Behavioural Brain Research

Behavioural Brain Research 198 (2009) 420 428 Contents lists available at ScienceDirect Behavioural Brain Research journal homepage: www.elsevier.com/locate/bbr Research report Fast and slow brain rhythms in rule/expectation violation tasks: Focusing on evaluation processes by excluding motor action Gabriel Tzur, Andrea Berger Department of Psychology and Zlotowski Center for Neuroscience, Ben-Gurion University of the Negev, Beer Sheva 84105, Israel article info abstract Article history: Received 4 September 2008 Received in revised form 10 November 2008 Accepted 18 November 2008 Available online 3 December 2008 Keywords: ERN FRN EEG ERP ACC Conflict detection Error detection Theta rhythm has been connected to ERP components such as the error-related negativity (ERN) and the feedback-related negativity (FRN). The nature of this theta activity is still unclear, that is, whether it is related to error detection, conflict between responses or reinforcement learning processes. We examined slow (e.g., theta) and fast (e.g., gamma) brain rhythms related to rule violation. A time frequency decomposition analysis on a wide range of frequencies band (0 95 Hz) indicated that the theta activity relates to evaluation processes, regardless of motor/action processes. Similarities between the theta activities found in rule-violation tasks and in tasks eliciting ERN/FRN suggest that this theta activity reflects the operation of general evaluation mechanisms. Moreover, significant effects were found also in fast brain rhythms. These effects might be related to the synchronization between different types of cognitive processes involving the fulfillment of a task (e.g., working memory, visual perception, mathematical calculation, etc.). 2008 Elsevier B.V. All rights reserved. 1. Introduction From early childhood through adulthood and old age, a person needs to adapt and adjust to the surrounding. Monitoring of self-performance, meaning the capability to evaluate outcomes of self-actions and differentiate between correct and erroneous information, is a crucial process to this adjustment. Electro-physiological and brain-imaging studies have suggested that monitoring of selfperformance, such as detecting an error response or evaluating outcomes and feedbacks, is related to theta activity involving the anterior cingulated cortex (ACC) [1 6]. Event related potentials (ERPs) studies of self-performance monitoring have investigated two main ERP components: the errorrelated negativity (ERN) and the feedback-related negativity (FRN). Both components seem to involve the ACC and have been related to theta activity (4 8 Hz) [5 8]. The ERN is a negative component over the medial frontal cortex that follows error commission in choice reaction tasks, even in the absence of explicit performance feedback [3,5,7,9 13], and the FRN relates to a negative electrical deflection similar to ERN, that follows feedback associated with unfavorable outcomes (e.g., winnings/losses) [6,8,14]. [15] suggested that both Corresponding author at: Department of Psychology, Ben-Gurion University of the Negev, P.O.B. 653, Beer Sheva 84105, Israel. Tel.: +972 8 6477757; fax: +972 8 6472072. E-mail address: andrea@bgu.ac.il (A. Berger). ERN and FRN components are functionally similar, and reflect the operation of an error-processing system [3,9 11,13,15]. However, other studies have suggested that the role of these components is not exclusively related to error detection, and involves response conflict monitoring [1,6,16] or/and reinforcement learning signals [8,14,17]. Moreover, a study of Yeung et al. [6] suggests that during errors, conflict arises between the executed incorrect response and activation of the correct response due to ongoing stimulus evaluation. Therefore, the response conflict theory can account for both error and conflict detection tasks [6]. Nevertheless, this theory cannot explain the finding of FRN in the absence of overt responses, as has been reported by Yeung et al. [8]. In their study the results were explained by adopting the reinforcement learning theory, which suggests that the ACC involves processing motivationally significant information concerning rewards and punishments, and therefore can explain these components even in the absence of overt responses [8]. However, a new discrepancy arises in this case, since the view of the negative medio-frontal components as related to motivational processes cannot explain the appearance of such components in situations where there is no motivational aspect, such as in pure conflict situations eliciting the N2 component in the flanker task [6,18]. A recent study of Tzur and Berger [19] showed that rule-violation tasks, such as distinguishing between correct and incorrect simple mathematical equations (e.g., 1 + 2 =, correct solution 3 or incorrect solution 8 ), are related to theta activity (4 8 Hz) that seems to involve the ACC. Their time frequency analysis suggested 0166-4328/$ see front matter 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.bbr.2008.11.041

