You are on page 1of 27

The Effect of Unfocused Direct and Indirect Written Corrective Feedback on Rewritten

Texts and New Texts: Looking into Feedback for Accuracy and Feedback for Acquisition
Author(s): FLORENTINA NICOLÁS-CONESA, ROSA MARÍA MANCHÓN and LOURDES
CEREZO
Source: The Modern Language Journal, Vol. 103, No. 4 (Winter 2019), pp. 848-873
Published by: Wiley on behalf of the National Federation of Modern Language Teachers
Associations
Stable URL: https://www.jstor.org/stable/45286641
Accessed: 11-09-2023 03:51 +00:00

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

National Federation of Modern Language Teachers Associations, Wiley are collaborating


with JSTOR to digitize, preserve and extend access to The Modern Language Journal

This content downloaded from 168.195.52.137 on Mon, 11 Sep 2023 03:51:20 +00:00
All use subject to https://about.jstor.org/terms
The Effect of Unfocused Direct and
Indirect Written Corrective Feedback
on Rewritten Texts and New Texts:
Looking into Feedback for Accuracy
and Feedback for Acquisition
FLORENTINA NICOLÁS-CONESA, 1 ROSA MARÍA MANCHÓN,2 and LOURDES CEREZO3
1 University of Murcia, Faculty of Education, Campus deEspinardo, 30100 Murcia, Spain Email : florinc@um.es
2 University of Murcia, Faculty of Arts, Campus de la Merced, 30071 Murcia, Spain Email: manchon@um.es
3 University of Murcia, Faculty of Arts, Campus de la Merced, 30071 Murcia, Spain Email: lourdesc@um.es

This study attempts to add new empirical evidence on the language learning ( opera tionalized and mea-
sured in terms of several dimensions of accuracy) that may derive from accessing and processing written
corrective feedback (WCF) on one's own writing. The research questions examined potential interac-
tions between type of WCF (direct vs. indirect), type of errors (grammar vs. nongrammar), and the per-
spective of feedback (accuracy vs. acquisition) in a single research design. To this end, 46 English majors
at a Spanish university participated in a pretest-posttest design, with 2 intervention groups (those re-
ceiving direct or indirect WCF and asked to process it via written languaging) and a control group (who
wrote and rewrote their texts without the help of WCF but also engaged in languaging) . The analyses
conducted show (a) limited appropriation of the WCF received, (b) positive short-term and longer term
benefits resulting from the combined effect of WCF and written languaging, and (c) differential effects
of type of WCF on error types. These findings are discussed from the perspective of current theory and
research on the manner in which WCF may contribute to language learning in terms of the dual distinc-
tion between feedback for accuracy and feedback for acquisition.

Keywords: feedback for accuracy; feedback for acquisition; unfocused feedback; written languaging; error
types

THIS ARTICLE SEEKS TO CONTRIBUTE TO study: its ecological validity and its focus on the
the growing theoretical and empirical work on efficacy of unfocused WCF on overall measures of
the manner in which writing and written correc- accuracy as well as on distinct error types. Thus,
tive feedback (WCF) processing may contribute given several calls for greater ecological validity in
to improving writing skills and to language learn- the research domain (Rang 8c Han, 2015; Storch,
ing (see Bitchener & Storch, 2016; Manchón 8c 2010), our study was conducted in a real class-
Cerezo, 2018, for overviews). This potential con- room. Additionally, the study intended to shed
tribution is closely linked to two features of the light not only on short-term uptake of WCF, but
also on its potential retention (i.e., learning) over
The Modern Language Journal, 103, 4, (2019) time. Following Storch 8c Wigglesworth (2010),
DOI: 10.1 1 1 l/modl.12592 uptake in this study refers to appropriation of the
0026-7902/19/848-873 $1.50/0 WCF received in terms of changes (errors success-
© National Federation of Modern Language Teachers fully corrected) in rewritten texts, whereas reten-
Associations
tion refers to the "long-term impact of feedback"
[Correction added on 14th November 2019 after (p. 310), manifested in our case in a reduction in
first online publication: In Tables 5-7 (pp. 858-859), error-making over time (in terms of both global
the column header was corrected with "Unsuccessful
percentage of errors and error categories) .
Corrections"]

This content downloaded from 168.195.52.137 on Mon, 11 Sep 2023 03:51:20 +00:00
All use subject to https://about.jstor.org/terms
Fiorentina Nicolás-Conesa et al. 849
The design, therefore, included the analysis of duced earlier, and it should also be linked to Nor-
texts rewritten after receiving WCF (the most fre- ris and Ortega's (2003) caution regarding "the
quent approach in WCF studies to date), and the extent to which accuracy in production of L2
creation of two new pieces of writing 1 week and forms should be taken as reflection of IL [inter-
3 weeks after having received and processed WCF. language] development" (p. 737). We are fully
Following Bitchener and Ferris's (2012) recom- aware of their claim regarding the nonlinear re-
mendations, we investigated the effects of WCF on lationship between accuracy of production and
overall measures of accuracy as well as on specific L2 development and simply adopt the distinc-
error types. tion between feedback for accuracy and feedback
The study was theoretically and pedagogically for acquisition as a heuristic to look into short-
motivated. From the theoretical angle, it intended term and longer term potential learning effects of
to test the effects of WCF in terms of the dual WCF, which, nevertheless, is also related to Norris
distinction between feedback for accuracy and and Ortega's distinction between emergence and
feedback for acquisition (Manchón, 201 la) . From accuracy.
a pedagogical angle, and as a secondary aim, In terms of feedback for accuracy, some stud-
its findings might be useful in designing peda- ies have shown the beneficial effects of feedback
gogical interventions in instructional settings in to improve the accuracy of revised texts (e.g.,
which second language (L2) users simultaneously Ferris 8c Roberts, 2001; Truscott & Hsu, 2008;
develop their L2 writing abilities (they learn to Van Beuningen, de Jong, 8c Kuiken, 2008). Rang
write) and engage in writing practice as a means and Han's (2015) recent meta-analysis offers a
to improve their L2 competencies (they learn the clear picture of extant empirical work in the do-
L2 through writing; Manchón, 2011a, 2011b). Of main: WCF (direct and indirect, focused and un-
particular pedagogical significance was the ques- focused) can lead to improved accuracy, although
tion of which errors benefit from which type of the observed effects are low to moderate. How-
WCF in the short term as well as the longer term, ever, as rightly noted by Truscott (2007), writers'
a key pedagogical concern we shall come back to incorporations of teacher corrections when revis-
at different points in the article. ing a text (equivalent to uptake, as defined be-
In what follows, we review several research fore) cannot be equated with the autonomous
strands that are relevant to our own study in terms production of correct L2 forms in new texts
of the aims pursued and methods employed. (or retention, as defined before), which would
be indicative of the effectiveness of WCF for
acquisition.
REVIEW OF RELEVANT RESEARCH
From theoretical, empirical, and applied per-
spectives, the key issue of debate is whether im-
Feedback for Accuracy and Feedback for Acquisition
proved accuracy in revised texts can be taken
The aims of the study are linked to the er- as evidence of short-term learning, which could
ror correction debate, associated primarily with eventually lead to consolidation of targeted L2
Truscott (1999, 2007) and Ferris (1999, 2004). forms. Truscott and Hsu (2008) concluded that
While there is general agreement that WCF re- successful error reduction of revised texts in their
sults in greater grammatical accuracy in imme- English-as-a-foreign-language (EFL) setting was
diate revisions (with diverse sets of variables not a predictor of L2 development (i.e., decrease
mediating potential effects), the empirical evi- of errors in new texts), while Van Beuningen, de
dence is differentially assessed for longer term Jong, and Kuiken (2012) found that unfocused
language-learning effects. The debate has re- feedback led to increased accuracy in an L2 con-
cently been reinterpreted as a dichotomy between text not only during revision but also in new
feedback for accuracy and feedback for acquisi- pieces of writing.
tion (Manchón, 2011a). Feedback for accuracy Another issue that has been the focus of
entails the provision of WCF to encourage accu- scrutiny in WCF research is whether diverse forms
racy in learners' use of the L2 shortly after feed- of feedback are more propitious for uptake or
back processing, whereas feedback for acquisition retention. The main distinction investigated (see
aims at fostering long-term language learning by Liu & Brown, 2015) is between direct and in-
involving students in feedback processing, detec- direct WCF, that is, between providing learners
tion of errors, self-reflection on errors (hence en- with the correct form or leaving them to pro-
couraging noticing with understanding) , and new vide the correct form themselves, with or without
output. The distinction is closely linked to the metalinguistic (ML) support from the teacher or
dichotomy between uptake and retention intro- researcher.

This content downloaded from 168.195.52.137 on Mon, 11 Sep 2023 03:51:20 +00:00
All use subject to https://about.jstor.org/terms
850 The Modern Language Journal 103 (2019)
The effectiveness of different feedback types feedback in the short term (1 week) and in the
in the short or long term has been researched long run (4 weeks) . However, since none of these
in both unfocused and comprehensive stud- studies yielded significant differences between
ies (those targeting more than a few selected the effectiveness of unfocused direct and indi-
error types in research settings) and focused rect feedback, no firm conclusions can be drawn
research (those targeting one or two specific about the potential of unfocused WCF for L2
errors) . Regarding unfocused WCF, recent stud- development.
ies have shown that direct WCF is more effec- As for focused WCF, some one-shot feedback
tive than other, less explicit1 WCF types. Vyatk- studies conducted with ESL learners (Bitchener,
ina (2010), for example, examined the effects of 2008; Bitchener 8c Knoch, 2008, 2010a; Sheen,
three kinds of WCF with varying degrees of ex- 2007) and EFL students (Stefanou, 2014) exam-
plicitness (direct, indirect with coding, and in- ined the effects of various combinations of ex-
direct without coding) in beginning learners of plicit feedback types (direct error correction and
German. The participants received feedback in oral or written ML comments or explanations) .
consecutive treatments (five 2-draft compositions Mixed findings were reported in L2 contexts,
of 70 words). Results showed a slight advantage and no differences between feedback groups were
for direct feedback in immediate revisions but no found in the study conducted in an EFL set-
advantage for any of the three WCF types in the ting. Another group of one-shot feedback stud-
long term (16 weeks). Qualitative data suggested ies compared the effectiveness of ML feedback
that "more implicit feedback types (coded and un- to direct or indirect WCF in ESL (Bitchener &
coded) may be confusing for learners and prompt Knoch, 2010b; Shintani 8c Ellis, 2013) and EFL
them to make wrong guesses" (Vyatkina, 2010, settings (Guo, 2015; Rummel, 2014; Shintani,
p. 17). In Chandler's (2003) study with 20 inter- Ellis, 8c Suzuki, 2014). Again, mixed findings were
mediate English second language (ESL) learners, reported in L2 contexts, whereas in FL settings,
direct feedback and underlining were found to explicit WCF (direct and written ML feedback)
be more effective than error codes. In contrast, groups outperformed those receiving less explicit
Ferris (2006) found that ESL university students feedback types (underlining and error codes;
benefitted from the feedback they received on 15 Guo, 2015) and direct feedback groups outper-
common error categories over a semester, and in- formed other treatment groups (e.g., ML: Shin-
direct WCF resulted in an increase in accuracy tani et al., 2014; indirect feedback and written ML:
and in better long-term results than direct feed- Rummel, 2014). Conflicting findings have been
back. According to Ferris, these results confirmed reported even by the same group of researchers
the value of guided learning and problem-solving in different settings: Whereas Shintani and Ellis
(Lalande, 1982) provided by less explicit forms (2013; ESL) found an advantage for ML feedback
of WCF. The conclusions drawn about indirect compared to direct feedback, the results in Shin-
and direct feedback in Ferris's research should tani et al. (2014; EFL) showed an advantage for
be taken with caution because indirect feedback direct WCF over ML feedback.
was provided incidentally by teachers and did not In sum, studies on different types of focused
form part of the research design at all. On bal- WCF have reported mixed findings. The conflict-
ance, thus, there are no firm conclusions about ing results obtained in WCF studies justify the rele-
the relative merits of both feedback types on ac- vance of examining potential variables mediating
curacy and language learning. its effectiveness, as discussed next.
In the previously mentioned studies, the
participants received feedback in consecutive
Variables Mediating Potential Effects of Written
treatments rather than in one session, which Corrective Feedback
disallows comparing these findings with studies
involving only one-shot feedback treatment. To The available research shows that potential
date, the few unfocused studies involving one- short- or longer term effects of WCF are mediated
shot feedback treatment have reported conflict- by a host of feedback-, task-, and learner-related
ing findings about the effectiveness of different factors. These are the three main sets of inde-
feedback types. Thus, Truscott and Hsu (2008) pendent variables targeted in empirical studies to
did not find unfocused indirect feedback effec- date, but they have not always been explored si-
tive for shortterm development (1 week) in high multaneously and in their interrelationships, an
intermediate EFL learners, while Van Beunin- important issue to be borne in mind, as explained
gen et al. (2008, 2012) showed in an L2 setting later. Although some of these sets of variables were
the effectiveness of unfocused direct and indirect not targeted in our own study, in what follows we

