Participants were 529 Spanish undergraduates (367 women, 69.40%, and 162 men, 30.60%) aged between 18 and 55 years, M = 21.55, SD = 4.39, recruited from different faculties in a private Spanish university: Physical Therapy (n = 125, 23.63%), Social Education (n = 101, 19.10%), Occupational Therapy (n = 85, 16.07%), Psychology (n = 80, 15.12%), Law Sciences (n = 49, 9.26%), Speech Therapy (n = 46, 8.70%), Podiatry (n = 26, 4.91%) and Educational Sciences (n = 17, 3.21%). Most of the participants were unmarried (n = 506, 95.65%), followed by those who was married or lived with a common-law partner (n = 22, 4.16%), and by divorced (n = 1, 0.19%). All participants reported to be Catholics and believers. Participation was voluntary and participants did not receive any compensation for their contribution.

Religious Orientation Scale (ROS) (Batson & Ventis, 1982). We used the Ramírez (2006) Spanish translation, a 31-item scale (ROS-31 henceforth): 8 for the intrinsic scale, 11 for the extrinsic, and 12 for the questionnaire, which is responded to on a Likert-type scale (1 = Strongly disagree; 9 = Strongly agree). Ramírez (2006) obtained Cronbach’s alphas of 0.89, 0.73, and 0.72, respectively. In the current study, the internal consistency was very good for the intrinsic scale, α = 0.94, good for the quest, α = 0.83, and acceptable for the extrinsic, α = 0.77.

Participants were recruited in a Spanish university, and they were asked for their informed consent and filled out a protocol that included the Spanish adaptation of the ROS-31 (Ramírez, 2006), under the supervision of the authors. Anonymity was guaranteed, doubts were resolved, and the need to give honest answers was highlighted, in order to maximize the validity of the information. The average time to fill out the protocol was 30 minutes.

A Confirmatory Factor Analysis (CFA) of the ROS-31 was performed. Because the model showed an inadequate fit, a Principal Component Analysis (PCA) extraction method and Equamax rotation method were carried out in the randomly selected 50% of the sample (Browne & Cudeck, 1993). In addition to both the Kaiser-Mayer-Olkin measure and Bartlett’s sphericity test, the following criteria were taken into account: (1) the component loading of the items should be ≥ 0.40; (2) the items which loaded ≥ 0.40 in two or more components would be removed; (3) the component in which at least 3 items did not load ≥ 0.40 would be dismissed.

Then, a CFA of the final solution was carried out in the other 50% of the sample. Because it was not possible to assume multivariate normality, robust estimation was used (Satorra & Bentler, 2001). Fit indices included the Comparative Fit Index (CFI) and Tucker-Lewis Index (TLI) (a score between 0.90 and 0.95 indicates reasonable model fit), and Root Mean Square Error of Approximation (RMSEA) (a score lower than 0.050 is optimal, a score between 0.050 and 0.080 suggests an acceptable fit, a score between 0.080 and 0.10 suggests a mediocre fit, and a score higher than 0.10 indicates an unacceptable fit) (e.g., Hair, Anderson, Tatham, & Black, 2006).

The reliability of the obtained model for the ROS was analyzed by estimating the Cronbach’s alpha. A reliability index ≥ 0.70, 0.80, or 0.90 can be interpreted as acceptable, good, or excellent, respectively. Also, following the suggestions of experts (e.g., Brown, 2015), the Composite Reliability (CR) was calculated for each scale of the obtained model for the ROS.

For the PCA, descriptive statistics, correlations, and estimation of the internal consistency, the SPSS Statistics 22.0 program for Windows (IBM, 2013) was used. Interpretation of effect sizes for r was based on Cohen (1988). For the CFA, the EQS 6.1 for Windows (Bentler, 2006) was used.

Descriptive statistics of the ROS-31 items are shown in Table 1. A CFA showed an inadequate fit of this model: SBχ² ₍₄₃₄₎ = 2221.85, p < 0.01, CFI = 0.75, TLI = 0.75, RMSEA = 0.088 (90 % CI [0.085, 0.092]).

Table 1 Table 1 Means and standard deviations of the ROS-31 items Note.

In parenthesis, the religious orientation: (E) = Extrinsic; (I) = Intrinsic; (Q) = Quest.

A PCA extraction method with Equamax rotation method was performed on the ROS-31 (Table 2) with the randomized 50% of the sample (N = 264; 194 women, 73.5%, and 70 men; M _age = 22.34, SD = 5.78). The final solution showed a three-component model: Component l was composed by 8 items (2, 7, 8, 9, 12, 19, 21, and 27; 25.86% of the variance); component 2 comprised 7 items (5, 6, 11, 20, 24, 26, and 31; 16.83% of the variance); component 3 included 6 items (14, 22, 23, 25, 28, and 30; 13.42% of the variance). These components were called Intrinsic Religious Orientation (IRO), Quest Religious Orientation (QRO), and Extrinsic Religious Orientation (ERO), respectively. These components showed positive correlationships with an intermediate effect size (Cohen, 1988): r _(IRO-ERO) = 0.46, r _(IRO-QRO) = 0.39, and r _(ERO-QRO) = 0.48. The obtained scale was called Religious Orientation Scale-21 Items (ROS-21).

Table 2 Table 2 Principal Component Analysis of the ROS-31 Note.

Extraction method: Principal Component Analysis. Rotation method: Equamax with Kaiser normalization. EV = Explained variance. In parenthesis, the religious orientation: (E) = Extrinsic; (I) = Intrinsic; (Q) = Quest.

^a Rotation converged in 11 iterations.

^b Rotation converged in 7 iterations.

^c Rotation converged in 7 i iterations.

