Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2011 May 1.
Published in final edited form as: Cogn Affect Behav Neurosci. 2010 May;10(2):279–297. doi: 10.3758/CABN.10.2.279

Moment-to-moment fluctuations in fMRI amplitude and inter-region coupling are predictive of inhibitory performance

Srikanth Padmala, Luiz Pessoa
PMCID: PMC2946362  NIHMSID: NIHMS200314  PMID: 20498351

Abstract

We investigated how moment-to-moment fluctuations in fMRI amplitude and inter-regional coupling are linked to behavioral performance during a stop-signal task. To quantify the relationship between single-trial amplitude and behavior on a trial-by-trial basis, we modeled the probability of successful inhibition as a function of response amplitude via logistic regression analysis. At the group level, significant logistic slopes were observed in, among other regions, the inferior frontal gyrus (IFG), caudate, and putamen, all bilaterally. Furthermore, we investigated how trial-by-trial fluctuations in responses in “attentional regions” covaried with fluctuations in “inhibition-related” regions. The coupling between several fronto-parietal attentional regions and the right IFG increased during successful vs. unsuccessful performance, suggesting that efficacious network interactions are important in determining behavioral outcome during the stop-signal task. In particular, the link between responses in the right IFG and behavior were moderated by moment-to-moment fluctuations in evoked responses in the left intraparietal sulcus.

Introduction

Response inhibition, the ability to suppress actions that are no longer behaviorally relevant or contextually appropriate, is a key function of the human executive control system. This function has been investigated behaviorally, with monkey physiology, and with human ERPs and fMRI by using go/no-go (Casey et al., 1997; Eimer, 1993; Kalaska and Crammond, 1995) and stop-signal (Aron et al., 2007; Boucher et al., 2007; Li et al., 2006; Logan, 1994; Logan and Cowan, 1984) tasks. Response inhibition is believed to involve “control regions” in prefrontal cortex and both lesion and fMRI studies have suggested that the inferior frontal cortex (IFC), especially on the right hemisphere, is centrally involved in this function (Aron et al., 2003; Rubia et al., 2003), a notion that is supported by transcranial magnetic stimulation (TMS) studies (Chambers et al., 2007; Chambers et al., 2006). Other studies in the literature have provided evidence for the involvement of additional brain structures in response inhibition, including the pre-supplementary motor area, superior/medial prefrontal cortex, and precentral gyrus (Chen et al., 2009; Floden and Stuss, 2006; Li et al., 2006; Nachev et al., 2007; Picton et al., 2007). In addition to cortical structures, several subcortical areas have been linked to response inhibition, including the caudate, putamen (Eagle and Robbins, 2003; Li et al., 2008b; Vink et al., 2005), and the subthalamic nucleus (Aron and Poldrack, 2006). The latter, in particular, has been suggested to be part of a “hyperdirect” pathway that includes the IFC and is critical for motor inhibition. Taken together, response inhibition appears to engage a broad constellation of cortical and subcortical sites that are recruited in order to cancel a prepotent response when inhibition is called for (Chambers et al., 2009).

Recent studies have also made the case that network interactions subserve behavioral performance during response inhibition, revealing that multiple “inhibition-related” brain regions simultaneously contribute to this type of behavior (Duann et al., 2009). More generally, successful performance during response inhibition is behaviorally challenging and depends on several processes, including perceptual processing and attention, in addition to inhibitory mechanisms per se. Consistent with this notion, a recent MEG study revealed that fluctuations of sensory processing linked to both go and stop stimuli impact inhibitory performance during a stop-signal task (Boehler et al., 2009). In a related fashion, several studies have reported the involvement of parietal regions during response inhibition (Garavan et al., 1999; Hester et al., 2009; Liddle et al., 2001), although the exact nature of their involvement remains to be determined. One possibility is that their role is attentional and, in particular, that they reflect trial-by-trial fluctuations in the allocation of “resources” (see Leber et al., 2008) that are needed for successful behavioral performance during demanding tasks. Overall, interactions between diverse brain regions involved in multiple aspects of task performance may be present during successful response inhibition.

The goals of the present investigation were two-fold. First, our aim was to characterize how evoked fMRI responses are linked to behavioral outcome during a response inhibition task on a moment-to-moment basis. Traditionally, both in human fMRI studies and in monkey physiology, two conditions are compared by testing for differences in the associated mean responses. A complementary approach is to investigate how trial-by-trial fluctuations in evoked responses are linked to behavioral performance. Although the information conveyed by these two types of analyses is related to each other, the trial-by-trial analysis offers a potentially valuable way to quantify the predictive relationship between response magnitude and observed behavior (Fig. 1A). For instance, logistic regression analysis can be used to model the probability of a dichotomous behavioral variable (e.g., detected vs. undetected stimulus, correct vs. incorrect performance) as a function of single-trial responses. Trial-based approaches have been used fruitfully in studies investigating perceptual decisions (Padmala and Pessoa, 2008; Ress et al., 2000), in addition to more cognitive tasks (Leber et al., 2008; Pessoa et al., 2002). Our objective here was to investigate its use in a demanding cognitive task so as to further describe the viability of this analysis strategy when fMRI signals are employed – which are typically viewed as offering considerably less specific information relative to spike data, for instance.

Figure 1.

Figure 1

Open in a new tab

Overall logic of trial-by-trial analysis. (A) The top row illustrates a schematic fMRI time series, with responses during successful and unsuccessful trials indicated in red and blue, respectively. The right panel shows a schematic representation of the behavior. If critical trials (stop trials, in the present case) are sufficiently spaced apart, the trial-by-trial variability in the magnitude of the evoked response can be quantitatively linked to the probability of successful performance via logistic regression analysis (bottom, right). This analysis complements more standard ones based on mean responses (bottom, left). (B) Trial-by-trial analysis can also take into consideration signals from multiple regions. For example, the link between single-trial responses in ROI2 and behavior can be evaluated as a function of the signal strength in ROI1 to assess the presence of a moderation-like pattern. In this case, the slope of the ROI2-to-behavior relationship (as in part A) would depend on the strength of the evoked responses in ROI1.

In the present study, subjects performed a stop-signal task in which they were required to withhold responding upon hearing an auditory cue among a rapid stream of visual go trials (Fig. 2). Because such stop trials were infrequent and spaced apart in time, trial-by-trial behavioral outcome, namely successful vs. unsuccessful inhibition, was linked to single-trial response amplitude. Typically, previous neuroimaging studies of the stop-signal task have employed fast event-related designs and reported that mean responses between successful and unsuccessful stop trials differ in the extent to which they engage the right IFC and various other regions implicated in response inhibition. By using trial-based analysis, we aimed to further characterize the relationship between fMRI responses and observed behavior and, in particular, to test whether parametric increases in response amplitude could be quantitatively linked to the probability of successful response inhibition. More generally, quantifying the strength of this relationship allows one to compare the predictive power across different brain regions that are engaged during response inhibition. In this manner, a more complete characterization of the extent to which specific regions contribute to successful task performance is provided. Furthermore, research studies comparing inhibitory performance between groups (e.g., ADHD vs. controls) or between conditions within the same participants (e.g., reward vs. no-reward (Padmala and Pessoa, 2009)), can employ trial-based analysis to quantify potential changes in the relationship between moment-to-moment fluctuations in fMRI responses and behavior – which can be done both at the group and individual levels.

Figure 2.

Figure 2

Open in a new tab

Stop-signal task paradigm. During go trials, subjects responded to the go signal (circle or square?), whereas during stop trials, they were instructed to withhold their motor response (signaled by an auditory cue). The stop signal followed the go stimulus after a variable-length delay, the stop signal delay (SSD), which was updated based on a staircase procedure that maintained behavioral performance at approximately 50% correct.

A second goal of the present study was to investigate “functional interactions” between brain regions during response inhibition (Duann et al., 2009). We reasoned that because successful performance during response inhibition depends on the effective allocation of attention, the relationship (or “coupling”) between responses in attentional regions in frontoparietal cortex (Corbetta and Shulman, 2002; Kastner and Ungerleider, 2001) and responses in regions more directly implicated in response inhibition (such as the right IFC) would vary as a function of behavioral performance, namely successful vs. unsuccessful inhibition. We therefore sought to investigate inter-region signal relationships and their link to behavior on a trial-by-trial basis. For instance, if responses in the right IFC are predictive of successful performance, how is this relationship dependent on, namely moderated by, fluctuations in responses in brain regions important for attention? In general, because network interactions are believed to subserve behavioral performance, developing trial-by-trial analyses to include signals from multiple regions will, we contend, prove to be an invaluable tool in characterizing the brain bases of behavior (Fig. 1B).

Methods

Subjects

Thirty-five volunteers (22 ± 3 years old; 19 females) participated in the study, which was approved by the Institutional Review Board of Indiana University, Bloomington. All subjects were in good health with no past history of psychiatric or neurological disease and had normal or corrected-to-normal vision. All participants gave informed written consent. One participant's data were removed from the analysis because of unusually poor performance (70% correct on go trials).

