Strategic Self‐Regulation for Speaking English as a Foreign Language: Scale Development and Validation

Teaching English to speakers of other languages is challenging, particularly in an environment without sufficient input and output to support learners’ English as a foreign language (L2) development (Uztosun, 2020, 2021). For most students in non-target language environments, they rely primarily on classroom teaching to learn another language. However, classroom-based L2 learning is far from enough for individuals’ continuous L2 speaking ability development. In light of the positive correlations between self-regulated learning (SRL) and L2 achievements (Oxford, 2017), a more feasible approach to sustaining the development of L2 speaking skills is perhaps to foster learners’ SRL ability. As a result, learners with competent SRL strategies may become more successful in L2 learning and thus their L2 speaking skills. Although an increasing number of studies have examined SRL strategies in L2 settings, research within this line of inquiry centers more on SRL writing and reading strategies (Teng & Zhang, 2021). There is a lack of understanding of other L2 strategic skills such as speaking and listening, particularly SRL speaking strategies. To date, Uztosun’s study (2020) is perhaps the earliest attempt to investigate into SRL speaking strategies. Informed by Pintrich’s (2004) SRL model, Uztosun (2020) developed a 20-item scale for measuring self-regulated motivation for improving L2 English speaking. The scale contains four factors, including task value activation, regulation of learning environment, regulation of affect, and regulation of classroom environment. However, Uztosun’s (2020) SRL speaking strategies mainly focus on self-regulated motivation without taking a multidimensional approach to examining learners’ SRL speaking strategies (Oxford, 2017). As evidenced in the literature, SRL “addresses the interactions of cognitive, motivational, emotional, and contextual factors rather than their isolated contributions to learning” (Teng & Zhang, 2021, p. 3; Schunk & Greene, 2018). Furthermore, L2 strategy use is a metacognitive, motivational, and behavioural active process (e.g., Teng & Zhang, 2021; Zimmerman, 1986). Therefore, it would be more insightful to explore SRL speaking strategies from a multidimensional perspective. To bridge the gap, this study aims to develop a more comprehensive scale for measuring SRL speaking strategies by drawing on Oxford’s (2017) strategic self-regulation (S2R) model. According to Oxford’s S2R model, there are cognitive, motivational, social, and affective domains of SRL strategies, and the four S2R domains contain upper-level metastrategies and lower-level concrete strategies. Specifically, metastrategies refer to individuals’ paying attention to, planning for, organizing, monitoring, and evaluating their cognition, motivation, contexts, communication, and culture (CCC), and affect. There are four categories of metastrategies, including metacognitive, metamotivational, metasocial, and meta-affective strategies. Under the guide of metastrategies, learners will consciously and/or unconsciously develop their detailed cognitive, motivational, social, and affective SRL strategies. For example, learners equipped with adequate SRL speaking strategies may pay careful attention to their speaking development by setting up goals, creating conducive environments, and monitoring and evaluating their cognitive, motivational, social, and affective strategy sets. Please note that the present study focuses on the lower-level concrete SRL strategies rather than the upper-level abstract metastrategies in the process of developing strategic self-regulation for speaking English as a foreign language scale (S2RS-EFL). One of the main reasons is that the nature of a scale development is to establish a detailed, robust, and feasible instrument that can be operationalized to measure latent variables/factors. It is, therefore, more practical to examine those lower-level concrete SRL speaking strategies. Additionally, the developed lower-level or fine-grained strategies will be more informative, serving as an explicit guide for learners to diagnose and navigate their use of SRL speaking strategies. A total of 507 first year college students voluntarily participated in the study. Thirteen participants were deleted from the data pool given that they provided the same answer for all questionnaire items. Among the valid participants, 300 were males and 193 were females. Their ages ranged from 17 to 23 was an average of 18.27 years old (SD = .65). Their average English score was 132.86 out of 150 (SD = 6.50), suggesting that they had a good command of English. The scores were collected from their national college entrance examination (NCEE, or gaokao/高考 in Chinese) for the English subject. The NCEE is a standardized examination of academic performance in mainland China. The instrument used for data collection was adapted from Cohen and Chi’s (2006) speaking strategy inventory and Teng and Zhang’s (2016) SRL writing strategies questionnaire in reference to Oxford’s (2017) cognitive, motivational, social, and affective domains of S2R model (see Appendix 1). Specifically, the adapted questionnaire included two sections: a background information section and a scale of strategic self-regulation for speaking English as a foreign language (S2RS-EFL). The initial version of the S2RS-EFL was composed of 52 SRL speaking strategies in a 7-point Likert scale ranging from 1 = not true of me at all to 7 = very true of me. The cognitive domain of SRL speaking strategies includes cognitive processing (4 