Assessing Facets of Mindfulness in the Context of Work: The Mindfulness@Work Scale as a Work-Specific, Multidimensional Measure of Mindfulness

Overview of Existing Mindfulness Measures and Their Applicability in the Work Context

Research into mindfulness in the context of work has predominantly been conducted with the MAAS (Mesmer-Magnus et al., 2017). The fact that with 15 items it is relatively short, has good psychometric properties, and that items do not involve Buddhist concepts or wording and are easy to understand for meditation-naive participants likely has contributed to its popularity in the work-related mindfulness literature. However, the MAAS is a unidimensional scale focusing on the awareness component of mindfulness exclusively. Mindfulness scholars have therefore expressed concerns about the construct validity of the MAAS as it may not sufficiently capture all essential aspects of mindfulness (Bergomi et al., 2012; Grossman, 2011). Another point of concern is that the MAAS consists of reversed scored items exclusively. It has therefore been criticized that the MAAS captures an “agitated lack of attentiveness to daily life” (p. 405; Grossman, 2008) and operationalizes mindfulness as the converse of inattentiveness or absent-mindedness (Bergomi et al., 2012; Grossman, 2008). Recently, a similar one-dimensional scale has been developed, capturing the awareness component at work (Zivnuska, Kacmar, Ferguson, & Carlson, 2016).

Originating predominantly from the clinically oriented mindfulness literature, a number of mindfulness scales have been developed in the past that represent different subfacets of mindfulness (for overviews see Bergomi et al., 2012; Sauer, Walach, Schmidt, Hinterberger, Lynch, Büssing, & Kohls, 2012). Yet, all of them have shortcomings regarding their usefulness in work settings. Some have a clear clinical orientation, such as the SMQ (Chadwick, Hember, Symes, Peters, Kuipers, & Dagnan, 2008) that focuses on how people deal with distressing thoughts, making the scale useful for the treatment of mental health problems. Notably, although the SMQ includes items representing different subfacets of mindfulness, it only has a single-factor structure and does therefore not yield any facet scores. Similarly, the primary purpose of the multi-faceted CAMS/CAMS-R (Feldman, Hayes, Kumar, Greeson, & Laurenceau, 2007) is the use in psychotherapy, particularly the treatment of depression (Hayes & Feldman, 2004). Accordingly, various items refer to a tendency of preoccupation and worrying (e.g. “I am preoccupied by the future”; “I am preoccupied by the past”) (Bergomi et al., 2012). Furthermore, due to a strong higher-order factor and low internal consistencies of subfacets, the measure yields a total mindfulness score but no facet scores. Although being truly multi-faceted, the KIMS (Baer et al., 2004) builds on a conceptualization of mindfulness as described in dialectical behavioral therapy (Bergomi et al., 2012) and therefore also has a predominantly clinical orientation. The TMS (Lau, Bishop, Segal, Buis, Anderson, Carlson, Shapiro, Carmody, Abbey, & Devins, 2006) has been designed to measure mindful states retrospectively in reference to mindfulness meditation practice (Baer, 2011). It is therefore not suitable as a general measure of trait mindfulness in the context of work. The FMI (Walach et al., 2006) has roots in Buddhist psychology (Sauer et al., 2012) and contains items that are unsuitable to be presented to participants in a professional work context. Such items include “I see how I create my own suffering”; “I notice how my emotions express themselves through my body”. Similar concerns can be raised with items from the PHLMS (Cardaciotto, Herbert, Forman, Moitra, & Farrow, 2008) such as “When I walk outside, I am aware of smells or how the air feels against my face”, the KIMS (Baer et al., 2004) “When I take a shower or bath, I stay alert to the sensations of water on my body”, and the FFMQ (Baer et al., 2006) “I notice how foods and drinks affect my thoughts, bodily sensations, and emotions”. Another concern regarding the applicability of the FFMQ is the unclear factor structure (Baer et al., 2006; Williams, Dalgleish, Karl, & Kuyken, 2014; van Dam, Earleywine, & Danoff-Burg, 2009). The five-factor structure emerging in meditator samples could not be replicated in non-meditators. Furthermore, items belonging to the observing facet have been shown to function fundamentally differently in samples with meditators and non-meditators suggesting that they are interpreted in different ways by these groups of participants (Baer, 2011; Bergomi et al., 2012). A mindfulness measure whose factor structure changes depending on meditation experience is of concern when one wants to use this scale in samples drawn from the general working population that typically includes a wide range of participants with different degrees of meditation experience. Finally, with 39 items it is relatively long, further limiting the applicability in the work context. Taken together, although these scales have proven utility in the mindfulness meditation and clinical mindfulness literature, their usefulness and applicability in the work context is limited. In the present work, we therefore set-out to develop a work-related multidimensional mindfulness scale.

