Open Journal of Modern Linguistics
2012. Vol.2, No.2, 57-70
Published Online June 2012 in SciRes (http://www.SciRP.org/journal/ojml) http://dx.doi.org/10.4236/ojml.2012.22008
Copyright © 2012 SciRes. 57
Corrective Feedback, Negotiation of Meaning and Grammar
Development: Learner-Learner and Learner-Native Speaker
Interaction in ESL
Satomi Kawaguchi, Yuan Ma
Humanities and Communication Arts, University of Western Sydney, Sydney, Australia
Email: {s.kawaguchi, Ningyue.Ma}@uws.edu.au
Received February 24th, 2012; revised March 24th, 2012; accepted April 2nd, 2012
This study aims to investigate the role of corrective feedback and negotiation of meaning within an Inter-
actionist Approach (Long, 1996) in native speaker-Second Language learner and L2 learner-L2 learner
interactions. While negotiation of meaning (NoM) and corrective feedback (CF) between native and non-
native speakers has been shown to be helpful for the nonnatives, it remains unclear whether CF and NoM
between learners of equivalent or different proficiency produce greater negotiation of meaning and suc-
cessful uptake of corrective feedback compared to the more traditional native-nonnative interaction. The
key issue in this study is whether CF and NoM in different interactional combinations of interlocutors
make a difference, in quantitative and qualitative terms. The study adopts a pretest-treatment-posttest de-
sign with six participants: two native English speakers, two Chinese L1 NNSs of high English proficiency
level (NNS High) and two Chinese L1 NNSs of low English proficiency level (NNS Low). These infor-
mants generated 14 different dyads and produced 2377 turns while engaging in task-based interaction. By
introducing the notions of group (i.e., NS-NNS versus NNS-NNS groups), combinations (e.g., NS-NNS
High versus NNS High-NNS Low), and dyads, it is possible to compare results across groups, combina-
tions and individuals. Results confirm that CF and NoM happen in NNS-NNS interaction yet they differ,
qualitatively and quantitatively, according to the type of combination. Significantly, the best rate of suc-
cess was obtained in the combination of learners with different proficiency levels i.e., the NNS High-NNS
Low combination. In addition, error rates decreased from pre-test to post-test in all learners, especially
NNS Low, which lends support to the notion that CF and NoM promote second language development
also in interaction between learners.
Keywords: Corrective Feedback; Negotiation of Meaning; Grammar Learning; ESL; Interaction
Introduction
This study investigates corrective feedback (CF) and nego-
tiation of meaning (NoM) occurring in the interaction in Eng-
lish between: 1) native speaker (NS) and second language (L2)
learner; 2) two L2 learners of the same proficiency level and; 3)
two L2 learners of different proficiency levels. The current
study is conducted within the interaction approach (Long 1996,
Gass & Mackey, 2007) and it aims to examine whether there
are qualitative and/or quantitative differences among interac-
tions involving speakers with different language proficiency.
The study further investigates whether negotiated interactions
promote grammatical development in L2 learners. For the last
two decades in the field of second language learning and
teaching, the role of interaction in language acquisition has
been researched extensively (e.g., Gass, 1997; Long, 1996).
Many studies that analyzed L2 learner’s interaction with native
speakers found that: CF that L2 learner receives and NoM that
L2 learner participates in help the learner improve his/her L2
language skills (Oliver, 1995; Pica, 1994; Sheen, 2004). Fewer
research works examined interactions between nonnative
speakers (NNS) (Chiba, 2010; Mackey, Oliver, & Leeman,
2003; Varonis & Gass, 1985) while no previous studies inves-
tigated the effects of negotiation of meaning between L2 learn-
ers of different proficiency levels. Also, many studies have
reported that the majority of CF/NoM in L2 learners’ interac-
tion relates to meaning consultation, rather than L2 learners’
errors, (e.g. Iwasaki & Oliver, 2003). This poses the question as
to whether CF promotes L2 learner’s grammar learning. Only a
handful of studies (e.g., Mackey, 1999) broached this issue and
the answer to this question remains unclear.
Communicative language teaching, which creates ample op-
portunities for a learner-centered interaction, is the most widely
used teaching method since the 1980s (Richards & Rodgers,
2001). This method has the advantage of providing ranging
opportunities for L2 learners to interact with their peers and
their teachers by using communicative language tasks and con-
sequently, promoting effective language acquisition (Finocchi-
aro & Brumfit, 1983). It is important to explore the potential for
learning and development among learners themselves, particu-
larly since most class time in communicative teaching/learning
environments is devoted to learner-learner activities such as
role play, pair work, simulation, language games and interac-
tional conversation (Harmer, 2007).
In this study, a total of six adult informants were recruited:
two NSs of English and four ESL learners of Chinese first lan-
guage background. Among the four ESL learners, two are at
high proficiency level (International English Language Testing
System (IELTS) 7.0) and the other two are at low level (IELTS
S. KAWAGUCHI, Y. MA
4.5). These participants form 14 interaction dyads. CF and
NoM will be examined in detail for learner-learner interaction
focusing on selected grammatical items, and then their learning
effect will be compared with learner-NS interaction. In order to
examine the effectiveness of CF and NoM, four English gram-
matical phenomena are examined in this study. They are: 1)
verb past tense, 2) noun plural form, 3) subject-verb agreement
and 4) question constructions. Unlike other studies which tend
to present aggregated results as a group (such as NS, NNS),
detailed analysis of both group-wise and as an individual results
will be shown. The research questions motivating this study
are:
1) Does CF and NoM occur in learner-learner interaction? If
so, are they comparable to native speaker (NS)-learner interac-
tion?
2) Does CF and NoM differ qualitatively and quantitatively
in different combination of interactions involving NS and dif-
ferent levels of ESL learners?
3) Can negotiated interaction promote grammar development
in ESL learners?
The remainder of this paper is organized as follows. In the
next section, we present the theoretical background and litera-
ture review followed by the methodology in this study. Next,
results and discussion are presented. The last section presents
our conclusion.
Theoretical Background
Input, Interaction and Output
In second language acquisition (SLA), input, interaction and
output are regarded as the three important elements of success-
ful L2 learning (Long, 1996; Gass, 2003; Swain, 1995). In the
following, we will address these key notions which are most
relevant to our study.
Input. Input plays an important and necessary role in lan-
guage acquisition because it provides language-specific infor-
mation for learners to interact. Many researchers investigate
input from different point of views. For instance, Ellis (2002)
emphasises the importance of frequency in SLA. Gass (2003),
on the other hand, believes that input provides positive evi-
dence of what is possible within a language. Positive evidence
attained from input is able to help L2 learners with their lan-
guage acquisition. In addition, the negative evidence obtained
through interaction, a type of information given to L2 learners
about their incorrect use of an utterance either in explicit form
or in implicit form, is also believed to be conducive to L2
learning. On the other hand, universal grammar (UG) approach
looks at input from yet a different perspective. Schwartz (1993),
for example, argues that input, as a secondary role, interacts
with an innate system affecting acquisition. We adopt Gass’
(2003) view in this study. Input can be regarded as the first step
to learn a second language. It provides positive evidence for
learners to notice and interact. Positive evidence, which hap-
pens in the input process, consists of a set of well-formed
sentences that learners are exposed to. It further provides
opportunities for learners to monitor their own utterances or
written words and consequently modify their input informa-
tion.
Interaction. Long’s (1996) updated interaction hypothesis
suggests that “negotiation of meaning, and especially negotia-
tion work that triggers interactional adjustments by the NS or
more competent interlocutor, facilitates acquisition because it
connects input, internal learner capacities, particularly selective
attention, and output in productive ways (pp. 451-452).” Thus,
negotiated interaction directs learners’ attention to their non-
target like production or problematic expressions. They may
notice the types and amount of differences between the lan-
guages they used and the language a native speaker used. They
may also notice their gap between their intended content of the
speech and their ability in expressing it, due to their lack of
language knowledge. The interaction could also draw learners’
attention to something new, such as a new vocabulary, gram-
matical item, or syntax.
Now we look at CF which helps language learner to acquire
the language they are learning through interaction. CF is a
form-focused instruction that gives to the SL learners on their
non-target like form either in speech or in writing (Li, 2010).
The purpose of providing such corrective instruction is to draw
L2 learners’ attention to their non-target like production and
assist their L2 learning. Long (1996) defines that CF is com-
posed of two categories. One category is explicit CF: an overt
correction and metalinguistic explanation given to the learner
about his/her wrong output. The other category is implicit CF,
which can be divided into two components, recast and NoM.
Recast refers to the target-like reformulations provided to a
learner’s incorrect utterance. For instance, from our data,
NNS’s ungrammatical utterance, “..two floors is enough” re-
ceived recast, “two floors are enough”.
