'i.’ ELIII
3 Sun}... w...

 

 

 

 

 

ESV§ fi¥mznub§£wn

ms

t 700%
c93539/

This is to certify that the
thesis entitled

ANONYMITY AND NEGATIVE FEEDBACK IN CMC

presented by
YOKO HANGUI

has been accepted towards fulfillment
of the requirements for the

MA. degree in TESOL

W M‘ X52”

Major Professor’s §ingaTure

 

July 6, 2004

 

Date

MSU is an Affirmative Action/Equal Opportunity Institution

—.-.—»-.-._.-.- .n-o-o-.----.-.-.-o--c-.-n-p-.-o-------u--

 

t' 777‘—

 

LIBRARY '
Michigan State
University

 

 

 

PLACE IN RETURN Box to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

JUL 1 o 200

 

04 3ll N

 

 

 

 

 

 

 

 

 

 

 

 

 

6/01 c:/CIRC/DateDue.p65—p.15

AN ONYMITY AND NEGATIVE FEEDBACK IN COMPUTER-MEDIATED
CMMUNICATION

By

Yoko Hangui

A THESIS

Submitted to
Michigan State University
In partial fulfillment of the requirements
for the degree of

MASTER OF ARTS

Department of Linguistics, Germanic, Slavic, Asian, and African Languages

2004

ABSTRACT

AN ONYMITY AND NEGATIVE FEEDBACK IN COMPUTER-MEDIATED
COMMUNICATION

By

Yoko Hangui

The present study investigated whether and how anonymity influences the provision and
use of negative feedback in task-based computer-mediated communication (CMC). A
total of 32 dyads, involving native speakers (N Ss) and normative speakers (NNSs) of
English, participated in a two-way communication task. Two synchronous online chat
programs were chosen in order to manipulate anonymity: one for the identifiable
condition, and the other for the anonymous condition. The results of chi-square tests
show that, whereas there were no differences in the amount of the provision of negative
feedback and use of opportunities between the identifiable condition and the anonymous

condition, there was a significant difference in the number of NNS reactions to NS

response between two conditions.

Cepyright by
‘Yokoliangui

2004

ACKNOWLEDGMENTS

I would like to thank Dr. Susan Gass and Dr. Charlene Polio for their comments on the
contents of this thesis and suggestions on editing. I would also like to thank the Michigan
State University ELC students, the Michigan State University students, the University of
Aizu students, and the JET teachers who participated in the research for their cooperation.
I am also grateful for the help from Laura Carrion, Graham Reeves, and Miki Motohashi

in the process of coding the data.

iv

TABLE OF CONTENTS

LIST OF TABLES .................................................................................. vii
LIST OF FIGURES ............................................................................... ix
INTRODUCTION .................................................................................. 1
Positive Evidence and Negative Evidence ..................................................... 1
Interaction as a Source of Negative Evidence ............................................. 6
Feedback and Attention ........................................................................ 8
Computer-Mediated Communication and SLA ........................................... 9
Interaction in CMC ............................................................................. 14
Four Features of CMC ......................................................................... 17
Anonymity ....................................................................................... 18
The Purpose of the Present Study ............................................................. 19
METHOD ........................................................................................... 20
Research Questions ............................................................................. 20
Variables and Hypothesis ..................................................................... 20
Experimental Design Overview ............................................................ 23
Participants ....................................................................................... 24
Participants in Japan25
Participants in the United States .................................................. 25

Task Design ..................................................................................... 26
Tasks ...................................................................................... 26
Materials ......................................................................................... 27
Chat Software 1: MSN Messenger version 6.1 ..................................... 27

Chat Software 2: Activeworlds ....................................................... 28
Procedures ........................................................................................ 30
Coding ............................................................................................ 30
Chat Scripts ............................................................................. 30

Off-task Discourse ..................................................................... 32

Data Coding ............................................................................. 32
RESULTS ............................................................................................ 43
Negative Feedback .............................................................................. 44
NNS Reactions to Negative Feedback ....................................................... 57
Summary of Results ............................................................................ 65

Footnotes .......................................................................................... 81

APPENDIX ........................................................................................... 83

REFERENCES ..................................................................................... 86

LIST OF TABLES

Table l - Dyads in the identifiable and the anonymous conditions .......................... 44
Table 2 - Non-target-like turns with and without opportunities for negative feedback. . . .45

Table 3 - Negative feedback to non-target-like production in the identifiable and the
anonymous dyads ................................................................................. 46

Table 4 - Provision of negative feedback in the case that opportunities were provided
......................................................................................................... 47

Table 5 - Negative feedback provision to non-target-like turns without opportunities for
negative .............................................................................................. 48

Table 6- Negative feedback with and without opportunities for NNS reactions

......................................................................................................... 49
Table 7 - Distribution of types of negative feedback ........................................... 52
Table 8 - NNS reactions to negative feedback .................................................. 58
Table 9 - NNS reactions in the case of opportunities provided .............................. 59

Table 10 - NNS reactions to negative feedback with no opportunities provided

......................................................................................................... 60
Table 11 - Types of NNS reactions ............................................................ 61
Table 12 - NNS reactions to each feedback type ............................................... 63

vii

Table 13 — “Modify” type of reactions to negative feedback types ........................ 64

Table 14 — “Other” type of reactions to negative feedback types .......................... 64

Table 15 - NNS reactions to each feedback type ............................................. 74

viii

LIST OF FIGURES

Figure 1 - MSN Messenger version 6.1 interface ............................................. 28

Figure 2 - Activewords interface ............................................................ 29

Figure 3 - Non-target-like turns with and without opportunities for negative feedback

....................................................................................................... 45
Figure 4 - Provision of negative feedback to non-target-like production .................. 46
Figure 5 - Provision of feedback in the case that opportunities were provided .......... 47

Figure 6 - Negative feedback provision to non-target-like turns without opportunities

for negative feedback ............................................................................ 49
Figure 7 - Negative feedback with and without opportunities for NNS reactions ...... 50
Figure 8 - Distribution of types of negative feedback .................................... 52
Figure 9 - NNS reactions to negative feedback ............................................. 58
Figure 10 - NNS reactions in the case of opportunities provided .......................... 59
Figure 11 - NNS reactions to negative feedback with no opportunities provided ...... 6O

ix

Introduction
Positive Evidence and Negative Evidence

It is commonly accepted that input is important for language acquisition to occur;
however, there are different theoretical perspectives regarding the exact role of input in
the process of language acquisition. One perspective regards input as the major factor
while another has the view that input works within the framework of internally
programmed Universal Grammar. Nonetheless, regardless of the different views, input is
considered to be an important construct in language acquisition.

Although the importance of input is uncontroversial both in first language acquisition
(FLA) and second language acquisition (SLA), the kinds of linguistic input that play a
role in language acquisition have been debated. Theoretically in language acquisition
there are at least two kinds of evidence available for language learners: positive evidence
and negative evidence. Positive evidence is input, which comprises the set of well-formed
sentences to which learners are exposed. Negative evidence is the data that provide
information to the language learner about what is not allowable in the target language.
With respect to evidence requirements for language acquisition, two positions view the
roles of two kinds of evidence differently. One position strongly supports the role of
innate knowledge in language acquisition. The opposing position considers the effects of
linguistic environmental variables on language development as well as the role of innate
knowledge.

From a nativist perspective, which supports the role of innate knowledge, only
positive evidence constitutes input for language learners. Grimshaw and Pinker (1989),

Pinker (1989), and Beck and Eubank (1991) argue that the role of negative evidence in

language acquisition should not be assumed unless four conditions are met: (a) Negative
evidence exists, (b) it exists in a usable form, (c) it is used, and (d) it is necessary for
acquisition.

There have been a number of studies exploring negative evidence. In FLA, early
definitions limited negative evidence to explicit prohibitions or corrections. Thus, very
little negative evidence was found in conversation between mothers and children (Brown
& Hanlon, 1970). There is some research investigating the existence and use of negative
evidence. Replicating the study of Brown and Hanlon (1970), Hirsh-Pasek, Treiman, and
Schneiderman (1984) found that mothers are more inclined to repeat children’s
ungrammatical sentences than grammatical sentences. They suggested that the cues that
distinguish between well-formed and ill-formed sentences exist in the language learning
environment. Demetras, Post, and Snow (1986) expanded the definition of negative
feedback. They investigated whether mothers responded differently to their children’s
utterances with explicit and implicit feedback. They found that mothers are more likely to
provide implicit feedback or request clarification of children’s ill-formed utterances than
of the well-formed ones. Bohannon and Stanowicz (1988) observed adult-child
conversation for evidence of adults’ differential responding to the children’s syntactic,
phonological, and semantic errors. They found that adults are likely to provide the correct
syntactic form or pronunciation immediately after the child error has been uttered. Farrar
(1992) investigated the effects of recasts on the morpheme acquisition of children, and
found there were corrective recasts following ungrammatical sentences and children

imitated the grammatical morphemes contained in the corrective recasts. Thus, these

findings suggest that there is negative evidence provided in child language acquisition
and children notice the information and make use of it.

In SLA, while there are those who support the nativist perspective, a number of
studies that contribute to the four conditions for negative evidence to be assigned a role in
language acquisition have been conducted. Early studies defined only explicit error
corrections as negative evidence in SLA as well. Implicit negative feedback was not
taken into consideration. Chun, Day, Chenoweth, and Luppescu (1982) investigated
native speaker (NS) - nonnative speaker (NN S) interaction. They examined natural
conversation and found that only 8.9 % of errors were corrected by NSs.

On the other hand, recent research includes implicit negative feedback, such as recasts
and negotiation strategies, which involve repetition, confirmation checks, and
clarification requests. A large amount of research has revealed that negative feedback
obtained during negotiation work or elsewhere may be facilitative of L2 development.

There are many studies that explored the existence and the use of negative feedback,
which are the first two requirements for negative evidence Oliver (2000) examined the
provision and use of negative feedback in the context of NS-NNS dyads and the context
of classrooms with both adults and children. The results indicate that learners in any
condition received negative feedback and used the feedback provided to them. There was
no significant difference according to age, but learners ignored negative feedback more in
the dyad context than the classroom context. Mackey, Oliver, and Leeman (2003)
compared interaction between NS-NNS and NNS-NNS adult and child dyads. They
found that negative feedback exists in all dyad types, where at least 30 % of errors

resulted in negative feedback, much of which led to modified output. Further, they found

that NSs provided significantly more feedback than NNSs in adult dyads. This finding
seems consistent with the previous literature. Pica, Lincoln-Porter, Panios, and Linnell
(1996) compared the interaction of NNS-NNS with the interaction of NS-NNS. They
found that the interaction of NNS-NNS does not provide as much feedback as interaction
with NSs does.

Thus, it seems that the first and the second requirements for negative evidence, that is,
the existence and the use of negative evidence, have been met in SLA. Furthermore, not
only existence and use in various contexts, but also the different forms of negative
feedback, have been explored. With regard to the forms of negative feedback, which is
the third requirement for negative evidence, Long (1996) suggests that the forms of
negative evidence can be either preemptive or reactive. Reactive negative evidence, that
is, negative feedback, can be either explicit or implicit.

There is some observational research investigating the forms of negative feedback
found during conversational interaction. Oliver (1995) examined the pattern of
interaction in child NS-NNS dyads. Two forms of negative feedback were identified:
recasts and negotiation strategies, the latter of which includes repetition, clarification
checks, and comprehension checks. The findings were that NSs respond differently to the
grammaticality and ambiguity of their NNS peer’s utterances, that NS responses were
triggered by the type and complexity of NNS errors, and finally that negative feedback
was incorporated by the NNSs into their interlanguage system. Another observational
study was classroom research conducted by Lyster and Ranta (1997). They transcribed
four French immersion classes in Canada and identified six different types of negative

feedback used by teachers. The six types were explicit correction, recasts, clarification

requests, metalinguistic feedback, elicitation, and repetition. They found that there was
negative feedback provided to learners and that the most frequent type of negative
feedback in classrooms was recasts even though they did not elicit learners’ immediate
incorporation. Leeman (2003) compared four different types of interactional input,
including recasts, negative evidence, enhanced salience of positive evidence, and
unenhanced positive evidence (control), to investigate the elements that can account for
the benefits of recasts. She found that only recast and enhanced salience groups
significantly outperformed negative evidence and unenhanced groups, concluding that the
utility of recasts is derived in part from enhanced salience of positive evidence, and that
the implicit negative evidence may not be a crucial factor. Mackey, Gass, and
McDonough (2000) investigated the perception of negative feedback. In order to examine
leamer’s interpretation of interactional feedback, the technique they used was stimulated
recall. They found that learners were accurate in perceiving lexical and phonological
feedback as such, while they were not accurate in identifying morphosyntactic feedback
as such. Given that learners perceived some feedback as feedback, this suggests that the
forms of interactional feedback are potentially usable even if not actually used during
interaction.

The effects of negative feedback on linguistic development according to type have
also been studied. To measure linguistic development, these studies were conducted with
a pretest-posttest design. Long, Inagaki, and Ortega (1998) investigated the role of
implicit negative feedback, recasts in particular, for L2 Japanese and Spanish learners.
They found that adults can learn from implicit negative feedback and recasts can be more

effective than models. Mackey and Philp (1998) investigated the effect of recasts on

learners’ interlanguage development over time as well as the nature and content of
learner response to recasts. They suggested that recasts may be beneficial for short-term
development even if it is not incorporated into immediate learner response. Iwashita
(2003) examined the role of task-based interaction in grammatical development. She
found that although NS interactional moves containing positive evidence were 10 times
more frequent than those containing implicit negative feedback, only learners who had an
above-average score on the pretest benefited from the positive evidence provided. In
contrast, implicit negative feedback had beneficial effects on short-terrn development of
the grammatical targets regardless of the learner’s current mastery of the target structures.
Recasts were found to have a larger impact than other conversational moves on short-
terrn L2 grammatical development.

In sum, many studies have examined negative evidence across languages, ages, and
interlocutor types. The results of studies suggest that in SLA, negative evidence exists, its
forms are potentially usable, it is used, and lastly it seems to contribute to language
development.

