Role Ambiguity: Translation to Spanish and Analysis of Scale Structure

Abstract

The main objective of this study is to examine the structure of the Role Ambiguity Framework and a validation into Spanish of the Role Ambiguity Scale with professional players. We employed a sample of 604 male and female professional athletes, aged between 15 and 38 years. To find the best solution of the questionnaire’s factor structure, a set of different models was tested. Data analysis showed that bifactor-ESEM (exploratory structural equation modeling) solution provided the best degree of fit to the data. Furthermore, reliability and invariance across gender were also demonstrated, and the scale showed good concurrent validity. Results suggest that the structure of the Role Ambiguity is composed with one second-order factor and three first-order factors: scope-behavior, evaluation, and consequences. In addition, the adaptation of this scale to Spanish is a valid and reliable instrument to measure role ambiguity in male and female players.

Keywords

psychometric properties questionnaire role ambiguity

Group processes in the context of work teams has been an essential area of analysis to optimize performance. There are different lines of work, including the analysis of roles within a group (Beauchamp & Bray, 2001; Bosselut, Heuzé, & Sarrazin, 2010; Leo, González-Ponce, Sánchez-Miguel, Ivarsson, & García-Calvo, 2015). A large number of investigations in the work setting have been conducted about the theory of role, especially about role ambiguity (Kahn, Wolfe, Quinn, Snoek, & Rosenthal, 1964). The most comprehensive review of this construct was carried out by Kahn et al. (1964) in relation to the work and organizational context. Role ambiguity has been defined as a lack of clear, consistent information about the expectations associated with one’s position (Kahn et al., 1964). This concept has normally been studied as individuals’ perceptions when performing work functions independently of their coworkers (Jackson & Schuler, 1985). However, in conditions in which interdependence is an essential characteristic to develop the role, such as the sport context, role ambiguity should be further examined (Beauchamp & Bray, 2001; Eys & Carron, 2001).

Based on Kahn et al. (1964), Beauchamp, Bray, Eys, and Carron (2002) developed a conceptual model to study role ambiguity in sports teams. They developed the Role Ambiguity Scale (RAS) to attempt to measure this variable in sports. In spite of the fact that this tool has been extensively employed, to our knowledge, there are few studies that have corroborated the validity and reliability of the scale, and there is some divergence in the results found (Bosselut et al., 2010). Therefore, the goal of this study is to examine the structure of the role ambiguity framework.

Conceptual Model of Role Ambiguity

Beauchamp et al. (2002) conceptualized role ambiguity as multidimensional and situational. First, Beauchamp et al. (2002) described four dimensions of role ambiguity that may exist in teams of interdependent sports (Eys & Carron, 2001; Kahn et al., 1964). They refer to the lack of clarity associated with (a) the scope of responsibilities, which represents a lack of clear information regarding the general extent of the athletes’ responsibilities; (b) role behaviors, which reflect a lack of clear information regarding what behaviors are necessary to fulfill the athletes’ responsibilities; (c) role evaluation, which represents a lack of clear information about how the athletes’ role performance will be evaluated; and (d) role consequences, which reflect a lack of information about the consequences of not fulfilling role responsibilities. Second, these four manifestations can appear in different behavioral contexts (Eys & Carron, 2001), but research in the sport sphere has been limited to attack and defense, the two main contexts in which players perform a certain role (Bosselut et al., 2010).

Drawing on their conceptual model of role ambiguity, Beauchamp et al. (2002) developed the RAS. This scale has 40 items, with five items for each of the dimensions in the attack and defense phases, with higher scores reflecting lower role ambiguity and vice versa. The construct validity of this instrument was studied with confirmatory factor analysis (CFA), the results of which supported a four-factor structure in both contexts—attack and defense—with acceptable values of internal consistency for each one of the four subscales. This instrument has been used in numerous studies, finding that role ambiguity is related to other variables, such as role efficacy (Beauchamp et al., 2002; Beauchamp, Bray, Fielding, & Eys, 2005), player satisfaction (Eys, Carron, Bray, & Beauchamp, 2003; Høigaard et al., 2010), leadership (Beauchamp, Bray, Eys, & Carron, 2005), coach competence (Bosselut, Heuzé, Eys, Fontayne, & Sarrazin, 2012), and cohesion (Bosselut, McLaren, Eys, & Heuzé, 2012; Eys & Carron, 2001).

These results have shown the robustness of the validity of the RAS, but the relation of these dimensions with other variables is not very clear, such as, for example, with anxiety (Beauchamp, Bray, Eys, & Carron, 2003), intention of staying on the team (Eys, Carron, Bray, & Beauchamp, 2005), collective efficacy (Leo, González-Ponce, Sánchez-Miguel et al., 2015), or intra-team communication (Cunningham & Eys, 2007). Hence, it is interesting to continue to examine in greater depth the validity of the RAS.

Study Motivation

This study is an attempt to solve some of the limitations of the RAS, which have been examined in depth by Bosselut et al. (2010) in French culture. Among these limitations, the most outstanding is that, in spite of the fact that the conceptual model establishes a hierarchical organization of the latent variables with four first-order factors and one second-order factor, in the original version of the RAS, CFA was performed with a sole structure with four first-order factors, without taking into account other possibilities of a hierarchical nature. Therefore, it would be interesting to examine the validity of the instrument through different factorial structures, as this may help to better understand the behavior of the different items in each one of the factors and in the global factor of role ambiguity. For this purpose, in this study, we integrated classic (CFA) and emergent models (exploratory structural equation modeling [ESEM], bifactorial, and bifactorial ESEM) for the research of the psychometric multidimensionality of the construct (Marsh, Morin, Parker, & Kaur, 2014; Morin, Arens, & Marsh, 2016; Myers, Martin, Ntoumanis, Celimli, & Bartholomew, 2014). Through this study, we propose various first- and second-order structures to attempt to solve the limitation indicated previously by Bosselut et al. (2010).