Stimuli and behavioral task

We employed a stop-signal task to investigate the neural correlates of response inhibition (Fig. 2). We used a simple choice-reaction time task, which included both go and stop trials. Each go trial started with the presentation of a simple shape-stimulus and participants were asked to indicate “circle” or “square” via button-press on an MR-compatible response box by using the index or middle finger of their right hand. Participants were instructed to respond as soon as possible during the presentation of the shape stimulus (trials with reaction time longer than 1 sec were treated as incorrect trials). Following the visual stimulus, subjects viewed a blank screen for 1000 ms. Stop trials were identical to go trials, except that a brief tone (300 ms) was played after a variable stop-signal delay (SSD) relative to the onset of the go stimulus, which indicated that participants should withhold their response (the initial value of the SSD was set to 250 ms). The SSD was adjusted dynamically throughout the experiment, such that if the participant successfully inhibited their response on a stop trial, the SSD was increased by 50 ms on a subsequent stop trial, and if the participant failed to inhibit their response, the SSD was reduced by 50 ms on a subsequent stop trial (Logan et al., 1997; Rubia et al., 2003). This staircasing procedure ensured that participants were successfully inhibiting their response on approximately 50% of the stop trials. Participants were instructed to respond as quickly and accurately as possible and were asked to inhibit their response upon hearing a tone that followed the initial shape stimulus. They were also told that sometimes it might not be possible to successfully inhibit their response and that, in such cases, they should simply continue performing the task. Overall, the importance of going and stopping was stressed equally. Participants performed a short practice run (approximately four minutes) during the initial anatomical scan (see below) to familiarize themselves with the task.

Each participant performed four runs. Each run contained a total of 150 trials, out of which there were 126 (84%) go trials and 24 (16%) stop trials. Trial order was randomized but fixed across participants; no stop trials occurred during the last six trials of each run. Go and stop trials contained circle and square shape stimuli in equal proportion. The data from a second condition (presented in separate experimental runs) involving reward were not analyzed here and will be discussed elsewhere. The runs from this second condition were identical in length to runs reported here, and alternated with the main ones (the order was counterbalanced across participants).

Because standard slow event-related designs would be problematic in the context of fast-paced stop-signal tasks and given that the primary interest of this study was to perform trial-by-trial analysis, approximately 75% of stop trials (i.e., 72 out of 96 trials) were spaced apart at least 10 sec from other stop trials (with a minimum of 5 go trials in between) to allow for the estimation of single-trial responses (considering go trials as baseline condition; see below). The remaining stop trials were presented in close temporal succession so as to prevent participants' expectancies from developing, but were not included in the trial-by-trial analysis.

MR data acquisition

MR data were collected using a 3 Tesla Siemens TRIO scanner (Siemens Medical Systems, Erlangen, Germany). Each scanning session began with a high-resolution MPRAGE anatomical scan (TR = 1900 ms, TE = 4.15 ms, TI = 1100 ms, 1 mm isotropic voxels, 256 mm field of view). Subsequently, in each functional run, 153 volumes were acquired with a TR of 2000 and TE of 25 ms and consisted of 34 axial-slices with a thickness of 3.8 mm and an in-plane resolution of 3.8 × 3.8 mm (240 mm field of view).

Behavioral data analysis

As stated above, the SSD was adjusted dynamically to yield an inhibition success rate of approximately 50%. The stop-signal reaction time (SSRT), which provides an estimate of the “inhibitory reaction time”, was calculated by subtracting the average SSD from the median RT during correct go trials, following the race model (Logan and Cowan, 1984).

General fMRI data analysis

Pre-processing of the data was done using tools from the AFNI software package (Cox, 1996) (http://afni.nimh.nih.gov/afni). The first 3 volumes of each functional run were discarded to account for equilibration effects. The remaining volumes were slice-time corrected and spatially registered for motion correction to the volume acquired closest in time to the particular subject's high-resolution anatomy. The functional data were then normalized to Talairach space (Talairach and Tournoux, 1988) and spatially smoothed using a Gaussian filter with a full-width at half maximum of 7.6 mm (i.e., two times the voxel dimension). Finally, the signal intensity of each voxel was scaled to a mean of 100.

Voxelwise analysis

Voxelwise analyses were run to determine regions of interest (see below). Each participant's fMRI data were analyzed using standard multiple linear regression. A linear model was defined for each participant that included three regressors corresponding to the three main event types, namely successful stop trials (SUCC), unsuccessful stop trials (UNSUCC), and an event type that included all incorrect go trials (INC). All regressors were convolved with a canonical hemodynamic response function (Cohen, 1997) to account for the low-pass properties of fMRI responses. Constant, linear, and quadratic terms were included for each run separately (as covariates of no interest) to model drifts of the MR signal. Correct go trials were not modeled explicitly and constituted the implicit baseline in the model. This type of baseline condition has been used successfully in several fMRI studies of the stop-signal task (Chamberlain et al., 2009; Rubia et al., 2003; Rubia et al., 2007). In this manner, all parameter estimates reported in this study should be interpreted relative to the responses evoked by correct go trials.

Region of interest analysis

To maximize statistical power, we focused our analysis on a set of regions of interest (ROIs) that have been reported consistently in the previous response inhibition literature (as listed in Table 2); note in particular that regions such as the posterior cingulate, anterior cingulate, and anterior insula have been implicated in, among others, error-related processing during these tasks. Specifically, the precise location of these ROIs was determined based on general task-related activation, formally defined by the contrast vector cselection = [1 1 0]T (where T denotes the transpose operation), corresponding to the conditions SUCC, UNSUCC, and INC, respectively, at a p value of .005, corrected for multiple comparisons according to a false discovery rate procedure (Genovese et al., 2002). Individual ROIs were drawn using a sphere of 5-mm radius centered at the peak voxel of each cluster (defined at the group level). For each individual, a representative time series for the ROI was then defined by averaging across all of the voxels. Before doing so, the variance explained by incorrect go trials was removed from a voxel's time series (slow-varying drifts in MR signal were likewise removed). A convenient way to implement this procedure is available via the 3dSynthesize program in AFNI. When our analyses were repeated without removing the variance related to incorrect go trials (which were fewer than 2.5% overall), nearly identical results were observed.

Table 2.

ROI based trial-by-trial logistic analysis linking fMRI signals to behavior (peak Talairach coordinates are provided for each ROI)

Location X y z fMRI slope fMRI × SSD slope *UNSUCC (t)
SUCC > UNSUCC
Parietal
Intraparietal sulcus R 28 −58 42 0.33 (p < .05) −2.45 (p = . 20) 3.47 (p < .005)
Inferior Parietal lobule R 36 −37 44 0.89 (p < .005) 0.2 (p = .94) 1.08 (p = .29)
Frontal
Inferior Frontal gyrus R 48 26 20 0.23 (p < .05) −2.14 (p = .16) 1.1 (p = .28)
L −48 26 14 0.28 (p < .05) −1.74 (p = .14) −3.68 (p < .005)
Superior Frontal gyrus L −19 19 49 0.46 (p < .005) 2.18 (p = .14) −4.64 (p < .001)
Precentral gyrus L −34 −11 45 0.40 (p < .05) −1.01 (p = .71) 4.91 (p < .001)
Supplementary Motor area R/L 0 −4 52 −0.06 (p = .65) −2.51 (p = .11) 7.37 (p < .001)
Subcortical
Putamen R 21 4 4 0.76 (p < .005) −0.5 (p = .80) 2.51 (p < .010)
L −19 4 4 0.68 (p < .005) −0.09 (p = .96) 2.03 (p < .050)
Caudate R 11 4 9 0.24 (p < .05) −0.36 (p = .85) 4.62 (p < .001)
L −10 4 8 0.29 (p < .05) −0.12 (p = .94) 4.22 (p < .001)
Subthalamic nucleus R 10 −14 −3 0.04 (p = .80) −0.03 (p = .99) 3.46 (p < .005)
UNSUCC > SUCC
Posterior Cingulate R/L 0 −25 28 −0.37 (p < .005) −1.84 (p = .26) 8.46 (p < .001)
Anterior Cingulate R/L 0 20 28 −0.42 (p < .005) −3.24 (p < .05) 8.64 (p < .001)
Anterior Insula R 35 14 6 −0.23 (p = .22) −5.37 (p < .01) 13.88 (p < .001)
L −29 17 7 −0.2 (p = .23) −5.63 (p < .01) 13.96 (p < .001)
Open in a new tab

SUCC: successful inhibition; UNSUCC: unsuccessful inhibition.

*

The last column displays the standard contrast between unsuccessful trials and baseline (as suggested by a reviewer).

The above selection criterion was employed because it was orthogonal to the main contrast of interest in our study, namely the contrast of successful and unsuccessful trials, specifically ctest = [1 −1 0]T (note that cselectionT ctest = 0). In this manner, the selection and test criteria were independent, avoiding potential biases in the ROI analysis. In addition, the staircasing procedure guaranteed that an approximately equal number of trials were obtained for each trial type of interest (Kriegeskorte et al., 2009).

We further simulated the effect of the ROI selection procedure by generating 500,000 Gaussian random samples (M = 0, SD = 1) with 36 trials per condition (the mean number of “spaced apart” trials per condition observed during the fMRI experiment). For every sample, we contrasted SUCC and UNSUCC trials via logistic regression analysis to generate the distribution of logistic regression slopes based on this “noise” distribution. These results were then compared to the distribution of slopes obtained when our selection criterion was first adopted (i.e., when the selection contrast was statistically significant). No selection bias was observed, as evidenced by the complete overlap of the two distributions (Supplementary Fig 1).

For illustration of average evoked responses for SUCC and UNSUCC conditions, for each ROI (see Figs. 3–6), selective averaging was performed using responses to stop trials that were spaced at least 10 sec apart, as in trial-based analysis (see below).

Figure 3.

