Professional Documents
Culture Documents
Loewen-Philp-2006
Loewen-Philp-2006
Loewen-Philp-2006
Classroom: Characteristics,
Explicitness, and Effectiveness
SHAWN LOEWEN JENEFER PHILP
Department of Applied Language Studies and Department of Applied Language Studies and
Linguistics Linguistics
University of Auckland University of Auckland
PB 90219 PB 90219
Auckland Auckland
New Zealand New Zealand
Email: s.loewen@auckland.ac.nz Email: j.philp@auckland.ac.nz
THEORETICAL ARTICLES AND RECENT RE- on form as “overtly draw[ing] student’s attention
views of empirical research involving instructed to linguistic elements as they arise incidentally in
second language acquisition (SLA) have provided lessons whose overriding focus is on meaning and
strong support for the assertion that learners ben- communication” (pp. 45–46). He emphasized the
efit from some kind of focus on linguistic form importance of attention to form in the context
in classrooms, in addition to meaning-focused of a primary focus on meaning (see also Long
communication (Doughty & Williams, 1998b; El- & Robinson, 1998). Such a focus on form is of-
lis, 2001; Long, 1991, 1996; Long & Robinson, ten identified with the incidental feedback that
1998; Morris & Tarone, 2003; Norris & Ortega, teachers provide students during communicative
2000; Spada, 1997). Long (1991) defined focus language use (Ellis, 2001; Ellis, Basturkmen, &
Loewen, 2001; Long, 1991, 1996).
The Modern Language Journal, 90, iv, (2006) In instructional settings, corrective feedback
0026-7902/06/536–556 $1.50/0 provided as learners attempt to use the target
C 2006 The Modern Language Journal language (TL) verbally1 ranges in explicitness
Shawn Loewen and Jenefer Philp 537
and serves to identify problems of either appro- 2004; Iwashita, 2003; Mackey & Philp, 1998; Philp,
priateness or accuracy in communication (Ellis, 2003), few have gone beyond description to con-
2001; Seedhouse, 1997; van Lier, 1988). At one sider the effectiveness of recasts in instructional
end of a continuum, explicit feedback is overtly settings (exceptions include Doughty & Varela,
corrective, marking the learner’s production as 1998; Lyster, 2004; Mackey, in press; Nabei, 2002).
nontarget-like. Such feedback may include pro- Furthermore, the majority of these studies have in-
vision of the target-like form in addition to met- vestigated what Ellis (2001) referred to as planned,
alinguistic “comments, information or questions rather than incidental , focus on form.4
related to the well-formedness of the learner’s
utterance” (Lyster & Ranta, 1997, p. 47). The DEFINING CHARACTERISTICS
provision of explicit information about the prob-
lematic linguistic form (termed inform by El- The characteristics and contexts of second lan-
lis et al., 2001; Loewen, 2004) is apparent in guage (L2) recasts have been defined in different
Example 1.2 (See Appendix A for transcription ways in prior scholarly literature (Nicholas, Light-
conventions.) bown, & Spada, 2001). In the present study, we fo-
Such explicit feedback switches the focus of in- cus on the incidence and effectiveness of recasts
teraction from meaning to form. In contrast, im- in the adult L2 classroom. In instructional con-
plicit feedback retains the focus on meaning by texts, recasts represent a form of feedback that is
implying rather than overtly stating the existence pedagogically expeditious: A recast is time-saving,
of an error (Ellis, Loewen, & Erlam, 2006). This less threatening to student confidence, and less
use of implication can involve eliciting the correc- disruptive of the flow of interaction than, for
tion from the student(s) themselves, for example, example, elicitation of self-repair. Unlike explicit
via full or partial repetition of the learner’s ut- correction, recasts maintain the focus on mean-
terance (S: “yesterday she go,” T: “she?”; see Lyster, ing. Recasts also allow the teacher to maintain
1998). Implicit feedback can also take the form control. It is perhaps for these reasons that re-
of recasts, in which the target-like form is pro- casts are so frequent in many language classrooms
vided, not elicited. Recasts are TL reformulations (Long, in press).
by the interlocutor of a learner’s nontarget-like In defining recasts, we take as central, first,
utterances that retain the central meaning while that recasts are generally provided incidentally
changing the form of the utterance (Long, 1996), in the course of meaning-focused interaction
as shown in Example 2. The recast functions both in response to nontarget-like learner utterances.
to confirm the meaning of the student’s utterance Second, recasts retain the central meaning of the
and to correct the form. learner’s utterance while changing lexical, mor-
Recasts may draw the learners’ attention to the
phosyntactic, or phonological form (Long, 1996).
inconsistency between their utterances and the
Finally, recasts provide both positive evidence (ex-
TL forms (Long & Robinson, 1998). In instruc-
emplifying what is possible in the TL) and neg-
tional contexts, recasts are usually incidental to
ative feedback (implying what is not possible)
the discourse; that is, they are not preplanned by
rather than providing overt correction (stating
the teacher, but arise in response to the language
what is not possible; see Long & Robinson, 1998;
the student produces. In many meaning-oriented
Nicholas et al., 2001).5
language classrooms, recasts comprise a large per-
centage of the types of corrective feedback of-
fered by teachers (Donato, 1994; Doughty, 1994; POTENTIAL BENEFITS AND LIMITATIONS
Ellis et al., 2001; Loewen, 2002, 2004; Lyster & OF RECASTS
Ranta, 1997; Oliver, 2000; Panova & Lyster, 2002;
Seedhouse, 1997).3 Although numerous experi- In both first language (L1) and L2 acquisition
mental studies have shown beneficial effects of re- research, two characteristics common to all re-
casts in interactional contexts (Han, 2002; Ishida, casts, juxtaposition and contingency, ostensibly
Example 1
Student (S): uh didn’t work well (·) it must be rip=ded rip=ded
Teacher (T): so you need a noun now more explicit feedback
S: it must be rip=ded
T: it must be a rip off
538 The Modern Language Journal 90 (2006)
Example 2
S: to her is good thing (·) to her is good thing
T: yeah for her it’s a good thing recast
S: because she got a lot of money there
facilitate noticing problematic forms in the 1994, 2001; Schmidt & Frota, 1986). Schmidt
learner’s production (Long, 1996; Nicholas et (2001) described noticing as a “first step in lan-
al., 2001; Philp, 2003; Saxton, 1997). First, re- guage building” (p. 31) that is, in integrating
casts serve to juxtapose the incorrect forms with new language within the interlanguage system.
the correct forms, which provides both a model He claimed that it is only what the learner no-
and a contrast with the learner’s nontarget-like tices about the input that holds potential for learn-
utterance because they immediately follow the ing. In particular, learners need to notice the gap
learner’s ill-formed utterance (Saxton, 1997). Re- between their interlanguage forms and the TL
casts are contingent on what the learner pro- form (Doughty, 2001; Ellis, 1991; Gass & Varonis,
duces and, for this reason, are congruent with the 1994; Philp, 2003). Thus, in exploring the ben-
learner’s production and are not just the teacher’s efits of recasts, it is important to consider what
own pedagogic focus.6 The timing of recasts is im- aspects of recasts highlight the changes made to
portant because it is while focused on meaning nontarget-like production. Certain elements may
and use that the learner receives feedback on how increase the salience of the recast, that is, help to
to communicate that meaning (Doughty, 2001; draw the learner’s attention to the recast form.
