Language mindsets,perceived instrumentality,engagement and graded performance in English as a foreign language students

1 Language mindsets

Mindsets refer to the beliefs about the nature of intelligence and are broadly categorized as entity beliefs and incremental beliefs (Dweck, 2006). The former describes personal propensities (e.g. intelligence) as fixed, whereas the latter defines personal traits as malleable (Dweck, 2006). Relevant studies showed that students’ mindsets were significantly related to their engagement with learning and academic achievement (e.g. Blackwell, Trzesniewski, & Dweck, 2007).

Recently, mindsets have also been examined in terms of students’ beliefs about foreign language learning, labeled as language mindsets (e.g. Ryan & Mercer, 2012). Relevant studies revealed that students’ language mindsets were significantly related to their motivations for language learning, for engagement in classes, and for academic achievement (e.g. Lou, Masuda, & Li, 2017). Nonetheless, the generalizability of the results of these studies is highly restricted because they contained a limited number of participants and lacked a comprehensive theoretical conceptualization of language mindsets in relation to various learning-related constructs (e.g. student engagement).

Indeed, a comprehensive conceptualization of students’ language mindsets has only recently been developed. Specifically, Lou and Noels (2016, 2017) described students’ language mindsets based on the three factors: beliefs about general language intelligence (GLB), beliefs about second language aptitude (L2B), and beliefs about age sensitivity in language learning (ASB). The GLB refers to beliefs about whether general language ability is fixed or malleable, whereas the L2B refers to beliefs about whether the ability to learn a foreign language is fixed or can be altered. The ASB, however, refers to beliefs about whether language ability is malleable up to a certain age and becomes fixed afterwards or whether language learning is malleable throughout the lifespan. Lou and Noels (2016) further designated these factors based on the ‘incremental’ and ‘entity’ axis and obtained six factors accordingly: entity beliefs about general language intelligence, incremental beliefs about general language intelligence, entity beliefs about second language aptitude, incremental beliefs about second language aptitude, entity beliefs about age sensitivity in language learning, and incremental beliefs about age sensitivity in language learning.

The results of the studies conducted by Lou and Noels (2016, 2017) demonstrated that students’ language mindsets could be reliably examined based on the aforementioned factors. Furthermore, Noels and Lou (2015) also found that students with incremental mindsets endorsed learning goals and reported greater mastery goals in addition to less helplessness in failure situations, regardless of their actual language competence level. In contrast, students with entity mindsets adopted performance goals and reported a higher level of anxiety.

Notably, although the sample of these studies contained either native or non-native English-speaking students, none of the studies addressed EFL students attending an IEP for a year at the tertiary level of education. However, an IEP by its nature provides an intense language learning context, particularly for non-native English-speaking students in many educational contexts (e.g. Chinese educational context; Peng, 2007). Therefore, the present study was conducted based on the non-native English-speaking students enrolled in an IEP to investigate the roles of language mindsets in a highly relevant context that requires EFL students to achieve their language-specific learning goals more effectively. Due to its uniqueness and comprehensiveness, the conceptualization of language mindsets (Lou & Noels, 2016) was used in the present study.

2 Student engagement

Student engagement can be defined as ‘the quality of a student’s connection or involvement with the endeavor of schooling and hence with people, activities, goals, values, and place that compose it’ (Skinner, Kinderman, & Furrer, 2009, p. 494). Considerable research demonstrated that student engagement was significantly related to academic achievement as well as the improvement of learning through equality, social justice, and curricular relevance (e.g. Quaye & Harper, 2014).

Previous studies on student engagement were conducted mostly based on cognitive engagement and, to a lesser extent, emotional and behavioral engagement. However, student engagement is hypothesized as an inclusive term of cognitive, emotional, and behavioral engagement, which particularly reflects diverse, yet interrelated, facets of the learning process (e.g. Trowler, 2010). Accordingly, cognitive engagement refers to ‘effective strategy use, commitment to learning, and flexibility in problem-solving, while emotional engagement refers to students’ affective reactions in the classroom, including interest, boredom, happiness, sadness, and anxiety’ (Fredricks, Blumenfeld, & Paris, 2004, p. 63). Behavioral engagement ‘concerns involvement in learning and academic tasks such as effort, persistence, concentration, attention, asking questions, and contributing to class discussion’ (Fredricks et al., 2004, p. 62).

Recently, the concept of agentic engagement has also been introduced in the relevant literature to define ‘students’ constructive contribution to the flow of the instruction they receive’ (Reeve & Tseng, 2011, p. 258). Specifically, agentic engagement refers to ‘the process in which students intentionally and somewhat proactively try to personalize and otherwise enrich both what is to be learned and the conditions and circumstances under which it is to be learned’ (Reeve & Tseng, 2011, p. 258).

Reeve and Tseng (2011) studied the four aspects of engagement (i.e. cognitive, emotional, behavioral, and agentic) with a sample of high school students from Taiwan and reported that agentic engagement was both a separate and a central construct that significantly and positively projected students’ academic achievement. Furthermore, studies reveal that the four aspects of engagement could provide a reliable and valid framework in samples other than Taiwanese students. For example, Eren (2013) showed that the four aspects of engagement framework had good fit to the data that were obtained based on a large sample of university students (i.e. pre-service teachers) from Turkey (n = 521).

