This study has the overarching aim of examining whether affective consequences can account for cognitive and behavioral consequences of inattention. To do so, we examine common dimensions or characteristics of inattention, particularly in the MW domain and recurrent thoughts. Manifestations of inattention included in this study are construed as attentional lapses, daydreaming, mindfulness, rumination, reflection, worry, postevent processing, ADHD inattentiveness, and counterfactual thinking. Measures of attentional lapses, daydreaming, and mindfulness (reversed scored) have commonly been used to assess MW on a trait level, though each scale likens MW slightly differently: daydreaming is thinking that is shifted away from a current task or an external situation, typically thought of as more intentional than mind wandering (Singer and Antrobus, 1963); mindfulness is the state of being attentive and aware of the present moment (Brown and Ryan, 2003); attentional lapses, or absent-mindedness, are demonstrated by minor negative repercussions of being momentarily off-task (e.g., walking into a room and forgetting what one intended to do; Carriere et al., 2008).

Measures of recurrent thoughts are described as follows: Rumination is repetitively thinking about an experience in a negative light after the experience occurred (Nolen-Hoeksema and Morrow, 1991) or about the level of neurotic self-consciousness; reflection measures the level of intellectual self-consciousness to a given situation (Trapnell and Campbell, 1999); worry is repetitively thinking about future life events with marked anxiety about uncertainty of outcomes (Roemer and Borkovec, 1993); post-event processing is repeatedly thinking about a perceived inadequacy in a previous social situation (McEvoy and Kingsep, 2006); inattentiveness is difficulty concentrating and hyperactivity-impulsivity is characterized by restlessness, fidgeting, and behavioral inhibition, which may be reflective of poor executive functioning or mental control; counterfactual thinking is conceiving of alternatives to how reality unfolded in the recent past, such as “if only” thoughts (Rye et al., 2008).

The first main objective is to explore how these manifestations of inattention converge and overlap with each other; i.e., similarities and differences that may be demonstrated in a certain number of latent constructs. The second main objective is to determine how much of the relationship between MW and low or anxious mood may be explained by recurrent thoughts.

Six hundred and fifty-five participants were recruited using Amazon’s Mechanical Turk (MTurk), an Internet crowd-sourcing marketplace where researchers can post studies for participants to complete. Data collected via MTurk is comparable to data collected by traditional means with regards to participant characteristics, attrition, effect sizes, validity, and reliability (e.g., Casler et al., 2013; Schleider and Weisz, 2015; Ramsey et al., 2016). Moreover, MTurk has the advantages of having stable access to a diverse group of individuals (e.g., Buhrmester et al., 2011) as well as being more cost-effective than laboratory studies (Mason and Suri, 2012; Schleider and Weisz, 2015). Previous MW studies have found the medium to be well-suited to collecting data (e.g., Carriere et al., 2013).

Several steps were taken to maximize data validity based on recommendations in the literature (e.g., Aust et al., 2013; Rouse, 2015; Thomas and Clifford, 2017). Ex ante exclusion involved initial screener questions consisting of the two eligibility criteria, via a drop-down list to select age and primary language, which needed to be above 18 and English in order to proceed with the study. The study was described as a “psychology survey about your thought patterns, self-reflection, and attention” and described it as a “UBC-based study: make ratings about your thought patterns, a 30–40-min survey,” which addresses one of MTurkers’ complaints of unclear task description such as not providing an estimated time commitment. The participant requirement standard was set to the highest available on MTurk: only those with a system qualification of >99% approval rate on more than 10,000 “Human Intelligence Tasks” can access the study. Participants who met those criteria were asked to complete a set of questionnaires hosted on the online platform, Qualtrics. The time estimate of 30–40 min was based on timing three fluent English-speaking undergraduate students; therefore, for a broader audience, 1 h was allotted to complete all questionnaires. Each participant was paid $1.75 based on studies at the time of data collection that evaluated the balance of compensation, data collection rate, motivation, and data quality (Bohannon, 2011; Buhrmester et al., 2011; Litman et al., 2015). Ex post exclusion involved eliminating from analyses participants who: (1) did not score 3 out of 3 on attention check items that were embedded in three questionnaires (e.g., for this question, please select the response option “frequently”), (2) had more than 10 measure items missing, and (3) were outliers in response times, i.e., responding too quickly or too slowly. Although other screeners could be included, such as comprehension and/or open-ended questions, we opted for keeping with the above checks in order to minimize de-incentivizing participants’ efforts and time. Notably, our data was collected in 2016, prior to the observed spike in fraudulent responses by use of virtual private servers (see Dennis et al., 2020) in the summer of 2018 (Chmielewski and Kucker, 2020; Kennedy et al., 2020). Six hundred and twelve participants remained and were included in the final data analyses of this study.