NoM is the strategy that language users employ for commu-
nication until successful comprehension is achieved. The
strategies for NoM include clarification requests, confirmation
checks, comprehension checks, and repetition (Oliver, 1995).
Clarification requests are the expressions used to ask L2 learn-
ers to clarify their preceding utterances by asking a question
such as, “what does it mean?”. Confirmation checks refer to the
expressions used by the listener to make sure that an utterance
he/she heard or understood is correct. For example, a speaker
might ask “do you mean …?” to confirm his/her correct under-
standing of the interlocutor’s utterance. Comprehension checks
are utterances used to check whether an interlocutor has under-
stood. “Do you understand?” is an expression used for com-
prehension check. Repetitions include both the speaker and
the interlocutor’s exact repetitions of lexical items from their
own preceding utterances within five speaking turns and they
also include repetition of the other’s wrong utterances (Pica &
Doughty, 1985). Figure 1 summarises each type of CF and
also gives a clear view of the relationship between CF and
NoM.
Figure 1.
CF and NoM (based on Long, 1996).
Copyright © 2012 SciRes.
58
S. KAWAGUCHI, Y. MA
Output. Output, either in speaking or writing, is claimed to
push L2 learners to think about the syntactic use of language
(Swain, 1995). Output, thus, has a potentially significant role in
the development of syntax and morphology due to its require-
ment for learners to produce more comprehensible and accurate
output. As a way for learners to encourage more accurate pro-
duction, output forces L2 learners to contemplate with more
effort on their own problematic utterances during interaction.
Thus, input, interaction and output are three interrelated steps
for successful L2 learning.
Previous Studies
Interactionists believe that negotiated interaction can provoke
L2 learners’ attention to non-target like forms, including vo-
cabulary, morphology, or syntax, and thus promote SLA (Lys-
ter & Ranta, 1997; Mackey, 1999). Li’s (2010) meta-analysis
has further confirmed the effectiveness of CF in SLA. Through
the examination of 33 studies on the effects of CF, he found a
“medium overall effect for corrective feedback and the effect
was maintained over time (p. 309).”
In addition, many studies compare the effects of different
types of CF received by learners (e.g., Ellis, Loewen, & Erlam,
2006; Panova & Lyster, 2002; Iwashita, 2003; Philp, 2003).
Some studies suggest that learners could benefit more from
explicit rather than implicit CF (e.g., Carroll & Swain, 1993;
Ellis, Loewen, & Erlam, 2006; Kim & Mathes, 2001; Varnos-
fadrani & Basturkmen, 2009). Moreover, among the different
types of implicit CF, Lyster (1998, 2004) mentions that recasts,
occurring after grammatical errors, assist L2 learning. This
claim is further supported by studies, such as Long, Inagaki, &
Ortega (1998), Braidi (2002), Hauser (2005), and Nassaji
(2009). Besides classroom interaction, some recent studies have
investigated technology-oriented interaction between native
speakers and L2 learners, such as CF and NoM through eTan-
dem interaction (Bower & Kawaguchi, 2011; Iwasaki & Oliver,
2003), CF via instant messenger (Sotillo, 2005) and NoM
through synchronous text chat and videoconferencing (Zhao &
Angelova, 2010). All findings support the contention that CF
and NoM are crucial for L2 learners.
Moreover, many researchers have investigated the effective-
ness of negotiated interaction between adult NS and non-native
speakers (NNS) (e.g., Brock, Crookes, Day, & Long, 1986;
Nassaji, 2007; Pica, Young, & Doughty, 1987). Oliver (1998)
explores the effects of peer learning in school aged-children
involving NS-NNS and NNS-NNS combinations. She found
that similar to adults, children used variety of negotiation
strategies during their interaction which could facilitate L2
learning, while other studies (e.g., Lyster & Ranta, 1997;
Mackey & Oliver, 2002; Mackey, Oliver, & Leeman, 2003;
Oliver, 2002) examine the interaction between adult NS and
child NNS. Many more studies explore the interaction in
learner-learner combination. For instance, Chiba (2010) inves-
tigated whether NoM differ quantitatively or qualitatively be-
tween two advanced NNSs (NNS-NNS) and a native speaker
with advanced NNS (NS-NNS), interacting in Japanese. She
found that NNS-NNS combination provided more opportunities
of CF and NoM in comparison to NS-NNS combination. Other
studies, such as Mackey, Oliver, & Leeman (2003) and Varonis
& Gass (1985) show positive outcome in adult peer interaction.
However, unlike our study, all adult learners recruited were
from the same L2 language proficiency levels.
Furthermore, some studies explore the link between negoti-
ated interaction and grammatical development. Mackey (1999)
claims that conversational interaction facilitates L2 learners
with their acquisition of English question formations. Ellis,
Loewen & Erlam (2006) found that explicit feedback was more
effective than implicit feedback on the acquisition of English
past tense -ed. Yang & Lyster (2010) explore the effects of
form-focused practice and feedback on acquisition of English
regular and irregular past tense forms and found that prompts
were effective in increasing accuracy. Iwashita (2003) exam-
ines the role of task-based interaction in Japanese L2 gram-
matical development. She concluded that implicit negative
feedback had a beneficial effect on the short-term development
of target grammatical structures but recasts had a larger impact.
Despite the large number of studies in CF and NoM as re-
viewed above, there is no study to date that looks at qualitative
and quantitative differences of interaction involving different
proficiency levels of the L2 learner. Our study aims to fill this
gap.
Methods
In order to address the three research questions presented
above, we conduct a study involving six participants: two na-
tive speakers (NS) of English, and four nonnative speakers
(NNS) of Chinese L1 background (two high and two low profi-
ciency levels) forming 14 dyads (see below). Three communi-
cative tasks are used to elicit interactional conversation to ex-
plore CF and NoM occurring in different dyads: a board game,
interactional conversation and picture describing and drawing.
The grammatical items focused on in this study are: noun plural
forms, verb past tense, subject-verb agreement, and question
formation. During their interaction, learners are instructed to
provide feedback to their partners, especially on errors. More
detailed instructions given to the learners are presented under
Tasks below. All conversations are audio-recorded and tran-
scribed for further analysis.
Participants. All six participants, all female, are students at
the University of Western Sydney, Australia. Two participants
are native speakers of English respectively undertaking Master
of Social Science and Master of Teaching. The other four par-
ticipants are nonnative English, Chinese L1 background so that
issues of first language influences are controlled. Of the four
NNSs, two of them are of high proficiency level, abbreviated as
NNS (High), undertaking Master of Interpreting and Transla-
tion, which requires an IELTS score of 7.0 for UWS entry. The
other two are NNS of lower proficiency level, abbreviated as
NNS (Low), currently enrolled in English for Academic Prepa-
ration course level 1, which requires an IELTS score of 4.5 for
enrolment. Their age range was from 18 to 28 years. The six
participants in this study are represented as NS1, NS2, NNS
(High)1, NNS (High)2, NNS (Low)1 and NNS (Low)2.
Groups, Combinations and Dyads
The notions of groups, combinations and dyads are used for
data analysis in order to investigate performances relating to CF
and NoM both as a group and individual. Most studies on CF
and NoM present aggregate results as groups (such as NS, NNS,
etc). By utilising the notion of groups, combinations and dyads,
our study is able to show results by individuals as well as
groups.
Copyright © 2012 SciRes. 59
S. KAWAGUCHI, Y. MA
Groups. The six participants are divided into two different
types of groups in terms of interaction: 1) NS-NNS and 2)
NNS-NNS. The purpose of this division is to answer the first
research question: whether CF and NoM occur in learner-
learner group.
Interaction Combinations. The six participants form five
different interaction combinations: 1) NS-NNS (Low); 2) NS-
NNS (High); 3) NNS (Low)-NNS (High); 4) NNS (Low)-NNS
(Low) and; 5) NNS (High)-NNS (High). The purpose of exam-
ining different combinations is to investigate the extent and the
variety of CF provided, whether the participants can notice their
non-target like utterance, and whether the interaction can pro-
mote grammar learning according to different combinations.
Furthermore, the purpose of NNS (Low)-NNS (High) combina-
tion is to investigate how much the NNS (Low) and the NNS
(High) notice their own and their partners’ non-target like ut-
terances and whether this type of combination assists their
grammar learning when they are grouped with a partner with
different L2 proficiency levels. The last two combinations,
NNS (Low)-NNS (Low) and NNS (High)-NNS (High), have
similar purpose to the NNS (Low)-NNS (High) combination,
but in this case each participant is grouped with a partner from
the same L2 proficiency level.
Dyads. The six participants from 14 different dyads. The
purpose of using dyads is to observe participant’s linguistic
behaviour individually. By analysing negotiated interaction
happening in each dyad, we intend to determine the range of
grammatical errors participants make, and the range of negotia-
tion strategies they would adopt for their partners. Table 1
illustrates the relationship of groups, combinations and dyads.
