Reconceptualizing the Relationship between Anxiety, Mindfulness, and Cognitive Control

Abstract

Prior research has provided initial support for the claim that cognitive control mediates the relationship between anxiety and mindfulness; however, findings are often inconsistent. In this review, we argue that the inconsistency may be due to a lack of both conceptual and methodological precision in terms of how anxiety, cognitive control, and mindfulness are operationalized and assessed, and that this imprecision may be a critical source of study confounds and ambiguous outcomes. We unpack this argument by first decomposing anxiety, cognitive control, mindfulness, and relevant experimental paradigms into key dimensions in order to develop a non-unitary, multi-dimensional taxonomy of these constructs. Subsequently, we review and reinterpret the prior experimental literature, focusing on studies that examine the relationship between anxiety and cognitive control, mindfulness and cognitive control, and the three-way relationship between anxiety, mindfulness, and cognitive control. Across the reviewed studies, there was great variation in the dimensions being examined and the behavioral and/or neural measures employed; therefore, results were often mixed. Based on this review of literature, we propose a conceptually and methodologically precise framework from which to study the effects of mindfulness on cognitive control in anxiety. The framework theoretically aligns anxiety dimensions with specific mindfulness states and interventions, further suggesting how these will impact specific cognitive control dimensions (proactive, reactive). These can be assessed with experimental paradigms and associated behavioral and neural metrics to index the relevant dimensions with high precision. Novel experimental studies and tractable research designs are also proposed to rigorously test this theoretical framework.

1. Introduction

Anxiety is broadly characterized by a range of cognitive, emotional, and somatic symptoms (Burdwood et al., 2016), including worry or future-oriented repetitive thinking, muscle tension, restlessness, fatigue, hypervigilance, shortness of breath, alterations in heart rate, dizziness, and sweating (Craske et al., 2009; Szuhany & Simon, 2022). Research has focused on elucidating the neurocognitive mechanisms that may account for anxiety symptoms. For example, it has been suggested that attentional biases towards threat-related information (e.g., hypervigilance, difficulty disengaging attention from threat, and avoidance of threat) may contribute to the etiology and maintenance of anxiety (Gupta et al., 2019). However, influential frameworks such as Attentional Control Theory (Eysenck et al., 2007) have suggested that anxiety is associated with more general deficits in attentional and cognitive control (Berggren & Derakshan, 2013). Indeed, both electrophysiological and neuroimaging studies have provided strong evidence for altered processing during cognitive control paradigms in anxiety (Paulus, 2015).

Given the role of attentional and cognitive control deficits in anxiety, interventions that target and improve attentional and cognitive control may be particularly effective for improving anxiety symptoms (Zhang & Xiang, 2023). Mindfulness-based interventions (MBIs) may fit this characterization well, given that mindfulness has been popularly defined as “paying attention in a particular way: on purpose, in the present moment, and nonjudgmentally” (Kabat-Zinn, 1994). Moreover, meta-analytic evidence suggests that MBIs are particularly effective at reducing certain anxiety symptoms in clinical (Fumero et al., 2020; Ghahari & Mohammadi-Hasel, K. Malakouti, S. K. Roshanpajouh, 2020) and healthy (Chiesa & Serretti, 2009; Khoury et al., 2015) populations. Likewise, neuroimaging studies have suggested that mindfulness meditation improves attentional and cognitive control via changes in interactions between frontoparietal and default mode brain networks (Malinowski, 2013; Y. Y. Tang et al., 2015). Taken together, findings suggest that mindfulness may mitigate anxiety through an improvement in attentional and cognitive control.

Prior research has provided initial support for the claim that attentional and cognitive control mediate the relationship between anxiety and mindfulness (Jankowski & Bąk, 2019; MacDonald & Olsen, 2020); however, findings are often inconsistent. In this review, we argue that the inconsistency may be due to a lack of both conceptual and methodological precision in terms of how anxiety, cognitive control, and mindfulness are operationalized and assessed, and that this imprecision may be a critical source of both study confounds and ambiguous outcomes. We unpack this argument by first decomposing anxiety, attentional and cognitive control, mindfulness, and relevant experimental paradigms into key dimensions (Section 2). In subsequent sections, we review and reinterpret the prior experimental literature, focusing on studies that examine the relationship between anxiety and cognitive control (Section 3), mindfulness and cognitive control (Section 4), and the three-way relationship between anxiety, mindfulness, and cognitive control, (Section 5). For these sections, rather than a systematic review or quantitative meta-analysis, we conducted a narrative review that aimed to illustrate the current state of the relevant literature; consequently, we did not attempt to be exhaustive in identifying studies. Nevertheless, potentially relevant studies were identified through literature searches conducted in Google Scholar and PubMed using various combination of the following terms, with no limit on publication date: “anxiety”, “mindfulness”, “proactive control”, “reactive control”, or “cognitive control”. Studies had to be peer reviewed, and care was taken to minimize any potential biases in the papers selected to be included for discussion. Based on this review, we then propose a conceptually and methodologically precise framework from which to study the effects of mindfulness on cognitive control in anxiety (Section 6). Critically, this framework theoretically aligns anxiety dimensions with specific mindfulness states and interventions, further suggesting how these will impact proactive and reactive dimensions of cognitive control. We discuss how these dimensions can be assessed with experimental paradigms and associated behavioral and neural metrics to index the relevant dimensions with high precision.

2. A Dimensional Perspective on Anxiety, Cognitive Control, Mindfulness, and Experimental Paradigms

Here, we highlight key dimensions of anxiety, attentional and cognitive control, mindfulness, and experimental paradigms in order to develop a non-unitary, multidimensional taxonomy applied to these four constructs (see Figure 1). It is important to note that the list of dimensions discussed herein is non-exhaustive; indeed, these four constructs may be decomposed into various alternative dimensions. However, our primary goal here is to highlight the importance of acknowledging the multifaceted nature of these constructs when conducting research in each domain.

Figure 1. A non-unitary, multidimensional taxonomy of anxiety, cognitive control, mindfulness, and experimental paradigms.

A summary of the decomposition of anxiety, cognitive control, mindfulness, and experimental paradigms into key dimensions.

2.1. Anxiety Dimensions

Treatment of anxiety as a unitary construct has led to mixed and contradictory findings (Sharp et al., 2015), likely because anxiety can be conceptualized in many ways. For example, anxiety is often conceptualized as a clinical disorder. Anxiety disorders are highly prevalent and one of the leading causes of disease burden worldwide (Konnopka & König, 2020). Anxiety disorders, such as generalized anxiety disorder (GAD), social anxiety disorder (SAD), and specific phobia, are associated with hypervigilance to potential threat in preparation for future danger, cautious or avoidant behaviors (American Psychiatric Association, 2013), and delayed disengagement from threat (Amir et al., 2003).

However, anxiety can also be conceptualized as a trait, particularly when referring to subclinical (or non-diagnosed) populations. Trait anxiety is defined as an individual’s tendency to appraise situations as threatening, avoid anxiety-provoking situations, and demonstrate high baseline physiological arousal (Elwood et al., 2012; Knowles & Olatunji, 2020). It has been proposed that trait anxiety, a stable personality characteristic, can also serve as an important vulnerability factor for the development of anxiety disorders (Andrews, 1991; Calvo & Cano-Vindel, 1997; Elwood et al., 2012; Sandi & Richter-Levin, 2009). In a model outlining the proposed pathway from high trait anxiety to the development and maintenance of anxiety-related psychopathology, Knowles & Olatunji (2020) propose that trait anxiety drives cognitive biases, such as increased attention toward threat. These cognitive biases lead to increased negative emotional experiences and maladaptive safety behaviors which decrease immediate anxiety but maintain anxiety symptoms over the long term.

Anxiety can also be conceptualized as a state. Whereas trait anxiety is a more stable personality feature, state anxiety has been defined as aversive, conscious feelings of nervousness, tension, apprehension, and worry experienced in the immediate moment, which are associated with arousal of the autonomic nervous system (Spielberger et al., 1983). Importantly, both trait and state anxiety can be decomposed into dimensions. In line with a more transdiagnostic approach, it has been argued that anxious apprehension and anxious arousal are psychologically and physiologically distinct dimensions of trait anxiety that are associated with distinct neural mechanisms (Burdwood et al., 2016; Heller et al., 1997; Nitschke et al., 1999; Sharp et al., 2015). Anxious apprehension is marked by a propensity to engage in repetitive negative thinking, which can also be thought of as an enduring pattern of state worry, whereas anxious arousal consists of an enduring pattern of hypervigilance, sympathetic nervous system hyperarousal to mild stressors, and state fear (Sharp et al., 2015). Although previous research has used terms such as “worry” synonymously with anxious apprehension and “panic” or “fear” synonymously with anxious arousal, these terms better represent state anxiety phenomena and do not sufficiently characterize the neuropsychological patterns that comprise trait anxiety dimensions of anxious apprehension and anxious arousal (Sharp et al., 2015). Thus, we classify state “worry” and “panic/fear” as dimensions of state anxiety. The anxiety dimensions discussed herein are depicted in Figure 1.

