Perception of specific military skills – the impact of perfectionism and self-efficacy

The ideal conduct of military duties requires abilities such as being able to handle stress and bearing with uncertainty in different situations. The primary focus of military training is the building of confidence and high standards of military performance to endow the soldiers with the ability to cope with combat contexts and complete their missions (Driskell et al. 2006; Krueger 2008). In preparations for specific deployments, military personnel are trained in the skills needed to cope with that particular mission (Delahaij et al. 2006). Specific military skills are thus closely related to operational and combat tasks, which often demand coping with stress and decision-making skills under pressure, and specific military skills have thus been suggested to be the core elements of operational conduct (Johansen et al. 2014).

Solberg (2007) developed an instrument to measure specific military skills intended to measure the skills and abilities required under operational and combat circumstances. The instrument measures specific military skills by way of self-report and is more accurately a measure of perceived specific military skills. The instrument comprises three subscales: Individual Coping Capacity (ICC), Cooperation in Difficult Situations (CDS), and Motivation to Achievement (MA). These factors are all regarded as important skills in military operations. The enhancement of military performance and skills can be achieved by improving either selection procedures so that individuals who possess the desirable attributes and traits are selected for military training or the training of desirable skills and qualities.

Although research on military performance has expanded during the past few years, less research has been carried out on the specific military skills needed to obtain successful performance and especially what enhances the development of these military skills. A study on specific military skills found that operational identity predicted 6% of the variance in these specific military skills, whereas personality traits explained as much as 42% of the variance in specific military skills (Johansen et al. 2014). Emotional stability, intellect/openness, and hardiness were found to be significant individual predictors of specific military skills. Intellect/openness was the strongest predictor of specific military skills, and it has been suggested that this reflects the importance of the ability to use creative and perhaps unorthodox approaches in ambiguous conditions with a high degree of uncertainty (Johansen et al. 2014). Since intellect/openness implies intellectual curiosity, openness to new ideas and a willingness to try new approaches (Costa and McCrae 1992), openness has been suggested to facilitate learning and adaptability, which are valuable qualities in military training (Johansen et al. 2014). Emotional stability may also contribute to the development of specific military skills because these skills are closely related to coping with stress and handling decision-making under stressful situations (Johansen et al. 2014). In addition to predict specific military skills, hardiness has also been found to be related to leadership skills and leadership performance in both theoretical and practical military contexts (Bartone et al. 2009; Eid and Morgan 2006). Because most studies in this field have been cross-sectional, longitudinal studies are warranted in order to make assumptions about the development of specific military skills.

Knowledge about factors affecting the training and development of military skills is of paramount importance for the armed forces in order to optimize the training in the military academies and subsequently enhance operational readiness. Several studies have identified both personality traits and self-efficacy beliefs as predictors of academic and work performance (Barrick and Mount 1991; Judge et al. 2007; Poropat 2009; Salgado 1998), and there are reasons to believe that these factors will also affect the development of military skills.

Realistic practical training, as well as exposure, has been shown to have a significant impact on cadets’ skills and ability to cope with interrogation in a simulated prisoner-of-war exercise, in which those who participated in the practical training showed significantly better results compared to those with only theoretical preparations (Laberg et al. 2000). The positive effect of pretraining was explained by two mechanisms: the development of effective schemas (i.e., mental representations) for the situation and higher self-efficacy expectations that assisted them in coping. Self-efficacy beliefs have been suggested to play an important role in the actual coping ability in extreme situations in which expectations of success are highly related to the outcome in the specific situation (Bandura 1982, 1997, 2014).

Previous researches on personality traits in relation to military selection have primarily focused on personality variables such as the Big Five and hardiness (Hystad et al. 2011; Johansen et al. 2014), although other personality traits/variables may also have importance. One such factor could be adaptive perfectionism. Because adaptive perfectionism has been found to contribute to greater achievement and motivation (Accordino et al. 2000; Hamachek 1978), and to be associated with personal control, resourcefulness, self-esteem, and adaptive learning strategies (Flett et al. 1994; Flett et al. 1991a, 1991b; Mills and Blankstein 2000), we assumed that adaptive perfectionism may be relevant in relation to the development of military skills.

