Development of the Beliefs About Yoga Scale

Abstract

Beliefs about yoga may influence participation in yoga and outcomes of yoga interventions. There is currently no scale appropriate for assessing these beliefs in the general U.S. population. This study took the first steps in developing and validating a Beliefs About Yoga Scale (BAYS) to assess beliefs about yoga that may influence people’s engagement in yoga interventions. Items were generated based on previously published research about perceptions of yoga and reviewed by experts within the psychology and yoga communities. 426 adult participants were recruited from an urban medical center to respond to these items. The mean age was 40.7 (SD = 13.5) years. Participants completed the BAYS and seven additional indicators of criterion-related validity. The BAYS demonstrated internal consistency (11 items; α = 0.76) and three factors emerged: expected health benefits, expected discomfort, and expected social norms. The factor structure was confirmed: χ² (41, n = 213) = 72.06, p < .001; RMSEA =.06, p = .23. Criterion-related validity was supported by positive associations of the BAYS with past experiences and future intentions related to yoga. This initial analysis of the BAYS demonstrated that it is an adequately reliable and valid measure of beliefs about yoga with a three-factor structure. However, the scale may need to be modified based on the population to which it is applied.

Introduction

Empirical research supporting the benefits of yoga is increasing. Yoga has been shown to be effective for the management of chronic health conditions (e.g., asthma, back pain, cancer, depression, heart disease) and for improving overall quality of life.^1–6 Additionally, yoga is gaining popularity in the United States. Recent survey data indicated that the national use of yoga has increased from 5.1% of the population in 2002 to 6.1% in 2007.⁷ Thus, approximately 13 million adults were participating in yoga in 2007, and this number is likely to be even higher now. However, it would seem that a great number of people who could benefit from yoga are not receiving its benefits. Therefore, it is important to understand why some people do decide to participate in yoga while others do not.

Beliefs about yoga may contribute to this decision. Social learning theory, the integrative model of behavior, and the theory of reasoned action all posit that behavioral beliefs—defined as generalized expectancies about the outcome of a specific behavior—are a necessary component in understanding behavioral decision making.^8–12 Identifying behavioral beliefs about yoga would be a step toward understanding how people decide whether to participate in yoga.

Three published measures, all developed by researchers in India, assess constructs similar to beliefs about yoga. The Kaivalyadham Yoga Attitude Scale ¹³ and a modified version of this scale, the P.G.I. Yoga Attitude Scale,^{14, 15} are only available in the Hindi language. Another scale, the Yoga Attitude Scale (YAS), has been translated into English.¹⁶ However, the YAS consists of items that are not likely to be appropriate for the general person’s understanding of yoga in the United States (e.g., “I believe in my inner strength, hence I need not practice yoga”^{16, p. 49}). Additionally, the items in the YAS are more complex than is recommended, with an overall Flesch-Kincaid Readability Grade-level of 7.6 and some “double-barreled” wording.¹⁷ If used in the United States, the YAS would likely have low predictive value given cultural differences and the complex wording of the scale.

The current study took the first steps toward developing and validating a scale to assess behavioral beliefs about practicing yoga in the general U.S. population.^{17, 18} Knowledge of beliefs about yoga will lead to an increased understanding of who decides to practice yoga and why. This scale could also be applied to improve recruitment and retention in yoga intervention research and to better address the beliefs of new yoga students in general.

Methods

Participants

A convenience sample of 426 participants was recruited from outside the cafeteria at a medical center in New York City. The appropriate institutional ethical approval was obtained prior to commencing recruitment, and informed consent was obtained from each participant. Study personnel at a table asked passersby to share their thoughts about yoga in a quick survey. Participants completed a 44-item questionnaire without identifying information in exchange for a small incentive (candy). The questionnaire was written in English. Therefore, those who were unable to read and write in English were excluded. A majority of the sample recruited was female (74%), highly educated (81% college-graduates), and married or living with a partner (54%), with a mean age of 40.7 (SD = 13.5) years. 62% of the sample self-identified as White, 19% Black, 9% Asian/Indian, 6% Asian or Pacific Islander, and 4% other; ethnically, 17% were Hispanic. Additionally, 77% were employees of the hospital, 12% visitors, 8% students, and 3% patients.

