w.
. 7h.
. anew.
.Vw .

 

 

.
tan“...
Hungarian"

s2

3 w. ..
ugamuﬁ

ﬁuwmﬁﬁ

 

 

3 .

.é... ,2.

 

 

Lab

A Oi O LIBRARY
Michigan State
Unixersity

 

 

 

This is to certify that the
dissertation entitled

THE NOTICING AND EFFECT OF TEACHER FEEDBACK IN ESL
CLASSROOMS

presented by

WEIQING WANG

has been accepted towards fulﬁllment
of the requirements for the

Curriculum, Teaching, and

PhD. degree 'n Educational Policy

 

 

Li" a ,-
L -a- ~-—-~- 7

 

Major Professor's Signature
\
M" AL (1 g j: 29

Date

 

MSU is an Affirmative Action/Equal Opportunity Employer

 

 

 

 

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

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5108 K:IProilec&Pros/ClRC/DateDue.indd

 

 

THE NOTICING AND EFFECT OF TEACHER FEEDBACK IN ESL CLASSROOMS

By

Weiqing Wang

A DISSERTATION

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

DOCTOR OF PHILOSOPHY

Curriculum, Teaching, and Educational Policy

2009

 

ABSTRACT

THE NOTICIN G AND EFFECT OF TEACHER FEEDBACK IN ESL CLASSROOMS

By

Weiqing Wang

This study examined learners’ noticing of teacher feedback and their language
improvement from three aspects: the characteristics of feedback episodes, the
nonlinguistic cues in teacher feedback, and the metalanguage in teacher feedback. The
data came from classroom observations, stimulated recall interviews, and two tailor-made

individualized tests: an immediate test and a delayed test.

A total of nine classes were observed, with four to six observations in each class. Lessons
were both audio and video taped. Both the immediate test and stimulated recall interview
were carried out one to three days after the observation(s) from which they were
developed while the delayed test was administered 14 to 16 days after the observation(s)
from which it was developed. Altogether, there were 1434 feedback episodes, 279
stimulated recall comments, 272 test items in the immediate test, and 282 test items in the

delayed test.

The results suggest that all the seven characteristics of feedback episodes examined had a
statistically signiﬁcant effect on both the occurrence and successful uptake but only some

of them showed such an effect on the rate of teacher-feedback-related comments and

 

 

learners’ performance in the two tests. Some nonlinguistic cues in teacher feedback also
signiﬁcantly affected the occurrence and successfulness of uptake, teacher-feedback-
related comments, and test results but mostly at the general level. Most of the speciﬁc
types of nonlinguistic cues did not distinguish themselves from others. Metalanguage in
teacher feedback had some effect on learners’ noticing of the feedback and language

improvement too but to a lesser degree and not always in a positive way.

 

 

C0pyright by
Weiqing Wang
2009

 

ACKNOWLEDGEMENTS

This dissertation is impossible without the support from professors, friends, and my

family. I would like to express my deepest appreciation to:

Professor Lynn Fendler, my advisor, and Professor Shawn Loewen, the co-director of this
dissertation, for their careful guidance in every step I took, their detailed comments in
every draft I submitted, and their continuous assurance and encouragement during

difficult times.

My committee members Professor Tom Bird and Professor Charlene Polio for their

insightful challenges and incisive suggestions.

My colleagues Anny Fritzen, Wenxia Wang, and Joan Zou for spending their valuable

time coding part of my data for inter-rater reliability check.

The anonymous teachers and students for allowing me to observe their lessons and for

making room in their tight schedule to do interviews and to take tests with me.

My friends Shih-pei Chang, Hui J in, and Epimenio Torres Jr. for their encouraging words

and generous help in the whole study.

My family for their understanding and patience throughout this long strenuous process.

 

TABLE OF CONTENTS

LIST OF IGURES ............................................................................................................ xiv
KEY TO ACRONYMS ..................................................................................................... xv
CHAPTER 1
INTRODUCTION ............................................................................................................... 1
CHAPTER 2
LITERATURE REVIEW AND RESEARCH QUESTIONS .................................. 5
2.1 Introduction ........................................................................................................ 5
2.2 Mapping the big picture: input, output, and interaction ............................ 5
2.2.1 Input ......................................................................................................... 6
2.2.2 Output ...................................................................................................... 8
2.2.3 Interaction .............................................................................................. 11
2.3 Ipoking closer: feedback and the measurement of the effectiveness of
feedback ........................................................................................................... 12
2.3.1 The role of feedback in second language acquisition ..................... 12
2.3.2 The measurement of the effectiveness of feedback ........................ 13
2.3.2.1 Tailor—made individualized tests .................................... 14
2.3.2.2 Learner uptake ........................................................ 18
2.4 Looking in particular: noticing and the measurement of noticing ............... 20
2.4.1 Attention and noticing in second language acquisition ................... 21
2.4.2 The measurement of noticing ................................................ 23
2.4.2.1 Stimulated recall ...................................................... 24
2.4.2.2 Learner uptake ......................................................... 28
2.5 Pulling it together: factors that may affect leamers’ noticing of feedback and
the effectiveness of feedback ......................................................... 33
2.5.1 Characteristics of feedback ................................................... 33
2.5.2 Nonlinguistic cues in feedback ............................................... 37
2.5.3 Metalanguage in feedback ..................................................... 40
2.6 Summary ................................................................................ 42
2.7 Research questions ..................................................................... 44
CHAPTER 3
METHODOLOGY ................................................................................. 46
3.1 Introduction ............................................................................. 46
3.2 Teaching context ....................................................................... 46
3.3 Participants .............................................................................. 47
3.4 Instruments .............................................................................. 48

vi

3.4.1 Observations .................................................................... 49

3.4.2 Stimulated recall interviews .................................................. 50
3.4.3 Language testing ............................................................... 53
3.5 Procedures .............................................................................. 57
3.6 Data analysis ........................................................................... 60
3.6.1 The coding of observation data ............................................... 60
3.6.1.1 The coding of the characteristics of feedback episodes ......... 60
3.6.1.2 The coding of nonlinguistic cues ................................... 71
3.6.1.3 The coding of teachers’ metalanguage ............................ 84
3.6.1.4 The coding of the occurrence and successfulness of uptake. . ..86
3.6.2 The coding of stimulated recall comments ................................. 93
3.6.2.1 The coding of overall recall comments ............................ 93
3.6.2.2 The coding of language-related comments ........................ 96
3.6.2.3 The coding of target-feature-related comments .................. 98
3.6.3 The coding of test results .................................................... 100
3.6.4 The reliability of coding ...................................................... 101
3.6.5 Statistical analysis ............................................................ 102
3.7 Summary .............................................................................. 105
CHAPTER 4
CHARACTERISTICS OF FEEDBACK EPISODES, THE NOTICIN G AND THE
EFFECT OF TEACHER FEEDBACK ......................................................... 106
4.1 Introduction ........................................................................... 106
4.2 Distribution of feedback episodes by characteristics ............................ 106
4.3 Characteristics of feedback episodes and learner uptake ........................ 108
4.4 Characteristics of feedback episodes and recall comments ..................... 119
4.5 Characteristics of feedback episodes and test results ........................... 125
4.6 Review of results ..................................................................... 132
4.7 Discussion ............................................................................. 134
4.8 Summary .............................................................................. 154
CHAPTER 5
NONLINGUISTIC CUES, THE NOTICING AND THE EFFECT OF TEACHER
FEEDBACK ....................................................................................... 1 56
5.1 Introduction ........................................................................... 156
5.2 Occurrence and distribution of nonlinguistic cues ............................... 156
5.3 Nonlinguistic cues and learner uptake ............................................. 158
5.4 Nonlinguistic cues and recall comments .......................................... 165
5.5 Nonlinguistic cues and test results ................................................. 169
5.6 Reviewofresults ............. ........................ 174
5.7 Discussion ............................................................................. 176
5.8 Summary .............................................................................. 192

vii

CHAPTER 6
METALANGUAGE, THE NOTICING AND THE EFFECT OF TEACHER

FEEDBACK ........................................................................................ 194
6.1 Introduction ........................................................................... 194
6.2 Occurrence and distribution of metalanguage .................................... 194
6.3 Metalanguage and learner uptake .................................................. 197
6.4 Metalanguage and recall comments ................................................ 200
6.5 Metalanguage and test results ...................................................... 202
6.6 Review of results ..................................................................... 204
6.7 Discussion ............................................................................. 205
6.8 Summary .............................................................................. 214
CHAPTER 7
CONCLUSION .................................................................................... 215
7.1 Introduction ........................................................................... 21 5
7.2 Summary and implications ......................................................... 215
7.2.1 Summary ....................................................................... 215
7.2.2 Implications .................................................................... 216
7.3 Reﬂection on the present study ..................................................... 221
7.4 Suggestions for future research ..................................................... 224
7.5 Concluding remarks .................................................................. 225
REFERENCES .................................................................................... 227

viii

Table 1

Table 2

Table 3

Table 4

Table 5

Table 6

Table 7

Table 8

Table 9

Table 10

Table 11

Table 12

Table 13

Table 14

Table 15

Table 16

Table 17

Table 18

Table 19

Table 20

Table 21

LIST OF TABLES

Coding scheme for the characteristics of feedback episodes ..................... 62
Examples for the coding of the characteristics of feedback episodes. . .. . . . . ....7O
Coding scheme for learner uptake .................................................... 87

Examples for the coding of the occurrence and successfulness of uptake. . ....90

Examples for the coding of test results ............................................. 100
Reliability of coding ................................................................... 102
Treatment of variables in chi-square analysis ..................................... 103
Distribution of feedback episodes by characteristics .............................. 107
Overview of the occurrence and successfulness of uptake ...................... 109
Type and the occurrence of uptake ................................................. 110
Type and the occurrence of uptake residuals ...................................... 110
Type and the successfulness of uptake ............................................. 110
Type and the successfulness of uptake residuals .................................. 110
Source and the occurrence of uptake ................................................ 11 1
Source and the occurrence of uptake residuals .................................... 111
Source and the successfulness of uptake ........................................... 112
Source and the successfulness of uptake residuals ................................ 112
Response and the occurrence of uptake ............................................ 113
Response and the successfulness of uptake ........................................ 113
Linguistic focus and the occurrence of uptake ................................... 114
Linguistic focus and the occurrence of uptake residuals ........................ 114

ix

Table 22
Table 23
Table 24
Table 25
Table 26
Table 27
Table 28
Table 29
Table 30
Table 31
Table 32
Table 33
Table 34
Table 35
Table 36
Table 37
Table 38
Table 39
Table 40
Table 41
Table 42
Table 43

Table 44

Linguistic focus and the successfulness of uptake .............................. 114

Linguistic focus and the successfulness of uptake residuals ................... 114
Directness and the occurrence of uptake .......................................... 1 15
Directness and the successfulness of uptake ..................................... 116
Complexity and the occurrence of uptake ........................................ 117
Complexity and the successfulness of uptake .................................... 117
Emphasis and the occurrence of uptake ........................................... 118
Emphasis and the successfulness of uptake ...................................... 118
Distribution of overall comments .................................................. 120
Distribution of language-related comments ....................................... 120
Distribution of target-feature-related comments .................................. 120
Type and teacher-feedback-related comments .................................... 121
Type and teacher-feedback-related comments residuals ........................ 121
Source and teacher-feedback-related comments .................................. 122
Response and teacher-feedback-related comments .............................. 122
Linguistic focus and teacher-feedback-related comments ....................... 123
Directness and teacher-feedback-related comments .............................. 123
Complexity and teacher-feedback—related comments ............................ 124
Emphasis and teacher-feedback-related comments ............................... 125
Overview of test results .............................................................. 126
Type and test results .................................................................. 127
Source and test results ................................................................ 128
Response and test results ............................................................. 128

Table 45

Table 46

Table 47

Table 48

Table 49

Table 50

Table 51

Table 52

Table 53

Table 54

Table 55

Table 56

Table 57

Table 58

Table 59

Table 60

Table 61

Table 62

Table 63

Table 64

Table 65

Table 66

Table 67

Linguistic focus and test results ..................................................... 129

Directness and test results ............................................................ 130
Complexity and test results ........................................................... 130
Emphasis and test results ............................................................. 131
Overall results by the characteristics of feedback episodes ..................... 132
Occurrence of nonlinguistic cues .................................................... 157
Distribution of different types of nonlinguistic cues ............................. 15 7
General paralinguistic cues and the occurrence of uptake ....................... 159
General paralinguistic cues and the successfulness of uptake ................... 159
Type of paralinguistic cues and the occurrence of uptake ....................... 160
Type of paralinguistic cues and the successfulness of uptake .................. 160

Type of paralinguistic cues and the successfulness of uptake residuals ....... 161

General extralinguistic cues and the occurrence of uptake ...................... 162
General extralinguistic cues and the successfulness of uptake .................. 162
Type of extralinguistic cues and the occurrence of uptake ...................... 163
Type of extralinguistic cues and the occurrence of uptake residuals .......... 163
Type of extralinguistic cues and the successfulness of uptake ................. 163

Type of gestures and the occurrence of uptake ................................... 164
Type of gestures and the successfulness of uptake ............................... 164

General paralinguistic cues and teacher-feedback-related comments ......... 166
Type of paralinguistic cues and teacher-feedback-related comments ....... z .167
General extralinguistic cues and teacher-feedback-related comments. . . . . ....167
Type of extralinguistic cues and teacher-feedback-related comments. . . . . ....168

xi

Table 68

Table 69

Table 70

Table 71

Table 72

Table 73

Table 74

Table 75

Table 76

Table 77

Table 78

Table 79

Table 80

Table 81

Table 82

Table 83

Table 84

Table 85

Table 86

Table 87

Table 88

Table 89

Table 90

 

 

Type of gestures and teacher-feedback-related comments ...................... 169

General paralinguistic cues and test results ........................................ 170
Type of paralinguistic cues and test results ........................................ 171
General extralinguistic cues and test results ....................................... 172
Type of extralinguistic cues and test results ....................................... 172
Type of gestures and test results ..................................................... 173
Overall results by nonlinguistic cues ............................................... 175
Occurrence of general metalanguage ........................................ . ....... 194
Distribution of different types of metalanguage .................................. 195
Randomly selected teacher metalanguage by property and number of

teachers ................................................................................. 1 95
General metalanguage and the occurrence of uptake ............................ 198
General metalanguage and the successfulness of uptake ........................ 198
Type of metalanguage and the occurrence of uptake ............................ 199
Type of metalanguage and the occurrence of uptake residuals ................. 199
Type of metalanguage and the successfulness of uptake ........................ 199
General metalanguage and teacher-feedback-related comments ................ 201
Type of metalanguage and teacher-feedback-related comments ............... 201
General metalanguage and test results .............................................. 202
Type of metalanguage and test results .............................................. 203
Overall results by metalanguage ..................................................... 204
Uptake and teacher-feedback-related comments ..... ' ............................. 219
Successfulness of uptake and teacher-feedback-related comments ............ 220
Uptake and test results ................................................................ 220

xii

Table 91 Successfulness of uptake and test results ........................................... 220

xiii

Figure 1
Figure 2
Figure 3
Figure 4

Figure 5

Figure 6

 

 

LIST OF FIGURES

Main tenets of the interaction approach ............................................ 43
Measurement of the noticing and effect of teacher feedback ................... 48
Coding process of the observation data ............................................ 61
Coding process of stimulated recall comments ................................... 94

Overview of the characteristics of feedback episodes predicting (successful)
uptake, teacher-feedback-related comments, and correct test results. . . . . 1 33

Overview of nonlinguistic cues predicting (successful) uptake, teacher-
feedback-related comments, and correct test results ........................... 176

xiv

ASR

EFL

ESL

L1

L2

MA

NS

SLA

TESOL

 

KEY TO ACRONYMS

Adjusted Standardized Residuals
English as a Foreign Language
English as a Second Language
First Language

Second Language

Master of Arts

Native Speaker

Student

Second Language Acquisition
Teacher

Teaching English to Speakers of Other Languages

XV

 

CHAPTER 1

INTRODUCTION

Why can’t learners learn what teachers teach? What can help learners learn what teachers
teach? These questions have bewildered educators and researchers for years and years
and myriads of studies have been carried out to ﬁnd the answer. In the area of second
language acquisition, one notion that has often been used to account for the success or
failure of language teaching and learning is attention. Frequently associated with the
notion of attention is learners’ noticing of teacher feedback and the effectiveness of
teacher feedback. This is what the present study explored. Altogether, three factors were
examined: the characteristics of teacher feedback episodes, the nonlinguistic cues in
teacher feedback, and the metalanguage in teacher feedback. It is hoped that the results
would be able to shed some light on the forever puzzling mysteries of second language

teaching and learning.

There have been a plethora of studies that examined the role of the characteristics of
particular feedback types (e.g, Lyster, 1998a, 1998b, 2004; Sheen, 2006) or the
characteristics of whole feedback episodes (e. g., Ellis et al., 2001a, 2001b; Loewen, 2004,
2005). One strategy that has been frequently adopted by researchers to investigate the
issue is to audio-tape teacher-student interactions and look at learners’ verbal responses
to teacher feedback ‘as an indicator of the effect of the feedback. These studies have
contributed to our understanding of factors that may inﬂuence the effect of teacher

feedback. On the other hand, there is one aspect of teacher-student interactions that has

not received much attention so far. Just like in everyday conversations, oftentimes
learners’ responses to teacher feedback are nonverbal. In classroom settings, they might
respond to what the teacher is saying with nodding, writing in their books, or using other
body language. Such responses, although not as noticeable as verbal responses, constitute
part of teacher-learner interactions. Will the characteristics of teacher feedback (episodes)
show an effect on learners’ noticing and learning when learners’ nonverbal responses are
taken into account? The current study attempted to ﬁll in this gap of feedback research. I
used both audio and video recordings for data collection, thus making it possible for me
to identify learners’ nonverbal responses to teacher feedback along with their verbal

responses.

Another aspect of teacher feedback that has attracted researchers’ attention is teachers’
use of nonlinguistic cues. Although nonlinguistic cues have been suggested to be
facilitative of language learning for almost two decades now in the L1 literature (Farrar,
1992), it was not until recently that L2 researchers started to discuss it and/or examine it
in empirical studies (e.g., Davies, 2006; Faraco & Kida, 2008; Kida, 2008; Long, 2007;
Sime, 2006). The results from these studies are controversial. Some researchers found
that nonlinguistic cues had a positive effect on learning while others found that learners
did not notice them at all. Can nonlinguistic cues have an effect on learners’ noticing of
teacher feedback and their language improvement after all? If yes, to what degree? These
questions remain to be answered. The current study explored this area in a classroom
setting. Unlike most studies which looked at nonlinguistic cues from a single angle, I

looked at nonlinguistic cues from multiples perspectives, namely, learner responses to

teacher feedback, test results, and learners’ retrospective comments on their perception of

teacher feedback.

A third factor the current study strived to examine is the potential effect of metalinguistic
terms teachers used in their interactions with students. The use of metalanguage has been
under much debate for decades. To some metalanguage is not helpful to language
learning and may even be counteractive (e.g., Corder, 1973) while to others
metalanguage can serve as an important learning tool (e.g., Faech, 1985). Apart from
such theoretical discussions, there has been very little empirical research on the role of
metalanguage in language teaching and learning, especially in L2 literature. Among the
few studies that have been carried out, most examined metalanguage use in terms of how
teachers view it and when they use it. Little has been said about the actual effect of
metalinguistic terms on learners’ noticing of teacher feedback and second language
learning. Is there an effect at all? Do different kinds of metalanguage have different
effects? The present study sought preliminary answers to these questions, again by
looking at learner responses to teacher feedback, test results, and learners’ retrospective

comments on their perception of teacher feedback.

Revolving around the three factors described above, the next 6 chapters of the
dissertation are as follows: Chapter 2 reviews the literature on issues pertinent to the
present study, speciﬁcally, a few key concepts in second language acquisition, feedback
and the measurement of its effect on learning, and noticing and the measurement of

noticing. The research questions are then given based on the review of literature. Chapter

 

3 describes methodological issues, including the settings of the study, participant
information, data collection instruments and procedures, coding schemes, and statistical
analysis of the data. Chapters 4, 5, and 6 respectively present and discuss the results of
the study in terms of the characteristics of feedback episodes, nonlinguistic cues in
teacher feedback, and teachers’ use of metalanguage. Chapter 7 summarizes the results
from the previous three chapters, discusses their implications, reﬂects on the values and

limitations of the present study, and makes tentative suggestions on future research.

CHAPTER 2

LITERATURE REVIEW AND RESEARCH QUESTIONS

2.1 Introduction

This chapter reviews the literature and presents the research questions of the current
study based on the review. First, three key concepts in second language acquisition are
outlined: input, output, and interaction. Then, the constructs of feedback and noticing are
examined in three different sections: the role of feedback and its effectiveness, the role of
noticing and its measurement, and factors that may affect the noticing and effectiveness

of feedback. Finally, the research questions are put forward.

2.2 Mapping the big picture: input, output, and interaction

One approach to the study of language acquisition that has been substantially discussed in
SLA literature is the interaction approach (Gass, 1997, 2003; Gass & Mackey, 2006;
Gass & Selinker, 2001; Long, 1996). Associated with this approach are the concepts of
input, output, and interaction. The Input-Output-Interaction model has been considered to
be “the closest thing that we have to a ‘big’ theory to date” (Block, 2003, p. 26). Gass and
Mackey (2006) explain it succinctly: “the interaction approach considers exposure to
language (input), production of language (output), and feedback on production (through
interaction) as constructs that are important for understanding how second language
learning takes place” Q3. 3). Input, output, and interaction, then, are the three major ‘
constructs in the interaction approach to second language acquisition, and my study is

based on insights from the interaction approach.

2.2.1 Input

Simply speaking, input refers to what is available to the learner (Corder, 1967). In the L2
literature, almost all theories of second language learning recognize the signiﬁcance of
input in the acquisition process (Gass & Selinker, 2001). In the behaviorist
conceptualization of second language learning, input is of crucial importance because
imitation is viewed as the primary mechanism of language learning. In Krashen’s (1982,
1985) input hypothesis, comprehensible input, or the language information that is slightly
ahead of what the language learner currently knows, is central to all language acquisition.
Even in theories the focus of which is not speciﬁc input (e. g. interaction theories), input
is also attributed an important role. In terms of its nature, a distinction has been made

between input with positive evidence and input with negative evidence.

Positive evidence is the set of well-formed sentences the learner is exposed to while
negative evidence refers to information that is provided to the learner concerning the
incorrectness of an utterance (Gass, 2003). Whereas positive evidence has been
unanimously regarded as a necessary condition for language acquisition, the position of
negative evidence is more controversial. In the literature of ﬁrst language acquisition,
negative evidence has been suggested to be infrequent, often ignored, and therefore not a
sine qua non for language acquisition (Baker, 1979; Brown & Hanlon, 1970; Gass &
Selinker, 2001). Universal Grammar or “the system of principles, conditions, and rules
that are elements or properties of all human languages” (Chomsky, 1975, p.7) has been
proposed to explain the fact that ﬁrst language can be acquired despite the poverty of

evidence for impossible sentences. In second language acquisition, some researchers also

believe that positive evidence alone is enough for language acquisition to occur. Two
most typical examples are the above-mentioned behaviorist view of second language
acquisition and Krashen’s (1985) input hypothesis, both of which seem to assume that
acquisition will happen as long as there is the right amount and type of positive evidence.
Second language acquisition, however, has been noted to be different from ﬁrst language
acquisition in terms of ultimate attainment, learners’ prior knowledge, motivation, and so
on. Because of that, some researchers have assigned negative evidence to a more
important position. Schwartz (1992), for example, believes that negative evidence may
affect learners’ performance although it cannot permeate underlying competence. It has
even been observed that adults have to be exposed to negative evidence in order to
accomplish the goal of learning a second language (Birdsong, 1989; Bley-Vroman, 1989;

Gass, 1988; Gass & Selinker, 2001; Schachter, 1988).

There is mounting support for the claim that negative evidence can facilitate language
development (Long, 2007). One of the most widely cited study is White’s (1991)
investigation of how learners learn not to do something that is present in the native
language but not in the target language. The participants were French learners of English
who had to learn that English allows subject-adverb-verb order but does not allow
subject-verb-adverb-object order, which is allowed in French. Using a pretest, immediate
posttest, and delayed posttest design, White showed that negative evidence in the form of
correction promoted participants’ learning of adverb placement. In a follow-up study,
Trahey and White (1993) provided further evidence for this conclusion. Later studies

which examined the effect of corrective feedback (e.g., Ayoun, 2001; Gass & Veronis,

1989; Mackey, 2006; Pica et al., 1989; Trahey, 1996) have also lent support to White’s
ﬁndings. Some of these studies even showed that negative evidence can be more effective
in certain cases. A typical example is Iwashita’s (2003) investigation of the differential
effects of negative feedback and positive evidence on beginning Japanese as a second
language learners’ short-term grammatical development. The data were drawn from the
task-based interactions of 55 native-nonnative speaker dyads. Five types of interactional
moves were categorized either as negative feedback (recasts and negotiation) or as
positive evidence (completion, translation, and simple model). It was found that although
positive evidence occurred 10 times more than negative feedback, only learners who had
an above-average score on the pretest beneﬁted from it. Implicit negative feedback, by
contrast, had beneﬁcial effects on learners’ short-term grammatical development

irrespective of their current mastery level of the targeted language structures.

With all that said, the role of positive and negative evidence in second language

acquisition can be summarized in the following way:
positive evidence can reveal to learners the presence of information in the second
language that is different from the native language, but...negative evidence is
necessary to show what is not possible in the second language when it is possible
in the native language. (Gass & Selinker, 2001, p. 284)

2.2.2 Output

Output is the language produced by the learner (Gass & Mackey, 2006). It has been

claimed to be another condition for successﬁrl language learning. Output is regarded as a

way to practice new knowledge that has been learned. Another role it has been assigned

is that it provides a forum for learners to receive more input by eliciting feedback. The

 

concept of output did not receive heavy attention until Swain’s (1985) proposal of the
output hypothesis. This hypothesis is a response to the fact that students in Canadian
immersion programs demonstrated weaknesses in grammatical accuracy even though
they received rich comprehensible input. It holds that if students are to be both ﬂuent and
accurate in the target language, what they need is not only comprehensible input but also
opportunities to get feedback on what they produce. A key concept in the hypothesis is
comprehensible output or pushed output. What is meant by this notion is that learners are
“forced” to produce more accurate language as a necessary part of making themselves
understood. Swain explains it this way: when learners experience comprehension
problems, they will be pushed to make their output more precise, coherent, and accurate,
and thus be forced “to move from semantic processing to syntactic processing” (p. 249).
In a more recent article, Swain (2005) speciﬁes three functions of learner output: the
noticing function, the hypothesis-testing function, and the metalinguistic function. The

design of my study is based on these understandings of input and output.

While producing language, learners may notice that they do not know how to say what
they want to say or that they only have partial knowledge to express themselves. If this
gap is (made) explicit enough, it will raise learners’ awareness of their language problems.
For example,
(1) S: Iwent to a...a zoo.

T: What 200?

S: A 200 with lots of sea animals and ﬁsh.

T: An aquarium?
S: Oh, you call that an aquarium?

In this scenario, the student’s response “you call that an aquarium?” is a good sign of
selective attention and learning. It is through her language production and the feedback it

elicits that she notices her lack of the vocabulary word “aquarium.”

Sometimes learners are not sure of certain language structures. By producing language,
they can try out their hypotheses about such structures. With the feedback from
interlocutors conﬁrming or disconﬁrming their hypotheses, they may anive at a
conclusion. In the following example, a Chinese ESL learner is not sure of the English
word “eye—opening,” the literal translation of which into Chinese is “opening-eye,” so she
tries it out:
(2) S: This is a...opening-eye experience for me.

T: You mean an eye-opening experience?

S: Oh, eye-opening, not opening-eye?
Through this short conversation, the learner might realize that literal translation is not

going to work here. This realization is achieved by her trying her hypothesis about the

word “opening-eye” and the negative-evidence feedback the teacher gives her.

According to Swain, output also enables learners to control and internalize linguistic
knowledge and causes them to reﬂect on their own use of the target language, which is a
possible step toward language acquisition. For example:
(3) S: I saw two deers in the woods.

T: You saw two deer?

S: Oh, deer is an irregular plural form?

From her meta talk, the learner is obviously thinking metalinguistically about the plural

form of the word “deer”. The difference between her utterance and that of the teacher has

10

triggered this reﬂection. As a result, she has a better chance to learn the correct plural

form of “deer”. Without her producing “deers”, this might not have happened.

Since the proposal of the output hypothesis, various studies (e.g., Ano, 1998;
McDonough, 2005; Swain, 1997; Swain & Lapkin, 1995) have lent support to it.
Discussing how output provides a forum for language learning, Gass and Selinker (2001)
list four functions of output: (1) testing hypothesis, (2) receiving crucial feedback for the
veriﬁcation of these hypotheses, (3) developing automaticity in interlanguage production,
and (4) forcing a shift from more meaning—based processing of the second language to a
more syntactic mode. These functions, although expressed in a different way, are in

keeping with Swain’s output hypothesis.

2.2.3 Interaction

Decades ago, Wagner-Gough and Hatch (1975) argued that conversational interaction is
the basis for the development of syntax, not just a forum for the practice of grammatical
structures. It is now commonly accepted in the ﬁeld of SLA that interaction and learning

are closely related.

Conversational interaction can be deﬁned as the exchanges between/among learners or
between/among learners and native speakers where negotiation of meanings occurs (Gass
& Mackey, 2006). Central to this deﬁnition is the notion of negotiation of meaning,
which refers to instances in which participants interrupt the ﬂow of the conversation in

order for both parties to understand what the conversation is about (Gass & Selinker,

ll

2001). The role of negotiation of meaning in interaction is illustrated in Long’s (1996)
frequently cited statement:
Negotiation of 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. (p. 414)
From this statement, it can be seen that negotiation of meaning involves both input and
output, two of the most important factors in language learning as discussed above. Given
that, it is not surprising that Long’s (1983, 1985, 1996) interaction hypothesis has been
said to subsurne aspects of both Krashen’s (1982, 1985) input hypothesis and Swain’s
(1985, 1995, 2005) output hypothesis (Gass & Mackey, 2006). Gass (1997) has also
argued for the facilitating role of interaction through negotiation of meaning. In her view,
focused negotiation work in interaction can orient learners’ attentional resources to what

they do not know or do not know well, thus serving as a priming device to set up the

initial step of learning.

2.3 Looking closer: feedback and the measurement of the effectiveness of feedback

The role of interaction in SLA has invited a large body of research (e. g., Ellis et al., 1994;
Gass & Veronis, 1994; Polio & Gass, 2005). A concept that has received heavy attention
in interaction research is feedback. This includes both the effect of feedback on learners’

language improvement and the measurement of the effect.

2.3.1 The role of feedback in second language acquisition
Feedback in conversational interactions can provide input to learners either as positive

evidence or negative evidence or a combination of the two, depending on the nature of

12

the feedback. Upon receiving feedback, learners may try to correct themselves. The new
utterance can be regarded as pushed output, which can then help language learning by
realizing the three functions illustrated by Swain (2005). Long (2007), in a new book
about problems in SLA, devotes a whole chapter to the role of recasts, a type of implicit
negative feedback. After a detailed review of both cross-sectional and longitudinal
studies, Long concludes that, although they are not necessary for acquisition, recasts
“appear to be facilitative, to work better than models, and to do so incidentally, without
interrupting the ﬂow of conversations and participants’ focus on message contents” (p.
94). This facilitative role of feedback in language learning is also reﬂected in negotiation:
Negotiation serves as a catalyst for change because of its focus on incorrect forms.
By providing learners with information about incorrect forms, negotiation enables
learners to search for additional conﬁrmatory or nonconﬁrrnatory evidence. (Gass
& Selinker, 2001, p. 283)
While the focus of this claim is negotiation, it points to the importance of feedback. To a
large extent, feedback is what initiates changes in learners’ interlanguage systems.
Without feedback, learners may not notice the gaps or holes in their L2 knowledge, may
not modify their language production, and may not get the input they need in order to
conﬁrm or disconﬁrm their hypotheses about a language structure. Both the
aforementioned studies that examined the role of negative evidence and those reviewed

later about the effectiveness of feedback (see Section 2.5) strongly suggest that feedback

plays a signiﬁcant role in second language learning.

2.3.2 The measurement of the effectiveness of feedback
Loewen (2004) provides a list of the methods that have been utilized to measure the

effectiveness of recasts (and other types of feedback) in classroom contexts: learner

l3

uptake, uptake charts following a lesson, spontaneous production of recast forms elicited
in task-based interactions, elicited production of targeted forms in tailor-made
individualized tests, and indirect measures such as grarnmaticality judgment tests that
seek to access learners’ underlying knowledge of a form. All these methods have their
own advantages and disadvantages. For the purpose of the current study, only two

methods are reviewed: tailor-made individualized tests and learner uptake.

2.3.2.1 Tailor-made individualized tests

Tailor-made individualized tests have been discussed and adopted by a number of
researchers who study interaction (e.g., Loewen, 2002, 2005; Loewen & Philp, 2006;
Nabei & Swain, 2002; Swain & Lapkin, 1998; Williams, 2001). Since they are used in
different contexts by different researchers, no universal rules have been established about
how to develop such tests and what forms they should take. Overall, tailor-made
individualized tests tend to require learners to correct their nontarget-like utterances on
which they have previously received feedback (Loewen & Philp, 2006). Test items are
created from forms that arise in language episodes which involve test takers. Below is an

example which shows how tailor-made test items are developed.

(4)
Language episode:
OL: He want to hitchhike. . .travel and—
L: travel to Europe. . .hitchhiking? What hiking?
OL: Hitchhiking, you know?
L: no.
OL: Like you don’t have car but I want to move another city and my thumb —
L: - Oh! OK hitch. How do you spe-no — how do you pronounce it?
OL: Hitchhike.
L: Hitchhike.
Test item:

Lena wanted to travel to the next town but she had no car and no money for the

14

bus. So she decided to . She walked to the side of road and put out her
thumb. Soon a car stopped and gave her a ride. (From Williams, 2001 , p. 331)

In the language episode, a learner initiates a question about a vocabulary word and
another learner supplies the information. The test item accordingly assesses whether the

initiator of the question can recall the word by ﬁlling in a blank.

Like any language tests, tailor-made individualized tests are not perfect. One of the
biggest disadvantages of such tests is that since test items are based on language forms
that incidentally occur in language episodes, they cannot be predicted or are hard to
predict. As a result, it is difﬁcult to design pretests. In the absence of pretests, however,
the tests cannot provide information about learners’ previous knowledge of the forms
tested and thus cannot differentiate between the acquisition of new knowledge and the
consolidation of latent knowledge. As Loewen (2002) argues, for individualized testing,
“the FFE [focus-on-form episode] itself has to serve as a type of pretest, indicating
students’ lack of prior knowledge or ability regarding the targeted linguistic item” (p. 83).
On the other hand, since they are speciﬁc to particular language forms and particular
learners receiving particular feedback, such tests can at least “inform us about leamers’
ability to use demonstrably problematic forms subsequent to the provision of feedback”
(Loewen & Philp, 2006, p. 542). Individualized testing therefore can serve as a great tool

to measure the effectiveness of feedback.

One issue that is pertinent to any type of language testing is construct validity. In the area
of SLA, however, there is a widespread failure to consider the construct validity of

testing instruments (Ellis, 2005). As a relatively new testing method, and because of their

15

complexity and variation in terms of their contents, format, etc., tailor-made
individualized tests have received even less attention with regard to their construct
validity. One study that deserves special attention in this respect is Loewen’s (2002)
study of incidental focus on form. In order to design a valid and reliable testing
instrument, Loewen conducted two pilot studies and performed a validity and reliability
analysis on the individualized test items in his ﬁnal study. Despite such careful design
and analysis, Loewen still showed a concern with the construct validity of his tests,
saying that “Another limitation related to the testing concerns whether learner
performance on the test items are representative of learner performance in other contexts”
(p. 287). When “other contexts” are classroom interactions, Loewen’s concern can be
interpreted as the mismatch between learner performance in the tests and learner
performance in the classroom. Speciﬁcally, the tests most likely allowed learners to
access their explicit L2 knowledge. However, this knowledge could be much less
accessible during spontaneous classroom interactions due to the different settings and
objectives of learner language production. This discrepancy between test performance
and classroom performance points to two important issues concerning tailor-made
individualized tests: what to test and how to test. Both are crucial questions in examining
the construct validity of such tests. The what question can virtually be traced to the
relationship between implicit knowledge and explicit knowledge while the how question

is a matter of how to operationalize the two types of knowledge for research purpose.

Ellis (2005) illustrates the relationship between implicit knowledge and explicit
knowledge from three different interface theories. The noninterface position largely

rejects the possibility of explicit knowledge transforming directly into implicit knowledge

16

and vice versa. The strong interface position claims that the two types of knowledge can
be derived from or converted into each other. The weak interface position acknowledges
the possibility of explicit knowledge becoming implicit but with some limitation on when
or how this can take place. Despite the fact that there are different and conﬂicting
positions about the role played by explicit knowledge in the acquisition of implicit
knowledge, it is widely accepted that explicit knowledge can contribute to performance.
A test that supposedly measures learners’ implicit knowledge, therefore, may in actuality
measure some aspects of learners’ explicit knowledge. Meanwhile, it has been found that
different performance tasks are likely to induce L2 learners to draw differentially on their
implicit and explicit knowledge (Bialystok, 1982, cited in Ellis, 2005). This means that

the two types of knowledge can probably be measured in different ways.

Based on the characteristics of implicit knowledge and explicit knowledge, Ellis (2005)
proposes seven criteria for assessing L2 implicit knowledge and explicit knowledge:
degree of awareness, time available, systematicity, certainty, focus of attention,
metalinguistic knowledge, and leamability. More speciﬁcally, when a task assesses
learners’ implicit knowledge, learners make use of “feel” when responding to the task;
they are pressured to perform the task online; they provide consistent answers; and they
are highly certain of their responses; the primary focus of the task is meaning; it does not
require learners to use metalinguistic knowledge, and the knowledge involved favors
early learning. In contrast, when a task assesses learners’ explicit knowledge, learners
provide responses to the task using rules; they are not pressured by time when performing

the task; they provide variable responses; and they have a low degree of certainty in their

17

responses; the primary focus of the task is form; it encourages learners to use
metalinguistc knowledge; and the knowledge involved favors late, form-focused
instruction. In a psychometric study which aimed to develop a battery of tests that would
provide relatively separate measures of implicit and explicit knowledge, Ellis (2005)
found that most of these criteria proved true although some of them did not receive clear
evidence or were not supported because of various factors such as learners’ random

behavior.

Going back to Loewen’s (2002) concern about his individualized tests, it suggests that the
items in such tests should tap into learners’ implicit knowledge (e. g, knowledge accessed
during classroom interactions and online production), not just their explicit knowledge
(e. g, knowledge accessed in decontextualized tests). This requires investigators to

minimize the mismatch between the testing situation and classroom interactions.

2.3.2.2 Learner uptake

As early as the 19803, the term “uptake” was proposed by Allwright (1984) to refer to
what learners report they have learned from a lesson. More recently, Lyster and Ranta
(1997) deﬁne uptake as “a student’s utterance that immediately follows the teacher’s
feedback and that constitutes a reaction in some way to the teacher’s intention to draw
attention to some aspect of the student’s initial utterance” (p. 49). Uptake is regarded as
an important observable source for understanding the impact of feedback ( Nabei &
Swain, 2002), hence it has been used to measure the effectiveness of feedback in various

studies (e.g., Ellis et al., 2001a, 2001b; Loewen, 2002, 2004; Lyster, 1998a; Lyster &

18

 

 

Ranta, 1997; Panova & Lyster, 2002; Sheen, 2006. For more detail about these studies,
see Section 2.5). According to Ellis et al. (2001a), as also discussed in Basturkrnen et a1.
(2002) and Loewen (2004), “there are theoretical grounds for believing that uptake might
contribute to acquisition” (p. 287). Two reasons are given for this claim. First, uptake as a
response to teacher feedback helps learners practice language items and automatize the
retrieval of them. Second and more importantly, by producing uptake, learners attempt to
use forms they have previously used incorrectly or received information about. Uptake in
this sense is a type of pushed output, which allows learners to process syntactically rather
than semantically. Uptake therefore “may create the conditions needed for language

acquisition to occur” (Ellis et al., 2001a, p. 287).

On the other hand, some researchers (e. g., Long, 2007; Ohta, 2000) argue that uptake is
merely a discourse phenomenon that may not necessarily be related to the acquisition
process. Uptake is an optional move (Ellis et al., 2001a; Loewen, 2004; Sheen, 2006).
Learners may or may not respond to the linguistic information provided to them. At times
they may choose not to produce uptake even if there is an opportunity for them to do so.
This may happen when other interactional tasks are in priority or when they think it is
unnecessary to respond. There are also occasions where learners do not have the chance
to respond to the feedback they receive. Many studies (e.g., Loewen, 2002, 2004; Lyster,
1998a; Oliver, 2000) have shown that teachers may continue their turn after giving
learners feedback, allowing no time for learners 'to respond to the feedback. Uptake is
therefore not a guarantee of acquisition (Williams, 2001). Even researchers who have

repeatedly used uptake in their studies have been very cautious with the term. Ellis et al.

19

 

(2001a), for example, reiterate that “uptake cannot be viewed as evidence that acquisition
has taken place. . .nor. . .necessary for acquisition to take place” (p. 287). In view of such
concerns, uptake alone may not be sufﬁcient to examine the effectiveness of feedback

and may work better together with other measures.

2.4 Looking in particular: noticing and the measurement of noticing
The important role of feedback in second language learning is evident. The degree of the
effectiveness of feedback, however, is more complicated. To a large extent it depends on
learners’ noticing of feedback and yet it is not always clear whether learners can perceive
feedback as expected (Gass & Selinker, 2001; Hawkins, 1985). Below is an example
where the learner fails to recognize a clariﬁcation request as negative evidence.
(5) S: One day, the fairy, sting the magic wand to Cinderalla.
T: Sorry?
S: One day, the fairy sting the magic wand to Cinderalla.
T: Ok.
S: Cinde, ah, Cinderaella changed into, the beautiful girl. (Laugh) Ah, and, the,
Cin, Cinderella went to the palace by coach. The, the prince fall in love at a
ﬁrst glance.
T: Sorry?
S: Ah, the prince fall in, falled falled in love Cinderella at a ﬁrst glance. And they
dance, they danced... Ah, Cinderella have, Cinderella have to go home.
(Takashima, 1995, cited in Gass and Mackey, 2006)
In this example, although the learner might have ﬁnally been alerted to the use of past
tense as indicated in his struggle to change “fall” into “falled” and “dance” into “danced”,
he does not make any change after the ﬁrst “sorry”. He probably has assumed that the

teacher does not understand him simply because of his low voice, unclear pronunciation,

or broken discourse.

20

Given the complexity of noticing, this section ﬁrst brieﬂy talks about the role of attention
and noticing in second language acquisition, and then discusses the measurement of

noticing.