Mindfulness and Components of Mindfulness

Although there is not one universally accepted definition of mindfulness (Van Dam et al., 2018), mindfulness scholarship has highlighted the role of distinct components of mindfulness including acceptance (Bishop et al., 2004), describing, acting with awareness (Baer et al., 2004), non-judgment, or non-doing (Leary & Tate, 2007). In the present study, we chose to follow the most widely accepted multidimensional operationalization of mindfulness (cf. Mattes, 2019; Quaglia, Braun, Freeman, McDaniel, & Brown, 2016) as it is assessed in the Five Facet Mindfulness Questionnaire (FFMQ; Baer et al., 2006). The FFMQ includes the five dimensions of (a) observing, (b) describing, (c) acting with awareness, (d) nonreactivity to inner experience, and (e) nonjudging of experience. Describing, acting with awareness, nonreactivity, and nonjudging components have been found to emerge consistently in meditator as well as in non-meditator samples, have been shown to load onto a higher-order mindfulness factor, and were associated with theoretically related constructs (e.g. psychological symptoms) in meaningful ways (Baer et al., 2006, 2008; Williams et al., 2014).

Arguably, the most central component of mindfulness is awareness of internal and external experiences. “Describing” builds on this central component and captures the ability to describe, label, or note observed phenomena by applying words. Simply put, describing involves the ability to “tune in” to one's current experience and find the words to describe it.

Rather than doing multiple things at the same time, and constantly losing focus, “acting with awareness” involves being attentive and fully engaged in one's current activity. Mindfulness has sometimes been referred to as “single tasking” and can be considered the antithesis of multitasking. A mindful person is able to remain focused and to redirect attention back to the original object of attention when distracted (Bishop et al., 2004).

Mindfulness is based on the principle that it is not aversive experiences themselves that are problematic, but rather the way people relate to these states. By means of nonreactivity, a more tolerant and flexible relationship with difficult experiences can be cultivated. “Nonreactivity to inner experience” involves taking an accepting stance towards thoughts and feelings by allowing them to come and go without reacting to them or getting caught up in them. Nonreactivity can be considered a self-regulation strategy where the individual refrains from acting impulsively by taking an observing stance towards feelings, thoughts and urges.

Finally, “nonjudging” means experiencing thoughts and feelings without judging them or criticizing oneself for having them. Instead of judging and labeling experiences in a dichotomous way, mindfulness involves paying full attention to external events and internal thoughts, feelings, and sensations with a non-judgmental attitude of curiosity and acceptance (Baer, 2003). In this way, mindfulness has been suggested to cultivate a more compassionate stance towards the self and others (Shapiro & Schwartz, 2000). Nonreactivity and nonjudging capture the attitudinal foundations of mindfulness (Peters, Smart, Eisenlohr-Moul, Geiger, Smith, & Baer, 2015) that describe how attention is paid to present-moment experiences, a fundamental aspect of mindfulness that needs to be considered in addition to attentional components (i.e., act with awareness) to encompass the essence of mindfulness (Bishop et al., 2004; Shapiro et al., 2006).

Bear et al.’s (2006) study showed that these four components were internally consistent and moderately intercorrelated, suggesting distinct but related constructs (see also Williams et al., 2014). Furthermore, these four subdimensions of mindfulness have been shown to be differentially related to psychological outcomes (Baer et al., 2008; Cash & Whittingham, 2010).

Notably, a dimension termed “observing” has also been proposed as a component of mindfulness in the literature (Dimidjian & Linehan, 2003). Observing refers to monitoring or attending to various internal (e.g. cognitions, bodily sensations) and external stimuli (e.g. sounds, smells) in the present moment. Observing originates from what mindfulness practitioners are encouraged to do when practicing mindfulness, namely paying careful attention to internal sensations (e.g. their location, intensity) and sounds (e.g. their pitch, volume) (Baer et al., 2004). However, empirical investigations of the dimensionality of the FFMQ have shed doubt on the construct validity of the observing facet in trait measures of mindfulness (cf. Rudkin, Medvedev, & Siegert, 2018). Although developed as a five-factor questionnaire, confirmatory factor analyses of the FFMQ repeatedly revealed the best fit for a four-factor hierarchical structure (Baer et al., 2006; Williams et al., 2014). Furthermore, items of the observing facet showed different loadings on a higher-order mindfulness factor depending on the type of sample used (Williams et al., 2014) and the observing facet did not show consistent convergent validities with other mindfulness facets (Siegling & Petrides, 2016). Moreover, patterns of relationships of the observing facet with well-being and strain outcomes differed in strength and sometimes even sign between non-meditator and meditator samples (Baer et al., 2008; Rudkin et al., 2018). Repeatedly, the observing subfacet has yielded no or even positive correlations with psychological symptoms and strain in non-meditator samples (Rudkin et al., 2018). Such observations are not unique to the observing facet of the FFMQ but have also been documented for observing items of other mindfulness questionnaires (Rudkin et al., 2018). Seeking to offer explanations for these counterintuitive findings, researchers have suggested that items included in observing facets are interpreted differently by meditators and non-meditators (Rudkin et al., 2018; Williams et al., 2014). Specifically, participants familiar with mindfulness meditation may correctly understand observing items as reflecting pure awareness without evaluation or judgement. In contrast, for participants without meditation experience, attending to and observing experiences might involve evaluation and judgement, which stands in contrast to mindful observing and reflects maladaptive, exaggerated, and self-critical self-focus instead (Rudkin et al., 2018).