Procedure

This study followed the eight steps for scale development as presented by DeVellis ¹⁷ and further recommended by Worthington and Whittaker.¹⁸ The procedure for addressing these steps is discussed in detail below.

1. Clearly specify what is being measured. Beliefs about practicing yoga are defined here as the expectancies about outcomes from yoga practice.¹² Beliefs about practicing yoga were assessed instead of beliefs about yoga in general because beliefs about performing a behavior are more likely to be related to the actual behavior than beliefs about the general idea of yoga.^{10, 11}

2. Generate an item pool. Thirty items were generated from a published qualitative study that used focus groups to identify benefits, barriers, and cues to action of yoga practice in the United States. ¹⁹ Thus, these items are likely shaped by U.S. culture. Common themes identified by this study were then worded in the form of an expectancy or belief (e.g., “If I practiced yoga, I would be embarrassed in the class”). An effort was made to write items that were worded simply and directly.¹⁸ The Flesch-Kincaid Readability Grade-level for the item pool was 4.9.

3. Determine the format for measurement. Items were rated on a Likert scale ranging from 1 (extremely unlikely) to 7 (extremely likely). This measurement approach aimed to capture salient beliefs ²⁰ and allow for more statistically sophisticated analyses of the measure, while limiting participant burden.²¹

4. Have experts review the initial item pool. Both health psychology experts (Montgomery, Schnur) and yoga instructors (Daly, Johnson Mork, Sohl) were consulted to review the original item pool and to improve the quality of the items.¹⁷

5. Consider inclusion of validation items. Additional items were included in the questionnaire that assessed experience with and impressions of yoga. These items, listed below, served as a check of criterion-related validity.¹⁷

   Attitude towards yoga was assessed with one item: “If I practiced yoga, it would be (circle a number closer to the adjective that you believe best describes your thoughts about practicing yoga),” on a scale from 1 (harmful) to 7 (beneficial).

   Intention to attend a free class was assessed with one item: “Assuming you have access to a free yoga class in the future, what is the probability that you would attend?” on a scale from 1 (extremely unlikely) to 7 (extremely likely).

   Experience with yoga was assessed with a yes/no item: “Have you ever done yoga before?”

   Satisfaction with previous yoga experience was assessed by asking participants “How satisfied were you with your most recent yoga experience?” from 1 (not satisfied at all) to 7 (as satisfied as I could be).

   Recent practice of yoga was assessed with two yes/no items: “Have you done a yoga video in the past month?” and “Have you attended a yoga class in the past month?”

   Willingness to pay was asked directly: “How much would you be willing to pay for one yoga class?”

   Salient beliefs were captured with an open-ended item that asked participants to complete the sentence “If I practiced yoga, I…”. This item was designed to capture any prevalent beliefs about yoga that may have been overlooked in the item generation process.

   Demographic information was also collected, including age, sex, ethnicity, marital status, level of school completed, and affiliation with the hospital (e.g., employee, patient, visitor, student).

6. Administer the items. Scale items were administered to a self-selected sample of participants who agreed to complete a questionnaire (as described above). The relatively large sample size (N = 426) should lessen the influence of individual subject variance.¹⁷

7. Evaluate the items. Items were first evaluated for skewness and kurtosis; all items were normally distributed. Negatively worded items were reverse scored so that a high score on the measure assessed positive beliefs about yoga. Half of the sample was randomly selected for exploratory factor analysis. Based on the results of these exploratory analyses, a confirmatory factor analysis was validated using the data from the remaining half of the sample.

8. Optimize the scale length. The initial exploratory factor analysis used Principal Component Analysis (SPSS 15.0), the Promax Rotation Method with Kaiser Normalization with an oblique solution, and allowed for the number of factors to be determined by Eigenvalues >1. The items were theoretically created to capture expected benefits, negative outcomes, normative beliefs, and efficacy beliefs about the practice of yoga; however, we did not impose an a priori factor structure. Guidelines were set for systematically eliminating items as previously recommended¹⁸ and implemented.²² Items were removed based on the following criteria in the order presented: communality less than 0.04; negative-factor loading; factor loading less than 0.4; less than three items on a factor; conceptual inconsistency. An exploratory confirmatory factor analysis was further conducted on the first sample using structural equation modeling (LISREL 8.80) to establish a clear factor structure. Additional items were dropped within this analysis as indicated by the highest modification indices until there were no remaining significant standardized residuals. Lastly, this model was evaluated with a confirmatory factor analysis using the data from the validation sample.

