Search tips
Search criteria 


Logo of wtpaEurope PMCEurope PMC Funders GroupSubmit a Manuscript
Cereb Cortex. Author manuscript; available in PMC 2017 April 13.
Published in final edited form as:
PMCID: PMC5390852

Human Subthalamic Nucleus Theta and Beta Oscillations Entrain Neuronal Firing During Sensorimotor Conflict


Recent evidence has suggested that prefrontal cortical structures may inhibit impulsive actions during conflict through activation of the subthalamic nucleus (STN). Consistent with this hypothesis, deep brain stimulation to the STN has been associated with altered prefrontal cortical activity and impaired response inhibition. The interactions between oscillatory activity in the STN and its presumably antikinetic neuronal spiking, however, remain poorly understood. Here, we simultaneously recorded intraoperative local field potential and spiking activity from the human STN as participants performed a sensorimotor action selection task involving conflict. We identified several STN neuronal response types that exhibited different temporal dynamics during the task. Some neurons showed early, cue-related firing rate increases that remained elevated longer during high conflict trials, whereas other neurons showed late, movement-related firing rate increases. Notably, the high conflict trials were associated with an entrainment of individual neurons by theta- and beta-band oscillations, both of which have been observed in cortical structures involved in response inhibition. Our data suggest that frequency-specific activity in the beta and theta bands influence STN firing to inhibit impulsivity during conflict.

Keywords: conflict, deep brain stimulation, impulsivity, inhibition, theta oscillations


Optimal decision-making in the face of conflict depends on our ability to make deliberate rather than impulsive choices. Indeed, many pathological disorders, such as attention deficit hyperactivity disorder, obsessive compulsive disorder, and even schizophrenia, are associated with impaired response inhibition (Lipszyc and Schachar 2010). Inhibiting impulsive choices requires time, and even simple decisions that involve sensory conflict elicit longer response times relative to decisions without conflict (Eriksen and Eriksen 1974). Although a large body of evidence has identified various cortical regions involved in these processes (Botvinick et al. 2004), recent computational models have increasingly implicated the basal ganglia as the final point of convergence for conflict triggered response inhibition (Frank 2006). Specifically, these models posit that successful action selection relies on an intact subthalamic nucleus (STN) that can adjust the temporal dynamics of motor control during conflict, thus delaying movements until sufficient information to select the correct option has been integrated (Frank 2006; Bogacz and Gurney 2007; Wiecki and Frank 2013). Consistent with these models, the STN has been shown to interact with cortical structures involved in response inhibition (Aron et al. 2007; Herz et al. 2014; Zavala et al. 2014). Furthermore, disruption of STN activity via deep brain stimulation (DBS) is associated with altered cortical activity (Ballanger et al. 2009; Cavanagh et al. 2011; Swann et al. 2011) and impulsivity (Baunez and Robbins 1997; Frank et al. 2007; Hälbig et al. 2009; Cavanagh et al. 2011; Coulthard et al. 2012; Obeso et al. 2014).

The precise neural mechanisms responsible for the STN’s involvement in the delaying of responses in the face of conflict, however, remain unclear. Experimental studies have demonstrated conflict-related increases in STN oscillatory power in the theta (Cavanagh et al. 2011; Fumagalli et al. 2011; Zavala et al. 2013, 2014) and beta (Kühn et al. 2004; Brittain et al. 2012; Leventhal et al. 2012; Ray et al. 2012; Alegre et al. 2013; Bastin et al. 2014; Benis et al. 2014) bands. Similar increases in theta and beta oscillatory power have also been observed over cortical areas that mediate response inhibition such as the medial prefrontal cortex (mPFC; Cohen and Cavanagh 2011; Cavanagh et al. 2012) and the inferior frontal gyrus (IFG; Swann et al. 2009). As mPFC and IFG provide direct connections to the STN (Nambu et al. 2002; Aron et al. 2007), these results together suggest that coherent oscillations could reflect the coordination of neural activities between these structures and the basal ganglia during response inhibition (Alegre et al. 2013; Zavala et al. 2014). Traditionally, however, it is the firing rate dynamics of the basal ganglia that are assumed to be responsible for determining whether or not a response is executed (Albin et al. 1989; DeLong 1990). Several studies have demonstrated that STN spiking activity increases during motor control (Magariños-Ascone et al. 2000; Williams et al. 2005), and is further modulated when subjects are asked to prevent or delay responses (Isoda and Hikosaka 2008; Zaghloul et al. 2012; Schmidt et al. 2013; Bastin et al. 2014) or when they make decisions during high levels of doubt (Burbaud et al. 2013).

As both STN oscillatory power and spiking activity appear related to action selection, an important yet unanswered question is how to reconcile the classic neuronal firing rate model (Albin et al. 1989; DeLong 1990) of basal ganglia architecture with any role played by cortical and subcortical neuronal oscillations. This question has remained elusive partially because most studies involving behaviorally relevant recordings from the human STN are usually restricted to either local field potential (LFP) or single-unit recordings alone. Here, we directly study these neural mechanisms together and how they converge on the STN to help resolve conflict. We examined simultaneously recorded LFP and spiking activity from the human STN as participants undergoing DBS surgery for Parkinson’s disease performed an Eriksen flanker task. We identified several neuronal spiking response patterns in the STN, and specifically examined interactions between spiking activity and changes in the LFP theta and beta frequency bands.

Materials and Methods

Intraoperative Task and Recordings During DBS Surgery

We captured intraoperative recordings in 15 participants undergoing DBS surgery of the STN for Parkinson’s disease. The study was conducted in accordance with an NIH IRB approved protocol, and all subjects gave their written informed consent to take part in the study. Subjects received no financial compensation for their participation. Parkinson’s medications were stopped on the night before surgery (12 h preoperatively). We captured recordings while participants were alert, at rest, and in an OFF state while in the operating room.

As per routine DBS surgery, we used intraoperative microelectrode recordings to identify the STN based on firing rate and pattern (increased spiking activity and background noise relative to the more dorsal zona incerta and thalamus). We captured intraoperative recordings using targeting electrodes comprised of a pair of macro- and microelectrode contacts (Alpha Omega, Co., Alpharetta, GA, USA). Each macroelectrode contact was positioned 3 mm dorsal to the corresponding microelectrode contact. We simultaneously advanced 3 targeting electrodes, separately spaced 2 mm apart, during each recording session (placed along a central, 2 mm lateral, and 2 mm anterior trajectory; Fig. 1A). Raw signals were sampled at 1.5024 and 24.0345 kHz from macro- and microelectrode contacts, respectively, and stored using a MicroGuide Pro data acquisition system (Alpha Omega Co.). Mean coordinates of the central microelectrode recording sites during the behavioral paradigm, referenced to the mid-commissural point, were x = 11.1 ± 0.2, y = 3.6 ± 0.5, and z = 4.9 ± 0.6 for left electrode recordings, and x = −11.8 ± 0.2, y = 3.2 ± 0.6, and z = 4.5 ± 0.6 for right electrode recordings. Mean coordinates of the implanted DBS electrodes (Medtronic model 3389) were x = 9.8 ± 0.3, y = 3.9 ± 0.6, and z = 6.4 ± 0.5 for the left STN, and x = −10.8 ± 0.3, y = 3.4 ± 0.6, and z = 6.3 ± 0.5 for the right STN. These coordinates correspond to left and right STN on the Schaltenbrand–Wahren brain atlas. Microelectrode recording sites during the behavioral task were 2.6 ± 0.2 mm (56.4 ± 3.6% of STN length) ventral to the dorsal border of the STN, as determined by the clinical microelectrode recordings. All spiking activity captured from microelectrode recordings and used in these analyses were from locations judged to be within the STN based on the intraoperative clinical identification of the STN. Because the LFP macroelectrodes were positioned 3 mm dorsal to the microelectrodes, not all LFP recordings were made directly in the STN. All LFP and spike-phase–locking analyses were restricted to only macroelectrode contacts that were positioned within the STN.