Interaction as a Source of Negative Evidence

From the perspective of those who support the effects of linguistic environmental
variables on language acquisition, conversational interaction plays an important role. In
the interactionist paradigm, whereby it is believed that conversational interaction plays a
role in language acquisition, it is important that interaction is not considered as a cause of
language development, but as a “priming device” for language learning (Gass, 1997).
Long (1996) also discusses the facilitative role of interaction in the updated version of the

Interaction Hypothesis:

. negotiation for meaning, and especially negotiation 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)

In addition to the theoretical support for the role of interaction, empirical studies
support the role of interaction as a facilitator for language learning (Gass, Mackey, &
Pica, 1998, for review). Some research suggests that interaction facilitates language
development. Gass and Varonis (1994) investigated the relationship among input,
interaction, and production. They manipulated the conditions for input: modified or
unmodified, and for interaction: interactive, or noninteracive. They found that there is a
positive influence of interaction on a learner task performance. Polio and Gass (1998),
replicating the study of Gass and Varonis (1994), found interaction induced NSs’
modified output, and led to better comprehension of NNSs. Mackey and Philp (1998)
found positive developmental effects of interaction with intensive recasts. In the intensive
recast treatment, the trained NSs provided recast to any non-target-like utterance given by
the NNSs as much as possible. Their results suggest that for more advanced learners,
interaction with intensive recasts may be more beneficial. Mackey (1999) investigated the
development of question formation. She found a positive relationship between interaction
and development. Learners with structure focused interaction moved along a
developmental path more rapidly than learners without it. Thus, it can be said that the
findings of empirical studies lend support to the facilitative role of interaction.

In relation to negative evidence, one of the reasons that interaction is beneficial in

language learning concerns the fact that learners can receive feedback during interaction.

In other words, conversational interaction possibly avails language learners with both
positive and negative evidence. Thus, conversational interaction can be said to not only
play a facilitative role in language acquisition, but work as a source of negative evidence.
Feedback and Attention
Given that positive and negative evidence are available in the linguistic environment
for language learners, it might be important to discuss the cognitive process of learners
that associates positive and negative evidence with learning. What helps language
learners internalize linguistic input available in the environment into their interlanguage
system is an issue to concern. Gass, Svetics, and Lemelin (2003) discuss the concept of
attention in SLA, summarizing the literature. They propose that attentional devices might
enable humans to “turn in” some stimuli and “turn out” others in language processing.
Further, the importance of feedback has also been discussed in relation to attention (Gass
et al., 1998; Gass, et al., 2003). Pica (1994) considers the availability of feedback in
negotiated interaction regarding attention. She suggests that negotiation offers feedback
to learners, and that feedback leads learners to pay attention to L2 versions of their
interlanguage utterances and heightens their awareness of their own interlanguage
system. Gass et al.(2003) refer to the issue of error correction (negative feedback) and
discuss the relationship between attention and negative evidence. They state that:
...corrective feedback and negative evidence, both of which entail directed
attention to specific parts of language, are not static phenomena: the learner ’5
internal mechanism interacts with external factors. (p. 507)
It might be said that error correction (negative feedback) is important in that it can act as

a catalyst to draw learners’ attention to form.

There is some classroom research on error correction with respect to attention.
Research such as focus on form, or form-focused instruction (Spada, 1997), has
investigated the role of feedback in relation to attention.

Spada and Lightbown (1993) examined the effects of form-focused instruction and
corrective feedback on the use of interrogatives in an oral communication task. They
found that form-focused instruction and corrective feedback contribute to second
language development in the short and long term. Doughty and Varela (1998)
investigated the effect of implicit focus on form technique on past tense in a
communicative language classroom. The results reveal that learners in the treatment
group improved both in accuracy and the total number of attempts at past tense reference.
Summarizing the literature, Long (2000) argued that by briefly shifting leamer’s attention
to linguistic features during meaning-focused tasks, learners can focus on form. In other
words, feedback can play a role in drawing learners’ attention to a problem area in
language learning. Thus far, linguistic environments have been proven to play an
important role in language acquisition. The role of linguistic environments does not seem
to be limited to face-to-face communication. Computer-mediate communication might
also be considered to be potential environments in language acquisition.
Computer-Mediated Communication and SIA

Advanced computer-mediated communication (CMC) technologies have had an
impact on practices in communication. CMC technologies today afford us with not only
text tools, but also audio and visual tools. Furthermore, with the advent of the Internet,
CMC technologies enable learners to communicate with anyone in any place in

synchronous or asynchronous modes. Synchronous CMC is communication where a

message is sent and received instantaneously, such as chatting online, while
asynchronous CMC is communication where there is an interval when a sender sends a
message and when a receiver retrieves the message, such as email.

CMC seems to have great potential for language teaching and learning. Kern (1995)
compared language production in CMC with production in oral discussion. He found that
there was a striking difference in quantity of production. The CMC group outperformed
the oral discussion group. Beauvois (1992) observed the computer-assisted classroom
discussion of foreign language classes and found that typically reticent students and
anxious or shy learners tend to participate in the CMC discussion more readily and more
often than in the regular classroom. She also argues that CMC creates a relatively non-
threatening atmosphere in the classroom. Chun (1994) found that students developed
discourse skills and interactive competence. Furthermore, she found that CMC is a less
stressful environment for learning, where learners are under neither time pressure to
respond nor psychological pressure of making a mistake and looking foolish. In addition,
another advantage is that CMC seems to have both informal and formal L2 environments;
that is, it has an informal L2 environment because it provides extensive exposure to the
L2 as spoken by native speakers within authentic communicative settings with many
opportunities for productive use of the L2, and it also has a formal L2 environment, since
it also provides focus on form and corrective feedback to the L2 learner (Cziko, in press).

Synchronous CMC technologies have been increasingly used for language teaching.
Unique features of synchronous CMC have also been discussed. It has been said that
synchronous CMC discourse exhibits features of both oral and written language. As

Smith (2003) summarizes, the real-time nature of the communication, the ability to

10

provide stress to words and phrases, the use of the first person, and the clear informality
are the characteristics similar to spoken language. It is similar to written language in that
it lacks intonation, it has a permanent record of the discourse, it has increased lexical
density, and it uses punctuation and textual formatting in messages. The unique features
for synchronous CMC discourse are simplified registers, abbreviations, simplified syntax,
the acceptance of surface errors, and the use of symbols and emoticons to express
emotion. With respect to turn-taking, it includes a high degree of disrupted turn
adjacency and more overlaps than in face-to-face communication. In her research
investigating interactional coherence, Herring (1999) found that responses are often
separated from the turns they are responding to, topics tend to decay quickly, and
multiple, overlapping exchanges often share the same channel. Only one interactant at a
time can use a given channel. The other interactant can read the message only when the
message is sent by his/her interactant hitting the enter key on the keyboard of the
computer.

It might be interesting to look at an example of how turn-taking can be disrupted. The
following data, chat scripts taken from the data of the present study, shows an example of
a disrupted turn adjacency. A NS and a NNS are engaged in the communicative task,
where they are comparing very similar pictures. They are talking about a female in their
pictures. Example 1 (a) illustrates the conversational data that appear in the screen of

their computers.

11

(1) Example of a disrupted turn adjacency

(a)

NNS:

NS:

NS:

NS:

NS:

woman hair is wave?
hmm

I would have said 'curl
'curls'

girl hair color is black??

811W]

Is it shoulder length hair?

thanks.

yap..
Then probably that’s not really a difference

(NS 12, NNS 14, identifiable)

In most synchronous CMC technologies, every time a user hits the enter key on the

keyboard, the message pops up in the screen of interactants. In Example 1 (a), the second,

the third, and the fourth lines started with “NS,” which means that a NS typed a message,

“hmm,” once and hit the enter key to send this message to his/her interactant, and then

the NS typed “I would have said 'curl” and hit the enter key, and typed “'curls'” similarly.

Although this looks like three utterances, it can be said that the second, the third, and the

fourth lines comprise one communicative purpose. Fragmentation like this was often

observed both in the NNS and NS data of the present study.

12

Furthermore, turn-taking in CMC has been discussed in the CMC literature, where
disrupted turn adjacency is considered as the rule rather than exception (Smith, 2001 ,
2003; Herring, 1999). Looking at the example closely enables us to realize that not only
turn-taking is violated, but also the sequence is incoherent. This happens because of the
one-way nature of CMC. We cannot be certain from the data of the present study about
whether or not each participant is really typing at the same time. Example 1 (b) illustrates
a possible interpretation of the turn coherence of the same data with Example 1 (a), based
on a speculated time frame. When both the NNS and the NS are thought to be typing
simultaneously, their production is placed on the same line.

(1) Example of a disrupted turn adjacency
(b)
Time NNS NS
1 woman hair is wave?

2 girl hair color is black?? hmm

3 I would have said 'curl

4 'curls'

5 ahW Is it shoulder length hair?

6 thanks.

7 yap..

8 Then probably that’s not

really a difference

13

As seen in Example 1(b), speculating the timing when the participants typed a message
can give us a possible explanation of disrupted turn adjacency. In Time 1 only a NNS
typed a message. In Time 2, however, a NS started typing a corresponding message to the
first turn of the NNS, meanwhile the NNS did not wait for his/her interactiant’s reply but
typed the further question, “girl hair color is black??” After the NNS read the NS’s
utterance on the hair style of a female in the picture, the NNS typed in Time 5,
“ahW,” and in Time 6, “thanks..,” which is possibly responding to the NS
utterance Time 2 through 4. At the same time, in Time 5 the NS moved to the next
question, which resulted in a delayed reply from the NNS in Time 7.

Furthermore, communication is carried out in slow motion in online chatting
(Beauvois, 1992). This could be very beneficial since learners are provided more time for
processing while reading and typing messages, though the “feel” of the interaction
remains similar to that of face-to-face oral interaction (Smith, 2001, 2003). Hence, it
seems that those features unique to CMC are beneficial for language learning.
Furthermore, CMC technologies carry possibilities for classroom and nonclassroom use.
Interaction in CMC

Thus far, given its advantages, CMC technologies seem to have potential as a tool for
communication. One concern is whether CMC technologies afford an environment that
can be facilitative for language learning as face-to-face communication does. In other
words, whether similar features of interaction to ones in face-to-face communication can
be found is in question.

A common account for the benefits of conversational interaction in face-to-face

communication is that conversational interaction has such features that interaction

14

involving NNSs includes negotiation of meaning frequently, that interaction provides
learners with linguistically modified input, such as repetition or simplification, and
interactionally modified input, such as comprehension checks or clarification requests,
both of which make input more comprehensible, and that interaction provides learners
with opportunities for comprehensible output, where learners may move from semantic
processing to syntactic processing.

Interaction in CMC can also be said to facilitate second language acquisition. There is
evidence of comprehensible input, modified output, and negotiation of meaning observed
in CMC as well (Blake, 2000; Kitade, 2000; Pellettieri, 2000). Kitade (2000) investigated
text-based CMC, and argues that a lack of non-verbal cues may facilitate negotiation of
meaning because communication has to rely merely on verbal correspondence in the
CMC environment. Toyoda and Harrison (2002) examined negotiation of meaning
occurring between NSs and NNSs of Japanese in the CMC environment. They examined
what difficulties resulted in “triggers” 2 in negotiated interaction in CMC. Smith (2001,
2003) examined NNS-NNS interaction in task-based CMC and found that learners indeed
negotiate for meaning in the CMC environment when nonunderstanding similar to face-
to-face communication occurs. Negotiated interaction occurred similarly in CMC, but
there were some differences from face-to-face communication. He compared his CMC
model with the face-to-face model established by Varonis and Gass (1985), and found
more frequent occurrence of NNS reactions to the NS response in CMC negotiated
interaction than in face-to-face communication. In addition, interaction in CMC is also

beneficial because learners can monitor and edit their messages (Ortega, 1997; Pellettieri,

2000).

15

In addition to the similarity found between face-to-face communication and CMC,
features unique to CMC were also found. CMC technologies have logging capabilities,
which make it possible for participants to see the conversation on the screen while they
are doing text-based chatting. The conversation data can also be saved very easily and
can be retrieved and reviewed by learners and teachers. Blake (2000) argues that these
logs could be valuable resources for the enhancement of the learners’ interlanguage.

Interactions in CMC have been studied by many researchers. However, not many
studies have investigated negative feedback in CMC. In her observational study,
Beauvois (1992) refers to error correction in CMC. In observing computer-assisted
classroom discussion, she found that students were becoming indifferent to the
appropriate usage of the target language as the semester progressed. Pellettieri (2000)
examined task-based synchronous CMC interaction between NNSs of Spanish. She found
that corrective feedback was given to lexical mophorsyntactic, and semantic errors in
CMC, and that there was incorporation of both explicit and implicit feedback, though the
rate was very low. Iwasaki and Oliver’s (2003) empirical study investigated negative
feedback in a CMC laboratory setting. They explored communicative interactions
between N85 and NNSs of Japanese with an online text-chat program. They had the
participants have “free online chat conversation.” They found that negative feedback
indeed exists in CMC, and it was used by learners, although the proportion of negative
feedback to the number of non-target-like turns was lower than that found in the previous

studies of face-to-face communication.

16

Four Features of CMC

Provided that CMC technologies afford potential use in language acquisition, a
theoretical examination of features of CMC technologies might be required. That is, the
question is whether and how the features specific to CMC influence the environment of
SLA. Smith, Alvarez-Torres, and Zhao (2003) provide a theoretical view for CMC
technologies. They discuss the features of CMC technologies and their impact on
language learners’ online interaction. They state that CMC technologies should be
considered not as a uniform entity but viewed in terms of their own individual
characteristics, and they delineated the four fundamental qualities: temporality,
anonymity, modality and spatiality. In terms of temporality, as discussed, communication
can take place in two fundamentally different modes: asynchronous and synchronous.
Modality concerns information presentation. Some CMC technologies afford only written
texts, while the others can include audio, video, and graphics. Spatiality in face-to-face
communication is concerned with proximics, such as seating arrangement, whereas
spatiality in CMC environments can be manipulated with graphical online chat programs
by controlling the interface’s spatial arrangement. For instance, participants were
instructed that they could get closer to or farther from a Webcamera with MSN
Messenger, and that they could get closer to or step away from their interactant’s avatar
by using the arrow keys with Activeworlds. In terms of anonymity, CMC technologies
also differ in the degree to which participants’ identities are revealed to one another.
Complete anonymous communication is possible in CMC, while CMC also supports
communication where identities are more obvious. Anonymity is less controlled in email.

