Abstract

Humans are unique in being able to reflect on their own performance. For example, we are more motivated to do well on a task when we are told that our abilities are being evaluated. We set out to study the effect of self-motivation on a working memory task. By telling one group of participants that we were assessing their cognitive abilities, and another group that we were simply optimizing task parameters, we managed to enhance the motivation to do well in the first group. We matched the performance between the groups. During functional magnetic resonance imaging, the motivated group showed enhanced activity when making errors. This activity was extensive, including the anterior paracingulate cortex, lateral prefrontal and orbitofrontal cortex. These areas showed enhanced interaction with each other. The anterior paracingulate activity correlated with self-image ratings, and overlapped with activity when participants explicitly reflected upon their performance. We suggest that the motivation to do well leads to treating errors as being in conflict with one's ideals for oneself.

Introduction

Humans can reflect on themselves and their achievements. They can set themselves goals in relation to their expectations and monitor their progress toward them. In other words they are “self-motivated.” We know something of the brain mechanisms of motivation in other primates, for example that activity can be enhanced the greater the reward in the anterior cingulate cortex (Shidara and Richmond 2002; Amiez et al. 2006), the dorsolateral prefrontal cortex (Leon and Shadlen 1999), and the parietal cortex (Platt and Glimcher 1999). But experiments on how the behavior of animals is regulated according to external rewards cannot tell us much about how humans regulate their behavior according to their own ideals.

An early study (McClelland et al. 1953) demonstrated that the motivation of participants to perform well was influenced by the task instructions given. If the participants were told that the test they were taking measured their intelligence they used more achievement related words when writing a short story after the test. A recent study using similar instruction manipulations showed an increase in task performance, as well as an increase in cardiovascular and electrodermal reactivity (Gendolla and Richter 2005). We set out to study motivation by imaging the brain while the participants were tested on a cognitive task. The participants performed the n-back working memory task for letters (Fig. 1a). This task has been much studied in imaging experiments (Braver et al. 1997). We manipulated motivation by giving different instructions to 2 groups. We told one group (Group High_M) that we were going to assess their memory and that memory is related to intelligence. We told the other group (Group Low_M) that we were piloting the task so as to optimize certain task parameters. As revealed from participants’ ratings of motivation (Fig. 1b), the first group regarded successful performance a valuable ability.

Figure 1.

(a) In the n-back task a sequence of letters was presented on a screen. For each letter that they saw the participants pressed a button. If a current letter also appeared 3 letters back (the n − 3 runs) or one letter back (the n − 1 runs) the participant make a yes response; otherwise they made a no response. Each letter was presented for 500 ms. (b) Ratings given by participants in the 2 groups in answer to the question on how important it had been for them to perform well on the task. Group High_M rated significantly higher (P < 0.016) that it was important to perform well.

Figure 1.

(a) In the n-back task a sequence of letters was presented on a screen. For each letter that they saw the participants pressed a button. If a current letter also appeared 3 letters back (the n − 3 runs) or one letter back (the n − 1 runs) the participant make a yes response; otherwise they made a no response. Each letter was presented for 500 ms. (b) Ratings given by participants in the 2 groups in answer to the question on how important it had been for them to perform well on the task. Group High_M rated significantly higher (P < 0.016) that it was important to perform well.

People have beliefs about their own ability to handle a task, and so their performance is assessed partly in terms of their own internal self-evaluations (Ryan 1982; Bandura 1989), which may involve self-esteem, fear of academic standing, or social prestige. The motivation of an individual is especially stimulated when a task provides people with the opportunity to stretch their capacities (Csikszentmihalyi 1975). Studies investigating neural processes of mental state attributions have found a particular role for the anterior paracingulate cortex. Enhanced activation in this region is observed when people take a first-person perspective, judge whether adjectives relate to themselves or not, and in memory processes of items encoded in a self-relevant manner (Kelley et al. 2002; Macrae et al. 2004; Fossati et al. 2004; Ochsner et al. 2005). This led us to predict that activity in the anterior paracingulate cortex would also be increased when participants perform a cognitive task and their self-evaluation is enhanced.

In order to investigate this, we needed to match performance for the 2 groups. Unless we did so, we could not know if enhanced activity in Group High_M reflected a difference in self-involvement or a difference in performance. For this reason we used a staircase method where we altered the rate of stimulus presentation to match performance between the groups. We set the performance level to 75% correct.

Materials and Methods

Participants

Twenty-six healthy, right-handed volunteers took part in the study. In Group High_M there were 13 participants with a mean age of 23.4 ± 4.0 years (3 males, 10 females); in Group Low_M there were 13 participants with a mean age of 24.0 ± 4.3 years (5 males, 8 females). The participants gave written informed consent. The study was approved by the joint ethics committee of the Institute of Neurology and University College London Hospital, London, UK. Another 5 participants were scanned but excluded from the analysis because their performance did not reach the performance level we had set up. In addition, 2 participants were excluded before the data analysis because the manipulation failed to work. On the question asking how important it was to do well on the task, one of the participants in Group High_M only gave a rating of 3 out of 9; the other participant, from Group Low_M, gave a rating of 9 out of 9.