Figure 1
Sensorimotor conflict task during recordings captured from the human STN.

Behavioral Task

Participants performed an Eriksen flanker sensorimotor action selection task (Fig. 1A; Eriksen and Eriksen 1974) in the intraoperative environment on a testing laptop (PyEPL, Geller et al. 2007). On each trial, a white fixation dot initially appeared in the center of the screen for 900 ms. We subsequently displayed flanking arrows, all pointing in the same leftward or rightward direction, to the left and right of the fixation dot. Following the flanking arrows by 100 ms, we presented a target arrow between the flanking arrows above the fixation dot. Target arrows could point in the same direction (low conflict) or in the opposite direction (high conflict) as the flanking arrows. Arrows were displayed on the screen until the participant indicated the direction of the target arrow or until 4 s elapsed. Participants indicated their responses by moving a digital joystick to the left or to the right. Participants made all movements with the hand contralateral to the side of intraoperative recording for that session. Though 7 of the 15 patients had tremor as one of their primary symptoms, this did not interfere with their ability to control the joystick. Patients with a tremor that was severe enough to interfere with their control of the joystick were unable to perform the task and therefore excluded from the study. Trials were separated by a blank screen displayed for a random duration of 1000 ± 100 ms. Throughout an experimental session, we balanced the number of low and high conflict trials, and the number of trials requiring leftward and rightward movements. Participants completed 3 blocks of 40 trials each during each experimental session. We discarded all incorrect trials as well as all trials with response times <300 ms and >1500 ms.

On the day before surgery, participants performed a complete session to familiarize themselves with the task. During the operative procedure, most participants performed one session while recordings were captured from the left and right STN. Five participants did not complete the second session because of fatigue. Of the 10 participants who performed the task during both the left and right STN recordings, 4 of them showed a steep drop off in performance during the second session, likely also due to fatigue. To exclude, in an unbiased manner, sessions in which the participants did not perform the task correctly, we calculated the inverse of the response times to normalize their distributions (Carpenter and Williams 1995), and then used an unpaired t-test to compare the inverse response times between low and high conflict trials. We excluded all experimental sessions that did not exhibit a significant difference between low and high conflict trials (6 sessions excluded). Thus, only 6 participants met the behavioral criteria for both the left and the right STN, resulting in 19 total intraoperative STN recordings included in the analysis (Table 1). Two of these 6 participants performed an additional third session during recordings from one side. The third session was recorded at a slightly different depth than the second session so as to record the activity of additional cells. For 1 of these 2 participants, we only retained LFP power and phase analysis from the first recording session so as to remain consistent with the remaining participants. For the second participant, however, we retained data from the second session because the macroelectrode contacts during the first session were not within the STN. Because we performed spike firing rate analyses on individual cells, we included cells from both sessions for spiking analysis. Three months after surgery, 14 participants performed 2 additional sessions of the task, first with their DBS stimulators turned on and then with their stimulators turned off (see Supplementary Table 1 for stimulation parameters). Patients were given approximately 5 min to acclimate to the off state. These postoperative sessions were performed while patients were on their regular Parkinson’s medications.

Table 1
Clinical details

LFP Power and Phase

All analyses were performed using custom MATLAB scripts (Mathworks, Natick, MA, USA). We extracted LFP activity by bandpass filtering macroelectrode signals between 1 and 500 Hz, notch filtering at 60 Hz, and downsampling all LFP signals to 1 kHz. We then referenced macroelectrode signals to the common average of all simultaneously recorded macroelectrode contacts to yield 3 referenced monopolar channels per STN. Prior to any subsequent analysis, we discarded all trials that exhibited a clear artifact in the LFP trace. We also performed all analyses using bipolar referencing between macroelectrode contacts, and confirmed nearly identical conflict-related differences in theta power and intertrial phase consistency (ITPC). We investigated phase reversals in the average event-related potentials using bipolar referencing, and found that 13 of the 15 STNs we included in our macroelectrode LFP analysis showed a phase reversal in at least one bipolar pair.

To obtain magnitude and instantaneous phase information in the frequency domain, we convolved the macroelectrode LFP signal from each trial with complex-valued Morlet wavelets (wave number 5). We used 47 logarithmically spaced (8 scales/octave) wavelets between 2 and 107 Hz and convolved each wavelet with 2000 ms of LFP data from each trial. For cue-locked analyses, we analyzed LFP signals from 500 ms before to 1500 ms following target arrow presentation. For response-locked analyses, we analyzed LFP signals from 1000 ms before to 1000 ms after the response. We used a 1500-ms buffer on both sides of the clipped data to eliminate edge effects. We squared the magnitude of the continuous-time wavelet transform to generate a continuous measure of instantaneous power. We z-scored power for each channel separately for each frequency using the mean and standard deviation of the power recorded from that channel during all fixation periods (baseline; 1000–200 ms before target arrow onset). For each STN recording session, we averaged the z-scored power from macroelectrodes that were within the STN as identified during the operative procedure.

To assess differences in spectral power between high and low conflict trials across participants, we first calculated the mean difference in z-scored power for each time point and for each frequency separately across all low and high conflict trials for each experimental session and averaged across STN recording sessions. All comparisons between high and low conflict trials were made without segregating into trials involving leftward and rightward movements. We assigned a P-value to each time–frequency point by comparing the true difference with the distribution of differences from 1000 surrogate datasets generated by permuting the low and high conflict trials prior to averaging across trials and sessions. The P-values were then corrected for multiple comparisons using exceedance mass testing (Maris and Oostenveld 2007), which involved integrating the excess mass of supra-threshold clusters in each permutation and recording the most significant cluster. The top 5% of this distribution then determined the corrected threshold for image-wise significance in the real dataset.

To assess for differences in ITPC, we defined phase values for a single time point across all trials as unit vectors on the complex plane, and separately calculated the mean vector length (ITPC) across all low and high conflict trials. We separately calculated differences in ITPC between low and high conflict trials at every time point and frequency for each channel, and then averaged differences across channels within each STN. We used the non-parametric clustering-based procedure described above to assess whether differences in ITPC were significant across participants.

We also analyzed conflict-related differences in evoked activity. We first calculated the mean event-related potential separately for the low and high conflict trials, and then calculated the power spectrum of the resulting average event-related potentials. We z-scored the resulting evoked spectrograms for the low and high conflict trials by the mean evoked response during the baseline period (1000–200 ms before target arrow onset) for all trials. We averaged the z-scored high and low conflict spectrograms across channels, and then across STNs. We assessed for significant differences in the evoked activity across participants by using the same nonparametric clustering-based permutation procedure described above.

To confirm that the observed power increases were indeed due to changes in oscillatory power, in each participant we separately subtracted the low and high conflict evoked spectrogram from the time–frequency spectrogram calculated during each low and high conflict trial, respectively, to generate spectrograms of induced activity. We performed a similar calculation on the baseline period (1000–200 ms before target arrow onset) from all trials to generate a mean and standard deviation of induced baseline activity. We z-scored the induced spectrograms from all low and high conflict trials with this mean and standard deviation. We averaged the z-scored high and low conflict spectrograms across channels, and then across STNs. We assessed for significant differences in the induced activity across participants by using the same nonparametric clustering-based permutation procedure described above.