items), remembering (6 items), assistance seeking (4 items), idea planning (3 items), goal-based monitoring and evaluation (4 items), and self-reflection (3 items). The motivational domain of SRL speaking strategies includes interest enhancement (3 items) and motivational self-talk (7 items). The social domain of SRL speaking strategies includes interactional practice (4 items), peer learning (3 items), feedback management (4 items), and environment control (3 items). The affective domain of SRL speaking strategies includes anxiety control (4 items). To ensure the content validity, experts on self-regulated learning were invited to go through the initial pool of SRL speaking strategies in the S2RS-EFL. The screened data were first divided into two random samples for exploratory factor analysis (EFA) and confirmatory factor analysis (CFA), respectively. The homogeneity of the two samples was examined by a Chi-square test of independence (in terms of sex), independent samples t-tests (in terms of age and L2 English scores) and Mann-Whitney U tests (in terms of the Likert scale questionnaire of SRL speaking strategies). Afterwards, one sample of 244 participants was subjected to the EFA to extract underlying variables of S2R speaking strategies. Another sample of 250 participants was subjected to the CFA to confirm the measurement model of SRL speaking strategies extracted in the EFA stage. Indices including the ratio of chi-square to its degree of freedom (χ2/df) < 3, the root mean square error of approximation (RMSEA) <.08, the standardized root mean square residual (SRMR) <.08, and the comparative fit index (CFI) >.90 were adopted for model fit examination (Hair et al., 2014; Hu & Bentler, 1999). In terms of homogeneity, the Chi-square test of independence (χ2(1) = 1.02, p = .31) found no sex difference between the EFA (Male = 143, Female = 101) and CFA (Male = 157, Female = 93) samples. The subsequent independent samples t-tests showed no significant differences between the two samples’ age (t(492) = .005, p = .42) and L2 English scores (t(492) = 1.26, p = .52). Specifically, the average ages of the two samples were both 18.27 (SDEFA = .58, SDCFA = .70). The average L2 English scores were 133.92 (SD = 7.52) for the EFA sample and 132.49 (SD = 5.81) for the CFA sample. Lastly, the Mann-Whitney U tests did not detect any significant differences between the two samples’ self-reported use of the proposed thirteen SRL speaking strategies with p values ranging from .07 to .98. The medians of the two samples’ SRL speaking strategies ranged from 2 (not true of me) to 6 (true of me) and the mean scores from 2.30 to 5.65. In terms of normality, the skewness values of the two samples were between −1.33 and 1.27, and the kurtosis values were between −1.18 and 2.72. They were smaller than the respective cutoff values of |3| and |10| for skewness and kurtosis, suggesting that the univariate normalities of the participants’ responses were supported (Kline, 2016). EFA was performed by using maximum likelihood estimation with promax rotation. The EFA results showed that Kaiser-Meyer-Olkin (KMO) value was .879, suggesting the sample size was adequate for the analysis. After the examination of the item loadings, three items of the peer learning variable and one item of the interactional practice variable were found to have cross-loadings or low loadings. Moreover, the initially proposed remembering variable and idea planning variable were grouped in the same construct but with high item loadings. To minimize the change of the original scale, we only deleted the five items with loading issues. A second round of EFA was performed after deleting the five items. The results showed that the rest 47 items loaded high on 11 variables (see Table 1). Specifically, the 11 extracted SRL speaking strategies were cognitive processing (4 items, α = .744), remembering and planning (9 items, α = .892), assistance seeking (4 items, α = .700), goal-based monitoring and evaluation (4 items, α = .854), and self-reflection (3 items, α = .851) from the cognitive domain. Interest enhancement (3 items, α = .818) and motivational self-talk (7 items, α = .935) from the motivational domain. Interactional practice (3 items, α = .801), feedback management (4 items, α = .853), and environment control (3 items, α = .857) from the social domain. Anxiety control (4 items, α = .775) from the affective domain. Further, the Cronbach α reliability results showed that the overall reliability of the proposed SRL speaking strategies scale (i.e., S2RS-EFL scale) and the reliabilities of the 11 subscales or variables all met the .70 threshold (Pallant, 2016). CFA with maximum likelihood estimation was performed to cross-validate the extracted 11-factor construct of S2R speaking strategies in the EFA stage. The CFA results revealed that the model fit indices were satisfactory with χ2/df = <2.028, RMSEA = .046, SRMR = .055, and CFI = .917 (see Figure 1 for standardized results). Additionally, standardized factoring loadings were all higher than the .30 threshold, suggesting that items had salient loadings on corresponding factors (Raykov & Marcoulides, 2008). The CFA results also confirmed the discriminant validities of the 11 SRL speaking strategies. As shown in Figure 1 and Table 2, the correlation coefficients between factors ranged from r = .09 between goal-oriented monitoring and evaluation (GME) and assistance seeking (AS) and between AS and self-regulation (SR) to r = .71 between anxiety control (AC) and interest enhancement (IE), suggesting reasonable discriminant validity (Kline, 2016). In other words, although the 11 SRL speaking strategies