G. Tzur, A. Berger / Behavioural Brain Research 198 (2009) 420 428 421 a phase-lock increase in theta power for incorrect solutions compared to correct ones. Tzur and Berger [19] suggested that this effect might be related to a violation of expectation, that is, a conflict arising between the expected rule (e.g., 1+2=3 ) and the presented information, which violated that rule (e.g., 1+2=8 ). This idea was based on the conflict monitoring theory, according to which, as mentioned, the ACC monitors for the presence of conflict between simultaneously active but incompatible processing streams [1,6,16,20,21]. Tzur and Berger [19] proposed that this view of the ACC function, as reflected in theta activity, should be expanded and include not only the monitoring of response-conflicts, but also the monitoring of conflict between expectations. Nevertheless, it could still be argued that the theta activity observed by Tzur and Berger [19] relates to motor response planning and/or executing processes, since the task that was used included a verification motor response (e.g., press key 1 for the correct solution and key 4 for the incorrect solution). Therefore, the first aim of the present study was to compare the effect of a phase-lock increase in theta activity (4 8 Hz) for an incorrect solution in a rule-violation task [19], to the one found even in the complete absence of overt responses [22]. Finding similar patterns of theta activity would strongly support the idea that this phase-lock theta activity relates also to an evaluation process, regardless of any motor response planning and/or execution. Participants were presented with mathematical equations, and were asked to distinguish between correct and incorrect solutions without any overt response, that is, just by looking passively at the mathematical exercises and their presented solutions (this will be referred to as the passive group). The second aim was to evaluate similarities found between time frequency analyses of rule-violation tasks (from this study) and analyses related to the ERN and FRN components [5,7,8,14]. Finding such similarities would suggest that these neural activities might reflect the operation of a generic evaluation mechanism, meaning that the cognitive processes which are related to the ERN and FRN [5,15,17] are also involved in situations of ruleviolation. The EEG (electroencephalogram) data collected from the passive group was partially obtained from the study of Berger et al. [22], and was analyzed with wider time frequency analyses (i.e., analyzing the relative power and phase synchrony of frequency bands ranging from 1 to 95 Hz). These analyses were used also on the EEG data collected from Tzur and Berger [19] (this EEG data will be referred to as the active group), which were compared to the passive group analyses. We hypothesized that the time frequency analyses of both groups (passive and active) would reflect similar neural processing patterns, related to an increase in theta frequency band (4 8 Hz) phase-lock power and phase synchrony, for incorrect solutions compared to correct ones. We used a time frequency decomposition analysis, from which one can obtain estimates of instantaneous power, that is, energy at different frequencies [23], and inter-trial phase synchrony, that is, consistency of oscillation onset across trials [24]. This was done on a wide frequency band (1 95 Hz), including upper gamma, which to our knowledge has not yet been used to examine rule-violation tasks. This new approach of time frequency analyses has presented a novel view of understanding brain activity related to cognitive processes, beyond the classic averaged ERPs [25,26]. Examining neural synchronization and its related energy at a wider frequency band (1 95 Hz) should contribute to a better understanding of these neural cognitive processes. Cohen et al. [14] reported an increase in power and phase synchrony in both theta and gamma frequency bands (over the medial frontal cortex) when participants received negative feedback (i.e., FRN) on their actions. Following this and our second aim, we expected to find an increased power and phase synchrony in the gamma frequency band as well. 2. Materials and methods 2.1. Participants There were 2 groups of participants: passive group 28 participants (20 females and 8 males), with a mean age of 24.16 years (SD = 2.35); active group (from Tzur and Berger, [19]) 17 participants (14 females and 3 males), with a mean age of 23.8 years (SD = 1.3). All participants were right-handed and were students at Ben- Gurion University of the Negev. They were all healthy with no history of neurological illnesses and had normal or corrected-to-normal vision. Participants gave informed consent and participated in the study as partial fulfillment of course requirements. 2.2. Procedure Participants were presented with 360 trials (plus 30 practice trials) of simple mathematical equations (addition or subtraction), which were followed by either correct solutions (180 trials) or incorrect solutions (180 trials). Within the incorrect solution condition, there were three possible levels of deviation, appearing with equal probability. For example, for the equation 1+2=, the incorrect solution could be either 4 (L1), 6 (L3) or 8 (L5). The number of positive and negative deviations (of incorrect solutions from correct ones) was equal. Equations that had identical operands (e.g., 3 + 3, 4 + 4) were excluded. The 360 trials were presented in a random order in four blocks (45 correct and 45 incorrect trials in each block). Each trial began with a fixation point (500 ms), followed by an equation (1,500 ms), then a black screen (600 ms for baseline calculation), and ended with a solution (1,500 ms). Random inter-trial intervals (ITIs; 200/400/600 ms) were inserted in order to reduce a monotonous task rhythm. Participants were seated 60 cm in front of a computer monitor and asked to be as relaxed as possible in order to reduce muscle tension. They were told at the beginning of the experiment that they were participating in cognitive research in the field of numerical processing, and that they would be presented with simple mathematical equations followed by either correct or incorrect solutions. They were asked to silently distinguish between correct and incorrect solutions, that is, just by looking at exercises and the presented solution without any overt response. Except for the absence of an overt response, this passive procedure is identical to the active one used by Tzur and Berger [19]. 