Recasts, Prompts and Noticing: A Comparative Study

Noticing has been regarded as an important theoretical construct in the mechanism of how corrective feedback (CF) facilitates second language acquisition. However, to date, only a paucity of CF research has examined noticing triggered by different types of CF (i.e., recasts vs. prompts). The study is intended to fill in the gap by examining the relationships between type of CF and level of noticing. To that end, 105 Malaysian ESL learners were asked to perform four communicative tasks during which recasts and prompts were provided contingent upon the encounter of past tense errors. To assess noticing, the study has employed a triangulated method using multiple elicitation procedures including diary writing, stimulated recall, and exit questionnaire. The results of the study revealed that both recasts and prompts were able to induce noticing the corrective intent, noticing the target of CF or form, noticing the gap, and noticing the rule. However, contrary to what was expected, recasts were able to promote higher levels of noticing across all noticing categories. Moreover, the greatest difference between recasts and prompts was found in noticing the gap. The study suggests that CF that provides exemplars of the target linguistic feature may promote higher levels of noticing the gap which may, in turn, increase the effectiveness of CF in L2 acquisition. the immediate effects of the CF treatment. After the administration of the post-tests, about 40 of the experimental participants were asked to perform stimulated recall interviews. The delayed post-tests were administered approximately two weeks after the completion of the CF treatment, to measure the long-term effects of CF. Immediately after the delayed post-tests, participants answered the exit questionnaire, which aimed to probe any extra-experimental factors that might have affected the results of the study.


INTRODUCTION
Noticing is a crucial factor in second language acquisition (Gass, 1997(Gass, , 2003Schmidt, 1995Schmidt, , 2001Swain, 1985Swain, , 1995Swain, , 2005 and an important theoretical construct that supports the effectiveness of corrective feedback (Loewen, 2012). To what extent second language (L2) learners are able to benefit from interactional input including corrective feedback is mediated by the cognitive activity of noticing. Gass (1988Gass ( , 1991Gass ( , 2003 believes that corrective feedback (CF) triggers noticing of the form and/or the meaning and the mismatch between learners' non-target-like form and the targetlike form, which eventually leads to grammar restructuring. Long (1996) proposed that conversational interaction can raise learners' awareness of language, resulting in an increased attention to form and can heighten the inclination of noticing the mismatches between the non-target-like form in learner utterances and the target-like form in the modified output. The Noticing Hypothesis by Schmidt (1995, p. 20) states that "what learners notice in input is what becomes intake in learning".
Previous classroom observational or descriptive studies (Lyster, 1998a(Lyster, , 1998bLyster & Ranta, 1997;Panova & Lyster, 2002;Sheen, 2004) and classroom experimental studies (Ammar & Spada, 2006;Lyster, 2004;Yang & Lyster, 2010) appear to concur that an explicit type of feedback (prompts) seems to be more effective than an implicit type of feedback (recasts). However, in these studies, little is known about learners' cognitive reactions to CF, that is to what extent learners were able to recognize the corrective intent of the teacher's feedback, and notice the gap between their erroneous production form and the target-like form embedded in CF.
Different types of CF may be associated with different levels of noticing (Kim & Han, 2007;Lyster, 1998aLyster, , 1998bLyster & Ranta, 1997;Mackey, 2006;Mackey et al., 2007;Mackey et al., 2000;Nassaji, 2009;Panova & Lyster, 2002;Yoshida, 2010). However, to date, few CF studies have examined the relationships between type of corrective feedback and level of noticing, specifically whether a more explicit type of CF leads to a higher level of noticing.
As pointed out by Lyster et al. (2013), more research that conceptually defines the construct of noticing in a more refined manner is warranted. The present study has attempted to fill in the gap by examining a range of noticing categories triggered by recasts and prompts which includes: a) Noticing the corrective intent (Carroll, 1995(Carroll, , 2001Lyster, 1998a;Lyster & Ranta, 1997;Roberts, 1995). b) Noticing the form/target (Carroll, 1995(Carroll, , 2001Schmidt, 1995Schmidt, , 2001. c) Noticing the gap (Doughty, 2001;Gass, 1988Gass, , 1991Gass, , 2003Long, 1996Long, , 2007Pica, 1994;Schmidt & Frota, 1986). d) Noticing the rule (Robinson, 1995;Schmidt, 1995Schmidt, , 2001. The objective of the study is to investigate if a more explicit type of corrective feedback, prompts, leads to a higher level of noticing than a more implicit type of CF, recasts. The study is aimed at addressing the following research questions: 1. Do recasts and prompts induce noticing? 2. Do prompts induce higher levels of noticing than recasts, measured by: a) total frequency of noticing, and b) hierarchical order of noticing?

