Learners’ Beliefs as Mediators of What Is Noticed and Learned in the Language

Each teacher was assigned to a treatmentcondition that fit his corrective feedback style, and each providedfeedback in response to errors with the past tense and questions inthe past.. T

Learners’ Beliefs as Mediators of What

Is Noticed and Learned in the Language Classroom

EVA KARTCHAVA AND AHLEM AMMAR

Universit
e de Montr
eal

Montr
eal, Canada

The goal of this study was to determine whether learner beliefsregarding corrective feedback mediate what is noticed and learned inthe language classroom The participants were four groups of high-beginner college-level francophone English as a second languagelearners and their teachers Each teacher was assigned to a treatmentcondition that fit his corrective feedback style, and each providedfeedback in response to errors with the past tense and questions inthe past Participants (N= 197) completed a beliefs questionnaire,and half (n= 99) took part in the intervention that followed Beliefswere probed using a 40-item questionnaire, and average belief scoreswere calculated for each learner These were then correlated bothwith the noticing reported on an immediate recall measure and withthe test scores on picture description and spot-the-differences tasks.The results reveal four common beliefs, two of which mediated thenoticeability of the supplied feedback, but none of which impactedthe learning outcomes

doi: 10.1002/tesq.101

Individual differences (IDs)—such as intelligence, language aptitude,motivation, risk taking, and beliefs—are thought to influence andeven to predict second language (L2) learning success (Breen, 2001;D€ornyei, 2005; Fox, 1993; Gardner & MacIntyre, 1992; Horwitz, 1985,1999; Sawyer & Ranta, 2001) However, little is known about theireffect on the processes of L2 acquisition and, more specifically, abouttheir impact on the ways learners process language instruction (Sheen,2011) Corrective feedback (CF) is an integral part of language studyand allows teachers to provide information about the grammaticality

of a learner’s oral and written production In particular, CF is said tofacilitate learners’ noticing the difference between their incorrectutterance and the target form, leading to L2 development (Schmidt,

1990, 1995) For feedback to be effective, however, learners need to

recognize its didactic focus (Carroll, 1997) by understanding that theteacher’s correction is targeting the form, and not the meaning, of anill-formed utterance As such, it makes sense to question whether IDsinfluence the noticeability and thus the effectiveness of oral CF.Studies on the impact of IDs on the effectiveness of oral CF haveconsidered differences in learners’ age (Mackey & Oliver, 2002); profi-ciency (Ammar & Spada, 2006; Philp, 2003); and attention, memory,and language aptitude (Ammar & Sato, 2010; Mackey, Philp, Egi, Fujii,

& Tatsumi, 2002; Trofimovich, Ammar, & Gatbonton, 2007), thoughthese investigations are scarce (Russell & Spada, 2006) Together, theysuggest that age, proficiency, and a number of cognitive factors affectlearners’ ability to notice and, as a result, profit from oral feedback,more specifically from recasts, that is, a CF technique that providesthe target form in response to an error In terms of age, childrenbetween 8 and 12 years old are able to respond to feedback on ques-tions in dyadic interactions with adults and to benefit from it sooner(leading to more immediate changes in their restructuring and inter-language) than their adult counterparts (Mackey & Oliver, 2002) Asfor proficiency, high-level learners tend to notice and benefit fromrecasts more readily than low-proficiency learners (Ammar & Spada,2006; Philp, 2003) Finally, it has been demonstrated that learners with

a large working and phonological memory (Mackey et al., 2002) aswell as a broad attention span (Ammar & Sato, 2010) are more likely

to notice recasts than prompts (i.e., CF techniques that cue the lem in the learner’s utterance and encourage self-correction)

prob-Only two affective variables, anxiety and learner attitudes, have beeninvestigated in relation to CF effectiveness, but not to its noticeability(Sheen, 2008, 2011) These two variables, along with analytical ability(a cognitive factor), were examined in relation to both oral and writ-ten feedback The results show that although all three factors medi-ated the effectiveness of different types of CF, their impact depended

on the mode in which the feedback was delivered and on the specific

CF type Of the affective factors, anxiety proved to be a variable in oralfeedback, but did not play a role in written feedback Learner attitudestowards CF, on the other hand, figured much more in the case of writ-ten than oral feedback These attitudes measured the degree to whichthe participants were willing to accept feedback and whether they saw

