A Study of the Role of Metatalk in the Acquisition of

Second Language Grammar: An Empirical

Investigation of the Output Hypothesis

Wataru SUZUKI
Graduate School of Education, Tohoku University

Abstract

This article reports on the results of a study conducted with Japanese high school students
learning English as a foreign language (EFL). The participants completed a sentence
combination task in which they were required to use a relative pronoun to combine two
independent sentences into a coherent whole and then studied a relevant input. Learners
retrospective reports were analyzed with respect to Swains (1985, 1995, 1998) output hypothesis.
Results indicate (1) the subjects tend to produce metatalk regarding English language rather than
English grammar, (2) producing output enables learners to employ syntactic processing during the
subsequent input, and (3) the quantity of metatalk may be less important as an aid to the
development of L2 grammar knowledge than the quality of metatalk. The results are also
discussed in reference to the theoretical underpinnings of the output hypothesis and second
language learning.

1. Introduction

During the last two decades, a controversy has existed about the possible effects and limitations
of linguistic input and output with regard to second language (L2) learning. Krashen (1985) argues
that the cause of L2 acquisition is language input that is understood by learners; in other words, only
when learners are engaged in receiving comprehensible input language acquisition can occur. On the
other hand, Swain (1985) has proposed the output hypothesis and argues that comprehensible input is
necessary but insufficient, stressing the essential role of comprehensible output in L2 learning.
Swain (1995, 1998) proposed three functions of output in L2 learning; noticing, hypothesis
formation and testing (hypothesis testing), and metatalk. Initially, producing output gives language
learners an opportunity to notice that they do not know how to convey the meaning they intend.
Secondly, learners may utilize their output as a way to try out hypotheses about the structure of L2.
Lastly, as learners reflect on their own target language use, their output serves the metalinguistic
function of enabling them to control and internalize linguistic knowledge. Since it has been
hypothesized that metatalk, including the explicit statement of rules and the use of metalinguisitic
terminology, may be a primary source of L2 learning (Swain, 1998), this paper focuses on metatalk in
which learners reflect upon their own target language use.
Several researchers have presented empirical evidence that supports the output hypothesis (e.g.,
Izumi and Bigelow, 2000; Sugiyama, Itagaki, and Inose, 2002; Swain and Lapkin, 1995). Swain and
Lapkin (1995), using quantitative analysis of metatalk, showed that when learners face difficulties in
output-oriented tasks, they may be engaged in various cognitive processes that can result in the
internalization of new linguistic knowledge, or the consolidation of existing knowledge. The research
conducted by Izumi and Bigelow (2000) provides further support for the claims proposed by the
output hypothesis. These researchers, focusing on past hypothetical conditional sentences,
investigated whether output would alter the learners subsequent input processing and promote
linguistic development. On the basis of their findings, they suggested that extended opportunities to
produce output promote noticing of the mismatch between their own output and the target language
input. Sugiyama et al (2002) asked Japanese EFL learners to solve a scrambled sentence task, and to
report on how they solved the problem and on how they compared their own solutions with the correct
one, which was provided as explicit feedback. They suggest that EFL learners should, to a certain
extent, be knowledgeable about grammar before they are able to metatalk about their target language.
The primary concern, based on the consideration of the results of the prior theoretical and
empirical studies, is whether the use of metatalk helps Japanese EFL learners to solve grammar
exercise. In order to investigate the question in an empirical manner, this paper analyzes learner
metatalk about language and grammar while learners perform grammar exercises.

2. Study Methodology

2.1 Subjects
Subjects were recruited from a pool of students at a Japanese high school. They had all
experienced at least 4 years of formal English language education in Japan. Sixty-five students who
participated in the study completed all the treatments and the post-test.

2.2 Materials
Two types of relative clauses were selected; a sentence in which the relative clause has the same
subject in a main clause (SU), and a sentence in which relative clause has the same object in a main
clause (DO). According to Keenan and Comires (1977) Noun Phrase Accessibility Hierarchy, it is
hypothesized that it is easier to solve SU material than to solve DO material. Two examples and their
subsequent correct answers were designed and exemplified in the following.

For SU exercise
Task: We want a teacher. She can speak English and French well.
Correct answer: We want a teacher who can speak English and French well.
For DO exercise
Task: I like the girl. I happened to see her at the supermarket.
Correct answer: I like the girl who (whom) I happened to see at the supermarket.