This content downloaded from 168.195.52.137 on Mon, 11 Sep 2023 03:51:20 +00:00
All use subject to https://about.jstor.org/terms
Florentina Nicolás-Cortesa et al. 851
offer a brief review of the relevant research to bet- is that more treatable errors are more amenable
ter situate our own investigation. to error correction than less treatable errors
(Bitchener, 2008; Bitchener, Young, 8c Cameron,
Feedback-Related, Varìables. These include (a)
2005; Ferris, 2006; Ferris 8c Roberts, 2001; Sheen,
type of WCF, namely direct/indirect or focused/
2007; Shintani 8c Ellis, 2013; Shintani et al., 2014) .
unfocused feedback (as reviewed in Bitchener 8c
Importantly, for our current purposes, Van Be-
Storch, 2016; Rang 8c Han, 2015) with or without
uningen et al. (2012) have more recently studied
ML explanations (cf. Ellis 8c Shintani, 2013; Shin-
the aforementioned interaction between types of
tani et al., 2014), and (b) frequency of the WCF,
error correction and types of error by grouping
that is, whether feedback is provided on one or
errors in their L2 Dutch participants' written out-
several occasions (see Rang 8c Han, 2015, for a re-
put in "three superordinate categories: (a) lexical
view) . Divergent findings are typically attributed
errors: word choice; (b) grammatical errors: word
to a presumed interaction between the feedback
form (e.g., verb tense, singular/ plural), word or-
provided and the target linguistic structure, which
der, incomplete sentences, and addition or omis-
points to the relevance of looking into the inter-
sion of a word; and (c) orthographical errors:
action of potentially relevant variables when ascer-
spelling, punctuation, and capitalization" (p. 12).
taining short- or longer term WCF effects.
However, in discussing the effectiveness of unfo-
Task-Related Varìables. The variables investi- cused WCF in relation to error types, the authors
gated so far include type of texts written or task basically referred to grammatical (i.e., category
conditions, that is, whether task completion and [b]) and nongrammatical (i.e., categories [a] and
WCF processing was undertaken individually or [c] and pragmatic errors). They concluded "that
collaboratively (see Bitchener 8c Storch, 2016). both grammatical and nongrammatical errors are

Learner-Related Variables. The variables that amenable to CF, but they benefit from different
types of corrections: Direct correction is better
have been found to mediate the efficacy of
suited for grammatical errors and indirect correc-
WCF include L2 proficiency (with a general ten-
tion is better suited for nongrammatical errors"
dency for greater positive effects as proficiency in-
(Van Beuningen et al., 2012, pp. 32-33). Their
creases; see review in Rang 8c Han, 2015), gram-
overall conclusion was that distinct error types are
matical sensitivity together with knowledge of
amenable to WCF, although learners benefitted
metalanguage (Stefanou 8c Révész, 2015), and in-
from different types of feedback. This assertion
ductive language learning capacity (Sheen, 2007) .
has nevertheless been questioned by Bitchener
Some researchers (e.g., Ferris, 2002) claim that
and Ferris (2012), who have noted that investi-
direct feedback is more beneficial to low L2
gating broad error groups will not help discover
proficiency students, while indirect feedback is
the specific error types that benefit from receiving
more effective in advanced learners, as low L2
WCF; for that reason, they recommended look-
proficiency learners may lack the ML aware-
ing into the effects of unfocused WCF on specific
ness to interpret indirect feedback (Ferris, 2004;
error types. We followed this recommendation in
Hyland 8c Hyland, 2006).
our own study. We developed a more fine-grained
Given the array of variables mediating potential
coding scheme for specific error types than that
effects of WCF and the likely interaction among
of Van Beuningen et al. (2012), as detailed in the
them, future research on the effects of WCF must
Method section.
go beyond single univariate designs and must ex-
plore them in relation to error types.
RESEARCH QUESTIONS

Written Corrective Feedback and Error Types


While many studies have shown that focused
WCF can lead to L2 development, few have inves-
Ferris (1999) introduced the distinction be- tigated the potential longer-term effects of unfo-
tween treatable, or rule-governed, errors (e.g., cused WCF or its effects on revision (feedback for
verb tenses or plurals), and untreatable, or id- accuracy) and on new texts (feedback for acquisi-
iosyncratic, errors (e.g., lexical choices or incor- tion). Moreover, available findings on the relative
rect word order) and suggested that untreatable effectiveness of direct and indirect WCF are still
errors might be best addressed by using direct inconclusive, and studies investigating the value
WCF, while indirect WCF seems to be more ap- of unfocused WCF for different error types have
propriate for addressing treatable errors. Once been relatively scarce. Following Bitchener and
again, research is not conclusive on that point, al- Storch 's suggestion (2016) that further work is
though, as could be expected, a general finding needed on the "potentially moderating effect of

This content downloaded from 168.195.52.137 on Mon, 11 Sep 2023 03:51:20 +00:00
All use subject to https://about.jstor.org/terms
852 The Modern Language Journal 103 ( 201 9)
both cognitive and noncognitive factors/variables 2015), and differentiates it from earlier work
on a learner's (...) use of written CF in revised that was more illustrative of laboratory-type con-
and new pieces of wridng" (p. 128), the present ditions. Furthermore, following previous recom-
study is an attempt to examine "the potential com- mendations (Bitchener & Ferris, 2012), we also
plex interaction of variables" (p. 129) that may addressed the focused-unfocused issue alongside
contribute to the effectiveness of different WCF questions such as the interaction between error
types. Accordingly, the following research ques- categories and WCF types (direct and indirect) .
tions guided our study: We did so by investigating which type of WCF, di-
rect or indirect, led to greater error reduction
RQ1. Do different types of unfocused WCF in terms of broad error categories (e.g., gram-
(direct and indirect) lead to differences mar vs. nongrammar errors) and specific error
in overall measures of accuracy in rewrit- types (e.g., verb tense, article usage, word choice,
ten texts?
or prepositions). As such, we also followed re-
RQ2. Do different types of unfocused WCF cent recommendations to opt "for studies that
(direct and indirect) lead to differences investigate the effects of WCF (...) on different
in accuracy regarding error types in structures that vary not just in terms of whether
rewritten texts?
they are rule based but also in their complexity"
RQ3. Do different types of unfocused WCF (Shintani et al., 2014, p. 109).
(direct and indirect) lead to differ-
ences in overall measures of accuracy in
posttests and delayed posttests? Participants and Setting
RQ4. Do different types of unfocused WCF
Participants were 46 (13 males and 33 females)
(direct and indirect) lead to differences
first-year EFL undergraduates (mean age = 18)
in accuracy regarding error types in
enrolled in a semester-long composition course
posttests and delayed posttests?
as part of their degree in English Studies. On
average, they had been learning EFL for about
METHOD
12 years prior to entering university and, ac-
cording to the Oxford Placement Test (OPT; Al-
In order to shed light on the likelihood of L2
lan, 2004), their proficiency corresponded to a
development through WCF, we followed Bitch-
B1 Threshold-Independent User (intermediate level) .
ener and Ferris's (2012) proposal that the design
We selected that level so that learners could pro-
of WCF studies ought to involve four or more
cess the feedback they received and also engage
tasks: A pretest, rewriting of the pretest, a posttest,
with it during the languaging session.
and a delayed posttest. Our study involved rewrit-
Participants were divided into groups of simi-
ing a previously produced text (Narrative Topic
lar L2 proficiency and writing ability on the ba-
1: pretest) after receiving one of two forms of
sis of the results obtained in the OPT, the over-
WCF (direct or indirect) as compared to a con-
all percentage of errors on the pretest, and the
trol group that received no WCF. In addition, all
number of words written in the first essay (see
participants wrote two new texts at different time
Table 1).
intervals. Taking into account the characteriza-
tion of explicitness of WCF introduced earlier, the
two forms of WCF provided differed in terms of Design and Data Collection Procedures
their explicitness: The direct WCF was more ex-
plicit given that the correction was provided (see A mixed design with one intrasubject factor
Appendix A) , whereas in the case of indirect WCF, (time, with four values: pretest, rewriting, posttest,
the exact location of each error together with a and delayed posttest) and one intersubject fac-
ML code indicating its nature were provided. tor (WCF, with three values: unfocused direct
We opted for unfocused WCF due to the du- WCF, unfocused indirect WCF, and no feedback)

bious ecological validity of research on focused was used. This resulted in two treatment groups
WCF (Storch, 2010): Focused WCF laboratory and one control group (CG): The treatment
studies may satisfy the rigor of researchers but groups received either unfocused direct (direct
may not be relevant to current pedagogy (Bruton, group [DG]) or indirect WCF (indirect group
2009), to the needs of classroom teachers (Evans, [IG]); the CG received no feedback prior to
Hartshorn, & Allen, 2010), or to L2 learners them- rewriting their texts. Because of ethical consider-
selves. This decision makes the study more repre- ations, the CG received WCF after completing the
sentative of real classroom practice (Liu & Brown, study.

This content downloaded from 168.195.52.137 on Mon, 11 Sep 2023 03:51:20 +00:00
All use subject to https://about.jstor.org/terms
Florentina Nicolás-Conesa et al. 853
TABLE 1
Initial L2 Proficiency Level (Oxford Placement Test Score), Errors, and Number of Words in the Pretest

IG (»=15) DG (n = 14) CG(»=17) v Kruskal-Wallis ........


95% CI 95% CI 95% CI Tesť
M (SD) LCL UCL M (SD) LCL UCL M (SD) LCL UCL p rj2
L2 profi- 130.27 122.04 138.49 130.50 121.77 139.23 129.47 122.14 136.80 .96 .00
ciency (14.85) (15.11) (14.25)
Errors 12.43 9.84 15.02 13.11 11.16 15.06 13.27 10.99 15.54 .64 .02
(total (4.68) (3.38) (4.42)
percent-
age)
Number of 337.53 290.36 384.70 349.21 299.37 399.05 342 301.85 382.15 .78 .01
words (85.18) (86.32) (78.09)
Note. IG = indirect group; DG = direct group; CG = control group; M = Mean; SD = standard deviation; CI =
confidence interval; LCL = lower confidence limit; UCL = upper confidence limit.
aThe effect size for Kruskal-Wallis Tests was computed by dividing Chi square value by N- 1.