it as helpful and important; their perceptions towards grammaticalaccuracy were also investigated In terms of CF type, learners withlower anxiety outperformed those with higher anxiety in the case oforal metalinguistic CF (operationalized as the teacher’s provision ofthe correct form following the error, together with a metalinguisticexplanation; Sheen, 2011), but anxiety was not a factor in the effective-ness of oral recasts (defined as “a teacher’s reformulations of a

student’s erroneous utterance, without changing the meaning of thestudent’s original utterance in the context of a communicative activ-ity”; Sheen, 2011, p 62) Similarly, learner attitudes mediated gainscores for the learners in the oral metalinguistic group, but not forthose in the oral recast group These findings suggest that affectivevariables influence the effectiveness of CF in the classroom, but it isstill not clear whether they impact the noticeability of oral feedback.With this in mind, the current study was designed to determinewhether the noticeability and benefits of feedback are dependent ondifferences in learner beliefs about corrective feedback.

The consideration of learner beliefs in general is important becausethey “have been recognized as learner characteristics to count withwhen explaining learning outcomes” (D€ornyei, 2005, p 214) and havebeen empirically shown to be constant among learners and consistentacross different language groups (Horwitz, 1985, 1987, 1988) Investi-gation into learner beliefs about CF in particular is necessary because,

to date, no published studies have considered learners’ perceptions ofwhat feedback is, how it is best delivered, and the factors that make iteffective Although learner beliefs have been claimed to underlie manyaspects of learner behaviour and learning outcomes, no studies havelooked into learner beliefs about CF independently from other lan-guage-related constructs; rather, they have allotted the concept second-ary importance (e.g., Loewen et al., 2009; Schulz, 1996, 2001) In fact,Loewen et al (2009) show that learners view grammar instruction and

CF as distinct categories and call for future research to consider thisdifferentiation Finally, there have been a number of calls for secondlanguage acquisition (SLA) research to consider IDs as mediators ofwhat CF is noticed and consequently learned from in the classroom(e.g., Lyster & Saito, 2010; Russell & Spada, 2006) Russell and Spada(2006) observe that

few studies have investigated the impact of individual learner factors inrelation to CF [and that] until more studies are done to isolatethese variables and investigate them in a series of studies in classroomsand laboratories, they remain compelling arguments without adequatesupporting evidence (p 155)

Research that has considered the noticeability of CF has primarilyfocused on recasts, suggesting that the learners’ ability to recognizetheir corrective intent is limited by error type (Mackey, Gass, &McDonough, 2000), length (Philp, 2003) and explicitness of the recast(Ammar & Sato, 2010); proficiency level (Philp, 2003); working mem-ory capacity (Mackey et al., 2002); and attention-switching ability(Ammar & Sato, 2010) Comparisons of noticing between recasts andprompts are rare; only two studies have broached the subject (Ammar,

2008; Mackey et al., 2000) Although an early attempt of this tion was inhibited by an unbalanced distribution of the techniques(Mackey et al., 2000), Ammar’s (2008) systematic comparison ofrecasts to prompts reveals that the latter is more noticeable than theformer Yet it is not clear whether beliefs affect learners’ ability tonotice CF.

evalua-The effectiveness of recasts and prompts seems to be context cific Although early laboratory studies (e.g., Leeman, 2003; Mackey &Philp, 1998) found that recasts facilitate L2 grammar learning, investi-gations that compared recasts to other feedback types either foundrecasts more effective (Long, Inagaki, & Ortega, 1998; Mackey & Philp,1998) or yielded no differences between the CF types (e.g., Lyster &Izquierdo, 2009; McDonough, 2007) In the classroom-based studies,however, prompts yielded the most gains (e.g., Ammar & Spada, 2006;Ellis, Loewen, & Erlam, 2006; Lyster, 2004; Yang & Lyster, 2010) Thefact that recent laboratory studies have failed to differentiate the effec-tiveness of the CF techniques suggests that recasts and prompts differ