2.2. Attentional and Cognitive Control Dimensions

The terms “attention control”, “cognitive control”, and “executive function” are often used interchangeably to reflect the collection of processes involved in generating, maintaining, and updating task goals, as well as the way in which these task goals are used to modify attentional biases (Burgoyne & Engle, 2020; Diamond, 2013; Gratton et al., 2018). However, “attention control” is often used when describing within-trial or trial-by-trial variations in the direction of attention in response to changing task demands, while “cognitive control” is often used to refer to the set of mechanisms that are deployed in tasks that demand flexibility at one or multiple levels (Badre, 2024). In contrast, “executive function” may be more broadly construed as also implicating long-term goal representation and nesting or interleaving of goals set for different time scales (Gratton et al., 2018). Recognizing that these terms have tended to be used semi-interchangeably, hereafter, we will primarily utilize the term cognitive control to refer to this set of processes, since it better reflects our theoretical preferences.

Research has suggested that cognitive control should not be viewed as a unitary process, but rather as an emergent function born of the interaction of a number of elementary components (dimensions). Indeed, extant theoretical models have proposed different, though often partially overlapping, subdivisions (for a review, see Gratton et al., 2018). For example, in a classic paper, Miyake et al. (2000) used a structural equation modeling approach to demonstrate that executive function (EF) could be parsed into three components: updating (constant monitoring and rapid addition/deletion of working memory contents), shifting (switching flexibly between tasks or mental sets), and inhibition (deliberate overriding of dominant or prepotent responses). In a newer, unity/diversity framework (Miyake & Friedman, 2012), however, each EF ability (e.g., updating) can be decomposed into what is common across all three EFs, or unity (termed common EF), and what is unique to that particular ability, or diversity (e.g., updating-specific ability). As such, the current unity/diversity framework focuses on the elements that may more cleanly map onto the underlying cognitive processes (common EF, updating-specific, and shifting-specific abilities) and seeks to specify their underpinnings. It is important to note that the inhibition-specific component is absent because, once the unity (common EF) is accounted for, there is no unique variance left for the inhibition-specific factor (Miyake & Friedman, 2012).

As another example, the Dual Mechanisms of Control (DMC) framework (Braver, 2012) proposes that a core component of cognitive control function is its inherent variability, which manifests intra-individually (i.e., state related or task related), inter-individually (i.e., trait related), and between-groups (i.e., related to changes in brain function or integrity in different populations). The DMC framework proposes that this variability may arise from qualitative distinctions in temporal dynamics across two modes of cognitive control: proactive and reactive. In the proactive control mode, goal-relevant information is actively maintained in a sustained manner, prior to the occurrence of cognitively demanding events, to optimally bias attention, perception, and action systems in a goal-driven manner. This form of “early selection” is thought to be associated with sustained and/or anticipatory activation of the lateral prefrontal cortex (PFC). The reactive control mode, however, is mobilized only as needed and in a just-in-time manner, such as after a high interference event is detected. This “late correction” mechanism is thought to be reflected in transient activation of the lateral PFC, along with a wider network of additional brain regions, including both the fronto-parietal (FPN) and cingulo-opercular (CON) networks. Therefore, while proactive control relies upon the anticipation and prevention of interference prior to its occurrence, reactive control relies upon the detection and resolution of interference following its onset. As the two modes are associated with complementary advantages and limitations, successful cognition likely depends upon some mixture of both strategies (Braver, 2012).

In the unity/diversity framework, the Common EF factor reflects individual differences in the ability to maintain and manage goals and use those goals to bias ongoing processing (Friedman & Miyake, 2017; Miyake & Friedman, 2012). Friedman & Miyake (2017) note that the conceptualization of Common EF shares many similarities with proactive control in the DMC framework¹. While the DMC framework focuses on the temporal dynamics of cognitive control, Braver (2012) also suggests that there may be stable individual differences in the tendency to adopt a proactive mode based on cognitive abilities; thus, Common EF may be related to stable biases in the balance between proactive and reactive control across tasks.

The role of PFC mechanisms within the unity/diversity framework also bears similarities to how these mechanisms are conceptualized within the DMC framework. Herd et al. (2014) presented a biologically based neural network model of task switching that showed individual differences in two separable components of switch costs – Common EF and Shifting-Specific. The authors proposed that individual differences in the Common EF factor in large part reflect goal maintenance: the ability to actively maintain goals and goal-related information, often in the face of interference, and to use these goals to bias ongoing processing (i.e., top-down attention and task control). In contrast, they proposed that the Shifting-Specific component reflects flexibility: the ease of transitioning to new task set representations. Results were consistent with the hypotheses that variation specific to task switching (i.e., Shifting-Specific) may be related to uncontrolled, automatic persistence of goal representations, whereas variation general to multiple EFs (i.e., Common EF) may be related to the strength of PFC representations and their effect on processing in the remainder of the cognitive system². However, within the Herd et al. (2014) model, the temporal dynamics of PFC representations are relatively unspecified in relationship to the DMC framework. Accordingly, we first decompose cognitive control into the unity/diversity framework’s Common EF, updating-specific, and shifting-specific components. Subsequently, we consider proactive and reactive control as a potential further subdivision associated with Common EF, though this is not a strong theoretical commitment (indicated through dashed lines connecting the two levels). The dimensions of cognitive control discussed herein are depicted in Figure 1.

2.3. Mindfulness Dimensions

Similar to anxiety and cognitive control, mindfulness has been conceptualized in terms of distinct dimensions (Lin, Tang, et al., 2022). These dimensions map onto the distinct ways of operationalizing mindfulness for research investigation: (1) as a dispositional trait (i.e., an individual’s natural propensity to be mindful, considered to be a stable personality characteristic that varies across people), (2) as a psychological state (i.e., the degree that an individual is mindful at a singular time point, subject to temporal and situational variation), (3) as a skill (i.e., the capacities or level of expertise that an individual has acquired as a function of sustained and long-term mindfulness practice), and (4) as a training method or intervention (i.e., involving prospective practice, during which contemplative techniques are deliberately applied to cultivate the attitudinal qualities of mindfulness).

Mindfulness training can also be conceptualized in different ways, ranging from formal meditation practice (e.g., meditation retreats) to mindfulness-based interventions, or MBIs (Lin, Tang, et al., 2022). Further, MBIs themselves can be decomposed. Commonly employed and researched MBIs, such as Mindfulness-Based Cognitive Therapy (MBCT; Segal et al., 2013) and Mindfulness-Based Stress Reduction (MBSR; Kabat-Zinn, 1990), incorporate a variety of distinct practices and train a variety of different cognitive and affective skills that are presented together under the umbrella term “mindfulness” (Britton et al., 2018). Standard MBIs intermix two distinct practices: focused attention (FA) and open monitoring (OM); yet, the integrated delivery of these two practices may impede understanding of their practice-specific effects or mechanisms of action within MBIs (Britton et al., 2018). While FA meditation entails voluntary focusing of attention on a chosen object, OM meditation involves non-reactive monitoring of the content of experience from moment to moment (Lutz et al., 2008).

Recently, however, studies have “dismantled” MBCT into the separate core practices of FA and OM in order to assess the mechanisms of each practice independently and in comparison to the combination (i.e., MBCT) in randomized controlled trials (Britton et al., 2018; Cullen et al., 2021). The results from such dismantling studies have shown that FA practice engages mechanisms related to attention control, while OM practice engages mechanisms related to emotional non-reactivity (Britton et al., 2018). It is important to note that FA and OM can be parsed not only as specific components of mindfulness training, but also as psychological states that can be elicited through engagement with mindfulness practices (Lin, Tang, et al., 2022). The dimensions of mindfulness discussed herein are depicted in Figure 1.

2.4. Experimental Paradigms

Cognitive control is frequently assessed using tasks, or a particular set of experimental paradigms. These experimental paradigms are designed to manipulate the contribution of distinct cognitive control processes to task performance while controlling for sensory, memory, and/or motor contributions (Gratton et al., 2018). Many tasks have been used to assess various aspects of cognitive control. While various attentional and cognitive control tasks exist, they are often not highly correlated with one another, which may reflect the multifaceted nature of attentional and cognitive control (P. G. Williams et al., 2017), as described above. For example, the Stroop task is a measure of selective attention and inhibitory control, the AX-Continuous Performance Test (AX-CPT) is a measure of context processing and inhibitory control, cued task-switching paradigms have been used to assess multitasking, and the Sternberg item recognition paradigm has been used to assess working memory (R. Tang et al., 2023). The Eriksen flanker task (Eriksen & Eriksen, 1974), a classic measure of attention control (Burgoyne et al., 2023), has also been used, among many others.

