Neural and behavioral correlates of selective stopping: Evidence for a different strategy adoption
Introduction
Response inhibition, defined as the ability to suppress unwanted thoughts and actions, plays a fundamental role acting as a basement for more complex cognitive capabilities (Verbruggen and Logan, 2008). Moreover, it allows people to flexibly adapt their behavior depending on current goals. The importance of this fundamental ability emerges clearly when considering its impairment in several neurological diseases and psychiatric disorders, such as Huntington's disease (Beste et al., 2008), obsessive-compulsive disorder (Bannon et al., 2002), attention-deficit/hyperactivity disorder (López-Martín et al., 2015) or substance abuse (Fillmore and Rush, 2002, Monterosso, 2005).
In order to investigate response inhibition, several experimental tasks have been designed. In particular, the so called go-no go and stop signal paradigms have widely been used to characterize the processes involved in response inhibition (Bari and Robbins, 2013, Chambers et al., 2009, Huster et al., 2013). In these tasks, a response tendency is first induced to participants (go condition), which has to be suddenly interrupted (no-go/stop condition). As a consequence the targeted response inhibition process is triggered. However, these two tasks differ in the specific nature of inhibition: the go/no-go task demands to withhold a prepotent but not yet initiated response, whereas the stop-signal task requires cancelling an already initiated response. Specifically, in the stop-signal task participants are usually asked to make button press responses to go stimuli and to interrupt their responses whenever a stop-signal is presented shortly after them. The main parameter of this task is the so called stop signal delay (SSD), which indexes the lag between the go stimulus and the stop-signal. Crucially, the probability of committing an inhibition error can be manipulated by modifying the duration of this delay, which allows for the calculation of the stop signal reaction time (SSRT). This has been considered to be a precise measure of the time that takes to inhibit a response. It is important to note that SSRT is an index of the several processes involved in the successful interruption of a motor response, which at least include the encoding of the stopping stimulus and the subsequent interruption of the ongoing response (Boucher et al., 2007). This estimation relies on the assumptions made by the horse-race model proposed by Logan and Cowan (1984) that describes the relative finishing time of the go (go RT) and stop (stop-respond RT) processes. Specifically, the go RT reflects the finishing time of the go process, whereas the stop-respond RT represents go RT on those trials with stop signals in which individuals fail to inhibit the response. In order to simplify the formal model of the estimation of SSRT, it is assumed that going and stopping processes (which are both active in stop trials) are independent (Logan and Cowan, 1984, Verbruggen and Logan, 2008; see also the interactive model proposed by Boucher et al., 2007). This means that the go RT distribution does not change once go responses have been initiated in stop signal trials. Importantly, this proposal assumes that a commission error will be made when going process finishes the race before stopping process. Thus, the horse-race model predicts that mean RTs on failed stop trials should be faster than mean RTs on go trials (Logan and Cowan, 1984, Verbruggen and Logan, 2009b). Finally, the model posits that the latency of the inhibition process can be measured based on the signal-response distribution by estimating the RT that represents the 50% likelihood of emitting a response and making a subsequent subtraction of the SSD from that value.
Several fMRI studies have tried to identify the neural basis of response inhibition. The results of these studies suggest that medial frontal regions (primarily, the pre-supplementary motor area; pre-SMA), the ventrolateral prefrontal cortex (inferior frontal gyrus) and the basal ganglia are particularly involved in inhibitory control (Chikazoe, 2010, Horn et al., 2003, Li et al., 2006, Li et al., 2008, Liddle et al., 2001). Of note, the chain of processes involved in the successful inhibition of a dominant response tendency is considered to last just a few hundred milliseconds (Huster et al., 2013). Therefore, the use of techniques with a high temporal resolution -such as event-related potentials or ERPs- may be suitable to complement fMRI data by providing insights on the temporal course of the processes associated with response inhibition. In this sense, prior ERPs studies with go/no-go and stop-signal tasks have already shown that two frontocentral components -the N2 (200–400 ms) and the P3 (300–600 ms)- are of particular interest when investigating response inhibition (Albert et al., 2013, Huster et al., 2013, Ramautar et al., 2004). Although the precise functional meaning of these components is still debatable, recent evidence suggests that the onset of the P3 may be a reliable index of response cancellation processes (Dimoska et al., 2006, Kok et al., 2004, Wessel and Aron, 2015). Moreover, P3 onset latency (defined as the temporal point at which differences between stop and go trials reached significance) has been found to match the time of the end of the stop process (as measured by the end of the SSRT) in the stop-signal paradigm (Wessel and Aron, 2015). In contrast, P3 peak latency is thought to reflect evaluative and reinforcement processes, which occur once the response has been successfully interrupted (Huster et al., 2013). With respect to the N2, it is however unclear whether this component reflects response interruption or other processing stages prior to response cancellation, such as perceptual mismatch, novelty processing or conflict detection (Folstein and Van Petten, 2008).