Recasts vs. Prompts
Recasts and prompts are two corrective feedback types that have received prominence in comparative studies of CF, whether in a classroom or laboratory setting. In general, recasts and prompts are distinctive in terms of their explicitness, depending on whether there is an overt indication that some sort of linguistic errors has occurred in the learner's utterance. Thus, recasts are often regarded as an implicit feedback while prompts an explicit feedback.
In addition to classifying corrective feedback on the continuum of implicitnessexplicitness, Ellis (2006) classified different CF types on the dimension of inputproviding and output-prompting. In other words, CF, such as recasts and explicit correction, that provides a model of the correct form is classified as input-providing, while CF, such as clarification requests, prompts, that do not supply the correct model is labelled output-prompting CF. On this dimension, recasts and prompts vary considerably in that recasts provide the correct form in their reformulation (inputproviding CF). On the other hand, the correct form is usually withheld from the learner in prompts (output-prompting CF). In the SLA (second language acquisition) literature, recasts are defined as "the teacher's reformulation of all or part of a student's utterance, minus the error" (Lyster & Ranta, 1997, p. 46). In contrast, prompts are usually taken to mean either metalinguistic feedback and/or elicitation (Lyster, 2004).
There is a plethora of theories that support the role of noticing in corrective feedback. The Interaction Hypothesis (Long, 1996) proposes that interactional feedback can raise learners' awareness of language, resulting in an increase of attention to form and can heighten the inclination of noticing the mismatches between the nontarget-like form and the target-like form in the modified output. Based on this hypothesis, many interactionist researchers have ascribed the facilitative effects of CF to cognitive processes, in particular noticing induced by CF during interaction (Ellis, 1991;Gass, 1997;Long, 1996;Pica, 1994).
The role of noticing in the Interaction Hypothesis is corroborated by the Noticing Hypothesis (Schmidt, 1990) which argues that awareness at the level of noticing is the necessary and sufficient condition for acquisition to take place. Schmidt (1990) claims that input will turn into intake if and only if learners notice the input they are exposed to. In addition, the role of noticing in L2 development is also grounded on the Comprehensible Output Hypothesis (Swain, 1985(Swain, , 1995(Swain, , 2005. According to this theoretical perspective, learners when producing the output may become aware of, or notice, the linguistic problem in their output, triggered either by internal or external feedback, which pushes them to modify their output. In general, second language researchers have conceived that an explicit type of corrective feedback tends to promote higher levels of noticing than a more implicit type of CF (Ammar, 2008;Havranek, 1999;Lyster et al., 2013;Mackey et al., 2007;Nassaji, 2009;Panova & Lyster, 2002). However, evidence that supports this claim remains scarce. Moreover, it is uncertain as to which category of noticing is inclined to occur in response to a particular type of corrective feedback.

Factors Influencing the Noticing of Corrective Feedback
A small number of studies have investigated factors that influence learners' noticing of CF including the length of CF, the number of changes, the linguistic target of CF, and learners' developmental level. To date, research has found that shorter recasts with five or fewer morphemes tend to lead to a higher tendency of noticing the corrective function of recasts in learners (Philp, 2003). In addition, previous CF studies have reported that learners were able to recognize the corrective function of recasts which involve a single change more accurately than recasts with multiple changes (Egi, 2007;Loewen & Philp, 2006;Philp, 2003;Sheen, 2006). In the same vein, Kim and Han (2007) discovered that learners were able to notice the gap between the erroneous form and the target-like form better for recasts that contain one change to learners' erroneous utterance. With respect to the linguistic target of CF, several studies have shown that learners were more accurate in noticing recasts of lexical and phonological errors than recasts of morphosyntactic errors (e.g., Kim & Han, 2007;Mackey et al., 2000). Similarly, Trofimovich et al. (2007) showed that learners reported more noticing the gap for lexical errors than morphosyntactic errors following recasts.
On the other hand, there are a handful of studies that examined how cognitive factors such as working memory (WM), attention control, and analytical ability affect noticing. Mackey et al. (2002), for example, found that the relationship between verbal reports of noticing and WM scores was only marginally significant. In a similar vein, Trofimovich et al. (2007) found no significant relationships between phonological memory, working memory, attention control, and analytical ability and noticing. However, a survey of the literature has uncovered that individual learners' developmental level pertaining to the target of CF is a crucial factor that affects learners' ability of noticing (Carpenter et al., 2006;Ellis & Sheen, 2006;Philp, 2003;Schmidt, 1990;Trofimovich et al., 2007).