Researchers have therefore concluded that the observing facet is not well suited to assess mindfulness in non-meditator samples or to track changes in mindfulness during participation in mindfulness training programs (Williams et al., 2014) and even suggested dropping this facet from mindfulness measures altogether (Baer, 2011; Williams et al., 2014), arguing that “observing items may provide misleading or confusing findings in non-meditating samples” (Baer, 2011, p. 253). In the present work, we therefore focused on developing a work-related mindfulness scale capturing only the four dimensions of describing, nonreactivity, nonjudging, and act with awareness.

Convergent and Discriminant Validity

To further assess the convergent and discriminant validity of the M@Work scale, we chose constructs that are theoretically distinct, albeit conceptually related to mindfulness.

Criterion-Related Validity

Another important step in the development of a scale is providing criterion-related validity evidence, indicating the extent to which a construct is related to theoretically derived criteria (Cronbach, 1955; Hinkin, 1998). We therefore investigated the relationship of the M@Work Scale with a wide variety of criteria that are relevant in the context of work. In doing so, we included well-being-, motivation-, and relationship-related criteria.

Incremental Validity—The Benefits of a Comprehensive Operationalization of Mindfulness

Differential Validities of Subfacets of Mindfulness

As outlined in the introduction, the third goal of the present work was to test whether a multi-faceted measure of mindfulness at work helps unravel differential validities of subdimensions of mindfulness in predicting work outcomes. This may deepen theoretical insights into why and how mindfulness relates to specific work outcomes (Liang et al., 2017; Mesmer-Magnus et al., 2017). While it is beyond the scope and feasibility of the present paper to develop specific hypotheses of differential relationships of mindfulness sub-facets with all 21 outcomes, we focus on patterns of relationships for subgroups of outcome variables that can be expected on the basis of mindfulness theory and previous empirical findings in the mindfulness literature.

Overview of Scale Development Phases and Validation Strategy

The development and validation of the M@Work Scale involved several phases. In the Pilot Phase, we developed a large pool of items, conducted preliminary analyses to assess content validity, reduce the number of items and adapt items. In Phase 1, the final set of items was selected, and factorial structure and psychometric properties were assessed. Phase 2 served to validate the M@Work Scale by assessing convergent and discriminant and criterion-related validity. In Phase 3, we assessed the incremental validity of the M@Work Scale over the most widely used unidimensional scale of mindfulness in the context of work. Finally, in Phase 4, we assessed differential validities of subfacets of mindfulness.

Study 1

Study 1 relied on an English-speaking MTurk sample of people who worked at least 10 hours a week (n = 252). The study was advertised as a 10-minute study on work and well-being. Participants could earn $1 in return. According to information provided by TurkPrime, 86 percent of participants were Caucasian, 17 percent were Black/African American, 7 percent Asian, and 1 percent Multiracial. A total of 254 started the survey, two of which were omitted from analysis as they stopped filling in the survey after the first questions. Participants held different types of jobs including self-employed people, clerks, accountants, and sales representatives.

Study 2

Study 2 consisted of two German convenience samples of working adults recruited by two research assistant teams. Both teams used a range of different recruitment strategies including their personal network, advertisements via social networking sites such as Facebook, Instagram, and the German career platform Xing. The snowballing technique was used to increase outreach (King, O’Rourke, & DeLongis, 2014). Furthermore, in Sample A employees of a German engineering company were invited for participation via the company's HR department. In Sample B flyers were handed out individually. Requirement for participation was employment during the time of data collection and a minimum of 10 work hours per week. No compensation was offered in return for participation.

Study 3

Study 3 was an English-speaking sample collected via TurkPrime. It was advertised as a study on how individuals experience their work. The study took 10 minutes on average, and participants were offered $1 in return. Individuals who were between 19 and 66 years of age and employed for at least 10 hours a week were eligible for the study. A total of 249 participants completed the survey. According to information provided by TurkPrime, 76 percent of participants were Caucasian, 7 percent Hispanic/Latino, 7 percent Black/African American, 7 percent Asian and 2 percent chose the answer option “other”. Thirty-one participants either failed to answer correctly to at least one of two attention check items or took less than 4 minutes to take the survey and were therefore omitted from analyses, resulting in a final sample of 218 participants. Type of jobs included in the sample were manifold, including secretaries, software developers, consultants, sales managers, and teachers.

