46), the ventrolateral PFC bilaterally (area 47/12), the left SMA (area 8), and the lateral premotor cortex bilaterally (area 6) (Fig. 2 and Table 4). Subcortically, a significantly increased activity was found in the caudate nucleus, bilaterally as well as in the left thalamus.

Table 4

Random same versus control condition

Anatomical areas Stereotaxic coordinates 
 x y z t values Cluster size 
Dorsolateral PFC (area 9/46) 
    Left −50 30 28 8.34 >10000 
Ventrolateral PFC (area 47/12) 
    Left −30 24 10 5.71 >10000 
    Right 30 28 4.50 680 
SMA 
    Left −4 16 50 5.27 3216 
Lateral PFC (area 6) 
    Left −24 10 50 4.31 >10000 
    Left −42 38 7.36 >10000 
    Right 26 54 5.02 1784 
Caudate nucleus (body) 
    Left −18 −2 22 4.89 3976 
    Right 18 −6 22 5.47 4560 
Thalamus 
    Left −10 −20 14 4.33 3976 
Anatomical areas Stereotaxic coordinates 
 x y z t values Cluster size 
Dorsolateral PFC (area 9/46) 
    Left −50 30 28 8.34 >10000 
Ventrolateral PFC (area 47/12) 
    Left −30 24 10 5.71 >10000 
    Right 30 28 4.50 680 
SMA 
    Left −4 16 50 5.27 3216 
Lateral PFC (area 6) 
    Left −24 10 50 4.31 >10000 
    Left −42 38 7.36 >10000 
    Right 26 54 5.02 1784 
Caudate nucleus (body) 
    Left −18 −2 22 4.89 3976 
    Right 18 −6 22 5.47 4560 
Thalamus 
    Left −10 −20 14 4.33 3976 

CONTINUOUS SHIFT versus RANDOM SAME

We found significant activation in the left dorsolateral PFC (area 9/46) and in the right caudate nucleus (Table not shown).

RANDOM SHIFT versus RANDOM SAME

Only a single significant peak was observed in the left dorsolateral PFC (area 9/46) for this particular contrast (Table not shown).

Correlation analysis for CONTINUOUS SHIFT Condition

We observed a significant correlation between the level of BOLD signal and increasing trial position in the CONTINUOUS SHIFT condition, in the left dorsolateral PFC (area 46), ventrolateral PFC bilaterally (area 47/12), and left posterior PFC (area 8) as well as the left inferior frontal junction (area 6, 8, 44) (Fig. 3 and Table 5). Subcortically, significant activation was found in the right thalamus and in the left caudate nucleus. No frontal or subcortical activity correlated significantly with a decrease in trial position.

Table 5

Correlation analysis for the continuous shift condition

Anatomical areas Stereotaxic coordinates 
 x y z t values Cluster size 
Positive peaks (correlating with increasing trial position) 
    Dorsolateral PFC (area 46) 
        Left −46 34 18 4.13 7312 
    Ventrolateral PFC (area 47/12) 
        Left −26 20 14 3.92 7312 
        Right 28 26 12 4.03 392 
    Posterior PFC 
        Left (area 8) −52 18 32 4.46 7312 
        Left (area 6/8/44) −42 38 4.73 sc 
    Thalamus 
        Right 16 −14 18 4.20 576 
    Caudate nucleus 
        Left −18 −8 22 3.45 1032 
        Left −18 −20 20 4.11 sc 
Anatomical areas Stereotaxic coordinates 
 x y z t values Cluster size 
Positive peaks (correlating with increasing trial position) 
    Dorsolateral PFC (area 46) 
        Left −46 34 18 4.13 7312 
    Ventrolateral PFC (area 47/12) 
        Left −26 20 14 3.92 7312 
        Right 28 26 12 4.03 392 
    Posterior PFC 
        Left (area 8) −52 18 32 4.46 7312 
        Left (area 6/8/44) −42 38 4.73 sc 
    Thalamus 
        Right 16 −14 18 4.20 576 
    Caudate nucleus 
        Left −18 −8 22 3.45 1032 
        Left −18 −20 20 4.11 sc 
Figure 2.

