A junior female volleyball team was tested. There were 19 athletes in the team, along with 3 coaches. The athletes’ average age was 15.63 (SD=1.32), and their average year of training was 3.12 (SD=1.02). The coaches were all males (age range 25–38years). Two athletes quit the team during the study, so that there were complete data of 17 athletes. The athletes and coaches provided informed consent prior to participation. The institutional ethics committee approved the study. The results were kept confidential, only used for the research. The experimental procedure was conducted individually to each athlete by the experimenter (i.e., the second author who worked in the team as a sport psychologist at the time) in a quiet room in the training center of the team. The athletes were informed that if feeling uncomfortable they could skip a test or quit the procedure at any time and their data would not be revealed to the coaches or anyone else outside the research group. The athletes were fully debriefed about the study after the intervention and approved their data to be analyzed and published.

We performed EC interventions on the athletes twice a week for 6weeks. Each athlete received EC interventions for two randomly selected coaches and no intervention for the other coach. In this way, the same athletes’ relationships with the EC-intervened and non-intervened coaches were compared, so that we could minimize possible influence of other factors on the coach–athlete relationships. Before (Week 0), during (Week 4), and at the end of the EC (Week 6), as well as post-EC (Week 6+2, i.e., 2weeks after the EC ended), the following measures were tested on the athletes (see Figure 1). We measured the athletes’ affective associations with the coaches by using biases in an emotion recognition test (Maner et al., 2005; Krems et al., 2015) and an implicit association test (IAT; Greenwald et al., 2003; Karpinski and Steinman, 2006). We measured the athletes’ explicit evaluation of their relationships with the coaches by using the Coach–Athlete Relationship Questionnaire (CART-Q; Jowett and Ntoumanis, 2004; Zhong and Wang, 2010), in which the athletes explicitly give ratings to a series of evaluative statements about the relationships. We also tested the athletes’ reaction time and physiological responses of automatic nervous system (indexed by GSR and HRV; Quintana et al., 2012) when viewing the coaches’ facial images in a face orientation judgment task.

We analyzed both the individual effects of the EC intervention on each athlete and the overall effects across the athletes. We examined the individual effect by visually inspecting the progress of each athlete’s relationship with each coach along with the EC intervention. And we computed the percentages of athletes showed improvement in each measure at the end of the EC intervention (i.e., Week 6) compared to before the intervention (i.e., Week 0). We conducted 4 (testing time: Week 0, Week 4, Week 6, Week 6+2)×2 (intervention status: EC intervention, no intervention) mixed two-way analyses of variance (ANOVAs) on each measure to analyze the overall effects of the EC intervention with time.

We visually inspected the variation of each measure regarding each athlete’s relationship with each coach along with the EC intervention. Figures 2, 3 depicted the progresses of two athletes’ relationships with the coaches with and without the EC intervention. The athletes’ progresses showed similarities and differences.

Specifically, on the emotion recognition biases, many athletes showed a relatively high rating with a large fluctuation in happiness. For example, Athlete 2 exhibited an increase in the happiness rating for the two coaches involved in EC, while there was no increase (and even a temporary decrease at Week 4) in the happiness rating for the coach that was not involved in EC (Figure 3A). Complementary to the changes in the happiness rating, Athlete 2 exhibited a remarkable increase in the anger rating for the coach without EC at Week 4, which extended to Week 6 and Week 6+2, while the anger rating for one EC coach remained at the lowest level and the anger rating for the other EC coach temporarily increased at Week 4 yet then decreased to the lowest level (Figure 3B). Athlete 1 exhibited continuous increase in the happiness rating for one EC coach while the happiness rating for the other two coaches fluctuated in an increasing trend (Figure 2A); meanwhile, the anger ratings for the three coaches fluctuated in a decreasing trend (Figure 2B). On the ratings of disgust, sadness, fear, and surprise, most athletes scored at low levels with small variation (see Figures 2C–F, 3C–F as examples).

