Measuring military university Students’ motivational goals in the domain of physical exercise

The original partition of two dimensions (learning vs. performance), which were considered relatively opposing, has been refined by the identification of more-specified dimensions and a point of view that recognizes simultaneous emphasis of multiple goals (Pintrich 2000, 2003; Elliot 2005; Kaplan and Maehr 2007). More in-depth reviews of goal research have been conducted and reported elsewhere (e.g. Urdan 1997; Vandewalle 1997; Duda and Whitehead 1998; Jagacinski and Duda 2001; Hulleman et al. 2010); in this article, my purpose is to focus on instrumentation that draws from and illustrates the same theoretical and empirical backgrounds of achievement goal research. Achievement goals have been labelled and operationalized in varying ways in prior studies, but as the two ‘primary orientations’ accentuate similar key contents in generic terms (cf. Ames and Archer 1987, 1988; Elliot 2005), these dimensions are referred to as mastery and performancegoals or goal orientations in this study (except when citing the original terms becomes necessary).

First, in chronological order, two types of performance strivings were distinguished: focus on performance-approach and performance-avoidance goals (the so-called trichotomous model; cf. Elliot and Harackiewicz 1996; Middleton and Midgley 1997; also refer Skaalvik 1997). Performance-approach goals include demonstrating competence and outperforming others, while the performance-avoidance goals refer to avoiding failure or verdicts of unskilfulness.

Second, in addition to mastery goals, which refer to the learning focus (learning goals or task involvement; cf. Nicholls 1984; Dweck 1986), an additional class of mastery-avoidance goals has been proposed. This dimension defines competence in absolute/intrapersonal terms, but which is negatively valenced (2×2 model; refer Elliot and McGregor 2001).

Recently, it has been suggested that achievement strivings could be conceptualized based on three standards of competence references (task, self and other), valenced in terms of approach and avoidance tendencies (3×2 model; refer Elliot et al. 2011).

Different achievement goal dimensions have also been applied in motivational research in the field of sports and physical education. Following the work of goal orientation theorists, Duda (1992) proposed that people have motivational tendencies also related to sports in terms of dispositional task and ego goal emphasis. Duda and Whitehead (1998: 22) referred to these goal categories as distinct ways in which people process certain activity. Different perspectives and contexts have led to varied definitions and operationalizations, as well as a variety of labels (Duda and Whitehead 1998: 23). Altogether, Nicholls and Duda (cf. Duda and Nicholls 1992) modified an instrument measuring dispositional goals in classroom settings so that it could be applied in a sports context. In this most significant work, Duda and Nicholls (1992) differentiated between the two original goal dimensions, i.e. task and ego, in Nicholls’ terms. The trichotomous model of achievement goals has also been examined in the physical field. For example, Shen et al. have used adapted measures into contexts of softball and physical education classes. Their studies have shown that the proposed dimensions (mastery, performance-approach and performance-avoidance), after being adapted to the physical context, were structurally identical to the original scales (Shen et al. 2007; Shen et al. 2009). Later on, Wang et al. implemented the 2×2 achievement goal framework to examine adolescents’ motivation in physical education classes. Their results supported the hypothesized four-factor structure consisting of mastery-approach, performance-approach, mastery-avoidance and performance-avoidance goals (Wang et al. 2007).

In this study, I follow the works of Nicholls et al. in the conceptualization of dispositional goal orientations (Nicholls 1989; Nicholls et al. 1989) and aim to extend prior research by examining adult students’ motivation in physical exercise using contemporary instrumentation.

To start with, it seems that mostly prior instrumentations are focused on, and thus limited in, context and terms related to competitive sports or to young peoples’ experiences in physical education in school. As is discussed in Duda and Whitehead (1998), the reference of success in competitive sports is anchored to the concept of winning. With regard to classroom or the class context in school, a person’s goal endorsement is reflected in his/ her stance in relation to evaluation practices and pedagogical delivery (e.g. Ames 1992). In recreational sport or individuals’ physical training, goal structures need to be differently anchored or referenced than in competitive sports or classroom work. In practice, it is necessary to express items in terms that are relevant to how adults approach their personal physical exercise.

This study adopts the operationalization of personal achievement goal orientations, proposed by Niemivirta (2002), as the baseline.

The original instrumentation also includes a fifth dimension, work avoidance orientation, which includes aims of avoiding effort and evading challenges, instead of striving for competence or any specific relevant outcomes. This scale was omitted as the very idea of avoiding effort is not relevant to the context of physical exercise; if one was only trying to avoid input of effort, one would altogether avoid exercise. Perhaps the concept of avoidance in this context could be addressed by examining some sort of levels of effort one is willing to input for certain levels of outcomes: this remains to be examined in future research. The idea and different wordings for adaptation were initially tried out in Visa Eloranta’s Bachelor’s Thesis (National Defence University, Helsinki, Finland; 2015), but the adaptation of items and the data of this study are independent from his work.