Descriptive analyses and criterion-related validity were evaluated with descriptive statistics, Pearson’s correlations, and independent-samples t-tests (SPSS 15.0). All variables were normally distributed and there were no outliers. Additionally, the open-ended item was rated by two independent coders (Daly and Suslov) and Cohen’s kappa coefficient was calculated (SPSS 15.0) to determine inter-rater reliability.

Results

Factor Analyses

Using 30 items, the initial exploratory analysis resulted in eight factors (KMO = 0.84, Bartlett’s Test of Sphericity, χ² (435) = 1997.37, p <.001). As dictated by the guidelines outlined above, three items (I would injure my joints or muscles; I would not be following my religious beliefs; It would cost too much money) were removed because they each loaded solely on separate factors (see discussion for further consideration of these items). Rerunning the same analysis resulted in a six-factor structure. Two items had factor loadings of less than 0.40 (There would be all kinds of people in a class; My friends would think I was strange) and were thus also removed. The item I would meet friends in a class was the sole item on a factor. The item I would feel overweight in a class had a low factor loading. Two items, Most people who are important to me would approve and I would have to do unrealistic pretzel-like postures were cut because they were not conceptually consistent with the other items on their respective factors. Removing these items resulted in a five-factor structure; however, two of the factors were highly related and conceptually similar. A factor analysis was conducted using structural equation modeling in the same sample to determine if placing these items on the same factor was supported statistically.

This analysis consisted of 21 items divided into four factors. Further items were deleted in the following order as suggested by the largest modification indices (I would be able to do it; It would be boring; I could find a class I would like; The postures taught could be adapted for my body; The instructor would be accepting of my skill level; It would be difficult to find a good instructor; It would give me time to myself; I would improve my balance; It would take up too much time; I would be less stressed). Each change resulted in a significant improvement as indicated by the χ² change. This resulted in an adequately fitting 11-item, three-factor model χ²(41, n = 213) = 67.03, p = .006; RMSEA = .06, p = 0.35, with good internal consistency (see Table 1).

Table 1.

The Beliefs About Yoga Scale (BAYS) with factor loadings.

If I practiced yoga …			Sample 1	Sample 2
			(n = 213)	(n = 213)
Factor 1:	Expected benefits	α =	0.83	0.84
	It would help me focus.		0.74	0.79
	It would improve my overall health.		0.69	0.89
	It would help me gain self-awareness.		0.84	0.78
	It would help me sleep better.		0.71	0.71
	I would become more flexible.		0.50	0.51
Factor 2:	Expected discomfort	α =	0.67	0.63
	The teacher would make me uncomfortable. (−)		0.63	0.46
	I wouldn’t be good at it. (−)		0.51	0.62
	I would be embarrassed in class. (−)		0.80	0.75
Factor 3:	Expected social norms	α =	0.62	0.48
	There would only be women in a class. (−)		0.57	0.27
	There would only be “new age” people in a class. (−)		0.61	0.60
	I would have to be more flexible to take a class. (−)		0.56	0.67
	Total α =		0.76	0.75

Note: (−) reverse scored items

The independent confirmatory factor analysis with the second half of the sample also supported this model, χ²(41, n = 213) = 72.06, p = .002; RMSEA = .06, p = .23. The item loadings and Cronbach’s alphas for the second sample are also displayed in Table 1. The intercorrelations among and internal consistency of the resulting three factors are shown in Table 2. The 11 items on the final scale had a Flesch-Kincaid Readability Grade-level of 5.1.

Table 2.

Intercorrelations and internal consistency among the factors in confirmatory sample.

