The brain adapts to control our behavior in different ways depending on the specific situation, which is particularly useful when deciding how to interpret conflicting sets of information. The 'Stroop task' is a classic demonstration of this process. In this task, individuals are shown words where the color and the meaning of the text conflict – for example, the word 'green' is written in blue. When asked what the color of the text is, individuals must suppress the instinct to read the word. This causes them to make more mistakes and take longer to decide on an answer than when they perform the same task using words that have no conflict (for example, when “red” is written in red).

Previous work has suggested that several regions within part of the brain called the frontal cortex play a role in this cognitive control process. However, the relative contributions of each of these regions, and the order in which they are activated, remain unclear. This is in part due to the fact that accurately measuring the electrical activity of the frontal cortex requires implanting electrodes into the brain.

Tang et al. took advantage of a rare opportunity to record this activity from a group of patients who had electrodes implanted in their frontal cortex to treat epilepsy. The electrical signals recorded by these electrodes as the subjects performed the Stroop task revealed that four regions in the frontal cortex altered their activity during trials where the color and the meaning of a word conflicted. These responses corresponded with the subject’s reaction time, changed depending on the exact nature of the task, and even reflected the subjects’ errors. These responses arose at different times in different regions, allowing Tang et al. to suggest how signals flow through the frontal cortex during cognitive control.

In the future it will be important to further understand how the regions of the frontal cortex identified by Tang et al. interact with each other and to establish their roles in cognitive control. These observations could then be used to produce a theoretical framework that describes how the brain adapts behavior to different circumstances.

Flexible control of cognitive processes is fundamental to daily activities, including the execution of goal-directed tasks according to stimulus inputs and context dependencies. An important case of cognitive control arises when input stimuli elicit conflicting responses and subjects must select the task-relevant response despite competition from an often stronger but task-irrelevant response (Miller, 2000; Miller and Cohen, 2001). A canonical example of this type of conflict is the Stroop task: subjects are asked to name the font color of a word where the semantic meaning conflicts with the color signal (e.g. the word 'red' shown in green versus red). Such incongruent inputs lead to longer reaction times, attributed to weaker signals (color processing) that must be emphasized over the automatic processing of word information (Stroop, 1935). The Stroop task is frequently used in cognitive neuroscience and clinical psychology and forms the foundation for theories of cognitive control.

Neurophysiological, neuroimaging, and lesion studies have ascribed a critical role in cognitive control to networks within frontal cortex (Miller, 2000; Miller and Cohen, 2001), yet the neural circuit dynamics and mechanisms responsible for orchestrating control processes remain poorly understood. Lesion studies (Cohen and Servan-Schreiber, 1992; Perrett, 1974), human neuroimaging measurements (Egner and Hirsch, 2005a; MacDonald, 2000), and macaque single unit recordings (Johnston et al., 2007) implicate the dorsolateral prefrontal cortex (dlPFC) in providing top-down signals to bias processing in favor of the task-relevant stimuli (Botvinick et al., 2001; Miller and Cohen, 2001). The medial frontal cortex (mFC) also participates in cognitive control, possibly in a conflict monitoring capacity (Botvinick et al., 2001; Ridderinkhof et al., 2004; Rushworth et al., 2004). Recordings and lesions studies in the macaque anterior cingulate cortex (ACC) (Ito et al., 2003; Nakamura et al., 2005) suggest that ACC neurons are principally involved in monitoring for errors and making between-trial adjustments (Brown and Braver, 2005; Ito et al., 2003; Johnston et al., 2007; Rothé et al., 2011)—an idea that has received support by a recent study in the human ACC (Sheth et al., 2012). Recent work has also demonstrated that the supplementary motor area and the medial frontal cortex play an important role in monitoring for errors (Bonini et al., 2014). An alternative and influential theoretical framework posits that the ACC monitors for potential conflicts and subsequently directs the dlPFC to engage control processes (Botvinick et al., 2001; Shenhav et al., 2013). Several human neuroimaging studies are consistent with this notion (Botvinick et al., 1999; Kerns, 2006; Kerns et al., 2004; MacDonald, 2000) but the relative contributions of dlPFC, mFC, and ACC to cognitive control remain a matter of debate (Aarts et al., 2008; Cole et al., 2009; Fellows and Farah, 2005; Mansouri et al., 2007; Milham and Banich, 2005; Milham et al., 2003; Rushworth et al., 2004).

Previously, some neuroimaging studies have suggested that these frontal cortex regions can be differentiated based on the presence or absence of conflict signals (MacDonald, 2000). The challenge in dissociating the relative roles of these regions during Stroop-like tasks is that increased task difficulty recruits a host of executive functions (attention, decision-making, uncertainty, cognitive control). These functions are associated with neural activity spanning tens to hundreds of milliseconds and the underlying dynamics are difficult to untangle with the low temporal resolution of existing neuroimaging techniques (Shenhav et al., 2013). Human single neuron studies provide millisecond resolution but have focused on individual regions (Sheth et al., 2012). We took advantage of the high spatiotemporal resolution of intracranial recordings in human epilepsy patients and the ability to record simultaneously from multiple regions to directly investigate the dynamics of conflict responses during cognitive control. We hypothesized that subregions of frontal cortex could be differentiated based on the temporal profile of their conflict responses. We recorded intracranial field potentials from 1397 electrodes in 15 subjects while they performed the Stroop task or a variation in which they were asked to read the word instead of focusing on its color.

We observed conflict-selective activity throughout several regions in frontal cortex: ACC, mFC, dlPFC, and also orbitofrontal cortex (OFC). Several analyses link these signals to cognitive control. Neural responses increased for incongruent compared to congruent trials, and these signals correlated with behavioral reaction time, depended on the task, and exhibited adaptation over trials. We compared pairs of simultaneously recorded electrodes to disassociate these different regions based on the timing of these conflict responses rather than their presence or absence. Conflict responses emerged first in the ACC and subsequently emerged in dlPFC and mFC and finally in OFC. These observations propose a plausible flow of signals underlying cognitive control.

