Assessing Perceived Barriers to Meditation: the Determinants of Meditation Practice Inventory-Revised (DMPI-R)

Abstract

Objectives

Engaging in meditation on a regular basis has been shown to enhance well-being. However, barriers to adopting it as a health behavior are poorly understood. The Determinants of Meditation Practice Inventory (DMPI) is an existing scale designed to measure perceived barriers to meditation. However, it was developed without factor analyses; thus, the dimensionality and construct validity of overall scale and/or subscale scores are unknown. Using factor analyses and tests of convergent validity, the present study explored the psychometric properties of the DMPI and presents a revised, psychometrically valid scale (The Determinants of Meditation Practice Inventory-Revised; DMPI-R).

Methods

Adult participants living in the USA (n = 621) provided data through an online survey platform. Extensive exploratory factor analyses were conducted (n = 311) and followed by confirmatory factor analysis (n = 310) on the best-fitting model. Convergent validity was estimated using the full sample data.

Results

Five items were removed because they demonstrated high residual variances and cross loaded onto multiple factors. Relationships among the remaining items were best explained by a four-factor structure with the following subscales: low perceived benefit, perceived inadequate knowledge, perceived pragmatic barriers, and perceived sociocultural conflict. Convergent validity was evidenced by associations between subscale scores and experiential avoidance, distress tolerance, and curiosity.

Conclusions

The multifactor structure of the DMPI-R indicates that there are multiple classes of perceived barriers on which people can vary. Validity analyses suggest that the DMPI-R is a promising measure of perceived barriers to meditation among North American adults.

Meditation is a mental training technique that can be employed to benefit psychological and physical well-being (Kabat-Zinn 1982, 2003). Results from a recent meta-analysis of 209 studies involving 12,145 participants suggest that meditation training is more effective in ameliorating physical and mental health symptoms than active control treatments like relaxation training, and equally effective as cognitive-behavioral therapy (Khoury et al. 2013). Further, engaging in meditation has been shown to enhance positive psychological qualities and experiences like self-compassion (Orzech et al. 2009), positive affect (Jislin-Goldberg et al. 2012) and self-esteem (Pepping et al. 2013). Meditation practices comprise core components of programs like mindfulness-based cognitive therapy for depression (MBCT; Segal and Teasdale 2013) and mindfulness-based stress reduction (MBSR; Kabat-Zinn 1982, 2009), as well as third wave therapies like dialectical behavioral therapy (Linehan et al. 2008). Individuals can also access meditation outside the clinical setting; for instance, through classes, community-based sanghas, and retreat centers (Ostafin et al. 2006).

Although engaging in a regular meditation practice confers many benefits, it appears that there are also barriers to regular practice. The basic instructions for meditation training are simple, but as noted by Kabat-Zinn (2014, p. 341), “meditation is not for the faint-hearted.” Barriers occurring in meditation practice are written about in the centuries-old Buddhists texts from which the present-day, secularized meditation training practices are derived (Bodhi 2005). Such texts include descriptions of unpleasant physical and emotional states arising during the practice, such as restlessness, sleepiness, and doubt, as well as challenging emotional states such as anger (Bodhi 2005). Contemporary scientific reports also suggest that there are barriers to practicing meditation, many of which overlap with those described in traditional Buddhist texts about meditation (Bodhi 2005). Participant reports of difficulties, challenges, and doubts related to meditation have been reported in qualitative work with college students (Sears et al. 2011), psychotherapists and therapy clients (Horst et al. 2013), practicing meditators (Lomas et al. 2015), health professionals (Cohen-Katz et al. 2005), and a variety of patient groups (Malpass et al. 2012).

Several prominent health behavior change theories (Ajzen 1991; Maiman and Becker 1974, Prochaska 2013) propose that individuals who consider adopting a new health behavior will perceive barriers that impede developing intentions and taking action to change. These barriers could involve social norms, low self-efficacy beliefs, negative outcome expectations, perceived inconveniences, and others. Further, barriers are thought to vary based on the type of health behavior one seeks to adopt (Prochaska 2013). Although there is a well-developed body of literature discussing barriers to health behaviors like physical exercise, healthy eating, and nonsmoking (Schutzer and Graves 2004; Shepherd et al. 2006; Twyman et al. 2014), barriers to meditation use are poorly understood. Given that meditation practice has been established as an evidence-based, health-enhancing behavior in recent decades, empirical studies exploring barriers to its use are needed.

A small number of investigations have been conducted exploring barriers to meditation. Empirical data obtained through these studies suggest that individuals who try meditation often question the adequacy of their meditation knowledge and skills. For instance, college students progressing through a semester-long mindfulness course reported doubting the correctness of their meditation knowledge or technique (Sears et al. 2011), as did adults of low socioeconomic status (SES) immediately following participation in two brief mindfulness exercises (Spears et al. 2017). Experienced meditators interviewed in a qualitative study (Lomas et al. 2015) also reported doubting their meditation skills.

