The current study employed fMRI to investigate for the first time how the executive control network is modulated by SOA in a Stroop task. Of particular interest were 1) the neural effects of SOA on interference and facilitation effects; 2) response priming in negative SOAs; and 3) the effects of blocked SOA presentation on strategic orientation of attention. To briefly summarize the results that will be discussed at length in the next sections, four areas in the executive control network were sensitive to trial-specific SOA effects on interference. An overall ANOVA investigating the global, congruency-independent effects of SOA demonstrated that the RIFG was sensitive to response priming effects in negative SOAs, whereas the right superior parietal lobe (BA 7) was sensitive to attentional effects of blocked SOA presentation.
SOA modulation of interference and facilitation effects
Previous work with SOA manipulation in the Stroop task has documented varying amounts of interference and facilitation in each SOA. Behaviourally, the current data replicated prior observations that, in a manual task, peak interference occurred at the −200 ms SOA and was also significant at the 0 ms SOA [38–40]. Facilitation was similar between the −400 ms and −200 ms SOAs, which also replicates previous literature [38, 41, 43, 44]. Importantly, the effects of SOA on the magnitudes of interference and facilitation effects suggest differences in executive control strategies or recruitment in each SOA, which was investigated with fMRI for the first time in the current study.
Analyses indicated that areas of the brain participating in Stroop effects were not strongly modulated by SOA. However, three areas of the cognitive control network were sensitive to the effects of SOA on interference: the right superior parietal lobe (BA 7), RCZ (BA 8), and superior frontal gyrus (BA 9). Percent signal change analyses indicated that these areas showed greater BOLD change for the 0 ms and −200 ms SOAs than the −400 ms SOA, suggesting sensitivity to the magnitude of cognitive conflict. Specifically, this indicates that areas involved in performance monitoring (RCZ), conflict resolution (superior frontal gyrus/BA 9), and task-relevant attentional control (superior parietal lobule/BA 7) were most affected by interference during simultaneous stimuli presentation in the 0 ms SOA. As this SOA showed comparatively smaller behavioural interference effects, the stronger recruitment of these areas may reflect more efficient conflict processing. In contrast, the reduced activation in the −200 ms SOA illustrates that conflict resolution mechanisms were not engaged as efficiently, generating larger behavioural effects. In sum, this demonstrates that SOA significantly affected the recruitment of the cognitive control network during interference, as predicted.
When investigating Stroop, interference, and facilitation effects in each SOA individually, the 0 ms SOA showed a traditional recruitment of the executive control network for Stroop and interference effects, including the RCZ, LMFG, LIFG, and right superior parietal lobe, in line with previous literature [1, 3, 4, 12, 21, 23, 24]. The −200 ms SOA activated these same areas but to a lesser extent, again suggesting a less-efficient recruitment of cognitive control which generated increased behavioural interference. Therefore the two most cognitively-demanding SOAs activated a similar neural network, but the amount of activation was modulated by SOA.
Despite the relatively reduced activation in the conflict contrasts of the −200 ms SOA, when directly comparing the congruency conditions this SOA showed heightened ACC and LMFG (BA 9) activation in all congruencies, including the control condition. Previous research has reported that the ACC and prefrontal cortex are sensitive to the amount of conflict in a task [29] and that activation can be enhanced with task difficulty across the entire task rather than on a trial-by-trial basis [59]. The observed ACC and LMFG activation therefore suggests that cognitive control was enhanced throughout the −200 ms SOA block and in all congruencies due to the heightened cognitive demands in this SOA.
Overall, there was a disparity between the behavioural and neural effects: the 0 ms SOA elicited stronger brain activity yet experienced smaller behavioural conflict effects, while the opposite was true for the −200 ms SOA. This negative association of behavioural and neural responses has been reported previously [60–63] and suggests that successful cognitive control requires more extensive activation of the executive control network to reduce behavioural conflict effects.