Spiking Activity

We extracted spiking activity by bandpass filtering microelectrode recordings between 0.3 and 3 kHz and resampling the filtered signals at 24 kHz. We identified individual spike clusters offline for each recording using principal component analysis (Plexon Offline Sorter, Inc., Dallas, TX, USA). Spikes were detected using a voltage threshold method. The average threshold used was 3.16 ± 0.12 standard deviations away from the mean voltage, and the average neuron showed a mean peak voltage of 5.20 ± 0.28 standard deviations away from the mean voltage. We classified a spike cluster as belonging to a single neuron if <2% of the spike events in that cluster fell within a refractory period of 1.5 ms and if all spike waveforms clustered together in principle component space. However, single-unit isolation is difficult to achieve in the STN (Weinberger et al. 2006; Sharott et al. 2014), and it is possible that some of the units we recorded reflected the activity from more than one neuron. Out of the 32 microelectrode recordings that demonstrated spiking activity and had appropriate behavior, 6 recordings demonstrated a clear separation of 2 waveforms in principle component space and clearly reflected 2 spike clusters. We therefore exported data from each of these 6 recordings as 2 separate clusters, although only at most one of each of these 2 clusters exhibited task-related increases in firing and was used in our subsequent analyses. Henceforth, we will refer to the 38 exported clusters as cells. The average number of cells per STN recorded was 2 ± 0.40 cells.

To determine whether an individual cell was responsive during the task, we divided the period between target arrow onset and response into 3 equally sized temporal epochs for each trial based on that trial’s response time. We additionally defined a baseline epoch during the fixation period (1000–200 ms before target arrow onset). We then compared spiking activity during each of the 3 temporal epochs with activity during the baseline epoch across all trials using a paired t-test. If any of the 3 temporal epochs exhibited a significant difference in activity (P < 0.016, Bonferroni-corrected across epochs), we considered the cell to be responsive. We assessed responsiveness by investigating 3 separate temporal epochs because some cells only exhibited brief changes in firing that would otherwise be missed if we averaged over the entire period. To determine if a cell was directionally tuned, we divided trials into rightward and leftward movements of a joystick, and separately compared spiking activity in each temporal epoch between both movements. We considered any cell that exhibited a significant direction-related difference in spiking activity in any of the 3 temporal epochs (P < 0.016, Bonferroni-corrected) to be directionally tuned.

To identify any consistencies in the temporal firing rate dynamics observed during the task, we calculated the continuous-time firing rates for each cell by smoothing the spike train from each trial (1 ms bins) with a Gaussian kernel (standard deviation 100 ms). Though temporal smoothing increases the signal-to-noise ratio of the firing rate data, it comes at a cost of decreased temporal precision. To generate a z-scored firing rate, we compared continuous-time firing rates for each trial with the mean and standard deviation of the firing rates during the baseline period (1000–200 ms before target arrow onset) and then averaged across trials. To determine if different cells clustered into populations that exhibited similar temporal dynamics, we identified the time point at which each cell’s average firing rate reached a maximum for both the cue-aligned and the response-aligned data. We used k-means clustering to divide the cells into different clusters based on their peak response times. By measuring the relative distance of each point within a cluster to the points in the neighboring clusters (silhouette value, Rousseeuw 1987), we were able to assess clustering using a variable number of clusters (from 2 to 10 clusters). This revealed that the optimal number of clusters was 2 as this produced the highest mean silhouette value (mean silhouette = 0.82 ± 0.04).

To determine if continuous-time z-scored firing rates exhibited a significant difference between high and low conflict trials across cells, for every cell we calculated the mean difference in firing rate at every time point between trial types. We then calculated the average difference across all cells within a given group (early cells, late cells, and down cells) and compared the difference with an empiric distribution generated by permuting the low and high conflict trial labels 1000 times. We then corrected for multiple comparisons using the same nonparametric clustering-based procedure. We also identified time points during which the average firing rate was significantly (P < 0.05, one sample t-test) greater than zero across trials for individual cells and across cells for the group-level analysis. For the group-level analysis, we determined if differences in onset or offset times were significant between trial types by permuting the low and high conflict trial labels 1000 times.

Spike-Phase Interactions

We calculated spike-phase interactions for each cell by first filtering LFP signals within each frequency band of interest and then extracting instantaneous phase information using a Hilbert transform. We only included cells that had macroelectrode recordings within the STN for these analyses. As conflict-related theta-band activity has been observed in the STN in frequencies as low as 2 Hz (Cavanagh et al. 2011), we used the following frequency bands: theta (2–8 Hz), alpha (9–14 Hz), and beta (15–30 Hz). We tabulated the instantaneous phase of the macroelectrode LFP signal during all spiking events during each trial (from target arrow onset until response) captured on the associated microelectrode. We calculated the mean vector length of phases across all spike events in all low and high conflict trials, separately yielding a spike-phase–locking value for each trial type (rlow and rhigh). We normalized the resulting spike-phase–locking values by permuting phase information across trials. Whereas in the true case, spike times from a given trial were assigned an instantaneous phase from that same time point in the same trial, in each permuted case we assigned to each spike time an instantaneous phase from the same time point in a different trial drawn at random from the pool of similar conflict level trials (i.e., rlow values were normalized by low conflict trials and rhigh values were normalized by high conflict trials). We permuted phase information separately for low and high conflict trials to prevent any biases that may emerge from nonuniform phase distributions caused by differences in ITPC. In this manner, we normalized spike-phase–locking values observed for a given trial type by the probability of observing spike-phase locking by chance given the distribution of phases and spike times observed during that trial type. We permuted phase information 1000 times resulting in a distribution of 1000 surrogate rlow and rhigh values for each cell. We compared the true rlow and rhigh with the mean and standard deviation of the distribution of permuted values to generate a normalized spike-phase–locking value for each cell and for each trial type (Rlow and Rhigh).

To test for differences in spike-phase locking between trial types, we compared the distribution of normalized values between low and high conflict trials across cells using a paired t-test. We also confirmed differences between trial types by calculating the pairwise phase consistency for low and high conflict trials (Vinck et al. 2010) as well as by using the wavelet transform to extract the phase information. To generate a continuous metric of spike-phase locking, we calculated the normalized mean vector length for each trial type during sliding 750 ms windows, stepped every 50 from 500 before to 1500 ms following cue presentation, and from 1000 before to 1000 ms following response. We determined if individual time points exhibited significant differences between trial types by using the same permutation and the nonparametric clustering-based procedure described above to correct for multiple comparisons.

To visualize spike-phase interactions in a single cell (Fig. 5A,B), we extracted the instantaneous phase information of the filtered theta (2–8 Hz) or beta (15–30 Hz) LFP signal captured from the corresponding macroelectrode using a Hilbert transform. We identified all of the troughs of the oscillation observed between the target arrow onset and the motor response in each high conflict trial. We then calculated the continuous-time firing rate in a 700-ms (theta) or 400-ms (beta) temporal epoch surrounding each trough by convolving the spike train (1 ms bins) with a Gaussian kernel (standard deviation 15 ms). We used a 15-ms kernel in order to maintain the temporal resolution necessary to discern spike-phase interactions. We detrended the continuous-time firing rates during each epoch, and then averaged across all epochs. To visualize theta- and beta-band activity, we also extracted the raw macroelectrode LFP trace in the same epochs and averaged across all epochs during all high conflict trials.

Figure 5
Theta and beta oscillations entrain STN cells during conflict.


Behavioral Task

We captured simultaneous micro- and macroelectrode recordings from the STN during DBS surgery for Parkinson’s disease in 15 participants [5 males; 58.1 ± 2.67 (mean ± SEM) years old] as they performed an Eriksen flanker sensorimotor task (Fig. 1A). Patient details are summarized in Table 1. Because DBS electrodes were implanted bilaterally in all participants, some participants performed the task twice yielding 25 total experimental sessions. During each session, we captured activity from 3 macro-/microelectrode pairs to investigate STN LFP and spiking activity (Fig. 1A).