Pseudonyms seem to be preferred in online chat environments.

17

Anonymity

I will focus on the anonymity feature in the present study. The debate on anonymity is
not limited to CMC technologies. The previous research on anonymity has found its
impact on human behavior in several ways. Anonymity seems to induce more aggressive,
even abnormal, behaviors. There is research investigating the relationship between
anonymity and driving behavior (Ellison, Govern, Petri, & Figler, 1995). They
investigated how anonymity influences the driving behavior of those who drive
convertibles or 4x43 with the tops down (identifiable condition) and with the tops up
(anonymous condition). They found supporting evidence for the previous findings that
anonymity facilitates aggression. Anonymity also influences honesty. Bump (1990)
examined the effects of use of pseudonyms in class discussion using networked
computers. Participants felt that the primary advantage of pseudonyms was that they
allowed true feelings to be expressed without fear of future recrimination. The results
revealed that with pseudonyms computer-assisted discussion increases individuality and
self-disclosure. Makkai and MacAllister (1991) investigated the relationship between
anonymity and accuracy on sensitive questions. They asked the respondents the same
questions in a sealed booklet and in a standard face-to-face interview. The results
revealed that anonymity contributed to accuracy. Furthermore, anonymity seems to lead
to more critical feedback. Antonioni (1994) investigated the effects of feedback
accountability on upward appraisal ratings. He found that subordinates who used the
accountability procedure rated their managers significantly higher than the subordinates
who responded anonymously. Robinson (1995) investigated how anonymity influences

students’ teacher rating. The results suggest that students’ ratings of teaching

18

performance suffer from significant validity problems and that performance indicators are
likely to provide misleading information. That is, the results indicate that in students’
teacher performance ratings differed more when they responded anonymously than when
they responded in the identifiable condition. Hence, it seems that anonymity could create
more aggressive behavior, make people more critical, raise self-disclosure, make people
honest, and so forth. Looking at anonymity in association with CMC technologies, Zhao
(1998) points out that regardless of the participant’s knowledge of each other’s identity
and personal acquaintances, the electronic medium deindividualizes the communication,
rendering it more socially and psychologically anonymous.

Given a context for teaching and learning, in light of the previous studies of
anonymity on psychological and behavioral consequences, Zhao (1998) explored the
effects of anonymity on CMC collaborative learning. He examined whether anonymity
would enhance the critical nature of student peer review in email. He found that
anonymity encouraged participants to be more critical during peer review, yet it led them
to work less. Anonymity encouraged learners to focus more on fellow students’ texts
while engaged in peer editing sessions, but yielded less helpful and lower quality reviews
than reviews in the identifiable condition.

The Purpose of the Present Study

The focus of the present study is to explore the potential of the CMC environment for
language acquisition. In order to explore this, there are issues that should be taken into
account. Whether the CMC technologies, which make it possible to control features such
as anonymity according to purposes, influence the linguistic environment should be

considered. If so, it is important to consider whether the manipulated feature brings any

19

beneficial or harmful outcomes for language learning. Motivated by the previous
literature, I chose anonymity as the feature for manipulation in the CMC environment.
Negative feedback, one of the important components of the linguistic environment in
SLA, is chosen as one of the dependent variables. Reaction to negative feedback is the
other dependent variable.
Method

Research Questions

In order to investigate whether and how anonymity influences conversational
interactions between native speakers and learners of English, the following research
questions were examined.
1. Does anonymity facilitate more negative feedback from a native speaker of English to

an L2 learner?
2. Does anonymity facilitate more reactions from an L2 learner to negative feedback?
3. Do the characteristics of negative feedback and reactions to negative feedback differ

according to anonymity?
Variables and Hypotheses

As discussed above, many researchers in second language acquisition agree that
interaction facilitates learning (Long, 1996; Gass, 1997; Pica, 1994; Mackey, 1999; and
others). Also, there are many studies investigating the effects of interactional feedback
(Gass & Varonis, 1994; Mackey & Philp, 1998; Lyster & Ranta, 1997). Because CMC
technologies allow us to choose the degree to which users’ identities are revealed to one
another (Smith et al., 2003), the question is whether and how anonymity influences NS-

NNS interaction.

20

The purpose of the present study was to investigate the influence of anonymity
on feedback. Thus, two conditions were examined: the anonymous condition and the
identifiable condition. Smith et al. (2003) state that synchronous CMC technology can be
realized in multiple forms and shapes and they should be considered from the viewpoint
of its own individual characteristics and their combinations. The different characteristics
delineated by Smith et al. (2003) are: temporality, anonymity, modality, and spatiality. In
the present study, in order for anonymity to be salient, the other three characteristics were
controlled in the two conditions. That is, in terms of temporality, in both conditions
synchronous online chat was carried out. In terms of modality, participants used the text-
chat function in both conditions. With regard to spatiality, unlike text-based chat
program, where participants do not have any control over spatiality, graphical chat
programs allow participants to control the spatial arrangement of the settings, their
avatars, and other objects (Smith et al., 2003). Avatars are computer-graphic visual
representations of people who are using chat software. Avatars can communicate with
each other by gesturing as well as by chatting. Two programs were chosen for the present
study: a graphical chat program with a video function, where participants can see their
partners’ faces, and one with an avatar function, where participants can not see their
partners’ faces, but see their partners’ avatars. Thus, participants can manipulate spatial
distances between themselves and their partners either with the avatar function by
moving their avatar toward or away fi'om their partner’s avatar, or with video by getting
their face closer to or farther from the Webcamera. Anonymity was determined by
whether they can identify their partners’ face. In the anonymous condition, participants

were not able to see their partner’s face in the screen, while in the identifiable condition

21

they were able to see their partner in it. They were asked to use pseudonyms during the
experiments in both conditions. As discussed, many previous studies of CMC
technologies on anonymity agree that anonymity in CMC might allow a potential
freedom from prejudice (Herring, 1996), less risk and less pressure (Smith et al., 2003),
and more distance between participants (Kern, 1998). Zhao (1998) investigated the
effects of anonymity during peer review. He found that anonymity could be a “double-
edged sword.” Anonymity encouraged participants to be more critical during peer
reviews, but they worked less. The participants focused more on students’ journals, but
yielded less helpful reviews in the anonymous condition than the identifiable condition.
Based on the previous literature, the hypotheses to the research questions were
established.

Since in the anonymous condition, native speakers do not have to see their partners’
emotional reactions, they might feel more comfortable providing negative feedback.
Thus,

Hypothesis 1: Native speakers provide more negative feedback in the

anonymous condition than in the identifiable condition.
Also, when nonnative speakers are provided negative feedback in the anonymous
condition, it might be likely that they do not feel a responsibility to react to their native
speaker partners. Similarly, it might be possible that nonnative speakers feel responsible
to react to the negative feedback given by their native speaker partners in the identifiable
condition. Thus,

Hypothesis 2: Normative speakers provide fewer reactions to the negative

feedback in the anonymous condition than the identifiable condition.

22

Furthermore, given that anonymity helps separate the message from its messenger,
objectify knowledge and make it easier to criticize, and also creates a situation in which
people could work less, based on the previous literature, a third hypothesis might be
proposed.
Hypothesis 3: The characteristics of negative feedback and reactions differ
between the anonymous condition and the identifiable condition.

Experimental Design Overview

The data collection for the present study was conducted in two locations, following
the same procedure3. The procedure consists of three stages: The training session,
Experiment One, and Experiment Two. Ahnost all participants had experienced online
chat. In the training session, participants were introduced to the interfaces of two online
chat programs, MSN Messenger version 6.1 and Activeworlds, learning how to use both.
This training session took between thirty minutes to one hour. These chat programs were
chosen because they are different in terms of anonymity. Activeworlds, which is a three
dimensional chat program (See Figure 2 in Materials for a screen shot), was used for the
anonymous condition because one cannot see one’s chat partner’s face in the screen,
while MSN Messenger version 6.1 was used for the identifiable condition for its
Webcamera function (See Figure 1 in Materials), where one can see one’s partner’s facial
expressions in the screen. The participants learned how to use the functions of these
programs and how to save the chat data. In the two experiments, participants were
randomly assigned to form NS-NNS dyads. All participants had different dyad partners in
these experiments. For example, a dyad of participants who carried out a task in

Experiment One had a different partner than the participants in Experiment Two. A two-

23

way communication task was used in this study. Each participant of a dyad had a picture
(See Appendix) that was similar to his/her partner’s picture but slightly different. There
were ten differences between their pictures. The dyads worked together to identify the
differences between their pictures through online chatting. The present study was
conducted in a within-participant design, where all the participants experienced two
different conditions. All participants carried out tasks in two different conditions: the
anonymous condition, and the identifiable condition. The tasks were counterbalanced,
half of the dyads completed a task in an anonymous condition first, and the other half
completed in an identifiable condition first. Also, each participant used a different task
picture in the two experiments in counterbalanced order. Each session lasted

approximately 50-60 minutes. The chart below illustrates the overall procedure of the

 

 

 

 

 

 

 

 

present study.

Experiments

Training Session Experiment One Experiment Two

Introduce CMC interfaces Anonymous Identifiable
Identifiable Anonymous

Participants

The participants in the present study were native speakers of English and normative
speakers of English. Both native speakers and normative speakers were recruited and
participated in two locations: Japan and the United States. Both the participants in Japan
and those in the United States volunteered to participate in the present study". All

nonnative speakers were classified as having lower-intermediate and intermediate

24

 

proficiency levels in English by their language programs. The native speakers and
normative speakers who participated in the present study had never met each other either
online or in person before the experiments.
. Participants in Japan

The nonnative speakers of English (N =6) were first-year Japanese students who were
studying at the same college in Japan. Five were male and one was female. The average
age was 20.0 years old (ranging from 18 to 22 years old). They were taking the same
English course. The native speakers of English (N=6) were all from the Japan Exchange
and Teaching (JET) program and were teaching English at junior high and/or high school
in Japan from six months to one year and a half. Two were male and four were female.
The average age was 26.25 years old (ranging from 24 to 32 years old). Of the total
number of dyads, seven dyads were mixed-sex dyads.

Participants in the United States

Normative speakers of English (N =1 0) were Korean students who enrolled in the same
level of an intensive English program at a university in the United States. Four were male
and six were female. Based on some students’ ages that were available, the average age
was around 21 years old. The native speakers of English (N=10) were third-year and
fourth-year students who were studying at the same university as the normative speakers.
Six were male and four were female. The average age of the native speaker participants
was 21.3 years old (ranging from 19 to 23 years old). With regard to the sex of the dyads,

of the total of twenty dyads, ll dyads were same-sex dyads.

25

Task Design

In order to investigate how anonymity influences negative feedback, two conditions
were established: the anonymous condition, and the identifiable condition. A within-
participant design was chosen in both contexts. Participants were assigned to form NS-
NNS dyads, participated in a task under one of the two conditions, and they were
assigned another dyad partner to complete the second task under the other condition. Two
pictures were used for data elicitation. Conditions and pictures were counterbalanced in
both contexts.

Tasks

In order to compare the amount and nature of negative feedback from native speakers
of English in two conditions, the present study was carried out in a task-based design.
Pica, Kanagy, and Falodun (1993) mentioned two important features for a ‘task.’ The
first is that tasks are oriented toward goals, and the second feature is work or activity. In
terms of learner production, particularly modification on interlanguage in response to NS
feedback, Pica et al. (1993) say that the task requires learners and NS interlocutors to take
turns drawing and describing pictures for each other to replicate, such as the information
gap task, have been useful. Also, the jigsaw task should elicit the most negotiation of
meaning. The jigsaw task is also a two-way communication task, where interactants hold
portions of a totality of information that must be exchanged and manipulated toward a
single task goal.

I chose a two-way communication task for NS-NNS dyads. “Spot the difference,” a
problem-solving task, which carried two features Pica et al. mentioned, was selected as a

task for this study. In “spot the difference,” each interactant of a dyad has a picture that is

26

similar to his or her partner’s picture but slightly different. They work together to identify
the differences between their pictures through interaction. In this task, interactants have
shared access to the information needed for task completion.

The present study investigated negative feedback from NSs of English, rather
than focusing on specific features of interaction, such as negotiation of meaning or
learner output, a problem-solving task. “Spot the difference” was chosen for a feasible
data collection method. Two pictures (See Appendix) were used for the tasks5 . The
pictures used for this study were “Train station,” and “Gas station,” taken from Ur
(1998). Participants were asked to do the task through only text-chat, and they were not
allowed to see each other’s picture.

Materials

Two chat programs were used for the present study for two different conditions: The
identifiable condition and the anonymous condition. As discussed earlier, there are four
features in CMC: anonymity, temporality, modality, and spatiality. In order to keep the
anonymity feature salient, the other three features should be as similar as possible. MSN
Messenger version 6.1 and Activeworlds were selected to satisfy this.

Chat Sofiware I : MSN Messenger version 6.]

MSN Messenger version 6.1 was chosen as chat software for the identifiable condition.
MSN Messenger version 6.1 can be used with a Webcam and it allows users to see each
other in a small video window while chatting (See Figure 1). According to Cziko and
Park (2003), who reviewed six free programs for synchronous audio and video
communication via the Internet, MSN Messenger was among the best of the programs

reviewed, and permits senders to monitor their own video as well as their correspondent’s

27

video in a screen. Although it also allows the audio function, the audio was not used
during the present study, but a Webcam function and a text-chat function were used. In
the interface, there is a chat window in the center where you can see the chat
conversation between you and your partner. To the right of the chat window, you can see
your partner’s face in the small video window. Below the chat window, there is a text
window, where you can type what you want to say. To the right of it, there is also a small

video window, in which you can see your face.