spe-in the type of learnspe-ing opportunities they afford (Lyster & Izquierdo,2009; Lyster & Saito, 2010) That is, whereas recasts provide correctreformulations of the ill-formed utterance and invite learners to inferthe corrective intent, prompts overtly signal the presence of an error,cue its location, and push students to self-correct Still, there is noindication of whether learner beliefs about CF impact what is learned

in the classroom

In light of this, it is necessary to examine learner beliefs regarding

CF and to determine whether differences in learner perceptions ate their ability to notice and learn from the feedback they receive inthe classroom If beliefs influence L2 learning, then there could be alink between beliefs and noticing and between beliefs and learning.Hence, this study poses the following research question: Do learnerbeliefs about CF mediate their noticing and learning of L2 norms?

medi-METHODOLOGY

Participants

Three highly experienced English as a second language (ESL)teachers and their 197 high-beginner college students (mean age:20.75 years old), who came from intact classes and spoke French astheir first language, participated in this study All 197 participantscompleted a beliefs questionnaire (see the Learner Beliefs subsection

in the Data Collection Measures section for details), and 99 from sixintact classes took part in the intervention to ensure experimental

comparability Table 1 breaks down the number of learners acrossgroups and teachers The participants’ prior exposure to ESL instruc-tion amounted to 120 hours in primary school and 670 hours in highschool The college English classes they were enrolled in met once aweek for 3 hours, 2 of which were spent in the classroom and 1 in thelanguage laboratory The courses were taught by the same teachers forthe entire term.

The teachers, bilingual in English and French, were observed andinterviewed prior to the investigation in order to identify if and howeach provided CF The three teachers addressed most of the learners’errors but did so using different methods Whereas one teacher(Albert) responded to errors primarily with recasts, another (Brian)showed a clear preference for prompts The third teacher (Charles)consistently alternated between recasts and prompts Because it wasnot possible to find another teacher from the same college whoprovided no feedback, the first author taught the control group

Corrective Feedback Conditions

The CF conditions were operationalized in accordance with the CFtypes documented by Lyster and Ranta (1997) and Sheen (2004).Recasts were operationalized as the teacher’s reformulation of a lear-ner’s incorrect utterance The Recast teacher was allowed to react with

a full, partial, interrogative, or integrated reformulation For example,

in response to a student’s utterance*He go to the movies yesterday, any ofthe following approaches could be adopted:

Full reformulation: Okay He went to the movies yesterday

Partial reformulation: (He) Went

Interrogative reformulation: Where did you say he went yesterday?

Integrated reformulation: He went to the movies yesterday Did he

go alone or with someone?

Prompts were defined as techniques that elicited the correct formfrom the learner These techniques included (1) repetition, wherethe teacher repeated the student’s incorrect utterance, either as awhole with rising intonation or partly by zooming in on the errorwhile withholding the correct form; (2) elicitation, where the teacherrepeated part of the learner’s utterance and paused at the error toprovide a clue as to the problem, as well as to invite the student

to self-repair; and (3) metalinguistic information, where the teacherprovided metalinguistic clues but did not provide the correct form,thus pushing the learner to self-correct Hence, the Prompt teachercould adopt any of the following in response to *He go to the moviesyesterday:

Full repetition: He go to the movies yesterday?

Partial repetition: Go yesterday? Go?

Elicitation: He what [stressed] yesterday?

Metalinguistic information: It happened yesterday So what should

we say? (How do we form the past in English?)

The Mixed group’s teacher was asked to alternate between recastsand prompts as equally as possible during the activities The need for

a combination of techniques stems from Lyster and Ranta’s (1997)early observation that language teachers regularly use combinations offeedback types to address learners’ errors As such, the use of a mixedgroup is warranted not only to provide pragmatically justified evidencefor the noticeability and effectiveness of CF types, but to also answerthe call of “moving away from dichotomous comparisons of CF strate-gies that isolate CF from other relevant instructional variables andtowards an examination of combinations of CF types that more closelyresemble teachers’ practices in classroom setting” (Lyster, Saito, &Sato, 2013, p 30)