On the IAT score, most athletes exhibited increases for the EC coach and no changes for the no-EC coach (e.g., Athlete 1 as depicted in Figure 2G). Additionally, Athlete 2’s IAT score on the no-EC coach exhibited a decrease at Week 4 (Figure 3G), consistent with the athlete’s decreased happiness rating and increased anger rating with the no-EC coach at Week 4.

On the CART-Q questionnaire, most athletes rated high scores with all the three coaches. Athlete 1 exhibited a small but steady increasing trend in the CART-Q for all the three coaches (Figure 2H). Athlete 2 exhibited an increasing trend in the CART-Q for the EC coaches, while the score remained unchanged for the no-EC coach (Figure 3H). Interestingly, Athlete 2’s explicit evaluation of the relationship with the no-EC coach (i.e., CART-Q score) remained unchanged at a high level even at Week 4, while the implicit measures of the relationship (i.e., happiness rating, anger rating, IAT) exhibited a substantial decrease.

Athletes’ physiological responses upon seeing the coaches’ faces exhibited no clear patterns. For example, Athlete 1’s HRV with all the three coaches fluctuated (Figure 2I), whereas Athlete 2’s HRV remained largely stable except a sharp decrease from Week 0 to Week 4 for the no-EC coach (Figure 3I). Athlete 1’s GSR for all the three coaches increased from Week 0 to Week 6 and then decreased at Week 6+2 (Figure 2J), while Athlete 2’s GSR for all the three coaches decreased at Week 4 and Week 6 and then increased at Week 6+2 (Figure 3I).

We calculated the percentages of athletes that showed increases at Week 6 (i.e., at the end of EC) in comparison with Week 0 (i.e., before EC) for the coaches with or without the EC intervention in each measurement (see Table 1). The results showed the trends that EC led to improvements in a considerable portion of athletes on the happiness rating, the IAT, and the CART-Q measurements. In contrast, a large portion of athletes showed increases on the anger and the disgust ratings for the coaches without EC, while only a few athletes showed such increases for the coaches with EC.

We examined the overall effects of EC on the athletes’ biases in recognizing coaches’ emotions by conducting a set of 4 (testing time: Week 0, Week 4, Week 6, Week 6+2)×2 (intervention status: EC intervention, no intervention) mixed two-way ANOVAs separately on the ratings of happiness, anger, disgust, sadness, fear, and surprise, with the testing time as the within-participants variable and the intervention status as the between-participants variables. The results are listed in Table 2.

The key results were that the Testing Time × Intervention Status interactions were significant for happiness, anger, and disgust ratings (ps<0.015; see Figure 4). Further analysis showed that for the EC-intervened coaches, the main effect of testing time on happiness was significant, F(3, 135)=3.577, p=0.021, η_p²=0.186, showing a higher happiness rating at Week 4 than Week 0 (3.32 vs. 2.50, p=0.012) while no significant difference in the comparisons between the other weeks (ps>0.141). For the non-intervened coaches, the main effect of testing time on happiness was not significant. On the other hand, for the non-intervened coaches, the main effect of testing time on anger [F(3, 135)=5.589, p=0.002, η_p²=0.263] and disgust [F(3, 135)=10.480, p<0.001, η_p²=0.401], both showing higher ratings at Week 4, 6, and 6+2 in comparison with Week 0 (2.61, 2.52, 2.22 vs. 1.57, p=0.005, 0.008, 0.025; 1.83, 2.00, 1.47 vs. 1.24, p=0.002, p<0.001, p=0.033, respectively for anger and disgust). None of the other comparisons were significant (ps>0.313). For the EC-intervened coaches, the main effects of testing time on anger and disgust were not significant. On the sadness, fear, and surprising ratings, the ANOVAs did not yield any significant main effects or interactions (ps>0.120). The results demonstrated that for the coaches with EC intervention, the happiness association increased with time while the anger and disgust associations remained unchanged, whereas for the coaches without EC intervention, the anger and disgust associations deteriorated with time while the happiness association remained unchanged (see Figure 4).

