Narcissism, beliefs about controlling interpersonal style, and moral disengagement in sport coaches

We tested the relations among narcissism (including both its adaptive and maladaptive facets), effectiveness and normalcy beliefs about controlling interpersonal style, controlling coach behaviours, and moral disengagement in sport coaches. Participants were 210 sport coaches, representing a variety of sports and levels of coaching. Coaches completed a multi-section questionnaire assessing the study variables. Path analyses revealed that global narcissism and maladaptive narcissism were positively associated with controlling coach behaviours. Furthermore, effectiveness and normalcy beliefs about controlling interpersonal style were positively associated with controlling coach behaviours, while controlling coach behaviours were positively associated with coach moral disengagement. Finally, adaptive narcissism had an indirect effect on controlling coach behaviours via effectiveness beliefs about a controlling interpersonal style. These ﬁ ndings contribute to the literature on antecedents and outcomes of controlling coach behaviours, as reported by coaches.


Introduction
Coaches are key authority figures in sport, hence, the interpersonal styles they utilise when communicating with their athletes can play a critical role in shaping athletes' psychological experiences in sport. Although some interpersonal styles can be beneficial in that they support athletes' psychological needs, other styles can be controlling and have the potential to undermine athletes' psychological needs and well-being (Bartholomew, Ntoumanis, Ryan, Bosch, & Thøgersen-Ntoumani, 2011). A theoretical framework for studying a controlling interpersonal communication style is self-determination theory (SDT; Ryan & Deci, 2017). According to SDT, such a style reflects a set of behaviours whereby the agent (e.g. coach) acts in pressuring or coercive ways, imposing ways of thinking, feeling, and behaving upon their athletes (Bartholomew, Ntoumanis, & Thøgersen-Ntoumani, 2009). A controlling interpersonal style has been associated with negative athlete outcomes, such as psychological need frustration, ill-being, and athlete moral disengagement (Curran, Hill, Ntoumanis, Hall, & Jowett, 2016;Healy, Ntoumanis, van Zanten, & Paine, 2014;Hodge & Gucciardi, 2015). Although considerable research efforts have expended on how controlling coaching can shape athletes' experiences, there is limited evidence on the role of personality antecedents of a controlling coach interpersonal style (Occhino, Mallett, Rynne, & Carlisle, 2014). We focus on this putative antecedent category, and in particular on coach narcissism.
Perhaps unsurprisingly, narcissism has been linked with negative leadership qualities and lack of leadership effectiveness (Grijalva, Harms, Newman, Gaddis, & Fraley, 2015;Schoel et al., 2015;Sedikides & Campbell, 2017). Narcissistic leadership has also been recently explored within the coaching domain. Matosic et al. (2017) recruited coaches from a variety of sports (e.g. swimming, football) and levels (e.g. national, international). Coaches responded to scenarios in which they experienced self-threat. Coaches higher (compared to those lower) in narcissism reported that they would implement more often controlling behaviours toward their athletes, such as yelling, belittlement, or guilt-inducement. In another study, Matosic, Ntoumanis, Boardley, Stenling, and Sedikides (2016) also sampled coaches and athletes from a variety of sports and levels. Coaches higher in narcissism were perceived as more controlling by their athletes. In line with literature on narcissistic leaders (Schoel et al., 2015), Matosic et al. explained their results by arguing that coaches high in narcissism behave in an authoritarian manner, take advantage of others, are hypersensitive to criticism, and use guilt-inducing tactics to express their disappointment to seemingly underperforming athletes. These coaches implement the abovementioned controlling strategies in order to gain self-enhancement benefits, such as admiration and reflected glory, as well as to establish authority and superiority over their athletes (Mathieu & St-Jean, 2013;Woodman et al., 2011). Matosic et al. (2016Matosic et al. ( , 2017) examined narcissism at the global level. However, narcissism has also been differentiated in terms of its adaptive and maladaptive facets (Barry & Malkin, 2010). Adaptive narcissism pertains to viewing oneself as authoritative and self-confident, whereas maladaptive narcissism pertains to feeling entitled, being motivated to gain status over others, and seeking attention or admiration. More relevant to the objectives of the current study, adaptive narcissism is unrelated to social misconduct (e.g. aggression) when controlling for the "effects" of maladaptive narcissism, whereas maladaptive narcissism is positively related to social misconduct (Barry, Frick, & Killian, 2003;Barry, Pickard, & Ansel, 2009). As such, it is possible that maladaptive, but not adaptive, narcissism is associated with controlling coach behaviours. In addition, global narcissism could be associated with controlling coach behaviours due to its maladaptive facet.