Linguistic Targets

The past tense and questions in the past were chosen as the guistic targets for the current study because both (1) are problem-atic for ESL learners, regardless of first language (L1); (2)represent different levels of complexity (DeKeyser, 1998, 2005); (3)are subject to L1 interference (Ammar, Sato, & Kartchava, 2010;Collins, 2002); (4) occur frequently in the input (Doughty & Varela,

lin-1998), facilitating their elicitation during communicative tasks(McDonough, 2007); and (5) have been shown to be good candi-dates for learner improvement when they are targeted with CF (forquestions, see Mackey, 2006; Mackey & Philp, 1998; McDonough,2005; for past tense, see Doughty & Varela, 1998; Ellis et al., 2006;Yang & Lyster, 2010) Research has also shown that the noticing of

CF leads to language development, though this largely depends onthe technique employed and on the nature of the error (Ammar,2008; Ammar & Sato, 2010; Mackey et al., 2000)

Gatbon-Data Collection Measures

Learner beliefs To uncover learner beliefs about CF, a two-partquestionnaire was created In Part 1, demographic information on theparticipants was gathered, including their linguistic background Foreach of the languages spoken, the participants then indicated (1) wherethey learned it (i.e., classroom, home, other); (2) the number of yearsthey had been speaking it; and (3) how well they spoke, wrote, listened,and read in each using a scale of poor to excellent Part 2 of the question-naire consisted of 40 statements about CF, which were based on theoreti-cal and empirical findings in the CF literature (e.g., Horwitz, 1988, 1999;Mohamed, 2011; Schulz, 1996, 2001) Specifically, the statements, therationale for which is described next, centered on the (1) importanceand (2) expectations for feedback; the (3) mode, (4) timing, and (5)amount of CF; as well as (6) recasts, (7) prompts, and (8) the manner inwhich and (9) by whom the CF should be delivered On a scale of 1 to 5(1 = strong disagreement and 5 = strong agreement) the participants indi-cated the degree to which they agreed with each statement The ques-tionnaire was written in French, and the items were randomized prior tothe administration during Week 1 of the term

Rationale for the questionnaire statements Researchers (e.g., Long,1996; Schmidt, 1990, 1995, 2001) and L2 learners (Chenoweth, Day,Chun, & Luppescu, 1983; Schulz, 1996, 2001) agree that CF is impor-tant in language learning In fact, learners across different contextsexpect to have their oral and written errors corrected (Schulz, 1996,2001; Sheen, 2011) and in large amounts (Chenoweth et al., 1983; Jean

& Simard, 2011) If they are not, their motivation to learn an L2 maydecrease and they may question their teacher’s credibility (Horwitz,1990; Schulz, 1996, 2001) In terms of timing, although some methodol-ogists believe that CF interrupts the communicative flow (e.g., Bartram

& Walt, 1991; Harmer, 2007) and recommend that teachers addresserrors at the end of a task (e.g., Willis, 1996; Hedge, 2000), learnersvalue teachers who provide immediate feedback to oral errors (Brown,2009) It is not clear how much feedback is necessary to affect learning,but methodologists promote and some learners prefer (Lasagabaster &Sierra, 2005) selective correction (e.g., Harmer, 2007), suggesting thatonly errors that are prevalent and impede the message transmissionshould be treated Regarding learner preferences for specific CF tech-niques, the research is scarce, with studies showing that lower profi-ciency learners favor prompts over recasts (Mohamed, 2011; Yoshida,2008), but that more advanced-proficiency learners prefer recasts toprompts (Brown, 2009) Similarly, little is known about learners’ prefer-ences for who should correct—the teacher or the students WhereasHendrickson (1980) argues for learners’ involvement in correctingtheir own errors, Hedge (2000) claims that this might not always be pos-sible due to the learners’ inability or unwillingness to effect correction.Noticing Immediate recall (administered during class activities)and lesson reflection sheets (completed at the end of each session)were used to measure the noticeability of the supplied feedback Theimmediate recall consisted of the first author lifting a red card follow-ing some CF instances, prompting each learner to write down (inEnglish or French) their thoughts about what was happening in theclass The exact instructions for the task were as follows: “Each timeyou see the red card, write what you are thinking in relation to the les-son.” Examples of the immediate recall corrective episodes are pre-sented in Table 2, and the breakdown of the number of episodesacross the three groups is shown in Table 3 To ensure a comparabledistribution of errors corrected with recasts and prompts, the totalnumber of episodes and distractors was increased for the Mixedgroup The distractors were used to divert the participants’ attentionand were limited to disciplinary comments and task instructions.Whereas the immediate recall protocols were analyzed for the types ofnoticing reported and to calculate learner average noticing scores, the

lesson reflection sheets, adapted from Mackey (2006), were analyzedqualitatively to provide additional evidence of noticing To ensure thatthe coding was representative, an independent rater analyzed and cate-gorized part of the data; inter-rater reliability was 93% based on simpleagreement.