Perfectionism can be defined as “the striving for flawlessness” (Flett and Hewitt 2002, p. 5). This trait can be viewed as adaptive in a military context, where preciseness is emphasized as extremely important (The Norwegian Defence University College 2014). In military training, in which drilling is important in order to make certain behaviors automatic and rapid (Flin 1996), one may expect a perfectionist to strive for more precision and to show greater endurance in the pursuit of perfection (e.g., to train more) and hence be able to perform the drilled exercise with greater precision and less flaws compared to a non-perfectionist. Thus, perfectionism may be a particular adaptive trait within a military context. On the other hand, military skills often involve coping with stress and decision-making under high degrees of uncertainty in combat situations, so one may assume that perfectionism and striving for flawlessness could be maladaptive or counterproductive in such situations.

There has been little systematic study of perfectionism among military personnel. Among the general population, perfectionism has previously been viewed as a positive factor in adjustment or achievement (Hamachek 1978) but has also been linked to negative outcomes such as guilt, shame, feelings of failure, indecisiveness, procrastination, and low self-esteem (Flett et al. 1989; Pacht 1984). Most researchers in the field have now come to recognize that perfectionism is a multidimensional construct where the different dimensions have distinct prediction patterns, e.g., having positive or negative aspects (Bieling et al. 2004; Chang 2000; Enns et al. 2001; Terry-Short et al. 1995). Adaptive perfectionism is associated with (high) standards modified in accordance with the situation and matched to the person’s limitations and strengths, striving for success, a sense of self that is independent of performance, the timely completion of tasks, and balanced thinking. These attributes are also considered to be important in a military context. One type of adaptive perfectionism has been labeled as self-oriented perfectionism, which is defined as unrealistic high standards and perfectionistic motivation for the self (Hewitt and Flett 1991). Self-oriented perfectionism has shown a strong correlation with personality traits such as achievement striving, dutifulness, and self-discipline (Hill et al. 1997). These factors are also considered to be important in military training. Adaptive perfectionism has further been identified as a likely antecedent of effective coping, in which the individual’s perfectionistic standards are assumed to lead to the effective regulation of one’s behavior and hence more effective coping (Gaudreau and Antl 2008).

Socially prescribed perfectionism refers to the belief that significant others expect oneself to be perfect (Hewitt and Flett 1991), and this dimension of perfectionism is associated with excessive concerns over mistakes and has also been linked to maladjustment (Slaney et al. 2002). Consequently, socially prescribed perfectionism is often denoted as maladaptive perfectionism. Higher levels of socially prescribed perfectionism is further associated with more severe forms of psychopathology, and socially prescribed perfectionism incorporates perceptions of disturbed social relations, as well as elements of learned helplessness and a-motivation (Hewitt et al. 1991). Maladaptive perfectionism is associated with inflexibly high standards, which are often beyond what is expected, such as fear of failure, a focus on avoiding errors, a sense of self-worth dependent on performance, and a black and white thinking pattern (perfection vs. failure) (Enns and Cox 2002). These attributes may be assumed as less adaptive in a military context.

Based on what is presented earlier and because self-oriented perfectionism is often associated with perseverance and effort and potentially with a consecutive performance, such an endorsement might be adaptive from the point of developing military competence. If high levels of self-oriented perfectionism are associated with striving for better quality on a training task and getting better at doing the task correctly without flaws, one may therefore assume adaptive perfectionism to have a positive impact on the development of specific military skills, whereas high levels of maladaptive perfectionism can be assumed to have a negative impact. However, for individuals characterized by high levels of adaptive perfectionism, there may still be a cost to well-being in the striving for perfection.

While perfectionism is viewed as a personality trait and is assumed to be relatively stable across time and situations, self-efficacy is a more situation-specific variable that has been shown to predict coping behavior when facing challenging situations and environmental demands (Bandura 1986, 1989; Jerusalem 1990). Self-efficacy refers to the beliefs an individual holds in his or her own capability to organize, function and exercise control over events that affect his or her life (Bandura 1997). Self-efficacy has been found to play a key role in stress reactions and is related to the quality of coping in stressful and threatening situations (Bandura 1988, 1997; Lazarus and Folkman 1984). As military operations usually involve high degrees of stress and possibly threatening situations, the levels of self-efficacy will probably affect how soldiers handle and cope with stress in military combat.