Extending the work of Matosic et al. (2016Matosic et al. ( , 2017, in this study we tested potential relations between narcissism and controlling behaviours via coaches' effectiveness and normalcy beliefs about controlling interpersonal style, respectively. Such beliefs have been previously examined as potential antecedents of controlling behaviours (Reeve et al., 2014), and, as such, may constitute an explanatory mechanism for coaches' use of controlling behaviours. Effectiveness beliefs (Reeve et al., 2014) refer to how successful or impactful an interpersonal style is judged by individuals in positions of authority (e.g. coaches, teachers). Normalcy beliefs refer to how normative (i.e. common, accepted, or expected) an interpersonal style is judged by individuals in positions of authority. Both effectiveness and normalcy beliefs about controlling interpersonal style are positively associated with use of controlling behaviours by teachers (Reeve et al., 2014). One reason for this association is that teachers think controlling behaviours (e.g. offering rewards) promote students' engagement (Boggiano, Barrett, Weiher, McClelland, & Lusk, 1987). Another reason is that teachersespecially those in schools characterised by competition, external evaluation, and strict time constraintsregard controlling behaviours as the norm (Barrett & Boggiano, 1988). What is considered as normative may also be considered effective, and therefore teachers who endorse normalcy and effectiveness beliefs about controlling interpersonal style view controlling strategies as acceptable (Reeve et al., 2014). By implication, coaches who consider controlling interpersonal style as effective may also consider it as a norm, and will therefore be likely to enact controlling behaviours when interacting with their athletes.
We examined whether effectiveness and normalcy beliefs about controlling interpersonal style represent mechanisms through which narcissism may be associated with coaches' use of controlling behaviours. This process has the potential to explain why coaches high in narcissism report more frequent engagement in controlling behaviours (Matosic et al., 2016(Matosic et al., , 2017. Specifically, coaches high in narcissism may hold favourable effectiveness and normalcy beliefs regarding controlling interpersonal style, and this allows them to view controlling behaviours as legitimate and justifiable. Consistent with this contention, higher levels of global, adaptive, and maladaptive narcissism have been positively related to normalcy beliefs regarding aggression and bullying (e.g. social exclusion, verbal threat), and these beliefs have been linked to stronger engagement in such behaviours (Blinkhorn, Lyons, & Almond, 2016;Onishi, Kawabata, Kurokawa, & Yoshida, 2012). For example, in a school setting, narcissistic individuals are more likely to be aggressive when perceiving higher levels of classroom norms for aggression (Onishi et al., 2012). Additionally, adaptive and maladaptive narcissists engage in more aggressive and bullying behaviours, respectively, because they believe these behaviours are acceptable and normative (Ang, Tan, & Mansor, 2011;Blinkhorn et al., 2016). However, although adaptive narcissism appeared to be positively linked to antisocial behaviour, the effects may be due to the confluence of this construct and that of maladaptive narcissism (i.e. maladaptive narcissism was not covaried out from adaptive narcissism). A recent meta-analysis of the narcissism and leadership literature further bolsters the relevance of effectiveness beliefs (Grijalva et al., 2015). The meta-analysis reported positive relations among global, adaptive, and maladaptive narcissism with self-reported leadership effectiveness. It also showed that global narcissists engaged in aggressive behaviours as a means of influencing and guiding others. Given the established links between aggressive and bullying behaviours and controlling coach behaviours (Bartholomew et al., 2009), we surmise that a similar process operates between narcissism (global, adaptive, maladaptive) and controlling coach behaviours via effectiveness and normalcy beliefs about controlling interpersonal style.
As well as aiming to understand more deeply antecedents of controlling coach behaviours, we investigated coaches' moral disengagement as a potential outcome of controlling coach behaviours. SDT literature has found that controlling coach behaviours may lead to detrimental outcomes (Ntoumanis, 2012). A group of detrimental outcomes that has been scarcely examined refers to morality-related, and, as such, we focus on coach moral disengagement.
Moral disengagement is a collective term for eight psychosocial mechanisms (e.g. moral justification, displacement of responsibility, attribution of blame) that allow people to justify or rationalise inappropriate behaviour (Bandura, 2002). These mechanisms facilitate such conduct by reducing or eliminating the emotional consequences that normally follow one's untoward action, and would ordinarily deter it. Importantly, moral disengagement can be used socially to justify or rationalise one's harmful conduct to others (Bandura, 2016). As such, coaches who behave in a controlling manner may engage in moral disengagement to justify or rationalise their controlling behaviours to others. Thus, higher frequency of controlling coach behaviours may be associated with increased moral disengagement. To date, researchers have reported a positive relation between athletes' perceptions of controlling coach behaviours and athlete moral disengagement (Hodge & Gucciardi, 2015), but the relation between controlling coach behaviours and coach moral disengagement has not been addressed.