2.3 Procedures
The experiment was carried out in three phases. First, subjects were asked to perform a sentence
combination task using either who or which (hereafter, task 1), and then to write, as fully and
precisely as they could, a retrospective report on how they performed the exercise. Second, the
subjects were asked to study four examples of similar sentence combinations, and then to reflect in
their writing on the relationship between task 1 and the examples presented. Finally, the subjects were
asked to perform another sentence combination task using who or which (hereafter, task 2). Each
subject worked on items that included the same relative pronoun in both tasks. Subjects were allowed
to take as much time as they wished. On average, they were able to finish all sections in less than 40
minutes.

2.4 Scoring and analysis

This study focused on 65 retrospective reports reflecting on how the subjects both solved the
sentence combination task (first report), and how they studied four examples (second report). The
metatalk was classified into three categories and a code was given to each category of the metatalk:
metatalk about Japanese language (JMT), metatalk about English language (EMT), and metatalk
about English grammar (GMT) (Itagaki, Sugiyama, and Kubota, in press). Five graduate students
(including the author) coded these categories. Below are the three following examples.
Metatalk about Japanese language (JMT) includes Japanese words, phrases, clauses or sentences
manifested in subjects' reports which are simple Japanese translations of English words, phrases,
clauses, or sentences in the target sentences. An example is Watashi-wa sonosyozyuzo-ga sukida (I
like the girl) in the following part of the report: Watashi-wa sonosyouzozto-ga sukida noimiwo
kanngaeta (I just wondered how I should translate I like the girl into Japanese).
Metatalk about English language (EMT) includes English words, phrases, clauses or sentences
manifested in the subjects' reports. Two examples are which and who in the following part of the
report: who-wa hitonotokidakara which-to kangaeta (I wondered if I should use which, because
who represented human).
Metatalk about English grammar (GMT) includes any grammatical term manifested in subjects'
reports (e.g., subject, tense, object, relative pronouns). An example is subject in the
following report: Who-ga syugo no yakuwari wo suru (I wondered that the role of who is
subject).

3. Results

3.1 Percentage of Correct Answers

As a preliminary analysis, mean percentages of correct answers were obtained for task 1 and task
2. Table 1 displays the percentage of the sentence combination in task 1 and task 2.

Table 1. Mean Percentages of Correct Answers for the Two Sentence Types
Sentence type Task 1 (%) Task 2 (%) Mean (%)
SU 57.6 60.6 59.1
DO 28.1 53.1 40.1
Mean (%) 43.1 56.9 50.0
Note: SU- Subject relative clauses, DO- Direct Object relative clause

As predicted (Keenan and Comire, 1977), the percentages of correct answers for SU were greater
in task 1 and task 2 (57.6% and 60.6%, respectively) than those of DO (28.1% and 53.1%,
respectively).

3.2 Amounts of Metatalk in the Subjects' Retrospective Reports

In order to further investigate learners metatalk, the subjects were divided into two groups; one
group who succeeded in task 1 (CG: correct group) and the other group who failed in task 1 (IG:
incorrect group). Figure 1 displays the CGs mean amount of metatalk in SU, whereas Figure 3
displays the CGs mean amount of metatalk in DO. Likewise, Figure 2 represents the IGs mean
amount of metatalk in SU, whereas Figure 4 represents IGs mean amount of metatalk in DO. There
are five major findings discussed below.
 Figure1 CG's Mean Amount of Metatalk in Figure 2 IG's Mean Amount of Metatalk in
SU SU
Mean Amount of Metatalk

Mean Amount of Metatalk

6
5
5
4 J MT
4 J MT
3 EMT
3 EMT
2 GMT
2 GMT
1 1
0 0
report1 report2 report1 report2

Figure 3 CG's Mean Amount of Metatalk in DO Figure 4 IG's Mean Amount of metatalk in DO

Mean Number of Metatalk

6
Mean Number of Metatalk

6
5 5
4 J MT 4 J MT
3 EMT 3 EMT
2 GMT 2 GMT
1 1
0 0
report1 report2 report1 report2