We recorded field potentials from 15 epilepsy patients implanted with intracranial electrodes in frontal cortex as they performed the Stroop task (Figure 1, Supplementary file 1). After 500 ms of a fixation cross, subjects were presented with one of three words (Red, Blue, Green), which were colored either red, blue, or green. We refer to congruent trials (C) where the font color matched the word (60% of the trials) compared to incongruent trials (I) where the font color conflicted with the word (40% of the trials). Within each trial type, the word-color combinations were counter-balanced and randomly interleaved. The stimuli were presented for 2 s (in two subjects, for 3 s). Subjects were asked to respond verbally and either name the color (Stroop task), or read the word (Reading task) in separate blocks. Performance during congruent trials was essentially at ceiling (Figure 1—figure supplement 1).

Figure 1 with 1 supplement see all

Download asset Open asset

Figure 1—source data 1

Behavioral data.

https://doi.org/10.7554/eLife.12352.004

Download elife-12352-fig1-data1-v2.zip

An ANOVA conducted on subjects’ performance with stimulus type (congruent or incongruent) and task (Stroop or Reading) as repeated measures revealed a significant interaction between stimulus type and task (F = 22.9, P < 0.001). For the Stroop task, subjects made more errors during incongruent trials (average error rate: 5 ± 3%, P < 0.001 paired t-test), as demonstrated in previous studies (Bugg et al., 2008; Egner and Hirsch, 2005b; Kerns et al., 2004). There was no difference in the number of error trials during the Reading task (P = 0.76, paired t-test). Subsequent analyses focused on correct trials only unless otherwise stated. Subjects’ reaction times also had a significant interaction between stimulus type and task (F = 65.2, P < 10^–5, ANOVA). Consistent with previous observations (Stroop, 1935), subjects’ response times during the Stroop task were delayed for incongruent trials compared to congruent trials (Figure 1B, average delay: 215 ± 93 ms, P < 0.001, paired t-test, see also Figure 1—figure supplement 1C for individual subject data). The reaction time delays were shorter in the Reading task (Figure 1C, average delay: 22 ± 31 ms, P = 0.02, paired t-test). Trial history also has a strong effect on reaction time (known as Gratton effect in the literature [Gratton et al., 1992)]). A repeated measures ANOVA revealed an interaction between previous and current trial type (F = 19.5, P < 0.001). Incongruent trials that were preceded by a congruent trial (cI trials) elicited slower reaction times compared to incongruent trials that were preceded by an incongruent trial (iI trials) (Figure 1D, average reaction time difference: 34 ± 14 ms, P = 0.03, paired t-test). A similar Gratton effect was found for iC versus cC trials (Figure 1D, average reaction time difference: 72 ± 136ms, P < 0.001, paired t-test).

We recorded intracranial field potentials from 1397 electrodes (average 93 ± 31 electrodes per subject) while subjects performed the Stroop and Reading tasks. The number of electrodes per subject and the location of these electrodes were strictly dictated by clinical needs. Therefore, there was a wide distribution of electrode locations, as is typical in this type of recordings (Liu et al., 2009). We excluded electrodes in epileptogenic regions. We focused on the neural signals in the gamma band (70–120 Hz) given their prominence in sensory, motor and cognitive phenomena (Crone et al., 1998a; Liu et al., 2009; Oehrn et al., 2014); results for other frequency bands are shown in Figure 2—figure supplement 2 and Figure 4—figure supplement 1 and 2. Presentation of the visual stimuli evoked rapid and color/word selective neural responses in visual cortical areas within 200 ms of stimulus onset, as expected from previous studies (e.g. Liu et al., 2009]). Other electrodes were activated for different motor (verbal) outputs (e.g. Bouchard et al., 2013; Crone et al., 1998a).

Conflict responses in frontal cortex

We focused on 469 electrodes located in areas within frontal lobe which have been previously implicated in executive function: medial frontal cortex (mFC, n = 111), orbitofrontal cortex (OFC, n = 156), dorsolateral prefrontal cortex (dlPFC, n = 168) and the anterior cingulate cortex (ACC, n = 34). We applied a non-parametric analysis of variance (ANOVA) to measure whether and when the physiological responses differed between congruent and incongruent trials. An electrode was considered conflict-selective if the F-statistic was greater than a significance threshold computed by a permutation test with P = 0.001 for 50 consecutive milliseconds (Materials and methods). The latency was defined as the first time of this threshold-crossing.

Figure 2 shows an example electrode from the left Anterior Cingulate Cortex that responded differentially between congruent and incongruent trials during the Stroop task. These signals were better aligned to the speech onset than to the stimulus onset, as shown in the response-aligned view (compare Figure 2A–C with Figure 2D–F). During the Stroop task, the response-aligned signals were significantly stronger for the incongruent (brown) trials compared to the congruent (black) trials (Figure 2D, P < 10^–5, ANOVA), and were invariant to the particular word/color combinations (Figure 2G). Incongruent trials could be discriminated from congruent trials at a latency of 669 ± 31 ms (mean ± s.e.m.) before the onset of the response (Figure 2D). This conflict response was also specific to the Stroop task; there was a significant interaction between congruency and task (F = 13.5, P = 0.007, ANOVA). The same stimuli did not elicit differential activity during the Reading task (Figure 2F). We assessed the correlation between the neural signal strength and behavioral reaction times in single trials. The maximal gamma power during each incongruent trial (using the average gamma power yielded similar results) was positively correlated with the behavioral reaction times (Figure 2H, ρ = 0.25, P = 0.02).

Figure 2 with 2 supplements see all

Download asset Open asset

Figure 2—source data 1

Conflict-selective electrode data.

https://doi.org/10.7554/eLife.12352.007

Download elife-12352-fig2-data1-v2.zip

Any differences between congruent and incongruent trials in the stimulus-aligned analyses can be confounded by the reaction time differences; therefore, we focus subsequent analyses on the response-aligned signals. More example electrodes are shown in Figure 2—figure supplement 1 (dlPFC) and Figure 2—figure supplement 2 (OFC).