are correlated, they are distinct constructs measuring what they are supposed to measure. This study reported on the development and validation of the strategic self-regulation for speaking English as a foreign language (S2RS-EFL) scale with its theoretical framework underpinned by Oxford’s (2017) S2R model. The EFA and the CFA analyses resulted in an 11-factor S2RS-EFL scale with 48 items covering cognitive, motivational, social, and affective domains of SRL speaking strategies. In terms of the cognitive domain of SRL speaking strategies, there are cognitive processing, remembering and planning, assistance-seeking, goal-based monitoring and evaluation, and self-reflection. Specifically, cognitive processing refers to learners’ intentional check of their speech production. Remembering and planning refers to learners’ attempts to memorize useful words and expressions for speech planning. Assistance seeking refers to “the ways that learners employ when coping with problems during speaking” (Sun et al., 2016, p. 599). Goal-based monitoring and evaluation refers to learners’ efforts to set up goals to monitor and evaluate their progress in L2 speaking (Teng & Zhang, 2016). Self-reflection refers to learners’ endeavours to review their L2 speaking for future improvement. In terms of the motivational domain of SRL speaking strategies, there are interest enhancement and motivational self-talk. Specifically, interest enhancement refers to learners’ efforts to make their L2 speaking improvement more enjoyable. Motivational self-talk refers to learners’ attempts to encourage themselves to improve their L2 speaking. In terms of the social domain of SRL speaking strategies, there are interactional practice, feedback management, and environment control. Interactional practice refers to learners’ attempts to interact with others to learn and to communicate. Feedback management refers to learners’ intention to rely on others’ comments and suggestions to improve their L2 speaking. Environment control refers to learners’ endeavors to create appropriate learning environments to improve their L2 speaking. In terms of the affective domain of SRL speaking strategies, there is only one variable - anxiety control which refers to learners’ efforts to regulate their anxiety for better L2 speech performance. One possible explanation for the limited number of affective SRL speaking strategies is that the strategies in the present study are mainly based on the literature without considering learners’ actual beliefs on affective S2R speaking strategies. Future research may bridge the gap by examining the construct of affective SRL speaking strategies. The present study has some theoretical, practical, and pedagogical implications. Theoretically, the development of the S2RS-EFL scale, to some extent, render evidence to Oxford’s (2017) S2R model that cognitive, motivational, social, and affective metastrategies/strategies are the four domains needed for effective learning. The CFA results of the present study also support Oxford’s (2017) argument that the four domains of strategies are distinctive and yet correlated. Practically, the S2RS-EFL scale can be adopted to measure students’ use of self-regulated cognitive, motivational, social, and affective speaking strategies. As a result, the measurement may help students realize the shortcomings of their SRL speaking strategies use. Consequently, students with more knowledge and awareness of SRL speaking strategies may be more willing to adjust their use of strategies to achieve their learning goals. Pedagogically, the S2RS-EFL scale may be insightful for teachers’ classroom instruction. With reference to the self-reported S2RS-EFL results of their students, teachers can intentionally adjust their teaching to enhance their students’ use of SRL speaking strategies. As Wolters and Benzon (2013) pointed out, “knowing what strategies are preferred or used most often by students within more authentic academic contexts provides insight into which ones might best be used as the target of instructional interventions” (p. 201). There are limitations of the present study that can be informative for future research. First, a self-reported questionnaire might not be able to establish a full profile of learners’ use of SRL speaking strategies in reality, particularly when learners fail to realize and/or recall strategies that they have used. Future research is recommended to use different methods for data collection, such as stimulated recall after completing a task, reflection journals, and interviews. Second, the participants in the study were university freshmen only in China. More research is needed with different populations to generalize the findings of the present study. Third, this study only focuses on the construct validity of S2RS-EFL scale. Future research may consider examining the correlations between these strategies and speaking proficiency to check the predictive validity of the instrument. The authors declare that there is no conflict of interest regarding the publication of this paper. This work was supported by the National Social Science Foundation of China under Grant 19CYY008 and the Fundamental Research Funds for the Central Universities. The author also would like to thank anonymous reviewers for their constructive feedback on the paper. Peijian Sun is an Associate Professor of Applied Linguistics at Zhejiang University. His research focuses on L2 teaching and learning, teacher education, and educational technology. His publications have appeared in various journals such as Computer Assisted Language Learning, TESOL Quarterly, and System. Email: [email protected].