Figure 3

Open in a new tab

Trial-by-trial responses during response inhibition. (A, E) The probability of successful inhibition was modeled as a function of single-trial fMRI magnitude. Representative individual fits are shown for the right inferior frontal gyrus (IFG) and left precentral gyrus regions. Data points on the logistic plots represent the proportion of SUCC trials in each bin of the fMRI amplitude. (B, F) Logistic slope values for all participants. (C, G) Logistic regression fit at the group level. Data points on the logistic plots represent the proportion of SUCC trials in each bin of the fMRI amplitude. (D, H) Average evoked responses as a function of trial type. Error bars indicate standard within-subject errors (Loftus and Masson, 1994). SUCC: successful stop trial; UNSUCC: unsuccessful stop trial.

Figure 6.

Figure 6

Open in a new tab

Trial-by-trial responses during response inhibition. (A, E) The probability of successful inhibition was modeled as a function of single-trial fMRI magnitude. Representative individual fits are shown for the anterior cingulate gyrus (ACC) and posterior cingulate gyrus (PCC) regions. Unlike in Figures 3–5, the probability of a failure increased with fMRI signal amplitude. (B, F) Logistic slope values for all participants. (C, G) Logistic regression fit at the group level. (D, H) Average evoked responses as a function of trial type. Error bars indicate standard within-subject errors (Loftus and Masson, 1994). SUCC: successful stop trial; UNSUCC: unsuccessful stop trial. Data points on the logistic plots in parts A, E, C, and G represent the proportion of SUCC trials in each bin based on fMRI response magnitude.

Trial-based analysis

To quantify the link between fMRI amplitude and behavior at the individual level, trial-by-trial fMRI analysis was performed on the time series from each ROI, for each individual. We modeled the probability of success during a stop trial by performing standard logistic regression analysis (Hosmer and Lemeshow, 2000) based on single-trial fMRI response amplitude. Because the SSD varied for each trial due to the staircasing procedure, these values were also entered into our model, together with the interaction between SSD and fMRI amplitude (to avoid strong correlations between the interaction and remaining terms, the fMRI and SSD variables were initially mean corrected). Thus, the probability of success was modeled by

Pr(yi=1)=logit−1(b0+b1fMRIi+b2SSDi+b3(SSDi×fMRIi)),

where y is the behavioral outcome (success:1, failure: 0), the function logit−1 transforms continuous values to the range (0,1), which is necessary for probabilities (note that this is simply a logistic “sigmoidal” curve), fMRI is the response amplitude, SSD is the staircasing delay, and i is a trial index. For plotting purposes (see Figs. 3–6), only the influence of fMRI responses is illustrated (thus keeping our plot two-dimensional). In the plots of individual data fits (Figs. 3–6, parts A and E), data were binned into 14 bins based on fMRI amplitude such that 5 trials were present in each bin (except the last one which had 7 trials); data points represent the proportion of SUCC trials in each bin. A similar strategy was used for group fits (Figs. 3–6, parts C and G), but in this case data were pooled across subjects and 24 bins were used with approximately 100 trials in each bin (see below, for the methods employed in the group fits). Finally, note that whereas a linear regression fit to the data is conceivable (see, e.g., Fig. 3C), a logistic-type fit is required given the binary nature of the data (success/failure).

Response strength for each trial was indexed by considering activations around the expected peak time, specifically the time points at 4, 6, and 8 sec following trial onset (only trials sufficiently spaced apart were considered). Because responses were somewhat variable in their timing (see Figs. 3–6), the peak response during the 4–8 sec window was averaged with the second-largest response in the same window (only consecutive points were averaged). For instance, if the peak occurred at 6 sec and the second largest response was at 4 sec, responses at 4 and 6 sec were averaged; if the second largest response was at 8 sec, responses at 6 and 8 sec were averaged. Critically, very similar results were obtained when only the peak response between 4–8 sec was used. For regions that exhibited deactivations relative to baseline (e.g., left superior frontal gyrus), we averaged the two consecutive responses with the largest responses in absolute value (i.e., response minima).

To assess the link between single-trial amplitude and behavioral performance at the group level, multilevel logistic regression analysis (Gelman and Hill, 2007) was performed via the method of generalized estimating equations (GEE) (Zeger and Liang, 1986). GEE is a powerful statistical technique that is often used with longitudinal and correlated data, especially when the data are binary. GEE combines the information from all of the participants and determines an overall population-level logistic function taking into account the correlated structure of the data within participants. As above, we modeled the probability of success during a stop trial as a function of response amplitude, SSD, and the interaction term. The slope of the logistic fit corresponding to response amplitude indicates the strength of the association between fMRI responses and behavior at a single-trial level (at the group level). To implement GEE, the GEEQBOX toolbox (Ratcliffe and Shults, 2008) was used in Matlab (Mathworks, Natick, MA).

Calculation of average predictive comparisons based on fMRI amplitude

Logistic regressions are nonlinear on the probability scale such that a constant fMRI signal difference does not correspond to a fixed change in probability – the gain will be greater at the “middle” portion of the sigmoidal curve where the slope is steepest. To provide a measure of the increase in probability of success as a function of fMRI signal changes, for each individual, we compared probabilities when activation changed from “medium” to “high” levels, where “medium” corresponded to the amplitude when Pr(y = success) = 0.5 and “high” corresponded the mean response amplitude of the trials with the 10% highest responses. We defined “high” in this manner to minimize the effects of unusually high single-trial responses if the largest single trial (i.e., the max) were picked. Determining the predictive difference in probabilities between these two cases (“medium” and “high”) is admittedly arbitrary but provides an effective way to summarize how changes in fMRI responses affect the ability to predict successful performance (Gelman and Hill, 2007). A complementary summary strategy is provided by the slope of the logistic fit at the group level, as evaluated via GEE (see above), a strategy that takes into account all trials (from all participants). However, because the logistic fit represents a relationship that is nonlinear, the slope value is informative mostly about the region of greatest probability change (e.g., similar changes in fMRI amplitude at “low” values of response amplitude do not change the probability of behavior by much).

More precisely, the average (across trials) predictive difference in probability for a given individual was given by 1/n ΣΔi, where

Δi=logit−1[b0+b1high+b2SSDi+b3(SSDi×high)]−logit−1[b0+b1medium+b2SSDi+b3(SSDi×medium)],

n is the number of trials and i is a trial index (all trials must be considered because SSD varied as a function of trial; for further details, see Gelman and Hill, 2007). This predictive difference in probability was computed for every participant and then averaged across subjects to provide a final value.

Functional connectivity analysis

We investigated functional interactions between attention- and inhibition-related regions. We use “attention-related” and “inhibition-related” as convenient shorthand terms, without implying that the associated regions are exclusively linked to these functions – in particular, the right IFC has been implicated in several executive functions (Brass et al., 2005), including more reactive/exogenous attentional effects (Corbetta and Shulman, 2002; Pessoa and Ungerleider, 2004). Based on the regions existing literature, the following attentional regions were probed: intraparietal sulcus (IPS, right: x = 28, y = −58, z = 42; left: x = −29, y = −58, z = 43), inferior parietal lobule (IPL, right: x = 36, y = −37, z = 44; left: x = −34, y = −37, z = 38), and frontal eye field (FEF, right: x = 28, y = −8, z = 51; left: x = −24, y = −13, z = 54). For these regions, ROIs were defined as described previously, namely by assessing general task-related activation and by creating representative time series. Likewise, the following inhibition regions were probed: right inferior frontal gyrus (IFG), bilateral caudate, and bilateral putamen (we did not include the left IFG because it exhibited a different response pattern compared to the preceding regions; see Fig. 4D).

Figure 4.

Figure 4

Open in a new tab

Trial-by-trial responses during response inhibition. (A, E) The probability of successful inhibition was modeled as a function of single-trial fMRI magnitude. Representative individual fits are shown for the left inferior frontal gyrus (IFG) and left superior frontal gyrus regions. (B, F) Logistic slope values for all participants. (C, G) Logistic regression fit at the group level. (D, H) Average evoked responses as a function of trial type. Note that evoked responses decreased relative to the go-trial baseline. Error bars indicate standard within-subject errors (Loftus and Masson, 1994). SUCC: successful stop trial; UNSUCC: unsuccessful stop trial. Data points on the logistic plots in parts A, E, C, and G represent the proportion of SUCC trials in each bin based on fMRI response magnitude.

Initially, we investigated whether the correlations between trial-by-trial responses in inhibition-related regions were more strongly correlated with signals in fronto-parietal attentional regions during successful vs. unsuccessful trials. Prior to computing the correlations, response strengths were mean centered, such that only deviations from the mean were considered. Note also that because we employed a slow event-related design, the correlations involved a single measure of response strength per trial (i.e., correlations did not involve the entire time series; see Zhou et al., 2009. To compare the difference in correlations between SUCC and UNSUCC conditions at the group level, correlations were initially transformed (via Fisher's Z-transform) and then compared via a paired t test.

Relationship between fronto-parietal responses and the strength of the right IFG to behavior relationship

We investigated the relationship between trial-by-trial fluctuations in responses in attentional regions (IPS, IPL, and FEF) and the strength of the association between right IFG and behavior. To do so, we probed whether the trial-by-trial relationship between the right IFG and behavior depended on the magnitude of IPS, IPL, or FEF (each tested in a separate analysis). To increase statistical power, we pooled data from all participants and binned trials based on the strength of fronto-parietal activation independent of behavioral responses. For each bin, all trials were employed to determine the logistic regression slope between right IFG response strength and behavior (successful vs. unsuccessful). The regression slopes were then correlated with the median amplitude of fronto-parietal responses in each bin (see Fig. 10B for further details). A total of 22 bins were used to partition the range of fronto-parietal responses into approximately equal number of trials per bin. Note that the results were robust with respect to the specific number of bins employed when we partitioned the data into 10 to 25 bins.

Figure 10.