Contacting painful internal states during meditation presents a barrier to many meditators, a finding which parallels the barriers to psychotherapy literature documenting the desire to avoid painful emotional material as a barrier to help-seeking (Komiya et al. 2000). As reported in Lomas et al. (2015), one experienced meditator stated, “you’re coming face to face with your own heart and mind, fear, anger, hatred, confusion, frustration and anxiety, all the difficult emotions…That’s the whole point…It was certainly challenging” (p. 853). In the same study, other participants reported feeling that meditating could even intensify difficult emotions like anxiety. Similarly, nursing professionals participating in an 8-week MBSR course and qualitative study (Cohen-Katz et al. 2005) reported that contacting painful emotions during the practice was challenging. They also described difficulty experiencing other unpleasant states like restlessness, sleepiness, and physical pain. College students trying to learn mindfulness meditation (Sears et al. 2011) and experienced meditators (Lomas et al. 2015) also described experiencing physical pain, tiredness, and boredom. Data obtained through a meta-ethnography (Malpass et al. 2012) documenting the experiences of individuals struggling with physical disease (e.g., cancer, chronic pain, HIV) participating in MBSR and MBCT programs also included accounts of participants struggling with, and sometimes feeling overwhelmed by, difficult emotions.

Practical barriers like time issues and logistical problems have also been identified as barriers to meditation practice. Nursing professionals participating in MBSR identified work-related scheduling conflicts, weather/driving issues and trouble finding time to do the MBSR homework as obstacles (Cohen-Katz et al. 2005). Time barriers have also been reported in qualitative studies of healthcare providers (Banerjee et al.2017) and mental health treatment-seeking adults of low SES (Spears et al. 2017) participating in mindfulness-based practices. As reported by Sears (2011), college students commonly reported having a lack of time and motivation to meditate.

Participants in several studies have reported noticing that their mind wanders frequently during meditation, and conceptualized this experience as a barrier to practice (Banerjee et al. 2017; Sears et al. 2011; Spears et al. 2017). Other obstacles reported in the empirical literature include doubting the value or benefits of meditation, feeling weird or strange while meditating, having unmet expectations related to the practice, and viewing meditation sessions as failures when unable to achieve a valued goal like relaxation (Malpass et al. 2012). Finally, feeling guilty for taking the time for oneself to meditate has also presented a barrier for some individuals (Cohen-Katz et al. 2005).

The Determinants of Meditation Practice Inventory (DMPI) was designed to assess perceived barriers to meditation among individuals without substantive prior meditation experience, or meditation novices (Williams et al. 2011). The DMPI item pool was developed through a literature review and interviews with experienced meditation instructors who were teaching in the United States (USA). In the interviews, meditation teachers were asked to describe their students’ perceived barriers to practice. The authors identified several barriers through the literature review and teacher interviews, which they then grouped into three content domains to guide item generation: perceptions and misperceptions, pragmatic concerns, and sociocultural beliefs. Of note, the content domains were not conceptualized as single factors, as each contained multiple constructs within it (see Williams et al. 2011, p. 17). Content validity of each item generated was established using iterative feedback from a panel of experts (i.e., clinicians, teachers, and meditation practitioners) and 10 community members. Through this process, the initial item pool was reduced to 17 items, which were then administered to a sample of 150 self-identified caregivers living in the northeastern USA who reported no prior experience with meditation. Caregivers were selected as the study population because they are known to be a highly stressed population who are likely to benefit from meditation.

Using the data provided by caregivers, Williams et al. (2011) reported an internal consistency reliability estimate of 0.87 and concluded that the items hold together in a cohesive fashion. They then used a total score to estimate test-retest reliability and construct validity. However, the authors did not employ factor analyses. Therefore, the dimensionality of the scale is unknown, and the validity and reliability estimates reported in Williams et al. (2011) may be inaccurate. The use of a total score rests on an untested assumption that the DMPI is a unidimensional scale (i.e., that barriers to meditation is a single, unified construct). However, it is possible that perceived barriers to meditation is a multidimensional construct, with novice meditators viewing barriers as falling into multiple specific domains. In such a case, a total score would not accurately measure perceived barriers and subscale scores would be needed to measure each domain. There are also cases where responses on psychometric scales are impacted by a general factor as well as multiple specific factors (i.e., measures characterized by a bi-factor structure). In this case, perceived barriers would be measured by a total score in addition to subscale scores. To psychometrically validate the DMPI, it is necessary to establish its dimensionality through factor analyses and establish the construct validity of the emergent scale and/or subscale scores. Additionally, it would benefit the DMPI to conduct construct validity tests on a larger, more diverse sample (i.e., one not restricted to caregivers) to enhance generalizability.

