Neural Correlates Of Stopping And Changing Responses: Evidences From Spatiotemporal Analysis Of ERPs

ObjectiveResponse inhibition is a central component of executive control and is referred to as the cancellation of planned or pre-potent actions that are inappropriate in a given context. For example, in the prototypical stop signal task (SST), participants were required to respond as quickly as possible to a go stimulus; a stop signal was occasionally present following the go stimulus, instructing the withdrawal of the ongoing response. One’s ability to suppress obsolete action, which can be assessed by the stop signal reaction time (SSRT), is critical in everyday life. Deficits in response inhibition, namely impulsivity, may be the most compelling characteristic of many psychopathological disorders, such as attention deficits/hyperactivity disorder, obsessive-compulsive disorder. Alternatively, one might think of situations calling for a change of motor plans such that the prepared response has to be withheld but another has to be excuted instead. The question remains controversial whether the same brain mechanisms are underlying the inhibitory action control required in stopping and changing responses.Converging evidence from previous lesion, animal, transcranial magnetic stimulation, and neuroimaging studies aimed at isolating response-inhibition-related regions have shown that successful inhibition consistently activated a fronto-bansal-ganglia circuit mainly in the right hemisphere (response inhibition network, RIN), including the right inferior frontal gyrus (rIFG), the pre-supplementary motor area, the subcortical basal ganglia and so on. It is claimed that there are two types of stopping mechanisms in general, i.e., global and selective. Behaviorally global inhibition, as addressed in the classic stop-signal paradigm, suppresses the motor system in a global way, e.g., stopping a finger action was associated with the inhibition of the irrelevant leg; whereas behaviorally selective inhibition is more relevant and targeted. Interestingly, selective inhibition could be achieved either by stopping one action while keeping another action, or by stopping all actions and re-initiating a new action. The latter was commonly studied in a stop-change paradigm, an extension of the stop-signal paradigm. In this paradigm, change trials were added to the go and stop trials and participants had to change their initial response to another response once a go stimulus was followed by a change signal. That is, the change trials not only call for suppression of an ongoing response as the stop trials, but also require subsequent behavioral adjustment. The change signal reaction time (CSRT) can be estimated from the horse-race model, in a similar way with SSRT. However, studies using the stop-change paradigm observed contradictory data. Several functional magnetic, resonance imaging (fMRI) researches found that the rIFG, a core part of the RIN, was activated in both stop and change trials, supporting the notion that inhibitory process functions similarly. On the contrary, some event-related potentials (ERP) studies showed that the amplitude of the lateralized readiness potential was subthreshold, and the inhibition-related frontal N2 was absent in change trials as compared with stop trials, suggesting that different inhibitory mechanisms were involved. Therefore, whether re-engagement of another action as in change trials modifies the inhibitory process and its neural correlates is still an open question. It seems reasonable that, by obtaining both the spatial and temporal information of neural activities simultaneously, the above issue may be better addressed. For example, it is possible that the same inhibition-related cortical regions are activated by both stop and change inhibition but in the same or different processing latencies, or distinct cortical areas are activated but within the same or different time range.Furthermore, to gain more insight into the similarity and/or differences between stop and change inhibition, participants’ control modes should be examined. Generally, action control can be either reactive and automatic, or proactive and prepared. Reactive control is driven by the salient external signal, while proactive control is triggered by internal goals. Proactive control or advance preparation has already been shown to modulate the neural underpinnings of stop inhibition. Jahfari et al.’s fMRI study applied a probability-cued stop signal task, in which a cue indicating the likelihood of the following stop signal. A typical slowing effect of reaction time was revealed in the proactive go trials, reflecting participants’ adjustment of response strategy. Interestingly, the RIN was pre-activated by the cue, but its activity declined during the actual stop trials. Similar findings were observed by other fMRI studies. In parallel, some ERP studies also revealed that the amplitude of the frontal N2/P3 decreased with the increasing likelihood of a stop trial. As to change inhibition, little is known about how its behavioral and neural correlates were modulated by proactive control, and whether such modulations were similar with stop and global inhibition.By applying spatiotemporal analysis of ERP, the present study aimed to investigate whether the same inhibition-related neural network was activated in change inhibition as in classic stop inhibition. Based on the horse-race model, the ERP scalp map differences between successful inhibition and correct go trials were computed to best isolate the neural underpinnings of the inhibitory process. A statistical parametric mapping technique was used for ERP data points at all scalp electrodes and latencies, in order to statistically infer the precise spatiotemporal characteristics of functional neural networks involved in response inhibition. MethodsTwenty healthy undergraduate or graduate students (12 females; age range:23-26 years; mean age:25 years, Standard Deviation:1 years) served as the paid participants. All were right-handed and gave written informed consent.A go stimulus was consisted of three arrows against a black background. The central arrow was always gray. The flanker arrows were made gray to indicate no cue (33%), and were colored green, red or blue to indicate the cue (67%) of a subsequent double go, stop or change signal. The go, stop and change signal was a green arrow, a red ’×’ and a blue arrow in reversed direction respectively, requiring continuing, stopping and changing the ongoing response. In case that the participants waited for the signal intentionally, blank black screen was also