One useful classification of cognitive control tasks has emerged from the Miyake et al. (2000) study described in the previous section, in which the authors used factor analyses to examine individual differences in performance across a range of cognitive control tasks. This resulted in three major task groupings: Inhibition tasks, which require avoiding a dominant or prepotent response (e.g., antisaccade, stop-signal, and Stroop tasks); Updating tasks, which require continuously updating the contents of working memory, adding new information and removing no-longer-relevant information (e.g., keep track, letter memory, and spatial 2-back tasks), and Shifting tasks, which require switching between two subtasks according to a cue that appears before each trial (e.g., number-letter, color-shape, and category-switch tasks) (Friedman & Miyake, 2017; Miyake et al., 2000; Miyake & Friedman, 2012). In the newer unity/diversity framework described in Miyake & Friedman (2012) and Friedman & Miyake (2017), however, unity is captured with a common factor (Common EF) explaining shared variance in all nine EF tasks studied, while diversity is captured by orthogonal factors (Updating-specific and Shifting-specific) that capture remaining correlations among the updating and shifting tasks, respectively, once the Common EF variance is removed.

The DMC framework described in the previous section decomposes cognitive control into proactive and reactive control modes (Braver, 2012); thus, another potentially useful classification of cognitive control tasks may be to group tasks based on modes they are best suited to assess. However, Braver et al. (2021) notes that most studies of proactive and reactive control have focused on a single task or a limited cognitive domain, thus precluding a rigorous test of the validity and the domain-generality of cognitive control modes across multiple cognitive domains. In order to address this issue, Braver and colleagues developed the Dual Mechanisms of Cognitive Control (DMCC) task battery, which includes four cognitive tasks (Stroop, AX-CPT, Cued Task-Switching, and Sternberg WM), one for each of four distinct cognitive domains (selective attention, context processing, multitasking, and working memory, respectively), that are theoretically optimized to capture variability in proactive and reactive control (Braver et al., 2021; R. Tang et al., 2023). Specifically, the battery includes three variants of each task representing (1) a baseline condition that does not bias the adoption of proactive or reactive control; (2) a condition that shifts individuals towards proactive control; and (3) a condition that independently engages the reactive mode of control.

While these tasks provide a powerful experimental methodology from which to assess cognitive control, it is important to note that they are not without their limitations. For example, an unavoidable quality of EF tasks is task impurity; EFs involve controlling lower-level processes, so EF tasks must include nonexecutive processes that could influence performance in addition to the EF of interest (Friedman & Miyake, 2017). Likewise, the tasks in the DMCC battery do not provide an exhaustive set of those potentially useful for examining proactive and reactive control. Furthermore, tasks may be used on their own, in behavioral studies, or concurrently with neural measures such as EEG and fMRI, which may lead to differing results across studies. The experimental paradigm dimensions discussed herein are depicted in Figure 1.

3. Anxiety and Cognitive Control

Many theories relating anxiety to attentional and cognitive control have been proposed, but one of the most influential frameworks is Attentional Control Theory (ACT; Eysenck et al., 2007), which accounts for the effect of trait anxiety on cognitive performance (Eysenck et al., 2023). The ACT framework proposes that anxiety impairs efficient functioning of the goal-directed attentional system and increases the extent to which processing is influenced by the stimulus-driven attentional system. Along with decreasing attentional control, anxiety increases attention to threat-related stimuli, whether internal (e.g., worrisome thoughts) or external (e.g., threatening task-irrelevant distractors). In ACT, high levels of trait anxiety are postulated to impair processing efficiency (the relationship between the effectiveness of performance and the effort or resources spent in task performance) more than performance effectiveness (quality of performance) because of the use of compensatory strategies (e.g., enhanced effort; increased use of processing resources) (Eysenck et al., 2007, 2023). A meta-analysis of self-report and behavioral studies has provided broad support for ACT (Shi et al., 2019). Additionally, Eysenck et al. (2023) found support from neuroimaging (EEG and fMRI) studies for a proposed neurofunctional account of ACT which emphasizes the differential involvement of three brain networks—FPN, CON, and default mode network (DMN)—during task performance by individuals with high and low trait anxiety or worry. The ACT framework is not explicitly aligned with the DMC framework. Nevertheless, as discussed further below, it is possible to draw a rough correspondence between trait anxiety-related impairment in functioning of the goal-directed processing system within ACT as being consistent with a shift away from proactive control, from the perspective of the DMC framework.

To first better flesh out this point, in this section, we review studies that link dimensions of anxiety and cognitive control (specifically, proactive and reactive control dimensions) to demonstrate how different conceptualizations of these constructs, as well as the experimental paradigms used to study them, may impact findings and their interpretation. Within the DMC framework, it has been proposed that anxiety is associated with a reduction in the capacity to actively maintain cognitive task goals in working memory because this capacity is taken up with a sustained internal attentional focus towards task-unrelated thoughts (i.e. worries and rumination) or an external focus toward unpredictable threats in the environment (Braver, 2012). This implies that anxiety will be associated with a shift towards primary utilization of reactive control and away from proactive control (Braver, 2012). We evaluate this claim across anxiety dimensions and experimental paradigms below.

3.1. Studies of Trait Anxiety

Several studies have examined the effect of trait anxiety on proactive and reactive control, with some supporting the claim that trait anxiety impairs proactive control. Some of these studies have utilized variants of the color-word Stroop task (Stroop, 1935), in which top-down selective attention is required to focus processing on the task-relevant font color of printed words while ignoring the irrelevant but otherwise dominant word name. Here, the primary index of cognitive control is the “Stroop interference effect,” which contrasts incongruent (word name indicates a different color than the font color, e.g., BLUE in red font) and congruent (word name matches font color, e.g., BLUE in blue font) trials (Braver et al., 2021; R. Tang et al., 2023). For example, in a behavioral study, Kalanthroff et al. (2016) investigated the effect of task-irrelevant emotional distractors on attentional proactive control and its interaction with trait anxiety using a Stroop task variant. In this variant, target stimuli were preceded by brief (neutral vs. aversive), irrelevant emotional distractors, thus affecting proactive control (i.e., allocated in advance of the stimulus to inhibit automatic but incorrect responses) rather than diverting attention from the relevant stimulus at hand. The authors hypothesized that proactive task control would remain high in low-trait-anxious (LA) participants, whom they predicted would be able to inhibit the emotional distractor. However, the authors hypothesized that the emotional effects would be more potent for high-trait-anxious (HA) participants, whom they predicted would show reduced proactive task control. Indeed, results showed that, while the negative emotional distractors had little effect on the Stroop RT in low-trait-anxious (LA) participants, they had a significant effect in HA participants resulting in a slowdown after negative emotional distractors, for both congruent and incongruent Stroop stimuli, but not for neutral Stroop stimuli. These results are consistent with the idea that trait anxiety influences the interaction between irrelevant emotional stimuli and proactive control; while LA participants can maintain intact proactive task control, anxious individuals show difficulty filtering out emotional distracters and are thus impaired in their proactive task control.

Studies employing EEG and fMRI methodologies have also provided support that trait anxiety impairs proactive control. Krug & Carter (2012) conducted an investigation of individual differences in trait anxiety using an emotional facial Stroop task during fMRI. Participants were instructed to indicate whether face stimuli were neutral or fearful in emotion while ignoring the accompanying “neural” or “fearful” word. Critically, they performed high expectancy (HE) (65% incongruent trials) and low expectancy (LE) (35% incongruent trials) versions of the task, hypothesizing that, with a high expectancy for conflict, participants would engage proactive control, orienting attention more strongly to the task-relevant color and away from the word, even prior to stimulus onset. The authors focused on high conflict cI (incongruent trials preceded by congruent trials) neutral face-fearful word trials, as participants with higher trait anxiety may have difficulty disengaging from the task-irrelevant “fearful” word, particularly when it is not expected and control processes are not yet recruited. Results were indeed consistent with the interpretation of higher trait anxiety being associated with alterations in proactive control. Specifically, in the high expectancy task, on high conflict trials with task-irrelevant emotional information, behavioral impairments (slower response time and decreased accuracy) were observed in high trait-anxious individuals, along with reduced activity in left ventrolateral prefrontal cortex, anterior insula, and orbitofrontal cortex. The results suggest that individual differences in anxiety may be associated with expectancy-related strategic control adjustments, particularly when emotional stimuli must be ignored.

Another frequently used cognitive control task is the Eriksen flanker task (Eriksen & Eriksen, 1974), which requires that participants quickly and accurately identify a target letter in the middle of a letter string. The target stimulus is surrounded by non-target stimuli, corresponding to either the same response as the target (i.e., congruent trials; e.g., HHHHH) or to the alternative response (i.e., incongruent trials; e.g., SSHSS). Incongruent, but not congruent, trials elicit response conflict, thus requiring enhanced control to deliver the correct response. For example, Schmid et al. (2015) used the Eriksen flanker task along with concurrent EEG recording in order to investigate the neurocognitive processes through which trait social anxiety relates to task performance. The study focus was on the relative contributions of proactive control, theoretically associated with top-down regulation and dorsolateral prefrontal cortex (dlPFC) activity, and reactive control, theoretically associated with conflict monitoring and dorsal anterior cingulate cortex (dACC) activity. The EEG measures provided continuous indices of sustained dlPFC engagement throughout the task, indicated by left frontal EEG asymmetry, to index proactive control; conversely, event-related potential (ERP) indices of conflict-related CON activity, indexed by the response-locked N2 ERP, or N2r, were used to index reactive control. Results showed that greater left prefrontal EEG activity predicted better behavioral control for all participants; however, in high social anxiety participants only, greater N2r responses also predicted behavioral control. These results suggest that low social anxiety individuals engaged a proactive control process, driven by dlPFC activity, whereas high social anxiety individuals relied additionally on a reactive control process, driven by conflict-related dACC activity.