Social cognitive theory has identified self-efficacy as the most powerful self-regulatory mechanism in affecting behavior (Bandura 1997), and several studies have demonstrated that task-specific efficacy beliefs are crucial in the prediction of performance in a given situation (Stajkovic and Luthans 1998). Self-efficacy has repeatedly been shown to be related to task performance in a variety of settings (Tannenbaum et al. 1991), e.g., leadership self-efficacy has been shown to mediate leader effectiveness in military contexts (McCormick 2001; Ng et al. 2008). As personality research has highlighted the importance of motivational processes and self-efficacy is a central motivational construct in behavioral prediction, Ng et al. (2008) proposed that leadership self-efficacy is a key motivational mechanism that links leaders’ broad personality traits to leader effectiveness. Similarly, we may expect that military cadets’ self-efficacy is a central motivational mechanism in the training of specific military skills, linking personality traits such as perfectionism to military skill performance. Consequently, one may expect that both levels of self-efficacy and perfectionism will affect the development of military skills, e.g., cadets with greater levels of self-efficacy and adaptive perfectionism will show a greater improvement of military skills during military academy training.

The aim of the present study was threefold. First, we wanted to explore the stability of specific military skills, i.e., how ICC, CDS, and MA develop during three years of training in a military academy. We hypothesized that three years of training in the academy will produce a positive increase in these skills over the three-year period. Second, we wanted to investigate whether the initial levels of adaptive and maladaptive perfectionism in cadets impact the development of military skills over time. We hypothesized that the cadets scoring high on adaptive perfectionism prior to training will show a greater positive development on military skills during the three years of training at the military academy. Third, we wanted to investigate whether the cadets’ levels of self-efficacy impact the levels of specific military skills. We hypothesized that levels of self-efficacy will predict levels of military skills over the three years.

The data for this study were collected as part of the Norwegian Military Academy Study, 2007–2011. Cadets attending the three different military academies in Norway (the Naval Academy, the Air Force Academy, and the Army Academy) were selected on both physical and psychological parameters, with the aim of the Norwegian Military Academy Study being to gain comprehensive knowledge about the cadet’s development on a range of physiological and psychological measures during their three-year military academy education program. A total of 330 cadets from two different cohorts were asked to participate, and a total of 286 cadets (response rate of 86.7%) completed the self-report questionnaire at baseline (T1) during the first week of school. The majority of the cadets were men (258 men vs. 28 women), ranging in age from 19 to 37 years (M = 23.2 years, SD = 2.92). The cadets were asked to respond to the questionnaire on three subsequent occasions during their three-year education: at the end of the first year (T2), at the end of the second year (T3), and at the end of the third year (T4). At T2, 235 cadets had completed the questionnaire (17.8% dropout). At T3, 162 cadets completed the perfectionism measure (31.1% dropout from T2), and finally at T4, 159 cadets completed the perfectionism measure (1.9% dropout from T3).

The study was approved by the Norwegian Social Science Data Service, and participation in the study was voluntary.

Specific military skills were measured by the self-report instrument Military Skills and Ability (MSA) (Solberg 2007), which comprises measures of the particular skills and abilities required under operational and combat circumstances. The instrument contains 20 statements rated on a Likert scale ranging from 1 (very bad) to 5 (very good), in which respondents evaluate their handling of military challenges. The MSA comprises three subscales: ICC (e.g., “Keeping a cool head in stressful situations” or “My ability to handle stress and difficult situations”), CDS (e.g., “Willingness to act on command in emergency situations” or “My ability to motivate and inspire others in difficult situations”), and MA (e.g., “My motivation to participate in peacekeeping operations abroad”). The analyses were based on the scores on the subscales, and former studies have revealed a satisfactory internal consistency, with a Cronbach’s alpha >0.70 (Laberg et al. 2002; Moldjord et al. 2015; Solberg 2007).

Adaptive and maladaptive perfectionism was measured using two dimensions/subscales of the Multidimensional Perfectionism Scale (MPS; Hewitt and Flett 1991): self-oriented perfectionism and socially prescribed perfectionism. Each subscale contained 15 items, in which respondents were asked to indicate to what extent they disagree/agree on a 7-point Likert scale (1 – totally disagree and 7 – totally agree). Examples of items from the self-oriented subscale are “When I am working on something, I cannot relax until it is perfect” and “One of my goals is to be perfect in everything I do”, while examples from the socially prescribed subscale include “I find it difficult to meet others’ expectations of me” and “Anything that I do that is less than excellent will be seen as poor work by those around me”.