Analogous to the analyses above, we conducted a 4 (testing time)×2 (intervention status) mixed two-way ANOVA on the D score in the SC-IAT task. The main effects of testing time and intervention status were significant, and more importantly, the Testing Time × Intervention Status interaction was significant (p<0.001, see Table 3). Further analysis showed that for the EC-intervened coaches, the Ds at Week 4, 6, and 6+2 were all significantly higher than that at Week 0 (0.32, 0.62, 0.30 vs. 0.04, p<0.001, p<0.001, p=0.003, respectively), and the D at Week 6 was significantly higher than that at Week 4 (0.62 vs. 0.32, p=0.001) and that at Week 6+2 (0.62 vs. 0.30, p=0.001), while there was no significant difference between Week 4 and Week 6+2. For the non-intervened coaches the D score did not vary with time, F(3, 135)=1.940, p=0.136, η_p²=0.110. The results demonstrated that the athletes’ implicit associations with the coaches become more and more positive as the implementation of the EC intervention, and then descended after the intervention stopped (see Figure 4).

Analogous to the analyses above, we conducted a 4 (testing time)×2 (intervention status) mixed two-way ANOVA on the average score of CART-Q. The results showed that the Testing Time × Intervention Status interaction was significant (p=0.037), while the main effects were not significant (ps>0.181; see Table 3). Further analysis showed that for the EC-intervened coaches, the CART-Q scores at Week 6 and Week 6+2 were significantly higher than that at Week 0 (6.44, 6.41 vs. 6.09, p=0.025, 0.020, respectively), while there was no significant difference between Week 4, 6, and 6+2 (ps>0.075). For the non-intervened coaches, the CART-Q score did not vary with time, F(3, 48)=0.263, p=0.852, η_p²=0.016. The results demonstrated that the EC intervention improves the athletes’ explicit evaluation of their relationships with the coaches (see Figure 4). We also performed repeated measure ANOVAs on the three dimensions of the CART-Q (i.e., emotion, thoughts, and behaviors) separately. The result patterns of the three dimensions were all similar to that of the total score, showing trends of improvement for the EC coaches and no improvement for the no-EC coach. Yet none of the effects reached significant (ps>0.111) in the dimensions of thoughts and behaviors. In the emotion dimension, the main effect of testing time was significant, F(3, 147)=3.958, p=0.009, η_p²=0.075, and the Testing Time × Intervention Status interaction approached significant, F(3, 147)=2.378, p=0.072, η_p²=0.046, suggesting that the emotion dimension may be relatively sensitive to the EC intervention.

We firstly subtracted the athletes’ reaction time, GSR, and HRV when judging the direction of the bar from those when judging the orientation of the coaches’ faces, so as to obtain the measures reflecting the effects of the coaches’ faces. Then, analogous to the analyses above, we conducted 4 (testing time)×2 (intervention status) mixed two-way ANOVAs on these measures. The Testing Time×Intervention Status was not significant for any of the measures (ps>0.231; Table 3), indicating that the EC intervention did not lead to significant changes in the athletes’ reaction time and physiological responses when viewing the coaches’ faces.

The present study illustrated the importance of affective associations in coach–athlete relationships. As suggested by the interdependence theory of interpersonal relationship (Kelly and Thibaut, 1978) and the dual-process models (Smith and DeCoster, 2000; Gawronski and Bodenhausen, 2006; Hicks and McNulty, 2019), in daily interactions with coaches, athletes may automatically track their positive and negative experiences. If a coach’s presence is frequently accompanied with positive/negative events and affect, athletes may establish associations between the coach and the positive/negative affect. These affective associations are automatically activated whenever people encounter or merely think about the partner, and thus impact their judgments in the interpersonal interactions (McNulty and Olson, 2015; Hicks and McNulty, 2019). If an athlete has established a negative association with a coach, the athlete would be biased to judge that the coach is expressing anger and disgust to him/her even when the coach is in a neutral state.