     

 

Emit" i: ‘5“ “-343
were) IND flirts) y—Mp me
g” a a: a e . . , M“
0 MIME WNW Ififfil‘QD 'f‘LitFTJG)
“.Iflhmm-fllu- u

_ a
Ewihmfiohotnoileom t Wake-n out! CEMLSU}. infirm». *r‘Jflll (Ame) Illul‘UEE
U-

n N w h, .0... n

aunhmaomimucon mums Wake-m E¢lhmb§o

Bantams-homemaker“ 0 Woman oifl Eilfiltbt. illmfll. ‘FV‘JEIL (Nico) IILUEE
U.

.ulh'n33fltlhcllullcom l3 Wobclnl can maintainer.

 

InlhnflskOholmcilcon mantra) Wobcun E$lhb$bto

.w'flhnaaflhctnailcan i Wobcam 0.771 CEMWM. “BEND”. $192» (Aha) filletz
U-

 

n . .... - ,, ..n w L. ..-— ----'---' - airman awe) L311!" “the. H
in MIND) bath,
Err-mun was: mainstream.

“withhaaelfihotnmlcom [I Nobel!» 0&5: AU) Ilfifiifllkbk-

Await» :pv Irma) -
l

  

 

Figure I . MSN Messenger version 6.1 interface

Chat Software 2: Activeworlds
Software selected for the anonymous condition was Activeworlds, which is
commercially available online chat software (www.activeworlds.com). This software

allows one to view three-dimensional worlds where there are three-dimensional objects

28

and avatars available (See Figure 2). With avatars, they can turn, wave, jump, spin, etc...
They can choose any avatars they want to be, such as a female, a male, or a bird, among
more than 10 other avatars. They can move toward or away from their partners. On their
right side of the screen, there is a window for displaying World Wide Web (WWW)
pages with Internet Explorer, which enables one to ask for help online. There is a 3-D
window in the center, below which there is a chat window. Users can communicate to
multiple people in the environment, or individuals. All the chat conversations appear in
the chat window below the 3-D window. For the present study, I created 10 worlds. Each

dyad was assigned a world to enter with passwords.

    

  

 

Ifiila Ielemt 301m Qatims 3m Lmh‘fivatu visually nah Hub

«4 Mia am to a. [:0]
ME]-

 
  

 

  
   
  
 
  
 
  
   

             
     
   
 
 
 

 

U:
KLDK 1
LHW 0 a.
test 0
aunt] a
grab] a
@WHJU 0 Need a little help? Looking fii
33W”? 0 technical question? Let us if-..
fluid} 0 ,
amend! 0 Activeworlds provides a was
an“); 0 cu to draw from. Regardles
“maj 0 customgrfrzssistanc: rglqulijre
., , , covere m searc a e a‘
33:; 3 [natural lon 0f:|::;:at “:39 yourself at hale! Have experienced Technical Supp
uwldfi 0 [u m] : ilel Io Yoko Hanxul Iolcou to ISll's w, encourage you m use 0.
Kl ICII Mr I d ' online customer service rest
5 Support Department, as cha
5 reaches our inboxl ,
i- w l [2]

 

 

 

 
    

 

 

 

Figure 2. Activeworlds interface

29

Procedures

The first experiments were conducted in Japan. The researcher met participants and
had an approximately one-hour training session with them before the experiments.
Participants were introduced to the CMC interface, the functions of a Webcamera, how to
save the chat data, the task and so forth. Participants were given two pictures in different
envelopes. They were asked not to look at the pictures until the experiments. On the
experiment day, NS participants met in a lab and NNS participants were at their home
with their home computers. They completed two tasks with different dyad partners and
with different chat programs in a counterbalanced design. They sent the chat data they
saved to the researcher by email.

The experiments in the United States followed next. All the participants used their
home computers this time. They also had a one-hour online training session with the
researcher before the experiments. The experiments in the United States was not
conducted in a day, instead it was done in the duration of a month. Also, they completed
two tasks with different dyad partners and with different chat programs in a
counterbalanced design. They sent the chat data to the researcher by email.

Coding

Chat Scripts

Smith (2003) states that one of the advantages of computer-mediated communication
is that the “logging” capabilities of most CMC programs enables researchers to capture
and readily access interaction, in addition to that the computer as a data collection
instrument seems to be less intrusive in many ways than traditional procedures for

recording. As he mentions, all the chat conversations were collected electronically with a

30

logging function. The data were analyzed based on turns. Regarding turn-taking in
computer-mediated communication, most CMC systems are of a one-way nature, where
only one interactant at a time can travel a given channel. It is technically impossible for
the addressee to respond while the message is being written (Herring, 1999; Smith,
2001). Herring (1999) also found a high degree of disrupted adjacency of turns,
overlapping exchanges, and topic decay as features of computer-mediated
communication. She says that violations of sequential coherence are the rule rather than
the exception in CMC. One-way nature of CMC causes adjacency pairs of turns, because
the users create irrelevant messages. Based on Smith’s (2003) study, a turn was counted
each time there was a transfer of the “floor” from one participant to the other. The chat
data used in the present study consist of the first 80 turns from each dyad in the data
collected in Japan, and the first 100 turns from a dyad in the United States data, in order
to compare the data between two conditions across two contexts. The averages of the
numbers of turns are 97 in the Japan data, and 180 in the United States data. There seems
to be several possible reasons for the difference in the averages of the total turns between
the data collected in Japan and the United States. The proficiency levels of NNSs might
have affected the number of turns and the NS variables, such as age or teaching
experience, could also have affected it, since when the data compared, it seems that the
conversation went on faster in the United States data than in the Japan data. The
minimum number of turns for the task completion or within the time limit in the Japan
data was 81 turns, while the minimum number in the United States data was 102 turns.

Thus, considering the minimum numbers of turns for each context, 80 turn from the data

31

collected in Japan and 100 turns from the data collected in the United States comprised
the corpus. The corpus for the present study was comprised of 2960 turns in total.

Off-task Discourse

Following Smith (2001), off-task discourse was not included in the 80 turns in the
Japan data and 100 turns in the United States data. He found that off-task discourse,
which consisted of greetings, or small talk, was found at the beginning of most chat
scripts and after a dyad completed the task, and no negotiated interaction occurred during
off-task discourse. Although the present study was not investigating the negotiation
routine, in order to keep the contents of interaction similar in two conditions, off-task
discourse was not included in data.

Data Coding

I will discuss the coding procedures for the data concerning the first and the second
research questions. Research questions, 1 and 2, are presented again below.
1. Does anonymity facilitate more negative feedback from a native speaker of English to

an L2 learner?
2. Does anonymity facilitate more reactions from an L2 learner to negative feedback?

Mackey et al. (2003) investigated the effects of interlocutor type on provision and
incorporation of feedback in face-to-face communication. They established three possible
categories for turns of N83 and NNSs of English: Initial learner utterances, interlocutor
response to non-target-like learner utterances, and learner response to feedback. Iwasaki
and Oliver (2003) also used a similar categorization. In their study, they investigated
provision and incorporation of native speakers’ negative feedback to their nonnative

speaker interlocutors in interactions in online chat. They coded the data based on turns.

32

The data of the present study were also analyzed based on these previous studies, and the
following three categories were established.

Initial nonnative speaker turns NNS turns were categorized as either target-like or non-
target-like. As in the previous studies, only non-target-like turns were analyzed since the
purpose of the present study was to examine the amount and nature of negative feedback.
Target-like turns consist of turns that are acceptable for N83 of English. Turns including
surface spelling errors that obviously occurred by typing errors independent from NNSs’
phonological problems were considered as target-like. Also, self-corrections within the
same turn, or in the subsequent turn without negative feedback, were classified as target-
like turns. Non-target-like turns included one of the following features with the examples
below: (1) unacceptable linguistic elements, (2) typographical errors that were difficult to
distinguish from spelling, lexical, and grammatical errors, and (3) obviously incomplete
turns by normative speakers. As in the following example, incomplete turns, where words
or phrases were provided but it is difficult to interpret what the proposition of the turn is,
were often observed.

Non-target-like turn examples

(I) unacceptable linguistic element
NNS: there 3 bag

NS: yes, 3 bags.

(NS 14, NNS 14, Anonymous)

33

(2) typographical errors that are difficult to distinguish from spelling,
lexical, grammatical error
NNS: curely?
NS: Ya, it’s curly.
(NS 10, NNS 10, Identifiable)
(3) Obviously incomplete turn
NNS: that’s ok
NNS: okok!
NNS: a woman
NS: ok, then near the 3rd and 4h train area, there is a woman.

(NS 13, NNS 15, Identifiable)

Furthermore, after non-target-like turns were identified, the non-target-like turns were
classified as either non-target-like turns with or without opportunities for negative
feedback. A NNS turn, which is ending with an utterance including non-target-like forms,
is a non-target-like turn with a feedback opportunity. If a NNS kept the floor by changing
topics, or adding new information to the original non-target-like forms, it was considered
as a non-target-like turn without feedback opportunities.

Native speaker response to non-target-Iike turns Native speaker responses to the
normative speakers’ non-target-like turns were identified and classified as either ignoring
non-target-like turns, or providing negative feedback. Negative feedback in the present
study was comprised of either explicit feedback, or implicit feedback, including recasts,

confirmation checks, and clarification request. That is, if NS turns followed NNS non-

34

target-like turns by providing evidence that NNS turns were not acceptable, the turns
were coded as negative feedback. Unlike face-to-face communication, negative feedback
was not necessarily immediately following non-target-like turns, because of its disrupted
turn adjacency. If there was no evidence that nonnative speaker turns were unacceptable,
the turns were coded as no feedback. Below are the example of NS response with
negative feedback to non-target-like turn and that of NS response without negative
feedback to non-target-like turn..

(2) Example of NS response with negative feedback

Non-target—like NNS turn: NNS: two window.

Do you see the window side of the

clock?
NS response with negative NS: What do you mean?? There are
feedback: two windows under the clock and
next to the door.

(NS 6, NNS 4, Anonymous)
(3) Example of NS response without negative feedback
Non-target-like NNS turn: NNS: Are there two train in your
picture?
NS response with no NS: Yes.
feedback:
(NS 5, NNS 5, Identifiable)
Furthermore, NS feedback was categorized as either negative feedback with the

opportunity to produce modified output or one without the opportunity for modified

35

output. If a N8 let his or her partner take the floor after they give negative feedback, it is
counted as negative feedback with the opportunity to produce modified output, while if a
NS keeps the floor by changing topics, or continuing talking, it is considered as no
opportunity for modified output. Although CMC has the nature of disrupted turn
adjacency, the provision of opportunity was determined based on whether there was topic
change within a turn. In other words, when a NNS read a NS turn including negative
feedback, there is an opportunity for the NNS to react if the NS turn ended with negative
feedback. However, if the NS turn included different semantic units at the end in addition
to negative feedback, it does not seem that there is an opportunity for the NNS to react to
the NS response. Examples 4 and 5, which illustrate negative feedback with ad without
opportunity for NNS reactions, are as follows.

(4) Example of negative feedback with no opportunity

Non-target-like NNS turn: NNS: and he is wearing hat

Negative feedback with no NS: he might be wearing a hat but he’s got

opportunity: the dog in one hand on a leash and a

suitcase in the other.
(N81 1, NNS ll, Identifiable)

(5) Example of negative feedback with the opportunity

Non-target-like NNS turn: NNS: the light shape is circle?

Negative feedback with NS: what do you mean?

opportunity:

(NS 13, NNS 15, Identiifable)
Nonnative speaker reaction to feedback The NNS turns following the negative

feedback were coded whether the normative speaker reaction was responding to the

36

negative feedback by modifying their original non-target-like output, incorporating the

negative feedback, and noticing that their original forms were non-target-like, or not.

Unlike in face-to-face communication, NNS reactions to feedback do not have to appear

immediately after the negative feedback in CMC because of its nature. The delayed

reactions that could be considered as reactions to the negative feedback were also

counted as well.

(6) Example of the NNS reactions (Incorporation)

Non-target-like NNS turn: NNS:

NNS:

Negative feedback NS:

NNS reaction to negative NNS:

feedback

37

1 bottom, look like mother, and sun
they hold hand

OK, and the son?

mother is right, son is left

(NS 10, NNS 8, Identifiable)

(7) Example of the NNS reaction (noticing the non-target-like-ness)

Non-target-like NNS turn:

Negative feedback

NNS reaction to negative

feedback

NNS reaction to negative

feedback

NNS: woman hair is wave?
NS: hmm
I would have said 'curl
'curls'
NNS: girl hair color is black??
ahW
NS: Is it shoulder length hair?

NNS: thanks.

(NS 12, NNS 14, Identifiable)

Two English NS graduate students in the TESOL program were chosen as

coders. They had a training session together first, and then coded the data separately.

Despite the training session, the intercoder reliability was not high enough. We had a

retraining session, and they redid the data coding. In the second time, the intercoder

reliability reached a higher level. A sample of 25 % of the data was examined for

reliability. The intercoder reliability was 98.4% for non-target-like turns, 98.4 % for

interlocutor response to non-target-like turns, and 100% for nonnative speaker response.

When there were discrepancies, the third coder, the researcher, determined the category.

In the course of data coding, in determining the provision of opportunity in particular,

there were some problems, which resulted from the disrupted turn adjacency. .

Sometimes NNS utterances were difficult to determine whether they were continuing

38

talking or whether they were responding to the previous utterance of the native speaker.
Below are the examples of with and without opportunity for negative feedback.
(8) Example of the NNS turn with opportunity for negative feedback
NS: what do you mean forward?
NS: The car separates three parts.
NNS: Second parts have four pieces of luggage.
NNS: ok?
(NS 4, NNS 5, Anonymous)
(9) Example of the NNS turn without opportunity for negative feedback
NS: on the right is a woman
NS: he has black hair
NNS: she wear black skirt
NNS: ok
(NS 8, NNS 10, Anonymous)
As mentioned earlier, each of the examples above is considered as one turn in the present
study. The problem was whether or not “ok” in the first example and the second example
function similarly. Because of the nature of the CMC environment it is possible that you
are not sure what your partner is going to say or ask. It is possible to interpret “ok?” in
Example 8 as that NNS was confirming his or her information since a question mark is
provided, while in “ok” in Example 9 NNS was possibly responding to the former
question of NS since it does not include a question mark . Thus, although it is not
possible to determine the intention of participants, the researcher and coders discussed

and agreed that “ok?,” “right?,” and “yes?” were confirmation checks, whereas “ok,”

39

“right,” and “yes” were responding to the previous NS utterance. Therefore, the latter are
the turns without the opportunity for either negative feedback. This definition of the
opportunity provision was used for the NS turns as well.