Location of the frontal and striatal peaks. The anatomical MRI is the average of the T1 acquisitions of the 15 participants transformed into standard stereotaxic space. (A) The “top left” panel shows a coronal section through the mid-dorsolateral PFC at Y = +30 mm. (B) The “top right” panel shows a coronal section through the mid-ventrolateral PFC at Y = +26 mm. (C) The “bottom” panel displays the body of the caudate nucleus through a coronal section at Y = −6 mm.

Figure 2.

Location of the frontal and striatal peaks. The anatomical MRI is the average of the T1 acquisitions of the 15 participants transformed into standard stereotaxic space. (A) The “top left” panel shows a coronal section through the mid-dorsolateral PFC at Y = +30 mm. (B) The “top right” panel shows a coronal section through the mid-ventrolateral PFC at Y = +26 mm. (C) The “bottom” panel displays the body of the caudate nucleus through a coronal section at Y = −6 mm.

Figure 3.

Correlation of the evolution of the BOLD signal according to the trial position. Each box represents the evolution of the BOLD signal according to the trial position for 2 prefrontal regions and the caudate nucleus. The red upward arrow represents a significant increase of activity across trial, while the blue downward arrow represents a significant decrease in the BOLD signal, as participants performed a specific condition. B, bilateral; L, left; DLPFC, dorsolateral PFC; VLPFC, ventrolateral PFC.

Figure 3.

Correlation of the evolution of the BOLD signal according to the trial position. Each box represents the evolution of the BOLD signal according to the trial position for 2 prefrontal regions and the caudate nucleus. The red upward arrow represents a significant increase of activity across trial, while the blue downward arrow represents a significant decrease in the BOLD signal, as participants performed a specific condition. B, bilateral; L, left; DLPFC, dorsolateral PFC; VLPFC, ventrolateral PFC.

Correlation Analysis for the SAME RULE Condition

We observed a significant correlation between the level of the BOLD signal and the increased trial position (as the time since the last set shift increases) in the premotor cortex bilaterally and the left insula (Fig. 3 and Table 6). Significant activation correlated with a decreasing trial position (as the time since the last set shift decreases) in the left dorsolateral PFC (area 9/46), the left cingulate cortex, as well as subcortically, in the right putamen, and the caudate nucleus bilaterally.

Table 6

Correlation analysis for the same rule condition

Anatomical areas Stereotaxic coordinates 
 x y z t values Cluster size 
Positive peaks (correlating with increasing trial position) 
    Premotor cortex (area 6) 
        Left −54 5.04 6088 
        Right 52 −4 4.23 3504 
    Insula 
        Left −40 −12 4.31 6088 
Negative peaks (correlating with decreasing trial position) 
    Dorsolateral PFC (area 9/46) 
        Left −50 30 26 3.88 2360 
    Putamen 
        Right 24 12 4.44 856 
    Caudate nucleus 
        Left −18 26 4.05 1256 
        Right 18 −2 28 3.90 640 
    Cingulate cortex 
        Left −2 −30 28 4.92 2144 
Anatomical areas Stereotaxic coordinates 
 x y z t values Cluster size 
Positive peaks (correlating with increasing trial position) 
    Premotor cortex (area 6) 
        Left −54 5.04 6088 
        Right 52 −4 4.23 3504 
    Insula 
        Left −40 −12 4.31 6088 
Negative peaks (correlating with decreasing trial position) 
    Dorsolateral PFC (area 9/46) 
        Left −50 30 26 3.88 2360 
    Putamen 
        Right 24 12 4.44 856 
    Caudate nucleus 
        Left −18 26 4.05 1256 
        Right 18 −2 28 3.90 640 
    Cingulate cortex 
        Left −2 −30 28 4.92 2144 

Discussion

In the present study, we looked at the difference of the frontostriatal patterns of activation in different contexts of rule implementation. In particular, we examined the brain activity during the 4 conditions: one in which a different classification rule was applied on each one of the 12 consecutive trials, another in which the same rule was applied also for 12 consecutive trials, one in which a shift between 2 rules was needed after every 2, 3, or 4 trials, and the control condition in which the target card had to be matched with one of the reference cards. Relative to the control condition, the left dorsolateral PFC and the left posterior PFC exhibited increased activity in all conditions (i.e., CONTINUOUS SHIFT; ALTERNATING, including the random same and random shift trials; and SAME RULE conditions) (Fig. 2A). Significant increases in activity in the depth of the horizontal ramus of the Sylvian fissure, involving area 47/12 of the ventrolateral PFC and the caudate nucleus, was observed during the sporadic shifting (RANDOM SHIFT) and the CONTINUOUS SHIFT conditions versus the CONTROL (Fig. 2). Of particular interest, we found the significantly increased activity in the same 2 regions in a series of trials in between shifts (i.e., RANDOM SAME vs. CONTROL condition), and this pattern of activity decreased when executing the same rule as the number of consecutive trials increased, that is, as the rule gets more familiar.