We labeled every trial as a low or high conflict trial, depending on whether a cue arrow displayed centrally on the screen was congruent or incongruent with flanking arrows. We found that in 19 experimental sessions (13 participants), participants exhibited significantly greater response times (P < 0.05, unpaired t-test) during high conflict compared with low conflict trials. In these sessions, participants were significantly more likely to make a correct movement during low conflict (98 ± 1%) compared with high conflict trials (91 ± 2%; t(18) = −4.36, P = 0.001; Fig. 1B). We retained correct trials in these sessions for subsequent analyses. During the retained sessions, participants responded with a joystick movement 840.10 ± 43.83 ms following cue presentation during high conflict trials and 704.98 ± 39.65 ms following cue presentation during low conflict trials (Fig. 1B). We found that displaying high and low conflict trials did not have a significant effect on the response time of the subsequent trial regardless of whether the subsequent trial involved high or low conflict (across subjects: t(18) = 0.2, P = 0.81 and t(18) = 1.8, P = 0.09, respectively; within subject: P > 0.05 for all subjects for both low and high conflict subsequent trials; Gratton et al. 1992). We therefore treated all correct trials identically regardless of conflict in the previous trial.

Fourteen of the participants performed the task 2 additional times, once with the implanted DBS stimulator turned on and once with the stimulator turned off, during an outpatient follow-up 3 months following surgery. These participants exhibited significantly higher error rates during high conflict trials, significantly more errors when their stimulators were on, and a significant interaction between stimulation and conflict (ANOVA, within-subject repeated-measures, conflict × stimulation: conflict F1,12 = 7.74, P = 0.016; stimulation F1,12 = 7.37, P = 0.018; interaction F1,12 = 9.08, P = 0.01; Fig. 1C). Reaction time differences also showed a main effect of conflict and stimulation (conflict F1,12 = 101.94, P < 0.001; stimulation F1,12 = 14.67, P = 0.002), but we did not find a significant interaction between the two (interaction F1,12 = 1.2, P = 0.29).

Macroelectrode LFP Power and Phase Changes

We investigated changes in spectral power in the macroelectrode LFP signal, and found that both low and high conflict trials exhibited similar decreases in beta and increases in gamma oscillatory power during the task (Fig. 2A). Theta-band activity, however, showed significantly greater power during the high conflict trials compared with low conflict trials following cue presentation (P < 0.05, permutation test; for data time-locked to the response, see Supplementary Fig. 1A). We confirmed that these changes reflect induced theta oscillatory activity, rather than changes in the evoked event-related potential (P < 0.05, permutation test; see Materials and Methods). We also calculated the macroelectrode LFP ITPC for every frequency and found that high conflict trials had significantly greater theta-band phase consistency relative to low conflict trials following cue presentation (P < 0.05, permutation test; Fig. 2B). We did not find significant differences in theta-phase consistency between high and low conflict trials when we time-locked trials to the motor response (Supplementary Fig. 1B), indicating that the observed changes in phase consistency were specific to the cue (Zavala et al. 2013).

Figure 2
STN theta-band activity is higher during conflict.

Microelectrode Single-Unit Firing Changes

We extracted spiking activity from microelectrode recordings in the STN while participants engaged in the task and identified 38 neuronal firing clusters, which we will henceforth refer to as cells [33.7 ± 3.2 spikes/s (mean ± SEM)]. We examined the spiking activity following cue presentation and found 22 cells that individually exhibited a significant (P < 0.016, paired t-test, see Materials and Methods) increase in spiking activity between cue presentation and the response. Nine of these cells exhibited significant differences in firing rates that depended on the direction of movement of the contralateral hand (Supplementary Fig. 2). Seven additional cells responded with a significant (P < 0.016) decrease in firing between cue presentation and the response (down cells, Supplementary Fig. 3), and 9 cells did not demonstrate a significant change. We focused our subsequent analyses on the 22 cells exhibiting a significant increase in spiking activity during the task.

Representative spiking activity from a single cell exhibiting a significant increase in firing during the task is shown in Figure 3A. Spiking activity for this cell is time-locked to the target arrow onset. Representative spiking activity from a different cell is shown in Figure 3B and exhibits a peak in firing during the motor response. To examine whether these cells represent distinct response types, we identified principal components of the continuous-time firing rate that accounted for at least 10% of the variance across all 22 of the cells that exhibited a significant increase in spiking activity during the task (Fig. 3C; Narayanan and Laubach 2009). One principal component accounted for 45% of the variance across cells and exhibited an increase that peaked around the response, whereas a second component accounted for 34% of the variance and exhibited a peak that occurred immediately following cue presentation. These 2 principal components suggested that different cells (or neuronal firing clusters) may preferentially respond at different times during the task. We therefore identified the time of peak firing for cells relative to the target arrow onset and to the motor response. We used these peak response times to divide these 22 cells into 2 populations using k-means clustering (Fig. 3C; see Materials and Methods). One group of cells (n = 8) exhibited spiking activity on average that peaked immediately following cue presentation (early cells), and a second group of cells (n = 14) exhibited spiking activity on average that peaked around the motor response (late cells; Fig. 3D,E). We defined the physical location of each cell as the relative depth between the dorsal and ventral border of the STN, and found that early cells were found in a more ventral location (t(20) = 2.19, P = 0.04, unpaired t-test; Fig. 3F). Though our clustering procedure suggests that at least 2 separate response patterns were present in the data, not all cells could be unambiguously categorized as early or late responders.

Figure 3
Task-related increased spiking activity in STN cells.

To explore the effect of conflict on spiking activity, we separately examined spiking activity during high and low conflict trials for early and late cells. Spiking activity for a representative early cell demonstrated an initial significant increase (P < 0.05, one sample t-test) during both high and low conflict trials that returned to baseline prior to the response. This increase was maintained for a longer period of time during high conflict trials (~650 ms for high conflict vs. 450 ms for low conflict trials, Fig. 4A). We found that this distinction between high and low conflict trials was consistent as 7 of the 8 early cells maintained their firing rates above zero for a longer period of time during the high conflict condition. Next, we averaged the firing rate across all early cells and found that the firing rate after the initial peak was significantly higher for the high conflict trials (P < 0.05, permutation test; Fig. 4B, left; for data aligned to the response, see Supplementary Fig. 4). We calculated the time points at which average spiking activity was significantly greater than baseline across all early cells for the low and high conflict trials separately. There were no significant differences in the onset of average spiking activity (P > 0.05, permutation test), but low conflict trials exhibited spiking offset that significantly preceded spiking offset during high conflict trials (~450 vs. 750 ms after the cue presentation, P < 0.01, permutation test).

Figure 4
Conflict-related differences in increased spiking activity in STN cells.

In contrast to the early cells, the late cells showed no conflict-related differences in the activity of both low and high conflict trials peaked at the time of the response (Fig 4C; for data aligned to the target arrow onset, see Supplementary Fig. 4). We found no significant differences in spiking onset times relative to cue presentation in the average spiking activity across this population of cells. When locked to the response, average spiking activity during high conflict trials tended to exhibit an increase above baseline at a time earlier than during low conflict trials (~1000 vs. 550 ms), but this difference in spiking onset was not significant (P = 0.13, permutation test).

LFP–Single-Unit Interactions During Conflict

To explore whether the observed changes in oscillatory LFP activity were linked to the changes in firing rates, we investigated the interaction between the oscillatory phase of the macroelectrode LFP signal and the microelectrode spiking activity. In a single cell or neuronal firing cluster, we observed that spiking activity occurred more frequently during the rising phase of a theta oscillation (2–8 Hz; Fig. 5A, top). To quantify this relation, we tabulated the instantaneous theta phase for all spiking events that occurred during high conflict trials, and found that the distribution of phases was significantly less uniform than the distribution that would occur by chance (Rhigh = 2.90, P = 0.002; see Materials and Methods; Fig. 5A, bottom). In another cell or neuronal firing clusters, we observed that spiking activity occurred near the peaks of a beta oscillation (15–30 Hz; Fig. 5B, top). As in the theta-locked cell, the distribution of instantaneous phases of the beta oscillation captured at each spiking event in the beta-locked cell was significantly less uniform than the distribution that would occur by chance (Rhigh = 3.74, P < 0.001; Fig. 5B, bottom).