Second, two types of performance goal strivings are distinguished: (a)performance-approach and (b) performance-avoidance orientations (e.g. Elliot and Harackiewicz 1996; Middleton and Midgley 1997; also self-enhancing and self-defeating ego orientations, as in Skaalvik 1997). The performance-approach orientation reflects the aim of demonstrating competence or appearing better than others, while the performance-avoidance orientation has a focus on avoiding judgments of incompetence or appearing worse than others.

The participants comprised 139 (aged 20–23 years) second-year students at the Finnish National Defense University (NDU). In Finland, higher education in military science is provided by the NDU, which trains officers for the Finnish Defence Forces. The military students are recruited after they have completed their military service, which is obligatory for Finnish men (conscription) and voluntary for Finnish women. The annual intake of the NDU is approximately 140 and only 3%–6% of the students were female at the time the measurements of this study were conducted. The selection process consists of a preselection phase based on applications, prior education and military service records, followed by a series of psychological and ability tests, as well as physiological screening. Undergraduate studies (Bachelor of Military Science) last 3 years and include a combination of academic and vocational studies. Part of their training is, naturally, very physical and students need to complete certain fitness tests during their studies in order to graduate.

Achievement goal orientations were assessed using a modified questionnaire (Niemivirta 2002; also refer Tuominen-Soini et al. 2012; Pulkka and Niemivirta 2013; Tapola et al. 2014, for studies that have used the original instrument in Finnish), which in this study included four types of orientations: mastery-intrinsic (focus on learning new things and developing competence), mastery-extrinsic (focus on learning and mastery but with extrinsic criteria such as grades), performance-approach (focus on outperforming others) and performance-avoidance orientation (focus on avoiding judgments of incompetence).

The adaptation of the instrument included revising the wording of original items by replacing anchor terms such as ‘studies’ and ‘school’ with, e.g. ‘when I exercise’ or ‘in physical training’ and so on. Furthermore, the criterion, reference or projected outcome was modified from, e.g. ‘grades’ to ‘results’, appearing ‘competent’ into ‘fit’ and so on. Following this, the initial instrument included altogether 12 items for the chosen four goal orientation dimensions: mastery-intrinsic orientation (three items, e.g. ‘I train to improve and develop my fitness level.’), mastery-extrinsic orientation (three items, e.g. ‘Getting good results is important for me.’), performance-approach orientation (three items, e.g. ‘An important goal for me when I exercise is to do better than other people.’) and performance-avoidance orientation (three items, e.g. ‘It is important for me not to fail in front of other people when I exercise.’). The paper-and-pencil questionnaire (in Finnish) was administered to the students by the author himself. Students were informed of the purposes of the study, on data handling procedures as well as on anonymity, and they were instructed to leave an item unanswered if they did not understand it or found that they could not answer it. Cadets rated each statement on a Likert scale ranging from 1 (not true at all) to 7 (very true).

As this study applies the existing instrument to a significantly different field of achievement, and to a more domain-specific level, thorough examination of the structure is deemed necessary. Following analytical strategies that have been used in the validation of a new instrument (e.g. Vandewalle 1997) and research on dimensionality of achievement goal orientations (e.g. Zweig and Webster 2004; Attenweiler and Moore 2006), the following procedure was implemented. A series of explorative maximum-likelihood factor analyses with varimax rotation, as implemented in SPSS 23, was conducted for all items, and item analyses were used to examine the consistency of the resulting scales. The structural validity was tested with confirmatory factor analysis with Mplus Statistics software (Muthén and Muthén 1998–2009). Thus, the analytical strategy was to extract a plausible solution and to compare it with theoretical suggestions as well as fit to data.

A series of factor analyses indicated that, initially, the resulting structure deviated to some extent from the proposed dimensions, and that several parallel paths of iterations were plausible. Factors were extracted for each model based on inspection of the factor structure and the contents of items. Results from the Kaiser-Meyer-Olkin Measure of Sampling Adequacy and Bartlett’s test of sphericity indicated that the data matrix was appropriate for factor analysis.

After excluding one item from the performance-approach orientation, based on slightly low communality, the initial solution suggested two factors, of which the first included items from the mastery-intrinsic and mastery-extrinsic scales. This factor reflected striving for good results and optimal development. The second factor included items on performance-avoidance orientation and performance-approach scales. This factor reflected striving to outperform others and avoiding appearing less fit or less able than others. Altogether, this solution, which included 11 of the original 12 items, suggested two combined mastery and performance dimensions. Item analysis showed that the internal consistency of the proposed scales was good. However, the goodness-of-fit test indicated that the structure was significantly dissimilar to the data (p<0.001). This solution is labelled as two-dimensionalfor comparison.

As the aim of this study was to explore the dimensionality of goal preferences in physical training, theoretically grounded compositions and reasonable modifications were examined.

