1. Introduction
The timing of corrective feedback (CF), alternatively called feedback timing, refers to the choice of a timepoint for providing corrections on second language (L2) errors or making comments on the appropriacy of L2 learners' verbal or nonverbal behaviors. A typical distinction related to the notion of feedback timing is between immediate and delayed feedback, but what constitutes immediate or delayed has been interpreted and defined in different ways. In one stream of research, immediate feedback is operationalized as feedback provided during a learning task and delayed feedback as feedback provided after a task is completed (Arroyo & Yilmaz, 2018Footnote *; Li Zhu & Ellis, 2016a*; Quinn, 2014*). One methodological variation in this distinction is interim feedback, which is provided after the first task is completed and before the second task is started (Li, Li, & Qian, under review). Interim feedback is relevant or possible when multiple tasks are performed. It refers to feedback provided during the interval(s) between tasks. Interim feedback is different from delayed feedback in that the latter refers to feedback provided after the task (if there is only one task) or all tasks (if there are multiple tasks) are completed and there is no further task performance following the feedback session. This way of conceptualizing feedback timing is based on the positioning of feedback during a task cycle, instead of the proximity to errors. Another way to examine feedback timing is to distinguish feedback provided immediately after an error is made and feedback delayed until a later time in the instructional cycle, such as one week later (Lavolette, Polio, & Kahng, 2015*). In this case, both immediate and delayed feedback can occur either during or after the completion of a learning task. A third way is to define feedback timing options in terms of their relation to instruction, namely whether feedback is provided immediately after explicit instruction or at a later stage after learners complete some practice activities (Fu & Li, 2022*). It should be clarified that this way of operationalizing feedback timing is markedly different from that in other studies in that it focuses on feedback's relation to instruction instead of errors. To conclude this section, it is necessary to point out that the conceptualization and operationalization of feedback timing should be reconsidered in L2 research. Feedback timing is not merely a matter of the length of interval or the distance between errors and feedback, and other parameters of the instructional system where errors occur are also involved or relevant, such as the distance between feedback and instruction, the positioning of feedback in a task cycle (such as within, after, or between tasks), and so on. These parameters are important because they contribute to the effectiveness of different timing options. Despite the variation in the operationalization of feedback timing, we argue that it is a unified construct that is theoretically justifiable, empirically examinable, and pedagogically valuable.
What theoretical perspectives are there on feedback timing? Most L2 theories do not make explicit claims about feedback timing, but their claims about how learning occurs support the superiority of immediate feedback over delayed feedback (Li, 2020*). According to the Behavioristic approach to L2 learning, language learning is habit formation, and errors must be corrected immediately before turning into bad habits. Based on the Interaction Hypothesis (Long, Reference Long2015), feedback should be provided during interaction to allow learners to make an immediate comparison between the erroneous and correct forms while engaged in a communicative task. The Sociocultural Theory holds that feedback must be tailored in an ongoing fashion during the interaction between a novice and an expert. Thus, both the Interaction Hypothesis and the Sociocultural Theory (Aljaafreh & Lantolf, Reference Aljaafreh and Lantolf1994) claim that feedback provided during interaction is more effective than feedback provided after interaction. Skill Acquisition Theory (DeKeyser, Reference DeKeyser, VanPatten and Williams2015) posits three stages for learning: declarative knowledge, proceduralization, and automatization. Declarative knowledge refers to knowledge about language and is obtained through instruction. Declarative knowledge is proceduralized through application of the knowledge in skill-specific practice activities and automatized through repeated practice. Feedback can be provided immediately after instruction or delayed until a later time, and in this context, feedback timing is determined in terms of feedback's proximity to instruction. It can be argued that feedback can reinforce and solidify declarative knowledge if it is adjacent to instruction, and that the effects of feedback are likely compromised if it is delayed or disjointed from instruction (Fu & Li, 2022*).