For all 17 cells that both demonstrated an increase in spiking activity during the task and had coincident LFP macroelectrode recordings within the STN, we calculated spike-phase locking for the theta (2–8 Hz), alpha (9–14 Hz), and beta (15–30 Hz) frequency ranges during high and low conflict trials. We found that across these cells, the extent of spike-phase locking during high conflict trials between the target arrow onset and the motor response was significantly greater than during the low conflict trials in the theta (t(16) = 2.51, P = 0.023) and beta (t(16) = 3.01, P = 0.008) bands (Fig. 5C). These differences arose because the theta and beta bands showed significantly nonzero spike-phase locking during the high conflict trials (theta: t(16) = 2.30, P = 0.018; beta: t(16) = 2.96, P = 0.005) but not during the low conflict trials (theta: t(16) = −1.44, P = 0.92; beta: t(16) = 0.18, P = 0.43). We confirmed these differences by quantifying spike-phase locking using pairwise phase consistency (Vinck et al. 2010) and found significantly greater spike-phase locking during high compared with low conflict trials in the theta and beta bands (P < 0.05). We did not observe any significant spike-phase locking in the alpha band. When we examined time evolving spike-phase locking, we found significantly greater theta and beta spike-phase locking during high conflict trials compared with low conflict trials that was specific to the target arrow onset (P < 0.05, permutation test; Fig. 5D,E and Supplementary Fig. 5).

Because the spiking responses of individual cells suggested the presence of both early and late cell types, we examined whether the observed spike-phase interactions affected these populations differently. A two-way ANOVA of spike-phase locking between the cue and motor response did not reveal a significant conflict-by-cell type interaction for either the theta (conflict F1,15 = 6.63, P = 0.01; cell type F1,15 = 0.71, P = 0.41; interaction F1,15 = 1.01, P = 0.31) or beta (conflict F1,15 = 8.48, P = 0.01; cell type F1,15 = 1.25, P = 0.28; interaction F1,15 = 0.35, P = 0.56) bands. We also did not find evidence for significant spike-phase locking in the down cells (P > 0.05), suggesting that spike-phase locking was specific to conflict in the cells exhibiting an increase in spiking activity during the task.


Consistent with previous studies (Zavala et al. 2015),we found increased theta oscillatory power and related changes in spiking activity in the STN during perceptual decisions that involved high conflict. We also found 3 different neuronal firing patterns, some of which were differentially modulated by conflict during the task. Importantly, by directly measuring simultaneous LFP and spiking activity in the human STN, our data demonstrate significant interactions between changes in the theta and beta bands and spiking activity during high conflict trials. Taken as a whole, our data provide insights into the neural mechanisms underlying the STN’s role in mediating actions during conflict and add to a growing body of evidence demonstrating the importance of the STN in effective decision-making (Frank et al. 2007; Hälbig et al. 2009; Cavanagh et al. 2011; Obeso et al. 2014).

Though theta oscillations in the basal ganglia have traditionally been associated with pathologic tremors (Deuschl et al. 1996; Magariños-Ascone et al. 2000), a growing body of evidence suggests that both mPFC and STN theta oscillations may play a physiologic role in encoding conflict. Both sites demonstrate conflict-related increases in theta-band activity (Cavanagh et al. 2011, 2012; Cohen and Cavanagh 2011; Fumagalli et al. 2011; Brittain et al. 2012; Zavala et al. 2013, 2014; Cohen and van Gaal 2014), altered theta-band activity has been associated with impulse control disorders (Cavanagh et al. 2011; Rodriguez-Oroz et al. 2011; Rosa et al. 2013), and mPFC theta oscillations drive those of the STN during conflict (Zavala et al. 2014). Furthermore, previous studies suggest that STN DBS both alters activity in the mPFC (Ballanger et al. 2009) and reverses the relation between mPFC theta oscillations and conflict-induced slowing (Cavanagh et al. 2011). Taken together, these data suggest that synchronized theta oscillations, and the attendant synchronized patterning of neural activity that underlies them, may play a role in how conflict-related information is conveyed. Our data build upon these studies by demonstrating that theta-band oscillations are associated with changes in the firing rate dynamics in the STN during action selection in the presence of conflict.

Our data also demonstrate conflict-related spike locking to beta-band activity in neurons exhibiting increases in firing during the task. Increases in beta-band power have been observed during response inhibition (Kühn et al. 2004; Brittain et al. 2012; Leventhal et al. 2012; Ray et al. 2012; Alegre et al. 2013; Bastin et al. 2014; Benis et al. 2014) and are thought to underlie the excessive inhibition of movements seen in Parkinson’s disease (Hammond et al. 2007). Conversely, as we observed in our data, drops in basal ganglia beta-band power may allow cells to increase their firing rate in a movement-related manner (Courtemanche et al. 2003). The enhanced beta-band entrainment we observed during conflict suggests a possible link between changes in beta oscillations and task-related increases in firing rate. We hypothesize that while decreases in beta-band power may allow spiking activity to increase, and therefore movement to occur, beta-band spike-phase locking may play a role in delaying when this happens during conflict.

Within the neurons that showed an increase in firing during the task, we observed spiking activity that peaked throughout the decision period. We used a clustering procedure to categorize these cells into 2 different populations, although we cannot categorically exclude the possibility that their spiking responses may be drawn from a more continuous distribution. Nevertheless, dividing cells through clustering does offer some possible insights into the mechanisms of action selection during conflict. We identified several STN neurons that seemed to exhibit a firing rate increase that was more clearly related to the stimulus onset. Computational models have predicted that early increases in STN spiking activity during a decision may represent an antikinetic signal, conveyed directly from the cortex to the STN, that initially halts all motor responses at the beginning of each decision (Nambu et al. 2002; Frank 2006). Our finding that these early cells demonstrated a longer firing rate increase during conflict is consistent with these models. Our data further demonstrate greater theta and beta spike-phase locking during high conflict compared with low conflict trials. One possibility, then, is that greater oscillatory entrainment may be important in maintaining elevated STN firing rates during conflict in these early cells, thereby prolonging the inhibition of movements.

We also identified STN neurons that exhibit increases in spiking activity that seemed more closely related to the execution of the motor response. These late firing cells did not demonstrate any conflict-related differences in their firing rate increases. The timing of their response may therefore entirely depend on external inputs that determine when movements should occur. Whether these inputs arise from other populations of STN neurons, such as the early cells, or other structures outside the STN is unknown. Notably, the traditional rate models of basal ganglia architecture assign STN firing an antikinetic role (Albin et al. 1989; DeLong 1990), yet the movement-related increases we observed suggest a more complex organization (Goldberg et al. 2013). One possibility is that basal ganglia circuits may be organized in multiple parallel motor loops with a center-surround architecture (Mink 1996; Nambu et al. 2002). Given this model, the increases in late cell firing may reflect the inhibition of actions that might interfere with the execution of the appropriate response. Consistent with a center-surround interpretation, we found a subset of neurons that decreased firing during the task and another subset of neurons that exhibited significant differences in spiking activity depending on the direction of movement.

Although our data suggest that the spiking activity of some STN neurons is related to conflict, whereas the spiking activity of others is not, the distinction between early and late cells and the timing of their peak spiking activity is not entirely clear. Previous work has suggested a dorsal-motor, ventral-cognitive segregation of the STN (Rodriguez-Oroz et al. 2001, 2011; Mallet et al. 2007; Lardeux et al. 2009). We did find that the early cells that were more responsive to conflict were located more ventrally, yet we also found late cells throughout the STN. Furthermore, we found that all cells exhibiting significant increases in firing during the decision period demonstrated significant spike-phase locking to both theta-band and beta-band oscillations in the presence of conflict, and that there was not a significant difference in phase locking between cell types. Hence, increased oscillatory entrainment may be the neural mechanism that dictates the interaction between these populations and the precise timing of movements during conflict.