The following measures are solely based on participants’ self-report; therefore, we are assessing participants’ own reflections on all their thought tendencies and correlates, rather than a direct measurement of inattention per se.

To identify which unique aspects of inattention converge with each other, exploratory factor analysis (EFA) was used. EFA re-arranges variables into a small set of clusters based on shared variance and allowing underlying constructs to be isolated. Shared variances infer common underlying factors that influence two or more surface constructs, whereas nonshared variances infer specific factors are unique to a certain surface construct (Yong and Pearce, 2013). An EFA parceled by measures was conducted to examine which manifestations of inattention included in our study (i.e., rumination, reflection, worry, postevent processing, inattentiveness, counterfactual thinking, daydreaming, mindfulness, and attentional lapses) cluster together using mean scale scores.

To determine whether recurrent thoughts mediate the relationship between MW and depressive and anxiety symptoms, structural equation modeling (also referred to as latent variable path analysis) was conducted. Although this model implies directionality, we cannot claim causation due to a cross-sectional study design. The main reasons for using SEM as opposed to regression are twofold. First, SEM allows a model to reflect underlying conceptual theory about latent constructs, and not only observed variables as in a regression. Secondly, SEM uses latent variables to account for measurement error, which also allows multiple indicators of the same construct, whereas multiple indicators in a regression model may cause collinearity problems (e.g., Gunzler et al., 2014). This SEM is not to validate factors as often done in confirmatory analysis, rather, to explore and establish covariation among the constructs of interest.

Findings from the above EFA were used to construct latent variables in the model to avoid issues of multicollinearity. Given that the measures that strongly loaded on Factor I (i.e., MAAS, ARCES, and CSS) assess the behavioral and cognitive consequences of inattention, we describe Factor 1 as “inattentive consequences” of MW. Although DFS is a commonly used MW measure, it weakly loaded on all factors and was therefore excluded. The a priori construct of recurrent thoughts loaded across two factors, separable into constructs of repetitive negative thoughts (RNTs; Factor II) and contemplative/reflective thoughts (Factor III). However, the latter construct was excluded altogether due to CTNES loading weakly on all factors and retaining only RRS-Refl and RRQ-Refl to form a latent variable on Factor III would result in model underidentification. Moreover, contemplative/reflective thinking is not central to the current research question and was therefore excluded. In turn, the remaining measures that strongly loaded on Factor II (i.e., RRQ-Rum, RRS-Brood, PSWQ, and PEPQ-R) are consistent with capturing RNTs (Ehring and Watkins, 2008; McEvoy et al., 2013; Zetsche et al., 2018), and thus we describe Factor II as RNTs. The two outcome variables were depressive symptoms (CES-D) and anxiety symptoms (STAI).

Although MW research appear to assume that we have limited control over thoughts that enter and remain in our minds, evident in self-report measure items, distinguishing deliberate versus spontaneous MW has been useful in showing some potential differences in mood correlates.