In order to measure self-efficacy, a 5-item self-efficacy scale specifically developed to measure self-efficacy in a military context was used (Buch et al. 2015). The use of general self-efficacy instruments was discouraged, with Bandura (2012, p. 15) arguing and stating that “there is no single all-purpose measure of self-efficacy with a single validity coefficient” and general decontextualized measures of self-efficacy usually bear a weak relationship to the domain-related self-efficacy beliefs and to behavior and are consequently less predictive, respectively. Hence, this specific self-efficacy measure was developed to capture self-efficacy beliefs regarding the ability to successfully complete the military training education at the military academy. Items were designed to measure the perceived capability to manage process activities likely to be of importance, such as managing tough times, making it through one’s studies and performing well. Sample items included “will manage to complete the military training” and “will achieve a result I can be proud of”. Respondents indicated how much they agree/disagree on a 7-point Likert scale, ranging from 1 (totally disagree) to 7 (totally agree). Previous studies have reported a satisfactory internal consistency, with Cronbach’s alphas ranging from 0.83 to 0.89 (Buch et al. 2015; Fosse et al. 2015).

Latent growth-curve models (LGCMs) within a structural equation modeling (SEM) framework (Rabe-Hesketh and Skrondal 2012) were used to explore whether cadets’ specific military skills develop during their time at the military academy. Separate LGCMs were estimated for the three skill dimensions of ICC, CDS, and MA. Our starting point in each analysis was a simple growth-curve model that estimated latent intercept and latent slope growth factors. These are known as fixed effects, as they summarize the average initial level (intercept) and the overall growth trajectory (slope). To identify the intercept, all loadings from the latent variable to the observed skills variables were constrained to 1. The observed skill variables were also regressed on the latent slope variable, with loadings fixed to the time point t_i (where i = 0, 1, 2, and 3).

LGCMs also estimate the variance of the fixed effects, known as random effects. These random effects summarize the individual differences from the overall growth trajectory and the average intercept. A significant variation around the fixed effects allows us to explore covariates that can help us explain the variation. In this study, we included self-oriented perfectionism and socially prescribed perfectionism as time-invariant predictors of both the latent intercept and slope. They are time invariant because we believe perfectionism to be a relatively stable characteristic that stays constant across time points. Conversely, self-efficacy was included as a time-varying predictor, thereby allowing self-efficacy to have a different effect at each time point.

Figure 1 presents a diagram of our full LGCM. The LGCM analyses were performed using Stata 13 with the full-information maximum likelihood estimator to be able to use all the available data (Arbuckle 1996). Stata does not give a z-value for the variance components that can be used for significance testing because of the boundary issue associated with the estimates (i.e., the variance cannot be less than zero) (Acock 2013). However, 95% confidence intervals (CIs) were computed, which were just as informative, if not more so, as a conventional p-value. CIs were used in lieu of p-values in all text and tables that describe the LGCM analyses.

Fig. 1

Confirmatory factor analyses (CFAs) were conducted on all measures and time points to check for construct validity. Specific military skills were treated as correlated three-factor models, perfectionism was treated as a two-factor model, and self-efficacy was treated as a one-dimensional factor model. A summary of the results from these factor analyses is presented in Table 1.

Tab. 1

Confirmatory factor analyses of specific military skills, self-efficacy and perfectionism.

For specific military skills, all indicators loaded onto their parent factor at Time 1 (mean λ = 0.60, range: 0.32–0.91), Time 2 (mean λ = 0.61, range: 0.31–0.94), Time 3 (mean λ = 0.60, range: 0.35–0.92), and Time 4 (mean λ = 0.62, range: 0.25–0.95), and all loadings were statistically significant. The fit indices shown in Table 1 all point toward less than perfectly fitting models based on the usual recommendations of RMSEA ≤ 0.08 and CFI ≥ 0.90 (Kline 1998; McDonald and Ho 2002). The modification indices suggested that some improvements in model fit could be gained by removing some factor loadings or allowing for cross-loadings. Nevertheless, we chose not to do any changes, because there was no consistent pattern to the suggested modifications across measurement occasions. The reliability coefficients (Cronbach’s alpha) for ICC at the four measurement points were 0.83, 0.83, 0.83, and 0.84; for CDS, they were 0.67, 0.69, 0.71, and 0.75 and for MA, they were 0.78, 0.82, 0.78, and 0.76.