First, it was deemed necessary to examine whether the inclusion of the rather novel (in this context) mastery-extrinsic orientation caused the unexpected structural results via its close relation to mastery-intrinsic orientation. Thus, it was tested whether the trichotomous model could be observed in the data. A factor analysis that was conducted on only nine items from mastery-intrinsic, performance-approach and performance-avoidance orientations yielded precisely the expected three dimensions, with only some slight crossloading, for two items, between performance-approach-and performance-avoidance-focused factors. Goodness-of-fit test (p=0.123) indicated that the solution was not significantly different from the data. Item analysis showed that the internal consistency of the proposed scales was good for the first two factors but only moderate (0.68) for the third factor, which included the performance-approach items. This solution is labelled as three-dimensional for further analysis.

Second, a forced four-factor model was examined. Following the former model, items theoretically pertaining to the mastery-extrinsic orientation were added. The rationale for this was to find out whether the implemented partition would produce two independent factors for the two mastery-focused orientations. The extracted four factors each included the original corresponding items of the proposed four dimensions – with the exception of one item from the mastery-intrinsic orientation (‘To develop as much as possible is an important goal for me in my training.’) loaded into the factor that included all mastery-extrinsic items. Based on the item analyses, and for the purposes of clarity, this item was excluded from further analysis. It may be that the actual wording of the item ‘as much as possible’ is associated more with external criteria than simply ‘personal best’, which was the intention behind this item. This modified solution yielded four factors that included items from the original corresponding scales. Goodness-of-fit test (p=0.166) indicated that the solution was not significantly different from the data. Item analysis showed that the internal consistency of the proposed scales was good. This solution is labelled reduced four-dimensional for further analysis.

Factor loadings, corresponding items and statistics from the item analysis for these three factor solutions are reported in Table 1.

The proposed models were tested for structural validity using confirmatory factor analysis. In the measurement model and consequent test models, each factor was specified to predict corresponding items, and initially the error terms of items were uncorrelated. Chi-square statistics, comparative fit index (CFI, cut-off value > 0.95), and standardized root-mean-square residual (SRMR, cut-off value < 0.08) were used to evaluate model fit following the two-index strategy proposed by Hu and Bentler (1998, 1999) for smallish samples. These indices are recommended instead of, e.g. root-mean-square error of approximation (RMSEA). which tends to indicate falsely poor fit for small samples (refer also Kenny et al. 2015).

First, the measurement model, i.e. the proposed four-dimensional model with original item assignment, was tested. Second, the two-dimensional model with combined mastery and performance scales was tested. Third, the three-dimensional model with mastery-intrinsic, performance-approach and performance-avoidance scales was tested, and finally, the reduced four-dimensional model with the additional mastery-extrinsic scale was tested.

Within each model, first, one minor modification was needed to the proposed model (i.e. freeing the error covariance between one pair of similarly worded items of the mastery-intrinsic orientation scale). Second, regarding the two-dimensional model, one more similar modification was necessary (in two items of the performance-approach scale). Results indicate that the most acceptable fit was achieved with the three-dimensional and reduced four-dimensional models, as the cut-off values were met. Goodness-of-fit criteria are reported in Table 2.

Following these three-factor solutions, items were then used to construct composite variables for further examination. These scales were labelled based on the contents of the items and the original scales as follows: (i) two-dimensional solution: 2Dmastery orientation (Factor 1, six items) and 2Dperformance orientation (Factor 2, five items); (ii) three-dimensional solution: 3Dmastery-intrinsic orientation (Factor 2, three items), 3Dperformance-approach orientation (Factor 3, three items) and 3Dperformance-avoidance orientation (Factor 1, three items) and (iii) reduced four-dimensional solution: R4Dmastery-intrinsic orientation (Factor 3, two items), R4Dmastery-extrinsic orientation (Factor 1, three items), R4Dperformance-approach orientation (Factor 4, two items) and R4Dperformance-avoidance orientation (Factor 2, three items).

In terms of the relations between variables (Table 3), first, the two dimensions of the two-factor model did not correlate with each other. For the three-factor model, the 3Dmastery-intrinsic orientation and 3Dperformance-approach orientation were positively correlated, as were the two 3Dperformance orientations. With respect to the reduced four-factor model, the two R4Dperformance orientations were again correlated but more strongly than in the 3D model. The two R4Dmastery orientations had differential correlation pattern: the R4Dmastery-intrinsic orientation correlated with R4Dmastery-extrinsic orientation but was not related to the two performance orientation factor, whereas the R4Dmastery-extrinsic orientation was positively correlated with R4Dperformance-approach orientation. This suggests that the partition of the mastery and performance dimensions into two components potentially reveals significant relationships that aggregating them might mask. The patterns of correlations are consistent with those in prior studies (Tuominen-Soini et al. 2012; Pulkka and Niemivirta 2013; Tapola et al. 2014).

With respect to descriptive statistics, the mastery-intrinsic orientation had the highest mean, followed by mastery-extrinsic orientation (in the R4D model) and performance orientations, in which the approach dimension had a higher mean. Moreover, the performance-approach orientation and performance-avoidance orientation had significantly different mean levels. This was also observed for the two differentiated mastery dimensions, although not as clearly.