On the one hand, as mentioned previously, depression, anxiety, and stress appear to be more common in spontaneous MW compared to deliberate MW (Seli et al., 2019); however, evidence remains mixed in that individuals who have more severe depressive symptoms mind-wander more frequently both with and without meta-awareness (Nayda and Takarangi, 2021), implying deliberate and spontaneous MW, respectively. Spontaneous MW is also associated more strongly with ADHD symptomatology and associated impairments more so than deliberate MW (Seli et al., 2015). In the current study, both MW/inattentiveness (Factor I) and RNT (Factor II) comprises measures with items that suggest unintentional or spontaneous thinking, whereas reflection/introspection (Factor III) comprises measures with items that suggest intentional or deliberate analytical thinking. Moreover, the measures that constitute Factor III appear to have a weaker correlation with negative mood symptoms, largely driven by RRQ-Refl, which appears to contain items with higher face validity in inquiring about reflection and contemplation independently from low mood compared to RRS-Refl. Futuremeasures or studies of inattention may be more cognizant of such an underlying assumption of awareness or intentionality, and use scales such as the brief Mind Wandering: Deliberate scale and Mind Wandering: Spontaneous scale, which clearly demarcates differences between intentionality of MW via self-report (e.g., “I allow my thoughts to wander on purpose” vs. “I find my thoughts wandering spontaneously”; Carriere et al., 2013), or the Rumination scale, which demarcates intrusive and deliberate/constructive rumination (e.g., “I thought about the event when I did not mean to” vs. “I have tried to make something good come out of my struggle”; Calhoun et al., 2000).

On the other hand, rumination and MW research findings appear to be more consistent, suggesting that intentional thoughts may not be systematically associated with negative outcomes unlike unintentional thoughts. Ruminative thoughts have been characterized as “brooding” rumination and “reflective pondering” (Treynor et al., 2003) as well as “intrusive” and “deliberate” rumination (e.g., Taku et al., 2009). Reflective pondering and deliberate rumination tend to be purposeful, constructive problem-solving, qualities of intentional thinking, whereas brooding and intrusive rumination tend to be passive, non-constructive dwelling, qualities of unintentional thinking. Considered together, intentional rumination (reflective and deliberate) appears to have a stronger positive correlation with post-traumatic growth (Taku et al., 2009; Stockton et al., 2011), in contrast, unintentional rumination (brooding and intrusive) appears to have a stronger positive correlation with negative affect (Moberly and Watkins, 2008), depression (Burwell and Shirk, 2007), and suicidal ideation (Chan et al., 2009). Moreover, a small body of research distinguishes between “cued” and spontaneous counterfactual thoughts, noting that more depressed individuals may have lowered ability to spontaneously develop alternative courses of action (via counterfactual thinking) compared to their less depressed counterparts (Beldarrain et al., 2005). Although many of the recurrent thoughts included in this present study have not explicitly distinguished between intentional and unintentional aspects of the constructs, the second set of findings suggest that consequential differences between MW and recurrent thoughts are less likely to be due to intentionality dimensions, which may find its mechanisms explained in executive functioning.

The finding that MW/inattentive consequences and RNTs fell into different factors suggest there may be meaningful differences between the two broader constructs, a similar inference with the intentionality domain above. Further, the finding that RNTs at least partially mediate the relationship between MW and both depression and anxiety suggests that executive mechanisms implicated in MW could overlap with executive mechanisms implicated in RNTs. But if so, to what extent might this possibility align with predominant theories linking executive function to MW, inattention, and RNTs?

Executive functioning models in MW and ADHD do have some convergence, even though they do not thoroughly overlap partially due to MW being a relatively narrow cognitive phenomenon, whereas ADHD involves a large set of symptoms with inattentiveness and hyperactivity/impulsiveness subtypes. Notably, despite a handful of studies showing associations between MW and one or more domains of ADHD (i.e., inattention, hyperactivity, impulsivity; e.g., Biederman et al., 2017; Jonkman et al., 2017; Arabaci and Parris, 2018), none of these relationships have been specifically linked with executive functioning and/or mechanistic explanations of these processes (Lanier et al., 2021). Nonetheless, the nature of inattention can be conceptualized as a broad spectrum. In particular, MW has been associated with poorer performance on a concurrent task and has been explained by either MW consuming working memory resources (Smallwood and Schooler, 2006), or MW representing a brief lapse in executive control (McVay and Kane, 2010). ADHD executive functioning deficits have been explained by the Inhibition model, which proposes that a primary impairment of inhibition sets the stage for secondary executive functioning deficits to surface, such as working memory (Barkley, 1997). ADHD executive functioning deficits have also been explained by the Cognitive-Energetic Model, which proposes an interplay among three levels of difficulties: (1) a lower-level attentional control such as encoding, (2) a second-level energetic system related to arousal, and (3) a higher-level executive functioning system. This highest level has been described as a system that maintains “problem sets” for goal-directed behaviors (Sergeant, 2000).