Using the aforementioned criteria, we identified n = 51 conflict selective frontal cortex electrodes during the Stroop task, with contributions from 13 subjects (Supplementary files 2 and 3). These electrodes were distributed throughout different subregions within frontal cortex (Figure 3A). To evaluate whether random variation in the signals could give rise to apparent conflict-selective electrodes, we randomly shuffled the congruent/incongruent trial labels 10,000 times and applied the same statistical criteria (Materials and methods). Across our population, we found n = 4.4 ± 0.03 false positive electrodes (mean ± s.e.m., out of 469 electrodes), which corresponds to a false discovery rate (FDR) of q = 0.01, which is significantly less than our observation of n = 51 electrodes. The number of conflict-selective electrodes within each subregion was significantly greater than expected by chance (Figure 3B, P < 0.01, all regions). We repeated the analyses during the Reading task. In contrast with the Stroop task, we only observed n = 3 conflict-selective frontal cortex electrodes during the Reading task (out of 469 electrodes), a number that is within the false positive rate.

Figure 3

Download asset Open asset

Figure 3—source data 1

Population gamma-power data.

https://doi.org/10.7554/eLife.12352.011

Download elife-12352-fig3-data1-v2.zip

To account for within-subject and across-subject variation, we used a multilevel model (Aarts et al., 2014) to conduct a group analysis of the physiological responses, with electrodes nested within subjects (Materials and methods). Across the population, we observed a significant interaction between the factors congruency and task on the gamma power (χ²=9.2, P = 0.002). Consistent with the single electrode examples, gamma power was greater for incongruent compared to congruent trials, but only during the Stroop task (Figure 4A, Stroop: P < 10^–3, Reading: P = 0.56). We computed the average response in each region (Figure 4B). Each electrode’s response was normalized by dividing the power during incongruent trials by the power in congruent trials (dividing the brown curve by the black curve in Figure 2), computing the logarithm and finally pooling within each region. The pooled responses in the OFC are visually less compelling (Figure 4B, bottom right subplot) due to the heterogeneity in the latency of the individual electrodes but the responses in the OFC were as vigorous as the ones in other areas (e.g. Figure 2—figure supplement 2). Similar conclusions were reached when plotting the pooled responses aligned to stimulus onset (Figure 4—figure supplement 3).

Figure 4 with 3 supplements see all

Download asset Open asset

Figure 4—source data 1

Population gamma-power data.

https://doi.org/10.7554/eLife.12352.013

Download elife-12352-fig4-data1-v2.zip

Error monitoring signals

The conflict responses reported above are based on correct trials only. Yet, error monitoring has also been ascribed to frontal cortical circuits (Bonini et al., 2014; Shenhav et al., 2013; Yeung et al., 2004). To investigate whether the same electrodes responding to conflict are also involved in successful error monitoring, we analyzed the neural signals during self-corrected trials. In these trials, subjects initially made an erroneous response and rapidly corrected themselves with the right answer. Given the high performance level of all subjects, the number of such trials is low. However, these trials are particularly interesting because we can be certain that there was successful error detection (as opposed to error trials without any self-correction). An example self-corrected trial from the ACC electrode shown previously is illustrated in Figure 5A. The subject initially made an incorrect response (green), which was rapidly followed with the correct response (red). Increased gamma power was observed after onset of the erroneous response. In contrast, the following corrected behavioral response exhibited no such post-response signal. Additionally, these error-monitoring signals were not observed in correct incongruent trials (Figure 2D), and were consistent across the n = 11 self-corrected trials for this subject (Figure 5B, P = 0.001, signed rank test). Another example electrode is shown in Figure 5C–D. There were only two subjects contributing n = 7 conflict-signaling electrodes that had a sufficient number of self-correction trials (greater than five trials) for this analysis. For each electrode, we compared the difference in neural signals during the one-second post-response window between the initial error and the following self-correction. Of those n = 7 electrodes, n = 5 electrodes showed evidence of error monitoring (Figure 5E, P < 0.05, sign-rank test). Although the number of electrodes and trials in this analysis is small, these results provide a direct correlate of error monitoring signals. Furthermore, these results highlight that the same electrodes that respond to conflict leading up to the behavioral response can also show post-response error monitoring.

Figure 5

Download asset Open asset

Figure 5—source data 1

Data for self-correcting trials.

https://doi.org/10.7554/eLife.12352.018

Download elife-12352-fig5-data1-v2.zip

Regional differences in conflict response latencies

We observed conflict-selective responses in the anterior cingulate cortex, medial frontal cortex, dorsolateral prefrontal cortex and orbitofrontal cortex. To examine the dynamics of cognitive control orchestrating the transformation of conflicting visual signals to motor outputs, we compared, across those four regional groups, the latencies relative to behavioral response onset at which the congruent and incongruent trials could be discriminated. Comparing latencies across regions is difficult, especially across subjects with varying reaction times. For a controlled and direct comparison, we restricted the analysis to compute the latency differences between pairs of simultaneously recorded electrodes. This within-subject pairwise analysis had increased power to examine the relative dynamics between frontal lobe areas (Figure 6). The relative latencies were significantly different across the regions (P = 0.01, permutation test, post-hoc testing was controlled for multiple comparisons using the Benjamin-Hochberg procedure, Materials and methods). Conflict responses in the ACC preceded those in all the other frontal lobe regions, followed 207 ± 40 ms later by dorsolateral prefrontal cortex and 388 ± 83 ms later by medial frontal cortex. Signals in orbitofrontal cortex emerged 319 ± 78 ms after dlPFC. This entire processing cascade took approximately 500 ms. For comparison, subjects’ behavioral reaction times to incongruent trials were 1105 ± 49 ms. The latency difference between ACC and dlPFC is based on 6 electrode pairs: one ACC electrode and six simultaneously recorded dlPFC electrodes. There was only one pair of simultaneous recordings between ACC and OFC and we do not report this value in Figure 6. The other region comparisons have contributions from multiple electrodes in multiple subjects (Supplementary file 3). These results suggest a temporal hierarchy of cognitive control mechanisms culminating in speech onset.