The ventrolateral PFC and the caudate nucleus were activated significantly in the 3 following contrasts: CONTINUOUS SHIFT, RANDOM SHIFT, and RANDOM SAME (each vs. the CONTROL condition) but not in the SAME RULE versus the CONTROL condition. The ventrolateral PFC was reported to be involved in the active selection process associated with the active retrieval, in which there is an active comparison of the stimulus in mind with the targeted stimuli leading to object selection, reducing the amount of ambiguity between the stimuli (Cadoret et al. 2001; Petrides 2002; Kostopoulos and Petrides 2003). Previous neuroimaging studies have shown the contribution of mid-ventrolateral PFC not only in the controlled retrieval of information within the memory, as opposed to a more automatic memory processing (Dove et al. 2006), but also when a specific array of stimuli need to be identified regardless of the sequence in which they were presented (Jonides et al. 1993; Owen, Evans, et al. 1996). As for the caudate nucleus, neuroimaging studies have shown the involvement of this subcortical structure in the working memory tasks in which a manipulation of information was required as opposed to the maintenance/retrieval of the same material (Lewis et al. 2004), or in the planning of complex sequences of actions (Baker et al. 1996; Owen et al. 1996).

The activation of both the ventrolateral PFC and caudate nucleus was also shown in negative feedback processing in the WSCT, that is, when a set shift is required (Monchi et al. 2001). Monchi et al. (2006) examined specifically the events within a new card-sorting task (the Montreal Card Sorting Task) and found that these regions were significantly activated when a set shift needs to be planned leading to the use of a new rule of classification. They also showed that the ventrolateral PFC is involved in the retrieval of a rule independently of whether a set shift needs to be performed, but that the caudate nucleus is only involved when a set shift needs to be self-generated (Monchi et al. 2006). In the present study, it should be noted that, while the rule for classification is given to the participant, the response card that needs to be selected is not unlike the CONTINUOUS SHIFT condition of the Montreal Card Sorting Task (MCST), where both the rule and the card were implicitly suggested by the task (Monchi et al. 2006). This explains why the caudate nucleus was significantly activated in all the shifting conditions of the present study but was not in the CONTINUOUS SHIFT condition of the MCST. Most importantly, the new finding of the present study is that ventrolateral PFC and caudate nucleus had significant activation in the contrast RANDOM SAME versus the CONTROL condition similarly to the 2 set-shifting versus CONTROL conditions. In these trials, participants had to apply the same rule for only 2, 3, or 4 consecutive trials following a set shift. As predicted, our results showed that regions associated with the set shifting, such as the ventrolateral PFC and the caudate nucleus, are still significantly activated during the contrast between the RANDOM SAME vresus CONTROL conditions, most likely because the rule is not yet clearly established unlike the SAME RULE condition where such patterns of significant activity were not observed versus the CONTROL condition. Indeed, our present functional neuroimaging results support the behavioral notion of interference arising from the competing rules within a specific task (Allport and Wylie 2000; Wylie and Allport 2000). The switch cost associated with the set shift is measured as the difference in reaction time from nonshifting trials subtracted from set-shifting trials. However, the difference in reaction time was shown to go beyond the actual set shift and would last for several trials after the set shift. This temporary elevation in reaction time is described to be linked with the “task-set reconfiguration” in which a new rule of classification is applied for the first few times. This effect in reaction time was previously observed in tasks in which a rule needed to be applied for several trials. Indeed, Allport and Wylie reported that the participants responded faster in repeated trials of the same task in which only one condition was needed to be performed, as opposed to the repeated trials of the same task in which more than one condition was needed to be executed by the participants. In fact, this discrepancy in reaction times could be linked to the settling of a new rule following a shift. In our study, our behavioral results are in agreement with this notion. The reaction time observed in the RANDOM SAME trials was significantly greater than that in the SAME RULE condition. In fact, the CONTROL and the SAME RULE conditions were the 2 conditions with the fastest reaction time. Moreover, our neuroimaging results showed the regions associated with a set shift continued to be solicited beyond the actual set shift, which is what we observed in the contrast RANDOM SAME versus CONTROL condition. Furthermore, considering that the SAME RULE condition is an extended version of the RANDOM SAME trials, we can assume that there is a decrease in activity in these regions following a certain number of trials. Indeed, in another analysis not shown here, we contrasted the RANDOM SAME trials with only the 4 first trials of the SAME RULE condition blocks to see whether the resulting activity is the same. None of our regions of interest were significantly activated or deactivated bringing further evidence that the decrease of the prefrontal regions in the SAME RULE condition is due to the continuous execution of a same rule of classification for a longer period (i.e., 4 trials or more using the same rule).