However, some results appear to contradict the prediction that anxiety is associated with primary utilization of reactive control (Braver, 2012). For example, Zhang & Xiang (2023) investigated whether trait anxiety modulates reactive control using the color-word Stroop with ERPs. The authors manipulated item-specific proportion congruence (ISPC), as the ISPC effect is theoretically predicted to elicit reactive control (Bugg & Crump, 2012; Bugg & Hutchison, 2013; Jacoby et al., 2003). To manipulate ISPC, certain colors occur with low proportion congruence (e.g., items appearing in green font will be mostly incongruent [MI]), while other colors occur with high proportion congruence (e.g., items appearing in red font will be mostly congruent [MC]). Through this manipulation, strong associations are predicted to develop between the critical color feature (e.g., green) and increased control demands (i.e., high interference), leading to more effective goal retrieval and utilization when this feature is detected (i.e., post-stimulus onset, thus, reactively). The authors demonstrated that individuals exhibiting high levels of trait anxiety experienced a more pronounced effect on response time due to the ISPC compared to individuals with low levels of trait anxiety, suggesting that trait anxiety influenced reactive control. During the EEG analysis, they also found a considerably larger ISPC effect associated with high trait anxiety compared to low trait anxiety, specifically concerning the slow potential (SP) component. Therefore, the authors state that they observed both behavioral and EEG evidence that trait anxiety impairs reactive control.

A closer look at the Zhang & Xiang (2023) results shows that their findings may be more complex. Behaviorally, high trait anxiety participants had a larger Stroop effect than low trait anxiety participants in the MC condition; in this condition, both proactive and reactive control are thought to be at low levels, so the greater Stroop effect likely reflects an overall impairment in cognitive control rather than a specific reactive control deficit. In the MI condition, reactive control is engaged, requiring the quick detection that specific MI colors are associated with conflict; however, the Stroop effect did not differ between the high and low trait anxiety groups in the MI condition. Additionally, in terms of accuracy, high trait anxiety participants displayed a reduced Stroop effect; this suggests a change in the speed-accuracy tradeoff function reflecting increased response caution. Together, these behavioral patterns suggest that high trait anxious participants may actually demonstrate stronger reliance on reactive control, rather than impaired reactive control. In terms of the EEG results, there appeared to be a heightened response to MI incongruent items in high trait anxious participants, which could also reflect stronger reactive control rather than impairment.

The studies presented above suggest that trait anxiety is associated with alterations in both proactive and reactive control. However, the observed findings were not fully consistent across studies, likely resulting from variations in the trait-anxious samples employed (e.g., populations with trait anxiety versus trait social anxiety, differences in how high and low trait anxious groups were formed) and study design (i.e., different tasks used to assess cognitive control, the use of behavioral measures and/or neural measures). Additionally, these studies did not account for the fact that trait anxiety itself can be decomposed into anxious apprehension and anxious arousal dimensions, and participants’ state anxiety and anxiety disorders were not considered. Thus, in the following subsections, we also review studies examining the effect of trait anxiety dimensions, anxiety disorders, and state anxiety on proactive and reactive control.

3.2. Studies of Trait Anxiety Dimensions

Some studies conducted in trait-anxious populations have specifically focused on the effects of trait worry on cognitive control. As discussed above, worry is the core feature that best characterizes the dimension of anxious apprehension; thus, “trait worry” may be reflective of “anxious apprehension”. For example, Forster et al. (2015) examined whether heightened worrying is secondary to deficits in the frontal cortical control of attention or an independent feature of anxiety. The researchers examined this question within the context of an fMRI study of the Sustained Attention to Response Task (SART). In the SART (Robertson et al., 1997), participants must respond quickly but accurately to a series of “Go” stimuli while withholding responses to occasionally presented “No Go” stimuli. Forster et al. (2015) postulated that the infrequent nature of the No Go stimuli required proactive control of sustained attention across Go trials in order to maintain task goals (i.e., to avoid responding so quickly that it is difficult to withhold a response when a No Go stimulus occurs). The authors also measured individual differences in trait anxiety and worry using two distinct self-report measures. Three key hypotheses were investigated: (1) trait anxiety would be associated with impoverished proactive maintenance of sustained attention, reflected by reduced DLPFC activation and reduced connectivity between DLPFC and thalamo-striatal regions across SART “Go” trials; (2) trait anxiety would be independently linked to increased DLPFC engagement during off-task thought, with this being accompanied by increased DLPFC-Default Mode network connectivity (projected to be observed to the greatest extent in blocks containing commission errors); and (3) individual differences in worry would be positively correlated with the extent of DLPFC engagement in off-task thought, and DLPFC–default mode connectivity, but would be orthogonal to individual differences in frontal engagement in attentional control.

The findings of Forster et al. (2015) revealed that trait anxiety, but not worry, was associated with impoverished recruitment of frontal regions to support the proactive control of sustained attention. Importantly, in task blocks containing commission errors, both trait anxiety and worry were linked to increased DLPFC functional connectivity with Default Mode regions implicated in self-referent processing, providing support for the proposal that increased trait anxiety and worry-related DLPFC activity during blocks containing commission errors may reflect spontaneous off-task self-referent thought. However, unlike trait anxiety, worry was not linked to reduced frontal-striatal-thalamic connectivity, impoverished frontal recruitment, or slowed responding during blocks without commission errors, contrary to accounts proposing a direct causal link between worry and impoverished attentional control. The Forster et al. (2015) results suggest that the anxious apprehension dimension of trait anxiety is not specifically associated with proactive control impairments; however, it is important to note that, in this study, there was no attempt to dissociate general trait anxiety from trait worry. Further, the marker of proactive control used (i.e., activity on Go trials in SART blocks relative to activity on Go trials in Control blocks) is not the most precise measure of sustained activity.

3.3. Studies of Anxiety Disorders

Studies have also contrasted the effects of anxiety disorders with trait worry (reflecting the trait anxiety dimension of anxious apprehension) on cognitive control. Many of these studies have used the AX-CPT, in which participants view a series of sequentially-presented letters on a computer screen (A, B, X, Y) organized as cue-probe pairs (Servan-Schreiber et al., 1996). Participants must press the “target” button when they see the letter X, but only if the X is immediately preceded by an A (AX trial type); in all other cases, participants must press the “non-target” button (Hallion, Tolin, & Diefenbach, 2019). AX trials occur with high frequency, leading to strong cue–probe associations (Braver et al., 2021). There are various means of indexing proactive and reactive control in the AX-CPT depending on the specific task variant utilized. Most commonly, performance on AY trials is treated as an index of proactive control, reflecting preparatory tendencies elicited by the A-cue for an expected target response, while BX performance is treated as an index of reactive control, reflecting the need to quickly override the target response associations elicited by the X-probe. For example, Hallion, Tolin, & Diefenbach (2019) used a novel emotional variant of the AX-CPT (i.e., adapted to include negative emotional and neutral distractor word stimuli) to characterize cognitive control over neutral and emotional distractors in clinically significant GAD and as a function of trait worry. Surprisingly, they found that GAD was associated with enhanced cognitive control in the context of emotional distraction, which they further interpreted as specific to proactive control. This enhancement was observed relative to neutral distraction (within-subjects) and relative to healthy and non-GAD (OCD) clinical controls (between-subjects). Enhancement was specific to GAD versus trait worry, although there was some evidence linking higher trait worry to better cognitive control specifically in participants with GAD. The authors suggest that findings support models conceptualizing worry as an intentional (albeit maladaptive) cognitive control or emotion regulation strategy that is actively maintained by individuals with GAD in order to avoid engaging with more distressing emotional information.

In another study, Hallion, Tolin, Billingsley, et al. (2019) investigated the relationship between proactive and reactive cognitive control using a non-emotional version of the AX-CPT while also measuring subjective attentional symptoms (attentional focusing and trait worry) in a mixed clinical sample of individuals with GAD and/or OCD and a comparison sample of healthy controls. The authors examined the proactive behavioral index (PBI), which provides a ratio of the extent to which participants used a more effortful proactive control strategy versus a less effortful reactive control strategy by contrasting AY and BX trial performance (Chiew & Braver, 2013). Results from Hallion, Tolin, Billingsley, et al. (2019) showed that clinical participants reported more severe attentional symptoms in daily life relative to healthy controls, but performed comparably, and in some cases outperformed healthy controls in terms of response time. Furthermore, the relationship between attentional symptoms and response time varied by clinical status, such that attentional symptoms (specifically, focusing and worry) were associated with slower responding on BX trials and lower PBI in healthy participants, but faster responding and greater effort in clinical participants (although the pattern in clinical participants was nonsignificant after correction for multiple comparisons). The authors state that this preliminary evidence suggests that differences in task effort in anxious versus healthy adults may relate to subjective attentional symptoms in GAD and OCD. It is important to note that, while the authors used the PBI as an index of task effort, it more accurately reflects the tendency to use proactive versus reactive control. For example, the PBI was reduced in healthy subjects with high trait worry, which could be interpreted as a shift away from proactive control and toward reactive control. Likewise, the authors associated BX performance with proactive control and AY performance with reactive control, which is somewhat opposite to how these measures have typically been treated in the literature.