2.4.1 Attention and noticing in second language acquisition

Attention has been identiﬁed as an important cognitive process in second language
acquisition, a process that “encodes language input, keeps it active in working and short-
term memory, and retrieves it from long-tenrr memory” (Robinson, 2003, p. 631).
Noticing is the part of the attentional system that involves the detection and consequent
registration of stimuli in memory (Philp, 2003; Robinson, 1995; Tomlin & Villa, 1994).
The role of attention and noticing in selecting input for L2 learning is a contentious issue
in the SLA literature. Some researchers believe that attention is not important in second
language acquisition. A typical example is Krashen. Krashen (1982) made a distinction
between language acquisition, a subconscious process, and language learning, a
conscious process. He believes that only the acquired system can be used to produce
language while the learned system can only serve as a monitor for the former. Unlike
Krashen, some researchers have seen an important role of attention and noticing in SLA.
Long (1996), for example, argues that selective attention along with learners’ L2
processing capacity mediates the L2 acquisition process. VanPatten (1994) demonstrated
the role of attention in L2 acquisition by showing how a Spanish learner starts to use
subjunctive mood in his own speech after he notices it in others’ speech. Schmidt (2001),
one of the strongest proponents for the role of attention and noticing, believes that “the

concept of attention is necessary in order to understand virtually every aspect of second

21

language acquisition” (p. 3). In his famous noticing hypothesis (1990, 1993, 2001), it is
proposed that the subjective experience of “noticing” is the necessary (and sufﬁcient)
condition for the conversion of input to intake; noticing is therefore the ﬁrst step of
language learning. A series of studies have been conducted around the issue of attention
and noticing (e.g., Kim, 1995; Leow, 1997, 2000; Robinson, 1996, 1997; Rosa & O’Neill,
1999; Schmidt & Frota, 1986). Although some studies lend less support to the noticing
hypothesis, the cumulative ﬁndings show that attention and noticing are important in
SLA. It has been argued that even if noticing is not necessary for L2 learning, it is
facilitative. Even Gass (1997), who attributes some learning to universal grammar and
argues against the claim that all L2 learning requires attention, stresses that her argument
is not intended to lessen the importance of attention. In simple words, “there does not
appear to be any evidence at all against the...claim that people learn about the things they
attend to and do not learn much about the things they do not attend to” (Logan et al.,

1996, cited in Schmidt, 2001, p. 30).

In the SLA literature, the term “noticing” has been associated with and distinguished
ﬁom other concepts such as awareness, alertness, consciousness, and detection. Robinson
(1995), for example, distinguishes between noticing and detection on the basis of
awareness, deﬁning noticing as detection with awareness and rehearsal in short-term
memory. For the purpose of the current study, noticing refers to detection accompanied
by lesser and greater degrees of awareness (Philp, 2003). The various levels of awareness;

however, were not explored respectively.

22

2.4.2 The measurement of noticing

In the SLA literature, noticing has been measured through both online procedures and
ofﬂine procedures. Online procedures or concurrent reports take various forms such as
think-aloud protocols (e.g., Alanen, 1995; Loew, 1997, 2000), private speech (e.g., Ohta,
2000), online uptake charts (e.g., Mackey et al., 2001), and online learning journals (e. g.,
Mackey, 2004, cited in Egi, 2004). Ofﬂine procedures or retrospective reports also take a
variety of forms such as post-treatrnent questionnaires (e.g., Mackey et al., 2002 ;
Robinson, 1995), stimulated recall (e. g., Egi, 2007a, 2007b; Mackey et al., 2000; Swain
& Lapkin, 1998, 2001), offline uptake charts (e. g., Slimani, 1989), prompted repetition of
feedback (e.g., Philp, 2003), and diaries (e.g., Schmidt & Frota, 1986). When
investigating learners’ noticing of feedback and its impact on learning, it is important to
make sure that the measures of noticing “should accurately capture learners’ cognitive
processes while neither facilitating nor hindering learning” (Egi, 2004, p. 243). As
researchers strive for this goal, concerns have been raised about both online and ofﬂine
procedures (e.g., Mackey, 2006; Schmidt 2001). For example, online think-aloud
protocols require learners to report their mental processes under temporal and
communicative pressure. This can lead to the underreporting of noticing and thus fail to
capture certain aspects of the cognitive processes. Offline procedures such as diaries and
questionnaires have their own problems too. Tomlin and Villa (1994) point out that the
cognitive processing of input “takes place in relatively brief spans of time, seconds or
even parts of seconds” (p. 185). However, diaries and questionnaires might span a much
longer time. By the time learners write their diary or do the questionnaire, they might

have forgotten what they were thinking. Such reports of noticing, like online think-aloud

23

protocols, may also lead to underreporting. In view of such problems, Mackey (2006) and
Mackey and Gass (2006) suggest that researchers should triangulate methods of

collecting noticing data to obtain a picture as full as possible.

The advantages and drawbacks of verbal reports have been substantially discussed by Egi
(2004, 2007a, 2008). Two major issues concerning the validity of verbal reports in Egi’s
discussion are reactivity and veridicality. Reactivity concerns the effect of verbalization
on participants’ task performance. If verbal reports affect participants’ task performance,
they are reactive. If not, they are non-reactive. Veridicality concerns the accuracy of
verbal protocols as a reﬂection of learners’ cognitive processes. If verbal reports fail to
capture participants’ thoughts or include thoughts that did not take place, they are non-
veridical. In line with the principle that verbal reports should not inﬂuence participants’
performance while accurately reﬂecting their cognitive processes, “non-reactive and
veridical reports are considered to be valid data” (Egi, 2004, p. 245). For the purpose of
the current study, two noticing measures are discussed in view of their reactivity and

veridicality: stimulated recall and again learner uptake.

2.4.2.1 Stimulated recall

Stimulated recall is a retrospective data collection method in response to the ongoing
shift toward examining cognitive factors in second language research (Gass & Mackey,
2007). It is often conducted by giving participants stimuli and asking them to comment
on their thoughts during a previous task. Such comments allow researchers to gain access

to participants’ internal thinking processes and gather information that might be

24

unavailable from observation data. Like other retrospective verbal reports, it has been

examined with respect to reactivity and veridicality.

Reactivity is not an issue for stimulated recall when it is administered after a task (Egi,
2004). When it is carried out before a task, however, it can be reactive due to at least two
reasons: stimulus presentation and verbalization (Egi, 2008). While recall cues such as
video or audio recordings of participants’ task performance can help them more
completely recall their cognitive processes and thus mitigate the problems caused by
memory decay, they can also serve as additional L2 input. As a result, learners are
exposed to the same L2 input twice, once during the task and once during the recall. This
double-input exposure (Leow, 2002) may reinforce learners’ previous learning or give
them time to process input that they have not previously processed. The second source of
reactivity is the act of verbalization. Egi (2004) explains it with Swain’s (1985, 1995,
2005) output hypothesis. For one thing, verbalization of the cognitive process gives
learners additional opportunities to practice the target language, which can promote
ﬂuency. For another thing, verbalization gives learners opportunities to reﬂect on their
own language use, thus fulﬁlling the metalinguistic function of pushed output. For such
reasons, both the stimuli and verbalization can affect learners’ test performance. In terms
of veridicality, there are also two potential sources for the non-veridicality of stimulated
recall. The ﬁrst source again is stimulus presentation. As described above, the recall cues
provide learners with additional input. This double exposure to the same input may allow
learners to reconstruct their thoughts at the time of recall rather than retrieving

information from long-term memory (Egi, 2004; Gass & Mackey, 2000). The second

25

source is what Egi (2004) calls the interviewer effect. Stimulated recall is often carried
out in an interview format. The presence of the interviewer may lead participants to
report what they believe is of interest to the interviewer and/or avoid reporting their
errors in order to sound intelligent (e.g. Jourdenais, 2001; Norris,l990, cited in Egi,
2004). In brief, stimulated recall faces reactivity and veridicality issues depending on

when and how it is conducted.

In the area of SLA, although stimulated recall has been adopted by many researchers to
measure cognitive processes such as noticing and awareness, its validity is under-
explored (Egi, 2004, 2008). One study that has shed light on the reactivity of stimulated
recall is the one by Adams (2003). Fifty-six L2 Spanish learners collaboratively wrote a
story in pairs. They were then randomly assigned to three goups: a control group, a
noticing-only group, and a noticing-plus-stimulated-recall group. The control group
repeated the writing task without additional treatment. The noticing-only group compared
their original essay and a reformulated version of their essay. In addition to text
comparison, the noticing-plus-stimulated-recall group recalled their thoughts at the time
they made the text comparisons. Both experimental groups then individually wrote the
story again. Results indicated that the noticing-plus-stimulated-recall group incorporated
reforrnulations in the second writing signiﬁcantly more than the noticing-only group. It
was concluded that learners’ participation in the stimulated recall had beneﬁted their
learning, suggesting that stimulated recall in this study'was reactive. Egi’s (2007a) study
provided support for the “contaminating effect” of stimulated recall. Two groups of

learners were asked to report their interpretation of recasts provided during

26

conversational interactions: an immediate report group and a stimulated recall group. The
immediate report group reported any thoughts they had during a brief conversational turn
as prompted by a native speaker interlocutor (for details about immediate reports, see E gi,
2004, 2007a and Philp, 2003). The stimulated recall group conducted a recall interview
after the interaction sessions while watching their own videotaped interactions and
communicative tasks. The two groups did not show signiﬁcant differences on the
immediate posttest, which was administered before the stimulated recall session.
However, in the delayed posttest, which was administered after the stimulated recall
session, the stimulated recall group signiﬁcantly outperformed the immediate report
group. This change suggests that the stimulated recall interview had inﬂuenced learners’
performance. Like Adams’ study, this study shows that stimulated recall can be reactive

if it is conducted before a testing event.

Compared with its reactivity, the veridicality of stimulated recall has received less
attention, probably because one of the major criticisms of delayed retrospective reports,
memory decay, is alleviated by stimulus presentation. To look into this issue, we can turn
to another piece of research by Egi (2004). Although the main purpose of the study is to
explore the use of immediate retrospective verbal reports, it also has indications for the
veridicality of stimulated recall. The design of the study is similar to but different from
the one described above. It followed a pretest-treatment—posttest sequence, but there was
no delayed posttest. Participants were also divided into two groups: an immediate report
group and a stimulated recall group. Similarly, the immediate report group recalled their

thoughts about language episodes immediately after a brief conversational turn as

27

prompted by an auditory stimulus while the stimulated recall group recalled their
thoughts while watching videotaped treatment sessions after the completion of the
treatment and posttest. Analysis of the data suggested that stimulated recall was less
effective in capturing learners’ noticing due to various factors such as memory decay,
multiple exposures to the same input, and interviewer effect. This result suggests that

stimulated recall is not always veridical.

In summary, although stimulated recall can be argued to elicit valid data when carefully
operationalized (Egi, 2004), there are problems researchers need to be careful with
whether it is conducted before or after testing events. Measures should be taken to reduce

or circumvent its reactivity and non-veridicality.

2.4.2.2 Learner uptake

Uptake in its traditional deﬁnition as verbal report is not free from reactivity and non-
veridicality. For example, in the use of online uptake charts, learners’ normal learning
activities might be interrupted, which may in turn affect their test performance. In the use
of ofﬂine uptake sheets, there might be new information that learners have noticed but
are not able to recall at the moment of reporting. In either case, learners may report what
they have not actually comprehended in order to appear intelligent. Such reports can

provide researchers with false information about noticing and learning.

In its more recent deﬁnition by Lyster and Ranta (1997, see Section 2.3.2.2), uptake is a

leamer’s response to the teacher’s corrective move, the ﬁnal objective of which is the

28

leamer’s attention to his or her erroneous utterance. The production of uptake, successful
uptake in particular, may therefore indicate that the learner has indeed attended to his or
her erroneous utterance as expected. In the SLA literature, many researchers have
referred to uptake as an indication of noticing. Lightbown (1998), for example, claims
that “a reformulated utterance from the learner gives some reasons to believe that the
mismatch between learner utterance and target utterance has been noticed, a step at least
toward acquisition” (p. 193). Believing that uptake constitutes one type of pushed output,
Loewen (2004) observed that “This pushed output, then, may be an indication of
noticing” (p. 157). Loewen and Philp (2006), talking about the relationship between
uptake and the noticing of recasts, stated that “the production of successﬁrl uptake
provides an indication that the learner has noticed the recast” (p. 542). Sheen (2006),
drawing on previous discussions about uptake as an indication of noticing and studies
that probed into the relationship between uptake and the noticing or interpretation of
recasts, concluded that “uptake/repair can serve as one measure of learner noticing and

thus have potential for language acquisition” (p. 368).

To me, uptake has methodological advantages that other noticing measures may not have.
First, unlike think-aloud protocols, uptake is part of interactional discourse and does not
require learners to stop the task at hand to report what is going on in their mind. The
occurrence of uptake therefore will not interrupt the learning process and affect ultimate
test performance. For the same reason, the reactive effect posed by the verbalization of
cognitive processes can be somewhat circumvented. Furthermore, uptake does not

involve stimulus presentation. This helps to solve the double-input exposure problem

29

caused by recall cues. In terms of veridicality, uptake has its strengths too. Uptake occurs
online. As a result, memory loss is not a problem. Since no interviewer is involved,
learners do not need to provide information that they think is of interest to the interviewer
but that does not actually reﬂect their real thinking. Interviewer effect therefore is not an
issue either. In brief, although uptake may not provide information that directly describes
the cognitive process, it can at least give researchers clues about whether noticing has

happened and sometimes, what has been noticed.

Apart from theoretical discussions, empirical evidence for the potential of uptake as a
measure of noticing can be found in both L1 and L2 research. In ﬁrst language
acquisition, it has been observed that children frequently repeat recasts (See Long &
Robinson, 1998). Repetition constitutes a type of uptake. Although learners may repeat
teachers’ model utterances as nothing more than a rote language behavior (Mackey &
Oliver, 2002), it should not be ruled out that sometimes repetition does result ﬁom
learners’ noticing of the corrective nature of recasts and the errors in their initial
utterances. In second language acquisition, one noteworthy study pertinent to uptake as a
noticing measure is Mackey et al’s (2000) investigation of learners’ recognition of the
target of feedback. The participants ﬁrst ﬁnished a spot-the-difference-task and then
watched the videotape of their performance and made comments on it. In a post hoc
analysis, the researchers explored the relationship between learners’ perception about
feedback and their immediate uptake of the feedback during the interaction. It was found
that learners’ stimulated recall reports generally revealed accurate perception about

feedback for which they had uptake at the time of the interaction. Speciﬁcally, in cases

30

where learners produced uptake, they also seemed to be accurate in their perception of
what the feedback was about. When they did not react to the feedback in the discourse,
they generally did not perceive it as feedback. This ﬁnding provides strong support for

uptake as an indication of noticing.

Han (2002) drew on learner uptake of feedback to look at noticing. In her small-scale
empirical study about the impact of recasts on tense consistency in L2 output, Han gave
an example about how a learner’s tense consistency in an oral narrative improved after
she received recasts. Han discussed this change with reference to uptake and noticing.
In producing the oral narrative above, Jee-Young received six recasts, four of
which targeted her incorrect use of tense. Her uptake suggests that her attention
was indeed drawn, albeit to a varying extent, to the discrepancies between her
own output and the input provided by the recasts. (p. 561)
Here Han’s conclusion about the learner’s noticing of the differences between the tense
use in her own production and that in the recasts was drawn from the leamer’s uptake.

Although uptake was not clearly stated as the measuring tool of noticing in the study, it

can be seen that Han was using it to make inferences about noticing.

Han (2002) is not the only one who interpreted uptake as evidence for noticing. In a study
about the effectiveness of recasts in different instructional settings, Lyster and Mori
(2006) concluded that “the overall higher proportion of uptake and repair following
feedback in JI [Japanese immersion] classrooms ...suggests that JI students were
predisposed to noticing the corrective purpose of recasts” (p. 292). This statement
suggests that Lyster and Mori also treated uptake as an indication of noticing in their

study. A study which took learner uptake as the primary measure of noticing is Song

31

(2007), who argued that uptake “may be an effective way of assessing noticing” (p. 6)
given its role as pushed output. Operationalizing noticing with uptake, Song found that
learners’ noticing of recasts was constrained not only by the number of changes in recasts,

but also by the linguistic domains of those changes.

Despite the belief in the potential of uptake as an indication of noticing, the use of uptake
as a measurement for noticing is a contentious issue for the same reasons as described in
Section 2.3.2.2: its optional nature and sometimes the lack of opportunities for its
production. As a result, while successﬁrl uptake can provide evidence for noticing, the
reverse does not necessarily hold true (Sheen, 2006). Furthermore, one may argue that
since uptake is an internal part of interactional discourse and a type of pushed output, it

inevitably has an effect on testing results and hence is inevitably reactive.

In summary, both stimulated recall and uptake have advantages and disadvantages as
measures of noticing. Stimulated recall may directly provide speciﬁc information about
learners’ internal thought processes, but may be reactive and non-veridical due to factors
such as stimulus presentation, verbalization, memory decay, and interviewer effect.
Uptake may circumvent validity problems caused by these factors, but may not provide
complete and direct information about cognitive processes due to its optional nature and
sometimes the lack of opportunities for learners to produce it. Taken separately, neither
stimulated recall nor uptake can be assumed to measure noticing with complete accuracy.
Taken together, however, they can be supplementary to each other. Uptake may provide

information that learners have forgotten even with recall cues while stimulated recall may

32

provide information that is not observable from uptake. In the hope to get a ﬁrller picture
of learners’ noticing of teacher feedback, and also to answer the call for method
triangulation in collecting noticing data, both stimulated recall and learner uptake were

adopted in the present study.

2.5 Pulling it together: factors that may affect learners’ noticing of feedback and the
effectiveness of feedback

Having discussed the role of noticing and feedback in second language acquisition, this
chapter now turns to factors that might affect learners’ noticing of feedback and the
effectiveness of feedback. Among the various factors, three are discussed: the
characteristics of feedback, the nonlinguistic cues in feedback, and the metalanguage in
feedback. It should be noted that in most of the studies mentioned below, uptake was

used as a major indicator for the effectiveness of feedback, not a noticing measure.

2.5.1 Characteristics of feedback

The effectiveness of feedback has been found to be related to the speciﬁc features of
feedback. A study by Loewen and Philp (2006) revealed that recasts with stress, recasts
with declarative intonation, recasts in extended episodes and recasts with only one
change led to more successful learner uptake and higher test scores. Sheen (2006)
examined seven characteristics of single-move recasts: mode, scope, length, reduction,
number of change, type of change, and linguistic focus. It was found that recasts that
were short, declarative, reduced, repeated, with a single error focus, and involving

substitutions were positively related to learner uptake and repair.

33

These studies were focused on the characteristics of a particular type of feedback, namely
recasts. With the rise of focus on form research, there have been a series of studies which
investigated the effect of the characteristics of focus-on-form episodes (e.g., Ellis et al.,
2001a; 2001b; Loewen, 2003, 2004, 2005). A focus-on-form episode often consists of an
occasional shift of attention to linguistic code features (Long & Robinson, 1998). It is
triggered by comprehension or production problems, whether they are perceived or
anticipated (for a detailed discussion, see Williams, 2005). As a result, focus on form
often engages different types of teacher feedback. Studies on the characteristics of focus
on form, therefore, shed light on the characteristics of various feedback types. Two of the
most important studies are Ellis et al. (2001a) and Loewen (2004). In Ellis et al., four
characteristics of focus-on-forrn episodes were investigated: source, complexity,
directness, and linguistic focus. Results indicated that there was a higher rate of uptake in
negotiation of meaning than form and that it occurred more loften and was more
successful in complex than simple episodes. Directness and linguistic focus, on the other
hand, were not signiﬁcant factors although uptake was more successful in episodes
involving pronunciation than those involving vocabulary. Using a similar method,
Loewen (2004) investigated the occurrence of uptake and its successfulness in relation to
the characteristics of incidental focus on form. He found that complex, immediate, and
elicit feedback led to a high level of uptake, and the successfulness of uptake was

associated with complex, code-related,'immediate, reactive, heavy, and elicit feedback.

34

All the studies reviewed above demonstrate that the characteristics of feedback have an
impact on the effectiveness of feedback. In view of the important role of noticing in
second language learning, the characteristics of feedback may also be related to noticing.
In actuality, some researchers have investigated this relationship. Following his ﬁnding
that recasts led to a low level of uptake, Lyster (1998b) explored the reason why it was as
such. Two characteristics of recasts were examined: whether they were isolated or
incorporated, and whether they were declarative or interrogative. Recasts were coded into
four categories: isolated and declarative, isolated and interrogative, incorporated and
declarative, and incorporated and interrogative. Lyster found that recasts were distributed
in equal proportions and fulﬁlled identical functions with noncorrective repetitions and
that teachers frequently used positive feedback to express approval of the content of
learners’ messages irrespective of well-formedness. These ﬁndings suggest that the
corrective reformulations entailed in recasts may not always be perceived by learners
because they can be easily overridden by their functional properties in meaning-based

classrooms.

Although Lyster’s main objective was to investigate “aspects of L2 classroom discourse
that may minimize the perceptual salience of recasts and thereby limit their propensity to
be noticed as negative evidence” (p. 56), noticing was not directly operationalized in the
study. This task was carried over by Egi (2007b), who explored how linguistic targets,
length, and number of changes might affect learners’ noticing and interpretation of
recasts. Recasts of morphosyntactic and lexical errors were provided to learners during

task-based activities. Information of their noticing and interpretation of recasts was

35

gathered with immediate recall and stimulated recall. Results indicated that recasts were
occasionally interpreted as responses to content when they were long and substantially
different from learners’ problematic utterances. In contrast, when recasts were short and
closely resembled the original utterances, learners were signiﬁcantly more likely to attend
to the linguistic evidence. The patterns were observed in both morphosyntactic and
lexical recasts. These ﬁndings suggest that length and number of changes might partially

determine the explicitness of recasts and thus affect learners’ abilities to interpret them.

A study about learners’ interpretation of feedback that has been frequently cited is
Mackey et a1. (2000) as described in Section 2.4.2.2. The characteristic the researchers
focused on was the linguistic target of interactive feedback. Drawing on stimulated recall
comments, they found that learners were more accurate in their perception of lexical,
semantic, and phonological feedback than morphosyntactic feedback. In a more recent
study carried out in Arabic foreign language classrooms, Mackey and other colleagues
(2007) again examined learners’ interpretation of feedback in relation to its linguistic
target. Learners engaged in lessons on a range of linguistic targets (e.g. phonology,
morphology/lexis, and syntax) in a number of different ways. After the lessons, both
teachers and students participated in a stimulated recall session. Their comments were
then analyzed to provide information about whether the learners understood the
intentions of the teachers who provided the corrective feedback. The researchers found
that learners’ perception and teachers’ intention about the linguistic target of corrective
feedback overlapped the most when the feedback concerned lexis and was provided

explicitly.

36

To sum up, learners’ noticing and perception of feedback can be inﬂuenced by its
characteristics such as length, number of changes, and linguistic targets. Many studies
which examined learners’ noticing and perception of feedback concentrated on the
characteristics of one particular type of feedback or one feature of different types of
feedback. It is interesting to explore learners’ noticing of feedback in relation to different

characteristics across different types of feedback.

2.5.2 N onlinguistic cues in feedback

A second factor that may affect the effectiveness of feedback is nonlinguistic cues,
including paralinguistic cues and extralinguistic cues. In his book which examines
literacy from multidimensional and interdisciplinary perspectives, Kucer (2005) gives a
clear deﬁnition of the two terms in oral language. Paralinguistic cues are features that are
part of language but are not linguistic in nature, for example, pitch, intonation, stress, and
rhythm. Extralinguistic cues are those that often accompany the use of language but are
not part of language, for examples, facial expressions, gestures, physical movements,
settings, and objects in the environment. While Kucer classiﬁes nonverbal behavior as
extralinguistic cues, not all researchers do so. Davies (2006), for example, categorizes
focus-on-forrn episodes with body language as paralinguistic focus on form. Gullberg and
McCafferty (2008) argue that to study gesture is to study the paralinguistic modes of
interaction. The word “paralinguistic” as used by these researchers would be the word
“extralinguistic” as used by Kucer. The present study adopted Kucer’s deﬁnition of

paralinguistic and extralinguistic cues.

37

Talking about learners’ noticing of recasts, some researchers have suggested that
paralinguistic and extralinguistic cues might provide additional communicative clues to
learners. Such observations have come from both L1 and L2 acquisition literature. In the
L1 acquisition literature, Farrar (1992) suggests that paralinguistic cues might contribute
to children’s elimination of certain incorrect rules. In the SLA literature, Long (2007)
observes that there is some suggestive evidence that subtle prosodic and extralinguistic
cues can help with the disambiguation of recasts. Such suggestive evidence can be found
in recast studies. Sheen (2004), for example, in a study which examined the similarities
and differences among teachers’ use of corrective feedback and learners’ uptake and
repair across four instructional settings found that some teachers in New Zealand ESL
and Korean EFL classrooms recast with rising intonation and emphasis, which might
have partially increased the salience of recasts and hence learners’ noticing of recasts.
Loewen and Philp’s (2006) study conﬁrmed the effect of intonation and stress. In
examining the effect of recasts, they found that declarative intonation and stress, among
other characteristics of recasts, predicted successful uptake whereas interrogative
intonation was predictive of test scores. As for extralinguistic cues, Sirne (2006) explored
the meanings that EFL learners attributed to teachers’ gestures. Using a stimulated recall
protocol, she demonstrated that learners generally attended to teachers’ gestures, and the
gestures generally helped to clarify verbally expressed meanings. Davies (2006)
examined the effect of body language (deﬁned by David as paralinguistic cues) on
learners’ production of uptake by comparing implicit episodes with or without

extralinguistic cues. It was found that episodes with extralinguistic cues tended to result

38

in more learner uptake than topic continuation while those without extralinguistic cues

tended to lead to more topic continuation than learner uptake.

The role of nonlinguistic cues in learning and whether learners attended to them at all,
however, are contentious. There is counterevidence too. Carpenter et a1. (2006) compared
the role of linguistic contexts and nonverbal cues in advanced ESL learners’
interpretation of recasts. Two types of videos clips with recasts and repetitions were
created: one with both learner utterances and the responses and one with the responses
only. Two groups of learners were asked to view the videos and indicate what responses
they were hearing. It was found that the utterance-response group was signiﬁcantly more
successful than the response-only group in recognizing recasts, indicating that the
immediate discourse context played an important role in learners’ perceptions of recasts.
The think-aloud protocol from a subset of 14 participants, on the other hand, suggested
that among a total of 252 comments there was only one clear reference to the use of a
facial expression as a clue to the nature of a recast. The researchers hence concluded that
the learners, even those who could not draw on linguistic context, did not use nonverbal
cues to interpret recasts. There have also been studies which suggest that even when
learners attend to nonlinguistic cues, the effect is not always positive. In a descriptive
study, F araco and Kida (2008) found that while teachers’ gaze worked at the dynamic
level of interaction and their gestures functioned like a metalinguistic gloss, visual
information could also lead to misunderstandings. For example, when teachers softened

their corrective moves by breaking their gaze toward a learner or when they produced the

39

same gesture upon both the repetition and correction of a learner mistake, learners might

not understand the corrective nature of the teacher response.

Altogether, there is no agreement on the role of paralinguistic and extralinguistic cues in
learners’ noticing of feedback and hence the effectiveness of feedback. Although there is
suggestive evidence that it might affect noticing and learning, more research is needed to

explore it further.

2.5.3 Metalanguage in feedback

There are both proponents and opponents of metalanguage use in the language classroom.
Many researchers are strongly against terminology use in language teaching. Corder
(1973), for example, believes that the level of abstraction created by terminology adds to
learners’ burdens. Halliwell (1993) speculates that terminology can only enable learners
to talk about language rather than to use it for communication. Unlike Coder and
Halliwell, other researchers have assigned terminology to a more important role. Berman
(1979) argues that grammatical labels provide a shortcut to all kinds of devious
circurnlocutions. Faech (1985) observes that metalanguage is a monitor for language
production and serves as an important heuristic tool for learners to elicit information
about a language. In the call for studies of explicit language talk, Borg (1998, 1999)
conducted two empirical studies which examined teachers’ use of metalanguage and
factors that impinged on their use of metalanguage. From his ﬁndings, Borg (1998) '

concluded that metalanguage can make students aware of the language system they are

40

learning and how it compares with their ﬁrst language. Moreover, it also helps students

become aware of their own linguistic needs.

In recent years, many researchers have referred to metalinguistic feedback in their studies
about the effectiveness of feedback (e.g., Lyster & Ranta, 1997; Panova & Lyster, 2002;
Sheen, 2004), which often involves the use of metalanguage. Metalinguistic feedback has
been found to be more successful than recasts in eliciting learner uptake and repair in
some studies. Nevertheless, metalanguage is not the (only) focus of investigation in these
studies. One study which did focus on this aspect of interactive feedback is Basturkrnen
et al.’s (2002) examination of teachers’ and students’ use of metalanguage and its

relationship with learner uptake.

The study was carried out in two communicative classrooms, one of the intermediate
level and one of the pre-intermediate level. The researchers found that the use of
metalanguage and the occurrence of uptake varied in different types of focus-on-form
episodes. The two occurred together in 50.3% of student-initiated episodes but only over
5% of teacher-initiated and reactive episodes. Further analysis revealed that the
relationship between the use of metalanguage and successful uptake also differed
according to different types of focus-on-form episodes. While metalanguage and
successful uptake occurred together in over 44% of student-initiated episodes, the
percentages of such occurrence in teacher-initiated and reactive episodes were as low as
27.3% and 9.5% respectively. Compared with discussions about the use of grammatical

terminology, Basturkrnen et al’s study looked into a broader scope of metalanguage by

41

 

situating its use in a more inclusive environment, focus-on-form episodes, which might
contain not only grammar but also other language units such as vocabulary and

pronunciation.

The uptake studies mentioned above provide indicative evidence that metalanguage use
may affect learners’ noticing of feedback and the effectiveness of feedback. However,
there is no general conclusion due to the scarcity of research in this area. More studies are
needed to provide empirical evidence. Metalanguage use is therefore another aspect of

feedback the current study aimed to explore.

2.6 Summary

This literature review covers some of the most important components of SLA: input,
output, interaction, feedback, and attention and noticing. In the teaching and learning
processes, these components are intertwined. Through the negotiation of meaning in
interaction, learners produce the target language and receive feedback from the
interlocutor. When the input in the feedback is noticed and attended to, learners may
improve their language production, which in turn may affect interaction and the ultimate
learning. Recently, Mackey and Polio (2009) nicely present these intertwining tenets of
the interaction approach in a diagram (Figure 1). In addition to cognitive factors, they
include social factors such as motivation. Limited by time and resources, the current

study only focused on: learners’ noticing of feedback and its effect on learning.

42

 

Figure 1 Main tenets of the interaction approach

 

 
  
 

 

 
 
 
  
 
 
 

 

 

 

 

 

 

 

 

   

 
 
  

  

 

 

 

 

    
     
     
     
     
    
  
   
 
 
 

    
 

  

 

 

 

 
  
 

 

 

 

 

    

 

 
   
 

 

 

 

 
    

Feedback
Corrective feedback
can occur during
Social factors interaction, often in
Social factors, - the form of
including ' - -
motivation, can Diggitfo—i-lhigd .
affect access to facilitates SL A, Learmng
input, type and '\ perhaps through What counts
frequency of drawing attention to as evrdence
feedback, and mismatches. of learmng rs
Willingness to \ open to
produce output, dgscussron,
as well as the ut many
attention learners Input researchers
pay to language. \ Attention Exposure to input believe thlzlrt
Attention is a (stimuli) is essential but :ﬁlazrzrrgsain
not sufﬁcient for .
central Ian a e ac ' 't' p roductron,
component in gu g . qursr ion. comprehensi
L2 Interaction can help on or
Cognitive development. make mp 1‘“ more Q awareness
factors Interaction helps comprehensrble. indicate
Individual factors / draw learners’ H some
such as working attention to ASPCCIS 0f the input evidence of
memory and / features in the that are processed learning.
developmental input, while by the learner can Clearly,
level can affect producing serve as POSIUVe long-term
the amount of OUIPUL 0’ when evidence. If delayed post-
attention a learner receiving attended t°»_ they tests and
can pay to input ./ feedback. can serve as intake. evidence of
and feedback, as change
well as the kinds across a
of output they variety of
produce. Such contexts rs
factors might also Output the best
inﬂuence the Production 0f evidence of
types of input '/ _ output during learning.
learners receive : mteractron allows
from their for hypothesrs
interlocutors. testing and the
development of
automaticity and
can lead to leamers’

 

 

   

attending to form.

  

(Mackey & Polio, 2009, p. 5)

43

2.7 Research questions
The current study explores learners’ noticing of teacher feedback and the effectiveness of
teacher feedback in relation to the characteristics of feedback episodes, nonlinguistic cues,
and metalanguage. The research questions are:
1. Do the characteristics of teacher feedback episodes affect learners’ noticing and
learning?
2. Do the nonlinguistic cues in teacher feedback affect learners’ noticing and
learning?

3. Does the metalanguage in teacher feedback affect learners’ noticing and learning?

In many of the studies reviewed above, only corrective feedback was examined. In this
study, however, teachers’ feedback about questions initiated by learners or by teachers
themselves was also investigated. A feedback episode is therefore deﬁned as a sequence
beginning with an erroneous utterance, a query by the learner, or a question by the
teacher, followed by feedback provided by the teacher, and ending with the learner’s
reaction to the teacher’s feedback where applicable (adapted from Lyster and Ranta,
1997). It is important to point out that the term “feedback episode” as deﬁned here is not
exactly the same as the term “focus-on-form episode” as used in the studies reviewed
above. Focus on form in those studies occurred in a discourse that was primarily meaning
centered; the attention to language form was incidental and transitory; and a variety of
language forms could be attended to in a single lesson (Ellis et al., 2001a, 2001b).'ln the
current study, apart from such focus on form activities, there were also lessons where pre-

selected forrns were taught in communicative activities, for example, talking in pairs

44

 

about one’s summer plan with the “be going to” structure; describing a timeline in groups
to practice past tense; and playing a game as a whole class using adverbial clauses. On
such occasions, the discourse is not strictly meaning centered but more form oriented; the
attention to language form is preplanned and proactive rather than incidental and
transitory; and only a limited rather than a large number of language forms might be
attended to in a single lesson. In the effort to disentangle the myth of focus on form, Ellis
et a1. name such attention to language form as the stretched version of Long’s (1991)
focus on form. The current study does not concern itself so much with the nature of the
discourse. Any language episode involving teacher feedback, as long as there was an
observable learner problem, was considered as a feedback episode. To put it simply, the
term “feedback episode” in this study embraces both focus on form in its original
deﬁnition and focus on form in its stretched deﬁnition. Revolving around this notion of
feedback episode, the next ﬁve chapters detail the design and ﬁndings of the current

study.

45

CHAPTER 3

METHODOLOGY

3.1 Introduction

This chapter outlines the methodological issues of the study. These include the settings of
the study, the general information of the participants, the instruments and procedures of
data collection, and the schemes of data analysis. In addition, the results of reliability

check are presented and the statistical analysis of the data is introduced.

3.2 Teaching context

The study took place in an intensive English program at a large university in the United
States. The program is designed to help students learn communication and academic
skills that are necessary in daily life and academic work. Students are put into different
course levels according to the results of placement tests. The lessons at each level cover
grammar, reading, writing, speaking, and/or listening skills. The courses offered are
based on a communicative curriculum. In this study, four integrated listening and
speaking classes, two integrated reading and writing classes, two academic reading
classes, and one grammar class were observed. In each class, learners were involved in
all kind of communicative activities, discussing and comparing answers in groups,
summarizing texts and stories in pairs, doing information gap activities, to just name a
few. Even the grammar class, as the teacher put it: “is surprisingly communicative even

though the students need a lot of help” (Observation note, April 14, 2009).

46

 

3.3 Participants

A total of eight teachers and nine classes participated in the study, with one teacher
teaching two parallel classes. All the teachers except one have English as their native
language. The one exception is a teacher whose ﬁrst language is Spanish. However, this
teacher grew up and went to school in the US. Her English is therefore like that of a
native speaker. Among the eight instructors, six are females and two are males. Their
teaching experience varies ﬁom 1.5 years to 37 years, with an average 10.3 years. At the
time of the research, four of the teachers had a MA degree in TESOL; two held an
elementary and middle school teaching certiﬁcate; one had a teaching English as a

foreign language certiﬁcate; and still another one was working on her MA degree.

A total of 117 students signed the consent form. Fifteen students who chose not to do so
were excluded from the study. The 117 students, with an average age of 20.4, are ﬁ'om a
variety of L1 backgrounds, including Chinese, Arabic, Korean, and Japanese. All of them
had studied English before they came to the US, with an average 7.3 years of previous
English study. The majority of students had arrived in the US not too long, with an
average 5.7 months of stay. Only a few had been in the country for 12 months or more.
Most of the students enrolled into the intensive English program to improve their English
either to qualify for admission to regular American academic institutions or to return to

their home country for ﬁrrther career development.

47

3.4 Instruments

According to the research questions, two variables were measured in the present study:
learners’ noticing of teacher feedback and the effectiveness of the feedback. In response
to the call for triangulating data collection methods, each variable was operationalized
with two measures. As shown in Figure 2, the noticing of feedback was examined with
learners’ stimulated recall comments and the effectiveness of feedback was assessed with
individualized tests. Learner uptake was used as an additional measurement for both

noticing and learning. All three measures were developed from teacher feedback episodes.

Figure 2 Measurement of the noticing and effect of teacher feedback

 

 

 

 

 

 

 

 

 

Stimulated recall
, , , , ’ ' interview
Noticing of r ’ '
feedback \
II: Learner uptake Feedback
x’ ’ episodes

 

 

 

 

 

 

 

 

Effectiveness of ’ ' ’
feedback

 

 

 

‘ ~ , Individualized
tests

 

 

Accordingly, three instruments were used in the study: classroom observations,
stimulated recall interviews, and language tests; Classroom observations'provided data
for learner uptake as well as the characteristics of feedback episodes, the occurrence of

nonlinguistic cues, and teachers’ metalanguage use. The stimulated recall interviews and

48

language tests were also based on the observations. Below is a detailed illustration of the

three instruments.

3.4.1 Observations

Classroom observations, as shown in the studies reviewed in Chapter 2 (e. g., Basturkrnen
et al., 2002; Loewen & Philp, 2006; Lyster & Ranta, 1997; Oliver & Mackey, 2003), are
a data collection technique widely adopted in SLA. As Gass and Mackey (2007) put it,
observations “allow researchers to gather detailed data on the events, interactions, and
patterns of language use within particular foreign and second language classroom
settings” (p. 165). In the present study, four to six observations, depending on teachers
and students’ consent, were made in each classroom. During the observations, I served as
a non-participant observer and did not participate in any of the classroom activities. My
job was to monitor the recording instruments and to take notes. A Sony digital voice
recorder with a clip-on microphone attached to the teacher was used to record the oral
exchanges in the classrooms. As Loewen (2004) notes, this arrangement can record all
teacher-student interactions, whether the students interact with the teacher as a whole
class, in small groups, or one on one. However, it does not capture nonverbal interactions,
which is an important component I intended to examine. To tackle this problem, in
addition to the voice recorder, a Sony high deﬁnition video camera with a built-in
speaker was set up wherever the teacher suggested was more convenient and less

obtrusive.

49

3.4.2 Stimulated recall interviews

Stimulated recall interviews, along with uptake from the observation data, were major
tools to collect noticing data. One to two recall interviews were conducted in each class
depending on the number of observations that were allowed by the teacher and learners.
In each class a minimum of four students were invited to participate in this session,
depending on the circumstances and time frame. In order to mitigate the reactive effect of
factors such as double-input exposure and verbalization, data from the interview
observation(s) were not reused for testing. That is, test items were developed from other
observations. In line with the recency rule proposed by Gass and Mackey (2000, 2007)
and to reduce the non-veridicality caused by memory decay, the interviews were
conducted within three days after the observation(s). It has been suggested that recall may
be more successful when the stimuli are presented in the same way they were presented
in the previous task so that the sequence of the cues is most likely to be consistent with
memory indexes in long-term memory (Egi, 2004). For this reason, it would be best to
have learners view the recordings of whole or big chunks of class periods. However, due
to time constraints, and also in the effort not to burden learners with too much extra work,
excerpts of the video recordings of classroom interactions (i.e. feedback episodes) were
used to serve as recall cues. To examine learners’ noticing of teacher feedback in
different circumstances, clips from a variety of settings were selected, some from
interactions between teachers and individual students, some from group work, and some
from whole class discussions. Only excerpts of both good audio and visual quality~were
selected. It was hoped that high quality stimuli would be able to help learners retrieve

their memory more easily and more thoroughly, thus revealing more about learners’

50

 

cognitive processes and enhancing the veridicality of the noticing data. Because one of
the objectives of the study is to examine the effect of nonlinguistic cues on learners’
noticing of teacher feedback and yet nonlinguistic cues did not occur frequently in every
observation, episodes with nonlinguistic cues received special attention in the excerpt
selection process and were selected for use as long as they are of good quality. However,
this is not an issue for teachers’ metalanguage use now that metalinguistic terms occurred

much more frequently than nonlinguistic cues.

After the excerpts were selected, learners who were engaged in these episodes were
invited to view the video clips and report what they were thinking. On some occasions
observers of an episode1 were also asked to participate in the interview session. This
generally happened with episodes from group work or whole class discussion. It was
hoped that observers’ reports could help reveal learners’ noticing of teacher feedback
when they were not directly engaged in teacher-student interactions. To reduce
interviewer effect, I did a pre-viewing debrieﬁng before each interview, explaining to
learners, with speciﬁc examples, that they should report their then-thinking, not their
now-thinking, and that they did not need to worry about what I thought of their
performance or reports. They were also informed that they could stop the tape whenever
they had a comment to make. During the interview, at points where there were

communication breakdowns, for example, when the learners looked confused, when there

 

' The observer of a feedback episode refers to a learner who is not directly involved in a teacher-
student interaction but still experiences it. For example, in a whole class discussion, a learner
might initiate a question which other learners turn out not to understand either. As the teacher
gives feedback to the question-initiator, other learners hear it too and hence are observers of the
episode.

51

 

was a long pause in the conversation, and when the learners stuttered, I also stopped the

tape to ask them what they were thinking at the moment.

Ericsson and Simon (1993) suggest that participants should only be asked to verbalize
information that has been attended to. Only such information can be encoded in memory
and available for verbal reports. They also recommend that the instructions for recall
should avoid a request for reasoning and elaboration from respondents. Such requests
may result in participants altering their thoughts in order to generate the information
requested. For these reasons, I only asked questions such as “What were you thinking
then?” or “Any thoughts there?” not questions such as “Why were you thinking so?” or
“Could you tell me more about that?” When learners said they were not thinking of

anything, I did not push them to think further.

One thing I noticed in the recall session is that with English as a second language, some
learners, especially those at a lower proﬁciency level, had difﬁculty expressing
themselves on some occasions. For example, a few learners were not able to remember a
particular word and so asked “How do you say. . .?” or “What’s the opposite word of . . .?”
In this case, I gave them some language help as I saw necessary. Meanwhile, I took
caution not to give them extra information. I hoped this approach would help prevent the
unnecessary loss of valuable data due to learners’ limited proﬁciency level. Another thing
that stood out in the recall session is that about 60% of the learners who did the interview
reported that their native language is Chinese, which happens to be my native language

too. To reduce non-veridicality to the largest extent, and also given that the purpose of

52

the recall interview is to obtain noticing data, not to test learners’ language skills, these
learners were allowed to speak Chinese if they wanted to. It was hoped that this would

give them a chance to express their thoughts more thoroughly and more accurately.

3.4.3 Language testing

Loewen (2002) developed a battery of individualized tests in his study on the
effectiveness of incidental focus on form. Given the careful design of his study, his
testing method was adopted in the current study. Altogether, there were six types of test
items: 1) supply meaning, 2) supply word/phrase, 3) correction, 4) pronunciation, 5)

spelling, and 6) supply information.

In Supply Meaning, learners were asked to provide the meaning or deﬁnition for a word
or phrase. For example:

(6) Please tell me the meaning of the word: Lower.

“Lower” had been discussed in a feedback episode where the learner asked the teacher

what the word meant and the teacher explained it with synonyms and gestures.

In Supply Word/Phrase, I provided the meaning of a word/phrase and learners were asked
to give the word/phrase. For example:

(7) When the quality of the soil gets worse and worse, or there is less and less
nutrition in the soil, we say the soil is losing F

 

Here the expected answer is “fertility”, which had been discussed in a feedback episode.
In this type of tests, the ﬁrst letter of the word was given in order to ensure that learners

would not give another equally correct word.

53

In Correction, I read incorrect sentences to learners and explicitly asked learners to
correct them. For example:
(8) The following sentence is incorrect or inappropriate. Please listen carefully and
tell me how you could make it better.
During the spring break, I will go to many place in California.

The inaccurate form “many place” had been corrected in a teacher-student interaction.