Research question 3 concerns the characteristics of NS responses to non-target-like
turns and NNS reactions. The coding criteria for further examination were established.
First, NS responses to non-target-like turns were classified according to negative
feedback types. Negotiation strategies, such as clarification requests and confirmation
checks, and recasts are not mutually exclusive categories. For example, confirmation
checks, normally considered negotiation moves, can also contain recasts (Mackey et al.
2003). Thus, the following five categories were identified in the present study: explicit
negative feedback, recasts, confirmation checks, clarification requests, and a combination
of a recast and a confirmation check. Example 10 shows the combination type of a
confirmation check and a recast.

(10) Example of a combination type of a confirmation check and a recast
NNS: employee =man with black hat
NNS: he is inserting gas
NS: is the employee pumping the gas?
(N813, NNS 13, Anonymous)
What follows is the examination of NNS reactions to negative feedback. Following
Mackey and Philp (1998), whereby they examined responses to recasts, NNS reactions to
NS responses were categorized into four different categories: (a) continue, (b) repeat, (c)

modify, and ((1) other.

40

When NNSs do not modify their non-target-like forms, repeat their interlocutor’s
utterance, or show any reaction to their NS interlocutor’s negative feedback, but they
continue the task, it is considered as the “continue” type. Example 11 illustrates the
“continue” type of the NNS reaction. The NNS uttered a non-target-like form, “he also
wearing a coat,” which is followed by the NS’s recast. However, the NNS did not react to
the NS’s recast, but continued the conversation.

(I 1) Example of the “continue” type of the NNS reaction

NNS: he also wearing a coat

NS: yes he’s wearing a coat
NNS: mine is he only holding a stick
(NSlO, NNS 10, Identifiable)

When NNSs repeat the recast or the combination type of NS response, this is considered
as “repeat” type. Example 12 is taken from Mackey and Philp (1998). The NNS’s non-
target-like utterance received negative feedback from the NS, “Is she having ftm?” The
NNS repeated after the NS with target-like form, “having fun.” Below is the example of

the “recast” type of negative feedback.

41

(12) Example of the “repeat” type of the NNS reaction
NNS: oh (..) she go to the zoo and she is she fun?
NS: Is she=?
NNS: fun
NS: is she having fun?
NNS: having fun
NS: yeah yeah
(data from Mackey & Philp, 1998, p. 344)
Further, when NNSs modify their original non-target-like forms after NS response, this is
counted as the “modify” type. Example 13 is taken from the present study. The NNS’s
non-target-like form, “scart,” is corrected by the NS. The NNS noticed he/she caused
confusion on the NS by a spelling mistake, and modified his/her original non-target-like
form into target-like one, “skirt.”
(13) Example of the “modify” type of the NNS reaction
NNS: yeah, but the mother (left) wears black scart
NS: OK, mine doesn’t have a scarf. Good job. ’V‘ 6
NNS: nonono, I meah skart
skirt
hahaha
(NS 10, NNS 8, Anonymous)
Lastly, if some NNSs show any reactions, indicating their noticing their non-target-like-
ness of their original forms, this is also considered as reaction, and categorized into the

“other” type. Example 14 is taken from the present study. As seen, when the NNS’s

42

question, “can you see grandfather,” received negative feedback from the NS, the NNS
reacted by apologizing although he/she did not repeat or incorporate the original non-
target-like form immediately.
(14) Example of the “other” type of the NNS reaction
NNS: Can you see grandfather??
(2 turns)
NS: what do you mean when you say “can you see grandfather”?
NNS: sorry.
(NS 14, NNS 14, Anonymous)
For the intercoder reliability of analyses for Hypothesis 3, a sample of 25% of the data
was examined by two trained graduate students who have taken the second language
acquisition course. The intercoder reliability was 100%.
Results
Table 1 shows the descriptive analysis of the data collected from the total of 32 dyads
interactions. The means, standard deviations, and ranges for each group are presented.
A chi-square procedure was used to compare negative feedback and reactions to
negative feedback between the anonymous and the identifiable conditions. This method

investigates the relationship of cell probabilities in nominal data.

43

Table 1

Dyads in the identifiable and the anonymous conditions: Means, standard

deviations, and ranges in ()

 

 

 

Identifiable Anonymous
M SD M SD
Turns 150.3 149.9
Non-target-like turns 16.88 8.45 15.94 6.65
(6-34) (5-31)
Feedback
. . 14.25 6.88 13 5.1
opportunrtres
(4-27) (5-20)
Feedback 2.69 2.47 2.56 2.25
(0-9) (0-9)
Reaction
. . 1.88 2 1.81 1.47
opportunities
(0-3) (0-5)
Reactions 0.19 0.40 0.63 0.72
(0-1) (0-2)
Negative Feedback

Hypothesis I predicted that the larger amount of negative feedback from NSs would
be found in the anonymous condition than in the identifiable condition. Before comparing

the amount of feedback between two conditions, the environment for negative feedback

44

was examined. All non-target-like turns were examined, in terms of whether they

provided opportunities for negative feedback or not. Table 2 and Figure 3 illustrate the

proportion of non-target-like turns with and without opportunities for negative feedback.
Table 2

Non-target—Iike turns with and without opportunities for negative feedback

 

 

Identifiable Anonymous
Feedback opportunities 228 (84.4 %) 208 (81.6 %)
No opportunities 42 (15.6 %) 47 (18.4 %)

{(1, N=525) = 0.770, ns

 

 

‘
O
O

90
80
70
60
50
4O
30
20
1 0

 

Ill Feetback
opportunities
I No opportuninites

 

 

    

Percentage of nontargetlike turns

Identifiable Anonymous

 

 

 

Figure 3. Non-target-like turns with and without opportunities for negative feedback

As seen in Table 2, in comparison of the non-target-like turns with and without
opportunities for feedback between two conditions, there was no statistical difference in
terms of opportunities for negative feedback between two conditions. Thus, the

anonymity variable is independent from the opportunities for negative feedback.

45

Next, the NS responses to all non-target—like turns, that is, whether negative feedback
was provided to non-target-like turns, were tabulated (See Table 3 and Figure 4).

Table 3

Negative feedback to non-target-Iike production in the identifiable and the

 

 

anonymous dyads
Identifiable Anonymous
Feedback 43 (15.9 %) 41 (16.1%)
No fwdback 227 (84.1%) 214 (83.9%)

{(1, N=525)= 0.002, ns

 

 

‘
O
O

11] Feetback
I No feedrack

Percentage of nontargetlike turns

 

 

 

 

Figure 4. Provision of negative feedback to non-target-like production

As in Figure 4 and Table 3, the comparison between the identifiable and the anonymous
conditions illustrates that negative feedback was provided in both conditions, and that the
frequency is not different from each other. Instead, a very similar percentage of negative

feedback was observed. There was no significant difference between the identifiable and

46

the anonymous dyads in terms of provision of negative feedback. This means that in the
present study, anonymity did not influence the provision of negative feedback.

Table 4 and Figure 5 show the interactions between the opportunities for negative
feedback and the provision of feedback across two conditions. There was no statistical
difference in the interaction between negative feedback provision, and negative feedback
opportunities. The distribution is almost the same between two conditions.

Table 4

Provision of negative feedback in the case that opportunities were provided

 

 

Identifiable Anonymous
Feedback 35 (15.4 %) 30 (14.4 %)
No feedback 193 (84.6 %) 178 (85.6 %)

{(1, N=436) = 0.074, ns

 

 

100
90
80
70
60
50
40
30
20
10

ll] Feedrack
No feetback

 

Percentage of feedrack opportunities

Identifiable Anonymous

 

 

 

Figure 5. Provision of feedback in the case that opportunities were provided

47

Next, the NS behavior in the case that no opportunities for negative feedback were
provided was also examined. The relationship between no opportunities for negative
feedback and feedback provision is illustrated in Table 5 and Figure 6. Negative feedback
was provided in both conditions even though there were no opportunities for negative
feedback. Negative feedback was observed more frequently in the anonymous dyads,
although the difference between the two conditions was not statistically significant.
Nonetheless, the difference of the amounts of feedback between two conditions in
percentage is greater when there were not negative feedback opportunities than when
there were negative feedback opportunities. In comparison, when there were no negative
feedback opportunities, there was less than one fifth of non-target-like turns with
feedback occurring in the identifiable dyads, whereas there were more than one fifth with
feedback occurring in the anonymous dyads.

Table 5

Negative feedback provision to non-target-Iike turns without opportunities for

 

 

negative feedback
Identifiable Anonymous
Feedback 8 (19 %) 11 (23.4%)
No feedback 34 (81 %) 36 (76.6%)

12(1, N = 89) = 0.25, ns

 

48

 

0
00°C

[I] Feedreck
I No feerback

‘Nw#OIG~l@
@000

Percentage of no opportunities
0

 

O

Identifiable Anonymous

 

 

 

Figure 6. Negative feedback provision to non-target—like turns without opportunities

for negative feedback

Hypothesis 3 predicted that the characteristics of negative feedback would differ
between the identifiable dyads and the anonymous dyads. To explore the nature of
feedback between two conditions, negative feedback provided opportunities for NNS
reactions was examined (See Table 6 and Figure 7).

Table 6

Negative feedback with and without opportunities for NNS reactions

 

 

Identifiable Anonymous
Feedback with opportunities 30 (69.8 %) 29 (70.7 %)
Feedback without opportunities 13 (30.2 %) 12 (29.3 %)

x2 (1, N = 84) = 0.009, ns

 

49

 

    
 
  

III Feetback with
opportunities
I Feedrack without

opportunities

Percentage of all feedrack

 

Identifiable Anonymous

 

 

 

Figure 7. Negative feedback with and without opportunities for NNS reactions

Table 6 and Figure 7 illustrate that there is not a significant difference in the nature of
feedback between two conditions in terms of opportunities. Of all feedback, 69.8 % of
negative feedback in the identifiable dyads provided opportunities for NNS reactions.
70.7% of feedback provided opportunities for reactions in the anonymous dyads. There
was no statistical significance between these two conditions

The examination of negative feedback in terms of the amount of its provision and its
concomitant opportunities suggests that there are no significant differences between the
identifiable condition and the anonymous condition. However, in order to explore the
nature of negative feedback further, the linguistic environments of negative feedback
were analyzed from the framework of types of interactional moves.

In order to examine the nature of negative feedback more closely, the residual parts of
data that were not used for the previous analysis, were taken into consideration. With
regard to the database for the further analysis, all the data collected were analyzed, not

just the first 80 or 100 turns as discussed in the Method section. That is, more data were

50

added to the corpus than for the prior analysis of negative feedback. The corpus for the
prior analysis was comprised of 2960 turns involving the first 80 turns of the data
collected in Japan, and the first 100 turns in the United States data. For the analysis
within a framework of the interactional move types, the interactions beyond 80 turns in
the Japan data and beyond 100 turns in the United States data were also included, since
all the dependent variables in interaction were coded by coders and the intercoder
reliability reached a high level, as discussed in the Method section.

As illustrated in Table 7, all the negative feedback instances were classified into five
categories: explicit feedback, recasts, recasts and confirmation checks, confirmation
checks, and clarification requests. The primary negative feedback type was recasts for
both the identifiable and the anonymous conditions. There is a tendency observed in both
conditions that the recast type was most frequent. A combination of a recast and a
confirmation check is the second most frequent. This condition was followed by the
confirmation check type and the clarification request type. The explicit feedback type
occurred the least fiequently. In addition to the fact that the tendency was the same
between two conditions, the frequency of negative feedback in each type was also
similarly observed. Among all the feedback types, there were slightly more occurrences
in percentage for the explicit type and the recast type in the data of the anonymous dyads
than in the identifiable dyads, whereas there were slightly more occurrences for the
confirmation check type and the clarification request type in the identifiable dyads data. It
must be noted that the differences between the two conditions were noted in only a small
number of instances, especially for the explicit feedback (N=2, for each condition). The

distribution of negative feedback type is illustrated by the graph in Figure 8.

51

Table 7

Distribution of types of negative feedback

 

 

 

 

Identifiable Anonymous
Explicit fwdback 2 (3 %) 2 (4 %)
Recasts 42 (61 %) 36 (63 %)
Recastsandconfirmation checks 13(19 %) 11(19 %)
Confirmation checks 6 (9 %) 4 (7 %)
Clarification requests 6 (9 %) 4 (7%)
70

Percentage of feedback
N 01) -h 0|
0 O O O

.-
O

 

 

Ell Identifiable
ll Anonymous

 

 

Figure 8. Distribution of types of negative feedback

What follows is a further examination of the nature of negative feedback. Example 15

is the recast type, which is the most frequent type of feedback (more than 60 % of the

time), for both conditions. In Example 15, which is taken from the identifiable dyad, the

52

NNS did not provide an article before a noun, “center.” The NS reformulated the
sentence into a correct form.
(15) Example of the “recast” type of the NS response
NNS: Yes Yes. Mine, too.
Is there man in center?
walk with a stick.

NS: Yes. There is a man in the center. He is walking a dog and
walking to the left side of the picture. He is wearing a hat and
carrying a suitcase in his left hand.

(NS 2, NNS 2, Identifiable)
Example 16 below is also example of a recast, which is taken from an anonymous dyad.
(16) Example of the “recast” type of the NS response
NNS: the guy who is pumping gas have a mustache.
igyes
NS: yes he has a muctache
(NS 8, NNS 11, Anonymous)
The next frequent type was the combination. of a recast and a confirmation check. This
type occurred 19 % of the time in both conditions. In Example 17, taken fi'om the
identifiable condition, and Example 18 fi'om the anonymous condition, the NSs are

confirming their understanding, while reformulating the NNS utterances.

53

(17) Example of the “combination of a recast and a confirmation check” type

NNS: mom hold bag on left
NS: so the mom and kid both have a bag in their left hand?

