Servicios personalizados
Inglés (pdf)
Articulo en XML
Referencias del artículo
Traducción automática
Enviar articulo por email
Citado por SciELO 
Citado por Google 
Similares en
SciELO 
Similares en Google 
Permalink
Introduction
Currently, the construct achievement goal (AG) is defined as goals based on competence used to guide behaviour (Elliot, 1999). Over the last three decades, the AG theory has evolved from a dichotomic perspective (Ames, 1992), to the 3x2 framework (Elliot, Murayama, & Pekrum, 2011), via the trichotomic view (Elliot & Harackiewicz, 1996) and the 2x2 structure (Elliot & McGregor, 2001). First, mastery or task goals focused on competence standards based on the task or intrapersonal (self-referenced), while performance or ego referred to interpersonal competence standards or normative. The second step produced the trichotomic model, considering two types of valence: approach goals, focused on acquiring positive consequences, and avoidance goals, focused on avoiding negative consequences, which produced two types of performance goals (performance-approach and performance-avoidance). A few years later, Elliot & McGregor (2001) proposed a third change named 2x2, since mastery goals, like performance goals, were divided in approach and avoidance. As a result of the combination of definition (intrapersonal/normative reference) and valence (approach/avoidance), the 2x2 framework produced four types of AG: mastery-approach (intrapersonal and positive), performance-approach (normative and positive), mastery-avoidance (intrapersonal and negative) and performance-avoidance (normative and negative).
The fourth step is the current 3x3 AG model ((Elliot, Murayama, & Pekrum, 2011; Murayama, Elliot, & Friedman, 2012), which establishes three standards to define competence: absolute (task), intrapersonal (self), and normative (others). Mixing these three standards (definition) with the two types of valence, Elliot et al. (2011) determined the six goals of this framework: task-approach (Tap), oriented to achievement based on the task (i.e. “finish the task correctly”); task-avoidance (Tav), focused to avoid incompetence based the task (i.e. “avoid making the taks incorrectly”); self-approach (Sap), focused on achievement based on the self (i.e., “performing better than before”); self-avoidance (Sav), oriented to avoiding incompetence based on me (i.e., “avoid performing worse than before”); others-approach (Oap), oriented to competence based on others (i.e., “performing better than others”); and others-avoidance (Oav), focused on avoiding incompetence based on others (i.e., “avoid performing worse than others”).
Besides being a framework to study individual differences in students’ motivation, AG theory also helps assess class context influence on motivation and learning (Meece, Anderman, & Anderman, 2006). In educational contexts, class achievement framework (CAF) has been defined as the environmental emphasis relative to competence, which is observed via class practice and the specific messages the teachers send to their students (Ames, 1992). Considering that students’ perceptions on CAF (psychological environment) play a key role in their goal adoption, research on CAF has focused on students’ perceptions instead of the environment itself. It could be expected that the evolution of the concepts linked to environmental factors which lead individuals to interpret competence and adopt different goals could happen parallel to the personal AG construct. However, the evolution of the environmental framework has been limited. From the dichotomic perspective, research focused on two types of structures: mastery CAF, which plays the class emphasis on mastery, comprehension and personal improvement, and performance CAF, whose emphasis is on skill and competition in class (Ames & Archer, 1988). Teachers who promote mastery CAF provide students with stimulating and multifaceted learning opportunities, focusing on effort, support student autonomy, stimulate students to use self-referenced criteria and believe that errors are an integral part of the learning process. In contrast, teachers who use performance CAF tend to group students based on competence, appraise them when they are successful over others, give privileges to students with better grades and use assessment procedures which encourage students to compare with their class mates (Kaplan, Middleton, Urdan, & Midgley, 2002).