Figure 6

Download asset Open asset

Figure 6—source data 1

Data for region latency comparisons.

https://doi.org/10.7554/eLife.12352.020

Download elife-12352-fig6-data1-v2.zip

We used intracranial field potentials to measure the dynamics of conflict responses across frontal cortex leading up to the behavioral response in the Stroop task. Previous physiological and functional neuroimaging studies have documented the involvement of multiple of these frontal cortex areas in the Stroop or similar tasks (Botvinick et al., 1999; Kolling et al., 2012; MacDonald, 2000; Niendam et al., 2012; Oehrn et al., 2014; Sheth et al., 2012). The intracranial field potential recordings reported here show conflict-selective signals in ACC (e.g. Figure 2), dlPFC (e.g. Figure 2—figure supplement 1), mFC (e.g. Figure 4B) and OFC (e.g. Figure 2—figure supplement 2). The mFC and dlPFC have been previously implicated in cognitive control, and these structures are extensively connected to the rest of frontal cortex areas (Ridderinkhof et al., 2004). The role of the OFC in cognitive control during Stroop-like tasks has not been reported previously, possibly because of technical challenges in neuroimaging near this area (Weiskopf et al., 2006).

We presented several lines of evidence that demonstrate that these conflict-selective physiological signals are relevant for behavior during the Stroop task. Longer behavioral reaction times were correlated with greater gamma power on a trial-by-trial basis during the Stroop task but not during the Reading task, even after accounting for trial history and for differences between congruent and incongruent stimuli (Figure 2H, 4C). The same identical stimuli can elicit a range of behavioral reaction times and this internal degree of conflict can be captured, at least partly, by the strength of gamma power in frontal cortex in each trial.

The neural correlates of behavioral adaptation (Gratton effect [Gratton et al., 1992]) were observed in the ACC, consistent with prior studies based on human single neuron recordings (Sheth et al., 2012), neuroimaging (Botvinick et al., 1999; Kerns, 2006) and also in accordance with the behavioral effects of ACC resection (Sheth et al., 2012). Conflict responses throughout the other frontal cortex regions also demonstrated the neural Gratton effect, suggesting a more distributed network involved in across-trial adaptation than previously hypothesized. The physiological responses in these areas were stronger in cI trials (incongruent trials that were preceded by congruent trials) than iI trials (Figure 4D). While the increased activity in cI trials compared to iI trials is consistent with neuroimaging studies (Botvinick et al., 1999), single neuron recordings in a different Stroop-like task report the opposite relationship (iI > cI) (Sheth et al., 2012). These differences point to potentially interesting distinctions between the activity of individual neurons and coarser population measures that warrant further investigation.

Another discrepancy between neuroimaging studies and single unit recordings is the presence of conflict responses and error signals. Single unit recording in macaque ACC typically find error monitoring signals but not conflict-selective responses (Cole et al., 2009; Emeric et al., 2010; Ito et al., 2003; Taylor et al., 2006), see however (Ebitz and Platt, 2015), whereas human neuroimaging studies report both types of signals in ACC. There has been significant debate concerning whether action monitoring and conflict detection represent distinct processes (Carter et al., 1998; Carter et al., 2000; Nee et al., 2011; Swick and Turken, 2002). Because both processes may co-occur on the same trials, high temporal resolution is required to disassociate the two computations. A recent human intracranial study has found error signals in supplementary motor area and medial frontal cortex (Bonini et al., 2014), and a human single unit study reported conflict signals in ACC (Sheth et al., 2012). The current work demonstrates the coexistence of both error signals and conflict signals. The analysis of the few self-correction trials in our data suggests that the same areas responsible for pre-behavioral conflict signals can also produce post-behavioral response error-monitoring signals (Figure 5). In addition, the relative timing of the conflict and error signals surrounding the neural responses confirms computational predictions based on a connectionist architecture to explain the mechanisms of conflict (Yeung et al., 2004) and scalp EEG studies (Hughes and Yeung, 2011). These results are consistent with computational models suggesting that these signals may represent a general error-likelihood prediction, of which conflict and error detection are special cases (Brown and Braver, 2005).

It has been suggested that ACC and supplementary eye field neurons in macaque monkeys respond to specific stimulus and/or behavioral combinations but are not directly modulated by conflict (Cole et al., 2009; Nakamura et al., 2005). At the level of the intracranial field potentials reported here, the modulation of conflict trials observed in the four frontal cortex regions could not be ascribed to specific stimulus or behavioral responses (e.g. Figure 2G) and were also task dependent (compare Figure 2A versus 2C). In these patients, we did not have access to single neuron responses and we therefore cannot rule out the possibility that individual neurons show distinct patterns of responses that are averaged out at the field potential level.

Besides the high gamma band, we also observed conflict responses in the beta and theta bands, but not the low gamma band (e.g. Figure 4—figure supplement 1). Previous work has suggested differential roles for distinct oscillatory components of the local field potential (Cavanagh and Frank, 2014; Kahana et al., 2001; Ullsperger et al., 2014; von Stein and Sarnthein, 2000). There were clear differences in the type of information conveyed by distinct frequencies components. Lack of significant correlations with reaction time in the theta and beta bands suggests that the gamma band better captures the behavior. Additionally, conflict responses were characterized by increased power in the gamma band, but decreased power in the theta and beta bands (Figure 4—figure supplement 1). Previous scalp EEG recordings (Cavanagh and Frank, 2014; Ullsperger et al., 2014; van Driel et al., 2015) have demonstrated that conflict and/or error trials elicit increased theta power, suggesting potentially interesting differences in how theta is captured across spatial scales. We also observed a decrease in beta power, which is consistent with previous studies that correlate frontal cortex activation with desynchronization in the beta band and increased synchronization in the gamma bands (Crone et al., 1998a; Crone et al., 1998b). Differences across tasks, recording methods, and targeted regions should be interpreted with caution. The roles of different oscillatory components in neocortex are not clearly understood. One possibility is that lower frequency bands reflect the summed dendritic input of the nearby neural population (Logothetis et al., 2001; Mitzdorf, 1987) and can act as channels for communication (Cavanagh and Frank, 2014), whereas higher frequency bands represent the population spiking rate (Buzsaki et al., 2012; Ray and Maunsell, 2011). Along these lines, we speculate that the theta desynchronization we observe could reflect a reduction of inputs, leading to inhibition of the prepotent but erroneous response.