2.3. Electroencephalogram (EEG) recording The EEG was recorded from 128 scalp sites using the EGI Geodesic Sensor net and system [27]. Electrode impedances were kept below 40 k, an acceptable level for this system [28]. All channels were referenced to the Cz channel and data was collected using a 0.1 100 Hz bandpass filter. Signals were collected at 250 samples per second and digitized with a 16-bit A/D converter. 2.4. Time frequency analysis Time frequency analysis of the data was conducted using a wavelet-based analysis [23,24]. Before the wavelet analysis, each participant s raw (0.1 100 Hz) EEG data was segmented into trials, time-locked to the presentation of the solution. The segmented data was inspected for artifacts (e.g., bad-channels resulting from channel-saturation, muscle movement, etc.) while excluding channels within each segment that exceeded the fast average amplitude of 200 V or the differential average amplitude of 100 V. Segments having 10 or more bad channels were excluded, and segments with fewer than 10 bad channels were included after replacing the bad-channel data with spherical interpolation of the neighboring channel values. Prior to wavelet analysis, the data of each trial was re-referenced to the average of all of the sensors at each time point. For calculating the phase-lock power values, trials were averaged into correct and three incorrect (i.e., L1, L3, L5) conditions (stimulus-locked to the solution presentation) [29]. For calculating the total power (that is, phase- and non-phase-lock) and phase synchrony values between trials [24,30], trials were kept unaveraged. Following this, a family of Morlet wavelets was constructed at intervals of 0.5 Hz frequency, ranging from 1 to 95 Hz. Our wavelet family was computed using a f 0/ f ratio of 7 [23,31]. The power values (i.e., squared amplitude) and phase synchrony values (range from 0 no synchrony, to 1 full synchrony) were normalized with respect to a 200 to 0 ms pre-solution baseline. The time frequency analysis was conducted for the frequency bands raging from 1 to 45 Hz and 65 to 95 Hz, excluding the 45 65 Hz band, since this was in the range of our electrical power network frequency and might have been vulnerable to electromagnetic interference (EMI). The statistical analyses were done on the mean of a group of four channels, located between Cz and Fz of the 10 20 system (of electrode placement). This localization is comparable to the ERN and FRN components [5,8,14], see Fig. 1 (top row). The wavelet power and phase synchrony analyses of both groups (active and passive) were conducted in the following way: For each condition (i.e., correct, L1, L3, L5), the adaptive-mean (calculated from a time-window of ±50 ms centered around a local maxima) of the power and phase synchrony from each frequency band

422 G. Tzur, A. Berger / Behavioural Brain Research 198 (2009) 420 428 Fig. 1. Grand averaged voltage distribution in two-dimensional scalp topographic maps (top row) and the ERP mean of the group of four channels (bottom row) of 17 participants from the active group (filtered with 4 12 Hz bandpass). Greater negative voltage distributions (top row, circled in white) and ERP (bottom row) are seen for the incorrect solution conditions compared to the correct one (about 255 ms after the solution presentation) over the medial frontal cortex. This topographical location is similar to the location of the ERN and FRN components. (i.e., delta: 1 4 Hz, theta: 4 8 Hz, alpha: 8 12 Hz, beta: 12 30 Hz, lower gamma: 30 45 Hz; medial gamma: 65 80 Hz and upper gamma: 80 95 (the gamma band was divided into three equal bands for a better resolution of the data)) was extracted from a 0 600 ms time-window for each participant. This time-window captures the theta effects seen in the time frequency analyses of the passive and active groups (Figs. 2 4), and is compatible with theta effects reported in error detection and evaluation tasks [14,19]. The extracted values of the power (total and phase-lock) and phase synchrony were then analyzed separately using repeated measures analysis of variance (ANOVAs) [32] with the solution conditions and the frequency bands as withinsubject variables (significance level was set to.05). Within each of the solution frequency interaction effects, a planned comparison (i.e., A vs. B) was conducted comparing correct (A) vs. incorrect (B) conditions for each frequency band separately. This was done in order to evaluate which of the Fig. 2. Spectral phase-lock power distribution (time-window of 0 600 ms) in two-dimensional scalp topographic maps of the active (top row) and passive (bottom row) groups incorrect condition (L5). The color spectrum indicates the relative power intensity (i.e., percentage of total power) of the investigated frequencies. That is, dark areas indicate low power, whereas light areas denote high power. A relative increase in theta power is seen over the medial-frontal cortex for both groups. These topographical locations and time course are compatible with Fig. 1 and similar to the location of the ERN and FRN components.