Study 4

Study 4 was an English-speaking sample collected by contacting subscribers to a positive psychology website via mail. It was advertised as a study on work experiences, well-being and work behavior. The study took approximately 10 minutes and participants were offered participation in a raffle of five $50 Amazon gift vouchers. The study invitation was mailed to about 90,000 individuals of which 2,644 started taking the survey. Fourteen cases were removed because they dropped out before filling in the M@Work scale items. In addition, 28 cases were removed because they indicated not to be native English speakers and not to have understood the items due to their English language skills. An additional 22 cases were deleted because they indicated to be retired, students, or unemployed, resulting in a final sample of 2,580 participants. Participants were from Asia, North America, South America, Africa, Australia, and Europe and held a wide range of different jobs including teachers, psychotherapists, sales managers, therapists, social workers, nurses, or accountants.

Study 5

Study 5 was an English-speaking sample collected via the same website using the same procedure as described in Study 4. To minimize the burden for participants and keep questionnaires short, we created three questionnaire versions. All three versions assessed the same demographics and the M@Work Scale. In addition, they contained different outcome measures. Demographics and the M@Work Scale were therefore completed by the full sample while outcome measures were completed by different subsamples. Notably, in versions A and B of the questionnaire we used a 5-point agreement scale. To assess potential differences in item characteristics and factor structure due to a different answer scale, a 6-point frequency scale was used in Version C.

Overall, 1,948 individuals started taking the survey. A total of 931 dropped out, did not complete the M@Work Scale and were therefore deleted. In addition, 12 cases were dropped because participants indicated to have poor English language skills and 18 were dropped because they indicated to be retired, students, or unemployed, resulting in a final sample of 987 participants. Participants were from Asia, North America, South America, Africa, Australia, and Europe and held a variety of jobs including nurses, clinicians, psychologists, counsellors, clerks, HR managers, teachers, and professors.

Study 6

To supplement cross-sectional findings from Studies 1 to 5, Study 6 was a two-wave study with a time lag of 3 weeks. The sample was English-speaking and data was collected with Prolific. Participants were offered £0.70 for completion of the first survey, £0.75 for filling in the second survey and a bonus payment of £0.40 if both surveys were completed. Of the 410 participants that filled in the time 1 survey, 8 were dropped while of the 378 people filling in the time 2 survey, 3 participants were dropped because they failed to fill in the attention check correctly. This resulted in a final sample of 370 individuals with matched valid surveys at time 1 and time 2. Participants were predominantly from the United Kingdom and had a wide variety of jobs.

Pilot Phase: Initial Item Development and Preliminary Item Reduction and Revision

Focusing on the above-mentioned theoretical components of mindfulness, the two authors independently generated English items reflecting the four facets of mindfulness of describing, act with awareness, nonjudging of experience, and nonreactivity to inner experience. We focused on these four mindfulness facets as they have yielded a clear factor structure and patterns of relationships with outcome variables in meditator and non-meditator samples in previous research (Baer et al., 2006; Williams et al., 2014). Considering repeated criticism of existing mindfulness scales consisting of reversed items exclusively (e.g. Grossman, 2011) and recommendations from measurement experts to include items worded in different directions (Weijters, Baumgartner, & Schillewaert, 2013), we generated items that were worded in the direction of mindfulness as well as reversed items. In doing so, items were either newly developed or selected from existing mindfulness scales (e.g. FFMQ, MAAS, KIMS) and adapted to the work context. This resulted in a pool of 69 items. 3

To establish content validity and reduce the number of items, seven mindfulness experts, including practitioners and/or researchers, were provided with a brief definition of every mindfulness facet and were asked to rate the adequacy of each item to assess this facet in the context of work on a 5-point Likert scale ranging from not adequate to very adequate (for a similar procedure see Brown & Ryan, 2003). Based on the V statistic (Aiken, 1985), we only retained items that were rated highly and consistently across raters. Furthermore, highly similar items were dropped. The remaining 31 items 4 were presented to seven English native speakers unfamiliar with mindfulness training to assess the extent to which the items were clear and understandable. Again, the V statistic was used to identify unclear items. As a result, some items were revised to improve item wording.

These items were then used in a pilot study (Study 1), with which we performed an exploratory factor analysis (EFA) using maximum likelihood extraction and oblique rotation. A parallel analysis (Horn, 1965; O’Connor, 2000; 31 items, n = 252, 200 correlation matrices) suggested the extraction of four factors. An inspection of the factor loadings yielded a pattern that poorly mirrored the expected factor structure. Specifically, while positively worded nonreactivity and act with awareness items loaded on the first factor, negatively worded nonreactivity, nonjudging, and act with awareness items loaded on the second factor, describing items loaded on the third factor, and the fourth factor consisted of positively worded nonjudging items. This pattern suggests that reversed-keyed items distorted the expected factor structure, a common finding in factor models including items varying in coding direction (Weijters & Baumgartner, 2012).