Midley et al. (2000) applied the AG trichotomic model to the class context, recognizing three CAF: mastery CAF, where the class environment is focused on academic work to finish the task and develop competence in relation to oneself (intrapersonal and task based competence); performance-approach CAF, where the class environment is oriented towards academic work to show competence in relation to others (normative competence); and performance-avoidance CAF, where the class environment is oriented on academic work to avoid showing incompetence in relation to others (normative incompetence). Probably, in the lower educational levels there is not much performance-avoidance CAF because teachers tend to focus on finishing tasks and learning. However, as students go up in secondary or college education performance-avoidance CAF is more evident (Karabenick, 2004; Murayama & Elliot, 2009). Peng and Cherng (2005) extended Elliot and McGregor’s (2001) 2x2 framework to the CAF and build a four-factor structure, more suitable to secondary and baccalaureate Taiwanese students: mastery-avoidance CAF refers to the class environment focused on academic work to avoid showing incompetence compared to oneself (intrapersonal and task-based incompetence). Unfortunately, task-avoidance CAF and performance-avoidance CAF have been scarcely researched and only a few studies have adopted the four-factor CAF (Lin & Cherng, 2006, 2007; Peng, Cherng, Chen, & Lin, 2013). To date, no studies have explored the CAF using the 3x2 AG framework.
Contextual factors have been research as one of the AG antecedents within the hierarchical model of achievement motivation (Elliot, 1999). It has been proposed that the CAF can influence students’ AG (Ames, 1992, Maehr & Midgley, 1996). Church, Elliot, & Gable (2001) y Michou, Mouratidis, Lens, & Vansteenkiste (2013) showed that mastery CAF can facilitate the adoption of mastery approach goals, performance-approach CAF promote performance-approach goals, and performance-avoidance CAF promote the adoption of performance-avoidance goals.
Murayama and Elliot (2009) proposed a framework to assess the connections between CAF, AG and achievement results which included three models (direct effect, indirect effect and interaction effect). The direct effect model considers that the CAF influences the more relevant achievement results, even more than the personal AG (Ames & Archer, 1988; Gutman, 2006; Karabenick, 2004; Lau & Nie, 2008; Nolen, 2003; Turner et al., 2002; Urdan, Midgley, & Anderman, 1998; Wolters, 2004). The indirect effect model hypothesizes that the CAF influences indirectly the more relevant achievement results via its impact on the adoption of personal AG. This model highlights the mediating role of the AG in the CAF-achievement outcomes connection, since CAF influences AG adoption and this influences achievement outcomes. Most studies have supported the indirect effect model (Bong, 2005; Greene, Miller, Crowson, Duke, & Akey, 2004; Miki & Yamauchi, 2005; Patrick, Ryan, & Kaplan, 2007; Yamauchi & Miki, 2003). Finally, the interaction effect model points out that CAF moderate the influence of AG on the relevant achievement results. It highlights the interactive role of personal AG and CAF. It is believed that the influence of AG adoption changes depending on the type of CAF promoted. However, this model has been scarcely researched (Lau & Nie, 2008; Linnenbrink, 2005; Murayama & Elliot, 2009). These three models are not incompatible, but they have been researched only from the trichotomic framework (Murayama & Elliot, 2009; Schwinger & Stiensmeier-Pelster, 2011) and the results have been inconclusive. This is why the present study has focused on the assessment of the three models from the 3x2 AG framework.
Mastery-approach CAF have been linked to more adaptive variables: intrinsic motivation, self-efficacy, persistence and effort, challenge seek, less disruptive behaviours, positive affect and enthusiasm, positive grades and a decrease in depressive symptoms over time. Performance-approach CAF have produced contradictory results: greater exam anxiety, disruptive conducts, limited effort in difficult tasks, cheating and depressive symptoms (Linnenbrink, 2005; Meece et al., 2006; Murayama & Elliot, 2009; Urdan & Midgley, 2003; Wolters, 2004). Finally, the effects of performance-avoidance and mastery avoidance CAF have not been assessed.