The block-wide facilitation ANOVA also demonstrated that a cluster in the right inferior parietal lobe (BA 40) was sensitive to SOA effects on facilitation, and percent signal change analyses confirmed that this area showed greater signal change for the −200 and −400 ms SOAs compared to the 0 ms SOA. This mirrors the behavioural data, which showed large facilitation effects for the negative SOAs but virtually no facilitation for the 0 ms SOA, and also supports previous literature finding similar facilitation effects across negative SOAs [38, 41, 43, 44].
When extracting the percent signal change from this cluster in the facilitation effects, the results showed that the negative SOAs showed larger signal change in right BA 40/2 for the control condition than the congruent condition, whereas the 0 ms SOA showed similar levels of signal change for both congruencies. As mentioned in the Introduction, the parietal lobes are involved in top-down attentional control towards the task-relevant target or attribute [17–20]. It may be that in negative SOAs, pre-exposure of the control stimulus allows the semantic system to evaluate the stimulus and determine that the symbol string has no meaning, such that when the colour appears attention can be more efficiently directed to the target stimulus; this could explain the greater activation of the parietal lobe in response to the control stimulus compared to the congruent stimulus. In contrast, in the 0 ms SOA, simultaneous presentation of stimuli requires that the word be evaluated at the same time the colour is being processed, which may interrupt this efficiency of the parietal lobe. This is a tentative interpretation, however, and more research is needed to fully evaluate the neural correlates of SOA effects on facilitation. Nevertheless, the current results confirm that SOA manipulation does modulate facilitation effects, both behaviourally and in the brain.
In sum, the cognitive control network was sensitive to trial-specific effects of SOA on interference. Specifically, three regions of the network were most active in the 0 ms SOA, leading to correspondingly smaller behavioural interference effects. In contrast, the −200 ms SOA experienced comparatively less neural activation, suggesting less-efficient cognitive control which led to larger behavioural interference effects. This therefore demonstrates that SOA modulates the conflict-processing demands of the executive control network and suggests that short pre-exposure of the word in the −200 ms SOA disrupts the efficient processing of this system.
Response priming effects in negative SOAs
Appelbaum et al. [40] have suggested that negative SOAs create a response priming effect by pre-activating response selection, which generates larger behavioural interference and facilitation effects compared to the 0 ms SOA. This study explored the neural representation of these response priming effects in the −200 ms and −400 ms SOAs. The block-wide SOA analysis identified two regions that were modulated by the global effects of SOA: the RIFG and the right superior parietal lobe. As will be argued here, the RIFG was involved in response priming effects.
In the SOA-specific analyses, the Stroop and interference contrasts in the −200 ms SOA elicited RIFG activation to a greater extent than the 0 ms SOA; additionally, the −400 ms SOA activated the RIFG in the interference contrast. As mentioned in the Introduction, the RIFG has been implicated in response inhibition (i.e. inhibiting pre-potent motor responses, as in a no-go paradigm; [13–16]). The activation of this area in negative SOAs suggests its involvement in response priming effects; specifically, the fact that RIFG activation occurred in Stroop and interference contrasts in negative SOAs suggests that this area is involved in applying response inhibition after incorrectly-primed response selection.
To illustrate, in incongruent conditions the pre-exposed word primes (incorrect) response selection, which must then be overcome (via response inhibition mechanisms in the RIFG) to make a correct response to the colour. This would explain why the −200 ms SOA generates larger interference and facilitation effects: the need for response inhibition in incongruent conditions leads to longer incongruent RTs and consequently larger behavioural interference effects relative to the other conditions. In congruent conditions, however, the primed response preparation leads to faster RTs and increased behavioural facilitation effects. Response priming in the RIFG can therefore explain the larger interference and facilitation effects observed in the −200 ms SOA, as observed in the current data and in previous research [38–40].
In contrast, the −400 ms SOA generated large behavioural facilitation effects but no interference, which contradicts the proposal that response priming increases both interference and facilitation effects. In direct comparisons of the individual congruencies in the neural data, the −400 ms SOA showed more RIFG activation in the control condition compared to other SOAs. This suggests that the RIFG cannot purely reflect response inhibition in this SOA, because a response cannot be primed in the control condition as it does not contain semantic information.