Data collection occurred in five stages (see without disrupting the class dynamics. Other ad-
Figure 1) over 5 weeks during which partici- ditional reasons for using languaging were the re-
pants wrote their first text (pretest) , received WCF ported reactivity of think-aloud protocols to inves-
(treatment groups) or no feedback (CG), lan- tigate WCF processing (see Sachs 8c Polio, 2007;
guaged about their errors in their original texts Wigglesworth 8c Storch, 2012), the scarce investi-
(with or without the availability of WCF for the gation on written languaging to investigate WCF
intervention and control groups, respectively), (Suzuki, 2012, 2016), and the alleged suitability
rewrote their initial text (rewriting) and, finally, of languaging to delve into cognitive activity. The
wrote two more texts on two new topics (posttest potential combined effect of WCF and languag-
and delayed posttest) . The essays were composed ing in our data is certainly a concern, given that
under individual writing conditions to encourage we did not include a group that received WCF and
self-reflection processes about the feedback re- did not engage in languaging; however, our de-
ceived before rewriting their texts and writing up sign did allow us to disentangle the potential ef-
new texts. We are aware that this data collection fect of WCF given the existence of a control group
procedure may have resulted in an unintended that languaged about their errors but received no
practice effect but saw no other way of collect- WCF.

ing data to test the effect of the feedback on new Our data collection protocol followed Truscott
pieces of writing. This is an important method- and Hsu's (2008) advice that the interval be-
ological consideration in classroom-based studies tween revised and new texts should neither be
that look into the effect of WCF on learning and too short (to avoid fatigue) nor too long (so
one that is undoubtedly compensated by the eco- as not to "overwhelm the effects of the single
logical validity of research conducted in real class- episode of error correction and revision," p. 295).
rooms with real students. The study was intentionally similar in time span
As just mentioned, the intervention and con- to that of previous studies (i.e., Shintani & El-
trol groups differed in whether or not they re- lis, 2013; Shintani et al., 2014; Stefanou, 2014;
ceived WCF but they all rewrote their texts af- Truscott 8c Hsu, 2008; Van Beuningen et al.,
ter having engaged in written languaging, which 2008, 2012) to facilitate the comparison of results.
refers to "the process of making meaning and The English-for-academic-purposes (EAP) course
shaping knowledge and experience through lan- lasted 16 weeks and the participants received dif-
guage" (Swain, 2006, p. 98). As detailed elsewhere ferent types of feedback, and hence data collec-
(Cerezo, Manchón, 8c Nicolás-Conesa, 2019), our tion with a larger interval was precluded. During
choice of written languaging resulted from practi- the study, the teacher (a research-team member)
cal and methodological considerations. From the and the English language teacher deliberately re-
first angle, as our study was conducted in a real frained from providing grammar instruction or
classroom, written languaging allowed us to col- any other language feedback on the students'
lect data from a large number of participants texts.

This content downloaded from 168.195.52.137 on Mon, 11 Sep 2023 03:51:20 +00:00
All use subject to https://about.jstor.org/terms
854 The Modern Language Journal 103 (2019)
FIGURE 1
Timing and Order of Data Collection

Note. WCF = written corrective feedback; Ss = students. Time spent on task is indicated in parentheses.

Three narrative tasks were used (see quality of writing, to guarantee similar levels of
Figure 1): Narrative Topic 1 (pretest), Narra- difficulty in terms of content, and to facilitate
tive Topic 2 (posttest), and Narrative Topic 3 comparability across tasks. The tasks were not
(delayed posttest). Participants had 50 minutes counterbalanced, and we acknowledge this as
to plan and write each text, as this is the regular a potential limitation of the study. However, as
procedure for essay writing in language courses noted by one reviewer, the writing task is affected
in the participants' degree studies. As they had no by topics, and we did take measures to control
access to external resources, the time allocated topics for the three prompts.
to plan and write the amount of text required After having written the first text (Narrative
(about 300 words) proved adequate. All partici- Topic 1; pretest), the two feedback groups re-
pants managed to produce a text within the word ceived unfocused feedback (direct or indirect).
limits indicated. Additionally, this 50-minute pe- During the languaging session, all participants
riod allowed time both for data collection and for had 50 minutes to go over their errors and process
the participants' task involvement and resulting feedback (feedback groups) or to identify their
problem-solving behavior while composing in a own errors so as to reflect on them (CG). The
real classroom environment, a time condition participants were given the following instructions
that has been posited to be closely linked to po- based on Suzuki (2012): "Have a look at your es-
tential language-learning outcomes of L2 writing say and identify your errors. Why is each linguistic
and feedback processing (Bitchener & Storch, form incorrect? Explain it."
2016; Manchón & Cerezo, 2018). In all three cases (DG, IG, CG), the languaging
The three prompts were slightly different so session consisted of copying each error from the
that the tasks could be considered new ones but, pretest on a separate piece of paper, providing a
at the same time, they were sufficiently similar to correction (already given in the feedback, for the
allow comparisons (see Appendix B) : Participants DG) , a ML code (already given in the feedback,
were requested to narrate personal experiences, for the IG), and a ML explanation (not present
ranging from a negative experience (Narrative in the feedback provided to either of the two feed-
Topic 1; pretest), to a positive one (Narra- back groups) .
tive Topic 2; posttest), to an experience that To assure ecological validity, and in con-
made them change their opinion about someone trast to previous studies (e.g., Truscott 8c Hsu,
(Narrative Topic 3; delayed posttest). All tasks 2008; Van Beuningen et al., 2008, 2012), our
were related to personal experiences to avoid the participants had access to their original texts
influence of prior knowledge on the quantity or without corrections during the rewriting of the

This content downloaded from 168.195.52.137 on Mon, 11 Sep 2023 03:51:20 +00:00
All use subject to https://about.jstor.org/terms
Florentina Nicolás-Conesa et al. 855
pretest. In addition, rewriting the original texts of errors (just one or multiple errors) included
while looking at teacher feedback could have re- in each T-unit or the length of T-units. As a re-
sulted in an exercise of merely identifying the er- sult, short clauses are considered disproportion-
ror and changing it without showing understand- ately accurate in contrast to long clauses. For the
ing of the error or the feedback, at least in the analysis of error types (broad and narrow coding) ,
case of the DG. It was expected that during the we used the raw number of errors within each task
rewriting session (50 minutes; 2 days after the lan- and group.
guaging session), the participants could identify As for languaging, we developed a coding
in their original uncorrected essays the errors that scheme (see Appendix D) based on Leow's (2015)
they remembered from the languaging session so operationalization of depth of processing so as
as to rewrite their texts more accurately. One week to distinguish learners' cognitive effort when pro-
after having rewritten their texts, all participants cessing the WCF received (intervention groups) ,
wrote a new text (Narrative Topic 2; posttest). or when reflecting on their own self-identified er-
Two weeks after the posttest, all participants com- rors (CG). Five different levels of analysis were
posed another new text (Narrative Topic 3; de- specified. These levels distinguished students'
layed posttest) . awareness of errors at the level of noticing (Level
1 ) , at the level of reporting (Levels 2 and 3) and at
the level of understanding (Levels 4 and 5) . Three
Data Coding and Analysis Procedures
researchers set up these levels through iterative
Contrary to what is often the case in ap- analysis of the data. One researcher coded all the
plied linguistics, two coders (experienced EFL languaging data on two separate occasions, which
teachers, one of them being one of the re- resulted in an intrarater reliability of .78 (Cron-
searchers) coded all the errors together for all bach's alpha).
coding schemes to ensure maximum reliability We conducted nonparametric statistics
(100%), bearing in mind the highly detailed cod- (Kruskal-Wallis, Mann-Whitney U tests, Wil-
ing schemes employed. In cases of disagreement coxon Signed Rank test, and Friedman tests)
between raters, errors were discussed until con- given our small sample and the nonnormal dis-
sensus was reached. tribution of the data. Kruskal-Wallis tests were
Two types of coding were implemented (see Ap- conducted to compare the percentage of errors,
pendix C) : a broad coding, partially based on Van the number of error types, and the decrease of
Beuningen et al. (2012), distinguishing grammar errors across the three groups. When significant
(morphological and syntactic) and nongrammar differences across the three groups were found,
errors (lexis, spelling, punctuation, and other mi- Mann-Whitney U tests were used to find possi-
nor errors); and a narrow coding, partly based ble significant differences between two specific
on Ferris et al. (2013) and partly data-driven. Fol- groups. Friedman tests were conducted to com-
lowing Van Beuningen (2011), we further dis- pare the decrease of errors across tests within
tinguished among errors targeted in the WCF each group. Subsequently, Wilcoxon Signed Rank
that were successfully corrected (equated with up- tests were used to find if there were significant
take), uncorrected, unsuccessfully corrected, and differences between specific tasks within each
deleted. group.
As in previous studies (e.g., Truscott 8c Hsu,
2008; Van Beuningen et al., 2008, 2012), we com-
RESULTS
puted the error rate percentage for the three
pieces of writing: (total number of errors/ total
RQ1: Unfocused Direct and Indirect Written Corrective
number of words) x 100. For the rewritten texts,
Feedback and Overall Accuracy in Rewńtten Texts
we computed the error correction percentage:
(number of errors successfully corrected, unsuc- There were statistically significant differences
cessfully corrected, uncorrected, or deleted/ total among the three groups in the percentage of er-
number of corrected errors) x 100. The compu- rors successfully corrected and left uncorrected
tation of the number of errors per 100 words was (see Table 2), as well as a trend toward signifi-
an objective measure of analysis that allowed us to cance in the percentage of errors unsuccessfully
compare our results with the findings of previous corrected. As expected, the two treatment groups
research. T-units were not used in our study be- corrected more errors successfully than the CG,
cause their validity to measure linguistic accuracy although no significant differences between the
has been questioned (Skehan 8c Foster, 2007). two feedback groups were found (see Table 3).
This measure does not control for the number In addition, the CG had significantly more

This content downloaded from 168.195.52.137 on Mon, 11 Sep 2023 03:51:20 +00:00
All use subject to https://about.jstor.org/terms
856 The Modern Language Journal 103 ( 201 9)

1* O O l>IU
IS A% •*
.52
s Eř^ "f
COcdTt«
cmHcoo ^ rH CM
Š
2 00 CD C^f

ķO A ■«. O O O iT3 U

ö g;
g
(D

'G
OpS
O

ĒČ G
CG

-, Tfi
pJ CD m
-, CM Oí CN
O Oy G
U

S u
d Ů S ^ 2 fe >

OC^I IO .J>
, CO
IO IG
J>(M(M

o»_) ii CD
^ CM
2O TfSCOCO
IIH.
O . ^ , Mjî CO l-H
III
•J H lO (J
""
¿ X> u^
'i
U CD
O

mmN

O s^co
u * s^co
^ iri
CO O CSC lO flj
|s00C 1>
s c03
ug
u *Q
CM X

CD
od
^ cd^ h ^ hc c§
03 1% 00 CO
^ O l-H O
u
Um
2 00 i> p

, ^ * "il
u
ú «I
, ^ §8 g S S « I

ub
%

^ §§§ i
ič G
J

H-,-,OhOí
m l>
oí CO
o H- I
HI
'''1

^"f 'G
D CN
y CDCDTř
CM H 5
03 'G

G
O
•-Hm»oCM
o cmm8:oo
s

cG
a
03 hJ
io
, *C CO "«f l> ai C/3

^ ^ CM O
J co (M <D
U ai, 03
-riCM
coCN 00 <j
r-H- G 111 I
Č
2
C/3

C/3
11
g» V -
C- oo i> i.
fi CO (M OJ II
O r>T ci ai ^ *""! •- i
S
ïr

1
b «-H
CMXOiS
O
Ē
G
C

«<CO 1-1 O IO Tf b^.