(NS 9, NNS 9, Identifiable)

(18) Example of the “combination of a recast and a confirmation check” type
NNS: aright bot wears black jacket
yeah
NS: They are both wearing black jackets?
(NS 12, NNS 14, Anonymous)
The confirmation check type and the clarification request type follow the combination
type. The confirmation check type was with negative feedback occurring at 9% of the
time in the identifiable dyads, 7 % in the anonymous dyads. Example 19 illustrates the
case, where the NS is confirming his/her understanding to the NNS. In Example 20, in
the example the clarification request type, NNS was testing his/her hypothesis for a
lexical problem. The NS indicates that his/her non-understanding, though he/she noticed
what the NNS was trying to say after he/she provided negative feedback.
(1 9) Example of the “confirmation check” type
NNS: do u have line
NS: line?

(NS 7, NNS 7, Identifiable)

54

(20) Example of the “clarification request” type
NNS: I have sign.
like foster..poster
NS: ?
I'm not sure what you mean by 'foster poster'?
Oh, poster.
(NS 12, NNS 14, Identifiable)
Examples 21 through 24 illustrate the explicit feedback type. Only two instances for each
condition occurred in the explicit type of feedback. Since the number of tokens was the
same between two conditions, all the instances were examined. Example 21 and 22 are
from the identifiable dyads.
(21) Example of the explicit negative feedback type
NNS: The back of the fence
There are woods
NS: woods?
NNS: How can I explain?
NS: behind the fence +—

(NS 7, NNS 7, Anonymous)

55

(22) Example of the explicit negative feedback type
NNS: woman hair is wave?
NS: hmm
I would have said ‘curl 4—
‘curls’
(NS 12, NNS 15, Identifiable)
In Example 21, the explicit feedback from NS was requested by the NNS’s preceding
utterance, “How can I explain?” In other words, this explicit feedback occurred with the
NNS’ utterance triggering. On the other hand, the explicit feedback in Example 22 looks
more spontaneous, although it might be considered as a suggestion, rather than an overt
correction. Examples 23 and 24 were excerpts from the anonymous dyads.
(23) Example of the explicit negative feedback type
NNS: next #2 there are something??
in the back~
NS: you mean like, in the distance <—
yes
it looks like a white building

(NS 4, NNS 5, Anonymous)

56

(24) Example of the explicit negative feedback type
NNS: I am disappointing.
(4 turns)
NS: In English you say....I am disappointed. Not (ing)
I am disappointed. This is disappointing. Do you understand
(NS 14, NNS 14, Anonymous)
In contrast to Examples 21 and 22, Examples 23 and 24 can be considered as more overt
and direct. In Example 24, it is clear that the NS intentionally indicated what was
incorrect and provided the targetlike version for the NNS.
NNS Reactions to Negative Feedback

Hypothesis 2 predicted that fewer reactions fi'om NNSs would be found in the
anonymous condition than the identifiable condition. In the course of exploring the nature
of negative feedback, the environment for NNS reactions had been examined. That is,
whether NS’s negative feedback provided the opportunities for NNS reactions was
illustrated in Table 6 and Figure 7. As seen, there was no statistically significant
difference in terms of the opportunities for NNS reactions.

The frequency of NNS reactions to negative feedback is tabulated in Table 8 and
Figure 9. In the identifiable dyads, NNSs reacted to NS negative feedback 7.0 % of the
time, whereas NNSs reacted 24.4 % in the anonymous dyads. Since the number in a cell
was less than 5, Fisher’s exact test was used instead of Pearson Chi-Square test. Fisher’s
exact probability was 0.036. The difference between the identifiable and the anonymous

dyads was statistically significant.

57

Table 8

NNS reactions to negative feedback

 

 

Identifiable Anonymous
Reaction 3 (7.0%) 10 (24.4 %)
No reaction 40 (93.0%) 31 (75.6 %)

Fisher’ s exact probability = 0.036 < 0.05

 

 

1351 No reaction

Percentage of feedrack

 

 

Identifiable Anonymous

 

 

Figure 9. NNS reactions to negative feedback

The opportunities provided with negative feedback were used for NNS reactions were
examined (See Table 9 and Figure 10). While the opportunities for reactions were used
10 % in the identifiable dyads, 31 % of the opportunities were used in the anonymous
dyads. Fisher’s exact test was used. The Fisher’s exact probability was 0.057 (p > .05).
This result suggests that even though the probability of the difference is not significant, it
approaches significance. As Table 9 and Figure 10 illustrate, NNSs in the anonymous

dyads reacted more frequently than NNSs in the identifiable dyads when the

58

opportunities for reactions were provided. That is, NNSs in the anonymous dyads used
the opportunities more frequently than the identifiable dyads.
Table 9

NNS reactions in the case of opportunities provided

 

 

Identifiable Anonymous
Reaction 3 (10 %) 9 (31 %)
No reaction 27 (90 %) 20 (69 %)

Fisher’s exact probability = 0.057 > 0.05, ns

 

 

I No reaction

Percentage of reaction oppotunities

 

Identifiable Anonymous

 

 

 

Figure 10. NNS reactions in the case of opportunities provided

In contrast, the frequency of NNS reactions when no opportunities for reactions were
provided was compared between two conditions (See Table 10 and Figure 11). When
there were no opportunities for reactions, both groups of dyads did not repeat after their

NS interlocutor. There was not a significant difference between two conditions.

59

Table 10

NNS reactions to negative feedback with no opportunities provided

 

 

Identifiable Anonymous
Reaction 0 (0 %) 1 (0.4 %)
No reaction 240 (100 %) 225 (99.6 %)

Fisher’s exact probability =0.485 > 0.05, ns

 

 

..a
N
0

§

I Reaction
I No reaction

A
0

Percentage of feedback without
opportunltlee
'3 8 8

 

Identifiable Anonymous

 

 

 

Figure 11. NNS reactions to feedback with no opportunities provided

Hypothesis 3 concerns the characteristics of NNS reactions. To explore the nature of
NNS reactions, the reaction types were also analyzed7. The categories were continue,

repeat, modify, and other. The reaction types were tabulated in Table 11.

60

Table 11

 

 

 

Types of NNS reactions
Identifiable Anonymous
Continue 61 (89.7 %) 39 (76.5%)
Repeat 0 (0 %) 0 (0%)
Modify 6 (8.8 %) 8 (15.7%)
Other 1 (1.5 %) 4 (7.8%)

 

Table 11 includes one case, where NNS used two different types of reactions to one
negative feedback instance (See Example 25).
(25) Example of two reactions to one negative feedback type
NNS: Can you see grandfather??
(2 turns)
NS: what do you mean when you say “can you see grandfather”?
NNS: sorry. <—
N8: the people on mine are small, i can’t see any faces
NNS: mine is 12
NS: ok
NNS: in front of pictures, there is grandfather with stick~ *—
(NS 14, NNS 14, Anonymous)
Interestingly, there was no case for the repeat type observed. The most commonly
observed was the case that NNSs continued the conversation without reacting to negative

feedback from NSs. The next commonly observed reaction was the modify type. In

61

Example 26, the NNS modified his/her original utterance, “newt,” indicated by the NS
negative feedback.
(26) Example of the “modify” type of the NNS reaction
NNS: maybe in front of sign, there are 2people
the next 2people the newt 2people
right?

NS: in front of the sign that says TIMETABLE there are 2 people, (a man
and a woman, the man has one bag, the woman has a bag and a purse
newt?

NNS: next sorry
(NS 7, NS 7, Identifiable)
In one case in the identifiable dyads and four cases in the anonymous dyads, the NNS
showed that they noticed that their turns included non-target-like forms, but the NNS did
not modify their original utterances, for example saying by “sorry,” “ah~, thanks,” “I
want to say it” etc.(See Example 27).
(27) Example of the “other” type of the NNS reaction
NNS: I am disappointing.
(4 tur118)
NS: In English you say....I am disappointed. Not (ing)
I am disappointed. This is disappointing. Do you understand
NNS: Sorry.

(NS 4, NNS 5, Anonymous)

62

Next, the relationship between the types of negative feedback and NNS reactions were
examined (See Table 12).
Table 12

NNS reactions to each feedback type

 

 

 

Identifiable Anonymous
Explicit feedback 1 1
Recasts 1 2
Recast and confirmation check 0 3
Confirmation checks 2 2
Clarification requests 3 3

 

It could be noted that whereas a combination of a recast and a confirmation check type
did not induce NNS reactions in the identifiable dyads, it was generally frequent in the
anonymous condition, although this should be cautiously discussed because of its small
instances.

Table 13 and 14 illustrate the relationship between the types of negative feedback and

the types of NNS reactions.

63

Table 13

“Modifii ” type of reactions to negative feedback types

 

 

 

Identifiable Anonymous
Explicit feedback 0 0
Recasts 1 l
Recast and confirmation checks 0 2
Confirmation checks 2 2
Clarification requests 3 3

 

Table 14

“Other type ” of reactions to negative feedback types

 

 

Identifiable Anonymous
Explicit feedback 1 1
Recasts 0 1
Recast and confirmation checks 0 1
Confirmation checks 0 0
Clarification requests ‘ 0 1

 

As in Table 13, the distribution of NNS reactions to each type of negative feedback is not
very different between the two conditions, except that the combination type induced two
instances of NNS reactions in the anonymous dyads, while no instances were found in the

identifiable dyads. Table 14 indicates that the NNSs in the anonymous dyads showed

64

their noticing to negative feedback to more various types of negative feedback in the
present study.
Summary of Results

As can be seen in Table 1 through Table 14, and in Figure 3 through 11, there was no
statistical difference between the identifiable condition and the anonymous condition, in
terms of provision of negative feedback. In contrast, in terms of NNS reactions to
negative feedback, the difference between the identifiable condition and the anonymous
condition was statistically significant. Contrary to Hypothesis 2, the NNS reacted more
frequently in the anonymous condition. With regard to the nature of negative feedback,
the tendency for the frequent types of negative feedback was not very different, while
when the two conditions were compared, the frequency within negative feedback types
between two conditions was slightly different. The anonymous dyads outnumbered in the
explicit feedback type and the recast type, while the identifiable dyads outnumbered in
the confirmation check type and the clarification request type. With regard to the nature
of NNS reactions, NNS showed that they noticed non-target-like forms in their utterances
more in the anonymous condition than in the identifiable condition.

Discussion

One of the goals of the present study was to investigate whether anonymity influences
negative feedback in CMC. Anonymity in the present study was manipulated in terms of
face visibility; that is, whether one can see his/her interlocutor’s face during interaction
was controlled. The finding of the present study indicates that negative feedback to non-
target-like forms is provided during a task-based online chat activity both in the

identifiable condition and the anonymous condition. However, the amount of negative

65

feedback observed in the present study was rather small. The percentages of negative
feedback provided in the present study were 15.9% of the time in the identifiable
condition, and 16.1% of the time in the anonymous condition. In the previous literature,
Mackey et al. (2003) found that 46 % of non-target-Iike turns involved negative feedback,
Oliver (1995) observed that 61% of NNS error turns received implicit negative feedback,
and Oliver (2000) also found negative feedback 47 % of the time to non-target-like
utterances. These high percentages of negative feedback provision were found in face-to-
face communication. Given that face-to-face communication includes phonological
feedback in a large amount (Mackey et al., 2000), the low percentage of negative
feedback observed in the present study might not be too surprising, since the present
study of text-based online chat did not include phonological elements. However,
compared to the research on negative feedback in text-based CMC by Iwasaki and Oliver
(2003), where negative feedback was observed 21.58% of the time, the percentage found
in the present study is still small.

There are several possibilities to explain the low rate of negative feedback in the
present study. First, it would be possible that the task type might have influenced
negative feedback provision. Iwasaki and Oliver (2003) had their participants have
conducted “free online conversation,” while the present study adapted a two-way
communication task to compare the identifiable condition with the anonymous condition.
The difference might be explained by Pica et al.’s (1993) claim that although the opinion-
exchange task does not seem to provide opportunities for comprehension, feedback, and
modified production as much, opportunities may be provided more than in the other task

types. Smith (2003) investigated the effects of the tasks on negotiated interaction in CMC.

66

He compared the jigsaw task with the decision-making task in the CMC environment.
With the tasks manipulated in terms of lexical saliency, he found that the decision-
making task elicited more negotiated interaction, which is inconsistent with the previous
literature on face-to-face communication. Not many studies have been done on the task
effects in CMC, the task effects on negative feedback in CMC in particular. Thus, it
might not be clear yet if task types matter for negative feedback provision. This awaits
future research. Second, the language should be taken into consideration. The present
study was conducted in the contexts of English as a second language and a foreign
language, while Iwasaki and Oliver (2003’) used the context of Japanese as a foreign
language. The difference in provision of negative feedback to non-target-like forms might
be attributed to the nature of the language. Acceptability of non-target-like forms might
possibly differ across languages, negative feedback to which might also differ. Lastly, the
variables concerning participants should also be taken into consideration. For example,
gender, proficiency level of learners, and age of participants need to be considered.

With regard to the impact of the anonymity variable on the provision of negative
feedback between the identifiable condition and the anonymous condition, no statistical
difference was observed in the present study. Rather, the amount of negative feedback
between the two conditions was almost the same. This could indicate that NNSs possibly
receive a certain amount of negative feedback regardless of anonymity (i.e., face
visibility) in the task-based CMC, and that the amount of negative feedback does not
necessarily differ according to anonymity.

This does not support the prediction that there would be more negative feedback in the

anonymous condition. Although the previous research on anonymity proposes that

67

anonymity has features, such as reducing social responsibility and relieving one from
social pressure, and that anonymity presumably brings more aggressive behaviors, more
honest self-expression, and more criticism to others, it might not necessarily be that
anonymity helps provide more negative feedback, because anonymity may be
multidimensional (Holden, Magruder, Stein, Sitarenios, and Sheldon, 1999). Another
possibility is that since the electronic medium per se deindividualizes the communication,
making it more socially psychologically anonymous (Zhao, 1998 for a summary), the
anonymity manipulation in the present study (i.e., face visibility) might presumably not
have given a significant impact on negative feedback provision.