There are several constructs that might relate to perceived barriers to meditation and be useful in estimating construct validity. As evidenced by qualitative data (Cohen-Katz et al. 2005; Lomas et al. 2015; Malpass et al. 2012), a common barrier that new and experienced meditators experience is an unwillingness to contact difficult emotions and experiences, which is often referred to in the clinical literature as experiential avoidance. Experiential avoidance involves engaging in excessive negative evaluation of unwanted private events (e.g., thoughts, feelings) and having a general unwillingness to experience such events (Hayes et al. 2004). Distress tolerance, a related construct defined as the capacity to tolerate negative emotional states (Simons and Gaher 2005), may similarly relate to perceived barriers. Curiosity, or the extent to which people hold a general willingness to embrace uncertainty and to seek novel experiences (Kashdan et al. 2009), may relate to perceived barriers as suggested by data from the literature on barriers to professional help-seeking. Empirical findings suggest that individuals high in emotional openness tend to hold more favorable attitudes towards seeking therapy relative to those lower in emotional openness (Komiya et al. 2000). Like meditation, engaging in psychotherapy involves contacting and exploring internal experiences, both negative and positive. In a similar manner, it may be that those who are open and unresisting towards their experiences hold more favorable perceptions of meditation.

The purpose of the present study was to psychometrically validate the DMPI using factor analyses and tests of construct validity. Given that Williams et al. (2011) did not propose a specific factor structure and that little is known about barriers to meditation, we conducted extensive exploratory factor analyses allowing for one-, two-, three-, and four-factor models as well as two-, three-, and four-factor bi-factor models. Exploratory analyses were followed by confirmatory factor analysis on the best-fitting model. Convergent validity was tested with correlations between DMPI scores and experiential avoidance, distress tolerance, and curiosity. It was hypothesized that perceived barriers to meditation would correlate positively with experiential avoidance and negatively with both distress tolerance and curiosity. DMPI scores were also correlated with scores on five items written for the present study assessing perspectives on practicing meditation, specifically, personal interest in meditating, perceived difficulty of learning meditation, and perceived benefit of meditation. It was expected that participants reporting more favorable views of meditation on these items would also report lower levels of perceived barriers to practice.

Method

Participants

The total study sample consisted of adult Internet users (n = 621). Recommendations are that exploratory and confirmatory factor analyses be performed on data from different samples (Worthington and Whittaker 2006); in accord, the total sample was split in two by chronological order of study completion (i.e., the first half of completers were included in sample 1 and the subsequent in sample 2). Sample 1 (EFA sample) included 311 adults (41% male, 58% female, 1% transgender; mean age = 34.8 years). Racial and ethnic characteristics of the sample were the following: White/European American (78%), African American/Black (9%), Hispanic/Latino (6%), Asian American/Pacific Islander (3%), Asian-Indian/Pakistani (1%), Native American/Native Alaskan (1%), Biracial/Multiracial (1%), and Other (1%). The most commonly reported religious identity of participants was Christianity (26% Protestant; 21% Catholic), followed by Agnosticism (15%), Atheism (13%), and “spiritual but not religious” (11%). Other religion categories endorsed were non-identification with any religion (7%), others not listed in the survey (5%), Judaism (1%), Hinduism (1%), Islam (0.3%), and Buddhism (0.3%).

Sample 2 (CFA sample) consisted of 310 adults (29% male, 70% female and 0.3% transgender; mean age = 35.7 years). Racial and ethnic characteristics of the sample were the following: White/European American (75%), African American/Black (8%), Hispanic/Latino (a) (5%), Asian American/Pacific Islander (4%), Asian-Indian/Pakistani (.3%), Native American/Native Alaskan (2.3%), Biracial/Multiracial (5%). and Other (1%). The most commonly reported religious identity of participants was Christianity (25% Protestant; 19% Catholic), followed by Agnosticism (13%), Atheism (10%), and “spiritual but not religious” (12%). Other religion categories endorsed were non-identification with any religion (11%), others not listed in the survey (7%), Judaism (1%), Islam (1%), and Buddhism (2%).

Procedure

Participants were recruited via Amazon’s mechanical Turk (mTurk), an online survey administration platform. MTurk participants are typically paid between 10 and 15 cents for completing a battery of surveys (Buhrmester et al. 2011). All mTurk users (i.e., individuals living in the USA as well as international users) were eligible. MTurk users were invited to participate with a recruitment paragraph describing the study as addressing experience with and attitudes about meditation. After providing informed consent, participants completed the survey in a single sitting. Participants were compensated on average 26.6 cents per survey. All responses were downloaded through Qualtrics software and de-identified. The study was approved by the University of Maryland Institutional Review Board.

