Translation as a catalyst for foreign language learning: a self-regulated learning approach mediated by instructor feedback and peer collaboration

The findings of this study are aligned with the three RQs. In particular, section 4.1 answers RQ1 on whether the trainees developed and sustained SRL. Section 4.2 answers RQ2 by explaining whether peer collaboration, instructions, and feedback fostered FL learning and SRL. Meanwhile, section 4.3 answers RQ3 on whether the trainees’ FL proficiency increased during the study.

Trainees’ self-regulated learning practices

Trainees developed and sustained self-regulated learning during the computer-assisted translation course

The trainees’ SRL schedules, as logged on the online form, were downloaded each week and documented. The data analysis indicated a gradual increase in their total SRL hours during the study. The number of hours increased from Week 1 (86 hours) and plateaued around Week 5, fluctuating between 108 hours and 42 minutes and 111 hours and 15 minutes for the remaining 10 weeks of the study, as outlined in Fig. 3.

Weekly number of hours and minutes spent on self-regulated learning.

The data illustrated that the trainees continued with SRL during the study. In particular, it was found that SRL time did not decrease on weekends or during weeks when trainees project managed rather than translated respective chapters.

SRL time was also investigated based on the trainees’ undergraduate majors to determine whether their backgrounds influenced their SRL practices. The analysis showed that the trainees who did not major in either English or translation at the undergraduate level spent more hours on SRL than their peers who majored in English and translation. In particular, while those with an English and translation background spent less than 80 hours a week on SRL, the rest spent over 80 hours on it, as indicated in Table 2.

An independent t-test computed in SPSS to determine the statistical differences among the various majors showed a significant statistical variance between the trainees who majored in English (M = 4449.14, SD = 407.07) and those who majored in translation (M = 4695.83, SD = 258.84), p = <0.001. There was also a practical difference between the two variables given that the effect size (Cohen’s d = 569.04) was significantly high. In contrast, no statistical differences (p = <0.008) were found between those who majored in English and other disciplines (TCM [M = 5328.50, SD = 224.15]) and those who majored in tourism and management (M = 5318.00, SD = 364.88), despite a considerably high practical significance judging from the high effect size (Cohen’s d = 429.43). There was also no statistical significance (p < 0.006) between the SRL time of translation and trainees who majored in other disciplines, despite a high effect size (Cohen’s d = 324.1). This finding showed that, to a significant extent, the trainees’ profiles—based on their undergraduate majors—determined their SRL practices.

Trainees spent more time on self-regulated learning when not translating

Trainees’ number of SRL hours and their weekly workload (i.e., the number of words they translated for that week) were compared to determine whether they developed and sustained SRL practices during the study. The goal was to determine whether their SRL practices changed based on their translation workload. The data indicated that the trainees’ SRL time did not decrease even during the weeks they had no assigned translation tasks. For instance, on Week 5, set aside for the trainees to revise their translations, they spent 108 hours and 42 minutes on SRL—that is, 5 hours and 34 minutes on average per trainee. The data showed that overall, the trainees spent between 4 hours and 38 minutes and 6 hours and 41 minutes (weekly average = 5 hours and 47 minutes) on SRL during the 14 weeks of the study.

Furthermore, it was found that the trainees spent more time on SRL for weeks without an assigned translation task. As mentioned earlier, each group had to project manage the weekly translation task. The data indicated that the average number of SRL hours on weeks with no translation task was 7 hours and 3 minutes, that is, 1 hour and 17 minutes above the weekly average. The average SRL time per trainee during the weeks of no assigned translation is outlined in Fig. 4.

Self-regulated learning time during weeks with and without assigned translation tasks.

A t-test computed in SPSS unveiled a statistical variance between the average SRL time when trainees had an assigned translation task (M = 342.33, SD = 34.46) and that when they had no assigned translation task (M = 424.14, SD = 99.52), p = <0.001 (statistically significant). In addition, there was practical significance between the two variables, with a moderately high Cohen’s d (0.7.16). Therefore, the finding confirmed that the trainees’ SRL did not depend on their translation workload. In other words, the trainees continued to learn the FL regardless of the translation task.