First, as illustrated in Figures 1 - 4, the greatest amount of metatalk occurred in EMT (metatalk
about English language). However, although the amount of EMT dropped considerably from report 1
to report 2, the amount of GMT (metatalk about English grammar) remained relatively stable.
Secondly, according to Figure 1 and Figure 2, in SUs report 1 the amount of metatalk did not depend
on the group; the difference in the amount of metatalk (CG: 8.63 and IG: 5.57) is not statistically
significant, which indicates that IG students could not succeed in their tasks, regardless of the amount
of metatalk. Thirdly, according to Figure 1 and Figure 2, in SUs report 2 the amount of metatalk did
not depend on the group; the difference in the amount of metatalk (CG: 3.74 and IG: 4.71) is not
statistically significant. However, the difference in accuracy in task 2 (CG: 84% and IG: 29%) is
statistically significant (t=3.66, p<0.01). This indicates that IG students could not succeed in their
tasks, regardless of the amount of metatalk. Fourthly, according to Figure 3 and Figure 4, in CGs
report 1 the amount of metatalk did not depend on the group; the difference in the amount of metatalk
(CG: 7.33 and IG: 8.00) is not statistically significant. In other words, although IG students
verbalized their language and grammar, it did not guarantee success in the task. Finally, according to
Figure 3 and Figure 4, in DOs report 2 the amount of metatalk depended on the group; the difference
in the amount of metatalk (CG: 1.22 and IG: 3.91) is statistically significant (t=3.02, p<0.01).
However, the difference in accuracy in task 2 (CG: 100% and IG: 35%) is also statistically significant
(t=3.97, p<0.01). In other words, although IG students verbalized their language and grammar much
more than CG students did, it did not guarantee success in the task.

4. Discussion and Conclusion

The five main results above will be discussed in conjunction with the research question described
earlier and with the theoretical underpinnings of the output hypothesis and second language learning.

4.1 Types and Amount of Metatalk

Judging from the results above, the subjects tended to produce metatalk about English language
rather than English grammar. This seems inconsistent with theoretical and empirical studies of the
output hypothesis which states that language output allows learners to engage in syntactic processing
rather than semantic processing. Krashen (1998) also criticizes the output hypothesis in that many of
the instances of metatalk in Swain and Lapkin (1995) are not grammatical but lexical. However, the
subjects tended to produce metatalk about grammar in report 2 much more than they did in report 1.
This indicates that producing output may enable learners to employ syntactic processing in the
forthcoming input, which is consistent with the previous studies of the output hypothesis (Izumi and
Bigelow, 2000). In other words, the activity of producing output in this experiment may prompt the
learners to consciously recognize some of their grammatical problems; it may bring to their attention
something they need to discover in relevant input (Qi and Lapkin, 2001). It is therefore suggested that
the process of producing output and receiving relevant input (e.g., feedback) may be fundamental in
facilitating L2 acquisition (Swain, 1998).
As demonstrated in the analysis above, the subjects in IG could verbalize their explicit and
metalinguisitic knowledge as much as the subjects in CG could, while they produced output and
received relevant input. There are a number of possible interpretations for this finding. One
interpretation could be that students with certain characteristics such as a higher level of motivation
tend to use more explicit and metalinguistic knowledge. Another interpretation could be that explicit
knowledge, which can be verbalized but also remains on the surface of their existing knowledge, may
be relatively useless in producing output. This type of knowledge, called inert knowledge, does not
actively interact with a learners existing knowledge (Bereiter, 2002). It is therefore suggested that the
quantity of metatalk may be less important as an aid to the development of L2 grammar knowledge
than the quality of metatalk. However, even if learner metatalk leads to the incorrect solution, there is
no doubt that new knowledge has been created through a search of their own existing knowledge. In
this sense, metatalk may be a source of L2 learning (Swain, 1998),