Hypotheses
We first tested a model in which global narcissism predicted controlling behaviours via effectiveness and normalcy beliefs about controlling interpersonal style. This model expands on Matosic et al. (2016Matosic et al. ( , 2017 who obtained a positive link between global narcissism and controlling coaching behaviours. Similar to Barry et al. (2003) and Barry and Malkin (2010), we report the results for overall narcissism first, followed by a more elaborate version of that model that differentiates between adaptive and maladaptive narcissism. In these two models, we hypothesised that global and maladaptive, but not adaptive, narcissism would be positively and directly associated with controlling coach behaviours. We also hypothesised that effectiveness and normalcy beliefs about controlling interpersonal style would be positively associated with controlling coach behaviours, and that controlling coach behaviours would be positively associated with coach moral disengagement. Finally, we hypothesised that global, adaptive, and maladaptive narcissism would be positively linked with controlling coach behaviours indirectly, via both effectiveness and normalcy beliefs about controlling interpersonal style (Figures 1 and 2).

Method Participants
Participants were 210 coaches (164 men, 46 women) from a variety of team (e.g. football, rugby) and individual (e.g. swimming, athletics) sports, as well as levels of competition (e.g. national, international, regional). Coaches' ages ranged from 18 to 88 years (M = 35.76, SD = 13.53; 23 participants did not report their age). Coaches had on average 12.99 (SD = 9.59) years of coaching experience and were predominantly White British (83.10%).

Narcissism
We assessed coach (global) narcissism with the 40-item, forced-choice Narcissistic Personality Inventory (NPI; Raskin & Terry, 1988). We opted for the NPI over other measures of narcissism (e.g. Narcissistic Admiration and Rivalry Questionnaire; Back et al., 2013), because the NPI is the standard scale to assess grandiose narcissism (Boldero, Bell, & Davies, 2015;Miller, Lynam, & Campbell, 2016a, 2016b, and this would allow us to compare our results with relevant findings in the literature. For each item, participants chose between a narcissistic (e.g. "I think I am a special person") and a non-narcissistic (e.g. "I am no better or no worse than most people") statement. Scores range from 0 to 40, with higher scores reflecting higher levels of narcissism. Evidence support the NPI's construct validity and internal consistency in the sports domain (Roberts et al., 2018;Woodman et al., 2011).
Narcissism has been subdivided into two facets, adaptive and maladaptive (Barry et al., 2003;Barry & Malkin, 2010). As per Barry et al. (2003) and Barry and Malkin (2010), we calculated adaptive narcissism scores by averaging items of the Authority (e.g. "I like to have authority over others") and Self-sufficiency (e.g. "I always know what I am doing") subscales of the NPI. Further, we calculated maladaptive narcissism score by averaging items from Exploitativeness (e.g. "I can make anybody believe anything I want them to"), Entitlement (e.g. "I expect a great deal from other people"), and Exhibitionism (e.g. "I really like to be the centre of attention") subscales of the NPI (Barry et al., 2003;Barry & Malkin, 2010). All subscales have good construct validity and internal consistency (Barry et al., 2003;Barry & Malkin, 2010).