As it has been argued in the relevant literature, despite the important role that student engagement beliefs play on the learning processes of the individuals, limited attention has been given to student engagement in the EFL contexts (Dinçer et al., 2019). In relation, rather than constructing direct links among the aforementioned variables, parallel links were addressed to understand the role of engagement in EFL learning. For instance, Chen and Kraklow (2015) argued that when EFL learners’ motivations and engagement were investigated, their overall motivation and regulation significantly predicted the learners’ behavioral engagement. Conversely to the aforementioned limited findings on the role of engagement in the language learning contexts, Noels, Lascano, and Saumure (2019) reported that despite the positive relations of motivational orientations, language learners’ engagement decreased over time. Therefore, the need for implementing a multivariate relation among related constructs (e.g. language mindsets, instrumentality) to explore the probable reciprocity of the relevant constructs and student engagement in the EFL contexts emerges. Based on the relevance of examining student engagement in addressing its role in the academic achievement of learners, thus, in the present study, the four aspects of student engagement framework were used to examine the EFL students’ engagement in English classes.

3 Perceived instrumentality

Perceived instrumentality reflects the value aspect (i.e. utility value/perceived usefulness of the task) of the expectancy-value theory of achievement motivation (Wigfield & Eccles, 2000). Utility value refers to ‘how a task fits into an individual’s future plans’ (Wigfield & Eccles, 2000, p. 72). Thus, perceived instrumentality can be described as the extent to which individuals perceive their current actions as instrumental in reaching their future goals (Bembenutty, 2010). According to Phalet, Andriessen, and Lens (2004), the motivational force of achieving future goals depends highly on the perceived instrumentality on the content being studied. Likewise, considerable research showed that perceived instrumentality was significantly and positively related to students’ valuing for their studies (e.g. Husman & Lens, 1999), and academic engagement (e.g. Horstmanshof & Zimitat, 2007). Moreover, there is evidence indicating that high levels of perceived instrumentality could positively relate to individuals’ knowledge-seeking behaviors (e.g. Nebus, 2004), epistemic curiosity (Eren, 2011), and knowledge constructing strategies (e.g. Hilpert et al., 2012).

Conversely, only a very small number of studies examined the relationship between non-native English-speaking students’ perceived instrumentality and graded performance. For example, based on a sample of vocational high school students from Turkey, Özçetin and Eren (2010) showed that, regardless of the effects of the demographic variables (e.g. gender), students’ perceptions of instrumentality regarding English classes significantly and positively predicted their attitudes towards English classes and graded performance. Likewise, based on a sample of middle school students from Iran, Rostami, Hejazi, and Lavasani (2011) found that students’ perceptions of instrumentality regarding English classes played a significant role in explaining their graded performance.

It should be noted that perceived instrumentality could be explored both through a one-factor structure (e.g. Özçetin & Eren, 2010; Rostami et al., 2011) and a multi-factor structure (e.g. Husman & Lens, 1999). For example, Miller, DeBacker, and Greene (1999) used a perceived instrumentality scale (PIS) to examine students’ evaluation of the academic tasks in terms of instrumentality to attain their personal goals. The results revealed that perceived instrumentality could be reliably examined by a single-factor structure. However, Husman and Lens (1999) examined perceived instrumentality through a two-factor structure; exogenous instrumentality, which refers to ‘a task that is useful for jumping hurdles but not necessarily useful for fully realizing a long-term goal’, and indigenous instrumentality, which refers to ‘understand[ing] the ways in which gaining competence in a task can help reach future goals’ (Husman et al., 2004, p. 66). Furthermore, Simons, Dewitte, and Lens (2000) demonstrated that perceived instrumentality could be investigated through a four-factor structure: proximal utility-internal regulation, proximal utility-external regulation, distal utility-internal regulation, and distal utility-external regulation. Because the present study focused on EFL students’ overall perceived instrumentality regarding English classes in the context of an IEP, perceived instrumentality was examined as a unidimensional (single-factor) construct (Miller et al., 1999).

1 Context and participants

In Turkey, at most of the universities (n = 73), a year-long IEP apart from departmental studies is offered in undergraduate programs (National Measuring, Selection, and Placement Centre, 2016). Thus, university students attending IEPs are required to pursue their tertiary education by means of attending a program in English-medium-instruction, which requires the students surpassing in academic English to begin, pursue and complete undergraduate and lifelong academic studies. The major aim of the IEP is to increase students’ EFL proficiency in order to use the English language effectively in their departmental studies. The responsibility of the planning, programming, and implementing the IEPs is conducted by the School of Foreign Languages (SFL) as official academic units of the universities.

The IEP in the current context is offered both on a voluntary basis, at any time throughout the tertiary education, if students are willing to attend, and on a mandatory basis, before entering the departments that use English as the language of instruction (e.g. chemistry). Hence, students who pass the proficiency/placement examination and who enter the department can attend the IEP for a year and can then continue with their departmental studies after completing the one-year program. The IEP consists of 40 hours per week of integrated skills-based English language classes focused on speaking, listening, reading, writing, and grammar. The classes are arranged to encourage students to develop both their written and oral communication skills employing a variety of activities (e.g. note-taking strategies).