Although the RIFG has been specifically implicated in response inhibition, previous investigations of the right posterior ventrolateral prefrontal cortex, which includes the RIFG, have indicated that this area is involved more generally in updating action plans, a function which includes, but is not limited to, response inhibition [13–16, 49, 64, 65]. The current data in the −400 ms SOA support this more general role of the RIFG in action updating. To illustrate, although the pre-exposure of the word primes response selection, the long pre-exposure may allow sufficient time to fully inhibit the motor response, as the word is a non-target stimulus: this would explain the lack of behavioural interference in the −400 ms SOA. If the primed response is fully inhibited, this would also predict a reduction in facilitation effects; however, facilitation is increased in this SOA. Therefore in addition to response inhibition, the RIFG may also perform more general action updating, as proposed by previous literature, which readies the motor system to make a response. If response preparation mechanisms are primed in a −400 ms SOA, upon subsequent colour presentation the system benefits from the convergent information in the congruent condition (therefore generating large facilitation effects) but the incongruent condition does not cause any additional conflict (resulting in little or no interference). In both congruencies, similar brain regions are active, which may explain the lack of neural differences between these conditions in the current −400 ms SOA data.
Thus, the current data can be explained by assuming that in the −200 ms SOA the RIFG is engaged primarily for response inhibition in incongruent conditions, as a result of the response priming effect, whereas in the −400 ms SOA the RIFG is involved in more general action updating. Importantly, in the −200 ms SOA the increased interference occurs because the response priming effect does not have enough time to be resolved. Lexical access occurs approximately 200 ms after word onset (e.g. [66, 67]), meaning that the colour appears at the same approximate time that semantic activation occurs in this SOA, leaving little extra time for stimulus suppression before conflict arrives. As a result, there is not enough time to overcome the response priming in the incongruent condition before the colour arrives, creating conflict and requiring the RIFG to perform response inhibition. In contrast, in the −400 ms SOA there is ample time for both semantic activation of the word and subsequent suppression of the primed response (via the RIFG), which explains the lack of behavioural interference. As well as inhibiting the primed response, the RIFG also performs a more general function of action updating, priming the system to make a motor response. This is a tentative explanation, as the RIFG has also been implicated in other cognitive functions such as reorienting [68], the detection of salient cues [15], and stopping motor actions [27]. However, the current data fits best with an explanation of the RIFG as involved in response inhibition and updating action plans [13–16, 49, 64, 65]. Therefore the current data supported Appelbaum et al. [40]’s proposal of response priming with word pre-exposure and also provided additional knowledge of how this mechanism functions in each SOA.
Effects of blocked SOA presentation
The current study additionally investigated whether blocked SOA presentation would create a global effect of attentional orientation such that the temporal predictability could be used to direct attention to the upcoming target stimulus [41, 48]. Such effects should be apparent across the entire block. As mentioned, the global (congruency-independent) analysis of SOA effects revealed two clusters of activation: the RIFG (which has been attributed to response priming effects of response inhibition and action updating [13–16, 49, 64, 65]) and the right superior parietal lobe (BA 7). As BA 7 is involved in top-down attentional control [17–20] this area may have been sensitive to attentional control effects resulting from the temporal predictability of blocked SOAs.
It was expected that attentional control effects would be most prominent in negative SOAs, as the word pre-exposure might act as a temporal cue that the target colour would soon appear. However, the percent signal change analyses illustrated that BA 7 was most active for the 0 ms SOA. This could suggest that when stimuli are simultaneously presented, attention to the relevant stimulus (the colour) is enhanced in order to facilitate response selection. For example, Egner and Hirsch [61] have suggested that conflict resolution proceeds via amplification of task-relevant attributes; enhanced attentional control in the 0 ms SOA may therefore indicate a strategy of directed attention towards the colour in order to overcome the effects of the distracting word stimulus. In contrast, the pre-exposure of the word in the negative SOAs may disrupt this process, leading to less activation in BA 7. Interestingly, the enhanced activation of this area for the 0 ms SOA mirrors the findings of Appelbaum et al. [48], who reported a larger Ninc in the 0 ms SOA with blocked SOA presentation.