The leamer’s task was to listen to the sentence and then correct it.

In pronunciation tests, learners were asked to read a sentence that contained a targeted
word and then read the word in isolation. For example:
(9) Please read aloud the following sentence and word.
The word passionate means very eager or exited.
Eager
Here the sentence is very similar to the one the learner had produced in a feedback

episode, where the word “eager” was pronounced incorrectly and was recast by the

teacher.

In spelling tests, learners were asked to orally spell a word that had received attention in a
feedback episode. For example:

(10) Please spell the following word: sibling

In a previous teacher-student interaction, the learner had received feedback from the

teacher about the spelling of the word “sibling”.

54

Items which required learners to supply information were often associated with feedback
episodes where students asked metalinguistic questions. The prompts for such items
might be questions such as “Is the word ‘lower’ positive or negative?” This question was
developed from an episode where the learner consulted the teacher whether lower is

positive or negative.

Similar to Loewen’s study, there was no pretest but an immediate test and a delayed test
based on different observations. However, three changes were made as I saw necessary in
the context of the current study. In Loewen’s study, only students who initiated a
question were tested on a particular linguistic feature. In the current study, I noticed that
in many cases when learners were discussing questions, they decided together that they
needed the teacher’s help. Even though only one student initiated a question, other group
members would listen and respond to the teacher. For example:
(1 1) Maj: T, we don’t understand about me too.

T: Oh me too is useful. So what are you going to do this summer? Are you going

to do anything the same?

Maj: (Shakes head)

T: Nothing? What about this one? (Points at Muh’s paper) Are you going to study

English?

Maji: Yeah, but I didn’t write it.

T: Oh ok, but still in a conversation, if he said I’m going to study English, you can

say [me too].

Muh: [Me too]. Yeah. (Nods)

T: Yeah. It’s true you didn’t write it, but that’s ok, so me too. Understand?

Maj: Yeah I understand. (Nods)
In this example, the question initiator uses the word “we” when asking the question,
suggesting that both he and his partner need help with the phrase “me too.” Although

Muh is not the one who asks the question, he listens to the teacher’s explanation and

responds to her by saying “me too” and nodding. This learner therefore also experiences

55

the feedback episode. This is also the case when the teacher initiated a question and more
than one student showed or said they did not know the answer or when a group made the
same mistake and the teacher corrected them all. On such occasions, the student(s) who

was (were) not the initiator of a question was (were) also tested on the same item.

In the case of supplying meaning and spelling, Loewen did not provide a full sentence
containing the targeted word or phrase to test takers. This is probably a good way to
prevent learners from doing guess work according to the context and more precisely tap
into their implicit knowledge about the targeted linguistic features. As I attempted to ask
learners such questions, however, it was found that learners sometimes talked about
another word or phrase as a result of homophones or similar words. For example:
(12) Please spell the following word: phase
Among seven learners who were asked this question, four gave the word “face” without
hesitation. In order to make the prompt clearer, it was decided that a sentence containing
the targeted word or phrase could be provided when necessary. The prompt of the above
example thus would become:

They’re in the second phase of love. Tell me the meaning of the word: phase.
By providing the sentence, learners would hopefully be able to see the “phase” I was
asking about is not the body part “face”, which would not make sense in this context. In
order to minimize additional comprehension problems, the sentence was generally very
sirrrilar to, sometimes the same as, the one that occurred in'the original feedback episode.
On the other hand, it was realized that the addition of such sentences, especially in the

case of supplying meaning, could easily increase the chance of wild guesses. To address

56

this problem, the sentence was made as short as possible and contained as little additional

information as possible.

In order to ensure the reliability and validity of his tests, Loewen made a great effort to
design test items that reﬂect the modality of focus-on-form episodes in real-time
classroom interactions (e. g., all test types were oral rather than written). To further
address the problems incurred by the mismatch between test and non-test situations, an
additional change was made in line with one of Ellis’ (2005) most important criteria for
the assessment of implicit and explicit knowledge, timing. As Ellis observes (see 2.3.2.1),
a test task that taps into a leamer’s implicit knowledge does not allow the learner to spend
an extended time planning a response. This proved true in Ellis’ (2005) psychometric
study. Following this rule, learners were encouraged to answer the questions as quickly
as they could. In addition, the time learners spent answering each question was recorded.
With that said, however, there is no guarantee that the tests indeed or merely measured
learners’ implicit knowledge. As Ellis observes, “even if task conditions that inclined
learners to use one type of knowledge in preference to the other could be identiﬁed, it
would be impossible to construct tasks that would provide pure measures of the two types

of knowledge” (p. 153).

3.5 Procedures
Before starting formal observations in a class, I distributed and collected the consent form:
as well as a small background survey in the class. With the teacher’s consent, 1 then

stayed in the classroom for an hour or two to get familiar with the teacher’s teaching style

57

and to select possible positions to put the video camera. This also gave me the
opportunity to talk to learners and to try to establish a rapport with them, which could

hopefully help increase the ecological validity of the study.

Data collection in each class took three to four weeks, depending on how many times
teachers and students agreed to have me with them. There was no ﬁxed schedule due to
the complexity of particular contexts. Sometimes learners spent all their class time taking
a unit test. Sometimes the majority of class time was devoted to silent writing. Sometimes
teachers were out of town for a conference. On the other hand, both the tests and
stimulated recall interviews, especially the latter, were highly time sensitive.
Consequently, when anything unexpected happened, the observation and the test or
interview based on it would be cancelled and new arrangements would be made
accordingly. Because the tests and interviews were based on different observations, such

adjustments should not severely affect the results.

Generally, when there were six observations, Observations 1 and 2 would serve as the
basis of the delayed test, Observations 3 and 4 would serve as the basis of the stimulated
recall interviews, and Observations 5 and 6 would serve as the basis of the immediate test.
When there were four to ﬁve observations, the delayed test would be developed from
Observation 1, the recall cues for the interviews would be selected from Observation 2,
and the immediate test would be developed from Observation 3. Observations 4 and 5

would supply additional material for the tests and interviews and I decided whether more

58

test items were needed and whether more recall interviews should be conducted

according to what I had already.

The immediate test was administered one to three days after the observation from which
it was developed while the delayed test was administered 14 to 16 days after the
observation from which it was developed. As Loewen (2002) notes, giving a range of
days for the immediate test and delayed test is to allow some leeway for administering
tests. For example, if students cannot take the immediate test on one day, they can take it
in the other two days. Ideally, the stimulated recall interviews should be conducted right
after an observation. However, this turned out to be very difﬁcult in the current study. For
one thing, it was not quite possible for me to identify feedback episodes and select recall
cues during the observation: it took quite a while even just to import the video to a laptop
and to convert it to an easily viewable and editable format. For another thing, learners
often had other classes to attend right after an observation or they were often too tired
after a long day and were reluctant to do the interview. More often than not, it was even
not easy to arrange a meeting time in the following day(s). For such reasons, it was
decided that a three-day range was also necessary for the stimulated recall interviews.
Meanwhile, every effort was made to shorten the time between the observation and the
interview. The average time span is about 48 hours. If a learner for some reason was not
able to take the tests or to do the interviews within the three-day deadline, the tests and

interviews would be cancelled.

59

 

3.6 Data analysis

Three kinds of data were analyzed: classroom observations, stimulated recall comments,
and test results. The observation data were coded in terms of the characteristics of
feedback episodes, the nonlinguistic cues in teacher feedback, the metalanguage in
teacher feedback, and learner uptake. The stimulated recall comments were coded
according to their relevancy to language in general, the target features in the stimuli, and
teacher feedback in the stimuli. Finally, learner responses in the immediate test and
delayed test were coded as correct, incorrect, or partially correct. After all data were
coded, a random sample of about 10% of each data set was coded by a second rater and

chi-square tests were performed where necessary and possible.

3.6.1 The coding of observation data

The ﬁrst step of coding the observation data is to identify all episodes that contain teacher
feedback. These feedback episodes were then coded according to their characteristics, the
nonlinguistic cues teachers used, the metalanguage teachers used, and learner responses
to the feedback. Some of these variables were then ﬁrrther analyzed. The coding process

is shown in Figure 3.

3.6.1.1 The coding of the characteristics of feedback episodes

As shown in the literature review, R. Ellis and colleagues (e.g., Ellis et al., 2001a; 2001b;
Loewen, 2003, 2004, 2005) have developed 'a coding scheme to examine the
characteristics of focus-on-form episodes. A careful look at these characteristics (see

Table 1) reveals that all of them revolve around feedback. It is no exaggeration to say that

60

 

feedback is at the core of focus on form. The characteristics of focus-on-form episodes
therefore can also apply to feedback episodes. In the current study, the coding scheme in

Loewen (2004) was adopted to examine the general characteristics of feedback episodes.

Figure 3 Coding process of the observation data

 

Transcription of
recordings

1

Feedback
episodes

/ \

 

 

 

 

 

 

 

Characteristics of

Non-linguistic

Occurrence of

feedback episodes cues Teachers’ tak t k
(type, source, (paralinguistic metalanguage “Pno Egg}: 6
response, linguistic cues, use (present, no ’
focus, directness, extralmgurstrc absent) opportunity)

 

 

 

 

 

 

 

 

complexity, emphasis) cues)

ParalinguistiCV Metalanguage Uptake
V V

 

 

Extralinguistic cues

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Word stress, Gestures, Technical
rising whole body terms only, Successful,
intonation, acting, head non-technical partially
elrcrtrng stop, movements, terms only, successful,
dragging vorce, , techmcal unsuccessful,
mimicking facral terms plus acknowledge,
sound, expressrons, non-technical inconclusive
combination combination terms
Gestures
r
Iconics,
metaphorics,
deictics,
beats,
combination

 

 

 

61

 

As illustrated in Table 1, seven characteristics were examined: type, source, response,
linguistic focus, directness, complexity, and emphasis. In his study, Loewen (2004)
included uptake and successful uptake as two characteristics of focus on form. In the
present study, these two variables were coded separately because they were important

tools to examine both learners’ noticing of teacher feedback and their language

 

 

improvement.
Table l Coding scheme for the characteristics of feedback episodes
Characteristics Deﬁnition Categories
How the feedback T eacherirnrtrated: teacher starts a question
Type . . . . . Student-mrtrated. student starts a query
episode rs mrtrated . .
Reactive. teacher responds to erroneous utterance
Code: inaccurate use of language forms without
communication problems
What causes the Message: problems understanding meamng
Source . . . . . .
feedback episode Semantic: negotiation involves meamng of a
word/phrase but there’s no communication
breakdown
Provide: teacher recasts or provides information
Res use How the feedback is about language form
p0 given Elicit: teacher strategically draws out language
information from student
Vocabulary
Linguistic What language aspect is Grammar
focus targeted at Pronunciation
Spelling
Directness How explicit the Direct: explicit (e.g., metalingual explanation)
feedback is Indirect: implicit (e.g., recasts)
Com lexi t How many feedback Simple: one feedback move
p y moves there are Complex: more than one feedback move
. . Combmatron Of Light: Indirect and simple
Emphasrs complexrty and . .
d' Heavy. Direct and/or complex
rrectness

 

Adapted from Loewen (2004, p. 165-166)

62

Type

Type refers to how a feedback episode is initiated. This characteristic is classiﬁed into
three categories: teacher-initiated, student-initiated, and reactive. In a teacher-initiated
episode, the teacher starts a language question and gives learners feedback if they can not
provide an answer to the question or provide an incorrect answer. In a student-initiated
episode, the learner starts a query about a language feature and the teacher gives feedback
to the learner. In a reactive episode, the learner produces an erroneous utterance and the

teacher responds to the error. Below are three examples which show the three types of

episodes.

(13) T: About the football game. It’s coed. Do you know what coed means?
Class: No.
T: It means boys and girls together.
Class: Ah.

(14) Lofa: What to say when there are three semesters?

T: So if there were three, they would be called trimesters.

Lofa: Yeah. (Smiles)
(15) T: Anybody have something different? Le?

Le: She don’t want children.

T: Ok, the ﬁrst thing is she doesn’t want children. (Writes on board)

Le: (N o more response)
Here the ﬁrst episode is initiated by the teacher asking the class about the meaning of the
word “coed” while the second episode is initiated by a learner asking the teacher about
the word to describe three semesters. These two episodes are teacher-initiated and
student-initiated respectively. In the third example, the learner uses the wrong form of the

helping verb “do” and the teacher recasts him. The teacher’s feedback is a reaction to the ‘

leamer’s response. It is therefore a reactive episode.

63

Source

This characteristic describes what causes a feedback episode. It is put into three
categories: code, message, and semantic. A code-related episode results from the
inaccurate use of language forms without communication problems. The purpose of
teacher feedback is to increase learners’ language accuracy. A message-related episode
results from the inaccurate use of language forms with communication problems. The
purpose of the episode is to solve comprehension problems. In a semantic episode, there
may not be a communication breakdown; the learners might have understood the gist of
the discourse but wants to know the meaning or usage of a speciﬁc word or phrase.
Below are three examples that show the three types of sources.

(16) Vah: And the two couple, one couple, they’re enjoying, they’re enjoying their
love.
T: Ok.
Vah: Yeah. They close their eyes. So they’re blind.
T: Ok, they’re blind-, they’re blindly in love, ok.
Vah: They’re blindly in love.

(17) Happ: Fashion trend. (“Fashion” sounds like “ﬁshing”)
T: Huh?
Happ: Fashion. (“F ashion” still sounds like “ﬁshing”)
T: Fishing? (Hand gestures ﬁshing)
Happ: Fashion. (A little bit clearer)
T: Fashion, fashion trend, yeah. Eh ladies you know about fashion trends?

(18) T: Cute.
Tai: It’s me.
Vah: You!
Class: (Laugh)
T: Usually they say girls are cute, boys are, babies are cute, but boys are usually
not cute.
Vah: Pet, pet.
T: Pets are cute. So the way they look. All right?
Ara: Yeah.
Tai: (No response)

64

The three episodes are code-related, message-related, and semantic respectively. In the
ﬁrst episode, the learner is describing a picture to the teacher. There does not seem to be
a communication problem. The teacher recasts the learner to help him express his ideas in
a better way. Different from the teacher in the ﬁrst episode, the teacher in the second
episode does seem to have difﬁculty understanding what the learner is saying because of
his problematic pronunciation of the word “fashion”. The teacher attempts to solve the
comprehension problem with clariﬁcation requests. In the third episode, the learner uses
the word “cute” to describe himself, which the teacher explains is inappropriate. There is
no communication breakdown. The teacher’s feedback is a response to the fact that the

learner does not have a full understanding of how to use the word “cute”.

Response
Response refers to how the teacher’s feedback is given. It is classiﬁed into two categories:
provide and elicit. When the response type if provide, the teacher recasts the learner’s
erroneous utterance or provides information about a language feature. When the response
type is elicit, the teacher strategically draws out language information from the learner.
Below are two examples which show the two types of response.
(I 9) Yiw: It’s almost a pa-, pa- (Not being able to pronounce the word “passionate”)
T: Passionate.
Yiw: Passionate about his girlfriend.
(20) , Jacq: He thinks studying English is not good enough, so he went to here to study
English in the ELC.
T: He thinkS (Stresses “s”) studying English is not good enough back home?

(Rising intonation) That’s why he came here?
Jacq: He thought.

65

In the ﬁrst episode, the learner can not pronounce the word “passionate”. The teacher
simply models it for her. In the second episode, the learner uses the wrong tense of the
word “think”. Instead of providing the correct form, the teacher repeats the learner’s
erroneous utterance in a rising intonation, stresses the third person “s” in “thinks”, and
asks a question in the past tense. With such strategies, he successfully draws out the

correct form of “think” from the learner.

Linguistic focus
Linguistic focus refers to what language aspect is targeted at in a feedback episode. It can
be vocabulary, grammar, pronunciation, or spelling. In vocabulary episodes, the linguistic
focus can be the meaning of words, phrases, idioms, and sentences, word choice, non-
target derivations of words, pragmatic aspects of words or phrases such as their
appropriateness in speciﬁc social contexts, whether a word or phrase is positive or
negative, etc. In grammar episodes, the linguistic focus can be determiners, prepositions,
pronouns, word order, tense, verb morphology, auxiliaries, subject-verb agreement,
plurals, negation, question formation, plural —s, sentence construction, comparative and
superlative, part of speech, etc. In pronunciation episodes, the linguistic focus may
involve segmental and superasegrnental aspects of the phonological system such as the
pronunciation of words, stress, and intonation. In spelling episodes, the linguistic focus is
the orthographic form of words. Below are four examples which show the four types of
linguistic focus.
(21) Lyb: Like loosen the soil. Remove the soil. (Hand gestures loosening)

T: Move the soil?

Lyb: Loosen like eh exchange. (Hand gestures exchanging)
T: Not. . .Mm, yeah they need to tear the soil. (Hand gestures tearing)

66

 

Lyb: Yes.
(22) Fa: They came back in the evening and noticed that the milk was disappeared.

T: The milk HAD disappeared. (Stresses “had”)

Fa: Had disappeared.
(23) Gra: But they stopped dating because there was no chemistry- / 'tsemistri/

T: Chemistry.

Gra: Chemistry between them. (/'kemistri/)
(24) T: Jealous, to be envious.

Ara: How to spell it?

T: E-N-V-I-O-U-S.

Ara: N-V. (Writes on paper)

T: V-I-O-U-S.

Ara: (Finishes writing on paper without saying anything more)
In the ﬁrst episode, the teacher and the learner are seeking the right word to describe an
action. In the second episode, the teacher corrects the leamer’s inappropriate use of a
helping verb. In the third episode, the teacher is helping the learner with the
pronunciation of a word. In the fourth episode, the teacher and the learner are talking

about the spelling of the word “envious”. Accordingly, the linguistic focus of the four

episodes is vocabulary, grammar, pronunciation, and spelling respectively.

Directness

This characteristic describes how explicit teacher feedback is. There are two categories in
it: direct and indirect. Direct feedback is explicit, for example, explicit correction and
metalinguial explanation. Indirect feedback is implicit, for example, recasts, clariﬁcation
requests, and repetitions. Below are two examples for direct feedback and indirect
feedback.

(25) Zhou: What does that mean? The ﬁrst word.

T: Yes, eh testosterone. That’s a male hormone. That’s eh, something that’s inside
your body that the men have, not the women. Just men.

67

Zhou: Oh.
(26) Mei: (Talking about what is in a picture) And gas masker, masker.
T: Yeah, gas masks.
Mei: Masks.
In the ﬁrst example, the teacher explicitly explains the meaning of a word the learner asks

about. The feedback is therefore direct. In the second example, the teacher corrects the

learner’s erroneous production by recasting him. The feedback is therefore indirect.

Complexity
This characteristic involves the number of teacher feedback moves in a feedback episode.
There are two categories in it: simple and complex. When there is only one feedback
move by the teacher, it is simple. When there is more than one feedback move by the
teacher, it is complex. Below are two examples showing the complexity of feedback
episodes.
(27) Hasa: Eh can you spell pursue please?

T: P-U-R-S-U-E, pursue.

Hasa: P-U-R-S-U-E.
(28) Xin: What is interrninal?

T: Interrninal?

Xin: Yeah.

Lo: Not a word.

T: No. That’s not a word. (Shakes head)

Xin: Means forever.

T: Interrninal?

Xin: Eterrninal.

T: Eternal?

Xin: Yeah, eternal.
In the ﬁrst example, there is only one feedback move by the teacher. In the second

example, there are four. Accordingly, the ﬁrst episode was coded as simple while the

second one was coded as complex.

68

Emphasis
The last characteristic is a combination of complexity and directness. There are two
categories in it: light and heavy. Light episodes are indirect AND simple while heavy
episodes are direct OR complex or both. Below are three examples illustrating whether an
episode is light or heavy.
(29) T: What kind of work do they do?

Casu: A lawyer.

T: Yeah, they’re lawyers. [And]...

Casu: [And] they took whatever cases.
(30) Lofa: Can I say too many garbages?

T: Too much garbage.

Lofa: Too much garbage.

T: It’s not a count noun, so you can’t say garbages. Just

garbage.

Lofa: Too much garbage.

T: Yeah.
(31) Rouf: Which form of balance should I use?

T: Eh you want a noun. Something that’s not balanced. If something is not

balanced, it’s out of balance, it’s imbalanced, it’s unbalanced. That’s ok. Global

words, word forms.

Rouf: (Nods)
In the ﬁrst example, the teacher corrects the learner by recasting her inappropriate use of
the word “lawyer”. There is only one feedback move and it is implicit. This episode was
therefore coded as light. In the second example, the teacher provides a direct answer to
the leamer’s question and there is more than one feedback move. This episode was
therefore coded as heavy. In the third example, the teacher explicitly explains the usage

of the word “balance”. Although there is only one feedback move, it is direct. This

episode was therefore also coded as heavy.

69

The two examples in Table 2 illustrate how the same feedback episode was coded
according to the seven different characteristics. In the ﬁrst episode, the teacher asks a
question about the word “generate” and explains it when learners’ responses indicate that
they do not understand it. This episode was categorized as message-related because it
involves the meaning of a word with an apparent comprehension problem. It is teacher-
initiated because it is the teacher who starts the question. The feedback type is provide
because the teacher directly tells learners what the word means. The linguistic focus is
vocabulary because it involves word meaning. It is direct because the teacher explicitly

gives learners the answer. It is complex because there are two feedback moves by the

Table 2 Examples for the coding of the characteristics of feedback episodes

 

 

Episodes Characteristics Categories
(32) T: And what does generate mean? Type Teacher-initiated
Adu: Generate? Source Message
Lyb: Oppress, oppression. Response Provide
T: It just means make. It can produce. Linguistic focus Vocabulary
Adu: Produce. Directness Direct
T: That’s another way of saying it. Complexity Complex
Lyb: Produce. Emphasis Heavy
Type Reactive
(33) Lai: 1 was tooking an English class while Daniel 50““ C9df’
. . Response Ellcrt
was tookmg Spanish. . . .
_ . . . . . ngurstrc focus Grammar
T. Tooklng? (Rrsmg mtonatron) . .
Lai' T akin Directness Indirect
' g. Complexity Simple
Emphasis Light

 

teacher. It is heavy because the teacher’s feedback is both direct and complex. In the
second episode, the learner uses the wrong form of the verb “take”- but there does not
seem to be any cormnunication problem. Instead of explicitly correcting her, the teacher

brings out the correct form from the learner by repeating the incorrect form in a rising

70

 

intonation with one single move. This episode is therefore reactive, code-related, indirect,

simple, and light. The feedback type is elicit, with a focus on grammar.

3.6.1.2 The coding of nonlinguistic cues

As shown in Figure 3, two types of nonlinguistic cues were identiﬁed: paralinguistic cues
and extralinguistic cues. After all the episodes were classiﬁed as being with or without
the two types of cues, those with either type were further classiﬁed into more speciﬁc

categories.

The coding of paralinguistic cues

In the ﬁelds of speech communication and nonverbal communication, much attention has
been given to paralinguistic features. However, as in many other ﬁelds, there does not
seem to be a unanimous agreement on the deﬁnition and boundary of the term. Trager
(1958, cited in Jenkins & Parra, 2003), for example, uses the term “paralanguage” to refer
to features such as voice set (physiological and physical aspects), voice quality (pitch
range, rhythm, articulation, resonance, and speed), and vocalization (nonlexical sounds).
Martin (1981), on the other hand, classiﬁes many of these features into the prosodic
domain (e.g., stress and intonational pitch direction) and paralinguistic domain (e. g.,
loudness, pitch, tempo, voice quality, and voice characterizers). Due to the inconsistency
of deﬁnition, researchers have used the word “paralinguistic” in different ways.
Reviewing previous prosody and discourse studies, Shriberg et a1. (1998) found that the
acoustic-prosodic features covered in these studies include pitch range, intonational

contour, declination patterns, utterance duration, preboundary lengthening phenomena,

71

pause patterns, speaking rate, and energy patterns. Given the speciﬁcs of the current
study, it was decided that instead of using the coding scheme of a particular researcher, a
few features derived from Shriberg et al.’s list would be examined, namely: word stress,
rising intonation, eliciting stop, dragging voice, and the combination of the other
categories. The combination category was not excluded because the combined effect of

various categories might be different from that of a particular category by itself.

Word stress
With this type of paralinguistic cues, the teacher stresses a particular letter, word, or
phrase, which is often crucial to learners’ understanding of a linguistic feature. For
example:
(34) Nass: Could you tell me test score please?

T: MY test score. (Stresses “my”) Could you tell me my test score please?

Nass: My test score.
In this example, the learner leaves out the pronoun “my” in his sentence. The teacher

recasts him and stresses the word “my” in her feedback to the learner. The learner notices

that and adds “my” to his new utterance.

Rising intonation
Rising intonation is mostly used when the teacher is indicating that the learner’s utterance
is erroneous and wants to elicit the correct form from the learner or to draw the learner’s
attention to the correct form. For example:
(35) T: What kind of activities are these?

Hasa: Entertainment.

T: They’re entertainment. Anything else?
Hasa: To waste their time.

72

T: To waste their time? (Rising intonation)
Class including Hasa: To spend their time.

In this example, the teacher is leading a discussion after the class watched a sitcom
episode where a family is entertaining themselves. One learner chooses the wrong word
for spending time. The teacher repeats the inappropriate sentence in a rising intonation.
Upon that, the learner realizes his mistake and corrects himself together with other

students.

Eliciting stop
In episodes with eliciting stop, the teacher ﬁrst gives the learner speciﬁc hint(s) and then
stops on pm'pose for the learner to provide the complete answer. For example:
(36) T: Now how do you hit somebody with the back of your hands?
Chub and Pyo: (Silence, 3 seconds)
T: S- (Hand gestures slapping someone with back of hand and stops on purpose)
Chub: Whoa, slap?
T: Ok, yeah. (Nods)
Pyo: Ah!
In this example, the learners have difﬁculty ﬁnding the right word to describe the action
of hitting somebody with the back of one’s hand. The teacher tries to prompt them to give
the word “slap” by giving the ﬁrst letter of the word. With that hint, one learner ﬁgures

out the right answer and the other acknowledges it.

Dragging voice
Dragging voice is generally used in two cases. On some occasions, the learner produces a
problematic utterance and the teacher makes a short utterance such as “well”, “um”, “eh”,

and “bababa” with a lengthened ending voice to show his/her disagreement but without

73

 

providing any speciﬁc hint. On other occasions, the teacher pauses on purpose after
saying part of a word or phrase before ﬁnishing the whole utterance. The purpose is
presumably to draw the leamer’s attention to what is coming up next. Below are two
examples illustrating how the dragging voice technique is used under the two different
circumstances.
(37) Lai: I saw Monique at the club last night.

Lin: What did she do there?

T: Mm-- (Drags voice)

Lin: What was she doing there?

T: Yeah, what was she doing there?
(38) T: Climate change. Ok, what kind of climate change?

Ru: Drough. (/draut/)

T: Dr-- (drags voice) «ought. Say it. Drought.

Class including Ru: Drought.
In the ﬁrst example, the learner uses the wrong tense for the verb “do”. The teacher
simply says mm- in a dragging voice without any speciﬁc hint. Upon that the learner
realizes his error and gives the correct form in no time. In the second example, the learner
mispronounces the vowel in the word “drought”. When giving her feedback, the teacher

drags his voice after pronouncing the consonant cluster /dr/. With the teacher’s

encouragement, the learner repeats it correctly in the choral response.

Mimicking sound
With this type of paralinguistic cues, the teacher mimicks the sound of a thing or the
noise from an action. For example:

(39) T: Siren.
Hasa: No, no. I don’t know.
T: Police. (Irnitates siren)
Class: Oh.

Chen: Fire alarm.

74

 

T: Like the ﬁre alarm, police siren.
Sa: Don’t do that.

T: Don’t do that?

Sa: Don’t say that.

T: Don’t say that? Why?

Yiw: We have a ﬁre alarm last night.

(40) Tal: What is splash?
T: Splash? Psh- (Makes splashing noise and hand gestures water splashing) Water
splashes.
Tal: Oh. (Writes on paper)

In the ﬁrst example, the teacher imitates the sound of a siren, a thing. In the second

example, the teacher imitates the sound of water splashing, an action. In both episodes, a

real sound is mimicked.

Combination
On many occasions, the teacher uses more than one paralinguistic cue in the same
episode, especially rising intonation and word stress. For example:
(41) Joy: They’re talking about how many camels does...

T: How many camels DOES? (Stresses “does” and rising intonation)

Joy: Eh do eh [does]

T: [Does]?

Joy: Does, does, does he has.

T: He HAS? (Stresses “has” and rising intonation)

Joy: He have. How many does he have.
In this episode, there are two combinations of rising intonation and word stress. The ﬁrst
one occurs when the learner puts a plural noun and a singular helping verb together
(which turns out to be correct when he continues the sentence). The second one occurs

when the learner uses the helping verb “does” and yet still uses the third person “has” for

the same subject. In both cases, the teacher uses two types of paralinguistic cues.

75

 

The coding of extralinguistic cues

The coding of extralinguistic cues is a stepwise procedure. After all episodes with any
extralinguistic cues were sorted out, they were further categorized into ﬁve types: hand
gestures, whole body acting, head movements, facial expressions, and combination. Hand
gestures were then further coded as iconics, metaphorics, deictics, beats, or combination.
Combination again was included because the use of different types of extralinguistic cues
or gestures might have a different effect on learners’ noticing and learning from a

particular type of extralinguistic cues or gestures.

Discussing his concept of implicit communication, Mehrebian (1972) gives a list of
kinesic behaviors: facial expressions, gestures, posture, head movements, gaze, and
position. This list provided a basis for the coding scheme for the type of extralinguistic
cues in the current study although the terminology is slightly different and not every

category in the list was examined.

Gestures

Gestures are the movements of the ﬁngers, the arms, and the hands, “the spontaneous,
unwitting, and regular accompaniments of speech” (McNeil, 1992, p. 3). This is one of
the most frequently used extralinguistic cues in the classrooms observed. For example:

(42) Smal: What’s a pad?
T: What’s that?
Smal: The yellow pad. What’s that? P-A-D.
T: B-A-D?
Class: P-A-D.
T: P-A-D. Oh pad. So it’s a small writing book. Small writing book.
Noa: Like notes.
T: Yeah. Notepad and notebook. It’s the same thing. A pad just means like a

76

rectangular card thing. (Hand gestures “rectangular”)
Smal: (N 0 response)

In this episode, the learner asks about the meaning of the word “pad”. In addition to
verbal explanation, the teacher also uses both hands to form a rectangular, which is the

normal shape of a pad.

Whole body acting

With this type of extralinguistic cues, the teacher acts out what is being discussed. The
body parts that are used are not just the hand or arm. It often involves the whole body,
with the term in its broad sense (i.e., not exactly every part of the body). For example:

(43) T: Perk up.
Class: (Silence, 7 seconds)
T: Perk up. (Pauses, 2 seconds) Ok, students, if you, if you’re like this. . .(Acts
being listless) But if you perk up. . .(Acts perking up)
Hasa: Wake up?
T: (Acts perking up again) You see? The differences the way you are? In one, if
you’re not perky, you’re like. . .(Acts being listless again) But when you perk
up. . .(Acts perking up a third time) You see how big your eyes get?
Hasa: (Laughs)
T: Did I perk you up? You know, everybody looks like this. . .(Acts being listless
a third time) And then everybody looks at me. . .(Acts perking up a fourth time)
And-
Mar: Perk up.
T: You perk up.

In this example, the teacher is explaining the meaning of the phrase “perk up”. While
doing so, she acts out how it is like when one perks up by showing the change of the head,
the eyes, the arms, and the torso. It was not coded as hand gesture but whole body acting

because the movement of the arms is only part ofthe acting.

77

Head movements

In one of his books about gestures, McNeil (2005) points out that some head movements
can be considered as the extension of gestures. For example, when the anatomical hand is
immobilized or engaged, one might use the head as a third hand (e.g., pointing in a
direction). In the current study, no such head movements were found. All the head
movements are either the emblematic nodding or shaking. Below are two examples
where the teachers nod or shake their head.

(44) Anh: I’m going to the bank. Can I say I will go to the bank?
T: I will go is ok. (Nods) The reason they choose I’m going to is because usually

for plans we use going to. But it’s ok, the grammar is ok if you say I will. (Nods)
Anh: (Nods)

(45) T: (Discussing paragraphs in book) However, the example provides hope that
careﬁrl planning and creative thinking can lead to solutions to many of them. Ok,
quick quiz, what is them?

Adu: Cities?

T: Cities? No. (Shakes head)
Adu: Eh people.

T: No, not people. (Shakes head)
Lyb: Example.

T: No.

Lyb: No? (Looks at book again)
Roma: Problem.

T: Problem. Right? Solutions to many of them, solutions to many of the problems.
Adu: Oh.

Lyb: Problem.

Others: (Write in book)

In the ﬁrst episode, the teacher nods twice with additional verbal information to conﬁrm
the learner’s language hypothesis. In the second example, the teacher shakes head twice

when learners are trying to guess the antecedent of a pronoun.

78

Facial expressions

Sometimes teachers use facial expressions to conﬁrm or disconﬁrm learners’ utterances

or opinions or to show that they have problems understanding what learners are saying

because of a problematic utterance. For example:

(46) T: Now, give me some adjectives for Hasa (a student in the class).

(47)

Chen: Handsome.

Mar: Gorgeous.

Gra: Fun.

T: Fun, thank you, that’s a good one.

Zhou: Stupid.

T: Zhou, how many times have I told you? (Shows helplessness on face)
Class: (Laughs)

Zhou: Oh, silly.

T: Yeah, silly is better.

Cadiz Is there agency adoption? (“Agency” pronounced as /ei'd3 ansi/)
T: Uh? (Looks puzzled)

Cadiz Agency. (/ei'd3ansi/)

T: Uh? (Looks puzzled)

Cadiz Agency. (/ei'd3ansi/)

T: A-? (Looks puzzled and drags A sound)
Cadiz Agency. (/ei'd3ansi/)

T: (Silent and still looks puzzled)

Cadiz A- (Tries to spell the word)

Lai: Agency. (/'eid3ansi/)

T: Ok, the adoption agency.

Cadiz Yeah.

In the ﬁrst example, the learner gives a pragrnatically inappropriate word to describe a

classmate. While giving verbal feedback, the teacher also shows helplessness on her face,

which in a sense strengthens her disagreement and helps to prompt the learner to give the

right word. In the second example, the teacher is not able to understand what the learner

is saying because of a pronunciation problem. Four out of ﬁve teacher turns, he shows a

puzzled look on his face, which, together with his short utterances, pushes the learner to

repeat his utterance four times.

79

 

Combination
Like paralinguistic cues, teachers sometimes use more than one type of extralinguistic
cues in the same episode. In this case, the multiple types of extralinguistic cues were
coded as combination. For exarnplez
(48) T: Do you know the meaning of desire?
Sar and others: No.
T: Desire means want. You want something.
Sarz (Indistinguishable)
T: Hmm?
Sar: (Indistinguishable)
T: Reversible?
Sarz Vegetable.
T: Vegetable?
Sar: Yeah.
T: No. (Shakes head) No, desire is something you want. (Points at heart) So I
want, I want more ﬁ'uit. That’s the meaning.
Class including Sar: (N 0 response)
In this example, the teacher uses two types of extralinguistic cues, head movement

(shaking head) and gesture (pointing) when explaining the meaning of the word “desire”.

The type of extralinguistic cues in this episode is therefore combination.

The coding of gestures

Various coding systems for gestures have been proposed over the years. One of the most
frequently adopted is the one established by McNeil (1992). In this scheme, there are four
major categories: iconics, metaphorics, deictics, and beats. The current study also
followed this coding tradition. In addition to the four major categories, the use of multiple
types of gesture in the same episode was coded as a ﬁfth category, combination, again for
its potential to exert an inﬂuence on learners’ noticing and learning different from a

particular type of gesture by itself.

80

Iconics

Iconic gestures present images of concrete entities and/or actions. The form of the gesture

and/or its manner of execution embodies the picturable aspects of semantic contents.

Below is an example where the teacher uses iconic gestures to explain a word.

(49) Rusz What, what’s a dumpster?
T: A dumpster is like a big, big, big box made of metal. Like this big. (Opens
arms to show big) You dump trashes. (Hand gestures dumping) And a big
machine comes and does this (hand gestures picking up things), wu-, wu-. Ok,
and people were arrested... (Leaves no time for the learner to respond)

In this episode, the teacher uses three gestures as he explains the word “dumpster”, a

tangible object with a shape, color, and size in real life. Each gesture in turn represents a

concrete thing or action, with the ﬁrst one describing the size of a dumpster, the second

one picturing the dumping action, and the third one depicting the picking up action.

Metaphorics
Metaphoric gestures present images of abstract ideas or concepts. One example McNeil
(1992) gives is that the speaker appears to be holding an object as if presenting it but the
meaning is not presenting an object but rather holding an idea or memory or some other
abstract ‘object’. Below is an example from the current study where a metaphoric gesture
is used.
(50) Sal: (Looks puzzled) Might, may, more problem, more possibility...
T: May is a little more possibility. A little. (Hand gestures “a little” by putting two
ﬁngers close) ‘
Sal: A little? What’s more?

T: More, may.
Sal: May, more possibility.

81

In this episode, the teacher puts two ﬁngers together as if two things were close to each
other side by side. What she is really trying to convey is not the short distance between
two objects but the idea of “a little”. The concept expressed in the gesture is therefore

abstract rather than concrete.

Deictics
Deictic gestures are pointing with the ﬁnger or other body parts. These gestures can refer
to both the concrete and the abstract. In the current study, most deictic gestures fall into
the former type. For example:
(51) Bia: Tell me the homework.
T: Just tell me the homework? Not if you want to be polite. Like I can say to you
tell me the home work. (Points at self and then students)You can’t say to me tell
me the homework.
Bia: (Writes on book)
In this episode, the teacher points at herself and then her students as she explains the

pragmatic aspects of the imperative. Both references, the teacher and the students, are

concrete entities.

Beats

Beats are mere ﬂicks of the hand(s) up and down or back and forth that seem to ‘beat’
time along with the rhythm of speech. Beat may look simple but the meaning behind it
can be complex. It often signals the temporal locus in speech of something the speaker
feels is important with respect to the larger discourse (McNeil, 1992). Below is an
example where beats are used.

(52) Cadiz While he was studying at the ELC, he was playing basketball.

82

T: While he WAS studying, but he still IS. That’s when we can’t use the past
because he IS still studying and he still IS playing basketball. That’s present.
(Stresses “was” and “is” and moves right hand up and down)
Cadiz (Nods)
In this example, the teacher is explaining why the learner should use the present tense
“is” instead of the past tense “was” in his sentence. Every time he utters these words with

a stress, he also moves his right hand up and down. The “verbal stress” from his mouth

and the “nonverbal stress” of his hand work together to make these words stand out.

Combination
Sometimes teachers use more than one of the four types of gestures described above. In
this case, the different types of gestures were coded together as combination. For
example:
(53) Lyb: Excuse me, can you explain the second one? (Points at book)
T: (Looks at book) Ok, please pick it up and put it on. So it should go between
pick and up (points at book) because that verb is a phrase. It’s more than a word.
So for some reason, I’m not sure why we do it this way, but whenever we have a
phrasal verb and we’re changing that noun to a pronoun we put the pronoun
between the two parts of the phrasal verb. (Hand gestures “between” by raising
and opening two ﬁngers)
Lyb: (Writes in book)
In this episode, the teacher is explaining the position of the pronoun “it” in phrasal verbs.
As she talks about it, she ﬁrst points at the leamer’s book to show where it should go, and
then opens two ﬁngers to show its position. The ﬁrst gesture is deictic and the second one

is iconic. Now that two types of gestures are used, this episode was coded as having a

combination of gestures.

83

3.6.1.3 The coding of teachers’ metalanguage

J arnes (1999) indirectly deﬁnes the concept of metalanguage when explaining
metalanguage awareness. According to J arnes, “meta” refers to the ability to use language
to talk about language in the linguistic sense and cognition about cognition in the
psychological sense. That is, linguistic metalanguage awareness means knowing what
something is called, for example, part of speech and type of sociolects, while
psychological metalanguage awareness means cognition about one’s own knowledge,
whether one can talk about it or not. In the current study, only metalanguage in the

linguistic sense was examined.

Each episode was classiﬁed as being with or without metalanguage in the feedback. The
metalanguage was then coded with a scheme adapted from Basturkrnen et a1. (2002).
Two types of metalanguage were identiﬁed: technical terms and non-technical terms.
Technical terms are those likely to be found in grammar books or linguistic references
and often used by a small population such as teachers and linguists; non-technical terms
are those that are commonly used by a larger population, including students (Basturkrnen
et al., 2002). Some of the examples for technical terms are “preposition”, “irregular verb”,
and “yes/no question”. Some of the examples for non-technical terms are “word”, “mean”,
and “say”. Like Basturkrnen et al., if the same term occurred multiple times in the
feedback of the same episode, it was counted only once. However, when different terms
were used in the feedback of the same episode, the occurrence of each term was counted

once. The following example illustrates how the metalanguage in teacher feedback was

coded:

84

(54) Sal: What’s the difference between he and to be?

T: To be is the inﬁnitive. Be is the base form, base form be.

Sal: (Writes on paper)
In the teacher’s feedback, there are two different technical terms which can be found in
linguistic references: “inﬁnitive” and “base form”. Although they occur in the same
teacher turn, they were both counted. On the other hand, the term “base form” was only

counted once even though it occurs twice in the same teacher turn. Ultimately, there are

two occurrences of teacher metalanguage in this episode.

After all metalanguistic terms were labeled, the feedback episodes were further analyzed
into three categories according to the properties of the terms: those with only technical
terms, those with only non-technical terms, and those with both technical and non-
technical terms. For exarnplez
(55) Ye: Iwere-

Tz I WAS, were is plural. (Stresses “was”)

Ye: Yeah, I was doing my homework.
(56) Lo: White and black. Write them white and black.

T: We say black and white.

Lo: Oh ok. Black and white.

(5 7) Wa: Riot, I don’t know how to pronounce riot. Is it riot? (/ra:t/)
T: What’s the word? Ri--ot, ri--ot. Two syllables. (Drags voice between /rai/ and

/at/)
Wa: Riot.

T: But people say it quickly, riot.

Wa: Riot, riot.
In the ﬁrst episode, the teacher uses the term “plural”, a technical term which can be
found in most grammar glossaries. In the second episode, the teacher uses the wor “say”,

a common word which occurs very often in daily conversations. In the third episode, the

teacher uses two terms: “word” and “syllable”, with “word” as a non-technical term and

85

“syllable” as a technical term. Teachers’ metalanguage in the three episodes was
therefore categorized as technical term only, non-technical term only and technical plus

non-technical term.

As Basturkrnen et a1. (2002) note, “it is realised that the distinction between technical and
non-technical terms is not always straightforward and some metalinguistic terms fall
somewhere between the two ends of the continuum” (p5). To avoid over-subjectivity, I
consulted three grammar books: Portable English Handbook (Herman, 1982),
Introducing English Grammar (Leech, 1992), and A Glossary of English Grammar
(Leech, 2006). Now that the metalanguage examined in the current study is not just
grammar terms, I also consulted a native speaker and an ESL speaker about non-grarnmar
terms that fall between the technical and non-technical continuum. Both parties provided
their insights on the property of these terms according to their “feel” and daily language
experience. It was hoped that the knowledge and perspectives of “outsiders” would help
objectify my decision-making. When a term was still difﬁcult to decide after this, I would

discuss it with another SLA researcher.

3.6.1.4 The coding of the occurrence and successfulness of uptake

In coding learner uptake, a scheme was again adapted from Loewen (2004) as shown in
Table 3. In the ﬁrst step, depending on whether learners responded to teacher feedback
and whether they had the chance to do so, three categories were identiﬁed: uptake, no
uptake, and no opportunity. Any response to teachers’ feedback, even ambiguous

expressions such as “Oh” or “Yeah”, was coded as uptake. Otherwise they were coded as

86

no uptake or no opportunity. The difference between no uptake and no opportunity is that
in no uptake learners have the opportunity to respond but choose not to do so, while in no
opportunity learners have no chance to respond because of factors such as topic
continuation. It is important to note that now that a video camera was used in the current
study, the recordings show that many times learners responded to teachers’ feedback by
acting, nodding, or writing with or without linguistic utterances. Such reactions were also
learners’ responses to teachers’ feedback. Acting, nodding and writing, even when there

was no verbal utterance accompanying them, were therefore counted as uptake.