While we observed conflict responses throughout frontal cortex, the spatiotemporal resolution of our intracranial recordings allowed us to separate regions by the latency at which conflict-selective responses emerge with respect to speech onset. By comparing pairs of simultaneously recorded electrodes, we found that conflict responses in the ACC lead the dlPFC by ~200 ms. Medial frontal cortex is anatomically close and extensively connected to the ACC, and the two regions are often grouped together (Cavanagh et al., 2009; Ridderinkhof et al., 2004). Yet, conflict responses in the mFC trail the ACC by hundreds of milliseconds, suggesting an important distinction between the two regions (Rushworth et al., 2004). The relative latency measurements place the OFC at the bottom of this cascade. The hierarchical cascade of processes described here is consistent with predictions from mechanistic models of cognitive control (e.g. see Figure 2 in Shenhav et al, Neuron 2013). In particular, stimulus related signals are evident along the ventral visual stream early on and feed onto frontal cortex, where we find that ACC activity precedes activity in other frontal regions, followed by dlPFC, and finally mFC, and OFC.

Since the local field potential pools over many neurons, latency measures can be influenced by a variety of factors, such as the proportion of neurons selective for conflict and their laminar organization. Yet, at least in the ACC, the temporal profile of conflict responses we observed is similar to responses from human single unit recordings (Sheth et al., 2012). The relatively long delays between regions are also particularly intriguing. There are monosynaptic connections that link these four regions within frontal cortex and yet, it takes 100–200 ms to detect the relative activation between these areas (Figure 6).

Daily decisions require integration of different goals, contexts, input signals, and the consequences of the resulting actions. The current study provides initial steps to elucidate not only which brain areas participate in cognitive control on a trial-by-trial basis but also their relative interactions and differential roles. The relative latency measurements and correlations between neural activity and reaction time provide a framework to constrain theories of cognitive control, and propose a plausible flow of conflict responses through frontal cortex.

1. 1. Aarts E
   2. Roelofs A
   3. van Turennout M

   (2008) Anticipatory activity in anterior cingulate cortex can be independent of conflict and error likelihood

   Journal of Neuroscience 28:4671–4678.

   https://doi.org/10.1523/JNEUROSCI.4400-07.2008

   • Google Scholar
2. 1. Aarts E
   2. Verhage M
   3. Veenvliet JV
   4. Dolan CV
   5. van der Sluis S

   (2014) A solution to dependency: using multilevel analysis to accommodate nested data

   Nature Neuroscience 17:491–496.

   https://doi.org/10.1038/nn.3648

   • Google Scholar
3. 1. Bonini F
   2. Burle B
   3. Liegeois-Chauvel C
   4. Regis J
   5. Chauvel P
   6. Vidal F

   (2014) Action monitoring and medial frontal cortex: leading role of supplementary motor area

   Science 343:888–891.

   https://doi.org/10.1126/science.1247412

   • Google Scholar
4. 1. Botvinick M
   2. Nystrom LE
   3. Fissell K
   4. Carter CS
   5. Cohen JD

   (1999) Conflict monitoring versus selection-for-action in anterior cingulate cortex

   Nature 402:179–181.

   https://doi.org/10.1038/46035

   • Google Scholar
5. 1. Botvinick MM
   2. Braver TS
   3. Barch DM
   4. Carter CS
   5. Cohen JD

   (2001) Conflict monitoring and cognitive control

   Psychological Review 108:624–652.

   https://doi.org/10.1037/0033-295X.108.3.624

   • Google Scholar
6. 1. Bouchard KE
   2. Mesgarani N
   3. Johnson K
   4. Chang EF

   (2013) Functional organization of human sensorimotor cortex for speech articulation

   Nature 495:327–332.

   https://doi.org/10.1038/nature11911

   • Google Scholar
7. 1. Brown JW
   2. Braver TS

   (2005) Learned predictions of error likelihood in the anterior cingulate cortex

   Science 307:1118–1121.

   https://doi.org/10.1126/science.1105783

   • Google Scholar
8. 1. Bugg JM
   2. Jacoby LL
   3. Toth JP

   (2008) Multiple levels of control in the stroop task

   Memory & Cognition 36:1484–1494.

   https://doi.org/10.3758/MC.36.8.1484

   • Google Scholar
9. 1. Buzsáki G
   2. Anastassiou CA
   3. Koch C

   (2012) The origin of extracellular fields and currents — EEG, ECoG, LFP and spikes

   Nature Reviews Neuroscience 13:407–420.

   https://doi.org/10.1038/nrn3241

   • Google Scholar
10. 1. Canolty RT
    2. Edwards E
    3. Dalal SS
    4. Soltani M
    5. Nagarajan SS
    6. Kirsch HE
    7. Berger MS
    8. Barbaro NM
    9. Knight RT

    (2006) High gamma power is phase-locked to theta oscillations in human neocortex

    Science 313:1626–1628.

    https://doi.org/10.1126/science.1128115

    • Google Scholar
11. 1. Carter CS
    2. Braver TS
    3. Barch DM
    4. Botvinick MM
    5. Noll D
    6. Cohen JD

    (1998) Anterior cingulate cortex, error detection, and the online monitoring of performance

    Science 280:747–749.

    https://doi.org/10.1126/science.280.5364.747

    • Google Scholar
12. 1. Carter CS
    2. Macdonald AM
    3. Botvinick M
    4. Ross LL
    5. Stenger VA
    6. Noll D
    7. Cohen JD

    (2000) Parsing executive processes: strategic vs. evaluative functions of the anterior cingulate cortex

    Proceedings of the National Academy of Sciences of the United States of Americs 97:1944–1948.