Controlling coach behaviours
We assessed controlling coach behaviours using the 15-item Controlling Coach Behaviors Scale (CCBS; Bartholomew, Ntoumanis, & Thøgersen-Ntoumani, 2010). Responses ranged from 1 (strongly disagree) to 7 (strongly agree), with higher scores reflecting more controlling behaviours. For the purposes of this study, we modified the CCBS to refer to coach self-perceptions (e.g. "I try to control what athletes do during their free time;" for a similar approach, see Stebbings, Taylor, & Spray, 2011) rather than athlete perceptions (e.g. "My coach tries to control what I do during my free time"). Evidence supports the modified scale's construct validity and internal consistency (α = .83; Stebbings et al., 2011).

Controlling interpersonal style beliefs
For each of the controlling coaching behaviours items (CCBS; Bartholomew et al., 2010), the participants were asked to answer two questions about how effective and two questions about how normative the behaviour captured by this item was. The effectiveness and normalcy beliefs regarding controlling interpersonal style were assessed via a questionnaire developed by Reeve et al. (2014). Two items measured coaches' effectiveness beliefs about a controlling interpersonal style (i.e. "How effective would this approach to coaching be in terms of motivating and engaging your athletes?" and "If you coach this way, how much would your athletes benefit in terms of learning and achievement?"). For effectiveness beliefs, responses ranged from 1 (extremely ineffective, it would not work at all) to 7 (extremely effective, it would certainly work) for the first item, and from 1 (no benefit at all) to 7 (a great deal of benefit) for the second item. Additionally, two items measured coaches' normalcy beliefs about controlling interpersonal style (i.e. "Does this approach describe what the other coaches you know and work with do as coaches?" and "How typical or common is this approach to coaching for the coaches you know and work with?"). For normalcy beliefs about controlling interpersonal style responses ranged from 1 (no, not at all) to 7 ("yes, very much") for the first item, and from 1 (extremely atypical, uncommon) to 7 (extremely typical, common) for the second item. The scale has good construct validity and internal consistency (e.g. Reeve et al., 2014).

Moral disengagement
We assessed moral disengagement using the 8-item Moral Disengagement in Sport Scale-Short (MDSS-Short; Boardley & Kavussanu, 2008). A sample item is: "Shouting at the opponent is okay as long as it does not end in violent conduct." Responses ranged from 1 (strongly disagree) to 7 (strongly agree). Evidence supports this scale's construct validity and internal consistency (α = .87; Hodge & Gucciardi, 2015). As the scale was originally validated with athletes, we examined its factorial validity with the present sample of coaches using confirmatory factor analysis. Initial specification of the 8-item unidimensional model from Boardley and Kavussanu (2008) resulted in poor fit, χ 2 (20) = 78.7, p ≤ .01, CFI = .84, RMSEA = .12, SRMR = .07. However, specification of a model that accounted for a significant correlation between the error terms of two items identified through model misfit statistics resulted in very good model fit, χ 2 (19) = 35.1, p ≤ .05, CFI = .96, RMSEA = .06, SRMR = .06. Specifying correlated errors when present is important to prevent possible inaccurate parameter estimates (see Kline, 2015).

Procedures
Following approval from the ethics committee of the first author's institution, we recruited coaches via national governing bodies, sport club websites, social media, and personal contacts. We explained the purpose and procedure of the study to coaches via email or in person. We emphasised that their participation was voluntary and all information would be confidential. Prior to completing the 15-min online (collected via the Lime Survey online application) or hardcopy (collected in person) multi-section questionnaire, we provided coaches with a consent form (online or face-to-face). We received 204 online and 11 hardcopy responses. Out of 215 participants, three were duplicates, one was not based in the United Kingdom, and one requested withdrawal. Thus, the final data set consisted of 210 participants (199 online and 11 hardcopy responses). Upon completion of the survey, participants were able to enter a prise draw. We randomly selected two participants to win a £50 Amazon voucher each as compensation.