An important theoretical question concerns the processes underlying the experimental conditions typically used in studies with stop signal tasks. Notably, conclusions on the neural basis of stopping have been mainly drawn by comparing brain activity associated with successful stop trials and successful go trials, under the assumption that these conditions essentially differ in the activation of brain networks underlying response interruption. However, several studies have criticized this contrast for not being specific enough (Albert et al., 2013, Boehler et al., 2010, Dimoska et al., 2006, Etchell et al., 2012, Li et al., 2006, Sharp et al., 2010). In this sense, it should be noted that stop and go conditions not only differ in the involvement of response cancellation mechanisms per se but also in other processes (e.g., perceptual and cognitive). To try to solve this problem, in some studies successful stop trials were compared to unsuccessful stop trials. However, this solution has shown to be too restricted, since response interruption is involved to some extent in both conditions (Boehler et al., 2010). Moreover, compared with successful stopping, inhibition failures are associated with emotional processes (e.g., frustration) and error monitoring (Li et al., 2006). Finally, according to the assumptions made by the horse-race model, failed stop trials would result from a delay in the initiation of the stopping process compared to successful stop trials. This would make successful and failed stop trials difficult to compare.
In an attempt to isolate the brain activity specifically associated with the cancellation of an already initiated response, a new condition called ignore (or continue) has recently been introduced in the stop-signal paradigm. Although this new control condition still differs from the stop condition in decision-making processes (the ignore condition could be less relevant to the current task program than the stop one) and learning-related processes (the ignore condition leads to fewer errors than the stop one), it minimizes the aforementioned limitations by controlling both the novelty effect and sensory properties (Albert et al., 2013, Sharp et al., 2010, Etchell et al., 2012). Specifically, the sequence of events in the ignore condition resembles that of the stop condition since both trials start with a go stimulus which is followed by a signal stimulus after a delay. However, as opposed to the stop condition, participants in the ignore condition should continue responding after the signal stimulus as if it was a go trial. Moreover, the probabilities of occurrence of ignore and stop trials are kept equal, ruling out the possibility that activation differences between these two conditions reflect novelty/oddball processing. To date, only few fMRI (Boehler et al., 2010, Sharp et al., 2010), and ERP (Etchell et al., 2012) studies have followed this new approach.
In the behavioral domain, the results of studies that included ignore trials have led to the definition of the concept of selective inhibition (Bissett and Logan, 2014). It refers to the implementation of an inhibition that occurs only under specific circumstances related to certain stimulus features or to the response demanded by task instructions. Specifically, the inclusion of ignore trials in the so-called stimulus-selective stopping task would trigger slightly different processes (e.g., perceptual discrimination between stop and ignore signal) than those involved in the simple stop signal paradigm. Remarkably, it is assumed that in the stimulus-selective stopping paradigm participants selectively interrupt their responses to stop but not to ignore signals once the signal has been discriminated. However, as we will discuss later, recent evidence suggests that some participants perform the stimulus-selective stopping task by inhibiting selectively whereas others do not (because they stop their response whenever a signal occurs -ignore or stop-, and then restart the cancelled response if the signal presented was an ignore one Bissett and Logan, 2014). Interestingly, it is possible to identify the strategy adopted by each participant to perform a stimulus-selective stopping task (whether subjects inhibit selectively or not) by comparing RTs during no signal (go), ignore and stop trials (see decision matrix in Bissett & Logan, 2014, p. 457). Moreover, the independence between going and stopping assumed by the horse-race model can be tested separately for each strategy, which provides important information to establish the validity of the SSRT calculation.