    https://doi.org/10.1073/pnas.97.4.1944

    • Google Scholar
13. 1. Cavanagh JF
    2. Cohen MX
    3. Allen JJB

    (2009) Prelude to and resolution of an error: EEG phase synchrony reveals cognitive control dynamics during action monitoring

    Journal of Neuroscience 29:98–105.

    https://doi.org/10.1523/JNEUROSCI.4137-08.2009

    • Google Scholar
14. 1. Cavanagh JF
    2. Frank MJ

    (2014) Frontal theta as a mechanism for cognitive control

    Trends in Cognitive Sciences 18:414–421.

    https://doi.org/10.1016/j.tics.2014.04.012

    • Google Scholar
15. 1. Cohen JD
    2. Servan-Schreiber D

    (1992) Context, cortex, and dopamine: a connectionist approach to behavior and biology in schizophrenia

    Psychological Review 99:45–77.

    https://doi.org/10.1037/0033-295X.99.1.45

    • Google Scholar
16. 1. Cole MW
    2. Yeung N
    3. Freiwald WA
    4. Botvinick M

    (2009) Cingulate cortex: diverging data from humans and monkeys

    Trends in Neurosciences 32:566–574.

    https://doi.org/10.1016/j.tins.2009.07.001

    • Google Scholar
17. 1. Crone N
    2. Miglioretti DL
    3. Gordon B
    4. Lesser RP

    (1998a) Functional mapping of human sensorimotor cortex with electrocorticographic spectral analysis. II. event-related synchronization in the gamma band

    Brain 121:2301–2315.

    https://doi.org/10.1093/brain/121.12.2301

    • Google Scholar
18. 1. Crone N
    2. Miglioretti DL
    3. Gordon B
    4. Sieracki JM
    5. Wilson MT
    6. Uematsu S
    7. Lesser RP

    (1998b) Functional mapping of human sensorimotor cortex with electrocorticographic spectral analysis. i. alpha and beta event- related desynchronization

    Brain 121:2271–2299.

    https://doi.org/10.1093/brain/121.12.2271

    • Google Scholar
19. 1. Destrieux C
    2. Fischl B
    3. Dale A
    4. Halgren E

    (2010) Automatic parcellation of human cortical gyri and sulci using standard anatomical nomenclature

    NeuroImage 53:1–15.

    https://doi.org/10.1016/j.neuroimage.2010.06.010

    • Google Scholar
20. 1. Ebitz RB
    2. Platt ML

    (2015) Neuronal activity in primate dorsal anterior cingulate cortex signals task conflict and predicts adjustments in pupil-linked arousal

    Neuron 85:628–640.

    https://doi.org/10.1016/j.neuron.2014.12.053

    • Google Scholar
21. 1. Egner T
    2. Hirsch J

    (2005a) Cognitive control mechanisms resolve conflict through cortical amplification of task-relevant information

    Nature Neuroscience 8:1784–1790.

    https://doi.org/10.1038/nn1594

    • Google Scholar
22. 1. Egner T
    2. Hirsch J

    (2005b) The neural correlates and functional integration of cognitive control in a stroop task

    NeuroImage 24:539–547.

    https://doi.org/10.1016/j.neuroimage.2004.09.007

    • Google Scholar
23. 1. Emeric EE
    2. Leslie M
    3. Pouget P
    4. Schall JD

    (2010) Performance monitoring local field potentials in the medial frontal cortex of primates: supplementary eye field

    Journal of Neurophysiology 104:1523–1537.

    https://doi.org/10.1152/jn.01001.2009

    • Google Scholar
24. 1. Fellows LK
    2. Farah MJ

    (2005) Is anterior cingulate cortex necessary for cognitive control?

    Brain 128:788–796.

    https://doi.org/10.1093/brain/awh405

    • Google Scholar
25. Book

    1. Goldstein H

    (2011)

    Multilevel Statistical Models, 4th Edn Chichester

    West Sussex: Wiley.

    • Google Scholar
26. 1. Gratton G
    2. Coles MGH
    3. Donchin E

    (1992) Optimizing the use of information: strategic control of activation of responses

    Journal of Experimental Psychology: General 121:480–506.

    https://doi.org/10.1037/0096-3445.121.4.480

    • Google Scholar
27. 1. Hughes G
    2. Yeung N

    (2011) Dissociable correlates of response conflict and error awareness in error-related brain activity

    Neuropsychologia 49:405–415.

    https://doi.org/10.1016/j.neuropsychologia.2010.11.036

    • Google Scholar
28. 1. Ito S
    2. Stuphorn V
    3. Brown JW
    4. Schall JD

    (2003) Performance monitoring by the anterior cingulate cortex during saccade countermanding

    Science 302:120–122.

    https://doi.org/10.1126/science.1087847

    • Google Scholar
29. 1. Johnston K
    2. Levin HM
    3. Koval MJ
    4. Everling S

    (2007) Top-down control-signal dynamics in anterior cingulate and prefrontal cortex neurons following task switching

    Neuron 53:453–462.

    https://doi.org/10.1016/j.neuron.2006.12.023

    • Google Scholar
30. 1. Kahana MJ
    2. Seelig D
    3. Madsen JR

    (2001) Theta returns

    Current Opinion in Neurobiology 11:739–744.

    https://doi.org/10.1016/S0959-4388(01)00278-1

    • Google Scholar
31. 1. Kerns JG
    2. Cohen JD
    3. MacDonald AW
    4. Cho RY
    5. Stenger VA
    6. Carter CS

    (2004) Anterior cingulate conflict monitoring and adjustments in control

    Science 303:1023–1026.

    https://doi.org/10.1126/science.1089910

    • Google Scholar
32. 1. Kerns JG

    (2006) Anterior cingulate and prefrontal cortex activity in an FMRI study of trial-to-trial adjustments on the simon task

    NeuroImage 33:399–405.

    https://doi.org/10.1016/j.neuroimage.2006.06.012

    • Google Scholar
33. 1. Kolling N
    2. Behrens TEJ
    3. Mars RB
    4. Rushworth MFS

    (2012) Neural mechanisms of foraging

    Science 336:95–98.

