The term ‘cognitive control’ refers to a broad array of cognitive situations in which distracting information must be ignored, a habitual response must be overcome, or one must switch between varying mental sets. In the cognitive psychology literature, cognitive control is labeled more formally as ‘executive function’, a category which spans a number of cognitive functions such as working memory, response selection and/or suppression, and conflict detection and resolution. A vast amount of literature has been dedicated to understanding the cognitive and neural mechanisms of these various aspects of executive control, and with the emergence of neuroimaging technologies such as positron emission tomography (PET), magnetoencephalography (MEG) and functional magnetic resonance imaging (fMRI), this literature has grown enormously.
Previous work using fMRI has identified an extensive network of executive control consisting of regions across the prefrontal and parietal cortices that participate in a range of cognitive functions. For example, the rostral cingulate zone (RCZ; located along the borders of Brodmann areas (BAs) 6, 8, 32 and 24 in the medial frontal cortex) is thought to be involved in performance monitoring
. A subset of this region, the anterior cingulate cortex (ACC: Brodmann areas (BAs) 24/32) is clearly involved in executive control but is highly debated regarding its precise function, being implicated in processes such as conflict monitoring
[2–4], top-down regulation of conflict
, and anticipatory adjustments in control
, to name a few (see
[1, 7, 8] for reviews). Other frontal areas such as the middle and inferior frontal gyri, including the dorsolateral prefrontal cortex (DLPFC; BAs 9/46) are involved in conflict processing and regulation of executive control
[2–4, 7]. The left inferior frontal gyrus (LIFG) is believed to implement cognitive control via suppression of irrelevant semantic information
[9–12], while the right inferior frontal gyrus (RIFG) is involved in inhibitory control, specifically response inhibition
[13–16]. The inferior (BAs 39/40) and superior (BA 7) parietal lobes are involved in top-down visuospatial control of attention towards the task-relevant target or attribute
[17–20]. Other areas of the prefrontal cortex such as the premotor (BA 6) and frontopolar (BA 10) cortices are also involved, as are subcortical structures such as the thalamus and caudate (e.g.
[1, 21, 22]; see
[23, 24] for meta-analyses).
Although a number of alternative theories exist regarding the precise function of these structures, especially the ACC (e.g.
[6, 7, 25]), DLPFC (e.g.
), and RIFG (e.g.
[15, 27]), these areas are reliably activated for a spectrum of executive control functions, including working memory, cognitive flexibility, vigilance or sustained attention, and – importantly for this study – inhibition of prepotent behaviours and the management of cognitive conflict
. The activation and recruitment of this executive control system is also affected by various task parameters, such as the context, magnitude, and nature of cognitive conflict (e.g.
[19, 28–30]). This malleability of the executive control network highlights its dynamic, moment-to-moment recruitment of different conflict processing strategies. The current study specifically explored how this network is modulated by SOA manipulation in the Stroop task.
This paradigm presents a colour word printed in coloured ink, and asks subjects to ignore the word and respond to its ink colour
. Interference arises in incongruent conditions (e.g. blue printed in red ink, correct response “red”) due to the conflicting semantic and response information, and longer reaction times (RTs) arise because the automatic process of word reading must be overcome in order to name the colour. The Stroop task recruits the canonical executive control network, generating stronger activation for incongruent trials (e.g.
[24, 32–35]). Many variations of the Stroop task have been employed with fMRI to investigate the precise function of executive control structures (e.g.
[19, 36, 37]).
One notable variation is stimulus onset asynchrony (SOA) manipulation, which spatially separates the colour and word stimuli (e.g. a coloured rectangle surrounding the word) and presents them at different times in order to gain temporal information on colour and word interference. A ‘negative SOA’ presents the irrelevant stimulus (e.g. the word) before the relevant target stimulus (the colour) at a specific interval. For example, a negative 200 ms SOA (‘-200 ms SOA’) pre-exposes the word for 200 ms before the colour appears. A ‘0 ms SOA’ presents the word and colour simultaneously, as in a traditional Stroop task. Typically, the strongest interference effects (incongruent minus control) occur at −200 ms to 0 ms SOAs
[38–41]. Interference is decreased, but remains significant, at negative SOAs out to −400 ms. Facilitation effects (control minus congruent) are generally found for negative SOAs and not for positive SOAs beyond +200 ms
[38, 39, 41, 42], but facilitation typically does not differ across negative SOAs
[38, 41, 43, 44]. At the 0 ms SOA, some researchers report significant facilitation effects
 and some do not
[38, 39, 41, 42].
The Stroop task is traditionally administered with a verbal response, in which participants name the colour of the ink aloud. A manual response modality, which uses a button-press instead of a vocal response, results in decreased (but still significant) interference effects
, as well as faster reaction times overall
[36, 45–47]. SOA has been found to elicit different patterns of interference effects depending on the response modality: vocal responses elicit the maximum amount of interference at a 0 ms SOA (e.g.
[38, 41]), whereas manual responses shift the peak of interference to the −200 ms SOA due to the faster manual response time
[39, 40]. In terms of facilitation effects, manual and vocal responses appear to have similar effects on facilitation magnitude across SOAs, with significant facilitation effects for negative SOAs and no facilitation for positive SOAs
SOA variation has proven to be a useful manipulation of the Stroop task because it provides temporal information about the speed of processing of the two conflicting stimulus dimensions. To investigate these temporal effects further, recent studies have employed Stroop SOA manipulation with electroencephalography (EEG), to investigate how pre- or post-exposure of the word affects conflict-related ERP components
[40, 42, 48]. These studies have demonstrated that the onset and duration of the Ninc or N450 (an ERP component thought to be indicative of conflict detection) is modulated by SOA, and that this component is sensitive to conflict across a variety of task designs and conflict demands. However, the current study is the first to explore the neural effects of Stroop SOA manipulation (using −400 ms, -200 ms, and 0 ms SOAs) on the activation and recruitment of the executive control network using fMRI. Based on prior research, this study addressed three specific cognitive aspects of SOA manipulation.