In support of the interpretation proposed in the previous paragraph, the correlation analysis looking at the BOLD activity, with respect to the trial position in the CONTINUOUS SHIFT condition, revealed that the ventrolateral PFC and the caudate nucleus are significantly more activated as the condition evolved across time. This may explain why significant activity was observed in the caudate nucleus in the CONTINUOUS SHIFT versus RANDOM SAME, while it was not observed in the RANDOM SHIFT versus the RANDOM SAME. Indeed, the RANDOM SHIFT trials correspond to the first trials from the CONTINUOUS SHIFT condition. On the other hand, when correlating increasing trial position for the SAME RULE condition, no significant peak was observed in the dorsolateral PFC or the caudate nucleus as an increasing number of trials were performed within the condition. As discussed earlier, the ventrolateral PFC and the caudate nucleus regions have an important role in the set-shifting process, and it would be expected that they are increasingly solicited as more and more set shifts are required. Furthermore, both caudate nucleus and putamen correlated significantly with the reverse order of the trials, that is as the trials became closer to the first trial of the SAME RULE condition and also therefore to the last set shift, which comes from the preceding condition. Interestingly, at the beginning of the SAME RULE condition, the striatum is significantly active as the contrast between the RANDOM SAME and the SAME RULE condition also revealed. As the participants go through the condition, the activity in these regions decreases across time. In the SAME RULE condition, the frontostriatal network activity decreases significantly, as matching within the same rule is performed many times. This is likely due to the fact that the rule becomes clearly established and without interference, unlike the RANDOM SAME trials. This result complements nicely the contrast between the RANDOM SAME versus CONTROL condition and definitely helps clarify on how the brain deals functionally when a set shift is performed.

With regards to the ventrolateral PFC, as mentioned above, we found significant increase of activity in the correlation analysis, as trials increased in CONTINUOUS SHIFT condition only. This observation can be explained by the increasing level of ambiguity caused by 1) the changing stimulus that needs to be selected, but even more importantly by 2) the changing rule that needs to be considered on each trial to execute the proper selection. Indeed, as the number of trials being executed increases within this condition, the use of previous rules interfere gradually with the task at hand which, in turn, requires greater demands on the active retrieval process mentioned above. Significant activation was found in the ventrolateral PFC for SAME RULE versus CONTROL conditions. This can be explained that while the same rule is used throughout this condition, participants still need to actively compare the target stimulus with the reference stimuli at the level of the intrarule selection (e.g., green vs. red for color), in order to execute the proper selection on each trial. However, since the rule remains the same throughout the condition, the level of ambiguity at the interrule level (e.g., color, shape, or form) remains stable across the different trials of the condition, which could explain the lack of correlation between the ventrolateral PFC activity and increasing trials in the SAME RULE condition.

As mentioned above, the dorsolateral PFC was active during all of our experimental conditions versus the CONTROL condition. Significantly increased activity in the dorsolateral PFC correlated with an increased trial position. Previous lesion studies in monkeys and functional neuroimaging experiments have led to the proposal that the dorsolateral PFC is involved in the monitoring of events within the working memory and not the maintenance of information per se (Petrides 2000). According to this hypothesis, the dorsolateral PFC has an important role in keeping track of different events that may occur in different contexts. In our task, the rule of classification is given at the beginning of each trial. The proper rule of classification needed to be monitored in the working memory is to be then applied correctly to successfully complete a trial. Indeed, since participants were trained according to the 3 rules of classification, they needed to keep track of the proper rule given prior to the trial, especially given that they were not aware of the existence of specific conditions. Indeed, the dorsolateral PFC was required to keep track of these changing rules for all of the active conditions whether during set shifting or while using an ongoing rule. It is interesting that the increase in dorsolateral PFC activity correlated with an increased trial position in the CONTINUOUS SHIFT condition, while it correlated with a decreased trial position in the SAME RULE condition. With increasing trials in the CONTINUOUS SHIFT condition, the online monitoring demands increase as the classification rule constantly changes. On the other hand, in the SAME RULE condition, as the number of trials increases and the classification rule remains the same and the demand for online tracking diminishes.