Interaction Schedule. Interactions with 14 dyads are con-
ducted over three weeks. In the first week, the learners in the
same proficiency level interact with each other, i.e., NNS (Low)
1-NNS (Low)2 and NNS (High)1-NNS (High)2. Over the fol-
lowing two weeks, interaction sessions of remaining 12 dyads
are organized. For these weeks, four learners will interact with
either NS or other learners of different proficiency levels. Every
learner engaged in five dyads during this period. For example,
NNS (Low)1 participated in: 1) NNS (Low)1-NNS (Low)2, 2)
NS1-NNS (Low)1, 3) NS2-NNS (Low)1, 4) NNS (Low)1-NNS
(High)1 and 5) NNS (Low)1-NNS (High)2. The data gained
Table 1.
The relationship of groups, combinations and dyads.
Group Combination Dyad
NS1-NNSL1, NS1-NNSL2
NS-NNS (Low)
NS2-NNSL1, NS2-NNSL2
NS1-NNSH1, NS1-NNSH2
NS-NNS
NS-NNS (High)
NS2-NNSH1, NS2-NNSH2
NNSL1-NNSH1
NNSL1-NNSH2
NNSL2-NNSH1
NNS (Low)-NNS (High)
NNSL2-NNSH2
NNS (Low)-NNS (Low)NNSL1-NNSL2
NNS-NNS
NNS (High)-NNS (High) NNSH1-NNSH2
from interaction with these 14 dyads over the three weeks will
be utilised to analyse CF and NoM according to different
groups/combinations.
Pre-test, Treatments and Post-test Design. In order to ex-
plore whether negotiated interaction promotes grammatical
development of ESL learners, this study adopts a pre-test,
treatment and post-test design. The data collected in the first
week will be regarded as the baseline of the four learners and
will function as a pre-test. Once all dyads complete their inter-
actions in week 2 - 3 (these are the “treatment” sessions), a
post-test will be conducted in week 4. The post-test involves
learners of the same language proficiency level performing the
interaction once again. Taking NNS (Low)1 as an example, she
is asked to interact with NNS (Low)2 for the first time in week
1, and then she is involved in four sessions with two NNS
(High)s and two NSs respectively in week 2 - 3. In week 4, she
participates in the post-test with NNS (Low)2. Her grammatical
error rates from both pre-test, treatment and post-test will be
compared to examine whether there is any improvement.
Tasks
Three tasks are utilized to create contexts for interaction be-
tween interlocutors. They are: 1) board game; 2) an interac-
tional conversation and; 3) describing and drawing pictures.
During interaction activities, every participant is given a new
topic to ask questions about and a new picture to describe.
Hence, none of the participants are exposed to the same topic or
picture twice. Up to 45 minutes are required for each of the
dyads to complete the three tasks.
Board game. The board game (Appendix A) aims to elicit
past tense marking in the verb. This game follows a common
board game procedure where two participants move a counter
from the “START” to the “FINISH” by throwing a dice to de-
cide how many spaces one can move. Between the “START”
and “FINISH”, a past-tense question is listed in each space.
Each participant is required to read out the question on which
her counter landed and answers it by using past tense forms.
The participant has to provide an appropriate past tense form of
the verb to answer the question, as the auxiliary verb in English
question carries the tense and the verb that needs to be in the
infinite form (e.g. What did you watch on TV yesterday?). The
partner’s duty is to provide CF when the interlocutor fails to
use the correct past tense. They continue to move forward in
turns, until either one reaches the “FINISH”. There are 28
questions on the board game and each participant answers ap-
proximately 10 questions.
Interactional conversation. This task aims to elicit all
grammatical items focused on this study, e.g., verb past tense,
noun plural forms, S-V agreement and question constructions.
In this task, one participant in the pair chooses a topic from 40
options and asks questions to the other. Some of the topics are:
“old school days” and “your best friend”. Participants are in-
structed to ask at least 10 questions to each other, and discover
as much information from her partner in the designated three
minutes. During the conversation, participants are instructed to
identify the errors that their partners made, and provide CF.
Describing and drawing pictures. The third task involves
describing and drawing pictures (Appendix B provides an ex-
ample). Similar tasks have been used and tested in a number of
research studies (Mackey, 1999; Oliver, 1995). In this study,
one of the participants in the dyad is asked to describe the pic-
Copyright © 2012 SciRes.
60
S. KAWAGUCHI, Y. MA
ture as clearly as possible, and the partner draws it, based on the
verbal information provided. During the activity, they are also
instructed to identify their partners’ errors, and provide feed-
back when necessary. This task mainly elicits the grammatical
items of subject-verb agreement, noun plural forms and ques-
tion formations.
Data Analysis Procedure
Analysis of NoM and CF for NS-NNS and NNS-NNS
groups. In order to address the first research question, this
study analyses turn takings that happened between each NS-
NNS and NNS-NNS groups. The definition of a turn in this
study follows that provided by Di Biase (2000): an utterance
made by a speaker before the other participant in the interaction
produces a response. The data analysis includes total number of
turns, total number of NNS turns, total number of NNS tar-
get-like turns and non-target like turns, code-switching turns,
turn counts on CF and NoM, and the number of explicit CF and
recasts.
To analyse turns that involve NoM, Varonis and Gass’ (1985)
model of NoM is adopted, which has also been used in other
studies investigating similar issues (e.g., Blake, 2000; Lee,
2006; O’Rourke, 2005). According to their model, instances of
NoM consist of up to four consecutive turns as can be appreci-
ated in Table 2.
NoM typically involves a sequence of trigger-signal (of
communication problem)-response-reaction to response in the
interaction. An example of trigger, signal, response and reac-
tion to response from the NNS (Low)1-NNS (High)1 dyad is
provided in (1). As (1) of the current data shows, a NoM often
happens when the Hearer encounters communication problem.
The problematic utterance is called trigger and it often involves
utterances containing grammatical errors, incomprehensible or
ambiguous utterances. Once the Hearer notices these triggers,
he/she would use strategies to solve the communication break-
down. The strategies the Hearer uses are defined as signals.
Signals usually take the form of implicit CF, including NoM
and recasts. Signal then leads to a response from the Speaker.
The response can be a modified output or a target-like refor-
mulation or further explanation of an ambiguous utterance.
Finally, the “reaction to response” is observed, where the
Hearer indicates whether or not she has understood the “re-
sponse”.
1) Trigger (Speaker: NNS (High)1) How many garages do
you want?
Signal (Hearer: NNS (Low)1) Garage? What?
Response (Speaker: NNS (High)1) The room for you to put
the car in.
Reaction to response (Hearer: NNS (Low)1) Aha.. Maybe
three or four.
Analysis of NoM and CF Concerning Different Combina-
tions. The data are analyzed to address the second research
Table 2.
Varonis and Gass’ (1985) model of negotiation of meaning.
Trigger Resolution
Trigger (Utterance
involving problem(s)) Signal Response Reaction to response
Speaker Hearer Speaker Hearer
question asking whether there is any qualitative/quantitative
difference in different combinations of interactions. In order to
record the number of CF and NoM in each combination pre-
cisely, the NNS turns are divided into two categories: NNS
target like turns and NNS non-target like turns. This type of
division is based on the system utilized by Iwasaki & Oliver
(2003). The non-target like turns can further be divided into
two categories: non-target like turns related to NNS grammati-
cal errors, and turns unrelated to NNS grammatical errors.
Grammatical errors include verb tense, articles, subject-verb
agreement, plural form and question formations, etc. Non-target
like turns unrelated to grammatical errors mainly include pro-
nunciation, and lexical errors. In order to find the quantitative
differences according to different combinations, the number of
each combination’s CF and NoM is first compared. Then,
analysis comparing the number of CF related to grammar
among the six combinations is conducted.
As for the qualitative differences, distribution of each CF
(such as explicit CF, recast and NoM) provided to the inter-
locutor is examined. Further, in order to analyse which combi-
nation shows a better learning result, all the negotiation in-
stances will be classified into successful and unsuccessful ne-
gotiation instances. A successful negotiation instance indicates
that a participant notices the CF provided for her, and then she
produces modified output or acknowledges a correction. On the
contrary, if a participant ignores the CF or fails to notice modi-
fied output then the negotiation instance is considered to be an
unsuccessful one. Furthermore, if a participant produces a
wrong CF, it is also regarded as an unsuccessful negotiation
instance. Examples of successful and unsuccessful negotiations
from our data are offered in (2) and (3) respectively. In (2),
NNS (Low)1 noticed her error on the verb tense of buy and
corrected to bought after NNS (High)1’s signal. In contrast,
NNS (Low) failed to notice her grammatical error of there have
after she received a recast from NNS (High)1 saying there are
four trees in (3). She continued her interaction without suc-
cessfully modifying the output.