Several assessment instruments have been proposed to assess CAF in educational contexts. From a dichotomic perspective, Ames and Archer (1988) designed a questionnaire to assess students’ perceptions of two CAF dimensions: mastery and performance. They identified group of dimensions of the class climate (i.e., definition of success, reasons to be satisfied, reasons for the effort, evaluation criteria...) related to the adoption of each goal orientation. From the trichotomic perspective, the Patterns of Adaptive Learning Strategies (PALS; Midgley et al., 2000) acknowledged three environments based on students’ perceptions: mastery-oriented, performance-approach orientation and performance-avoidance orientation. In sport and physical education contexts Newton, Duda and Yin (2003) validated the Perceived Motivational Climate Sport Questionnaire (PMCSQ). Papaioannou, Tsigilis, Kosmidouy, & Milosis (2007) developed a tool from the trichotomic perspective, and recently, Méndez-Giménez, Cecchini, & Fernández-Río (in press) assessed construct validity of the 3x2 Motivational Climate Questionnaire in physical education.
Since there is no assessment instrument for academic contexts in line with the latest steps, 3x2 AG, which incorporate there definition criteria (task, self, other) and the two valences (approach-avoidance), it seems relevant to design and validate a tool to assess the six environmental constructs on CAF. Murayama et al. (2012) emphasized that the assessment of the influence of the mastery-avoidance CAF could be worthy, because teaching strategies that promote mastery-avoidance goals (i.e., mastery-avoidance CAF) are common (i.e.: “try to avoid making mistakes”). This mastery-avoidance CAF could have a substantial impact in the learning process. Similarly, the last theoretical step in 3x2 AG suggests that mastery structures should be divided based on the task and the self, which remains unexplored.
In the present study two goals were set: a) examine the structural validity of the 3x2 achievement class structure questionnaire (ACSQ 3x2), and b) assess the direct, indirect and interaction relations between 3x2 achievement class structures, 3x2 AG and self-determined motivation in secondary education and baccalaureate students. The first hypothesis was that the results will dissociate task-based CAF into task, self and others, both approach and avoidance. The second hypothesis was that CAF will not be direct and indirect predictor of self-determined motivation, but also that interactions will emerge between personal AG and CAF. Based on previous research (Lau & Nie, 2008; Méndez-Giménez, Cecchini, Fernández-Río, Méndez-Alonso, & Prieto-Saborit, 2017; Murayama & Elliot, 2009), the third hypothesis was that both mastery-approach CAF and Tap and Sap will positively predict self-determined motivation, while others-approach CAF and Oap will predict it negatively. Regarding the indirect effects, the fourth hypothesis was that approach goals will mediate, at least partially, the relation between approach goals and self-determined motivation. Finally, based on the scarce research on the interaction model, no hypothesis was established.
Method
Participants
2284 students included in 148 secondary education classes from 34 schools of 17 Spanish provinces agreed to participate. A convenience sampling technique was utilized. The sample’s gender distribution was 51.6% men and 48.4% women. They ranged in age from 12 to 17 years old (M = 14.31; SD =1.15). The electronic format used for the questionnaire forced participants to answer all the questions to proceed. Therefore, no questionnaire had to be disregarded. Prior to data analysis, several questionnaires were eliminated for different reasons (duplicates, item errors…).
Assessment instruments
3x2 Achievement goals. The Spanish validated version (Méndez-Giménez et al., 2017) of the 3x2 Achievement Goal Questionnaire (Elliot et al., 2001) was used. It contains 18 items grouped in six factors: task-approach goals (Tap), focused on goal achievement based on the task (i.e., “do the task correctly”), task-avoidance goals (Tav), focused on avoiding incompetence based on the task (i.e., “avoid doing the task incorrectly”), self-approach goals (Sap), focused on competence based on oneself (i.e., do it better than before”), self-avoidance goals (Sav), focused on avoiding incompetence based on the self (i.e., “avoid doing it better than before”), others-approach goals (Oap), focused on competence based on others (i.e., “do it better that others”), and others-avoidance goals (Oav), focused on avoiding incompetence based on others (i,.e., “avoid doing it worse than others”). In the validated version (Mendez-Gimenez et al., in press), Cronbach’s alphas ranged from .75 to .89. Participants responded in a 7-point likert scale from 1 (absolutely not true to me) to 7 (absolutely true to me).