Second, the discriminant capacity of each item within each factor has not been reported in most studies, so we lack information about the weights of the items within the factors. The results of the different studies show a tendency to obtain high correlations among the four dimensions (Beauchamp, Bray, Eys, et al., 2005; Cunningham & Eys, 2007; Eys et al., 2003; Eys et al., 2005), both in the range of role responsibilities and behaviors and in role evaluation and role consequences. This may mean that there is a conceptual overlap in some of the dimensions of role ambiguity. In fact, the French version of the RAS proposes joining the dimensions of Scope and Behavior, given the high correlations among the factors. As commented above, the use of new factor analyses may clarify some of these issues.

Third, we note that Bosselut et al. (2010) proposed the development of the RAS in a general game context and they compared it with attack and defense proposed in the original version of Beauchamp et al. (2002). The results show that both questionnaires are valid and reliable. If we take into account that it is difficult to specify roles in the game phases and that the players’ roles are not only attack and defense, but instead, up to four phases can be identified (attack, defense, offensive transition, defensive transition; Hewitt, Greenham, & Norton, 2016), a general evaluation of ambiguity may be sufficient. Moreover, a high number of items are employed, and many of them have a similar definition and some of them are meant to be reverse coded due to their negative phrasing (e.g., “I am unclear about the breadth of my responsibilities”). Regarding the validation of Bosselut et al. (2010), they decided to eliminate the negative items from their results because they did not contribute to the measure of role ambiguity. Previously, other authors defended the idea of removing these items (Eys, Carron, Bray, & Brawley, 2007) and even reducing the number of items of the scales to favor the measurement of several variables at the same time or a variable at different times (Buton, Fontayne, Heuzé, Bosselut, & Raimbault, 2007; Leo, González-Ponce, Sánchez-Oliva, Pulido, & García-Calvo, 2015). Therefore, in this study, the adapted RAS only has general positively stated items with regard to team play, and a reduced number of items, based on expert judgment, as we had no prior data of previous studies about the factor loadings of each item on each factor.

Fourth, as with other instruments related to group processes (Group Environment Questionnaire; Carron, Widmeyer, & Brawley, 1985), which underwent an exhaustive validity process concerning the type of sport, participants’ age, gender, sport level, culture, or language, this instrument was originally validated with a very concrete group of participants (e.g., male secondary school rugby players). As noted by Beauchamp et al. (2002), new studies that examine in greater depth the concept of role ambiguity in different contexts are recommended. Hence, this study offers a validation in a different sport modality, such as soccer, with adult participants of both genders, at another level of competition, such as high performance, in a culture or language, such as Spanish.

In view of the limitations to measure role ambiguity in the sporting context, the main objective of this study is to examine the structure of the Role Ambiguity Framework and to adapt and validate the RAS in the Spanish sport context with professional players of both genders. For this purpose, factor structure, internal consistency values, discriminant validity, concurrent validity (with cohesion; Bosselut, McLaren et al., 2012; Eys & Carron, 2001; Mellalieu & Juniper, 2006), and gender invariance were analyzed. In addition, this research extends the knowledge about the psychometric characteristics of the RAS, because the use of different factor analyses (CFA, Hier-CFA, bifactor-CFA, ESEM, Hierarchical-ESEM [H-ESEM], and bifactor-ESEM) recently proposed (Marsh et al., 2014; Morin et al., 2016; Myers et al., 2014) allow us to validate this instrument with a measurement model adjusted to reality.

Method

Participants

The study sample was made up of male and female soccer players who belonged to 31 professional teams. The participants were 375 male players and 229 female players, aged between 15 and 38 years (M = 24.34, SD = 4.03) who played in national Spanish competitions, specifically the Women’s 1st Division and the Men’s 2nd B Division. We used intentional sampling to select the participants: 13 female teams and 18 male teams agreed to participate.

Instruments

Role ambiguity

To assess the players’ perceptions of role ambiguity, we used a scale adapted from the RAS developed by Beauchamp et al. (2002), which measures various dimensions: scope of responsibilities, behaviors in fulfilling role responsibilities, evaluation of role performance, and consequences of not fulfilling role responsibilities. To prepare the instrument, according to the indications about the content validity index (Lynn, 1986), a group of three experts on the topic performed the translation, adaptation, and reduction of the items for each factor. First, all items were translated and adapted. An example of role ambiguity includes “I am clear about the different responsibilities that make up my role”; in Spanish “Yo tengo claras las diferentes responsabilidades que forman parte de mi rol.” Second, the group of experts selected the items that best represented role ambiguity. In the first step, the negative items (Eys et al., 2007) were eliminated taking into account the values obtained in the validation of Bosselut et al. (2010). In the second step, each of the experts chose those three items that best identified the dimension represented. This decision was taken (a) because some of the items of the original instrument are very similar; for example, “I understand the extent of my responsibilities” and “I understand the scope of my responsibilities”; or “I understand the criteria by which my role responsibilities are evaluated,” and “The criteria by which my role is evaluated are clear to me”) and (b) to adapt the scale to the professional sport context (Buton et al., 2007; Leo, González-Ponce, Sánchez-Oliva, et al., 2015). The experts considered that the repeated items did not provide relevant information to the scale, and their removal did not affect the objective of the study. In addition, data on previous studies were not available to enable us to select items with higher factor loads. Finally, each of the expert’s contributions was assessed and agreed upon, and the final content for the questionnaire was selected (see Table 1). We obtained 12 items grouped in four factors, like the original scale. Players responded to all items on a 9-point scale ranging from strongly disagree (1) to strongly agree (9). Thus, higher ratings of agreement indicated greater role clarity and, hence, less role ambiguity.

Table 1.

Role Ambiguity Scale in Spanish (English Version).