	1	2	3
1. Expected benefits	--
2. Expected discomfort	0.31	--
3. Expected social norms	0.24	0.65	--
Cronbach’s α	0.83	0.67	0.62

Descriptive Statistics

The mean total score on the BAYS was 55.62 (SD = 8.58) out of a possible range of 11–77, with higher scores indicating more positive beliefs. 60% of the sample had tried yoga before and were, on average, satisfied with this experience (M = 5.31, SD = 1.60). In the past month, 8.9% had used a yoga video and 12.9% had attended a yoga class. Overall, participants indicated that they would be likely to attend a free yoga class (M = 5.44, SD = 1.80) with 73.7% indicating a 5 or higher on this 7-point scale. Participants also viewed yoga as generally beneficial (M = 6.29, SD = 1.06) and were willing to pay a mean of $17.39 (SD = 14.06) for one yoga class. Females (M = 57.50, SD = 8.62) had significantly more positive beliefs about yoga than did males (M = 53.98, SD = 7.86; t(366)=3.51, p < .001). The total BAYS score was also significantly positively correlated with age (r = .10, p < .05), while no significant association was found between the BAYS and ethnicity, marital status, or affiliation with the medical center.

Criterion-Related Validity

The relationship between the BAYS and the following items evaluated criterion-related validity: positive attitudes about yoga (r = .52, p < .001) and the intention to attend a free class (r = .47, p < .001) were significantly associated with higher scores on the BAYS (positive beliefs). Furthermore, those who had tried yoga before had higher scores on the BAYS (M = 58.31, SD = 8.00) than those who did not (M = 53.76, SD = 8.79; t(366) = 5.07, p < .001). Of those who had tried yoga, satisfaction with their previous yoga experience was significantly related to the total BAYS score (r = .51, p < .001). Participants who had practiced yoga by video (M = 60.74, SD = 6.81) or attended a class (M = 61.76, SD = 7.26) in the past month also had higher scores on the BAYS (video: t(363) = 2.99, p < .01; class: t(363) = 4.66, p <.001) than those who did not do a video (M = 56.24, SD = 8.62) or attend a class (M = 55.86, SD = 8.48). Additionally, the BAYS was significantly associated with willingness to pay for yoga (r = 0.16, p < .01).

Open-Ended Responses

We asked participants to complete the sentence, “If I practiced yoga, I…” to ensure that the most prevalent beliefs were captured by the BAYS. Most participants reported a perceived benefit (92.0% of the responses). Benefits included overall well-being 30.0%; stress management 28.9%; physical benefit 25.6%; enjoy yoga 4.2%; spiritual benefit 3.3%, all of which was captured by Factor 1. Two other emerging themes indicated that some participants expected to be embarrassed (3.1%) or to get hurt (2.0%), captured by Factor 2, and only three participants indicated initial responses (<1.0%) that would fall into Factor 3. Inter-rater reliability for these analyses was high (kappa = 0.87, 95% CI: 0.86–0.88).

Discussion

Overall, initial analyses of the BAYS demonstrated that it is a reliable three-factor measure of beliefs about yoga. The three factors that emerged were: expected benefits, expected discomfort, and expected social norms. Criterion-related validity was supported by positive associations of the BAYS with impressions of and experience with yoga. Significant associations were found between the BAYS and demographic variables, providing direction for future research. Additionally, responses from the open-ended question verified the content of the scale items.

The current sample had generally positive beliefs about practicing yoga. This is likely partially explained by the result that 60% of the current sample had tried yoga, as previous research has demonstrated that experience influences beliefs.²³ Although this high percentage of experience with yoga may be viewed as a limitation of our sampling method, it also allowed for known-groups validation analyses, a type of criterion-related validity.¹⁷ That is, those who participated in yoga were likely to have more favorable beliefs about yoga than the general public. Satisfaction with previous yoga experience resulted in an even stronger relationship with the BAYS than previous experience alone.

Significant associations between the BAYS and indicators of attitude and intention also supported the criterion-related validity of this measure. Such associations support the idea that the beliefs assessed by the BAYS will fit the integrative model of behavior prediction.⁹ In the integrative model, behavioral beliefs inform attitudes that, combined with other factors, predict intentions of behavior. Future research should consider the relationship between beliefs and such other factors (e.g., environmental constraints).