3x2 achievement class structure. The 3x2 Achievement Class Structure Questionnaire (ACSQ-3x2), developed from the 3x2 Motivational Climate in physical education (PE) (Méndez-Giménez et al., in press) was used. Items were modified to refer to general academic contexts, including the heading to make participants focus on the different subjects assessed. The questionnaire includes several items that represent the CAF generated by the teacher in class (perceived by the students) and incorporates the ideas of the 3x2 framework (Elliot et al., 2011). It has 18 items grouped in six CAF (three items on each construct) with the following stem: “In class, the goal of the teacher is that each student...” (Annex 1). Task-approach (CAFTap): “...perform correctly the tasks”, task-avoidance (CAFTav): “...avoid performing wrong the tasks”, self-approach (CAFSap): “...perform the tasks better that before”, self-avoidance (CAFSav): “...avoid performing the tasks worse than before”, others-approach (CAFOap): “...perform the tasks better than others”, and others-avoidance (CAFOav): “...avoid performing the tasks worse than others”. Participants responded in a 5-point likert scale from 1 (absolutely not true to me) to 5 (absolutely true to me). In the original study Cronbach’s alphas were .82, .72, .79, .78, .84, .77 for CAFTap, CAFTav, CAFSap, CAFSav, CAFOap, and CAFOav, respectively.
Motivation in the academic context. The Spanish validated version (EME-E; Nuñez, Martín-Albo, Navarro, & Suárez, 2010) for non-university educational contexts of the Academic Motivation Scale (AMS; Vallerand et al., 1992) was used. It includes 7 subscales: amotivation (i.e., “I had good reasons to go to school, but now I ask myself if it is worth continue”), external regulation (i.e., “to obtain a more prestigious job”), introjected regulation (i.e., “because I want to probe myself that I can pass”), identified regulation (i.e., “because I believe that the education I receive in my school will improve my working competence”), intrinsic motivation tom knowledge (i.e., “because I feel pleasure and satisfaction when I learn new things”), intrinsic motivation to achievement (i.e., “for the pleasure that I feel when I pass the exams”), and intrinsic motivation to stimulating experiences (i.e., “because it is stimulating to read about things that I find interesting”). Each subscale has four items that ask about the reasons why students go to school. Participants responded in a 7-point likert scale from 1 (absolutely not true) to 7 (absolutely true). Cronbach’s alphas were .82, .75, .82, .73, .86, .88, .73, respectively. A self-determination motivation index (SDI) proposed by Vallerand (1997) was obtained using the following equation: (2 x (IM to knowledge + IM to achievement + IM to stimulating experiences)/3) + Identified Reg.( - ((Introjected Reg. + External Reg.)/2) + (Amotivation x 2)(.
Control variables. Prior research has showed that educational stage and gender can be important predictors of motivation and outcomes variables (Eccles & Midgley, 1989; Hyde & Durik, 2005). Therefore, this information was obtained and assessed. Educational stage was coded as: 1 = grades 7-8, 2 = grades 8, 9, 10. Males were coded as 1, and females as 2.
Procedure
Researchers contacted the schools’ directors to share the goals of the project and they signed a written consent. After a full review of all the existing CAF and motivational climate questionnaires, Méndez-Giménez et al.’s questionnaire (2017) was modified to assess 3x2 CAF. Two pilot studies were conducted to know if secondary education students were able to understand the meaning of the different items. Prior to test administration, teachers were informed of the protocol to use to fill the questionnaires in class and online through google forms (30 minutes aprox.). Participants had access to the questionnaire via a link provided by the researchers. They were told that it was voluntary, that their responses were anonymous, and that they would not affect their grades. They were also asked to respond honestly. All data was assessed using SPSS, 20.0. and EQS 6.2.