Ambiguity Related to Scope of Responsibilities

1. Entiendo el alcance de mis responsabilidades [I understand the scope of my responsibilities]

5. Conozco todas mis responsabilidades [I understand all of my responsibilities]

9. Tengo claras las diferentes responsabilidades que componen mi rol dentro de este equipo [I am clear about the different responsibilities that make up my role]

Role Behavior Ambiguity

2. Conozco como ajustar mi comportamiento para cumplir con mi rol [I understand what adjustments to my behavior need to be made to carry out my role]

6. Comprendo lo que debo hacer para llevar a cabo mis funciones [I understand the behaviors I must perform to carry out my role]

10. Están claros los comportamientos que tengo que realizar para cumplir con mi papel en este equipo [It is clear what behaviors I should perform to fulfill my role]

Role Evaluation Ambiguity

3. Entiendo como se evalúa mi rol [I understand how my role is evaluated]

7. Tengo claro como son evaluadas las responsabilidades que conlleva mi rol [It is clear to me how my role responsibilities are evaluated]

11. Los criterios por los cuales se evalúan mis funciones durante el juego los tengo claros [The criteria by which my role is evaluated are clear to me]

Role Consequences Ambiguity

4. Tengo claro qué ocurriría si no cumpliera con las responsabilidades de mi rol [It is clear to me what happens if I fail to carry out my role responsibilities]

8. Entiendo las consecuencias de fracasar en mis funciones [I understand the consequences of failing to carry out my role responsibilities]

12. Sé lo que pasaría si no llevo a cabo las responsabilidades de mi rol [I know what will happen if I don’t perform my role responsibilities]

Cohesion

To assess cohesion, we used a Spanish adaptation of the Group Environment Questionnaire (Carron et al., 1985) developed by Leo, González-Ponce, Sánchez-Oliva, et al. (2015). This instrument has 12 positive items, grouped into four factors: Group Integration–Task (GI-T, three items, for example, “The team members unite their efforts to achieve the goals during training and matches”), Group Integration–Social (GI-S, three items, for example, “Our team members like to get together in situations other than training and matches”), Individual Attraction to Group–Task (ATG-T, three items, for example, “I can do my best on this team”), and Individual Attraction to Group–Social (ATG-S, three items, for example, “The teammates make up one of the most important social groups I belong to”). We used the cohesion global factor to assess concurrent validity. Agreement with items is rated on a 9-point Likert-type scale, ranging from strongly disagree (1) to strongly agree (9).

Procedure

The study received ethical approval from the university. All participants were treated according to American Psychological Association ethics guidelines regarding consent, confidentiality, and anonymity of responses. The study used a cross-sectional design. A single measurement was performed in the first third of the season to ensure that the teams had competed in various official matches and there was enough time to form the group to be able to assess role ambiguity.

We developed a protocol to ensure that data collection would be similar for all the participants involved in the investigation. First, club officials (i.e., coaches, psychological services) were contacted to request their supervision of the investigation and their consent. We also informed the athletes that their participation was voluntary and their responses would be treated confidentially. We provided a letter to the parents of minor athletes, informing them of the goals and requesting their consent for their children to participate in the study. Assessments were conducted in the changing room without the presence of the coach, in an appropriate climate that allowed the players to concentrate without any distractions. The questionnaires required approximately 10 min to complete. The main investigator was present while the subjects completed the questionnaires.

Data Analysis

The psychometric properties of the Role Ambiguity Scale in Spanish (RAS-S scale) were analyzed with Mplus 7.3 (Muthén & Muthén, 2015) and the robust maximum likelihood estimator (MLR). MLR estimator provides standard errors and fit indices that are robust to the Likert-type nature of the items and non-normality.

With the aim of testing the best model, we followed the recommendations of Morin et al. (2016). This process involved testing eight nested models: (a) one global first-order CFA factor (First-Order-CFA); (b) four correlated first-order CFA factors (C4-First-Order-CFA); (c) three correlated first-order CFA factors (C3-First-Order-CFA), where items were restricted to load on their specific factor and the three and four first-order factors were allowed to correlate; (d) hierarchical CFA (H-CFA), where, the three first-order factors were allowed to load on the second-order factor; (e) bifactor-CFA (B-CFA), with all items specified as loading on the general factor and on the specific factors, with no cross-loadings allowed across the other two factors. All factors were specified as orthogonal, with the aim of ensuring that the interpretability of the solution was in line with bifactor assumptions (Chen, West, & Sousa, 2006); (f) first-order ESEM, with oblique target rotation (Asparouhov & Muthén, 2009), where all the main loadings were freely estimated, while all cross-loadings were specified to be close to zero (Morin et al., 2016), and the three first-order factors were allowed to correlate; (g) H-ESEM, estimated using ESEM-within-CFA (Morin et al., 2016), through the exact values of the nonstandardized loadings and cross-loadings estimated from the ESEM model as initial values; and (h) bifactor-ESEM (B-ESEM), estimated following the ESEM approach, and using orthogonal target rotation.

To assess the models’ fit, as given moderately large samples, the chi-square can be oversensitive to even minor model misspecifications, the following common goodness-of-fit indices were considered: the comparative fit index (CFI), the Tucker–Lewis Index (TLI), the standardized root mean residual (SRMR), and the root mean square error of approximation (RMSEA). Values greater than .90 and .95 for the CFI and TLI are considered adequate and excellent, respectively, whereas values smaller than .08 or .06 for the SRMR and RMSEA support acceptable and excellent model fit, respectively. We also used three indices for the information criteria: the Akaike information criterion (AIC), the Bayesian information criterion (BIC), and the sample size adjusted BIC (ABIC).

Finally, we analyzed the invariance of the factor structure as a function of participants’ gender. By means of this technique, we could confirm that the designed instrument behaves similarly for each group. That is, this analysis allows us to confirm that the psychometric properties of the instrument are invariant by the gender. Thus, the possible differences between the different models can be tested.

Results

Factorial Structure and Reliability

We tested the first- and second-order structures proposed by Beauchamp et al. (2002), used in their original instrument, and the structures proposed by Bosselut et al. (2010) to confirm the best factor structure. First, we performed three CFAs: (a) First-Order-CFA, (b) C4-First-Order-CFA, and (c) C3-First-Order-CFA, to test which of these three structures had better fit indexes. Given that C3-First-Order-CFA was the one with the best fit, then we developed an H-CFA model from this structure. Furthermore, we included different structures proposed by Morin et al. (2016): B-CFA, ESEM, 7 H-ESEM, and B-ESEM, based on structure C3-First-Order-CFA, with the aim of better knowing the psychometric properties of the RAS (see Figure 1).

Figure 1.

Graphic representation of the alternative models considered in this study.