2) Trigger: NNS (Low)1 Three. What did you buy yesterday?
I buy bread. I buy ticket.
Signal: NNS (High)1 So you buy?
Response: NNS (Low)1 Err…no, I bought
Reaction to response: NNS (High)1 Yeah, good.
3) Trigger: NNS (Low)1 Hmm. In front, there have..in the
north, no, in the front there have four trees.
Signal: NNS (High)1 There are four trees.
Response: NNS (Low)1 Yes. And four children and one
woman sit on the floor. And there have a big cloth
Analysis of L2 grammar development through NoM and
CF. The error rate on each grammatical item in question (i.e.,
past tense, plural forms, S-V agreement and question construc-
tion) is calculated with pre-test, treatment and post-test for the
comparison.
Results
To answer Research Questions 1 and 2, that aim to investi-
gate differences/similarities according to the groups and com-
binations, all data except the data from the post-test are ana-
lysed both qualitatively and quantitatively. Meanwhile, to an-
swer Research Question 3, data from pre-and post-tests are
analysed to examine whether CF has an effect on grammar
learning.
Copyright © 2012 SciRes. 61
S. KAWAGUCHI, Y. MA
Copyright © 2012 SciRes.
62
Data Summary According to Two Groups
Our data according to two different groups, (i.e. NS-NNS
and NNS-NNS groups) are first summarised. The turn count
from each group are presented in Table 3. The average number
of turns per session (i.e., per dyad) is also listed, as numbers of
dyads involved in these two combinations are different. There
are eight dyads in NS-NNS group (i.e., NS1-NNS (High)1,
NS1-NNS (High)2, NS1-NNS (Low)1, NS1-NNS (Low)2,
NS2-NNS (High)1, NS2-NNS (High)2), NS2-NNS (Low)1 and
NS2-NNS (Low)2)) and six dyads in NNS-NNS group (i.e.,
NNS (High)1-NNS (High)2, NNS (Low)1-NNS (Low)2, NNS
(High)1-NNS (Low)1, NNS (High)1-NNS (Low)2, NNS (High)
2-NNS (Low)1 and NNS (High)2-NNS (Low)2)). The NS-NNS
group produced 1322 turns in total (165.3 turns per session)
while NNS-NNS group produced 1055 turns in total (175.8
turns per session). Hence, the NNS-NNS group produced
slightly more turns per session than the NS-NNS group to per-
form the same tasks.
Table 4 displays breakdown of NNS turns. In the NS-NNS
group, the total of 661 NNS turns are divided into 595 tar-
get-like turns (90%) and 66 non-target like turns (10%). On the
other hand, the total number of NNS target-like turns in NNS-
NNS group is 875 (83%) and the number of non-target like
turns is 136 (13%). Code-switching occurred only with NNS-
NNS group, with a percentage of 4%.
Table 5 presents the frequency of CF happened during in-
teraction by group. In the NS-NNS group, there are 21 in-
stances of CF on grammar error (session average 2.6), 17 in-
stances of other errors (session average 2.1) and 51 instances of
meaning consultations (session average 6.4). On the other hand,
in the NNS-NNS group, CF on grammar error, on other errors
and on meaning consultations are 34 (session average 5.7), 10,
(session average 1.7) and 52 (session average 8.7) respectively.
The NNS-NNS group interaction produced therefore more CF
instances than the NS-NNS group.
In order to check whether there are any differences in inter-
action groups, the total number of turns devoted to explicit CF,
implicit CF and unrelated to CF are quantified and displayed in
Table 6. As for the NS-NNS group, 94 out of 1322 turns are
devoted to recasts (session average 11.8 turns) and 257 turns to
NoM (session average 32.1 turns). Interstingly, no turns were
devoted to explicit CF. The NNS-NNS, on the other hand, de-
voted 3 turns to explicit CF (session average 0.5 turns), 99 turns
to recasts (session average 16.5 turns) and 284 turns to NoM
(session average 47.3 turns). All of them are higher than those
of NS-NNS group. Table 6 also presents the turn count of CF
converted to every 100 turns. This conversion is necessary for
Table 3.
Summary turn counts of each group.
Group NS turns NNS turns Total
8 dyads total 661 (50%) 661 (50%) 1322 (100%)
NS-NNS
Session average82.6 82.6 165.3
6 dyads total N/A 1055 (100%) 1055 (100%)
NNS-NNS
Session averageN/A 175.8 175.8
Table 4.
Breakdown of NNS turns.
Group Target like turns Non-target like turns Code-switching turns Total
8 dyads total 595 (90%) 66 (10%) 0 (0%) 661 (100%)
NS-NNS
Session average 74.4 8.3 0 82.6
6 dyads total 875 (83%) 136 (13%) 44 (4%) 1055 (100%)
NNS-NNS
Session average 145.8 22.7 7.3 175.8
Table 5.
CF and NoM on grammar errors, other errors and meaning consultations.
Groups On grammar errors On other errors On meaning consultations Total
Explicit CF 0 0 0 0
Recast 14 10 0 24
NoM 7 7 51 65
NS-NNS (8 dyads)
Subtotal (SA) 21 (2.6) 17 (2.1) 51 (6.4) 89 (11.1)
Explicit CF 1 0 0 1
Recast 20 4 0 24
NoM 13 6 52 71
NNS-NNS (6 dyads)
Subtotal (SA) 34 (5.7) 10 (1.7) 52 (8.7) 96 (16.0)
Total of 14 dyads (SA) 55 (3.9) 27 (1.9) 103 (7.4) 185 (13.2)
SA = Session Average.
S. KAWAGUCHI, Y. MA
Table 6.
Turns devoted to explicit and implicit CF and NoM.
Implicit CF
Group Explicit CF
Recasts NoM
Unrelated to CF Total Turns
8 dyads total 0 94 257 971 1322
Session average 0 11.8 32.1 121.4 165.3
NS-NNS
Per 100 turns 0 7.11 19.44 73.5 100
6 dyads total 3 99 284 669 1055
Session average 0.5 16.5 47.3 111.5 175.8
NNS-NNS
Per 100 turns 0.3 9.38 26.92 63.4 100
running a chi-square test to examine whether or not there is any
statistically significant difference between the NS-NNS and the
NNS-NNS groups. As there was no instance of explicit CF with
NS-NNS group and only one instance with NNS-NNS group,
we do not include explicit CF for the statistical analysis: we test
only implicit CF. In other words, this analysis examines
whether there is any statistical difference between the L2 learn-
ers’ interaction with NS and their interaction with other NNS in
terms of their opportunities for recast and NoM. The results
from the chi-square test indicated that there was no significant
difference between the two groups (Yate’s p-value = 0.8357, χ2
= 0.043. df = 1). This suggests that L2 learners experience simi-
lar opportunities for CF/NoM when they interact with other L2
learners as when they interact with NSs.
Data Summary Concerning the Five Combinations
This subsection presents quantitative and qualitative analyses
of the five different combinations. For the quantitative analysis,
the number of CF items recorded from each combination is first
examined. Second, the number of CF related to NNS errors
(both grammar errors and other errors) and the number of CF
unrelated to NNS errors, i.e., meaning consultations, is com-
pared. For the qualitative analysis, who initiated CF and NoM
in different combinations is further examined. The qualitative
outcomes of CF and NoM, namely whether negotiated items are
successfully taken up by the interlocutor or ignored (i.e. unsuc-
cessful) is also analysed.
Quantitative Analysis. Table 7 summarises the number of
each CF item recorded for the five different combinations. Note
that NNS (Low)-NNS (Low) and NNS (High)-NNS (High)
involve one dyad each, while all other combinations involve
four dyads, constituting 14 dyads all together. In order to com-
pare across the combinations, number counts per dyad are also
provided. When the proficiency levels of the learners in an
interaction are the same, more CF and NoM are observed with
NNS (Low)-NNS (Low) (16 times) than NNS (High)-NNS
(High) combinations (8 times). As for the interactions involving
participants of different proficiency levels, NNS (Low)-NNS
(High) scored highest with both recasts (5.5) and NoM (12.3)
per dyad. Also, total number of CF was highest with NNS
(Low)-NNS (High) per dyad (18.0) among the five combina-
tions. In order to address whether NNS received CF when they
produced errors, CF on errors were specifically examined.
In Table 8, the number of all CF, including recast and NoM,
received on errors and the total number of errors are listed. The
number in the bracket indicates the percentage of errors that
received CF. The percentages enable the comparison of NNS’s
opportunities of receiving CF across different combinations
regardless of the number of dyads involved in each combination.