To address this issue we consulted the item development literature and implemented the following changes to our item pool. First, we dropped or reformulated items with particle negations (i.e., using the particle “not”) following recommendations to avoid using negated items when constructing reversed items (Weijters & Baumgartner, 2012; Weijters et al., 2013). As recommended, however, we retained reversed items with antonymic expressions, for example, “When I have negative feelings at work, I get angry or disappointed that this happens to me” (Weijters & Baumgartner, 2012; Weijters et al., 2013). Finally, because of relatively low reliability of some facets and to be able to choose from a larger pool of items, a few new items were developed assessing the nonreactivity and nonjudging facets of mindfulness. This resulted in a set of 36 parallel German and English items that were the starting point for a series of studies aiming to further reduce the number of items and establish the dimensionality, convergent, discriminant, criterion-related, and incremental validity of the M@Work Scale and differential validities of its subscales.

Phase 1: Final Item Selection, Factor Structure and Psychometric Properties

The revised set of 36 items was used in Study 2. Using Sample A, we inspected item characteristics and conducted a series of analyses with the goal of selecting a smaller subset of items that adequately captures all four dimensions of mindfulness (describing, nonreactivity, nonjudging, and act with awareness). Specifically, we first fitted single-factor models using CFA to identify items with loadings lower than .5 on the intended factor (for a similar approach see Sonnentag & Fritz, 2007). As a result, 10 items were removed. 5 Next, we inspected means and skewness of items and removed one item with a particularly high mean of 4.08 and skewness of −1.02. With the remaining 25 items, we performed a 5-factor CFA specifying the four a priori dimensions of mindfulness and loadings of all reversed items on a method factor. We included the method factor in order to account for method effects of reverse-keyed items (Weijters et al., 2013). The four substantive factors were allowed to correlate among each other but not with the method factor (Motl & DiStefano, 2002). This model yielded a good fit (χ² = 314.059, df =255, p < .01; CFI = .969; TLI = .964; RMSEA = .029). As we aimed for a parsimonious scale, three more items were dropped that were similar in item wording to other items of the same dimension.

Using the remaining 22 items (see Table 3), we conducted another confirmatory factor analysis (CFA). We specified a five-factor model consisting of the four expected substantive factors representing the four sub-facets of mindfulness and a method factor for reversed items. This model yielded a good fit (χ² = 238.919, df =190, p < .01; CFI = .970; TLI = .964; RMSEA = .030). Intercorrelations between factors ranged from .26 (describing with nonreactivity) to .48 (describing with nonjudging and with act with awareness) and are depicted in Table 5. 6

We also tested alternative CFA models (see Table 4), including a two-factor model in which all items loaded on a single substantive factor and on a method factor (Model B), and two three-factor models: Model C1 consisted of the method factor, a factor on which all describing, nonjudging and act with awareness items (the three most highly correlated factors) loaded on one and nonreactivity on another factor. Model C2 consisted of the method factor, a factor with loadings from all describing and nonjudging items and another factor representing awareness and nonreactivity items. Furthermore, a four-factor model was specified (Model D), with a combined describing and nonjudging factor, and separate act with awareness and nonreactivity factors. Finally, we tested a higher-order model (Model E) which corresponded to Model A and included a higher-order factor that loaded on the four mindfulness factors. Results of nested model comparisons are depicted in Table 4, revealing that the a priori model (Model A) fit the data significantly better than all alternative models. Overall, these results provide evidence for the construct validity of our measure.

The two best fitting models (Model A and Model E) were cross-validated in all other studies comprising six different samples. Results are depicted in Table 4 and revealed that across all samples both factor solutions provided good fit to the data with CFI values ranging from .923 to .970 and RMSEA values ranging from .062 to .033. In one sample Model A and E did not differ significantly in terms of fit, while in the other samples Model A displayed a slightly but significantly better fit than Model E. Table 5 displays factor intercorrelations between the four subfacets of the M@Work scale in all samples (based on Model A). These were moderately high.

As can be seen from Table 5, the four subfacets of the 22-item version M@Work Scale displayed good internal consistency, with Cronbach's alpha values ranging from .71 to .87. Also the overall scale displayed good internal consistency (Cronbach's alpha ranging from .86 to .91; see Table 2). Almost identical values were obtained when using omega as an alternative reliability index (Dunn, Baguley, & Brunsden, 2014). We also assessed the test-retest reliability of the M@Work scale in Study 6, a two-wave study including a time lag of 3 weeks. Test-retest reliability was good, with values of .79 for the overall scale and values of .60 (nonjudgement), .64 (describing), .71 (nonreactance), .74 (act with awareness) for the respective subscales.