Data analysis
Confirmatory factor analysis. Since data did not follow a normal distribution (Kolmogorov-Smirnov, p < .05), the program EQS 6.2. (Bentler, 2005) was used to conduct a confirmatory factor analysis based on Satorra-Bentler chi-square (S-Bχ2; Satorra y Bentler, 1994) and robust standard estimates. Evaluation of the sample’s goodness-of-fit data was performed using multiple criteria (Byrne, 2008): the Comparative Fit Index (*CFI; Bentler, 1990), the Root Mean-Square Error of Approximation (*RMSEA; Browne & Cudeck, 1993), and the Standardized Root Mean Square Residual (SRMR). The *CFI represents the CFI robust version calculated on the S-Bχ2 statistical basis. The *RMSEA is the robust version of the usual RMSEA and it takes into account the approximation error in the population (Browne & Cudeck, 1993). To complete the analysis, the confidence interval to 90% provided by *RMSEA (Steiger, 1990) was also included. Finally, the SRMR, the average standardized residual value, was also used.
Comparison with alternative models. Following Elliot et al.’s procedure (2011), additional analysis were conducted to confront the fit of the hypothesized model with six alternative models: (a) the 2x2 model, where others-achievement structures loaded on the latent hypothetical factors, while task and self-achievement structures loaded on the common latent factors; (b) the trichotomic model, where others-achievement loaded on its hypothesized latent factor, but task and self-achievement loaded together in a latent factor; (c) the dichotomic model, where others-achievement loaded in a latent factor, and task and self-achievement loaded together in a common latent factor; (d) the Tap/Tav (task-approach/task-avoidance) model, where all items loaded in their hypothesized latent factors, except task-approach and task avoidance which loaded together in a common latent factor; (e) the Sap/Sav (self-approach/self-avoidance) model, where all items loaded in their hypothesized latent factors, except self-approach and self-avoidance which loaded together in a common latent factor; and (f) the Oap/Oav (others-approach/others-avoidance) model, where all items loaded in their hypothesized latent factors, except others-approach and others-avoidance which loaded together in a common latent factor. The Akaike (AIC) criterion was used to compare the hypothesized model and the alternative models (Kline, 2005). Scores for an alternative model in the chi-square difference significantly larger than cero indicate that the alternative model produces a worse fit than the hypothesized model. Lower AIC scores indicate a better fit.
Linear hierarchical model analysis. Linear hierarchical models (Raudenbush & Bryk, 2002) or multilevel (Goldstein, 2003) were designed to assess data when cases are grouped in large information units and means are considered in the lowest (students) and highest levels (classes). MIXED SPSS (18.0) procedure was used to adjust the models (Pardo, Ruiz, & San Martín, 2007). Following Elliot and Maruyama (2009), the process was divided in several steps:
First, as a preliminary analysis, the unconditioned or null model was examined to estimate the variances between classes and in the class (Hofmann, Griffin, & Gavin, 2000). CAF and SDI significant variation among classes were explored. The intraclass correlation coefficient was also obtained (ICC).