Table 3 Coding scheme for learner uptake

 

Deﬁnition Categories

 

Uptake: student responds to feedback

No uptake: student has the chance to respond to
feedback but does not respond

No opportunity: student does not have the chance to
respond to feedback

Occurrence of Whether student
uptake responds to feedback

Successful uptake: student incorporates correct form
in production

Partially successful uptake: Student incorporates
some of the correct form in production but not all of
it so that production is still not completely correct.
Unsuccessful uptake: student tries but fails to
incorporate correct form in production
Acknowledge: Student simply acknowledges
feedback with utterances such as oh and year or by
nodding.

Inconclusive: unclear whether student successfully
incorporates the correct form in production

How the quality of
student response to
feedback is

Successfulness of
uptake

 

Adapted from Loewen (2004, p.166)

The second step is the coding of the successfulness of uptake. When learners
incorporated the correct form in their production, the response was coded as successful
uptake. When learners tried to incorporate the correct form but only did part of it, the

response was coded as partially successﬁrl uptake. When learners tried but totally failed

87

l

to incorporate the correct form, it was coded as unsuccessful uptake. When learners
reacted to teachers’ feedback by simply saying “Yeah”, “Oh”, etc, the response was
coded as acknowledge. Acknowledge was not operationalized as unsuccessful uptake
because although it is possible that learners did not understand the feedback but simply
said something to indicate that they were listening, it is also possible that they might have
indeed understood the correct form but decided not to take the trouble to incorporate it in
a new language production. A ﬁfth category was inconclusive. In inconclusive uptake, it
is difﬁcult to determine if the correct form was produced or understood due to the quality
of the recordings, choral responses, etc. For exarnplez

(58) T: Eh unadom.
Gra: Yes
Hasa: No.
T: What is that? Gra, you said yes, what’s unadom?
Gra: Eh... (Chuckles)
T: Ok. Students, you know, you know, what are these? (Touches earrings) Mik?
Hasa: Earrings.
T: Mik?
Mik: I don’t know.
T: That’s why I’m trying to help you understand. What’re these?
Class: Earrings.
T: Why do you believe I put the earrings on? Why?
Hasa: To look, to look beautiful.
T: To adorn. Ok?
Class: 0h.
T: To make me look nice. Now if I take them off, then I would be doing what?
Class: (Indistinguishable)
T: Then I would be what?
Mar: Gorgeous. (Laughs)
T: Unadom.
Hasa: No, no.
T: When you adorn, it’s that you put things on to make your, to make yourself
look nice. Ok? But when you take them off, then it’s [unadom]. (Shows putting
on and taking off earrings)
Class: [Unadorn].

88

In this example, there are a few points where more than one learner responds to the
teacher at the same time and it is impossible to distinguish who are speaking. Hasa
indicates that he does not understand the word “unadom”. Even though he participates in
the discussion all the way toward the end of the episode, it is unclear from neither the
audio tape nor the video tape if he is one of those students who ﬁnally say the word

together with the teacher.

The examples in Table 4 illustrate how the occurrence and the successfulness of uptake
were coded. In Example 59, the learner has the opportunity to respond to the teacher’s
feedback but he chooses to talk about something else. In Example 60, the learner does not
get the chance to respond to the teacher’s recast as a result of the teacher’s topic
continuation. The uptake move for these two episodes was therefore coded as no uptake
and no opportunity respectively. In the next four examples, learners do respond to
teachers’ feedback, but in different ways. In Example 61, the learner successfully
incorporates the correct form in his production. In Example 62, the learner is able to give
a partially correct answer at the teacher’s elicitation but still leaves out the object of the
word “move” after the teacher gives him the correct form. His ﬁnal production therefore
is still grammatically ﬂawed. In Example 63, the learner fails to produce the correct form
after a few tries. In Example 64, the learner responds to the teacher’s feedback with a
simple “ah” and it is not clear if he really understands it. Based on the different kinds of
response moves by learners, the successfulness ofzuptake in these episodes was coded as

successful, partially successful, unsuccessful, and acknowledge respectively.

89

Table 4 Examples for the coding of the occurrence and successfulness of uptake

 

Episodes

Occurrence of
uptake

Successfulness of
uptake

 

(59) T: Think about your day. When do you have ten or
ﬁfteen minutes when you’re not necessarily doing
something that you can work on vocabulary?

Saudz In the bus.
T: On the bus.
Saudz Waiting before the class.

(60) Rusz Grinding to a halt. (/’grindilj/)
T: Ok, grinding to a halt. Grinding. How many
people here know what grinding means?

(61) Squz One people speak one people write?
T: One PERSON can speak. (Stresses “person”)
Squz Oh one person. I’m sony.

(62) Matoz The answer is let’s go? (Could you please
help me move the sofa?)
T: Let’s go is like let’s go away outside. (Hand
gestures going away) Let’s- (Stops)
Matoz Move.
T: Move it. Yeah.
Matoz Let’s move.

(63) Ne: I would some coffee.
T: I would some coffee? (Rising intonation) You
need a verb. I would like.
Ne: I would s-, I would some...
T: (Shakes head) When you use, when you use
would, would is like a helping verb, so you have to
have a main verb.
Ne: I would some coffee?
T: (Shakes head and shakes hand) Would is not by
itself. It’s not alone.
Ne: (Nods) But I. . .some kind of...

(64) T: So what kind of strategies have we used so far
today?
Storz Strategy?
T: Strategies means a way of doing something, a

plan for how to do it.
Storz Ah. (Writes in notebook)

No uptake

No opportunity

Uptake

Uptake

Uptake

Uptake

Successful

Partially successful

Unsuccessful

Acknowledge

 

90

In the coding of the successfulness of uptake, one thing that deserves a special note is
again students’ nonverbal responses to teacher feedback, including acting, nodding and
writing. On some occasions, learners’ nonverbal responses clearly indicate that they have
understood the feedback. In this case, the uptake was coded as successful. For example,

(65) T: How about beady? Beady?
Class: (Silence, 4 seconds)
T: Beady? Ok guys, watch, Tona, watch. (Acts) Look at my eyes. Now if I put
them. . .(Acts) So when I, when I put them like this and I go. . .(Acts)
Class: (Laugh)
Sa: (Irnitates)
T: Sa! But that’s beady. See the eyes? They get like a little like this like, but yeah
the eyes they’re going like this like looking all over the place. (Acts) Beady, ok?
Beady, beady.
Class: Beady, beady eyes.

In this episode, even though it is not clear if Sa is among the students who repeat the
word “beady” in the end, her imitation of beady eyes suggests that she understands what
the teacher is saying. The uptake move for Sa in this episode was therefore coded as

successful instead of inconclusive.

On other occasions, learners responded to teacher feedback by nodding. This might
happen with or without verbal utterances. When there was a verbal utterance, the
successfulness of the uptake was decided by the verbal utterance. When there was no
verbal utterance, the nodding was coded as acknowledge. For example:

(66) Xin: Pot of money means a big amount of money?
T: Sorry?
Xin: Pot of money.
T: Pot of money? Where do you see that?
Xin: (Shows T word in book) Pot of money.
Ira: Pot of money.
Xin: Refers to big sum of money?
T: No, not necessarily a lot. I think she actually means a physical pot. (Hand
gestures “pot”) She has some sort of bowl or pot or jar that she puts money into.

91

\l

But it doesn’t mean something necessarily physical. It could also mean just, just a
collection of money. (Hand gestures “collection”) When you play poker, the
money you put in the middle when you bet, that’s the pot. The winner takes the
money. It’s not really a pot. (Hand gestures “pot”) It’s just a pile of that money.
(Hand gestures “pile”)
Xin: (Nods)
T: Everyone just puts money into the pot. We call that a pot. Yeah.
Xin: (Nods)
(67) Bibo: ATM card, you can eh take your money in ATM machine?
T: Take it FROM the ATM machine. (Stresses “from”)
Bibo: Yeah yeah yeah. (N ods)
(68) Lyb: 'Yes, the farm now belongs the animal.
T: Belongs TO the animal, right? (Stresses “to”) Belongs to the animal.
Lyb: Belongs to the animal. (Nods)
In Example 66, Xin responds to the teacher’s explanation of the phrase “pot of money”
with nodding without saying a word. The uptake in this episode was therefore coded as
acknowledge. In the other two examples, nodding is both accompanied by an utterance.
In Example 67, the utterance is the typical acknowledgement token “yeah” while in
Example 68, the student correctly incorporates the teacher’s feedback in her new

utterance. The uptake in these two episodes was therefore coded as acknowledge and

successful uptake respectively.

The coding of students’ writing move with verbal utterances is similar to that of nodding
with verbal utterances, but writing without verbal utterances is different from nodding
without verbal utterances. When students start to write after a feedback move, it is not
clear whether they are writing what the teacher is saying or something that is not related
to the particular episode. Writing alone was therefore coded as inconclusive uptake. For
example:

(69) Rusz What do you call this? (Points at picture)

92

T: Ok, I’ll say it exactly. It’s called a riot baton or riot stick. Riot stick or riot
baton or a night stick. They hit people with a night stick. (Hand gestures hitting)
Rusz (Similes and writes on paper)
In this episode, the student writes on a piece of paper after the teacher’s feedback. Even
though it is quite possible that she is writing down riot baton or night stick, there is no

observable evidence that this is indeed what she is writing. The uptake in this episode

was therefore coded as inconclusive.

3.6.2 The coding of stimulated recall comments

The analysis process of the comments made by student participants in the stimulated
recall interviews is shown in Figure 4. First, the overall recall comments were classiﬁed
into ﬁve categories: language-related comments, non-language-related comments, no
thought, no memory, and unclassiﬁable comments. Next, language-related comments
were further classiﬁed into target-feature-related, non-target-feature-related, and
inconclusive comments. Finally, target-feature-related comments were put into three
categories: teacher-feedback-related, non-teacher-feedback-related, and inconclusive

comments.

3.6.2.1 The coding of overall recall comments
In coding the overall recall comments, a scheme was adapted from Egi (2004). The
comments were classiﬁed into ﬁve categories: language-related comments, non-

language—related comments, no thought, no memory, and unclassiﬁable comments.

93

Figure 4 Coding process of stimulated recall comments

/7

 

 

Recall
comments

 

 

 

Language-
related
comments

 

Non-
language-
related
comments

 

 

 

 

I

 

 

 

No
thought

 

 

 

No
memory

 

 

Unclassi-
ﬁable
comments

 

/\

 

 

 

Target-feature-

related
comments

 

 

Non-target-
feature-related
comments

 

Inconclusive

 

 

 

[\i\

 

 

 

 

Teacher- Non-teacher-
feedback-related feedback-related
comments comments

 

 

 

Inconclusive

 

 

 

Language-related comments

This category was operationalized as any comment that involves linguistic items or
language structures with or without participants’ reference to their own error production

or the teacher’s feedback. The comments range from simple observational remarks to

those involving linguistic analyses. For example:

(70)

(71)

I forgot what’s that. The teacher remembered to us imperative.

I just concentrate on the two words and I found the differences that when we say
pay BACK, it’s not a compound. There’s a blank between it. So I just considered
it as a verb and preposition. When we don’t link those two together, it’s probably
the verb. But payback is actually one word. So despite of the verb, it should be the

noun. When it’s a compound, it’s a noun.

94

 

 

In these two examples, the ﬁrst comment is a simple observation of what the participant
heard. In the second comment, the learner made a detailed analysis on the spelling,

pronunciation, structure, and part of speech of “pay bac ” and “payback”.

N on-language-related comments

In this category, participants talked about topics that are not related to language issues,
for example, the task at hand, other learners’ behavior in class, teachers’ teaching method,
etc. For example:

(72) I was thinking I disagree with the article in the book.

(73) I think Kuwa makes a lot of questions. I said, ok Kuwa, you make a lot of
questions. Some are boring. Some were not that helpful for me.

(74) She’s only speaking to one, two, three, four, four people. She must speak to all
these people. But maybe other people understand. Maybe she asked other people
already. I feel bad if the teacher didn’t explain to everybody. I took a course in
education in my country. I know some education. Yeah, that’s I was thinking.

In these three examples, the ﬁrst one is an opinion statement about a question the class

was discussing; the second one is a remark about another learner’s action in class; and the

third one is a comment on the way the teacher was talking to students. All three

comments are somewhat related to the lesson but none is related to language issues.

No memory
This category refers to participants’ report that they do not remember their thoughts
regarding a particular feedback episode. For example, one learner said:

(75) I don’t remember. Sorry.

95

No thought
This category refers to participants’ report that they had no speciﬁc thoughts during a
particular feedback episode. For example, a learner said:

(76) I don’t think nothing because I was bored. That’s enough.

Unclassifiable comments
This category was operationalized as any comment that cannot be classiﬁed into any of
the categories above. For example:

(77) Any thought? It’s very interesting. I didn’t so relax. . .(inaudible) because me and
Rin, especially the student sitting at the lastest place in the class.

In this example, it is hard to understand what the learner was saying and whether he was
talking about the lesson or something else. This comment was therefore coded as

unclassiﬁable.

3.6.2.2 The coding of language-related comments
After language—related comments were identiﬁed, they were further categorized into

target feature-related, non-target-feature-related, and inconclusive comments.

Target-feature-related comments

This category was operationalized as any comment that is related to the language feature
which had been targeted at by the teacher in a feedback episode, with or without the
learner’ reference to the feedback. For example: . '

(78) Yes, I can remember. She is explain two words, what’s the difference between the
words. The ﬁrst word is “give” and the second word is “taking”. She said many

96

students is confused, so she explain with example. I think this is a very good
example, very simple example.

In this comment, the learner was talking about the difference between “imply” and
“infer”, two words the teacher had explained in detail in a feedback episode, and how the

teacher explained it. It was therefore coded as target-feature-related.

N on-target-feature-related comments

This category was operationalized as comments that are related to a language feature
which was not the target of the teacher’s feedback or comments where the learner was
talking about the same language item as the teacher but with a different linguistic focus.
For example:

(79) About the word. I didn’t know this word at all but I wanted to get it from word
root, preﬁx or sufﬁx. With S-O-L, I thought of “solid”. Then I thought of “solo”.
“Solo” means alone. So I thought this word might be related to that. Then “tude”,
I don’t know why I associated with “attitude”. So I thought “solitude” might mean
an attitude or idea.

In the original feedback episode, the teacher was giving feedback about the spelling of

the word “solitude”, but in the comment the learner was talking about the meaning of the

word. Since the learner’s focus is different from that of the teacher, it was coded as non-

target-feature—related.

Inconclusive

This category was operationalized as comments that do not fall into either of the two
categories above. For example: i i
(80) He said something from. I told him this is not polite. I don’t know this meaning.

Don’t say your name. (Draws on paper) Don’t say from, just put something. He
told me to use from only when I write ﬁiend or family.

97

This comment comprises six separate sentences which do not seem to be closely
connected. It can be seen that the student was talking about language use, but it is very
difﬁcult to tell whether he was talking about the word “sincerely”, which was the target

of the original feedback episode.

3.6.2.3 The coding of target-feature—related comments
Finally, target-feature-related comments were put into three categories: teacher-feedback-

related, non-teacher—feedback-related, and inconclusive.

Teacher-feedback—related comments

This category was operationalized as any comment where students referred to teacher

feedback or things that they thought of as the result of the feedback. For example:

(81) I didn’t know about that word. Me and my partner asked this question about this
word. And then she explained more speciﬁcally, and at that time she gave us
example, so I could understand more easily.

(82) When I heard of the word (classic), I thought of classic music. And then he said
classic coke. That reminded me of the NCA I’d been watching. Coca cola did a
lot of commercials there. So I was thinking of the coca cola commercials.

In the ﬁrst comment, the learner explicitly referred to the teacher’s use of examples and

how that helped with his understanding of a word. In the second comment, although the

learner got “carried away” to talk about the coca cola commercials instead of the meaning
of the word “classic”, which the teacher was explaining in the original episode, the

thought of the coca cola commercials is a result of hearing the teacher’s use of the classic

coke example. It was therefore still coded as teacher-feedback-related.

98

Non-teacher—feedback-related comments

In contrast to teacher-feedback-related comments, non-teacher-feedback-related

comments are those without learners’ reference to teacher feedback or things they

thought of as the result of the feedback. For example:

(83) Yesterday I was thinking, at that time, I was thinking about the spelling of this
word because we was, we were, me and my partner, we were struggling and have
a little problem how to write this word, solitude. That’s it. Yeah, I was thinking
how to write the spelling of the word.

In this example, the learner didn’t say anything about the teacher’s feedback about the

spelling of the word “solitude” but how he himself was trying to ﬁgure it out. It was

therefore coded as non-teacher-feedback—related.

Inconclusive

Similar to comments which are neither target-feature-related nor non-target-feature-
related, this category was operationalized as comments that are neither teacher-feedback-
related nor non-teacher-feedback-related. For example:

(84) It’s the same word in Spanish, motivation. I was thinking, that’s the same word in
Spanish!

In the original episode, the teacher and the class were talking about the meaning of the
word “motivation”. In the comment, the learner was talking about the similarity between
the English word and the Spanish word. It is not clear if she was talking about the
meaning or the pronunciation of the word or both. Nor is it clear if the student’s thought
is a result of her hearing the sound of the word or the teacher’s explanation. It was

therefore classiﬁed as inconclusive.

99

3.6.3 The coding of test results

The results from the immediate and delayed test were analyzed with Loewen and Philp’s

(2006) coding scheme. Learners’ answers to the test items were sorted into three

categories: correct, incorrect, and partially correct. When the answer a learner provided

matched the correct form that had been targeted at in a feedback episode, it was rated as

correct. When a learner failed to provide the correct form that had been targeted at, the

answer was rated as incorrect. When a leamer’s answer showed some improvement

toward the target form but still needed repair, it was rated as partially correct. The

examples in Table 5 show how the test results were coded.

 

 

Mato: Terrible.

be to look good.

100

Table 5 Examples for the coding of test results
Episodes Prompts Responses Ratings
(85) Lyb: My husband studied
in Florida.
T: Oh really? The following sentence is
Lyb: Yeah. Before ten years. inappropriate. Please listen My
T: He studied there ten years carefully and tell me how you husband
AGO. (Stresses “ago”) could make it better. studied in Correct
Lyb: Yes. Florida ten
T: Not before ten years. Ten My husband studied in Florida years ago.
years AGO. (Stresses “ago”) before ten years.
Lyb: Ah ten years ago, yes
yes.
. . . 9 , .
(gagingat 1s thrs. What S thrs Please read the following sentence The word
Rouf: Skull. (/skul/) and “”0“" 3.1m"
strll Incorrect
T: A skull. Ok' DO we Everyone has a skull pronounced
f) 0
have a skull. (lskAU). Skull as /skul/
Class. Yes.
(87) Matoz It is going to not to The following Sims“? rs Th
be to look good. inappropriate. ease rsten enew.
. , . . carefully and tell me how you building rs .
T. It s gorng to look temble. . . Partrally
. . could make it better. gorng to be
Mato. Temble. correct
T' It is going to look terrible to 100k
' ' The new building is going to not to terrible.

In the ﬁrst example, the learner successﬁrlly corrected an ungrarnmatical sentence in the
way the teacher did in the feedback episode. In the second example, the learner
pronounced the word “skull” as /sku1/ again as he did during the classroom interaction. In
the last example, the teacher made three changes to the learner’s original sentence:
deleting the negation word “not”, deleting the “be” verb, and changing the adjective
“good” into “terrible”. In the test, the learner deleted “not” and changed “good” into
“terrible” but kept the “be” verb, thus producing a sentence that is better but still
ungrammatical. The rating for the answers to the three questions is therefore correct,

incorrect, and partially correct respectively.

3.6.4 The reliability of coding

In order to increase the reliability of coding, a second rater coded 11.2% of the feedback
episodes according to the seven characteristics, nonlinguistic cues and metalinguistic
terms in teacher feedback, and the occurrence and successfulness of uptake. 10.8% of the
stimulated recall comments and 11.9% of the test results were also coded by a second
rater. The percentage of agreement for each variable is shown in the Table 6. It ranges

from 78.1% for the source of feedback episodes to 98.3% for test results.

101

Table 6 Reliability of coding

 

 

Variables Percentage of agreement
Type 84.4%
Source 78.1%
Response 81 .9%
Linguistic focus 88.8%
Directness 90.0%
Complexity 90.0%
Emphasis 93.1%
General paralinguistic cues 97.5%
Type of paralinguistic cues 88.0%
General extralinguistic cues 98.1%
Type of extralinguistic cues 93.7%
Type of gestures 83.3%
General metalanguage 86.9%
Type of metalanguage 89.4%
Occurrence of uptake 91.3%
Successfulness of uptake 87.1%
Recall comments 93.3%
Test results 98.3%

 

3.6.5 Statistical analysis

After the coding of all data, Pearson’s chi-square tests were performed with SPSS 12.0 to
see if there is a signiﬁcant relationship between the dependent variables (the occurrence
of uptake, the successfulness of uptake, test results, and stimulated recall comments) and
the independent variables (the characteristics of feedback episodes, the nonlinguistic cues
in teacher feedback, and the metalanguage in teacher feedback). The conventional alpha
value of .05 was selected as‘the cutting point. In order to meet the requirement of chi-
square tests that no more than 20% of expected cell counts should be less than 5, a trial

test was performed for each pair of dependent and independent variables. According to

102

the results of the trial tests, categories with small counts were either excluded or
conﬂated with other categories as necessary and appropriate. Table 7 shows the changes
of the categories within variables. All changes are noted again in or at the bottom of
corresponding tables except three, namely, the combination of partially successful and
unsuccessful uptake, the combination of inconclusive teacher-feedback-related comments
and the category of “other”, and the combination of partially correct and incorrect test
results. These three were highly ﬁequent ones. It would be repetitive if they were stated

again and again.

Table 7 Treatment of variables in chi-square analysis

 

 

Variables Categories Treatment
T Teacher-initiated, student- No chan e
ype initiated, reactive g
Semantic and message combined when
Source Code, message, semantic dependent variable is delayed test results and
recall comments
Response Provide, elicit No change
Linguistic Vocabulary, grammar, Spelling excluded when dependent variable is
focus pronunciation, spelling immediate test results and recall comments
Directness Direct, indirect No change
Complexity Simple, complex No change
Emphasis Light, heavy No change
(General)
paralinguistic Present, absent No change
cues
. . Mimicking sound excluded from all chi-square
Word stress, rrsrng . . .
Type of . . . . . tests, combmatron excluded when dependent
. . . intonatron, elrcrtmg stop, . .
paralinguistic dra in voice mimickin variable rs the occurrence and successfulness of
cues gg g ’ g uptake; rising intonation excluded when

sound, combination

dependent variable is immediate test results

 

103

Table 7 (cont’d)

 

(General)
extralinguistic
cues

Type of
extralinguistic
cues

Gestures

(General)
metalanguage

Type of
metalanguage

Occurrence of
uptake

Successfulness of
uptake

Recall comments

Test results

Present, absent

Gestures, whole body
acting, head movements,
facial expressions,
combination

Iconics, metaphorics,
deictics, beats,
combination

Present, absent

Tech-only, non-tech-only,
tech+non-tech

Uptake, no uptake, no
opportunity

Successful, partially
successful, unsuccessful,
acknowledge,
inconclusive

Teacher-feedback-related,
non-teacher-feedback-
related, inconclusive,
other

Correct, incorrect,
partially correct

No change

Facial expressions excluded when dependent
variable is successfulness of uptake and test
results; Whole body acting excluded when
dependent variable is test results

Combination excluded when dependent variable
is immediate test results; beats and combination
excluded when dependent variable is delayed
test results

No change

No change

No change

Partially successful and unsuccessful combined;
both excluded when dependent variable is test
results and when independent variable is type of
paralinguistic cues and type of extralinguistic
cues

Inconclusive combined with “other” in all chi-
square tests

Partially correct and incorrect combined in all
chi-square tests

 

While a chi—square value can tell whether the relationship between two variables is

statistically signiﬁcant or not, it does not specify how much each category contributes to

the rejection of the null hypothesis. In order to get this information, adjusted standardized

residuals (expressed in the present study as ASR) were computed along with chi-squares. .

The conventional 2.0 in absolute value was selected as the cutting point. Now that

adjusted standardized residuals were used to ﬁnd the major contributors of signiﬁcant

chi-square values, they were not examined when a chi-square value is greater than .05

104

(See Hinkle et al., 2003). To economize language and save space, the term “adjusted
standardized residuals” is presented as “residuals” in tables and table titles. For the same
reason, when a variable (dependent or independent) only has two categories in it and
when the chi-square is 2x2, only the adjusted standardized residuals of one category are
presented and noted now that the residuals of the other category always have the same

absolute values but only with the opposite property.

3.7 Summary

In summary, this chapter addressed methodological issues such as instruments and
procedures, detailed the coding schemes for the observation data, learners’ stimulated
recall comments, and test results, and reported the reliability of coding. In addition, it
brieﬂy explained the statistical analysis of the whole data set. The next three chapters

report and discuss the ﬁndings from the analysis.

105

CHAPTER 4
CHARACTERISTICS OF FEEDBACK EPISODES, THE NOTICING AND THE

EFFECT OF TEACHER FEEDBACK

4.1 Introduction

This chapter addresses the ﬁrst research question: Do the characteristics of teacher
feedback episodes affect learners’ noticing and learning? After presenting the distribution
of feedback episodes according to their general characteristics, I report and discuss the
relationship between these characteristics and learners’ responses to teachers’ feedback,
learners’ comments in the stimulated recall interviews, and results from both the

immediate test and delayed test.

4.2 Distribution of feedback episodes by characteristics

This section reports the distribution of feedback episodes according to their general
characteristics: type, source, response, linguistic focus, directness, complexity, and
emphasis. As shown in Table 8, in terms of type, there is no big difference among the
three kinds of feedback episodes, with reactive episodes (35.1%) slightly more than
student-initiated episodes (31.4%) and teacher-initiated episodes (33.5%). In terms of
source, both code-related episodes (59.8%) and message-related episodes (36.8%)
occurred noticeably more often than semantic episodes (3.4%) and code-related episodes
noticeably more often than message-related episodes. In terms of teachers’ response type,
the percentage of provide is as high as 92.1% while that of elicit is as low as 7.9%, with

an 84.2% difference between the two. In terms of linguistic focus, vocabulary comes ﬁrst

106

Table 8 Distribution of feedback episodes by characteristics (N =l434)

 

 

Characteristics Number Percentage
Type

T-initiated 481 33.5%

S-initiated 450 3 1 .4%

Reactive 503 35.1%
Source

Code 858 59.8%

Message 527 36.8%

Semantic 49 3.4%
Response

Provide 1321 92.1%

Elicit 113 7.9%
Linguistic focus

Vocabulary 712 49.7%

Grammar 532 37.1%

Pronunciation 152 10.6%

Spelling 38 2.6%
Directness

Direct 1107 77.2%

Indirect 327 22.8%
Complexity

Simple 572 39.9%

Complex 862 60.1%
Emphasis

Light 182 12.7%

Heavy 1252 87.3%

 

with a percentage of 49.7%; grammar comes next with a percentage of 37.1%;
pronunciation comes third with a percentage of 10.6%; and spelling comes last with a
percentage as low as 2.6%. In terms of directness, the majority of feedback episodes are

direct (77.2%) rather than indirect (22.8%). In terms of complexity, 39.9% of the

107

episodes are simple and 60.1% are complex. Last, in terms of emphasis, the percentage

for heavy episodes is 87.3% while that of light episodes is only 12.7%.

To sum up, reactive, code-related, direct, complex, and heavy episodes occurred more
often than teacher/student-initiated, message-related/semantic, indirect, simple, and light
episodes in the classrooms observed in this study. In most of the feedback episodes,
teachers provided direct information instead of eliciting answers from learners. The

linguistic focus of teachers’ feedback is mainly vocabulary and grammar.

4.3 Characteristics of feedback episodes and learner uptake

This section reports the relationship between the general characteristics of feedback
episodes and learner uptake. After an overview of the occurrence and successfulness of
uptake, the two variables are presented together according to each characteristic in the

order of type, source, response, linguistic focus, directness, complexity, and emphasis.

Overview of the occurrence and successfulness of uptake

Table 9 shows the occurrence and successfulness of uptake. In 65.2% of a total of 1434
episodes, learners responded to teachers’ feedback in some way. In 21.3% of the episodes,
they did not produce any uptake even though there was a chance for them to do so. In
13.5% of the episodes, they did not have the opportunity to respond. When uptake did
occur, 45.2% was successful; 0.6% was partially successful; 1.2% was unsuccessful,

37.6% fell into the acknowledge category, and 15.3% was inconclusive.

108

Table 9 Overview of the occurrence and successfulness of uptake

 

 

 

Number Percentage
Occurrence of uptake
Uptake 935 65.2%
No uptake 306 21.3%
No opportunity 193 13.5%
Total 1434 100%
Successfulness of uptake
Successful 423 45.2%
Partially successful 6 0.6%
Unsuccessful 1 1 1.2%
Acknowledge 352 37.6%
Inconclusive 143 15.3%
Total 935 100%
Type

Tables 10, 11, 12, and 13 illustrate the occurrence and successfulness of learner uptake
according to the type of feedback episodes. Table 10 shows that learners responded to
teacher feedback in 54.7% of teacher-initiated episodes, 71.6% of student-initiated
episodes, and 69.6% of reactive episodes. Adjusted standardized residuals in Table 11
reveal that uptake was signiﬁcantly more frequent in student-initiated and reactive
episodes while no uptake and no opportunity were signiﬁcantly more frequent in teacher-
initiated episodes. Table 12 shows that when uptake did occur, 45.2% was successful in
teacher-initiated episodes, 30.4% was successful in student-initiated episodes, and 58.9%
was successful in reactive episodes. Adjusted standardized residuals in Table 13 reveal
that reactive episodes led to signiﬁcantly more successful uptake, student-initiated
episodes led to signiﬁcantly more acknowledge, and teacher-initiated episodes led to

signiﬁcantly more inconclusive uptake. Chi-square tests indicate that both the occurrence

109

and successfulness of uptake are signiﬁcantly related to the type of feedback episodes

(p=.000 in both cases).

Table 10 Type and the occurrence of uptake

 

 

 

Number of Occurrence of uptake
episodes Uptake No uptake No opportunity
T-initiated 481 263 (54.7%) 122 (25.4%) 96 (20.0%)
S-initiated 450 322 (71.6%) 97 (21.6%) 31 (6.9%)
Reactive 503 350 (69.6%) 87 (17.3%) 66 (13.1%)

 

x2(4, n=l434)=49.523, p=.000

Table 11 Type and the occurrence of uptake residuals

 

 

Uptake No uptake No opportunity
T-initiated -5.9 2.6 5.1
S-initiated 3.4 .1 -4.9
Reactive 2.6 -2.7 -.3

 

Table 12 Type and the successfulness of uptake

 

 

 

Number of Successfulness of uptake
uptake Successful Unsuccessful Acknowledge Inconclusive
T-initiated 263 119 (45.2%) 3 (1.1%) 89 (33.8%) 52 (19.8%)
S-initiated 322 98 (30.4%) 4 (1.2%) 167 (51.9%) 53 (16.5%)
Reactive 350 206 (58.9%) 10 (2.9%) 96 (27.4%) 38 (10.9%)

 

X2(6, n=935)=69.521, p=.000

Table 13 Type and the successfulness of uptake residuals

 

 

Successful Unsuccessful Acknowledge Inconclusive
T-initiated .0 -1 .0 -1 .5 2.4
S-initiated -6.6 -1 .0 6.5 .7
Reactive 6.5 1 .8 -5 .0 -29

 

Source
Tables 14, 15, 16, and 17 illustrate the occurrence and successfulness of learner uptake

according to the source of feedback episodes. Table 14 shows that learners responded to

110

teacher feedback in 69.8% of code-related episodes, 56.4% of message-related episodes,
and 79.6% of semantic episodes. Adjusted standardized residuals in Table 15 reveal that
uptake was signiﬁcantly more frequent in code-related and semantic episodes while no
uptake and no opportunity were signiﬁcantly more frequent in message-related episodes.
Table 16 shows that when uptake did occur, 51.8% was successful in code-related
episodes, 33.7% was successﬁrl in message-related episodes, and 33.3% was successful
in semantic episodes. Adjusted standardized residuals in Table 17 reveal that code-related
episodes led to signiﬁcantly more successful uptake while message-related and semantic
episodes led to signiﬁcantly more acknowledge. There is no signiﬁcant difference among
the three in terms of unsuccessful and inconclusive uptake. Chi-square tests indicate that
both the occurrence and successfulness of uptake are signiﬁcantly related to the source of

feedback episodes (p=.000 in both cases).

Table 14 Source and the occurrence of uptake

 

 

 

Number of Occurrence of uptake
episodes Uptake No uptake No opportunity
Code 858 599 (69.8%) 169 (19.7%) 90 (10.5%)
Message 527 297 (56.4%) 132 (25.0%) 98 (18.6%)
Semantic 49 39 (79.6%) 5 (10.2%) 5 (10.2%)

 

x2(4, n=1434)=34.345, p=.000

Table 15 Source and the occurrence of uptake residuals

 

 

Uptake No uptake No opportunity
Code 4.5 -1 .9 -4.0
Message -5.4 2.6 4.3

Semantic 2.2 -l .9 -.7

 

lll

Table 16 Source and the successfulness of ruitake

 

 

 

Number of Successfulness of uptake
uptake Successﬁrl Unsuccessful Acknowledge Inconclusive
Code 599 310 (51.8%) 13 (2.1%) 175 (29.2%) 101(16.9%)
Message 297 100 (33.7%) 3 (1.0%) 155 (52.2%) 39 (13.1%)
Semantic 39 13 (33.3%) 1 (2.6%) 22 (56.4%) 3 (7.7%)

 

X2(6, n=935)=52.206, p=.000

Table 17 Source and the successfulness of uptake residuals

 

 

Successful Unsuccessful Aclcnowledge Inconclusive
Code 5.3 1.1 -7.1 1.8
Message -4.8 -1.3 6.3 -1.3
Semantic -1.5 .4 2.5 -1.3

 

Response

Tables 18 and 19 illustrate the occurrence and successfulness of learner uptake according
to the response type of feedback episodes. Table 18 shows that learners responded to
teacher feedback in 62.4% of provide episodes and 98.2% of elicit episodes. Adjusted
standardized residuals in the table reveal that uptake was signiﬁcantly more frequent in
elicit episodes while no uptake and no opportunity were signiﬁcantly more frequent in
provide episodes. Table 19 shows that when uptake did occur, 38.7% was successful in
provide episodes and 93.7% was successful in elicit episodes. Adjusted standardized
residuals in the table reveal that elicit episodes led to signiﬁcantly more successful uptake
while provide episodes led to signiﬁcantly more acknowledge and inconclusive uptake.
There is no signiﬁcant difference between the two in terms of unsuccessful uptake. Chi-
square tests indicate that both the occurrence and successfulness of uptake are

signiﬁcantly related to the response type of feedback (p=.000 in both cases).

112

Table 18 Response and the occurrence of uptake

 

 

 

Number of Occurrence of uptake
episodes Uptake No uptake No opportunity
Provide 1321 824 (62.4%) 304 (23.0%) 193 (14.6%)
Residuals
(Provide) -- -7.7 5.3 4.4
Elicit 113 111 (98.2%) 2 (1 .8%) 0 (0.0%)

 

x2(2, n=1434)=59.045, p=.000

Table 19 Response and the successfulness of uptake

 

 

 

Number of Successfulness of uptake
uptake Successful Unsuccessful Acknowledge Inconclusive
Provide 824 319 (38.7%) 15 (1.8%) 349 (42.4%) 141 (17.1%)
Residuals
(Provide) -- -10.9 .0 8.1 4.2
Elicit 111 104 (93.7%) 2 (1 .8%) 3 (2.7%) 2 (1.8%)

 

x2(3, n=935)=121.207, p=.000

Linguistic focus

Tables 20, 21, 22, and 23 illustrate the occurrence and successfulness of learner uptake
according to the linguistic focus of feedback episodes. Table 20 shows that learners
responded to teacher feedback in 60.7% of vocabulary episodes, 68.8% of grammar
episodes, 69.1% of pronunciation episode, and 84.2% of spelling episode. Adjusted
standardized residuals in Table 21 reveal that uptake was signiﬁcantly more frequent in
grammar episodes while no opportunity was signiﬁcantly more frequent in vocabulary
episodes. Spelling episodes also resulted in signiﬁcantly more uptake and meanwhile
were least likely to result in no uptake. Table 22 shows that when uptake did occur,
38.4% was successful in vocabulary episodes, 42.6% was successful in grammar episodes,
85.7% was successful in pronunciation episodes, and 34.4% was successful in spelling
episodes. Adjusted standardized residuals in Table 23 reveal that pronunciation episodes

led to signiﬁcantly more successful uptake, vocabulary led to signiﬁcantly more

113

acknowledge, and spelling led to signiﬁcantly more inconclusive uptake. There is no
signiﬁcant difference among the four in terms of unsuccessful uptake. Chi-square tests
indicate that both the occurrence and successfulness of uptake are signiﬁcantly related to

the linguistic focus of feedback episodes (p=.000 in both cases).

Table 20 Linguistic focus and the occurrence of uptake

 

 

 

Number of Occurrence of uptake
episodes Uptake No uptake No opportunity
Vocabulary 712 432 (60.7%) 160 (22.5%) 120 (16.9%)
Grammar 532 366 (68.8%) 118 (22.2%) 48 (9.0%)
Pronunciation 152 105 (69.1%) 25 (16.4%) 22 (14.5%)
Spelling 38 32 (84.2%) 3 (7.9%) 3 (7.9%)

 

X2(6, n=1434)=26.145, p=.000

Table 21 Linguistic focus and the occurrence of uptake residuals

 

 

Uptake No uptake No opportunity
Vocabulary -3.6 l .0 3.7
Grammar 2.2 .6 -3.8
Pronunciation 1 .1 -l .6 .4
Spelling 2.5 -2.1 -1.0

 

Table 22 Linguistic focus and the successfulness of uptake

 

 

 

Number of Successfulness of uptake
uptake Successful Unsuccessful Acknowledge Inconclusive
Vocabulary 432 166 (38.4%) 5 (1.1%) 196 (45.4%) 65 (15.0%)
Grammar 366 156 (42.6%) 10 (2.8%) 140 (38.3%) 60 (16.4%)
Pronunciation 105 90 (85.7%) 2 (1.9%) 13 (12.4%) 0 (0.0%)
Spelling 32 11 (34.4%) 0 (0.0%) 3 (9.4%) 18 (56.3%)

 

X2(9, n=935)=130.089, p=.000

Table 23 Linguistic focus and the successfulness of uptake residuals

 

 

Successful Unsuccessful Acknowledge Inconclusive
Vocabulary -3.9 -l .4 4.5 -.2
Grammar -l .3 1 .7 .3 .7
Pronunciation 8.8 .1 -5.7 -4.6
Spelling -1 .3 -.8 -3.4 6.5

 

114

Directness

Tables 24 and 25 illustrate the occurrence and successfulness of learner uptake according
to the directness of feedback episodes. Table 24 shows that learners responded to teacher
feedback in 62.0% of direct episodes and 76.1% of indirect episodes. Adjusted
standardized residuals in the table reveal that uptake was signiﬁcantly more ﬁ'equent in
indirect episodes while no uptake was signiﬁcantly more frequent in direct episodes.
There is no signiﬁcant difference between the two in terms of no opportunity. Table 25
shows that when uptake did occur, 33.2% was successful in direct episodes and 78.3%
was successful in indirect episodes. Adjusted standardized residuals in the table reveal
that indirect episodes led to signiﬁcantly more successful uptake while direct episodes led
to signiﬁcantly more acknowledge and inconclusive uptake. There is no signiﬁcant
difference between the two in terms of unsuccessful uptake. Chi-square tests indicate that
both the occurrence and successfulness of uptake are signiﬁcantly related to the

directness of feedback episodes (p=.000 in both cases).

Table 24 Directness and the occurrence of uptake

 

 

 

Number of Occurrence of uptake
episodes Uptake No uptake No opportunity
Direct 1107 686 (62.0%) 269 (24.3%) 152 (13.7%)
Residuals (Direct) -- -3.9 4.3 .4
Indirect 327 249 (76.1%) 37 (11.3%) 41 (12.5%)

 

x2(2, n=1434)=19.835, p=.000

115

Table 25 Directness and the successfulness of uptake

 

 

 

Number of Successfulness of uptake
uptake Successful Unsuccessful Acknowledge Inconclusive
Direct 686 228 (33.2%) 12 (1.7%) 309 (45.0%) 137 (20.0%)
Residuals
(Direct) -- -11.5 -1 .4 7.5 6.5
Indirect 249 195 (78.3%) 5 (2.0%) 43 (17.3%) 6 (2.4%)

 

x713, n=935)=145.059, p=.000

Complexity

Tables 26 and 27 illustrate the occurrence and successfulness of learner uptake according
to the complexity of feedback episodes. Table 26 shows that learners responded to
teacher feedback in 57.8% of simple episodes and 70.2% of complex episodes. Adjusted
standardized residuals in the table reveal that uptake was signiﬁcantly more frequent in
complex episodes while no opportunity was signiﬁcantly more frequent in simple
episodes. There is no signiﬁcant difference between the two in terms of no uptake. Table
27 shows that when uptake did occur, 38.1% was successful in simple episodes and
49.2% was successful in complex episodes. Adjusted standardized residuals in the table
reveal that complex episodes led to both signiﬁcantly more successﬁrl and unsuccessful
uptake while simple episodes led to signiﬁcantly more acknowledge. There is no
signiﬁcant difference between the two in terms of inconclusive uptake. Chi-square tests
indicate that both the occurrence and successfulness of uptake are signiﬁcantly related to
the complexity of feedback episodes (p value is .000 for the occurrence of uptake

and .001 for the successﬁrlness of uptake).

116

Table 26 Complexity and the occurrence of uptake

 

 

 

Number of Occurrence of uptake
CPiSOdCS Uptake No uptake No opportunity
Simple 573 331 (57.8%) 124 (21.6%) 118 (20.6%)
Residuals (Simple) -- -4.8 .2 6.5
Complex 861 604 (70.2%) 182 (21.1%) 75 (8.7%)

 

x2(2, n=1434)=44.227, p=.000

Table 27 Complexity and the successfulness of uptake

 

 

 

Number of Successfulness of uptake
uptake Successful Unsuccessful Acknowledge Inconclusive
Simple 331 126 (38.1%) 2 (0.6%) 149 (45.0%) 54 (16.3%)
Residuals
(Simple) -- -3.3 -2.1 p 3.4 .6
Complex 604 297 (49.2%) 15 (2.5%) 203 (33.6%) 89 (14.7%)

 

x2(3, n=935)=17.720,p=.001

Emphasis

Tables 28 and 29 illustrate the occurrence and successfulness of learner uptake according
to the emphasis of feedback episodes. Table 28 shows that learners responded to teacher
feedback in 61.0% of light episodes and 65.8% of heavy episodes. Adjusted standardized
residuals in the table reveal that there is no signiﬁcant difference between the two in
terms of uptake and no uptake but no opportunity was signiﬁcantly more frequent in light
episodes. Table 29 shows that when uptake did occur, 64.0% was successful in light
episodes and 42.7% was successful in heavy episodes. Adjusted standardized residuals in
the table reveal that light episodes led to signiﬁcantly more successful uptake while
heavy episodes led to signiﬁcantly more inconclusive uptake. There is no signiﬁcant
difference between the two in terms of unsuccessful uptake and acknowledge. Chi-square

tests indicate that both the occurrence and successﬁrlness of uptake are signiﬁcantly

ll7

related to the emphasis of feedback episodes (p value is .014 for the occurrence of uptake

and .000 for the successfulness of uptake)

Table 28 Ermihasis and the occurrence of uptake

 

 

 

Number of Occurrence of uptake
episodes Uptake No uptake No opportunity
Light 182 111 (61.0%) 34 (18.7%) 37 (20.3%)
Residuals (Light) - -1.3 -.9 2.9
Heavy 1252 824 (65.8%) 272 (21.7%) 156 (12.5%)

 

x2(2, n=1434)=8.569, p=.014

Table 29 Emphasis and the successfulness of uptake

 

 

 

Number of Successfulness of uptake
uptake Successful Unsuccessful Acknowledge Inconclusive
Light 111 71 (64.0%) 2 (1.8%) 35 (31.5%) 3 (2.7%)
Residuals (Light) -- 4.2 .0 -1.4 -3.9
Heavy 824 352 (42.7%) 15 (1.8%) 317 (38.5%) 140 (17.0%)

 

x2(3, n=935)=24.068, p=.000

To sum up, the descriptive and inferential analysis in this section reveals that all the
seven characteristics of feedback episodes signiﬁcantly affected both the occurrence and
successfulness of uptake (p<.05). Adjusted standardized residuals indicate that code-
related, elicit, indirect, and complex episodes resulted in both signiﬁcantly more uptake
and signiﬁcantly more successful uptake (ASR>2.0). In addition to episodes with these
characteristics, light episodes also resulted in signiﬁcantly more successﬁrl uptake.
Regarding the type and linguistic focus of feedback episodes, while student-initiated and
spelling episodes led to signiﬁcantly more, learner uptake, reactive and pronunciation

episodes led to signiﬁcantly more successful uptake.