Specifically, there is evidence of three distinct strategies that can be used to accomplish a stimulus-selective stop task (Bissett and Logan, 2014): Stop then Discriminate (StD), independent Discriminate then Stop (iDtS) and dependent Discriminate then Stop (dDtS). In the StD strategy, subjects inhibit an already initiated response whenever a signal occurs (either stop or ignore), and subsequently discriminate the signal (restarting the cancelled go response if the signal is an ignore). The use of this strategy is typically reflected in shorter RTs for stop compared to no-signal trials (the context independence assumption between going and stopping is therefore kept), as well as in longer RTs for ignore compare to no-signal trials (because subjects stop non-selectively to both stop- and ignore-signals, and then they have to restart the response on ignore-signal trials). Thus, participants who used the StD strategy might not be doing what would be expected (inhibit their responses selectively). Given that discrimination of the signal occurs after the interruption of the ongoing response, SSRTs computed for subjects who use this strategy are thought to be similar to those reported in simple stop-signal tasks. Thus, in the StD strategy, SSRTs can be calculated using no-signal RT distributions, as in classic stop signal tasks.
By contrast, in the iDtS and dDtS strategies, subjects discriminate the signal first (either ignore or stop) and, subsequently, they cancel the already initiated go response if the signal was a stop. By adopting these strategies, participants would be thus performing the task as expected, since they stop their responses selectively. However, while signal discrimination does not interfere with going process in the iDtS strategy, it does so in the dDtS strategy. For this reason, the context independence assumption of the horse-race model is preserved in the former but not in the latter strategy. Thus, the iDtS strategy is associated with shorter RTs for stop than for no-signal and with similar RTs for ignore than for no-signal. This pattern of results suggest that go processing is not affected by the requirement to discriminate the signal, which may be attributed to the fact that stop and ignore signals are easy to discriminate. Therefore, SSRTs can be calculated using no-signal RT distributions in this strategy. The dDtS yields similar or even larger RT for stop compared to no-signal RT, and larger RTs for ignore compared to no-signal RT. These results suggest that discriminating between stop and ignore signals slows the going process whenever a signal occurs (either stop or ignore). Therefore, a certain degree of interaction is assumed in the dDtS strategy, possibly because stop and ignore signals are difficult to discriminate. As a consequence, the assumption of the independent race model (Logan and Cowan, 1984) would be violated, which invalidates the commonly used method for estimating SSRTs using no-signal RT distributions.
Interestingly, Bissett and Logan (2014) showed that it is possible to induce participants to adopt any of the three strategies by modifying the relative probabilities of stop and ignore trials. Thus, the same individual could use several strategies depending on the context of the task. However, when the stop and the ignore trials are presented equally often to participants (as in previous neuroimaging studies aiming to disentangle novelty from inhibition processes), there is not a priori assumption about which of the three strategies will be adopted. Thus, brain activity found in prior fMRI and ERP studies with selective stopping tasks reflected the performance of participants that may have been adopting different strategies to complete the task. It seems therefore crucial to detect and categorize the strategy used by each subject before analyzing their brain activity. Otherwise, “some subjects may not be doing what researchers think they are doing” (i.e., inhibit selectively: Bissett & Logan, 2014, p. 457).
The main aim of this study was to examine the neural correlates of response interruption for the different strategies that participants used to accomplish a stimulus-selective stop-signal task. To fully take advantage of the precise temporal resolution of the EEG, we conducted mass univariate analyses on ERP data (Groppe et al., 2011). This approach allowed us to precisely determine both the onset and the end of the processes related to response cancellation for each strategy. Source localization techniques were also employed to estimate the neural underpinnings of the effects observed at the scalp level. Importantly, an appropriate estimation of SSRTs was computed based on integration method, taking into account the strategy adopted and its consequences for horse-race model assumptions. First, we identified and classified the strategy used by each participant following the criteria proposed by Bissett and Logan (2014). We then examined the functional comparison that better isolated the brain activity specifically associated with response interruption in each strategy (successful stop vs. go, successful stop vs. failed stop, or successful stop vs. ignore). Finally, we also considered whether or not the behavioral estimation of the precise timing of response interruption process (i.e., SSRT) matches the time course of the electrophysiological differences for each of these contrasts within each strategy.
