The interjection of goal-irrelevant information with goal-relevant information is referred to as cognitive interference. For instance, while trying to concentrate on your job, you may have to inhibit the habitual tendency to check your Facebook feed. Successful interference resolution depends on flexible cognitive control that suppresses goal-irrelevant inputs, while selecting and organizing goal-relevant inputs.

The multi-source interference task (MSIT) is a cognitively demanding well established paradigm for assessment of cognitive interference. In the MSIT, stimuli (e.g., the digits “1”, “2”, or “3”, or a letter “X” or a digit “0”) are organized into groups of three and participants are required to recognize a unique target among the three items under congruent and incongruent conditions [1, 2]. The spatial position of the unique target matches its correct button-press response in the congruent condition (e.g., “1XX” or “100”, the unique targets were “1” and the button was responded at the 1st position) and is in conflict with its correct button-press response in the incongruent condition (e.g., “331”, the unique targets were “1” but the button was responded at the 3nd position). In the MSIT, an interference effect is indexed by the difference in reaction time between incongruent and congruent conditions.

In an initial pilot imaging study of MSIT performance, Bush et al. reported that the dorsal anterior cingulate cortex (dACC) was reliably activated at either the individual- or group-level in the incongruent condition, compared with congruent condition, indicating that the dACC is important for interference processing [1]. Likewise, imaging studies with both youth and adults have shown increased activation in the dACC during MSIT performance and such dACC activity correlated with interference- and error-processing [3, 4]. Moreover, studies examining female twins [5] and subjects diagnosed with attention deficit hyperactivity disorder (ADHD) [6] have provided the evidence indicating that MSIT-related dACC activation may be attributable to genetic factors. Clinical studies have associated dACC dysfunctions with MSIT-related cognitive interference in patients with pediatric obsessive–compulsive disorder (OCD) [7], schizophrenia [8], and posttraumatic stress disorder (PTSD) [9, 10], suggesting that dACC abnormalities may contribute to cognitive difficulties.

Cingulate-frontal-parietal (CFP) cognitive-attentional networks have also been reported to be widely and significantly activated by MSIT [11, 12]. In a sample of younger and older adults, interference process on MSIT was associated with activation of the fronto-parietal and basal ganglia networks [13]. Patients with ADHD have been reported to show dysfunction of CFP cognitive-attention networks and abnormal ACC activity during interference processing [14, 15]. Also, relative to healthy controls, patients with chronic low back pain have been reported to have decreased MSIT-related activation in structures of the CFP network, including the dorsolateral prefrontal cortex, dACC, and superior parietal cortex [16]. Patients with OCD have been reported to exhibit functional abnormalities in the cingulate-frontal circuits, insular cortex and the putamen when performing the MSIT [17–19]. These findings could help to explain the inhibitory control deficits in OCD.

The MSIT interference effects on cortical activity in the aforementioned studies were variable, perhaps due to differences in study design and sample characteristics. Hence, a quantitative assessment of brain network activity in MSIT is needed. In the present study, we applied a meta-analytic approach to synthesize the published MSIT-fMRI studies with the aim of clarifying the locations of generators of interference processing during MSIT performance. We used effect-size signed differential mapping (ES-SDM) as the meta-analytic toolbox [20–22]. The ES-SDM is a reliable quantitative voxel-based meta-analytic method, which allow to integrate statistical parametric maps and peak coordinates. The meta-analytic method has to be superior to other coordinate-based meta-analytical methods owing to its ability to enable reconstruction of both positive and negative coordinate in the same map, leading to a signed differential map and keeping a special voxel from wrongly arising to be positive and negative at the same time [23]. It provides Jackknife sensitivity and heterogeneity analyses to further confirm the replicability of voxel-based meta-analytic findings. In this meta-analysis, we expected to demonstrate replicable brain activation patterns associated with MSIT interference processing within the dACC and in the CFP network.

To our knowledge, this is the first report of a voxel-based meta-analysis that identified MSIT-associated functional brain activation. Robustness analyses confirmed that the significance of two major activation clusters involving the dACC, MPFC, SMA, right insula, right IFG, and right PUT was reliable and robust during comparison between incongruent and congruent conditions.