Effects of Type of Corrective Feedback on Noticing
Previous studies have shown that CF provided when learners are engaged in communicative tasks during classroom interaction aids learners to notice their errors. For example, Sakai (2004) examined the relative efficacy of recasts and models in promoting noticing of errors in university Japanese EFL learners. The results revealed that recasts were relatively more effective in promoting noticing of errors than models. However, there was no significant difference between the two types of CF in their learning outcomes. One plausible explanation is that both CF treatments, recasts, and models supplied learners with positive evidence of what is acceptable in the target language.
In another study, Sakai (2011) investigated the effects of recasts and no CF on learners' noticing of errors. Analyses of stimulated recall reports showed that recasts were effective in promoting noticing of errors. The verbal reports also revealed that learners were able to notice their errors with or without the provision of CF. The findings appear to support the Comprehensible Output Hypothesis which posits that in the course of producing L2 output, learners may at times become aware of their own linguistic problems. Noticing their own linguistic problems will then encourage them to modify their output which results in L2 development. By isolating noticing triggered by recasts and noticing resulting from the production of L2 output, Sakai (2011) showed that recasts were able to provide both positive and negative evidence.
In another CF noticing study, Kartchava and Ammar (2014) compared noticing triggered by three corrective feedback types (i.e., recasts, prompts, and a combination of recasts and prompts) in a quasi-experimental study. The results showed that learners treated with prompts or prompts mixed with recasts were able to promote more noticing than recasts. A possible explanation for this is that prompts which push learners for self-repair make the corrective intent of CF more salient, thus increasing learners' noticing of the error or corrective intent.
In a more recent study, Fujio (2017) explored the cognitive processes underlying Japanese foreign language learners' responses to recasts and prompts. Analyses of stimulated recall reports revealed that a range of cognitive activities was linked to successful self-repairs following prompts, including recognition of error, search for knowledge, retrieval of knowledge, and correct applications of grammatical rules. As for recasts, the study found that recasts were able to promote noticing activities such as recognition of error, recognition of the corrective function of recasts, and recognition of recasts as the correct form and awareness of the correct conjugation rule, which lead to successful repairs. The results of the study imply that different types of CF may be associated with different noticing processes.
As with any verbal reports elicited by retrospective methods, such as stimulated recall, the absence of verbal reports may not constitute the absence of noticing. In the case of prompts, it is rather surprising to note that the prompt group in the study (Fujio, 2017) did not report on noticing the correct conjugation rule. This was despite the fact that the operationalization of prompts included metalinguistic information in the study. This phenomenon perhaps can be explained by two factors. The first is that the study employed an open recall to elicit learners' thought processes when exposed to feedback, where learners were given the free rein to report on their noticing activities or the lack of them during the feedback episode. Second, the cognitive orientation (i.e., rule-based vs. meaning-based) of the learners in the prompt group could have influenced whether they noticed the correct conjugation rule when treated with prompts. Equally noteworthy is that although the verbal reports showed that recasts provided positive evidence to learners upon the encounter of the non-target-like polite past adjective form, there was no report of noticing the gap. This was probably due to the methodological procedures used to elicit verbal reports.
In summary, while CF noticing studies have focused mainly on recasts, the number of research studies that analyse noticing induced by other CF techniques, such as prompts, remains relatively smaller. Fewer still are research studies that compare noticing of CF induced by different types of corrective feedback, such as recasts versus prompts. On the basis of the small number of comparative CF noticing studies, coupled with a wide array of methodological variations across studies, such as the type of CF, operationalization of CF, the linguistic target of CF, noticing elicitation procedures, noticing categories, and instructional context, no firm conclusions with regard to the relationship between the type of CF and noticing have been reached. Perhaps what is needed is more research that examines the effects of CF techniques on noticing categories that are conceptually defined in a more refined manner.

Participants
The 105 participants of the study were learners at a government secondary school in Sabah, a Malaysian state located in the northern part of Borneo. Based on the grades they attained in their English Language Examination in the Lower Secondary Assessment (a national public-school examination taken in the third year of secondary school), half of the learners were assessed as being at the high proficiency level, and the other half at a low proficiency level. All the learners would have studied English through classroom instruction for an average duration of nine years (six years in primary school and three in secondary school). The majority of the learners came from two first language (L1) backgrounds, Malay and Chinese. For the experimental groups, the 16-year-old learners were equally distributed into two CF groups of recasts and prompts, with equal numbers of high proficiency and low proficiency students in each group.

Procedures
The study adopted a quasi-experimental design. The procedures of the quasiexperimental research are shown in Figure 1. The quasi-experimental study was conducted in a language laboratory for 14 weeks. Prior to the study, participants underwent training in diary writing where they were asked to make diary entries based on the corrective feedback, they received in their weekly English Language lessons three times per week, over a duration of four weeks. In the fourth week, they were also asked to attend an explicit teaching session on the English past tense where they were taught the form, meaning, and use of the simple past. At the end of the fourth week, the pre-tests were administered.
The treatment phase began in the fifth week and lasted for a duration of seven weeks. During the treatment phase, participants were asked to perform four communicative tasks which aimed to elicit the correct use of the simple past tense in English. The CF treatment was provided to the experimental groups according to the type of CF designated to participants, while the control group performed the communicative tasks without receiving any CF. Immediately after the completion of each treatment session, participants were instructed to reflect on the CF episodes and write their diary entries.
In the following week, participants were given post-tests to assess the immediate effects of the CF treatment. After the administration of the post-tests, about 40 of the experimental participants were asked to perform stimulated recall interviews. The delayed post-tests were administered approximately two weeks after the completion of the CF treatment, to measure the long-term effects of CF. Immediately after the delayed post-tests, participants answered the exit questionnaire, which aimed to probe any extra-experimental factors that might have affected the results of the study.