Behavioral Task

The participants were tested on the n-back task (Fig. 1a). They were given both the 3-back and the 1-back tasks for letters. Sequences of letters were presented on a screen, and for each letter the subject pressed a button. If the letter appeared 3 letters back (on the n − 3 runs) or one letter back (on the n − 1 runs) the participant made a yes response; otherwise they made a no response. The yes response was made by pressing the button corresponding to the index finger, the no response by pressing the button corresponding to the middle finger. A block consisted of 19 letters, each letter being visible for 500 ms. Within a block there were 30% hits.

The interstimulus interval (ISI) was varied between runs of the n-3 back task, depending on the performance of the participant. This was done to equate performance between subjects. During scanning we titrated the ISI for each participant so as to achieve a level of roughly 75% correct performance. The ISI ranged from 500 to 2500 ms: the rate was increased if performance on the last block exceeded 75% and it was decreased if it was less than 75%. To control for the rate of presentation of the letters between subjects presentation on the n − 1 was yoked to the previous block of trials on the n − 3 task. There were 4 sessions in the functional magnetic resonance imaging (fMRI) scanner. The first session was a practice session so as to familiarize the participants with the scanner environment. The following 3 sessions each consisted of 10 pairs of a block of n-3 and a block of n-1.

Before the scanning, the participants practiced each task for 10 blocks. Following the practice, we informed them about the study. Half of the participants (Group High_M) were told that we were assessing their working memory capacity and that working memory is related to intelligence. The other half of the participants (Group Low_M) were told that we were piloting a task so as to optimize certain parameters. Participants were randomly assessed into either of the 2 groups.

Given that we predicted that enhanced self-motivation would lead to activation in the anterior paracingulate cortex, we introduced an additional condition, independent of the memory task, in which members of Group High_M explicitly reflected on their task performance. The reason was that we wanted an independent measure of where the activations lay for self-reflection. A question was presented after every second n − 3 block. Participants answered questions related to their performance such as “My concentration was good” or “My performance was poor,” whereas Group Low_M rated parameters of the task, like “The letters were easily visible” or “The task was fast.” They answered the questions according to a 1–4 scale, where 1 meant not at all. The question was visible on the screen for 10 s. Within each session there were also 10 blocks of rest where a fixation cross appeared on the screen for 10 s.

After fMRI scanning the participants answered questionnaires on trait anxiety (Spielberger 1983), depression (Burns 2007), as well as questions on how important it was to succeed on the task and how difficult they found the task to be. To evaluate participants’ self-image in relation to being intelligent we administered rating scales where the participants rated 10 positive statements all associated with being intelligent, such as “I am clever” and “I am inventive.” The ratings were done on a 1–10 scale where 1 corresponded to not at all, and 10 corresponded to yes indeed. They also rated 10 negative statements, that is, associated with being stupid, such as “I am foolish” and “I generally appear confused.” In addition, for each statement the participants rated how important the attribute was to them (1–10). We calculated the product of the rating on each statement and corresponding importance factor. Subsequently, we subtracted the average score on positive statements with the average score on negative statements. High scores correspond to a positive self-image.

The group comparison tests reported in Table 1 are 2-sided t-tests.

Table 1

Group comparisons on behavioral measures

 Group High_M (mean ± SD) Group Low_M (mean ± SD) Group comparison 
Motivation to do well on task 6.6 ± 1.5 5.0 ± 1.6 P < 0.016 
Perceived task difficulty 6.8 ± 1.8 6.9 ± 1.1 P > 0.79 
Performance n − 3 75.8 ± 7.5% 73.2 ± 4.2% P > 0.27 
Reaction time on the n-back tasks 571.6 ms ± 24.1 649.5 ms ± 44.0 P > 0.13 
Reaction time on question ratings 2.03 s ± 0.15 2.12 s ± 0.13 P > 0.66 
Self-image score 34.8 ± 19.8 40.5 ± 20.7 P > 0.47 
Trait anxiety score 44.1 ± 3.4 43.6 ± 5.9 P > 0.81 
Depression score 15.0 ± 7.9 14.0 ± 7.4 P > 0.74 
 Group High_M (mean ± SD) Group Low_M (mean ± SD) Group comparison 
Motivation to do well on task 6.6 ± 1.5 5.0 ± 1.6 P < 0.016 
Perceived task difficulty 6.8 ± 1.8 6.9 ± 1.1 P > 0.79 
Performance n − 3 75.8 ± 7.5% 73.2 ± 4.2% P > 0.27 
Reaction time on the n-back tasks 571.6 ms ± 24.1 649.5 ms ± 44.0 P > 0.13 
Reaction time on question ratings 2.03 s ± 0.15 2.12 s ± 0.13 P > 0.66 
Self-image score 34.8 ± 19.8 40.5 ± 20.7 P > 0.47 
Trait anxiety score 44.1 ± 3.4 43.6 ± 5.9 P > 0.81 
Depression score 15.0 ± 7.9 14.0 ± 7.4 P > 0.74 

Imaging Procedures

Imaging was performed using a 3 Tesla scanner (Siemens Allegra: Erlagen, Germany). The functional images sensitive to blood oxygen level–dependent (BOLD) contrasts were acquired as gradient-echo, echo-planar T2*-weighted images. Image volumes of the whole brain were built up from contiguous axial slices (n = 40) time repetition = 2.6 s, time echo = 10 ms, matrix size 64 × 64, slice thickness 2 mm with 1-mm gap. A high-resolution, 3-dimensional gradient-echo T1-weighted anatomical image was also collected for each participant.

fMRI Data Analysis

Image processing and analysis were done using the SPM5 software (http://www.fil.ion.ucl.ac.uk/spm/, Wellcome Trust Centre for Neuroimaging, London, UK). The first 5 volumes were discarded, and the remaining volumes realigned to the first volume to correct for head movements. Subsequently, the volumes were coregistered and normalized to the standard space (Talaraich and Tournoux 1988) using the Montreal Neurological Institute reference brain. The time series were smoothed spatially with an isotropic Gaussian filter of 8-mm full width at half maximum.