For example, in NNS (Low)-NNS (Low) combination, there
were 58 grammatical errors but only three errors received CF
(5% of total number of errors). As for NNS (High)-NNS (High)
and combinations involving NS, i.e., NS-NNS (High) and NS-
NNS (Low), rates of CF on NNS errors unrelated to grammar are
much higher than CF related to grammar. For example in NS-
NNS (High) combination, 38% of grammatical errors received
CF but it was 56% with errors unrelated to grammar. In contrast,
in NNS (Low)-NNS (High), NNS (High) provided very high rate
of CF on NNS (Low)’s grammatical error (i.e., 64%). Among all
combinations, NNS (Low)-NNS (High) is the only combination
where the rate of CF on grammatical error (64%) is much
higher than CF on errors unrelated to grammar (36%). However,
NNS (Low) in the same combination did not provide CF at all
on NNS (High)’s grammatical errors and other errors.
Table 7.
Corrective feedback: five different combinations.
Implicit CF
Combinations Explicit CF
Recasts NoM Total
NNSL-NNSL (1 dyad)0 2 14 16
NNSH-NNSH (1 dyad)0 0 8 8
NS-NNSL (4 dyads) 0 (0) 16 (4.0) 35 (8.8)51 (12.8)
NS-NNSH (4 dyads) 0 (0) 8 (2.0) 30 (7.5)38 (9.5)
NNSL-NNSH (4 dyads)1 (0.3) 22 (5.5) 49 (12.3)72 (18.0)
Numbers in the bracket indicate average frequency counts per dyad.
Table 8.
NNS’s errors and CF on errors.
Combinations Grammar errors Other errors
NNSL-NNSL 3/58 (5%) 1/14 (7%)
NNSH-NNSH 2/11 (18%) 1/8 (13%)
NS-NNSL 12/32 (38%) 12/29 (41%)
NS-NNSH 6/16 (38%) 5/9 (56 %)
NNSL-NNSH
(NNSL’s errors corrected by NNSH) 29/45 (64%) 8/22 (36%)
(NNSH’s errors corrected by NNSL) 0/13 (0%) 0/3 (0%)
Number of CF/number of errors (rate of CF on errors).
Copyright © 2012 SciRes. 63
S. KAWAGUCHI, Y. MA
Qualitative analysis. In order to examine qualitative differ-
ences in the combinations involving speakers of different profi-
ciency levels, we focused on: 1) who initiated recast and NoM
and; 2) outcomes of CF in five different combinations.
Figures 2(a) and (b) show clear differences as to who initi-
ated recast and NoM depending on the proficiency level of the
interlocutor. It is interesting to observe that the speaker of
higher proficiency level in all combinations consistently show
higher frequencies of initiating recasts than the speakers of
lower level while this observation is reversed with NoM. For
example, in NNS (Low)-NNS (High) interaction, each partici-
pant initiated recast 0 and 22 time respectively. In the same
interactional combination, NNS (Low) and NNS (High) initi-
ated NoM 29 and 20 times respectively. When examining the
obtained data, the less proficient speaker in a combination ac-
tively asked questions to clarify the interlocutor’s utterances
while the more proficient speakers tried not to disturb the flow
of conversation. In order to attract the less proficient speaker’s
attention without disrupting the conversation flow, the more
proficient speaker were more likely to use recast as a strategy to
(a)
(b)
Figure 2.
(a) Initiation of recast; (b) Initiation of negotiation of
meaning.
correct the errors in the interlocutor’s utterance. In this way, the
more proficient speaker avoided imposing additional burden on
the less proficient speaker to clarify her ambiguity in speech.
This shows a clear qualitative difference in the initiation of CF
according to the level of proficiency of the speaker.
Table 9 presents outcomes of CF in the five combinations. It
is evident that NNS (Low)-NNS (High) combination has the
highest success rate (74%) followed by NS-NNS (Low) com-
bination (69%), and then NS-NNS (High) (66%), NNS (High)-
NNS (High) (50%). NNS (Low)-NNS (Low) combination fol-
lows well behind (25%). This suggests that NNS (Low) NNS
(Low) interaction may not provide an effective linguistic envi-
ronment for language learning.
Grammatical Development through Corrective
Feedback
This sub-section focuses on whether negotiated interaction
on errors leads to grammar learning in L2 learners. We showed
above that most CF involved in the interactional sessions were
on meaning consultation rather than on errors. One might ask
whether this means that CF on grammar errors plays a trivial
role for L2 learners in their grammar development. First we
examine L2 learners’ grammatical errors and CF on their errors
individually over pre-test and treatment period where all 14
dyads were activated. Note that every learner engaged in five
dyads during this period. In Table 10, the frequency counts and
percentages of all CF received on errors are summarised. We
found that the two NNS (Low) learners shared a similar pattern
and so did the two NNS (High) learners. Both NNS (Low)1 and
NNS (Low)2 received the highest rate of CF on errors when
they interacted with NNS (High) (70% and 62% respectively)
followed by NS (17% and 27% respectively). In contrast, they
received the lowest rate of CF from the other NNS (Low) (i.e.,
1% and 7% respectively). NNS (Low)1 and NNS (Low)2 re-
ceived CF at the rates of 32% and 34% respectively on errors in
total through five dyads. On the other hand, the two NNS (High)
learners made much smaller numbers of errors in the first place
as comparison to NNS (Low) learners. Both NNS (High) learn-
ers received CF on errors by 38% respectively from NS while
they received much less CF from the other NNS (High) (17%
and 11% respectively). It is noteworthy that both NNS (High)
received no CF from NNS (Low). Overall, the two NNS (High)
learners received 24% and 18%, respectively, of CF on errors
through the five dyads in which they participated. From this
analysis, it has become clear that NNS (Low) benefited the most
when interacting with NNS (High) while NNS (High) with NS.
Next we look at outcomes of CF on grammatical errors. Ac-
cording to Table 11, although frequency counts of CF on
Table 9.
Outcomes of CF of the five combinations.
Successful Unsuccessful Total
NNSL-NNSL (1 dyad) 4 (25%) 12 (75%) 16 (100%)
NNSH-NNSH (1 dyad) 4 (50%) 4 (50%) 8 (100%)
NS-NNSL (4 dyads) 35 (69%) 16 (31%) 51 (100%)
NS-NNSH (4 dyads) 25 (66%) 13 (34%) 38 (100%)
NNSL-NNSH (4 dyads) 53 (74%) 19 (26%) 72 (100%)
Copyright © 2012 SciRes.
64
S. KAWAGUCHI, Y. MA
Copyright © 2012 SciRes. 65
Table 10.
NNS’s grammar errors and corrective feedback on grammar errors.
Interaction Combination
L2 Learner
NNSL-NNSL NNSH-NNSH NS-NNSL NS-NNSH NNSL-NNSH Total
NNS (L)1 2/18 (1) N/A 5/30 (17) N/A 16/23 (70) 23/71 (32)
NNS (L)2 1/15 (7) N/A 7/26 (27) N/A 13/21 (62) 21/62 (34)
NNS (H)1 N/A 1/6 (17) N/A 3/8
(38)
0/3
(0) 4/17 (24)
NNS (H)2 N/A 1/9 (11) N/A 3/8
(38)
0/5
(0) 4/22 (18)
Number of CF/number of errors (percentages).
Table 11.
Outcomes of CF on grammar errors.
Learner Source of CF Successfully Uptaken Ignored Total CF
from same level 0 2 2
from NS 4 1 5
from different level 13 3 16
NNS (L)1
Total 17 (74%) 6 (26%) 23 (100%)
from same level 0 1 1
from NS 4 3 7
from different level 9 4 13
NNS (L)2
Total 13 (62%) 8 (38%) 21 (100%)
from same level 0 1 1
from NS 2 1 3
from different level 0 0 0
NNS (H)1
Total 2 (50%) 2 (50%) 4 (100%)
from same level 1 0 1
from NS 2 1 3
from different level 0 0 0
NNS (H)2
Total 3 (75%) 1 (25%) 4 (100%)
grammatical errors are not large, both NNS (Low)1 and NNS
(Low)2 showed high successful uptaking rates of CF (74% and
62% respectively). As for NNS (High)1 and NNS (High)2,
success rates of negotiated interaction were 50% and 75% re-
spectively. However, given total counts of CF they received
were only four each, it is reasonable to conclude that the role of
CF on grammatical errors for advanced learners is much
smaller than for less advanced learners. For advanced learners,
CF on meaning consultation seems to be more significant.
In order to investigate whether negotiated interaction pro-
motes grammatical development, as mentioned in the methods
section, our study adopted a pre-test, treatment and post-test
design. Table 12 provides the frequency of each NNS’s gram-
mar errors on verb past tense, noun plural forms, S-V agree-
ment, question constructions and other grammar errors accord-
ing to pre-test and post-test. An example of “other errors” re-
lates to the choice of definite and indefinite article. In the table
the numbers before and after the slash indicate the frequency of
errors and contexts respectively. Error percentages are provided
in brackets.