Notably, recent reports on the state of English in higher education in Turkey revealed that ‘students enter preparatory school with low English proficiency levels and low motivation. Preparatory school classes do not fully address these problems as the curriculum is perceived to be lacking in relevance’ (British Council, 2015, p. 110). This phenomenon indicates that it is highly important to focus on how EFL students’ motivation to engage with learning EFL can be increased throughout IEPs. In fact, this is a serious concern in many countries that EFL proficiency level is categorized as either low or very low (e.g. Japan, Brazil, Thailand, China, Albania, and Turkey) in terms of EFL learning (English First, 2019). Thus, it can be claimed that the context of the study (i.e. EFL students from Turkey) is highly relevant to the international concerns of EFL learning.

Specifically, a total of 526 EFL students (330 female) were conveniently sampled from the IEP (approximately 800 students) of the SFL at a large university (approximately 33,000 undergraduate students) located in the north-west of the Black Sea region in Turkey. The students ranged in age from 17 to 36 years old (M = 19.05; SD = 1.56). Of the students, 357 were grouped as A (low level of language proficiency), 37 as B (mid-level of language proficiency), 46 as C (high level of language proficiency), and 86 as R (repeating students), according to their initial language proficiency levels determined by the SFL at the beginning of the semester via a comprehensive placement test. Accordingly, 94 of the students reported that they had attended language classes in addition to their high school education, whereas 448 of the students reported that they had not attended an extra language course other than the one offered at the high schools. Additionally, 78 students reported that they had attended an IEP, whereas 448 of the participants reported that they had not been enrolled in an IEP. The students came from 18 different departments (e.g. chemistry, psychology, international relations, English language teaching).

2 Research instruments

3 Procedure

The data were collected during the fall semester of the 2017–18 academic year by the researchers. The LMI, FASEQ, and PIS were applied to the participants who volunteered to participate in the study (n = 526; participation rate = 66%) at the end of one class hour. Instructions included the aim of the study and brief explanations on how to respond to the items in the research instruments. These instructions were also read aloud at the beginning of the process. The administration process lasted approximately 20 minutes. Ethical approval was obtained from the ethics committee of the university, where the study was carried out.

4 Data analysis

No missing data were observed. This can be due to the fact that all of the participants voluntarily participated in the study. Thus, the data collected from the participants were all included in the data analyses. Before addressing the research questions, the possible effects of demographic variables on the research variables were examined by conducting a series of univariate analyses of covariance (ANCOVA) and multivariate analyses of covariance (MANCOVA). In the analyses, gender, initial English language proficiency level, private English language coursework before enrolling in the IEP, and completion of English language classes during high school before enrolling in the IEP were determined as the independent variables, and age was determined as a covariate, whereas the research variables were determined as the dependent variables.

The results of these analyses were investigated in terms of the magnitudes of partial eta-square (η²_p) coefficients because they are less sensitive to sample size than p values (Ferguson, 2009), and those coefficients that were smaller than 0.06 were interpreted as negligible (Richardson, 2011). The results demonstrated that, with one exception (i.e. initial English language proficiency levels), the effects of the demographic variables (i.e. gender and age) on the research variables were trivial (all η²_p coefficients were equal to or smaller than 0.02). Thus, gender and age were not included in the analyses that were conducted to address the research questions. However, initial English language proficiency levels were included in the regression and structural equation modeling (SEM) analyses because they were considerably related to graded performance (η²_p = 0.09). Likewise, the results of the pairwise comparisons revealed that the achievement scores of the Level A students were lower than the Level C students but higher than the Level R students; the achievement scores of the Level B students were lower than the Level C students but higher than the Level R students; the achievement scores of the Level C students were the highest. On the other hand, Level R students had the lowest scores among all the groups. In fact, it has long been acknowledged that students’ initial foreign language proficiency significantly and positively relates to their later foreign language proficiency (e.g. Cummins, 1984); hence, these results are not unexpected.

Dummy coding procedure (also known as indicator coding procedure; see Pedhazur, 1997) was used to code the categories of the initial English language proficiency levels (Level R was described as the reference category). Accordingly, three binary variables were created, each of which contained dummy coded (1 and 0) categories of initial English language proficiency levels (i.e. A, B, and C). However, Level R was not coded because SPSS and AMOS automatically described the non-coded category as the reference category. The effects of students’ initial English language proficiency levels on their graded performance were not discussed due to the scope of the study and were not shown in the Tables for presentation clarity (but see Figure 2). Notably, the effects of binary variables on dependent variables can be reliably examined through the SEM analysis using the maximum likelihood (ML) method of estimation as long as binary variables are determined as independent variables and the possible distributional violations are controlled by using robust methods (e.g. percentile bootstrap) (Hancock & Mueller, 2006; Hoyle, 2011).

OPEN IN VIEWER

Figure 2.

The mediation model.

For the first research question, a latent factor correlation analysis and a series of multiple regression analyses, in which all the latent research variables (i.e. language mindsets, perceived instrumentality, and engagement) and graded performance as an observed variable were allowed to correlate with each other, were conducted. As aforementioned, the p values are highly sensitive to the sample size, and significant p values are likely to be found in large samples even when the magnitudes of correlation coefficients are negligible (Sullivan & Feinn, 2012). Hence, the magnitudes of the correlation coefficients were considered to reliably evaluate the importance of the relationships among the research variables based on Cohen’s (1988) guidelines.