Invariance of Factor Structure across Participants with and without Meditation Experience

Considering shortcomings of previous multi-faceted scales that yielded different factor structures in meditation versus non-meditation samples (cf. Bergomi et al., 2012), we sought to establish measurement invariance of the M@Work Scale across these two groups of participants. In Study 5, participants therefore indicated their experience with mindfulness meditation on a scale ranging from 1 (no experience), 2 (I have tried it in the past but stopped), 3 (I practice occasionally, i.e., a couple of times per month), 4 (I practice regularly, i.e., once per week) to 5 (I practice almost every day). Using this self-report item, we assigned participants indicating a 1 or 2 to the no-meditation group and people indicating a 3, 4, or 5 to the meditation group. To test for measurement invariance of the M@Work Scale across these two groups, we first conducted separate CFAs of Model A in the no-meditation and in the meditation group (van de Schoot, Lugtig, & Hox, 2012). Model fit was good in both models (no meditation group, n = 978: χ² =432.627, p < .001; CFI = .96, TLI = .95, RMSEA = .04, SRMR = .04; meditation group, N = 1,602: χ²=505.169, p < .001; CFI = .97, TLI = .97, RMSEA = .03, SRMR = .03). Next, we tested for scalar invariance constraining factor loadings and intercepts to be equal across groups in a multigroup CFA. The model yielded a good fit (χ² = 1005.454, p < .001, CFI = .97, TLI = .96, RMSEA = .03, SRMR = .04) suggesting that both the meaning of mindfulness and levels of underlying items are equal across both groups (van de Schoot et al., 2012).

Phase 2: Validation of the M@Work Scale

Criterion-Related Validity

Interpersonal Relations.

The M@Work Scale was indeed significantly positively related to OCB-I (r = .14, p < .05) and to team-member exchange (r = .30, p < .001) and negatively to enacted incivility (r = −.34, p < .001). While all subdimensions were significantly related to team-member exchange and enacted incivility in expected directions, only the nonreactivity dimension was significantly related to OCB-I. Hypothesis 9 was thus partially supported.

Phase 3: Incremental Validity over One-Dimensional Mindfulness Measures

Incremental validity was investigated by means of a hierarchical regression analysis in which the MAAS was entered in the first step and the M@Work Scale facets were entered in the second step. As can be seen from Table 7, out of a total of 16 investigated relationships, the M@Work scale displayed significant incremental validity above and beyond the MAAS for 13 relationships. The incremental amount of variance that was explained by the M@Work Scale ranged from 4 percent (family to work conflict) to 12 percent (psychological work strain). Considering Cohen's rule of thumb (Cohen, 1988), the incremental validity explained by the M@Work Scale over the MAAS corresponded to a medium effect (R² = .09–.25) for seven outcome variables: job satisfaction, psychosomatic complaints, psychological work strain, exhaustion, disengagement, team-member exchange, and intrinsic motivation. Only for life satisfaction, sleep quality and OCB-I, the M@Work Scale did not explain incremental variance above and beyond the MAAS. Hypothesis 10 was thus predominantly supported.

Phase 4: Differential Validities of Subfacets of Mindfulness

To assess the differential validity of mindfulness dimensions, we conducted two sets of analyses. First, we conducted relative importance analyses to examine the extent to which each of the four sub-facets of the M@Work Scale drives the prediction of work outcomes. Relative importance analysis has been argued to overcome shortcomings of a traditional regression approach in appropriately partitioning variance to various predictors and provides an accurate account of the contribution each predictor makes in explaining criterion variance (Tonidandel & LeBreton, 2011). Specifically, we used relative weights analysis providing information on the total amount of variance explained by all four facets jointly (total R²) and on the impact of each facet relative to the other facets (Tonidandel & LeBreton, 2011; 2015). Table 8 depicts the rescaled relative weights (i.e., scaled as the percentage of explained variance). Second, we conducted a traditional hierarchical regression analysis to demonstrate (a) that the act with awareness component explains unique variance in positive psychological states and in enacted incivility above and beyond the other three components, and (b) that nonreactivity and nonjudgement explain unique variance in indicators of psychological distress above and beyond the other two components.

As can be seen from Table 8, nonjudging and/or nonreactivity had stronger relative weights in predicting indicators of psychological distress (i.e., negative affectivity, perceived stress, psychosomatic complaints, psychological work strain, exhaustion, and disengagement) than the other two mindfulness components. Furthermore, the amount of incremental variance explained in these outcomes by nonreactivity and nonjudgement was higher than the amount of incremental variance explained by act with awareness. Conversely, relative to the other components, act with awareness had the largest relative weight in predicting positive psychological states (i.e., positive affectivity, vitality, job satisfaction, and work engagement) and enacted incivility. In addition, the incremental validity explained by the act with awareness component over the remaining components was larger than the incremental validity explained by nonjudging and nonreactivity. An exception to this general pattern was life satisfaction for which nonjudging had a larger relative weight than act with awareness. Taken together, Hypotheses 11 and 13 were fully supported while Hypothesis 12 was predominantly but not fully supported.

Psychometric Properties and Criterion-related Validity of the M@Work Scale

Results across the seven samples supported the psychometric properties of the M@Work scale as a multidimensional scale with good internal consistency and retest reliability of the respective dimensions and the overall score. A series of confirmatory factor analyses revealed that a correlated first-order factors model fit the data well. A similar model including an additional higher-order mindfulness factor also yielded good fit statistics in all samples. However, the higher-order model only fit the data equally well as the correlated first-order factors model in one sample while the higher-order model yielded a slightly worse fit than the correlated factors model in the other samples. Taken together, findings thus lend support to the multi-faceted nature of the M@Work Scale and provide mixed support for the existence of a higher-order mindfulness factor.