It was concluded that these findings provided evidence of constructivism, given that the trainees’ SRL schedules proved that they took responsibility for their learning by actively choosing what, when, and how long to study (Zimmerman, 2000). Their personalization of the learning process—a vital element of constructivism—fully manifested as they tailored their learning to their individual needs based on their strengths in English. For example, P18, who spent 10 hours and 36 minutes on SRL, revealed that they spent between 60 and 90 minutes per day improving their grammar, vocabulary, and delivery errors. In contrast, they spent 120 to 180 minutes per week improving their writing errors. The emphasis on writing errors reflected the trainee’s number of FL errors.

More precisely, based on the trainer’s instructions and feedback, P18 worked on wordy sentences, passive and active sentences, and sentence clarity (writing); prepositions, misplaced adverbs, dangling modifiers, and punctuation (grammar); colloquial expressions (delivery); and synonyms (vocabulary). The ability to plan, prioritize, and make decisions about their learning, perceived as core tenets of constructivism, were perceived as evidence of self-reflection and autonomy. Furthermore, the finding that the trainees remained engaged on weekends and when they had no translation tasks suggested they had set goals and reflected on how to achieve them.

Contribution of peer collaboration and instructor’s feedback and instructions

The trainer’s instructions and feedback, self-motivation, and peer assistance were crucial factors for trainees’ self-regulated learning development and sustainment

In the post-study questionnaire, the trainees responded to an open-ended question to rank, in order of importance, the factors they believed enhanced their SRL development and sustainment during the 14 weeks of the study. A thematic analysis of the data indicated that overall, most participants believed the trainer’s feedback and instructions (63%), feedback (58%), self-motivation (78%), help from / influence of peers (26.3%), and group collaborative atmosphere (26.3%) ranked top on the list of factors that fostered SRL, as indicated in Fig. 5. At 78%, self-motivation was the most important factor, followed by the trainer’s feedback and instructions.

Ranking of factors that enhanced self-regulated learning.

Unexpectedly, the trainees’ SRL development was also found to be influenced by the volume of their coursework during the semester (21.3%) and their internship (21.3%) schedules.

It was equally relevant to uncover whether the factors influencing the trainees’ SRL practices differed based on their profiles (i.e., undergraduate majors). The analysis indicated that English majors ranked the trainer’s feedback (71.4%) first on their list. Meanwhile, translation majors ranked self-motivation (100%) first, and the other majors ranked both the trainer’s instructions and self-motivation (83.3%) at the top of their list, as shown in Table 3.

This finding was consistent with research (Ping, 2012; Zimmerman & Kitsantas, 1999) undertaken within the SRL framework indicating that self-motivated language learners often develop more learning strategies to outperform less self-motivated learners.

Trainees perceived the trainer’s feedback and instructions to be extremely useful in foreign language learning and self-regulated learning

In the post-study questionnaire that assessed their perceptions of the course, the trainees were asked to rate the feedback and instructions provided by the trainer. Notably, they received feedback when the instructor revised their translations during in-class sessions, via the class instant messaging platform, and through personalized sessions targeting trainee-specific problems. Therefore, the questionnaire requested the trainees to determine the extent of the feedback’s and instructions’ usefulness on a scale from 1 (not useful at all) to 10 (extremely useful).

The data analysis indicated that most trainees ranked the usefulness of feedback between 8 and 10, perceiving it to be significantly useful. Specifically, 12 (63%), 3 (16%), and 4 (21%) trainees ranked the trainer’s feedback at 10, 9, and 8, respectively, as presented in Fig. 6.

Trainees’ perceptions of the usefulness of the trainer’s feedback.

Further analysis of the data revealed that the trainees who did not major in English and translation at the undergraduate level perceived the feedback to be most useful. All six of them (100%) ranked the usefulness at 10 (extremely useful), as opposed to 4 out of 6 (66.7%) and 2 out of 7 (28.6%) of the trainees who majored in Translation and English, respectively.

Concerning the instructions provided by the trainer during the study, the trainees were unanimous regarding the extent of its usefulness, with the majority ranking their perception at 9 (53%), 10 (42%), and 8 (5%), as indicated in Fig. 7.

Trainees’ perceptions of the usefulness of the trainer’s instruction.