Data analyses
In preliminary analyses, we calculated means, standard deviations, correlations, and tested for internal reliabilities, as well as univariate and multivariate normality (i.e. skewness and kurtosis), using SPSS 22.0 software. We averaged scores in all subscales we used. We then evaluated the main study hypotheses by conducting path analyses with maximum likelihood (ML) estimation using Mplus 7.2 software (Muthén & Muthén, 1998. We assessed model fit using the χ 2 goodness-of-fit index, root mean-square error of approximation (RMSEA), comparative fit index (CFI), Tucker-Lewis index (TLI), and square root mean residual (SRMR). CFI and TLI values exceeding .95 are indicative of good fit, while SRMR and RMSEA values≤ .08 and .06, respectively, are considered satisfactory (Hu & Bentler, 1999). We calculated indirect effects using bias-corrected (BC) bootstrapped 95% confidence intervals (CIs) with 5000 resamples, as recommended by Preacher and Hayes (2008). We report the standardised version of specific indirect effects and their BC-CIs. A 95% CI not containing zero indicated a statistically significant indirect effect (Preacher & Hayes, 2008).

Results
We present descriptive statistics, Cronbach Alpha's (α) coefficients, and inter-correlations for all study variables in Table 1. All the variables had high internal consistency and were normally distributed (skewness range: −.238-.706, kurtosis range: −1.36-−.001). Correlation coefficients were in the expected direction (see Table 1) and ranged in effect size from small to large (Cohen, 1988

Direct and indirect effects
We conducted path analyses to test our models (Figures 3 and 4). The fit indices for our first a priori hypothesised model indicated good model fit: χ 2 (3) = 3.27, p = 0.35, CFI = 1.00, TLI = .99, RMSEA = .02, SRMR = .03. As shown in Figure 3, global narcissism was positively associated with controlling coach behaviours, but not to effectiveness and normalcy beliefs about controlling interpersonal styles. Effectiveness and normalcy beliefs about controlling interpersonal styles were positively associated with controlling coach behaviours. Finally, controlling coach behaviours was positively associated with coach moral disengagement. In the first model, the proposed indirect effect between global narcissism and controlling behaviours via effectiveness and normalcy beliefs about controlling interpersonal style was not significant (Table 2). Additionally, the indirect effects of narcissism on moral disengagement via effectiveness belief and controlling  The fit indices for our second a. priori hypothesised model also indicated good model fit: χ 2 (4) = 6.28, p = 0.18, CFI = .98, TLI = .94, RMSEA = .05, SRMR = .03. As shown in Figure  4, adaptive narcissism was positively associated with effectiveness beliefs, but was unrelated to normalcy beliefs about controlling interpersonal style. Also, adaptive narcissism was not directly related to controlling behaviours. Maladaptive narcissism was not associated with either of the beliefs, but had a direct significant positive link with controlling behaviours. Additionally, effectiveness and normalcy beliefs about controlling interpersonal style were positively associated with controlling coach behaviours. Finally, controlling coach behaviours was positively associated with coach moral disengagement. In the second model, the total indirect effect between adaptive narcissism and controlling coach behaviours via effectiveness and normalcy beliefs about controlling interpersonal style was statistically significant. The indirect effect accounted for 81.58% of the total effect (Preacher & Kelley, 2011). In addition, the specific indirect effect between adaptive narcissism and controlling coach behaviours via effectiveness beliefs about controlling interpersonal style was statistically significant. This specific indirect effect explained 56.58% of the total effect (see Table 2). No other significant indirect effects emerged.

Discussion
We advanced prior research on coaching from a SDT perspective by testing models that linked antecedents (global, adaptive, and maladaptive narcissism; effectiveness and normalcy beliefs about controlling interpersonal style) and consequences (moral disengagement) of coaches' controlling behaviours. We obtained support for all our direct-effect hypotheses such that: (a) global and maladaptive, but not adaptive, narcissism were positively associated with controlling coach behaviours, (b) effectiveness and normalcy beliefs about controlling interpersonal style were positively associated with controlling coach behaviours, and (c) controlling coach behaviours were positively associated with coach moral disengagement. However, only the indirect effect of adaptive narcissism on controlling coach behaviours via effectiveness beliefs about controlling interpersonal style was supported. Stated otherwise, adaptive narcissism was positively associated with controlling behaviours through effectiveness beliefs about controlling interpersonal style. Indirect effects of maladaptive narcissism on controlling coach behaviours via effectiveness beliefs about controlling interpersonal style, and adaptive and maladaptive narcissism on controlling coach behaviours via normalcy beliefs about controlling interpersonal style, respectively, were not supported.