Gass & Varonis, 1994; Long, 1996). Thus, the po- We will discuss these characteristics in a later
tential of interaction to facilitate acquisition lies section.
both in timing and context. Interactional modifi- Although according to some researchers,
cations, such as those in Example 3, draw learn- recasts facilitate noticing linguistic items in pro-
ers’ attention to problems in their interlanguage ductive ways, other researchers take a more pes-
at the very point when the learners’ production simistic view of their effectiveness (e.g., Lyster,
or representation of the L2 is at odds with the TL 1998; Panova & Lyster, 2002). In instructional set-
form (Ellis, 1991; Faerch & Kasper, 1983; Gass & tings, the transience of recasts may mean that they
Varonis, 1994; Long, 1996; Mackey, 1999, in press; pass unnoticed. In addition, as other learners take
Pica, 1994). the subsequent turn or as the teacher moves on
It has been hypothesized that such feedback al- to the content at hand, the opportunity to notice
lows the learner to map form to meaning, rather or incorporate the recast verbally disappears. In
than focusing on form as a discrete element light of research in an immersion context, with rel-
(Doughty, 2001). Doughty observed that, because atively large numbers of young learners (in classes
explicit correction breaks into the learner’s en- of 24–30 students), Lyster (1998) suggested that
coding of an utterance, it is intrusive to learning, feedback options more explicit than recasts are
whereas implicit correction allows the learner to preferable. He discovered that teachers left many
integrate new linguistic information as he or she errors uncorrected and also repeated correct ut-
continues to speak. terances as often as they recast incorrect utter-
It is important to point out that recasts may ances. Given these similar response patterns for
be of benefit only if they are noticed by the both correct and incorrect utterances, Lyster and
learner (Philp, 2003; Schmidt, 1990, 1993, 2001). Ranta (1997) concluded that recasts, because they
Long (1996) noted that the selective attention are implicit, are unlikely to benefit learners, who
of the learner is of fundamental importance in may experience difficulty in differentiating posi-
the connection between conversational interac- tive and negative feedback.
tion and acquisition (Ellis, 1991; Gass, 1991; Gass The ambiguity of recasts in instructional con-
& Varonis, 1994; Long, 1996; Schmidt, 1990, texts has been noted often (B. Allwright, 1984;
Example 3
S: yeah and uh when they went to the bank (·) the gateway and they stu- stu- in
T: got stuck recast
S: got stuck on the rova (·) ro: (revolving door)
Shawn Loewen and Jenefer Philp 539
R. Allwright, 1975; Calvé, 1992; Chaudron, 1977; feedback in promoting uptake, although the ef-
Doughty & Williams, 1998a; Fanselow, 1977; Long, fects are by no means negligible (usually reported
in press; Lyster & Ranta, 1997; Morris & Tarone, at about 30% despite occasional/frequent lack
2003; Nicholas et al., 2001). Lyster (2004) assessed of opportunity for uptake, see Oliver, 2000). Al-
the effectiveness of recasts over prompts, that is, though uptake following an elicitation generally
elicitation of self-repair, following a 5-week period demonstrates whether a learner has grasped the
of classroom-based instruction in eight fifth-grade problem or not, this result is not necessarily true
classes. On written tasks, the group receiving of uptake following recasts. Indeed, Mackey and
form-focused instruction (FFI) with prompts out- Philp (1998) suggested that responses to recasts
performed the group receiving FFI with recasts. are not indicative of the effects of recasts on L2
On oral tasks, these groups performed similarly, acquisition. In experimental research involving
and all FFI groups outperformed the control dyadic task-based interaction with intensive re-
group. The superiority of explicit instruction may casts, these investigators uncovered no difference
be particularly characteristic of immersion classes in spontaneous production of targeted forms for
with young learners, in which there is little or no immediate or delayed posttests between L2 learn-
attention to form and all focus is on content. In ers who responded to recasts and those who did
this case, it seems, young learners are unlikely to not. Although pushed output and, by implication,
attend to form. Yet, in language classrooms, there other types of moves that promote pushed output
is an underlying, pervasive understanding that may be useful in acquisition (Swain, 1995), the
the main purpose is to learn the TL rather than utility of these methods does not negate the effec-
any academic content. In principle, this under- tiveness of recasts (Long, in press).
standing orients the attention of all participants When considering the limitations of recasts,
to form, even in communicatively oriented class- one should also note that, depending on the tar-
rooms (Batstone, 2002). In summary, although geted form, they may be differentially beneficial
recasts can be ambiguous as corrective feedback, (Ellis & Sheen, in press; Long, 1996; Mackey, Per-
the likelihood of their effectiveness depends on due, & McDonough, 2000). In the few experimen-
various factors: the classroom context (including tal studies that have included pre- and posttest
both the age of participants and the extent to measures, recasts have proved to be beneficial for
which language is a focus of study), the context of some particular forms but not for other forms7
the recast within the discourse, and variable ele- (Iwashita, 1999; Long, Inagaki, & Ortega, 1998;
ments of the recast itself, which we will discuss in Mackey & Philp, 1998; Mackey, Philp, Egi, Fu-
the next section of this article. jii, & Tatsumi, 2002; Ono & Witzel, 2003; Ortega,
A second limitation of recasts, which Lyster and 1999). There is some indication that recasts may
colleagues (Lyster, 1998; Lyster & Ranta, 1997; be of greater benefit for salient, meaning-bearing
Panova & Lyster, 2002) pointed out, concerns forms than for nonsalient forms (Ortega & Long,
learner repair. According to these researchers, 1997; Long, in press). Studies that have examined
recasts are less beneficial than elicitation moves learners’ perceptions of recasts as corrective feed-
because, with recasts, the teacher solves the prob- back indicate differences attributable to linguistic
lem for the learners and the learners are under no focus (Mackey et al., 2000).
compulsion to modify their own utterance or to
produce “pushed output” (Swain, 1985, p. 249).