For example, in pre-test, NNS (Low)1 created nine contexts
for the past tense but failed to provide appropriate verb forms
four times, i.e., 44% error rate. In post-test, the same learner did
not make any errors for the eight past tense contexts produced.
Thus her error rates decreased from 44% in pre-test to 0% in
post-test. Both NNS (Low) improved substantially as the de-
crease in their error rate from pre- to post-test shows with all
four grammatical phenomena investigated in this study. Espe-
cially the error rates on past tense and plural forms decreased
dramatically from pre- to post-test. For example, error rates on
plural forms for NNS (Low)1 decreased from 80% to 0% and
NNS (Low)2 from 43% to 10% respectively. The sole excep-
tion regards S-V agreement with NNS (Low)1 which registered
a 9% error rate with pre-test and 14% error rate with post-test.
S. KAWAGUCHI, Y. MA
Table 12.
NNS’s errors on past tense, plural forms, S-V agreement and Question forms in pre-test and post-test.
Past tense Plural forms S-V agreement Question forms Other errors
NNS
PRE POST PRE POST PRE POST PRE POST PRE POST
Low 1 4/9 (44) 0/8 (0)4/5 (80) 0/8 (0)1/11 (9) 2/14(14)3/13(23)1/13(8) 6/9 (67) 4/8 (50)
Low 2 2/8 (25) 0/9 (0)3/7 (43) 1/10 (10) 3/9 (33)2/17(12)4/11(36)1/15(7) 3/5 (60) 2/10(20)
High 1 0/16 (0) 0/15 (0)1/15 (7) 1/12 (8)2/24 (4) 0/22 (0) 1/18 (6) 0/21(0) 2/16 (13) 1/17 (6)
High 2 2/18 (11) 1/19 (5)2/16 (13) 0/15 (0)3/27 (11) 0/23(0)1/22(5)0/23(0) 1/20 (5) 1/14(7)
Number of errors/number of contexts (percentage error), PRE = Pre-test, POST = Post-test.
As for the two NNS (High), their error rates were already very
low at pre-test, nevertheless it also decreased across the board
at post-test to reach, mostly 0%. Again there was one exception,
this time with plural forms produced by NNS (High)1, which
went from 7% at pre-test to 8% at post-test. Hardly a significant
change.
Table 13 summarises each learner’s number of errors, num-
ber of contexts and error percentages for each of the all gram-
matical phenomena in pre-test, treatment and post-test sessions.
The changes in NNS’s error rates in this table are graphically
represented in Figure 3. It is evident that the two NNS (Low)
made a dramatic decrease in their error rates as they progressed
through the sessions.
Discussion
We organize the discussion section around our research
questions.
Research question 1. Whether CF and NoM happen in
learner-learner combination. According to Table 6, total turns
of the two groups, NS-NNS and NNS-NNS groups, are 1322
Table 13.
NNS’s total grammar errors and number of contexts.
Learner Pre-test Treatments Post-test
NNS (Low)1 18/47 (38%) 53/205 (26%) 7/51 (14%)
NNS (Low)2 15/40 (38%) 47/183 (26%) 6/61 (10%)
NNS (High)1 6/89 (7%) 11/306 (4%) 2/87 (2%)
NNS (High)2 9/103 (9%) 13/317 (4%) 2/94 (2%)
Number of errors/number of contexts (percentage error).
Figure 3.
Error rates in pre-test, treatment sessions and post-test.
(session average 165.3) and 1055 (session average 175.8) re-
spectively while turns devoted to CF, which is the sum of ex-
plicit and implicit CF in Table 6, are 351 (27%) for NS-NNS
group and 386 (37%) for the NNS-NNS group. Therefore,
NNS-NNS group produced more opportunities for CF. This
confirms that CF and NoM occur in learner-learner combina-
tion. The chi-square test of turn frequencies devoted to recast
and NoM suggests that NS-NNS group and NNS-NNS group
were not significantly different, suggesting that ESL learner’s
interaction with another learner provides an interactional envi-
ronment comparable to interaction with a NS. This result is in
line with other studies (Lyster & Ranta, 1997; Mackey &
Oliver, 2002; Nassaji, 2007; Oliver, 1995, 1998).
Concerning the NNS-NNS groups in particular, it was found
that NNS (Low) received CF more frequently when she was
paired with a more competent interlocutor rather than when she
interacted with NS or another learner of the same proficiency
level. In the example (4), following NNS (Low)2’s usage of the
non-targeted utterance there have, NNS (High)2 immediately
produced a reformulation there is, allowing the NNS (Low)2 to
notice and correct her error. This shows that a more competent
L2 speaker facilitates language acquisition for a less competent
L2 speaker.
4)
NNS (Low)2: and above the stove, there have a picture.
NNS (High)2: There is a picture. There is.
NNS (Low)2: There is.
Occasionally, the NNS-NNS group may also use their native
language to resolve communication problems, as observed in
(5). NNS (Low)1 intended to express the word branch, how-
ever, she appears to have confused the two words branch and
breach. She produced a wrong utterance, triggering NNS
(Low)2 to double check the information. NNS (Low)1 then
used their mutual L1, Chinese, to explain the meaning of the
word. NNS (Low)2 subsequently understood the meaning of
the words and continued with her drawing task. The fact NNSs
can negotiate the meaning in their native language to solve
communication breakdown appears to be one of the reasons
why NNS(H)-NNS(L) group has a higher success rate for CF
than NS-NNS(H/L).
5)
(NNS (Low)1 is describing a picture to her partner)
NNS (Low)1: He catches a breach. Catch..
NNS (Low)2: breach?
NNS (Low)1: what I mean is a branch
Research question 2. Whether CF and NoM differ qualita-
tively and quantitatively in different combination of interac-
Copyright © 2012 SciRes.
66
S. KAWAGUCHI, Y. MA
tion. First, the number of CF in each combination and their
outcomes was analysed. Our data shows that NNS (Low)-NNS
(High) combination has a higher numbers of recast and NoM
than any other combinations (see Table 7). Further, the only
explicit CF instance was also found to occur in this combina-
tion. In addition, NNS (High) produced more recasts than NSs.
For example, NNS (High) produced recasts 22 times in NNS
(Low)-NNS (High) combination while NS produced recasts 15
times in NS-NNS (Low) combination (see Figure 2(a)).
Analysis of NoM also revealed interesting qualitative differ-
ences in the outcome among different interactional combina-
tions (see Table 9). The current study shows that NNS (Low)-
NNS (High) combination outperformed any other combinations,
with a success rate of 74%. NS-NNS (High) (66%) and NS-
NNS (Low) (69%) groups followed shortly behind, while NNS
(Low)-NNS (Low) (25%) group fell considerably behind. Va-
ronis and Gass (1985) found that more NoM occurs in NNS-
NNS interaction than in NS-NS or NS-NNS interactions. The
current study further suggests that NNS (High)-NNS (Low)
provides the best opportunities for NoM. Further, it shows that
more competent speakers were more likely to initiate recasts
while less competent speakers tend to initiate NoMs.
Long (1996) claims that “negotiation of meaning, and espe-
cially negotiation work that triggers interactional adjustments
by the NS or more competent interlocutor, facilitates acquisi-
tion (pp. 451-452)”. However, no studies have investigated
NoM between the NNS of different proficiency levels. Hence,
the current study throws further light on the precise meaning of
Long’s statement since it finds that NNS benefits best when
interacting with NNS of higher proficiency level rather than
with NS or learners of similar proficiency level. Several reasons
may provide an explanation for this finding. First, NSs tend to
dismiss NNS’ grammatical errors or ambiguous utterances.
They appear to focus more on conveying meaning. This seems
to parallel first language acquisition studies where caretakers
were found to correct children when the choice of vocabulary
or sentence meaning was wrong but not when making gram-
matical errors (e.g., Brown & Hanlon, 1970). The following
excerpt in (6) exemplifies this point in L2:
6) (performing “describing a picture” task)
NNS (Low)2: Next to the lamp stand, there has a sofa,
three-seated sofa.
NS2: What kind of sofa? Three-seated sofa?
NNS (Low)2: Yes, three-seated sofa. And the sofa with the
two little arm, two arm, yeah.
NS2: Yeah.
NNS (Low)2 in the above example made a number of gram-
matical errors (as underlined), however, NS did not call atten-
tion to these errors. NSs tend to ignore grammatical errors as
long as the content of the message is understood. On the con-
trary, NNS (High) are shown to be more grammatically sensi-
tive than NS. NNS (High)s pay more attention to their partners’
grammatical errors (See the example in (4) above). NNS
(High)2 immediately highlighted her partner’s incorrect output
there have, which the NS2 failed to do.