Importantly, findings also confirmed the invariance of the factor structure across participants with and without meditation experience. Such measurement invariance is a crucial prerequisite when one wants to use a scale with samples drawn from the general working population where some participants might, while others might not, have experience with mindfulness meditation. It is also a necessary precondition for studies aiming to track changes in mindfulness as a result of a mindfulness training intervention where researchers need to ensure that it is not the meaning and interpretation of items that change over time but the standing of participants on the construct as a result of training.

Correlations of the M@Work Scale with the MAAS, emotional intelligence, neuroticism, conscientiousness, and with present time perspective provided support for the convergent and discriminant validity of the M@Work Scale. In line with the theoretical links between mindfulness and these constructs, correlations were strong and significant without exceeding .70, speaking to the conceptual distinctiveness of the M@Work Scale from these other constructs and measures. Consistent with theoretical work on mindfulness proposing positive links with employee well-being, motivation, and interpersonal work relations (Good et al., 2016; Shapiro et al., 2015), the M@Work Scale was significantly related to a broad range of well-being outcomes, with most effect sizes being medium (r = .30) to large (r = .50; Cohen, 1992). Moreover, the M@Work Scale was positively related to indicators of interpersonal work relations, for example, team-member exchange, and positively to autonomous forms of motivation such as intrinsic motivation and learning-goal orientation, while negatively to the two types of performance goal orientations. With the exception of OCB-I, effect sizes were medium to large, demonstrating the good criterion-related validity of the M@Work Scale.

Theoretical Contributions

Apart from developing a reliable and valid multi-faceted mindfulness scale that is applicable to the work context, our findings contribute to mindfulness theory and research in several ways.

First, although mindfulness scholarship predominantly holds a multi-faceted perspective on mindfulness (e.g. Baer et al., 2006; Bishop et al., 2004; Shapiro et al., 2015), the most widely used measure of mindfulness in the context of work, the MAAS, is unidimensional and focuses exclusively on the attention and awareness component of mindfulness, while ignoring attitudinal aspects of mindfulness that characterize the quality of attention and are therefore a fundamental part of the overall construct of mindfulness (Shapiro et al., 2006). Although research using this unidimensional operationalization has provided important first insights into the role of mindfulness for work outcomes, the effects of mindfulness for work outcomes might have been underestimated by equating a sub-aspect with the overall construct and ignoring the contribution of other aspects that are inherent to mindfulness. By investigating the incremental validity of the multidimensional M@Work Scale over the unidimensional MAAS, the present paper addressed this issue using 16 different outcomes across the three outcome domains of well-being, motivation, and interpersonal relations. Overall, findings confirmed that the M@Work Scale displayed incremental validity in predicting a range of well-being, motivational, and relational outcomes above and beyond the unidimensional MAAS. For many outcomes, including job satisfaction, perceived stress, psychosomatic complaints, psychological work strain, burnout, team-member exchange and intrinsic motivation, the size of the incremental variance explained even amounted to a medium effect (Cohen, 1988), speaking to the importance of embracing a multi-faceted perspective on mindfulness. The incremental validity was predominantly driven by the attitudinal subfacets of the M@Work Scale that are not covered by the MAAS—that is, nonreactivity and nonjudging, and to a lesser extent also by describing. However, it is notable that the act with awareness component of the M@Work Scale also contributed significantly to the incremental validity of the M@Work scale over the MAAS, especially for work-related outcomes (i.e., job satisfaction, work engagement, disengagement, work to family conflict, and enacted incivility). This is likely to be explained by the contextualization of the M@Work Scale and it is in line with findings from the personality psychology literature showing that contextualized trait measures show stronger validities than non-contextualized trait measures when predicting outcomes from the same context (Bing et al., 2004; Lievens et al., 2008).

