CBA refers to any assessment embedded in classroom instruction, either explicit or implicit, regarded as in opposition to traditional large-scale tests external to the classroom (Turner, 2012). Technically, it is viewed as a process that “teachers and students use in collecting, evaluating, and using evidence of student learning” (McMillan, 2013, p. 1). In this sense, CBA is a process-based practice rather than a simple assessment instrument. CBA is broadly conceived as serving two purposes: summative assessment (SA) and formative assessment (FA), often referred to as assessment of learning and assessment for learning, respectively (Brookhart, 2004; Wu et al., 2021c). SA is primarily designed to ‘elicit evidence regarding the amount or level of knowledge, expertise or ability’ (Wiliam, 2001, p. 169) for administrative or reporting purposes, with a focus on scores. FA, on the other hand, focuses on using assessment to promote student learning (Black and Wiliam, 2009). It emphasizes the provision of descriptive feedback (rather than scores) as evidence about student learning, through which students are informed about their strengths and weakness, and scaffolded to close the gap between their current and desired performance (Sadler, 1989). Students, in particular, play a vital role in the assessment process, whereby they have an awareness of learning goals and assessment criteria, and actively engage themselves in self-and peer assessment (Pryor and Crossouard, 2008). Assessment conducted in this way is seen to have great potential in developing students’ capacity to self-regulate their learning (Assessment Reform Group, 2002; Panadero et al., 2018). Notwithstanding the two different purposes for assessment, the boundary between SA and FA is not as clear cut as usually represented because the same assessment information can be used for different purposes at different times (Rea-Dickins, 2007). It is how the information is used which provides the key distinction. In some cases, SA can be used formatively and FA can be used to serve a summative function (Dixson and Worrell, 2016; Dolin et al., 2018). Summative tests, for example, which typically take place at the end of a unit or term to record learning attainments, can be used to adjust teaching and improve students’ learning in the future. FA practices, such as self-and peer assessment, have the potential to generate a lot of data about students’ progress, which can be recorded and further used for summative reporting purposes. More recently, scholars have pointed out that, in an assessment for learning culture, all assessments, including those for administrative and reporting purposes, can and need to be implemented with a central aim of facilitating learning (Black and Wiliam, 2018; Davison, 2019).

Researchers have put forward a number of frameworks casting light on how to translate CBA for learning principles into practice. These include Black and Wiliam (2009) five strategies of formative assessment, Hill and McNamara (2011) framework for CBA processes, Davison and Leung (2009) framework for teachers-based assessment, and more recently learning-oriented assessment related frameworks (Carless, 2011; Turner and Purpura, 2016). Despite their differences, all the frameworks emphasize three instructional processes that are critical to student learning improvement, i.e., where learners are going, where they are in their learning, and how to get there (Wiliam and Thompson, 2008). Davison and Leung (2009) framework is particularly operational as it conceptualizes CBA into a cycle of four steps, addressing the process-based feature of CBA. It provides a working approach for the analysis of teachers’ classroom-embedded assessment practice, and thus was selected as the analytical framework for the present study.

The first step of Davison and Leung (2009) framework, planning assessment (PA), focuses on clarifying learning goals and assessment criteria to ensure that students have an awareness of where they are going. Selecting assessment methods that suit the needs of students is also an important element of this step. The second step, collecting learning evidence, is to collect instructionally tractable evidence of student learning, which can be achieved through various methods, such as spontaneous assessment opportunities (SAO), planned assessment opportunities, and formal assessment tasks (FATs; Hill and McNamara, 2011; Turner and Purpura, 2016). In the next step, making professional judgments (MPJ), collected learning evidence is interpreted in relation to established standards (i.e., criterion-referenced assessment) or students’ progress made over time (i.e., pupil-referenced assessment). The final step, providing appropriate feedback, moves students forward through descriptive feedback that enables students to recognize their learning gap and monitor their own learning to close such gap. These four steps of CBA are interrelated with one another, rather than separated, and need to be implemented in a holistic way to fulfil the overriding aim of supporting student learning.