Latent factor correlation analysis was conducted using the ML method of estimation from AMOS 20 (Arbuckle, 2011), whereas the multiple regression analyses were conducted using the linear regression module of the SPSS. With one exception (i.e. graded performance), all the variables were described as the latent factors in the correlation and SEM analyses as they were obtained through the self-report measures. For the second research question, a structural model was created to examine the indirect effects of language mindsets on graded performance through perceived instrumentality and four aspects of engagement based on the results of the regression analyses. All the significant links between the research variables that were detected through the regression analyses were examined in the structural model to validate mediation accurately (MacKinnon, Fairchild, & Fritz, 2007).

Using the ML method of estimation from AMOS 20 (Arbuckle, 2011), a SEM analysis was conducted to examine whether the structural model fits to the current data in terms of robust fit indices: comparative fit index (CFI), Tucker–Lewis index (TLI), root mean square error of approximation (RMSEA), and standardized root mean square residual (SRMR). Notably, the SEM analysis and latent factor correlation analysis were conducted based on the percentile bootstrap method (1000 bootstrap samples; 95% CI), which is a non-parametric resampling method, to control for possible distributional violations as well as to reduce possible Type I error rates (Hayes & Scharkow, 2013). Indeed, both parametric and non-parametric bootstrap methods could be used for estimating standard errors (Efron & Tibshirani, 1993). Nonetheless, because parametric bootstrap methods are less likely to be immune to systematic measurement error than non-parametric bootstrap methods (e.g. Kolen & Brennan, 2004), a non-parametric bootstrap method (i.e. percentile bootstrap) was used in the SEM and latent factor correlation analyses.

1 Preliminary analyses

Confirmatory factor analyses (CFAs) were conducted to examine the construct validity of the research instruments. The CFI, TLI, RMSEA, and SRMR were used to assess the fit of the data (Schumacker & Lomax, 2016). The CFI and TLI values between 0.90 and 0.95 indicate an acceptable, above 0.95 indicate a good fitting model (Marsh, Hau, & Wen, 2004; Schumacker & Lomax, 2016). In contrast, an RMSEA value less than or equal to 0.05 indicates a good, 0.08 or less indicates an acceptable fitting model. ¹ Similarly, an SRMR value less than or equal to 0.05 indicates good, 0.10, or less indicates an acceptable fitting model (Pituch & Stevens, 2016). Cronbach’s coefficients alphas were also computed to examine the internal reliability of the research instruments (DeVellis, 2017). The results are presented in Table 1.

OPEN IN VIEWER

Table 1.

The summary of the confirmatory factor analyses and reliability analyses.

Scale	Factor	χ²(df)	CFI	TLI	RMSEA[90% CI]	SRMR	α
LMI		247.99(117)	0.969	0.959	0.046[0.038/0.054]	0.040
	GLB-Entity						0.86
	GLB-Incremental						0.76
	L2B-Entity						0.50
	L2B-Incremental						0.75
	ASB-Entity						0.71
	ASB-Incremental						0.86
Alternative model 1 of the LMI		558.87(127)	0.896	0.875	0.080[0.073/0.087]	0.085
	Entity beliefs						0.81
	Incremental beliefs						0.86
Alternative model 2 of the LMI		572.44(128)	0.893	0.872	0.081[0.075/0.088]	0.087
	Language mindsets						0.89
FASEQ		452.73(199)	0.945	0.936	0.049[0.043/0.055]	0.047
	Agentic engagement						0.77
	Behavioral engagement						0.87
	Emotional engagement						0.80
	Cognitive engagement						0.82
PIS		8.59(3)	0.997	0.991	0.060[0.014/0.108]	0.011	0.92

Note. CFI = comparative fit index. TLI = Tucker–Lewis index. RMSEA = root mean square error of approximation. SRMR = standardized root mean square residual. LMI = the language mindset inventory. FASEQ = the four aspects of engagement scale. PIS = the perceived instrumentality scale. Alternative model 1of the LMI: second-order factor model with two latent factors. Alternative model 2 of the LMI: second-order factor model with singe latent factor. ASB = age sensitivity in language learning. GLB = general language intelligence. L2B = beliefs about second language aptitude. α = Cronbach’s alpha.

As seen in Table 1, the six-factor model with 18 items (i.e. LMI) had a good fit to the current data in terms of fit indices. With one exception (i.e. L2B-Entity), all the alphas were acceptable (DeVellis, 2017). The exception was due to a scale item which had a content-specific explanation (i.e. to a large extent, a person’s biological factors, e.g. brain structures, to determine his or her abilities to learn new languages), which could have created confusion and a response bias among the participants. Thus, the second language entity beliefs factor that the exceptional item belonged to was not included in later analyses.

Two alternative models were also tested to examine whether the second-order factor models of the LMI had better fit to the current data than the first-order factor model. In the first model (i.e. second-order factor model with two latent factors), subscales of the LMI were diversely predicted by two latent factors. Specifically, the GLB-Entity, L2B-Entity, and ASB-Entity subscales were predicted by one latent factor (i.e. entity beliefs); whereas the GLB-Incremental, L2B-Incremental, and ASB-Incremental subscales were predicted by another latent factor (i.e. incremental beliefs). In the second model (i.e. second-order factor model with a single latent factor), all the subscales were predicted by one latent factor (i.e. language mindsets). The results demonstrated that neither the first model nor the second model had an acceptable fit to the current data (Table 1). Thus, the first-order factor model (i.e. the six-factor model with 18 items) was used in the present study.