Target Linguistic Feature
The target linguistic feature in this study is the English past tense, which encompasses the use of the regular and irregular past tense forms. The simple past tense was chosen based on several reasons. First, verb tense errors are amenable to corrective feedback (Ferris, 1999;Ferris et al., 2000;Ferris & Roberts, 2001). Second, it is a common linguistic error committed by Malaysian English as a Second Language (ESL) learners (Maasum et al., 2012;Ting et al., 2010).

Treatment Tasks
Four 20-minute treatment tasks which allowed the use of the past tense were devised for the administration of treatment in the research. The treatment tasks consisted of three narrative tasks. These were a story-retelling of Cinderella, Johnny's most productive weekend and The day when I lost something I loved very much, and a role-play.

Measures of Learning
Three different tests were employed to measure learners' accuracy in the use of the simple past tense in obligatory contexts. The tests consisted of an oral production test, a written production test, and an untimed grammaticality judgment test. In the oral production test, learners were required to retell a story after listening to an audio of a real-life story. In the written production test, learners were required to produce a 200word essay on the topic 'How I spent my last school holidays', beginning with "In the first few weeks of the school holidays…". The untimed grammaticality judgment test contained 24 statements that target the use of the simple past, and the learners were asked to judge the grammatical accuracy of the statements. The pre-tests and post-tests of the three assessment tests differ only in the order of the past tense items, while the delayed post-tests contain not only the treatment items but also novel items.

Elicitation Measures of Noticing
Noticing is operationalized as the availability of noticing reports during a corrective feedback episode, whether in written or spoken form. To strengthen the veridicality and the reliability of noticing reports, the study employed a triangulated method in assessing learners' noticing of corrective feedback. The noticing reports were collected using three elicitation measures including diary writing, stimulated recall interviews, and exit questionnaires.

Diary writing
Diary writing is a retrospective and introspective method used to explore learners' cognitive processes that take place during task-based interactions. After completing each communicative task, participants were instructed to reflect on the feedback events and make a diary entry based on seven questions that aim to probe learners' noticing of teacher feedback (see Appendix A for Diary Writing Guidelines).

Stimulated recall interview
In the present study, stimulated recall interview is defined as a retrospective and introspective method designed to probe learners' noticing of teacher feedback. Using video recordings as the stimuli, the stimulated recall interviewer prompted learners with questions that probe into learners' thoughts during the time of feedback.

Exit questionnaire
The exit questionnaire is an awareness questionnaire which was administered after the experiment. It was designed to investigate whether participants were aware of the purpose of communicative tasks and to elicit what learners noticed during the experiment.

Coding of diary data
The noticing data derived from diary writing were coded in a dichotomous manner according to the noticing categories as shown in Table 1.

Noticing category Description a) Notice the corrective intent
A verbal report of noticing the teacher's corrective intent (i.e., an affirmative answer) or noticing one has made a mistake/an error or a description of the noticing experience. b) Notice the target/form of teacher feedback A verbal report of noticing the target/form of teacher feedback (i.e., an affirmative answer), or a verbal report of stating the target/form of teacher feedback or a description of the noticing experience. c) Notice the gap A verbal report of noticing the difference between the incorrect form the learner produced and the correct form provided/prompted by the teacher (i.e., an affirmative answer) or a description of the noticing experience. d) Notice the rule A verbal report of noticing the rules or a description of the rules related to the use of the past tense (e.g., for regular past forms, add -ed to the stem; for irregular past forms, change the vowel of the stem; or any relevant information related to the usage of the past tense).

Coding of stimulated recall data
The stimulated recall data were transcribed and coded according to these categories: a) Noticing the form: [+F] (Schmidt, 1995, p. 29) was operationalized as a "verbal reference to the target structure (i.e., the past tense verb form) without any mention of rules". b) Noticing the rule: [+R] (Schmidt, 1995, p. 29) was operationalized as a description of the rule that governs the correct usage of the target structure including what is required of learners (i.e., to change the base form to the past tense to reach the correct verb form in obligatory past-time contexts). c) No awareness report: [NAR]. Learners who showed no verbal reports of awareness were those mentioning the task and the advantages of the task they had just completed without any mention about the target structure or the rule that governs the correct usage of the target structure.