Given Swain’s (1985, 1995, 2000) output hypoth- CHARACTERISTICS OF RECASTS THAT MAY
esis, opportunity for self-repair is an important INFLUENCE BENEFITS
consideration. However, although elicitation pro-
vides greater opportunity for either self-repair or This section considers various characteristics
other-repair, this opportunity is not always real- of recasts that may influence their effectiveness
ized because self-repair requires at least latent in the classroom. The instructional context af-
knowledge of the targeted linguistic form (Long, fects both the provision and recognition of re-
in press). In such cases, recasts may follow unsuc- casts as feedback (Morris & Tarone, 2003; Oliver
cessful self-repair, and the teacher’s recast resolves & Mackey, 2003; Seedhouse, 1997; van Lier, 1988).
the knowledge gap (Loewen, 2002). Although recasts are incidental and provide im-
Many researchers have used the learner’s imme- plicit negative feedback, they may range in de-
diate response to feedback, or uptake, as a measure gree of explicitness and salience. The number of
of effectiveness (e.g., Chaudron, 1977; Loewen, feedback moves, prosodic cues, repetition, length
2004; Lyster, 1998; Lyster & Ranta, 1997). Recasts of recast, number of changes, and segmentation
are generally less effective than other forms of are all features that can affect the degree of
540 The Modern Language Journal 90 (2006)
implicitness in a recast. We will discuss these char- nior high school science classes with a total of 34
acteristics in turn. students, employed corrective recasts, in which the
recasts were first preceded by the repetition of the
Number of Feedback Moves error. In their view, prior repetition of the error
reduces the ambiguity of the recast as corrective
One way that recasts become more salient is feedback and draws the learner’s attention to the
through the number of feedback moves (also re- problematic form. Their study revealed a clear ef-
ferred to as “response moves,” Loewen, 2002, p. fect for corrective recasts on both oral and written
119). Following a learner’s nontarget-like utter- production, which was sustained over time. The
ance, a teacher may provide more than one kind findings suggest a beneficial effect in the combi-
of feedback move, for example, an elicitation fol- nation of drawing learners’ attention to the error
lowed by a recast, as shown in Example 4. and then recasting; they also demonstrate the pos-
In this case, the student first heard the correct sibility of focusing on both content and form.
word bullet in the preceding turn (l. 2), yet pro-
duces blood (l. 3). The teacher elicits self-repair Prosodic Emphasis
from the student (l. 4), who unsuccessfully mod-
Another way in which the recaster may cue the
ifies his original pronunciation (l. 5). After the
learner to the particular problem is by means of
student fails to self-repair, the student’s attempt is
prosodic emphasis, whereby a particular word or
recast by the teacher; the student then repeats it
morpheme is stressed, as in Example 5. This tech-
in the following turn (l. 7). Clearly, the student is
nique is used particularly for phonological prob-
led through a series of moves to notice the recast
lems, but as this example shows, it can also be used
bullet, and he successfully uptakes or incorporates
for morphosyntactic items.
into his production the correct linguistic form.
Although a number of studies have noted the
It can be argued that the recast in Example 4 is
presence of prosodic stress (Chaudron, 1977;
more explicit as corrective feedback than are iso-
Sheen, 2004), with the exception of Chaudron’s
lated recasts.
early research, this feature has not been examined
As Seedhouse (1997) and van Lier (1988) re-
as a variable in classroom studies of recasts.
minded us, recasts arise within an entire body of
discourse, and it is important, therefore, to con- Intonation
sider them within the interactional organization
of the classroom. When a recast is part of a length- Recasts provided in the context of conversa-
ened “focus on form episode” (FFE; Ellis et al., tion are often part of negotiation sequences and
2001, p. 294) with more than three turns, as in function as confirmation checks, with interroga-
Example 4, the FFE may have added value; that tive intonation (i.e., “Is this what you mean?”), as
is, there is potentially increased learner engage- in Example 6. In such cases, it may be unclear
ment with the targeted form (Fortune & Thorp, whether the interlocutor fails to understand the
2001). Implicit negative feedback may give rise to speaker or is simply correcting the form of the
an extended sequence of turns, in which the fo- utterance. An interrogative recast is ambiguous
cus is no longer on the task but on form itself, in as corrective feedback because the learner may
other words, negotiation of form (Swain & Lap- interpret it either as corrective or as a request
kin, 2001), as is apparent in Example 4. to confirm the intended meaning. In Example
Recasts may also be given increased value within 6, perhaps it was the latter because the learner
the same turn. Doughty and Varela (1998), in a simply repeats rather than modifies his initial ut-
quasi-experimental study involving two intact ju- terance.
Example 4
1 T: you have to tell this story to S okay (·) not your story (·) you’re
2 telling the story “girl had bullet in her scalp”
3 S: the title of the story is girl had blood in her scalp
4 T: blood? elicitation
5 S: bloot
6 T: bullet bullet= recast
7 S: =bullet bullet in her scalp is about is about
Shawn Loewen and Jenefer Philp 541
Example 5 Example 7
S: some people have racism S1: yeah
T: some people ARE racist stressed recast S2: no racist not like uh racist
S: are racist T: racism segmented recast
S1: racis this country
In instructional contexts, a recast is more fre- the recast, the more likely one is to notice it
quently delivered as a declarative (authoritative) (Egi, 2004; Philp, 2003). The distinction between
statement, as shown in Example 5, than as a ques- length of recast and segmentation is an important
tion, with rising intonation (Lyster, 1998; Sheen, one because segmentation potentially pinpoints
in press), as shown in Example 6. The didactic the problem for the learner. Although segmented
force of a recast is, thus, signaled by intonation, recasts are always short, not all short recasts are
with declarative recasts being arguably more ex- necessarily segmented.
plicit as corrective feedback than interrogative re- It appears, then, that in the classroom, the am-
casts because they do not suggest a choice for the biguity of recasts can be greatly reduced by the
learner to accept or refute. Alternatively, declara- phrasal, prosodic, and discoursal cues that teach-
tive recasts may be interpreted in the same way as ers provide. Nicholas et al. (2001), in their review
a teacher’s repetition of learners’ TL forms (see of the literature on recasts, recognized the degree
Lyster, 1998) and, thus, may function just as im- to which recasts may vary:
plicitly for the learner.