Findings concerning the nature of negative feedback seem interesting as well. The
results that show a relationship between opportunities for negative feedback and negative
feedback provision between the identifiable and the anonymous conditions suggest that
use of opportunities for negative feedback might not be categorical depending on
anonymity. In comparison to the provision of negative feedback when the opportunity
was provided and when the opportunity was not provided, there was no significant
difference between the identifiable and the anonymous conditions, in terms of use of
opportunities.

Next, in terms of types of negative feedback, as shown in the previous section,
negative feedback in the present study was found in two forms with five types stemming
fi'om them. In the present study, negative feedback is comprised of explicit negative
feedback and implicit negative feedback. Implicit negative feedback comprises three
groups: negotiation strategies, recasts, and a combination of a recast and one of the

negotiation strategies. Negotiation strategies include two types: confirmation checks and

68

clarification requests. A combination type consists of a recast and a confirmation check.
The results indicate that all types of negative feedback used for classification in the
present study were identified in both the identifiable and the anonymous condition. Also,
though the distributions were similar, there was a tendency for interactions in the
anonymous condition to emphasize explicit feedback and recasts and for interactions in
the identifiable conditions to make use of negotiation strategies when negative feedback
according to each type was compared.

In the course of analyzing the nature of negative feedback types, I explored the
relationship between negative feedback type and the anonymity variable. As shown in
Table 7 and Figure 8 in the previous section, with regard to negotiation strategies, both
confirmation checks and clarification requests were found more frequently in the
identifiable condition than in the anonymous condition. A combination type that includes
a confirmation check and a recast was equally frequent between the identifiable and the
anonymous condition. Recasts, another type of implicit negative feedback, and explicit
negative feedback were more frequent in the anonymous condition.

Although it is not possible to determine the reason for this from the data of the present
study, it could be said that the features of anonymity might have affected the
sociolinguistic factors of interaction. In the identifiable condition, the participants were
able to see each other. That is, NSs could possibly see their interlocutor’s facial
expressions and emotional reactions when they would provide negative feedback.
However, since the procedure of the present study did not provide the evidence that the
participants actually saw each other with video or saw their partner’s avatar during the

task, it might be necessary to investigate the eye placement of the participants during the

69

task-based CMC so as to examine how face visibility affects CMC interaction.
Negotiation strategies, such as confirmation checks and clarification requests, seem to
require more interlocutor involvement in the interaction, since negotiation strategies can
lead to elicit interlocutor output, while explicit negative feedback and recasts seem to
have a relatively one-way nature, and they seem to require less involvement therefore.
The face visibility might have helped NSs in the identifiable condition provide negative
feedback that sounds more interactive to the interlocutors, and face invisibility led NSs in
the anonymous condition to provide negative feedback that is less interactive. That is,
NSs might have felt secure providing negative feedback that is likely to elicit
interlocutor’s output when they see their interlocutor’s face.

I have argued that face visibility helps interactants be more involved. An interesting
example in the data from the identifiable condition supports this. There was a focus on
form that seems to be attributed somewhat to face visibility, as a result of which, explicit
feedback was provided by a NS. As seen in the previous section, there was the same
number of instances for the explicit negative feedback type between both the identifiable
and the anonymous conditions. However, when the environment of the interaction was
examined in depth, in the identifiable condition, the negative feedback was elicited by the

NNS’s behavior (See Example 28).

7O

(28) Example of the explicit negative feedback type
NS: but no pattern behind the bench fence?
NNS: yeb
NNS: the back of the fence

there are woods
NS: woods?

NNS: How can i explain?

NS: behind the fence «—
NNS: yeb
NNS: do u hve?

In Example 28, the NS’s confirmation check, “woods?,” triggered negotiation between
the NNS and the NS, which might have made the NNS realize that his/her turn included a
non-target-like form, although it is difficult to determine where he/she actually paid
attention from this data. The NNS, who seems to be noticing the gap, asked the NS for
feedback, instead of modifying or repeating his/her original non-target-like utterance. As
discussed before, there is a possibility that face visibility in the identifiable condition led
the NNS to use the negotiation strategies that require more interactants’ involvement,
which resulted in explicit negative feedback fiom the NS in this case. It might be noted
that, as with this case, although there is explicit negative feedback provided, the linguistic
environment should be taken into consideration.

Furthermore, how NSs view their negative feedback act might also be considered.
Although negative feedback in the present study was interpreted as the behavior that

might be promoted by the aggressive feature of anonymity, it is not necessarily true that

71

NSs perceive the negative feedback act as aggressive. Rather it might also be possible
that they consider the negative feedback act as helpful. Therefore, it is would be useful to
investigate how the NSs perceive the negative feedback act in the future research.

Thus far, given that there was no significant difference between the identifiable
condition and the anonymous condition in terms of negative feedback provision in the
present study, the results do not support the prediction that anonymity would facilitate
negative feedback. However, it is clear that negative feedback exists in CMC, though the
amount might not be as large as the previous literature suggests. Also, an analysis of the
nature of negative feedback could lead us to think that anonymity might have an impact
on the nature of negative feedback. There is room for future research with different task
types, languages, and different CMC interfaces. In addition, the interactants’ perception
on anonymity and its influence on their linguistic behavior could be explored in the
future.

Another goal of the present study was to investigate whether anonymity influences
NNSs’ reaction to negative feedback. Although it should be cautiously discussed because
of the small number of instances of NNS reactions, it is clear from the results of the
present study that NNSs used negative feedback in both the identifiable and the
anonymous conditions. In face-to-face communication, Mackey et al. (2003) found that
negative feedback with opportunities for modified output involved modified output 34 %
of the time, and Oliver observed 35 % (1995) and 33 % (2000). Although the rate is not
as high as that in face-to-face communication, the previous literature on interaction in

CMC also found a certain amount of reaction to negative feedback in CMC. NNS

72

reactions to negative feedback occurred 25.6 % of the time in CMC (Iwasaki & Oliver,
2003).

The result of the present study is quite interesting. The prediction that NNSs would
provide less reaction to negative feedback in the anonymous condition was not supported.
In comparison of occurrences of NNS reactions to negative feedback when opportunities
for NNS reactions were provided, in the identifiable condition NNS reactions occurred
10 % of the time. In contrast, in the anonymous condition the NNS reactions occurred
31 % of the time (p = 0.057 > 0.05)8.. Further, in comparison of the number of NNS
reactions to all negative feedback including both when opportunities were provided and
when opportunities were not provided, only 7 % of the time NNS reactions in the
identifiable condition happened, while reactions to negative feedback occurred 24.4 % of
the time in the anonymous condition, and the difference between the two conditions was
statistically significant (p = 0.036 < 0.05). There are a couple of possibilities for the
significant difference. First, it is possible that sociolinguistic factors might play a role.
The previous literature on CMC says CMC carries an anonymous feature, and there is
more learner participation in CMC (Kern, 1995). Face invisibility might presumably have
promoted this feature, which possibly resulted in more NNS reactions. In contrast, face
visibility in the identifiable condition might have provided nonverbal cues, such as facial
expressions, which possibly resulted in fewer NNS reactions. In other words, in the
anonymous condition, since NNSs knew that their NS interlocutors could not see their
reaction through the screen, they might have felt necessary to show their reactions
somehow, whereas in the identifiable condition, since interactants can see their partner

through the screen, NNSs did not have to worry about their NS interlocutor’s perception

73

of their reactions. Another possibility is that NNSs in the anonymous condition might
have paid attention to form more than NNSs in the identifiable condition, perhaps
because an anonymous feature in CMC, that is, face invisibility in this case, reduced
pressure for communication, and provided a more relaxed environment. However, these
explanations are only speculative. Future research on sociolinguistic factors is required to
determine the reasons.

There were also interesting findings in analyzing the nature of NNS reactions. In
examining the nature of reactions, 1 explored the types of negative feedback and the
following NNS reactions. Table 15 illustrates a comparison of NNS reactions according
to the negative feedback type.

Table 15. NNS reactions to each feedback (we

 

 

 

 

Identifiable Anonymous
Feedback type reaction / feedback ( % ) reaction / feedback ( % )

Explicit feedback 1 /2 (50 %) 1 /2 (50%)
Recasts 1 /42 (2.4%) 2 / 36 (5.6%)
A recast and 0/ l3 (0%) 3 / 11 (27.3%)
a confirmation check

Confirmation checks 2 / 6 (33.3%) 2 / 4 (50%)
Clarification requests 3 / 6 (50%) 3 / 4 (75%)

 

Although it should be discussed with caution because of the small number of instances,
as seen in Table 15, all the NNS reaction rates in the anonymous condition are higher

than those in the identifiable condition, except with explicit feedback. Among them, it is

74

noticeable that 3 out of 11 instances (27 .3%) of the combination type of negative
feedback resulted in NNS reactions in the anonymous condition, while none resulted in
NNS reactions in the identifiable condition. It is plausible that psychological factors
might interact with the type of negative feedback to which NNSs react. As discussed
before, I speculated that the numbers of NNS reactions to negative feedback differ
between the identifiable and the anonymous conditions, perhaps because anonymity, that
is, face invisibility, promotes more participation for NNS reactions. Given that NNS
reactions occurred more frequently to all types of negative feedback in the anonymous
condition in the present study, it might be that this feature of anonymity promoted NNSs’
reactions in the anonymous condition. In addition, given that the identifiable condition
did not lead reactions to a combination type, whose possible reaction could be yes/no
answer, it might be that face visibility discouraged the NNSs to react verbally since there
were nonverbal cues available through the screen in the identifiable condition, which
might have resulted in the fact that the NNSs did not pay attention to form.

Furthermore, it might be interesting to look at the types of NNS reactions closely.
NNSs reactions were categorized into either the “continue” type (no reaction), the
“repeat” type, the “modify” type, or “other.” As illustrated in Table 11 in the Results
section, in either the identifiable or the anonymous condition, the “continue” type of
reactions comprised the largest percent of all the reaction types. No “repeat” type was
found in both conditions, perhaps because interactants can see all the chat conversations
on the screen in text-based chat CMC so that they can reread anytime they want.
Repetitions of negative feedback might sound redundant during the task, where

interactants work for a clear goal. With regard to the “modify” type, comparing Tables 13

75

and 14, six out of seven instances (85.7 %) of the total reactions were categorized as the
“modify” type in the identifiable condition, while in the anonymous condition, eight
cases out of twelve (66.7 %) were of the “modify” type. One case out of seven, 14.3 %
of all reactions in the identifiable condition, and four cases out of 12, 33.3 % of all
reactions in the anonymous condition, involved “other”, which does not fit either
continue, repeat, or modify types (See Table 11). In this the “other” type, NNSs did not
modify but showed their attention to negative feedback provided (See Examples 29
through 33 below). Example 29 is the only example observed in the identifiable condition,
and 30 through 33 were found in the anonymous condition.
(29) Example of the “other” type of the NNS reaction
NNS: woman hair is wave?
NS: hmm
I would have said 'curl
'curls'
NNS: girl hair color is black??
ahW (—
NS: Is it shoulder length hair?

NNS: thanks. <—

(NS 12, NNS 14, Identifiable)
In Example 29, although interrupted by the disrupted turn adjacency, it seems the NNS
noticed that his/her original turn, “woman hair is wave?” was non-target-like when he/she
received negative feedback, “I would have said ‘curls’.” The NNS expressed thanks to

the NS’s negative feedback.

76

(30) Example of the “other” type of the NNS reaction
NNS: I am disappointing.
(3 “1318)
NS: In English you say....l am disappointed. Not (ing)
I am disappointed. This is disappointing. Do you
understand
NNS: Sorry. <—
(N S 4, NNS 5, Anonymous)
In Example 30, the NNS apologized when he/she received explicit negative feedback
from the NS.
(31) Example of the “other” type of the NNS reaction
NNS: post? I see porl
(4 turns)
NS: Yeah. I think I misunderstood before. You could see a
pole? Then there is no difference.
NNS: I ,m sorry speling miss m(_ _)m9 t—
(N S 6, NNS 4, Anonymous)
Example 31 illustrates that the NNS noticed his/her non-target-like form, “I see porl,”

was non-target-like, when negative feedback was provided and then he/ she apologized.

77

(32) Example of the “other” type of the NNS reaction

NNS:

NS:

NNS:

NS:

NS:

lefi side of enter

pump on the left

yes the pump on the left

right pump

do you have?

wait wait, lets start over

i'm confused

ok

sorry

it's ok

oh, ok, I missed it before, sorry,
I do se SALE

ok

to the lefi of the entrance

yes <—
thanks.

I want to say it..

AA

(NS 6, NNS 4, Anonymous)

In Example 32, negative feedback was provided to the original NNS non-target-like form,

“left side of enter” through negotiation of meaning. The NNS noticed the negative

feedback.

78

(33) Example of the “other” type and “modify” type of the NNS reactions
NNS: Can you see grandfather??
(2 turns)
NS: what do you mean when you say "can you see
grandfather"?
NNS: sorry. <—
(3 tumS)
NNS: in front of picture, there is grandfather with stick~ (—
(NS14, NNS 14, Anonymous)
In Example 33, the NNS first apologized when negative feedback was provided, then the
NNS tried to modify the original non-target-like form. As seen in the examples, the NNSs
reacted by expressing either thanks or apologies. The examples of “the other” type of
NNS reactions indicate that NNSs noticed the negative feedback. Thus, it is clear that in
the present study NNSs reacted to negative feedback by showing awareness, as well as by
modifying their original non-target-like forms.

