One seminal study on oral CF was carried out by Lyster and Ranta (1997), identifying six types of oral CF: recasts, explicit correction, metalinguistic clues, repetition, elicitation, and clarification requests. According to whether the correct form was directly provided or withheld, they further categorized these types of CF into two broad groups: reformulations (i.e., recasts and explicit correction), and negotiation of forms (i.e., elicitation, metalinguistic clues, clarification requests, and repetition), also referred to as prompts (Lyster, 2004). Another way to segment these CF types was given by Ellis et al. (2006), who classified them into explicit and implicit types according to whether the learner was drawn to their error overtly. In this segmentation, implicit CF constitutes clarification requests, repetition, recasts, and elicitation, while explicit CF includes metalinguistic clues and explicit correction.

Despite the existence of different classifications, Lyster and Ranta’s taxonomy has been widely used to code CF. To date, extant CF research falls mainly into two paradigms: observational and experimental (Li and Vuono, 2019). Research in the observational line has focused on describing CF patterns in terms of the distribution of different types of CF and their links to learner uptake (Brown, 2016; Wang and Li, 2020), while experimental research has emphasized CF efficacy in regards to whether CF is conducive to L2 learning and which type is most effective (Li, 2010; Lyster et al., 2013; Nassaji, 2017). In summary, both lines of research have shown a clear tendency for the efficacy of CF on L2 learning and also noted a host of mediating factors contributing to the patterns and efficacy of CF (Nassaji and Kartchava, 2020), of which is teacher and student preferences for CF.

Early research on teachers’ and students’ preferences for CF was mainly conducted with broader surveys of general beliefs about L2 teaching and learning, using Likert-type questionnaires that included only a few items related to CF (Schulz, 2001; Davis, 2003; Brown, 2009). However, more recent research has focused exclusively on CF beliefs in terms of attitudes, perceptions and preferences, the results of which have shown that teachers’ and students’ beliefs do not always align (Lee, 2013; Kaivanpanah et al., 2015; Roothooft and Breeze, 2016). For instance, Ha and Nguyen (2021) found that students are disposed toward having CF for all error types and desired to use peer correction during class. However, teachers were found to have less confidence in peer correction and more reluctance in giving CF. There are also discrepancies between teachers’ and students’ beliefs concerning CF timing, as Ha et al. (2021) reported that Vietnamese students appreciated immediate CF while their teachers were less favorable about this CF techniques out of concern for students’ emotional state and disruption of CF on the flow of students’ speech.

In addition, discrepancies have also been identified between teachers’ and students’ preferences for different types of CF. For instance, Yoshida (2008) showed that Japanese EFL learners preferred to be given time to rethink and self-repair their errors, as opposed to their teachers’ preference for recasts. ESL teachers in Lee (2013) also favored recasts, but their advanced-level learners preferred explicit correction. This also held true for Roothooft and Breeze (2016), who revealed that Spanish learners of EFL appreciated explicit CF in terms of explicit correction and metalinguistic clues, but their teachers preferred elicitation. However, unlike the studies reviewed above, Kaivanpanah et al. (2015) found that Iranian EFL learners showed a positive view of recasts and metalinguistic clues, while their teachers tended to use different types of CF depending on the circumstances. Taken together, this research shows that teachers and students in different contexts may hold varied preferences for CF types.

Explanations for these differing preferences have been related to a number of contextual variables, proficiency being one of them. Brown (2009) revealed high-proficiency students were more favorable to being corrected indirectly than lower-proficiency students. Similar findings were also found by Kaivanpanah et al. (2015), who noted that Iranian learners with higher proficiency responded more positively to elicitative types of CF that required self-correction compared to those with lower proficiency. Another study worth noting is Zhang et al. (2022), which compared CF beliefs between high- and low- proficiency CSL learners and found that low-proficiency learners preferred delayed correction more than those with high proficiency. Furthermore, proficiency also affects learners’ preferences for CF regarding error types, as Yang (2016) indicated that intermediate CSL learners are more confident in the efficacy of clarification requests on pronunciation errors than beginning learners. In addition, individual learning experiences (Schulz, 2001; Loewen et al., 2009), instructional settings (Lee, 2013; Kaivanpanah et al., 2015), and learner affect (Yoshida, 2008) also have contributed to the variance of CF-related preferences. This again highlighted the context-specific nature of beliefs, which suggests a need for more inquiries to examine the effects of other factors on teacher and learner CF preferences.