Second, the interaction effect model was assessed. Two analyses were conducted: first, a regression model with random coefficients, and later, a model of slopes as results. The regression analysis with random coefficients tries to assess which part of the variability in the class (level 1 variability) could be explained by the AG. To correctly assess the relation between AG and SDI is mandatory to obtain a regression equation for each group and asses how the intersections and the slopes of those equations change. With this procedure, it is assumed not only that classes can have a different SDI, but also that the relation SDI and personal AG can be different depending on the grade (different slopes). The models that include this type of variation are called random coefficients models, because they let both coefficients (intersection and slope) change randomly from class to class. To do this, a model was built with the six AG as simultaneous predictors of the SDI. Slopes were allowed to change through classes, and later, the non-significant random effects were eliminated for statistical reasons (Raudenbush & Bryk, 2002). Independent variables scores were re-scaled, subtracting the mean of the class group (central scores), to know the transversal interactions (Enders & Tofighi, 2007). To make the coefficient β0 clear enough, researchers usually re-scale independent variables scores, subtracting the mean (using differential scores, central, instead of the direct ones). Therefore, the mean scores were subtracted from the variable SDI. Grade level (1 = grades 7-8, 2 = grades 8, 9, 10) was included as a predictor of the intersections, while gender was kept as predictor of the fixed effects. Both variables remained non-central to facilitate the interpretation of the results. Next, a regression analysis of means and slopes as results was conducted to find which variables could signal that variability. The 3x2 CAF were added to the previous model and the interactions between CAF and personal AG. Non-significant predictors were omitted from the model. To assess the type of interactions, the significance of the slope in the CAF scores was tested, a standard deviation above and below of the samples’ mean (Bauer & Curran, 2005).
Third, the direct effects model was examined with the same model described in the previous paragraph, but re-scaling the AG scores, subtracting the mean of all participants to every score (scores centred in the total mean), because it is more appropriate when a level two predictor is more interesting and the level one effects must be controlled. Grade level and gender were controlled in the analysis.
Finally, indirect effects were examined using a multilevel mediation analysis (Krull & MacKinnon, 2001). Level one predictors were centred in the global mean. Grade level and gender were controlled in the analysis. Three phases were conducted: 1) a model was developed with the CAF as class predictors of the SDI; 2) a second model was developed with the CAF as class predictor of the personal AG, and 3) a third model was developed where both the CAF and the personal AG dimensions which had been significant for the SDI were again included as predictors of this dimension. The term intersection was used as a random coefficient.
Results
Descriptive analysis and bivariate correlations of the 3x2 achievement class structure
Table 1 shows descriptive statistics, Cronbach’s alphas and bivariate correlations of the six CAF dimensions
Control Variables
Regarding grade level, no significant differences were found in the SDI. On the contrary, gender produced significant differences in the SDI, with males showing higher scores than females.
Confirmatory factor analysis
Comparisons between models indicated that the hypothesized model provided a better fit to the data than any of the alternative models (Table 2). This results support the 3x2 CAF model and the need to differentiate between task and self.
Descriptive analysis and bivariate correlations between 3x2 personal achievement goals and SDI
Table 3 shows descriptive statistics and bivariate correlations between variables in class students’ scores.
Unconditioned model
Results from the preliminary analyses showed that all CAF and SDI dimensions changed significantly between classes: CAFTap, χ2(147) = 318.62, p < .001; CAFTav, χ2(147) = 272.06, p < .01; CAFSap, χ2(147) = 294.83, p < .001; CAFSav, χ2(147) = 238.94, p < .01; CAFOap, χ2(152) = 218.35, p < .01; CAFOav, χ2(147) = 225.75, p < .01, SDI, χ2(147)= 402.35, p < .001. The ICC were 6.9%, 4.3%, 6.7%, 3.1%, 3.1%, 3.4%, 11.3%, respectively.
Random coefficients regression analysis
For SDI, Tap and Sap were positive predictors, while Tav and Oap were negative predictors. Tap’s slope changed significantly between classes (Table 4), which indicates that in some of them the effect of Tap was positive and in others neutral or negative.
Means and slopes as outcomes regression analysis
The connection between Tap and SDI changed based on the CAFOap (Table 5). Slope analysis revealed that Tap was a positive predictor of SDI in a weak CAFOap (estimated beta = .26, p < .001) and this relation was strengthen in a strong CAFOap (estimated beta = .37, p < .001). Figure 1 shows a graphical resume of the slope results.
Direct and indirect effects
Effect direct model. Results showed that CAFSap and CAFOap were significant predictors of students’ SDI, CAFSap was a positive predictor (γ02 = 3.28, p < .001), while CAFOap was a negative predictor (γ03 = -1.28, p < .01).