The difference in motivation between the groups could either be reflected as a difference in state, that is, sustained activation throughout the experiment, or be reflected as a difference specific to the events of making either correct or error responses. To test which of these models that best explains our data we set up a mixed design (Otten et al. 2002; Dosenbach et al. 2006). For each n-back memory condition, we defined covariates for the transient activation in response to the motor responses separately for correct and incorrect responses, as well as a covariate representing each memory trial as an epoch. In addition, the question was modeled as a 5-s epoch with its onset locked to the presentation of the question. Thus, the regressors account for variance unique either to the event-related activity or the state-related activity. The transient activation was modeled with a delta function, whereas the epochs were modeled with a boxcar regressor. Statistical parametric maps of t-statistics were calculated for the 7 condition specific effects within a general linear model. All epochs and events were convolved with a canonical hemodynamic response function. The data were high-pass filtered with a frequency cut-off at 128 s.

We performed 4 contrasts of interest. We did these first for each participant separately, followed by a between subjects random-effects analysis based on summary-statistics from the subtraction images created from each subject. On the second level, we used a 2-sample t-test to compare the 2 groups. For the n-back task contrasts a regressor was included that accounted for the difference in ISI between participants.

Firstly, we were interested to see whether there was any difference between the 2 groups in sustained activity throughout the memory performance. For this, we compared the activation during the 3-back task and 1-back task.

Secondly, we were interested to see whether there was any difference between the 2 groups when performing a high load working memory task compared with a low load working memory task on correct performance (Group High_M vs. Group Low_M [3 − n correct response vs. 1 − n correct response]).

Thirdly, we investigated whether there was any difference between groups on error trials compared with correct performance (Group High_M vs. Group Low_M [errors vs. corrects]).

Last, we contrasted the activity for the ratings (Rating of performance for Group High_M vs. Ratings of the task parameters for Group Low_M). This comparison gives areas that are involved in reflecting on one's own performance.

In addition to these contrasts of interest, we analyzed similar effects between the 2 groups in performing the 3 − n task versus the 1 − n task for correct performance using all participants where the 2 groups were modeled as separate conditions (Group High_M + Group Low_M [correct 3 − n vs. correct 1 − n]). These activations reflect performance of the working memory performance. We then used these as a mask (at a level of P < 0.01) to see whether any of the activations for the contrast of Group High_M vs. Group Low_M (errors vs. corrects) lay in task related areas (Table S1). We also created a contrast for the main effect of making errors for all participants where the 2 groups were modeled as separate conditions (Group High_M + Group Low_M [errors vs. corrects]). This was used to asses if activations specific to Group High_M lay within general error-related activity.

Finally, we looked for psychophysical interactions (PPIs) (Friston et al. 1997) as a consequence of the increased intrinsic motivation in Group High_M. We selected a seed region (6-mm radius) in the paracingulate cortex (coordinate 10 46 36) that was significant in the contrast (Group High_M vs. Group Low_M [errors vs. corrects]). We extracted the time series from this region for each participant using the first eigen-time series (principal component). The PPI regressor was calculated from the product of the mean-corrected paracingulate activity and a vector coding for the differential effect of error and correct responses. This latter vector was defined as 1 for errors made on the n − 3 and n − 1 tasks, and −1 for correct responses made on the n − 3 and n − 1 tasks. This means that our analysis of effective connectivity was specific for context-dependent paracingulate influences that occurred over and above any task effects and context-independent paracingulate influences. Thus, this PPI analysis reveals which areas show activation patterns that covary with paracingulate activity where the context is errors versus corrects. Subsequently, we asked whether there were differences between the psychophysiological interactions (PPIs) in the 2 groups. Here the 2 groups form different contexts, high in self-motivation or not. We carried out a 2-sample t-test (random-effects analysis) to compare the interactions in Group High_M and Group Low_M.

In Table 2 we report the pattern of activation that was significant on the set-level (P < 0.0001). Set-level inference refers to the statistical inference that the number of clusters observed in the activation profile is highly unlikely to have occurred by chance (Friston et al. 1996). The peaks of activation reported in the table are taken from an uncorrected threshold of P < 0.001 at voxel-level, cluster size > 5 voxels.