G. Tzur, A. Berger / Behavioural Brain Research 198 (2009) 420 428 423 Fig. 3. An example of total power (top), phase-lock power (middle) and phase synchrony (bottom row) time frequency plots (from the Fz channel) of one individual participant from the active group. A relative increase in both phase-lock power and phase synchrony is seen mostly in the theta band (4 8 Hz) for the incorrect condition (L5, right column) compared to the correct one (left column). frequency bands expressed power and phase synchrony differences between the correct and incorrect conditions. Whenever a significant difference was found, two additional planned orthogonal comparisons L1 (A) vs. (L3 and L5) (B) and L3 (A) vs. L5 (B) were conducted sequentially only if the previous comparison reached statistical significance. This was done in order to find out whether greater deviations of the incorrect solution from the correct one (i.e., L3 and L5) were related to greater power and/or phase synchrony increases than for smaller deviations (i.e., L1). 3. Results 3.1. Active group In the phase-lock power planned comparisons analysis, only in the theta (4 8 Hz) frequency band did the first two sequential comparisons (i.e., correct vs. incorrect and L1 vs. (L3 and L5)) reach statistical significance (Table 1). These results are consistent with those of Tzur and Berger [19], and indicate that only the theta effects depended on the degree of deviation of the incorrect solution from the correct one, showing greater power for greater deviations (Fig. 5, middle-left, and Table 1). Nevertheless, a significant increase for the incorrect condition compared to the correct one was found also in the upper-gamma (80 95 Hz) frequency band. In the total power analysis, a greater increase for the incorrect solution is also seen in the theta band (Fig. 5, upper-left), but the effects did not reach statistical significance (Table 1). Moreover, the phase synchrony planned comparisons analysis revealed a significant increase in phase synchrony for the incorrect condition compared to the correct one (Fig. 3, bottom and Fig. 5, bottom-left) in all frequency bands except for the delta band (i.e., theta, alpha, beta and gamma) (Table 1). Both phase-lock power and phase synchrony patterns across the frequency bands (Fig. 5, middle/lower-left, and Table 1) indicate dominant effects in the theta band that were dependent on the deviation degree of the

424 G. Tzur, A. Berger / Behavioural Brain Research 198 (2009) 420 428 Fig. 4. An example of total power (top), phase-lock power (middle) and phase synchrony (bottom row) time frequency plots (from the Fz channel) of one individual participant from the passive group. Similar to the active group (Fig. 3), a relative increase in both phase-lock power and phase synchrony is seen mostly in the theta band (4 8 Hz) for the incorrect condition (L5, right column) compared to the correct one (left column). incorrect solution from the correct one, showing greater phase-lock power and phase synchrony for greater deviations. 3.2. Passive group Phase-lock power and phase synchrony patterns across the frequency bands (Fig. 5, middle/lower-right) indicate dominant effects in the theta band, showing greater phase-lock power and phase synchrony for incorrect solutions compared to correct ones, similar to the active group (Fig. 5, middle/lower-left). This similarity can also be seen in Figs. 3 and 4. An increase in phase-lock power and phase synchrony were found in the incorrect condition compared to the correct one in all frequency bands except for the delta band (i.e., theta, alpha, beta and gamma) (Table 2). Planned comparisons between the incorrect conditions (i.e., L1 vs. (L3 and L5) and ) reached statistical significance only in the upper gamma band (Table 2). The total power analysis indicated a greater increase for the incorrect solution only in the beta and lower/upper gamma bands (Table 2). However, it seems that in the passive group the power (total and phase-lock) and the phase synchrony values were to some extent lower than in the active group, especially in the theta band (Fig. 5). In order to test this difference, a group (active and passive) categorical variable was added to the ANOVA. Within each of the group frequency interaction effects, a planned comparison (i.e., A vs. B) was conducted comparing differences in power and phase synchrony between the passive (A) and the active (B) groups. As shown in Table 3, the major decrease is found in the theta activity. In this frequency band both power (total and phase-lock) and phase synchrony indicate a significant decrease in the passive group compared to the active one. A significant decrease in phase-lock power and phase synchrony is found also in the delta band of the