Recasts are distinguished from other kinds of focus-
on-form procedures because they are not explicit, do
Segmentation not isolate the features of language form that are the
focus of the feedback, and do not interrupt—even
Another feature of recasts in classrooms is seg- briefly—the flow of meaningful interaction. The ex-
mentation. Previously identified as partial recasts ception may be recasts that are accompanied by some
(Roberts, 1995) and recasts with reduction (Lyster, sort of overt signal. (pp. 234–235)
1998; Sheen, in press), segmentation may also re-
duce the ambiguity of a recast, as in Example 7, in The effectiveness of recasts is likely to be affected
which the teacher segments the problematic form by the features described above and other factors,
(racist) and recasts it in isolation (racism). such as degree of difference between the recast
In previous studies, segmentation varied ac- and the nontarget-like utterance.
cording to the type of classroom, representing
from 22%–24% (Chaudron, 1977; Lyster, 1998) MEASURING THE EFFECTIVENESS OF
to 50% (Roberts, 1995) to 84% (Sheen, 2004) of RECASTS
recasts provided.
To date, the effectiveness of recasts in class-
room contexts has been measured in various ways
Number of Changes and Length of Recast (Nicholas et al., 2001), including (a) the learner’s
immediate response to the recast, that is, uptake
Other factors that appear to affect learners’
(e.g., Lyster, 1998; Lyster & Ranta, 1997; Panova &
noticing and responses to recasts center on the
Lyster, 2002); (b) the use of uptake charts follow-
relationship of the recast to the original utter-
ing a lesson (Nabei, 2002); (c) spontaneous pro-
ance, including such factors as number of changes
duction of the recast form elicited in task-based
(i.e., the degree of difference between the recast
interaction (e.g., Egi, 2004; Lyster, 2004; Mackey
and the original utterance) and length of the
& Philp, 1998); (d) elicited production of the
recast. Previous experimental research has sug-
targeted form in tailor-made individualized tests
gested that the fewer the changes and the shorter
(Egi, 2004; Loewen, 2002, 2005; Nabei & Swain,
2001); and (e) use of indirect tests, such as gram-
maticality judgment tests that seek to access learn-
Example 6 ers’ underlying knowledge of a form (Ellis et al.,
2006; Long et al., 1998; Lyster, 2004; Nabei, 2002).
T: okay, S do you know this story? Although each of these methods has disadvan-
S: finish tages, the use of successful uptake as a measure
T: finished? interrogative recast
of effectiveness is particularly problematic in the
S: finish
case of recasts. Uptake does not indicate the status
542 The Modern Language Journal 90 (2006)
of the learner’s knowledge: whether it reflects the problematic forms subsequent to the provision of
acquisition of new knowledge or the activation of feedback.
latent knowledge (Long, in press). In addition,
the noticing of recasts is not contingent on up- RESEARCH NEEDED
take (Mackey & Philp, 1998). In interaction, re-
casts require neither a response nor a self-repair The distinctions discussed in the preceding sec-
from the learner; rates of repair following re- tion suggest greater variety among recasts in in-
casts vary considerably, with low rates reported in structional settings than has been acknowledged
some school settings (Havranek, 2002; Lyster & in research (Ellis & Sheen, in press; Sheen, in
Ranta, 1997) and higher rates reported among press), although earlier research recognized that
adults and young children (Braidi, 2002; Ellis there are such distinctions. Chaudron (1977), for
et al., 2001; Oliver, 1995). Nevertheless, the pro- example, distinguished among recasts with dif-
duction of successful uptake provides an indica- ferent features, such as emphasis (stress) and re-
tion that the learner has noticed the recast (Ellis duction (segmentation). Recasts may become less
et al., 2001; Lightbown, 1998; Loewen, 2004). implicit as a consequence of the turns that pre-
The term uptake has also been used to refer cede or follow them and by either prosodic or
to uptake charts in which learners register after segmental cues the teacher provides within them.
a lesson the linguistic forms noticed during the Therefore, when considering the effectiveness of
lesson (B. Allwright, 1984; Ellis, 1994; Slimani, recasts, it is important to investigate the impact
1989); however, such charts do not measure use of their different characteristics (Nicholas et al.,
per se and require the learner to articulate what 2001).
was noticed. Spontaneous production of the re- Conflicting findings across classroom-based
cast item in pre- and posttests has the advantage of and experimental studies on the benefits of re-
reflecting acquisition, yet it is best suited to the use casts for L2 acquisition call for further, more de-
of preselected forms because incidental focus on tailed research, particularly in classroom contexts
form is unpredictable (Swain, 2001). Similarly, in- (Nicholas et al., 2001). More work is needed that
direct tests, such as grammaticality judgment tests, examines not only the differences between recasts
in which learners evaluate the correct or incorrect and other types of feedback, but also the differ-
use of pretargeted forms, necessitate preselection ences within recasts, including the range of im-
of the linguistic items. plicitness of the recast. This research should be
The use of tailor-made tests, in which individ- coupled with investigations of the effectiveness of
ual test items are made for each student depend- recasts in terms of learners’ subsequent use of
ing on the forms arising in FFEs specific to each the linguistic items rather than through uptake
student, allows for the testing of incidental focus alone. The purpose of the present research was
on form (Egi, 2004; Loewen, 2002, 2005; Nabei to examine (a) the nature of recasts provided
& Swain, 2001; Swain & Lapkin, 1998, 2001). in an instructional context and (b) the effect
These tests tend to require the learner to cor- of different aspects of recasts on L2 short-term
rect the nontarget-like utterances on which they learning.
had previously received corrective feedback. Such
tests assess the effectiveness of incidental focus RESEARCH QUESTIONS
on form in terms of learners’ ability to recall
The study addressed three sets of research ques-
the linguistic information provided in the recasts
tions. The first two research questions were de-
(Loewen, 2002, 2005). However, in the absence
scriptive:
of pretests, such measures cannot provide infor-
mation about learners’ previous knowledge of the 1. What is the incidence of recasts, informs,
forms and, thus, cannot differentiate between the and elicitations in adult L2 classrooms?
acquisition of new knowledge and the consolida- 2. What are the characteristics of recasts in
tion of latent knowledge. As Loewen (2002) ar- these adult L2 classrooms?
gued, for individual testing, “the FFE itself has The next pair of research questions addressed
to serve as a type of pre-test, indicating students’ the effectiveness of recasts as compared to other
lack of prior knowledge or ability regarding the response moves in terms of posttest performance
targeted linguistic item” (p. 83). Although these and successful uptake:
tests may not capture acquisition of new linguis- 3. Are recasts associated with higher accuracy
tic knowledge (Long, in press), they can inform scores on posttests than are other types of correc-
us about learners’ ability to use demonstrably tive feedback?