The data was computed in SPSS to determine Pearson’s correlation coefficient between the feedback and instructions and the trainees’ SRL time. There was no significant correlation. Furthermore, a t-test did not unveil statistical differences between the trainees’ perceptions of instructions (M = 9.42; SD = 0.84) and feedback (M = 9.42; SD = 0.51).

Self-regulated learning development and sustainment also depended on peer collaboration

Qualitative data was collected when the trainees were requested to state the extent to which their SRL practices hinged on the actions of their peers. This data was presumed to be instrumental in designing future courses. According to our data analysis, all (100%) trainees claimed their peers influenced their SRL experience. The four areas with the most influence were proximity (some peers lived in the same dormitory) and mutual encouragement, group leaders’ constant reminders, co-scheduling of SRL time (co-learning), and coalescence around the same FL challenges (peers who had the same FL errors reached out to help each other). In particular, 13 trainees believed their SRL practice was influenced by proximity with their peers, and 11 believed their group leaders’ constant reminders were the most influential. Meanwhile, 10 trainees believed co-scheduling with peers and coalescing around the same FL challenges helped improve their SRL experience during the study, as illustrated in Fig. 8.

How peers influenced trainees’ self-regulated learning development and sustainment.

A further breakdown of the data indicated that the trainees who majored in English benefited most from proximity to peers (83.3%) and co-scheduling (83.3%). Those who majored in translation benefited most from proximity to peers (66.7% of the trainees), while the trainees who majored in other disciplines benefited most by coalescing around the same FL challenges (100%), as indicated in Fig. 9.

Perceived peer influence based on trainees’ undergraduate majors.

From the analysis, it is evident that the trainees with a non-linguistic background (i.e., who majored neither in English nor translation) found it most useful to gravitate toward each other in search of a solution to their FL challenges. In our estimation, this was a form of collaborative problem-solving (Nelson, 2013) resulting from a CAT course design focusing on SRL.

It was found that the answers to RQ2 reinforced the link with constructivism. First, by providing feedback and instructions, the trainer played a role in helping the trainees develop self-regulation (Oliver, 2000; Sun & Wang, 2020). The instructions provided scaffolds—that is, crucial guidance the trainees needed to construct a new understanding, develop and sustain SRL, and foster FL knowledge. Furthermore, constructivism emphasizes intrinsic self-motivation, which has been found to be a determinant in developing and maintaining SRL. In addition, the peer collaboration foregrounded constructivism, which emphasizes the social nature of learning. In particular, the finding that trainees with a non-English and translation background depended more on their peers reinforced the link with constructivism, which perceives learning as a process occurring within a community (Vygotsky, 1978) where learners support each other along the way.

Foreign language error categories and trainees’ performance

Overall, the more words trainees translated, the fewer foreign language errors they made

Our data analysis indicated that the trainees’ FL errors decreased as they translated more book chapters. The data was obtained by dividing the total words translated per chapter by the number of errors. According to the analysis, at the beginning of the study, the trainees made 1 error per 12.7 FL words translated (ratio = 1:12.7). However, as they translated more chapters, the number of FL words containing errors systematically increased. As illustrated in Fig. 10, the last chapter contained 1 error per 27.4 FL words translated (ratio = 1:27.4), an increase of 216.6%.

Number of foreign language words translated per error.

Comparatively, among the four broad error categories, our data analysis indicated that vocabulary witnessed the sharpest increase, from one error per 104.3 words translated at the beginning of the study to one error per 201.3 words translated at the end of the study. However, the change fluctuated, plummeting to one error per 73.7 words translated in Chapter 5, as indicated in Fig. 11.

Number of foreign language words translated per one vocabulary error.

Writing errors decreased as more words and chapters were translated. For instance, one grammar error per 21.8 words was observed when the book’s first chapter was translated. Then, the number of FL words translated per error rose unsteadily to 32.1 in Chapter 6 before gradually dropping to 28.2 by the end of the study (See Fig. 12).

Number of foreign language words translated per one grammar and writing error.

Regarding grammar errors, it was found that trainees committed one error per 47 words translated at the beginning of the study. Though the errors per word translated fluctuated during the research, there was a higher trend toward the end, where one error was made per 60 words translated, followed by a gradual increase to 61.2 words translated, as indicated in Fig. 12.