118

4.4 Characteristics of feedback episodes and recall comments

This section reports the relationship between the general characteristics of feedback
episodes and learners’ comments in the stimulated recall interviews. After an overview of
the overall recall comments, the frequency of comments related to teacher feedback is
presented in relation to each characteristic in the order of type, source, response,

linguistic focus, directness, complexity, and emphasis.

Overview of recall comments

Tables 30, 31, and 32 show the nature of learners’ comments during the stimulated recall
interviews and therein their attention to teacher feedback. As Table 30 shows, among a
total of 279 episodes, learners were thinking about language issues 86.0% of the time. In
6.5% of the episodes they were thinking about non-language—related issues. In 3.9% of
the episodes they were not thinking of anything. For the rest of the episodes, they either
didn’t remember what they were thinking or made a comment that was impossible to
classify. Table 31 shows that among a total of 240 language-related comments, 97.5% are
target-feature-related, 1.6% are non-target-feature-related, and 0.8% are inconclusive.
Table 32 speciﬁcally shows learners’ attention to teacher feedback. Among a total of 234
target-feature-related comments, 85.9% are related to teacher feedback, 12.8% are not
related to teacher feedback, and the rest are inconclusive. While the different tiers of
analysis provide an overview of all the recall comments, only the last one directly
addresses learners’ noticing of teacher feedback, one of the two major objectives of the
study. The statistical analysis of recall comments therefore only focused on target-

feature-related comments. Meanwhile, other comments, namely, non-language-related

119

comments, no thought, no memory, unclassiﬁable comments, non-target-feature-related
comments, and inconclusive language-related comments, compose an important part of
the recall interviews. Instead of totally excluding these categories from the chi-square
analysis, they were all considered together with inconclusive target-feature-related

comments as one category: other.

Table 30 Distribution of overall comments

 

 

 

 

 

 

 

 

 

Total number of Language- Non-language- No No .
comments related related thought memory Unclassrﬁable
279 240 1 8 1 1 7 3
(86.0%) (6.5%) (3.9%) (2.5%) (1 . 1%)
Table 31 Distribution of language-related comments
Total number of language-related Target-feature- Non-target-feature- .
Inconclusrve
comments related related
2 40 234 4 2
(97.5%) (1.6%) (0.8%)
Table 32 Distribution of target-feature-related comments
Total number of target-feature- Teacher-feedback- Non-teacher- Inconclusive
related comments related feedback-related
234 201 30 3
(85.9%) (12.8%) (1 .3%)
Type

Tables 33 and 34 show the rate of teacher-feedback-related comments in relation to the
type of feedback episodes. Table 33 shows that 74.8% of recall comments are related to
teacher feedback for teacher-initiated episodes, 76.0% are related to teacher feedback for
student-initiated episodes, and 60.9% are related to teacher feedback for reactive episodes.
Adjusted standardized residuals in Table 34 indicate that learners talked more (although

not signiﬁcantly) about teacher feedback after viewing student-initiated episodes and

120

signiﬁcantly more about other issues after viewing reactive episodes. Chi-square test

results suggest that the rate of teacher-feedback-related comments was signiﬁcantly

affected by the type of feedback episodes (p=.041).

Table 33 Type and teacher-feedback-related comments

 

 

 

 

 

Total number of Teacher-feedback- Non-teacher- Other
comments related feedback-related
T-initiated 115 86 (74.8%) 12 (10.4%) 17 (14.8%)
S-initiated 100 76 (76.0%) 12 (12.0%) 12 (12.0%)
Reactive 64 39 (60.9%) 6 (9.4%) 19 (29.7%)
x2(4, n=279)=9.988, p=.041
Table 34 Type and teacher-feedback-related comments residuals
Teacher-feedback-related Non-teacher-feedback-related Other

T-initiated .9 -.1 -.9
S-initiated 1 .2 .5 -1 .8
Reactive -2.4 -.4 3.1

 

Source

Table 35 shows the rate of teacher-feedback-related comments in relation to the source of

feedback episodes. A total of 68.1% of recall comments are related to teacher feedback

for code-related episodes and 76.7% are related to teacher feedback for message-related

episodes. Chi-square test results suggest that the rate of teacher-feedback-related

comments was not signiﬁcantly affected by the source of feedback episodes (p=.198).

121

Table 35 Source and teacher-feedback-related comments

 

 

Total number of Teacher-feedback— Non-teacher- Oth
comments related feedback-related er
Code 144 98 (68.1%) 17 (11.8%) 29 (20.1%)
Message 135 103 (76.7%) 13 (10.0%) 19 (13.3%)

 

X2(2, n=279)=3.240, p=.198 (Semantic episodes were conﬂated with message-related
episodes due to low expected cell counts)

Response

Table 36 shows the rate of teacher-feedback-related comments in relation to the response
type of feedback episodes. A total of 72.0% of recall comments are related to teacher
feedback for provide episodes and 72.2% are related to teacher feedback for elicit
episodes. The two are virtually the same. Chi-square test results suggest that the rate of
teacher-feedback-related comments was not signiﬁcantly affected by the response type of

feedback episodes (p=.997).

Table 36 Response and teacher-feedback-related comments

 

 

Total number of Teacher-feedback- Non-teacher- Other
comments related feedback-related
Provide 261 188 (72.0%) 28 (10.7%) 45 (17.2%)
Elicit 18 13 (72.2%) 2 (11.1%) 3 (16.7%)

 

X2(2, n=279)=.006, p=.997

Linguistic focus

Table 37 shows the rate of teacher-feedback-related comments in relation to the linguistic
focus of feedback episodes. A total of 76.0% of recall comments are related to teacher
feedback for vocabulary episodes, 63.1% are related to teacher feedback for grammar

episodes, and 70.0% are related to teacher feedback for pronunciation episodes. Chi-

122

square test results suggest that the rate of teacher-feedback-related comments was not

signiﬁcantly affected by the linguistic focus of feedback episodes (p=.227).

Table 37 Linguistic focus and teacher-feedback-related comments

 

Total number of Teacher-feedback— Non-teacher-

 

comments related feedback-related Other
Vocabulary 175 133 (76.0%) 18 (10.3%) 24 (13.7%)
Grammar 65 41 (63.1%) 7 (10.8%) 17 (26.2%)
Pronunciation 30 21 (70.0%) 4 (13.3%) 5 (16.7%)

 

X2(4, n=270)=5.648, p=.227 (Spelling was excluded due to low expected cell counts)

Directness

Table 38 shows the rate of teacher-feedback-related comments in relation to the
directness of feedback episodes. A total of 73.5% of recall comments are related to
teacher feedback for direct episodes and 61.8% are related to teacher feedback for
indirect episodes. Adjusted standardized residuals indicate that learners talked more
(although not signiﬁcantly) about teacher feedback after viewing direct episodes and
signiﬁcantly more about other issues after viewing indirect episodes. Chi-square test
results suggest that the rate of teacher-feedback-related comments was signiﬁcantly

affected by the directness of feedback episodes (p=.029).

Table 38 Directness and teacher-feedback-related comments

 

Total number Teacher- Non-teacher-

 

of comments feedback-related feedback-related Other
Direct 245 180 (73.5%) 28 (11.4%) 37 (15.1%)
Residuals . .
(Direct) -- 1 .6 .9 -2.6
Indirect 34 21 (61.8%) 2 (5.9%) 11 (32.4%)

 

x2(2, n=279)=7.105, p=.029

123

Complexity

Table 39 shows the rate of teacher-feedback-related comments in relation to the
complexity of feedback episodes. A total of 62.1% of recall comments are related to
teacher feedback for simple episodes and 74.4% are related to teacher feedback for
complex episodes. Adjusted standardized residuals indicate that learners talked
signiﬁcantly more about teacher feedback after viewing complex episodes and
signiﬁcantly more about other issues after viewing simple episodes. Chi-square test
results suggest that the rate of teacher-feedback-related comments was signiﬁcantly

affected by the complexity of feedback episodes (p=.049).

Table 39 Complexity and teacher-feedback-related comments

 

 

Total number Teacher- Non-teacher- Other
of comments feedback-related feedback-related
Simple 58 36 (62.1%) 6 (10.3%) 16 (27.6%)
Residuals
(Simple) -- -2.0 -.1 2.4
Complex 221 165 (74.7%) 24 (10.9%) 32 (14.5%)

 

x2(2, n=279)=6.045, p=.049

Emphasis

Table 40 shows the rate of teacher-feedback-related comments in relation to the emphasis
of feedback episodes. A total of 53.8% of recall comments are related to teacher feedback
for light episodes and 72.9% are related to teacher feedback for heavy episodes. Adjusted
standardized residuals indicate that learners talked more (although not signiﬁcantly).
about teacher feedback after viewing heavy episodes and signiﬁcantly more about other

issues after viewing light episodes. Chi-square test results suggest that the rate of teacher-

124

feedback-related comments was signiﬁcantly affected by the emphasis of feedback

episodes (p=.OO7).

Table 40 Emphasis and teacher-feedback-related comments

 

 

Total number Teacher- Non-teacher- Other
of comments feedback-related feedback-related
Light 13 7 (53.8%) 0 (0.0%) 6 (46.2%)
Residuals (Light) -- -l .7 -1.2 3.1
Heavy 266 194 (72.9%) 30 (11.3%) 42 (15.8%)

 

x2(2, n=279)=10.033, p=.007

To sum up, the descriptive and inferential analysis in this section reveals that among the
seven characteristics of feedback episodes, only the type, directness, complexity, and
emphasis of feedback episodes signiﬁcantly affected the rate of teacher-feedback-related
comments (p<.05). In other words, source, response, and linguistic focus did not
signiﬁcantly affect the rate of teacher-feedback-related comments. Adjusted standardized
residuals indicate that complex episodes led to signiﬁcantly more teacher-feedback-
related comments. Learners also made more comments about teacher feedback after
viewing student-initiated, direct, and heavy episodes although the values of the residuals

are not up to 2.0.

4.5 Characteristics of feedback episodes and test results

This section reports the relationship between the general characteristics of feedback
. episodes and test results. After an overview of the general test results, results from both
the immediate test and delayed test are presented together according to each characteristic
in the order of type, source, response, linguistic focus, directness, complexity, and

emphasis.

125

Overview of test results

Table 41 shows the general results of the two tests. In the immediate test, learners
provided a correct answer for 64.7% of the total items, an incorrect answer for 30.9% of
the total items, and a partially correct answer for 4.4% of the total items. In the delayed
test, the percentage of correct answers (53.2%) declined by 11.5% while the percentages
of incorrect answers (42.2%) and partially correct answers (4.6%) increased by 11.3%

and 0.2% respectively.

Table 41 Overview of test results

 

 

 

Immediate test Delayed test
Number Percentage Number Percentage
Correct 176 64.7% 150 53.2%
Incorrect 84 30.9% 1 19 42.2%
Partially correct 12 4.4% 13 4.6%
Total 272 100% 282 100%

 

Type

Table 42 is an illustration of test results according to the type of feedback episodes. In the
immediate test, the percentage of correct answers is 69.0% for teacher-initiated episodes,
72.9% for student-initiated episodes, and 56.0% for reactive episodes. Adjusted
standardized residuals suggest that learners provided signiﬁcantly more correct answers
for test items developed from student-initiated episodes and signiﬁcantly less correct
answers for items developed from reactive episodes. In the delayed test, the percentages
of correct answers for the three types of feedback episodes are very close to one another,

with 55.7% for teacher-initiated episodes, 50.6% for student-initiated episodes, and

126

52.3% for reactive episodes. Chi-square analysis indicates that the type of feedback

episodes signiﬁcantly affected the results of the immediate test (p=.030) but not the

results of the delayed test (p=.772).

Table 42 Type and test results

 

 

 

 

 

 

Immediate test Delayed test
Number Residuals Number
of test Correct Incorrect of test Correct Incorrect
. (Correct) .
items items
T- 71 49 8 22 1 1 5 64 51
initiated (69.0%) ' (31%) (55.7%) (44.4%)
S. . 85 62 2.0 23 79 4O 39
rmtrated (72.9%) (27.1%) (50.6%) (49.4%)
. 65 51 46 42
Ram“ ”6 (56.0%) '2'6 (43.9%) 88 (52.3%) (47.7%)
x2(2, n=272)=7.030, p=.030 x’(2, n=282)=.517, p=.772
Source

Table 43 is an illustration of test results according to the source of feedback episodes. In

the immediate test, the percentage of correct answers is 63.3% for code-related episodes,

67.4% for message-related episodes, and 66.7% for semantic episodes. In the delayed test,

the percentage of correct answers is 55.4% for code-related episodes and 49.5% for

message-related episodes. Chi-square analysis indicates that the source of feedback

episodes did not signiﬁcantly affect the results of either test (p value is .796 for the

immediate test and .342 for the delayed test).

127

Table 43 Source and test results

 

 

 

 

 

 

Immediate test Delayed test
Numl’e’ 0f Correct Incorrect Number Of - Correct Incorrect
test rtems test rtems

l 12 '65 98 79
C°d° 177 (63.3%) (36.8%) 177 (55.4%) (44.6%)

58 28 52 53
Message 86 (67.4%) (32.5%) 105 (49.5%) (50.5%)
Semantic 9 6 3 Conﬂated with message due to low

(66.7%) (33.3%) expected cell counts
x2(2, n=272)=.455, p=.796 x2(1, n=282)=.904, p=.342

Response

Table 44 is an illustration of test results according to the response type of feedback
episodes. In the immediate test, the percentage of correct answers is 63.7% for provide
episodes and 75.0% for elicit episodes. In the delayed test, the percentage of correct
answers for the two is 51.9% and 70.0% respectively. Chi-square analysis indicates that
the response type of feedback episodes did not signiﬁcantly affect the results of either test

(p value is .269 for the immediate test and .118 for the delayed test).

Table 44 Response and test results

 

 

 

 

 

Immediate test Delayed test
N‘m’l’e’ 0f Correct Incorrect Numi’e’ 0f Correct Incorrect
ICSI Items tCSI Items
. 158 90 136 126

”“1"" 248 (63.7%) (36.3%) 262 (51 .9%) (48.1%)

. . 18 6 14 6
Elm" 2" (75.0%) (25.0%L 20 (70.0%) (30.0%)

x’(1, n=272)=l .221, p=.269 x20, n=282)=2.443, p=.l 18

 

Linguistic focus
Table 45 is an illustration of test results according to the linguistic focus of feedback
episodes. In the immediate test, the percentage of correct answers is 68.8% for

vocabulary episodes, 67.0% for grammar episodes, and 48.7% for pronunciation episodes.

128

In the delayed test, the percentage of correct answers for the three is 46.8%, 64.2%, and
52.9% respectively. The percentage of correct answers for spelling episodes is 26.7%.
Adjusted standardized residuals suggest that learners provided signiﬁcantly more correct
answers for test items developed from grammar episodes and signiﬁcantly less correct
answers for items developed ﬁ'om spelling episodes. Chi-square analysis indicates that
the linguistic focus of feedback episodes signiﬁcantly affected the results of the delayed

test (p=.013) but not the results of the immediate test (p=.063).

Table 45 Linguistic focus and test results

 

 

 

 

 

Immediate test Delayed test
Number Number Residuals ~
of test Correct Incorrect of test Correct Incorrect
. . (Correct)
rtems rtems
86 39 58 66
V°°abulary 125 (68.8%) (31.2%) 124 (46.8%) '1'8 (53.3%)
69 34 70 39
Gramma’ 103 (67.0%) (33.1%) 109 (64.2%) 2'8 (35.8%)
. . 19 20 18 16
Pmmmc’a’m” 39 (48.7%) (51.3%) 34 (52.9%) '0 (47%)
. Excluded due to low 4 11
Spelhng expected cell counts 15 (26.7%) ‘2'1 (73.3%)
X2(2, n=267)=5.526, p=.063 X2(3, n=282)=10.860, p=.013
Directness

Table 46 is an illustration of test results according to the directness of feedback episodes.
In the immediate test, the percentage of correct answers is 68.4% for direct episodes and
53.0% for indirect episodes. Adjusted standardized residuals suggest that learners
provided signiﬁcantly more correct answers for test items developed from direct episodes.
In the delayed test, the percentage of correct answers is 54.3% for direct episodes and

49.2% for indirect episodes. Chi-square analysis indicates that the directness of feedback

129

episodes signiﬁcantly affected the results of the immediate test (p=.023) but not the

results of the delayed test (p=.484).

 

 

 

 

 

 

Table 46 Directness and test results
Immediate test Delayed test
. Number
1:12:13?me Correct 1:52:23; Incorrect of test Correct Incorrect
items
. 141 65 121 l 02
Due“ 206 (68.4%) 2'3 (31.5%) 223 (54.3%) (45.7%)
. 35 31 29 30
“um 66 (53.0%) '2'3 (47.0%) 59 (49.2%) (50.9%)
x2(1, n=272)=5.202, p=.023 x2(1, n=282)=.489, p=.484
Complexity

Table 47 is an illustration of test results according to the complexity of feedback episodes.

In the immediate test, the percentage of correct answers is 54.5% for simple episodes and

69.6% for complex episodes. In the delayed test,

the percentage of correct answers for the

two is 45.5% and 58.2% respectively. Adjusted standardized residuals suggest that

learners provided signiﬁcantly more correct answers for test items developed ﬁom

complex episodes in both the immediate test and delayed test. Chi-square analysis

indicates that the complexity of feedback episodes signiﬁcantly affected the results of

both tests (p value is .015 and .037 respectively).

Table 47 Complexity and test results

 

 

 

Immediate test Delayed test
Number . Number .
of test Correct Resrduals Incorrect of test Correct Resrduals Incorrect
. (Correct) . (Correct)
rtems rtems
. 48 40 5 1 61
SM?” 88 (54.5%) ‘2'4 (45.5%) 1 12 (45.5%) '2'1 (54.5%)
128 56 99 71
ample" 18" (69.6%) 2'4 (30.4%) 170 (58.2%) 2'1 (41.8%)

 

x2(1, n=272)=5.880, p=.015

x20, n=282)=4.374, p=.037

 

130

Emphasis

Table 48 is an illustration of test results according to the emphasis of feedback episodes.
In the immediate test, the percentage of correct answers is 32.3% for light episodes and
68.9% for heavy episodes. Adjusted standardized residuals suggest that learners provided
signiﬁcantly more correct answers for test items developed from heavy episodes. In the
delayed test, the percentage of correct answers is 41.2% for light episodes and 54.8% for
heavy episodes. Chi-square analysis indicates that the emphasis of feedback episodes
signiﬁcantly affected the results of the immediate test (p=.000) but not the results of the

delayed test (p=.134).

Table 48 Emphasis and test results

 

 

 

 

 

Immediate test Delayed test
Number Of Correct Resrduals Incorrect Number Of Correct Incorrect
test rtems (Correct) test rtems

. 10 21 14 20
“gm 31 (32.3%) '4'0 (67.7%) 34 (41.2%) (58.8%)

166 75 136 112
Heavy 24’ (68.9%) 4'0 (31.2%) 248 (54.8%) (45.1%)

x2(1, n=272)=16.130, p=.000 x2(1, n=282)=2.242,p=.134

 

To sum up, the descriptive and inferential analysis in this section reveals that the results
of the immediate test were signiﬁcantly affected by the type, directness, complexity, and
emphasis of feedback episodes (p<.05). Adjusted standardized residuals indicate that
learners provided signiﬁcantly more correct answers for items developed from student-
initiated, direct, complex, and heavy episodes (ASR>2.0) than those from teacher-
initiated/reactive, indirect, simple, and light episodes. The results of the delayed test, on
the other hand, were signiﬁcantly affected only by the linguistic focus and complexity of

feedback episodes. Adjusted standardized residuals indicate that learners provided

131

signiﬁcantly more correct answers for items developed from grammar and complex

episodes than those from vocabulary/pronunciation/spelling and simple episodes.

4.6 Review of results

The results presented above show that all seven characteristics of feedback episodes had
a statistically signiﬁcant bearing on the occurrence and successfulness of learner uptake.
In terms of teacher-feedback-related recall comments, however, only type, directness,
complexity, and emphasis showed a statistically signiﬁcant effect. These four
characteristics were also shown to have signiﬁcantly affected the results of the immediate
test. Results of the delayed test, on the other hand, were only signiﬁcantly inﬂuenced by
linguistic focus and complexity. The table below is a summary of the overall results in
this section. The plus sign symbolizes a statistically signiﬁcant relationship while the
minus sign represents a statistically non-signiﬁcant relationship. This convention also

applies to the summary tables in Chapters 5 and 6.

Table 49 Overall results by the characteristics of feedback episodes

 

Occurrence Successfulness of Feedback-related Immediate Delayed

 

of uptake uptake comments test test

Type + + + + -
Source + + - - -
Response + + - - -
ngurstrc + + - _ +
focus

Directness + + + + -
Complexity + + _ + + +
Emphasis + + + + -

 

132

Figure 5 is an overview of the characteristics that predicted uptake, successful uptake,
teacher-feedback-related comments, and correct test results. Speciﬁcally, student-initiated,
code-related, elicit, spelling, indirect, and complex episodes led to more uptake. Reactive,
code-related, elicit, pronunciation, indirect, complex, and light episodes led to more
successful uptake. Teacher-feedback-related comments were only predicted by complex
episodes. As for test results, more correct answers resulted from student-initiated, direct,
complex, and heavy episodes than from teacher-initiated/reactive, indirect, simple, and
light episodes in the immediate test and more from grammar and complex episodes than

from . y], loyvlling and simple episodes in the delayed test.

Figure 5 Overview of the characteristics of feedback episodes predicting (successful)
uptake, teacher-feedback-related comments, and correct test results

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Stimulated recall comments
Teacher-
Leamer uptake fee dback-
S' .. (1 Complex ’ related
-C-£etrate comments
-Elicit
-Spelling —’ Uptake
-Indirect
-Complex
Language tests
-Reactive TS: . Correct
-Code mrtrated answers in
-Elicit -D1rect —> immediate
-Pronunciation _, Successful -C0mplex test
-Indirect uptake L
-Cornplex
-Light Correct
-Grammar answers in
-Complex delayed
test

133

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

4.7 Discussion

This section discusses the results presented in the previous sections. The occurrence and
successfulness of uptake are examined together with learners’ stimulated recall comments
and test results to explore learners’ noticing of teacher feedback and the consequent
language improvement. Due to the cell count requirement of chi-square tests, not all the
categories of each variable were included in the results. Some categories are therefore
discussed solely according to descriptive statistics. On the other hand, an examination of
every category regardless of their relative importance would make the discussion
extremely long and tedious. The discussion therefore focuses on categories that stand out
from the results. Also, in the effort to interpret the results, new calculations were made
when necessary. These rules also apply to the discussion in Chapters 5 and 6 (in Chapter
5, there are also variables which did not receive chi-square analysis due to low expected

cell counts).

One thing the present study is different from previous studies examining the
characteristics of feedback episodes (e.g., Ellis et al., 2001a, 2001b; Loewen, 2004, 2005)
is the operationalization of uptake and the classiﬁcation of the successfulness of uptake.
In those studies, uptake only included learners’ verbal responses to teacher feedback. It
was classiﬁed as either successful or unsuccessful in inferential statistical analysis, with
unsuccessful uptake covering other types of uptake such as acknowledgement or
inconclusive uptake. In the present study, the use of a video camera made it possible for
me to look at learners’ nonverbal responses to teacher feedback, for example, nodding,

writing, acting, and pointing (See chapter 3 for detail). Acting and pointing occurred in a

134

small number of cases and were coded as successful uptake as I saw appropriate.
Nodding and writing alone (i.e., not including those accompanied by verbal responses
such as “oh” and “yeah”), on the other hand, occurred frequently across the nine classes.
Nodding was coded as acknowledge and writing was coded as inconclusive uptake.
Consequently, there is a high percentage of both acknowledge (37 .6%) and inconclusive
uptake (15.3%). Given the big percentage of these response types and the possibility that
they might make a difference to the overall results, it was decided that they should not be
excluded from the statistical analysis and that they should receive sufﬁcient attention in

the discussion of the results.

Before scrutinizing the detailed results, it is necessary to take a look at the overall results.
Among a total of 1434 episodes, 65.2% resulted in some kind of uptake and 45.2% of the
uptake was successful. In Ellis et. al (2001a) and Loewen’s (2004) studies, the level of
both uptake and successﬁrl uptake was above or close to 70%. One possible explanation
for the lower level of uptake in the present study is the opportunities learners were given
to respond to teachers. In the study of Ellis et al. (2001a), no opportunity was not
reported in the result. It is not clear whether it was coded as no uptake or not included at
all. In Loewen’s study, the overall percentage of no opportunity was 9.5%, 4% lower than
that of the present study (13.5%). This means learners in the present study had fewer
opportunities to respond to teachers. It is possible that they would produce a higher level
of uptake if given more opportunities. As for the lower level of ‘successful uptake, it

could be because the inclusion of nonverbal uptake in the present study largely increased

135

the amount of acknowledge and inconclusive uptake, which in turn lowered the

percentage of successful uptake.

Regarding the effect of speciﬁc characteristics on the occurrence and successfulness of
uptake, the three studies are not identical in the number of characteristics examined. In
addition to all the seven characteristics in the present study, Loewen also examined the
timing of uptake (i.e., when teacher feedback is given). Ellis et al., on the other hand,
only looked at type, source, directness, complexity, and linguistic focus. Moreover, the
design of the studies is not exactly the same either. In addition to the differences
described in the previous paragraph, the object of examination is different too. Both Ellis
et a1. and Loewen’s studies explored the effect of focus on form. Activities with a focus
on pre-targeted forms were excluded from the analysis. The episodes they examined
therefore were strictly mearring-based. The present study, in contrast, investigated the
effect of feedback episodes, which as illustrated in Chapter 1 is more of an umbrella
concept. Activities which aimed to promote language accuracy were included in the
analysis. The episodes examined therefore are partially meaning-focused and partially
form-focused. Due to these differences, it is not easy to compare the speciﬁcs about each
characteristic. One thing these studies do have in common is that some characteristics,
complexity, for instance, have a signiﬁcant effect on the occurrence and successfulness of

uptake.

With respect to testing, no test was reported in either Ellis et a1. or Loewen’s study.

Another study by Loewen (2005), however, did report the test results based on the same

136

data as his 2004 study. The overall percentage of correct answers was 43.3%, which was
15.5% lower than the percentage in the present study (58.8%). The higher percentage of
correct answers in the present study could be a result of the inclusion of test items
developed from form-focused activities. Meanwhile, both studies found that the
percentage of correct answers in general was lower in the delayed test than in the
immediate test (dropped by 8.3% in Lowen’s study and 11.5% in the present study). This
is a natural result of memory decay. Finally, concerning the stimulated recall comments,
none of the other three studies used it as an instrument. It is carefully examined in the

detailed discussion of each characteristic below.

The ﬁrst characteristic examined is type. It describes how a feedback episode is initiated.
Chi-square analysis indicates that the type of feedback episodes signiﬁcantly affected
both the occurrence and successfulness of uptake (p=.000 in both cases). In terms of
uptake, student-initiated episodes resulted in the highest level of uptake (71 .6%) while
teacher-initiated episodes led to the lowest level of uptake (54.7%), with reactive
episodes in the middle (69.6%). This result could have been caused by the nature of the
language problems. in the episodes. In student-initiated episodes, the questions learners
asked were real problems learners thought they needed to know. Learners therefore might
be more responsive to teachers' feedback. In teacher-initiated episodes, however, the
problems were based on teachers' assessment of learners’ knowledge. Learners might not
have these problems at all. They therefore might be less active in their interactions with
teachers. This could be the reason why teacher-initiated episodes also have a signiﬁcantly

higher level of no uptake (ASR=2.6) than the other two types of episodes. In reactive

137

episodes, even when there was a problem, learners might not feel the urgency to attend to
it. They therefore might not take as much initiative to respond to the feedback either.
Given these differences of the problems involved in the different types of episodes, it is

not strange that student-initiated episodes resulted in more uptake.

The initiation learners might take in student-initiated episodes could lead one to expect
more successful uptake in these episodes too. This nevertheless is not true. It turned out
that reactive episodes resulted in the highest level of successﬁrl uptake while student-
initiated episodes led to the lowest level of successful uptake (ASR value is 6.5 and -6.6
respectively). Student-initiated episodes also resulted in the highest level of acknowledge
(ASR=6.5). A look at the speciﬁcs of student-initiated episodes and reactive episodes
reveals part of the possible reasons why this is so. In a lot of student-initiated episodes
(51.6%), learners asked about the meaning of a word or asked about a word they needed
to express themselves. More often than not, they simply acknowledged the feedback once
they could continue with their interactions with the teacher or peers or when they could
continue the task at hand, hence leading to more acknowledge and less successful uptake.
Different from student-initiated episodes, a good percentage of reactive episodes (58.6%)
involve pronunciation problems. In these episodes, teachers often provided the correct
pronunciation and learners often simply repeated teachers right after the feedback,

leading to more successful uptake (see more in the discussion of linguistic focus).

The type of feedback episodes was also found to be signiﬁcantly related to teacher-

feedback-related comments (p=.041). To be speciﬁc, student-initiated episodes led to the

138

highest percentage of teacher-feedback-related comments (76.0%) while reactive
episodes led to the lowest percentage of teacher-feedback-related comments (60.9%).
This again could be explained by the reasons given above -- the nature of the problems
involved in the feedback episodes and how learners viewed it. That is, learners might pay
more attention to the feedback about a problem they were experiencing in student-
initiated episodes and less attention to the feedback about a problem initiated or detected
by other people, in this case, the teacher. As for the reason why reactive episodes, not
teacher-initiated episodes, led to the lowest percentage of teacher-feedback-related
comments, it could be because learners thought a question initiated by the teacher might
be of some importance and therefore paid more attention to it whether they understood it

or not.

Conceming test results, chi-square analysis indicates that there is a statistically signiﬁcant
difference among the three types of feedback episodes in the immediate test (p=.031).
Learners provided signiﬁcantly more correct answers for test items developed from
student-initiated episodes (ASR=2.0) and signiﬁcantly fewer correct answers for items
developed from reactive episodes (ASR=-2.4). This result corresponds well with the
results ﬁ'om the stimulated recall interviews -- more teacher-feedback-related comments
associated with student-initiated episodes and fewer teacher-feedback-related comments
associated with reactive episodes. The heavier attention to teacher feedback in student-
initiated episodes could have resulted in better performance with items developed from

these episodes. As for the delayed test, it was not signiﬁcantly affected by the type of

139

feedback episodes, probably because student-initiated episodes lost their advantage as a

result of memory decay.

The second characteristic examined is source. It describes the origin of a feedback
episode. Chi-square analysis indicates that the source of feedback episodes had a
signiﬁcant effect on both the occurrence and successfulness of uptake. Code-related
episodes led to signiﬁcantly more uptake (ASR=4.5) and more successful uptake
(ASR=5.3) than message-related episodes (ASR value is -5.4 for uptake and -5.3 for
successﬁrl uptake). Semantic episodes also led to signiﬁcantly more uptake than
message-related episodes, but only half of that of code-related episodes (ASR=2.2).
Moreover, just like message-related episodes, it led to less successful uptake (ASR=-1.5)
than code-related episodes as well. Semantic episodes are those where learners
understood the overall discourse but did not know the meaning of particular words or
phrases (Loewen, 2002). These episodes therefore are also meaning-based. If semantic
episodes were to be treated as part of message-related episodes as Ellis et al. (2001a) did,
it could be said that code-related episodes in general led to more (successful) learner
responses than message-related episodes. This could be due to two reasons. First, in
message-related episodes, learners were primarily concerned with communication. Once
the meaning of a language form was clearly conveyed and the communication breakdown
was ﬁxed, learners would continue with their interactions with teachers or peers. This
could be the reason why there are signiﬁcantly more acknowledge (ASR=6.3) and no '
uptake (ASR=2.6) in message-related episodes. In code-related episodes, however,

learners were primarily concerned with the accuracy of language forms. There was no

140

pressing need to continue a conversation and they could afford to uptake teacher
feedback and try to get it right. Another possible explanation is that because this is an
intensive English program and there were some grammar components in the majority of
the classes observed, learners might want to make sure they understand the target features

and would respond to teachers more (correctly) in episodes involving language forms.

According to what has been said, one might wonder whether learners would pay more
attention to code-related feedback than message-related feedback. Analysis of the
stimulated recall comments, however, shows that this is not the case. Although source
was shown not to signiﬁcantly affect the rate of teacher-feedback-related comments
(p=.198), the percentage of teacher-feedback-related comments is 8.6% higher for
message-related episodes than code-related episodes (76.7% and 68.1% respectively). It
is hard to tell exactly why this is so, but it does warn us that uptake is an optional move
(Ellis et al., 2001a; Loewen, 2004; Sheen, 2006), that less uptake does not necessarily
mean less noticing, and that learners might attend to teacher feedback even if they simply

acknowledge it or do not respond to it at all.

Regarding the test results, no statistically signiﬁcant relationship was observed between
source and test results (p>.05 in both the immediate test and delayed test). Given that
there is no statistically signiﬁcant relationship between source and teacher-feedback-
related comments either, a tentative conclusion can be drawn that because source did not
affect learners’ noticing of teacher feedback, it did not affect their learning. A careful

look at the percentages of correct answers in the immediate test shows that learners did

141

slightly better with test items developed from message-related episodes and semantic
episodes than those developed from code-related episodes (67.4% and 66.7% versus
63.3%). In the delayed test, however, learners provided slightly more correct answers for
items developed from code-related episodes than those from message-related episodes
(55 .4% versus 49.5%). One possible explanation for this change is that learners had been
re-exposed to some language forms that had been treated in code-related episodes before
the delayed test was administered. This is possible now that there were some focus
grammar activities or review activities in the lessons observed. As for target features
treated in message-related episodes, even though learners might also run across the same
language forms again, the chance is slimmer now that the language foci of these episodes

are generally more extensive and are not the focus of a lesson like grammar rules.

The third characteristic examined is response. It describes how teacher feedback is given.
Chi-square analysis indicates that there is a statistically signiﬁcant relationship between
teachers' response type and the occurrence of uptake. Adjusted standardized residuals
reveal that the level of uptake is signiﬁcantly higher in elicit episodes (ASR=7.7) while
the level of no uptake and no opportunity is signiﬁcantly higher in provide episodes
(ASR value is 5.3 and 4.4 respectively). There are two possible reasons why elicit
episodes led to more learner responses. One is that when teachers elicited answers from
learners, learners were “forced” to give an answer. The “forced” answer often constituted
learner uptake. Even though learners were not able to provide an answerall the time, the
expectation for them to do so largely increased the chance of learner uptake. Another

reason might be that many of these learners, aged around 20, were anxious to impress the

142

teacher in elicit episodes. As one learner reported in a stimulated recall interview: "I want
to let her know what I know...”, learners often tried to get the right answer when they
were given the opportunity to do so. This motivation to be the "knowing student" could
have also contributed to the higher level of uptake in elicit episodes. When teachers
provided answers, on the other hand, learners did not have to say anything if they did not
want to. Neither were there so many opportunities for them to impress teachers. This
might explain why there is a much higher percentage of no uptake (23.0%) and no

opportunity (14.6%) in provide episodes.

In terms of the successfulness of uptake, chi-square analysis indicates that it was also
signiﬁcantly affected by teachers' response type. Adjusted standardized residuals reveal
that elicit episodes resulted in signiﬁcantly more successﬁrl uptake (ASR=10.9) while
provide episodes resulted in signiﬁcantly more acknowledge and inconclusive uptake
(ASR value is 8.1 and 4.2 respectively). This result could also be explained by learners'
positions in the two types of episodes. In elicit episodes, often times learners would
manage to get the correct answers with the teachers’ and sometimes classmates’ help.
The correct answers would be considered as successful uptake. In provide episodes, on
the other hand, learners were often in a more passive position and might only give a
simple response without striving to use the right language form. The fact that the simple
provision of the correct form or language information as compared to elicitation led to a
surprisingly higher percentage of acknowledge (42.4% as against 2.7%) and inconclusive

uptake (17.1% as against 1.8%) could be evidence for this.

143

Given that elicit episodes led to both signiﬁcantly more uptake and successful uptake,
one might think they would also lead to signiﬁcantly more teacher-feedback-related
comments. Nevertheless, chi-square analysis indicates that there is no statistically
signiﬁcant difference between provide episodes and elicit episodes in terms of the recall
comments. A possible explanation for this result is that when teachers tried to elicit
answers from learners, learners took it as part of natural classroom interactions, not an
effort to draw their attention to a particular language form; they therefore did not pay
special attention to the feedback in elicit episodes. On the other hand, descriptive
statistics reveal that elicit episodes did result in a slightly higher percentage of comments
related to teacher feedback, meaning that learners could have paid slightly more attention

to the feedback in these episodes.

Attention has been claimed to be necessary (but not necessarily sufficient) for learning to
happen (Schmidt, 1990, 1993, 2001). The slightly more attention in elicit episodes should
have led to more correct answers in tests. This proved true in both the immediate test and
delayed test. In both tests, learners provided more correct answers for test items
developed ﬁom elicit episodes than those from provide episodes (75.0% versus 63.7% in
the immediate test and 70.0% versus 51.9% in the delayed test) although the difference
between the two is not statistically signiﬁcant (p value is .269 in the immediate test

and .118 in the delayed test).

The fourth characteristic examined is linguistic focus. It describes what language aspect

is targeted at in a feedback episode. Adjusted standardized residuals indicate that

144

grammar and spelling episodes resulted in signiﬁcantly more uptake (ASR values is 2.2
and 2.5 respectively) while vocabulary episodes resulted in a signiﬁcantly higher level of
no opportunity (ASR=3.7). As mentioned earlier, grammar practice activities were
included in the current data set. Since grammar was the focus of these activities, teachers
might give learners more opportunities to respond to their feedback, thus leading to more
uptake. As for spelling episodes, most of them (78.9%) were initiated by a learner with an
inquiry such as "How do you spell...?" As illustrated in the discussion of the type of
feedback episodes, learners as question initiators might be more responsive to teacher
feedback. It is therefore not surprising that learners produced more uptake in spelling
episodes too. Different from grammar episodes, vocabulary episodes often occurred in
more meaning-based activities (e.g., talking about the differences between the banking
system of the United States and that of one's own country or discussing one's ﬁrst-day
experience in the United Sates). The major concern therefore could be the contents of
conversations, not language forms. Consequently, teachers might simply continue with a
topic or start a new one once the feedback was given, thus leading to a higher level of no
opportunity. Different from spelling episodes too, vocabulary episodes did not
necessarily have learners as the dominant initiators of questions. Teachers could either
ask a question or react to learners' mistakes. Accordingly, learners might not always be in
an active position and hence would not always respond to the feedback, thus leading to a
higher level of no uptake (ASR value is 1.0, not signiﬁcant but higher than all the other

three types of linguistic foci).

145

Concerning the successfulness of uptake, chi-square analysis indicates that it was
signiﬁcantly affected by linguistic focus (p=.000). One noticeable thing in the result is
that pronunciation episodes resulted in the highest level of successful uptake (ASR=8.8).
This could be explained by the nature of teachers' feedback. It was found that in an
overwhelming majority of pronunciation episodes (96.1%), teachers provided the right
answer when a problem was detected. With teachers’ modeling, learners could easily
repeat it and correct themselves. In episodes with vocabulary, grammar, or spelling as the
linguistic focus, however, there were a variety of response types and learners sometimes
could not simply repeat what teachers said. Another noticeable thing in the result is that
inconclusive uptake constitutes 56.3% of learner responses in episodes with spelling as
the linguistic focus. This could be explained by the fact that in 52.6% of spelling episodes
learners started to write instead of talking once the teacher started to give the spelling of a
word, which was coded as inconclusive. Still another noticeable thing in the result is that
learners simply acknowledged teachers' feedback in a good percentage of vocabulary
episodes (45.4%). A look at these episodes reveals that many of them (54.5%) concern
word meanings. As discussed earlier, on such occasions there was not so much of a need
for learners to incorporate the target feature right away as long as the task at hand could

continue. Consequently, learners simply acknowledged the feedback.

As mentioned in the discussion of source, simple acknowledgement of teacher feedback
could also be an indicator of attention. The fact that learners produced the highest -
percentage of acknowledge in vocabulary episodes does not mean they were not paying

attention to teachers’ feedback. Compared with grammar and pronunciation episodes

146

(spelling episodes were not included because of its low expected cell count in chi-square
analysis), learners made more teacher-feedback-related comments after viewing
vocabulary episodes. One possibility is that since a large number of vocabulary episodes
concerned word meanings, it was crucial for learners to pay attention to teachers’
feedback and to understand the feedback in order for an interaction to continue. In
pronunciation and grammar episodes, however, prolonging a short vowel (e. g., kid
pronounced as /ki:d/) in a particular context or using the wrong tense of a verb which was
modiﬁed by an adverbial phrase (e. g., I go to Walmart yesterday morning) might not
cause comprehension problems. Learners therefore might not pay as much attention to

the feedback.

The next thing that was looked at is test results. Chi-square analysis indicates that the
linguistic focus of feedback episodes did not signiﬁcantly affect the results of the
immediate test. Descriptive statistics, however, shows that the percentages of correct
answers for vocabulary and grammar items (68.8% and 67.0% respectively) are both
about 20.0% higher than pronunciation items (Spelling was again not included because of
its low expected cell counts in the chi-square analysis). This result seems somewhat
surprising given the fact that learners produced the highest percentage of successful
uptake in pronunciation episodes. One possible explanation for this is that the big
majority of pronunciation problems (89.5%) did not cause communication breakdowns.
Learners' attention to the feedback about these problems was therefore not so intense.
Consequently, their memory of the correct pronunciation did not last long. Vocabulary

problems, again as shown earlier, often involved word meanings and might interrupt

147

classroom interactions. When there was a communication breakdown, learners might pay
more attention to the problem and retain the correct language feature longer. As for
grammar episodes, again because the big majority of classes observed (seven out of nine)
incorporated some grammar components in their lessons and there were grammar
practice activities, learners could have paid more attention to the focus grammar rules and

hence did better with grammar tests too.

Different from the immediate test, chi-square analysis shows that the results of the
delayed test were signiﬁcantly affected by the linguistic focus of feedback episodes
(p=.013). One thing that stands out is that although the percentage of correct answers for
vocabulary and grammar items is virtually the same in the immediate test, the percentage
of vocabulary items dropped by 22% in the delayed test while that of grammar items only
dropped by 2.8%. This could be because while vocabulary feedback was more extensive,
spread over the lessons as the classroom interactions required, and might not come back
again afterwards, some grammar rules which were the focus of grammar lessons might
have been reviewed later in a new lesson. Learners therefore could have received more
feedback about grammar problems and hence were able to remember the correct forms
for a longer time. Another thing that stands out in the results of the delayed test is that
learners provided the lowest percentage of correct answers for spelling items. This could
be explained by the fact that in all the spelling episodes, teachers spelt a word letter by

letter. It was probably hard for learners to remember these discrete letters two weeks later.