Shawn Loewen and Jenefer Philp 543
4. Are recasts associated with higher levels of were collected because the study’s research ques-
successful uptake than are other types of correc- tions did not directly address proficiency level.
tive feedback?
The final research questions concerned the Observations
characteristics of recasts that may potentially af-
fect learning outcomes: The first researcher was present during all
5. Are particular characteristics of recasts asso- observations as a nonparticipant observer (van
ciated with successful uptake? Lier, 1988). The teachers and learners were in-
formed that the purpose of the research was to
6. Are particular characteristics of recasts asso- examine classroom interaction during meaning-
ciated with accurate posttest scores? focused lessons; however, they were not made
aware of the precise focus. For the purposes of
METHOD the study, meaning-focused activities were those
in which the primary goal of the activity was to
The data for the present study were part of a exchange information rather than to learn about
larger study (Loewen, 2002, 2004) involving the or practice specific linguistic forms (Pica, Kanagy,
observation of meaning-focused classroom activi- & Falodun, 1993). The types of activities observed
ties, the identification of FFEs in teacher–student included information and opinion gap tasks, story
interaction, and the testing of learners’ ability to narration tasks, and discussions relating to top-
recall the targeted linguistic items. Of the 1,373 ics such as travel, smoking, and life experiences.
FFEs identified in Loewen’s study (2002, 2004), a Any activities that were not meaning focused were
subset of 465 FFEs, which were used for testing, excluded from the analysis. A total of 32 hours
were selected for the current analysis. (1,917 minutes) of classroom interaction was ob-
served and recorded, with 17 hours (1,022 min-
Participants utes) providing the data for testing and, thus, the
data for the present study. For recording, a wire-
The research was conducted at a private lan- less cassette recorder with a clip-on microphone
guage school in Auckland, New Zealand. A total attached to the teacher was used.
of 12 classes with 12 different teachers and 118
learners participated. The teachers, 8 males and 4 Identification of Corrective Feedback
females, were all native speakers of English, with
varying levels of teaching experience. The major- Corrective feedback occurring in the lessons
ity of the learners (over 75%) were from Korea, was identified by the first author. In order to in-
China, Japan, and Taiwan. Of the learners, 56% clude all the discourse relating to the corrective
were female, and 44% were male. The learners in feedback, the first author transcribed the FFEs,
the 12 classes ranged in English language profi- consisting of “the discourse from the point where
ciency, as assessed by the school’s 1-hour, in-house the attention to linguistic form starts to the point
placement test, which had writing, grammar, and where it ends, due to a change in topic back to
oral interview components. Of the 12 classes, message or sometimes another focus on form”
2 classes were labeled Intermediate 1, 1 class was (Ellis et al., 2001, p. 294). In Example 8, the
Intermediate 2, 5 classes were Intermediate 3, one FFE, in lines 3–5, occurs within the context of
class was Intermediate 4, and 3 classes were Upper a meaning-focused activity, a class discussion, in
Intermediate. Although more specific informa- which one student was relating his experiences as
tion about the learners’ L2 proficiency might have a soldier. Within this discussion, the participants
been desirable (Thomas, 1994), no further data focused briefly on a linguistic item, when the
teacher responded to the learner’s incorrect use
of the preposition to by saying protection from and
the learner repeated the correct form. In Line
Example 8
1 S: when I was soldier I used to wear the balaclava
2 T: and why did you wear it S for protection from the cold or for another reason
3 S: just wind uh protection to wind and cold FFE
4 T: protection from
5 S: uh from wind and cold
6 T: right (·) okay not for a disguise
544 The Modern Language Journal 90 (2006)
6, the discussion turned back to the student’s TABLE 1
experiences. Reliability of Coding
In order to determine reliability in the identi-
fication of FFEs, a sample of 10.3% of the data Category Kappa
(three lessons totaling 199 minutes) was coded by
a research assistant, who was trained in the iden- Linguistic focus .955
tification of FFEs. The resulting agreement rate Length of recast 1.000
was 89.9%.8 Segmentation .965
Number of changes .966
Coding of Corrective Feedback Number of feedback moves .747
Successful uptake .899
In Loewen’s (2002, 2004) larger study, correc-
tive feedback was coded according to several dif-
ferent categories (e.g., inform, recast, and clari-
fication request); however, these categories were because intonation and prosodic emphasis were
combined into two overarching categories, provi- high-inference categories with lower reliability
sion and elicitation, depending on whether the scores (κ = .46 and .55, respectively), both
teacher provided linguistic information or at- researchers coded all the data for these cate-
tempted to draw out the information from the gories and negotiated any discrepancies by lis-
learner. In order to address the research questions tening to the tapes together until consensus was
for the current study, the following categories of reached.