Table 2

Group High_M versus Group Low_M [errors vs. corrects]

Area Side x y z t-score 
Paracingulate cortex (32) 10 46 36 3.92 
Pre-SMA −6 30 52 4.26 
Pre-SMA −6 12 60 3.57 
Cingulate sulcus 28 38 3.92 
Cingulate sulcus −8 46 22 3.40 
Superior frontal gyrus 20 42 40 3.90 
Superior frontal sulcus 26 14 58 5.37 
Middle frontal gyrus −46 12 50 4.54 
Middle frontal gyrus −34 20 54 3.96 
Inferior frontal gyrus (47) −52 18 4.58 
Inferior frontal gyrus (47) −50 24 −4 3.77 
Inferior frontal gyrus (47) −52 34 3.35 
Insula −34 10 20 3.91 
Orbitofrontal cortex −32 20 −14 3.52 
Orbitofrontal cortex 34 10 −18 3.46 
Inferior temporal cortex −52 −24 −28 3.52 
Anterior temporal cortex −38 22 −26 3.42 
Cerebellum −20 −46 −48 3.43 
Cerebellum −30 −72 −46 3.65 
Area Side x y z t-score 
Paracingulate cortex (32) 10 46 36 3.92 
Pre-SMA −6 30 52 4.26 
Pre-SMA −6 12 60 3.57 
Cingulate sulcus 28 38 3.92 
Cingulate sulcus −8 46 22 3.40 
Superior frontal gyrus 20 42 40 3.90 
Superior frontal sulcus 26 14 58 5.37 
Middle frontal gyrus −46 12 50 4.54 
Middle frontal gyrus −34 20 54 3.96 
Inferior frontal gyrus (47) −52 18 4.58 
Inferior frontal gyrus (47) −50 24 −4 3.77 
Inferior frontal gyrus (47) −52 34 3.35 
Insula −34 10 20 3.91 
Orbitofrontal cortex −32 20 −14 3.52 
Orbitofrontal cortex 34 10 −18 3.46 
Inferior temporal cortex −52 −24 −28 3.52 
Anterior temporal cortex −38 22 −26 3.42 
Cerebellum −20 −46 −48 3.43 
Cerebellum −30 −72 −46 3.65 

The histograms in Figure 4 represent task related BOLD signal changes for as estimated with the general linear model. Error bars represents the standard error. The parameter estimates are given for the voxel showing a peak difference in the between-group comparison. The baseline represents null events that were not modeled. These consist of period of rest where the participants were looking at a fixation cross in the center of the screen.

Results

Behavioral Results

We first checked that our manipulation had worked. The ratings for Group High_M on the importance of doing well on the task were significantly higher than the ratings for Group Low_M (Table 1; Fig. 1b). As a control, the participants were also asked to rate how difficult they found the task, and for this measure there was no difference between the 2 groups. Moreover, as seen in Table 1, there was no significant difference in performance level between the groups. However, Group High_M were able to achieve this level with shorter ISIs than Group Low_M (P = 0.059). This means that the instruction to Group High_M did have a behavioral effect. However, in the analysis of the imaging data, we were careful to control for this effect (see Methods and discussion).

We were concerned that in the allocation of participants to the 2 groups, there might be differences in emotional state between the groups. To check for this, we gave self-report questionnaires on trait anxiety and depressive features in the last week. There were no differences between the groups (Table 1).

Imaging Results

n-Back Task

There was no brain areas showing higher activation for sustained activity in Group High_M compared with Group Low_M. We therefore lowered the significance level to P < 0.001 uncorrected, and still failed to find differences for this comparison.

For correct trials on the n-back task, there were no differences in activation between Group High_M and Group Low_M (P < 0.001 uncorrected). However, on error trials compared with correct trials, Group High_M showed increased activation in a wide range of prefrontal areas in contrast to Group Low_M (Table 2; Fig. 2). On the medial surface there was increased activation in the paracingulate cortex (area 32) and cingulate sulcus. On the lateral surface the enhanced activation was found in the superior frontal gyrus, middle frontal gyrus and the inferior frontal gyrus (area 47). There was also more activation in the insular cortex and the orbital frontal cortex.

Figure 2.

Participants of Group High_M were told that the working memory capacity correlates with intelligence. On error trials Group High_M showed enhanced activation in medial frontal, lateral frontal and orbitofrontal cortex. The activations (errors vs. corrects) are shown for Group High_M and Group Low_M separately. The activation maps are taken from a threshold of P < 0.001 uncorrected at voxel level. The pattern of activation for Group High_M versus Group Low_M is significant at a set-level (P < 0.0001).

Figure 2.

Participants of Group High_M were told that the working memory capacity correlates with intelligence. On error trials Group High_M showed enhanced activation in medial frontal, lateral frontal and orbitofrontal cortex. The activations (errors vs. corrects) are shown for Group High_M and Group Low_M separately. The activation maps are taken from a threshold of P < 0.001 uncorrected at voxel level. The pattern of activation for Group High_M versus Group Low_M is significant at a set-level (P < 0.0001).

There was no activity in the anterior paracingulate cortex on correct trials as assessed for the 2 groups combined (see supplementary materialTable S1). Yet, this area was activated on error trials for the comparison of Group High_M with Group Low_M (Table 2). The peak lay rostral to the anterior cingulate cortex in which error-related activity could be found for both groups. To illustrate this fact, Figure 3 shows the medial activation when all participants (Group High_M + Group Low_M) made errors. It will be seen from the location of the cross that the paracingulate peak for Group High_M lay well in front of this.

Figure 3.

The cluster on the medial wall represents error-related activity as seen when the data for Group High_M and Group Low_M are added together. The red cross represents the peak in the anterior paracingulate sulcus (10 46 36) that was significantly more active in Group High_M than Group Low_M on error trials. The anterior cluster does not overlap with the posterior cluster.