Findings from these studies suggest that trait anxiety dimensions (i.e., trait worry/anxious apprehension) and anxiety disorders may be differentially associated with alterations in proactive control. This further reinforces the importance of carefully controlling the anxious population being studied when examining cognitive control impairments, as different anxiety dimensions may lead to differing results. However, as noted above, it is also important to consider that interpretational differences related to association of different experimental indices with proactive and reactive control may also have contributed to the varied results.

3.4. Studies of State Anxiety

Some studies have also examined the effects of state anxiety on cognitive control. For example, in a behavioral study, Yang et al. (2018) examined how state anxiety affected proactive control, using the AX-CPT, and reactive control, using the classic Stroop task. The authors observed that state anxiety, induced with a threat of shock manipulation, inhibited proactive control on the AX-CPT, but increased reactive control in the Stroop task. These results suggest that anxiety may impair proactive control in contexts requiring goal maintenance by occupying limited working memory capacity, whereas it may enhance reactive control via facilitated attention allocation to threat and engaging the conflict monitoring system to quickly modify behavior.

Additionally, Fales et al. (2008) examined the effects of both trait and state anxiety on neural mechanisms of cognitive control and performance efficiency. The authors used a mixed blocked/event-related fMRI design with the n-back working memory task in high and low trait anxious individuals to examine transient and sustained activity in the dorsolateral prefrontal cortex (DLPFC), as indices of reactive and proactive control, respectively. Moreover, the task was performed after the participants viewed videos designed to induce neutral or anxiety-related mood states. Following the neutral video, the high-anxious participants had reduced sustained but increased transient activation in working memory areas in comparison with low-anxious participants. The high-anxious group also showed extensive reductions in sustained activation of “default-network” areas (possible deactivation). Following the negative video, the low-anxiety group shifted their activation dynamics in cognitive control regions to resemble those of the high-anxious group. These results suggest that reduced cognitive control in anxiety might be due to a transient, rather than sustained, pattern of working memory recruitment, consistent with the suggestion by Braver (2012) that anxious individuals should have a greater tendency to utilize cognitive control in a reactive rather than a proactive manner.

The studies reviewed here suggest that state anxiety impacts proactive control and enhances reactive control. However, as the study designs differed (i.e., different tasks used to assess cognitive control, the use of behavioral measures and/or neural measures, different state anxiety inductions), future research should continue examining the effect of state anxiety on proactive and reactive control to corroborate these findings.

3.5. Summary and Recommended Research Directions

In this section, we reviewed studies linking different dimensions of anxiety, cognitive control, and experimental paradigms. Across the reviewed studies, there was great variation in the type of anxiety or the anxious population(s) being investigated (i.e., trait anxiety, dimensions of trait anxiety, anxiety disorders, state anxiety), the cognitive control mode(s) being examined (i.e., proactive control only, reactive control only, both modes), and the tasks being used (i.e., Stroop variants, flanker task, SART, AX-CPT, n-back working memory). Additionally, studies differed on their use of behavioral (i.e., RT) and/or neural (i.e., EEG, fMRI) measures. Therefore, it is unsurprising that findings were mixed and suggested that trait anxiety, dimensions of trait anxiety, anxiety disorders, and state anxiety are associated with different alterations in proactive and reactive control. These results exemplify the importance of carefully controlling for the dimensions of anxiety, cognitive control, and tasks in research.

Based on the framework described above, we recommend that future work more explicitly account for trait anxiety dimensions (i.e., anxious apprehension and anxious arousal) when investigating effects on cognitive control. It has been shown that failure to account for a distinction between these dimensions when examining anxiety in relation to executive function may produce important confounds (Sharp et al., 2015). For example, Warren et al. (2021) drew upon the Miyake et al. (2000) model to evaluate associations between components of EF and symptom dimensions of anxiety (anxious apprehension, anxious arousal) using factor analyses and structural equation modeling. They found clear dissociations between these two dimensions in factor loadings, with anxious apprehension showing greater specificity to the shifting component. In another example, research involving event-related brain potentials has revealed that anxiety is associated with enhanced error monitoring, as reflected in increased amplitude of the error-related negativity (ERN). Through meta-analysis and a critical review of the literature, Moser et al. (2013) highlighted that it is actually the anxious apprehension/worry dimension of anxiety which is most closely associated with error monitoring. Although, overall, anxiety demonstrated a robust, small-to-medium relationship with enhanced ERN, studies employing measures of anxious apprehension showed a threefold greater effect size estimate than those utilizing other measures of anxiety (Moser et al., 2013). The ERN has been linked to a neural generator within the ACC and wider cingulo-opercular network (CON) (LoTemplio et al., 2023), and so, at first blush, may seem more tightly linked to reactive control processes within the DMC framework. Nevertheless, the ERN itself appears to function not as a direct index of proactive or reactive control (which somewhat by definition refer to pre-response cognitive processes), but instead as an error-related learning signal (Olvet & Hajcak, 2008) that may have an influence on modulation of both control modes.

Another important research direction will be to disentangle whether state anxiety dimensions (i.e., worry, panic/fear) have similar or unique impacts on cognitive control from other dimensions of anxiety, such as trait anxiety and anxiety disorders. Indeed, a review of findings from studies examining cognitive control dysfunction in anxiety found that processing and performance alterations in individuals with anxiety during cognitive control tasks were primarily a consequence of anxiety-induced changes in rumination, worrying, attention, and inhibition (Paulus, 2015). Thus, cognitive control (dys)function may be highly dependent on the state of the individual, providing more arguments for considering dimensional approaches to psychopathology that are clearly linked to current states, rather than categorical approaches to delineate the underlying neural systems’ pathologies (Paulus, 2015). In order to more precisely examine state anxiety dimensions, it may be effective to experimentally manipulate worry and panic/fear via the use of state inductions. Worry inductions instruct participants to think in detail about a past episode that made them feel sad, anxious, or stressed, or something that may happen in the future that worries them (Ottaviani et al., 2016). Worry inductions have been frequently used in neuroimaging studies with GAD populations (Makovac et al., 2020) and have been shown to elicit perseverative negative cognition (Ottaviani et al., 2016). Conversely, fear inductions may use a variety of techniques, such as anticipated electric shock paradigms, viewing fearful stimuli, or imagining being injected by a needle or trapped in an elevator, and have been found to influence both autonomic responses and self-reported fear (Siedlecka & Denson, 2019).

The review of findings from studies examining cognitive control dysfunction in anxiety (Paulus, 2015) also found that the effects observed in behavioral metrics of proactive and reactive cognitive control were inconsistent, and task processing differences were more consistently observed in individuals with high trait anxiety or anxiety disorders in terms of neural metrics, such as those acquired in electrophysiological and fMRI studies. Thus, there is a clear need to incorporate more sensitive neuroimaging methods (e.g., EEG, fMRI), in addition to behavioral and self-report metrics, when studying cognitive control in anxiety.

4. Mindfulness and Cognitive Control

In this section, we review studies linking dimensions of mindfulness with cognitive control (specifically, proactive and reactive control dimensions) in order to demonstrate how different conceptualizations of these constructs may affect findings. It has been proposed that mindfulness can be a means of balancing the cognitive control system by focusing on the present moment without judgment, thereby encouraging individuals to be more flexible in using both the reactive and proactive control modes (Chang et al., 2018). We investigate this claim across mindfulness dimensions and experimental paradigms below.

4.1. Studies of Mindfulness Training

A few studies have investigated the effect of mindfulness training on cognitive control. Li et al. (2018) used a pre-post design to compare a wait-list control group with a group completing an 8-week mindfulness training program based on MBCT, employing the AX-CPT to assess proactive and reactive control and the Five Facet Mindfulness Questionnaire (FFMQ) (Baer et al., 2006) to assess mindfulness levels. The authors calculated the relative balance of proactive versus reactive control using the PBI measure (also utilized by Hallion, Tolin, Billingsley, et al. (2019)). Some evidence of enhanced reactive control in the training group was observed, in terms of improved BX trial performance, while the PBI showed a trend toward a shift toward proactive control. Additionally, awareness of the present in mindfulness, as reflected by scores on the Observing subscale of the FFMQ, were positively correlated with the proactive control mode, providing further evidence of a relationship between components of mindfulness and different aspects of cognitive control. The authors proposed that the effects of an 8-week mindfulness practice might help individuals in overcoming interference and effectively navigating daily life through its impact on the two control modes.