Q ^ §, r>T C ci ai C 2°, S *""! U •- i
(U
G

m co <U
O Tť
irir-H U G •S
!
-G oo iri Ģ
^ <U
cd
.52
cd
S
G ^ If) CM r-H Z
cr cl ^ ^ 1-J o CM OJ
oo » o co oí .. 1
G
O
'G r, CTI -t< io CO CM u
1
r, ^ CM CM CD ^

8
mB
G 0
cd
.z

ič O
C/3

O J*!
C /3
C/3

•a. o! Sì g 2
Dh

u
U
•s G 2

jj 00 rHqCM -1 CM 2
°.
^ ggu§I
n -1 co 5c
C/3

•a «Ž

2
" a.
^ ►-> co m -f cm
►-> ID -F M r-H 0 a

S
¿3
x X

#
1 G
oaNN11u»
^ j cm m m co H
Oh
'C S

íá j S I 5 * 8
o a N N , 1 , 1 7 1 u "5 » „
1 o "5
u
C/3

<u

Q
1

-5-
r} IfJ Ô S CO
~ ~*-»»
id
X .S

00 03 ^ 0 S s S3 .s
C/3

eS OJ

G
HH g"
g,tFC CO
C ^'T SJ .g
fe ^ 0 g SS gl
io oco^ u Ł...o
S

^ io
^ co
coco oo
co'G
AJ

M* »-H
O
•G II £•
u rn 'G
fS
u QS Ë Q-
«5 Q Ti jss -
£
CM CD -¡F CM H* G ®
Oh ^ O
S &* 3
s r>CM
lO -FCM
CD TřCD
O Ou<fļ
U ï G -G P ^
ã be W)
ws .SeS fe 2
§gë«i
ê
G Cj C M
O
•G u
Cj 2 C 5 C+-C
Ë ë 8 .s «
ca
U e2 <U CÄ3 u c«wii 11 ^
•G o
o Ti G -G N
CA73 CA O.
o *Ū 1-1

C Geo U o ÎS Ti G S w
G

Ë
«I w w
(M O
iv « .2A w u « P, w P ,*0
A C 73 w Geo I .2 P, fc w P I ,*0 II g •e h 3 u u u c« 11 5¿¿
o S H u r n G3
3Ï P caEHCOWHO^G 4JtaS'G aŁ â> o S S § H u s «j r n
il S öl! s
il H W
HW llg!ĚIglla P caEHCOWHO^G il taS'G 4J HU a cgsa

This content downloaded from 168.195.52.137 on Mon, 11 Sep 2023 03:51:20 +00:00
All use subject to https://about.jstor.org/terms
Florentina Nicolás-Conesa et al 857

¡P J
uncorrected errors than either the IG or the DG,
1 M oficiNHpqqo
CM

and the IG corrected more errors unsuccessfully


than the CG (see Table 3).

uJ o
- ^ ^av
D

§0 © © H©h ©
© DI©H DI r-I CO
Tf r- I CO
H 00 Tf
Tf yG
G

RQ2 : Unfocused Direct and Indirect Written Corrective T3


cC

î:
_U 2
Feedback and Error Types in Rewritten Texts
H t^coooirioif-icOTîi ^ ^ £
We found significant differences among the
three groups in the successful correction of both ^*5._)HiNOOOiHNCitO 3
grammar and nongrammar errors and in the
grammatical errors left uncorrected (see Table 4) .
l-l rH J
. cor-ioe^r^co-^CM II 11
« cor-ioe^r^co-^CM ori evi ^ aí • • * * II 11

Regarding successful correction of grammar IlcomqqnPOWO


g <5 d Tf co d m .G
en p
errors, both feedback groups were significantly
O
U ^^ oo^cû03^ooct>i> qj
^ - 1 eu
ihDO«^00(N^ y
better than the CG (IG vs. CG: U = 71.50,
)2 ^ ifi w iri w rH co dì G
Z = -2.13, p = .03, r = .38; DG vs. CG: U = 13

29.50; Z = -3.56, p < .001, r - .64). However, ce


e

the DG benefitted significantly more than the IG 000(0^tifì<0<N CJ o


r tOifliOrtlNHiflTt .

UJ
from the feedback received regarding grammar CM CM


(U= 57.50, Z = -2.08, p = .04, r = .39) since

**
Í3
JD

the DG participants successfully corrected nearly H. ifjCMint^oowop


(ÛCO^OOiflND mII
y TřrHMřiíNCfiWrt .J
half of the grammatical errors for which they had

¿5
- P ' <u
received feedback (49.56%), whereas the IG suc-
cessfully corrected 41.78% of their grammatical
errors (see Tables 5 and 6) . Both feedback groups 1-1
05 (J OiflmOTfOCOfM U
J ai i> »n ^ ' i-i
' 1-1

successfully corrected verb tenses, prepositions,


II QOiNOODcOWin^
g óoí^^t^ooéoo^Ç"53e
and verb forms, as well as unnecessary articles or
missing articles. In addition, the DG also tended O ^ ©co©^©©©i> ^
to successfully correct sentence structure errors. ^ h oi oi 00 H N H H g
The two feedback groups were also signifi-
II

U^
cantly better than the CG in the successful cor- tûWCOiOMWOOif) řA
iCOtOHHCOHH <J
rection of nongrammar errors (IG vs. CG: U ■W, ,-H _ r-t g <u

= 6.0, Z = -4.61, p < .001, r = .81; DG vs. e

2
II "1p
Hj OíOO^iflOffiOOf
j HNXheoOrtiû i)
cr

CG: U = 5.5, Z = -4.52, p < .001, r = .81).


Most of the nongrammar errors corrected suc-
P OOOH*i«WN
cessfully by all the groups were related to lexis a
(wrong words), spelling, and punctuation (see *5
6 (j6 JoqoHwoìCto
hmcooocoHOioiO5
Tables 5-7). P d N iri iri r-i ^
As for uncorrected errors, the CG left signifi-
" (MCM^COr^CDOCCO Q
^KS'^t^a>oi03aqa>oqi>
co dì iri iri ' >-h i-i 'y^ ,,
cantly more uncorrected errors about grammar
in comparison with the two feedback groups (CG M ^ cor^coi>i>©cor^
w H^OI^NOHO O .. KK
vs. IG: U= 48.50, Z= -2.99, p < .001, r= .53; »OÓSOÓWW-hh ÇJ
CG vs. DG: U = 58.50, Z = -2.41, p = .02, r = 2
a
e
X o
.43); such errors pertained to prepositions, arti- <Moa>i>Tt«ox>m Jrr
¡s
cles, word form, and verb tenses (rule-based er- c u

rors; see Tables 5-7) . £ S


-3

g & o3 £ £ O
II

I
OC
RQ3: Unfocused Direct and Indirect Written Corrective G
1g-iHI-IuI°
(-i - osESErtErtĘ H - u VT
Feedback and Accuracy in Posttests and Delayed C/3

2 Ē Í Ē «E S5 o
(-i b osESErtErtĘ eÍSÍESSÍ

Posttests Ř
S 2 Í o 2 o S o 2 o o fe
u
OZOZOZOZ ZT u
A Friedman test indicated that all groups re- Ê §
duced the percentage of errors across the three W -3
G

times of data collection in a statistically significant m "8 3 «« u


ISI2i§o
O
^U
manner: IG, x2(2) = 10.13, /> = .01; DG, x2(2) = rvi QJ

21, p < .001; CG, x2(2) = 10.71, p < .001 (see


3 rvi -g
I Hnuoüihío ™ -2»
Figure 2). n U c/îuPiuPuQu
¡5á U c/îuPiuPuQu Hnuoüihío <i -2» <i

This content downloaded from 168.195.52.137 on Mon, 11 Sep 2023 03:51:20 +00:00
All use subject to https://about.jstor.org/terms
858 The Modern Language Journal 103 (2019)
OO 1% Tf is 00 H f» rf CM X h ©©©©©© 10 © oo cm ® ©
^ (O ' CO CO cf) ' « (Û ' OO I> © ' OO ' CM © go h h © ' co ' HH oo CO Tf nq ' © hh
£ ' r-i r-i ' HH' ' lÔ © ' ' H4 « gc H HH ' ' H lô CO ' HH ifi
fc fc

tí U
-o T3
fc fc

J¿ rfc
"w 13
Q Û
k, N t H Tř M M» W H Tř Ifi k, Tf Tf CM hh -<f CO 00 © CM CO © IO

o m m oo ^ h « en o t t H
^~ Tf
Tf ocm
© ' 't-»
cvir>;«-H
» co h<eoo
w »ai ^^ ^^ o®®©®®
oo ©
CM hhen
«O © CM eo o 00
lii m
GOo hen
h®O
*H Vi
<3 C nS C
8 .2 SO
u -ö U 'S
uu
u ti uu
5 fc
CO

su Du
hh CO CM 60
k, © co © cm oo © Tf © © cm cm ©
h CM CO
k, O 1> tú O 60 CM 00 © CO Tf © «s|

rf* nO CM X CM CO IO CM O nO GM © © Tf Tf © X © Hl h X ^ (N H


fc r-H ^ CM IH ^
cO ^°° iri
rH oó
ifì 6(O»o
Tpco
ifiNbj©Clno
OX«-iCOhh'
h ©
kl
C
PN^©
ri ©
H© © HH
*Tft^ © I>
OO Tf©CM
fc HH r-H 60 CM HH 60
CMCM
HH©

-ö T3
u
fc
u U
1)

fc fc
o O
u u

J~ r» ifì
h IO
w Hen
IO h
O CM © Tf
N© X H-
coI cm £
3 «m h (N hJr»h©Tf
©O CO ©
t hh CMTfCO
60 ©
© CM
hh ©
CO00hH
IO

2 X ifl O if) ® O X X h X CM l* 2 oo Tf Tf oo © © no © © © © ©
o ^ Tf cm © Tf © © > ^ ifì ai te O ^ P IO © © © CO ©
•ao tj
•a ^ ©HH
ai © Tf
Tf coCM
oo h HH
w X òTf
hh tj HH HH f IN H Tf
•a O ^ Tf CM CM CO 00 00 ci ri CO 0C HH H

fc fc
o
u a
s

C/3 ^ mm
1

C/3 hH hh
8
ir V
í/5

u
H U «M
U r, r, ©
r-Hcm
CMco
CO©hh
©CM
cmCM
-i J>
© CM
co ©
CO©CO H r» Tf^HCOCMI%© X IN H IO Tf N
II
1-H
^ HH Tf Tf ri CM CM © hH © Tf CM 60 hh
e II

e
Q*
0 a*

6 w 1-1 U C 60 hh CM
J bc £ 2 BC 2
O Lr CO © CO CM -h Tf CM © I> CO CO CM 0 fc O Lr X CM H X © X IO © 00 00 CM
u
OJ
u £ Lr «N co X n- co co m © © 1-1 cm j> ^h -H fliļ_ C Lr '■S CM © © HH © © 60 © © © HH ©
u U Pm W
2
G
S
¿3 ¿3
X X

č EÏ

u
e
•a u
e
t> u
c
•a cuu
t>
e c
OJ

?
e

fc a
C

fc
O O

lH 'c? ,U u "e'
t« O c^u « .U
"e' O g
1

V) o -fc oIHX
1
G

^e U-»H G

c/5
hT^ u
tí? o C O T3
C/5

Ç =I§iI
1 t« -sill O c . hT
<L> <L>

tí? V) 1 -H o e § .8 T3 -»H S 6
t Ē o * 2 C 'S O T3 §
ř? 6

2 fc ¿¿ ^
fc g _T
g -o Ih
13 PP
u u
^ "Sa 2 fc«4H
fc «4Hoo"go-
"go- C
C -o
. -ofcfc
T3UU
fc

u s -eil _T I £ s M
fc

s0 -e
U 2
3 "B
I . Slii ss
„-o
c •§-S §s C^ ïuu sl-p§ cÄG iŁ ? T êî „-o S Ä I
Uh Uh
O O

O
'■ã
r9 oüSw2-?So o g g1
C

1g:aI§!i¡i§11
« -â
* Ķ «u2¿¿^Cu<üOb0C2uZ il ï 1 g s a ! s g i ¡ i I io ! g I

zI« ■SS.JīStSāflSsS
r9 * Ķ «u2¿¿^Cu<üOb0C2uZ :Sg|8ëir?^ôoië5 oüSw2-?So g g1
ag
3y
HU
2 'u a. © "P fc "-»aj'S 5 © u ^ "3
Z ^ Ci > > X O H ^ X fc O H
I 2 'u ^ '§ï"ê!g|'l!s.§|l a. © "P fc O "-»aj'S ^ 5 © u O ^ "3
n •Sīllg-S'lllā'Sl
« f2»«£ofS
This content downloaded from 168.195.52.137 on Mon, 11 Sep 2023 03:51:20 +00:00
All use subject to https://about.jstor.org/terms
Florentina Nicolás-Conesa et al 859
Tř oi (û ai w o o oí o N w w FIGURE 2
^ ^ Ol ' m W 00 00 Ifl N N 00 00 (O O
2 ^ ^ ' w oi (N w eó r-* oi ^H' 'en Percentage of Errors Across Time
Ê [Color figure can be viewed at
M
T3 wileyonlinelibrary.com]
u

jy
ID

® Cł <rf n © 00 <N n to Iii Ol (£> M N

© n io <3 «©(Min®--!»
.1 " ^ I-í
<2 e
" C^î lft 60 ^
l> " <3 "«1 " 60 tO rH Cvļ CO to

U -3
8 .s -3

uu

DO
^ h eM M n
k, ^ © CO l> CM © <M ^ O ^ O h K)

y> W 00 W ^ 00 ^ Ol Cl ^ Cl tO
y>
2 g©co35 © <3
oo i>
© i> « ri
©w d -in
h (oteto
^tetê
M
C oó^Hr-H
cm <£) cm^ ,-H Note. IG = indirect group; DG = direct group; CG =
T3

ä control group.
u

^ ) cm cu
Ë
O
Importandy, no statistically significant differ-
G r, ^ 00 r-HOlO^ to O) tß ^ Ifl
u

r ^ ) ^ co in m w eo m in w m co h h ences were found among groups in the percent-


age of errors, in the number of words written
in any of the tests (see Table 8), or in the de-
crease of the percentage of errors across tests (see
C
0ma> i>
bmcu riai oo
^N^ co
w to
cotocooutohio Table 9) . It should be noted that all the scores of
•3 cm co oó ' ' <<* ©co^^'iô the pretest represent the error scores before the
1
rewriting session. We show in Table 8 the num-
â
ber of words across tests for clarification purposes,
S
8 since a possible reduction in the number of words
V
i-H

o k, h to io t(* co c 5 ^ ^ w § could influence the number of errors made by


II

s participants in each group.