148

The ﬁfth characteristic examined is directness. It describes how explicit teacher feedback
is. Statistical analysis indicates that indirect episodes led to signiﬁcantly more uptake
(ASR=3.9) and successful uptake (ASR=11.5). Direct episodes, on the other hand,
resulted in signiﬁcantly more no uptake (ASR=4.3) and more acknowledge (ASR=7.5) as
well as inconclusive feedback (ASR=6.5). This result might have to do with the type of
feedback in different episodes. In indirect episodes, two major techniques were used:
recasts where teachers implicitly provided the correct form and elicit where teachers gave
linguistic and/or nonlinguistic cues to help learners ﬁgure out the correct answers
themselves. Elicit, as illustrated in the discussion of response type, led to a high
percentage of uptake and successﬁrl uptake. As for recasts, different from Lyster and
Ranta’s (1997) ﬁnding that they were ineffective in eliciting learner repair (successful
uptake), learners in the current study were able to correct themselves 62.5% of the time
when given the opportunity to do so, thus leading to a relatively higher level of uptake
and successful uptake. Now that both the two major components of indirect episodes led
to more uptake and successful uptake, it is not surprising that indirect episodes as a whole
led to a high level of. uptake and successﬁrl uptake. In direct episodes, on the other hand,
teachers often provided direct language information other than doing explicit error
correction. Some of the information teachers provided might not concern learners. For
example, learners might not be interested in the information about a teacher-initiated
question which they understood already before the feedback. They therefore might
choose to either avoid responding to the feedback or simply acknowledge it or respond to

it inconclusively.

149

About the stimulated recall interview, despite that fact that direct episodes have a much
lower percentage of uptake and successful uptake, its percentage of teacher-feedback-
related comments (73.5%) is 11.7% higher than that of indirect episodes (61.8%). This
could be because the feedback in indirect episodes is more implicit than the feedback in

explicit episodes, hence not so easy for learners to notice.

Concerning the test results, chi-square analysis indicates that they were signiﬁcantly
related to the directness of feedback episodes in the immediate test but not in the delayed
test (p value is .023 and .484 respectively). Descriptive statistics, however, shows that the
percentage of correct answers is higher for direct episodes in both tests (15.4% higher in
the immediate test and 5.1% higher in the delayed test). Given the crucial role of
attention in learning, this result is reasonable now that learners made signiﬁcantly more
teacher-feedback-related comments in the stimulated recall interviews after viewing

direct episodes.

The sixth characteristic examined is complexity. It describes how many feedback moves
there are in an episode. Chi-square analysis indicates that the complexity of feedback
episodes had a statistically signiﬁcant effect on all the ﬁve dependent variables examined:
the occurrence of uptake, the successﬁrlness of uptake, the rate of teacher-feedback-
related comments, the immediate test, and the delayed test (p value ranges from .000

to .049).

150

In terms of the occurrence of uptake, complex feedback led to 12.4% more uptake while
simple episodes led to 11.9% more no opportunity. Now that no opportunity is often not a
choice by learners (e. g., topic continuation by the teacher), the less frequent occurrence
of uptake in association with simple episodes does not indicate that learners were less
willing to respond to teachers in these episodes. What it could mean is that in simple
episodes teachers devoted less time to the linguistic form under discussion and thus

allowed fewer opportunities for learners to respond to them.

As for the successfulness of uptake, the percentage of successful uptake is 11.1% higher
in complex episodes than in simple episodes (49.2% and 38.1% respectively). This means
learners incorporated the correct language form in their language production more often
when there were multiple feedback moves in an episode. A possible explanation for this
is that multiple feedback moves constitute more input, thus enabling learners to

understand the target feature better and to use it correctly.

Concerning the stimulated recall comments, learners talked signiﬁcantly more about
teacher feedback after viewing complex episodes (ASR-=20), with a percentage of
teacher-feedback-related comments 12.6% higher than that of simple episodes (74.7%
and 62.1% respectively). This could be because the multiple feedback moves had helped

to draw more attention to the feedback from learners.

Regarding the test results, learners provided signiﬁcantly more correct answers for items

developed from complex episodes in both the immediate test and delayed test (ASR value

151

is 2.4 and 2.1 respectively). This shows that learners could have indeed paid more
attention to teacher feedback in complex episodes as reﬂected in the higher percentage of

teacher-feedback-related comments associated with these episodes.

The last characteristic examined is emphasis. It is the combination of directness and
complexity. Heavy episodes are either direct or complex or both while light episodes are
both indirect and simple. Emphasis was found to be signiﬁcantly related to the
occurrence of uptake, the successfulness of uptake, teacher-feedback-related comments,

and the results of the immediate test (p value ranges from .000 to .014).

In terms of the occurrence of uptake, heavy episodes led to slightly more uptake than
light episodes (65.8% and 61.0% respectively). Although not statistically signiﬁcant
(ASR=1.3), this result still could mean that learners were more responsive in heavy
episodes. Light episodes, on the other hand, led to a signiﬁcantly higher level of no
opportunity (ASR=2.9). Now that the feedback in light episodes is both indirect and
simple, this could be because many times teachers might not want to draw overtly heavy
attention to the target features in these episodes and purposeﬁrlly left no opportunities for

learners to respond to them.

In contrast to the occurrence of uptake, results from the successfulness of uptake indicate
that learners produced signiﬁcantly more successful uptake in: light episodes than in ‘
heavy episodes (ASR value is 4.2 against -4.2). This could be because light episodes only

contained one teacher feedback move; it was therefore easy for learners to incorporate the

152

target feature in their own utterance. In heavy episodes, by contrast, there are a big
percentage of episodes which involved multiple feedback moves (68.8%). Language
forms which require multiple feedback moves might be more complex and more difﬁcult
to take in. Learners therefore might choose to simply acknowledge teachers' feedback or
to respond to it inconclusively. This could be the reason why learners produced 7.0%
more acknowledge and signiﬁcantly more inconclusive uptake (ASR=3.9) in heavy

episodes.

Given that learners produced signiﬁcantly more successful uptake in light episodes, one
might expect that they would talk more about teacher feedback after viewing light
episodes and perform better with test items developed from light episodes. Analysis of
the stimulated recall comments and test results, however, reveals that this is not true. In
actuality, heavy episodes resulted in 19.1% more teacher-feedback-related comments
than light episodes (72.9% and 53.8% respectively). This could be because the multiple
feedback moves or/and the directness of the feedback in heavy episodes made the target
features more pronounced than those treated simply and indirectly in light episodes. It
proves once again that learners might pay attention to teacher feedback when their

responses to the feedback are simple acknowledgement or inconclusive uptake.

Concerning the test results, chi-square analysis shows that emphasis signiﬁcantly affected
the results of the immediate test but not the results of the delayed test. In the immediate
test, learners performed signiﬁcantly better with items developed ﬁ'om heavy episodes

than those developed from light episodes (ASR value is 4.0 against -4.0). In the delayed

153

test, although there is no statistically signiﬁcant difference between the two kinds of test
times, learners did get 13.6% more correct answers for items developed from heavy
episodes. The fact that learners performed better with test items developed from heavy
episodes in both tests goes well with the fact that learners talked more about teacher
feedback after viewing heavy episodes. It lends support to the assumption that more

attention leads to more learning.

4.8 Summary

In summary, the results reported in this chapter suggest that both the occurrence and
successfulness of uptake were signiﬁcantly affected by all the seven characteristics
examined. Code-related, elicit, indirect, and complex episodes led to both a higher level
of uptake and a higher level of successﬁrl uptake. About the other three characteristics,
student-initiated episodes, spelling episodes, and heavy episodes resulted in more uptake
while reactive episodes, pronunciation episodes, and light episodes resulted in more
successful uptake. Learners’ stimulated recall comments were signiﬁcantly affected by
four characteristics: type, directness, complexity, and emphasis. Student-initiated, direct,
complex, and heavy episodes are associated with a higher rate of teacher-feedback-
related comments. The results of the immediate test were also signiﬁcantly affected by
these same four characteristics in the same way, with the same four categories resulting
in more correct answers. The results of the delayed test were only signiﬁcantly affected
by linguistic focus and directness. Learners performed better with test items developed

from grammar and complex episodes. These results taken as whole show that the answer

154

to the ﬁrst research question is: yes, the characteristics of teacher feedback episodes

affect learners’ noticing and learning to a certain degree.

155

CHAPTER 5
NONLINGUISTIC CUES, THE NOTICING AND THE EFFECT OF TEACHER

FEEDBACK

5.1 Introduction

This chapter addresses the second research question: Do the nonlinguistic cues in teacher
feedback affect learners’ noticing and learning? After presenting the frequency of
nonlinguistic cues in teacher feedback and the distribution of their subcategories, I report
and discuss the relationship between nonlinguistic cues and learners’ responses to
teachers’ feedback, learners’ comments in the stimulated recall interviews, and results
from both the immediate test and delayed test. For purposes of this analysis,
“nonlinguistic” includes paralinguistic cues such as word stress and intonation, and

extralinguistic cues such as gestures and head movements.

5.2 Occurrence and distribution of nonlinguistic cues

This section reports the occurrence of general paralinguistic cues and general
extralinguistic cues as well as the distribution of their subtypes. Table 50 shows that
teachers used paralinguistic cues in 22.3% of a total of 1434 feedback episodes. The
percentage for episodes with extralinguistic cues is 40.7%. Table 51 shows that among a
total of 320 paralinguistic cues, 52.5% are word stress, 10.9% are rising intonation,
14.4% are eliciting stop, and 11.9% are dragging voice. Mimicking sound (3.1%) and
combination (7.2%) take the rest 10.3%. Concerning the type of extralinguistic cues,

among a total of 548 extralinguistic cues, 58.4% are gestures, 15.1% are head movements,

156

Table 50 Occurrence of nonlinguistic cues (N=l434)

 

 

Number Percentage
Paralinguistic cues
Present 320 22.3%
Absent 1 1 14 77.7%
Extralinguistic cues
Present 584 40.7%
Absent 850 59.3%

 

Table 51 Distribution of different types of nonlinguistic cues

 

 

Number Percentage
Type of paralinguistic cues
Word stress 168 52.5%
Rising intonation 35 10.9%
Eliciting stop 46 14.4%
Dragging voice 38 11.9%
Mimicking sound 10 3.1%
Combination 23 7.2%
Total 320 100%
Type of extralinguistic cues
Gestures 341 58.4%
Whole body acting 29 5.0%
Head movements 8 8 l 5.1%
Facial expressions 12 2.1%
Combination 1 14 19.5%
Total 5 84 100%
Type of gestures
Iconics l 39 40.8%
Metaphorics 38 1 1.1%
Deictics 99 29.0%
Beats 24 7.0%
Combination 41 12.0%
Total 341 100%

 

157

and 19.5% are combinations. Whole body acting (5.0%) and facial expressions (2.1%)
take the rest 7.1%. The distribution of different types of gestures is presented in the same
table to save space. Among a total of 341 gestures, 40.8% are iconic, 11.1% are
metaphoric, 29.0% are deictic, and 12.0% are combinations. With a percentage of 7.0%,

beats occurred most infrequently.

5.3 Nonlinguistic cues and Ieamer uptake

This section reports the relationship between the nonlinguistic cues in teacher feedback
and learner uptake. The occurrence and successfulness of uptake are presented together
according to each type of nonlinguistic cues and their subcategories in the order of
general paralinguistic cues, type of paralinguistic cues, general extralinguistic cues, type

of extralinguistic cues, and type of gestures.

General paralinguistic cues

Tables 52 and 53 illustrate the occurrence and successﬁrlness of uptake according to the
presence of general paralinguistic cues. Table 52 shows that learners responded to teacher
feedback in 73.1% of episodes with paralinguistic cues and 62.9% of episodes without
paralinguistic cues. Adjusted standardized residuals in the table reveal that uptake was
signiﬁcantly more frequent in episodes with paralinguistic cues while no opportunity was
signiﬁcantly more ﬁ‘equent in episodes without paralinguistic cues. There is no
signiﬁcant difference between the two in terms of no uptake. Table 53 shows that when
uptake did occur, 53.4% was successful in episodes with paralinguistic cues and 42.5%

was successful in episodes without paralinguistic cues. Adjusted standardized residuals in

158

the table reveal that episodes with paralinguistic cues led to signiﬁcantly more successful
uptake while episodes without paralinguistic cues led to signiﬁcantly more acknowledge.
There is no signiﬁcant difference between the two in terms of unsuccessful and
inconclusive uptake. Chi-square tests indicate that both the occurrence and successfulness
of uptake are signiﬁcantly related to teachers’ use of general paralinguistic cues (p value

is .000 for the occurrence of uptake and .006 for the successﬁrlness of uptake).

Table 52 General paralinguistic cues and the occurrence of uptake

 

 

 

Number of Occurrence of uptake
episodes Uptake No uptake No opportunity
Present 320 234 (73.1%) 62 (19.4%) 24 (7.5%)
Residuals
(Present) -- 3.4 -1 .0 -3.5
Absent 1114 701 (62.9%) 244 (21.9%) 169 (15.2%)

 

x2(2, n=1434)=15.578,p=.000

Table 53 General paralinguistic cues and the successfulness of uptake

 

 

 

Number of Successfulness of uptake
uptake Successful Unsuccessful Acknowledge Inconclusive
Present 234 125 (53.4%) 4 (1.7%) 66 (28.2%) 39 (16.7%)
Residuals ’
(Present) " 2'9 ‘-1 -3.4 .7
Absent 701 298 (42.5%) 13 (1.8%) 286 (40.8%) 104 (14.8%)

 

X2(3, n=935)=12.410, p=.006

Type of paralinguistic cues

Tables 54, 55, and 56 illustrate the occurrence and successfulness of uptake according to
the type of paralinguistic cues. Table 54 shows that learners responded to teacher
feedback in 70.8% of episodes with word stress, 80.0% of episodes with rising intonation,
73.9% of episodes with eliciting stop, and 76.3% of episodes with dragging voice. Table

55 shows that when uptake did occur, 44.9% was successful in episodes with word stress,

159

74.1% was successful in episodes with rising intonation, 63.6% was successful in
episodes with eliciting stop, and 62.1% was successful in episodes with dragging voice.
Adjusted standardized residuals in Table 56 reveal that episodes with rising intonation led
to signiﬁcantly more successful uptake while episodes with word stress led to
signiﬁcantly more acknowledge. There is no signiﬁcant difference among the four in
terms of inconclusive uptake. Chi—square tests indicate that while the occurrence of
uptake is not signiﬁcantly related to the type of paralinguistic cues, the successfulness of
uptake is (p value is .619 for the occurrence of uptake and .050 for the successfulness of

uptake).

Table 54 Type of paralinguistic cues and the occurrence of uptake

 

 

 

Number of Occurrence of uptake
episodes Uptake No uptake No opportunity
Word stress 168 119 (70.8%) 37 (22.0%) 12 (7.1%)
Rising intonation 35 28 (80.0%) 4 (11.4%) 3 (8.6%)
Eliciting Stop 46 34 (73.9%) 11 (23.9%) 1 (2.2%)
Dragging voice 38 29 (76.3%) 6 (15.8%) 3 (7.9%)

 

X2(6, n=287)=4.429, p=.619 (Mimicking sound and combination were excluded due to
low expected cell counts)

Table 55 Type of paralinguistic cues and the successfulness of uptake

 

 

 

Number of Successfulness of uptake
uptake Successful Acknowledge Inconclusive
Word stress 118 53 (44.9%) 45 (38.1%) 20 (16.9%)
Rising intonation 27 20 (74.1%) 4 (14.8%) 3 (11.1%)
Eliciting Stop 33 21 (63.6%) 8 (24.2%) 4 (12.1%)
Dragging voice 29 18 (62.1%) 5 (17.2%) 6 (20.7%)

 

X2(6, n=207)=12.536, p=.050 (Mimicking sound, combination, partially successful
uptake, and unsuccessful uptake were excluded due to low expected cell counts)

160

Table 56 Type of paralinguistic cues and the successfulness of uptake residuals

 

 

Successful Acknowledge Inconclusive
Word stress -3.1 3.0 .5
Rising intonation 2.2 -1.8 -.7
Eliciting Stop 1.2 -.8 -.7
Dragging voice .9 -1.6 .8

 

General extralinguistic cues

Tables 57 and 58 illustrate the occurrence and successﬁrlness of uptake according to the
presence of general extralinguistic cues. Table 57 shows that learners responded to
teacher feedback in 65.9% of episodes with extralinguistic cues and 64.7% of episodes
without extralinguistic cues. Adjusted standardized residuals in the table reveal that no
opportunity was signiﬁcantly more frequent in episodes without extralinguistic cues than
those with extralinguistic cues but there is no signiﬁcant difference between the two in
terms of uptake and no uptake. Table 58 shows that when uptake did occur, 38.4% was
successful in episodes with extralinguistic cues and 50.0% was successful in episodes
without extralinguistic cues. Adjusted standardized residuals in the table reveal that
episodes without extralinguistic cues led to signiﬁcantly more successful uptake while
episodes with extralinguistic cues led to signiﬁcantly more acknowledge. There is no
signiﬁcant difference between the two in terms of unsuccessﬁrl and inconclusive uptake.
Chi-square tests indicate that both the occurrence and successfulness of uptake are
signiﬁcantly related to teachers’ use of general extralinguistic cues (p value is .047 for

uptake and .005 for the successfulness of uptake)

161

Table 57 General extralinguistic cues and the occurrence of uptake

 

 

Number of Occurrence of uptake
CPiSOdCS Uptake No uptake No opportunity
Present 584 385 (65.9%) 135 (23.1%) 64 (11.0%)
Residuals (present) -- .5 1.4 -2.3
Absent 850 550 (64.7%) 171 (20.1%) 129 (15.2%)

 

x2(2, n=1434)=6.113,p=.047

Table 58 General extralinguistic cues and the successfulness of uptake

 

 

Number of Successfulness of uptake
uptake Successful Unsuccessful Acknowledge Inconclusive
Present 385 148 (38.4%) 7 (0.8%) 166 (43.1%) 64 (16.6%)
Residuals
(Present) -— -3.5 .0 2.9 .9
Absent 550 275 (50.0%) 10 (1 .8%) 186 (33.8%) 79 (14.4%)

 

x2(3, n=935)=12.645, p=.005

Type of extralinguistic cues

Tables 59, 60, and 61 illustrate the occurrence and successfulness of uptake according to
the type of extralinguistic cues. Table 59 shows that learners responded to teacher
feedback in 61.6% of episodes with gestures, 72.4% of episodes with whole body acting,
83.0% of episodes with head movements, 66.7% of episodes with facial expressions, and
64.0% of episodes with the combination of different types of extralinguistic cues.
Adjusted standardized residuals in Table 60 reveal that uptake was signiﬁcantly more
frequent in episodes with head movements, no uptake was signiﬁcantly more ﬁequent in
episodes with the combination of different types of extralinguistic cues, and no
opportunity was signiﬁcantly more frequent in episodes with gestures. Table 61 shows
that when uptake did occur, 33.3% was successful in episodes with gestures, 47.7% was
successﬁrl in episodes with whole body acting, 51.4% was successful in episodes with

head movements, and 37.1% was successful in episodes with the combination of different

162

types of extralinguistic cues. Chi-square tests indicate that while the occurrence of uptake
is signiﬁcantly related to the type of extralinguistic cues, the successfulness of uptake is
not (p value is .000 for the occurrence of uptake and .171 for the successfulness of

uptake).

Table 59 Type of extralinguistic cues and the occurrence of uptake

 

 

 

Number of Occurrence of uptake
episodes Uptake No uptake No opportunity
Gestures 341 210 (61.6%) 77 (22.6%) 54 (15.8%)
Whole body acting 29 21 (72.4%) 8 (27.6%) 0 (0%)
Head movements 88 73 (83.0%) 12 (13.6%) 3 (3.4%)
Facial expressions 12 8 (66.7%) 3 (25.0%) 1 (8.3%)
Combination 114 73 (64.0%) 35 (30.7%) 6 (5.3%)

 

X2(8, n=584)=30.270, p=.000

Table 60 Type of extralinguistic cues and the occurrence of uptake residuals

 

 

Uptake No uptake No opportunity
Gestures -2.6 -.4 4.5
Whole body acting .8 .6 -l .9
Head movements 3.7 -2.3 -2.5
Facial expressions .1 .2 -.3
Combination -.5 2.1 -2.2

 

Table 61 Type of extralinguistic cues and the successfulness of uptake

 

 

 

Numbgr of Successfulness of uptake
uptake Successful Acknowledge Inconclusive
Gestures 207 69 (33.3%) 97 (46.9%) 41 (19.8%)
Whole body acting 21 10 (47.7%) 9 (42.9%) 2 (9.5%)
Head movements 72 37 (51.4%) 25 (34.7%) 10 (13.9%)
Combination 70 26 (37.1%) 33 (47.1%) 11 (15.7%)

 

X2(6, n=370)=9.049, p=.171 (Facial expressions, partially successful uptake, and
unsuccessful uptake were excluded due to low expected cell counts)

163

Type of gestures

Tables 62 and 63 illustrate the occurrence and successfulness of uptake according to the
type of gestures. .Table 62 shows that learners responded to teacher feedback in 59.0% of
episodes with iconic gestures, 68.4% of episodes with metaphoric gestures, 64.6% of
episodes with deictic gestures, 66.7% of episodes with beats, and 53.7% of episodes with
the combination of different types of gestures. Table 63 shows that when uptake did
occur, 28.4% was successful in episodes with iconic gestures, 34.6% was successful in
episodes with metaphoric gestures, 28.6% was successful in episodes with deictic
gestures, 50.0% was successful in episodes with beats, and 52.4% was successful in

episodes with the combination of different types of gestures. Chi-square tests indicate

Table 62 Type of gestures and the occurrence of uptake

 

 

 

Number of Occurrence of uptake
episodes Uptake No uptake No opportunity
Iconics 139 82 (59.0%) 34 (24.5%) 23 (16.5%)
Metaphorics 38 26 (68.4%) 5 (13.2%) 7 (18.4%)
Deictics 99 64 (64.6%) 21 (21.2%) 14 (14.1%)
Beats 24 16 (66.7%) 6 (25.0%) 2 (8.3%)
Combination 41 22 (53.7%) 11 (26.8%) 8 (19.5%)

 

X2(8, n=341)=4.881, p=.770

Table 63 Type of gestures and the successfulness of uptake

 

 

 

Number of Successfulness of uptake
uptake Successful Acknowledge Inconclusive
Iconics 81 23 (28.4%) 42 (51.9%) 16 (19.8%)
Metaphorics 26 9 (34.6%) 12 (46.2%) ' 5 (19.2%)
Deictics 63 18 (28.6%) 28 (44.4%) 17 (27.0%)
Beats 16 8 (50.0%) 7 (43.8%) 1 (6.3%)
Combination 21 11 (52.4%) 8 (38.1%) 2 (9.5%)

 

X2(8, n=207)=9.791, p=.280 (Partially successful uptake and unsuccessful uptake were
excluded due to low expected cell counts)

164

that neither the occurrence nor the successfulness of uptake is signiﬁcantly related to the
type of gestures (p value is .770 for the occurrence of uptake and .280 for the

successﬁilness of uptake).

To sum up, the descriptive and inferential analysis in this section reveals that among the
ﬁve factors examined, three of them showed a statistically signiﬁcant effect on the
occurrence of uptake: general paralinguistic cues, general extralinguistic cues, and the
type of extralinguistic cues (p<.05). Adjusted standardized residuals reveal that learners
responded to teacher feedback signiﬁcantly more often in episodes where teachers used
general paralinguistic cues and where they used head movements (ASR>2.0). Regarding
the successfulness of uptake, among the ﬁve factors examined, general paralinguistic
cues, general extralinguistic cues, and the type of paralinguistic cues exerted a
statistically signiﬁcant effect on it. Adjusted standardized residuals indicate that learners
produced signiﬁcantly more successful uptake in episodes where teachers used general

paralinguistic cues and where they used rising intonation.

5.4 Nonlinguistic cues and recall comments

This section reports the relationship between the nonlinguistic cues in teacher feedback
and leamers’ comments in the stimulated recall interviews. The frequency of comments
related to teacher feedback is presented in relation to each type of nonlinguistic cues and
their subcategories in the order of general paralinguistic cues, type of paralinguistic cues,

general extralinguistic cues, type of extralinguistic cues, and type of gestures.

165

General paralinguistic cues

Table 64 shows the rate of teacher-feedback-related comments in relation to the presence
of general paralinguistic cues. A total of 77.1% of recall comments are related to teacher
feedback for episodes with paralinguistic cues and 70.3% are related to teacher feedback
for episodes without paralinguistic cues. Chi-square test results suggest that the rate of
teacher-feedback-related comments was not signiﬁcantly affected by the presence of

general paralinguistic cues (p=.493).

Table 64 General paralinguistic cues and teacher-feedback-related comments
Total number of Teacher-feedback- Non-teacher-feedback—

 

comments related related Other
Present 70 54 (77.1%) 7 (10.0%) 9 (12.9%)
Absent 209 147 (70.3%) 23 (11.0%) 39 ( 18.6%)

 

x2(2, n=279)=1.413, p=.493

Type of paralinguistic cues

Table 65 shows the rate of teacher-feedback-related comments in relation to the type of
paralinguistic cues. A total of 88.5% of recall comments are related to teacher feedback
for episodes with word stress, 88.9% are related to teacher feedback for episodes with
rising intonation, 92.9% are related to teacher feedback for episodes with eliciting stop,
85.7% are related to teacher feedback for episodes with dragging voice, 75.0% are related
to teacher feedback for episodes with mimicking sound, and 100% are related to teacher
feedback for episodes with the combination of different types of paralinguistic cues.
There is no case for the categories of “inconclusive” and “other”. Chi-square analysis was

not performed due to low expected cell counts.

166

Table 65 Type of paralinguistic cues and teacher-feedback-related comments

 

 

Total number of Teacher-feedback- Non-teacher-

comments related feedback-related
Word stress 26 23 (88.5%) 3 (11.5%)
Rising intonation 9 8 (88.9%) 1 (11.1%)
Eliciting Stop 14 13 (92.9%) 1 (7.1%)
Dragging voice 7 6 (85 .7%) 1 (14.3%)
Mimicking sound 4 3 (75.0%) 1 (25.0%)
Combination 1 l (100%) O (0%)

 

General extralinguistic cues

Table 66 shows the rate of teacher-feedback-related comments in relation to the presence
of general extralinguistic cues. A total of 77.0% of recall comments are related to teacher
feedback for episodes with extralinguistic cues and 62.5%% are related to teacher
feedback for episodes without extralinguistic cues. Adjusted standardized residuals
indicate that learners talked signiﬁcantly more about teacher feedback after viewing
episodes with extralinguistic cues. Chi-square test results suggest that the rate of teacher-
feedback-related comments was signiﬁcantly affected by the presence of general

extralinguistic cues (p=.011).

Table 66 General extralinguistic cues and teacher-feedback-related comments

 

 

Total number of Teacher- Non-teacher- Other
comments feedback-related feedback-related
Present 183 141 (77.0%) 13 (7.1%) 29 (15.8%)
Residuals (Present) -- 2.6 -2.7 -.8
Absent 96 60 (62.5%) i 17 (17.7%) 19 (19.8%)

 

x’(2, n=279)=9.005, p=.Oll

167

Type of extralinguistic cues

Table 67 shows the rate of teacher-feedback-related comments in relation to the type of
extralinguistic cues. A total of 76.4% of recall comments are related to teacher feedback
for episodes with gestures, 60.0% are related to teacher feedback for episodes with whole
body acting, 80.0% are related to teacher feedback for episodes with head movements,
50.0% are related to teacher feedback for episodes with facial expressions, and 80.0% are
related to teacher feedback for episodes with the combination of different types of
extralinguistic cues. Chi-square analysis was not performed due to low expected cell

COUJTIS.

Table 67 Type of extralinguistic cues and teacher-feedback-related comments
Teacher- Non-teacher-

 

 

Total number feedback- feedback- Inconclusive Other
of comments
related related

Gestures 106 81 (76.4%) 8 (7.5%) 2 (1.9%) 15 (14.2%)
:33: ”Ody 5 3 (60.0%) 0 (0%) 0 (0%) 2 (40.0%)
Head movements 20 16 (80.0%) 0 (0%) O (0%) 4 (20%)
Facial o 0 o o
expressions 2 1 (50.0 /o) 0 (0 /o) 0 (0 /o) 1 (50.0 /o)
Combination 50 40 (80.0%) 5 (10.0%) 1 (2.0%) 4 (8.0%)
Type of gestures

Table 68 shows the rate of teacher-feedback-related comments in relation to the type of
gestures. A total of 81 .6% of recall comments are related to teacher feedback for episodes
with iconic gestures, 62.5% are related to teacher feedback for episodes with metaphoric
gestures, 65.4% are related to teacher feedback for episodes with deictic gestures, 80.0%

are related to teacher feedback for episodes with beats, and 93.8% are related to teacher

168

feedback for episodes with the combination of different types of gestures. Chi-square

analysis was not performed due to low expected cell counts.

Table 68 Type of gestures and teacher-feedback-related comments

 

 

Total number of Teacher-feedback- Non-teacher- Other
comments related feedback-related
Iconics 49 40 (81.6%) 3 (6.1%) 6 (12.2%)
Metaphorics 8 5 (62.5%) 0 (0%) 3 (37.5%)
Deictics 26 17 (65.4%) 4 (15.4%) 5 (19.2%)
Beats 5 4 (80.0%) 0 (0%) 1 (20.0%)
Combination 16 15 (93.8%) 1 (6.2%) 0 (0%)

 

To sum up, in terms of nonlinguistic cues and stimulated recall comments, chi-square
analysis was only performed for the presence of general paralinguistic cues and general
extralinguistic cues. Between these two, only extralinguistic cues had a signiﬁcant effect
on the rate of teacher-feedback-related comments. That is, learners talked signiﬁcantly
more about teacher feedback after viewing episodes with extralinguistic cues. It was not
clear whether the type of paralinguistic cues, the type of extralinguistic cues, and the type
of gestures had signiﬁcantly affected the rate of teacher-feedback-related comments

because low expected cell counts made it impossible to perform valid chi-square tests.

5.5 Nonlinguistic cues and test results

This section reports the relationship between the nonlinguistic cues in teacher feedback
and test results. Results from both the immediate test and delayed test are presented
together according to each type of nonlinguistic cues and their subcategories in the order
of general paralinguistic cues, type of paralinguistic cues, general extralinguistic cues,

type of extralinguistic cues, and type of gestures.

169

General paralinguistic cues

Table 69 is an illustration of test results according to the presence of general
paralinguistic cues. In the immediate test, the percentage of correct answers is 68.9% for
episodes with paralinguistic cues and 63.1% for episodes without paralinguistic cues. In
the delayed test, the percentage of correct answers for the two is 56.3% and 52.0%
respectively. Chi-square analysis indicates that the presence of general paralinguistic cues

did not signiﬁcantly affect the results of either test (p value is .374 and .517 respectively)

Table 69 General paralinguistic cues and test results

 

 

 

 

Immediate test Delayed test
Number Of Correct Incorrect Number Of Correct Incorrect
ICSI rtems ICSI rtems
51 23 45 35
Present 74 (68.9%) (31.1%) 80 (56.3%) (43.8%)
125 73 105 97
Absent 198 (63.1%) (36.9%) 202 (52.0%) (48.1%)
x2(1, n=272)=.790, p=.374 x2(1, n=282)=420, p=.517

 

Type of paralinguistic cues

Table 70 is an illustration of test results according to the type of paralinguistic cues. In
the immediate test, the percentage of correct answers is 76.0% for episodes with word
stress, 66.7% for episodes with eliciting stop, 62.5 for episodes with dragging voice, and
87.5% for episodes with the combination of different types of paralinguistic cues. In the
delayed test, the percentage of correct answers for the four is 56.1%, 100%, 40.0%, and
83.3% respectively. The percentage is 28.6% for episodes with rising intonation and

66.7% for episodes with mimicking sound. Chi-square analysis indicates that the type of

170

paralinguistic cues did not signiﬁcantly affect the results of the immediate test (p=.588).

No chi-square test was performed for the delayed test due to low expected cell counts.

Table 70 Type of paralinguistic cues and test results

 

 

 

Immediate test Delayed test
Number Of Correct Incorrect Number Of Correct Incorrect
test rtems test rtems
19 6 32 25
W°rd Stress 25 (76.0%) (24.0%) 57 (56.1%) (43.9%)
. . . 20 10 2 0
thmg “01’ 30 (66.7%) (33.3%) 2 (100%) (0%)
Dragging 5 3 5 2 3
voice (62.5%) (37.5%) (40.0%) (60.0%)
Combination 7 l 6 5 1
(87.5%) (12.5%) (83.3%) (16.7%)
Rising Excluded due to low 7 2 5
intonation expected cell counts (28.6%) (71.4%)
Mimicking Excluded due to low 3 2 1
sound eigiected cell counts (66.7%) (33.3%)

 

x’(3, n=71)=1.925,p=.588

 

General extralinguistic cues

Table 71 is an illustration of test results according to the presence of general

extralinguistic cues. In the immediate test, the percentage of correct answers is 73.0% for

episodes with extralinguistic cues and 55.7 % for episodes without extralinguistic cues. In

the delayed test, the percentage of correct answers for the two is 60.0% and 47.1%

respectively. Adjusted standardized residuals suggest that learners provided signiﬁcantly

more correct answers for test items developed from episodes with extralinguistic cues in

both the immediate test and delayed test. Chi-square analysis indicates that the presence

of general extralinguistic cues signiﬁcantly affected the results of both tests (p value

is .003 and .023 respectively).

171

Table 71 General extralinguistic cues and test results

 

 

 

 

 

Immediate test Delayed test
N‘m‘ber Residuals In- Number Residuals In-
of test Correct of test Correct
. (Correct) correct . (Correct) correct
rtems rtems
103 38 77 50
Present 141 (73.0%) 3.0 (27.0%) 127 (60.6%) 2.3 (39.4%)
73 58 73 82
Absent 131 (55.7%) '3'0 (44.3%) 155 (47.1%) '2'3 (52.9%)
X2( 1, n=272)=8.925, p=.003 X2(1, n=282)=5.135, p=.023

 

Type of extralinguistic cues

Table 72 is an illustration of test results according to the type of extralinguistic cues. In
the immediate test, the percentage of correct answers is 73.6% for episodes with gestures,
63.7% for episodes with head movements, and 71.1% for episodes with the combination
of different types of extralinguistic cues. In the delayed test, the percentage of correct
answers for the three is 54.3%, 52.4%, and 75.8% respectively. Chi-square analysis
indicates that the type of extralinguistic cues did not signiﬁcantly affect the results of

either test (p value is .777 and .088 respectively).

Table 72 Type of extralinguistic cues and test results

 

 

 

 

Immediate test Delayed test
Number Number
of test Correct Incorrect of test Correct Incorrect
items items

64 23 38 32
Gesm’es 87 (73.6%) (26.4%) 70 (54.3%) (45.7%)

Head 11 7 4 21 11 10
movements (63.7%) (36.3%) (52.4%) (47.6%)

Combination 38 27 1 1 33 25 ' 8
(71.1%) (28.9) (75.8%) (24.2%)

X2(2, n=l36)=.504, p=.777 X2(2, n=124)=4.856, p=.088

 

(Whole body acting and facial expressions were excluded from both tests due to low
expected cell counts)

172

Type of gesture

Table 73 is an illustration of test results according to the type of gestures. In the
immediate test, the percentage of correct answers is 65.7% for episodes with iconic
gestures, 90.0% for episodes with metaphoric gestures, 84.0% for episodes with deictic
gestures, and 40.0% for episodes with beats. Adjusted standardized residuals suggest that
learners provided more (although not signiﬁcantly) correct answers for items developed
ﬁ'om metaphoric and deictic gestures and signiﬁcantly less correct answers for test items
developed from episodes with beats. In the delayed test, the percentages of correct
answers for iconic, metaphoric, and deictic gestures are 52.0%, 66.7%, and 55.0%
respectively. Chi-square analysis indicates that the type of gestures signiﬁcantly affected

the results of the immediate test (p=.029) but not the results of the delayed test (p=.748).

 

 

 

 

Table 73 Type of gestures and test results
Immediate test Delayed test
Number Residuals Number

of test Correct Incorrect of test Correct Incorrect

. (Correct) .

rtems rtems

. 23 l 2 1 3 12
km” 35 (65.7%) '1 '0 (34.3%) 25 (52.0%) (48.0%)
. 9 1 6 3
Metaphmcs 10 (90.0%) 1'4 (10.0%) 9 (66.7%) (33.3%)
. . 21 4 11 9
De’Ct’CS 25 (84.0%) 1'7 (16.0%) 20 (55.0%) (45.0%)
4 6 Excluded due to low expected

Beats 10 (40.0%) '2'3 (60.0%) cell counts

 

x’(3, n=80)=8.991,p=.029 x2(2, n=54)=.581, p=.748

 

(Combination was excluded from both tests due to low expected cell counts)

To sum up, the descriptive and inferential analysis in this section reveals that general

extralinguistic cues had a statistically signiﬁcant effect on the results of both the

173

immediate test and delayed test (p<.05), resulting in signiﬁcantly more correct answers
when they were used in teacher feedback (ASR>2.0). The type of gestures also
signiﬁcantly affected the immediate test, with deictic gestures associated with more
correct answers although the value of the adjusted standardized residual is not up to 2.0.
On the other hand, no such pattern was found in the delayed test. The type of
paralinguistic cues did not have any statistically signiﬁcant effect on the immediate test
and it is not clear whether it had an effect on the delayed test now that no chi-square
analysis was performed due to low expected cell counts. The other two factors, general
paralinguistic cues and the type of extralinguistic cues, did not show any statistically

signiﬁcant effect on either test.

5.6 Review of results

The results presented above show that the occurrence of uptake was signiﬁcantly affected
by the presence of general paralinguistic cues, the presence of general extralinguisticues,
and the type of extralinguistic cues but not by the type of paralinguistic cues and the type
of gestures. The successﬁllness of uptake was signiﬁcantly related to the presence of
general paralinguistic cues, the type of paralinguistic cues, and the presence of general
extralinguistic cues but not by the type of extralinguistic cues and the type of gestures. In
terms of the rate of teacher-feedback-related recall comments, while the presence of
general paralinguistic cues did not have any statistically signiﬁcant effect on it, the
presence of general extralinguistic cues did. It is not clear whether the type of
paralinguistic cues, the type of extralinguistic cues, and the type of gestures had any

statistically signiﬁcant bearing on the recall comments because no chi-square analysis

174

could be conducted with these three variables. As for the tests, results of the immediate
test were signiﬁcantly affected by the presence of general extralinguistic cues and the
type of gestures but not by the presence of general paralinguistic cues, the type of
paralinguistic cues, and the type of extralinguistic cues. The only variable that was shown
to have signiﬁcantly affected the results of the delayed test is again the presence of
general extralinguistic cues. It is not clear whether the type of paralinguistic cues had
such an effect on the delayed test because no chi-square analysis could be performed with
this variable. The table below is a summary of the overall results in this section. In
addition to the plus sign and minus sign, a question mark is used to indicate a statistically

unknown relationship.

Table 74 Overall results by nonlinguistic cues

 

 

Occurrence Successfulness Fiﬁzgk- Immediate Delayed
of uptake of uptake test test
comments

General

paralinguistic + + - - -
cues

Type of

paralinguistic - + ? - ?
cues

General

extralinguistic + + + + +
cues

Type of

extralinguistic + - ? - -
cues

Type of _ _ r) + ..
gestures

 

Figure 6 is an overview of nonlinguistic cues and type of nonlinguistic cues that predicted

uptake, successful uptake, teacher-feedback-related comments, and correct test results.

175

Speciﬁcally, episodes with general paralinguistic cues and head movements led to more
uptake. General paralinguistic cues, together with rising intonation, also contributed to
more successful uptake. By contrast, teacher-feedback-related comments and correct
answers for both the immediate test and delayed test were predicted by the presence of

general extralinguistic cues.

Figure 6 Overview of nonlinguistic cues predicting (successﬁll) uptake, teacher-
feedback-related comments, and correct test results

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Stimulated recall comments
Presence of Teacher-
Learner uptake general feedback-
extralinguistic '_* related
-Presence of cues comments
general
paralinguistic
cues " Uptake
-Head Language tests
movements
Presence of Correct
-Presence of general answers in
general extralinguistic —’ immediate
paralinguistic Successful cues test
—>
cues uptake
-Rising
intonation Presence of Correct
general answers in
extralinguistic _’ delayed
cues test

 

 

 

 

 

 

 

 

 

5.7 Discussion:
This section discusses the results presented in the previous sections. The occurrence and

successfulness of uptake are examined together with learners’ stimulated recall comments

176

and test results in relation to general paralinguistic cues, the type of paralinguistic cues,

general extalinguistic cues, the type of extralinguistic cues, and the type of gestures.

As shown in Chapter 2, very little research has been conducted in the ﬁeld of second
language acquisition to systematically examine the effect of nonlinguistic cues on
noticing and learning. The few studies that did make an effort to address this issue
(Carpenter et al., 2006; Davies, 2006; Faraco and Kida, 2008; Loewen & Philp, 2006;
Sheen, 2004; Sime, 2006) only focused on some aspects of nonlinguistic cues. Although
there is some counterevidence, the general conclusion ﬁ'om these studies is that
nonlinguistic cues have a positive effect on noticing and/or learning. Different from these
studies, the present study attempted to look at the effect of both general nonlinguistic
cues and speciﬁc types of nonlinguistic cues. It was found that both paralinguistic cues
and extralinguistic cues signiﬁcantly affected some of the dependent variables when
examined at the general level. This partially conﬁrms the ﬁndings from existing studies.
At the more speciﬁc level, there are no signiﬁcant differences among the different types
of nonlinguistic cues in most cases. Among the six types of paralinguistic cues, only
intonation had a strong positive effect on one of the ﬁve major dependent variables, the
successfulness of uptake, conﬁrming the ﬁndings of Loewen and Philp (2006) and Sheen
(2006). Among the ﬁve types of extralinguistic cues, only head movements had a strong
positive effect on another one of the ﬁve dependent variables, uptake. Gestures, a type of
extralinguistic cues which Faraco and Kida (2008) and Sime (2006) found to help clarify
verbally expressed meanings, did not stand out from the results. Below is a detailed

examination of nonlinguistic cues.

177

General paralinguistic cues refer to any voice feature that is part of language but are not
linguistic in nature. The word “general” is used because this term does not concern itself
with any speciﬁc kind of cues. It was found that both the occurrence and successfulness
of uptake were signiﬁcantly affected by paralinguistic cues (p value is .000 and .006
respectively). Learners produced signiﬁcantly more uptake and more successful uptake
when paralinguistic cues were present in teacher feedback (ASR value is 3.4 and 2.9
respectively). This result lends support to Lowen and Philp’s (2006) ﬁnding that prosodic
features such as emphasis and intonation can increase the effectiveness of recasts. It
means paralinguistic cues can help engage learners in teacher-student interactions and

push them to incorporate the correct form in their language production.

Although the presence of paralinguistic cues signiﬁcantly affected the occurrence and
successﬁllness of uptake, it did not show any statistically signiﬁcant effect on the rate of
teacher-feedback-related comments or the results of the two tests (p>.05 in all three
cases). One possible explanation is that the paralinguistic cues identiﬁed in the current
data set, along with those that were not examined (e.g., speech rate and voice quality),
averaged teachers’ speech so that they became less pronounced to learners. A second
possibility is that paralinguistic cues as subtle aspects of speech (Mehrebian, 1972) are
not as noticeable as verbal words themselves. On the other hand, compared with episodes
' without paralinguistic cues, learners did make a slightly higher percentage of teacher-
feedback-related comments after viewing episodes with paralinguistic cues (77.1% versus

70.3%) and provided a higher percentage of correct answers for test items developed

178

from episodes with paralinguistic cues in both tests (68.9% versus 63.1% in the
immediate test and 56.3% versus 52.0% in the delayed test). The slightly higher
percentages of teacher-feedback-related comments and correct answers, along with the
fact that there are signiﬁcantly more uptake and more successful uptake in episodes with
paralinguistic cues, indicate that paralinguistic cues have the potential to help arouse

learners’ attention to teacher feedback and improve learning.