Assessing YLLs warrants a great deal of attention because of their unique age-related characteristics, which are generally categorized into three types: growth, literacy and vulnerability (McKay, 2006). First, YLLs are undergoing cognitive, social and emotional, as well as physical growth, which is nonlinear and dynamic (Berk, 2017). For example, they have short attention span, usually love physical activities, and are developing social awareness and a sense of self-esteem, which generates a strong demand for teachers to select or conduct appropriate assessment tasks (Patekar, 2021). Second, compared with older or adult learners, YLLs are still developing literacy skills in their first language when they are learning a foreign language, and this can have both conflicting and constructing influence on their literacy skills in the foreign language (Butler, 2016). Third, YLLs are particularly vulnerable to adults’ praise and criticism concerning their assessment performance. Their experience with assessments can have a long-lasting impact on their learning motivation, self-confidence, and learning outcome (Butler, 2019).

Against the backdrop of the above unique characteristics of YLLs, researchers have proposed principles for effective assessment of YLLs (Edelenbos and Vinjé, 2000; Hasselgreen, 2000; Cameron, 2001). Some key principles are summarized as follows: assessment tasks should fit with YLLs’ learning experience, reflecting those activities conducted in class; traditional achievement tests should not be viewed as the only form of assessment, instead, alternative forms of assessment, such as student portfolios, self-and peer assessment, need to be promoted for YLLs; assessment processes should help YLLs to monitor their language learning and develop their self-regulation abilities. CBA has been recognized as a practical solution to address such principles. It incorporates clear clarification of learning goals, multiple assessment methods, and quality feedback toward learning goals. Through CBA, YLLs can become aware of goals, develop positive learning attitudes, and gradually develop a sense of control over their own learning (Butler, 2021).

Given that assessment of YLLs is a newly emerging field (Hasselgreen, 2012), many areas have been underexplored, among which is CBA for YLLs, in particular. Limited research on this area has, however, highlighted that the implementation of CBA in young learner contexts is less than straightforward. For example, teachers of young learners fail to make assessment criteria explicit (Hild and Nikolov, 2011), frequently apply traditional assessment methods (e.g., objective tests; Prošić-Santovac et al., 2019; Yan et al., 2021), provide mostly evaluative feedback (e.g., marks) (Brumen et al., 2009), and mainly use assessment for summative purposes (Rixon, 2016). Moreover, research suggests that much can constrain teachers’ CBA enactment, such as teachers’ low assessment literacy (Vogt and Tsagari, 2014), lack of training in assessment (Patekar, 2021), limited opportunities for professional development (Lee et al., 2019), tight curriculum content (Mak and Lee, 2014), and an examination-oriented culture (Kaur, 2021). It could thus be said that although CBA is regarded as an effective approach to the improvement of YLLs’ learning, the implementation of CBA in local contexts is a challenging endeavor.

Despite the aforementioned research efforts into the enactment of CBA in young learner contexts, teachers’ beliefs and practices about CBA for YLLs, as well as their relationship, remains largely under-investigated. Teacher beliefs, used interchangeably with teacher conceptions and teacher cognition, generally refer to “the unobservable cognitive dimension of teaching—what teachers know, believe and think” (Borg, 2003, p. 81). Teachers hold beliefs about various aspects of their work like teaching, learning, teachers, students, curriculum and materials (Borg, 2001), and their beliefs exist as a system wherein some core beliefs are stable and hard to change, while others are peripheral (Pajares, 1992). Studies have shown that teachers’ beliefs can exert a powerful influence on their instructional decisions (Burns et al., 2015; Li, 2020), though such beliefs may not always be reflected in their practices (Johnson, 1992; Dixon et al., 2011). Multiple factors, such as contextual complexities (e.g., class size, time constraints, authority’s influence), teachers’ teaching experience, and students’ needs, can determine the extent to which teachers can act according to their beliefs (Phipps and Borg, 2009; Roothooft, 2014). In this study, teachers’ CBA beliefs refer to teachers’ views toward the processes of CBA, including PA, collecting evidence, making professional judgments, and providing feedback, and teachers’ CBA practices are described as the enactment of these processes.