NSs, on the other hand, attend more to their partners’ errors
on word choice. They provide more feedback on their incorrect
word choice, incorrect word form and collocation errors. In
example (7), non-target word choice of very big was used by
NNS(L)2 to describe the Chinese culture. NS2 immediately
corrected this by uttering the word very rich.
7)
NS2: What is your favourite topic or subject in school?
NNS (Low)2: Hmm. I like Chinese. Because you know, Chi-
nese culture is very big.
NS2: Very rich.
NNS (Low)2: Yes. We enjoyed it a lot.
Second, NNSs would use their native language when com-
munication problems arise. Notice that code switching occurred
in their negotiated interactions: 44 turns (4% of total turns) in
NNS-NNS and 0 (zero) in NS-NNS (see Table 4) interactions.
Code switching may be an effective strategy to assist NNS to
overcome language difficulties as can be seen from the excerpt
in (8).
8)
NNS (Low)1: She…she was in China?
NNS (High)1: Yes, she is in China.
NNS (Low)1: Do you think it is right to use was”?
NNS (High)1: Is. She is still in China now.
NNS (Low)1: Oh, she is in China.
During their conversation task, NNS (Low)1 asked NNS
(High)1 about where her high school teacher is. NNS (Low)1
became confused when she heard NNS (High)1 use is, and
asked NNS (High)1 in Chinese whether it was right for her to
use was here. After NNS (High)1 told NNS (Low)1 in Chinese
that her teacher is still in China, NNS (Low)1 noticed that she
should use is instead of was. Resorting to code switching as-
sisted NNS (Low)1 to revise her grammar knowledge, which
allowed her to notice the difference between is and was. As
Poole (2005) suggests, code switching assists less competent
interlocutor to learn and respond quickly.
The above two reasons provide some clues as to why NNS
(Low)-NNS (High) combination performed better in CF related
to grammatical errors. To sum up, CF and NoM do differ
qualitatively and quantitatively in different combinations of
interactions. It was observed in this study that the NNS (Low)-
NNS (High) combination outperforms other combinations in
CF and NoM, especially in relation to successful outcomes of
CF and NoM.
Research question 3. Whether negotiated interaction pro-
motes grammar learning. Mackey (1999) claims that there is a
link between interaction and L2 grammatical development. This
leads to the prediction that grammatical errors would decrease
after the treatment. Indeed, we find that this is the case but we
also found, that the learner’s proficiency level is inversely pro-
portional to the level of improvement that can be obtained
through CF and NoM. This means, roughly, that the lower level
learners achieve the greatest benefit from such interaction as
Table 13 and Figure 3 show. An obvious reason for the
smaller decrease in error rates for NNS (High) is due to the
“ceiling effect”: their error rates at pre-test were already very
low, i.e., 7% with NNS (High)1 and 9% with NNS (High)2,
and decreased to 2% for both (approaching native speaker
rates). On the other hand, error rates decreased from 38% (18
errors out of 47 contexts) to 14% (7 errors out of 51 contexts)
for NNS (Low)1, and from 38% (15 errors out of 40 contexts)
to 10% (6 errors out of 61 contexts) for NNS (Low)2.
Conclusion
This study presented an in-depth analysis of Corrective
Feedback (CF) and Negotiation of Meaning (NoM) in NS-NNS
and NNS-NNS interactions. The 14 dyads contain interactions:
Copyright © 2012 SciRes. 67
S. KAWAGUCHI, Y. MA
1) between L2 learners from the same language proficiency
level; 2) between a native speaker and a L2 learner and; 3) be-
tween learners of different proficiency. This type of design has
never been tried in any previous studies.
We posited three research questions. The first asked whether
there is a difference between NS-NNS and NNS-NNS groups in
terms of CF and NoM. Our data suggest that NNS-NNS inter-
action provides, at least, a comparable linguistic environment to
NS-NNS. Indeed, the outcome of CF in NNS (Low)-NNS
(High) combination provided more opportunities for CF and
NoM than learner-native speaker combination. Thus, peer
learning provides excellent opportunities for second language
learning and teaching. The second question asked whether the
CF and NoM differ qualitatively and quantitatively in different
combinations of interaction. Results revealed that the best CF
and NoM outcomes for NNS (Low) were obtained in interac-
tion with NNS (High) rather than with natives. On the other
hand, for NNS (High), the best outcomes were obtained in in-
teraction with native speakers. As for quality, different strate-
gies were employed by the lower/higher proficiency level
speakers: the lower proficiency level learners resorted to NoM
to solve communication problems while the higher proficiency
learners used recast to focus interlocutor’s attention on errors.
This suggests that the difference in levels of proficiency con-
tribute to successful communication in different ways. The last
research question queried the relationship between negotiated
interaction and grammar learning. Results show that there is a
positive relationship: in the post-test all learners made fewer
grammatical errors across the various structures investigated in
comparison with the pre-test.
This study has practical implications. Given the transition in
pedagogy from individual to collaborative learning, L2 learners
are now encouraged to work either in groups or in pairs (Rich-
ards & Rodgers, 2001). This study supports the effectiveness of
collaborative learning. Some scholars assert that L2 learning
through communicative classes do not gain high level of L2
proficiency or complex form (Ellis, 1996; Higgs & Cliford,
1982; Van, 1988). They base their argument on the use of
choppy interlanguage and pidginized language by less profi-
cient L2 learners, and their use of CF and NoM to convey
meaning rather than error correction. However, combinations
such as NNS (Low) and NNS (High), are shown here to be
effective in grammar learning. In fact, in the current study NNS
(Low) received more CF from NNS (High) than from NS. NNS
(High) are more experienced in L2 learning and possibly more
sensitive to grammatical issues than natives: NNS (High) are
able to notice and correct their partners’ grammatical errors
through negotiated interaction. In contrast, combinations in-
volving NS may result in greater learning effect in pronuncia-
tion and vocabulary acquisition, rather than grammar learning.
There are some limitations with this study. Firstly, the num-
ber of participants was small, restricting broad generalisability
and extensive statistical treatment. Second, this study only ex-
amined NNS participants of high and low language proficiency
without including intermediate proficiency. Thus, the perform-
ance of intermediate level learners when they are grouped with
learners of other levels/native speakers and their results on CF
and NoM remains an open question. Lastly, a broader investi-
gation relating accuracy and specific items learned by different
dyadic combinations would require a more explicit grammatical
development framework such as, for instance, Processability
Theory (Pienemann, Di Biase, & Kawaguchi, 2005).
Acknowledgements
We would like to thank Dr Bruno Di Biase and Dr Kenny
Wang for their comments for earlier version of this paper. We
also thank Rio Yamaguchi for her editorial help. Remaining
errors are our responsibility.
REFERENCES
Blake, R. (2000). Computer mediated communication: A window on
L2 Spanish interlanguage. Language Learning Technology, 4, 120-
136.
Bower, J., & Kawaguchi, S. (2011). Negotiation of meaning and cor-
rective feedback in Japanese/English eTandem. Language Learning
& Technology, 15, 41-71.
Braidi, S. M. (2002). Reexamining the role of recasts in native speaker/
nonnative-speaker interactions. Language Learning, 52, 1-42.
doi:10.1111/1467-9922.00176
Brock, C., Crookes, G., Day, R., & Long, M. (1986). The differential
effects of corrective feedback in native speaker-nonnative speaker
conversation. In R. Day (Ed.), Talking to learn (pp. 229-236). Row-
ley, MA: Newbury House.
Brown, R., & Hanlon, C. (1970). Derivational complexity and order of
acquisition in child speech. In J. R. Hayes (Ed.), Cognition and the
development of language (pp. 11-53). New York: Wiley.
Carroll, S., & Swain, M. (1993). Explicit and implicit negative feed-
back: An empirical study of the learning of linguistic generalizations.
Studies in Second Language Acquisition , 15, 357-366.
doi:10.1017/S0272263100012158
Chiba, C. (2010). Negotiation of meaning in the conversation between
advanced learners of Japanese and native speakers of Japanese:
Comparative study on conversation between nonnative speakers and
native speakers. Japanese Language Education, 39, 138-141.
Di Biase, B. (2000). Second language acqusition notes and exercises.
Sydney: Language Australia.
Ellis, N. C. (2002). Frequency effects in language processing. Studies in
Second Language Acquisition, 24, 143-188.