Previous studies have shown the involvement of the basal ganglia in the associative learning tasks. More specifically, the ventral striatum was shown to be involved in the stimulus–reward association learning. Indeed, lesions in the ventral striatum in rodents impaired the learning of the coupling between a stimulus and a reward (Atallah et al. 2007). On the other hand, lesions in the dorsal striatum impaired the selection of the rewarded versus the unrewarded stimuli; however, the encoding of the stimulus–response association is still intact. In humans, neuroimaging studies have shown the involvement of the ventral striatum to correlate with the degree to which an implicit motor task is learned, especially in the early stage of a novel task (Reiss et al. 2005), regardless to whether it is reinforced with a reward or not. Conversely, the dorsal striatum has been significantly activated in tasks in which the stimulus–response contingency changes over time (Rogers et al. 2000; Grinband et al. 2006; Monchi et al. 2006). A recent study from MacDonald et al. (2011) has provided evidence that the ventral striatum plays a crucial role in the general associative learning, while the dorsal striatum, more specifically the caudate nucleus, is preferentially engaged during decision making, according to the integration of conflicting situational information (such as those that occur during a set shift). In our study, such conflicting information increases with a set-shift trial, and even more so if multiple shift trials occur consecutively. Indeed, the execution of a new rule significantly solicited the caudate nucleus and its level of activation increased over time (with increasing trial position) in the CONTINUOUS SHIFT condition, while it decreased over time in the SAME RULE condition.

As mentioned above, our study also tapped on the notion of extradimensional and intradimensional shifts (Owen et al. 1991). In the SAME RULE condition, participants select a different stimulus within the same rule or dimension at every trial (e.g., red, blue, green, or yellow if color is required), as opposed to the CONTINUOUS SHIFT condition where a stimulus using a different dimension or rule (e.g., shape, number, or color) needed to be selected. The CONTINUOUS SHIFT condition has a significantly greater reaction time as opposed to the SAME RULE condition supporting the idea that extradimensional shifts are more demanding than intradimensional shifts. In an intradimensional shift, subjects have to choose another stimulus, but within the same dimension (corresponding to a rule in the present study) as opposed to an extradimensional shift in which participants have to change from one task set (or rule) to another as well as one stimulus to another, increasing the level of cognitive control. Other studies have reported similar results in terms of reaction time differences in the dimensional shifts (Rogers et al. 2000; Hampshire and Owen 2006). In terms of neuroimaging data, Rogers et al. (2000) reported significant activity in the dorsolateral and posterior PFC regions, but did not observe activation in the ventrolateral PFC and caudate nucleus. This difference may be due to the block-design positron emission tomography setup which might not have been sensitive enough to capture the transient effect of the different events within a set shift as reported here. More recently, Hampshire and Owen (2006) showed significant activation in the ventrolateral PFC in extradimensional shifts using fMRI but did not report significant activity in the caudate nucleus possibly because it was not chosen as a region of interest in their study.

Conclusion

The present article contributes to further clarify the role of PFC and caudate nucleus in set shifting. One can argue that when a set shift is performed, there is still a fair amount of uncertainty associated with the newly used rule and the former rule of classification. The ventrolateral PFC and caudate nucleus activity seems involved in resolving this uncertainty. Once a rule is well established, this ambivalence disappears and the activity in the ventrolateral PFC and caudate nucleus decreases. As shown in the present study, regardless of whether the participants were required to change rule continuously or sporadically, the ventrolateral PFC and the caudate nucleus were significantly active. Importantly, these same structures were active during the execution of the same rule for a short period of trial, showing that the regions involved in a set shift are not only active during the set shift but also during the following trials.

The present findings have implications for many cognitive functional neuroimaging protocols in which participants are asked to switch frequently from one condition to another. This switching between conditions could activate regions related to the set shifting even if some precautionary steps were to be taken, such as removing the first trial after the set shift. We argue that a new rule needs to be applied on multiple occasions before the brain areas usually associated with set shifting significantly decrease in activation. Further studies should be conducted with this respect to clarify the different steps in the cognitive processes behind the set shifting until a rule is being fully established.