4.2 Cognitive Functions of Output

 How metatalk can be related to the two other functions of noticing and hypothesis testing should
be noted here. There are a number of possible interpretations. One interpretation could be that
metatalk includes the function of noticing and hypothesis testing (Itagaki et al, in press; Swain, 1998).
The majority of the metatalk produced during output in this study was concerned with subjects
searching for the right word (who or which). Theses subjects noticed the gap between what they
needed to use and what they were not sure of (Swain, 1998). In other words, the activity of producing
output in this experiment may have prompted the learners to consciously recognize some of their
linguistic problems. This may have triggered cognitive processes which generated linguistic
knowledge that was new for learners, or consolidated their own linguistic knowledge (Swain and
Lapkin, 1995). The learners then were trying out a hypothesis about the structure of the L2. The
learners tested the hypothesis against their own internalized knowledge. Accordingly, their ultimate
solutions must have been hypothetical ones (Itagaki et al, in press). Furthermore, since many
instances of the metatalk during receiving relevant input were grammatical, learners must have
reflected on their output while receiving relevant input. This reflection is based on the subjects
noticing and hypothesis testing. In other words, in metatalk, noticing and hypothesis testing may be
made available for inspection (Swain, 1999).
Another interpretation could be that different types of cognitive mechanisms may make different
functional contributions to linguistic output. In other words, each is more appropriate for some tasks
and conditions than for others, and none is superior in all situations. Noticing and hypothesis testing
seem to be more appropriate for producing output as process, whereas the metalinguisitic function is
more appropriate for reflecting on what is produced as product (Wells, 1999). Hence, for linguistic
knowledge to be internalized and restructured, noticing and hypothesis testing need to be triggered by
producing output (ongoing stage), and then learners must reflect, on the basis of what they have
noticed and hypothesized, in their output, on their existing knowledge while receiving feedback
(reflective stage). Further research should, by separating the two stages, investigate the potential
role of linguistic output in an empirical manner.

4.3. The Role of Metatalk in L2 Acquisition

It is hypothesized that metatalk, where students reflect consciously on the language they are
producing, may be a source of, or even an occasion for, second language learning (Swain, 1999).
However, recent SLA researchers have not agreed on the extent to which metatalk can help with L2
learning (Hu, 2002; Hulstjin, 2002). As can be seen in this study, it is assumed that the subjects,
regardless of whether they are IG or CG, have enough metalinguisitic knowledge to verbalize. In one
sense, metatalk cannot help L2 performance and competence (Krashen, 1985). There is, however, no
doubt that new knowledge has been created through a search of their existing knowledge. The present
paper hence assumes that metatalk facilitates L2 learning under some circumstances. It has been
suggested that metalinguistic knowledge is likely to result in a delayed effect rather than immediate
acquisition (Ellis, 2002), and that it is a form of knowledge to be used as a resource where and when
implicit knowledge is not available (Hulstjin, 2002). Although it has been argued that metalinguistic
knowledge merely plays a monitoring role in reading and writing (Krashen, 1985), research into the
effect of metatalk in an instructional context suggests that it will continue to have a high educational
value and priority, given the limited opportunities for receiving extensive input and producing output,
especially in Japan.
Linguistic knowledge is described as consisting of a number of skills, some involving formal
knowledge of structure and others involving communicative uses of language (Bialystok, 2001).
Consequently, the pedagogy defined as both grammatical knowledge and communicative knowledge
has been and continues to be an area of some controversy for SLA researchers. According to Swain
and Lapkin (1995), although knowledge building is a necessary activity carried out by an individual, it
has its purpose and its fullest realization in socially oriented creation (Wells, 2000). The empirical
data in this experiment does not discuss Swains recent studies (e.g., Swain, 2000). Her recent
research has focused on tasks as a stimulus for generating metatalk among students (e.g., Swain, 2001;
Swain and Lapkin, 1998; Lapkin, Swain, and Smith, 2002). Nevertheless, it is indicative of further
research of the output hypothesis. This study has shown that Japanese EFL learners could verbalize
their linguistic knowledge while performing a grammar exercise (Sugiyama et al, 2002). Metatalk
may assist L2 grammar learning in that a verbalized explanation may play a crucial role in the
acquiring and further understanding of L2 grammar (Swain, 1998, 1999). As this study indicates,
however, explicit knowledge that can be verbalized but remains inert does not guarantee success.
Grammar teaching should be, in both EFL and ESL (English as a second language) contexts, oriented
toward making grammatical knowledge deeply internalized into learners existing knowledge through
formal and communicative teaching.

Acknowledgement

I thank Nobuya Itagaki and Gordon Wells for their continuous assistance throughout the period I
worked on this paper, the students and teachers who participated in this study, Takatoshi Muramoto
and several anonymous reviewers for their many helpful suggestions. My thanks also go to Kiyomi
Hoshino and three anonymous graduate students for data analysis.