Inter-rater reliability for noticing data
For the diary writing data, the main researcher coded and scored all the diary writing scripts (see Appendix B for Samples of Coded and Scored Diary Writing Noticing Data). A second independent rater then coded and scored a randomly selected 25% of the diary writing data. To validate the diary writing data, the researcher crosschecked noticing reports in learner diaries with noticing reports of the same category assessed through other elicitation procedures (i.e., stimulated recalls and exit questionnaires). The first and second raters achieved 100% agreement in rating the diary writing data.
For the stimulated recall data, the main researcher transcribed and coded all the stimulated recalls made by the experimental groups. A second independent rater viewed, transcribed, and coded a randomly selected 25% of the stimulated recalls. Whenever there were any discrepancies with regard to the interpretation and coding of a type of noticing report, the raters would discuss until they achieved consensus. The first and second raters achieved 100% agreement in rating the stimulated recall data.

Statistical analysis
A two-way independence chi-square test was used to see whether there were statistical differences between the groups (i.e., recasts vs. prompts) in their total frequency of noticing. Another two two-way group independence chi-square tests were performed to see whether there were statistical differences between the groups (i.e., recasts vs. prompts) in awareness of lower order and awareness of higher order. There is a statistical difference between the two experimental groups in their total frequency of noticing, noticing of the lower order, and noticing of the higher order, if p < .05. Table 2 shows the frequencies of noticing by the treatment group elicited through diary writing and exit questionnaire. The findings of the study show that the mean scores of recasts were relatively higher than those of the prompts across all noticing categories. What is also noteworthy is that the differences in mean scores between recasts and prompts for noticing the corrective intent and noticing the target of CF were small, less than one point. However, the differences in mean scores between recasts and prompts for noticing the gap and noticing the rule were slightly greater, between one to two points. What is also striking is that there were more cases of no noticing reports for noticing the rule than noticing the target of corrective feedback for both the treatment groups.

Noticing Reports Elicited through Diary Writing and Exit Questionnaire
Meanwhile, Figure 2 shows the percentage of noticing for each noticing category within each treatment group. As shown in Figure 2, recasts were able to induce higher percentages of noticing compared to prompts across all noticing categories.

Noticing the Gap
The diary entries revealed that participants might not perceive noticing the gap in dichotomies. Therefore, in the exit questionnaire, noticing the gap was measured in ranks (1: noticing the difference in 0-20% of corrective feedback items, 2: noticing the difference in 21-40% of corrective feedback items, 3: noticing the difference in 41-60% of corrective feedback items, 4: noticing the difference in 61-80% of corrective feedback items, 5: noticing the difference in 81-100% of corrective feedback items), resulting in 20 as the maximum score for noticing the gap in the exit questionnaire. Table 3 shows the frequencies of noticing the gap elicited through the exit questionnaire by the treatment group.  The results of the study reveal that there was a significant difference between recasts and prompts in noticing the gap. For the 41-60% rank of noticing the gap, prompts outweighed recasts in number and scores whereas for the 81-100% category, recasts trumped prompts. As the latter carries more weightage, overall recasts outperformed prompts for the noticing the gap category in the exit questionnaire, consistent with the results found in diary writing. In another word, participants treated with recasts were able to notice the gap in a higher proportion of the corrective feedback items compared to participants treated with prompts. Figure 3 shows noticing the gap elicited through the exit questionnaire by recasts and prompts. With reference to Research Question 1, whether recasts and prompts induce noticing, the results of the study showed that both recasts and prompts were able to induce noticing, albeit at different frequencies across the noticing categories.  Language and Education, 8(2), 416-441, 2021

Noticing Levels Measured by Total Frequency of Noticing
For the total frequency of noticing, the median score, 31 was used as a cutting point to distinguish between the high total frequency of noticing and the low total frequency of noticing. The descriptive statistics for noticing levels according to the total frequency of noticing are shown in Table 4.

Noticing Levels Measured by Hierarchical Order of Noticing
The median was used as a cutting point to distinguish between high noticing of lower order (i.e., noticing the form) and low noticing of lower order as well as high noticing of the higher order (i.e., noticing the rule) and low noticing of the higher order. The descriptive statistics for noticing levels according to the hierarchical order of noticing are shown in Table 5. For noticing levels measured by total frequency of noticing, the results of the chi-square test showed that recasts were able to induce a higher total frequency of noticing compared to prompts. Likewise, for noticing levels measured by hierarchical order of noticing, the results of the chi-square tests revealed that recasts were able to induce a higher frequency of noticing of the lower order (NLO) compared to prompts. Similarly, the results showed that recasts were able to induce a higher frequency of noticing of the higher order (NHO) compared to prompts. The summary of the chisquare test results is shown in Table 6.
With reference to Research Question 2, whether prompts induce higher levels of noticing than recasts, measured by a) total frequency of noticing and b) hierarchical order of noticing, the findings of the study suggest that prompts do not induce higher levels of noticing than recasts, measured by a) total frequency of noticing and b) hierarchical order of noticing. Noticing the rule recorded a greater difference between recasts and prompts compared to noticing the target of CF/form Note: * The difference between different types of corrective feedback is significant.
Rs> Ps = Recasts reported higher levels of noticing than prompts.