Measures

Demographics

Participants provided data on demographic characteristics and previous engagement with meditation. To report prior experience, participants responded yes or no to the following item: “Do you currently or have you previously practiced meditation activities?” If they responded affirmatively, they were directed to another question asking them to describe their experience level as one of the following: (a) novice, minimal practice; (b) intermediate, some practice that might be intermittent; or (c) advanced, regular, and extensive practice.

Determinants of Meditation Practice Inventory

Participants completed the 17-item DMPI measure (see Williams et al. 2011). The DMPI was designed to assess perceived barriers to meditation among individuals without prior meditation experience. Items are scored on 5-point Likert-type scales (1 = strongly disagree, 5 = strongly agree). The scoring instructions are to sum all items to yield a total score (range = 17–85), with higher scores reflecting higher levels of perceived barriers to meditation.

Experiential Avoidance

Experiential avoidance was assessed with the Acceptance and Action Questionnaire II (AAQ-II; Bond et al. 2011), a 7-item questionnaire with a one-factor structure. The AAQ-II has demonstrated adequate internal consistency (mean Cronbach’s alpha coefficient = 0.84) and test-retest reliability (3 month = 0.81; 12 month = 0.79) in community samples. In the present sample, Cronbach’s alpha was 0.90. The AAQ-II has also demonstrated adequate convergent and discriminant validity in community samples (Bond et al. 2011). Items are scored on 7-point Likert-type scales (1 = never true, 7 = always true). Possible total scores range from 7 to 49; higher scores indicate greater experiential avoidance.

Curiosity

Curiosity was assessed with the Curiosity and Exploration Inventory-II (CEI-II; (Kashdan et al. 2009). This 10-item instrument is scored for two subscales reflecting its two-factor structure: stretching (motivation to seek out knowledge and new experiences) and embracing (a willingness to embrace the uncertain and unpredictable nature of everyday life). Items are summed to create subscale scores and summed to create a total score. Scores on the stretching and embracing subscales have been shown to correlate highly, and it is therefore recommended to use the total curiosity score (Kashdan et al. 2009). Thus, the curiosity total score was used in analyses. Participants respond to items on 5-point Likert-type scales (1 = very slightly or not at all, 2 = a little, 3 = moderately, 4 = quite a bit, 5 = extremely). Total scores range from 10 to 50; higher scores reflect greater levels of curiosity. The CEI-II has demonstrated good internal consistency (Cronbach’s alpha = 0.85) and convergent validity in student samples (Kashdan et al. 2009). In the present study sample, Cronbach’s alpha was 0.91.

Distress Tolerance

Distress tolerance was assessed via the total score of the Distress Tolerance Scale (DTS), a 15-item instrument. The structure of the DTS is characterized by one general distress tolerance factor and four first-order factors (Simons and Gaher, 2005). Participants respond to items on 5-point Likert-type scales (1 = strongly disagree, 2 = mildly disagree, 3 = feel neutral, 4 = mildly agree, 5 = strongly agree). When completing the DTS, participants are prompted to think about specific times when they feel distressed or upset, and then respond based upon how they experience and respond to such distress. The total distress tolerance score is calculated by averaging the means of subscale scores. Possible total scores range from 1 to 5, with higher scores indicating higher tolerance for emotional distress. Adequate internal consistency has been demonstrated for the total score (Cronbach’s alpha = 0.82) and the four subscale scores (Cronbach’s alpha coefficients ranging from 0.70 to 0.82) in student samples. In the present study, Cronbach’s alpha for the total score was 0.91.

Perspectives on Meditation

Five Likert-type items were written for the present study assessing perspectives on meditation practice: (1) “How likely are you to seek an opportunity to meditate in the near future?” (1 = not at all likely, 7 = extremely likely); (2) “How interested are you in doing meditation?” (1 = not at all interested, 7 = extremely interested); (3) “How interested are you in learning more about meditation?” (1 = not at all interested, 7 = extremely interested”; (4) “To what extent do you think meditating would help you?” (1 = not at all, 7 = a great deal); (5) How difficult do you think it would be to learn meditation?” (1 = not at all, 7 = extremely difficult). It was expected that participants reporting more favorable perspectives towards meditation on these items would report lower levels of perceived barriers. Specifically, scores on items 1–4 were hypothesized to correlate negatively with DMPI scores, and scores on item 5 were expected to correlate positively with DMPI scores.

Data Analyses

Two thousand, one hundred and fifty-three mTurk users consented to participate in the study and were compensated. In data cleaning procedures, we removed cases from the dataset in the following order and for the following reasons: (a) duplicate cases (483 cases), (b) international participant status (743 cases), (c) missing data on all DMPI items (95 cases), (d) inaccurate responses to attention check questions (i.e., “please select the number “2” below”; please select the number “54” below; [25 cases]), and (e) report of substantial prior meditation experience (131 cases). The DMPI (Williams et al. 2011) was designed to measure barriers to meditation among individuals living in the USA. In order that the present results would best generalize to this population, we excluded international respondents. The DMPI was also designed to measure perceived barriers among individuals with little or no prior meditation experience. We therefore excluded participants who reported practicing meditation at an intermediate or advanced level.