References
Bereiter, C. (2002). Education and mind in the knowledge age. Mahwah, NJ: Lawrence Erlbaum
Associates.
Bialystok, E. (2001). Bilingualism in development: Language, Literacy, and Cognition. Cambridge:
Cambridge University Press.
Ellis, R. (2002). Grammar teaching practice or consciousness-raising? In J.C. Richard and W.A.
Renandya (Eds.), Methodology in language teaching: An anthology of current practice (167-
177). Cambridge, UK: Cambridge University Press.
Hu, G. (2002). Psychological constraints on the utility of metalinguistic knowledge in second
language production. Studies in Second Language Acquisition, 24, 347-386.
Hulstjin, J. (2002). Towards unified account of the representation, processing, and acquisition of
second language knowledge. Second Language Research, 18, 193-223.
Itagaki, N., Sugiyama, M., and Kubota, Y. (in press). Analysis of metalinguisitic talk in grammar
exercise: on the basis of the theoretical framework of the output hypothesis.
Annual Review of English Language Education in Japan ,14.
Izumi, S., and Bigelow, M. (2000). Does output promote noticing and second language acquisition?
TESOL Quarterly, 34, 239-278.
Keenan, E., and Comire, B. (1977). Noun phrase accessibility and universal grammar. Linguistic
Inquiry, 8, 63-99.
Krashen, S.D. (1985). The input hypothesis: Issues and implications. London: Longman.
Krashen, S.D. (1998). Comprehensible output? System, 26, 175-182.

Lapkin, S., Swain, M., & Smith, M. (2002). Reformulation and the learning of French pronominal
verbs in a Canadian French immersion context. Modern Language Journal, 86, 485-507.
Qi, D.S. and Lapkin, S. (2001). Exploring the role of noticing in a three-stage second language writing
task. Journal of second language writing, 10, 277-303.
Sugiyama, M., Itagaki, N., and Inose, F. (2002). An empirical study of the output hypothesis:
"Metatalk," "Noticing," and "Hypothesis-testing." Bulletin of
Tohoku English Education Society 22, 87-99.
Swain, M. (1985). Communicative competence: Some roles of comprehensible input and
comprehensible output in its development. In S. Gass & C. Madden (Eds.), Input in second
language acquisition (235-253). Rowley, MA: Newbury House.
Swain, M. (1995). Three functions of output in second language learning. In G. Cook & B. Seidlhofer
(Eds.), Principle and practice in applied linguistics: Studies in honour of H.G. Widdowson (125-
144). Oxford: Oxford University Press.
Swain, M. (1998). Focus on form through conscious reflection. In C. Doughty & J. Williams (Eds.),
Focus on form in classroom second language acquisition (64-81). New York: Cambridge
 University Press.
Swain, M. (1999). Integrating language and content teaching through collaborative tasks. In C. Ward
& W. Renandya (Eds.), Language teaching: New insights for the language teacher (125-147).
Singapore: RELC. Reprinted in the Modern Language Teachers' Association of Western
Australian Newsletter, 3, 4-12.
Swain, M. (2000). The output hypothesis and beyond: Mediating acquisition through collaborative
dialogue. In P. Lantolf (Ed.), Sociocultural theory and second language learning (97-114).
Oxford: Oxford University Press.
Swain, M. (2001). Examining dialogue: Another approach to content specification and to validating
inferences drawn from test scores. Language Testing, 18, 319-346.
Swain, M., and Lapkin, S. (1995). Problems in output and the cognitive processes they generate: A
step towards second language learning. Applied Linguistics, 16, 371-391.
Swain, M., & Lapkin, S. (1998). Interaction and second language learning: Two adolescent French
immersion students working together. Modern Language Journal, 82, 320-337.
Wells, G. (1999). Dialogic inquiry: towards a sociocultural practice and theory of Education.
Cambridge: Cambridge University Press.
Wells, G. (2000). Dialogic inquiry in education: Building on the legacy of Vygotsky. In C.D. Lee and
P. Smagorinsky (Eds.) Vygotskian perspectives on literacy research (51-85). New York:
Cambridge University Press.