The goodness-of-fit statistics and information criteria of the estimated models for the RAS-S are displayed in Table 2. The First-Order-CFA, C4-First-Order-CFA, and C3-First-Order-CFA obtained an adequate fit to the data. However, C3-First-Order-CFA showed better fit to the data (CFI = .955, TLI = .941, RMSEA = .060, SRMR = .029) than C4-First-Order-CFA (CFI = .955, TLI = .938, RMSEA = .062, SRMR = .029) and First-Order-CFA (CFI = .929, TLI = .908, RMSEA = .076, SRMR = .040). Furthermore, H-CFA, with first-order factors grouped into three factors, presented the same results as the C3-First-Order-CFA.

Table 2.

Goodness-of-Fit Statistics and Information Criteria of the Estimated Models.

Model	χ²	p	df	CFI	TLI	SRMR	RMSEA [90% CI]	AIC	BIC	ABIC
First-Order-CFA	217.619	.000	51	.929	.908	.040	.076 [.065, .086]	20,710.440	20,879.988	20,756.180
C4-First-Order-CFA	152.734	.000	48	.955	.938	.029	.062 [.051, .073]	20,563.497	20,746.087	20,612.756
C3-First-Order-CFA	157.412	.000	51	.955	.941	.029	.060 [.050, .071]	20,564.702	20,734.250	20,610.442
H-CFA	157.412	.000	51	.955	.941	.029	.060 [.050, .071]	20,564.702	20,734.250	20,610.442
B-CFA	159.845	.000	46	.951	.930	.059	.066 [.055, .077]	20,560.699	20,751.984	20,612.303
ESEM	171.413	.000	33	.941	.882	.024	.086 [.073, .099]	20,524.392	20,772.193	20,591.243
H-ESEM	259.133	.000	33	.962	.928	.024	.106 [.094, .118]	20,520.576	20,759.683	20,585.082
B-ESEM	100.439	.000	24	.967	.910	.013	.075 [.060, .090]	20,406.890	20,693.818	20,484.297
Configural invariance	138.323	.000	49	.966	.908	.015	.080 [.064, .096]	20,281.143	20,850.422	20,434.555
Weak invariance	177.816	.000	80	.962	.938	.033	.065 [.053, .078]	20,297.443	20,732.007	20,414.552
Strong invariance	197.357	.000	92	.960	.942	.039	.063 [.051, .076]	20,288.474	20,670.890	20,391.529
Strict invariance	219.508	.000	100	.954	.940	.054	.065 [.053, .076]	20,331.279	20,678.929	20,424.965
Variance-covariance invariance	227.444	.000	110	.955	.946	.085	.061 [.050, .072]	20,351.837	20,656.031	20,433.813
Latent means invariance	227.444	.000	110	.955	.946	.085	.061 [.050, .072]	20,351.837	20,656.031	20,433.813

Note. CFI = comparative fit index; TLI = Tucker–Lewis Index; SRMR = standardized root mean residual; RMSEA = root mean square error of approximation; CI = confidence interval; AIC = Akaike information criterion; BIC = Bayesian information criterion; ABIC = adjusted BIC; CFA = confirmatory factor analysis; First-Order-CFA = one global first-order CFA factor; C4-First-Order-CFA = four correlated first-order CFA factors; C3-First-Order-CFA = three correlated first-order CFA factors; H-CFA = hierarchical CFA; B-CFA = bifactor-CFA; ESEM = first-order exploratory structural equation modeling; H-ESEM = hierarchical ESEM; B-ESEM = bifactor-ESEM.

Next, the models B-CFA, ESEM, H-ESEM, and B-ESEM were made with three factors, as this was the best solution found. Thus, the B-CFA solution obtained adequate values in different fit indices (CFI = .951, TLI = .930, RMSEA = .066, SRMR = .059). However, the ESEM (CFI = .941, TLI = .882, RMSEA = .086, SRMR = .024) and H-ESEM (CFI = .964, TLI = .923, RMSEA = .110, SRMR = .024) showed poorer values in some fit indexes. Finally, the B-ESEM obtained the best fit to the data (CFI = .967, TLI = .910, RMSEA = .075, SRMR = .013; see Table 2).

Tables 3 and 4 present the standardized factor loadings and the variances of First-Order-CFA, C4-First-Order-CFA, C3-First-Order-CFA, H-CFA, B-CFA, ESEM, H-ESEM, and B-ESEM models. In the First-Order-CFA, C4-First-Order-CFA, C3-First-Order-CFA, and H-CFA solutions (Table 3), all the constructs were well-defined by high and significant (p < .001) factor loadings (First-Order-CFA = .676-.884; C4-First-Order-CFA = .696-.890; C3-First-Order-CFA = .703-.890; H-CFA = .706-.895). In the H-CFA model, factor loadings between specific factors and the global factor ranged between .902 and .931. The B-CFA parameter estimates revealed that the three factors were well-defined (λ_{Scope-Behavior} = .556-.640; λ_Evaluation = .547-.921; λ_Consequences = .477-.807).

Table 3.

Standardized Factor Loadings (λ) and Uniquenesses (δ) for the Alternative Measurement Models.

Items	One-factor CFA		Four-factors CFA		Three-factors CFA		H-CFA		B-CFA
	S-Factor	δ	S-Factor	δ	S-Factor	δ	S-Factor	δ	G-Factor	S-Factor	δ
	λ(p)	δ	λ(p)	δ	λ(p)	δ	λ(p)	δ	λ(p)	λ(p)	δ
Scope
1	.695	.517	.696	.515	.703	.506	.706	.506	.658	.658	.134
5	.835	.303	.844	.287	.848	.281	.822	.281	.833	.120	.241
9	.871	.242	.864	.254	.872	.240	.884	.240	.873	.041	.236
Beha
2	.830	.311	.836	.300	.838	.298	.834	.298	.815	.241	.278
6	.850	.277	.856	.268	.861	.258	.837	.258	.862	−.011	.291
10	.884	.218	.880	.225	.888	.212	.895	.212	.900	−.024	.190
Eva
3	.676	.543	.770	.407	.770	.407	.773	.407	.638	.508	.336
7	.776	.398	.843	.289	.843	.289	.840	.289	.756	.337	.314
11	.788	.378	.882	.222	.882	.223	.883	.223	.774	.417	.227
Conse
4	.757	.427	.838	.298	.838	.298	.839	.298	.672	.554	.241
8	.734	.461	.831	.310	.831	.310	.831	.310	.700	.439	.318
12	.763	.417	.890	.208	.890	.209	.888	.209	.715	.547	.189
Specific factor loadings
		Sco-Beh	Eva	Conse
Hierarchical CFA		.931	.926	.902