Although these two models of MW and two models of ADHD are all distinct, they are nonetheless congruent in several critical ways. For one, the working memory consumption model of MW aligns with the Inhibition model directly via the proposed secondary deficiency of working memory, and likewise, the executive functioning lapse model of MW aligns via the primary deficiency of inhibition that leads to an overall executive functioning lapse. Moreover, both MW models align with and are relevant to the Cognitive-Energetic Model, which proposes executive management failures are related to a low ability to keep goals in mind, and therefore is relevant for working memory consumption as well as executive control more broadly.

Similar convergence is seen when considering theorizing on RNTs. To the point, there are four main RNT models related to executive functioning that can complement or extend MW models. Three of these models suggest that impaired executive functioning underlies RNTs and prevents regulation of thoughts, and in particular, through either lowered ability to (1) update working memory to include positive or goal-directed information (Whitmer and Gotlib, 2013), (2) inhibit negative or goal-irrelevant information from returning into working memory (Joormann and Gotlib, 2010), or to (3) shift attention or flexibly disengage from negative, goal-irrelevant information (Koster et al., 2011). The fourth model linking RNTs to executive function, however, suggests the reverse direction, and specifically, that RNTs deplete cognitive resources for optimal executive functioning, leading in turn to lowered ability to shift attention away from perceived threat and lowered inhibition of irrelevant information (Eysenck et al., 2007; Stefanopoulou et al., 2014). To be certain, while the first three models focused on rumination and the fourth model focused on worry, they may all apply to RNTs in general due to RNTs’ transdiagnostic nature (Ehring and Watkins, 2008; Wahl et al., 2019), frequent comorbidities for mood and anxiety disorders (McEvoy et al., 2013), and common co-conceptualization (see review Mennies et al., 2021).

Converging evidence comes from counterfactual thinking. Classified in this study as a type of recurrent thought, though ultimately excluded from the latent construct in analyses, counterfactual thinking also involves holding a different timeline than the present reality in working memory, comparing alternative courses of actions and outcomes, and actively monitoring various executive functions (e.g., Kahneman and Miller, 1986; Roese, 1997; Ridderinkhof et al., 2004; Van Overwalle, 2009; Barbey and Patterson, 2011; Van Hoeck et al., 2013). Along the same lines, adaptive qualities are evident in rumination, persistently seeking resolutions to personal concerns (Klinger, 1971, 2012), and in worry, alerting attention toward perceived threat (Hirsch and Mathews, 2012). The proposal that recurrent thoughts occur and are maintained to keep personally relevant issues in mind may be directly applicable in consuming excessive working memory in the working memory depletion model of MW. Furthermore, the maintenance of recurrent thoughts may be integrated into the executive functioning lapse model of MW, such that the attentional lapse is explained by goals of competing importance at any given moment—the current task at hand vs. personal concerns or perceived threats—rather than an executive “conflict,” rather than “failure,” per se.

Overall, executive functioning may be the essential mechanism that assists in switching from abstract thinking in recurrent thoughts to concrete problem-solving, accurately appraising potential threats, and increasing meta-awareness in recognizing and intentionally deciding upon current goals when our thoughts have drifted. Also given the finding that executive functioning deficits have been shown to moderate the association between RNTs and negative affect, i.e., those with weaker executive functioning had more RNTs and higher negative affect, but those with stronger executive functioning had lower negative affect even with presence of RNTs (Madian et al., 2019), future research may need to consider examining individual differences in higher vs. lower executive functioning related to MW or RNTs and negative mood to further illuminate executive functioning consequences of inattention.

The content of thoughts is often one of the most salient features when describing different types of inattentive thoughts; however, this qualitative nature of thoughts is not consistently assessed across measures. As we consider the relationship between mood and MW, naturally, we would expect the subject matter of our thoughts to influence how we feel. Many studies in earlier research have explored MW content using experience sampling or thought probes, in which participants are interrupted throughout a given period. At each interruption, they would report what they had been just thinking about—ranging from open-ended responses to assigning thoughts into at one or more categories of future/past, self/other, positive/negative (e.g., Killingsworth and Gilbert, 2010; Smallwood and O'Connor, 2011; Shrimpton et al., 2017). Although these studies that probe the content of in-the-moment MW thoughts are informative in the “what” of inattention, it appears that many MW-related questionnaires about thoughts lean more toward querying the “how” of inattention; that is, the quality and characteristics of these thoughts rather than specific content per se, which may be cognitively informative at a deeper mechanistic level.