Incagli et al. (2020) conducted a longitudinal study to investigate the effect of MBSR (control training group included) on neural as well as behavioral metrics of proactive and reactive cognitive control. Participants were tested on a modified AX-CPT with concurrent EEG recording before and after eight weeks of training. Unlike in the traditional AX-CPT, the frequency of A- and B-cues were equated within each block, allowing the authors to control for cue validity across trials and for potential confounding factors related to the infrequency/novelty of B-cues. The frequency of each probe type was also the same, placing a greater emphasis on cue information. No information about the probability of AX pairs was provided; thus, participants were not biased towards implementing a specific proactive strategy. Amplitude modulations of event-related potentials (ERPs) associated with cues and probes were examined. Results showed that, after the training, the MBSR group exhibited a significant reduction of errors on high conflict trials associated with both proactive and reactive control (i.e., AY and BX trials). Concurrently, the Contingent Negative Variation (CNV), an ERP index of anticipatory processes elicited by task cues, became more pronounced in the post-training session in the MBSR group only. In addition, an attenuated probe-locked N2 and an increased P3a component emerged. Taken together, the behavioral and electrophysiological results suggest that mindfulness practice enhanced the ability to implement both proactive and reactive cognitive control processes.

4.2. Studies of State and Trait Mindfulness

The effects of both state and trait mindfulness on cognitive control have also been examined. In Chang et al. (2018), the relationship between state mindfulness and cognitive control mode was examined with a brief (10-minute) mindfulness induction immediately prior to performance of the AX-CPT task, using a between-groups design comparing the mindfulness induction group to control (relaxation) and no-induction groups. The authors calculated the relative dominance of proactive versus reactive control using the PBI measure. The results showed overall faster reaction times in the mindfulness manipulation group and a greater balance between reactive and proactive control relative to the control groups, in which proactive control dominated. In a second study, the authors utilized the Mindful Attention Awareness Scale (MAAS) (Brown & Ryan, 2003) to measure trait mindfulness effects on AX-CPT performance. Similar to the state mindfulness effects, individuals high in trait mindfulness exhibited overall faster reaction times and a more balanced PBI relative to those with low trait mindfulness. These findings suggest that mindfulness impacts both reactive and proactive control, which leads to flexible cognitive control performance.

Aguerre et al. (2021) proposed that mixed findings regarding the relationship between mindfulness and cognitive control may arise from the incorporation of different questionnaires to assess mindfulness and from the use of single tasks to assess cognitive control. Thus, the authors used a multicomponent approach to investigate the extent to which dispositional mindfulness relates to dynamic use of control modes, including a composite based on two well-established trait mindfulness measures (MAAS and FFMQ) and two well-validated experimental tasks measuring proactive/reactive control modes (AX-CPT and Cued Task-Switching Paradigm [C-TS]). A multiple regression approach was utilized in which composite trait mindfulness scores were predicted by metrics from each task. The findings suggested that trait mindfulness was predicted by a greater balance between proactive and reactive control on the AX-CPT, primarily due to a reduction in proactive control (on AY trials), and a similar pattern of reduced proactive control in the C-TS (focusing on conditions in which proactive control was possible and/or encouraged due to cue availability during the preparatory period).

4.3. Summary and Recommended Research Directions

In this section, we reviewed studies linking different dimensions of mindfulness, cognitive control, and tasks. Across studies, there was variation in the type of mindfulness being studied (i.e., mindfulness training, state mindfulness, trait mindfulness). Additionally, studies differed in their use of behavioral and/or neural measures. However, a strength is that all studies used the AX-CPT, and all studies examined both proactive and reactive control modes, but mostly within single tasks and conditions (however, one study also employed different conditions from the C-TS). Findings generally suggested that trait mindfulness and mindfulness training are associated with enhancements in, and a more balanced use of, proactive and reactive control (sometimes due to reductions in proactive control). However, more work is needed to parse the dimensions of mindfulness, cognitive control, and experimental tasks in order to ascertain the robustness and generality of mindfulness-related effects.

It is critical that future work accounts for specific components of mindfulness training and state mindfulness (i.e., FA and OM) when investigating effects on cognitive control. Although Chang et al. (2018) did use a state induction design, it was not clear from the description whether it was specific to FA or OM. Previous work has highlighted the utility of experimentally inducing specific mindfulness states, such as FA and OM, to directly examine their proximal effects on attentional and cognitive control (Lin et al., 2024; Lin, Tang, et al., 2022). For example, in a sample of healthy participants with no prior meditation experience, Lin et al. (2024) implemented a state induction protocol designed to elucidate the neurobehavioral influence of discrete mindfulness states—FA and OM, comparing them both against an active control—on behavioral and event-related potential (ERP) indices of executive attention and error monitoring assessed during the Eriksen flanker task. Another critical feature was that a fully within-subject design was utilized; all participants performed each condition (in separate sessions with a counterbalanced order), thus increasing statistical power to detect state mindfulness effects. Surprisingly, OM selectively produced a more cautious and intentional response style, characterized by higher accuracy, slower RTs, and reduced P3 amplitude. However, this study utilized a flanker task not well-suited to capturing proactive and reactive control modes; thus, future work leveraging state induction designs would benefit from the use of more sensitive cognitive control tasks, such as the AX-CPT.

5. Anxiety, Mindfulness, and Cognitive Control

In Sections 3 and 4, we demonstrated that the way anxiety, mindfulness, cognitive control, and experimental paradigms are conceptualized can greatly influence findings. Mixed findings suggested that trait anxiety, dimensions of trait anxiety, anxiety disorders, and state anxiety are associated with different alterations in proactive and reactive control. Findings also suggested that trait mindfulness and mindfulness training are associated with enhancements in, and a more balanced use of, proactive and reactive control. These results suggest that mindfulness may mitigate anxiety through an improvement in attentional and cognitive control.

Some studies utilizing a dimensional approach have provided support for the claim that attentional and cognitive control mediate the relationship between anxiety and mindfulness. For example, MacDonald & Olsen (2020) examined different elements of attentional control (i.e., self-reported shifting and focusing attentional control) as mediators of the relationship between specific facets of dispositional mindfulness (i.e., observing, describing, acting with awareness, nonjudging, and nonreacting) and anxiety symptoms. Results showed that greater describing and nonreacting mindfulness skills indirectly predicted fewer anxiety symptoms in a sample of college students, in part through their effect on better focusing attentional control, even after controlling for shifting attentional control, as well as participant ethnicity and sex. Additionally, Jankowski & Bąk (2019) investigated the relationships between trait mindfulness (i.e., dispositional mindfulness facets of receptive attention and decentration, defined as the ability to observe one’s thoughts and feelings as temporary, objective events in the mind, as opposed to reflections of the self that are necessarily true (Fresco et al., 2007)), trait anxiety, and attentional control (assessed via self-report scale) using structural equation modeling and multi-model inference. Results showed that attentional control partially mediated the relationship between anxiety and decentration, but attentional control did not relate to receptive attention.

5.1. Summary and Recommended Research Directions

The studies summarized above provide initial support for the claim that attention and cognitive control mediate the relationship between anxiety and mindfulness, but also exhibit variation in the dimensions being studied. Additionally, the studies examined self-reported attentional control, which should not be characterized as a veridical index of cognitive abilities; rather, attentional control is most directly assessed via performance-based behavioral measures, such as cognitive tasks (P. G. Williams et al., 2017). Additionally, none of the studies examined proactive and reactive cognitive control modes or employed measures to assess brain function such as EEG or fMRI. Thus, future work should more carefully control the dimensions being studied and utilize both behavioral and neural methods to assess cognitive control.

An important research direction for future work linking components of anxiety and mindfulness with cognitive control will be to test for specific and dissociable effects of FA and OM training. While a wealth of research has focused on examining the effects of mindfulness-based interventions on anxiety (e.g., Chiesa & Serretti, 2009; Fumero et al., 2020; Ghahari & Mohammadi-Hasel, K. Malakouti, S. K. Roshanpajouh, 2020; Khoury et al., 2015), there is a need to further examine the effects of specific mindfulness practices, or stand-alone mindfulness exercises (SAMs) (i.e., mindfulness exercises not integrated into a larger therapeutic framework such as MBSR or MBCT), on anxiety. Blanck et al. (2018) conducted a systematic review and meta-analysis regarding the effects of SAMs on symptoms of anxiety and depression. Four categories of mindfulness exercises were identified: breathing meditation, body scan, sitting meditation, and sound scan. After exclusion of one outlier, SAMs had small to medium effects on anxiety and depression when compared with controls (summary effect estimates decreased, but remained significant when corrected for potential publication bias). These results suggest that regular performance of simple mindfulness exercises is beneficial, even without being integrated in larger therapeutic frameworks.

Although the meta-analysis by Blanck et al. (2018) did not specifically examine FA and OM practices as SAMs, other work has shown that FA and OM may have differing effects on anxiety. Cullen et al. (2021) assessed both short- and long-term between- and within-group differences in affective disturbance among FA, OM, and their combination (MBCT) within the context of a randomized controlled trial. The authors observed that FA outperformed the other two treatments on several indices of anxiety and showed statistically significant and clinically meaningful improvements that were rapid, sustained, and accompanied by minimal deteriorations.