Figure 10

Open in a new tab

Trial-by-trial interactions between right inferior frontal gyrus (IFG) and other fronto-parietal regions and their relationship to behavior. (A) We investigated whether the relationship between the right IFG and behavior varied as a function of the strength of activations in fronto-parietal regions (e.g., intraparietal sulcus, IPS). (B) Trial-by-trial fluctuations in the left intraparietal sulcus and the magnitude of the brain-behavior relationship in the right IFG. The schematic diagram highlights the main steps in assessing this relationship. The filled circles represent the magnitude of left intraparietal sulcus (IPS, bottom row) and right inferior frontal gyrus (R IFG, middle row) responses, as indicated via the height from the horizontal bars. Response strength in the left IPS was approximately Gaussian, as indicated via the bell-shaped curve. The top row illustrates idealized logistic regression fits showing an increase in the slope value as a function of the magnitude of evoked responses in the left IPS. The red triangles indicate the positions of the median response for each left IPS response bin, which were correlated with the slope of the logistic fit for that same bin. (C) The strength of the right IFG-to-behavior relationship (as indexed via the slope of the logistic fit) was correlated with the magnitude of evoked responses in the left IPS, such that the stronger the response in left IPS, the tighter the relationship between r IFG and behavior.

Results

Behavioral results

Behavioral results are summarized in Table 1 (results are for the same set of trials used in the fMRI analysis; see Methods). Mean reaction time on correct go trials was 487 ms and mean go error rate was 2.3%. As expected, because of the staircasing procedure, the inhibition rate during stop trials was approximately 50%. The critical behavioral index of response inhibition, SSRT, was on average 205 ms, in the range previously reported in studies of the stop-signal task (Colzato et al., 2007; Williams et al., 1999). Finally, the reaction time during UNSUCC trials was faster than those of correct go trials (t(33) = 5.74, p < .001), in line with predictions of the race model (Logan and Cowan, 1984).

Table 1.

Behavioral Results

Median Go RT (ms) 487.1 ± 18.6
Inhibition Rate (%) 50.3 ± 0.7
SSD (ms) 282.4 ± 25.1
SSRT (ms) 204.7 ± 10.0
UNSUCC RT (ms) 462.7 ± 17.5
Go Error Rate (%) 2.3 ± 0.3
Open in a new tab

Abbreviations: SSD: stop-signal delay; SSRT: stop-signal reaction time; UNSUCC: unsuccessful stop trial

Functional MRI results

To maximize statistical power, trial-by-trial analyses were restricted to a set of ROIs (Table 2; see Methods). For each ROI, we investigated the link between single-trial amplitude and behavioral performance at the individual level by performing logistic regression analysis. Accordingly, fMRI response amplitude was employed to predict whether participants were successful at withholding responses during stop trials. Positive slopes indicate that the probability of successful inhibition during a stop trial increased as a function of the amplitude of the evoked response. Examples of logistic fits at the individual level are shown in Fig. 3 for the right inferior frontal gyrus (IFG; A) and left precentral gyrus (E), together with bar plots showing slope values for all participants (B, F). Group-level logistic fits are shown in the third column (C, G); note that for the range of fMRI amplitudes considered, the figure illustrates the “linear part” of the logistic function). Finally, mean evoked responses are shown in the last column (D, H).

Both the right IFG and the left precentral gyrus have been advanced as important regions for response inhibition. In this context, the results of Fig. 3 can be summarized as follows. In both of these regions, mean evoked responses were stronger during SUCC vs. UNSUCC trials. Critically, as the single-trial response amplitude increased, the probability of successfully inhibiting a motor response increased. This relationship was observed at the level of the individual for most participants (in terms of the sign of the slope) and reliably at the group level. Note that even though a small number of trials were employed per participant, the fit was significant at the level of the individual for some of them.

Positive logistic slopes at the individual and group levels were also observed in two other PFC regions, namely left IFG and left superior frontal gyrus (Fig. 4). Unexpectedly, however, mean responses exhibited a different pattern when compared to those shown in Fig. 3. Specifically, both SUCC and UNSUCC trials displayed decreased activation relative to baseline (go trials).

Positive logistic slopes were also observed in the right inferior parietal lobule (IPL) and right intraparietal sulcus (IPS) (see Table 2 for complete results). Furthermore, significant group-level positive slopes were observed in subcortical regions, including caudate and putamen (both bilaterally), as illustrated in Fig. 5.

Figure 5.

Figure 5

Open in a new tab

Trial-by-trial responses during response inhibition. (A, E) The probability of successful inhibition was modeled as a function of single-trial fMRI magnitude. Representative individual fits are shown for the right caudate and right putamen regions. (B, F) Logistic slope values for all participants. (C, G) Logistic regression fit at the group level. (D, H) Average evoked responses as a function of trial type. Error bars indicate standard within-subject errors (Loftus and Masson, 1994). SUCC: successful stop trial; UNSUCC: unsuccessful stop trial. Data points on the logistic plots in parts A, E, C, and G represent the proportion of SUCC trials in each bin based on fMRI response magnitude.

It is of interest to consider the slope values for the regions discussed in the preceding paragraphs. Overall, the slope varied more than three-fold, with the lowest values being observed in the right IFG (.23) and right caudate (.24), and the highest values being observed in the right putamen (.76) and right inferior parietal lobule (.89).

Interestingly, some regions exhibited significant group-level slopes that were negative, including the posterior cingulate cortex and anterior cingulate cortex (ACC; Fig. 6). For these regions, increases in fMRI response strength were associated with a decreased probability of successful inhibition; or conversely, increases in fMRI strength were linked with an increased probability of error (i.e., failed inhibition).

We were interested in further quantifying the strength of the relationship between fMRI magnitude and the probability of successful inhibition. For instance, for the right IFG, when evoked responses changed from “medium” to “high” (see Methods), the probability of successful inhibition increased from 50% to 55.5 ± 1.89% (t(33) = 2.88, p < 0.01). Although this figure is relatively modest, plotting the values across individuals (Fig. 7A) shows that considerable inter-subject variability was present (SD = 11%). Thus, for instance, predictive probabilities for the 10 top participants averaged 69.1%. The distribution of predictive probabilities is also shown for the right caudate (Fig. 7B) and right putamen (Fig. 7C). For the former region, the average value was 53.9 ± 1.49% (t(33) = 2.62, p < 0.05; SD = 8.7%; 64.5% for the top 10 participants) and for the latter region, the average value was 60.5 ± 1.55% (t(33) = 6.75, p < 0.001; SD = 9.0%; 70.4% for the top 10 participants).

Figure 7.

Figure 7

Open in a new tab

Predictive probability based on fMRI signals. The bar plots illustrate changes in probability for all participants in the right inferior frontal gyrus (A), caudate (B), and putamen (C). Note that probabilities were sorted in descending order for plotting purposes. Blue bars indicate the participants with predictive probability of less than 50% (i.e., increases in fMRI response decreased the probability of success), black bars indicate those with values between 50 to 60%, and red bars indicate participants with change in predictive probability greater than 60%.

In addition to evaluating the slope corresponding to the magnitude of evoked fMRI responses, we tested the remaining two explanatory variables in the logistic regression model, namely SSD and the fMRI by SSD interaction. The SSD variable was not statistically significant for any of the regions, whereas the SSD by fMRI interaction was significant in the ACC and bilateral anterior insula (Table 2). For both of these regions, the link between fMRI responses and behavior depended on the delay, such that at high SSD values the logistic slope was steeper compared to the slope at low SSD values (Fig. 8).

Figure 8.

Figure 8

Open in a new tab

Interaction patterns between fMRI amplitude and stop-signal delay (SSD). The probability of successful inhibition was plotted as a function of single-trial fMRI magnitude at two different levels of SSD values, chosen such that they were one standard deviation above and below the mean respectively. Group level fits are shown for the anterior cingulate gyrus (A) and right anterior insula (B).

The final column of Table 2 displays regions that were reliably engaged during UNSUCC trials (relative to the go baseline). The subthalamic nucleus, a region previously implicated in some response inhibition studies (Aron and Poldrack, 2006), was engaged during these trials, as well as most brain regions (likely because of the unspecific nature of the contrast between UNSUCC and go trials).

Connectivity analysis

What is the relationship between inhibition- and attention-related regions and how do they vary as a function of behavioral performance? To investigate these questions, we evaluated the correlations between responses in inhibition- and attention-related regions (see Methods for the definition of these regions). Increased correlations for successful vs. unsuccessful performance were observed between the right IFG and bilateral IPS (right IFG and left IPS: t(33) = 2.88, p < .01; right IFG and right IPS: t(33) = 3.39, p < .005) and between right IFG and bilateral FEF (right IFG and left. FEF: t(33) = 2.54, p < .05; right IFG and right FEF: t(33) = 2.55, p < .05), as illustrated at the group (Fig. 9A) and individual level (Fig. 9B–C); significant differential correlations were not observed between other pairs of regions (e.g., IPS and caudate).

Figure 9.

Figure 9

Open in a new tab

Correlations between right inferior frontal gyrus (IFG) and other fronto-parietal regions. (A) At the group level, correlations between the right IFG and intraparietal sulcus (IPS) and between the right IFG and frontal eye field (FEF) increased during successful vs. unsuccessful inhibition. (B, C) These correlations are further illustrated for the left IPS and right FEF, respectively, for representative individuals. SUCC: successful stop trial; UNSUCC: unsuccessful stop trial.