Oh All groups reduced their overall percentage of

¡
errors from the pretest to the posttest, and from
0 »2 the pretest to the delayed posttest (see Table 10).

â
G
üH U E n T* n CO
B O r» Ol 00 CO CO ifl (O Tř Tf CO ^ © H
üH E r» ^ Tf 00 O Tf lfl oo N 00 to m IN N
CU W
RQ4: Unfocused Direct and Indirect Written Corrective
ô Feedback and Error Types in Posttests and Delayed

o ^
'ń Posttests
X

•3
U t)
tj
č

C
3 2?
g c No significant differences among groups were
s found in the mean number of grammar or non-
grammar errors made in any of the tests (see
1 u
«¿"e5
O ^¿y
£ Table 11). There were no significant differences
p O u„ ^ £
.S
&
V
<0
fi
& -tí-tí
Oc
^ O §o'Cü'C
-g cc «
among groups in the decrease of nongrammar
errors across time. However, significant differ-

St £ Ss 'S^ 4'EI ¿I i
&
p fi S Ē I Oc i -5 o 'C § ences among groups were found in the decrease

-e
(m

of grammar errors across time (see Table 12) and,

o Si
■§s5 Ís2» 4ìif§gl i ii
ê

? -M e sä fi,
m rather crucially, the statistical difference applied
Uh

only to the treatment groups (see Table 13). The


¡6
O

G
reduction of grammar errors in nearly all error
**ê
Wu .SfoyuCr^ I ^Sci1»
¡6 <3 I .SfoyuCr^ -M s g 5 S Í ê 1 I I -Sii e s
3Ē OïS^C^oÛliofiÎ^Z ^Sci1» types was significantly larger in the DG (pretest:
J2 2 "tj o. o "O ï ^ © "S £ 3 u ^ 'S
5y .trž H 2 CJ *2 oj ä o 4ü o« 3 2 o
5 "tj •öw'Sfec^-SöPüCÄS o. 2 o ^ © "S £ 3 u ^ 'S 335; delayed posttest: 183) compared to the IG
H <J Z <3 H &i > CJ ;£ c/3 oj O ä h o ^ t/5 o« Of 3 O 2 H o (pretest: 292; delayed posttest: 271; see Table 14).

This content downloaded from 168.195.52.137 on Mon, 11 Sep 2023 03:51:20 +00:00
All use subject to https://about.jstor.org/terms
860 The Modem Language Journal 103(2019)
1
3U
1*
T
iE
% «
£
C- C^I
c« CNļ O f-H
OO i- iOrHO
i- i O
cm o r-
c- o o

u o
E "f
¿2 -®L03
cOOí
00 00 O Tf
l> r-i I> 5c -o

î£ II^ U
CC
c

t Tf co oo Tf
co oj
u
o

03
j S *ntí
Mrļ
iE^ in
u

sIIÖtí
Ē
l6^ co
^ ^c4OHÕh
W3

rj ®î m co i-,

e
řS

p rj J 2 m » r>: co i-, ^

sge
g #2^H^Hglissi Ttii^oj
i
a
M

^ 2 ^^ w
w ^ow

5! _<N<OC^inint-»2°£25
_ ^ ^ ^ ^^ J-H U
fc OOiNpHOOtDlM^SS. OJ
^hc(ÌHNH^O)OI
25 2 & 2 28 2 o* <o ^ SO
w w ^ CO N W ^
O
II

rj o m •- 1
hJ

1 co 0> <0 u

^mmcc O
U^^^
O
II <S
p id ^ ^ CM O ^

00(J rr.
_j rr.
. 03 - 1oju
2 <N
oo _ 1- H

Ç- J 3 oá «5 g
•Ç0, _ r-H CV|

¿~~~§
2
C/3

(S

n
X „^ n„ ~^ M
co^tr.
~ "t~ oí ~ ©tiu "
C/3

^ oo
^ 2 d 2 d oo ín ti "
1
2
1 . " .- I
fr O) w m g w o C

^ o
íS cr

w ^ co m co t¿5

Q fe
OJ

G

II

n1 GM Tf QO
00Oh
C/3

O CM ^
U tí
'o

►X id cm X o
C/3

£5 O) i3
l-H
sIlUo^ O
fS

£O
CD

I O)
J Tf If) 03 S
io JUOl
00 ^to^ CO 0 «c/3
c/3
O-,

« -g
*c
u C/3

C/3

V Oh
g
Q

co
? ^O^ u
-wuC
8

o §00
Ć/3

o O feo
'S O
O ^"~"
2"_c oo
^ Cl
" cio Cl
H Tf
H ooTf
_ "~" Tf ai Tf <ji Tf o tío ocum 3cuļj3 t-<
t-< ļj be
c
I

'O
Lh
O
J

fr ^ ^ êEE * §
u
OJ

Xi
S
Ê

Z
o
w co w co O u
Tfr-HCN¡ÜCO^COt>.CS1 w g tí
Lh

T5 a£u O

G C/3


tí o« ^ OJ
OJ

£
3*

g I I <u
J3 J3
lililíteteuuto
O d otoO JŠ t* wiS I) i o

«t* O
wiSUe
G
t r<y JŠ OJ Oh
tí 8 H-> "tí
<4-1
^ Dh O QÛ,l d o Ph O Q « O CU Ue aj § OJ
•- P 4-1
O
'C 4-1 tí
tí ^ C tí
HJ tí
0ow
C ^ "tí

iâ I^ I*8
C/3

.5 tí OH

1o C«ÏÎc«
0J O


2 -S
bo
Il S as oj
oo £
W fí tí 3
tí o; tí 3 g G
5 HH
O oju oj
CQ b
< OJ
U Ph Ph
u
H Oh
SI H Oh HQ

This content downloaded from 168.195.52.137 on Mon, 11 Sep 2023 03:51:20 +00:00
All use subject to https://about.jstor.org/terms
Florentina Nicolás-Conesa et al. 861
3^S
•2 9(N O9CM®coPiO P
CO 9
f"»
1-
i*
¡2 -,•«-.-Ht^CMCOTfCM
Tf o oo co en o-,

^ O
3 v'
ļ? *. S8

U
NH ce
Tj
o I
iCMt^Tft^cOCM Oí

P 0> "
u

O II
□ lO 00 CO O rH rH U
►=(CM
Ol CM
CMCM
OHCM»o'CM
CM>- fi
iU
en co

^ CM CM
II
(I ^ Sq ^ $ Ol Ol g
3 j ® 2 2 :: 2 ^ §<

u

^
U5
^ a> CM
I ?g»£2>SäS°gS§Sg« .*

u
^ ¿ iq co

IO ^
v

HH U*
cu

G
<u

j Tf
CO CM cOrfìfì
co 00
cO ^e
T3

rj Is» Tf
enCMiri
^ rH H CM co eni hh
i><U0


^ CM i-H i-

"O -U-T

II «, §§
3vv
2 ^ W Tf QiUf->.11
2
CA

(S O

e
Q
^Ili H»qcMifìcoi>p2
CM 00 _i 11
ã s j 00 CO CO CM O QQ ÇJ
T3
U

So
Ćn
N ^ co CO
N^
8
II
S 9£929Ín9oo9ÊÍ°l£ S
^'''*'
G
O
X
0 J' ^oi^ooSoo^^^^^^ ' ' ' * ' C
T3
ce
1

fcS o
e

o3
o

niniqcq^coio
r^- co co oo en || "
vi

_j "

p s fe
cu
ex
3
^ 00 00 CA

^ IO
2 (S -"J «CMcoCMid-h co io i-H
.-< CM co ncoUUn
O CA

d* Q
G
G

ë
i
«-« i t>« 00 © if) lO O^
8 »-I ,-H H C
h
Il
" "9 Ï^COrnScOt-rH
*d co cri cmi ©ai^ fi
SIl ^rH S
ž 'S
M
a

9o
9?o
CA

0
- V II
___O
CA
CA -o II

8 §
u
§ ^5oò^«q®gq§flq^ ¿T
CA
CU

(X •fi
y ddÔMCOcûNUODflîW^ fa-
¡§! h^CM>"h^h3h2,h!£, g
e G
CA

"fi
2
§ be
a fe
en
m 'G 00 *■>
u
0 N CM CO W
I I .fa I g

Pn c3 c3 c3 .'l
cu
T3 Oh
ff 'C

ç-i-i-ia
II
u
e CA
O

"icicle
g Gi Io sS oI !2SIo 3
a-
(D (U

O Q
8

P
cu
ix CA
Oh

O Z O Z O Z fe
<U

1 ît» uI £
tu

I Ł .§ !
w -o 'S
fa
Uh
u 'S
O
5 g. .s
G
Uh
O fi

S Sa ^'gsa ^S
hH O s I ii
ïï S
II

¡I 2iS^
co co
ā|
?I JßCU
H Ü OH
îi ^^ ^^ 11 J3 Ou
H H G Oh
o3QQoQ.
£ <<

This content downloaded from 168.195.52.137 on Mon, 11 Sep 2023 03:51:20 +00:00
All use subject to https://about.jstor.org/terms
862 The Modern Language Journal 103 ( 201 9)
CM ^ co x © ^w X--H
X CVI
f-i
c- © ^ © ©

Vi

tS
Vi

_ rH co m
•äh _ O CO cvi

_ m in m o
rt; o H a

2
Vi

©©©
uo o »- j
Ł) ^ cvi r-í rH j
u
III
S

3s 2ss2 V5 C/5
c

H 00
Uh CU CU
S
s
O
_WVi 'C H "Ū
aj Vi ÛJ vi
CU

O 3 cd 3 (73
s _ B 3 Vi aj 3 Vi ÛJ &
ÇQ Vi TÎ <rrļ
3

'C O O O O 2
Cd OhÛPhÛ
$
Vi
CU ¿¿¿i o
Oh

o2222
Ë </> C/J C/3 C/5 Vi

2
T3
O W <U O <ü o
W Uh Uh Uh Uh I
G
cd Pm Pm Pu PU
<
V)

o
in oo o S
cu g
3 2
O
Uh M
Uh

O ^ in r-
^OI 0>
X IIIIIIII
rH (JU
cd
tuo
c S
o
O.
3
O
Vi
Uh
<u
O 10
& Uh

Uh
X 1
S "5
fc
3
Z
11

M
Uh O
G
O 0
Uh V
(U o.
X) 3 3
Ē 2
u lo
3
o

rH
o •-^
Z O
<u Um

O 1
■S
cd
G
3
Cvi •- O Uh cd I CO §

a^
•£ fe O O g II
fe OU. o
rH <U
O & Uh & g
3 1 ÇsgI aI p" S S ^ T*
S Uh
Ch U 2
2 cIO2
2 oc ss I p" fe S S r ^ "2Ž
< OJ
HO w Z HU

This content downloaded from 168.195.52.137 on Mon, 11 Sep 2023 03:51:20 +00:00
All use subject to https://about.jstor.org/terms
Florentina Nicolás-Conesa et al. 863
?wM<y
jT)CM
^ 0 (¿
jT)I>^ XH O!NrHO OO 1>
M Tf I> TfLO
LO CM
Tf 00
X ©rM-^O^
rH
O 2 ^ 00 2©
CM rHCM hh
£s
Q tor,r,toomo
<U o
cm xXX 00 m
i> r-H Tfi>OwOo O
Tfco
co CO
co CO

^ co^ Ox Hcm
x> rf lo
c£>x aix ©LO CObJ>
o ©io mGM
h© ©
^ £ CM CM --h' OI> LO ^ LO J> cri l> ^