One item belonging to the socially prescribed perfectionism dimension did not have a statistically significant factor loading. After removing this item, the two-factor model revealed an acceptable fit judged by RMSEA = 0.079 and an average factor loading of λ= 0.51 (range: 0.24–0.77). However, the comparative fit index (CFI) of 0.72 was less than optimal. The MPS is a well-established and much used instrument (Appelton et al. 2009; Frost et al. 1993; Kilbert et al. 2005), and for the sake of comparability to other studies, we decided not to do any further modifications. The Cronbach’s alpha values were satisfactory, with α = 0.84 for the self-oriented perfectionism subscale and α = 0.83 for the socially prescribed subscale.

Lastly, the CFAs with self-efficacy revealed that one item consistently performed poorly on all measurement occasions (i.e., item 4: “My willingness to obey orders in a threatening situation is ….”). We therefore reran the analyses with this item excluded. The average factor loading at Time 1 was λ = 0.79 (range: 0.64–0.88), at Time 2 was λ = 0.70 (range: 0.48–0.93), at Time 3 was λ = 0.79 (range: 0.67–0.89), and at Time 4 was λ = 0.74 (range: 0.60–0.87). The CFIs suggested well-fitting models at all time points, although the RMSEAs were less than optimal. The Cronbach’s alphas for the four-item measure in the present study were 0.87, 0.80, 0.87, and 0.82 at T0, T1, T2, and T3, respectively.

A one-way analysis of variance (ANOVA) was used to compare participants who contributed data on all measurement points (“completers”) with participants who participated at three, two, or one measurement point. The results indicated no statistically significant difference between completers and dropouts on age (F[3, 291] = 0.30, p = 0.83). A Pearson’s χ² test showed that a larger proportion of women (55%) than men (35%) were among the completers, χ²(1, N = 295) = 4.78, p = 0.03. It should also be noted that there were very few women altogether (n = 29).

The results from three one-way ANOVAs suggested that completers and dropouts differed on the initial ICC (F[3, 269] = 3.11, p = 0.03), CDS (F[3, 244] = 4.01, p = 0.01), and MA (F[3, 283] = 6.33, p < 0.001). Post hoc pairwise comparisons with Bonferroni corrections showed that completers had a higher initial ICC status than those who only participated at T0 (mean difference: 1.92, 95% CI: 0.11–3.37) but not those who participated at two or three measurement points. Completers also had higher initial scores on CDS than both participants who missed two (mean difference: 1.29, 95% CI: 0.09–2.49) or three (mean difference: 1.71, 95% CI: 0.03–3.39) measurement points but not those who missed only one. Finally, the post hoc tests showed that completers scored higher on initial MA than the participants who missed one (mean difference: 0.90, 95% CI: 0.08–1.72), two (mean difference: 1.07, 95% CI: 0.26–1.72), and three measurement points (mean difference: 1.34, 95% CI: 0.25–2.42).

Table 2 presents the results from the unconditional, simple growth-curve models for the three military skill dimensions. Except for MA, the latent slope growth factors were statistically significant in all instances. For ICC, the average growth trajectory of 0.80 (95% CI: 0.65–0.94) means that the expected increase in the ICC for each year is 0.80 points. The associated likelihood ratio of our model of χ²(5) = 6.735, p > 0.05 suggests that this estimated linear trajectory describes our observed data well. The additional measures of fit for the unconditional growth-curve model are presented in the top row of Table 3 (ICC model 1). Similarly, the expected increase in CDS each year was 0.46 (95% CI: 0.31–0.61). The associated model likelihood ratio was χ²(5) = 1.743, p > 0.05 (see CDS model 1 in Table 3 for additional measures of model fit).

Tab. 2

Unconditional growth models of specific military skills.

Tab. 3

Measures of model fit.