We further investigated the relationship between inter-regional evoked responses and behavioral performance in the following way. We reasoned that if fluctuations in responses in fronto-parietal regions are important for determining behavior, the strength of the predictive effect between inhibition-related regions and behavior should depend on the strength of these very fronto-parietal signals. In particular, we were interested in assessing how the slope of the logistic regression slope between the right IFG and behavior varied as a function of fronto-parietal responses (Fig. 10A). To do so, trials across participants were pooled together (to increase statistical power) and binned according to response magnitude independent of behavior (see Methods). The same trials were then used and trial-by-trial logistic regressions involving the right IFG and behavior were determined for every bin, as illustrated in Fig. 10B. Fig. 10C displays the results, which revealed that as response strength in the left IPS increased, the slope of the logistic regression was steeper (r(22) = .44, P < .05) – i.e., the link between brain responses in right IFG and behavior was tighter. This type of relationship of the right IFG to behavior was not observed when other fronto-parietal regions were considered (i.e., right IPS, bilateral IPL, and bilateral FEF; r's < .3). Furthermore, when the caudate or putamen were evaluated (in the place of the right IFG in the analysis of Fig. 10A), no significant results were detected (r's < .1). Finally, when the right IFG and the left IPS were swapped around in the analysis shown in Fig. 10A, no significant relationship was observed (r(22) = .11, P = .62), suggesting that the role of these two regions was not completely interchangeable in our task.

Discussion

In this study, we investigated the link between single-trial fMRI responses and behavioral performance during response inhibition. To do so, participants performed a stop-signal task with stop trials sufficiently spaced apart so as to allow the estimation of single-trial responses. The quantitative assessment of brain-behavior relationships revealed that several cortical and subcortical regions previously implicated in response inhibition parametrically predicted stopping performance on a moment-to-moment basis. Methodologically, our study revealed that the relationship between trial-by-trial responses and behavior could be detected under experimental conditions similar to those employed to investigate differences in mean responses. In particular, participants were scanned during only one session, unlike some other experiments in which multi-session data have been utilized (Lim et al., 2009; Padmala and Pessoa, 2008; Ress et al., 2000; Sylvester et al., 2007). Our results thus suggest that cognitive tasks are amenable to trial-by-trial analyses with fMRI (Leber et al., 2008; Pessoa et al., 2002).

The stop-signal task is a popular behavioral paradigm used to study response inhibition (Verbruggen and Logan, 2008). It allows for the estimation of the latency of the stop process, namely the “inhibitory reaction time” (i.e., SSRT), which is by definition, unobserved. In the present study, because we were interested in the variability in trial-by-trial responses, we did not employ a standard fast-event related design, but instead spaced stop trials sufficiently apart such that the hemodynamic responses to individual stop trials did not overlap each other. Although this strategy did not allow us to estimate responses to go trials (which comprised the baseline condition), the task structure may be better suited to study response inhibition than other fast event-related paradigms that include interspersed null events lasting, say, 2–6 sec, which may induce a more controlled mode of responding. In addition, the strategy of treating go trials as a baseline has been successfully used in the past with both go/no-go (Chikazoe et al., 2009; Garavan et al., 1999; Hester et al., 2009) and stop-signal studies (Chamberlain et al., 2009; Rubia et al., 2003; Rubia et al., 2007).

Previous neuroimaging investigations have adopted two alternative strategies of indexing response inhibition, specifically by either contrasting SUCC vs. go trials (Aron and Poldrack, 2006; Xue et al., 2008) or by contrasting SUCC vs. UNSUCC trials (Li et al., 2006; Rubia et al., 2003). The approach of the present study was to probe response inhibition on a trial-by-trial basis by employing logistic regression analysis in terms of a dichotomous behavioral variable (successful, unsuccessful), an approach qualitatively similar to the latter strategy above. The rationale for contrasting SUCC vs. go trials is, at times, that the outcome on a stop trial depends mainly on the speed of the go process; e.g., (Aron and Poldrack, 2006; Leung and Cai, 2007). Another reason for contrasting SUCC vs. go trials is that SUCC stop trials engage both go and stop processes, such that subtracting the go component would provide an indication of inhibition-related activity (Aron and Poldrack, 2006; Leung and Cai, 2007). In this case, however, the inhibition-related response is relatively unspecific as it includes contributions due to various other cognitive processes recruited during the stop-signal task, including “oddball” effects (because of infrequent stop trials), decision making, response selection, and conflict monitoring (Rubia et al., 2003).

The right IFG is believed to be an important brain region for response inhibition, as evidenced by lesion, fMRI, and TMS studies (Aron et al., 2003; Chambers et al., 2007; Chambers et al., 2006; Forstmann et al., 2008; Rubia et al., 2003). Our quantitative brain-behavior analysis is consistent with this notion. Although the magnitude of the relationship was modest at the group level, it was sizeable for several participants. Specifically, at the group level, the probability of successful inhibition increased to 55.5% for trials with “high” activation (i.e., top 10% percentile) relative to those trials for which subjects were at 50% performance. At the individual level, for the 10 participants with the strongest brain-behavior relationships, the same probability increased to 69.1% on average. It is noteworthy that both group and individual values are comparable to those obtained with neuronal data during perceptual decision tasks with monkeys (Britten et al., 1996; Purushothaman and Bradley, 2005), as well as other fMRI studies of decision making (Pessoa and Padmala, 2005). Somewhat surprisingly, the value of the slope of the fMRI signal in the right IFG was among the lowest observed. In the literature, there is still some dispute whether differences between SUCC and UNSUCC trials are present in the right IFG (Aron and Poldrack, 2006; Rubia et al., 2003). The relatively shallow, yet significant, logistic fit observed in our study characterizes how moment-to-moment changes in fMRI responses are linked to changes in behavior and is consistent with the previous studies that have reported differences between SUCC and UNSUCC trials in right IFG based on mean responses (Rubia et al., 2003). Finally, it is important to consider that the slope of the logistic fit reflects the collective contributions of evoked responses to successful inhibitory performance. And because successful behavior in the task relies on several types of processes, including those more closely linked to attention (see below), all of those contributions will influence the steepness of the slope.

Significant positive slopes were also observed in the left precentral gyrus, bilateral caudate and bilateral putamen. The left precentral gyrus has been implicated in response inhibition (Li et al., 2006) and is also involved in stimulus-response associations (Brass et al., 2009). Converging evidence also indicates the importance of the striatum in response inhibition. For instance, in a lesion study using the stop-signal task, lesions of the rat striatum increased the SSRT by 60% (Eagle and Robbins, 2003), and a recent stop-signal task study reported the involvement of the caudate during inhibition (Li et al., 2008b). Interestingly, hypoactivation in this structure has been documented in clinical populations with inhibitory deficits (Booth et al., 2005). Our results revealed that moment-to-moment fluctuations in response amplitude in these structures were quantitatively linked to the probability of successful inhibition during the stop-signal task.

We did not observe evidence for an association between fMRI responses and behavioral performance in the subthalamic nucleus, a subcortical region that has been recently implicated in response inhibition in humans (Aron and Poldrack, 2006; Ray et al., 2009). In our participants, this region was equally engaged during SUCC and UNSUCC trials. A possible reason for this negative result is related to the finding that inhibition-related regions may be engaged robustly during unsuccessful trials, too (see Table 2, last column) – but too late to countermand the go response (Garavan et al., 2002). If this is the case, it may be difficult to disentangle successful vs. unsuccessful trials based on fMRI signal amplitude (because of the sluggish nature of the BOLD signal). In a recent lesion study with rats, subthalamic nucleus lesions did not affect the SSRT (Eagle et al., 2008). Instead, the lesions affected stopping accuracy irrespective of the SSD, consistent with the notion that this region may be more strongly involved in the response selection component of the task and not inhibition per se (Eagle et al., 2008). A similar suggestion about the role of the subthalamic nucleus in response selection was made based on Parkinson's disease patients (van den Wildenberg et al., 2006). The above considerations need to be qualified by the fact that the subthalamic nucleus is a very small structure and that the site observed in the present study might actually not correspond to this nucleus. It is thus possible that the actual subthalamic nucleus was not robustly engaged by our task.

In the left frontal cortex, we observed significant positive logistic slopes in the IFG and superior frontal gyrus. Although the right IFC is more consistently reported to be involved in response inhibition, the IFC on the left hemisphere has been reported to exhibit differential SUCC vs. UNSUCC responses (Li et al., 2006; Rubia et al., 2007) and has been linked to response inhibition in a recent lesion study (Swick et al., 2008). The left superior frontal gyrus has also been shown to be involved during response inhibition (Li et al., 2006) and patients with left superior frontal cortex lesions exhibit poor performance during go/no-go tasks (Picton et al., 2007). It should be noted that although positive logistic slopes were observed for these regions, unexpectedly, a different pattern of mean responses was observed (relative to, say, those in the right IFG). Specifically, relative to the go baseline, stop-related responses were negative. Because responses to go trials could not be separately estimated in the present study (as they constituted the baseline), the interpretation of these decreased activations is unclear. For instance, it is possible that these regions were engaged by both go and stop processes, but additional studies are needed to answer this question.

Some regions exhibited significant logistic fits but with negative slopes, including the anterior and posterior cingulate cortices (ACC and PCC, respectively). In our design, a negative logistic slope indicated that trials with larger evoked responses were more likely to correspond to failed inhibitions (i.e., errors). The results of the ACC are thus of particular interest given the role of this region in predicting the likelihood of making an error (Brown and Braver, 2005) in particular, and error processing more generally (Dehaene et al., 1994; Gehring et al., 1993). The ACC results are also consistent with those obtained in previous response inhibition studies (Garavan et al., 2003; Hester et al., 2004; Hester et al., 2009; Rubia et al., 2003). The results of the PCC are also consistent with the literature reporting error-related responses during response inhibition (Menon et al., 2001) and/or could be linked to motivational aspects associated with making an error (Maddock, 1999).