Data cleaning procedures yielded 621 viable cases. The dataset was then split in half based on chronological order of data collection (see Participants section): The first half of the sample data was used in EFA (n = 311) and the subsequent half in CFA (n = 310). A recommended sample size for confirmatory factor analysis constitutes a 10:1 ratio of participants to parameter estimates (Worthington and Whittaker 2006); therefore, these sample sizes were likely adequate to perform the present analyses. All retained cases had complete data on the DMPI.

Full information maximum likelihood (FIML) exploratory factor analyses with oblique rotation were conducted allowing for one-, two-, three-, and four-factor exploratory models as well as two-, three-, and four-factor bi-factor models. Decisions on the optimal number of factors to extract were made using parallel analysis (Hayton et al. 2004). Fit indices along with recommended cutoff scores (i.e., SRMR < 0.08, RMSEA < 0.06, CFI > 0.95; Hu and Bentler, 1999) were reviewed for each exploratory model, as were the factor loadings, communalities, presence of cross loadings, and conceptual interpretability. Regarding factor loadings and communalities, recommendations to retain items with loadings at or above 0.40 and to remove items with communalities below 0.40 were used to guide decision-making (Worthington and Whittaker 2006). Low communalities suggest that latent factors of interest fail to explain a substantive portion of variance in the items; such items can be considered for deletion (Worthington and Whittaker 2006).

Full information maximum likelihood confirmatory factor analyses were performed on the best-fitting model. The confirmatory model was evaluated using recommended fit indices and cutoff scores (i.e., SRMR < 0. 08, RMSEA < 0.06, CFI > 0.95; Hu and Bentler 1999). Modification indices were considered when consistent with theoretical conceptualizations of DMPI items (Schreiber et al. 2006). All factor analyses were conducted using MPlus (Muthén and Muthén 2012). Convergent validity was estimated using the full sample data (n = 621) in SPSS v22.

Results

Exploratory Factor Analyses

Parallel analysis suggested that four-factor solutions were optimal. Therefore, solutions under closest consideration were a four-factor exploratory model and a bi-factor exploratory model with one general barrier to meditation factor and three correlated, specific factors. However, poor model fit was evident for these solutions (chi square = 212.08, df = 74, p < .001, CFI = .90, RMSEA = 0.08, 90% CI = 0.07–0.09; SRMR = 0.04). Both models were also characterized by several cross loadings and low factor loadings. Most notably, the communalities for several DMPI items were extremely low (i.e., less than .30). In accordance with recommended practices (see Worthington and Whittaker2006), low communality items were sequentially removed based upon the magnitude of each communality, beginning with the lowest. With each iteration, the parallel analysis was conducted, and the loadings, communalities, and fit indices associated with each solution were evaluated. This process was repeated until the communalities were acceptable, yielding 12 items. At this juncture, all communalities were above the cutoff level of 0.40 except for the item “meditation might be boring,” which approached 0.40 (i.e., 0.37). This item was retained due to its high factor loading onto a single factor and conceptual interpretability in conjunction with the other items. Next, the parallel analysis was re-run, which supported a four-factor solution. The best model was the four-factor model (chi square = 29.81, df = 24, p = .19; RMSEA = 0.03, 90% CI = 0.00–0.059; CFI = 0.99; SRMR = 0.02). Each item indicated one of four factors titled low perceived benefit, perceived inadequate knowledge, perceived pragmatic barriers, and perceived sociocultural conflict (see Table 1). Factor intercorrelations ranged from 0.04 to 0.27 (see Fig. 1).

Table 1.

Standardized Factor Loadings Obtained in the Best-fitting Exploratory Model

Item	Low Perceived Benefit	Perceived Inadequate Knowledge	Perceived Pragmatic Barriers	Perceived Sociocultural Conflict
1. I prefer to be accomplishing something.	0.60	−0.02	0.17	−0.08
2. Meditation might be boring.	0.69	0.18	−0.08	0.03
3. It is a waste of time to sit and do nothing.	0.72	0.04	0.03	0.19
4. I don’t know much about meditation.	0.02	0.98	−0.02	−0.05
5. There is no quiet place where I can meditate.	−0.11	0.13	0.67	0.10
6. I don’t have time.	0.25	−0.02	0.66	0.10
7. There is never a time when I can be alone.	0.01	−0.04	0.92	0.02
8. I wouldn’t know if I were doing it right.	0.00	0.63	0.07	0.05
9. I’m concerned meditation will conflict with my religion.	−0.03	−0.02	−0.02	0.79
10. My family would think it was unusual.	0.04	0.13	0.17	0.50
11. I don’t believe meditation can help me.	0.57	−0.03	−0.02	0.33
12. I wonder if meditation might harm me.	0.03	−0.03	0.01	0.74

Note: Italics are added to emphasize which items indicate which factor.