DISCUSSION
Overall, recasts were able to induce higher levels of noticing than prompts across all categories of noticing. As shown by the analysis of noticing data in Table 2, the difference between recasts and prompts lay primarily in the categories of noticing the gap and noticing the rule. Therefore, it is appropriate to focus the following discussion on reasons as to why these two treatment groups differed on noticing the gap and noticing the rule, in particular the reasons underlying why recasts induced higher levels of noticing the gap and noticing the rule than prompts in the experiment.
One of the plausible reasons is that participants treated with recasts (or recasts in short) were consistently exposed to exemplars (the target-like past tense forms) during the experiment period whereas participants treated with prompts were deprived of the well-formed target linguistic feature of CF. The adjacency of exemplars of past tense provided in the recasts and the non-target-like form of the past tense is likely to enhance the chances of noticing the gap between the learners' non-target-like interlanguage past tense and the L2 target-like past tense (Gass, 2003). Thus, the amount of noticing the gap among the recasts was observed to be higher than that reported by the prompts. Schmidt (2001) commented that the object that learners notice in the input is exemplars that abide by a certain principle or system, but not the principle or system itself. Once the noticing of these exemplars occurs, generalizations or induction across the exemplars can be formed without learners being aware of what the generalizations are.
Another possible reason is that noticing is affected by learners' prior knowledge. It is difficult for learners to achieve the correct past tense forms for lexical items or words which are not in their existing knowledge of L2, in particular for verb stems or past tense forms which are not in learners' wealth of lexicon. This phenomenon is prevalent when it comes to the accurate production of irregular pasts. This is supported by past tense acquisition theories, such as the words-and-rules theory (Pinker & Prince, 1994), which advocates that regular pasts are learned by rule and irregular pasts are learned by memorizing the irregular past forms. If learners were not able to produce the target-like form based on their non-target-like form, it is argued that in the absence of the target-like form, even when learners were treated with a more explicit corrective feedback (prompts), they were not able to compare the target-like form and their original non-target-like form, thus reducing their incidences of noticing the gap.
Prior knowledge or learners' existing knowledge of L2 is a crucial factor that determines the extent to which learners are able to notice the second language input they are exposed to. This claim is supported by Gass (1988Gass ( , 1997Gass ( , 2003 in her information processing framework of second language acquisition. Gass (1988Gass ( , 1997 proposes that apperception or noticing the gap is the catalyst or the first step towards restructuring the second language grammar. The process of noticing the mismatch between the learner's interlanguage form and the teacher's target language form is referred to as cognitive comparison (Doughty, 2001). Noticing the gap requires cognitive comparison. This means learners would have to compare incoming input with the representations stored in long-term memory or traces available in short-term memory. In order for learners to perform cognitive comparisons, learners need to possess a minimum number of linguistic resources (Doughty, 2001), without which it is uncertain as to how learners can become aware of the discrepancies between their interlanguage form and the L2 target-like form. In performing cognitive comparisons, learners are required to retrieve representations from long-term memory or traces of linguistic materials left in short-term memory. When learners were not able to retrieve linguistic resources from long-term memory or short-term memory due to a possible lack of prior knowledge, it is doubted how learners were able to notice the mismatch or the gap between their own erroneous form and the target-like L2 form. This probably explains why prompts in the current study reported a lower level of noticing compared to recasts.
An alternative explanation may be the recast group was made up of learners who were of a more advanced developmental level where the target linguistic feature was concerned. Conversely, it seems possible that more learners in the prompt group failed to notice the gap, not because they did not notice but they were developmentally not ready to notice the gap. Previous studies have shown that learners who are of a more advanced developmental level were able to perform noticing more accurately (Carpenter et al., 2006;Ellis & Sheen, 2006;Philp, 2003;Schmidt, 1990;Trofimovich et al., 2007). As Philp (2003) points out, the more linguistic resources the learner has, the better the ability to notice the gap. Schmidt (2001) comments that what is noticed in the input is exemplars of language rather than abstract rules that govern the use of the exemplars. If learners are aware of the rules, this implies that learners demonstrate awareness at the level of understanding, which differentiates between explicit learning based on conscious knowledge, insights, and hypotheses, and implicit learning based on unconscious generalization and abstraction processes (Schmidt, 1995). In the present study, noticing the rule involves either a brief description of the grammar rule that requires the use of the past tense in obligatory contexts or the change from the base form to the past tense. The higher levels of noticing the rule reported by participants treated with recasts could likely be attributed to the exemplars of past tense forms they were exposed to while completing the communicative tasks with the provision of recasts. In the treatment phase, participants who received the recast treatment were consistently and intensively exposed to exemplars of the past tense contingent upon the encounter of past tense errors in learner utterances. The repeated exposure to the instances of well-formed past tense forms is believed to have aided participants treated with recasts to be aware of the rule that underlies the correct usage of past tense forms, thus increasing their chances of noticing the rule, compared to participants treated with prompts who had access to exemplars only if they were successful in their attempts to produce self-repair. The findings of the study seem to suggest that it is the positive evidence in corrective feedback that contributes to noticing the rule when learners are exposed to the input. An alternative explanation is that some learners by default might be more meaning-oriented or grammar-oriented when solving communicative tasks. Hence, their attention was always focused on either meaning or form. In this case, learners' mental orientation may also possibly affect their report of noticing the rule.
The results of the current study, which showed recasts were able to promote higher levels of noticing the corrective intent than prompts are found to be contrasting to those reported in Kartchava and Ammar (2014). The difference in findings may be attributed to considerable variations in the operationalization of corrective feedback. For instance, recasts in the present study were operationalized as a full or partial reformulation of participants' erroneous use of the past tense while in Kartchava and Ammar (2014), recasts transcended beyond a full or partial reformulation of learners' non-target-like utterance to include interrogative reformulation as in "(1) Where did you say he went yesterday? (2) He went to the movies yesterday. Did he go alone or with someone?" (p. 433). Upon closer examination, interrogative reformulation differs from full or partial reformulation in that it directs learners' attention to content more than form. The variations in the operationalization of CF techniques may have influenced the noticeability of CF types, leading to conflicting results in different CF noticing studies. Furthermore, the amount of noticing triggered by different types of CF may be subject to the noticing category that is being measured. For the category of noticing the corrective intent, Kartchava and Ammar (2014) found that prompts outperformed recasts. However, it remains an empirical question if prompts were superior to recasts on the measurement of other noticing categories, such as noticing the gap.