Learning To measure learning outcomes, all groups completedtwo tasks (one per target) before and immediately after the

TABLE 2

Immediate Recall Corrective Episodes

CF type Past tense Questions

Recast “WE WENT TO THE AIRPORT” “DID SHE TALK WITH YOU?”

S1: What did you do when you woke up? T: Where did you meet Britney

Spears?

S2: Ahh, we don’t do anything special T: Shopping?

T: You did not do anything special? S1: Yeah [laughter] at the

T: You went to the airport? S2: She talk with you?

S2: No, we go to the airport

[looks at the teacher confused] Yes.

S1: No.

T: You went to the airport? T: Did she talk with you?

S2: Yes S1: No.

Prompt “WE TOOK THE AIRPLANE” “WHAT DID YOU DO?”

T: In the airplane you ate McDo?

[Class laughs.]

T: Questions ! S3: We bought the food and after

that we take airplane.

S1: What do you eat at McDonald’s? T: After that we (pauses, then gestures

the need for the past tense) “take” in

the past tense.

T: What (gestures the need for the past tense)

S3: Took S1: What did you eat?

T: Good.

TABLE 3

CF Immediate Recall Instances Across Groups

CF type Past tense Questions Distractors

intervention In the spot-the-differences task, used to elicit questions

in the past tense, student pairs received separate accounts of a tional character’s written biography, which differed in 10 ways Thestudents were instructed to ask each other a minimum of 10 questions

fic-to identify the differences The questions were written down by bothpartners, and the conversations were audio-recorded To elicit pasttense usage, the participants were asked to write a narrative of whathappened in a cartoon strip at a specified point in the past (yesterday,last week) To ensure linguistic uniformity, the learners had to incor-porate the supplied context-appropriate verbs (10 per strip) in theirstories at least once Of the 20 verbs, 4 reoccurred across the tests(enter, tell, leave, and go), 9 called for the regular past forms (enter,point, demand, park, deposit, climb, cross, walk, and stop), and 7 called forthe irregular forms (tell, leave, go, put, drive, meet, and come) All theverbs were telic and depicted accomplishment and achievement verbcategories that require the use of the past tense (Bardovi-Harlig, 1998;Collins, 2007) Two versions of the tasks were developed to counterthe test-retest effect

Data analysis To examine common themes in the participants’beliefs as a group, the responses on Part 2 of the questionnaire weresubjected to an exploratory factor analysis The suitability of the datawas confirmed by the Kaiser-Meyer-Oklin value of 82 and the statisti-cally significant Bartlett’s Test of Sphericity (v2 = 2968.542, p < 001).Cronbach’s alpha for the 40-item scale was 84, indicating internalitem consistency To determine to what extent the confirmed factorsdistinguished among the learners in all conditions, average scores foreach learner were generated to represent values for the identifiedbeliefs and to use in subsequent analyses investigating possiblerelationships between beliefs and learning and between beliefs andnoticing

Pearson analyses of the correlation between each learner’s beliefsscore and noticing score were conducted; the same was done for thebeliefs and learning relationship The noticing scores were calculated

by dividing the total number of times a learner reported noticing bythe total number of recall instances provided, which were then con-verted into percentages This analysis was carried out for each targetand feedback condition The scores were used to determine the differ-ential noticing of recasts, prompts, and the mixture of the two Simi-larly, percentage accuracy scores were computed for each target acrosstest times For the past tense, the total number of verbs accurately sup-plied in the obligatory contexts was divided by the maximum score of