Research on L2 teachers’ beliefs and practices related to CBA has produced mixed results. There is evidence of a powerful effect that teachers’ beliefs have on the way they implement CBA practices (e.g., Mui So and Hoi Lee, 2011; Wang, 2017; Prošić-Santovac et al., 2019; Wu et al., 2021b). Mui So and Hoi Lee (2011), for instance, investigated English-as-a-second-language (ESL) secondary school teachers’ beliefs and practices of CBA for learning and found that their assessment practices consistently reflected their beliefs about the purpose of CBA. Similar congruence between teachers’ CBA beliefs and practices was identified in EFL contexts (Zhou and Deneen, 2016; Wu et al., 2021c). However, more studies have reported discrepancies teachers’ CBA beliefs and practices (e.g., Xu and Liu, 2009; Chen et al., 2014; Gan et al., 2018; Nasr et al., 2018; Vattøy, 2020; Wang et al., 2020; Mäkipää, 2021). For example, in their study of two university EFL teachers’ enactment of CBA, Chen et al. (2014) found that the teachers expressed positive attitudes toward students’ involvement in the assessment process, whereas in their actual practice, they seldom engaged students in self-and peer assessment. More recently, Vattøy (2020) interviewed ten secondary EFL teachers in Norway and found misalignment between teachers’ beliefs and practices regarding formative teacher feedback. These studies have also shown that L2 teachers’ CBA beliefs and practices are affected by individual factors (e.g., students’ needs, core beliefs held by teachers and their teaching experience) and sociocultural factors (e.g., policy support, class size, time constraints, prescribed curriculum and assessment culture).

One significant gap emerges from the existing studies on L2 teachers’ CBA beliefs and practices is that a majority of the previous studies were concerned mainly with secondary and university teachers, and less space has been devoted to teachers of YLLs. More research is needed regarding teachers’ beliefs and practices pertinent to CBA in young learner contexts. As discussed earlier, YLLs set themselves apart from older learners due to their special age-related characteristics. While previous research has offered insights into how secondary and university language teachers perceive and enact CBA for older learners, whether teachers of YLLs manifest similar beliefs and practices about CBA remains unknown. Given that teachers’ beliefs are context-dependent (Yu et al., 2020), further investigation into teachers’ beliefs and practices about CBA in young learner contexts is essential. Another important gap is that previous studies on L2 teachers’ CBA beliefs and practices merely focused on one or two aspects of CBA, with an intensive discussion on self-and peer assessment and teacher feedback (e.g., Chen et al., 2014; Vattøy, 2020; Mäkipää, 2021). To address this gap, two recent studies, conducted by Wang et al. (2020) and Wu et al. (2021c) respectively, have attempted to investigated language teachers’ beliefs about CBA in a more comprehensive way. Nonetheless, both studies were contextualized within university EFL classrooms. Thus, the current study, situated in the Chinese context, aims to investigate primary school EFL teachers’ beliefs and practices relating to the holistic process of CBA, including PA, collecting evidence, MPJ, and providing feedback.

Our research instrument, the Primary School English Teachers’ CBA questionnaire, was developed in a larger study on the implementation of CBA for young learners in China (Yan, 2020). The questionnaire comprised four sections: Perceived-purpose Scale, Perceived-process Scale, Practice Scale, and Demographic Information. The present study focused on the Perceived-process Scale and the Practice Scale, which examined teachers’ beliefs about the processes of CBA and their self-reported CBA practices, respectively.

Following Dörnyei and Taguchi (2010) guidelines, we first identified potential constructs of the Perceived-process Scale and the Practice Scale with reference to Davison and Leung (2009) framework of CBA. As discussed earlier, four dimensions are included in this framework: PA, collecting learning evidence, making professional judgments and providing appropriate feedback. The dimension collecting learning evidence was designed in this study to emphasize three types of assessment methods: SAO, planned assessment opportunities and FATs. Given the importance of distinguishing descriptive feedback from evaluative feedback (Wiliam, 2010), both types of feedback were included in the dimension providing appropriate feedback. Thus, seven potential constructs were included in both the Perceived-process Scale and the Practice Scale: PA, using SAO, using planned assessment opportunities, using FATs, MPJ, providing descriptive feedback (PDF), and providing evaluative feedback (PEF).