added. In each trial, the go stimulus was present for 150 ms, followed by the 300 ms signal. The stimulus onset asynchrony (SOA) between the go stimulus and the signal was 175 ms (11.5%),200 ms (77%) and 225 ms (11.5%), interleaved randomly. The inter-trial interval was 1300 ms,1400 ms and 1500 ms with equal probability.The participants were instructed to press a left or right button as quickly as possible according to the direction of the central gray arrow, and to try their best to continue, stop or change the response once the go, stop or change signal was present. All trials were presented pseudo-randomly in each block, avoiding three successive same trials. They completed 20 blocks of 120 trials after a 10-min practice block. A 30 s rest was arranged after each block. The whole session lasted about 65 minutes.The Electroencephalogram (EEG) was continuously recorded at a sample rate of 1000 Hz with a 19-channel EEG amplifier (the Symtop Instrument(?)). The recording bandwidth was 0.5 to 100 Hz. The international 10-20 system was used with linked earlobes as reference. The electrode impedances were kept below 10 kΩ. The ERP epochs were extracted, including 100 ms pre-go-stimulus and 1000 ms post-go-stimulus. Trials contaminated with ocular, muscular or any other type of artifacts were detected with a ±70 μV threshold and were corrected automatically by principal component analysis. Only the correct trials with the 200 ms SOA from 240 trials of each response type (go, stop and change) in each cue type (un-cued and cued) condition were averaged. The baseline of ERP measurement was the mean amplitude of a 100 ms pre-go-stimulus interval. To facilitate the comparison with previous studies, the signal onset was defined as the ERP analysis onset.The traditional indices, SSRT in stop trials and CSRT in change trials were estimated for each subject in each separate cue condition according to the ’integration method’. The inhibition accuracies and reaction times were submitted to a two-way repeated measure ANOVA with the factors ’Response’(stop vs. change) and ’Cue’ (cued vs. un-cued). We statistically compared stopping and changing responses in an ANOVA with two factors ’Response’ (stop vs. change) and ’Cue’ (cued vs. un-cued). Furthmore, to describe the details in stopping and changing responses, the same two-way ANOVA was carried out to reveal the differences between stop vs. go and change vs. go, respectively. And two-tailed paired t-tests were carried out for further comparisons. The multichannel time series of F-value and t-value were used to generate topographical maps at each moment by an interpolation method of generalized cortical imaging technique. The significant threshold was set to 0.05.ResultsBehavior results Interestingly, a significant interaction between’Response’and ’Cue’was observed for inhibition accuracies (F(1,19)=51.45, P<0.01). Further analysis showed that the cue did not affect the stop inhibition accuracy (t(19)=1.53, P=0.14), but deteriorated the change inhibition accuracy (t(19)=4.82, P<0.01). In addition, accuracy of cued-stop trials was better than cued-change trials (t(19)=6.00, P<0.01), but the difference was insignificant for un-cued trials (t(19)=1.45, P=0.16).Inhibition reaction times indicated that change inhibition was modulated by the cue in a different way from stop inhibition, demonstrated by a significant interaction between ’Response’ and ’Cue’(F(1,19)=34.97, P<0.01). The SSRT in cued-stop trials, that was, with proactive control, was faster than that in uncued-stop trials (t(19)=7.38, P<0.01). However, such difference was weaker in the CSRT (1(19)=4.52, P<0.01). Besides, The SSRT was shorter than the CSRT in cued trials (t(19)=6.49, P<0.01), but they showed no significant difference in un-cued trials (t(19)=1.52, P=0.15).ERP results Importantly, the main effect of the ’Response’ in an ANOVA with the factors ’Response’(stop vs. change) and ’Cue’ (cued vs. un-cued) mainly differed in the right prefrontal region (F4, Fp2) and the bilateral parietal regions (C3, C4, P3, P4, Cz, Pz) during time period of 250～350 ms, corresponding with the behavioral inhibition reaction time. In the remaining time period (100-250 ms and 400-500 ms), there were limited significant electrodes concentrating on the occipital visual regions. Specifically, stop inhibition (the ’Response’ effect:stop vs. go) activated the right prefrontal region (F4, Fp2), while change inhibition (the ’Response’ effect:change vs. go) activated the bilateral frontoparietal region (F4, C3, C4, P3, P4, Fz, Cz, Pz). In addition, the comparisons between successful inhibition and correct go trials revealed common brain activities between stop and change inhibition. During the pre-response processes, the dorsomedial frontal region (Fz, Cz, Pz; 100-200 ms) and the right temporoparietal region (P4, T6, Pz; 200-250 ms) were activated similarly. Interestingly, during the post-response processes, the same right temporoparietal region (350～400 ms) and right prefrontal region (450～500 ms) were activated again. The SPM(t) showed the difference of bilateral parietal regions (P3, P4) existed in both uncued-and cued-stop vs. change conditions in 250-350 ms. In addition, the main effect of the cue lied in bilateral frontal region (F3, F4, C3, C4, Fz; 100-200 ms) and posterior fronto-parietal region (C3, C4, P3, P4, Pz; 200-300 ms). At last, there was few significant electrode in the interaction effect.ConclusionWe contrasted the ERP spatiotemporal characteristic between stopping and changing responses when participants completed a probability-cued stop-change signal task. At first, besides the rMFG and preSMA/SMA (RIN) in stop inhibition, the frontoparietal region (DAN) implemented changing response, supporting the view that change inhibition depended on a go 1 process, a stop process, and a go2 process, and simply activating a go2 process cannot stop the ongoing action. Second, similar regulation mechanisms by proactive preparation existed in stopping and changing responses. Third, both types of inhibition commonly recruited a series of control-related networks, including the dorsomedial frontal region (SN), the temporoparietal region (VAN) and the right prefrontal region (RIN). The repeated activation of the RIN/VAN in both response selection/inhibition and later response evaluation might also indicate the domain-generality and even process-generality feature. In summary, the findings here offered a new perspective on the neural mechanisms of response inhibition that there might be unique neural correlates involved in response inhibition processes under different behavior contexts.