10 and then multiplied by 100 If the same verb was used more thanonce, only its initial use was counted to offset overuse To account for

the different number of questions produced by each learner, the ber of correctly formed questions was divided by the total number ofquestions supplied and then multiplied by 100.

num-RESULTS

Learner Beliefs

The factor analysis revealed five components, explaining a total of44.3% of the variance Interpretation of the components proved diffi-cult because the output indicated that although the participants saw

CF as important and expected it in the L2 classroom, they appearedconfused as to how, when, and by whom they preferred to be cor-rected (a detailed description is available from the authors of this arti-cle) Hence, it was decided to run another analysis on only thoseitems (n = 26) that dealt with the expectation for and importance of

CF as well as the two CF techniques of interest (recasts and prompts).The resulting factor analysis (.85 Kaiser-Meyer-Oklin value and statisti-cally significant Bartlett’s Test of Sphericity value, v2 = 1821.755,

p< 001, Cronbach’s coefficient alpha 855) produced a component solution, explaining a total of 43% of the variance, withComponent 1 contributing 26.72%, Component 2 contributing 9.13%,and Component 3 contributing 7.15% Oblimin rotation helped ininterpreting the resulting factors, which were named using the highestloading items on each component (Pallant, 2007) Because the 16items that loaded on Factor 1 were concerned with the expectation of

three-CF (Questions 7, 16, 17, 20, 21, 22, 23, 25, 26, and 32) and withrecasts as the technique of choice (Questions 6, 11, 12, 14, 18, and40), this factor was labelled “Expectation of CF and Recasts as CFMethod.” Among the 6 items that loaded on Factor 2, 5 items (Ques-tions 3, 15, 33, 34, and 39) represented the belief that the best way toprovide CF is through prompts; 1 item (Question 36), with the lowestloading score, attributed the importance to recasts Because the major-ity of the items with high loadings spoke of prompts as the correctivetechnique of choice, this factor was named “Prompts as CF Method.”Finally, because the 2 items (Questions 2 and 35) that loaded on Fac-tor 3 pertained to the negative consequences that CF may yield, thisfactor was labelled “Negative Consequences of CF.” Table 4 shows theresulting factor loadings for the 26-item analysis It is important tonote that 2 items (Questions 8 and 19) did not load onto any of thethree factors

Calculation of average scores per factor for each learner proved ficult because both Factor 1 and Factor 2 were composed of diverging

dif-TABLE 4

Rotated Factor Loadings for Learner Beliefs About CF (26 Items, N = 197)

Item Factor 1 Factor 2 Factor 3

I Expectation of CF and recasts as CF method (26.72% of variance)

6 Provision of the correct form is

helpful for the beginner students.

(Fournir la forme correcte est b
en
efique

pour les
etudiants de niveau d
ebutant.)

.512

7 The correction of speaking errors

is necessary in an English class.

(La correction des erreurs orales est

indispensable en classe d’anglais.)

.528

11 Provision of the correct form is

the best technique to correct

vocabulary errors in English.

(Fournir la forme correcte est la

meilleure technique pour corriger les

erreurs de vocabulaire en anglais.)

12 Provision of the correct form is

the best technique to correct

grammatical errors in English.

(Fournir la forme correcte est la

meilleure technique pour corriger

les erreurs grammaticales en anglais.)

.567

14 In light of my oral errors in

English, I prefer that my teacher

explicitly lets me know that my

utterance is incorrect and that he

or she supplies the correct form.

(Face a mes erreurs orales en anglais,

je pr
efere que mon professeur

m’indique de fac ßon explicite que mon

enonc
e n’est pas acceptable et qu’il me

fournisse la forme correcte.)

.536

16 If my English teacher does not

correct my speaking errors, my

determination to learn English

will diminish.

(Si le professeur d’anglais ne corrige

pas mes erreurs orales, ma

d
etermination d’apprendre l’anglais

diminuera.)

.636

17 The English teacher must inform

the student of the aspects that he

or she must improve so that the

student acquires them.

(Le professeur d’anglais doit informer

l’
etudiant des aspects qu’il doit

am
eliorer pour que ce dernier arrive a

les ma ^ıtriser.)

.654

18 Provision of the correct form is

the best technique to correct

pronunciation errors in English.

.683

(Continued)