We also observed a significant SSD by fMRI interaction in the ACC. On average, the slope corresponding to the fMRI amplitude variable was negative in this region (higher activity predicted unsuccessful inhibition) and the relationship between fMRI amplitude and probability of inhibition changed based on SSD, such that it was stronger for higher SSD values compared to lower ones (see Fig. 8A). It is thus possible that activity in the ACC was related to the likelihood of making an error (Brown and Braver, 2005), because at longer SSD values subjects would have had a greater chance of committing an error (as the stop command was encountered “too late” for behavioral correction). It is conceivable, however, that the signals observed in the ACC were related to error detection/correction processes (Dehaene et al., 1994; Gehring et al., 1993). However, we believe this is a less likely possibility because errors committed at low SSD trials would have been more salient (because they are more easily correctable) and expected to generate stronger error-related responses in the ACC, just the opposite of the pattern observed in our data.

Greater bilateral anterior insula activation during unsuccessful compared to successful stop trials has been reported in several response inhibition studies (Hester et al., 2004; Li et al., 2006; Li et al., 2008a; Ramautar et al., 2006), but it is unclear whether this activity reflects anticipation and/or prediction of errors (Preuschoff et al., 2008), or if it is a consequence of committing errors and is linked to arousal/affective responses subsequent to the errors (Magno et al., 2006). Following the same logic as for the ACC above, our data suggest that the activity in the anterior insula, bilaterally, was more strongly related to predicting or anticipating errors. However, as in the case of the ACC, we cannot rule out post-error effects and future studies are needed to clarify the role of the anterior insula during failed inhibitions. No significant interaction effect was detected in the PCC, suggesting that the responses were related to preparing for and making the erroneous response and hence did not interact with SSD.

A previous study of the stop-signal task also investigated how inhibition-related signals depend on the SSD (Aron and Poldrack, 2006), and revealed that activations in the subthalamic nucleus and pre-supplementary motor area were larger for trials with longer SSDs; responses in the IFC did not depend on the SSD. In our study, we did not observe significant fMRI by SSD interactions in these regions. However, the comparison between the two studies is not direct because Aron and Poldrack (2006) employed a different contrast (SUCC vs. go) to index response inhibition.

Although several studies have reported the involvement of parietal regions during response inhibition (Garavan et al., 1999; Hester et al., 2009; Liddle et al., 2001), the exact nature of their involvement remains to be determined. One possibility is that their role is attentional and, in particular, that they reflect trial-by-trial fluctuations in the allocation of “resources” (Leber et al., 2008) that are needed for successful behavioral performance during demanding tasks. Consistent with this notion, a recent MEG study revealed that fluctuations of sensory processing linked to both go and stop stimuli impact inhibitory performance during a stop-signal task (Boehler et al., 2009). Here, we further investigated the role of attention-related processes during response inhibition by focusing on fronto-parietal regions known to have important roles in attention and that were robustly engaged by the present task – specifically, IPS, IPL, and FEF. Our analysis revealed that the “coupling” between these regions and the right IFG was increased during successful vs. unsuccessful performance, consistent with the notion that trial-by-trial fluctuations in attention are important in determining the relationship between the right IFG and response inhibition. Interestingly, not all fronto-parietal “attentional regions” behaved in the same way. Although responses in the right IPL exhibited the steepest logistic regression slope (i.e., largest predictive power in terms of successful vs. unsuccessful trials), no differential coupling between this region and the right IFG was observed. These findings suggest that interactions between the right IFG and attentional regions are relatively specific during the stop-signal task. Finally, coupling between fronto-parietal regions and the caudate and putamen did not differ as a function of behavioral performance – specifically, similar correlation values were observed during both successful and unsuccessful trials.

Evidence that the interactions between fronto-parietal attentional regions and the right IFG subserve behavior during the stop-signal task was supported by an additional analysis. We reasoned that if responses in attentional regions moderated the impact of the right IFG on behavior, the higher the responses in attentional regions the stronger the link between the right IFG and behavior (i.e., the steeper the logistic regression slope). Indeed, this pattern was observed when signals from the left IPS were considered (but not when other attentional regions were considered). Although analysis based on fMRI signals, naturally, cannot be used to imply causality, the observation that no correlation was detected when the left IPS and right IFG were swapped in their position in the analysis, suggests that the roles of the two regions was not completely exchangeable in our task. In a related fashion, in a recent stop-signal fMRI study, Duann and colleagues reported greater functional connectivity between the IFC and pre-supplementary motor area during SUCC compared to UNSUCC trials (Duann et al., 2009). Both of these regions are believed to be important nodes of the response inhibition network (Chambers et al., 2009), and our current connectivity analysis complements these findings by revealing greater functional coupling between “inhibition” and “attentional” regions during successful response inhibition.

In summary, by employing a trial-based event-related design, our study allowed us to quantitatively assess how moment-to-moment fluctuations in brain responses were associated with the probability of successful performance during a stop-signal task. Our findings revealed a network of brain regions whose response variability was tightly coupled with behavioral performance. Among these regions, we observed the right inferior frontal gyrus, a region that has been consistently implicated in response inhibition. Other notable regions included the caudate and putamen, bilaterally. Our findings thus support the notion that these regions are important sites of executive control during response inhibition. Furthermore, we investigated how trial-by-trial fluctuations in responses in attentional regions covaried with fluctuations in inhibition-related regions. Accordingly, the coupling between fronto-parietal attentional regions and the right IFG increased during successful vs. unsuccessful performance, suggesting that efficacious network interactions are important in determining behavioral outcome during the stop-signal task. In particular, the link between responses in the right IFG and behavior were moderated by moment-to-moment fluctuations in evoked responses in the left intraparietal sulcus.

Supplementary Material

Figure S1

Simulation study to assess potential selection bias. The distribution of estimated logistic regression slopes based on a noise distribution (i.e., no pre-selection; blue) was identical to the one obtained via the ROI procedure employed in the main study. The complete overlap indicates that no bias was incurred by our procedure.

Acknowledgements

We thank Hugh Garavan and other anonymous reviewers for valuable feedback and Andrew Bauer for assistance with figures. Support for this work was provided in part by the National Institute of Mental Health (R01 MH071589) and the Indiana METACyt Initiative of Indiana University, funded in part through a major grant from the Lilly Endowment, Inc.