Our findings are consistent with previous fMRI studies on MSIT indicating robust activation in the dACC, MPFC and SMA during interference processing when incongruent and congruent conditions are compared [1, 11, 28]. In the MSIT, subjects need to respond to the target while ignoring simultaneously presented unrelated information. Conflict is generated when the task-irrelevant information is incompatible with the target, thereby impeding the processing of task-relevant information. The dACC is recruited to monitor conflict. Higher dACC activity for incongruent trials has also been found in the flanker task [29, 30], Stroop task [30, 31], and Simon task [30, 32], providing further evidence for the supposition that the ACC is involved in detecting conflict in various interference tasks. Electrophysiological studies in both humans and monkeys have shown that dACC neurons firing rates increase during conflict processes and this increase is thought to promote ongoing behavioral adjustment [33–36]. Moreover, our findings are consistent with the conflict-monitoring hypothesis, which posits that increased ACC activity occurs when a high level of conflict is detected in incongruent trials, thereby recruiting top-down cognitive modulation to resolve the conflict and improve performance [37]. On the other hand, most imaging studies examining MSIT performance have found higher SMA activity in incongruent trials than in congruent ones and our meta-analysis results confirmed this conclusion. Anatomically, the SMA has ventral connections with the dACC [38]. Anatomically, the SMA has ventral connections with the dACC [38]. Thus, the SMA and dACC might work together to solve the interference challenge in the MSIT. Functionally, the SMA participates in movement planning and in action initiation and inhibition [38–40]. In other conflict tasks, researchers have also found that the SMA played a leading role in guiding the process of action-monitoring [41]. In a recent review of neuroimaging, electrophysiological, and stimulation studies of the SMA, Coull et al. proposed that the SMA may be involved in the cognitive development of a sensory representation of time, in addition to its aforementioned roles [42]. Altogether, the SMA is implicated in the process of deciding when to initiate an action or not. This possibility is supported by a prior electrophysiological study showing that neuronal activity in the SMA is associated with proactive and reactive behavioral control in a stop-signal task [43]. The SMA plays a proactive role in controlling arm movements to regulate motor readiness, and is involved in inhibiting arm movements in response to an unexpected stop signal. Accordingly, in the MSIT, after conflict is detected by the dACC, the SMA might be activated to plan movements and to establish flexible adaptive behavior.

An unexpected finding in our meta-analysis was a significantly active cluster involving the right IFG, right insula, and right PUT in comparisons between incongruent and congruent conditions. But previous studies employing the MSIT have found that CFP cognitive-attentional networks are reliably activated under these conditions. Although the result was not predicted, it is in agreement with a previously proposed role of the right IFG [44]. In a systematic review of a decade of literature regarding right IFG functions, Aron et al. found that the right IFG, together with one or more fronto-basal-ganglia network regions (including the PUT), may play a critical role in outright action-stopping in response to external stop or salient signals or internal goals [44]. The authors of other reviews of empirical electrophysiological and neuroimaging data from various inhibition paradigms (e.g., Stroop, Simon, and flanker tasks) have proposed that right IFG/basal ganglia pathways may contribute to goal-directed and habitual inhibition [45–47]. However, Bari and Robbins, who contributed a systematic summary of inhibition and impulsivity studies, suggested that the right IFG appears to be involved not only in the processing of response inhibition but also in the updating of goal-related plans of action [48]. According to these reviews, incongruent MIST trials produce more interference and inhibitory control than congruent trials due to the need to suppress distracting stimuli. Thus, interference may be resolved by engagement of the right IFG and PUT.

The insula is a commonly activated region in the go/no-go task, flanker task, and stimulus–response compatibility task, and insula activation has been shown to be related to interference resolution in each task [49]. Cai et al. examined causal interactions within core frontal-cingulate-parietal regions in the stop-signal task and the flanker task [50]. The strength of causal interaction between the right anterior insula and dACC was found to be greater under high cognitive control conditions than under low ones, and to be significantly associated with cognitive control ability indices in both the stop-signal task and the flanker task, suggesting that both the right anterior insula and dACC may be involved in cognitive control in various interference tasks. On the other hand, the insula and dACC are constituents of “salient network”, in which the right insula is thought to detect salient stimuli for recruitment of inhibitory control [51–56]. The salient feature is considered as a stimulus that is highlighted. The incongruent condition of MSIT, in which the target response is inconsistent with the target locations, has higher interference and stand out from the congruent one. Accordingly, in MSIT, the activation in right insula may involve in detecting interference and recruiting the interference-resolution.

In summary, our findings extend the results of prior MSIT studies, confirming that the dACC and prefrontal cortex are the main brain areas activated by MSIT performance. Our meta-analysis confirms cogently, for the first time, two robust activation clusters encompassing the dACC, MPFC, SMA, right IFG, right PUT, and right insula during MSIT performance. Compared to the congruent condition in the MSIT, the incongruent condition is characterized by more conflict and a greater need for cognitive control. On the basis of the functions of the aforementioned brain regions, we postulate that the right insula may send saliently relevant (high interference) signals to the dACC to be used to induce conflict monitoring, and to the SMA, right IFG, and PUT to be used for movement planning and inhibitory control, enabling goal-related flexible, adaptive behavior to be established. Hence, our findings indicate that a cingulate-frontal-striatum network and the right insula may serve as a critical brain circuit in interference resolution.