Forty-four items were initially generated for both scales, most of which were drawn from the Beliefs about Assessment and Evaluation questionnaire (Rogers et al., 2007), the Classroom Assessment questionnaire (Cheng et al., 2004), the “Learning How to Learn” project’s questionnaire (James and Pedder, 2006), and the Assessment Practices Inventory (Zhang and Burry-Stock, 2003). Some items were generated on the basis of the semi-structured interviews conducted with three primary school EFL teachers who did not participate in the main study. The first researcher interviewed the three teachers individually, during which an open, comprehensive topic was discussed: What do you think about the purposes of teachers’ classroom assessment and what assessment practices do you employ when assessing your students? The teachers’ responses were used as a source for the item pool. For example, when asking about assessment practices, one teacher described that “I frequently check whether students have mastered what they learned in class through classroom tests, dictation and recitation.” This guided the writing-up of the questionnaire items such as “Teachers collect evidence of learning through classroom tests” and “Teachers collect evidence of learning through dictation.” Two 6-point Likert scales were established to study teachers’ beliefs about CBA in the Perceived-process Scale (1 = Not important at all, 2 = Not important, 3 = Somewhat important, 4 = Important, 5 = Very important, 6 = Completely important), and their self-reported CBA practices in the Practice Scale (1 = Never, 2 = Very rarely, 3 = Rarely, 4 = Occasionally, 5 = Frequently, 6 = Always), respectively.

Finally, to examine the content validity, two experts and a group of postgraduates were invited to provide feedback on the content relevance, clarity and comprehensiveness of the two scales, based on which two items were dropped, retaining 42 items, and some items were revised. The items were originally generated in English, and then translated from English to Chinese by the first author and a doctoral student, using a back-translation method (Nunan and Bailey, 2009). The questionnaire was then piloted with 26 primary school English teachers, who did not participate in the main study and provided comments on questionnaire clarity and administration procedures. Modifications were made and a finalized questionnaire was obtained (Appendix A).

The first author collected the questionnaire data. A convenience sampling strategy that is commonly used in L2 research (Dörnyei, 2007) was used to collect questionnaire responses from teachers who were teaching EFL at primary schools in China. Specifically, the target participants were primary school EFL teachers from one municipality and one province in China, to whom the first researcher had access. To approach these participants, an information sheet was first sent to six primary school EFL teaching advisors in the two places, who had the responsibility of providing guidance to teachers’ daily teaching and were in charge of all teachers in their own districts. An online questionnaire invitation was then sent out to primary school teachers through the advisors. A total number of 312 questionnaires were received. Of those, 117 questionnaires with an obvious response set (i.e., almost the same answers for all the items) were excluded, leaving 195 valid responses.

As Table 1 shows, a majority of the teacher participants were female (96.9%, n = 189), and most teachers were aged under 40 years (72.9%, n = 142). Over three quarters of the teachers held a bachelor’s degree (78.5%, n = 153), with a small minority holding a master’s degree and a degree lower than the Bachelor (21.5%, n = 42). About half of the participants had 6 to 20 years’ experience of teaching EFL to primary school students (52.3%, n = 102); 33.8% (n = 66) had less than 5 years and 13.5% (n = 27) had over than 20 years. The number of teachers from public schools (88.2%, n = 172) was larger than that from private schools (11.8%, n = 23). Of the participants, 69.2% (n = 136) taught lower grade levels (Grades 1, 2, 3 and 4), and a similar number taught higher grade levels (Grades 5 and 6) (64.1%, n = 125). Most participants had completed a course on assessment or received training in assessment (64.6%, n = 126).