It may also be useful to implement FA and OM as state inductions to examine the effects of these states on cognitive control in anxiety. In a systematic review and meta-analysis, M. Williams et al. (2024) examined (1) the effect of acute mindfulness induction on state anxiety and attention among individuals with high anxiety, and (2) the impact of any predictors, mediators and moderators of these effects. Although only a small number of studies were included in the review, with high risk of bias and low certainty of evidence present, findings suggested that brief mindfulness induction exercises can reduce state anxiety and distress with moderate and small effects, respectively, and increase state mindfulness with large effects, when compared to non-therapeutic exercises. Additionally, there was limited support for behaviorally measured attentional gains in anxious individuals, but the authors suggest that further research is needed on this topic.

6. A Conceptually and Methodologically Precise Framework to Study the Effect of Mindfulness on Cognitive Control in Anxiety

Together, the prior literature clearly reinforces the need for conceptual and methodological precision when studying the effects of mindfulness on cognitive control in anxiety. Based on important insights from this review, we suggest that future work linking components of anxiety, mindfulness, cognitive control, and experimental paradigms should more carefully control: (1) the anxiety dimension being examined (i.e., anxiety disorder, trait anxiety, trait anxiety dimensions, state anxiety), (2) the mindfulness dimension being studied (i.e., trait mindfulness, state mindfulness, mindfulness training, mindfulness skill), (3) the cognitive control modes being investigated (i.e., proactive control, reactive control), (4) the cognitive control experimental paradigm being used (i.e., Stroop, AX-CPT, etc.), and, if possible, (5) the specific conditional focus of the experimental paradigm being used (baseline, proactive, reactive) in order to avoid confounds. Moreover, there is a clear need to incorporate sensitive neuroimaging methods (e.g., EEG, fMRI), in addition to behavioral and self-report metrics, to offer more precise insight into brain-behavior-clinical symptom relationships. Behavioral methods often lack the temporal resolution required to disentangle proactive- and reactive-control-related brain activity (Valadez et al., 2021). Nevertheless, studies examining the effect of mindfulness on cognitive control in anxiety have rarely employed brain imaging techniques that have high temporal precision, such as event-related potentials (ERPs), as these may exhibit a higher sensitivity to variation in cognitive control (Schröder et al., 2024).

To address these issues, we propose a conceptually and methodologically precise framework to study the effect of mindfulness on cognitive control in anxiety. The framework leverages the non-unitary, multidimensional taxonomy of anxiety, cognitive control, mindfulness, and experimental paradigms (see Figure 1) to theoretically align anxiety dimensions with specific mindfulness states and interventions, measuring the effects on cognitive control dimensions with experimental paradigms and associated behavioral and neural metrics which index them with high precision. To illustrate the utility of this framework, we introduce two theoretical claims inspired by this review below. The first theoretical claim explores the relationship between anxious apprehension, worry, proactive control, focused attention, and proactive control experimental paradigms. The second theoretical claim explores the relationship between anxious arousal, panic/fear, reactive control, open monitoring, and reactive control experimental paradigms.

To experimentally test the framework, we propose study designs that leverage intervention, state induction, and precision neuroscience approaches which can be used to test these theoretical claims. In an intervention approach, participants are assessed on cognitive control tasks before and after undergoing a mindfulness training program, such as an 8-week intervention. In a state induction approach, specific states are induced through standardized audio-guided instructions and are tested for their influence on cognitive processing by instructing participants to maintain the induced state while performing an immediate subsequent task. As an extension to the state induction approach, a more longitudinal “precision neuroscience” (Poldrack, 2017) approach can be used, in which a small sample of dedicated participants are recruited and tested repeatedly, across multiple experimental sessions, during which different mindfulness states are induced and then tasks are performed. This precision neuroscience approach allows researchers to collect enough data to test for effects within individuals (Gordon et al., 2017; Siegel et al., 2024).

6.1. Theoretical Claim 1: Anxious Apprehension, Focused Attention, and Proactive Control

The first theoretical claim explores the relationship between anxious apprehension, worry, focused attention, and proactive control, using experimental paradigms designed to assess this control mode sensitively and precisely (see Figure 2). Anxious apprehension has been associated with an initial failure of goal maintenance due to worry co-opting working memory, which is compensated for by temporary reactivation of task rules (Moser et al., 2013). Thus, a key novel hypothesis to be investigated is that anxious apprehension will primarily impact the proactive control mode. Conversely, FA practices have a high theoretical overlap with proactive control, building capacity in maintaining an active predefined goal representation (i.e., attend to breath/target and redirect attention when mind wanders) (Lin, Tang, et al., 2022). Therefore, we propose that FA practices may be specifically well-targeted to address proactive control impairments in individuals with anxious apprehension. Further support for the utility of FA for anxious apprehension is provided by resting-state fMRI findings, which demonstrated clear distinctions in DMN connectivity between this group and individuals with anxious arousal (Burdwood et al., 2016). The authors suggested that “…even at rest, individuals with high levels of anxious apprehension struggle to engage in present-focused, self-oriented thought…These findings therefore suggest that, when treating clients with symptoms of anxious apprehension, therapeutic strategies such as mindfulness that work to increase present-focused, self-referential thoughts, rather than focusing strictly on reducing maladaptive repetitive thought processes, may be most productive.” This recommended practice bears a striking resemblance to FA, which entails voluntary focusing of attention on a chosen object (Lutz et al., 2008).

Figure 2. Theoretical Claim 1.

Leveraging the non-unitary, multidimensional taxonomy of anxiety, cognitive control, mindfulness, and experimental paradigms, theoretical claim 1 links the dimensions of anxious apprehension, worry, proactive control, focused attention, and proactive control experimental paradigms.

This theoretical claim could be tested with several different study designs. For example, using an intervention approach, the effect of FA meditation training in participants with high anxious apprehension can be investigated by assessing neural and behavioral markers of proactive control. The FA training intervention can be contrasted with a non-FA control, and different participant groups can also be contrasted, both with varying degrees of specificity (e.g., ranging from wait-list groups to those given training on other forms of mindfulness, and from non-anxious participants to those with other forms of anxiety). Alternatively, using a state induction approach, one can examine the proximal effects of inducing FA states by having participants with anxious apprehension perform experimental tasks immediately after the induction, while brain activity and behavior are monitored. The FA state can be directly contrasted with other non-FA control states, or directly with state worry, through a parallel worry induction, as has been used in the prior literature (Makovac et al., 2020; Ottaviani et al., 2016). As an extension to this type of state induction design, a precision neuroscience approach can be used, in which inductions and task performance are repeated across multiple sessions in a small sample of dedicated participants (e.g., N = 10), thus allowing for within-subject manipulations of the type of induction, the experimental paradigm utilized, and the cognitive control mode being assessed. Alternatively, the same conditions could be repeated across sessions to estimate the reliability, replication, and precision of the induced effects.

Experimental paradigms that are theoretically optimized for sensitivity to proactive control may be most effective for detecting FA effects on anxious apprehension. In this case, a key advantage of the experimental paradigms developed for the DMCC battery is that different variants have been designed to be differentially sensitive to proactive and reactive control, and these variants can also be contrasted with baseline versions. The use of such precise variants may increase both sensitivity and specificity. For example, in a prior study, it was shown that the effects of individual differences in working memory capacity were found most clearly in the proactive condition of the AX-CPT when contrasted against baseline and reactive versions (Lin, Brough, et al., 2022).

With this type of study design used to investigate FA effects on proactive control in individuals with anxious apprehension, behavioral and neural measures should also be optimized to detect key markers of proactive control. Within the DMCC task battery, selective behavioral markers of proactive control have been identified, including the A-cue bias for the AX-CPT, the list-wide proportion congruent effect (on diagnostic items) for the Stroop task, and the mixing cost for Cued-TS (Braver et al., 2021; R. Tang et al., 2023). Neural markers of proactive control have also been identified for these and other tasks. For EEG, these neural markers include the P3 ERP component, thought to reflect working memory updating processes (Donchin, 1981), and the CNV ERP, associated with response expectation and preparation (Schröder et al., 2024); these ERPs may be observed following cue onset in the AX-CPT, for example. For fMRI, neural markers include increased anticipatory activation dynamics in lateral PFC (Braver et al., 2021), which can emerge at the time of contextual cues, for example, in tasks such as the AX-CPT (Braver et al., 2009). Proactive control is also associated with sustained or tonic activation within the lateral PFC (Braver et al., 2021), potentially reflecting a more persistent or extreme form of preparation that reduces control demands at the time when contextual cues are presented in tasks such as Cued-TS, AX-CPT, and Sternberg WM (alternatively, it may be required for active goal maintenance in tasks where trial-based contextual cues are not available, such as in the Stroop task) (Braver et al., 2009).