Research on feedback timing has examined the topic from the following perspectives: the impact of feedback timing options on learning gains, learner and teacher beliefs about the options of feedback timing, teachers’ feedback providing practices in the classroom, and students’ reactions. The research on the impact of feedback timing examines causal effects, involves systematic manipulation of instructional treatments, and administers pretests and posttests to evaluate treatment effects (Li, Reference Li, Gurzynski-Weiss and Kim2022a). These studies can be divided into several subcategories examining feedback timing in communicative tasks, L2 writing, and drill-type activities, and in terms of the relationship between individual difference factors and treatment effects. In this article, studies on beliefs, practices, and reactions are grouped together and are distinguished from studies examining the effects of feedback on learning gains. In the following sections, we provide a snapshot of the different strands of research on feedback timing before providing a timeline for the trajectory of the research.
1.1 Feedback timing in communicative tasks
Among experimental studies where communicative tasks were used, four involved face-to-face communication (Fu & Li, 2022*; Li et al., 2016a*; Quinn, 2014*; Rassaei, 2024*), three text chat (Arroyo & Yilmaz, 2018*; Henderson, 2020*; Henderson, 2021*), and one video chat (Canals et al., 2021*). One (Henderson, 2021*) investigated vocabulary learning and all other studies grammar. The operationalization of feedback timing varies between these studies. In the studies by Arroyo and Yilmaz (2018)*, Henderson (2021)*, Li et al. (2016a)*, Rassaei (2024)*, and Quinn (2014)*, feedback timing was defined as whether feedback was provided during or after communicative tasks. In Canals, Granena, Yilmaz, and Malicka (2021)*, the immediate feedback group received feedback during video chat, while the delayed feedback group received feedback 24 hours later by watching video playbacks of their own task performance where oral feedback was added as video overlays. In Fu and Li's (2022)* study, feedback timing was operationalized as whether learners received feedback while performing communicative tasks immediately after receiving grammar instruction on the target structure or at a later stage after completing some communicative tasks. Thus, in Fu and Li's (2022)* study, feedback timing was defined in terms of feedback's relation to the initial grammar instruction, and in both immediate and delayed feedback, feedback was provided during task performance. The kinds of feedback provided also varied between the studies. Corrective recasts (prompt + recast) were provided in Fu and Li (2022)*, Li et al. (2016a)*, and Quinn (2014)*; explicit correction was the corrective strategy in Arroyo and Yilmaz (2018)*, Canals et al. (2021)*, and Henderson (2021)*; and prompts were given in Rassaei (2024)*. Regarding research context, two studies (Fu & Li, 2022*; Li et al., 2016a*) were conducted in the classroom where students worked in groups and received feedback from the teacher, and other studies were carried out in the laboratory where interaction happened between a learner and a native speaker. The methodological variation of the studies may be partly responsible for some disparate findings, which are discussed below.
The above studies showed the following findings. First, in general, immediate feedback provided during task performance is more effective than delayed feedback provided after a communicative task is completed. The researchers have attributed the advantage of immediate feedback over delayed feedback to immediate feedback's affordance of opportunities for immediate comparisons between errors and feedback and for constant application of the knowledge learned from feedback in ongoing task performance. Such opportunities are missing in delayed feedback. Second, the conclusion about the advantage of immediate feedback over delayed feedback is equivocal because some studies (Henderson, 2021*; Quinn, 2014*; Rassaei, 2024*) did not find any advantage for immediate feedback. Factors responsible for the lack of differences between the two feedback types’ effectiveness include the provision of explicit instruction before the feedback treatment, the laboratory setting, ceiling effects (i.e., learners have too much previous knowledge), and salience of the linguistic target (i.e., immediate feedback works better for nonsalient structures). It is noteworthy that Henderson's (2021)* study focused on vocabulary, while other studies investigated grammar. Third, Fu and Li's (2022)* study shows that feedback immediately following instruction is more effective than feedback provided at a later stage of the instructional cycle, suggesting that feedback is likely more effective when it is closer to instruction, which may reinforce the effects of feedback. This study also suggests the importance of practice on the grounds that while learners had opportunities to practice and apply the learned knowledge after receiving immediate feedback, they did not have such opportunities in the delayed feedback condition which took place in the final treatment session. Fourth, the results should be interpreted by consulting the methodological features of the studies. The study on vocabulary learning (Henderson, 2021)* was conducted with a small sample size, with ten learners in each group, and ten targeted vocabulary items in the treatment and tests, raising concerns over the statistical power of the results and test validity. In the delayed feedback conditions of the studies based on text chat, learners were provided with a list of wrong sentences with corrections, and the extent to which they processed the feedback is uncertain – a point that also figured in the studies on L2 writing.