G. Tzur, A. Berger / Behavioural Brain Research 198 (2009) 420 428 425 Table 1 Active group simple comparison analyses. Frequency Comparison (A vs. B) Total power Phase-lock power Phase synchrony F-value p-value F-value p-value F-value p-value Delta 1 4 Hz Crr vs. Incrr (L1, L3, L5).090.767.387.542 1.148.299 L1 vs. (L3 and L5) Theta 4 8 Hz Crr vs. Incrr (L1, L3, L5).436.518 4.933.042 * 19.123.000 * L1 vs. (L3 and L5) 8.334.011 * 9.744.007 *.995.333 1.138.302 Alpha 8 12 Hz Crr vs. Incrr (L1, L3, L5) 3.611.076 2.082.168 85.404.000 * L1 vs. (L3 and L5).024.878 Beta 12 30 Hz Crr vs. Incrr (L1, L3, L5) 3.473.081 1.780.294 81.696.000 * L1 vs. (L3 and L5).590.453 Lower Gamma 30 45 Hz Crr vs. Incrr (L1, L3, L5) 3.278.089.575.459 148.178.000 * L1 vs. (L3 and L5) 0.285.600 Medial Gamma 65 80 Hz Crr vs. Incrr (L1, L3, L5).715.410 2.400.141 260.236.000 * L1 vs. (L3 and L5) 3.507.079 Upper Gamma 80 95 Hz Crr vs. Incrr (L1, L3, L5).707.412 6.220.024 * 179.879.000 * L1 vs. (L3 and L5) 3.042.100.099.757 Abbreviation: Crr = Correct, Incrr = Incorrect. * Indicates statistically significant effects (p <.05), B > A. passive group. However, an opposite effect is found in the power comparison in the lower gamma frequency band (Table 3). 4. Discussion In the present study we examined similarities of medial-frontal slow and fast brain rhythms found in rule-violation tasks involving motor responses (i.e., the active group) and those found in ruleviolation tasks with the absence of overt responses (i.e., the passive group). A time frequency decomposition analysis showed similar phase synchrony and power patterns for both groups (Figs. 3 5). We found a significant increase in both phase synchrony and phase-lock power, especially in the theta band (4 8 Hz) (Figs. 3 5) for the incorrect compared to the correct condition (Tables 1 and 2). This finding supports the idea that this theta activity also relates to evaluation processes, which are not exclusively related to motor response monitoring. Focusing on evaluation processes while excluding motor response processes is in line with the idea that the theta effect seen in rule/expectation violation tasks may be explained Table 2 Passive group simple comparison analyses. Frequency Comparison (A vs. B) Total power Phase-lock power Phase synchrony F-value p-value F-value p-value F-value p-value Delta 1 4 Hz Crr vs. Incrr (L1, L3, L5).016.899.680.416.193.664 L1 vs. (L3 and L5) Theta 4 8 Hz Crr vs. Incrr (L1, L3, L5).071.792 5.370.028 * 29.024.000 * L1 vs. (L3 and L5).092.764 1.378.251 Alpha 8 12 Hz Crr vs. Incrr (L1, L3, L5).003.953 8.339.008 * 73.782.000 * L1 vs. (L3 and L5).160.693.038.846 Beta 12 30 Hz Crr vs. Incrr (L1, L3, L5) 6.255.019 * 4.298.048 * 51.161.000 * L1 vs. (L3 and L5).001.981.461.503 3.636.067 Lower Gamma 30 45 Hz Crr vs. Incrr (L1, L3, L5) 6.297.018 * 22.421.000 * 194.913.000 * L1 vs. (L3 and L5).733.399 2.127.156 1.969.171 Medial Gamma 65 80 Hz Crr vs. Incrr (L1, L3, L5) 2.873.102 13.794.000 * 489.832.000 * L1 vs. (L3 and L5).908 349 1.493.232 Upper Gamma 80 95 Hz Crr vs. Incrr (L1, L3, L5) 4.753.038 * 16.788.000 * 166.983.000 * L1 vs. (L3 and L5) 2.812.105.221.642 7.219.012 * 5.291.029 * Abbreviation: Crr = Correct, Incrr = Incorrect. * Indicates statistically significant effects (p <.05), B > A.

426 G. Tzur, A. Berger / Behavioural Brain Research 198 (2009) 420 428 Fig. 5. Active (left column) and passive (right column) groups total power (top), phase-lock power (middle) and phase synchrony (bottom) in the different frequency bands for the different conditions. Greater phase-lock power and phase synchrony neural activities are found mostly in the theta band for the incorrect conditions. in terms of violation of expectation processes, that is, a conflict/mismatch arises between the expected rule (e.g., 1+2=3)and the presented violation (e.g., 1+2=8)[19]. Our present study is consistent with, and even broadens, Tzur and Berger s [19] suggestion on violation of expectation processes in rule/expectation violation tasks, suggesting that the theta activity found in ERN and FRN components and rule/expectation violation tasks might all be connected to a generic evaluation process that compares and analyzes the similarities and differences between an expected stimulus/action (e.g., feedback, a solution of a mathematical equation, or performing an incorrect action/response) and a presented/performed stimulus/action. This suggestion predicts that the larger the conflict/mismatch between the expected and the presented/performed stimulus/action is, the greater the neural Table 3 Passive (A) vs. active (B) group comparisons between groups. Frequency Total power Phase-lock power Phase synchrony F-value p-value F-value p-value F-value p-value Delta 1 4 Hz 1.841.182 9.418.004 * 31.368.000 * Theta 4 8 Hz 5.815.020 * 7.282.010 * 12.316.002 * Alpha 8 12 Hz 1.142.291.555.460.093.762 Beta 12 30 Hz 2.555.117 1.718.197 2.527.119 Lower Gamma 30 45 Hz 5.177.028 ** 7.611.009 **.000.979 Medial Gamma 65 80 Hz.106.746.357.553 3.317.076 Upper Gamma 80 95 Hz 1.532.223.602.442.219.642 Abbreviation: Pas Grp = Passive Group, Act Grp = Active Group. * Indicates statistically significant effects (p <.05), B > A. ** Indicates statistically significant effects (p <.05), A > B.