Unlike general paralinguistic cues, the type of paralinguistic cues is concerned with
different kinds of acoustic nonlinguistic cues. Chi-square analysis indicates that there is
no statistically signiﬁcant difference among different types of paralinguistic cues in terms
of the occurrence of uptake (p=.619). This means that the effect of different types of
paralinguistic cues on the occurrence of uptake is similar to one another. In terms of the
successfulness of uptake, however, learners produced signiﬁcantly more successful
uptake in episodes with rising intonation (ASR=2.2) and signiﬁcantly less successful
uptake in episodes with word stress (ASR=-3.1). In 66.7% of episodes with rising
intonation, learners made an erroneous utterance and teachers repeated the problematic
language form in order to prompt learners to correct themselves (e. g., I would some
coffee?) Rising intonation was therefore often present in eliciting episodes, which has
been shown to lead to a high level of successful uptake. Given that, it is reasonable that
episodes with rising intonation led to more successful uptake. Word stress, although also
used in some eliciting episodes (generally together with rising intonation), were largely
(94.0%) employed when teachers were providing linguistic information to learners (e. g.,

You have class in 30 minutes. IN 30 minutes. Not after). As discussed in Chapter 4,

179

provide episodes tend to result in less successful uptake and more acknowledge because
learners were in a more passive position. The frequent presence of word stress in
information provision episodes therefore could have resulted in the lower level of

successful uptake and higher level of acknowledge (ASR-=30).

Another noticeable thing about the type of paralinguistic cues is eliciting stop. Like rising
intonation, eliciting stop is also a technique teachers often used to elicit language
information from learners. However, its level of successful uptake is lower than that of
rising intonation (ASR value is 1.2 against 2.2). It is not clear exactly why this is so, but
it could be due to the nature of the two types of techniques. With rising intonation,
teachers expected learners to correct themselves and learners were supposedly able to get
the right form with the proper amount of help because they at least had had some partial
knowledge of the language form already. With eliciting stop, however, learners were
expected to provide language forms or information they might not know at all. For
example, one teacher wanted to elicit the word “abandon” from learners, so he gave the
ﬁrst part of the word aban- and then stopped .for learners to give the whole word. It turned
out that learners were not able to provide the rest of the word and the teacher had to
provide it himself. Different ﬁ'om episodes with rising intonation where learners had
produced an utterance already, in this case learners did not seem to have any idea of the
target feature at all. Given the nature of eliciting stop, it is not surprising that it led to a

lower level of successful uptake than rising intonation.

180

With respect to the stimulated recall interviews, no chi-square analysis was conducted
due to low expected cell counts. It is therefore not clear whether there is a statistically
signiﬁcant relationship between the type of paralinguistic cues and teacher-feedback-
related comments. The percentages of such comments, however, indicate that there is no
big difference among the different types of paralinguistic cues. For the three types that
have been discussed in terms of the successfulness of uptake (word stress, rising
intonation, and eliciting stop), the percentage of teacher-feedback-related comments is
very close (88.5%, 88.9%, and 92.9% respectively). This reinforces what was discussed
previously in Chapter 4: learners’ responses to teacher feedback other than successful

uptake (e. g., acknowledge and inconclusive uptake) can also mean noticing.

Concerning test results, no chi-square analysis was performed for the delayed test due to
low expected cell counts. It is therefore not clear whether the different types of
paralinguistic cues signiﬁcantly affected the test results. Chi-square analysis of the
immediate test indicates that it did not. Descriptive statistics shows that the combination
of different kinds of paralinguistic cues has the highest percentage of correct answers in
the immediate test (87.5%). This result, as suspected, could be because more than one
type of paralinguistic cues together could create a bigger effect than one single type of
paralinguistic cues. This hypothesis is upheld by the still relatively high percentage of
correct answers resulted from episodes with the combination of different types of
paralinguistic cues in the delayed test (83.3%) although this percentage is lower than that ‘
of eliciting stop. Moreover, there are only two items for eliciting stop, both of which

learners answered correctly. It is possible that learners could still perform better with

181

combination items if there were more eliciting stop items. As for word stress, although it
led to signiﬁcantly less successful uptake than eliciting stop (ASR value is -3.1 against
1.2), it resulted in 76.0% of correct answers, 9.3% higher than eliciting stop (rising
intonation was excluded from the chi-square analysis due to low expected cell counts).
The higher percentage of correct answers for word stress could be because learners had
paid attention to teacher feedback when they simply acknowledged the feedback or

responded to it inconclusively in episodes with word stress.

General extralinguistic cues refer to any body movement that accompanies verbal
feedback. The word “general” is also used to mean that the term does not concern itself
with any speciﬁc kind of bodily cues. Chi-square analysis indicates that there is a
statistically signiﬁcant relationship between extralinguistic cues and the occurrence of
uptake. Compared with episodes without extralinguistic cues, episodes with
extralinguistic cues led to slightly more uptake (65.9% versus 64.7%) and no uptake
(23.1% versus 20.1%) but a signiﬁcantly lower level of no opportunity (ASR=-2.3). To
put it differently, learners had signiﬁcantly more opportunities to respond to teachers in
episodes with extralinguistic cues, and yet they only produced slightly more uptake. On
more occasions they chose not to respond to teachers even when there was a chance
(ASR=1.4). With that said, it can be concluded that learners in the present study tended to
be less responsive to teacher feedback in episodes with extralinguisitc cues. This is totally
different from Davies’ (2006) ﬁnding that‘episodes with extralinguistic cues were likely
to result in more uptake than topic continuation (no opportunity). One reason for this

difference is that Davies classiﬁed learners’ responses to teacher feedback in a different

182

way. In Davies' study, learner responses were coded as “either uptake or topic
continuation, that is, when a learner notices an error and produces pushed output of an
acceptable form or when a learner fails to notice and continues with the topic” (p. 843).
His topic continuation refers to leamers’ continuation of a topic, which would most likely
be considered as no uptake in the present study. Moreover, Davies did not distinguish
between no uptake and no opportunity. It seemed that teachers’ topic continuation which
prevented learners from responding was not considered. In the current study, such
episodes were coded as no opportunity. These coding and operational differences
between the two studies could have inﬂuenced the calculation of different types of
learner uptake, thus leading to different conclusions. A second important reason is that
Davies included peer-repair as part of learner uptake while the current study only
included the response by learners who were involved in an episode. The exclusion of peer
repair in learner uptake could have signiﬁcantly reduced the amount of learner uptake in

the present study.

With respect to the successﬁllness of uptake, extralinguistic cues were shown to have a
statistically signiﬁcant effect on it (p=.005). Compared with episodes without
extalinguistic cues, episodes with extralinguistic cues led to signiﬁcant less successful
uptake (ASR=-3.5) but signiﬁcantly more acknowledge (ASR=2.9). It is difﬁcult to tell
why this is so. One possible explanation isthat more than half of extralinguistic cues
(51.9%) were used to explain word meanings in message-related episodes: Such episodes,
as have been repeatedly discussed, led to more acknowledge. Another possible

explanation is that teachers sometimes did not give ﬁlll verbal feedback when

183

extralinguistic cues clearly conveyed a meaning. The feedback is therefore partially
linguistic and partially extralinguistic. This could have made it more difﬁcult for learners
to incorporate the target feature in their language production. Learners hence may choose
to simply acknowledge it or write it down. For example,

(88) Biey: What is disperse?
T: Disperse, ok. What do you think, so, look at the sentence there. Who dispersed?
Police dispersed who?
Biey: The crowd.
T: The crowd. Ok, it could work either way. So what did the police want? They
want the crowd together or apart? Disperse. (Hand gestures “together” and “apart”
and then “disperse”)
Biey: Oh ok.

 

In this example, the teacher hand gestures the meaning of “disperse” without really
giving the learner a clear verbal deﬁnition of the word. The learner might have an idea
that it somewhat carries the meaning of the word “apart”, but it might be difﬁcult for her
to uptake the feedback by giving “disperse” a clear verbal deﬁnition herself or to use it in

an utterance. As a result, she simply acknowledges it.

The lower level of uptake and successful uptake of episodes with extralinguistic cues
might lead one to expect that these episodes would result in less noticing and learning.
Analysis of the stimulated recall comments and test results, however, shows that the
reverse is true. Adjusted standardized residuals indicate that learners talked signiﬁcantly
more about teacher feedback after viewing episodes with extralinguistic cues than after
viewing episodes without extralinguistic cues (ASR value is 2.6 versus -2.6). The
percentage of teacher-feedback-related comments of the former is 14.5% higher than that
of the latter. In both the immediate test and delayed test, learners did signiﬁcantly better

with test items developed from episodes with extralinguistic cues than those developed

184

from episodes without extralinguistic cues (ASR value is 3.0 versus -3.0 in the immediate
test and 2.3 versus -2.3 in the delayed test). The percentage of correct answers of the
former is 17.3% higher than that of the latter in the immediate test and 13.5% higher in
the delayed test. These results show that learners had paid more attention to teacher
feedback when extralinguistic cues were used and as a result more learning had happened.
Talking about the reasons why gestures, one type of extralinguistic cues, are relevant to
the study of second language acquisition, Gullberg (2006) argues that gestures constitute
input and thus can enhance learning. Hostetter and Alibali (2004, cited in Gullberg, 2006)
observe that gestures help capture attention, provide redundancy, or engage more senses
by grounding speech in the concrete, physical experience. These claims, although
speciﬁcally about gestures, can be extended to extralinguistic cues in general now that
other types of extralinguistic cues also constitute input and have similar properties. The
observations by Gullberg and Hostetter and Alibali therefore can also explain the results

of the recall interviews and the two tests in the present study.

Unlike general extralinguistic cues, the type of extralinguistic cues is concerned with
different kinds of bodily nonlinguistic cues. Chi-square analysis indicates that the type of
extralinguistic cues exerted a statistically signiﬁcant effect on the occurrence of uptake.
Adjusted standardized residuals reveal that among the ﬁve categories of extralinguistic
cues, head movements led to signiﬁcantly more uptake (ASR=3.7) while gestures led to
signiﬁcantly less uptake (ASR=-2.6). This seems to show that learners responded to
teachers more actively in episodes with head movements than in episodes with gestures.

A look at the category of no opportunity, however, shows that episodes with gestures

185

 

have the highest level of no opportunity while episodes with head movements have the
lowest level of no opportunity (ASR value is 4.5 against -2.5). This result points to the
possibility that the low level of uptake in episodes with gestures had been caused by the

lack of opportunity.

Another notable thing is that there is a signiﬁcantly higher level of no uptake in episodes
with the combination of different types of extralinguistic cues (ASR=2.1) than episodes
with any other single type of extralinguistic cues. It is not clear why this is so. One
possibility is that learners understood the feedback so well with the help of multiple types
of extralinguistic cues that they felt it unnecessary to respond to it. This may sound
irrational. Learners' recall comments, however, show that it is possible. For example,
after viewing an episode with multiple extralinguistic cues, a learner stated with a frown:
"I don't think nothing because I was bored. She explained it too much." The learner did
not clearly say what made the explanation “too much”, but the multiple extralinguistic

cues could be part of it.

In terms of the successfulness of uptake, it was not signiﬁcantly affected by the type of
extralinguistic cues. According to the percentages of successful uptake, head movements
resulted in the highest level of successful uptake (51.4%) while gestures led to the lowest
level of successful uptake (33.3%). The detailed transcription of the observation data
shows that head movements are either nodding or headshaking. Nodding was mostly used
when teachers were conﬁrming a learner’s correct production or language hypothesis

while headshaking was mostly used when teachers wanted to emphasize a crucial point in

186

their feedback. In either case, learners' responses to the feedback were often successful.
Below are two examples where teachers use head movements in their feedback.
(89) Lofa: Mrs. H.

T: Yes?

Lofa: Is my a pronoun?

T: Mm-hmm.

Lofa: Why a pronoun?

T: Mm, it is a pronoun. It’s a possessive pronoun. Yeah.

Lofa: Possessive pronoun.

T: Yeah, yeah, yeah. (Nods)
(90) Ne: (To T) So neither just for the future.

T: No no no, if the sentence is afﬁrmative, I am going to study English, you say

me too. I am also going to study English.

Ne: (Nods)

T: If I say I am NOT (Stresses “not” and shakes head) going to New York city...

Ne: Me neither.

T: Me neither. I am not going to New York.
In the ﬁrst episode, the teacher nods to conﬁrm the learner's correct response to her
feedback. In the second episode, the teacher shakes head to help stress a negation which
is crucial to the learner's understanding of the target feature. Upon that, the learner
provides the correct form before the teacher ﬁnishes her sentence. In both cases, learners'
responses to teacher feedback were coded as successful uptake. Gestures, on the other
hand, were used to explain word meanings 57.2% of the time. The feedback as discussed
above might be more difﬁcult for learners to uptake because gestures themselves partially
took the place of verbal explanation and no direct deﬁnitions of the words were provided.
This could explain why gestures resulted in a higher combined percentage of
acknowledge and inconclusive uptake (66.7%) than head movements (48.6%). Given the

situations where head movements and gestures were used, it is understandable that the

former resulted in more successful uptake.

187

With respect to the stimulated recall interviews, no statistical analysis was performed due
to low expected cell counts. What can be found from the available descriptive statistics is
that learners talked more about teacher feedback after viewing episodes with head
movements. Given that learners produced signiﬁcantly more (successful) uptake and
hence might be more responsive in episodes with head movements, this is hardly
surprising. One noticeable thing is that episodes with the combination of different types
of extralinguistic cues, which led to the highest level of no uptake and a low percentage
of successful uptake next to gestures, have a percentage of teacher-feedback-related
comments as high as that of episodes with head movements (80.0%). This means that
learners actually were paying as much attention to teacher feedback in these episodes as
in those with head movements. Another thing that stands out is that gestures, which have
both the lowest level of uptake and successﬁll uptake, also resulted in 76.4% of teacher-
feedback-related comments, only a few percent lower than head movements and
combination. This once again proves that uptake and successful uptake are not the only

indicators for noticing.

With regard to test results, the type of extralinguistic cues did not show any statistically
signiﬁcant effect on either test. Descriptive statistics reveal that learners did best (73.6%
of correct answers) with test items developed from episodes with gestures in the
immediate test. The percentage of correct answers is close for items developed ﬁ'om
episodes with the combination of different types of extralinguistic cues (71.1%). Items
from episodes with head movements, on the other hand, have the lowest percentage of

correct answers (63.7%). In the delayed test, gestures and combinations still have a

188

higher percentage of correct answers than head movements (54.3%, 75.8%, and 52.4%
respectively). The higher percentages of correct answers for gestures and combinations
correspond to the fact that both the two have a higher percentage of teacher-feedback-

related comments, proving that more noticing leads to more learning.

What is surprising is that head movements led to the highest level of uptake, successful
uptake, and teacher-feedback-related comments and yet resulted in the lowest percentage
of correct answers in both tests. One possible explanation for this result is that the
noticing aroused by head movements tends to only have a temporary effect. If so, it could
be because head movements, both head shaking and nodding, are among the most brief
and common extralinguistic cues and hence might not be as striking and impressive as

other more complicated but less used extralinguistic cues.

Another thing that deserves special attention is the difference between gestures and
combinations in terms of the change of correct answers in the two tests. While the correct
answers for test items developed from episodes with gestures dropped by 19.3%, the
correct answers for items developed from episodes with a combination of different types
of extralinguistic cues increased by 4.7%. This could be because combinations with
various extralinguistic cues aroused learners' attention to a deeper level and hence better
helped them retain their memory while gestures as a single type of extralinguistic cues
did not have an effect as long lasting as that of multiple extralinguistic cues. As for the

reason why there is an increase in correct answers for items developed ﬁ'om episodes

189

with multiple extralinguistic cues, it could be because such episodes have a delayed effect

on test results (for more discussion, see Gass, 1997; Mackey, 1999).

Type of gestures is a subcategory of extralinguistic cues which speciﬁcally refer to the
movement of the hand and arm. It was found that although the type of gestures
signiﬁcantly affected the results of the immediate test (p=.029), there is no statistically
signiﬁcant difference among the different types of gestures in terms of the occurrence of
uptake, the successfulness of uptake, or the delayed test (p>.-05 in all three cases). About
the stimulated recall comments, no chi-square analysis was conducted due to low

expected cell counts.

According to the descriptive statistics, two types of gestures distinguish themselves from
others: metaphorics and beats. Metaphorics led to the highest percentage of uptake
(68.4%) and yet a lower (but not the lowest) percentage of successful uptake (34.6%).
Metaphorics present “an image of the invisible -- an image of the abstraction” (McNeil,
1992, p.14). Even though they are pictorial, the contents are still abstract. Learners might
ﬁnd it more difﬁcult to understand the abstract contents and therefore were pushed to
respond to teachers, thus producing more uptake. On the other hand, exactly for the same
reason, learners might not be so sure about their own understanding of the contents and
therefore were reluctant to incorporate the target features in their language production,

thus leading to less successful uptake.

190

In terms of the stimulated recall comments, it was found that learners paid more attention
to teacher feedback when they felt they did not understand a language feature (e. g., “I
was listening to the teacher’s explanation because I didn’t know what to call the stick
either”; “I didn’t know economical means from money perspective, so I paid special
attention to it when the teacher explained it”). Now that metaphoric gestures were used to
explain abstract ideas or concepts that are more difﬁcult to understand, learners should
pay more attention to teacher feedback with this type of cues. Results from the stimulated
recall interviews, however, show that learners made the lowest percentage of comments
on teacher feedback and the highest percentage of comments about other issues after
viewing episodes with metaphoric gestures. It is difﬁcult to tell if this is mere chance or
for some unknown reasons due to the lack of inferential statistics. A possible explanation
is that as a result of a less thorough understanding of the abstract contents, learners also
found it difficult to talk about the feedback and therefore chose to talk about other issues
concerning themselves (e.g., what they were guessing, how they felt about the new
language feature, and whether they were checking a dictionary). The low percentage of
comments on teacher feedback in relation to metaphoric gestures therefore may not mean

learners did not pay much attention to the feedback.

Regarding the test results, learners provided the highest percentage of correct answers for
test items developed from episodes with metaphoric gestures in both the immediate test
and delayed test. If attention is a necessary condition for learning, this result proves that
learners not only paid attention to teacher feedback with metaphoric gestures, they also

paid more attention to it.

191

Another type of gestures that proved intriguing is beats. A glimpse at the results shows
that the percentage of uptake for episodes with beats (66.7%) is very close to the
percentage for episodes with metaphoric gestures. The percentage of successful uptake
for episodes with beats (50.0%), however, is much higher than the percentage for the
latter. With respect to the stimulated recall comments, episodes with beats led to 80% of
teacher-feedback-related comments, 17.5% higher than episodes with metaphoric
gestures. These results as a whole mean that learners might have paid more attention to
teacher feedback in episodes with beats. Learners’ performance with items developed
from such episodes therefore should also be better in the two tests. In the delayed test,
beats were excluded from the chi-square analysis due to low expected cell counts. Results
from the immediate test show that learners actually provided the lowest percentage of
correct answers for items developed from episodes with beats (40.0%), 50% lower than
those developed from episodes with metaphoric gestures. One possible explanation for
this surprising result comes from the nature of beats. McNeil (1992) notes that “Of all
gestures, beats are the most insigniﬁcant looking” (p. 15) despite its semiotic value. This
“insigniﬁcant look” might have impaired the ability of beats to help learners retain their
memory even though they might have helped learners notice teacher feedback at the

moment.
5.8 Summary

In summary, the results reported in this chapter suggest that the occurrence of uptake was

signiﬁcantly affected by general paralinguistic cues, general extralinguistic cues, and the

192

type of extralinguistic cues. The presence of general paralinguistic cues, the presence of
general extralinguistic cues (to a lesser degree), and the use of head movements led to
more uptake. The successfulness of uptake was signiﬁcantly affected by general
paralinguistic cues, the type of paralinguistic cues, and general extralinguistic cues.
Learners produced a higher level of successful uptake when general paralinguistic cues
were present, when rising intonation was used, and when general extralinguistic cues
were absent. Learners’ stimulated recall comments were found to be signiﬁcantly
affected by general extralinguistic cues. The presence of general extralinguistic cues
resulted in a higher rate of teacher-feedback-related comments. It is not clear whether
other variables have an effect on the comments due to lack of inferential information.
About testing, the results of the immediate test were signiﬁcantly affected by general
extralinguistic cues and the type of gestures. The presence of general extralinguistic cues
and the use of metaphoric gestures predicted more correct answers. The results of the
delayed test were only signiﬁcantly affected by general extralinguistic cues. Similar to
the immediate test, learners performed better with test items developed ﬁom episodes
with extralinguistic cues in teacher feedback. These results taken as whole show that the
answer to the second research question is: yes, the nonlinguistic cues in teacher feedback

affect learners’ noticing and learning to a certain degree, especially at the general level.

193

CHAPTER 6
METALANGUAGE, THE NOTICING AND THE EFFECT OF TEACHER

FEEDBACK

6.1 Introduction

This chapter addresses the third research question: Does the metalanguage in teacher
feedback affect learners’ noticing and learning? After presenting the frequency of general
metalanguage and the distribution of different types of metalanguage, I report and discuss
the relationship between metalanguage and learners’ responses to teachers’ feedback,
learners’ comments in the stimulated recall interviews, and results from both the
immediate test and delayed test. By metalanguage I mean the use of technical and non-

technical linguistic terms such as “adjective” and “spelling”.

6.2 Occurrence and distribution of metalanguage

This section brieﬂy reports the frequency of general metalanguage and the distribution of
different types of metalanguage. Table 75 shows that some sort of metalanguage, be it
technical or non-technical, was used in 71.3% of a total of 1434 feedback episodes. Table
76 shows that among a total of 1023 feedback episodes where metalanguage was used,
7.2% only carry technical terms; 60.2% only contain non-technical terms; and 32.6%

have both technical and non-technical terms.

Table 75 Occurrence of general metalanguage

 

 

Number Percentage
Present 1023 71.3%
Absent 41 1 28.7%
Total 1 434 100%

 

194

Table 76 Distribution of different types of metalanguage

 

 

Number Percentage
Tech-only 74 7.2%
Non-tech-only 61 6 60.2%
Tech+non-tech 333 32.6%
Total 1023 100%

 

Table 77 shows the distribution of 53 randomly selected terms according to their

properties and the number of teachers who used them in their feedback to students. A

quick look at the table reveals that terms that were used by more teachers are often non-

technical while terms that were used by fewer teachers are often technical. For example,

among the terms that were used by all eight teachers, only one is technical (noun); and

among the terms that were used only by one teacher, all are technical.

Table 77 Randomly selected teacher metalanguage by property and number of teachers

 

 

. N of
Terms Propertles Examples teachers
Call Non-tech T: When you do it very loudly, that’s called “slam”, ok? 8
Mean/meaning Non-tech T: No, scram means to leave fast. 8
Noun Tech T: So every adjective needs a noun, so American what? 8
. ' 9 ’
Say Non-tech 'ilt". Why does water borl. That s the way we want to say 8
. T: And the way it’s spelled now, it’s already lumped
Spell/spelling Non-tech together in one word. 8
Use Non-tech T: You CAN use there in a tag question. 8
Need Non-tech T: I don’t think you need to use the. 7
Same Non-tech T: Yeah, they kind of have the same meaning, but the 7
way that you used them are a little bit different.
Talk about Non-tech T: Sometimes we use this one to talk about religion. 7
Verb Tech T: Thank you for saying that ‘cause treat can be a verb. 7
Adjective Tech T: Adjective, necessary is an adjective. 7
. w ' 9 '
Describe Non-tech Lyn s fog rlght. You can descnbe today as a foggy 6
T: It sounds the same, but the very you’re thinking of
Sound Non-tech has an E...but this one here is a verb, meaning things can 6
be different.
Subject Tech T: Ok, your subject rs 1t. So this meamng of look 1S not 6

an action verb.

 

195

 

Table 77 (cont’d)

 

Way
Idea

Make
Part
Singular
Tense

Adverb

Go+preposition

Name

Present

Question
Writing
Answer
Letter
Object
Phrase
Syllable

Term
Base form

Intonation

Past tense

Vowel
Compound
noun

Count noun

Helping verb
Neutral

Noun phrase

Part of speech
Preﬁx
Pronoun
Simple present
Stress

Non-tech

Non-tech

Non-tech

Non-tech
Tech
Tech

Tech

Non-tech

Non-tech

Non-tech

Non-tech

Non-tech

Non-tech
Non-tech
Tech
Non-tech
Tech
Non-tech
Tech

Tech

Tech
Tech

Tech

Tech

Tech
Non-tech

Tech

Tech
Tech
Tech
Tech
Tech

T: The way it’s going to be used is, they’re going to say
by Jove.

T: Ok, the idea of hold back is to eh not to go forward,
all right?

T: T: How about this one? Can you make it into a
question?

T: 80, a sentence with two parts, we have clauses.

T: Inhabitants because people are plural or singular?

T: Like add —ed or. . .Change the tense.

T: Ok, again we have an adverb. Don’t forget the
adverb.

T: So it should go between pick and up because that
verb is a phrase.

T: T: To plus verb. Do you know the name?

T: But this one we can either use to talk about the
past. . .or to talk about the present because we can say I
am used to studying at night.

T: When you ask a question about the object, what do
you need?

T: My example is for speaking, but the grammar is for
writing also.

T: Possibly we usually use like a short answer.

T: No, S-U-E, three letters. Yeah.

T: T: We, we don’t say cut corners and then an object.
T: So do you know this phrase put down roots?

T: HealthiER right? It’s two syllables and ends in y.

T: No, it’s a legal term.

T: What’s before the base form?

T: But there is a question, so you got to do the question
intonation.
T: We need past tense, past tense. You said she gets.

T: Brevity. Totally different, different vowel sound,
T: No, househusband is just a compound noun.

T: It’s not a count noun, so you can’t say garbages. Just
garbage.

T: You’re missing your helping verb in the ﬁrst one.

T: This is more common, more natural.

T: So sometimes it’s a whole noun phrase, not just a
noun. _
T: Recently. Ok, what part of speech is recently?

T: What does pre- mean? Pre- is a preﬁx.

T: Ok, where is the pronoun of the sentence?

T: No no no. No, all people die. Simple present.
T: Mm-hmm. Same stress, economical.

kit (JIMMQQ

A

Nwwwmwwbhhbb

NNNNNNNNN

 

196

 

Table 77 (cont’d)

 

Sufﬁx Tech T: Informallty or formalrty. We put that stress rlght

before the sufﬁx.
Synonym Tech T: Yeah, kind of like complain. That’s a good synonym. 1
Past participle Tech T: That’s a past participle. 1
Past . , . . .
progressive Tech T. But let 3 just try past progressrve and srmple past. 1
Simple past Tech T: I don t know, but I like the way you said that, With 1
the srmple past.
Tag (question) Tech T: Sometimes, it doesn t have to have a negative in a 1
tag question.
Uncount Tech T: Less homework, uncount. 1

 

6.3 Metalanguage and learner uptake
This section reports the relationship between the metalanguage in teacher feedback and
learner uptake. The occurrence and successfulness of uptake are presented together

according to both general metalanguage and different types of metalanguage.

General metalanguage

Tables 78 and 79 illustrate the occurrence and successfulness of uptake according to the
presence of general metalanguage in teachers’ feedback. Table 78 shows that learners
responded to teacher feedback in 65.5% of episodes where teachers used metalanguage
and 64.5% of episodes where teachers did not use metalanguage. Adjusted standardized
residuals in the table reveal that while there is no signiﬁcant difference between the two
in terms of uptake and no uptake, no opportunity was signiﬁcantly more frequent in
episodes where teachers did not use metalanguage. Table 79 shows that when uptake did
occur, 58.1% was successful in episodes where metalanguage was absent from teacher
feedback and 40.1% was successful in episodes where metalanguage was present in

teacher feedback. Adjusted standardized residuals in the table reveal that episodes with

197

metalanguage in teacher feedback led to signiﬁcantly less successful uptake and
signiﬁcantly more inconclusive feedback. Chi-square tests indicate that both the
occurrence and successfulness of uptake are signiﬁcantly related to the presence of
general metalanguage (p value is .021 for the occurrence of uptake and .000 for the

successfulness of uptake).

Table 78 General metalanguage and the occurrence of uptake

 

 

 

Number of Occurrence of uptake
episodes Uptake . No uptake No opportunity
Present 123 670 (65.5%) 230 (22.5%) 123 (12.0%)
Residuals
(Present) -- .4 1 .7 -2.5
Absent 70 265 (64.5%) 76 (18.5%) 70 (17.0%)

 

X2(2,n=1434)=7.700, —.021
p—

Table 79 General metalanguage and the successfulness of uptake

 

 

 

 

Number of Successfulness of uptake
uptake Successful Unsuccessﬁll Acknowledge Inconclusive
Present 670 269 (40.1%) 13 (1.9%) 266 (39.7%) 122 (18.2%)
Residuals (Present) -- -4.8 .4 1.9 3.9
Absent 265 154 (58.1%) 4 (1.5%) 86 (32.5%) 21 (7.9%)

 

x2(3, n=935)=28.368, p=.000

Type of metalanguage

Tables 80, 81, and 82 illustrate the occurrence and successfulness of uptake according to
the type of metalanguage. Table 80 shows that learners responded to teacher feedback in
75.7% of episodes with technical terms only, 65.9% of episodes with non-technical terms
only, and 62.5% of episodes with technical plus non-technical terms. Adjusted
standardized residuals in Table 81 reveal that while uptake was more frequent (although

not signiﬁcantly) in episodes where teachers used technical terms only, no uptake was

198

signiﬁcantly more frequent in episodes where teachers used technical plus non-technical
terms and no opportunity was signiﬁcantly more frequent in episodes where teachers
used non-technical terms only. Table 82 shows that when uptake did occur, 44.6% was
successful in episodes where teachers used technical terms only, 40.6% was successful in
episodes where teachers used non-technical terms only, and 38.0% was successful in
episodes where teachers used both technical and non-technical terms. Chi-square tests I.

indicate that while the occurrence of uptake is signiﬁcantly related to the type of

 

metalanguage, the successfulness of uptake is not (p value is .027 for the occurrence of ~

uptake and .319 for the successfulness of uptake).

Table 80 Type of metalanguage and the occurrence of uptake

 

Occurrence of uptake

 

Number of episodes

 

Uptake No uptake No opportunity
Tech-only 74 56 (75.7%) 13 (17.6%) 5 (6.8%)
Non-tech—only 616 406 (65.9%) 126 (20.5%) 84 (13.6%)
Tech+non-tech 333 208 (62.5%) 91 (27.3%) 34 (10.2%)

 

x2(2, n=1023)=11.004,p=.027

Table 81 Type of metalanguage and the occurrence of uptake residuals

 

 

Uptake No uptake No opportunity
Tech-only 1.9 -1.1 -l .4
Non-tech-only .3 -l .9 2.0
Tech+non-tech -1 .4 2.6 -1 .2

 

Table 82 Type of metalanguage and the successfulness of uptake

 

 

 

Number of Successfulness of uptake
uptake Successful Unsuccessful Acknowledge Inconclusive
Tech-only 56 25 (44.6%) 0 (0.0%) 17 (30.4%) 14 (25.0%)
Non-tech-only 406 165 (40.6%) 6 (1.4%) 164 (40.4%) 71 (17.5%)
Tech+non-tech 208 79 (38.0%) 7 (3.4%) 85 (40.9%) 37 (17.8%)

 

X2(6, n=670)=7.020,p=.319

199

To sum up, the descriptive and inferential analysis in this section reveals that general
metalanguage exerted a statistically signiﬁcant effect on both the occurrence and
successfulness of uptake (p<.05). The presence of metalanguage in teacher feedback led
to slightly more learner uptake (ASR=.4) but signiﬁcant less successful uptake (ASR=-
4.8). The type of metalanguage also signiﬁcantly affected the occurrence of uptake but
did not have any statistically signiﬁcant effect on the successfulness of uptake. Adjusted
standardized residuals reveal that when the metalalanguage in teacher feedback was all
technical, learners produced more uptake. The value of the adjusted standardized residual

is not up to 2.0 but very close (ASR=1.9).

6.4 Metalanguage and recall comments

This section reports the relationship between the metalanguage in teacher feedback and
learners’ comments in the stimulated recall interviews. The frequency of comments
related to teacher feedback is presented in relation to both general metalanguage and

different types of metalanguage.

General metalanguage

Table 83 shows the rate of teacher-feedback-related comments in relation to the presence
of general metalanguage in teacher feedback. A total of 72.2% of recall comments are
related to teacher feedback for episodes where teachers used metalanguage and 70.6% are
related to teacher feedback for episodes where teachers did not use metalanguage. Chi-

square test results suggest that the rate of teacher-feedback-related comments was not

200

 

signiﬁcantly affected by the presence of general metalanguage in teacher feedback

(p=.700).

Table 83 General metalanguage and teacher-feedback-related comments

 

 

Total number of Teacher-feedback- Non-teacher— Other
comments related feedback-related
Present 245 177 (72.2%) 25 (10.2%) 43 (17.5%)
Absent 34 24 (70.6%) 5 (14.7%) 5 (14.7%)

 

x2(2, n=279)=.715, p=.700

Type of metalanguage

Table 84 shows the rate of teacher-feedback-related comments in relation to the type of
metalanguage. A total of 57.1% of recall comments are related to teacher feedback for
episodes where teachers only used technical terms, 74.8% are related to teacher feedback
for episodes where teachers only used non-technical terms, and 69.4% are related to
teacher feedback for episodes were teachers used both technical and non-technical terms.
Chi-square test results suggest that the rate of teacher-feedback-related comments was

not signiﬁcantly affected by the type of metalanguage (p=.303).

Table 84 Type of metalanguage and teacher-feedback-related comments

 

 

Total number of Teacher— Non-teacher- Other
comments feedback-related feedback-related
Tech-only 14 8 (57.1%) 1 (7.1%) 5 (35.7%)
Non-tech-only 159 1 19 (74.8%) 14 (8.8%) 26 (16.4%)
Tech+non-tech 72 50 (69.4%) 10 ( 13.9%) 12 (16.7%)

 

x2(4, n=245)=4.854, p=.303

To sum up, the descriptive and inferential analysis in this section reveals that neither

general metalanguage nor the type of metalanguage exerted a statistically signiﬁcant

201

 

effect on the rate of teacher-feedback-related comments (p value is .700 for general

metalanguage and .303 for the type of metalanguage).

6.5 Metalanguage and test results
This section reports the relationship between the metalanguage in teacher feedback and
test results. Results from both the irmnediate test and delayed test are presented together

according to general metalanguage and different types of metalanguage.

General metalanguage

Table 85 is an illustration of test results according to the presence of general
metalanguage in teacher feedback. In the immediate test, the percentage of correct
answers is 66.8% for episodes with metalanguage in teacher feedback and 56.4% for
episodes without metalanguage in teacher feedback. In the delayed test, the percentage of
the two is 54.8% and 47.7% respectively. Chi-square analysis indicates that the presence
of general metalanguage did not signiﬁcantly affect the results of either test (p value

is .147 for the immediate test and .311 for the delayed test).

 

 

 

Table 85 General metalanguage and test results
Immediate test Delayed test
N‘m’l’e’ 0f Correct Incorrect Number Of Correct Incorrect
ICSI rtems ICSI rtems
145 72 1 1 9 98
Present 217 (66.8%) (33.1%) 217 (54.8%) (45.1%)
31 24 31 34
Absent 55 (56.4%) (43.6%) 65 (47.7%) (52.4%)

 

x2(1, n=272)=2.101,p=.147

x2(1, n=282)=l .026, p=.3ll

 

202

Type of metalanguage

Table 86 is an illustration of test results according to the type of metalanguage in teacher
feedback. In the immediate test, the percentage of correct answers is about 67.0% for
episodes with all the three types of metalanguage. In the delayed test, the percentage of
correct answers is 66.7% for episodes with technical terms only, 44.9% for episodes with
non-technical terms only, and 69.2 for episodes with both technical and non-technical
terms. Adjusted standardized residuals suggest that learners provided signiﬁcantly more
correct answers for test items developed from episodes with both technical and non-
technical terms and signiﬁcantly less correct answers for items developed from non-
technical terms only. Chi-square analysis indicates that the type of metalanguage
signiﬁcantly affected the results of the delayed test (p=.002) but not the results of the

immediate test (p=.998).

Table 86 Type of metalanguage and test results

 

 

 

 

 

Immediate test Delayed test
Number Number Residuals
of test Correct Incorrect of test Correct Incorrect
. . (Correct)
rtems rtems
6 3 8 4
Tammy 9 (66.7%) (33.3%) 12 (66.7%) '8 (33.3%)
Non-tech- 90 45 57 70
only 135 (66.7%) (33.3%) 127 (44.9%) '3'5 (55.1%)
Tech+non- 73 49 24 78 54 3 2 24
tech (67.1%) (32.8%) (69.2%) ' (30.8%)
x2(2, n=217)=.005, p=.998 x2(2, n=217)=12.285, p=.002

 

To sum up, the descriptive and inferential analysis in this section reveals that general
metalanguage did not have any statistically signiﬁcant effect on the results of either test
(p>.05). Concerning the type of metalanguage, while it did not signiﬁcantly affect the

results of the immediate test, it did show a statistically signiﬁcant effect on the results of

203

the delayed test (p=.002), with episodes where both technical and non-technical terms

were used leading to signiﬁcantly more correct answers (ASR=3.2).

6.6 Review of results

The results presented above show that the occurrence of uptake was signiﬁcantly affected
by both the presence of general metalanguage and the type of metalanguage but the
successfulness of uptake was only signiﬁcantly affected by the former, not the latter. In
terms of the rate of teacher-feedback-related recall comments, neither the presence of
general metalanguage nor the type of metalanguage had a statistically signiﬁcant effect
on it. This is also the case with the results of the immediate test. The results of the
delayed test, on the other hand, were signiﬁcantly affected by the type of metalanguage
but not by the presence of general metalanguage. Despite the fact that both general
metalanguage and the type of metalanguage had a signiﬁcant effect on the occurrence of
uptake and the former also signiﬁcantly affected the successfulness of uptake, no
category predicted more uptake or more successful uptake. The correct answers in the
delayed test, however, were predicted by episodes with technical plus non-technical

terms. The table below is a summary of the overall results in this section.

Table 87 Overall results by metalanguage

 

 

Occurrence of Successﬁllness of Feedback- Immediate Delayed
related
uptake uptake test test
comments
General + + _ _ _
metalanguage
Type of + _ _ _ +

metalanguage

 

204

 

6.7 Discussion:
This section discusses the results presented in the previous sections. The occurrence and
successfulness of uptake are examined together with learners’ stimulated recall comments

and test results in relation to general metalanguage and the type of metalanguage.

As reviewed in Chapter 2, very little empirical research has been conducted to
speciﬁcally examine the effect of teachers’ metalanguage in second language classrooms.
Borg’s (1998, 1999) studies were mainly concerned with teachers’ beliefs about
metalanguage use. They therefore did not speak directly to the effect of metalanguage.
One group of studies which did touch upon metalanguage (e.g., Lyster & Ranta, 1997;
Panova & Lyster, 2002; Sheen, 2004) is those which investigated the effect of different
types of feedback. A general ﬁnding of these studies is that metalinguistic feedback, a
feedback type which contains metalanguage, is more successful than recasts in eliciting
learner uptake and repair. Results from the present study provide both evidence and
counterevidence for this ﬁnding. It was found that the presence of metalanguage in
teacher feedback did lead to more learner uptake but only slightly. On the other hand,
there was signiﬁcantly less successful uptake when teachers used metalanguage in their
feedback. This difference could have resulted from different teaching contexts, different
participants, and different research designs. Meanwhile, the lower level of uptake and
successful uptake in the present study somewhat echoes the ﬁndings by Basturkrnen et a1.
(2002). In that study, the researchers examined the effectiveness of metalanguage by
different types of focus-on-form episodes. The results they reported show that uptake and

successful uptake co-occurred with metalanguage mostly frequently in student-initiated

205

 

focus-on—form episodes, but the percentage was not very high (50.3% for uptake and
44.2% for successful uptake). Unlike the present study, Basturkrnen et a1. did not
investigate the relationship between the speciﬁc types of metalanguage and learner

uptake. This is discussed below in addition to general metalanguage.

General metalanguage refers to any language that is used to describe the language system.
The word “general” again is used to mean that the term does not concern itself with any
speciﬁc kind of metalinguistic terms. Chi-square analysis shows that the presence of
metalanguage had a signiﬁcant effect on both the occurrence and successﬁrlness of
uptake (p value is .021 and .000 respectively). In terms of the occurrence of uptake, the
percentage of uptake is virtually the same whether or not there was metalanguage in
teacher feedback (65.5% and 64.5% respectively). On the other hand, learners had
signiﬁcantly more opportunities to respond to teacher feedback when metalanguage was
present (the adjusted standardized residual of no opportunity is -2.5) and yet these
episodes resulted in a higher level of no uptake although it is not statistically signiﬁcant
(ASR=1.7). Taken together, these results indicate that learners were less responsive to
teacher feedback with the presence of metalanguage. This could be because some
metalanguage (e.g., grammar labels) had exerted a threatening effect on learners as the
teachers in Borg’s (1998, 1999) studies were afraid, and had prevented them ﬁom
responding to teacher feedback. It could also be because some metalanguage that learners
were not familiar with had added to their cognitive burdens as some‘researchers have
argued (e.g., Corder, 1973), and had also prevented learners from responding to the

feedback.

206

In terms of the successfulness of uptake, adjusted standardized residuals indicate that
learners produced signiﬁcantly less successful uptake when metalanguage was used in
teacher feedback (ASR=4.8). However, it led to signiﬁcantly more inconclusive uptake
(ASR=3.9) and a much higher level of acknowledge (ASR=1.9). This could also be
because metalanguage was threatening to learners and/or had added to their cognitive
burdens so that they decided to respond to the feedback by simply acknowledging it or

writing it down instead of verbally responding to it.

With respect to the stimulated recall comments, chi-square analysis indicates that the
presence of metalanguage did not signiﬁcantly affect the rate of teacher-feedback-related
comments (p=.700). On the other hand, the descriptive statistics reveal that the
percentage of teacher-feedback—related comments is slightly higher for episodes with
metalanguage in teacher feedback than those without metalanguage in teacher feedback
(72.2% versus 70.6% respectively). This means that learners could have paid more
attention to feedback with metalanguage although the difference is not signiﬁcant. It
could be because some metalanguage which is substantially different ﬁom everyday
language had helped to arouse learners' attention to the feedback with its markedness.
Another possibility is that some of the language features in episodes with metalanguage
were the focus of a lesson or unit (e. g., the grammar rules in grarmnar practice activities).

‘ They therefore received more attention from learners.

207

 

Regarding the test results, chi-square analysis indicates that the presence of metalanguage
did not signiﬁcantly affect the results of either the immediate test or the delayed test. The
percentages of correct answers, however, reveal that learners performed better with test
items developed from episodes with metalanguage in teachers' feedback than those
developed from episodes with no metalanguage in teachers’ feedback (66.8% versus
56.4% in the immediate test and 54.8% versus 47.7% in the delayed test). Given the role
of attention in learning, this could be because learners had indeed paid more attention to

teacher feedback with metalanguage as shown in teacher-feedback-related comments.