1.2 Feedback timing in L2 writing
Early research on feedback in L2 writing has primarily focused on its efficacy without considering feedback timing as an independent variable. This is primarily because the provision of CF in writing naturally occurs after the completion of the text, and it is often delayed until a later time. Teachers often face time constraints, as they must provide feedback to a large number of students, making immediate feedback nearly impossible.
Related to feedback timing in L2 writing is the issue of timeliness, which pertains to the interval between text completion and the delivery of written feedback. The importance of timely feedback is underscored in research on dynamic written feedback, where indirect, comprehensive feedback is delivered to regular paragraph writing (done in almost every class) completed within 10 minutes in English as a second language (ESL) college (Kurzer, Reference Kurzer2018) and university writing classes (Hartshorn et al., Reference Hartshorn, Evans, Merrill, Sudweeks, Strong-Krause and Anderson2010; Hartshorn & Evans, Reference Hartshorn and Evans2015). In dynamic feedback studies, feedback is returned to students in a timely manner – for example, in the following class period (Evans et al., Reference Evans, Hartshorn, McCollum and Wolfersberger2010; Hartshorn et al., Reference Hartshorn, Evans, Merrill, Sudweeks, Strong-Krause and Anderson2010). One recent study on dynamic feedback has addressed the feedback timing factor in graduate student writing (Eckstein et al., 2020)*. In the study, students in the timely feedback group received feedback biweekly throughout the 12-week course, while in the postponed feedback group feedback was delayed to the last two weeks of the semester. Analysis of grammatical accuracy, lexical and syntactic complexity, and fluency showed that timely and delayed written feedback did not significantly improve grammatical accuracy and lexical complexity. However, timely feedback enhanced fluency and syntactic complexity in writing. Based on the results of Eckstein et al. (2020)*, it can be concluded that while feedback timing may not have any effect on written accuracy, timely feedback provided in a regular, ongoing manner may enhance the linguistic complexity of graduate student writing. In Lavolette et al.'s (2015)* study, ESL students submitted their essays to an automated feedback tool either immediately or one to three weeks after completion of the writing task. No differences were found between the two feedback conditions.
The advent of technology has made feedback timing, particularly focusing on written feedback provided during and after writing (similar to oral communicative activities), a researchable variable. A few studies have looked into the effects of synchronous and asynchronous written feedback on L2 student writing (Cheng & Zhang, 2024*; Shintani, Reference Shintani2016; Shintani & Aubrey, 2016*). Synchronous versus asynchronous is a distinction often used to refer to whether two events happen simultaneously in computer-mediated interaction. Applied to feedback timing, synchronous feedback refers to feedback that is provided while the writer is engaged in the writing process or composing the text, in which case the feedback is synchronized with the writing behavior. Asynchronous feedback is separate from composition and is provided after the writing task is completed. In the context of feedback timing, synchronous feedback can be considered a type of immediate feedback whereas asynchronous writing can be equated with delayed feedback. Both Cheng and Zhang (2024)* and Shintani and Aubrey (2016)* found synchronous feedback more effective than asynchronous feedback, and the advantage of synchronous feedback was ascribed to the practice opportunities learners had to apply declarative knowledge when composing new sentences, learner agency, immediacy of feedback, and engagement. The benefits of synchronous written feedback are also underscored in Shintani (Reference Shintani2016), where English as a foreign language (EFL) university students received direct feedback asynchronously and synchronously in two separate classes. The study investigated how students utilized and responded to written feedback in the two different feedback timing conditions. Results showed that synchronous feedback promoted interaction between students and the feedback provider (the researcher in the study) during the writing process, and it facilitated noticing, internalization, and students’ self-correction of errors. Students who received asynchronous feedback, in comparison, had fewer opportunities for consolidation, as they might repeat the same errors until they received postponed feedback after the writing had been completed. Another benefit of synchronous feedback, as borne out in the findings, is that students could attend to form and meaning contiguously rather than separately, as in the asynchronous condition. Overall, Shintani's (Reference Shintani2016) study has pointed to the positive influence of synchronous feedback on the writing process and in promoting language learning.