Questionnaire data was analyzed using SPSS 24. Missing data, outliers, and normality distribution were checked first. Exploratory factor analysis (EFA) was then conducted to explore the underlying constructs of the Perceived-process Scale and the Practice Scale. Assumptions were checked through Kaiser-Meyer-Olkin’s (KMO) Test of Sampling Adequacy and Bartlett’s Test of Sphericity. Factors were extracted using principal axis factoring with non-orthogonal rotation (Direct Oblimin, δ = 0) solution. A factor loading with a minimum absolute value of 0.30 was required. The results indicated a 37-item seven-factor solution and a 31-item six-factor solution for the Perceived-process Scale and the Practice Scale, respectively, (underlying constructs of the perceived-process scale and practice scale).

To answer RQ1 regarding teachers’ beliefs about CBA, descriptive analysis was conducted to obtain a general understanding of teachers’ beliefs about the seven factors. And repeated-measures ANOVA was calculated to examine the differences of teachers’ beliefs among different factors. The same analyzes were run to answer RQ2 regarding teachers’ self-reported CBA practices. To answer RQ3 regarding teachers’ belief-practice relationship, correlation analysis and multiple regression analysis were carried out. Correlation analysis was performed to examine whether there existed a direct relationship, and the latter was performed to explore how well teachers’ CBA beliefs could predict their practices.

Descriptive data of teachers’ beliefs of the CBA processes are displayed in Table 4. The seven factors derived from the Perceived-process Scale all had a mean score higher than 4.0 (“Important”), indicating that the teachers believed that the various processes of CBA were important for the enhancement of student learning. Specifically, they agreed that PA and PDF were especially valuable, judging from the mean scores that exceeded 5.0 (“Very important”).

One-way repeated-measures ANOVA was conducted to examine differences among teachers’ beliefs regarding the seven factors. The assumption of normal distribution was satisfied given that the values of skewness (from −0.421 to 0.209) and kurtosis (from −0.251 to 1.037) fell within the cut-off values of|3.0|and|8.0|respectively (Kline, 2011). Mauchly’s Test showed that the assumption of sphericity had been violated, x² (20) = 68.888, p < 0.001, so Greenhouse–Geisser of Huynh-Feldt (ε = 0.909) was applied to adjust degrees of freedom. The results of ANOVA showed that there were significant differences in the importance attached to the CBA processes, F (5.289, 1026.070) = 42.694, p < 0.001, η² = 0.18. Follow-up pairwise comparisons with Bonferroni corrections indicated that all pairwise differences were significant (p < 0.05) excepted four paired comparisons, namely PA and PDF (p > 0.05), SAO and FAT (p > 0.05), SIA and PEF (p > 0.05), FAT and MPJ (p > 0.05). This suggested that the teachers considered PA and PDF to be the most important CBA processes in promoting student learning. As for the specific assessment methods for learning evidence collection, they placed greater emphasis on SAO (e.g., teacher questioning, observations) and FAT (e.g., classroom tests, textbook exercises, recitation), whereas a relatively lower preference was shown for SIA (e.g., self-assessment and peer assessment). They also believed strongly in the value of MPJ (e.g., comparing learning evidence against pre-set goals). Comparatively, from teachers’ perspective, SIA and PEF were the least important CBA processes.

Table 5 presents the descriptive data of teachers’ self-reported CBA practices. The mean scores for PA, SAO, MPJ and PDF exceeded 5.0 (“Frequently”) and those for SIA and FAT exceeded 4.0 (“Occasionally”). The results revealed that, on average, the teachers reported frequent-use of PA, SAO, MPJ and PDF and occasional-use of SIA and FAT.