Figure 3.

The cluster on the medial wall represents error-related activity as seen when the data for Group High_M and Group Low_M are added together. The red cross represents the peak in the anterior paracingulate sulcus (10 46 36) that was significantly more active in Group High_M than Group Low_M on error trials. The anterior cluster does not overlap with the posterior cluster.

We suggest that the anterior paracingulate activation reflects the increased tendency for group high in self-motivation to monitor their performance on error trials. In favor of this interpretation is the fact that activation in the paracingulate cortex has been widely reported during self-reflection (Amodio and Frith 2006). To confirm this association in our own results, we contrasted the activations when Group High_M answered questions on their own performance with the activations when Group Low_M answered questions about the task parameters. The differences lay in the anterior paracingulate cortex (10 46 24, t = 2.82), right ventral prefrontal cortex (area 47) (44 36 −4, t = 3.42) and the inferior temporal gyrus (66 −32 −14, t = 4.06).

We checked whether the paracingulate peak for reflection on one's own performance lay within the cluster that we found for the enhancement of activity in for Group High_M on error trials. We created a mask from the paracingulate cluster on the n-back task (Group High_M vs. Group Low_M [error vs. correct]). Within this mask, the activation for the ratings nearly reached statistical significance at a corrected level (P = 0.06). This is shown in Figure 4.

Figure 4.

There was an increased interaction between the anterior paracingulate sulcus and the right area 47 when Group High_M made errors as compared with Group Low_M. There were significant peaks of activity in the same regions when participants in Group High_M reflected on their own performance.

Figure 4.

There was an increased interaction between the anterior paracingulate sulcus and the right area 47 when Group High_M made errors as compared with Group Low_M. There were significant peaks of activity in the same regions when participants in Group High_M reflected on their own performance.

Moreover, there was a significant positive correlation between activity in the paracingulate peak (10 46 36) that was enhanced when Group High_M made errors, and self-image as rated by the participants (r2 = 0.39, P < 0.05 one-tailed test) (Fig. 5). This correlation partialed out the effect of the ratings for motivation.

Figure 5.

There was a significant positive correlation between activity in the paracingulate peak (10 46 36) that was enhanced when Group High_M made errors, and self-image as rated by the participants. This correlation partialed out the effect of the ratings for motivation.

Figure 5.

There was a significant positive correlation between activity in the paracingulate peak (10 46 36) that was enhanced when Group High_M made errors, and self-image as rated by the participants. This correlation partialed out the effect of the ratings for motivation.

Interregional Interactions Specific to Group High_M

Further, we investigated whether there were significant interregional interactions between the anterior paracingulate cortex (10 46 36) and other areas. We investigated if there were greater interactions for Group High_M than for Group Low_M. The analysis showed that for Group High_M the activity in the anterior paracingulate cortex interacted with activity in the right inferior prefrontal cortex (area 47) (48 22 −6, t = 4.37) and the right orbitofrontal cortex (26 40 −12, t = 3.87).

The peak in the inferior prefrontal cortex was close to the peak that we found when we compared the subjective ratings for Group High_M versus Group Low_M. To show this we used the cluster that we found in the PPI (Group High_M vs. Group Low_M) as a mask for the rating contrast. The activation for the rating turned out to be significant within this cluster (P < 0.05 corrected). Figure 4 shows that the 2 activations are close.

Discussion

Because we were working with people rather than animals, we were able to manipulate the desire to do well so as to fit with one's self-image. That our manipulation succeeded is shown by the fact that Group High_M differed from Group Low_M in the importance that they had felt of doing well on the task, as reported after scanning (Fig. 1b). This was true in spite of the fact that the groups did not differ in how difficult they found the task to be.

It is essential to note that the 2 groups also did not differ in the accuracy with which they performed this task. As explained, this was guaranteed by the design that we used. We were aware, for example, that activations in the anterior cingulate cortex might differ according to the number of errors made, or even the likelihood that errors might be made (Nieuwenhuis et al. 2007). The consequence of our design is that we manipulated motivation while controlling the performance as assessed by the percentage of errors. We achieved this performance matching by adjusting the stimulus presentation rate. This meant that the ISI differed between subjects. We controlled for this by comparing errors with correct trials the ISI was controlled for within subjects.

It may be because we equated for effort that we did not observe any increased sustained activity for Group High_M compared with Group Low_M. In a normal setting, the motivated group would have performed better (Gendolla and Richter 2005). Here, the groups found the task to be equally difficult, and would have needed to sustain their attention throughout. Nevertheless, the analysis of task events showed that the effect of manipulating self-motivation was dramatic (Fig. 2). When Group High_M made errors, there was enhancement in many brain areas, including the medial frontal cortex, and the superior, middle, inferior and orbitofrontal gyri, compared with Group Low_M (Table 2).

This enhancement was only found on error trials. The participants did not receive feedback. However, it is known that people are able to detect that they are making errors also without external feedback (Rabbitt 1966). This is common on tasks where fast responses are required. It would have been particularly easy to notice errors on the n − 1 task. Both groups made a little over 10% of errors in this condition (Group High_M 11.8%, Group Low_M 13.9%; P < 0.30). So it is not that participants in Group Low_M would have been unaware of making errors. Rather, errors took on a greater significance for members of Group High_M. Learning from errors is a critical feature of human cognition. It underlies our ability to tune behavior for optimal performance. Group High_M were primed to view their performance as being of particular importance for their self-image. We note that in a related EEG study Hajcak et al. (2005) found that manipulation of motivation influenced the ERN (error-related negativity) but not the CRN (correct response negativity). It has also been shown that error-related activity in the cingulate cortex predicts future behavioral adjustment (Kerns et al. 2004; Klein et al. 2007).