Consistent with previous findings (Baunez and Robbins 1997; Frank et al. 2007; Coulthard et al. 2012; Green et al. 2013; Antoniades et al. 2014; Obeso et al. 2014), when we disrupted the STN network postoperatively with DBS, participants exhibited more impulsive errors in this simple decision task. Notably, our subjects did not demonstrate any stimulation-related changes in conflict-related slowing of reaction times. During stimulation both low and high conflict reaction times were faster, most likely due to the improvements in motor control caused by DBS. Our finding that DBS made patients more error prone without changing their conflict-induced reaction time slowing adds to the complex story concerning STN DBS and impulsivity (Zavala et al. 2015). Indeed, several studies have shown that STN DBS actually leads to improved response inhibition in some cases (Schroeder et al. 2002; van den Wildenberg et al. 2006; Swann et al. 2011; Mirabella et al. 2012). One possibility is that STN DBS disrupts early processes that are involved in inhibiting impulsive errors, but actually improves later processes that are involved in the continual suppression of the incorrect response (Wylie et al. 2010). Our electrophysiological findings are consistent with these predictions as our early STN cells displayed conflict-related differences, and thus the disruption of these early cells may result in increased errors.

Although our data provide direct evidence that oscillatory activity in the STN entrains spiking activity during conflict, it is important to note several limitations of the study that may restrict its interpretation. First, our data were collected from the Parkinsonian STN in an intraoperative environment from patients in whom Parkinsonian medication was withheld. In the off-dopamine state, the Parkinsonian STN demonstrates altered beta-band activity (Levy et al. 2000, 2002; Hammond et al. 2007; Oswal et al. 2012) that may affect spike-phase locking within that frequency band. However, our data primarily contrast beta-band spike-phase locking during low and high conflict trials, suggesting that these differences may exist regardless of whether beta-band activity is disrupted. Furthermore, participants in our study exhibited behavioral response time differences between low and high conflict trials, similar to healthy controls (Eriksen and Eriksen 1974) and the LFP changes in theta power were similar to those we have previously observed in postoperative recordings in which patients were given their medication (Zavala et al. 2013, 2014). The second limitation concerns the low number of cells that were included in the analysis. Out of the 38 cells or neuronal firing clusters we recorded, only 29 were task responsive and only 22 showed an increase in firing during the task. We further divided our data by separating these 22 cells into early and late cells. As the STN is thought to receive inputs from various circuits involved in movement and cognition, it may be the case that several other response patterns coexist alongside the 3 patterns (early up, late up, and down cells) we have identified. This limitation, however, does not detract from our findings that some STN cells were robustly entrained to the theta and beta bands, but only during trials involving conflict. Finally, although we used a simple sensorimotor decision task that involved short response times, conflict-related changes have also been observed in the STN during more complicated decisions requiring choosing between different values (Cavanagh et al. 2011; Fumagalli et al. 2011; Zaghloul et al. 2012). Whether the neural mechanisms mediating value conflict are identical to those mediating sensorimotor conflict, as observed here, remains an open question.

Despite the limitations noted above, our results suggest that theta and beta oscillatory activity interact with STN neurons in a way that could delay movement during conflict. We found increases in theta oscillatory power and a related entrainment of neurons during high conflict trials. Parallel to this, we also found beta-band suppression and increases in beta entrainment. These changes occur in the presence of a more persistent elevation of discharge rates among early cells, thereby prolonging motor inhibition and buying time for the selection of the most appropriate behavioral response. Finally, both low and high conflict trials were accompanied by increases in late cell discharge rates that may be related to the execution of an action. These data therefore suggest the presence of STN neuronal entrainment to 2 separate frequency bands during conflict that may influence the timing of responses during action selection. Although the origins of 2 STN oscillatory activity still remain largely unknown, previous work has suggested that the STN may interact with several nodes of a cognitive control network in parallel (Aron et al. 2007; Wiecki and Frank 2013; Herz et al. 2014) and that theta and beta oscillations may help synchronize these nodes with the STN and with each other during conflict (Hanslmayr et al. 2007; Cavanagh et al. 2009; Swann et al. 2009, 2012; Cohen and Cavanagh 2011; Ruiz et al. 2011; Cohen and van Gaal 2014; Zavala et al. 2014). Although our data do not explicitly address whether synchronized theta- and beta-band activity between these structures affects spiking activity in the STN, they do raise that possibility by providing a link between conflict-related oscillatory changes and STN neuronal firing.

Supplementary Material

Supplementary material can be found at:

Supplementary information



This work was supported by the National Institutes of Health Intramural Research Program (B.Z., S.D., J.D., C.L., and K.Z.) as well as by the Medical Research Council (P.B.) and the National Institute for Health Research Oxford Biomedical Research Centre (P.B.).



We thank all of the patients who participated in the study.

Conflict of Interest: None declared.