Table 2 also shows a statistically significant variance for the random intercepts and random slopes for all three military skill dimensions. The variances around the overall linear slopes are 0.49, 0.45, and 0.12 for ICC, CDS, and MA, respectively. To give this variation some context, consider that the σ² = 0.45 around the slope of 0.46 for CDS equals a standard deviation of 0.67. An approximate 95% CI on the slope will therefore be 0.46 ± 2 × 0.67, an interval of –0.88 to 1.8. Thus, some individuals are expected to have a negative slope, whereas others are expected to have a relatively steep positive slope.

The covariance between the intercept and slope was nonsignificant in all three simple growth-curve models, hence suggesting that initial levels of military skills are not related to the growth trajectories of the same military skills.

Given the significant variance in the latent intercepts and slopes, we included self-oriented perfectionism and socially prescribed perfectionism as predictors to attempt to explain some of the variations. Both variables were centered on their means to make the intercept a meaningful value (i.e., as the initial skill level for individuals with average levels of perfectionism). The likelihood ratios and measures of the model fit presented in Table 2 (model 2) suggest a good fit to the data for all models. Both self-oriented and socially prescribed perfectionism significantly predicted the intercept for ICC and CDS but not for MA.

The estimates are presented in Table 4 and show that a one-unit increase in self-oriented perfectionism is associated with a 0.6 increase in the initial skill level for ICC, while a one-unit increase in socially prescribed perfectionism is associated with a 0.9 decrease. The estimated initial skill level for individuals’ average in perfectionism was 25.5 (i.e., the intercept, which is not shown in Table 4). This means that individuals with a self-oriented perfectionism score one standard deviation (i.e., 0.76) above the mean are estimated to have an initial skill level of 0.6 × 0.76 = 0.46 above this value, thus holding levels of socially prescribed perfectionism constant. Similarly, individuals with a socially prescribed perfectionism score one standard deviation (i.e., 0.63) above the mean are estimated to have an initial skill level of 25.5 – 0.9 × 0.63 = 24.93, holding self-oriented perfectionism constant. As a result, neither self-oriented perfectionism nor socially prescribed perfectionism seems to have a particularly strong effect on initial ICC scores. This is also corroborated by the R² value of.05, showing that the perfectionism variables combined only explained 5% of the variance in the initial skill level.

Tab. 4

Perfectionism as a predictor of the latent intercept and slope.

For CDS, the estimated initial level for individuals’ average in perfectionism was 25.7 (not shown in Table 4). A one-unit increase in self-oriented perfectionism is associated with a 0.46 increase in the initial skill level, while a one-unit increase in socially prescribed perfectionism is associated with a 1.35 decrease in the initial skill level. The two perfectionism dimensions combined explained 12% of the variance in the initial skill level (R² = 0.12).

None of the perfectionism variables significantly explained variance in any of the latent slopes.

In the final set of models, we added self-efficacy as a time-varying predictor. Unlike perfectionism, self-efficacy was not modeled to influence the overall growth trajectory but instead was allowed to directly influence the military skill dimensions at each time point. In this manner, time-varying predictors were allowed to have a different effect at each wave. The likelihood ratios and measures of the model fit shown in Table 3 (model 3) again suggest good fit to the data for all models. The regression coefficients presented in Table 5 show that self-efficacy significantly predicted ICC at all time points, with estimates ranging between 0.73 and 0.77. Self-efficacious cadets are therefore estimated to be higher in ICC at each time point than what is estimated by the linear growth curve. The effects of self-oriented perfectionism (b = 0.41, 95% CI: –0.10 to 0.91) on the latent intercept was rendered nonsignificant once the influence of self-efficacy on ICC at T0 was taken into account.

Tab. 5

Self-efficacy as a time-varying predictor of specific military skills.

Self-efficacy also significantly predicted CDS and MA at all time points, with the exception for T3. Point estimates ranged between 0.28 and 0.62 for CDS and between 0.24 and 0.35 for MA (see Table 5), again suggesting that self-efficacious cadets are estimated to be higher in military skills than what is estimated by the linear growth curve. Similar to the analysis of ICC, the effect of self-oriented perfectionism (b = 0.28, 95% CI: –0.17 to 0.72) on the latent intercept was rendered nonsignificant once the influence of self-efficacy was taken into account.