The slowest progress was witnessed in the delivery error category. Our data analysis indicated that the number of words containing one delivery error increased significantly when the first four chapters were translated. Thereafter, the number of translated words per error plunged (one error per 375.1 L2 words translated) in Chapter 6 and did not significantly increase until the end of the study, as outlined in Fig. 13.

Number of foreign language words translated per one delivery error.

The error data were computed in SPSS, and a t-test was performed to obtain the statistical and practical variance among the variables. Our analysis confirmed various statistical differences and large effect sizes, as demonstrated by the p-values and Cohen’s d values outlined in Table 4.

The data analysis indicated that delivery errors (M = 595.58; SD = 263.28) were the most dominant, followed by vocabulary (M = 137.33; SD = 43.19), writing (M = 49.32; SD = 6.97), and grammar (M = 27.53; SD = 4.07) errors. Regarding practical significance and based on Cohen’s d value, writing ranked first (Cohen’s d = 7.07), followed by grammar (Cohen’s d = 6.76), vocabulary (Cohen’s d = 3.31), and delivery (Cohen’s d = 2.26).

Grammar errors were the most predominant

Our data analysis indicated that grammar was the category with the most errors, accounting for 54% of all errors. This was followed by writing, which had 31% of the errors; vocabulary, which had 13% of the total errors; and delivery, with only 2% of the errors, as shown in Fig. 14.

Trainees’ foreign language errors per category.

Among the grammar errors, the wrong use of determiners accounted for the most errors (28.6%), followed by punctuation of compound and complex sentences (13.6%), noun numbers (10.7%), comma misuse (10.1%), wrong prepositions (7.6%), confused words (5%), subject–verb agreement (4%), and several others, as highlighted in Table 5.

Concerning writing errors, wordy expression errors (34%) were the most dominant, followed by passive misuse (32%), unclear sentences (19%), hard-to-read expressions (6%), confusing words (6%), ineffective/missing emphasis (2%), and outdated language (1%). Meanwhile, all errors classified under the vocabulary category were word choice errors, meaning the category had only one error type. In terms of delivery errors, our analysis indicated that most errors were tone errors (36%), followed by potentially sensitive language (27%), colloquialisms (23%), and incomplete sentence structures (14%).

Foreign language errors had a relationship with trainees’ undergraduate majors

The data analysis showed discrepancies in the trainees’ performance among the grammar, writing, vocabulary, and delivery error categories. A t-test computed in SPSS revealed significant statistical variances among the four variables of grammar (M = 28.30, SD = 8.26), writing (M = 48.04, SD = 8.84), vocabulary (M = 122.78, SD = 10.95), and delivery (M = 743.07, SD = 418.85), with all the variances statistically significant at p = <0.001.

However, our statistical analysis unveiled differences in FL errors based on the trainees’ undergraduate majors. In particular, it was found that trainees with other majors (TCM, tourism/management, international affairs, and accounting) performed better in the grammar category given that they had an average of one grammatical error per 37.3 FL words translated. In contrast, the trainees who majored in English had an average of one grammatical error per 24.5 FL words translated, and those who majored in translation had an average of one grammatical error per 23.8 FL words translated. A t-test confirmed the statistical difference among the variables, showing significant statistical variance among the data of English majors (M = 24.49, SD = 2.84), translation majors (M = 23.80, SD = 2.24), and trainees with other majors (M = 37.27, SD = 9.49), p = <0.001. In addition, the practical significance of the data was calculated by determining the effect sizes. Our analysis showed a high effect size for all variables, with a Cohen’s d of 2.66, 2.34, and 2.84 for trainees with other, translation, and English majors, respectively.

In contrast, it was found that trainees who majored in English and translation outperformed in the vocabulary and writing categories. According to the analysis, those who majored in English had an average of one vocabulary error per 143.3 FL words translated compared to those who majored in translation with one error per 116.6 FL words translated. In comparison, trainees with other majors had one error per 105.1 FL words translated.

An independent t-test confirmed the statistically significant (p = <0.001) differences among the trainees with other majors (M = 105.1, SD = 27.3), translation majors (M = 116.6, SD = 29.79), and English majors (M = 143.3, SD = 68.05). Notably, no statistical differences were found based on undergraduate majors in the writing category.