Additionally, the results further demonstrated that the four-factor model with 22 items (i.e. FASEQ) had an acceptable fit to the data (Table 1). Moreover, alphas were satisfactory for each subscale of the FASEQ. The results also revealed that the one-factor model with five items (i.e. PIS) had a good fit to the data. Notably, Cronbach’s coefficient alpha was highly satisfactory for the PIS (Table 1).

2 Latent factor correlation analysis

As seen in Table 2, with correlation coefficients ranged in magnitude from 0.22 to 0.35, incremental mindsets were moderately and positively related to the four aspects of engagement and perceived instrumentality. On the other hand, with correlation coefficients ranged in magnitude from –0.15 to –0.28, entity mindsets were somewhat weakly and negatively related to the four aspects of engagement and perceived instrumentality.

OPEN IN VIEWER

Table 2.

Descriptive statistics and latent factor correlations.

Variable	M (SD)	1	2	3	4	5	6	7	8	9	10	11
1. GLB-Entity	5.61 (2.49)	–
2. GLB-Incremental	12.47 (2.10)	−0.72	–
3. L2B-Incremental	12.46 (2.25)	−0.38	0.56	–
4. ASB-Entity	8.55 (2.67)	0.50	−0.42	−0.24	–
5. ASB-Incremental	11.39 (2.62)	−0.49	0.59	0.48	−0.73	–
6. Agentic engagement	16.33 (3.74)	−0.18	0.31	0.23	−0.20	0.25	–
7. Behavioral engagement	20.35 (3.27)	−0.28	0.32	0.22	−0.17	0.24	0.43	–
8. Emotional engagement	16.13 (2.74)	−0.28	0.35	0.35	−0.22	0.32	0.52	0.75	–
9. Cognitive engagement	30.27 (5.05)	−0.15	0.24	0.26	−0.22	0.27	0.54	0.63	0.66	–
10. Perceived instrumentality	22.13 (3.25)	−0.19	0.28	0.35	−0.19	0.24	0.36	0.58	0.53	0.56	–
11. Graded performance	67.54 (16.51)	−0.17	0.09	0.03	−0.16	0.14	0.28	0.24	0.25	0.17	0.22	–

Note. Coefficients equal to 0.14 and 0.15 are significant at p < 0.05 level of significance whereas coefficients smaller than 0.14 are non-significant. Correlations equal to 0.16 or above are significant at p < 0.01 level of significance. All p values were corrected based on the percentile bootstrap method. The lower and upper bounds of the correlation coefficients were not shown in the Table for presentation clarity. ASB = age sensitivity in language learning. GLB = general language intelligence. L2B = beliefs about second language aptitude.

The relationships between incremental mindsets and graded performance were weak and positive (r values were equal to or smaller than 0.14), whereas the relationships between entity mindsets and graded performance were weak and negative (r values were equal to or smaller than –0.17). Behavioral engagement, emotional engagement, and cognitive engagement were positively and strongly related to perceived instrumentality (r values were equal to or bigger than 0.53). However, agentic engagement was positively, yet moderately, related to perceived instrumentality (r = 0.36). The relationships between agentic engagement, behavioral engagement, emotional engagement, and graded performance were positive and moderate (r values were equal to or bigger than 0.24), while the relationship between cognitive engagement and graded performance was positive, yet quite weak (r = 0.17). Perceived instrumentality was moderately and positively associated with graded performance (r = 0.22).

3 Multiple regression analyses

Along with R² values ranging in magnitude from 0.10 to 0.32, the independent variables explained a significant amount of the total variance in the dependent variables (Table 3), indicating that it was reasonable to explore the relationships between the research variables. Multicollinearity was not a concern in the regression analyses because all of the computed variance inflation factors were smaller than 2.50, which was below the suggested rules of thumb, such as 4 or 10 (O’Brien, 2007).

OPEN IN VIEWER

Table 3.

The summary of the regression analyses.