A CFA carried out with the other 50% of the sample (N = 265; 173 women, 65.3%, and 92 men; M _age = 20.76, SD = 2.01) showed an adequate fit of the obtained model, SBχ² ₍₁₈₆₎ = 352.45, p < 0.01, CFI = 0.93, IFI = 0.93, RMSEA = 0.059 (CI 90% [0.049, 0.067]). All the parameters of the standardized equation showed acceptable values, p < 0.05 (Figure 1).

Figure 1

Note. Values at the top of each rectangle are R²; values at the left of each rectangle are errors

The IRO scale showed an excellent Cronbach’s alpha, α = 0.92, the QRO showed a good Cronbach’s alpha, α = 0.82, and the ERO showed an acceptable Cronbach’s alpha, α = 0.72. The right internal consistency of these scales was confirmed with the CR, which was 0.88 for the IRO scale, 0.76 for the QRO scale, and 0.76 for the ERO scale. Given that the minimum value considering suitable for CR is 0.70 (Hair, Hult, Ringle, & Sarstedt, 2017), the values found for each ROS-21 scale reflected an appropriate accuracy of these measures.

Regarding the results obtained in the present study, there are some differences with the Batson and Ventis (1982) model: item 15, which was included by Batson and Ventis (1982) in the intrinsic scale, was discarded in the present study; items 3, 29, 16, and 17, which were included by Batson and Ventis (1982) in the quest scale, were discarded in the present study; items 1, 4, 10, 13, and 18, which were included by Batson and Ventis (1982) in the extrinsic scale, were discarded in the present study; item 8, which was included by Batson and Ventis (1982) in the quest scale, was included in the intrinsic scale in the present study, probably because that item refers to meaning in life, which is significantly related to intrinsic religiosity (e.g., García-Alandete et al., 2013). In summary, concerning the Batson and Ventis (1982) model, all the three scales were reduced (especially both the quest and extrinsic ones), and the intrinsic scale included the item 8, in which religiosity is related to meaning in life.

Both the extrinsic and quest scales obtained in the present study were one-dimensional, rather than two-dimensional and three-dimensional, respectively, in contrast to other studies (e.g., Altemeyer & Hunsberger, 1992; Beck & Jessup, 2004; Brewczynski & MacDonald, 2006; Donahue, 1985; Gorsuch & McPherson, 1989; Gorsuch & Venable, 1983; Kirkpatrick, 1989; Leak, 2011; Neyrinck et al., 2010).

Regarding the internal consistency of the 21-item for the ROS obtained in the present study, the Cronbach’s alpha was higher for the intrinsic scale but lower for the extrinsic scale than in Allport and Ross (1967). Likewise, the Cronbach’s alpha was higher for both the intrinsic and quest scales than it was in Batson and Ventis (1982), and equal for the extrinsic scale. These data agree with previous studies (e.g., Batson & Schoenrade, 1991b; Darvyri et al., 2014; García-Alandete & Bernabé, 2013). Also, the CR values showed an appropriate accuracy of all the three religious orientation scales of the ROS-21.

In summary, as noted above, the 21-item model for the ROS is a scale with good structural validity and internal consistency.

Some limitations of this study should be pointed out. First, the sample was only composed of undergraduates, most of them young people. Future studies should consider more representative samples of the general population. Second, the sample was composed exclusively of Catholic people; it would be important to use more heterogeneous samples, in terms of belonging to different religious identities (Muslims, Buddhists, Christians, others) (e.g., Johnstone et al., 2012), and a cross-cultural perspective (e.g., Ghorbani, Watson, Gharamaleki, Morris, & Hood, 2002). Likewise, to use a larger sample would allow to carry out multi-group confirmatory factor analysis (e.g., men and women, age subsamples).

It is necessary to further examine the construct validity and reliability of the ROS (e.g., Beck, Baker, Robbins, & Dow, 2001; Beck & Jessup, 2004; Donahue, 1985), in order to consider the idiosyncrasies of different religions and sociocultural contexts in the assessment of religious orientations (e.g., Flere et al., 2008; Flere & Lavrič, 2008). It is also important to analyze the relationships between the religious orientations and other variables, such as the empathy and social desirability (Watson, Hood, Morris, & Hall, 1984), type of religious internalization (Ryan, Rigby, & King, 1993) or religious denomination (Maltby & Lewis, 1996), the field dependence/independence cognitive style (Barrett, Patock-Peckham, Hutchinson, & Nagoshi, 2005), rigidity (Maltby, 1998) and other personality variables (Maltby, 1999b) emotional intelligence (Liu, 2010) and self-determination (Neyrinck et al., 2010) among others, as well as the differences between the various religions (which might condition the meaning and comprehension of the items). More conservative and authoritarian religions, such as Islam, especially its more radical forms, might negatively influence the QRO and positively influence both the IRO and the ERO. By contrast, religions that recognize freedom of conscience and a more personalized religiosity, such as Lutheranism or Buddhism, might positively influence the QRO. In addition, the process of secularization and other cultural characteristics might also positively influence the QRO (in addition to atheism and religious indifference, of course; Cfr. Watson, Morris, & Hood, 1992), among others (Cfr. Kristensen, Pedersen, & Williams, 2002; Miner, 2008).

Finally, it seems to be particularly important to analyze whether the Quest orientation is a mature form of religiosity (Batson & Ventis, 1982), an expression of agnosticism (Beck & Jessup, 2004; Donahue, 1985), or a transitional form of religiosity related to doubt and the search for one’s personal identity in early adulthood (e.g. Donahue, 1985; Klaassen & McDonald, 2002; Kojetin, McIntosh, Bridges, & Spilka, 1987; Spilka, Kojetin, & McIntosh, 1985). It might also be important to clarify its relationships with Intrinsic and Extrinsic religious orientations.