G. Tzur, A. Berger / Behavioural Brain Research 198 (2009) 420 428 427 energy (power) and phase synchrony in the theta band (4 8 Hz) will be. This idea is fully consistent with findings showing amplitude sensitivity to more salient outcomes in both ERN and FRN components, that is, greater ERN and FRN amplitudes were found when outcomes were unexpected and/or unfavorable [33 35]. We believe this idea is consistent with all three existing explanations in the literature regarding the ERN: the conflict detection, the error detection and the reinforcement learning. The conflict view holds that the ACC monitors for the presence of conflict between simultaneously active but incompatible processing streams. It suggests that during errors, a conflict arises between the executed incorrect response and activation of the correct response, due to ongoing stimulus evaluation processes [1,6,16,20,21]. This is consistent with the idea that the bigger the conflict, the greater the ACC activity observed. According to our broader definition of conflict, it can occur when there is a mismatch between the perceived stimulus and an expected one. In our paradigm, the more discrepant the solution of the equation was from the expected solution, the stronger the theta activity that was obtained. The error view states that the ERN reflects ACC processing that is directly related to detecting the error. This theory predicts that ERN and ACC activity should increase directly with the dissimilarity of the error from the correct response [3,9 11,13,15]. Again, if broadening the definition of an error to include perceived errors that are not directly connected to the participants actions, our results are consistent with the idea that the bigger the error (the larger the discrepancy), the larger the ACC response. The proposed evaluation process could even be compatible with a reinforcement learning perspective, as it is capable of rapidly determining whether feedback is better or worse than expected, and encode this difference between expectations and actual outcomes as reward prediction errors. Single unit recording studies in nonhuman primates suggest that this is indeed the case, with more unexpected outcomes yielding larger neural responses in midbrain dopamine neurons [36]. The neural activity found in rule/expectation violation tasks seems to have many similarities with the neural activity found for the ERN and FRN components. As mentioned in the introduction, Tzur and Berger [19] suggested the ACC as a possible source for theta activity found in rule/expectation violation tasks, which is in line with the vast literature connecting the ACC and theta activity to cognitive processes related to error detection and conflict between competing cognitions and responses [1 3,5 7,16,20,21]. Moreover, the increase in neural power and phase synchrony found in the theta band for the incorrect condition (Figs. 3 5, and Tables 1 and 2) is compatible with the study of Cohen et al. [14], which reported similar effects in the theta and gamma bands when participants received negative feedback (i.e., FRN) on their actions. This is also in line with the studies of Luu et al. [5,7] that connected the ERN and FRN to an increase in theta amplitude and power. Furthermore, even the relative reduction found in theta activity for the passive group (i.e., no overt responses) compared to the active group (i.e., verification response) (Figs. 3 5 and Table 3) is in line with the study of Yeung et al. [8], which reported a reduction in the FRN amplitude when participants made no active choices and no overt actions/responses in simple monetary gambling tasks. These findings support the idea that these similarities in neural activities might reflect the operation of a generic mechanism, as was previously suggested for the ERN and FRN [5,15,17], while expanding it also to rule-violation tasks. Nevertheless, there are at least two possible explanations for the reduction in neural activity found in the passive group. First, it is possible that this reduction is related to the absence of motor responses, which may involve neural activity cooperation and synchronization between evaluation processes (e.g., violation of expectation) and motor actions processes (e.g., planning and/or execution). Second, the passive group reported that the task was very monotonous and that they had difficulty in staying engaged and committed to the task. This might have caused a reduction in the quality of their evaluation processes (e.g., less sensitive and thorough evaluations) and therefore a reduction in neural activity. Moreover, this idea can account for finding theta sensitivity to more salient violations only for the active group (Fig. 5 and Table 1), that is, only the active group showed greater power and phase synchrony in the theta band for more salient violations than lesser ones, suggesting a better evaluation process for the active