Antecedents of controlling coach behaviours
As expected and also previously found by Matosic et al. (2016Matosic et al. ( , 2017, coach global narcissism was moderately positively associated with controlling coaching behaviours. In line with literature on narcissistic leaders (Grijalva et al., 2015;Schoel et al., 2015;Sedikides & Campbell, 2017), such coaches may pressure their players to the limit in order for the coaches to gain self-enhancement benefits, such as admiration and reflected glory (Mathieu & St-Jean, 2013;Woodman et al., 2011). As hypothesised with regard to the direct effects, maladaptive, but not adaptive, narcissism was associated with controlling coach behaviours. Put otherwise, entitlement, exhibitionism, and exploitativeness, but not authority or self-sufficiency, are likely to explain the frequency of controlling coach behaviours. For example, coaches who feel entitled to demand a great deal from their athletes, require unconditional praise and admiration from them, are comfortable in "using" them, and pressurise hard their athletes to the limit of their performance in order to achieve their own (i.e. coaches') desired ends. Adaptive narcissism was not directly associated with controlling behaviours, as expected, when controlling for the effects of maladaptive narcissism (Barry et al., 2003). Adaptive narcissismbut not global or maladaptive narcissismwas positively indirectly associated with controlling behaviours via coaches' effectiveness beliefs about a controlling interpersonal style. Higher levels of adaptive narcissism in coaches were associated with stronger effectiveness beliefs about controlling interpersonal style, which in turn was associated with more frequent controlling behaviours. According to Barry and Malkin (2010), adaptive narcissists evaluate situations before taking action to ensure that they are confident of their success. Thus, it is possible that coaches with higher levels of adaptive narcissism tend to use controlling behaviours when they believe those behaviours are effective. In contrast, most of the effect of global and maladaptive narcissism on controlling behaviours was direct; effectiveness beliefs about controlling interpersonal style did not have unique predictive ability over and above narcissism. This could be because coaches high in maladaptive narcissism feel that they are entitled to use controlling behaviours over their athletes (in a demonstration of power over them), irrespectively of whether such behaviours are deemed as effective.
There were no significant indirect effects of global, adaptive, and maladaptive narcissism on controlling coach behaviours via coaches' normalcy beliefs about controlling interpersonal style, although those beliefs were positively associated with controlling behaviours, in line with findings from the education literature (Reeve et al., 2014). The non-significant indirect effects could be explained through global, adaptive, and maladaptive aspects of narcissism being linked with the need to be different from others (Raskin & Terry, 1988), making individuals high in these traits less inclined to be influenced by beliefs about norms. For example, coaches who believe they are extraordinary (i.e. adaptive trait) and who like to be the centre of attention (i.e. maladaptive trait) are disinclined from following the norm, as this practice may not benefit them directly.

Controlling coach behaviours and moral disengagement
As hypothesised, controlling coach behaviours were positively associated with coach moral disengagement. In other words, coaches who reported using more controlling coach behaviours were more inclined to morally disengage. Controlling coaches may use moral disengagement to justify and rationalise athletes' engagement in aggressive and transgressive behaviours, because they see the potential for competitive advantage stemming from such athlete behaviour. This novel finding is consistent with past research that has linked coaches' controlling behaviours with athletes' moral disengagement (Hodge & Gucciardi, 2015;Hodge & Lonsdale, 2011;Hodge, Hargreaves, Gerrard, & Lonsdale, 2013). Integration of relevant theory (Bandura, 2016) with the findings from the present research and those of Hodge and Gucciardi (2015), Hodge et al. (2013), and Hodge and Lonsdale (2011) is consistent with the possibility that controlling coaches promote athletes' moral disengagement through their own use of it. Further investigations that expand our model to examine whether coaches' use of moral disengagement fosters athlete moral disengagement are therefore encouraged.