The response that participants were willing to pay approximately $17 for a yoga class is consistent with the going rate for group yoga classes in New York City, which ranges from $10 to $25. Additionally, the significant association found between willingness to pay and the BAYS suggests that beliefs about yoga may be indicators of the perceived financial value of this intervention. The relationship between the BAYS and willingness to pay may be explored further with a more developed assessment of this construct.²⁴

The result that females reported more positive beliefs about yoga than males did is in accordance with a previous study that reported a higher prevalence of yoga practice among females.²⁵ Interestingly, the practice of yoga in the previous study sample was inversely associated with age, whereas more positive beliefs about yoga in our current sample was positively associated with age. This suggests that there may be important barriers to participation among older adults that should be further explored. The previously reported sample (N = 1593) also found differences in the practice of yoga based on demographic variables (e.g., race, geographic region, education) that were not found in the investigation of beliefs in the current sample. It is possible that demographic differences in yoga practice are related to factors other than beliefs about yoga. Future research should examine how beliefs about yoga differ among subgroups, how beliefs relate to the practice of yoga, and what other factors (e.g., cost, availability) influence yoga practice.

In response to the open-ended item, most participants indicated foremost that they expected to benefit if they were to practice yoga. This item also established that within this sample, there were no prevalent themes missing from the scale items.

In addition to helping researchers understand how people decide whether or not to practice yoga, this scale may help researchers and yoga instructors recruit participants and retain those who have already chosen to practice yoga. For example, the three identified factors of the scale (expected benefits, expected discomfort, and expected social norms) could guide how yoga interventions are described to potential participants. It could also be administered to new participants to guide discussion and practice. A yoga instructor who knows that a participant has the belief that she or he has to be more flexible to take a class may choose to focus on postures that will allow this participant to feel successful, or introduce a relevant concept such as contentment with the current state of your body. When an instructor is informed about a participant’s beliefs, and addresses them in a skillful and compassionate way, that participant may be more likely to have a positive experience and return for another class. A novel orientation session that addressed participants’ ambivalence about participation in an active intervention study has previously been shown to improve retention rates.²⁶ Therefore, it is likely that retention rates in a yoga intervention study would be similarly influenced by proactively addressing the beliefs of patients who enroll. Retention is especially important when it comes to therapeutic yoga interventions. Whereas many conventional therapeutic interventions, such as taking medication or undergoing surgery, require only passive and/or time-limited participation, yoga requires active and ongoing participation.

There are certain limitations to the present study. First, although we did recruit a racially and ethnically diverse sample, the participants were self-selected, living or working in an urban area, and tended to be highly educated. Thus, responses on scale items by participants in the present sample may not be representative of the general U.S. population. Certain scale items that were excluded during factor analysis may be more relevant in other populations. In future studies, we seek to continue to develop and validate this scale in samples with more heterogeneous characteristics (e.g., rural communities, older adults).

In addition, three items (I would injure my joints or muscles; I would not be following my religious beliefs; It would cost too much money) were removed because they loaded on independent factors. However, these items were individually predictive of some of the one-item indicators of criterion-related validity, and could capture important barriers to participating in yoga. Thus, future studies may choose to incorporate these three items in the total scale (α including 3 items = 0.77), and add items to further assess expectation of injury, religious beliefs, or cost to expand these factors. Also, one item on the third factor, There would only be women in a class, was not clearly confirmed in the second sample. Furthermore, this item loaded negatively on the scale—meaning it represented a negative belief, or concern, about yoga—and was not rated differently by males and females (p = .93). Future evaluation of the BAYS may reconsider the inclusion of this item (alternate results for the second sample without this item: subscale α = 0.57; total α = 0.76). Finally, the one-item assessments of criterion-related validity provide useful but limited information about how beliefs correspond to participation in yoga. Further evaluation of validity of the BAYS is needed. Future research should also explore how beliefs about yoga influence outcomes from yoga interventions, and whether addressing such beliefs can result in more positive outcomes.

In summary, the BAYS developed in the current study will be useful in understanding beliefs about yoga. Three underlying themes in these beliefs emerged in the factor analyses: health benefits, expected discomfort, and expected social norms. Understanding a population’s beliefs about yoga will further inform both the promotion and implementation of yoga interventions to potentially benefit a greater portion of the population.