The “how” of inattention—how thoughts characteristically unfold or progress overtime—is informative for understanding thoughts processes, such as how thoughts become fleeting, maintained, or intensified, or how certain thinking styles or “habits” may be formed (or learnt) based on how frequently these mechanisms are repeated. In contrast, the “what” of inattention—what the specific substance of the thoughts are about at any given moment—provides a cross-sectional observation of only the output emerged or materialized from various dynamic thought processes that are based on countless possible inputs and pathways. In other words, whether thoughts are persistent and repetitive or meandering and free flowing, or whether they are generalized across situations or limited to specific situations, may be more illuminating than the object or subject of inattentionbecause examining thought patterns reveal the means rather than the end outcome (i.e., thought content), and may unveil more about any given individuals’ mind.

Using questionnaires in the present study as exemplars, rather than inquiring about subject matter during inattention, the “how” of mind wandering/inattention is assessed by behavioral outcomes or impacts of inattention such as forgetting information or tasks, day-to-day recurring mistakes. Comparably, the “how” of RNT subscales or measures is assessed by thinking styles or patterns such as the frequency, unpleasantness, and perceived uncontrollability of such cyclical thoughts. The “how” of reflection/contemplation subscales is assessed by the level of enjoyment one derives from analytically introspecting. Overall, while self-report surveys may be simplified as generally assessing trait-based qualities, and self-report lab probes assessing state-based real-time thoughts, beyond a multimethod approach per se, researchers may appreciate the “how” data gathered that can generate and inform research questions about inattention and associated consequences in general.

From a statistical point of view, a broad limitation is that factor analyses and related model development require a considerable amount of subjectivity to determine fit into either novel or existing theory. Specific to SEM, researchers have been cautioned against using cross-sectional mediation analyses due to biased estimates produced (e.g., Maxwell and Cole, 2007; Maxwell et al., 2011). The current study aims to preliminarily test and demonstrate that the construct of MW may heavily overlap with construct(s) of recurrent thought, and this has been replicated yet again correlationally. Nonetheless, although our model fit the data well based on the preceding EFA conducted, the generalizability of our data will need to be tested through replication. In addition, the cross-sectional nature of this investigation is a limitation for future research to examine causality, that is, whether propensity to mind-wander precedes susceptibility to recurrent thoughts, and consequently leads to negative mood symptoms. A data-driven approach may be useful in exploring underlying constructs or characteristics of inattention. However, the findings and conclusions reported here are undoubtedly limited by measures included in the study, which was not comprehensive or an exhaustive list. For example, regarding MW measures, three commonly used scales were selected to understand how MW has been frequently operationalized, but future studies may consider including questionnaires that have been more recently developed, such as the Mind-Wandering Scale: Deliberate and the Mind-Wandering Scale: Spontaneous (Carriere et al., 2013). Such next-generation scales suggest further consideration of different construals of MW may be fruitful, in terms of how they may or may not be complementary. Regarding RNTs and other clinically based measures, future studies may also want to consider “intrusive thoughts,” typically associated with obsessive thoughts. Likewise, when studying MW and its relationship to negative mood, it may also be best practice to include internal thoughts that are strongly correlated with low mood such as RNTs included in the present study.

Moreover, the many recurrent thoughts included in this study are associated with negative affect, and future studies may consider adding to the affective range when exploring inattention, such as thought constructs that are more neutral or positive (e.g., fantasy proneness; Merckelbach et al., 2001). In extension to affect, another next step may be to investigate how various types of off-task thinking operate in relation to each other, or in the presence of each other; for example, while inattention in the form of recurrent thoughts, especially RNTs, may be more related to affect consequences, inattention in the form of planning may be more related to cognitive consequences. Lastly, a study involving self-report of cognition will fall prey to issues of limited meta-awareness; nonetheless, as researchers delve further into inattention and its correlates, the relation between MW and negative moods can be clarified, and the mind’s narrative thinking processes may be ultimately uncovered to reveal unhelpful patterns of thinking and ameliorate the multidimensional consequences of inattention.