CONCLUSION
In conclusion, the study found that both recasts and prompts were able to induce a range of noticing categories which encompasses noticing the corrective intent, noticing the form, noticing the gap, and noticing the rule. However, contrary to what was expected, recasts which are a more implicit type of CF were found to induce higher levels of noticing than prompts. Although it is sensible to attribute the higher levels of noticing reported by recasts to the fact that participants treated with recasts were exposed to exemplars that provided them with positive evidence (what is permissible in the target language); it is noted that other intervening factors might have a bearing on the levels of noticing. This included prior knowledge, which is assumed to be related to their proficiency level, the laboratory setting, and the target linguistic feature of CF. Similarly, the lower levels of noticing reported by prompts may be related to the lack of prior knowledge which is an important factor that affects noticing the gap (Gass, 1988(Gass, , 1997(Gass, , 2003. Therefore, the findings that recasts lead to higher levels of noticing need to be treated with caution given the possibility that there were other variables which could influence noticing levels. However, the results of the study indicate that prompts, which push learners to make self-correction, had a greater tendency to induce noticing the corrective intent. Recasts, on the other hand, which supply learners with the acceptable L2 form, showed a greater likelihood of promoting 'noticing the gap'. Furthermore, the findings of the study suggest that recasts are inclined to exhibit a greater tendency than prompts in promoting noticing the rule. It is hypothesized that under circumstances where learners are aware of the corrective intent and the target of CF, CF types that are input-providing, such as recasts, may induce higher levels of noticing contributed by noticing the gap and noticing the rule. While the current study has provided evidence to show that an implicit type of CF like recasts can potentially lead to a higher level of noticing compared to an explicit type of CF; the inherent properties of corrective feedback that lead to the higher total frequency of noticing, higher frequency of noticing the form/target of CF and higher frequency of noticing the rule still require further investigation.
The findings of the current study have revealed that recasts were able to induce higher levels of noticing the form, noticing the gap, and noticing the rule. Several pedagogical implications can be drawn from these findings. First, the findings of the study suggest the benefit of employing recasts in L2 classroom interaction. In addition to the non-obtrusive nature of recasts, and also the capacity of providing scaffolding in language learning, recasts are predicted to promote higher levels of noticing in particular noticing the gap, owing to the accessibility to the correct form of the target structure embedded in recasts. Second, as noticing is often conceived as the impetus for a restructuring of L2 grammar (Gass, 1988(Gass, , 1997(Gass, , 2003, and recasts were found to promote higher levels of noticing compared to prompts, the findings may serve as a basis for L2 teachers in making an informed decision with regards to the type of CF deemed suitable for L2 learners. Third, the study has lent evidence to debunk the myth that prompts tend to promote higher levels of noticing in L2 learners. This implies that low proficiency learners may not necessarily benefit better from prompts due to the absence of the target-like L2 form in prompts. Fourth, to heighten L2 learners' awareness of CF, we would like to propose a two-pronged noticing strategy which includes first, to enrich their existing knowledge, and second, to consolidate the input they are exposed to with the target linguistic feature. The strategy works towards increasing L2 learners' wealth of knowledge, which is expected to enhance L2 learners' ability to notice the gap in second language learning. Finally, as shown in this study, it is important for L2 instructors to design communicative tasks, which require learners to be productive in the output of the target structure, to allow their linguistic errors to surface before the provision of CF in second language classrooms.
It must be acknowledged that the present study has its limitations. First-person verbal reports were elicited by means of immediate recalls such as in diary writing, and delayed recalls such as in stimulated recall interviews and exit questionnaires. The longer time lapse between experiencing CF and the time that the verbal report was completed in the delayed recalls may have given rise to the threat of memory decay. This is why participants might have forgotten what they had noticed during the experiment phase. Moreover, the absence of verbal reports may not necessarily reflect the absence of noticing. For instance, participants who were not able to verbalize the rule could have been aware of the rule but unable to verbalize it. Getting participants to engage in communicative tasks without CF, as in the control group may also have increased participants' noticing of form. Due to time constraints, the study was not able to examine noticing resulting solely from the production of the target linguistic feature. Although the study has provided empirical evidence to support the claim that input-providing CF, recasts may promote higher levels of noticing compared to outputprompting CF, prompts, the findings of this study remain inconclusive. More studies are warranted to ascertain the relationships between type of CF and level of noticing. To that end, the present study may be replicated using other input-providing CF techniques, for example, models and explicit correction. In addition, future research may also investigate the relative effects of CF techniques on different categories of noticing, in particular noticing the gap and noticing the rule and the extent to which higher levels of these noticing categories lead to gains in second language acquisition.