Limitations and future directions
Our study was based on coach self-reports, which could have been influenced to some degree by socially desirable responding. However, broadly similar findings were reported by Matosic et al. (2016), who collected data from both coaches and athletes. Nevertheless, follow-up research may incorporate alternative or additional methods of assessing coach behaviours, including observational techniques (i.e. blind rating of coach behaviours), to guard against such influences (Smith et al., 2015). Also, our study used a cross-sectional design and hence our findings do not allow causal inferences. Longitudinal designs are needed to help identify the temporal sequencing of relations between variables. Another way forward would be to implement interventions designed to influence coach effectiveness beliefs and ensuing controlling behaviours in samples of coaches with varying levels of narcissism.
Interestingly, researchers have distinguished between narcissism admiration (e.g. striving for uniqueness, charmingness) and narcissism rivalry (e.g. striving for supremacy, aggressiveness) as the bright and dark sides of narcissism, respectively (Back et al., 2013). Arguably, these concepts are complementary to adaptive and maladaptive narcissism components (Back et al., 2013). The adaptive component of narcissism (e.g. "I have a natural talent for influencing people") is highly comparable to narcissism admiration (e.g. "Mostly, I am very adept at dealing with other people"), and the maladaptive component of narcissism (e.g. "I get upset when people don't notice how I look when I go out in public") is highly comparable to narcissism rivalry (e.g. "I react annoyed if another person steals the show from me"). The admiration-rivalry distinction is new in the narcissism literature, and its conceptual and operational overlap with adaptive-maladaptive narcissism should be explored in future research.
Future work could also investigate the relations between grandiose and vulnerable forms of narcissism and controlling interpersonal style (cf. Sedikides, Ntoumanis, & Sheldon, 2018). Our study has addressed the relations between grandiose narcissism (i.e. narcissistic personality trait) and its facets (i.e. adaptive and maladaptive narcissism) with controlling interpersonal style. No research, however, has examined pathological form of narcissism (i.e. vulnerable narcissism; Thomaes et al., 2018) within sport context. Such research on vulnerable narcissism (using the hypersensitivity narcissism scale; Hendin & Cheek, 1997) may provide new insights into narcissism in sport coaches. Finally, assessment of additional "dark" personality traits, such as psychopathy and Machiavellianism (Paulhus & Williams, 2002), as antecedents of controlling coach behaviours would be useful. Psychopathy and Machiavellianism share maladaptive characteristics with narcissism, such as striving for self-promotion, lacking empathy, engaging in aggressive behaviours, and failing to show organisational success (Eisenbarth, Hart, & Sedikides, 2018;Muris, Merckelbach, Otgaar, & Meijer, 2017).

Conclusion and implications
Our research makes several unique contributions to the literature, particularly in regards to understanding antecedents of a controlling interpersonal style. First, we replicated and extended previous findings by showing that global narcissism and its maladaptive facet qualify as antecedents of controlling coaching behaviours. Second, we illustrated a positive indirect effect between adaptive narcissism and controlling coaching via effectiveness beliefs about controlling coaching. Finally, we demonstrated that controlling coaching behaviours were positively associated with coaches' reports of moral disengagement.
Extending on Matosic et al. (2017), our findings could inform coach-focused education programmes that aim to promote adaptive coaching environments in sport. From a motivational perspective, literature identifies specific examples of controlling coach behaviours and ways in which they can be substituted by autonomy-supportive ones (Ntoumanis, Quested, Reeve, & Cheon, 2018). From a narcissism perspective, evidence outside sport indicates that increasing empathy (Hatcher et al., 1994;Hepper, Hart, & Sedikides, 2014) or self-affirmation (Thomaes, Bushman, Orobio de Castro, Cohen, & Denissen, 2009) can reduce narcissistic tendencies. Our findings showcase the potential for combining the two perspectives. One could develop coach-education programmes that reduce narcissistic tendencies, challenge beliefs regarding the effectiveness of controlling coaching behaviours, and train coaches to replace such behaviours with autonomy-supportive ones. In doing so, one might curtail coach moral disengagement, in light of evidence that moral disengagement is positively linked to antisocial sport behaviour (Boardley & Kavussanu, 2011).