Affective associations may further impact the athletes’ behaviors and thoughts toward the coaches. Affective associations are automatically activated in milliseconds without conscious awareness, and the activated affect directly signal to the individual how to behave with respect to the stimulus (Fazio, 1990). Athletes may automatically approach the coaches associated with positive affect and behave in a positive manner, whereas they may automatically start the ‘fight or flight’ response towards the coaches associated with negative affect. Moreover, the activated affect can also impact the information processing related to the stimulus and the cognitive beliefs about the stimulus (Kiviniemi et al., 2007). Specifically, the affect can bias information processing by guiding attention (Roskos-Ewoldsen and Fazio, 1992). When an athlete has formed negative associations with a coach, he/she may selectively process the coach’s negative information and neglect positive information, which will confirm and exacerbate the negative bias. His/her cognitive belief about the coach may also be attuned with the negative information to justify the negative affect, so that the athlete may consciously form the belief that the coach is a ‘bad’ person who deliberately makes him/her feel bad.

The present study suggests that EC may be adopted to modify affective associations and improve explicit evaluation of coach–athlete relationships. In the present EC procedure, a coach repeatedly appeared with objects of positive affect, which efficiently added weight to athletes’ positive associations with the coach and prevented negative associations from deterioration. Furthermore, according to the dual-process models (Kelly and Thibaut, 1978; Gawronski and Bodenhausen, 2006; Hicks and McNulty, 2019), the affective associations can be automatically activated and taken into account when athletes deliberately evaluate their relationship with the coaches. Thus, as EC modified the implicit associative structures, the athlete’s explicit evaluation showed corresponding changes (Gawronski and Bodenhausen, 2006; Zerhouni et al., 2018). It is noteworthy that whereas the intervention stemmed from the dual-process theories, the consistent changes in the implicit and explicit measures can also be explained by single-process theories (e.g., Cunningham et al., 2007; Fazio, 2007). According to the single-process theories, the construct validity of the implicit measures such as IAT is questionable (Van Dessel et al., 2020; Schimmack, 2021), as both the implicit and explicit measures may reflect largely overlapping or unitary mental content, rather than explicit measures tapping explicit constructs and implicit measures tapping implicit constructs (De Houwer, 2014; Kurdi et al., 2021). In the present study, multiple measures were adopted not to test the dual-process and single-process theories, but to extensively explore the potential effects of EC on coach–athlete relationships from different aspects. The consistent changes in the measures shed light on the growing body of research on developing novel interventions in close relationship (McNulty et al., 2017; Faure et al., 2020).

An interesting finding in the present study is that the anger and disgust associations increased over time for the coaches without EC intervention while remained unchanged for the EC–intervened coaches. The result implies that the coach–athlete relationships in sports teams may tend to decline over time if not managed with extra care. Similar findings have been reported in research on marital relationships, showing that many married people evaluate their relationships less positively over time. The decline persists even in spouses with motivations to perceive their relationships positively and cognitive strategies that appear to help them do so (Hicks and McNulty, 2019). So far, research on the development of coach–athlete relationships over time is limited. Consistent with the marital relationships research, a case study discovered a regressive spiral in coach–athlete relationships after an initial ‘honeymoon’ phase (Jowett, 2003). In marital relationships, spouses with more positive associative processes are less likely to experience the decline (McNulty et al., 2013). In coach–athlete relationships, this may also be the case. Therefore, an EC intervention may be a helpful and needful tool for maintaining positive coach–athlete relationships over time. On the other hand, it should also be noted that coach–athlete relationships differ from marital relationships in a number of ways. For instance, a coach–athlete relationship is always one of uneven power and it usually occur in a group situation. Thus, it is better for interventions to be adapted accordingly. Additionally, the present results exhibited a decreasing trend in several measures on coach–athlete relationships after the EC intervention ceased, whereas previous research suggested that the effects of EC are resistant to extinction when strong CS–US contingency is no longer presented (Díaz et al., 2005; Bolders et al., 2012; Hütter et al., 2012). There has been a controversy on the extinction of EC effects (Hofmann et al., 2010; Gawronski et al., 2015). The present trend implied that some daily experiences with coaches may add weight to the negative associations in the coach–athlete relationships. Therefore, in coach–athlete relationships, the effect of EC interventions (and other interventions as well) may need to be reinforced from time to time.