    https://doi.org/10.1126/science.1216930

    • Google Scholar
34. 1. Liu H
    2. Agam Y
    3. Madsen JR
    4. Kreiman G

    (2009) Timing, timing, timing: fast decoding of object information from intracranial field potentials in human visual cortex

    Neuron 62:281–290.

    https://doi.org/10.1016/j.neuron.2009.02.025

    • Google Scholar
35. 1. Logothetis NK
    2. Pauls J
    3. Augath M
    4. Trinath T
    5. Oeltermann A

    (2001) Neurophysiological investigation of the basis of the fMRI signal

    Nature 412:150–157.

    https://doi.org/10.1038/35084005

    • Google Scholar
36. 1. MacDonald AW
    2. Cohen JD
    3. Stenger VA
    4. Carter CS

    (2000) Dissociating the role of the dorsolateral prefrontal and anterior cingulate cortex in cognitive control

    Science 288:1835–1838.

    https://doi.org/10.1126/science.288.5472.1835

    • Google Scholar
37. 1. Mansouri FA
    2. Buckley MJ
    3. Tanaka K

    (2007) Mnemonic function of the dorsolateral prefrontal cortex in conflict-induced behavioral adjustment

    Science 318:987–990.

    https://doi.org/10.1126/science.1146384

    • Google Scholar
38. 1. Milham MP
    2. Banich MT
    3. Claus ED
    4. Cohen NJ

    (2003)

    Practice-related effects demonstrate complementary roles of anterior cingulate and prefrontal cortices in attentional control

    NeuroImage 18:483–493.

    • Google Scholar
39. 1. Milham MP
    2. Banich MT

    (2005) Anterior cingulate cortex: an fMRI analysis of conflict specificity and functional differentiation

    Human Brain Mapping 25:328–335.

    https://doi.org/10.1002/hbm.20110

    • Google Scholar
40. 1. Miller EK

    (2000) The prefrontal cortex and cognitive control

    Nature Reviews. Neuroscience 1:59–65.

    https://doi.org/10.1038/35036228

    • Google Scholar
41. 1. Miller EK
    2. Cohen JD

    (2001) An integrative theory of prefrontal cortex function

    Annual Review of Neuroscience 24:167–202.

    https://doi.org/10.1146/annurev.neuro.24.1.167

    • Google Scholar
42. 1. Mitra P
    2. Bokil H

    (2008)

    Oxford University Press

    Observed brain dynamics Oxford, Oxford University Press, New York.

    • Google Scholar
43. 1. Mitzdorf U

    (1987) Properties of the evoked potential generators: current source-density analysis of visually evoked potentials in the cat cortex

    International Journal of Neuroscience 33:33–59.

    https://doi.org/10.3109/00207458708985928

    • Google Scholar
44. 1. Nakamura K
    2. Roesch MR
    3. Olson CR

    (2005) Neuronal activity in macaque SEF and ACC during performance of tasks involving conflict

    Journal of Neurophysiology 93:884–908.

    https://doi.org/10.1152/jn.00305.2004

    • Google Scholar
45. 1. Nee DE
    2. Kastner S
    3. Brown JW

    (2011) Functional heterogeneity of conflict, error, task-switching, and unexpectedness effects within medial prefrontal cortex

    NeuroImage 54:528–540.

    https://doi.org/10.1016/j.neuroimage.2010.08.027

    • Google Scholar
46. 1. Niendam TA
    2. Laird AR
    3. Ray KL
    4. Dean YM
    5. Glahn DC
    6. Carter CS

    (2012) Meta-analytic evidence for a superordinate cognitive control network subserving diverse executive functions

    Cognitive, Affective, & Behavioral Neuroscience 12:241–268.

    https://doi.org/10.3758/s13415-011-0083-5

    • Google Scholar
47. 1. Nieuwenhuis S
    2. Forstmann BU
    3. Wagenmakers E-J

    (2011) Erroneous analyses of interactions in neuroscience: a problem of significance

    Nature Neuroscience 14:1105–1107.

    https://doi.org/10.1038/nn.2886

    • Google Scholar
48. 1. Oehrn CR
    2. Hanslmayr S
    3. Fell J
    4. Deuker L
    5. Kremers NA
    6. Do Lam AT
    7. Elger CE
    8. Axmacher N

    (2014) Neural communication patterns underlying conflict detection, resolution, and adaptation

    Journal of Neuroscience 34:10438–10452.

    https://doi.org/10.1523/JNEUROSCI.3099-13.2014

    • Google Scholar
49. 1. Perret E

    (1974)

    The left frontal lobe of man and the suppression of habitual responses in verbal categorical behaviour

    Neuropsychologia 12:323–330.

    • Google Scholar
50. 1. Ray S
    2. Maunsell JHR

    (2011) Different origins of gamma rhythm and high-gamma activity in macaque visual cortex

    PLoS Biology 9:e1000610.

    https://doi.org/10.1371/journal.pbio.1000610

    • Google Scholar
51. 1. Ridderinkhof KR
    2. Ullsperger M
    3. Crone EA
    4. Nieuwenhuis S

    (2004) The role of the medial frontal cortex in cognitive control

    Science 306:443–447.

    https://doi.org/10.1126/science.1100301

    • Google Scholar
52. 1. Rothe M
    2. Quilodran R
    3. Sallet J
    4. Procyk E

    (2011) Coordination of high gamma activity in anterior cingulate and lateral prefrontal cortical areas during adaptation

    Journal of Neuroscience 31:11110–11117.

    https://doi.org/10.1523/JNEUROSCI.1016-11.2011

    • Google Scholar
53. 1. Rushworth M
    2. Walton M
    3. Kennerley S
    4. Bannerman D

    (2004) Action sets and decisions in the medial frontal cortex

    Trends in Cognitive Sciences 8:410–417.

    https://doi.org/10.1016/j.tics.2004.07.009

    • Google Scholar
54. 1. Shenhav A
    2. Botvinick MM
    3. Cohen JD

    (2013) The expected value of control: an integrative theory of anterior cingulate cortex function

    Neuron 79:217–240.