We hypothesized that the successful stop versus ignore comparison would provide the best functional contrast to isolate activation related to response interruption. However, drawing upon recent behavioral data suggesting strategic heterogeneity in selective stopping, we also predicted that such differences would emerge in DtS strategies (since subjects who used the DtS strategies, inhibit their responses selectively to stop but not to ignore signals) but not in the StD strategy (since subjects who used this strategy inhibit nonselectively to both stop and ignore signal). Electrophysiological differences observed in the successful stop versus ignore contrast in DtS strategies would have to temporally coincide with the end of the SSRT. Finally, we hypothesized that the onset of the P3, as well as the medial and ventrolateral frontal regions locations, would be the neural correlates involved in the cancellation of an ongoing response.
Section snippets
Participants
Fifty seven right-handed undergraduate university students, with an age range of 18–39 (mean = 21.9; SD = 3), took part in this experiment. The study was approved by the local ethics committee and informed consent was obtained from each subject prior to the experiment. All participants reported normal or corrected-to-normal visual acuity and had no history of neurological or psychiatric disorders. Eight subjects were excluded from the final analysis, two due to low overall task accuracy (< 2.5 SDs
Behavioral results
As explained before, the strategy followed by each participant was estimated by comparing their mean no-signal (go) RT, stop-respond RT and ignore RT. The result of these analyses indicated that none of the subjects adopted an iDtS strategy to perform the task. Evidence for the use of the StD strategy was found in 30 out of the 49 subjects. Therefore, the remaining 19 subjects used a dDtS strategy. Repeated measures t-tests performed at group level corroborated this individual distinction. In
Discussion
In the current study we investigated for the first time the neural basis of selective stopping by categorizing the strategy that participants adopted to perform the task. We exploited the high temporal resolution of the ERPs by using an analytical approach that allowed us to accurately determine the timing (both the onset and the end) of the response interruption processing stage for each strategy and functional comparison. Also, by using source localization methods we identified the brain
Conclusions
In sum, the present results provide evidence for the neural correlates underlying behavioral strategy adoption in selective stopping tasks (Bissett and Logan, 2014). Behavioral results suggest that, even though the same task instructions were given to participants and stop and ignore trials were presented equally often, they could perform the task in at least two different ways: by inhibiting selectively to stop but not to ignore signals (dDtS strategy) or by inhibiting non-selectively to both
Acknowledgments
This work was supported by the Ministerio de Economía y Competitividad (MINECO/FEDER) of Spain (PSI2015-68368-P) and the Comunidad de Madrid (H2015/HUM-3327). The authors thank Frederick Verbruggen for providing us the Matlab script of STOP-IT, Patrick Bisset for his thoughtful and useful review of a previous version of the manuscript, and three anonymous reviewers for their constructive and helpful suggestions. The authors also wish to thank Sara López-Martín, Eva M. Moreno and Pedro R.
References (64)
- et al.
Spatiotemporal characterization of response inhibition
NeuroImage
(2013) - et al.
Inhibition and the right inferior frontal cortex: one decade on
Trends Cogn. Sci.
(2014) - et al.
Response inhibition deficits in obsessive–compulsive disorder
Psychiatry Res.
(2002) - et al.
Inhibition and impulsivity: behavioral and neural basis of response control
Prog. Neurobiol.
(2013) - et al.
Response inhibition in Huntington's disease—a study using ERPs and sLORETA
Neuropsychologia
(2008) - et al.
Pinning down response inhibition in the brain—conjunction analyses of the stop-signal task
NeuroImage
(2010) - et al.
Cortical activity during manual response inhibition guided by color and orientation cues
Brain Res.
(2009) - et al.
Insights into the neural basis of response inhibition from cognitive and clinical neuroscience
Neurosci. Biobehav. Rev.
(2009) - et al.
EEGLAB: an open source toolbox for analysis of single-trial EEG dynamics including independent component analysis
J. Neurosci. Methods
(2004) - et al.
Spatial pattern of cerebral glucose metabolism (PET) correlates with localization of intracerebral EEG-generators in Alzheimer's disease
Clin. Neurophysiol.
(2000)
The auditory-evoked N2 and P3 components in the stop-signal task: indices of inhibition, response-conflict or error-detection?
Brain Cogn.
“Shut up!” An electrophysiological study investigating the neural correlates of vocal inhibition
Neuropsychologia
Impaired inhibitory control of behavior in chronic cocaine users
Drug Alcohol Depend.
Putting the brakes on inhibitory models of frontal lobe function
NeuroImage
Contrasting network and modular perspectives on inhibitory control
Trends Cogn. Sci.