Note. Nonsignificant parameters (p ≥ .05) are in italics. CFA = confirmatory factor analysis; H-CFA = hierarchical CFA; B-CFA = bifactor-CFA; G-Factor = Global Factor; Scope = Ambiguity Related to Scope of Responsibilities; Beha = Role Behavior Ambiguity; Eva = Role Evaluation Ambiguity; Conse = Role Consequences Ambiguity; Sco-Beh = Ambiguity Related to Scope of Responsibilities and Role Behavior Ambiguity; S-Factor = Single-Factor.

Table 4.

Standardized Factor Loadings (λ) and Uniquenesses (δ) for the Alternative Measurement Models.

Items	ESEM				H-ESEM				B-ESEM
	Sco-Beh	Eva	Conse	δ	Sco-Beh	Eva	Conse	δ	G-Factor	Sco-Beh	Eva	Conse	δ
	λ(p)	λ(p)	λ(p)	δ	λ(p)	λ(p)	λ(p)	δ	λ(p)	λ(p)	λ(p)	λ(p)	δ
Scope
1	.658			.509	.590			.508	.664	.521			.286
5	1.056			.212	.914			.212	.858	.117			.222
9	.643	.235		.257	.570	.325		.257	.853	.061			.262
Beha
2	.758			.304	.669			.304	.813	.303			.247
6	.956			.218	.848			.219	.891	− .042			.172
10	.657			.230	.586	.312		.230	.880	− .010			.223
Eva
3		.828		.372		.845		.372	.650		.492		.303
7		.647		.312		.681		.312	.771		.316		.291
11		.746		.231		.771		.231	.785		.380		.226
Conse
4			.629	.336	.336		.519	.335	.743			.333	.329
8			.694	.346	.246		.566	.345	.722			.336	.342
12			1.107	.094			.893	.096	.753			.569	.107
Specific factor loadings
			Sco-Beh		Eva	Conse
Hierarchical ESEM			.801		.935	.786

Note. Nonsignificant parameters (p ≥ .05) are in italics. Target factor loadings are in bold. For clarity, cross-loadings below 0.20 are not displayed in the table (cf. Myers, Martin, Ntoumanis, Celimli, & Bartholomew, 2014). ESEM = exploratory structural equation modeling; H-ESEM = hierarchical ESEM; B-ESEM = bifactor-ESEM; Scope = Ambiguity Related to Scope of Responsibilities; Beha = Role Behavior Ambiguity; Sco-Beh = Ambiguity Related to Scope of Responsibilities and Role Behavior Ambiguity; Eva = Role Evaluation Ambiguity; Conse = Role Consequences Ambiguity; G-Factor = Global Factor.

The ESEM revealed an inadmissible solution as Item 5 had a factor loading above one point (1.056) in its factor (Table 4). The H-ESEM revealed positive and significant target factor loadings: Scope-Behavior (.508-.591), Evaluation (.560-.909) and Consequences (.420-.796). Furthermore, in the H-ESEM model, acceptable and positives factor loadings between specific factors and the global factor ranged between .519 and .914. The B-ESEM solution showed (a) all items had high and significant factor loadings on the global factor (.172-.342) except for Item 12, which was nonsignificant (.107); and (b) well-defined Evaluation (.316 to .492) and Consequences (.333 to .569). However, Scope-Behavior only presented two significant items (Item 1 = .521, Item 2 = .303). The other factors had greater factor loadings on the global factor.

Discriminant Validity, Concurrent Validity, and Internal Consistency

In terms of discriminant validity, the CFA solution showed higher factor correlations (r_{ScopeBehavior-Evaluation} = .862; r_{ScopeBehavior-Consequences} = .840; r_{Evaluation-Consequences} = .836) than the ESEM solution (r_{ScopeBehavior-Evaluation} = .808; r_{ScopeBehavior-Consequences} = .813; r_{Evaluation-Consequences} = .784). Furthermore, as indicated above, we introduced cohesion to confirm concurrent validity. Accordingly, all the factors of role ambiguity showed positive high correlations with cohesion (r_{ScopeBehavior-Cohesion} = .41; r_{Evaluation-Cohesion} = .42; r_{Consequences-Cohesion} = .37; r_{GlobalFactor-Cohesion} = .44). Finally, the values of internal consistency were adequate, such that Cronbach’s alpha coefficients were .93 for Scope-Behavior, .87 for Evaluation, .88 for Consequences, and .95 for the global factor.

Analysis of Factor Invariance

Measurement invariance across gender of the best model was tested through six different models: (a) configural invariance (loading pattern is similar in all groups but the magnitude of all parameters (e.g., loadings, intercepts, variances may vary); (b) weak invariance (factor loadings/cross-loadings are constrained to be equal as fit to the data and the fit of this model is compared with the baseline model); (c) strong measurement (factor loadings and item intercepts are constrained to be equal is fit to the data and compared against the weak measurement invariance model); (d) strict invariance (invariance of the factor loadings/cross-loadings, intercepts, and uniquenesses are constrained to be equal is fit to the data and compared with the strong measurement invariance model); (e) latent variance-covariance invariance (factor loadings/cross-loadings, intercepts, uniquenesses, and latent variances-covariances are constrained to be equal is fit to the data and compared against the strict measurement invariance model); and (f) latent means invariance (the factor loadings/cross-loadings, intercepts, uniquenesses, latent variances-covariances, and latent means are constrained to be equal is fit to the data and compared against the latent variance-covariance invariance measurement invariance model). Nested models were compared via consideration of ΔCFI, ΔTLI, and ΔRMSEA. Accordingly, a decrease in the CFI and TLI equal to or higher than .01 or an increase in the RMSEA equal to or higher than .015 indicates lack of invariance across groups (Chen, 2007; Cheung & Rensvold, 2002).