    https://doi.org/10.1016/j.neuron.2013.07.007

    • Google Scholar
55. 1. Sheth SA
    2. Mian MK
    3. Patel SR
    4. Asaad WF
    5. Williams ZM
    6. Dougherty DD
    7. Bush G
    8. Eskandar EN

    (2012) Human dorsal anterior cingulate cortex neurons mediate ongoing behavioural adaptation

    Nature 488:218–221.

    https://doi.org/10.1038/nature11239

    • Google Scholar
56. 1. Stroop JR

    (1935) Studies of interference in serial verbal reactions

    Journal of Experimental Psychology 18:643–662.

    https://doi.org/10.1037/h0054651

    • Google Scholar
57. 1. Swick D
    2. Turken AU

    (2002) Dissociation between conflict detection and error monitoring in the human anterior cingulate cortex

    Proceedings of the National Academy of Sciences of the United States of Americs 99:16354–16359.

    https://doi.org/10.1073/pnas.252521499

    • Google Scholar
58. 1. Taylor SF
    2. Martis B
    3. Fitzgerald KD
    4. Welsh RC
    5. Abelson JL
    6. Liberzon I
    7. Himle JA
    8. Gehring WJ

    (2006) Medial frontal cortex activity and loss-related responses to errors

    Journal of Neuroscience 26:4063–4070.

    https://doi.org/10.1523/JNEUROSCI.4709-05.2006

    • Google Scholar
59. 1. Tort ABL
    2. Kramer MA
    3. Thorn C
    4. Gibson DJ
    5. Kubota Y
    6. Graybiel AM
    7. Kopell NJ

    (2008) Dynamic cross-frequency couplings of local field potential oscillations in rat striatum and hippocampus during performance of a t-maze task

    Proceedings of the National Academy of Sciences of the United States of Americs 105:20517–20522.

    https://doi.org/10.1073/pnas.0810524105

    • Google Scholar
60. 1. Ullsperger M
    2. Danielmeier C
    3. Jocham G

    (2014) Neurophysiology of performance monitoring and adaptive behavior

    Physiological Reviews 94:35–79.

    https://doi.org/10.1152/physrev.00041.2012

    • Google Scholar
61. 1. Weiskopf N
    2. Hutton C
    3. Josephs O
    4. Deichmann R

    (2006) Optimal EPI parameters for reduction of susceptibility-induced BOLD sensitivity losses: a whole-brain analysis at 3 t and 1.5 t

    NeuroImage 33:493–504.

    https://doi.org/10.1016/j.neuroimage.2006.07.029

    • Google Scholar
62. 1. Yeung N
    2. Botvinick MM
    3. Cohen JD

    (2004) The neural basis of error detection: conflict monitoring and the error-related negativity

    Psychological Review 111:931–959.

    https://doi.org/10.1037/0033-295X.111.4.931

    • Google Scholar
63. 1. van Driel J
    2. Swart JC
    3. Egner T
    4. Ridderinkhof KR
    5. Cohen MX

    (2015) (no) time for control: frontal theta dynamics reveal the cost of temporally guided conflict anticipation

    Cognitive, Affective, & Behavioral Neuroscience 15:787–807.

    https://doi.org/10.3758/s13415-015-0367-2

    • Google Scholar
64. 1. von Stein A
    2. Sarnthein J

    (2000) Different frequencies for different scales of cortical integration: from local gamma to long range alpha/theta synchronization

    International Journal of Psychophysiology 38:301–313.

    https://doi.org/10.1016/S0167-8760(00)00172-0

    • Google Scholar

Author details

1. Hanlin Tang

   1. Program in Biophysics, Harvard University, Boston, United States
   2. Department of Ophthalmology, Boston Children's Hospital, Harvard Medical School, Boston, United States

   Contribution

   HT, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article

   Competing interests

   The authors declare that no competing interests exist.

2. Hsiang-Yu Yu

   1. Department of Neurology, Taipei Veterans General Hospital, Taipei, Taiwan
   2. National Yang-Ming University, Taipei, Taiwan

   Contribution

   H-YY, Performed the experiments, Commented and approved the manuscript, Acquisition of data

   Competing interests

   The authors declare that no competing interests exist.

3. Chien-Chen Chou

   1. Department of Neurology, Taipei Veterans General Hospital, Taipei, Taiwan
   2. National Yang-Ming University, Taipei, Taiwan

   Contribution

   C-CC, Performed the experiments, Commented and approved the manuscript, Acquisition of data

   Competing interests

   The authors declare that no competing interests exist.

4. Nathan E Crone

   Department of Neurology, Johns Hopkins School of Medicine, Baltimore, United States

   Contribution

   NEC, Performed the experiments, Commented and approved the manuscript, Acquisition of data

   Competing interests

   The authors declare that no competing interests exist.

5. Joseph R Madsen

   Department of Neurosurgery, Boston Children's Hospital, Harvard Medical School, Boston, United States

   Contribution

   JRM, Performed the neurosurgical procedures, Commented and approved the manuscript, Acquisition of data

   Competing interests

   The authors declare that no competing interests exist.

6. William S Anderson

   Department of Neurosurgery, Johns Hopkins Medical School, Baltimore, United States

   Contribution

   WSA, Performed the neurosurgical procedures, Commented and approved the manuscript, Conception and design, Acquisition of data

   Competing interests

   The authors declare that no competing interests exist.

7. Gabriel Kreiman

   1. Program in Biophysics, Harvard University, Boston, United States
   2. Department of Ophthalmology, Boston Children's Hospital, Harvard Medical School, Boston, United States
   3. Center for Brain Science, Harvard University, Boston, United States

   Contribution

   GK, Conception and design, Analysis and interpretation of data, Drafting or revising the article

   For correspondence

   gkreiman@gmail.com

   Competing interests

   The authors declare that no competing interests exist.

   "This ORCID iD identifies the author of this article:" 0000-0003-3505-8475

• 2,162

  Page views

• 535

  Downloads

• 23

  Citations

Article citation count generated by polling the highest count across the following sources: Crossref, Scopus, PubMed Central.