Unlike general metalanguage, the type of metalanguage is concerned with different kinds
of metalinguistic terms. Chi-square analysis indicates that there is a statistically
signiﬁcant relationship between the type of metalanguage and the occurrence of uptake
(p=.027). Speciﬁcally, episodes with only technical terms led to the highest level of
uptake as compared with episodes with non-technical terms only and episodes with both
technical and non-technical terms (75.7% versus 65.9% and 62.5%). This is somewhat
surprising given that technical terms should be more threatening to learners than other
types of metalanguage. A careful look at episodes with only technical terms shed some
light on this puzzle. Metalanguage in a majority of these episodes (78.4%) is terms that
occurred frequently across the nine classes (see Table 77), sometimes from both teachers
and learners, and hence might be familiar to learners. These include terms about part of
speech (e.g., noun, verb, adjective, adverb), number (e.g., plural, singular), simple verb

tense (e. g., present tense, past tense), etc. Below are a few examples:

208

 

(91) Xin: (Writes “garage sell”)
T: (Looks at cross puzzle) Sale, not sell. Sell is the verb but sale is the, the noun.
Sale, S-A-L-E.
Xin: S-A-L—E. (Writes on paper)
(92) Ye: Iwere...
T: I WAS, were is plural. (Stresses “was”)
Ye: Yeah, I was doing my homework.
(93) (Reading a story that happened in the past)
Cadiz You’re smiling.
T: Let’s keep it in past tense.
Cadiz (Pauses, 4 seconds) You were smiling. (Makes changes on book)
In the ﬁrst episode, the technical terms the teacher uses are “verb” and “noun”, which can
be found in the recordings of eight out of the nine classes. In the second episode, the

’9

teacher uses the term “plur , which the learner seems to understand by correcting
herself in her new utterance after the feedback. In the third episode, the teacher uses the
term “past tense”, which the learner also seems to understand by incorporating the correct
form in his new utterance. Some of the terms in these episodes have actually been used
by learners themselves (e.g., "So collapsing is still a verb?" "T, garage sale, plural or
single?" ). Now that most of the technical terms are simple ones familiar to learners,
learners might feel less threatened or burdened to respond to the feedback. In contrast to
episodes with technical terms only, many episodes with both technical and non-technical
terms cover complicated and infrequent terms too (e.g., “You CAN use ‘there’ in a tag

,9 ‘6

question, ...but if you look at your list of non-action verbs, the word ‘think’ is not
there. ..but in the extended present, you know...) These terms could have had an

intimidating effect on learners who may have been reluctant to respond to the teachers’

feedback.

209

 

About the successfulness of uptake, Chi-square analysis indicates that it was not
signiﬁcantly affect by the type of metalanguage (p=.319). Nonetheless, the descriptive
statistics show that episodes with only technical terms resulted in a higher percentage of
successful uptake (44.6%) than those with both technical and non-technical terms
(38.0%). This difference could also be explained by the above-mentioned nature of the
metalanguage in the two types of episodes. It is natural that learners would be more likely
to be able to repeat the feedback or incorporate it in their own production when they were

more familiar and felt more comfortable with the language in the feedback.

Although learners produced more (successful) uptake in episodes with technical terms
only, they did not seem to have paid more attention to teacher feedback in these episodes.
Analysis of the stimulated recall comments indicates that episodes with technical terms
only actually resulted in a lower percentage of teacher-feedback-related comments than
those with both technical and non-technical terms (57.1% versus 69.4%). This could be
because the less familiar or unfamiliar terms in episodes with both technical and non-
technical terms had helped arouse learners' attention since learners tended to pay more
attention to teacher feedback when the target feature was new to them (see the discussion
of Chapter 5). Another explanation is that teacher feedback in the majority of episodes
with technical terms only (77.0%) is simpler, generally with one or two teacher turns.

Attention to the target feature of these episodes was therefore brief.

The feedback in episodes with both technical and non-technical terms, by contrast, is

more complicated. Among a total of 333 episodes, 63.7% have more than two teacher

210

 

turns. Attention to the target feature of these episodes therefore could be longer and
maybe more intense. This also provides a good explanation for the reason why episodes
with both technical and non-technical terms resulted in a higher percentage of correct
answers than episodes with technical terms only in both the immediate test and delayed
test. In the immediate test, the difference between the two is very slight (67.1% versus
66.7%) and not statistically signiﬁcant (p=.998). In the delayed test, the percentage
difference is still not big (69.2% versus 66.7%). However, adjusted standardized
residuals show that episodes with both technical and non-technical terms resulted in
signiﬁcantly more correct answers than episodes with technical terms only (ASR value is

3.2 versus .8).

Up to now, the discussion about type of metalanguage has focused on the categories of
technical terms only and technical plus non-technical terms. The other category, non-
technical terms only, also provides some interesting results. In theory, feedback with non-
technical terms only should be the least confusing, the least threatening, and the least
burdening among the three types of metalanguage. Learners therefore should feel most
comfortable to respond to teachers in episodes with such terms. In reality, however,
episodes with only non-technical terms took the middle position in terms of both the
occurrence and successﬁllness of uptake, with less uptake and successﬁrl uptake than
episodes with technical terms only and more uptake and successful uptake than episodes
with both technical and non-technical terms. To ﬁnd out why there is a lower level of
uptake in episodes with non-technical terms only, it is necessary to look at the category of

no uptake and no opportunity. Adjusted standardized residuals reveal that among the

211

 

three types of metalanguage, non-technical terms only has a signiﬁcantly higher level of
no opportunity (ASR value is 2.0 against -1.4 for technical terms only and -l.2 for
technical plus non-technical terms). On the other hand, it has a lower level of no uptake
(ASR value is -1.9 against -1.1 for technical terms only and 2.6 for technical plus non-
technical terms). Although the absolute value of its adjusted standardized residual is not

up to the conventional 2.0, it is very close.

All this means that the lower level of uptake for non-technical terms only could have
been the result of fewer opportunities, not because learners chose not to respond to
teachers. There could be a variety of reasons why teachers gave learners fewer
opportunities when the metalanguage in the feedback was non-technical terms only. One
possibility is that teachers assumed that learners should have fewer problems with
feedback without technical terms and therefore did not want to spend as much time on
them. As for the reason why there is a lower level of successful uptake in episodes with
non-technical terms only, it could be that learners, believing that they understood the
feedback, thought they could afford not to use the target feature right after the feedback.
They therefore simply acknowledged it or responded to it in an inconclusive way. Again,

this could only be one of many possible explanations.

With regard to the stimulated recall comments, episodes with only non-technical terms in
teachers’ feedback resulted in the highest level of teacher-feedback-related comments
(74.8% against 57.1% for technical terms only and 69.4% for technical plus non-

technical terms). This means that learners had paid more attention to the feedback in

212

these episodes. Given that learners did not produce the highest level of uptake or
successful uptake in episodes with non-technical terms only and that they had frequently
responded to teacher feedback by acknowledging it or producing inconclusive uptake in
these episodes, this result again points to the possibility that other types of uptake (e. g.,
acknowledge, inconclusive uptake) can also be indicators of attention. It should be noted
that episodes with only non-technical terms in teacher feedback also resulted in some
unsuccessful uptake, but the percentage is very low (1.4%). The very fact that learners
tried to uptake the feedback is evidence that they were attending to it even though they

failed to understand or use it.

Concerning the results of the two tests, they also look intriguing when associated with
non-technical terms only. In the immediate test, learners' performance with test items
developed from episodes with only non-technical terms in the feedback is almost the
same with those developed from episodes with the other two categories of metalanguage.
In the delayed test, however, learners provided signiﬁcantly fewer correct answers for
items developed from episodes with only non-technical terms (ASR value is -3.5
against .8 for technical terms only and 3.2 for technical plus non-technical terms). The
percentage of correct answers for these items is more than 20% lower than its
counterparts. Given that episodes with only non-technical terms resulted in the highest
percentage of teacher-feedback-related comments, this is surprising. It is hard to pinpoint
‘ the exact reasons for this result, but a look at the nature of the attention learners paid to
teacher feedback might help. When learners only acknowledged teachers’ feedback or

responded to it inconclusively, no pushed output was produced even though they might

213

 

have attended to the feedback. The attention learners paid to the feedback therefore might
not be intense. As a result, this kind of uptake might not be as beneﬁcial to learning as
successful uptake, where learners tried to use the target feature in their language
production. Consequently, learners' memory of the target features in episodes with
acknowledge or inconclusive uptake decayed faster. If this were true, it is understandable
that episodes with non-technical terms only, which resulted in more acknowledge and

inconclusive uptake, would lead to less correct answers in the delayed test.

6.8 Summary

In summary, the results reported in this chapter suggest that the occurrence of uptake was
signiﬁcantly affected by two different factors: general metalanguage and the type of
metalanguage. Episodes with general metalanguage and episodes with only technical
terms resulted in a higher level of uptake. The successfulness of uptake was signiﬁcantly
affected by general metalanguage but not by the speciﬁc type of metalanguage. Learners
produced more successful uptake when metalanguage was not present in teachers’
feedback. Learners’ stimulated recall comments did not have any signiﬁcant relationship
with either general metalanguage or the speciﬁc type of metalanguage. This is also the
case with the results of the immediate test. The results of the delayed test, however, were
signiﬁcantly related to the type of metalanguage. Learners performed better with test
items developed from episodes with both technical and non-technical terms. These results
together show that the answer to the third research question is: yes, metalanguage in

teacher feedback has some effect on noticing and learning.

214

CHAPTER 7

CONCLUSION

7.1 Introduction
This ﬁnal chapter begins with a brief summary of the ﬁndings in previous chapters and a
discussion of the implications of these ﬁndings. It continues with a reﬂection on the

design of the study, and ﬁnally makes some tentative suggestions on future research.

7.2 Summary and implications
In this section, I brieﬂy summarize the general ﬁndings of the present study according to
the three research questions and discuss both the pedagogical implications and research

implications of these ﬁndings.

7.2.1 Summary
Three questions are raised in this study:
1. Do the characteristics of teacher feedback episodes affect learners’ noticing and
learning?
2. Do the nonlinguistic cues in teacher feedback affect learners’ noticing and
learning?
3. Does the metalanguage in teacher feedback affect learners’ noticing and learning?
In terms of the characteristics of feedback episodes, all seven characteristics examined
affected all or some of the dependent variables that measured noticing and learning (the

occurrence of uptake, the successfulness of uptake, learners’ stimulated recall comments,

215

 

and test results). This also holds true for the presence of general nonlinguistic cues and
some of their subcategories. As for teachers’ metalanguage and its subtypes, they were
found to have some effect on noticing and learning too although to a lesser degree and

not always in a positive way.

7.2.2 Implications

The ﬁndings presented above carry important implications for both classroom teaching
and feedback research. Although they may not apply to every teaching or research
context, these implications can be very helpful for teaching and research in a context

similar to that of the present study.

Implications for Teaching

As discussed in Chapters 4, 5, and 6, the ﬁndings from this study are not exactly the same
as the ﬁndings in previous studies. It is therefore dangerous to make generalizations.
However, some aspects of teacher feedback episodes have been repeatedly shown to be
more effective than others in eliciting (successful) uptake, promoting noticing, and

facilitating learning. Here I describe three ﬁndings that are reasonably plausible.

The characteristics of feedback episodes matter

The ﬁndings of this study align with previous studies to suggest that the characteristics of
feedback episodes play a role in the noticing and effectiveness of teacher feedback. One :
example is that elicit episodes are more effective than provide episodes in drawing

learner uptake and successﬁll uptake in both immersion programs (Lyster, 1998a; Lyster

216

and Ranta, 1997) and ESL classrooms (Ellis et al. 2001a; Loewen, 2004; and also the
present study). This means that learners might beneﬁt more from the feedback if teachers
try to prompt learners to get the correct language form instead of providing direct
information to them. Another example is that complex episodes are more facilitative than
simple episodes in increasing learners’ noticing of teacher feedback and the subsequent
learning. The ﬁndings in this study show that compared with simple episodes complex
episodes resulted in signiﬁcantly more uptake, successful uptake, teacher-feedback-
related comments, and correct answers in both the immediate test and delayed test. This
indicates that complex episodes are more effective in both drawing learners’ attention to
teacher feedback and their language improvement. Concerning characteristics whose
effect is more arguable, teachers can make decisions according to the speciﬁc teaching

context such as the teaching objective and student population.

Non-linguistic cues help

As for the use of nonlinguistic cues, there has been positive evidence from different
research areas, for example, applied linguistics, speech communication, and second
language acquisition too. Although nonlinguistic cues do not always help communication
(See Faraco & Kida, 2008), they in general are effective in increasing learners’ noticing
of feedback and subsequent language learning. In the present study, the presence of
general paralinguistic cues elicited signiﬁcantly more uptake and successful uptake and
the present of extralinguistic cues predicted signiﬁcantly more teacher-feedback-related
comments in the stimulated recall interviews and signiﬁcantly more correct answers in

both the immediate test and delayed test. These ﬁndings suggest that nonlinguistic cues

217

 

 

as a whole can promote learners’ noticing of teacher feedback and the effectiveness of the
feedback. Teachers therefore can incorporate nonlinguistic cues in their feedback to help

improve students’ learning.

Metalanguage is not always eﬂective

Finally, about metalanguage use, the results of the present study as well as the ﬁndings
by Basturkrnen et al.’ (2002) show that it can affect learners’ noticing of teacher feedback
and their language improvement in a certain way, but the overall effect is not very
impressive or even counteractive. In the present study, the adjusted standardized residuals
indicate that neither general metalanguage nor its subtypes predicted signiﬁcantly more
uptake, successful uptake, teacher-feedback-related comments, or correct answers in the
immediate test. Only episodes with both technical and non-technical terms predicted
signiﬁcantly more correct answers in the delayed test. Moreover, when teachers used
metalanguage in their feedback to learners, there was signiﬁcantly less successful uptake,
indicating that learners were signiﬁcantly less able to incorporate the correct language
forms in their new language production after receiving teacher feedback with
metalanguage. Teachers therefore need to take caution about what kind of metalanguage

to use and when to use it.

Implications for Research
Concerning the implications for research, I outline two issues below: uptake and teacher-
feedback-related comments. Both are discussed in terms of their validity and/or reliability

as a measure of noticing and learning.

218

 

Uptake needs to be supported by other measures

One thing that stood out throughout the previous three chapters is that the same category
of a variable (the characteristic of feedback episodes, nonlinguistic cues, or teachers’
metalanguage) sometimes had totally different effects on the different noticing and
learning measures. For example, while indirect episodes resulted in a signiﬁcantly higher
level of uptake and successﬁrl uptake, they led to a signiﬁcantly lower level of teacher-
feedback-related comments and correct answers in the immediate test. The rates of
uptake and successful uptake have been found to run counter to teacher-feedback-related
comments and test results. This raises a serious issue. Although uptake has been
reiterated by many researchers (Ellis et al., 2001a; Loewen, 2004; Sheen, 2006) to be an
optional move, it has been widely adopted in different feedback studies to examine the
effectiveness of feedback. The tables below show the results of chi-square analysis on the
rate of teacher-feedback-related comments and test results in relation to the occurrence
and successfulness of uptake. The p value in all cases are >.05, indicating that the
occurrence and successfulness of uptake did not predict the rate of teacher-feedback-
related comments or test results. This ﬁnding suggests that uptake and successful uptake
cannot always serve as valid measures for noticing and learning, they therefore should be

supported and supplemented by other measures.

Table 88 Uptake and teacher-feedback-related comments

 

 

Total number of Teacher- Non-teacher- Other
. comments feedback-related - feedback-related
Uptake 182 130 (71.4%) 20 (11.0%) 32 (17.6%)
No uptake 60 43 (71.7%) 7 (11.7%) 10 (16.7%)
No opportunity 37 28 (75.7%) 3 (8.1%) 6 (16.2%)

 

x2(4, n=279)=.421, p=.981

219

Table 89 Successfulness of uptake and teacher-feedback-related comments

 

Total number of Teacher-feedback- Non-teacher-

 

comments related feedback-related Other
Successﬁrl 79 59 (74.7%) 7 (8.9%) 13 (16.5%)
Acknowledge 67 48 (71.6%) 8 (11.9%) 11 (16.4%)
Inconclusive 34 21 (61.8%) 5 ( 14.7%) 8 (23.5%)

 

X2(4, n=180)=2.139, p=.710 (Unsuccessful uptake and partially successﬁil uptake were
excluded due to low expected cell counts)

Table 90 Uptake and test results

 

 

 

 

Immediate test Delayed test
N‘m‘l’e’ 0f Correct Incorrect Number Of Correct Incorrect
test 1tems test 1tems
108 54 90 82
Uptake 162 (66.7%) (33.3%) 172 (52.3%) (47.7%)
54 28 31 31
N° uptake 82 (65.9%) (34.1%) 62 (50%) (50%)
No - 28 14 14 48 29 19
opportunity (50%) (50%) (60.4%) (39.6%)
x’(2, n=272)=2.972, p=.226 x2(2, n=282)=1.312, p=.519

 

Table 91 Successfulness of uptake and test results

 

 

 

 

 

Immediate test ’ Delayed test
Number
Number Of Correct Incorrect of test Correct Incorrect
test 1tems -
1tems
57 18 52 40
Success“ 75 (76.0%) (24.0%) 92 (56.5%) (43.5%)
36 24 31 27
ACkDOWICdge 60 (60.0%) (40.0%) 58 (53.4%) (46.6%)
. 15 10 7 14
Inconcluswe 25 (60.0%) (40.0%) 21 (33.3%) (66720
x2(2, n=160)=4.650, p=.098 X2(2,n=171F3o711,P=~156

 

(Unsuccessful uptake and partially successﬁll uptake were excluded due to low expected
cell counts)

T ea‘cher-feedback-related comments can be a reliable data source
Another thing that stood out in the ﬁndings of this study is the relationship between

teacher-feedback-related comments and test results. Since the stimulated recall stimuli

220

and test items were developed from different observations, it is impossible to examine the
relationship between the two with chi-square analysis. The available information,
however, shows that the rate of teacher-feedback-related comments in general
corresponds well with test results. Although there are some exceptions, they are not
ﬁequent. Given the relationship between noticing and learning, this suggests that teacher—
feedback-related comments could be a reliable predictor of noticing and learning. In
general, when there are more teacher-feedback-related comments ﬁom a particular type
of feedback, there are also more correct answers for items developed from such episodes
in the immediate test and delayed test. One typical example is the presence of general
extralinguistic cues. That is, episodes with general extralinguistic cues led to signiﬁcantly
more teacher-feedback-related comments and learners provided signiﬁcantly more
correct answers for items developed ﬁ'om such episodes in the two tests. This result

means that extralinguistic cues predicted both more noticing and more learning.

7.3 Reﬂection on the present study

This section is a reﬂection on the design of the present study. The study is ﬂawed in
several ways. First of all, many during-observation factors that could have inﬂuenced the
results were not operationalized. For example, certain types of feedback episodes
occurred ﬁ'equently in some classes but rarely in other classes. Some teachers organized
more communicative activities in their lessons than other teachers. Learners’ proﬁciency
levels, which have been found to 'affect noticing and learning (Mackey & Philp, 1998;
Philp, 2003), were different from class to class. There were also big differences among

learners in the same class. For example, in some classes a few learners were always

221

 

interacting with teachers while others were always quiet. All these inter- and intra-class
variations could have played a role in the ﬁndings but not considered. Another thing the
study should have taken into account is post-observation factors. Data collection took
place in part of an intensive English program. Learners attended various classes between
the observations and the tests or stimulated recall interviews. Most learners had to ﬁnish
certain amount of homework. In a second language environment, they might also
improve their English when socializing with native speakers. All these activities might
have contributed to the observed effect of teacher feedback and yet not controlled or
documented. A third drawback of the study is that the analysis is not deep enough.
Inferential information was mostly drawn from chi-square tests and adjusted standardized
residuals. Due to the complexity of categories involved as well as limited resources, no
further analysis (e.g., regression) was made to better investigate how well the
independent variables predicted the dependent variables. Moreover, chi-square analysis
was not conducted with all variables due to its expected cell count requirement. A
detailed analysis of the stimulated recall comments was made but left out because it did
not directly answer the research questions. A fourth limitation of the study is the timing
of the stimulated recall interviews. As a result of the research setting and again lack of
resources, the stimulated recall interviews on average were conducted later than desired.
This could have caused more loss of memory, reducing the veridicality of the noticing
data. A ﬁfth thing that could have strengthened the study would have been to have
teacher interviews. Some studies (e. g., Mackey et al., 2007) have found that teachers’ ‘
views on giving feedback can help reveal the relationship between teachers’ intentions

and learners’ interpretations of the feedback. Due to the lack of data on teacher views, the

222

 

interpretation of the results in many cases had to draw on assumptions without enough
evidence. The addition of teacher interviews to the research protocol would have helped

to solve this problem.

Despite the various drawbacks, the study is strong and unique in many ways too. First of
all, all the data were based on natural classroom interactions. Because many factors (e.g.,
the type of lessons and activities) were beyond my control, the coding and data analysis
might not be as clean as that of laboratory studies. However, the data do show the
interactions in real life ESL classrooms. The ﬁndings therefore allow for a different
perspective on interaction, and they add some nuance to research using laboratory

settings.

Second, unlike other studies that only operationalized noticing with a single measure, for
example, uptake charts or think-aloud protocols, this study triangulated the measurement
of noticing with both stimulated recall comments and learner uptake. The triangulation of
two measures, one offline and one online, might have helped to elicit more thorough

noticing data.

Third, studies which used stimulated recall interviews to elicit noticing data were often
conducted in an experimental setting. In some studies, the stimuli contained speciﬁed
language structures; in some, teachers were told what kinds of feedback types were
desired. In the current study, the stimuli for the recall interviews were totally based on

natural classroom interactions. The teachers were not asked to teach any particular

223

language structure or use any particular feedback type just for the purpose of the study.
The stimuli used for the recall interviews were therefore more classroom-based and the
noticing data elicited ﬁ'om these stimuli might better reﬂect learners’ cognitive processes

in real time classroom interactions.

Fourth, apart from factors that have received heavy attention but still call for further
investigation -— the characteristics of feedback, this study also investigated factors that
have received little attention — nonlinguistic cues and teachers’ metalanguage use. The
inclusion of these factors aligned with and added to the ﬁndings of the few researchers

who have worked on them.

Finally, the use of video-recordings made a difference. Most, if not all, studies which
investigated teacher feedback and learner uptake used only audio-recordings of classroom
interactions. In this study, I used a video camera to record classroom interactions. The
videotapes helped me to capture the minute detail that might otherwise have slipped away.
Consequently, I was able to examine learners’ responses to teacher feedback with

gestures and other body language, which few researchers have looked at.

7 .4 Suggestions for future research
Based on what has been discussed in the previous sections, some tentative suggestions
can be made about future research on the noticing and effectiveness of teacher feedback.

These are listed below, without any more redundant discussion:

224

l.

Nonlinguistic cues play an important role in the noticing and effect of teacher
feedback and therefore deserve more attention. It would help to have a study in which
nonlinguistic cues are the primary focus of data collection and in which a well
established taxonomy is used to analyze the multiple effects of nonlinguistic cues.

Learners’ nonverbal responses to teacher feedback can make a difference to research
ﬁndings and therefore should be considered in feedback research. We need a study
that examines how learners’ nonverbal responses contribute to the dynamics of
teacher-student interactions, and how such responses help with the negotiation of

meaning and subsequent learning.

. When a study involves classroom observations, post-observation learning

opportunities could affect learners’ language development. Therefore, we need a
study that measures post-observation learning opportunities at various intervals and in
various language use situations.

As I mentioned, the inclusion of both teacher views and learner views can help
researchers better interpret learners’ perceptions of teacher feedback. It would be a
good idea to conduct a study similar to mine (but not necessarily identical) and add

the component of teacher interviews.

7 .5 Concluding remarks

In conclusion, the ﬁndings of the present study indicate that all the three factors
examined -- the characteristics of feedback episodes, nonlinguistic cues in teacher
feedback, and teachers’ use of metalanguage -- can affect learners’ noticing of teacher

feedback and their language improvement. Although the ﬁndings may not apply to every

225

teaching context, they carry practical implications for classroom teaching to a certain
degree. Moreover, while lending support to the ﬁndings of existing research, the study
also raised important methodological issues. One ﬁnding is that learner uptake alone is
not a sufﬁcient measure for the noticing and effect of teacher feedback. Therefore, it is
necessary to triangulate multiple measures in order to increase the validity of research
data. On the other hand, teacher-feedback-related comments from stimulated recall
interviews that were based on natural classroom interactions could be a reliable predictor
of noticing and learning. Such interviews therefore can be a valuable tool to collect

noticing and learning data.

226

 

REFERENCES

Adams, R. (2003). L2 output, reformulation, and noticing: Implications for IL
development. Language Teaching Research, 7, 347-376.

Alanen, R. (1995). Input enhancement and rule presentation in second language
acquisition. In R. Schmidt (Ed.), Attention and Awareness in Foreign Language
Learning and Teaching (Technical Report #9) (pp. 259-302). Honolulu, Hawai’i:
University of Hawai’i, Second Language Teaching and Curriculum Center.

Allwright, B. (1984). Why don’t learners learn what teachers teach? The interaction
hypothesis. In D. Singleton & D. Little (Eds.), Language Learning in Formal and
Informal Contexts (pp. 3-18). Dublin: Irish Association for Applied Linguistics.

Ano, K. (1998). A study of the output hypothesis: Cognitive processes of speaking a
foreign language. Journal of Japan-Korea Association of Applied Linguistics, 2,
1 75-204.

Ayoun, D. (2001). The role of negative and positive feedback in the second language
acquisition of the passé compose and imparfait. The Modern Language Journal,
85(2), 226-243.

Baker, C. L. (1979). Syntactic theory and the projection problem. Linguistic Inquiry, 10,
533-581 .

Basturkrnen, H., Loewen, S., & Ellis, R. (2002). Metalanguage in focus on form in the
communicative classroom. Language Awareness, 11(1), 1-13.

Berman, R. A. (1979). Rule of grammar or rule of thumb? International Review of
Applied Linguistics, 1 7(4), 279-302.

Birdsong, D. (1989). Metalinguistic Performance and Interlinguistic Competence. New
York: Springer.

Bley-Vroman, R. (1989). What is the logical problem of foreign language learning? In S.
Gass & J. Schachter (Eds), Linguistic Perspectives on Second Language
Acquisition (pp. 41-68). Cambridge, England: Cambridge University Press.

Block, D. (2003). The Social T urn in Second Language Acquisition. Edinburgh: EUP.

Borg, S. (1998). Talking about grammar in the foreign language classroom. Language
Awareness, 7(4), 159-175.

227

Borg, S. (1999). Teacher’s use of grammatical terminology in the second language
classrooms: A qualitative study. Applied Linguistics, 20, 95-126.

Brown, R. & Hanlon, C. (1970). Derivational complexity and the order of acquisition in
child speech. In J. Heyes (Ed.,) Cognition and the Development of Language
(pp. 155- 207). New York: Wiley.

Carpenter, H., Jeon, K. S., McGregor, D., & Mackey, A. (2006). Learners’ interpretations
of recasts. Studies in Second Language Acquisition, 28, 209-236.

Chomsky, N. (1975). Reﬂections on Language. New York: Pantheon.

Corder, S. (1967). The signiﬁcance of learners’ errors. International Review of Applied
Linguistics, 5, 161-170.

Corder, S. (1973). Introducing Applied Linguistics. Harmondsworth: Penguin.
Davies, M. (2006). Paralinguistic focus on form. TESOL Quarterly, 40(4), 841-855.

Egi, T. (2004). Verbal reports, noticing, and SLA research. Language Awareness, 13,
243-264.

Egi, T. (2007a). Recasts, learners’ interpretations, and L2 development. In A. Mackey
(Ed.), Conversational Interaction in Second Language Acquisition: A Series of
Empirical Studies (pp. 249-267). Oxford: Oxford University Press.

E gi, T. (2007b). Interpreting recasts as linguistic evidence: The roles of linguistic target,
length, and degree of change. Studies in Second Language Acquisition, 29(4),
511-537.

Egi, T. (2008). Investigating stimulated recall as a cognitive measure: Reactivity and
verbal reports in SLA research methodology. Language Awareness, 00(0), 1-17.

Ellis, R. (2005). Measuring implicit and explicit knowledge of a second language: A
psychometric study. Studies in Second Language Acquisition, 27, 141-172.

Ellis, R., Basturkrnen, H., & Loewen, S. (2001a). Learner uptake in communicative ESL
lessons. Language Learning, 51(2), 281-318.

Ellis, R. Basturkrnen, H. ,& Loewen, S. (2001b). Preemptive focus on form 1n the ESL
classroom. TESOL Quarterly, 35(3), 407-432.

Ellis, R. & Sheen, Y. (2006). Reexamining the role of recasts in second language
acquisition. Studies in Second Language Acquisition, 28, 575-600.

Ellis, R., Tanaka, Y., & Yamazaki, A. (1994). Classroom interaction, comprehension and

228

the acquisition of L2 word meanings. Language Learning, 44, 449-491.

Ericsson, K. & Simon, H. (1993). Protocol Analysis: Verbal Reports as Data (2nd ed.).
Boston: MIT Press.

F aech C. (1985). Meta talk 1n FL classroom discourse. Studies in Second Language
Acquisition, 7(2), 184-199.

F araco, M., & Kida, T. (2008). Gesture and the negotiation of meaning in a second
language classroom. In S. G. McCafferty & G. Stam (Eds), Gesture: Second

Language Acquisition and Classroom Research (pp. 280—297). London:
Routledge.

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

Gass, S. (1988). Integrating research areas: A framework for second language studies.
Applied Linguistic, 9, 198-217.

Gass, S. (1997). Input, Interaction, and the Second Language Learner. Mahwah, NJ:
Lawrence Erlbaum Associates.

Gass, S. (2003). Input and interaction. In C. Doughty & M. Long (Eds), The Handbook
of Second Language Acquisition (pp. 224-255). Malden, MA: Blackwell.

Gass, S. & Mackey, A. (2000). Stimulated Recall Methodology in Second Language
Research. Mahwah, NJ: Lawrence Erlbaum Associates.

Gass, S. & Mackey, A. (2006). Input, interaction and output. AILA Review, 19, 3-17.

Gass, S. & Mackey, A. (2007). Data Elicitation for Second and Foreign Language
Research. Mahwah, NJ: Lawrence Erlbaum Associates.

Gass, S. & Selinker, L. (2001). Second Language Acquisition. An Introductory Course
(2nd ed) Mahwah, NJ: Lawrence Erlbaum Associates.

Gass, S. & Veronis, E. (1989). Incorporated repairs in NNS discourse. In M. Einstein
(Ed.), Variation and Second Language Acquisition (pp. 71-86). New York:
Plenum.

Gullberg, M. & McCafferty, S. G. (2008). Introduction to gesture and SLA: Toward an
integrated approach. Studies in Second Language Acquisition, 30, 133-146.

Halliwell, S. (1993). Grammar Matters. London: CILT.

Han, Z. (2002). A study of the impact of recasts on tense consistency in L2 output.

229

 

TESOL Quarterly, 36(4), 543-572.

Hawkins, B. (1985). Is an “appropriate response” always so appropriate? In S. Gass & C.
Madden (Eds.), Input in Second Language Acquisition (pp. 162-178). Rowley,
MA: Newbury House.

Hinkle, D. E., Wiersma, W., & Jurs, S. G. (2003). Applied Statistics for the Behavioral
Sciences (5th ed.). Boston, MA: Houghton Miﬁlin Company.

Iwashita, N. (2003). Negative feedback and positive evidence in task-based interaction.
Studies in Second Language Acquisition, 25, 1-36.

James, C. (1999). Language awareness: Implications for the language curriculum.
Language, Culture and Curriculum, 12(1), 94-115.

Jenkins, S. & Parra, I. (2003). Multiple layers of meaning in an oral proﬁciency test: The
complementary roles of nonverbal, paralinguistic, and verbal behaviors in
assessment decisions. The Modern Language Journal, 87(i), 90-107.

Jourdenais, R. (2001). Protocol analysis and SLA. In P. Robinson (Ed.), Cognition and
Second Language Instruction (pp. 354-375). Cambridge: Cambridge University
Press.

Kida, T. (2008). Does gesture aid discourse comprehension in the L2? In S. G.
McCafferty & G. Stam (Eds), Gesture: Second Language Acquisition and
Classroom Research (pp. 131-156). London: Routledge.

Kim, H. (1995). Intake from the speech stream: Speech elements that learners attend to.
In R. Schmidt (Ed.), Attention and Awareness in Foreign Language Learning and
Teaching (Technical Report #9) (pp. 65-84). Honolulu, Hawai’i: University of
Hawai’i, Second Language Teaching and Curriculum Center.

Krashen, S. (1982). Principles and Practice in Second Language Acquisition. London:
Pergarnon.

Krashen, S. (1985). The Input Hypothesis: Issues and Implications. Oxford: Pergamon
Press.

Kucer, S. B. (2005). Dimensions of Literacy: A Conceptual Base for Teaching Reading
and Writing in School Settings (2nd ed.). Mahwah, N.J.: Lawrence Erlbaum

Associates. ‘

Leow, R. (1997). Attention, awareness and foreign language behavior. Language
Learning, 4 7, 467-505.

Leow, R. (2000). A study of the role of awareness in foreign language behavior: Aware

230

 

vs. unaware learners. Studies in Second Language Acquisition, 22, 557-5 84.

Leow, R. (2002). Models, attention and awareness in SLA: A response to Simard and
Wong’s ‘Alertness, orientation, and detection: The conceptualization of
attentional functions in SLA.’ Studies in Second Language Acquisition, 24, 113-
119.

Lightbown, P. (1998). The importance of timing in focus on form. In C. Doughty & J.
Williams (Eds.), Focus on Form in Classroom Second Language Acquisition (pp.
177-196). Rowley, MA: Newbury House.

Loewen, S. (2002). The occurrence and effectiveness of incidental focus on form in
meaning-focused ESL lessons. Unpublished Doctoral Thesis, The University of
Auckland, Auckland, New Zealand.

Loewen, S. (2003). Variation in the frequency and characteristics of incidental focus on
form. Language Teaching Research, 7(3), 315-345.

Loewen, S. (2004). Uptake in incidental focus on form in meaning-based ESL lessons.
Language Learning, 54(1), 153-188.

Loewen, S. (2005). Incidental focus on form and second language learning. Studies in
Second Language Acquisition, 27, 361-3 86.

Loewen, S. & Philp, J. (2006). Recasts in the adult English L2 classroom: Characteristics,
explicitness, and effectiveness. The Modern Language Journal, 90(iv), 536-556.

Long, M. (1983). Linguistic and conversational adjustments to non-native speakers.
Studies in Second Language Acquisition, 5, 177-194.

Long, M. (1985). Input and second language acquisition theory. In S. Gass & C. G.
Madden (Eds.), Input in Second Language Acquisition (pp. 377-393). Rowley,
MA: Newbury House.

Long, M. (1991). Focus on form: A design feature in language teaching methodology. In
K. De Bot, R. Ginsberg, & C. Kramsch (Eds.), Foreign Language Research in
Crosscultural Perspective (pp. 39-52). Amsterdam: Johns Benjamins.

Long, M. (1996). The role of the linguistic environment in second language acquisition.
In W. C. Ritchie & T. K. Bhatia (Eds. ), Handbook of Second Language
Acquisition (pp. 413-468). San Diego, CA: Academic Press.

Long, M. (2007). Problems in SLA. Mahwah, NJ: Lawrence Erlbaum Associates.

Long, M. & Robinson P. (1998). Focus on form: Theory, research and practice. In C.

231

Doughty & J. Williams (Eds.), Focus on Form in Classroom Second Language
Acquisition (pp. 15-41). Cambridge: Cambridge University Press.

Lyster, R. (1998a). Negotiation of form, recasts, and explicit correction in relation to
error types and learner repair in immersion classrooms. Language Learning, 48(2),
1 83-21 8.

Lyster, R. (1998b). Recasts, repetition, and ambiguity in L2 classroom discourse. Studies
in Second Language Acquisition, 20, 51-81

Lyster, R. (2004). Differential effects of prompts and recasts in fonn-focused instruction.
Studies in Second Language Acquisition, 26, 399-432.

Lyster, R., & Mori, H. (2006). Interactional feedback and instructional counterbalance.
Studies in Second Language Acquisition, 28, 269-300.

Lyster, R. & Ranta, L. (1997). Corrective feedback and learner uptake: Negotiation of
form in communicative classrooms. Studies in Second Language Acquisition, 20,
37-66.

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

Mackey, A. (2006). Feedback, noticing and instructed second language learning. Applied
Linguistics, 27(3), 405-430.

Mackey, A., Al-Khalil, M., Atanassova, G., Hama, M., Logan-Terry, A., & Nakatsukasa,
K. (2007). Teachers’ intentions and learners’ perceptions about corrective

feedback in the L2 classroom. Innovation in Language Learning and Teaching,
1(1),129-152.

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

Mackey, A., McDonough, K., Fujii, A., & Tatsumi, T. (2001). Investigating learners’
reports about the L2 classroom. International Review of Applied Linguistics, 39,
285-308.

Mackey, A. & Philp, J. (1998). Conversational interaction and second language
development: Recasts, responses, and red her-rings? The Modern Language :

Journal, 82(3), 338-356.

Mackey, A., Philp, J ., Egi, T., Fujii, A., & Tatsumi, T. (2002). Individual differences in

232

working memory, noticing of interactional feedback and L2 development. In P.
Robinson (Ed.), Individual Diﬂerences and Instructed Language Learning (pp.
181-209). Philadelphia: Benjamins.

Mackey, A. & Polio, C. (2009). Introduction. In A. Mackey & C. Polio (Eds.), Multiple
Perspectives on Interaction: Second Language Research in Honor of Susan M.
Gass (pp. 1-10). New York: Routledge.

Martin, H. R. (1981). The prosodic components of speech melody. The Quarterly Journal
of Speech, 6 7, 81-92.

McDonough, K (2005). Identifying the impact of negative feedback and learners'
responses on ESL question development. Studies in Second Language Acquisition,
27(1), 79-103.

McDonough, K. & Mackey, A. (2006). Responses to recasts: Repetitions, primed
production, and linguistic development. Language Learning, 56(4), 693 -720.

McNeil, D. (1992). Hand and Mind: What Gestures Reveal about Thought? Chicago:
The University of Chicago Press.

McNeil, D. (2005). Gesture and Thought. Chicago, IL: The University of Chicago Press.
Mehrebian, A. (1972). Nonverbal Communication. Chicago: Aldine-Atherton.

Nabei, T. & Swain, M. (2002). Learner awareness of recasts in classroom interaction: A
case study of an adult EFL student’s second language learning. Language
Awareness, 1 1(1), 43 -63.

Ohta, A. S. (2000). Re-thinking recasts: A leamer-centered examination of corrective
feedback in the Japanese language classroom. In J. K. Hall & L. Verplaeste (Eds.),
The Construction of Second and Foreign Language Learning through Classroom
Interaction (pp. 47-71). Mahwah, NJ: Erlbaum.

Oliver, R. (2000). Age differences in negotiation and feedback in classroom and pairwork.
Language Learning, 50(1), 1 19-151 .

Oliver, R. & Mackey, A. (2003). Interactional context and feedback in child ESL
classrooms. The Modern Language Journal, 87(iv), 519-533.

Panova, I: & Lyster, R. (2002). Patterns of corrective feedback and uptake in an adult
ESL classroom. TESOL Quarterly, 36(4), 573-595.

Philp, J. (2003). Constraints on “noticing the gap”: Nonnative speakers’ noticing of
recasts in NS-NNS interaction. Studies in Second Language Acquisition, 25, 99-
126.

233

Pica, T., Holliday, L., Lewis, N., & Morganthaler, L. (1989). Comprehensible output as
an outcome of linguistic demands on the learner. Studies in Second Language
Acquisition, II, 63-90.

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

Robinson, P. (1995). Attention, memory, and the “noticing” hypothesis. Language
Learning, 45(2), 283-331.

Robinson, P. (1996). Learning simple and complex second language rules under implicit,
incidental, rule search and instructed conditions. Studies in Second Language
Acquisition, I 8, 27-67.

Robinson, P. (1997). Individual differences and the fundamental similarity of implicit and
explicit adult second language learning. Language Learning, 4 7, 45-99.

Robinson, P. (2003). Attention and memory during SLA. In C. Doughty & M. Long
(Eds.), The Handbook of Second Language Acquisition (pp. 631-678). Malden,
MA: Blackwell.

Rosa, E. & O’Neill, M. (1999). Explicitness, intake, and the issue of awareness: Another
piece to the puzzle. Studies in Second Language Acquisition, 21, 511-556.

Schachter, J. (1988). Second language acquisition and its relationship to Universal
Grammar. Applied Linguistics, 9, 219-235.

Schmidt, R. (1990). The role of consciousness in second language learning. Applied
Linguistics, 11(2), 129-158.

Schmidt, R. (1993). Awareness and second language acquisition. Annual Review of
Applied Linguistics, I 3, 206-226.

Schmidt, R. (2001). Attention. In P. Robinson (Ed.), Cognition and Second Language
Instruction (pp. 3-32). Cambridge: Cambridge University Press.

Schmidt, R. & Frota, S. (1986). Developing basic conversational ability in a second
language: A case study of an adult learner of Portuguese. In R. Day (Ed), Talking
to Learn: Conversation in Second Language Acquisition (pp. 237-322). Rowley,
MA: Newbury House. ' ‘

Sheen, Y. (2004). Corrective feedback and learner uptake in communicative classrooms
across instruction settings. Language Teaching Research, 8(3), 263-300.

Sheen, Y. (2006). Exploring the relationship between characteristics of recasts and

234

learner uptake. Language Teaching Research, 10(4), 361 -392.

Shriberg, E., Bates, R., Stolcke, A., Taylor, R, Jurafsky, D., Ries, K., Coccard, N.,
Martin, R., Meteer, M., & van Ess-Dykema, C. (1998). Can prosody aid the
automatic classiﬁcation of dialog acts in conversational speech? Language and
Speech, 41 (3-4), 443-492.

Sime D. (2006). What do learners make of teachers’ gestures in the language classroom?
International Review of Applied Linguistics, 44, 21 1-230.

Slimani, A. (1989). The role of topicalization in classroom language learning. System, I 7,
223-234.

Song, S. (2007). Beginning ESL learners’ noticing of morphological and syntactic
changes in recasts. Teachers College, Columbia University Working Papers in
TESOL & Applied Linguistics, 7(1), 1-25.

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 (pp. 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 (pp. 125-144). Oxford: Oxford University Press.

Swain, M. (1997). The output hypothesis, focus on form and second language learning. In
V. Berry, B. Adamson, & W. Littlewood (Eds.), Applying Linguistics: Insights
into Language in Education (pp. 1-21). Hong Kong: English Centre, University of
Hong Kong.

Swain, M. (2005). The output hypothesis: Theory and research. In E. Hinkel (Ed.),
Handbook of Research in Second Language Teaching and Learning (pp. 471-484).
Mahwah, NJ: Lawrence Erlbaum Associates.

Swain, M. & Lapkin, S. (1995). Problems in output and the cognitive processes they
generate: A step toward second language learning. Applied Linguistics, I6, 371-
391.

Tomlin, R. & Villa, V. (1994). Attention in cognitive science and SLA. Studies in Second
Language Acquisition, 16, 185- 204.

Trahey, M. (1996). Positive evidence in second language acquisition: Some long-term
effects. Second Language Research, 12, l 1 1-139.

Trahey, M. & White, L. ( 1993). Positive evidence and preemption in the second language

235

classroom. Studies in Second Language Acquisition, 15, 181-204.

VanPatten, B. (1994). Evaluating the role of consciousness in SLA: Terms, linguistic
features, and research methodology. AILA Review, I I, 27-36.

Wagner-Gough, K. & Hatch, E. (1975). The importance of input in second language
acquisition studies. Language Learning, 25, 297-308.

White, L. (1991). Adverb placement in second language acquisition: Some effects of
positive and negative evidence in the classroom. Second Language Research, 7,
133-161.

Williams, J. (2001). The effectiveness of spontaneous attention to form. System, 29, 325-
340.

Williams, J. (2005). Forrn-focused instruction. In E. Hinkel (Ed.), Handbook of Research

in Second Language Teaching and Learning (pp. 671-691). Mahwah, NJ:
Lawrence Erlbaum Associates.

236

llllllllll