  • Albin RL, Young AB, Penney JB. The functional anatomy of basal ganglia disorders. Trends Neurosci. 1989;12:366–375. [PubMed]
  • Alegre M, Lopez-Azcarate J, Obeso I, Wilkinson L, Rodriguez-Oroz MC, Valencia M, Garcia-Garcia D, Guridi J, Artieda J, Jahanshahi M, et al. The subthalamic nucleus is involved in successful inhibition in the stop-signal task: a local field potential study in Parkinson’s disease. Exp Neurol. 2013;239:1–12. [PubMed]
  • Antoniades CA, Bogacz R, Kennard C, FitzGerald JJ, Aziz T, Green AL. Deep brain stimulation abolishes slowing of reactions to unlikely stimuli. J Neurosci. 2014;34:10844–10852. [PMC free article] [PubMed]
  • Aron AR, Behrens TE, Smith S, Frank MJ, Poldrack RA. Triangulating a cognitive control network using diffusion-weighted magnetic resonance imaging (MRI) and functional MRI. J Neurosci. 2007;27:3743–3752. [PubMed]
  • Ballanger B, van Eimeren T, Moro E, Lozano AM, Hamani C, Boulinguez P, Pellecchia G, Houle S, Poon YY, Lang AE, et al. Stimulation of the subthalamic nucleus and impulsivity: release your horses. Ann Neurol. 2009;66:817–824. [PMC free article] [PubMed]
  • Bastin J, Polosan M, Benis D, Goetz L, Bhattacharjee M, Piallat B, Krainik A, Bougerol T, Chabardès S, David O. Inhibitory control and error monitoring by human subthalamic neurons. Transl Psychiatry. 2014;4:e439. [PMC free article] [PubMed]
  • Baunez C, Robbins TW. Bilateral lesions of the subthalamic nucleus induce multiple deficits in an attentional task in rats. Eur J Neurosci. 1997;9:2086–2099. [PubMed]
  • Benis D, David O, Lachaux J-P, Seigneuret E, Krack P, Fraix V, Chabardès S, Bastin J. Subthalamic nucleus activity dissociates proactive and reactive inhibition in patients with Parkinson’s disease. NeuroImage. 2014;91:273–281. [PubMed]
  • Bogacz R, Gurney K. The basal ganglia and cortex implement optimal decision making between alternative actions. Neural Comput. 2007;19:442–477. [PubMed]
  • Botvinick MM, Cohen JD, Carter CS. Conflict monitoring and anterior cingulate cortex: an update. Trends Cogn Sci. 2004;8:539–546. [PubMed]
  • Brittain J-S, Watkins KE, Joundi RA, Ray NJ, Holland P, Green AL, Aziz TZ, Jenkinson N. A role for the subthalamic nucleus in response inhibition during conflict. J Neurosci. 2012;32:13396–13401. [PMC free article] [PubMed]
  • Burbaud P, Clair A-H, Langbour N, Fernandez-Vidal S, Goillandeau M, Michelet T, Bardinet E, Chéreau I, Durif F, Polosan M, et al. Neuronal activity correlated with checking behaviour in the subthalamic nucleus of patients with obsessive–compulsive disorder. Brain. 2013;136:304–317. [PubMed]
  • Carpenter RHS, Williams MLL. Neural computation of log likelihood in control of saccadic eye movements. Nature. 1995;377:59–62. [PubMed]
  • Cavanagh JF, Cohen MX, Allen JJ. Prelude to and resolution of an error: EEG phase synchrony reveals cognitive control dynamics during action monitoring. J Neurosci. 2009;29:98–105. [PMC free article] [PubMed]
  • Cavanagh JF, Wiecki TV, Cohen MX, Figueroa CM, Samanta J, Sherman SJ, Frank MJ. Subthalamic nucleus stimulation reverses mediofrontal influence over decision threshold. Nat Neurosci. 2011;14:1462–1467. [PMC free article] [PubMed]
  • Cavanagh JF, Zambrano-Vazquez L, Allen JJB. Theta lingua franca: a common mid-frontal substrate for action monitoring processes. Psychophysiology. 2012;49:220–238. [PMC free article] [PubMed]
  • Cohen MX, Cavanagh JF. Single-trial regression elucidates the role of prefrontal theta oscillations in response conflict. Front Percept Sci. 2011;2:30. [PMC free article] [PubMed]
  • Cohen MX, van Gaal S. Subthreshold muscle twitches dissociate oscillatory neural signatures of conflicts from errors. NeuroImage. 2014;86:503–513. [PubMed]
  • Coulthard EJ, Bogacz R, Javed S, Mooney LK, Murphy G, Keeley S, Whone AL. Distinct roles of dopamine and subthalamic nucleus in learning and probabilistic decision making. Brain. 2012;135:3721–3734. [PMC free article] [PubMed]
  • Courtemanche R, Fujii N, Graybiel AM. Synchronous, focally modulated β-band oscillations characterize local field potential activity in the striatum of awake behaving monkeys. J Neurosci. 2003;23:11741–11752. [PubMed]
  • DeLong MR. Primate models of movement disorders of basal ganglia origin. Trends Neurosci. 1990;13:281–285. [PubMed]
  • Deuschl G, Krack P, Lauk M, Timmer J. Clinical neurophysiology of tremor. J Clin Neurophysiol Off Publ Am Electroencephalogr Soc. 1996;13:110–121. [PubMed]
  • Eriksen BA, Eriksen CW. Effects of noise letters upon the identification of a target letter in a nonsearch task. Percept Psychophys. 1974;16:143–149.
  • Frank MJ. Hold your horses: a dynamic computational role for the subthalamic nucleus in decision making. Neural Netw. 2006;19:1120–1136. [PubMed]
  • Frank MJ, Samanta J, Moustafa AA, Sherman SJ. Hold your horses: impulsivity, deep brain stimulation, and medication in parkinsonism. Science. 2007;318:1309–1312. [PubMed]
  • Fumagalli M, Giannicola G, Rosa M, Marceglia S, Lucchiari C, Mrakic-Sposta S, Servello D, Pacchetti C, Porta M, Sassi M, et al. Conflict-dependent dynamic of subthalamic nucleus oscillations during moral decisions. Soc Neurosci. 2011;6:243–256. [PubMed]
  • Geller AS, Schlefer IK, Sederberg PB, Jacobs J, Kahana MJ. PyEPL: a cross-platform experiment-programming library. Behav Res Methods. 2007;39:950–958. [PMC free article] [PubMed]
  • Goldberg JH, Farries MA, Fee MS. Basal ganglia output to the thalamus: still a paradox. Trends Neurosci. 2013;36:695–705. [PMC free article] [PubMed]
  • Gratton G, Coles MGH, Donchin E. Optimizing the use of information: strategic control of activation of responses. J Exp Psychol Gen. 1992;121:480–506. [PubMed]
  • Green N, Bogacz R, Huebl J, Beyer A-K, Kühn AA, Heekeren HR. Reduction of influence of task difficulty on perceptual decision making by stn deep brain stimulation. Curr Biol. 2013;23:1681–1684. [PubMed]
  • Hälbig TD, Tse W, Frisina PG, Hollander E, Shapiro H, Tagliati M, Koller WC, Olanow CW. Subthalamic deep brain stimulation and impulse control in Parkinson’s disease. Eur J Neurol. 2009;16:493–497. [PubMed]
  • Hammond C, Bergman H, Brown P. Pathological synchronization in Parkinson’s disease: networks, models and treatments. Trends Neurosci. 2007;30:357–364. [PubMed]
  • Hanslmayr S, Pastötter B, Bäuml K-H, Gruber S, Wimber M, Klimesch W. The electrophysiological dynamics of interference during the stroop task. J Cogn Neurosci. 2007;20:215–225. [PubMed]
  • Herz DM, Christensen MS, Bruggemann N, Hulme OJ, Ridderinkhof KR, Madsen KH, Siebner HR. Motivational tuning of fronto-subthalamic connectivity facilitates control of action impulses. J Neurosci. 2014;34:3210–3217. [PubMed]
  • Isoda M, Hikosaka O. Role for subthalamic nucleus neurons in switching from automatic to controlled eye movement. J Neurosci. 2008;28:7209–7218. [PMC free article] [PubMed]
  • Kühn AA, Williams D, Kupsch A, Limousin P, Hariz M, Schneider G-H, Yarrow K, Brown P. Event-related beta desynchronization in human subthalamic nucleus correlates with motor performance. Brain. 2004;127:735–746. [PubMed]
  • Lardeux S, Pernaud R, Paleressompoulle D, Baunez C. Beyond the reward pathway: coding reward magnitude and error in the rat subthalamic nucleus. J Neurophysiol. 2009;102:2526–2537. [PubMed]
  • Leventhal DK, Gage GJ, Schmidt R, Pettibone JR, Case AC, Berke JD. Basal ganglia beta oscillations accompany cue utilization. Neuron. 2012;73:523–536. [PMC free article] [PubMed]
  • Levy R, Ashby P, Hutchison WD, Lang AE, Lozano AM, Dostrovsky JO. Dependence of subthalamic nucleus oscillations on movement and dopamine in Parkinson’s disease. Brain. 2002;125:1196–1209. [PubMed]