Previous research has shown that error type is a factor affecting learners’ perceptions of CF. For instance, a number of studies have reported a tendency for learners to perceive phonological and lexical CF more accurately than grammatical CF (Nabei and Swain, 2002; Yoshida, 2010). However, there are exceptions, as Mackey et al. (2007) reported that learners of Arabic were more likely to notice teachers’ CF of lexical and grammatical errors than phonological ones. In addition, there is also a discrepancy between learners’ perceptions of and teachers’ intentions for CF with respect to error types. This line of research has shown an overall tendency that teachers tend to provide more CF for grammatical errors, while learners perceive phonological and lexical errors more accurately (Mackey et al., 2000). Discrepancy is also noted between teachers’ preferences for and learners’ responses to CF. This was demonstrated by Lyster (1998) who discovered that teachers preferred to recast grammatical errors, whereas learners generated more successful grammatical repairs as a result of negotiation of forms. Nonetheless, CF research in the context of Chinese as a second language is rather limited, especially in terms of comparisons between teachers’ and learners’ preferences for different types of CF and the underlying reasons. This dearth of material makes the present study essential.

Mainstream CSL research has long focused on the characteristics of the Chinese language from a linguistic perspective. Over the past decade, with the growing number of Chinese language learners and the global demand for effective Chinese pedagogy both in and outside China (Orton, 2008; Moloney and Xu, 2015), increasing attention has been paid to issues related to classroom teaching and learning, CF being one of the most prominent. It has been found that as with other languages, recasts are also the most frequently used type of CF in CSL classrooms (Bao, 2019), and the efficacy of CF on CSL learning has also been confirmed (Lu and Gao, 2015). Regarding the efficacy of different types of CF, prompts have been found to be more effective than recasts in the long term (Cao and Mu, 2013), echoing the findings of Ammar and Spada (2006). Likewise, variance also exists, as some studies show that phonological errors received CF most (Bao, 2019), whereas other research reports that CSL teachers provide the largest amount of CF for grammatical and lexical errors (Duan and Sun, 2015). In addition, Yang (2016) explored CSL learners’ preferences for six types of CF in relation to specific error types and found learners’ general preferences for explicit correction, metalinguistic clues, and recasts across error types. This research casts some light on CF-related beliefs in the context of CF. However, research on comparing CSL teachers’ and learners’ preferences for CF is scant. This justifies the need for the present study, especially considering the features of the Chinese language and the source of Chinese teachers’ pedagogical beliefs.

It is widely accepted that the unique sound, grammar, and orthography systems of the Chinese language not only distinguish it from many other languages but present challenges for foreign learners, in particular its pronunciation and tone (Orton, 2008). These linguistic features might lead to different preferences for CF between CSL students and those of alphabetic languages. Additionally, it is well-known that Chinese teaching beliefs have been significantly influenced by Confucian thoughts (Hu, 2002). As a result of this influence, the teacher is seen as an authority figure, taking on primary responsibility to transmit knowledge and supervise learner performance, while students are expected to be obedient and maintain a high level of receptiveness (Li, 2003). This conceptualization of the teacher’s role may lead to Chinese teachers’ CF preferences differing from their counterparts under the Western education schema and students coming from a background unrelated to Confucian culture. The reasons outlined above provide another justification for the need for the present study by focusing on the following research questions:

This study was conducted at two southern Chinese universities, which were selected through convenience sampling, since the authors worked in one, and had a personal connection with the other. Both universities provided a Chinese language program with an aim to help foreign students’ improve their Chinese proficiency. The respective programs at both universities had the same curriculums, textbooks, teaching methods, and assessment techniques. Specifically, they included listening, speaking, reading, writing, and comprehensive Chinese courses. In order to accommodate individual participant proficiency, three levels of modules were established according to students’ placement tests; these were elementary, intermediate, and advanced, each of which had two subcategories (i.e., lower and upper). All modules constituted 4 to 6 teaching sessions per day. Teaching at both universities was delivered by full-time university teachers, part-time teachers, and second-year intern graduate students majoring in Teaching Chinese to Speakers of Other Languages. A series of textbooks titled ‘Developing Chinese’ was used for all courses. At the end of each semester there was a written test, which was designed by the corresponding teacher of each course.