Indirect effects model. Results from phase 1 (model with CAF as an in class predictor of SDI) revealed that only CAFSap (γ = 2.64, p < .001) and CAFOap (γ = -1.31, p < .01) were significant. In the subsequent analyses only these two CAF dimensions were used. Results from phase 2 (model with CAFSap and CAFOap as an in class predictor of personal AG) showed that CAFSap positively predicted Tap (γ = .71, p < .001), Tav (γ = .72, p < .001), Sap (γ = .92, p < .001), and Oap (γ = .49, p < .01), while CAFOap only positively predicted Oap (γ = .99, p < .001). Results from model 3 (CAF dimensions and personal AG who were significant were again included as predictors) showed that Tap, Tav, Sap and Oap were significant predictors of SDI. Tap (γ20 = .25, p < .01) and Sap (γ40 = 1.71, p < .001) were positive predictors, while Tav (γ30 = -.18, p < .01) and Oap were negative predictors (γ50 = -.12, p < .01).
Based on the previous results, the indirect effect of CAFSap was examined via Tap (1), Tav (2) MAY (3) and Oap (4) and the indirect effects of CAFOap via Oap (5), using the Sobel (1982) test (Krull & Mackinnon, 2001). The indirect effects were significant for (1): z = 2.32, p < .05, para (2): z = 2.14, p < .05, (3): z = 7.73, p < .001, (4): z = 2.01, p < .05, y (5): z = 2.68, p < .01. In all the cases, the mediation was partial because the negative influence of both structures (CAFSap and CAFOap) on SDI was maintained (as it was the case in the direct effects). Figure 2 shows direct, indirect effects and their interactions.
Note. CAFSap = Self-approach achievement class structure; CAFOap = Others-approach achievement class structure; SDI = Self-determined motivation index; Tap = Task-approach achievement goals; Tav = Task-avoidance achievement goals; Sap = Self-approach achievement goals; Oap = Others-approach achievement goals.
Discussion
The study had two goals: (a) examine construct validity of the ACSQ-3x2, and (b) assess the influence of CAF and 3x2 AG, jointly, in secondary education students self-determined motivation.
Regarding construct validity of the ACSQ-3x2, results showed that the six CAF represent different constructs and that the 3x2 model provides a better fit to the data than any of the six alternative models tested. Results are in line with those reported by Méndez-Giménez et al. (in revision) in the context of PE, and confirm hypothesis 1: the differentiation between task, self and others (both approach and avoidance). Findings suggest the extension of the 3x2 AG model to the environmental component generated by the teacher. Similarly, the ACSQ-3x2 could be considered a relevant tool to mind a gap in the existing literature on AG.
Participants reported higher levels of approach CAF, except in others-based. Regarding achievement class structures perceived by the students, the highest scores were obtained in CAFSap and CAFTap, and the lowest in CAFOap. Correlations between CAF were positive and moderate, in line with those found by Méndez-Giménez et al. (2017). Data analysis discriminated between CAF task-based o self-based, correlations between CAFTap - CAFSap and CAFTav - CAFSav were elevated (.74 and .71, respectively). Similar results have been obtained in previous studies on the 3x2 AG framework (Elliot et al., 2011). Bong (2009) highlighted higher correlations between AG reported by secondary tan university students.
Regarding the global effects of CAF and 3x2 AG, results allow us to support hypothesis 2. CAF was not only direct and indirect predictor of SDI, interactions between personal AG and CAF also emerged in the prediction. The preliminary analysis (unconditioned model or null) revealed that all CAF dimensions and the SDI changed significantly between classes. In contrast to the previous study of Maruyama and Elliot (2009), where no significant changes were found in the performance-avoidance CAF, in the present study significant differences appeared in the CAFOav. The different assessment tools used and cultural differences could explain the different outcomes. Maruyama and Elliot’s study (2009) was conducted on Japanese students. Prior research (Elliot, Chirkov, Kim, & Sheldon, 2001; Wang, Biddle, & Elliot, 2007) found that individuals from collectivist countries (Russia, Korea, Japan…) tend to adopt avoidance achievement goals than people from individualistic countries (North America and Europe). More research is needed to assess if these differences can also influence on the variability of CAF.