References

  1. Aron AR, Durston S, Eagle DW, Logan GD, Stinear CM, Stuphorn V. Converging evidence for a fronto-basal-ganglia network for inhibitory control of action and cognition. J Neurosci. 2007;27:11860–11864. doi: 10.1523/JNEUROSCI.3644-07.2007. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Aron AR, Fletcher PC, Bullmore ET, Sahakian BJ, Robbins TW. Stop-signal inhibition disrupted by damage to right inferior frontal gyrus in humans. Nat Neurosci. 2003;6:115–116. doi: 10.1038/nn1003. [DOI] [PubMed] [Google Scholar]
  3. Aron AR, Poldrack RA. Cortical and subcortical contributions to Stop signal response inhibition: role of the subthalamic nucleus. J Neurosci. 2006;26:2424–2433. doi: 10.1523/JNEUROSCI.4682-05.2006. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Boehler CN, Munte TF, Krebs RM, Heinze HJ, Schoenfeld MA, Hopf JM. Sensory MEG responses predict successful and failed inhibition in a stop-signal task. Cereb Cortex. 2009;19:134–145. doi: 10.1093/cercor/bhn063. [DOI] [PubMed] [Google Scholar]
  5. Booth JR, Burman DD, Meyer JR, Lei Z, Trommer BL, Davenport ND, Li W, Parrish TB, Gitelman DR, Mesulam MM. Larger deficits in brain networks for response inhibition than for visual selective attention in attention deficit hyperactivity disorder (ADHD) J Child Psychol Psychiatry. 2005;46:94–111. doi: 10.1111/j.1469-7610.2004.00337.x. [DOI] [PubMed] [Google Scholar]
  6. Boucher L, Palmeri TJ, Logan GD, Schall JD. Inhibitory control in mind and brain: an interactive race model of countermanding saccades. Psychol Rev. 2007;114:376–397. doi: 10.1037/0033-295X.114.2.376. [DOI] [PubMed] [Google Scholar]
  7. Brass M, Derrfuss J, Forstmann B, von Cramon DY. The role of the inferior frontal junction area in cognitive control. Trends Cogn Sci. 2005;9:314–316. doi: 10.1016/j.tics.2005.05.001. [DOI] [PubMed] [Google Scholar]
  8. Brass M, Wenke D, Spengler S, Waszak F. Neural correlates of overcoming interference from instructed and implemented stimulus-response associations. J Neurosci. 2009;29:1766–1772. doi: 10.1523/JNEUROSCI.5259-08.2009. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Britten KH, Newsome WT, Shadlen MN, Celebrini S, Movshon JA. A relationship between behavioral choice and the visual responses of neurons in macaque MT. Visual Neuroscience. 1996;13:87–100. doi: 10.1017/s095252380000715x. [DOI] [PubMed] [Google Scholar]
  10. Brown JW, Braver TS. Learned predictions of error likelihood in the anterior cingulate cortex. Science. 2005;307:1118–1121. doi: 10.1126/science.1105783. [DOI] [PubMed] [Google Scholar]
  11. Casey BJ, Trainor R, Orendi JL, Schubert AB, Nystrom LE, Giedd JN, Castellanos FX, Haxby JV, Noll DC, Cohen JD, et al. A developmental functional MRI study of prefrontal activation during performance of a Go-No-Go task. Journal of Cognitive Neuroscience. 1997;9:835–847. doi: 10.1162/jocn.1997.9.6.835. [DOI] [PubMed] [Google Scholar]
  12. Chamberlain SR, Hampshire A, Muller U, Rubia K, Del Campo N, Craig K, Regenthal R, Suckling J, Roiser JP, Grant JE, et al. Atomoxetine modulates right inferior frontal activation during inhibitory control: a pharmacological functional magnetic resonance imaging study. Biol Psychiatry. 2009;65:550–555. doi: 10.1016/j.biopsych.2008.10.014. [DOI] [PubMed] [Google Scholar]
  13. Chambers CD, Bellgrove MA, Gould IC, English T, Garavan H, McNaught E, Kamke M, Mattingley JB. Dissociable mechanisms of cognitive control in prefrontal and premotor cortex. J Neurophysiol. 2007;98:3638–3647. doi: 10.1152/jn.00685.2007. [DOI] [PubMed] [Google Scholar]
  14. Chambers CD, Bellgrove MA, Stokes MG, Henderson TR, Garavan H, Robertson IH, Morris AP, Mattingley JB. Executive “brake failure” following deactivation of human frontal lobe. J Cogn Neurosci. 2006;18:444–455. doi: 10.1162/089892906775990606. [DOI] [PubMed] [Google Scholar]
  15. Chambers CD, Garavan H, Bellgrove MA. Insights into the neural basis of response inhibition from cognitive and clinical neuroscience. Neurosci Biobehav Rev. 2009;33:631–646. doi: 10.1016/j.neubiorev.2008.08.016. [DOI] [PubMed] [Google Scholar]
  16. Chen CY, Muggleton NG, Tzeng OJ, Hung DL, Juan CH. Control of prepotent responses by the superior medial frontal cortex. Neuroimage. 2009;44:537–545. doi: 10.1016/j.neuroimage.2008.09.005. [DOI] [PubMed] [Google Scholar]
  17. Chikazoe J, Jimura K, Asari T, Yamashita K, Morimoto H, Hirose S, Miyashita Y, Konishi S. Functional dissociation in right inferior frontal cortex during performance of go/no-go task. Cereb Cortex. 2009;19:146–152. doi: 10.1093/cercor/bhn065. [DOI] [PubMed] [Google Scholar]
  18. Cohen M. Parametric analysis of fMRI data using linear systems methods. Neuroimage. 1997;6:93–103. doi: 10.1006/nimg.1997.0278. [DOI] [PubMed] [Google Scholar]
  19. Colzato LS, van den Wildenberg WP, Hommel B. Impaired inhibitory control in recreational cocaine users. PLoS ONE. 2007;2:e1143. doi: 10.1371/journal.pone.0001143. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Corbetta M, Shulman GL. Control of goal-directed and stimulus-driven attention in the brain. Nature Reviews Neuroscience. 2002;3:201–215. doi: 10.1038/nrn755. [DOI] [PubMed] [Google Scholar]
  21. Cox RW. AFNI, software for the analysis and visualization of functional magnetic resonance neuroimages. Comput Biomed Res. 1996;29:162–173. doi: 10.1006/cbmr.1996.0014. [DOI] [PubMed] [Google Scholar]
  22. Dehaene S, Posner M, Tucker D. Localization of a neural system for error detection and compensation. Psychological Science. 1994;5:303–305. [Google Scholar]
  23. Duann JR, Ide JS, Luo X, Li CS. Functional connectivity delineates distinct roles of the inferior frontal cortex and presupplementary motor area in stop signal inhibition. J Neurosci. 2009;29:10171–10179. doi: 10.1523/JNEUROSCI.1300-09.2009. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Eagle DM, Baunez C, Hutcheson DM, Lehmann O, Shah AP, Robbins TW. Stop-signal reaction-time task performance: role of prefrontal cortex and subthalamic nucleus. Cereb Cortex. 2008;18:178–188. doi: 10.1093/cercor/bhm044. [DOI] [PubMed] [Google Scholar]
  25. Eagle DM, Robbins TW. Inhibitory control in rats performing a stop-signal reaction-time task: effects of lesions of the medial striatum and d-amphetamine. Behav Neurosci. 2003;117:1302–1317. doi: 10.1037/0735-7044.117.6.1302. [DOI] [PubMed] [Google Scholar]
  26. Eimer M. Effects of attention and stimulus probability on ERPs in a Go/Nogo task. Biol Psychol. 1993;35:123–138. doi: 10.1016/0301-0511(93)90009-w. [DOI] [PubMed] [Google Scholar]
  27. Floden D, Stuss DT. Inhibitory control is slowed in patients with right superior medial frontal damage. J Cogn Neurosci. 2006;18:1843–1849. doi: 10.1162/jocn.2006.18.11.1843. [DOI] [PubMed] [Google Scholar]
  28. Forstmann BU, Jahfari S, Scholte HS, Wolfensteller U, van den Wildenberg WP, Ridderinkhof KR. Function and structure of the right inferior frontal cortex predict individual differences in response inhibition: a model-based approach. J Neurosci. 2008;28:9790–9796. doi: 10.1523/JNEUROSCI.1465-08.2008. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Garavan H, Ross TJ, Kaufman J, Stein EA. A midline dissociation between error-processing and response-conflict monitoring. Neuroimage. 2003;20:1132–1139. doi: 10.1016/S1053-8119(03)00334-3. [DOI] [PubMed] [Google Scholar]
  30. Garavan H, Ross TJ, Murphy K, Roche RAP, Stein EA. Dissociable Executive Functions in the Dynamic Control of Behavior: Inhibition, Error Detection, and Correction. Neuroimage. 2002;17:1820–1829. doi: 10.1006/nimg.2002.1326. [DOI] [PubMed] [Google Scholar]
  31. Garavan H, Ross TJ, Stein EA. Right Hemispheric Dominance of Inhibitory Control: An Event-Related Functional MRI Study. Proceedings of the National Academy of Sciences of the United States of America. 1999;96:8301–8306. doi: 10.1073/pnas.96.14.8301. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Gehring W, Goss B, Coles M, Meyer D, Donchin E. A neural system for error detection and compensation. Psychological Science. 1993;4:385–390. [Google Scholar]
  33. Gelman A, Hill J. Data Analysis Using Regression and Multilevel/Hierarchical Models. Cambridge University Press; 2007. [Google Scholar]
  34. Genovese CR, Lazar NA, Nichols T. Thresholding of statistical maps in functional neuroimaging using the false discovery rate. Neuroimage. 2002;15:870–878. doi: 10.1006/nimg.2001.1037. [DOI] [PubMed] [Google Scholar]
  35. Hester R, Fassbender C, Garavan H. Individual differences in error processing: a review and reanalysis of three event-related fMRI studies using the GO/NOGO task. Cereb Cortex. 2004;14:986–994. doi: 10.1093/cercor/bhh059. [DOI] [PubMed] [Google Scholar]
  36. Hester R, Madeley J, Murphy K, Mattingley JB. Learning from errors: error-related neural activity predicts improvements in future inhibitory control performance. J Neurosci. 2009;29:7158–7165. doi: 10.1523/JNEUROSCI.4337-08.2009. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Hosmer DW, Lemeshow S. Applied logistic regression. Second edn John Wiley & Sons, Inc.; New York: 2000. [Google Scholar]
  38. Kalaska JF, Crammond DJ. Deciding not to GO: neuronal correlates of response selection in a GO/NOGO task in primate premotor and parietal cortex. Cereb Cortex. 1995;5:410–428. doi: 10.1093/cercor/5.5.410. [DOI] [PubMed] [Google Scholar]
  39. Kastner S, Ungerleider LG. The neural basis of biased competition in human visual cortex. Neuropsychologia. 2001;39:1263–1276. doi: 10.1016/s0028-3932(01)00116-6. [DOI] [PubMed] [Google Scholar]