That Group High_M paid especial attention to errors is shown by the enhanced activations in the medial frontal cortex; the anterior cingulate sulcus and in the overlying pre-Supplementary motor area (preSMA) (Fig. 2). The cingulate coordinate is close to the cingulate peak at which (Carter et al. 1998) reported activity that was related to error monitoring on the Stroop task (4 25 43). It also lies just above the region to which (Debener et al. 2005) assigned the error-related negative response in a combined EEG and fMRI study (−2 20 30).

Figure 3 shows that there was activation in common for the 2 groups on error trials, however one interesting difference in activation between the groups was found in the anterior paracingulate cortex. There is extensive evidence that the anterior paracingulate cortex is activated when participants reflect on themselves (Kelley et al. 2002; Fossati et al. 2004; Macrae et al. 2004; Ochsner et al. 2005). Thus, we suggest that members of Group High_M treated errors as being in conflict with their wish to do well.

Consistent with our proposal is that there was also activation in the anterior paracingulate cortex when Group High_M rated how they were doing. The peak in the anterior paracingulate cortex was in the same region where we found enhanced activity for Group High_M on error trials (Fig. 4). Further support for the involvement of the anterior paracingulate cortex in self-monitoring was the correlation between the activity in this region and the rating of self-image reported by Group High_M (Fig. 5). The more positive self-image, the stronger was the activation of the paracingulate cortex when making errors on the memory task.

In previous studies investigating the role of the paracingulate cortex in self-reflection participants have explicitly been judging trait words (Kelley et al. 2002; Fossati et al. 2004; Macrae et al. 2004; Ochsner et al. 2005). Here we show that activity in area 32 is enhanced during a working memory task when the participants spontaneously self-reflect. This is so because people have beliefs about their own ability to handle a task, and so their performance is assessed partly in terms of their own internal self-evaluations (Ryan 1982; Bandura 1989). For example, women who are told that females perform badly on a maths task then go on to perform worse than they would otherwise do (Spencer et al. 1999). Our data do not tell us whether the enhanced response in Group High_M results from the mismatch between their performance and their intrinsic ideal, or from social embarrassment in performing poorly in public.

Our peak is located in the dorsal part of the rostral medial prefrontal cortex. The peak is similar to that reported by Ochsner et al. (2005) for reflected appraisal of self (−12 48 36), as well as that reported by Macrae et al. (2004) when participants were judging self-relevance of trait words (19 40 32). It is also close to activations found when people make inferences about another person's mental state (Kelley et al. 2002; Mitchell et al. 2002).It is true that many studies find enhanced activation in the ventral part of the anterior medial prefrontal cortex on self-referential tasks (Kelley et al. 2002; Fossati et al. 2004; Ochsner et al. 2004; Mitchell et al. 2005; Jenkins et al. 2008). However, the pattern of anatomical connections suggests a functional distinction between the ventral and the dorsal paracingulate cortex (Ochsner et al. 2005). Whereas the ventral part is interconnected with emotion regulating areas, the dorsal part is not (Ongur et al. 2003). Activations in the dorsal region reflect cognitive processes, as in reasoning about oneself and others (Ochsner et al. 2005). In the present study, people were told that their cognitive capacity was being evaluated, which evokes a mental state of “me as an intelligent person”. This is a cognitive rather than emotional self-schema.

The second peak for the questions lay in right inferior frontal cortex (area 47). This is the same area that showed up in the analysis of effective connectivity for the n-back task (Fig. 4). There was an interaction in Group High_M between the anterior paracingulate cortex and the right inferior frontal cortex (area 47). The activations lay in the horizontal ramus of the Sylvian fissure. Its location corresponds well to the activation reported when participants learned that they had made an error on the first shift trial on the Wisconsin Card Sorting test (Monchi et al. 2001). It is also in similar location of the activity found in a meta-study by Dosenbach et al. (2006) when participants made errors on cognitive tasks. Activation in the inferior frontal area 47 has also been reported when participants judged that negative traits were relevant to themselves, as well as that positive traits did not apply to themselves (Moran et al. 2006). Thus the area appears relevant for shifting set on cognitive tasks.

If errors took on a particular significance for Group High_M, we might expect them to react with more emotion. In Group High_M there were also enhanced activations in the anterior insula proper and the orbitofrontal cortex. Activation has been reported in the insula when participants were aware, as opposed to unaware, of making errors (Klein et al. 2007). There is extensive evidence to associate activation in the anterior insula with the experience of negative affective states (Craig 2002; Singer et al. 2004). Activity in the lateral orbitofrontal cortex has been reported when participants experience punishment, e.g. monetary loss (O'Doherty et al. 2001; Kringelbach and Rolls 2004). In our study there was a significant PPI between the anterior paracingulate cortex and the right lateral orbitofrontal cortex. We suggest that this relates to emotional responses that results from reflection on errors. This is consistent with the finding showing that when participants were told their abilities were being evaluated, there was an increase in cardiovascular and electrodermal reactivity (Gendolla and Richter 2005).