ACKNOWLEDGMENTS
This article reports part of the study conducted for the first author's Doctoral research. The study was made possible with a scholarship granted to the first author by Universiti Teknologi MARA, Malaysia. We would like to express our gratitude to the participating teachers and students in Sabah for their assistance and cooperation throughout the study. (pp. 171-195 To that end, I have decided to employ diary writing as a research tool to collect data. Diary entries written by students will be used as a means to explore the inner cognitive processes taken place during and immediately after feedback was given. Therefore, I would like to invite you to participate in this activity, which is scheduled to run for four weeks. At the end of each English lesson, you are required to reflect on what has happened during the class and write down your thoughts and feelings related to the classroom experience.

Conversational interaction in second language acquisition: A collection of empirical studies
Guidelines for keeping a language learning diary (adapted from Carroll, 1994): 1. What is a diary?
A diary is a record of your daily language learning experience. In order to do this, you need to reflect on your learning on a daily basis.

Why should I keep a diary?
There are several reasons why students are encouraged to keep a language learning diary. a) Keeping a diary provides you the opportunity to express yourself in written English on a regular basis, thus improving your writing skills. b) Through diary entries, you get to think and reflect on your learning experiences and progress in learning a second language. c) Through diary entries, you get to share your expectations and needs of learning a second language with your teacher.
3. What should I write in my diary?
At the end of each English lesson, write for at least 15 minutes on your thoughts and feelings about what you experienced in the classroom. Specifically, I would like you to reflect on the feedback episodes in the classroom and provide answers to the following questions: a) Was the feedback directed at you or your classmate? b) Whether you recognized the corrective intent of the teacher's feedback? c) i) Whether you noticed the target of the teacher's feedback? (notice the form) ii) Was the target of the teacher's feedback new to you or you have learned it before? d) Whether you were able to describe the difference between the incorrect form you produced and the correct form provided/prompted by the teacher? (notice the gap) e) Whether you were able to produce the correct form following the feedback?
(intake) f) Whether you were able to describe the grammar rule that helps you reach the correct form? (If your answer is positive, please state the rule that you knew) Please note that your answers to the questions above can be either positive (i.e., you did notice something and reported what you noticed) or negative (i.e., you did not notice anything and reported the reason why you did not notice). However, you are discouraged to provide just a 'yes' or 'no' answer. In order to help you remember what has happened during the class, you may choose to record your observations when the class is in progress.

How should I write my diary? a)
There is no fixed format as to how the layout of your diary entries should look like. However, every diary entry should state clearly the date and time of the class and the date and time when the diary entry is made. b) Every diary entry should be in the form of continuous writing, and not notes form. c) There is no rule as to how much you should write. The length of your diary entries depends largely on the amount of observation that you made in the classroom. d) Students are allowed to take notes if this helps them remember what has happened in the classroom interaction better. However, it is not advisable for a student to do so when he/she becomes the respondent to the teacher's corrective feedback due to the fact that it is likely to interfere with the response of your feedback. e) Students are free to make their own comments on the feedback events initiated by the teacher in the classroom. However, students should support their assertion with evidence or examples found in the class. f) Students are encouraged to make their diary entries immediately after the class or on the same day as the class as it seems easier to write when the impression of the classroom events is still fresh. g) Students should spend about 10-15 minutes to reflect and write on the learning experience in the classroom. h) The diary-writing activity is scheduled to run for four weeks. Table B2 continued… tense. But some word also not change. For example, the word put will also same when it is past tense.