corrective feedback, based on Loewen (2002) and
Ellis et al. (2001), were used: (a) elicit, an attempt
Testing
to get the learner to provide the correct linguistic
form; (b) inform, the provision of explicit infor- Four observation sessions were conducted for
mation about the problematic linguistic form; (c) each of the 12 classes, but only two of them were
recast, reformulation of all or part of the erro- used for testing purposes. The first author con-
neous utterance. structed individualized test items relating to the
Following the identification of recasts in the linguistic forms arising in the FFEs. Learners who
data, both authors independently coded all the were involved in the FFEs were tested on those
recasts for a variety of characteristics, as de- items. Every effort was made to test all FFEs that
scribed in Appendix B. Any differences in cod- occurred in the observations designated for test-
ing were resolved through negotiation. The cod- ing. The first author administered tests on two
ing categories were based, in part, on research different occasions: immediate (1 to 3 days after
by Lyster (1998), Lyster and Ranta (1997), Ellis the FFEs and testing the linguistics items targeted
et al. (2001), and Philp (2003). Finally, the first in the second observation) and delayed (13 to 15
author coded instances of uptake using the fol- days after the FFEs and testing the linguistic items
lowing categories, again based on work by Loewen targeted in the third observation). He conducted
(2002, 2004) and Ellis et al.: (a) successful uptake, the tests individually with each student, and au-
the learner incorporates the linguistic informa- diorecorded each testing session. A total of 73
tion provided in the feedback into his or her own of the 118 learners participated in FFEs with re-
production; (b) unsuccessful uptake, the learner casts and subsequently were tested on the linguis-
responds to the feedback but does not incorpo- tic items in the FFEs. The number of test items for
rate the linguistic information into his or her own each learner ranged from 0 to 7 items for the im-
production; (c) no uptake, the learner does not re- mediate test and from 0 to 9 items for the delayed
spond to the feedback; (d) no chance, the learner test, with an average of roughly 1.5 items for each
does not have the opportunity to respond to the test.9
feedback. Between the occurrence of the FFE and the test-
ing, learners continued their normal classroom
Reliability of Coding activities; however, it was not possible to docu-
ment the foci of the intervening lessons. Test
To estimate the reliability of coding, 25% of items were constructed directly from the tran-
the data was randomly selected and coded by scribed FFEs and were based on a template, in-
a third person, the research assistant. Interrater volving correction or pronunciation. For correc-
reliability scores were calculated using Kappa. tion test items, the researcher read aloud an in-
(The coefficients appear in Table 1.) However, correct sentence to the student with instructions
Shawn Loewen and Jenefer Philp 545
to correct the sentence. To validate the test items, Inferential Statistics
three SLA and testing experts compared the test
items with the original FFEs. (See Loewen, 2002, Raw frequencies and percentages were calcu-
for further description of the testing.) The fol- lated for the feedback and test response cate-
lowing is an example of a correction test item: gories, and the results for the 12 classes were com-
“The following sentence is incorrect or inappro- bined. Because the data consisted of frequency
priate. Please listen and tell me how you could counts of categorical data, Pearson’s chi-square
make the sentence better: I used to wear the bala- analysis was employed. An alpha level of p < .05
clava for protection to wind and cold.” The student’s was set for all tests. Because the analyses involved
production of the incorrect preposition to, de- variables with more than two coding categories
scribed in Example 6, is precisely the subject of the (i.e., the chi-square table was larger than 2 × 2),
test. adjusted standardized residuals of greater than
For pronunciation items, a learner first read 2.0 or less than −2.0 were used to identify sig-
aloud a sentence containing the targeted word nificant differences. All inferential statistics were
or phrase and then the targeted word or phrase performed using the Statistical Package for the So-
in isolation. The provision of the written form cial Sciences (SPSS) 10.0.
of the word may have provided the learner with In order to investigate the possibility of causal-
clues to pronunciation that were not present in ity, a binary logistic regression analysis was
the original oral FFE; however, this limitation was performed (Hatch & Lazaraton, 1991; Hos-
unavoidable, in light of the difficulty of eliciting mer & Lemeshow, 2000; Saito, 1999). The two
test items without some type of cue. Again, the test dependent variables were successful uptake and
items were based as closely as possible on the cor- accurate test scores,10 whereas the independent
responding FFEs. Thus, in Table 2, the student’s variables consisted of the various characteristics
mispronunciation of scalp was tested by asking of recasts described in Appendix B. In addition,
him to read his original utterance. successful uptake was included as an independent
The learners’ ability to produce correct re- variable in the regression analysis for accurate test
sponses was assessed according to the following scores. The method of entry for each independent
coding categories: (a) correct, the student pro- variable in the regression equation was forward
duced a response that correctly matches the tar- stepwise, which involves adding variables one at
geted linguistic item in the FFE; (b) incorrect, the a time, in an attempt to produce the best and
student did not correctly produce the linguistic most succinct model. Forward stepwise regression
item targeted in the FFE; (c) partially correct, the analysis was chosen over hierarchical analysis, in
student produced a response that improved on which the researcher stipulates the order of en-
the targeted linguistic error in some way but was try, because there was no preexisting rationale for
still not entirely accurate. ranking the independent variables. An alpha level
A random sample of 21% of the test items was of .15 was used to determine which variables to in-
scored by the research assistant, with a reliability clude in the model, as recommended by Hosmer
coefficient of κ = .850, which fell above the sug- and Lemeshow (2000). Finally, one assumption
gested .80 level for acceptability (Davies, Brown, of logistic regression analysis is that the indepen-
Elder, Hill, Lumley, & McNamara, 1999). dent variables are not highly intercorrelated. To
test this assumption, a correlation matrix was pro-
duced for the independent variables in the var-
ious analyses. The coefficients ranged from .029
TABLE 2 (for intonation and segmentation) to .583 (for
Pronunciation Example segmentation and prosodic emphasis), with eight
correlations being significant at the .001 level, two
Test Item FFE Being Tested at the .05 level, and five not significant. Thus, al-
though some multicollinearity existed, suggesting
Please read aloud the S1: the title of the story is caution in interpreting the results, the R 2 values
following sentence and girl has bullet in her were well below the critical .80 level proposed by
word. scamp Menard (1995).11
T: scalp
1. The title of the story S1: scalp
is girl has bullet in S2: scalp RESULTS
her scalp. T: scalp (·) this bit
2. Scalp S2: scalp yep The first research question concerned the na-
ture of corrective feedback in adult English as a
546 The Modern Language Journal 90 (2006)
TABLE 3 morphosyntactic errors each received roughly
Type of Feedback equal levels of feedback, at around 25%. In ad-
dition, about 15% of the recasts targeted a combi-
N Percent nation of linguistic features. As for the length of
the recasts, the majority (87%) had fewer than
Recast 228 49.03 five morphemes, and most recasts (69%) were
Inform 172 36.98 also segmented. Almost three fourths of all re-
Elicitation 65 13.98 casts involved only one change to the original tar-
Total 465 100.00 get utterance, and nearly two thirds of the recasts
were a single recast, while one third were com-
bined with other feedback moves. The majority of
second language (ESL) classes. Table 3 shows that recasts (85%) contained additional stress on the
there were 465 tested FFEs in the data. Nearly 50% targeted linguistic form, and most recasts (83%)
of feedback consisted of recasts, with 37% provid- contained declarative, rather than interrogative,
ing information (inform), and 14% eliciting a re- intonation.
sponse from the learners.