In conclusion, our results show that telling participants that the task is a measure of intelligence increases their responses to errors. We stress the fact that the paracingulate activation lay anteriorly on the medial prefrontal surface because Amodio and Frith (2006) have suggested that functionally there is an anterior to posterior gradient on the medial frontal wall. The proposal is that the ability to reflect on oneself depends on tissue that is anterior to the cortex that is involved in monitoring actions or experiencing bodily sensations. We suggest that participants in Group High_M were reflecting, not on a sensation, but on whether their performance matched their sense of themselves as intelligent people.

Only humans harbor such ideals. According to Bandura (1989) children acquire expectancies and standards from others during their development and education. They then regulate their own behavior according to these standards via self-reinforcement. In this way humans gain control over their own destiny. The present study suggests that the paracingulate area 32 is closely involved in monitoring our progress toward these ideals, thus matching what we do to what we hope for ourselves.

Supplementary Material

Supplementary material can be found at: http://www.cercor.oxfordjournals.org/.

Funding

Wellcome Trust.

We are grateful to Will Penny for programming the task, and Tom Schofield for technical support, and Ray Dolan, Chris Frith, and Hans Bengtsson for valuable discussions. Conflict of Interest: None declared.

References

Amiez
C
Joseph
JP
Procyk
E
Reward encoding in the monkey anterior cingulate cortex
Cereb Cortex.
 , 
2006
, vol. 
16
 (pg. 
1040
-
1055
)
Amodio
DM
Frith
CD
Meeting of minds: the medial frontal cortex and social cognition
Nat Rev Neurosci.
 , 
2006
, vol. 
7
 (pg. 
268
-
277
)
Bandura
A
Pervin
LA
Self-regulation of motivation and action through internal standards and goal systems
Goal concepts in personality and social psychology
 , 
1989
Hillsdale (NJ)
Erlbaum
(pg. 
19
-
85
)
Braver
TS
Cohen
JD
Nystrom
LE
Jonides
J
Smith
EE
Noll
DC
A parametric study of prefrontal cortex involvement in human working memory
Neuroimage.
 , 
1997
, vol. 
5
 (pg. 
49
-
62
)
Burns
D
Feeling good
 , 
2007
New York
HarperCollins
Carter
CS
Braver
TS
Barch
DM
Botvinick
MM
Noll
D
Cohen
JD
Anterior cingulate cortex, error detection, and the online monitoring of performance
Science.
 , 
1998
, vol. 
280
 (pg. 
747
-
749
)
Craig
AD
How do you feel? Interoception: the sense of the physiological condition of the body
Nat Rev Neurosci.
 , 
2002
, vol. 
3
 (pg. 
655
-
666
)
Csikszentmihalyi
M
Beyond boredom and anxiety
 , 
1975
San Francisco
Jossey-Bass
Debener
S
Ullsperger
M
Siegel
M
Fiehler
K
von Cramon
DY
Engel
AK
Trial-by-trial coupling of concurrent electroencephalogram and functional magnetic resonance imaging identifies the dynamics of performance monitoring
J Neurosci.
 , 
2005
, vol. 
25
 (pg. 
11730
-
11737
)
Dosenbach
NCF
Visscher
KM
Palmer
ED
Miezin
FM
Wenger
KK
Kang
HC
Burgund
ED
Grimes
AL
Schlaggar
BL
Petersen
SE
A core system for the implementation of task sets
Neuron.
 , 
2006
, vol. 
5
 (pg. 
799
-
812
)
Fossati
P
Hevenor
SJ
Lepage
M
Graham
SJ
Grady
C
Keightley
ML
Craik
F
Mayberg
H
Distributed self in episodic memory: neural correlates of successful retrieval of self-encoded positive and negative personality traits
Neuroimage.
 , 
2004
, vol. 
22
 (pg. 
1596
-
1604
)
Friston
KJ
Buechel
C
Fink
GR
Morris
J
Rolls
E
Dolan
RJ
Psychophysiological and modulatory interactions in neuroimaging
Neuroimage.
 , 
1997
, vol. 
6
 (pg. 
218
-
229
)
Friston
KJ
Holmes
A
Poline
JB
Price
CJ
Frith
CD
Detecting activations in PET and fMRI: levels of inference and power
Neuroimage.
 , 
1996
, vol. 
4
 (pg. 
223
-
235
)
Gendolla
GH
Richter
M
Ego involvement and effort: cardiovascular, electrodermal, and performance effects
Psychophysics.
 , 
2005
, vol. 
42
 (pg. 
595
-
603
)
Hajcak
G
Moser
JS
Yeung
N
Simons
RF
On the ERN and the significance of errors
Psychophysics.
 , 
2005
, vol. 
42
 (pg. 
151
-
160
)
Jenkins
AC
Macrae
CN
Mitchell
JP