U CM rM (M
Il U r-> oi M rH I- I r- ( Tf CM i- I
5
r* I cor.LOTfLOoîCMx Tf
r* Utototo o oTfloCD
ooWooMn
lo (N
o hh
»
O
LOiol>OCM
Tf1HI>CiXCM
© CM COCO
© CO CM00CO
CM©©
*¿ H O ^ O CM O l> X CO CO ^
*ß »-H CM CM r- i i- i CM LO i-H i-T

rH ^ t-H CO
w

Oír,.
to OìXCO
Tf oo OCOTfLO
lOCO Tf Tf X Tf O i-CM
(
K

X CM X (Û m CM
LO 1> X CO LO LO © CO D O X ©
7 «> oLO
w
coX> CM LOO) X ©l> rM O O ©
CMcoCMcm cm^ ai
CM CM CMai(No ^

rH rH
JS «

tor-H
XXTf rHCMrf
CMTf xa ^LO COLO X
<u o
Q to r, X CM CO CM I> CM

l> X
rHhHq rH
Tf0> Tf § N CM
CO CM
LO co

£ fr
^ CMtc
1-1 © aór^odaÌLO
Tf 2co
© i>
LOcm co
Il Q¿ CM CM , - ( 1- H Tf CM CM

w CM
-OS to
i-H CM
ce
O Li Tf CM rH rH rH O X LO CM © X LO
X Li to CM LO I> CM LO Tf LO O CO lO CM

£s;
« o 5I
co co co xx i-H 2© xxx 2
CO Tf rH ļ> X X ©O I> 1-H O ©
*i
ftoo
CMi> i>
CM Tf,-hph
Tf i> ** co ai co Tf ^
lOrHCM
^ e.toXCM
CMCi h
O)COO X LO
ł- I LO LOCOLOXf-H
XX X co
LO CM H ^X fifi
CM

so
so 11 3u
a

a CO
Tf CO OX LO
I-ILO
rHTf 2Tf
00 lC
,-ho iH,- ©i 0«4h
ex 2

7 'K ^OÌlOCMX I>CM CM ©


co Tf io CO ^ t ^^ M
cu CM CM *- I CM LOrHCM ^
fi

2
o

§ II
•O

Q to Lr
to to
LrCM
l> aCOCM
COa rH
TfXo CO
H dt^OCMLO
X O co X
CM ^h X_iHH _
fi
a
OC

gu
ooo oo©o x2 x^ cm a io oo
2 5^
¡a

^^^ ©(U
^oopLOioq CM rn ,-h
S

M
Se

u
to1-HXCMrHrH
CM O LO LO LO LO 1- I CM „ II^ „
Tf 00 Tf CM
£

^ tori- Tf o O ,-h ,-h Tf o CM LO LO Tf CO • -


CM rH ^3
CA

Oto CM O LO X rH CM CM O rH O X Tf X ^3 fi-1

¿1 to, o
&

^ woxo ocoqqcm
TfÖa©00© oo »oot^ o 2^^ fi
u fe

coCO
f-î lo Ö i>CO
q 00
Tf rH Tf
CO ufi
£ rHCMCMr-trH^H X CM -fi
Ê
W
Im
O

to,toxax cmXX
h Tf
LO cm a i>xx
i""» omo
CA

co (O xa CM h1
<U

CM rH X ¿¿
bo

ch
ch O o
S o .. fe
^_r fa2 ¡ bo
S
c
D
fi

2
2

-e ^_r fa t
d
(X

fi5 g 2 "g S 2 S g ctrSSg Oo;Sht.h


T3
q T3 q
Q* gf-N2CvS CCv
_ ¿3Cprt_Ł S?5 ?5 ŁS Oo;Sht.h H^X-^C
ö Ü CV fi fi ^
g
CA
-T T3 O H^X-^C fi tn
T* -Sá °rĶ Q* SWËË^^P f-N prt . S -T T3 S 'S S S i 2 O OOeguS* tn Ü CV
T* 2 u
" °rĶ t lia^lsgg-ggigli-s SWËË^^P . S S 'S S i 2 »'S OOeguS* g 'S g j «
31
H Hto
s? to
-s s-s Sl§ SISIFI i 1 'g f. i

This content downloaded from 168.195.52.137 on Mon, 11 Sep 2023 03:51:20 +00:00
All use subject to https://about.jstor.org/terms
864 The Modern Language Journal 103 ( 201 9)
The DG reduced errors related to árdeles, prepo- studies on feedback (Van Beuningen et al., 2012).
sitions, and verb forms. The IG slightly reduced However, our study advances knowledge regard-
errors related to articles and word form. All the ing the beneficial effect of WCF versus no feed-
errors that both feedback groups reduced across back in rewritten texts by showing what gets cor-
tests were rule based. rected, what does not get corrected, and why. As
an overall statement, L2 writers do not appropri-
DISCUSSION ate all the WCF they are provided with. In fact, re-
gardless of the type of WCF received, only around
We will discuss our findings using the dual dis- half of the total percentage of errors made in the
tinction of feedback for accuracy and feedback for pretest and targeted with unfocused direct or indi-
acquisition mentioned at the outset as well as in rect WCF were successfully corrected (IG: 45.15%;
the title of the article. DG: 51.03%). This tendency in our data for a lim-
ited appropriation of WCF may reflect the learn-

Feedback for Accuracy


ers' L2 proficiency level, which may determine
students' depth of ML reflection (Sheen, 2007)
Two main sets of data are worth reporting and when processing feedback and the corresponding
interpreting. First, our results reinforce previous retention of error corrections, as previous studies
findings (cf. Ferris & Roberts, 2001; Truscott 8c conducted with EFL learners at low L2 proficiency
Hsu, 2008; Van Beuningen et al., 2008, 2012) on levels obtained similar results (Nicolás-Conesa,
the role of WCF (direct and indirect) in improv- Roca de Larios, 8c Monteverde, 2017). Psycholin-
ing the overall accuracy of rewritten texts: Our guistic readiness, as noted by one reviewer, is an-
treatment groups improved the accuracy of their other potential explanation, although our data do
texts (including both grammar and nongrammar not allow us to provide any kind of robust specu-
errors) in a statistically significant manner as com- lations in this respect.
pared with the CG. Given that both the feedback Our findings further corroborate the overall
and control groups engaged in written languag- conclusion in Rang 8c Han's (2015) recent meta-
ing, the observed effects can be attributed to the analysis that the observed effects of WCF stud-
WCF received. However, the type of WCF pro- ies are low to moderate, probably because not
vided was found to influence the depth of pro- all the feedback provided is noticed and under-
cessing observed in the participants' written lan- stood by learners. An obvious implication for fu-
guaging data, the overall finding being that direct ture research would be to investigate in detail the
WCF resulted in deeper processing (Cerezo et al., diverse options that apparently emerge when L2
2019) . Accordingly, as astutely observed by one re- writers receive and process WCF, crucially includ-
viewer, our study does shed light on the combined ing situations in which the provision and process-
effect of WCF plus languaging. Given also that ing of WCF is an iterative process rather than just
all three groups were attending the EAP course, a one-off event in the language-learning and writ-
the differences observed can be safely attributed ing experience of additional language learners
to the combined effect of WCF plus languaging and users.
(intervention groups) or writing plus languaging In addition to limited appropriation of the
(CG). In any case, we would like to argue that WCF received, our data also show that different
research in this domain that pursues ecological types of errors appear to benefit differentially
validity has to be able to accommodate the out- from distinct types of WCF. Specifically, the DG
comes of conducting research in real classrooms showed significantly higher uptake of WCF re-
(which is where we would ultimately be interested lated to grammar errors than the IG. This finding
to know more about any language-learning po- contrasts with Van Beuningen et al. 's (2012) re-
tential that may derive from writing and WCF) as search, which reported no significant differences
compared to the strictures and controls required between the two feedback groups. However, Van
and expected in more experimentally controlled Beuningen et al. noted the same tendency regard-
conditions of laboratory-type studies. This is, in ing performance of the DG. Our IG participants
effect, one of the more evident outcomes of the may have experienced greater difficulties in un-
most recent disciplinary debate in this domain derstanding the corrections due to the implicit
(Manchón, 2020). nature of indirect WCF compared to the more
Importantly, no significant differences were explicit corrections given to the DG, as was also
found between the two feedback groups in the found by Vyatkina (2010) . This might also explain
overall uptake of the WCF received. This result why the IG provided significantly more wrong cor-
is in line with the findings of recent unfocused rections than the CG.

This content downloaded from 168.195.52.137 on Mon, 11 Sep 2023 03:51:20 +00:00
All use subject to https://about.jstor.org/terms
Florentina Nicolás-Conesa et al. 865
To sum up, in terms of feedback for accuracy, vided and the consequent data analyses imple-
our study sheds light on the possible reasons mented. The study participants received feedback
why different types of feedback are better than on a large variety of error types captured in 19 er-
no feedback for the improvement of accuracy ror codes. In as much as restricted forms of unfo-
of rewritten texts. Whether this applies to writ- cused WCF (as in Van Beuningen et al. 's 2008 and
ers of different proficiency levels, to tasks differ- 2012 studies) actually approximate the type of fo-
ent from those used in our study, or to situations cused feedback targeted in some previous studies,
in which writing and rewriting after receiving it is less suited to answering one of the key ques-
WCF are iterative processes rather than isolated tions in writing pedagogy: Which errors benefit
learning events are empirical questions worth from error correction, when, and why?
adding to future research agendas. We stated previously that all three groups of
writers reduced the number of errors with no sig-

Feedback for Acquisition


nificant differences among groups in the num-
ber of grammar or nongrammar errors made
All three groups reduced their overall percent- in the pretest, posttest, and delayed posttest.
age of errors across new tasks and there were no However, greater improvements were observed in
significant differences in such overall reduction the data of the participants that received direct
of errors among groups. These findings are in WCF, who significantly reduced the number of
contrast with previous studies (Van Beuningen grammar errors from the pretest to the delayed
et al., 2008, 2012). Given that our design did posttest compared to the IG. These findings re-
not include a group that engaged in repeated fute Truscott's (2007) contention that the effec-
writing over time but did not engage in written tiveness of feedback is restricted to nongrammar
languaging, our data does not allow us to offer errors and suggest that direct feedback is effec-
firm conclusions on the potential effect of mere tive in the long run for grammatical errors com-
writing practice over time, an area of research pared to indirect feedback, a significant expan-
that certainly needs further investigation from sion of Lalande's (1982) conclusion that indirect
the perspective of task repetition in the envi- feedback encourages students' problem-solving
ronment of writing (see Bygate, 2018; Manchon, and results in long-term development. Our con-
2014a, 2014b). tention (reinforced by the languaging data, as
The contrast between our findings and those mentioned before) is that the explicitness of di-
reported in Van Beuningen et al. (2008, 2012) rect feedback helped the participants in the direct
might be related to differences in design, particu- feedback group to better understand the errors
larly in terms of participants' characteristics and, in the languaging session and retain the correc-
especially, the nature of the writing tasks. Van Be- tions for the writing up of new texts. In fact, the
uningen et al. 's adolescent participants wrote a errors that were reduced across tests were rule
short email (120 words) to a friend in 20 minutes based (e.g., articles, prepositions, verb forms).
following previous explanations given by the re- This proposal is in line with views of some SLA re-
searcher. In contrast, our university participants searchers, who note the advantages of providing
were asked to write three narrative tasks about clear information about specific targeted struc-
different personal experiences (300 words) in a ture (s) (e.g., Bitchener 8c Knoch, 2010b; Cerezo
50-minute class session. Thus, there were differ- et al., 2019; Manchón, 2011b). Although the su-
ences in the type of tasks used (email versus nar- periority of direct WCF over indirect WCF is
rative tasks), the length of the essays (short vs. well established (e.g., Bitchener, 2008; Guo, 2015;
long texts, and shorter texts are likely to result in Sheen, 2007) , the effects of unfocused WCF await
fewer overall errors and error types; Liu 8c Brown, clarification. Here, our study offers empirical ev-
2015), and the time allotted for task completion idence that contradicts Truscott's claim (2004)
(20 minutes vs. 50 minutes) . Taken together, these that grammar correction cannot lead to learning.
task-related differences might explain higher re- Additionally - and we would argue, importantly -
tention of overall L2 corrections for the feed- the current study offers new insights into the
back groups across tasks in Van Beuningen et al. 's potential effectiveness of unfocused direct WCF
(2008, 2012) research. Future research agendas when writing up new texts. Previous studies had
should therefore focus on the interaction of the just shown the effect of (a) feedback on the over-
effects of WCF as well as the tasks and task condi- all accuracy of revised texts (e.g., Ferris 8c Roberts,
tions/instructions used. 2001; Truscott & Hsu, 2008), (b) focused feed-
Another important contribution of the present back on the overall level of accuracy in new texts
study pertains to the nature of the feedback pro- (e.g., Bitchener, 2008; Bitchener 8c Knoch, 2008),