Using EC to improve coach–athlete relationships in sports teams has its distinct advantages. Firstly, EC can impact the automatic processes that cannot be addressed by traditional interventions that address explicit behaviors only. EC strengthens positive affective associations in memory, which can further lead the individual to automatically perceive the stimulus in a more positive manner in future encounters (Gawronski and Bodenhausen, 2018). Secondly, EC does not cost much cognitive resource (Hütter et al., 2012). The EC intervention requires mere co-occurrence of the target person with valenced objects or events. In the present EC procedure, athletes were only conducting a simple search task while the EC produces effects incidentally through repeated paring of the coaches’ faces and positive stimuli.

In the present study, the EC intervention did not lead to significant changes in the athletes’ reaction time and physiological responses when viewing their coaches’ faces. It is possible that EC does not reduce perceived social threat from the coaches’ faces or longer EC interventions are needed for altering the reaction time and physiological responses. Additionally, the present study suggests that there may be limitations in adopting physiological measures to examine coach–athlete relationships, as the data may be noisy and prone to interferences from movement, sweat, or electromagnetic sources (Palumbo et al., 2017; Yetton et al., 2019). Moreover, the physiological data may not directly map to the people mental states, and different people may have different patterns of physiological responding (Vasilev et al., 2009; Calvo and D'Mello, 2010). Therefore, researchers in future studies may need to develop more delicate methods to collect and analyze the physiological data, and interpret the results cautiously.

Additionally, in order to control possible confounding factors regarding coach–athlete relationships, we tested athletes from one team and randomly assigned their multiple coaches to either the EC intervention condition or no intervention condition. This resulted in a limited number of participants in the study. The reliance upon a single team of athletes and a small number of target coaches limited the robustness and generalizability of the study. Thus, the present study is still exploratory. The interpretation of and the inferences from the results should be made cautiously. The present study can serve as a starting point for further intervention on larger scale. Future studies may extend this line of research by testing more athletes in more sports teams.

On the other hand, it should also be noted that in real life, each athlete is a unique individual, and so is each coach. Each coach–athlete relationship may evolve in a unique way based on various factors, such as the athlete’s skills and competitive level, the coach’s behavior style and authority, and the athlete’s and the coach’s personalities and experiences. Thus, in sport psychology practices, it is impractical to abstract the coach–athlete relationship and adopt a one-size-fit-all method to conduct interventions on all relationships. The present study suggests that EC may be effective for a considerable portion of athletes, and hence it could be a potential tool for sport psychologists to adopt when dealing with certain coach–athlete relationships. Nevertheless, individual effect needs to be stressed and individualized interventions are better to be developed for each relationship.

Future study may explore the ways to incorporate EC in sport psychology practices. For instance, EC may be combined with behavioral interventions to improve coach–athlete relationships. On one hand, interpersonal behaviors are essential for good relationships. Coaches need to improve their explicit behaviors towards athletes; otherwise the positive effects of EC may be gradually cancelled out by negative interpersonal interactions in daily life. On the other hand, people in interpersonal relationships are also directed by implicit processes occurring automatically without awareness (McNulty and Olson, 2015; Hicks and McNulty, 2019). Improvement in behaviors alone do not necessarily lead to improvement in emotions and relationship satisfaction (Williamson et al., 2016). Therefore, interventions may be most effective if they target both explicit behaviors and implicit affective associations.

The present study suggests that EC may be adopted as an effective intervention for coach–athlete relationships. It alters athletes’ affective associations with coaches to be more positive and improves their explicit evaluation of the relationships.