group compared to the passive one. This second suggestion is also in line with the idea that the orientation of pyramidal cells in the ACC could generate a medial frontal activity like the FRN, whereas cortical layers in the nearby cingulate gyrus and supplementary motor area (SMA) are oriented tangentially to the scalp, hence they would not be expected to produce a corresponding scalp potential [17], and therefore might be less related to motor response processes. Nevertheless, our findings cannot rule out the possibility that the enhancement in theta activity was partly related to motor activity involving the verification response in the active group. Moreover, the present study findings are consistent with the idea that both types of monitoring processes, that is, evaluation (discrepancy with expectancy) and response monitoring, contribute to the medial frontal engagement, and the idea of a generic mechanism for error detection and evaluation. When addressing evaluation processes based on expectation violation processes, as suggested in this study, we should also address additional parallel cognitive processes necessary for the fulfillment of the task, such as numerical and arithmetical calculation processing, comparing processes, working memory, visual processing, etc. These cognitive processes often interact with each other from distant parts of the brain, and therefore require synchronization and coordination. Recent important studies have suggested that theta activity modulates gamma activity [37] and that fast brain rhythms (above 20 Hz, that is, beta and gamma bands) enable a precise functional association between specific brain regions over short as well as longer distances [38,39]. Furthermore, it seems that neural activities of fast brain rhythms (i.e., beta and gamma) are most likely related to synchronization between brain regions/processes, and that neural activities of slow brain rhythms (i.e., theta and alpha) are associated with memory processes (e.g., working memory, memory consolidation, encoding and retrieval) [25,38 40]. These findings may relate to the idea of expectation violation processes, that is, a retrieval of the expected stimuli/action, while retaining and comparing it to the presented stimuli/action. This also may relate to neural energy (power) and phase synchrony effects found in the theta, alpha, beta and gamma frequency bands (Tables 1 and 2). For example, the theta and the alpha activities might be related to the memory and comparison processes that are involved in evaluating the magnitude of the conflict between the presented and expected stimuli/action, while the beta and gamma activities may involve synchronization between brain regions related to these memory and comparison processes and other parallel cognitive (e.g., perceptual integration, attention selection, mathematical calculation and response planning/execution) and affective processes (e.g., emotional evaluation) [41]. Nevertheless, these are post hoc interpretation that need to be investigated in future studies. Acknowledgments We would like to thank Michael I. Posner for his constructive suggestions and ideas, and Desiree Meloul for her professional and generous help.

428 G. Tzur, A. Berger / Behavioural Brain Research 198 (2009) 420 428 This study was supported by the United States Israel Bi-national Science Foundation, grants #2001047 and #2005133 to Dr. Andrea Berger and Prof. Michael I. Posner. References [1] 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;280:747 9. [2] Dehaene S, Posner MI, Tucker DM. Localization of a neural system for error detection and compensation. Psychol Sci 1994;5:303 5. [3] Gehring WJ, Goss B, Coles MGH, Meyer DE, Donchin E. A neural system for error detection and compensation. Psychol Sci 1993;4:385 90. [4] Klein TA, Endrass T, Kathmann N, Neumann J, von-cramon DY, Ullsperger M. Neural correlates of error awareness. Neuroimage 2007;34:1774 81. [5] Luu P, Tucker DM, Derryberry D, Reed M, Poulsen C. Electrophysiological responses to errors and feedback in the process of action regulation. Psychol Sci 2003;14:47 53. [6] Yeung N, Botvinick MM, Cohen JD. The neural basis of error detection: conflict monitoring and the error-related negativity. Psychol Rev 2004;111:931 59. [7] Luu P, Tucker DM, Makeig S. Frontal midline theta and the error-related negativity: neurophysiological mechanisms of action regulation. Clin Neurophysiol 2004;115:1821 35. [8] Yeung N, Holroyd CB, Cohen JD. ERP correlates of feedback and reward processing in the presence and absence of response choice. Cereb Cortex 2005;15:535 44. [9] Coles MGH, Scheffers MK, Holroyd CB. Why is there an ERN/Ne on correct trials? Response representations, stimulus-related components, and the theory of error-processing. Biol Psychol 2001;56:173 89. [10] Falkenstein M, Hohnsbein J, Hoormann J, Blanke L. Effects of crossmodal divided attention on late ERP components: II. Error