One finding of the present study is that the forms of negative feedback provided in
both the identifiable and the anonymous conditions in task-based CMC are usable, and
they are actually used. This could partially satisfy the first two of the four requirements
necessary for negative feedback to play a role in language acquisition (Grimshaw &
Pinker, 1989; Pinker, 1989; Beck & Eubank, 1991). The four requirements are: (a)
Negative feedback should exists, (b) exist in a usable form, (c) be used, and (d) be
necessary for language acquisition. However, it should be noted that the percentages of

provision and use of negative feedback to all non-target-like turns found in the present

79

study is relatively small, compared to the results of Iwasaki and Oliver (2003) in CMC
and to the previous literature in face-to-face communication. Although according to
Levine (1963), concept learning was possible even when corrective feedback was
provided on fewer than 25% of trials, it is unclear whether the amount of negative
feedback in the present study can be proven to be beneficial for learning. Mackey, et al.
(2003) state:

If we establish that feedback can promote development, we then need to

study whether more feedback leads to more development, whether there is a

minimal level that is sufficient, and whether timing, quality, type, output

opportunities, and context of feedback are relevant factors, as we suspect they

are. (pp. 56-57)

In sum, the present study investigated whether anonymity influences negative
feedback and reactions to it in task-based CMC. Several interesting findings were
observed. First, NNSs can receive negative feedback in task-based CMC regardless of
anonymity. Second, there were no significant differences in terms of the relationship
between opportunities and negative feedback. Third, there was a significant difference
dependent on anonymity in terms of reactions to negative feedback. The anonymous
condition resulted in more frequent reactions. Thus, although the amount is not as large
as in the previous literature regarding face-to-face communication, the present study
suggests that negative feedback exists, it is usable, and is actually used regardless of
anonymity in task-based CMC. It is difficult to determine the reasons for the results of

the present study, yet there seems to be room for sociolinguistic factors to play a role.

80

Further research should investigate the effects of anonymity on negative feedback with

different variables such as age, task type, and language in the CMC environment.

Footnotes

'~ is often observed in chat data of Korean students. It signifies making the last sound of a word longer.

2 In the examination of the negotiation of meaning in CMC, Toyoda and Harrison (2002) based their
study on the model of negotiation of meaning established by Varonis and Gass (1985). The components of
the model of negotiation of meaning are a trigger, an indicator, a response, and a reaction to response.
Toyoda and Harrison examined the causes of triggers.

3The data collection was done in two locations. Since there was no statistical difference between the two
populations across variables, the data collected from the two populations comprise one corpus for the
present study. For logistical reasons, in the experiments in Japan, the native speakers’ experiments occurred
in a computer lab, while the normative speakers used their home computers. In the experiments in the
United States, both the native speakers and the normative speakers used their home computers.

4Although the participants were volunteers, they were compensated with cash incentives of 15 dollars in
the US, and 2,000 yen in Japan (around 15 dollars at the time of the research) when they completed the
research participation.

’A pilot study was conducted proceeding the present study. In the pilot study, two NS-NNS dyads
conducted a two-way communication task with five pictures, in order to examine which picture might
possibly elicit more non—target-like forms from the NNSs. In determining the pictures, the vocabulary used
during the task, and learner proficiency levels were also taken into consideration.

6" A is an emoticon favored by Asian people and some NSs in the present study, which signifies smiling.

7The analysis for the reaction types included all the reactions in the database, not just 80 or 100 turns.

“Oliver (1995) examined the provision and use of negative feedback according to the age of the learners
and the context of the interaction. Her finding was that in an initial analysis, NNSs incorporated just 9.9 %
of recasts. However, in post hoc analysis, where she was looking at whether the tokens without
opportunities for the use of negative feedback were excluded, 35 % of negative feedback involved NNSs
reactions. Given that the definitions of an opportunity in the present study are different than her study, the

percentage of reactions found in the anonymous condition in the present study is relatively high. In her

81

study, NNSs incorporation (35 %) was counted when given the opportunity and when it was appropriate to
do so. For example, if the NS recast had been in the form of a yes/no question, then the most appropriate
response for the NNS was simply yes or no. In the present study, an opportunity was considered given
when NSs turns ended with negative feedback to non-target-like turns without changing topics.

9 mL _)m is an emoticon favored by Japanese learners signifying apologizing.

82

APPENDIX

83

Appendix

Picture set 1

 

 

 

 

' s
1 _ I...
..-_. .- fl“ 1,-..
..ij1\' . a— .— .9. :1...— out
I I l I
I e .
=3: .33. '
Ow .-
' “9 e--
e O-
0..
U.“
~

 

  
  

Q ill J , a n
‘ “WmI-Ji

ii IEDUJIIUDU

if _.
‘IIIIIIHIIIIII

71—31
mun”.

 

‘JII .: a“ "Is-I

4W

  

   

Picture set 2

 

 

 

 

 

 

 

 

85

REFERENCES

86

References

Antonioni, D. (1994). The efl‘ects of feedback accountability on upward appraisal ratings.
Personal Psychology, 47, 349-356

Beauvois, M. H. (1992). Computer-assisted classroom discussion in the foreign language
classroom: Conversation in slow motion. Foreign Language Annals, 25, 455-464

Beck, M., & Eubank, L. (1991). Acquisition theory and experimental design: A critique
of Tomasello and Herron. Studies in Second Language Acquisition, 13, 73-76.

Blake, R. (2000). Computer mediated communication: A window on L2 Spanish

interlanguage. Language Learning and Technology, 4, 1, 120-136. Retrieved June 19,
2004, from http://llt.msu.edu/vol4numl/blake/default.html.

Bohannon, J. N., & Stanowicz, L. (1988). The issue of negative evidence: Adult
responses to children’s language errors. Developmental Psychology, 24, 684-689

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. New
York: Wiley.

Bump, J. (1990). Radical changes in class discussion using networked computers.
Computers and the Humanities, 24, 49-65

Chun, A. E., Day, R R., Chenoweth, N. A., and Luppescu, S. (1982). Errors, interaction,
and correction: A study of native-nonnative conversations. TESOL Quarterly, 16,
537-547

Chun, D. M. (1994). Using computer networking to facilitate the acquisition of
interactive competence. system, 22 (1), 17-31

Cziko, G A. (in press). Electronic tandem language learning (eTandem): A third approach
to second language learning for the let century. CALICO Journal

Cziko, G. A., & Park, S. (2003). Internet audio communication for second language
learning: a comparative review of six' programs. Language Learning & Technology.
7(1), 15-27. Retrieved June 19, 2004, from
http://llt.msu.edu/vol7num1/review1/default.html.

Demetras, M. J ., Post, K. N., & Snow, C. E. (1986). Feedback to first language learners:

The role of repetitions and clarification questions. Journal of Child Language, 13,
275-292

87

Doughty, C. & Varela, E. (1998). Commucative focus on form. In C. Doughty & J.

Williams (Eds). Focus on form in classroom second language acquisition
(pp.l 14-138). Cambridge: UK

Ellison, P. A., Govern, J. M., Petri, H. L., & Figler, M. H. (1995). Anonymity and
aggressive driving behavior: a field study. Journal of Social Behavior and Personality,
10, 1, 265-272

Farrar, M. J. (1992). Negative evidence and grammatical morpheme acquisition.
Developmental Psychology, 28, 90-98

Gass, S. M. (1997). Input, interaction and the second language learner. Mahwah, NJ:
Lawrence Erlbaum.

Gass, S. M., & Varonis, E. M. (1994). Input, interaction, and second language production.
Studies in Second Language Acquisition, 16, 283- 302

Gass, S. M., Mackey, A., & Pica, T. (1998). The role of input and interaction in second
language acquisition: Introduction to the special issue. Modern Language Journal, 82,
299-3 05

Gass, S. M., Svetics. 1., and Lemelin, S. (2003). Difl’erential effects of attention.
Language Learning, 53, 3, 497-545

Grimshaw, J ., & Pinker, S. (1989). Positive and negative evidence in language acquisition.
Behavioral and Brain Sciences, 12, 341-342

Herring, S. (1996). Computer-mediated communication: linguistic, social and
cross-cultural perspectives. Philadelphia, PA: John Benjamin Publishing Company.

Herring, S. (1999). Interactional coherence in CMC. Journal of Computer-Mediated
Communication, 4 (4). Retrieved June 19, 2004, from
http://www.ascusc.org/jcmc/vol4/issue4/herring.html.

Hirsh-Pasek, K., Treiman, R., & Schneiderman, M. (1984). Brown and Hanlon revisited:
Mothers’ sensitivity to ungrammatical forms. Journal of Child Language. 11, 81-88

Holden, R. R., Magruder, C. D., Stein, S. J., & Sitareinios, G (1999). The effects of

anonymity on the Holden Psychological Screening Inventory. Personality and
Individual Dtfilerences, 27, 737-742

Iwasaki, J., & Oliver, R. (2003). Chat-line interaction and negative feedback. Australian
Review of Applied Linguistics, 17, 60-73

88

Iwashita, N. (2003). Negative feedback and positive evidence in task-based interaction:
Difl‘erential efi‘ects on L2 development. Studies in Second Language Acquisition, 25,
1-36

Kern, R. (1995). Restructuring classroom interaction with networked computers: Effects

on quantity and characteristics of language production. Modern Language Journal, 79,
457-476

Kern, R. (1998). Technology, social interaction and FL literacy. In J. Muyskenns (Ed.),

New ways of learning and teaching: focus on technology and foreign language
education (pp. 57-92). Boston: Heinle & Heinle Publishers.

Kitade, K. (2000). L2 learners’ discourse and SLA theories in CMC: Collaborative
interaction in Internet chat. Computer Assisted Language Learning, 13, 2, 143-166

Leeman, J. (2003). Recasts and second language development: Beyond negative evidence.
Studies in Second Language Acquisition, 25, 37-63

Levine, M. (1963). Mediating processes in humans at the outset of discrimination
learning. Psychological Review, 70, 254-276

Long, M. H. (1996). The role of the linguistic environ+ment in second language
acquisition. In W. C. Ritchie & T. K. Bhatia (Eds), Handbook of language acquisition,
Vol. 2: Second language acquisition (pp. 413-468). New York: Academic Press.

Long, M. H. (2000). Focus on form in task-based language teaching. In Lambert, R. D.,
& Shohamy, E. (Eds), Language policy and pedagogy. Essays in honor of A. Ronald
Walton. (pp. 179-92). Amsterdam and Philadelphia: John Benjarnins.

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

Lyster, R., & Ranta, L. (1997). Corrective feedback and leaner uptake. Studies in Second
Language Acquisition, 19, 37-66.

Mackey, A. (1999). Input, interaction, and second language development: An empirical
study of question formation in ESL. Studies in Second Language Acquisition, 21,
557-587.

Mackey, A., & Philp, J. (1998). Conversational interaction and second language
development: Recasts, responses, and red herrings? Modern Language Journal, 82,
338-356

Mackey, A., Gass, S., & McDonough, K. (2000). How do learners perceive interactional
feedback? Studies in Second Language Acquisition, 22, 471-497.

89

Mackey, A., Oliver, R., & Leeman, J. (2003). Interactional input and the incorporation of
feedback: An exploration of NS-NNS and NNS-NNS adult and child dyads. Language
Learning, 53, 35-66.

Makkai, T. & McAllister, I. (1991). Measuring social indicators in opinion surveys: A
method to improve accuracy on sensitive questions. Social Indicators Research, 27,
169-186

Oliver, R. (1995). Negative feedback in child NS-NNS conversation. Studies in Second
Language Acquisition, 17, 459-481

Oliver, R. (2000). Age differences in negotiation and feedback in classroom and pairwork.
Language Learning, 50, 1 l9-151

Ortega, L. (1997). Process and outcomes in networked classroom interaction: Defining
the research agenda for L2 computer-assisted classroom discussion. Language
Learning and Technology, 1, 1, 82-93. Retrieved June 19, 2004, from
http://llt.msu.edu/vollnuml/Ortegal.

Pellettieri, J. (2000). Negotiation in cyberspace: The role of chatting in the development
of grammatical competence in the virtual foreign language classroom. In M.
Warschauer & R. Kern (Eds), Network-based language teaching: Concepts and
practice, (pp. 59-86). Cambridge, UK: Cambridge University Press.

Pica, T. (1994). Research on negotiation: What does it reveal about second language
learning conditions, processes, and outcomes? Language Learning, 44, 493-527

Pica, T., Kanagy, R., & Falodun, J. (1993). Choosing and using communication tasks for
second language instruction and research. In G Crookes & S. M. Gass (Eds), Tasksand
language learning theory and practice (pp. 9-34). Clevedon, England: Multilingual
Matters.

Pica, T., Lincoln-Porter, F., Panios, D., Linnell, J. (1996). Language Learners’ Interaction:
How does it address the input, output, and feedback needs of L2 learners? TESOL
Quarterly, 30, 59-84

Pinker, S. (1989). Learnability and cognition: The acquisition of argument structure.
Cambridge, MA: MIT Press.

Polio, C., & Gass, S. M. (1998). The role of interaction in native speaker comprehension
of normative speaker speech. Modern Language Journal, 82, 308-319

Robinson, D. L. (1995). A critical analysis of students’ teacher ratings. Personality and I
ndividual Dtfierences, l9, 6, 945-947

90

Smith, B. (2001). Taking students to task; Task-based computer-mediated communication
and negotiated interaction in the ESL classroom. Unpublished doctoral dissertation,
University of Arizona, Tucson.

Smith, B. (2003). Computer-mediated negotiated interaction: An expanded model.
Modern Language Journal, 87(1), 38-57.

Smith, B., Alvarez-Torres, M. J ., & Zhao, Y. (2003). Features of CMC technologies and
their impact on language learners’ online interaction. Computer in Human Behavior, 19,
703-729.

Spada, N. (1997). Form-focused instruction and second language acquisition: A review of
classroom and laboratory research. Language Teaching Abstracts, 30, 73-87

Spada, N., & Lightbown, P. M. (1993). Instruction and the development of questions in
L2 classrooms. Studies in Second language acquisition, 15, 205-224

Toyoda, E., & Harrison, R. (2002). Categorization of text chat communication between
learners and native speakers of Japanese. Language Learning and Technology, 6, 1,
82-99. Retrieved June 19, 2004, fi'om http://llt.msu.edu/vol6num1/toyoda/.

Ur, P. (1988). Grammar practice activities: a practical guideline for teachers. Cambridge,
UK: University Press.

Varonis, E. M., & Gass, S. M. (1985). Non-native/non-native conversations: A model for
negotiation of meaning. Applied Linguistics, 6, 71-90

Zhao, Y. (1998). The effects of anonymity on computer-mediated peer review.
International Journal of Educational Telecommunications, 4(4), 311-345.

91

   

1|JJJJJJJJIJJJJJJJJJJJJI