Independent variable	Dependent variable	B	S.E.	β	t
GLB-Entity	Perceived instrumentality	−0.02	0.07	−0.01	−0.24
GLB-Incremental		0.12	0.09	0.08	1.33
L2B-Incremental		0.36	0.07	0.25	4.80***
ASB-Entity		0.01	0.06	0.01	0.18
ASB-Incremental		0.07	0.07	0.06	0.98
Model summary: R² = 0.11; F(5,520) = 12.92, p < 0.001
GLB-Entity	Agentic engagement	−0.00	0.08	−0.00	−0.02
GLB-Incremental		0.25	0.11	0.14	2.34*
L2B-Incremental		−0.05	0.09	−0.03	−0.54
ASB-Entity		0.01	0.07	0.00	0.07
ASB-Incremental		0.12	0.08	0.08	1.37
Perceived instrumentality		0.25	0.05	0.21	4.82***
Model summary: R² = 0.10; F(6,519) = 9.03, p < 0.001
GLB-Entity	Behavioral engagement	−0.16	0.06	−0.12	−2.62**
GLB-Incremental		0.13	0.08	0.08	1.58
L2B-Incremental		−0.11	0.07	−0.08	−1.64
ASB-Entity		0.07	0.06	0.05	1.15
ASB-Incremental		0.12	0.06	0.09	1.80
Perceived instrumentality		0.50	0.04	0.50	12.88***
Model summary: R² = 0.32; F(6,519) = 39.70, p < 0.001
GLB-Entity	Emotional engagement	−0.11	0.05	−0.10	−2.02*
GLB-Incremental		0.13	0.07	0.10	1.84
L2B-Incremental		0.06	0.06	0.05	0.97
ASB-Entity		0.06	0.05	0.06	1.21
ASB-Incremental		0.14	0.06	0.13	2.53*
Perceived instrumentality		0.33	0.03	0.39	9.83***
Model summary: R² = 0.28; F(6,519) = 34.24, p < 0.001
GLB-Entity	Cognitive engagement	0.09	0.10	0.04	0.90
GLB-Incremental		0.13	0.13	0.05	0.97
L2B-Incremental		0.10	0.11	0.04	0.89
ASB-Entity		−0.02	0.09	−0.01	−0.21
ASB-Incremental		0.21	0.10	0.11	2.04*
Perceived instrumentality		0.69	0.06	0.44	11.06***
Model summary: R² = 0.26; F(6,519) = 30.07, p < 0.001
GLB-Entity	Graded performance	−0.31	0.33	−0.05	−0.95
GLB-Incremental		−0.41	0.43	−0.05	−0.95
L2B-Incremental		−0.54	0.36	−0.07	−1.50
ASB-Entity		−0.48	0.30	−0.08	−1.60
ASB-Incremental		0.06	0.34	0.01	0.17
Perceived instrumentality		0.52	0.24	0.10	2.14*
Agentic engagement		0.51	0.19	0.12	2.68**
Behavioral engagement		0.48	0.28	0.09	1.71
Emotional engagement		0.41	0.33	0.07	1.24
Cognitive engagement		−0.10	0.16	−0.03	−0.58
Model summary: R² = 0.27; F(13,512) = 14.52, p < 0.001

Notes. *p < 0.05. **p < 0.01. ***p < 0.001. Unstandardized beta coefficients (B), standardized beta coefficients (β), t values, and standard errors (S.E.) of the unstandardized beta coefficients were reported in the Table. The effects of initial English language proficiency of the undergraduate students on graded performance were controlled in the analysis. However, they were not reported due to the scope of the study. ASB = age sensitivity in language learning. GLB = general language intelligence. L2B = beliefs about second language aptitude.

Specifically, incremental beliefs about second language aptitude significantly and positively predicted perceived instrumentality (Table 3). Incremental beliefs about general language intelligence significantly and positively predicted agentic engagement, whereas entity beliefs about general language intelligence significantly yet negatively predicted behavioral engagement and emotional engagement. Incremental beliefs about age sensitivity in language learning significantly and positively predicted emotional engagement and cognitive engagement. No other significant relationships between the language mindsets, perceived instrumentality, and four aspects of engagement were observed. Perceived instrumentality significantly and positively predicted the four aspects of engagement. Notably, agentic engagement and perceived instrumentality significantly and positively predicted graded performance while language mindsets, emotional engagement, behavioral engagement, and cognitive engagement did not (Table 3).

4 Structural equation modeling analysis

The results of the SEM analysis revealed that the structural model had acceptable fit to the data (χ²(821) = 1659.31; CFI = 0.922; TLI = 0.914; RMSEA = 0.044 [90% CI: 0.041/0.047]; SRMR = 0.055). As shown in Table 4, the results of the SEM analysis were almost identical to the results of the regression analyses in terms of (1) the direct effects of language mindsets on perceived instrumentality and four aspects of engagement; (2) the direct effects of perceived instrumentality on four aspects of engagement and graded performance; and (3) the direct effect of agentic engagement on graded performance. Likewise, language mindsets, four aspects of engagement, and perceived instrumentality explained a significant amount of the total variance in graded performance (R² = 0.26, p < 0.01).

OPEN IN VIEWER

Table 4.

The summary of the structural equation modeling analysis.

Predictor variable	Predicted variable	B	S.E.	β
L2B-Incremental	Perceived instrumentality	0.49	0.10	0.35**
GLB-Incremental	Agentic engagement	0.21	0.06	0.21**
Perceived instrumentality		0.38	0.09	0.32**
GLB-Entity	Behavioral engagement	−0.14	0.04	−0.18**
Perceived instrumentality		0.57	0.07	0.55**
GLB-Entity	Emotional engagement	−0.09	0.04	−0.15**
ASB-Incremental		0.08	0.03	0.13*
Perceived instrumentality		0.38	0.05	0.49**
ASB-Incremental	Cognitive engagement	0.11	0.04	0.14**
Perceived instrumentality		0.51	0.08	0.52**
Perceived instrumentality	Graded performance	3.10	1.25	0.12*
Agentic engagement		4.47	1.19	0.20**

Notes. *p < 0.05. **p < 0.01. Unstandardized beta coefficients (B), standardized beta coefficients (β), and bootstrap standard errors were reported in the Table. The effects of initial English language proficiency of the undergraduate students on graded performance were controlled in the analysis. However, they were not reported due to the scope of the study. All p values were corrected based on the percentile bootstrap method. The Lower and upper bounds of the standardized beta coefficients were not shown in the Table for presentation clarity. ASB = age sensitivity in language learning. GLB = general language intelligence. L2B = beliefs about second language aptitude.