Effectiveness of Recasts
elicitation responses were associated with signifi- with declarative intonation. The variables of mor-
cantly higher levels of successful uptake and sig- pheme length and segmentation were not signifi-
nificantly lower levels of unsuccessful uptake and cant predictors of successful uptake.
no chance for uptake, whereas inform responses
Test Scores
were associated with significantly lower levels of
successful uptake and significantly higher levels The logistic regression analysis for the test
of unsuccessful uptake, χ 2 (6, 465) = 32.629, p < scores, as summarized in Table 8, showed an over-
.001. all percentage accuracy of 60.9% and indicated
that the best model for predicting correct test
Successful Uptake
scores included intonation, morpheme length,
Assessing the ability of recasts to predict suc- and number of changes. Recasts with rising in-
cessful uptake and correct test scores entailed the tonation were over two times more likely to result
use of logistic regression analysis. The analysis in correct test scores than were those with declara-
for successful uptake, shown in Table 7, had an tive intonation. Additionally, longer recasts of five
overall percentage accuracy of 71.5% and indi- or more morphemes were half as likely to result
cated that the best model for predicting success- in correct test scores compared to shorter recasts,
ful uptake included prosodic emphasis, number with fewer than five morphemes. Similarly, recasts
of feedback moves, intonation, and number of with more than one change were half as likely
changes. Recasts in which the corrected error re- to result in correct test scores as those with only
ceived prosodic stress were 13 times more likely one change. The variables of successful uptake,
to result in successful uptake than unstressed re- prosodic emphasis, number of feedback moves,
casts. Extended FFEs were five times more likely and segmentation were not significant predictors
to result in successful uptake than were simple of correct test scores.
FFEs. In addition, recasts with more than one Summary
change were half as likely to result in successful
uptake as those with only one change, and recasts The present study provides evidence that teach-
ending in high-rising intonation were a third as ers make use of different kinds of negative
likely to be followed by successful uptake as those feedback in ESL adult classrooms, with half of all
TABLE 8
TABLE 7
Logistic Regression Results for Test Scores
Logistic Regression Results for Successful
Uptake
95% Confidence
95% Confidence Intervals
Odds Odds
Step 4 Ratio Lower Upper Sig. Step 3 Ratio Lower Upper Sig.
Stress 13.235 4.136 42.355 .000 Intonation 2.206 .899 5.409 .084
Intonation .351 .153 .807 .012 Morpheme
Changes .519 .257 1.049 .067 length .509 .204 1.272 .149
Number of feed- Number of
back moves 5.052 2.494 10.234 .000 changes .518 .265 1.013 .054
548 The Modern Language Journal 90 (2006)
feedback moves taking the form of recasts. In addi- in a variety of ways. One way in which the teach-
tion, recasts in this context tended to be short and ers made the function of the recast more explicit
delivered with declarative intonation. The prob- was by using intonation. It was far more common
lematic form tended to be stressed and presented for them to provide recasts with declarative in-
as a segment of the original utterance. There tonation (83%) than with interrogative intona-
were roughly equivalent proportions of recasts of tion, a technique relying on the authority of the
phonological, morphological/syntactic, and lexi- teacher.
cal forms. In 85% of cases, the targeted form in the recast
Second, although elicitation moves may result received prosodic emphasis. This prosodic em-
in greater amounts of successful uptake, in this phasis was often used for phonological changes
study, there were high rates of successful uptake (e.g., CONtent, COMpany), but not exclusively, as
following elicitations and recasts (83% and 60%, apparent in Example 9, which involves a lexical
respectively); however, this rate was significantly change. In this example, the teacher did not ini-
lower following informs (46%). In terms of recall tially understand but then provided a recast af-
of forms eliciting feedback, in posttest conditions, ter the student repeated the nontarget-like utter-
recasts were associated with a 50% success rate. ance. In this case, successful uptake followed the
There were no significant differences in accuracy recast.
among the feedback moves. As for the length of recasts, the majority (86%)
Third, certain characteristics of recasts signifi- were fewer than five morphemes in length. In
cantly predicted successful uptake and test scores: an experimental study involving native-speaker–
recasts with stress, recasts with declarative into- nonnative-speaker dyadic task-based interaction,
nation, recasts within extended FFEs, and recasts Philp (1998) reported significant differences in
with only one change were predictive of successful length of recast according to the level of the
uptake; interrogative recasts, shorter recasts, and learner; however, in the present study, it seems
recasts with only one change were predictive of more likely that length related to segmentation. In
accuracy in posttest performance. this data, over two thirds of all recasts segmented
the learner’s utterance, in other words, the re-
casts provided focus on the problem by isolating
DISCUSSION
the main form, as seen in Example 10.
Negative Feedback in the Classroom By isolating the verb, the teacher highlights
in initial position the insertion of be in the pas-
The first two research questions addressed the sive construction. This recast may represent a
nature of corrective feedback in the L2 classroom difference with those in dyadic task-based inter-
and the nature of recasts in particular. Recasts action that tended to involve the entire utter-
were high in frequency, which was not unexpected ance (Mackey, 1999; Mackey & Philp, 1998; Oliver,
in view of the high rates of recasting in other class- 1995; Philp, 2003).13 It also differs from Lyster and
rooms (Ellis et al., 2001; Havranek, 2002; Lyster Ranta’s (1997) findings involving young learners,
& Ranta, 1997). As noted by others (Doughty, in which only 24% of recasts were segmented. A
1994; Ellis et al., 2001; Long, in press), recasts higher incidence of segmentation in these ESL
may be favored in the discourse because they are a classrooms may reflect an underlying orientation
nonthreatening, unobtrusive means of supplying to form by the teacher, as well as the language-
correction and do not interrupt the flow of the related goals of the adult learners.
interaction. Recasts allow a teacher to maintain Another means by which recasts were made
focus on the task and to control the discourse. In more overt in the present study was through ex-
contrast, elicitation moves, which can be lengthy, tended episodes, as shown in Example 9; one third
depend on learners’ prior knowledge and may ul- of recasts appeared as part of longer FFEs, that
timately be unresolved in the L2. is, more than three turns. As noted earlier, some
Example 11
S: somebody steal my paper (·) stolen
T: someone stole your paper? Interrogative recast
Shawn Loewen and Jenefer Philp 551
separately, this was not the case. This finding may learner repetition of recasts may not be a good
indicate that successful uptake predicts test per- indication of learning. Rather, for recasts, intona-
formance for elicitation moves and informs, but tion, morpheme length, and number of changes
not for recasts. The finding adds further sup- apparently have a greater impact on the learners’
port to Mackey and Philp’s (1998) claim that, accuracy of recall in posttests than does uptake.
for recasts, responses are of little consequence; Short recasts with few changes and with interrog-
that is, the presence or absence of uptake, even ative rather than declarative intonation are likely
successful uptake, is not an indication of the to be beneficial.