Response inhibition and impulsivity: an fMRI study
Neuropsychologia
Electroencephalography of response inhibition tasks: functional networks and cognitive contributions
Int. J. Psychophysiol.
Deficits in response inhibition associated with chronic methamphetamine abuse
Drug Alcohol Depend.
A temporo-spatial analysis of the neural correlates of extrinsic perceptual grouping in vision
Neuropsychologia
Integration of fMR and simultaneous EEG: towards a comprehensive understanding of localization and time-course of brain activity in target detection
NeuroImage
Effects of stop-signal probability in the stop-signal paradigm: the N2/P3 complex further validated
Brain Cogn.
Are the neural correlates of stopping and not going identical? Quantitative meta-analysis of two response inhibition tasks
NeuroImage
Response inhibition in the stop-signal paradigm
Trends Cogn. Sci.
Models of response inhibition in the stop-signal and stop-change paradigms
Neurosci. Biobehav. Rev.
Evidence for capacity sharing when stopping
Cognition.
Selective stopping? Maybe not
J. Exp. Psychol. Gen.
An alternative method for significance testing of waveform difference potentials
Psychophysiology
Inhibitory control in mind and brain: an interactive race model of countermanding saccades
Psychol. Rev.
Dissociable roles of right inferior frontal cortex and anterior insula in inhibitory control: evidence from intrinsic and task-related functional parcellation, connectivity, and response profile analyses across multiple datasets
J. Neurosci.
Localizing performance of go/no-go tasks to prefrontal cortical subregions
Curr. Opin. Psychiatry
Prefrontal dynamics underlying rapid instructed task learning reverse with practice
J. Neurosci.
Rapid instructed task learning: a new window into the human brain's unique capacity for flexible cognitive control
Cogn. Affect. Behav. Neurosci.
Cited by (15)
Arcuate fasciculus architecture is associated with individual differences in pre-attentive detection of unpredicted music changes
2021, NeuroImageCitation Excerpt :To explore only reliable differences between standard and deviant ERP responses, as well as to properly control for Type I errors within the multiple comparisons carried out (Kilner, 2013), the ERPs from each of our six conditions were submitted to a repeated measures, two-tailed permutation test based on the tmax statistic (Blair and Karniski, 1993), using a family-wise alpha level of 0.05 by means of the Mass Univariate ERP Toolbox (Groppe et al., 2011). A time window between 50 and 350 ms was used based on the existing literature on MMN and P3a components (Tervaniemi et al., 2014; Vuust et al., 2016; Petersen et al., 2020), as performed in previous reports applying similar methodological approaches (Sanchez-Carmona et al., 2016; Scheer et al., 2016; Campos-Arteaga et al., 2020). Within the selected time-window, all time points at all 29 scalp electrodes were included in the test (i.e., 5226 total comparisons).
Stop! – The automatic tendency of action, inhibition and frontal activation in individuals with alcohol-use disorder in abstinence
2020, International Journal of PsychophysiologyThe influence of dopaminergic polymorphisms on selective stopping
2020, Behavioural Brain ResearchCitation Excerpt :In the iDtS, subjects discriminate the signal first before deciding to stop: if the signal is a stop signal, they stop; if the signal is an ignore signal, they complete their response without ever initiating the stop process. The dDtS is similar to the iDtS but here the requirement to discriminate stop and ignore signals interacts with the go process and slows the go reaction time (RT) whenever a signal occurs [14,15]. Therefore, both the iDts and the dDtS are strategies that reflect selectivity, as opposed to the StD.
Oscillatory brain mechanisms supporting response cancellation in selective stopping strategies
2019, NeuroImageCitation Excerpt :We found that a successful cancelation of an initiated response is specifically associated with an increased oscillatory activity in the high-beta frequency band in the strategy characterized by stopping selectively (dDtS), but not in the strategy characterized by stopping non-selectively (StD). These findings provide further neural support for the existence of different strategies for successfully performing stimulus-selective stopping tasks (Bissett and Logan, 2014; Sánchez-Carmona et al., 2016; Sebastian et al., 2017). Moreover, they provide evidence suggesting that high-beta oscillations in medial superior and middle frontal cortices may constitute an important neural marker of response cancellation.
Response inhibition in borderline personality disorder: Neural and behavioral correlates
2019, Biological Psychology