Nevertheless, the SEM analysis provided more refined results by demonstrating that (1) the relationship between incremental beliefs about second language aptitude and graded performance was fully mediated by perceived instrumentality; (2) the relationship between incremental beliefs about general language intelligence and graded performance was fully mediated by agentic engagement; and (3) the relationship between perceived instrumentality and graded performance was partially mediated by agentic engagement (see Figure 2 and Table 5).

OPEN IN VIEWER

Table 5.

The total, unmediated, and mediated effects of mindsets.

Predictor variable	Predicted variable	Total effect	Unmediated effect	Mediated effect
L2B-Incremental	Graded performance	0.06 [0.03/0.10]	–	0.06 [0.03/0.10]
GLB-Incremental		0.04 [0.02/0.08]	–	0.04 [0.02/0.08]
Perceived instrumentality		0.18 [0.09/0.27]	0.12 [0.03/0.21]	0.07 [0.03/0.12]

Note. Standardized effects were reported in the Table. The coefficients in brackets represent lower (before the slash) and upper (after the slash) bounds of the standardized effects. The coefficients were at least significant at p < 0.05 level of significance and were corrected based on the percentile bootstrap method. The effects of initial English language proficiency of the undergraduate students on graded performance were controlled in the analysis. GLB = general language intelligence. L2B = beliefs about second language aptitude.

Notably, the lower bounds of the mediated effects of incremental beliefs about second language aptitude, incremental beliefs about general language intelligence, and perceived instrumentality on graded performance were all greater than zero (Table 5). These results indicate that the mediated effects of incremental beliefs about second language aptitude, incremental beliefs about general language intelligence, and perceived instrumentality on graded performance were significant.

1 The relationships between the research variables

The results of the correlational analysis showed that regardless of their specific aspects (e.g. ASB, L2B, and GLB), incremental mindsets were moderately and positively related to the four aspects of engagement and perceived instrumentality, whereas entity mindsets were weakly and negatively related to the four aspects of engagement and perceived instrumentality. Language mindsets refer to the beliefs that students hold for the nature of language learning (Lou & Noels, 2016), and students’ language mindsets selectively relate to their engagement with and motivations for the language learning process (Lou & Noels, 2017). Thus, it can be understood why entity mindsets and incremental language mindsets were selectively related to the diverse aspects of engagement.

Notably, relevant literature suggests that increased instrumentality of foreign language learning is associated with greater motivation, amplified performance, and rigorous persistence (e.g. Simons et al., 2000). Indeed, the mentioned variables are not entirely free from the effects of language mindsets because if, for example, a student believes that her/his language ability is malleable, s/he may upgrade the value of exerting effort to learn EFL and may be more likely to persist in engaging with English classes. In turn, this outcome may strengthen the relationships between incremental language mindsets and perceived instrumentality of English classes because the contents of these classes are likely to be perceived by the students as instrumental in attaining their future goals. At this point, it makes sense to claim that the reverse would be true for students who adopt an entity mindset, which explains the weak and negative relationships between entity language mindsets, the four aspects of engagement, and perceived instrumentality.

Following the same line of argument, the significant and positive relationships between the four aspects of engagement and perceived instrumentality can also be inferred because students highly value their schoolwork (Husman & Lens, 1999) and are more task-oriented (Simons et al., 2000) when they perceive their learning-related efforts as instrumental to attain their future goals (Bembenutty, 2010). This phenomenon would be particularly true for EFL students due to the specific nature of the IEP context because EFL students can perceive the relevant learning activities in these kinds of language-specific contexts as highly instrumental in attaining their future goals (e.g. increasing proficiency in the English language) (Wigfield, Tonks, & Lutz-Klauda, 2009).

Additionally, the results of correlation analysis revealed that the relationships between behavioral engagement, emotional engagement, cognitive engagement, and perceived instrumentality were positive and strong, whereas the relationship between agentic engagement and perceived instrumentality was positive, yet moderate. Agentic engagement requires students to proactively and intentionally try to enhance ‘both what is to be learned and the conditions and circumstances under which it is to be learned’ (Reeve & Tseng, 2011, p. 258). As such, agentic engagement more or less contains other types of engagement (e.g. cognitive engagement) and refers to the diverse and challenging aspects of human agency such as self-reactiveness and self-reflectiveness (Bandura, 2006), which make agentic engagement more exceptional than other types of engagement. The challenging characteristics of agentic engagement potentially make it difficult for students to recognize the essential motivational role of agentic engagement in goal achievement (Reeve, 2012, 2013), and this, in turn, may somewhat weaken the positive relationship between agentic engagement and perceived instrumentality.

The results also showed that incremental language mindsets were positively and weakly related to graded performance, whereas entity language mindsets were negatively and weakly related to graded performance. These results were aligned with previous studies which showed that students’ mindsets were weakly related to their academic achievement (e.g. Bahník & Vranka, 2017). Most notably, the results demonstrated that the relationships between the four aspects of engagement and graded performance, as well as the relationship between perceived instrumentality and graded performance, were somewhat stronger than the relationships between language mindsets and graded performance. These correlation patterns highlight the possible mediating roles of perceived instrumentality and four aspects of engagement in the relationships between language mindsets and graded performance.