Table 2 shows the fit indices of the six compared models in the analysis of invariance by gender. First, the Configural Invariance was inadmissible because the residual covariance matrix was positive on account of a large and significant negative residual associated with Item 12 (“I know what will happen if I don’t perform my role responsibilities”). To rectify the negative residual, the residual variance for this item was constrained to > .0001 (Chen, Bollen, Paxton, Curran, & Kirby, 2001). Second, in the other models, adequate fit indices were obtained. Finally, the values found in CFI, TLI, and RMSEA in the different invariance models indicate that the RAS-S does not vary as a function of gender.

Discussion

The goal of the present study was to examine the structure of the Role Ambiguity Framework and to analyze the psychometric properties of the RAS in Spanish in a sample of professional players of both genders. The data analyzed and the results lead to some inferences about the structure of role ambiguity, the psychometric properties of this version of the RAS-S and about consistency of its validity and reliability. Accordingly, the results indicate that the scale is composed with one second-order global factor and three first-order factors. Furthermore, the scale has appropriate internal consistency, and discriminant and concurrent validity, and is also invariant as a function of the athletes’ gender. Therefore, the Spanish version of the RAS is a valid and reliable scale for the analysis of role ambiguity.

First, we analyzed the validity of the instrument by means of different factorial structures, integrating the classic (CFA) and emergent models (ESEM, B-CFA, and B-ESEM) with first- and second-order structures. We tested three first-order CFA models, with a single factor (this variable was used in various studies with a general factor; Beauchamp & Bray, 2001; Beauchamp, Bray, Fielding, et al., 2005; Leo, González-Ponce, Sánchez-Miguel, et al., 2015), with four correlated factors, as validated by the authors of the instrument (Beauchamp et al., 2002), and three correlated factors (as validated by Bosselut et al., 2010, due to the high correlations between two of its factors: scope of responsibilities and behavior). From the results of these models, we observed that the three correlated factor model presented the best fit indices, and moreover, the factor loadings on each of its factors were optimal, and the correlation among the factors led us to choose this model. Next, we conducted an H-CFA, with three first-order factors (in view of the prior decision) and one global second-order factor (as in the conceptual model of the construct), obtaining the same values as in the C3-First-Order-CFA.

Subsequently, different emerging models proposed by various authors were developed: B-CFA, ESEM, H-ESEM, and B-ESEM (Marsh et al., 2014; Morin et al., 2016; Myers et al., 2014), with the aim of analyzing the psychometric multidimensionality of the RAS. Taking all fit indices as reference, B-ESEM model obtained adequate fit indices (CFI, TLI, SRMR, and RMSEA) and the lowest score in the comparative indices (AIC, BIC, ABIC) and, therefore, it presented the best fit. These results are not consistent with the results obtained by Beauchamp et al. (2002) because the original structure from CFA consisted of four correlated factors. However, this structure is similar to the hierarchical model of the theory with first-order factors and a global second-order factor. The main difference is the fact that, in our model, there are only three first-order factors, as previously found by Bosselut et al. (2010) in the French version, and in the conceptual model, four factors are defined as forming part of role ambiguity. This may mean that there is conceptual overlap in two of the dimensions of role ambiguity, in this case, scope of responsibilities and behavior.

In our context, these results confirm that the B-ESEM provided a multidimensional model of the measurement scale more adjusted to reality, indicating that the RAS-S has a bifactor pattern with items loading relatively high on both the general and the specific factors, Scope-Behavior, Evaluation, and Consequences. Specifically, the three factors displayed relatively high loadings on the general factor and weak loadings on their specific factors. It should be noted that some of the items are nonsignificant (Items 5, 6, 9, and 10) on their factor, because they loaded higher on the global factor. Our results also revealed that some role ambiguity items reflected players’ role ambiguity more than their specific perception of scope of responsibilities and behavior, consequences, or evaluation. Yet, these observations also made sense (i.e., “I understand and have a clear idea of all of my responsibilities,” Items 5 and 9; or “I understand and have clear what behaviors I should perform,” Items 6 and 10), and might provide a stronger measure of global role ambiguity for this sample of professional players for whom role ambiguity is identified mainly through the scope of responsibilities and behaviors, which Bosselut et al. (2010) called the task factor. In this regard, it is perhaps not surprising for our results to suggest that scope of responsibilities and behavior ratings play such a central role in the definition of the global role ambiguity factor. In a professional sport context, having a clear idea of the tasks to be performed is one of the most important things. A lack of scope of responsibilities and behavior can create problems at all levels and hinder the perception of consequences and evaluation of role ambiguity. Clearly, the present results need to be replicated using more diversified samples of sports and levels of competition to test this interpretation.

Therefore, the role ambiguity construct seems to be defined by three different factors with a common core, the general factor. Previous research did not analyze multidimensionality simultaneously with the factor weights on global and specific factors, using instead only first- and second-order factors to test the multidimensionality of the RAS (e.g., Beauchamp et al., 2002; Bosselut et al., 2010). Thus, the B-ESEM approach extends our knowledge about the factor structure of the RAS. However, future research should examine the factor loadings on the three factors to determine whether they need to be revised.

As mentioned previously, in spite of the fact that the global factor seems very important, if the correlations between the CFA factors and the ESEM factors are taken into account, the correlation indices among factors were appropriate. In spite of the fact that the general factor has more weight than the three dimensions, it can be stated that the three factors are independent. So, this instrument can be considered to have discriminant validity. That is, all three dimensions are strongly represented by the global factor, and, in turn, each dimension clearly reflects a different dimension from the rest and contributes multidimensionality to the concept of role ambiguity. If the RAS factors are differentially related to other variables, it would be interesting for future research to provide a clearer view of the similarities and differences among the RAS factors in the sport context.

In this sense, regarding the analysis of concurrent validity, we noted that the three factors of role ambiguity and the global factor presented a significant correlation with cohesion, as previously established in different studies (Bosselut, McLaren, et al. 2012; Eys & Carron, 2001; Mellalieu & Juniper, 2006). Thus, we can state that the instrument presents appropriate concurrent validity.