Block-wide strategic attention effects were also identified in the direct comparisons of congruencies across SOA blocks: specifically, the −200 ms SOA showed more activation in the posterior cingulate. While being assigned to a number of cognitive roles, one function of the posterior cingulate is in anticipating the need to spatially allocate attention [69]. This could suggest an attentional priming effect in the −200 ms SOA such that the short pre-exposure of the word acted as a cue for attentional engagement. It is unclear why a similar effect did not occur in the −400 ms SOA; one possibility is that the longer word pre-exposure allowed ample time for the suppression of the word information, so attentional allocation was not prioritized.
In summary, the right superior parietal lobe was sensitive to the effects of blocked SOA presentation, demonstrating that attentional control was modulated by the global effects of SOA. These effects were enhanced for the 0 ms SOA, which could suggest that attentional mechanisms of conflict resolution were engaged during simultaneous stimulus presentation in the 0 ms SOA. In addition, SOA effects in the posterior cingulate in the −200 ms SOA could reflect an anticipation of attentional control.
In general, the fact that global effects of SOA were observed in regions involved in attentional control supports the proposal of strategic orientation of attention with blocked SOA presentation. However, these are ultimately tentative interpretations in light of the fact that a mixed-SOA comparison condition was not included in order to fully test the effects of strategic attentional orientation. For example, if blocked SOA affects attentional orienting towards the relevant dimension, this effect should be diminished with mixed SOAs, leading to smaller interference at negative SOAs as compared to blocked presentation. Therefore mixed SOA presentation might lead to very different effects, both in the behavioural data and in the neural recruitment of the attentional control network [48]. As this was the first study to use the Stroop SOA paradigm with fMRI, the comparison of blocked vs. mixed SOAs, and how this paradigm choice affects the recruitment of conflict processing mechanisms, requires further exploration.
Distractor suppression effects in negative SOAs
In addition to the reported effects of SOA on conflict processing, response priming, and attentional control, one additional finding was that the LIFG was generally more active across all congruencies for negative SOAs. Specifically, more LIFG activation was observed for the −400 ms SOA congruent and control conditions and the −200 ms SOA congruent condition as compared to the corresponding congruencies in the 0 ms SOA. Previous research has suggested that within the cognitive control network the LIFG performs suppression of irrelevant information (e.g. [12]); this finding of enhanced LIFG activation throughout the negative SOAs may therefore suggest a strategy of distractor suppression. For instance, at the time of word presentation in negative SOAs the word’s eventual congruency is unknown, as the colour has not yet appeared to cause conflict. Therefore the LIFG may be suppressing all pre-exposed information, as it is irrelevant to the task, in order to avoid potential conflict when the colour appears. Importantly, the control condition also elicited enhanced LIFG activation in negative SOAs, suggesting that this mechanism is neither conflict- nor linguistically-specific, but is a global strategy of task-irrelevant distractor suppression.
This proposal of a distractor suppression mechanism in negative SOAs suggests a strategy of proactive cognitive control, which draws a parallel to the dual mechanisms of control theory put forth by Braver and colleagues [70–72]. This theory proposes that cognitive control consists of two mechanisms: one reactive, which is a ‘late correction’ response that uses context information transiently to resolve conflict once it has occurred; and one proactive, which uses an ‘early selection’ strategy to actively sustain goal-relevant information and pre-emptively reduce control demands when conflict occurs. The fact that LIFG activation occurred across all congruencies in negative SOAs suggests a sustained activation of this structure, potentially through a mechanism of proactive cognitive control. In contrast, reactive control may be more characteristic of the 0 ms SOA, in which suppression must be activated anew on every trial. Although a tentative explanation, this proposal of distractor suppression by the LIFG suggests a proactive strategy employed to lessen the influence of the non-target stimulus and highlights the dynamic nature of the executive control system in response to various cognitive demands.