2 The mediating roles of perceived instrumentality and the four aspects of engagement

The results of the regression analyses revealed that entity beliefs about general language intelligence significantly, yet negatively predicted behavioral engagement and emotional engagement. Considering that students with entity beliefs about general language intelligence are less likely to persist in learning when they experience failure during the learning process (Dweck, 2006), it is highly possible that students with entity beliefs about general language intelligence would be reluctant to participate in class-related activities (Blackwell et al., 2007) and easily lose their interest and enjoyment (Fredricks et al., 2004) in language classes when they think that they may experience a failure. In turn, this phenomenon may lead students to degrade the importance of engagement with and perceived instrumentality of learning English. The reverse would be true for the students who hold incremental mindsets. Likewise, the results showed that incremental beliefs about second language aptitude significantly and positively predicted perceived instrumentality, incremental beliefs about general language intelligence significantly and positively predicted agentic engagement, and incremental beliefs about age sensitivity in language learning significantly and positively predicted emotional engagement and cognitive engagement. These results further underlined the motivational roles of incremental language mindsets in students’ engagement with language learning (Lou & Noels, 2016, 2017).

The results also showed that agentic engagement and perceived instrumentality significantly and positively predicted graded performance, whereas language mindsets, emotional engagement, behavioral engagement, and cognitive engagement did not. Considering the positive and significant relationship between perceived instrumentality and graded performance together with the significant and positive relationships between perceived instrumentality and the four aspects of engagement, it can be said that these results emphasize the prominent role of perceived instrumentality in student achievement (Greene et al., 2004; Özçetin & Eren, 2010) and motivation (Creten, Lens, & Simons, 2001). The prominent role of agentic engagement in explaining EFL students’ graded performance could be explained based on the fact that, when compared to other types of engagement, agentic engagement requires students to be more proactive and intentional during lessons (Reeve, 2012, 2013). Thus, it is highly probable that agentic engagement may considerably suppress the relationships between the other types of engagement and graded performance.

Most notably, the results of the SEM analysis confirmed that the relationship between incremental beliefs about second language aptitude and graded performance was fully mediated by perceived instrumentality. Additionally, the relationship between incremental beliefs about general language intelligence and graded performance was fully mediated by agentic engagement, and the relationship between perceived instrumentality and graded performance was partially mediated by agentic engagement. The full mediating role of perceived instrumentality in the relationship between incremental beliefs about second language aptitude and graded performance can be explicated based on the distinctive nature of the EFL students as attendees of an IEP. As mentioned previously, students with clear short-term and/or long-term goals are likely to focus on their progress in learning (Dweck et al., 2014) by considering the instrumentality of the content they are being exposed to (Miller & Brickman, 2004). Thus, it is highly possible that the EFL students who hold incremental beliefs about second language aptitude may conceive English classes as a prerequisite to achieving their professional/personal goals, and this, in turn, may motivate them to better consider the value of English classes in IEP contexts together with their future goals. In turn, this may positively relate to the EFL students’ graded performance due to the major aim of IEP (i.e. increasing students’ EFL proficiency).

Following this line of reasoning, the full mediating role of agentic engagement in the relationship between incremental beliefs about general language intelligence and graded performance can also be understood because considering English classes as a prerequisite to attain professional/personal goals may motivate EFL students to engage with English classes in a comprehensive manner, and this, in turn, may positively relate to their graded performance (Reeve & Lee, 2014). These explanations may also shed light on the partial mediating role of agentic engagement in the relationship between perceived instrumentality and graded performance because when EFL students consider that the existing language-related activities and course content are relevant to their future goals, they potentially engage more intentionally and proactively with the lessons throughout the IEP, which, in turn, may at least to some extent influence their graded performance.

3 Limitations and directions for future studies

This study has several limitations that would provide a basis for future studies. First, although the sample of the study was large enough to examine the relationships between the research variables, future studies should be conducted based on larger samples (n > 1000) of EFL students from various universities in order to investigate the generalizability of the current results.

Second, the correlational design of the study does not allow one to infer causality regarding the relationships between the research variables. Thus, experimental studies in which EFL students’ language mindsets are manipulated through effective methods (e.g. explaining the benefits of adopting incremental mindsets for effective language learning) may enable researchers to make causal inferences regarding the relationships between language mindsets and, for example, perceived instrumentality.

Third, with the exception of graded performance, the data were obtained based on self-report measures, indicating that the current results are more or less open to the effects of social desirability. Therefore, social desirability could be controlled in future studies. As the results of the preliminary analyses revealed, the possible effects of EFL students’ initial foreign language proficiency levels on their graded performance should also be considered in future studies.

Fourth, the possible effects of instructors’ language mindsets on EFL students’ language mindsets were not considered in the current study. However, by considering the mentioned effects of instructors’ language mindsets, future studies may uncover the effects of probable consonances/dissonances between the instructors’ and students’ language mindsets on the graded performance of EFL students.

Fifth, in the present study, the perceived instrumentality has been examined by using a single-factor scale (i.e. the PIS). However, as mentioned in Section II.3, it can also be examined by using multi-factor scales including, for example, both exogenous and indigenous instrumentality (Husman et al., 2004). Thus, future studies in which the mediating roles of diverse types of instrumentality in the relationships between language mindsets and graded performance are investigated may provide more comprehensive results regarding the current topic.

Finally, longitudinal studies are also needed to examine whether the possible changes in students’ language mindsets, perceived instrumentality, and four aspects of engagement significantly explain the changes in graded performance. Such studies may enlarge the current results by enabling researchers to examine possible reciprocal relationships between language mindsets, perceived instrumentality, four aspects of engagement, and graded performance.