Regarding the direct effects model, results indicated that CAFSap was a direct positive predictor of self-determined motivation, while CAFOap was a negative predictor. This trend is supported by previous research works conducted on different outcome variables such as academic achievement, commitment or intrinsic motivation (Lau y Nie, 2008; Meece et al., 2006; Murayama y Elliot, 2009). Contrary to our predictions, the CAFTap was not a direct predictor. Results indicate that from the two dimensions of mastery-approach CAF in the 3x2 framework, the CAFSap was the one that plays a central role in the direct connection to self-determined motivation. Environments that promoted self-referenced competence improvement were powerful motivational stimuli for these adolescents. Congruent with prior research (Church et al., 2001; Méndez-Giménez et al., 2017), Tap and Sap were positive direct predictors of self-determined motivation, while Tav and Oap were negative direct predictors. In conclusion, results partially supported hypothesis 3.
Regarding the indirect effects model, results revealed that CAFSap was a positive predictor of Tap, Tav, Sap and Oap, and CAFOap was a positive predictor only of Oap. Results indicated that goals have acted as partial mediators with the self-determined motivation. The connection between CAFSap and the rest of the goals emerged for the first time in the scientific literature. Further from our more restrictive prediction (hypothesis 4), CAFSap promoted the adoption of different types of goals, not only the matching (Sap), and all had a proximal influence in the motivational results. For its part, results of the relation CAFOap - Oap were convergent with previous studies who found that performance-approach CAF leads to the adoption of performance-approach goals (Schwinger, & Stiensmeier-Pelster, 2011; Urdan, 2004), and those who emphasize the maladaptive role of both CAF and personal performance-approach goals (Anderman, Cupp, & Lane, 2009; Givens-Rolland, 2012; Kaplan, Gheen, & Midgley, 2002; Lau, & Nie, 2008; Murdock, Hale, & Weber, 2001).
Regarding the interaction effects model, results showed that the relation between Tap and SDI changed according to the CAFOap and that relation was damaged in a CAFOap strong vs weak. Lau and Nie (2008) also examined context-individual interactions using a multilevel analysis y they warned that the approach-avoidance structures (CAFOap in this study) could be specially negative for those students who adopt strong performance-avoidance goals, because they promote the maladaptive effects of the relation between performance-avoidance goals an relevant achievement outcomes. Emphasizing achievement goals in class can aggravate the risk of this group of students, even when teachers focus on the approach dimension of performance. Those students who adopt low Tap and perceive a strong CAFOap are less self-determined motivated than those who also adopt low Tap, but perceived a weak CAFOap. Interestingly, students with high Tap scores seem to be the more resistant to the negative effects of CAFOap than those with lower Tap scores. Future research should deepen in this and other interactions between CAF and AG.
Results from the present study have important practical implications. It seems highly recommendable to promote self-referenced contexts to promote students’ self-determined motivation. Similarly, teachers should limit comparative contexts, contexts focused on competence achievement based on others, or avoidance structures for their negative outcomes. Finally, students’ Tap and Sap in educational contexts should be fostered, because these personal orientations are more adaptive and they protect from the negative effect of normative contexts.
A limitation of the present study is that students provided, at the same time, CAF and AG data. This could have produced a bias in the participants’ responses (social desirability or acquiescence) or an over or underestimation of the scores and the correlations among variables.
In conclusion, this study has re-affirmed within the 3x2 AG framework, Murayama and Eliot’s (2009) thesis that considering globally the three models (direct, indirect an interactive) represent a more profound and complete scenario of analysis that can focus on each of the models independently. Future research should continue assessing this integral framework of analysis, still underused.