In the same vein, the values of internal consistency of each one of the factors and the global factor were high (α > .70; Nunnally & Bernstein, 1994). These results are coherent with the original scale (Beauchamp et al., 2002) and in previous studies where this scale was used (Beauchamp, Bray, Fielding, et al., 2005; Bosselut et al., 2010; Bosselut, Heuzé, et al., 2012; Cunningham & Eys, 2007).

Finally, the adaptation of the RAS-S was invariant by the gender of our participants. We note that the values of ΔCFI, ΔRMSEA and ΔSRMR were lower than .01. Therefore, in accordance with Cheung and Rensvold (2002), the instrument is invariant and can be used correctly in either gender (Byrne, 2001). This expands our knowledge about the RAS, as this instrument has not been used in both genders.

Conclusions, Limitations, and Future Research

Ultimately, the results indicate that the structure of the role ambiguity is composed with one second-order global factor and three first-order factors: scope-behavior, evaluation, and consequences. Furthermore, the version of the RAS adapted to Spanish presents appropriate fit index values in the expected model. It also presents optimal values of internal consistency, appropriate discriminant and concurrent validity, and is invariant by gender. Therefore, the adaptation of the RAS to Spanish is a valid and reliable instrument to measure role ambiguity in male and female professional players.

From a practical perspective, the results imply that coaches and sports psychologists could use this instrument to measure the global factor and each one of the dimensions of role ambiguity in high-performance athletes because the players can clearly differentiate the subscales (Hu & Bentler, 1999). In spite of this, more tests on the scale are needed because the number of items was reduced to obtain an instrument adapted to the characteristics of the context and the sample. Furthermore, it would be necessary to examine the scale with additional athletes because these psychometric tests were carried out with soccer players.

Another limitation of the validation is that the participants were professionals, so it would be interesting to calculate the invariance with amateur groups. Future research should examine the psychometric properties in different sports and with different populations. It would also be interesting to use this scale in other cultures and languages to ensure a standard measuring instrument for the professional sphere.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the European Social Fund and Government of Extremadura (Counseling of Economy and Infrastructure) (PD12112).

Author Biographies

Francisco M. Leo is a full professor in the Faculty of Teacher Training, University of Extremadura, Spain. He works in the Didactic and Behavioral Analysis of Sport research group; his main areas of research are examining group processes (role, cohesion, collective efficacy) and coach behaviors in professional sport.

Inmaculada González-Ponce is a predoctoral researcher. She works in the Didactic and Behavioral Analysis of Sport research group in the Faculty of Sport Science, University of Extremadura, Spain. Her research interests are coaching competency and justice perceived of coaches and their relationship with group processes in professional sport.

David Sánchez-Oliva is a postdoctoral researcher. He works in the Self-Regulation in Physical Activity, Nutrition and Obesity research group in the Faculty of Human Kinetics, University of Lisbon, Portugal. His research revolves around the field of motivational predictors of physical activity, applying self-determination-theory to develop interventions in physical education and sport.

Juan J. Pulido is a predoctoral researcher. He works in the Didactic and Behavioral Analysis of Sport research group in the Faculty of Sport Science, University of Extremadura, Spain. His main research interests are programs for coaches in youth sport and examining motivation in sport teams.

Tomás García-Calvo is full professor of team management in high-performance sport in the Faculty of Sport Science, University of Extremadura, Spain. He directs the Didactic and Behavioral Analysis of Sport research group. He specializes in group processes in professional sport, and motivation at different contexts.

References

Asparouhov

Muthén

B. O.

(2009). Exploratory structural equation modeling. Structural Equation Modeling, 16, 397-438. doi:10.1080/10705510903008204

Beauchamp

M. R.

Bray

S. R.

(2001). Role ambiguity and role conflict within interdependent teams. Small Group Research, 32, 133-157. doi:10.1177/104649640103200202

Beauchamp

M. R.

Bray

S. R.

Eys

M. A.

Carron

A. V.

(2002). Role ambiguity, role efficacy, and role performance: Multidimensional and mediational relationships within interdependent sport teams. Group Dynamics: Theory, Research, and Practice, 6, 229-242. doi:10.1037//1089-2699.6.3.229

Beauchamp

M. R.

Bray

S. R.

Eys

M. A.

Carron

A. V.

(2003). The effect of role ambiguity on competitive state anxiety. Journal of Sport & Exercise Psychology, 25, 77-92. doi:10.1123/jsep.25.1.77

Beauchamp

M. R.

Bray

S. R.

Eys

M. A.

Carron

A. V.

(2005). Leadership behaviors and multidimensional role ambiguity perceptions in team sports. Small Group Research, 36, 5-20. doi:10.1177/1046496404266684

Beauchamp

M. R.

Bray

S. R.

Fielding

Eys

M. A.

(2005). A multilevel investigation of the relationship between role ambiguity and role efficacy in sport. Psychology of Sport and Exercise, 6, 289-302. doi:10.1016/j.psychsport.2004.03.002

Bosselut

Heuzé

J. P.

Eys

M. A.

Fontayne

Sarrazin

(2012). Athletes’ perceptions of role ambiguity and coaching competency in sport teams: A multilevel analysis. Journal of Sport & Exercise Psychology, 34, 354-364. doi:10.1123/jsep.34.3.345

Bosselut

Heuzé

J. P.

Sarrazin

(2010). Structure of the role ambiguity framework and validity in the French culture. Psychology of Sport and Exercise, 11, 471-478. doi:10.1016/j.psychsport.2010.06.001

Bosselut

McLaren

C. D.

Eys

M. A.

Heuzé

J. P.

(2012). Reciprocity of the relationship between role ambiguity and group cohesion in youth interdependent sport. Psychology of Sport and Exercise, 13, 341-348. doi:10.1016/j.psychsport.2011.09.002

10.

Buton

Fontayne

Heuzé

J. P.

Bosselut

Raimbault

(2007). The QAG-a: An analog version of the Questionnaire sur l’Ambiance du Groupe for measuring the dynamic nature of group cohesion. Small Group Research, 38, 235-264. doi:10.1177/1046496407300476

11.