Repetition suppression of ventromedial prefrontal activity during judgments of self and others
Proc Natl Acad Sci USA.
 , 
2008
, vol. 
105
 (pg. 
4507
-
4512
)
Kelley
WM
Macrae
CN
Wyland
CL
Caglar
S
Inati
S
Heatherton
TF
Finding the self? An event-related fMRI study
J Cogn Neurosci.
 , 
2002
, vol. 
14
 (pg. 
785
-
794
)
Kerns
JG
Cohen
JD
MacDonald
AW
Cho
RY
Stenger
VA
Carter
CS
Anterior cingulate conflict monitoring and adjustments in control
Science.
 , 
2004
, vol. 
303
 (pg. 
1023
-
1062
)
Klein
TA
Endrass
T
Kathmann
N
Neumann
J
von Cramon
DY
Ullsperger
M
Neural correlates of error awareness
Neuroimage.
 , 
2007
, vol. 
34
 (pg. 
1774
-
1781
)
Kringelbach
ML
Rolls
ET
The functional neuroanatomy of the human orbitofrontal cortex: evidence from neuroimaging and neuropsychology
Prog Neurobiol.
 , 
2004
, vol. 
72
 (pg. 
341
-
372
)
Leon
MI
Shadlen
MN
Effect of expected reward magnitude on the response of neurons in the dorsolateral prefrontal cortex of the macaque
Neuron.
 , 
1999
, vol. 
24
 (pg. 
415
-
425
)
Macrae
CN
Moran
JM
Heatherton
TF
Banfield
JF
Kelley
WM
Medial prefrontal activity predicts memory for self
Cereb Cortex.
 , 
2004
, vol. 
14
 (pg. 
647
-
654
)
McClelland
DC
Atkinson
JW
Clark
RA
Lowell
EL
The achievement motive
 , 
1953
New York
Appelton-Century-Crofts
Mitchell
JP
Banaji
MR
Macrae
CN
The link between social cognition and self-referential thought in the medial prefrontal cortex
J Cogn Neurosci.
 , 
2005
, vol. 
17
 (pg. 
1306
-
1315
)
Mitchell
JP
Heatherton
TF
Macrae
CN
Distinct neural systems subserve person and object knowledge
Proc Natl Acad Sci USA.
 , 
2002
, vol. 
99
 (pg. 
15238
-
15243
)
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
)
Moran
JM
Macrae
CN
Heatherton
TF
Wyland
CL
Kelley
WM
Neuroanatomical evidence for distinct cognitive and affective components of self
J Cogn Neurosci.
 , 
2006
, vol. 
18
 (pg. 
1586
-
1594
)
Nieuwenhuis
S
Schweizer
TS
Mars
RB
Botvinick
MM
Hajcak
G
Error-likelihood prediction in the medial frontal cortex: a critical evaluation
Cereb Cortex.
 , 
2007
, vol. 
17
 (pg. 
1570
-
1581
)
Ochsner
KN
Beer
JS
Robertson
ER
Cooper
JC
Gabrieli
JD
Kihsltrom
JF
D'Esposito
M
The neural correlates of direct and reflected self-knowledge
Neuroimage.
 , 
2005
, vol. 
28
 (pg. 
797
-
814
)
Ochsner
KN
Knierim
K
Ludlow
DH
Hanelin
J
Ramachandran
T
Glover
G
Mackey
SC
Reflecting upon feelings: an fMRI study of neural systems supporting the attribution of emotion to self and other
J Cogn Neurosci.
 , 
2004
, vol. 
16
 (pg. 
1746
-
1772
)
O'Doherty
J
Kringelbach
ML
Rolls
ET
Hornak
J
Andrews
C
Abstract reward and punishment representations in the human orbitofrontal cortex
Nat Neurosci.
 , 
2001
, vol. 
4
 (pg. 
95
-
102
)
Ongur
D
Ferry
AT
Price
JL
Architectonic subdivision of the human orbital and medial prefrontal cortex
J Comp Neurol.
 , 
2003
, vol. 
460
 (pg. 
425
-
449
)
Otten
LJ
Henson
RNA
Rugg
MD
State-related and item-related neural correlates of successful memory encoding
Nat Neurosci.
 , 
2002
, vol. 
5
 (pg. 
1339
-
1344
)
Platt
ML
Glimcher
PW
Neural correlates of decision variables in parietal cortex
Nature.
 , 
1999
, vol. 
400
 (pg. 
233
-
238
)
Rabbitt
PMA
Error correction time without external error signals
Nature.
 , 
1966
, vol. 
212
 pg. 
438
 
Ryan
RM
Control and information in the intrapersonal sphere: an extension of cognitive evaluation theory
J Pers Soc Psychol.
 , 
1982
, vol. 
43
 (pg. 
450
-
461
)
Shidara
M
Richmond
BJ
Anterior cingulate: single neuronal signals related to degree of reward expectancy
Science.
 , 
2002
, vol. 
296
 (pg. 
1709
-
1711
)
Singer
T
Seymour
B
O'Doherty
J
Kaube
H
Dolan
RJ
Frith
CD
Empathy for pain involves the affective but not sensory components of pain
Science.
 , 
2004
, vol. 
303
 (pg. 
1157
-
1162
)
Spencer
SJ
Steele
CM
Quinn
DM
Stereotype threat and women's math performance
J Exp Soc Psychol.
 , 
1999
, vol. 
35
 (pg. 
4
-
28
)
Spielberger
CD
Manual for the State-Trait Anxiety Inventory
 , 
1983
Palo Alto (CA)
Consulting Psychologists Press
Talaraich
J
Tournoux
P
Co-planar stereotaxic atlas of the human brain
 , 
1988
Stuttgart
Thieme