Fig. 1.

Visual representation of the final model with factor intercorrelations. *p < .05; **p < .01

Exploratory Bi-Factor Models

Two, three, and four-factor bi-factor models were examined using data obtained from the 12 items. Each of these models was characterized by low factor loadings, items cross loading onto more than one specific factor in addition to the general factor, and poor conceptual interpretability. Under closest consideration was the four-factor bi-factor; however, given the aforementioned limitations, it was gauged as inferior to the four-factor exploratory model. Thus, the four-factor EFA model was tested in confirmatory analyses.

Confirmatory Factor Analyses

Means, standard deviations, and correlations among items are shown in Table 2.

Table 2.

Inter-item Pearson’s correlations and descriptive statistics on data used in CFA

Item	Mean	SD	1	2	3	4	5	6	7	8	9	10	11	12
1. I prefer to be accomplishing something.	3.26	1.14	1.00
2. Meditation might be boring.	2.77	1.25	0.43	1.00
3. It is a waste of time to sit and do nothing.	2.34	1.15	0.43	0.70	1.00
4. I don’t know much about meditation.	3.20	1.16	0.29	0.41	0.34	1.00
5. There is no quiet place where I can meditate.	2.58	1.33	−0.01	0.20	0.17	0.17	1.00
6. I don’t have time.	2.87	1.33	0.17	0.35	0.39	0.21	0.48	1.00
7. There is never a time when I can be alone.	2.55	1.30	0.07	0.25	0.20	0.15	0.73	0.55	1.00
8. I would not know if I were doing it right.	2.94	1.24	0.27	0.35	0.31	0.59	0.30	0.23	0.25	1.00
9. I’m concerned meditation will conflict with my religion.	1.65	0.97	0.05	0.13	0.25	0.07	0.04	0.07	0.11	0.08	1.00
10. My family would think it was unusual.	2.10	1.14	0.17	0.26	0.33	0.21	0.14	0.20	0.17	0.25	0.42	1.00
11. I don’t believe meditation can help me.	2.26	1.15	0.22	0.49	0.46	0.25	0.11	0.21	0.20	0.20	0.34	0.34	1.00
12. I wonder if meditation might harm me.	1.60	0.95	0.01	0.12	0.25	0.14	0.05	0.13	0.08	0.13	0.65	0.39	0.37	1.00

Model fit for the four-factor model was marginally acceptable (chi square = 136.78, df = 48, p < .0001; RMSEA = 0.08, CI = 0.06–0.09; CFI = 0.91, SRMR = 0.08). Modification indices were reviewed for theoretical plausibility and estimated impact on model fit. Error terms were allowed to covary for “I wonder if meditation might harm me” and “I’m concerned meditation will conflict with my religion.” Similarly, error covariances were included for “There is never a quiet place where I can meditate” with “There is never a time when I can be alone.” These error covariances were added sequentially. The final model fit was acceptable (chi square = 98.06, df = 46, p < .0001; RMSEA = 0.06, 90% CI 0.04–0.08; CFI = 0.95, SRMR = 0.05). Standardized loadings ranged from 0.50 to 0.91 and were statistically significant at the < .0001 level(see Table 3). All factors were significantly correlated with one another (range = 0.27–0.55; p’s < .01). A visual representation of the final model is shown in Fig. 1.

Table 3.

Standardized factor loadings and standard errors obtained in CFA along with item means

Factor	Estimate	SE	Mean
Low perceived benefit
1. I prefer to be accomplishing something.	0.50	0.05	3.27
2. Meditation might be boring.	0.84	0.03	2.78
3. It is a waste of time to sit and do nothing.	0.83	0.03	2.34
4. I don’t believe meditation can help me.	0.57	0.06	2.26
Perceived inadequate knowledge
5. I don’t know much about meditation.	0.79	0.06	3.20
6. I would not know if I were doing it right.	0.75	0.06	2.94
Perceived pragmatic barriers
7. There is no quiet place where I can meditate.	0.53	0.07	2.59
8. There is never a time when I can be alone.	0.60	0.08	2.56
9. I don’t have time.	0.91	0.09	2.87
Perceived sociocultural conflict
10. I’m concerned meditation will conflict with my religion.	0.50	0.08	1.64
11. I wonder if meditation might harm me.	0.48	0.08	1.60
12. My family would think it was unusual.	0.82	0.09	2.10

All loadings were significant at the .0001 level

Subscale scores corresponding to the four factors were computed by averaging the scores on items representing the factor. Total DMPI scores were not computed because EFA and CFA results suggested that there was no general, over-arching perceived barriers to meditation factor.