Byrne

B. M.

(2001). Structural equation modeling with AMOS, EQS, and LISREL: Comparative approaches to testing for the factorial validity of a measuring instrument. International Journal of Testing, 1, 55-86. doi:10.1207/S15327574IJT0101_4

12.

Carron

A. V.

Widmeyer

W. N.

Brawley

L. R.

(1985). The development of an instrument to assess cohesion in sport teams: The Group Environment Questionnaire. Journal of Sport Psychology, 7, 244-266. doi:10.1123/jsp.7.3.244

13.

Chen

F. F.

(2007). Sensitivity of goodness of fit indices to lack of measurement invariance. Structural Equation Modeling, 14, 464-504. doi:10.1080/10705510701301834

14.

Chen

F. F.

Bollen

K. A.

Paxton

Curran

P. J.

Kirby

J. B.

(2001). Improper solutions in structural equation models: Causes, consequences, and strategies. Sociological Methods & Research, 29, 468-508. doi:10.1177/0049124101029004003

15.

Chen

F. F.

West

S. G.

Sousa

K. H.

(2006). A comparison of bifactor and second-order models of quality of life. Multivariate Behavioral Research, 41, 189-255. doi:10.1207/s15327906mbr4102_5

16.

Cheung

G. W.

Rensvold

R. B.

(2002). Evaluating goodness-of-fit indices for testing measurement equivalence. Structural Equation Modeling: A Multidisciplinary Journal, 9, 233-255. doi:10.1207/S15328007SEM0902_5

17.

Cunningham

I. J.

Eys

M. A.

(2007). Role ambiguity and intra-team communication in interdependent sport teams. Journal of Applied Social Psychology, 37, 2220-2237. doi:10.1111/j.1559-1816.2007.00256.x

18.

Eys

M. A.

Carron

A. V.

(2001). Role ambiguity, task cohesion, and task self-efficacy. Small Group Research, 32, 356-373. doi:10.1177/104649640103200305

19.

Eys

M. A.

Carron

A. V.

Bray

S. R.

Beauchamp

M. R.

(2003). Role ambiguity and athlete satisfaction. Journal of Sports Sciences, 21, 391-401. doi:10.1080/0264041031000071137

20.

Eys

M. A.

Carron

A. V.

Bray

S. R.

Beauchamp

M. R.

(2005). The relationship between role ambiguity and intention to return the following season. Journal of Applied Sport Psychology, 17, 255-261. doi:10.1080/10413200591010148

21.

Eys

M. A.

Carron

A. V.

Bray

S. R.

Brawley

L. R.

(2007). Item wording and internal consistency of a measure of cohesion: The Group Environment Questionnaire. Journal of Sport & Exercise Psychology, 29, 395-402. doi:10.1123/jsep.29.3.395

22.

Hewitt

Greenham

Norton

(2016). Game style in soccer: What is it and can we quantify it? International Journal of Performance Analysis in Sport, 16, 355-372.

23.

Høigaard

Fuglestad

Peters

D. M.

Cuyper

B. D.

Backer

M. D.

Boen

(2010). Role satisfaction mediates the relation between role ambiguity and social loafing among elite women handball players. Journal of Applied Sport Psychology, 22, 408-419. doi:10.1080/10413200.2010.495326

24.

Bentler

P. M.

(1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling: A Multidisciplinary Journal, 6, 1-55. doi:10.1080/10705519909540118

25.

Jackson

S. E.

Schuler

R. S.

(1985). A meta-analysis and conceptual critique of research on role ambiguity and role conflict in work settings. Organizational Behavior and Human Decision Processes, 36, 16-78. doi:10.1016/0749-5978(85)90020-2

26.

Kahn

R. L.

Wolfe

D. M.

Quinn

Snoek

J. D.

Rosenthal

R. A.

(1964). Organizational stress. New York, NY: John Wiley.

27.

Leo

F. M.

González-Ponce

Sánchez-Miguel

P. A.

Ivarsson

García-Calvo

(2015). Role ambiguity, role conflict, team conflict, cohesion and collective efficacy in sport teams: A multilevel analysis. Psychology of Sport and Exercise, 20, 60-66. doi:10.1016/j.psychsport.2015.04.009

28.

Leo

F. M.

González-Ponce

Sánchez-Oliva

Pulido

J. J.

García-Calvo

(2015). Adaptation and validation in Spanish of the Group Environment Questionnaire (GEQ) with professional football players. Psicothema, 27, 261-268. doi:10.7334/psicothema2014.247

29.

Lynn

M. R.

(1986). Determination and quantification of content validity. Nursing Research, 35, 382-386. doi:10.1097/00006199-198611000-00017

30.

Marsh

H. W.

Morin

A. J. S.

Parker

P. D.

Kaur

(2014). Exploratory structural equation modeling: An integration of the best features of exploratory and confirmatory factor analyses. Annual Review of Clinical Psychology, 10, 85-110. doi:10.1146/annurev-clinpsy-032813-153700

31.

Mellalieu

S. D.

Juniper

S. W.

(2006). A qualitative investigation into experiences of the role episode in soccer. Sport Psychologist, 20, 399-418. doi:10.1123/tsp.20.4.399

32.

Morin

A. J.

Arens

A. K.

Marsh

H. W.

(2016). A bifactor exploratory structural equation modeling framework for the identification of distinct sources of construct-relevant psychometric multidimensionality. Structural Equation Modeling: A Multidisciplinary Journal, 23, 116-139. doi:10.1080/10705511.2014.961800

33.

Muthén

L. K.

Muthén

B. O.

(2015). Mplus user’s guide (7th ed.). Los Angeles, CA: Author.

34.

Myers

N. D.

Martin

J. J.

Ntoumanis

Celimli

Bartholomew

K. J.

(2014). Exploratory bifactor analysis in sport, exercise, and performance psychology: A substantive-methodological synergy. Sport, Exercise, and Performance Psychology, 3, 258-272. doi:10.1037/spy0000015

35.

Nunnally

J. C.

Bernstein

I. H.

(1994). The assessment of reliability. In Nunnally

J. C.

Bernstein

I. H.

(Eds.), Psychometric theory (pp. 248-292). New York, NY: McGraw-Hill.