SOA effects on neural conflict and facilitation
First, the current study explored how the executive control network in the brain is modulated by conflict and facilitation effects in each SOA. Behaviourally, each SOA generates different magnitudes of interference and facilitation, with maximal interference at simultaneous presentation or short word pre-exposure (i.e. 0 ms or −200 ms) and significant facilitation at negative SOAs
[38–40, 42]. Furthermore, the Ninc ERP component is sensitive to conflict across a variety of task designs and conflict demands
[40, 42, 48]. This modulation of conflict and facilitation effects suggests the participation of different cognitive control mechanisms for each SOA. The primary aim of the current study was therefore to explore how these ‘trial-specific’ effects of SOA affected the activation of the executive control network.
Overall, typical executive control areas of the prefrontal cortex were expected to be elicited by incongruency in the 0 ms SOA (as this was analogous to a traditional Stroop task), such as the RCZ, left middle/medial frontal gyrus (LMFG), and LIFG (e.g.
[1, 3, 12, 21–23]), as well as parietal regions such as the left angular gyrus
[12, 32, 49, 50] and the inferior/superior parietal lobe
[17–20]. Activation in these areas was also expected for the −400 ms and −200 ms SOAs, although with potentially different extents and/or strengths of activation compared to the 0 ms SOA. For example, the executive control network has demonstrated stronger activation in the presence of more conflict (e.g.
), so increased behavioural interference in the −200 ms SOA may be reflected in stronger neural recruitment of these areas.
Response priming effects in negative SOAs
The second topic addressed in the current study regarded the effects of response priming in negative SOAs. Appelbaum et al.
 have proposed that in negative SOAs, word pre-exposure creates a priming effect by pre-activating response selection. In congruent conditions this accelerates processing time because the subsequently-presented colour matches the pre-activated information, leading to larger behavioural facilitation effects. In contrast, incongruent conditions require more conflict control to overcome or inhibit the primed response, increasing behavioural RTs and interference effects. Increased interference and facilitation effects have been previously documented at the −200 ms SOA, in line with this proposal of response priming effects
[39, 40]. The current study sought to establish the neural correlates of response priming effects in negative SOAs.
Response priming effects were expected in executive control areas linked to response preparation, such as the DLPFC
 or supplementary and cingulate motor areas
. This activation was predicted to be stronger in the −200 ms SOA, and potentially also the −400 ms SOA, compared to the 0 ms SOA. Furthermore, if the increased behavioural interference in the −200 ms SOA arises from the need to overcome the primed response in incongruent conditions, evidence of response priming may also be observed in areas linked to response inhibition, such as the RIFG
Attentional effects of blocked SOA presentation
Finally, the third aspect of SOA manipulation investigated in this study concerned the effects of blocked SOA presentation. Appelbaum et al.
 have recently observed different patterns of interference for blocked and mixed SOA presentation. Specifically, temporally-predictable SOAs, as in blocked presentations, may lead to a strategic orientation of attention which could modulate the amount of conflict experienced. In their EEG data, Appelbaum et al.
 demonstrated that although the Ninc tracked the onset of conflict across SOA manipulation, a larger Ninc component occurred in the 0 ms SOA when SOAs were blocked, whereas when SOAs were randomized a larger Ninc occurred in the −200 ms SOA. In blocks of negative SOAs, the pre-exposed word may have acted as an alerting cue for the upcoming target information, prompting participants to use this cue to strategically orient their attention towards the target stimulus. In contrast, in the 0 ms SOA this strategy could not be used, leading to larger interference effects. Therefore Appelbaum et al.
 proposed that the temporal predictability of blocked SOAs encourages an attentional orientation strategy. On the other hand, Roelofs
 has also investigated this issue of blocked versus mixed SOA presentation using a behavioural paradigm and reported that, although overall RTs were affected, no difference in interference patterns occurred between SOA presentation methods. This argues against such a temporal predictability effect, but it may be that the electrophysiological technique used in Appelbaum et al.
 was more sensitive to strategic attentional effects. The current study therefore also investigated global attentional effects of blocked SOA presentation.
If blocked SOA presentation engages strategic attentional processes, such block-wide SOA effects should be observable in all congruency conditions. The current study investigated these global (i.e. block-wide and conflict-independent) effects of strategic attentional control by first collapsing over congruencies and comparing SOAs, as well as comparing congruency conditions across SOAs (e.g. -400 ms control vs. 0 ms control). Global SOA effects on attentional orientation were expected in areas involved in top-down attentional control such as the right parietal lobe, specifically the angular gyrus (BA 40) and superior parietal lobe (BA 7;
[17–20]). Specifically, if subjects use the pre-exposed word in negative SOAs as a temporal cue, activation in these attentional control areas should be enhanced in the −200 ms and −400 ms SOAs compared to the 0 ms SOA.
In summary, the current study employed a Stroop task with negative SOA modulation in fMRI to explore how SOA affects the recruitment and performance of the executive control network. Of specific interest were 1) the effects of SOA on Stroop, interference, and facilitation effects in the brain; 2) response priming effects in negative SOAs; and 3) global effects of blocked SOA presentation on attention.