Construct Validity

It was hypothesized that AAQ total scores and DMPI scores would be positively correlated. This hypothesis was supported by significant, positive correlations between all DMPI subscales and AAQ total scores (see Table 4). It was also hypothesized that DMPI scores would negatively correlate with total scores on the CEI-II. These hypotheses were partially supported by significant, low-magnitude, negative correlations between low perceived benefit and perceived inadequate knowledge with CEI-II scores. Finally, it was hypothesized that DMPI scores would correlate negatively with DTS scores. This hypothesis was also partially supported by significant correlations between the subscales perceived inadequate knowledge and perceived sociocultural conflict with DTS scores (see Table 4).

Table 4.

Descriptive statistics and convergent validity coefficients for the Determinants of Meditation Practice Inventory-Revised

			Convergent validity coefficients
DMPI-R Subscale	No. of items	M (SD)	AAQ-II	CEI-II	DTS
Low perceived benefit	4	2.69 (0.91)	0.09*	− 0.08*	− 0.02
Perceived inadequate knowledge	2	3.16 (1.06)	0.16**	− 0.11*	− 0.13**
Perceived pragmatic barriers	3	2.70 (1.10)	0.10*	0.03	− 0.06
Perceived sociocultural conflict	3	1.79 (0.83)	0.19**	− 0.03	− 0.10*

n = 621

AAQ-II Acceptance and Action Questionnaire II, CEI-II Curiosity and Exploration Inventory II, DTS Distress Tolerance Scale

^*p < .05;

^**p < .01

Correlations were computed between DMPI subscales and Likert-type items written for the present study assessing perspectives on meditation. In support of the hypothesis that individuals reporting lower perceived barriers would report more favorable perspectives towards meditation, significant correlations were observed between each of these items and at least one of the four DMPI-R subscales (see Table 5).

Table 5.

Correlations between DMPI-R subscales and items assessing perspectives on meditation

	Low perceived benefit	Perceived inadequate knowledge	Perceived pragmatic barriers	Perceived sociocultural conflict
How likely are you to seek an opportunity to meditate in the near future?	− 0.46*	0.10	− 0.10	− 0.05
How interested are you in doing meditation?	− 0.55*	0.15*	− 0.02	− 0.18*
How interested are you in learning more about meditation?	− 0.51*	0.26*	0.01	− 0.14*
How difficult do you think it would be to learn meditation?	0.05	0.33*	0.13**	0.02
To what extent do you think meditating would help you?	− 0.51*	0.18*	0.04	− 0.21*

^*p < .05;

^**p < .01

Discussion

The purpose of the present study was to establish the psychometric properties of the DMPI using factor analyses and tests of construct validity. Given the rapid increase in the use of meditation as an evidence-based clinical intervention, coupled with emerging empirical and clinical evidence of barriers to meditation, the present study addresses an important gap in the literature. As discussed, the DMPI was initially developed by Williams et al. (2011) in the absence of factor analysis, and therefore, the dimensionality and construct validity of emergent scale and/or subscale scores were unknown. Results from the present study support the use of a revised 12-item scale (Determinants of Meditation Practice Inventory-Revised; DMPI-R) assessing the following four constructs: low perceived benefit, perceived inadequate knowledge, perceived pragmatic barriers, and perceived sociocultural conflict (see Supplemental Material). Results did not support the presence of a global factor representing perceived barriers but indicated that there are a number of relatively distinct types of perceptions of barriers on which people can vary.

Five items were deleted as part of exploratory factor analyses due to failure to load onto any single-factor and high residual variances. However, these items might indicate important attitudes about meditation which could be explored in future revisions or extensions of the DMPI-R. Perhaps there were insufficient numbers of items in the original pool (Williams et al. 2011) representing these attitudes, precluding their inclusion on the DMPI-R. Two deleted items in particular, “I am uncomfortable with silence” and “I can’t stop my thoughts,” are so frequently noted by meditation teachers as problematic beliefs among new meditators (Williams et al. 2011) that we recommend administering them to patients along with the DMPI-R to prompt clinical discussions. The distorted belief that one must “get past” a wandering mind in order to meditate has been noted by novice meditators in other studies as well (Banerjee et al. 2017; Sears et al. 2011; Spears et al. 2017), pointing to the importance of assessing this perceived obstacle.