  40. Kriegeskorte N, Simmons WK, Bellgowan PS, Baker CI. Circular analysis in systems neuroscience: the dangers of double dipping. Nat Neurosci. 2009;12:535–540. doi: 10.1038/nn.2303. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Leber AB, Turk-Browne NB, Chun MM. Neural predictors of moment-to-moment fluctuations in cognitive flexibility. Proc Natl Acad Sci U S A. 2008;105:13592–13597. doi: 10.1073/pnas.0805423105. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Leung HC, Cai W. Common and differential ventrolateral prefrontal activity during inhibition of hand and eye movements. J Neurosci. 2007;27:9893–9900. doi: 10.1523/JNEUROSCI.2837-07.2007. [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Li CS, Huang C, Constable RT, Sinha R. Imaging response inhibition in a stop-signal task: neural correlates independent of signal monitoring and post-response processing. J Neurosci. 2006;26:186–192. doi: 10.1523/JNEUROSCI.3741-05.2006. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Li CS, Yan P, Chao HH, Sinha R, Paliwal P, Constable RT, Zhang S, Lee TW. Error-specific medial cortical and subcortical activity during the stop signal task: a functional magnetic resonance imaging study. Neuroscience. 2008a;155:1142–1151. doi: 10.1016/j.neuroscience.2008.06.062. [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Li CS, Yan P, Sinha R, Lee TW. Subcortical processes of motor response inhibition during a stop signal task. Neuroimage. 2008b;41:1352–1363. doi: 10.1016/j.neuroimage.2008.04.023. [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Liddle PF, Kiehl KA, Smith AM. Event-related fMRI study of response inhibition. Hum Brain Mapp. 2001;12:100–109. doi: 10.1002/1097-0193(200102)12:2&#x0003c;100::AID-HBM1007&#x0003e;3.0.CO;2-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Lim SL, Padmala S, Pessoa L. Segregating the significant from the mundane on a moment-to-moment basis via direct and indirect amygdala contributions. Proc Natl Acad Sci U S A. 2009;106:16841–16846. doi: 10.1073/pnas.0904551106. [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Loftus GR, Masson ME. Using confidence intervals in within-subject designs. Psychonomic Bulletin & Review. 1994;1:476–490. doi: 10.3758/BF03210951. [DOI] [PubMed] [Google Scholar]
  49. Logan GD. On the ability to inhibit thought and action: A user's guide to the stop signal paradigm. In: Dagenbach D, Carr TH, editors. Inhibitory processes in attention, memory, and language. Academic Press; San Diego: 1994. pp. 189–239. [Google Scholar]
  50. Logan GD, Cowan WB. On the ability to inhibit thought and action: A theory of an act of control. Psych Rev. 1984;91:295–327. [Google Scholar]
  51. Logan GD, Schachar RJ, Tannock R. Impulsivity and inhibitory control. Psychological Science. 1997;8:60–64. [Google Scholar]
  52. Maddock RJ. The retrosplenial cortex and emotion: new insights from functional neuroimaging of the human brain. Trends in Neuroscience. 1999;22:310. doi: 10.1016/s0166-2236(98)01374-5. [DOI] [PubMed] [Google Scholar]
  53. Magno E, Foxe JJ, Molholm S, Robertson IH, Garavan H. The anterior cingulate and error avoidance. J Neurosci. 2006;26:4769–4773. doi: 10.1523/JNEUROSCI.0369-06.2006. [DOI] [PMC free article] [PubMed] [Google Scholar]
  54. Menon V, Adleman NE, White CD, Glover GH, Reiss AL. Error-related brain activation during a Go/NoGo response inhibition task. Hum Brain Mapp. 2001;12:131–143. doi: 10.1002/1097-0193(200103)12:3&#x0003c;131::AID-HBM1010&#x0003e;3.0.CO;2-C. [DOI] [PMC free article] [PubMed] [Google Scholar]
  55. Nachev P, Wydell H, O'Neill K, Husain M, Kennard C. The role of the pre-supplementary motor area in the control of action. Neuroimage. 2007;36(Suppl 2):T155–163. doi: 10.1016/j.neuroimage.2007.03.034. [DOI] [PMC free article] [PubMed] [Google Scholar]
  56. Padmala S, Pessoa L. Affective learning enhances visual detection and responses in primary visual cortex. J Neurosci. 2008;28:6202–6210. doi: 10.1523/JNEUROSCI.1233-08.2008. [DOI] [PMC free article] [PubMed] [Google Scholar]
  57. Padmala S, Pessoa L. Interactions between cognition and motivation during response inhibition. Neuropsychologia. 2009 doi: 10.1016/j.neuropsychologia.2009.10.017. [DOI] [PMC free article] [PubMed] [Google Scholar]
  58. Pessoa L, Gutierrez E, Bandettini PB, Ungerleider LG. Neural correlates of visual working memory: fMRI amplitude predicts task performance. Neuron. 2002;35:975–987. doi: 10.1016/s0896-6273(02)00817-6. [DOI] [PubMed] [Google Scholar]
  59. Pessoa L, Padmala S. Quantitative prediction of perceptual decisions during near-threshold fear detection. Proc Natl Acad Sci U S A. 2005;102:5612–5617. doi: 10.1073/pnas.0500566102. [DOI] [PMC free article] [PubMed] [Google Scholar]
  60. Pessoa L, Ungerleider LG. Top-down mechanisms for working memory and attentional processes. In: Gazzaniga MS, editor. The new cognitive neurosciences. 3rd Edition MIT Press; Cambridge, MA: 2004. pp. 919–930. [Google Scholar]
  61. Picton TW, Stuss DT, Alexander MP, Shallice T, Binns MA, Gillingham S. Effects of focal frontal lesions on response inhibition. Cereb Cortex. 2007;17:826–838. doi: 10.1093/cercor/bhk031. [DOI] [PubMed] [Google Scholar]
  62. Preuschoff K, Quartz SR, Bossaerts P. Human insula activation reflects risk prediction errors as well as risk. J Neurosci. 2008;28:2745–2752. doi: 10.1523/JNEUROSCI.4286-07.2008. [DOI] [PMC free article] [PubMed] [Google Scholar]
  63. Purushothaman G, Bradley DC. Neural population code for fine perceptual decisions in area MT. Nat Neurosci. 2005;8:99–106. doi: 10.1038/nn1373. [DOI] [PubMed] [Google Scholar]
  64. Ramautar JR, Slagter HA, Kok A, Ridderinkhof KR. Probability effects in the stop-signal paradigm: the insula and the significance of failed inhibition. Brain Res. 2006;1105:143–154. doi: 10.1016/j.brainres.2006.02.091. [DOI] [PubMed] [Google Scholar]
  65. Ratcliffe SJ, Shults J. GEEQBOX: A MATLAB Toolbox for Generalized Estimating Equations and Quasi-Least Squares. Journal of Statistical Software. 2008;25 [Google Scholar]
  66. Ray NJ, Jenkinson N, Brittain J, Holland P, Joint C, Nandi D, Bain PG, Yousif N, Green A, Stein JS, Aziz TZ. The role of the subthalamic nucleus in response inhibition: evidence from deep brain stimulation for Parkinson's disease. Neuropsychologia. 2009;47:2828–2834. doi: 10.1016/j.neuropsychologia.2009.06.011. [DOI] [PubMed] [Google Scholar]
  67. Ress D, Backus BT, Heeger DJ. Activity in primary visual cortex predicts performance in a visual detection task. Nature Neuroscience. 2000;3:940–945. doi: 10.1038/78856. [DOI] [PubMed] [Google Scholar]
  68. Rubia K, Smith AB, Brammer MJ, Taylor E. Right inferior prefrontal cortex mediates response inhibition while mesial prefrontal cortex is responsible for error detection. Neuroimage. 2003;20:351–358. doi: 10.1016/s1053-8119(03)00275-1. [DOI] [PubMed] [Google Scholar]
  69. Rubia K, Smith AB, Taylor E, Brammer M. Linear age-correlated functional development of right inferior fronto-striato-cerebellar networks during response inhibition and anterior cingulate during error-related processes. Hum Brain Mapp. 2007;28:1163–1177. doi: 10.1002/hbm.20347. [DOI] [PMC free article] [PubMed] [Google Scholar]
  70. Swick D, Ashley V, Turken AU. Left inferior frontal gyrus is critical for response inhibition. BMC Neurosci. 2008;9:102. doi: 10.1186/1471-2202-9-102. [DOI] [PMC free article] [PubMed] [Google Scholar]
  71. Sylvester CM, Shulman GL, Jack AI, Corbetta M. Asymmetry of anticipatory activity in visual cortex predicts the locus of attention and perception. J Neurosci. 2007;27:14424–14433. doi: 10.1523/JNEUROSCI.3759-07.2007. [DOI] [PMC free article] [PubMed] [Google Scholar]
  72. Talairach J, Tournoux P. Co-planar stereotaxis atlas of the human brain. Thieme Medical; New York: 1988. [Google Scholar]
  73. van den Wildenberg WP, van Boxtel GJ, van der Molen MW, Bosch DA, Speelman JD, Brunia CH. Stimulation of the subthalamic region facilitates the selection and inhibition of motor responses in Parkinson's disease. J Cogn Neurosci. 2006;18:626–636. doi: 10.1162/jocn.2006.18.4.626. [DOI] [PubMed] [Google Scholar]
  74. Verbruggen F, Logan GD. Response inhibition in the stop-signal paradigm. Trends Cogn Sci. 2008 doi: 10.1016/j.tics.2008.07.005. [DOI] [PMC free article] [PubMed] [Google Scholar]
  75. Vink M, Kahn RS, Raemaekers M, van den Heuvel M, Boersma M, Ramsey NF. Function of striatum beyond inhibition and execution of motor responses. Hum Brain Mapp. 2005;25:336–344. doi: 10.1002/hbm.20111. [DOI] [PMC free article] [PubMed] [Google Scholar]
  76. Williams B, Ponesse J, Schachar R, Logan G, Tannock R. Development of inhibitory control across the life span. Developmental Psychology. 1999;35:205–213. doi: 10.1037//0012-1649.35.1.205. [DOI] [PubMed] [Google Scholar]
  77. Xue G, Aron AR, Poldrack RA. Common neural substrates for inhibition of spoken and manual responses. Cereb Cortex. 2008;18:1923–1932. doi: 10.1093/cercor/bhm220. [DOI] [PubMed] [Google Scholar]
  78. Zeger SL, Liang KY. Longitudinal data analysis for discrete and continuous outcomes. Biometrics. 1986;42:121–130. [PubMed] [Google Scholar]
  79. Zhou D, Thompson WK, Siegle G. MATLAB toolbox for functional connectivity. Neuroimage. 2009;47:1590–1607. doi: 10.1016/j.neuroimage.2009.05.089. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Figure S1

Simulation study to assess potential selection bias. The distribution of estimated logistic regression slopes based on a noise distribution (i.e., no pre-selection; blue) was identical to the one obtained via the ROI procedure employed in the main study. The complete overlap indicates that no bias was incurred by our procedure.

NIHMS200314-supplement-Figure_S1.pdf (161.4KB, pdf)

RESOURCES