Funding

Natural Sciences and Engineering Research Council of Canada (grant No 327518 to O.M.).

The authors would like to thank all the participants, the staff of the functional Neuroimaging Unit at the CRIUGM for the practical help as well as support. Conflict of Interest: None declared.

References

Allport
A
Wylie
G
Monsell
S
Driver
J
Task-switching, stimulus-response binding, and negative priming
Control of cognitive processes: attention and performance XVIII
 , 
2000
Berkshire (UK): MIT Press
(pg. 
35
-
70
)
Amunts
K
Schleicher
A
Bürgel
U
Mohlberg
H
Uylings
HB
Zilles
K
Broca's region revisited: cytoarchitecture and intersubject variability
J Comp Neurol
 , 
1999
, vol. 
412
 (pg. 
319
-
341
)
Atallah
HE
Lopez-Paniagua
D
Rudy
JW
O'Reilly
RC
Separate substrates for skill learning and performance in the ventral and dorsal striatum
Nat Neurosci
 , 
2007
, vol. 
10
 (pg. 
126
-
131
)
Baker
SC
Rogers
RD
Owen
AM
Frith
CD
Dolan
RJ
Frackowiak
RSJ
Robbins
TW
Neural systems engaged by planning: a PET study of the Tower of London task
Neuropsychologia
 , 
1996
, vol. 
34
 (pg. 
515
-
526
)
Berman
KF
Ostrem
JL
Randolph
C
Gold
J
Goldberg
TE
Coppola
R
Carson
RE
Herscovitch
P
Weinberger
DR
Physiological activation of a cortical network during performance of the Wisconsin Card Sorting Test: a positron emission tomography study
Neuropsychologia
 , 
1995
, vol. 
33
 (pg. 
1027
-
1046
)
Cadoret
G
Pike
GB
Petrides
M
Selective activation of the ventrolateral prefrontal cortex in the human brain during active retrieval processing
Eur J Neurosci
 , 
2001
, vol. 
14
 (pg. 
1164
-
1170
)
Collins
DL
Neelin
P
Peters
TM
Evans
AC
Automatic 3D intersubject registration of MR volumetric data in standardized Talairach space
J Comput Assist Tomogr
 , 
1994
, vol. 
18
 (pg. 
192
-
205
)
Devlin
JT
Poldrack
RA
In praise of tedious anatomy
Neuroimage
 , 
2007
, vol. 
37
 (pg. 
1033
-
1041
)
Dove
A
Brett
M
Cusack
R
Owen
AM
Dissociable contributions of the mid-ventrolateral frontal cortex and the medial temporal lobe system to human memory
Neuroimage
 , 
2006
, vol. 
31
 (pg. 
1790
-
1801
)
Grinband
J
Hirsch
J
Ferrera
VP
A neural representation of categorization uncertainty in the human brain
Neuron
 , 
2006
, vol. 
49
 (pg. 
757
-
763
)
Hampshire
A
Owen
AM
Fractionating attentional control using event-related fMRI
Cereb Cortex
 , 
2006
, vol. 
16
 (pg. 
1679
-
1689
)
Jonides
J
Smith
EE
Koeppe
RA
Awh
E
Minoshima
S
Mintun
MA
Spatial working memory in humans as revealed by PET
Nature
 , 
1993
, vol. 
363
 (pg. 
623
-
625
)
Kostopoulos
P
Petrides
M
The mid-ventrolateral prefrontal cortex: insights into its role in memory retrieval
Eur J Neurosci
 , 
2003
, vol. 
17
 (pg. 
1489
-
1497
)
Lewis
SJ
Dove
A
Robbins
TW
Barker
RA
Owen
AM
Striatal contributions to working memory: a functional magnetic resonance imaging study in humans
Eur J Neurosci
 , 
2004
, vol. 
19
 (pg. 
755
-
760
)
MacDonald
PA
MacDonald
AA
Seergobin
KN
Tamjeedi
R
Ganjavi
H
Provost
JS
Monchi
O
The effect of dopamine therapy on ventral and dorsal striatum-mediated cognition in Parkinson's disease: support from functional MRI
Brain
 , 
2011
, vol. 
134
 (pg. 
1447
-
1463
)
Milner
B
Effects of different brain lesions on card sorting: the role of the frontal lobes
Arch Neurol
 , 
1963
, vol. 
9
 (pg. 
90
-
100
)
Monchi
O
Petrides
M
Petre
V
Worsley
K
Dagher
A
Wisconsin Card Sorting revisited: distinct neural circuits participating in different stages of the task identified by event-related functional magnetic resonance imaging
J Neurosci
 , 
2001
, vol. 
21
 (pg. 
7733
-
7741
)
Monchi
O
Petrides
M
Strafella
AP
Worsley
KJ
Doyon
J
Functional role of the basal ganglia in the planning and execution of actions
Ann Neurol
 , 
2006
, vol. 
59
 (pg. 
257
-
264
)
Nagahama
Y
Okada
T
Katsumi
Y
Hayashi
T
Yamauchi
H
Oyanagi
C
Konishi
J
Fukuyama
H
Shibasaki
H
Dissociable mechanisms of attentional control within the human prefrontal cortex
Cereb Cortex
 , 
2001
, vol. 
11
 (pg. 
85
-
92
)
Oldfield
RC
The assessment and analysis of handedness: the Edinburgh inventory
Neuropsychologia
 , 
1971
, vol. 
9
 