Ellis, R. (1998). Teaching and research: Grammar in teaching. TESOL
Quarterly, 32, 39-40. doi:10.2307/3587901
Ellis, R., Loewen, S., & Erlam, R. (2006). Implicit and explicit correc-
tive feedback and the acquisition of L2 grammar. Studies in Second
Language Acquisition, 28, 339-368.
doi:10.1017/S0272263106060141
Ellis, R., Tanaka, Y., & Yamazaki, A. (1994). Classroom interaction,
comprehension, and the acquisition of L2 word meanings. Language
Learning, 44, 449-491. doi:10.1111/j.1467-1770.1994.tb01114.x
Finocchiaro, M., & Brumfit, C. (1983). The functional-notional ap-
proach: From theory to practice. New York: Oxford University Press.
Gass, S. M. (1997). Input, interaction, and the second language learner.
Hillsdale, NJ: Erlbaum.
Gass, S. M. (2003). Input and interaction. In C. Doughty, & M. Long
(Eds.), The handbook of second language acquisition (pp. 224-255).
Oxford: Blackwell. doi:10.1002/9780470756492.ch9
Gass, S. M., & Mackey, A. (2007). Input, interaction, and output in
second language acquisition. In B. Vanpatten, & J. Williams (Eds.),
Theories in second language acquisition (pp. 175-200). London:
LEA.
Gass, S. M., & Varonis, E. M. (1994). Input, interaction, and second
language production. Studies in Second Language Acquisition, 16,
283-302. doi:10.1017/S0272263100013097
Harmer, J. (2007). The practice of English language teaching. Harlow,
England: Pearson Education.
Hauser, E. (2005). Coding “corrective recasts”: The maintenance of
meaning and more fundamental problems. Applied Linguistics, 26,
293-316. doi:10.1093/applin/ami010
Iwasaki, J., & Oliver, R. (2003). Chat-line interaction and negative
feedback. Occasional Thematic Issue of the Australian Review of
Applied Linguistics (ARAL), 17, 60-73.
Higgs, T. V., & Clifford, R. (1982). The push toward communication.
Copyright © 2012 SciRes.
68
S. KAWAGUCHI, Y. MA
Copyright © 2012 SciRes. 69
In T. V. Higgs (Ed.), Curriculum, competence, and the foreign lan-
guage teacher (pp. 57-136). Lincolnwood, IL: National Textbook
Co.
Iwashita, N. (2003). Negative feedback and positive evidence in task-
based interaction: Differential effects on L2 development. Studies in
Second Language Acquisition, 25, 1-36.
doi:10.1017/S0272263103000019
Kim, H., & Mathes, G. (2001). Explicit vs. implicit corrective feedback.
The Korea TESOL Journal, 4, 1-15.
Lee, L. (2006). A study of native and nonnative speakers’ feedback and
responses in Spanish-American networked collaborative interaction.
In J. Belz, & S. Thorne, (Eds.), Internet-mediated intercultural for-
eign language education (pp. 147-176). Boston: Thomson Heinle.
Li, S. F. (2010). The effectiveness of corrective feedback in SLA: A
meta-analysis. Language Learning, 60, 309-365.
doi:10.1111/j.1467-9922.2010.00561.x
Long, M. H. (1996). The role of the linguistic environment in second
language acquisition. In W. C. Ritchie, & T. K. Bhatia (Eds.), Hand-
book of second language acquisition (pp. 413-468). New York: Aca-
demic Press.
Long, M. H., Inagaki, S., & Ortega, L. (1998). The role of implicit
negative feedback in SLA: Models and recasts in Japanese and
Spanish. Modern Language Journal, 82, 357-371.
doi:10.1111/j.1540-4781.1998.tb01213.x
Lyster, R. (1998). Recasts, repetition, and ambiguity in L2 classroom
discourse. Studies in Second Lang uage Acquisition, 20, 51-81.
doi:10.1017/S027226319800103X
Lyster, R. (2004). Differential effects of prompts and recasts in form-
focused instruction. Studies in Second Language Acquisition, 26,
399-432. doi:10.1017/S0272263104263021
Lyster, R., & Ranta, L. (1997). Corrective feedback and learner uptake:
Negotiation of form in communicative classrooms. Studies in Second
Language Acquisition, 20, 37-66.
Mackey, A. (1999). Input, interaction, and second language develop-
ment: An empirical study of question formation in ESL. SSLA, 21,
557-587.
Mackey, A., & Oliver, R. (2002). Interactional feedback and children’s
L2 development. System, 30, 459-477.
doi:10.1016/S0346-251X(02)00049-0
Mackey, A., Oliver, R., & Leeman, J. (2003). Interaction input and the
incorporation of feedback: An exploration of NS-NNS and NNS-
NNS adult and child dyads. Language Learning, 53, 35-66.
doi:10.1111/1467-9922.00210
Nassaji, H. (2007). Elicitation and reformulation and their relationship
with learner repair in dyadic interaction. Language Learning, 57,
511-548. doi:10.1111/j.1467-9922.2007.00427.x
Nassaji, H. (2009). Effects of recasts and elicitations in dyadic interac-
tion and the role of feedback explicitness. Language Learning, 59,
411-452. doi:10.1111/j.1467-9922.2009.00511.x
Oliver, R. (1995). Negative feedback in child NS-NNS conversation.
Studies in Second Language Acquisition, 17, 459-481.
doi:10.1017/S0272263100014418
Oliver, R. (1998). Negotiation of meaning in child interactions. The
Modern Language Journal, 82, 372-386.
doi:10.1111/j.1540-4781.1998.tb01215.x
Oliver, R. (2002). The patterns of negotiation for meaning in child
interactions. The Modern Languag e Journal, 86, 97-111.
doi:10.1111/1540-4781.00138
O’Rourke, B. (2005). Form-focused Interaction in online tandem
learning. CALICO Journal, 22, 433-466.
Pica, T. (1994). Research on negotiation: What does it reveal about
second-language learning conditions, processes, and outcomes? Lan-
guage Learning, 44, 493-527.
doi:10.1111/j.1467-1770.1994.tb01115.x
Pica, T., & Doughty, C. (1985). Input and interaction in the communi-
cative language classroom: A comparison of teacher-fronted and
group activities. In S. M. Gass, & C. G. Madden (Eds.), Input in sec-
ond language acquisition (pp. 115-132). Rowley, MA: Newbury
House.
Pica, T., Young, R., & Doughty, C. (1987). The impact of interaction
on comprehension. TESOL Quarterly, 21, 737-758.
doi:10.2307/3586992
Pienemann, M., Di Biase, B., & Kawaguchi, S. (2005). Extending proc-
essability theory. In M. Pienemann (Ed.), Cross-linguistic aspects of
processability theory (pp. 199-251). Amsterdam and Philadelphia:
John Benjamins.
Poole, A. (2005). Focus on form instruction: Foundations, applications
and criticisms. The Re ading Matrix, 5, 47-56.
Richards, J. C., & Rodgers, T. S. (2001). Approaches and methods in
language teaching (2nd ed.). New York: Cambridge University Press.
doi:10.1017/CBO9780511667305
Schwartz, B. D. (1993). On explicit and negative data effecting and
affecting competence and linguistic behavior. Studies in Second
Language Acquisition, 15, 147-163.
doi:10.1017/S0272263100011931
Sotillo, S. (2005). Corrective feedback via instant messenger learning
activities in NS-NNS and NNS-NNS dyads. CALICO Journal, 22,
467-496.
Storch, N. (2002). Patterns of interaction in ESL pair work. Language
Learning, 52, 119-158. doi:10.1111/1467-9922.00179
Swain, M. (1995). Three functions of output in second language learn-
ing. In G. Cook, & B. Seidlhofer (Eds.), Principle and practice in
applied linguistics: Studies in honour of H. G. Widdowson (pp. 125-
144). Oxford: Oxford University Press.
Van Lier, L. (1988). What’s wrong with the classroom talk? Prospect, 3,
267-283.
Varnosfadrani, A., & Basturkmen, H. (2009). The effectiveness of
implicit and explicit error correction on learners’ performance. Sys-
tem, 37, 82-98. doi:10.1016/j.system.2008.04.004
Varonis, E. M., & Gass, S. M. (1985). Non-native/non-native conversa-
tion: A model for negotiation of meaning. Applied Linguistics, 6,
71-90. doi:10.1093/applin/6.1.71
Yang, Y., & Lyster, R. (2010). Effects of form-focused practice and
feedback on Chinese EFL learners’ acquisition of regular and irregu-
lar past-tense forms. Studies in Second Language Acquisition, 32,
235-263. doi:10.1017/S0272263109990519
Zhao, Y., & Angelova, M. (2010). Negotiation of meaning between
non-native speakers in text-based chat and videoconferencing. US-
China Education Review, 7, 12-26.
S. KAWAGUCHI, Y. MA
Appendix
A: Board Game
This board game is downloaded from website “eslbase”, which
claims that this resource could be photocopied for class use.
B: A Sample Pictures for Description
Copyright © 2012 SciRes.
70