1.3 Feedback timing in drill-type activities
Drill-type activities refer to learning activities or exercises that consist of discrete, isolated items, that are purely form-based (except for structured input activities in Henshaw's (2011)* study that were meaning-based), and that do not involve form-meaning mapping (e.g., multiple choice grammar or vocabulary exercises) (Li, Reference Li, Gudmestad and Edmonds2018). In these activities, feedback is provided on students’ answers either during the activity, in which case the feedback is considered immediate, or after all questions are answered, in which case the feedback is delayed. Four studies have been conducted on feedback timing in drill-type activities. Among them, two examined grammar learning and involved comprehension practice (Henshaw, 2011*; Lavolette, 2014*); two examined vocabulary learning and involved production practice (Kim & Webb, 2023*; Nakata, 2015*). All but one study consisted of a presentation or instruction stage followed by a practice stage where learners were required to retrieve knowledge learned from the instruction and received feedback on their answers. Lavolette (2014)* is the only study without pre-practice instruction. The results of the four studies are homogeneous: there is no difference between immediate and delayed feedback in their effectiveness in L2 learning. However, no conclusion can be reached based on available research because of the small number of studies and methodological issues such as small sample sizes (e.g., only ten learners in a participant group), risks for test validity (e.g., only five items in a test of treatment effects), and so forth.
1.4 Individual difference factors in feedback timing
Individual difference (ID) factors refer to learner traits, dispositions, and propensities that cause learners to vary and that are posited to have a direct or indirect impact on learning behaviours, processes, and outcomes (Li, Reference Li and Li2024; Li et al., Reference Li, Hiver, Papi, Li, Hiver and Papi2022). ID factors examined in feedback timing research include anxiety, declarative memory (associative memory), language aptitude, procedural memory, and working memory. Studies where immediate feedback was provided during and delayed feedback after a communicative task show the following findings. Language analytic ability, a component of language aptitude, was correlated with the effects of delayed feedback in Li et al. (2016)*, but it showed no significant effects in Arroyo and Yilmaz (2017)*. Working memory was predictive of the effects of immediate feedback in Li et al. (2019)* and Rassaei (2024)*, but it was not a significant predictor under any treatment condition in Henderson (2020)*. Fu and Li (2021, 2024)* investigated IDs’ roles in feedback provided immediately after explicit instruction and feedback provided after students completed some communicative practice. They found that the effects of immediate feedback were predicted by procedural memory and working memory while the effects of delayed feedback were associated with anxiety, declarative memory, and working memory. It is worth clarifying that in both feedback conditions, feedback was provided during rather than after task performance. A recent study by Yilmaz et al. (Reference Yilmaz, Granena, Canals and Malicka2024)* examined the associations between associative memory, a type of declarative memory and delayed CF embedded in the video recordings of learners’ task performance. The learners received recasts, explicit correction, or no feedback depending on their group assignment. Treatment effects were measured by means of an oral production test and a grammaticality judgement test. Both feedback groups outperformed the control group, and there was no difference between the two treatment groups. Associative memory was a positive predictor of the effects of the two feedback types but a negative predictor of no feedback, and the significant results were found for the grammaticality test, not the oral production test. As can be seen, IDs have a complicated relationship with treatment effects, which varies as a function of the processing demands of learning tasks. In general, analytic ability is important when there is a lack of external assistance such as explicit instruction or when there is not a requirement for production (output) (Li, Reference Li, Li, Hiver and Papi2022b); working memory plays a role in conditions where learners have a heavy processing burden (Li, Reference Li, Godfroid and Hopp2022c); procedural memory is involved when learners try to apply explicit knowledge; declarative memory surfaces when learners engage in processing and acquiring explicit knowledge; anxiety tends to cause a negative effect when learners’ anxiety is triggered by the cognitive demands imposed by a learning task such as when learners’ entrenched errors are corrected via delayed feedback.