Finally, clinically sensitive self-report measures can also be utilized to test the hypothesis that FA effects on proactive control mediate the beneficial effects of this dimension of mindfulness on anxious apprehension. Measures optimized to assess anxious apprehension, such as the Penn-State Worry Questionnaire (PSWQ) (Meyer et al., 1990), and state-related cognitive anxiety symptoms (i.e., worry, intrusive thoughts), such as the State–Trait Inventory for Cognitive and Somatic Anxiety (STICSA)—State Version (Ree et al., 2008) can be used, along with measures capturing related constructs such as affect and mood (e.g., Positive and Negative Affect Schedule [PANAS] (Watson et al., 1988)). As depression and anxiety are highly comorbid (Pollack, 2005), dimensional psychopathology research has also highlighted the importance of accounting for depression (for example, using the anhedonic depression subscale of the Mood and Anxiety Symptom Questionnaire (MASQ) (Clark & Watson, 1991; Watson et al., 1995)) in order to assess, control for, or remove variance associated with it (Snyder et al., 2023). For intervention designs, these clinically sensitive self-report measures should be administered at pre- and post-intervention assessments along with proactive control measurements, such that the two can be linked via correlation and causal mediation analyses. For induction designs, it would be particularly useful to assess state-related worry before and after inductions.

6.2. Theoretical Claim 2: Anxious Arousal, Open Monitoring, and Reactive Control

The second theoretical claim explores the relationship between anxious arousal, panic/fear, reactive control, open monitoring, and reactive control experimental paradigms (see Figure 3). Research has shown that anxious arousal, characterized by an enduring pattern of hypervigilance, sympathetic nervous system hyperarousal to mild stressors, and state fear (Sharp et al., 2015), engages neural mechanisms that implement bottom-up attention more easily to mildly threatening stimuli (Spielberg et al., 2013). Thus, a key novel hypothesis to be investigated is that anxious arousal may lead to overreliance on the more bottom-up reactive control mode, which is stimulus dependent and much more vulnerable to transient attentional capture or orienting effects that may disrupt the ability to trigger goal reactivation when necessary (Braver, 2012). Critically, OM practices, which involve non-judgmental monitoring of the contents of experience from moment to moment but without the need to focus on any specific object (Lutz et al., 2008), appear to share theoretical overlap with reactive control. Therefore, we propose that OM practice may balance the use of reactive control such that high anxious arousal participants are not overly reliant on this mode. Further support for the utility of OM for anxious arousal is provided by resting-state fMRI findings (Burdwood et al., 2016), which showed that, “Although anxious arousal is often viewed as a more externally focused type of anxiety, results indicate that it involves an excess of internally focused cognition at rest. In light of these findings, interventions for those with high levels of anxious arousal might also include an adapted mindfulness-based component in which patients are taught to calmly and curiously attend to both internal and external stimuli.” This recommended practice bears a remarkable resemblance to OM, which involves non-reactive monitoring of the content of experience from moment to moment (Lutz et al., 2008).

Figure 3. Theoretical Claim 2.

Leveraging the non-unitary, multidimensional taxonomy of anxiety, cognitive control, mindfulness, and experimental paradigms, theoretical claim 2 links the dimensions of anxious arousal, panic/fear, reactive control, open monitoring, and reactive control experimental paradigms.

Paralleling the theoretical claim regarding FA, anxious apprehension, and proactive control, the relationship between OM, anxious arousal, and reactive control can be examined with similar types of study designs. For example, using an intervention approach, one can examine the effect of an 8-week OM meditation training in participants with high anxious arousal. Alternatively, using a state induction approach, the OM state could be directly contrasted not only with non-OM control states, but also with induced panic or fear states. With either type of design, OM effects should be assessed via experimental paradigms that are optimized for sensitivity to reactive control, such as the reactive control variants of tasks included in the DMCC battery.

Behavioral and neural measures should also be optimized to detect key markers of reactive control, such as those associated with bottom-up reactivation of task goals, linked to the dynamics and timing of imperative stimuli associated with high control demands. Within the DMCC task battery, selective behavioral markers of reactive control have been identified, including BX interference effects for the AX-CPT, item-specific proportion congruence effects for the Stroop task, and recent negative effects in the Sternberg WM task (Braver et al., 2021; R. Tang et al., 2023). For EEG, neural markers of reactive control include the target- (or NoGo-) N2, related to response conflict monitoring, and the target- (or NoGo-) P3, related to motor inhibition; these may be observed after target onset in the AX-CPT, following a NoGo stimulus in a cued Go/NoGo task, or following a Stop-Signal stimulus in a Stop-Signal Task (Schröder et al., 2024). For fMRI, neural markers of reactive control include changes in event-related activity to probe stimuli associated with high control demands within fronto-parietal and cingulo-opercular regions associated with interference/conflict detection (e.g., ACC, medial frontal cortex) and/or episodic/associative cueing (e.g., posterior parietal cortex, medial temporal lobe) (Braver et al., 2021).

Again, clinically sensitive self-report measures can also be utilized to test the hypothesis that OM effects on reactive control mediate the beneficial effects of this dimension of mindfulness on anxious arousal. In this case, measures optimized to assess anxious arousal, such as the anxious arousal subscale of the MASQ (Clark & Watson, 1991; Watson et al., 1995), and state-related somatic anxiety symptoms (i.e., hyperventilation, palpitations, sweating), such as the STICSA—State Version (Ree et al., 2008), can be used. Again, here, it is important to assess and control for potential comorbidities, such as depression.

6.3. Contrasting Theoretical Claims 1 and 2

Given the parallel nature of the two theoretical claims, a potentially powerful approach would be to directly contrast them. In this case, a factorial design would be particularly effective, as the hypothesis can be most clearly specified as a two group (anxious apprehension, anxious arousal) x two mindfulness practice (FA, OM) x two cognitive control mode (proactive, reactive) design. However, as discussed above, it would be most scientifically rigorous to expand the design to be 3 × 3 × 3, by including a control group, intervention or state, and baseline (non-specific) measure of cognitive control. Of course, it is important to acknowledge that the increase in comprehensiveness and rigor does come at a substantial cost, given that even the basic design would include 8 cells, whereas the expanded design includes 27. Furthermore, a key point underlying the design of the DMCC task battery is that proactive and reactive control should be assessed through multiple different tasks to establish the generality of the inferred effects. Such an approach would further expand the experimental space of investigation. Nevertheless, it is important to note that a key advantage of the factorial design logic is that the full space of theoretically required conditions could be acquired in a modular, incremental fashion across successive studies (Lin, Tang, et al., 2022). Likewise, using a precision neuroscience state induction approach, a small sample of participants (varying on anxiety dimensions) could each take part in a multi-session study that allowed for successive sampling of the 3 inductions (FA, OM, control) by 3 cognitive control modes (proactive, reactive, baseline) across sessions. Furthermore, in such an approach, each session could allow for assessment of multiple cognitive control tasks (e.g., AX-CPT, Cued-TS) in different induction + task blocks.

As a tractable starting point for future research, we recommend that the first experimental tests of our framework focus on conditions that we believe will yield the clearest effects: (1) contrasting FA against an active (non-mindfulness) control intervention in individuals with anxious apprehension, using precise measures of proactive control and a within-subjects design; and (2) contrasting OM against an active (non-mindfulness) control intervention in individuals with anxious arousal, using precise measures of reactive control and a within-subjects design. Importantly, in both cases, the within-subjects design allows the mindfulness versus non-mindfulness conditions to be contrasted within each individual. Although important for future work, we do not necessarily recommend starting with a direct test of the FA versus OM contrast prior to finding positive evidence of FA – anxious apprehension – proactive control effects and OM – anxious arousal – reactive control effects, which could be examined in separate studies. In this manner, the two separate study designs described above would still allow for important qualitative (e.g., cross-study) comparisons to be made regarding observed findings. If our predictions are confirmed, follow-up work could then be conducted to pursue more direct FA versus OM (and proactive versus reactive) comparisons.

7. Conclusion

In this review, we have argued that research examining the effect of mindfulness on cognitive control in anxiety lacks conceptual and methodological precision, which can lead to confounds and ambiguous outcomes. We first discussed key dimensions of anxiety, attentional and cognitive control, mindfulness, and experimental paradigms (Section 2) in order to develop a non-unitary, multidimensional taxonomy of these constructs. Subsequently, we reviewed and reinterpreted the literature on studies linking dimensions of anxiety and cognitive control (Section 3), studies linking dimensions of mindfulness and cognitive control (Section 4), and studies linking dimensions of anxiety and mindfulness with cognitive control (Section 5). Across the reviewed studies, there was great variation in the dimensions being examined and the behavioral and/or neural measures employed; therefore, it is unsurprising that results were often mixed.

Together, the findings exemplify the importance of future research that carefully controls for: (1) key dimensions of anxiety, mindfulness, cognitive control, and experimental paradigms, and (2) the behavioral and neural measures that are being employed. Based on this review of the literature, we proposed a conceptually and methodologically precise framework which theoretically aligns mindfulness states and interventions with precise anxiety dimensions, measuring the effects on cognitive control dimensions with experimental paradigms and behavioral and neural metrics which index them with high precision (Section 6). We also proposed novel experimental studies and tractable research designs that would enable this theoretical framework to be rigorously tested. It is our hope that cognitive neuroscience researchers interested in anxiety, mindfulness, and cognitive control will be inspired by this new framework and join us in testing out these theoretical claims, with the goal of providing a new mechanistic understanding, as well as clinical benefits to individuals suffering from various problematic dimensions of anxiety.