  • Levy R, Hutchison WD, Lozano AM, Dostrovsky JO. High-frequency synchronization of neuronal activity in the subthalamic nucleus of Parkinsonian patients with limb tremor. J Neurosci. 2000;20:7766–7775. [PubMed]
  • Lipszyc J, Schachar R. Inhibitory control and psychopathology: a meta-analysis of studies using the stop signal task. J Int Neuropsychol Soc. 2010;16:1064–1076. [PubMed]
  • Magariños-Ascone CM, Figueiras-Mendez R, Riva-Meana C, Córdoba-Fernández A. Subthalamic neuron activity related to tremor and movement in Parkinson’s disease. Eur J Neurosci. 2000;12:2597–2607. [PubMed]
  • Mallet L, Schüpbach M, N’Diaye K, Remy P, Bardinet E, Czernecki V, Welter M-L, Pelissolo A, Ruberg M, Agid Y, et al. Stimulation of subterritories of the subthalamic nucleus reveals its role in the integration of the emotional and motor aspects of behavior. Proc Natl Acad Sci. 2007;104:10661–10666. [PubMed]
  • Maris E, Oostenveld R. Nonparametric statistical testing of EEG- and MEG-data. J Neurosci Methods. 2007;164:177–190. [PubMed]
  • Mink JW. The basal ganglia: focused selection and inhibition of competing motor programs. Prog Neurobiol. 1996;50:381–425. [PubMed]
  • Mirabella G, Iaconelli S, Romanelli P, Modugno N, Lena F, Manfredi M, Cantore G. Deep Brain stimulation of subthalamic nuclei affects arm response inhibition in Parkinson’s patients. Cereb Cortex. 2012;22:1124–1132. [PubMed]
  • Nambu A, Tokuno H, Takada M. Functional significance of the cortico–subthalamo–pallidal “hyperdirect” pathway. Neurosci Res. 2002;43:111–117. [PubMed]
  • Narayanan NS, Laubach M. Delay activity in rodent frontal cortex during a simple reaction time task. J Neurophysiol. 2009;101:2859–2871. [PubMed]
  • Obeso I, Wilkinson L, Casabona E, Speekenbrink M, Bringas ML, Álvarez M, Álvarez L, Pavón N, Rodríguez-Oroz MC, Macías R, et al. The subthalamic nucleus and inhibitory control: impact of subthalamotomy in Parkinson’s disease. Brain. 2014;137:1470–1480. [PubMed]
  • Oswal A, Litvak V, Sauleau P, Brown P. Beta reactivity, prospective facilitation of executive processing, and its dependence on dopaminergic therapy in Parkinson’s disease. J Neurosci. 2012;32:9909–9916. [PMC free article] [PubMed]
  • Ray NJ, Brittain J-S, Holland P, Joundi RA, Stein JF, Aziz TZ, Jenkinson N. The role of the subthalamic nucleus in response inhibition: evidence from local field potential recordings in the human subthalamic nucleus. NeuroImage. 2012;60:271–278. [PubMed]
  • Rodriguez-Oroz MC, López-Azcárate J, Garcia-Garcia D, Alegre M, Toledo J, Valencia M, Guridi J, Artieda J, Obeso JA. Involvement of the subthalamic nucleus in impulse control disorders associated with Parkinson’s disease. Brain. 2011;134:36–49. [PubMed]
  • Rodriguez-Oroz MC, Rodriguez M, Guridi J, Mewes K, Chockkman V, Vitek J, DeLong MR, Obeso JA. The subthalamic nucleus in Parkinson’s disease: somatotopic organization and physiological characteristics. Brain. 2001;124:1777–1790. [PubMed]
  • Rosa M, Fumagalli M, Giannicola G, Marceglia S, Lucchiari C, Servello D, Franzini A, Pacchetti C, Romito L, Albanese A, et al. Pathological gambling in Parkinson’s disease: subthalamic oscillations during economics decisions. Mov Disord. 2013;28:1644–1652. [PubMed]
  • Rousseeuw P. Silhouettes: a graphical aid to the interpretation and validation of cluster analysis. J Comput Appl Math. 1987;20:53–65.
  • Ruiz MH, Strübing F, Jabusch H-C, Altenmüller E. EEG oscillatory patterns are associated with error prediction during music performance and are altered in musician’s dystonia. NeuroImage. 2011;55:1791–1803. [PubMed]
  • Schmidt R, Leventhal DK, Mallet N, Chen F, Berke JD. Canceling actions involves a race between basal ganglia pathways. Nat Neurosci. 2013;16:1118–1124. [PMC free article] [PubMed]
  • Schroeder U, Kuehler A, Haslinger B, Erhard P, Fogel W, Tronnier VM, Lange KW, Boecker H, Ceballos-Baumann AO. Subthalamic nucleus stimulation affects striato-anterior cingulate cortex circuit in a response conflict task: a PET study. Brain. 2002;125:1995–2004. [PubMed]
  • Sharott A, Gulberti A, Zittel S, Jones AAT, Fickel U, Münchau A, Köppen JA, Gerloff C, Westphal M, Buhmann C, et al. Activity parameters of subthalamic nucleus neurons selectively predict motor symptom severity in Parkinson’s disease. J Neurosci. 2014;34:6273–6285. [PMC free article] [PubMed]
  • Swann N, Poizner H, Houser M, Gould S, Greenhouse I, Cai W, Strunk J, George J, Aron AR. Deep brain stimulation of the subthalamic nucleus alters the cortical profile of response inhibition in the beta frequency band: a scalp EEG study in Parkinson’s disease. J Neurosci. 2011;31:5721–5729. [PMC free article] [PubMed]
  • Swann N, Tandon N, Canolty R, Ellmore TM, McEvoy LK, Dreyer S, DiSano M, Aron AR. Intracranial EEG reveals a time- and frequency-specific role for the right inferior frontal gyrus and primary motor cortex in stopping initiated responses. J Neurosci. 2009;29:12675–12685. [PMC free article] [PubMed]
  • Swann NC, Cai W, Conner CR, Pieters TA, Claffey MP, George JS, Aron AR, Tandon N. Roles for the pre-supplementary motor area and the right inferior frontal gyrus in stopping action: electrophysiological responses and functional and structural connectivity. Neuroimage. 2012;59:2860–2870. [PMC free article] [PubMed]
  • van den Wildenberg WPM, van Boxtel GJM, van der Molen MW, Bosch DA, Speelman JD, Brunia CHM. Stimulation of the subthalamic region facilitates the selection and inhibition of motor responses in Parkinson’s disease. J Cogn Neurosci. 2006;18:626–636. [PubMed]
  • Vinck M, van Wingerden M, Womelsdorf T, Fries P, Pennartz CMA. The pairwise phase consistency: a bias-free measure of rhythmic neuronal synchronization. NeuroImage. 2010;51:112–122. [PubMed]
  • Weinberger M, Mahant N, Hutchison WD, Lozano AM, Moro E, Hodaie M, Lang AE, Dostrovsky JO. Beta oscillatory activity in the subthalamic nucleus and its relation to dopaminergic response in Parkinson’s disease. J Neurophysiol. 2006;96:3248–3256. [PubMed]
  • Wiecki TV, Frank MJ. A computational model of inhibitory control in frontal cortex and basal ganglia. Psychol Rev. 2013;120:329–355. [PubMed]
  • Williams ZM, Neimat JS, Cosgrove GR, Eskandar EN. Timing and direction selectivity of subthalamic and pallidal neurons in patients with Parkinson disease. Exp Brain Res. 2005;162:407–416. [PubMed]
  • Wylie SA, Ridderinkhof KR, Elias WJ, Frysinger RC, Bashore TR, Downs KE, van Wouwe NC, van den Wildenberg WPM. Subthalamic nucleus stimulation influences expression and suppression of impulsive behaviour in Parkinson’s disease. Brain. 2010;133:3611–3624. [PMC free article] [PubMed]
  • Zaghloul KA, Weidemann CT, Lega BC, Jaggi JL, Baltuch GH, Kahana MJ. Neuronal activity in the human subthalamic nucleus encodes decision conflict during action selection. J Neurosci. 2012;32:2453–2460. [PMC free article] [PubMed]
  • Zavala B, Brittain J-S, Jenkinson N, Ashkan K, Foltynie T, Limousin P, Zrinzo L, Green AL, Aziz T, Zaghloul K, et al. Subthalamic nucleus local field potential activity during the Eriksen Flanker task reveals a novel role for theta phase during conflict monitoring. J Neurosci. 2013;33:14758–14766. [PMC free article] [PubMed]
  • Zavala B, Zaghloul K, Brown P. The subthalamic nucleus, oscillations, and conflict. Mov Disord. 2015;30:328–338. [PMC free article] [PubMed]
  • Zavala BA, Tan H, Little S, Ashkan K, Hariz M, Foltynie T, Zrinzo L, Zaghloul KA, Brown P. Midline frontal cortex low-frequency activity drives subthalamic nucleus oscillations during conflict. J Neurosci. 2014;34:7322–7333. [PMC free article] [PubMed]