1.5 Beliefs, practices, and reactions
Li (2017)* defined feedback beliefs as “attitudes, views, opinions, or stances learners and teachers hold about the utility of CF in L2 learning and teaching and how it should be implemented in the classroom” (p. 143). Practices refer to how immediate and delayed feedback occurs in the classroom or whether teachers provide immediate or delayed feedback in their teaching. Reactions refer to affective responses after receiving immediate and/or delayed feedback. Li (2017), who synthesized all empirical evidence on learner and teacher beliefs about CF, identified only two studies examining students’ preferences regarding feedback timing, and the studies showed somewhat mixed results. However, studies that examined learners’ reactions to feedback treatments showed that learners favored immediate feedback instead of delayed feedback. For example, Quinn (2014)* showed that students preferred to be corrected during an activity rather than after an activity or after all activities are finished. Murphy, Mackay and Tragant (2023)* used WhatsApp – a mobile instant messaging app – to provide feedback during and after communicative tasks and found that all learners were positive about within-task feedback. Finally, students’ reactions to synchronous and asynchronous written feedback seemed unclear based on available evidence. According to Cheng and Zhang (2024)*, most L2 writers in the synchronous (6 out of 8) and asynchronous (5 out of 8) groups liked the feedback they received.
On the teacher's side, Li's (2017)* aggregation of the results of six studies showed that only 40% of the teacher participants in the studies agreed with providing immediate feedback. A recent study by Yuksel, Soruç, and McKinley (2023)* revealed that teachers’ ratings for immediate and delayed feedback were similar, suggesting that they did not favour either feedback type. These results, together with the results on students’ preferences, seem to demonstrate that there are disparities between teachers and students in their preferences: while students are overall positive about immediate feedback, teachers are hesitant. However, Roothooft (2014) found that teachers’ preferences depended on error type: they thought it necessary to provide immediate feedback on errors that impede communication and delayed feedback on other errors. Yuksel et al. (2023) compared teachers’ feedback providing practices in the classroom and their feedback beliefs and found incongruency between what teachers claimed and what they did in the classroom. Specifically, while they did not support immediate feedback in the survey, 66% of their feedback was immediate, and delayed feedback occurred in only 34% of the cases.
Several studies have examined pre-service teachers’ beliefs. However, these results should not be seen as representative of practicing teachers’ beliefs, as pre-service teachers are still students in training. These studies show that similar to students, pre-service teachers mostly favour immediate feedback (Kartchava et al., Reference Kartchava, Gatbonton, Ammar and Trofimovich2020). One interesting finding of Kartchava et al.'s (2020) study is that participants who had taken a second language acquisition (SLA) course were more positive about providing immediate feedback than those who had not. Thus, training may have led to a stronger faith in the significance of feedback, which should be provided immediately without delay in these pre-service teachers’ opinion.