One-way repeated measures ANOVA was applied to examine the differences among the six factors. The data met the assumption of normal distribution (the values of skewness ranging from −0.285 to 0.150, and the values of kurtosis from −0.704 to 0.467), but the assumption of sphericity was violated, x² (14) = 79.825, p < 0.001.Therefore, Greenhouse–Geisser of Huynh-Feldt (ε = 0.877) was applied. Significant differences were identified in the frequency of different CBA practices, F (4.277, 829.715) = 66.582, p < 0.001, η² = 0.256. Follow-up pairwise comparisons with Bonferroni corrections showed that all pairwise differences were significant (p < 0.05) except six paired comparisons, namely, SAO and FAT (p > 0.05), SAO and MPJ (p > 0.05), SAO and PDF (p > 0.05), FAT and MPJ (p > 0.05), FAT and PDF (p > 0.05), MPJ and PDF (p > 0.05). The results revealed that, according to the teachers’ self-report, PA was the most frequently used CBA practice. As for collecting learning evidence, the teachers frequently used assessment methods of SAO (e.g., oral questioning and observations) and FAT (e.g., classroom tests and recitation tasks), while SIA (e.g., self-assessment and peer assessment) was less frequently used. Meanwhile, they reported frequent use of MPJ (e.g., made judgments of students’ performance against learning objectives or their previous performance) and PDF (e.g., providing feedback to students to identify learning strengths and weaknesses). In general, SIA was the least frequently used CBA practice by the teachers.

In summary, Chinese primary school EFL teachers in this study showed strong positive attitudes toward CBA, perceiving that PA and PDF were the most important assessment processes. While they emphasized the value of using multiple assessment methods, they placed more importance on SAO and FATs than student-involving assessment opportunities. MPJ were also considered as imperative for improving student learning. PEF, as opposed to PDF, gained the least popularity among teachers. A similar pattern was identified in their self-reported CBA practices, as they reported the most frequent practice of PA, multiple use of assessment methods to collect students’ learning evidence (with a heavy reliance on SAO and FAT), as well as frequent practice of MPJ and PDF. Such a similar pattern seemed to indicate alignment between teachers’ beliefs about CBA and their assessment practices.

To examine the relationship between teachers’ beliefs and their self-reported practices related to CBA, correlation analysis was first conducted. As seen in Appendix B, each of the seven factors of teachers’ beliefs about CBA was positively and significantly correlated with the six factors of teachers’ self-reported CBA practices. Thus, six multiple regression analyzes using each of the six self-reported CBA practices as dependent variables were conducted, with the seven factors of teachers’ beliefs about CBA being set as predictors (see Table 6). The seven factors of teachers’ beliefs were found to have a significant effect on teachers’ CBA practices.

First, the results of multiple regression showed a significant model for the practice of PA, R² = 0.311, adjusted R² = 0.286, df = (7,187), F = 12.078, p < 0.001. The PA practice was predicted by the factors regarding teachers’ beliefs about PA (β = 0.369, p < 0.001) as well as PDF (β = 0.273, p < 0.01). Second, the results showed a significant model for the practice of SAO, R² = 0.331, adjusted R² = 0.306, df = (7,187), F = 13.225, p < 0.001. Such practice was predicted by teachers’ beliefs about SAO (β = 0.349, p < 0.001). Third, a significant model was identified for the practice of SIA, R² = 0.232, adjusted R² = 0.193, df = (7,187), F = 6.398, p < 0.001. Likewise, this practice was predicted by the factor of teachers’ beliefs about SIA. Fourth, a significant model was found for the practice of FAT, R² = 0.235, adjusted R² = 0.206, df = (7,187), F = 9.191, p < 0.001. In the same vein, the practice was predicted by teachers’ beliefs about FAT (β = 0.417, p < 0.001). Regarding the practice of MPJ, a significant model was also identified, R² = 0.299, adjusted R² = 0.273, df = (7,187), F = 11.410, p < 0.001). Such practice was predicted by teachers’ beliefs of MPJ (β = 0.253, p < 0.01) and PDF (β = 0.201, p < 0.05). Finally, a significant model was identified for the practice of PDF, R² = 0.376, adjusted R² = 0.352, df = (7,187), F = 23.044, p < 0.001), showing that the factor of teachers’ beliefs of PDF was a significant predictor (β = 0.491, p < 0.001).

Overall, the above results revealed strong relationships between Chinese primary school EFL teachers’ beliefs about CBA and their assessment practices. Their beliefs regarding PA, SAO, SAI, FAT, MPJ and PDF were found to be important predictors of the related assessment practices.