Second, the present paper provides insights into the specific role of subfacets of mindfulness in predicting work outcomes and thereby furthers our theoretical understanding of the active ingredients of mindfulness that drive relations with important work outcomes. While researchers have highlighted the need to get more insights into the distinct roles of sub-aspects of mindfulness for different work outcomes (Mesmer-Magnus et al., 2017), we are aware of only one study that has addressed this issue. Focusing on workplace aggression as an outcome, Liang and colleagues (2017) documented that awareness and acceptance components had differential predictive validities. With the present investigation, we extend their findings by considering other work-related outcomes and providing more evidence for differential predictive validities. Specifically, our findings revealed that it was the act with awareness facet that was related to subsequent levels of enacted incivility while the other facets were not. This is in line with findings from Liang and colleagues (2017) that only act with awareness was related to workplace aggression—another, more severe, interpersonal deviance outcome. Furthermore, our findings showed that the act with awareness component carries the strongest weight in explaining variance in positive well-being outcomes (i.e., positive affectivity, vitality, job satisfaction, and work engagement). In contrast, nonreactivity and nonjudging played an active role in reducing psychological distress (i.e., negative affectivity, perceived stress, psychosomatic complaints, psychological work strain, and burnout). These findings are in line with findings reported in the clinical literature showing that nonjudging and nonreactivity were inversely related to symptoms of anxiety and depression in treatment-seeking adults (Desrosiers, Klemanski, & Nolen-Hoeksema, 2013). Mindfulness scholars have highlighted the role of attention and attitude in mindfulness (Shapiro et al., 2006). While attention refers to the ability to pay attention to present-moment internal and external experiences, attitude refers to the quality of attention. While attention in general can be critical, mindful attention is characterized by being open, accepting and nonjudging. The present findings underline that the attitudinal qualities of mindful attention, that is, nonreactivity and nonjudgement, play a particularly important role in reducing psychological distress. In contrast, attentional aspects (i.e., act with awareness) seem to be more important in driving positive psychological (work) experiences and hampering interpersonal aggression. These findings not only help refine mindfulness theory, but may also be of particular utility when designing and implementing organizational mindfulness interventions targeting specific outcomes. Our findings suggest that while the focus should be on the act with awareness component when targeting the reduction of workplace deviance or seeking to improve work satisfaction and engagement, one should also focus on attitudinal components (i.e., nonreactivity and nonjudging) when targeting the reduction of work strain and distress.

The M@Work Scale

Findings presented above suggest that the M@Work Scale is a reliable and valid instrument that captures the four mindfulness facets of describing, nonreactivity, nonjudging, and act with awareness. The language used and contextualization makes it suitable to be used in the context of work and with working populations, both for research and practical applications. Designed as a multi-faceted scale, the primary purpose is the use of individual facet scores, but an overall score may also be used, depending on whether one is theoretically more interested in the overarching construct of mindfulness or in the role of its individual subcomponents. Practical applications are manifold: The scale may be used as an intake questionnaire for mindfulness training programs, to track (differential) changes in mindfulness facets as a result of training, as a tool for self-reflection as part of occupational health interventions, or for personnel development purposes.

As the FFMQ served as a guiding framework in developing our scale, the M@Work Scale naturally bears similarities with the FFMQ, but it also differs in some key ways: First and foremost, we did not include an observing facet, considering well-documented problems with the construct validity of this facet, especially in non-meditator samples. In doing so, we followed the advice of mindfulness measurement scholars to omit observing items in the development of mindfulness questionnaires that target diverse samples including meditators and non-meditators (Baer, 2011; Williams et al., 2014). Our scale therefore includes only four of the five facets of the FFMQ. As a result of this, the M@Work Scale displays measurement invariance across participants with and without meditation experience. Furthermore, the M@Work Scale does not include items without spiritual wording or a clinical orientation, it is contextualized to the work context, and with 22 items it is rather short, compared to the 39-item FFMQ. As can be seen from Table 3, of the 22 items included in the M@Work Scale, the final scale includes 10 items that were newly developed and 12 that were adapted from existing measures (i.e., the SMQ, the MAAS, and the KIMS), 10 of which are also part of the FFMQ.

Limitations and Future Directions

The present set of studies bears some limitations that should be considered when drawing conclusions from our findings. First, Studies 1 to 5 relied on cross-sectional self-report data, which may have led to biases due to common method variance and to an overestimation of covariation between the M@Work Scale and outcome measures (Podsakoff, MacKenzie, Lee, & Podsakoff, 2003). As the goal of the present work was to develop and validate a new scale, cross-sectional self-report methodology is a necessary and adequate first step in doing so. In order to preclude that our results were primarily driven by biases we conducted an additional multi-wave study (Study 6) thereby separating the measurement of the M@Work Scale from the assessment of outcome measures. Effects obtained and the amount of variance explained were similar to results of the cross-sectional findings.

Due to the correlational nature of our data, we abstained from drawing causal conclusions (Spector, 1994). However, theoretical arguments on the link between mindfulness and work outcomes (Good et al., 2016) and previous empirical evidence on effects of mindfulness interventions on work outcomes using randomized controlled trials (Hülsheger et al., 2013; Hülsheger, Feinholdt, & Nübold, 2015; Slutsky, Chin, Raye, & Creswell, 2019; Wolever et al., 2012) speak to a causal effect of mindfulness on many outcome measures. Notably, however, effects may also be reciprocal, which may be investigated by using cross-lagged panel designs in future research.

Conclusion

To conclude, the present research demonstrates that the M@Work Scale is a reliable and valid measure of mindfulness in the context of work. Being contextualized to the work context and multi-faceted it helps advance the literature on mindfulness at work by allowing researchers (a) to assess mindfulness comprehensively, and (b) to investigate the differential impact of subfacets of mindfulness for work outcomes. The M@Work Scale thereby lays important foundations for future work on mindfulness in organizations. It provides researchers with a means to dig deeper and generate novel theoretical insights into which specific components of mindfulness drive relations with well-being, relational and motivational outcomes and develop targeted interventions.