One study (Rolin-Ianziti, 2010)* used conversation analysis to provide a detailed analysis of episodes where teachers provided feedback after students completed communicative activities. Thus, it examines how delayed feedback is implemented by teachers. The study shows that an episode of delayed feedback normally starts with the teacher's initiation of the error correction sequence. There are multiple ways to initiate the sequence, such as by quoting the erroneous utterance (e.g. “You said…”), asking the learner to recall the erroneous utterance without providing it (e.g. “You talked about x. What did you say?”). When quoting an error, the teacher must decide the amount of context that is restored, such as the isolated error, the whole utterance, or the erroneous utterance plus the preceding and following utterances. After the error is presented, the teacher may correct the error in multiple ways, such as by providing the correct form, encouraging the learner to self-correct, eliciting the correct form from other learners, juxtaposing the wrong and correct forms and asking the learner to make a choice (e.g., “Is it x or y?”), and so forth. The teacher may also use a hybrid strategy consisting of different forms of correction depending on the nature of the error, such as providing the correct form for new linguistic knowledge and encouraging self-corrections for previously learned linguistic forms. After the error is corrected, the teacher may move on to the next error directly or ask the learner to repeat the correct form before proceeding to the next error. Thus, one contribution of this study is the discovery that delayed feedback can be more varied, elaborate, and engaging, unlike what has happened in current experimental research, where delayed feedback normally takes the form of a list of wrong and corrected sentences or isolated single-move error correction sequences similar to what happens in a grammar exercise. The researcher concludes that overall the teacher's delayed feedback falls into two types: “teacher initiated and teacher completed correction” and “teacher initiated and student corrected”. In both cases, it is the teacher who initiated the error correction sequence. The difference is that in the former case the teacher provided the answer while in the latter the teacher elicited the answer from the student. It is necessary to emphasize that this is an observational study that describes what occurs in the classroom (Li, Reference Li, Gurzynski-Weiss and Kim2022a; Li & Vuono, Reference Li and Vuono2019). Classroom observation studies are invaluable in L2 research because they explore phenomena that happen in the natural state and show features and variables that can be tested in subsequent experimental research (Mao et al., Reference Mao, Lee and Li2024). Therefore, classroom observation research has high ecological validity, and it also increases the ecological validity of experimental research by making what is tested experimentally relevant to L2 pedagogy.
1.6 Categories of feedback timing research
-
A. Feedback timing in communicative tasks
1. Face-to-face oral interaction
2. Text chat
3. Video chat
3. B. Feedback timing in L2 writing
1. Synchronous vs. asynchronous written feedback
2. Immediate vs. delayed feedback
3. Timeliness of feedback
C. Feedback timing in drill-type activities
1. Grammar
2. Vocabulary
D. Beliefs, practices, and reactions
1. Beliefs
2. Practices
3. Reactions
E. Individual difference factors in feedback timing
Shaofeng Li is Professor of Applied Linguistics at The Hong Kong Polytechnic University, where he conducts research and teaches courses on second language acquisition and language pedagogy. He received a Ph.D. in Second Language Studies from Michigan State University. Dr. Li has published on a wide range of topics including task-based language teaching and learning, corrective feedback, second language writing, research methods, meta-analysis, and cognitive and affective individual difference factors such as anxiety, motivation, language aptitude, and working memory. He is the founding editor and editor-in-chief of Research Methods in Applied Linguistics, the first and only journal focusing exclusively on research methods in applied linguistics. He is also the book review editor of TESOL Quarterly, and the co-editor-in-chief of Digital Studies in Language and Literature. He is included in the Stanford University list of the world's top 2% most influential scientists.
Ling Ou is Associate Professor of English in the School of Foreign Languages and Cultures, at Chongqing University, China. Her research focuses on curriculum development and language teaching. She has made significant contributions to both academic research and pedagogical innovation. She has authored a number of articles, directed six research projects, and edited 12 textbooks. Her work has been central to the university's educational reform, particularly in the area of foreign language education. She led the development of a school-based English course system and played an important role in advancing language pedagogy at the university.
Icy Lee is Professor of Education (TESOL & Language Education) at the National Institute of Education in Nanyang Technological University, Singapore. Her main research interests are second language writing and second language teacher education. Her publications have appeared in numerous international journals, such as Applied Linguistics, Journal of Second Language Writing, Language Teaching, Language Teaching Research, System, and TESOL Quarterly. She is former co-editor of the Journal of Second Language Writing and currently Principal Associate Editor of The Asia-Pacific Education Researcher and co-editor of the International Journal of Christianity & English Language Teaching.