Construct validity analyses supported the convergent and discriminant validity of the four subscales. As described, several qualitative studies have documented that contact with challenging internal experiences presents a barrier to meditation (Cohen-Katz et al. 2005; Lomas et al. 2015; Malpass et al. 2012). Relatedly, the tendency to avoid contact with internal experience has previously been identified as a barrier to professional help-seeking (Komiya et al. 2000). As such, it was hypothesized that those individuals who habitually attempt to avoid or attenuate unpleasant emotions, and/or those having a lower tolerance for emotional distress, would report higher perceptions of barriers to meditation relative to those who are more open to internal experience. In support of these predictions, positive correlations representing small to moderate effect sizes were observed between each subscale on the DMPI and experiential avoidance. Experiential avoidance was the one construct assessed in validity analyses that related to each type of perceived barriers, indicating that the tendency to avoid unpleasant internal experiences co-occurs with the tendency to hold a negative view of meditation across multiple domains.

Negative correlations representing small effect sizes were observed between two DMPI-R subscales (i.e., perceived inadequate knowledge and perceived sociocultural conflict) and distress tolerance, a measure of the capacity to tolerate negative emotions (Simons and Gaher 2005). It may be that perceiving oneself as inadequately prepared to meditate and as having a sociocultural conflict with meditation brings up negative affects (e.g., anxiety) more so than perceiving meditation as being of little benefit and oneself as lacking the time to meditate. Meanwhile, perceived inadequate knowledge was negatively related to curiosity. Perceiving oneself as lacking the knowledge to meditate may bring up a sense of uncertainty, and more curious individuals with higher uncertainty tolerance (Kashdan et al. 2009) may be less concerned about their knowledge level. Individuals higher in curiosity also tended to view meditation as being of greater potential personal benefit relative to individuals lower in curiosity. It appears that the open-mindedness characterizing curiosity may help people consider meditation as a potentially beneficial activity. Of note, correlations between DMPI-R subscales and the measures used to assess convergent validity were of small to medium magnitude, suggesting that the DMPI-R and these measures assess constructs that are related but also distinct. Conceptually, the DMPI-R is unique in that it assesses four distinct areas of perceived barriers to meditation, and thus has the capacity to support further research on perceived barriers and how to help patients overcome them (e.g., by educating patients about the benefits of meditation).

In addition to psychological constructs assessing the relationship with internal experience, participants also responded to several items written for the present study addressing perspectives on meditation. Correlations between these items and DMPI-R subscales bolster the construct validity of the DMPI-R as a measure of perceived barriers to meditation. Specifically, participants scoring higher on low perceived benefit also tended to report feeling less inclined to seek future opportunities to meditate and a lower interest in doing meditation. This is consistent with theories describing perceived benefits as important in health behavior adoption (Ajzen 1991; Prochaska 2013). Also, respondents who tended to perceive meditation as more difficult to learn also tended to view their knowledge of meditation as lacking and pragmatic barriers as problematic. These findings are consistent with theoretical frameworks identifying perceived difficulty of learning a health behavior as a barrier to change (Ajzen 1991; Prochaska 2013). Finally, respondents who scored higher on perceived sociocultural conflict also reported lower interest in learning more about meditation as well as seeking an opportunity to meditate. These results accord with theoretical and empirical literature describing sociocultural beliefs and norms as influential on health behavior engagement (Ajzen 1991; Ball et al. 2010; Prochaska 2013).

Limitations and Future Research

The DMPI-R provides a brief, parsimonious measure of perceived barriers to meditation in novice meditators and demonstrates a pattern of high factor loadings and acceptable convergent validity. An important direction for future extensions or revisions of the DMPI-R, and/or the development of new measures, is to explore potentially untapped constructs that deleted DMPI items may have represented. Further, the subscale perceived inadequate knowledge had only two items. Of the four subscales, perceived inadequate knowledge is most closely related to self-efficacy, a construct representing one’s perceived ability to carry out a task (Bandura 1982). Self-efficacy has been theorized and documented as a strong predictor of health behavior adoption (Bandura 2004; Strecher et al. 1986). In future extensions of the DMPI-R, meditation self-efficacy should be represented more thoroughly.

This study relied on an mTurk sample, and limitations of this approach should be noted. The use of mTurk in psychological research is controversial and suggested safeguards to promote data quality in mTurk studies have been offered (see Cheung et al. 2017). In accord, the present study included two instructed response items to detect inattentive responding, and cases demonstrating inattentiveness were removed. Also, some have argued that demand effects may be particularly liable to impact data in mTurk studies and recommend that researchers advertise their studies in general terms to reduce their influence (Paolacci and Chandler 2014). Consistent with this suggestion, the present study was advertised as an effort to understand attitudes about meditation rather than about barriers per se. Finally, given that mTurk samples do not represent the general US population (Paolacci et al. 2010), an important future direction for the DMPI-R is to explore its psychometric properties in additional samples (e.g., samples representing other diverse populations, nationally representative samples) to expand its generalizability. However, despite its limitations, the present research supports progress in understanding how to help people successfully adopt meditation, a practice empirically shown to enhance well-being, into daily life.