A total of 328 students and 46 teachers participated in this study. The students came from various cultures; over half of them (53.8%) were from Asia, with African students being the next largest group, and a small percent from Western nations. Among them, there were 193 females and 135 males. They aged from 18 to 39 years old, with a mean age of 34 years. In terms of Chinese proficiency, students varied in their HSK level¹, ranging from HSK1 to HSK6. Specifically, participants with HSK1-3 were classified as elementary (n = 120), those with HSK4–6 were treated as intermediate (n = 208). Despite these differences, all the participants at both universities studied in a similar language program. Meanwhile, they shared a target-language learning context, which exposed them to plentiful Chinese input and had access to use Chinese in their daily life. In addition, as reported in the interviews, participating students have not been treated differently in class regardless of their various backgrounds. Given these commonalities, participating students were treated as one group. Certainly, any differences will be reported.

As for the 46 teachers, 34 of them were female and the rest (n = 12) were male. All were full-time, and their age ranges from 25 to 60 years old with a mean age of 40 years. They all had more than 5 years of experiences in teaching CSL, classifying them as experienced teachers according to Gatbonton (2008). All majored in Chinese linguistics; some of these (n = 15) were at PhD level while the rest (n = 31) were Masters. As such, the teachers were also treated as one group.

In order to capture the varying and dynamic nature of beliefs, the present study drew on the methods of questionnaires and interviews.

In response to Question 1, SPSS22.0 was used to carry out descriptive statistics to calculate the mean scores of the 18 items asked in both questionnaires. To answer Question 2, an independent t-test was carried out to examine whether there are significant differences between the teachers’ and students’ CF preferences. As for Question 3, a qualitative analysis of the interview data was analyzed inductively and coding was grounded in the data. Firstly, we iteratively read through the students’ interview transcripts and used open coding to assign codes to pieces of data pertinent to the reasons for their CF preferences (Corbin and Strauss, 2008). These codes provided an index to store and retrieve the data by recursive examination. Secondly, through many rounds of repetitive reading of the transcripts and those assigned codes, we merged the related codes into categories and developed them into a number of themes. Once these themes were established, we pooled all data touching on these themes. Finally, we moved iteratively from the data to the established themes and conceptualized the ultimate themes for the current study through axial coding. The same coding process was also carried out on the teachers’ interview transcripts to identify the reasons underlying their CF preferences. To enhance the reliability of this coding, member checking was used by sharing the coding with each interviewee, who agreed with the interpretations of their responses.

As for this question, the mean scores for the eighteen questions asked in the teacher and student questionnaires were calculated (see in Table 3). The results indicated CSL students’ preferences for explicit correction and metalinguistic clues across error types, mean scores of which were all over 3.00 (out of a maximum score of 4.00). However, they were less disposed toward repetition, elicitation, and clarification requests, with repetition having the lowest rating, regardless of error types. This seems to suggest CSL students’ preference for explicit types of CF over implicit ones. In terms of recasts, the mean scores for all three error types were relatively high, with phonological error being 2.79 and grammatical and lexical errors both being over 3.00. This seems to indicate that CSL students somewhat favored recasts, especially for the correction of grammatical errors (M = 3.27).

With regard to CSL teachers, they showed strong preferences for recasts across error types, the mean scores all being over 3.00. However, they were less disposed toward explicit types of CF and prompts; the mean scores for all these types were around the baseline (M = 2.5).

To answer this question, an independent t-test was performed to examine whether there were significant differences between CSL teachers’ and students’ preferences for different types of CF regarding the three error types. The results identified 11 items that showed significant differences (see Table 4), which are only reported here briefly due to space limitations.

As can be seen, significant differences were found between CSL teachers and students with regard to their preferences for explicit correction, metalinguistic clues, and clarification requests across error types. Interestingly, as for these items, the mean scores from the CSL students were all higher than those from the teachers, indicating that CSL students have a greater preference for these types of CF than teachers, especially in terms of explicit correction and metalinguistic clues. Significant difference is also found in recasts with regard to phonological and lexical errors; in the case of both items, the CSL teachers were more confident than students. This indicates that CSL students approve more of explicit types of CF and clarification requests, while CSL teachers embrace recasts.

To answer this question, a qualitative analysis of the interview transcripts from the teachers and students was conducted, concentrating on the reasons for their preferences for oral CF. The results were presented, respectively.