This content downloaded from 168.195.52.137 on Mon, 11 Sep 2023 03:51:20 +00:00
All use subject to https://about.jstor.org/terms
866 The Modern Language Journal 103 ( 201 9)
or (c) unfocused direct feedback on the over- grammar in the long term in ESL (e.g., Bitchener
all level of accuracy in new texts or on gram- 8c Knoch, 2008; Sheen, 2007) and EFL settings
matical accuracy compared to a practice group (e.g., Guo, 2015; Stefanou, 2014). Additionally,
(Van Beuningen et al., 2012) but not to unfo- our findings could be taken into consideration in
cused indirect WCF group. Our study is the first future studies that follow Bitchener and Ferris's
to find evidence for the possible superiority of (2012) recommendation, mentioned in the intro-
unfocused direct WCF on the retention of cor- ductory section, to investigate the effects of WCF
rections on grammatical errors from a pretest not only on overall measures of accuracy across
to a delayed posttest compared to the effects time, but also on specific error types within gram-
of unfocused indirect WCF in the same writing mar and nongrammar categories.
conditions.
We would argue that the languaging session CONCLUSION
(50-minute session) made possible the notic-
ing (i.e., detection) and retention of corrections In line with the study's theoretical and peda-
needed for language learning (see Bitchener & gogical motivation, we conclude with comments
Storch, 2016) and more specifically for grammar along these two dimensions. From the perspec-
learning, since the retention of corrections was re- tive of theory and research, our study disaggre-
stricted to grammar (i.e., rule-based) errors in the gates the effects of WCF in terms of the dual dis-
long term. This tendency may not hold for lexis tinction of feedback for accuracy and feedback
because most of the lexical errors in both feed- for acquisition and in the directions for future
back groups were related to an untreatable error research that might signal. Our data show that
(wrong selection of words) , which could be more (a) writers appropriate only part of the feedback
difficult to process and retain when unfocused they are given, and hence only part of the WCF
feedback is provided. Perhaps previous studies received is effectively incorporated into rewrit-
were unable to detect the effects of unfocused di- ten texts, and (b) the combined effect of unfo-
rect feedback over indirect feedback for the re- cused direct WCF plus languaging is significantly
tention of grammar corrections across time due more beneficial for grammar learning in terms of
to the nature of the feedback processing session: immediate uptake and the decrease of grammar
These studies often included a rewriting session in errors across time in new written texts (delayed
which learners were provided with their original posttest) compared to unfocused indirect WCF
texts and the subsequent corrections. As a result, with languaging. It follows that one direction for
learners simply had to copy their corrected essays further studies is to investigate the interaction of
onto a new piece of paper on the same day (20-30- task-, learner-, and feedback-related variables,
minute session) that they received feedback (e.g., bearing in mind learners' elaborate noticing
Truscott 8c Hsu, 2008; Van Beuningen et al., 2008, of errors that can lead to language learning.
2012). This may have precluded the kind of elab- These interactions should be explored in ped-
orate noticing of errors (i.e., noticing with under- agogically relevant and ecologically valid con-
standing) purported to lead to acquisition (feed- texts where providing and processing feedback
back for acquisition; Manchón, 2011a). are part and parcel of the language and writ-
If the tendency observed in our data were sup- ing learning experience of L2 learners. Such
ported by other studies that include a feedback research also has to find ways of counteract-
processing session, it might be possible to arrive ing the threats to validity that naturally emerge
at an answer to the pedagogically relevant ques- in research looking into feedback for acquisi-
tion mentioned at the outset of the article, that is, tion in pedagogically relevant and ecologically
which errors benefit the most from which type of valid ways, always bearing in mind that, as noted
WCF. Our data distinctly point to the advantage in an earlier section, we need to acknowledge
of unfocused direct WCF over unfocused indirect in full the nature of classroom-based research
WCF for grammar errors across time after a rewrit- in our search for robust ways of providing
ing session in which students use their original valid answers to the empirical questions in the
essays without corrections. Furthermore, the suc- domain.
cessful uptake of grammar corrections would ap- From the perspective of pedagogy, three main
pear to facilitate long-term acquisition provided implications derive from both the design and the
there is a languaging session that makes possible results of our study. The first addresses the nec-
the noticing of errors. These findings coincide essary balance between targeting a range of er-
with those reported in previous studies about the rors in the WCF provided to L2 users and guard-
effectiveness of focused WCF for the retention of ing against unnecessary burdens on students'

This content downloaded from 168.195.52.137 on Mon, 11 Sep 2023 03:51:20 +00:00
All use subject to https://about.jstor.org/terms
Florentina Nicolás-Conesa et al. 867
WCF processing capacity. A related point might Bitchener, J., 8c Knoch, U. (2010a). The contribution of
be that focused feedback might be more appro- written corrective feedback to language develop-
priate for long texts, whereas unfocused WCF ment: A ten-month investigation. Applied Linguis-
tics, 31, 193-214.
might be more successfully appropriated when
Bitchener, J., 8c Knoch, U. (2010b). Raising the linguis-
provided on shorter texts (see also recent dis-
tic accuracy level of advanced L2 writers with writ-
cussion of various combinations in Lee, 2017).
ten corrective feedback . fournal of Second Language
Finally, it is worth reiterating the pedagogical Writing 19, 207-217.
relevance of the finding that unfocused direct Bitchener, J., 8c Storch, N. (2016). Written corrective feed-
WCF appears to be more beneficial than indirect back for L2 development. Bristol, UK: Multilingual
WCF for the appropriation of WCF on grammar Matters.

in terms of immediate uptake and longer term Bitchener, J., Young, S., 8c Cameron, D. (2005). The ef-
retention. fect of different types of corrective feedback on
We would not like to finish without acknowl- ESL student writing, fournal of Second Language
Writing, 14, 191-205.
edging the limitations of our research, especially
Bruton, A. (2009) . Designing research into the effects of
the practice effect that our design might have en-
grammar correction in L2 writing: Not so straight-
tailed, and the use of only one type of task. It is
forward. fournal of Second Language Writing, 18,
hoped that future studies can take into consider- 136-140.
ation these limitations as well as our findings and Bygate, M. (Ed.). (2018). Learning language through task
their interpretation when designing new studies repetition. Philadelphia/ Amsterdam: John Ben-
that can shed further light on the theoretical and jamins.
applied dimensions of the effects of WCF from Cerezo, L., Manchón, R. M., 8c Nicolás-Conesa, F.
the dual perspective of feedback for accuracy and (2019). What do learners notice while process-
feedback for acquisition. ing written corrective feedback? A look at depth
of processing via written languaging. In R. Leow
(Ed.), The Routledge handbook of second language
research in classroom learning (pp. 171-185). New
ACKNOWLEDGMENTS
York: Routledge.
Chandler, J. (2003). The efficacy of various kinds of er-

This study was financed by the Spanish Ministry of ror feedback for improvement in the accuracy and
Economy and Competitiveness (research grant FFI2016- fluency of L2 student writing, fournal of Second Lan-
79763-P) and by Fundación Séneca, Murcia, Spain (re- guage Writing, 12, 267-296.
search grants 19463/PI/14 and 20832/PI/18). Ellis, R., 8c Shintani, N. (2013). Exploring language peda-
gogy through second language acquisition research. Lon-
don: Roudedge.
Evans, N., Hartshorn, J., 8c Allen, E. (2010). Written cor-
NOTE
rective feedback: Practitioners' perspectives. Inter-
national Journal of English Studies, 10, 47-77.
1 Explicitness in this context refers simply to whether Ferris, D. R. (1999) . The case for grammar correction in
or not the correct form is explicitly provided, a relevant L2 writing classes: A response to Truscott (1996).
clarification to be made given that WCF (as opposed Journal of Second Language Writing, 8, 1-10.
to some forms of oral corrective feedback) is always ex- Ferris, D. R. (2002). Treatment of error in second language
plicit, as discussed in the relevant literature (see, e.g., student writing. Ann Arbor, MI: The University of
Polio, 2012). Michigan Press.
Ferris, D. R. (2004). The "grammar correction" debate
in L2 writing: Where are we, and where do we go
from here? (And what do we do in the meantime
REFERENCES
...?). Journal of Second Language Writing, 13, 49-62.
Ferris, D. R. (2006). Does error feedback help student
Allan, D. (2004). Oxford placement test. Oxford: Oxford writers? New evidence on the short- and long-term
University Press. effects of written error correction. In K. Hyland 8c
Bitchener, J. (2008). Evidence in support of written cor- F. Hyland (Eds.), Feedback in second language writ-
rective feedback. Journal of Second Language Writ- ing: Contexts and issues (pp. 81-104). Cambridge:
ing , 17 ; 102-118. Cambridge University Press.
Bitchener, J., 8c Ferris, D. R. (2012). Written corrective feed- Ferris, D., Liu, H., Sinha, A., 8c Senna, M. (2013). Writ-
back in second language acquisition and writing New ten corrective feedback for individual L2 writers.
York: Routledge. Journal of Second Language Writing, 22, 307-329.
Bitchener, J., 8c Knoch, U. (2008). The value of written Ferris, D., & Roberts, B.J. (2001). Error feedback in L2
corrective feedback for migrant and international writing classes: How explicit does it need to be?
students. Language Teaching Research, 12, 409-431. Journal of Second Language Writing, 10, 161-184.

This content downloaded from 168.195.52.137 on Mon, 11 Sep 2023 03:51:20 +00:00
All use subject to https://about.jstor.org/terms
868 The Modern Language Journal 103 ( 201 9)
Guo, Q. (2015). The effectiveness of vmtten CF for L2 de- Polio, C. (2012). The relevance of second language ac-
velopment: A mixed-method study of vmtten CF types, er- quisition theory to the written error correction de-
ror categories and proficiency levels (Unpublished doc- bate. Journal of Second Language Writing, 21, 375-
toral dissertation). Auckland University of Tech- 389.

nology, Auckland, New Zealand. Rummel, S. L. (2014). Student and teacher beliefs about
Hyland, K., 8c Hyland, F. (2006). Feedback on second written CF and the effect those beliefs have on uptake:
language students' writing. Language Teaching, 39, A multiple case study of Laos and Kuwait (Unpub-
83-101. lished doctoral dissertation). Auckland University
Kang, E., 8c Han, Z. (2015). The efficacy of written cor- of Technology. Auckland, New Zealand.
rective feedback in improving L2 written accuracy. Sachs, R., 8c Polio, C. (2007). Learners' uses of two types
Modern Language Journal, 99, 1-18. of written feedback on a L2 writing revision task.
Lalande, J. F. (1982). Reducing composition errors: An Studies in Second Language Acquisition, 29, 67-100.
experiment. Modern Language Journal, 66, 140- Sheen, Y. H. (2007). The effect of focused written cor-
149. rective feedback and language aptitude on ESL
Lee, I. (2017). Working hard or working smart. Com- learners' acquisition of articles. TESOL Quarterly,
prehensive versus focused written corrective feed- 41, 255-284.
back in L2 academic contexts. In J. Bitchener, N. Shintani, N., 8c Ellis, R. (2013). The comparative effect
Storch, 8c R. Wette (Eds.), Teaching writing for aca- of direct written corrective feedback and metalin-

demic writing to multilingual students (pp. 168-180) . guistic explanation on learners' explicit and im-
New York: Routledge. plicit knowledge of the English indefinite article.
Leow, R. (2015). Explicit learning in the classroom. A Journal of Second Language Writing, 22, 286-306.
student-centered perspective. New York: Routledge. Shintani, N., Ellis, R., & Suzuki, W. (2014). Effects of
Liu, Q., 8c Brown, D. (2015). Methodological synthesis written feedback and revision on learners' accu-
of research on the effectiveness of corrective feed- racy in using two English grammatical structures.
back in L2 writing. Journal of Second Language Writ- Language Learning, 64, 103-131.
ing, 30, 66-81. Skehan, P., 8c Foster, P. (2007) . Complexity, accuracy, flu-
Manchon, R. M. (2011a). The language learning po- ency and lexis in task-based performance: A meta-
tential of writing in foreign language contexts: analysis of the Ealing Research. In S. A. Van Daele,
Lessons from research. In M. Reichelt 8c T. Chi- F. Kuiken, M. Pierrard, 8c I. Vedder (Eds.), Com-
masko (Eds.), Foreign language writing: Research in- plexity, accuracy and fluency in second language use,
sights (pp. 44-64). West Lafayette, IN: Parlour learning and teaching (pp. 207-226). Brussels: Con-
Press. tactforum.