processing in choice reaction tasks. Electroencephalogr Clin Neurophysiol 1991;78:447 55. [11] Falkenstein M, Hoormann J, Christ S, Hohnsbein J. ERP components on reaction errors and their functional significance: a tutorial. Biol Psychol 2000;51:87 107. [12] Luu P, Tucker DM. Regulating action: alternating activation of midline frontal and motor cortical networks. Clin Neurophysiol 2001;112:1295 306. [13] Scheffers MK, Coles MGH. Performance monitoring in a confusing world: errorrelated brain activity, judgments of response accuracy, and types of errors. J Exp Psychol Hum Percept: Hum Percept Perform 2000;26:141 51. [14] Cohen MX, Elger CE, Ranganath C. Reward expectation modulates feedbackrelated negativity and EEG spectra. Neuroimage 2007;35:968 78. [15] Miltner WHR, Braun CH, Coles MGH. Event-related brain potentials following incorrect feedback in a time-estimation task: evidence for a generic neural system for error detection. J Cogn Neurosci 1997;9:788 98. [16] Botvinick MM, Braver TS, Barch DM, Carter CS, Cohen JD. Conflict monitoring and cognitive control. Psychol Rev 2001;108:624 52. [17] Holroyd CB, Coles MGH. The neural basis of human error processing: reinforcement learning, dopamine, and the error-related negativity. Psychol Rev 2002;109:679 709. [18] Eriksen CW, Eriksen BA. Target redundancy in visual search: do repetitions of the target within the display impair processing? Percept Psychophys 1979;26:195 205. [19] Tzur G, Berger A. When things look wrong: an ERP study of perceived erroneous information. Neuropsychologia 2007;45:3122 6. [20] Van-Veen V, Carter CS. The anterior cingulate as a conflict monitor: fmri and ERP studies. Physiol Behav 2002;77:477 82. [21] Van-Veen V, Holroyd CB, Cohenb JD, Stenger VA, Carter CS. Errors without conflict: implications for performance monitoring theories of anterior cingulate cortex. Brain Cogn 2004;56:267 76. [22] Berger A, Tzur G, Posner MI. Infant brains detect arithmetic errors. Proc Natl Acad Sci USA 2006;103:12649 53. [23] Samar VJ, Bopardikar A, Rao R, Swartz K. Wavelet analysis of neuroelectric waveforms: a conceptual tutorial. Brain Lang 1999;66:7 60. [24] Lachaux JP, Rodriguez E, Martinerie J, Varela FJ. Measuring phase synchrony in brain signals. Hum Brain Mapp 1999;8:194 208. [25] Fell J, Dietl T, Grunwald T, Kurthen M, Klaver P, Trautner P, et al. Neural bases of cognitive ERPs: more than phase reset. J Cogn Neurosci 2004;16:1595 604. [26] Makeig S, Westerfield M, Jung TP, Enghoff S, Townsend J, Courchesne E, et al. Dynamic brain sources of visual evoked responses. Science 2002;295:690 4. [27] Tucker DM. Spatial sampling of head electrical fields: the geodesic sensor net. Electroencephalogr Clin Neurophysiol 1993;3:154 63. [28] Ferree TC, Luu P, Russell GS, Tucker DM. Scalp electrode impedance, infection risk, and EEG data quality. Clin Neurophysiol 2001;112:536 44. [29] Tallon-Baudry C, Bertrand O, Delpuech C, Pernier J. Stimulus specificity of phaselocked and non-phase-locked 40 Hz visual responses in human. J Neurosci 1996;16:4240 9. [30] Trujillo LT, Allen JJB. Theta EEG dynamics of the error-related negativity. Clin Neurophysiol 2007;118:645 68. [31] Muthukumaraswamy SD, Johnson BW. Primary motor cortex activation during action observation revealed by wavelet analysis of the EEG. Clin Neurophysiol 2004;115:1760 6. [32] Kiebel SJ, Tallon-Baudry C, Friston KJ. Parametric analysis of oscillatory activity as measured with EEG/MEG. Hum Brain Mapp 2005;26:170 7. [33] Holroyd CB, Nieuwenhuis S, Yeung N, Cohen JD. Errors in reward prediction are reflected in the event-related brain potential. Neuroreport 2003;14:2481 4. [34] Nieuwenhuis S, Yeung N, Holroyd CB. Sensitivity of electrophysiological activity from medial frontal cortex to utilitarian and performance feedback. Cereb Cortex 2004;14:741 7. [35] Yasuda A, Sato A, Miyawaki K, Kumano H, Kuboki T. Error-related negativity reflects detection of negative reward prediction error. Neuroreport 2004;15:2561 5. [36] Fiorillo CD, Tobler PN, Schultz W. Discrete coding of reward probability and uncertainty by dopamine neurons. Science 2003;299:1898 902. [37] Canolty RT, Edwards E, Dalal SS, Soltani M, Nagarajan SS, Kirsch HE, et al. High gamma power is phase-locked to theta oscillations in human neocortex. Science 2006;313:1626 8. [38] Engel AK, Singer W. Temporal binding and the neural correlates of sensory awareness. Trends Cogn Sci 2001;5:16 25. [39] Varela F, Lachaux JP, Rodriguez E, Martinerie J. The brainweb: phase synchronization and large scale integration. Nat Rev Neurosci 2001;2:229 39. [40] Mormann F, Fell J, Axmacher N, Weber B, Lehnertz K, Elger CE, et al. Phase/amplitude reset and theta gamma interaction in the human medial temporal lobe during a continuous word recognition memory task. Hippocampus 2005;15:890 900. [41] Bush G, Luu P, Posner MI. Cognitive and emotional influences in anterior cingulate cortex. Trends Cogn Sci 2000;4:215 22.