1
(pg. 
97
-
113
)
Owen
AM
Roberts
AC
Polkey
CE
Sahakian
BJ
Robbins
TW
Extra-dimensional versus intra-dimensional set shifting performance following frontal lobe excisions, temporal lobe excisions or amygdalo-hippocampectomy in man
Neuropsychologia
 , 
1991
, vol. 
29(10)
 (pg. 
993
-
1006
)
Owen
AM
Doyon
J
Petrides
M
Evans
AC
Planning and spatial working memory: a positron emission tomography study in humans
Eur J Neurosci
 , 
1996
, vol. 
8
 (pg. 
353
-
364
)
Owen
AM
Evans
AC
Petrides
M
Evidence for a two-stage model of spatial working memory processing within the lateral frontal cortex: a positron emission tomography study
Cereb Cortex
 , 
1996
, vol. 
6
 (pg. 
31
-
38
)
Petrides
M
The role of the mid-dorsolateral prefrontal cortex in working memory
Exp Brain Res
 , 
2000
, vol. 
133
 (pg. 
44
-
54
)
Petrides
M
The mid-ventrolateral prefrontal cortex and active mnemonic retrieval
Neurobiol Learn Mem
 , 
2002
, vol. 
78
 (pg. 
528
-
538
)
Reiss
JP
Campbell
DW
Leslie
WD
Paulus
MP
Stroman
PW
Polimeni
JO
Malcolmson
KA
Sareen
J
The role of the striatum in the implicit learning: a functional magnetic resonance imaging study
Neuroreport
 , 
2005
, vol. 
16
 
12
(pg. 
1291
-
1295
)
Rogers
RD
Andrews
TC
Grasby
PM
Brooks
DJ
Robbins
TW
Contrasting cortical and subcortical activations produced by attentional-set shifting and reversal learning in humans
J Cogn Neurosci
 , 
2000
, vol. 
12
 (pg. 
142
-
162
)
Talairach
J
Tournoux
P
Co-planar stereotaxic atlas of the human brain. 3-Dimensional proportional system: an approach to cerebral imaging
 , 
1988
Stuttgart
Thiemes
van den Heuvel
OA
Groenewegen
HJ
Barkhof
F
Lazeron
RHC
van Dyck
R
Veltman
DJ
Frontostriatal system in planning complexity: a parametric functional magnetic resonance version of Tower of London task
Neuroimage
 , 
2003
, vol. 
18
 (pg. 
367
-
374
)
Worsley
KJ
Spatial smoothing of autocorrelations to control the degrees of freedom in fMRI analysis
Neuroimage
 , 
2005
, vol. 
26
 (pg. 
635
-
641
)
Worsley
KJ
Liao
CH
Aston
J
Petre
V
Duncan
GH
Morales
F
Evans
AC
A general statistical analysis for fMRI data
Neuroimage
 , 
2002
, vol. 
15
 (pg. 
1
-
15
)
Wylie
G
Allport
A
Task switching and the measurement of switch cost
Psychol Res
 , 
2000
, vol. 
63
 (pg. 
212
-
233
)