Manchon, R. M. (2011b). Writing to learn the lan- Stefanou, C. (2014). L2 article use for generic and specific
guage: Issues in theory and research. In R. plural reference: The role of written corrective feedback,

M. Manchon (Ed.), Learning-to-write and writing- learner factors and awareness (Unpublished doctoral
to-learn in an additional language (pp. 61-82). dissertation). Lancaster University, Lancaster, UK.
Philadelphia/ Amsterdam: John Benjamins. Stefanou, C., 8c Révész, A. (2015). Direct written correc-
Manchon, R. M. (2014a). The distinctive nature of task tive feedback, learner differences, and the acquisi-
repetition in writing. Implications for theory, re- tion of second language article use for generic and
search, and pedagogy. ELIA, 14, 13-41. specific plural reference. Modern Language Journal,
Manchon, R. M. (2014b). The internai dimension of 99, 263-282.
tasks: The interaction between task factors and Storch, N. (2010). Critical feedback on written correc-
learner factors in bringing about learning through tive feedback research. International Journal of En-
writing. In H. Byrnes & R. M. Manchon (Eds.), glish Studies, 10, 29-46.
Task-based language learning: Insights from and for Storch, N., 8c Wigglesworth, G. (2010). Learners' pro-
L2 writing (pp. 27-52). Philadelphia/ Amsterdam: cessing, uptake, and retention of corrective feed-
John Benjamins. back on writing. Studies in Second Language Acquisi-
Manchon, R. M. (Ed.). (2020). Writing and lan- tion, 32, 303-334.
guage learning: Advancing research agendas. Philadel- Suzuki, W. (2012). Written languaging, direct correc-
phia/Amsterdam: John Benjamins. tion, and second language writing revision. Lan-
Manchon, R. M., 8c Cerezo, L. (2018). Writing as lan- guage Learning, 62, 1110-1133.
guage learning. The TESOL encyclopedia of English Suzuki, W. (2016). The effect of quality of written lan-
language teaching (pp. 1-6). New York: Wiley. guaging on second language learning. Writing &
Nicolás-Conesa, F., Roca de Larios, J., 8c Monteverde, Pedagogy, 8, 461-482.
M. A. (2017). The language learning potential of Swain, M. (2006). Languaging, agency, and collabora-
writing through EFL students' processing of feed- tion in advanced second language proficiency. In
back. Porta Linguarum, 2, 187-200. H. Byrnes (Ed.), Advanced language learning: The
Norris,J., 8c Ortega, L. (2003). Defining and measuring contribution of Halliday and Vygotsky (pp. 85-108).
SLA. In C. Doughty 8c M. Long (Eds.), The hand- London: Continuum.
book of second language acquisition (pp. 717-761). Truscott,J. (1999). The case for "The case against gram-
Oxford, UK: Blackwell. mar correction in L2 writing classes": A response

This content downloaded from 168.195.52.137 on Mon, 11 Sep 2023 03:51:20 +00:00
All use subject to https://about.jstor.org/terms
Florentina Nicolás-Conesa et al 869
to Ferris. Journal of Second Language Writing, £,111- Van Beuningen, C. G., De Jong, N. H., 8c Kuiken, F.
122. (2008). The effect of direct and indirect correc-
Truscott, J. (2004) . Evidence and conjecture: A response tive feedback on L2 learners' written accuracy. ITL
to Chandler. Journal of Second Language Writing, 13, International Journal of Applied Linguistics, 156, 279-
337-343. 296.

Truscott, J. (2007). The effect of error correction on Van Beuningen, C. G., De Jong, N. H., 8c Kuiken, F.
learners' ability to write accurately. Journal of Sec- (2012). Evidence on the effectiveness of compre-
ond Language Writing, 16, 255-272. hensive error correction in second language writ-
Truscott, J., 8c Hsu, A. Y. P. (2008) . Error correction, revi- ing. Language Learning, 62, 1-41.
sion, and learning. Journal of Second Language Writ- Vyatkina, N. (2010). The effectiveness of written correc-
ing, 17, 292-305. tive feedback in teaching beginning German. For-
Van Beuningen, C. G. (2011). The effectiveness of compre- eign Language Annals, 43, 671-689.
hensive corrective feedback in second language writing Wigglesworth, G., & Storch, N. (2012) . What role for col-
(Unpublished doctoral dissertation) . University of laboration in writing and writing feedback. Journal
Amsterdam, the Netherlands. of Second Language Writing, 21, 364-374.

APPENDIX A

Example of Direct Correction

This content downloaded from 168.195.52.137 on Mon, 11 Sep 2023 03:51:20 +00:00
All use subject to https://about.jstor.org/terms
870 The Modem Language Journal 103(2019)
Example of Indirect Correction

APPENDIX B

Writing Prompts
Narrative Topic 1: Write about something negative that happened to you, something that went wrong.
Narrative Topic 2: Write about an experience of success that you had in your life.
Narrative Topic 3: Write about an experience that made you change your opinion about somebody.

This content downloaded from 168.195.52.137 on Mon, 11 Sep 2023 03:51:20 +00:00
All use subject to https://about.jstor.org/terms
Florentina Nicolás-Conesa et al. 871
APPENDIX C

Examples of Error Coding


Broad and Narrow Coding of Grammar and Nongrammar Errors

Broad Coding Narrow Coding Explanation Examples


Grammar Verb tense Verb tense is incorrect My face LOOKS so sad ... ("LOOKED")
errors
Verb form Verb form is incorrect . . . started TO GIVE . . . ("GIVING")
Word form Word form is incorrect . . . three WORSE months . . . ("WORST")
including wrong . . . my STRENGTHS decreased . . .
genitive use ("STRENGTH")
... to my beach'S house ... ("BEACH")
I remember OTHER day . . . ("ANOTHER")
Article Article is missing, . . . but THE most of them . . . (unnecessary
unnecessary, or article)
incorrect ... in A near future . . . (wrong article, "the")
... XXX teachers couldn't ... (missing article,
"the")
Plural Noun plural marker is ... all THIS things ("THESE")
missing, unnecessary, ... IRRELEVANTS things ("IRRELEVANT")
or incorrect . . . people that IS important ("ARE")
...at NIGHTS ("NIGHT")
Preposition Preposition is missing, . . . called TO my uncle . . . (unnecessary
unnecessary, or preposition)
incorrect . . . ON the third year . . . (wrong preposition,
"in")
. . . playing XXX one hour . . . (missing
preposition, "for")
Word order Word order incorrect . . . the doctors found IN MY LEG a bad thing . . .
for the sentence (" . . . a bad thing IN MY LEG")
(except when We spent THERE five days . . . ("five days
affecting sentence THERE")
structure)
Sentence Sentence structure I did THE BETTER THAT . . . (instead of, e.g., "I
structure error did my best")
... despite of HIM ILL ... (instead of, e.g., "his
being ill" or "the fact that he was ill")
Pronoun Pronoun is missing, . . . what you know WHAT is the . . . (unnecessary
unnecessary or pronoun)
incorrect . . . didn't say NOTHING to us (wrong
pronoun, "ANYTHING")
... XXX is hard (missing pronoun, "IT")
Conjunction Conjunction is missing, . . . AND unfortunately . . . (unnecessary
unnecessary or conjunction)
incorrect . . . there was a time THAT . . . (wrong
conjunction, "WHEN")
. . . family, XXX my friends . . . (missing
conjunction, "AND")
Determiner Determiner is missing, ... was under MY control (unnecessary
unnecessary or determiner)
incorrect . . . for THIS reason (wrong determiner,
"THAT")
. . . had XXX money (missing determiner,
"SOME")

This content downloaded from 168.195.52.137 on Mon, 11 Sep 2023 03:51:20 +00:00
All use subject to https://about.jstor.org/terms
872 The Modern Language Journal 103 ( 201 9)
Broad Coding Narrow Coding Explanation Examples
Nongrammar Wrong word (Lexis) Wrong word/expression . . . deep REMEMBERS . . .
errors Word meaning incorrect for ( "MEMORIES" )
the sentence or necessary . . . that WRONG experience . . .
word is missing ("NEGATIVE")
. . . start BY ZERO . . . ("FROM
SCRATCH")
. . . the TRAVEL of my dreams . . .
("TRIP")
Word choice (Lexis) Word choice not exactly ... be able to KNOW anything . . .
wrong but could be clearer ("UNDERSTAND")
or more appropriate ... a BIG effort . . . ("GREAT")
I didn't improve ANYTHING . . . ("AT
ALL")
Spelling Spelling or spacing error ... an EPILECTIC . . . ("EPILEPTIC")
A owl appeared . . . ("AN")
Punctuation Commas, full stops, ... to suffer bullying, every day . . .
semi-colons, capital (should be full stop, not comma)
letters ... a machine, which . . . (unnecessary
comma)
. . . today; Now . . . (no capital N in
"Now")
Apostrophe (Other Incorrect or missing . . . our parent's permission . . .
errors) apostrophe ("PARENTS' permission")
Referent (Other Pronoun reference is vague I don't like IT ... (it's
errors) or unclear unknown/unclear what "it" refers to)
Oral language (Other Language more appropriate WELL, I didn't sleep . . .
errors) to oral speech than formal PLUS, you couldn't choose . . .
writing

errors) ("because he started")


Redundant (Other Redundant language . . . because OF he started . . .

... we had a TRAFFIC accident ("an


accident")
When I was a little girl, WHEN I WAS 7
years old ... ("When I was 7 years
old...")

APPENDIX D

Examples of Students' Languaging and Coding

Error Correction Error Type Students' Languaging Coding


It sounded too TRUE Word choice REAL. I don't know Level 1: Level of
good to be REAL what is wrong here noticing (error
transcription alone)

York "arrived in" reporting (error


I CAME to New ARRIVED IN Wrong word CAME; I should write Level 2: Level of

transcription +
error correction)

This content downloaded from 168.195.52.137 on Mon, 11 Sep 2023 03:51:20 +00:00
All use subject to https://about.jstor.org/terms
Florentina Nicolás-Conesa et al 873
Error Correction Error Type Students' Languaging Coding
I NEVER WAS with I WAS NEVER Word order I NEVER WAS with her. Level 2: Level of
her The order of the reporting (error
sentence is wrong. transcription + error
category)
After THIS THESE Determiner I made mistakes in the Level 3. Level of
situations SITUATIONS use of determiners: I reporting (error
wrote "this" instead of transcription + error
"these." correction + error
category)
The two girls I WAS I HAD BEEN Tense The two girls I WAS with Level 4. Level of
with that morning that morning. Wrong understanding (error
verb tense because it is transcription^- error
an event that category + MLE)
happened in the past.
I should have used
past perfect.
I thought about my TO SIT Verb I thought about my final Level 4. Level of
final option, which option , which WAS understanding (error
was SITTING in a SITTING in a park: I transcription + error
park should have written TO
correction
SIT after the verb
+ MLE)
to be.

She has been my FOR 11 years Preposition SINCE 11 years: The Level 5. Level of
best friend SINCE preposition is wrong. understanding (error
I I years The correct transcription + error
preposition is FOR, correction + error
because it refers to the category + MLE)
duration of the action.

Note. MLE = metalinguistic explanation.

SUPPORTING INFORMATION

Additional supporting information may be found online in the Supporting Information section at the
end of the article.

This content downloaded from 168.195.52.137 on Mon, 11 Sep 2023 03:51:20 +00:00
All use subject to https://about.jstor.org/terms

You might also like