extent to which the learner notices the recast Further research that continues to take account
or benefits from it. Some studies suggest that of the specific classroom context in which recasts
successful uptake may be facilitative of acquisi- are provided should be fruitful. To investigate the
tion (Ellis et al., 2001; Lightbown, 1998; Loewen, relationship between implicit feedback and acqui-
2004; Lyster & Ranta, 1997), but they also acknowl- sition, future researchers should consider ways of
edge that successful uptake is not a measure of pretesting at least a subset of structures in or-
acquisition. der to enable tracking the effects of incidental
recasts. The use of classroom tasks that elicit spe-
cific structures or vocabulary items could help re-
CONCLUSION AND FUTURE RESEARCH veal the likelihood of recasts of certain forms (see
Doughty & Varela, 1998; Mackey, in press). Fur-
The present study examined three aspects of
thermore, pre- and posttest measures should elicit
recasts in 12 adult ESL classrooms: the provision
not only explicit, but also implicit knowledge of
of recasts as one of three types of corrective feed-
forms.
back to learners; the nature of recasts; and the ef-
Recasts are clearly commonplace in commu-
fectiveness of recasts. The results were compared
nicative language classrooms and appear to be a
with results of other studies examining recasts in
primary means by which teachers provide correc-
a range of contexts.
tive feedback to learners. However, as this study
The findings indicated that recasts were widely
indicates, recasts differ widely in terms of empha-
used and that they were beneficial at least 50% of
sis, focus, segmentation, intonation, prosodic em-
the time. Given that recasts in the classrooms tar-
phasis, and degree of explicitness (Ellis & Sheen,
geted a wide variety of items arising incidentally
in press) and accordingly, as suggested by the find-
from the discourse, the 50% success rate on tests
ings, in their effectiveness. What kinds of recasts
should be encouraging for classroom teachers. In
are beneficial to learners, and in what contexts?
conjunction with the results of other studies re-
Some researchers have suggested that recasts are
viewed earlier that point to the effectiveness of
likely to be more effective when accompanied by
recasts, this study suggests that this kind of feed-
“clues” that identify them as corrective feedback
back is likely to be productive for learners. Al-
to learners than as a response to the meaning of
though other forms of feedback have their bene-
their utterance (Doughty, 2001; Nicholas et al.,
fits (see Lyster, 1998, 2004), recasts represent a
2001). The explicit and refined descriptions of
form of feedback that is pedagogically expedi-
recasts in the classroom data in this study have
tious: A recast is time-saving, less threatening to
provided some evidence that the greatest benefit
student confidence, and less intrusive to the flow
of recasts for learners lies in their comparability
of interaction than, for example, elicitation of self-
with the learner’s initial utterance, not necessarily
repair. In addition, unlike explicit correction, re-
in relation to their explicitness.
casts maintain the focus on meaning. In this study,
recasts did not differ greatly from other types of
corrective feedback in terms of benefit, as mea-
sured by recall on posttests. They did differ, how- ACKNOWLEDGMENTS
ever, in the connection between learner response
and test performance. Whereas a relationship ex-
This research was supported, in part, by a scholarship
isted for other types of feedback between success-
from the Foundation for Science, Research and Tech-
ful uptake and subsequent successful recall, for
nology, New Zealand. An earlier version of this article
recasts, successful uptake was not a significant fac- was presented at the 2005 Task-Based Language Teach-
tor. This finding has implications for classroom- ing conference in Leuven, Belgium. We are indebted
based research on corrective feedback that seeks to research assistants Susan McKenna and Ute Knoch
to determine effectiveness through measures of for their invaluable assistance in coding data and edit-
uptake alone. It suggests that in the classroom, ing the article. We would also like to thank Rod Ellis,
552 The Modern Language Journal 90 (2006)
Alison Mackey, and the anonymous reviewers for their
insightful comments. REFERENCES
APPENDIX A
Transcription Key
Symbol Meaning
S Student
T Teacher
CAPITALS Emphasis
(laugh) Extra information
<> Inaudible
(·) Micropause
= Linked speech
? Rising intonation
- Interrupted speech
: Lengthening
[] Phonetic representation
[ Overlapping speech
556 The Modern Language Journal 90 (2006)
APPENDIX B
Recast Characteristics
Linguistic Focus
Lexical The recast provides a new or modified S: I tried to (·) tried to hiking
lexical item or phrase (open class T: you tried to go hiking
items, e.g., nouns, verbs, adverbs, S: yeah I will receive more salary
adjectives), including recasts of here
incorrect prefixes and suffixes T: a higher salary (·) a high salary
Phonological Recast modifies learners S: sightseeing tour [tor]
pronunciation of an item/items T: tour [tur]
Morphological/ Recast modifies both the morphology S: so the Japanese language can use
syntactic and the syntax of learner’s it (·) it’s no <but> to study
utterance T: can be used
Combination Recast includes multiple changes, S: and Japanese student some are
including either phonology or <> area too uh eh touris tourisity
lexical changes (can also include T: it’s very touristy
morphology, syntax)
Length of Recast
Fewer than five Length of the entire recast utterance S: Your children would be my
morphemes contains fewer than five neephees [nifiz]
morphemes T: nephews
Five or more Length of the entire recast utterance S: my brain will s- I seem will seem
morphemes contains five or more morphemes to be explode
T: my my brain seems to be exploding
Prosodic Emphasis
Unstressed Linguistic item that is recast is not
given atypical stress
Stressed Linguistic item that is recast is given S: when I I can laugh ha ha the
atypical stress, through pitch, emotional the emotional (·)
additional pausing and emphasis conTENT the emotional
conTENT
T: the emoti- the emotional CONtent
Segmentation
Segmented The recast provides a partial recast of S: why he why should he release<d>?
the learner’s utterance T: be released
Whole The recast is an entire recast of the S: but I I crashed my [ga]rage
whole trigger utterance T: you what? You crashed into your
[gæ]rage
Number of Changes
One change Recast includes one change to the S: yes I usually use the <r/l>ast one
learner’s trigger utterance T: the last one
Two or more Recast includes two or more changes S: damag-ed (·) wall damag-ed
changes to the learner’s trigger utterance T: the hotel wall was damaged
Intonation
Declarative The recast is provided with falling S: why he why should he
intonation as a declarative release<d>?
statement T: be released
Interrogative The recast is provided with rising S: somebody steal my paper (·)stolen
intonation as a question T: someone stole your paper?
(confirmation check)