”42(1153:
x114. _.

.v 3.
3.1.5:, .
all. 11

duavt..1l.
x .2. . .
z «

"3 r‘ '1 ‘
mag: *v

Am»?

‘ Fair!
1!: 9.3.0.;
::..Ine

r .- \ I311): 32.

v .. 9 :.ol.‘..

7|I;.3... a .
.3“!

3.1.3..

: N

. u.
:1.
«it. .u .2,

. .3215?
. cl: . 33.5.:
.3... .. 4. ~u$fi§v$xnfi
.V :33! i. .(
7:21,}! A a
.3352...

3.1.7.15... 10!.

I. 3,136.52... .4 a.
2: .07.... .1: In.
$3.513?!

‘ I J‘I‘mi’l‘. ‘Ql," .-OA.

: , . E :iili 2: 495.2..-
.2 av tuba. 074...... gaysnnwxrtfdbvvh.
D ’ii“ ‘0 I‘vgé‘

‘ . ‘ gt- b{

. ,ll. 2.7.1

t. .. Rana
Lt......n!.nn:
“H.531: V ‘
z ‘
2.. . .. -3
i1. .5...- aanfiuuuhuhv.
. 2.7 :itérat)...‘ .9 IJ
. 3;- . .. i: r .. 1...... . .
p L y n . ....... n ““92. man}...
. u. ..r.
z.

 

LianAmES

\l‘lllflilllllliljlllllll\

3 29 010

This is to certify that the

dissertation entitled

A ROLE FOR INDIRECT NEGATIVE EVIDENCE
IN SECOND LANGUAGE ACQUISITION

presented by

India Catherine Plough

has been accepted towards fulfillment
of the requirements for

m.

PhD degree in English

 

)iwflM/Vl/Uzu/L

Major professor

Date JUl)’ 15, 1994

 

MSUis an Affirmative Action/Equal Opportunity Institution 0- 12771

 

 

LIBRARY
Mlchlgan State
Unlverslty

 

 

 

 

PLACE ll RETURN BOX» man this ohookoutrom your rooord.
TO AVOID FINES town on or bdoro doto duo.

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

: fl : [:1
__—:ll:_—JL:-J
II I I,

MSU loAn Afflnnotlvo Action/Equal Opportunity Institution
W

 

 

 

 

 

 

 

 

Ins-9.1

A ROLE FOR INDIRECT NEGATIVE EVIDENCE
IN SECOND LANGUAGE ACQUISITION

By

India Catherine Plough

A DISSERTATION

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

DOCTOR OF PHILOSOPHY

Department of English

1 994

ABSTRACT

A ROLE FOR INDIRECT NEGATIVE EVIDENCE
IN SECOND LANGUAGE ACQUISITION

By

India Catherine Plough

The potential significance of indirect negative evidence (INE) has been all but
ignored in Second Language Acquisition (SLA) research. In general terms, this
dissertation addresses the Learnability Problem of language acquisition and the role that
INE may play as a partial solution to that Problem. Specifically, the purpose of this
dissertation is to investigate the relationship between the use of inductive inferencing and
the acquisition of the ungrammaticality of Subject-Adverb-Verb-Object (SAVO) word
order in French by native speakers of English. I argue that the use of INE is analogous
to the use of inductive inferencing, and therefore, the use of INE can be investigated via
an investigation of the use of inductive inferencing.

The database for the study consists of 45 native speakers of English enrolled in
first year French at Michigan State University. All volunteers took three pretests, two
of which tested their knowledge of adverb placement in French and one of which tested
their inductive reasoning ability. Subjects were divided into two groups based on class

section. For a two week period in Group 1 the teachers’ regular lessons were altered so

that, wherever possible, exercises contained adverbs. In Group 2, in addition to adverb
exercises, students received verbal inductive inferencing exercises in both French and
English. Immediately following the two week instructional period, all subjects were
posttested on their knowledge of adverb placement in French. Three weeks after the first
posttest, students were posttested again.

Manifested in the results of this study is the now well-known fact of the necessary
interrelatedness of the variables involved in SLA. Just as one variable cannot be
examined in isolation from others, so too, the cognitive processes which both affect and
are affected by these variables must be considered; the use of INE or inductive
inferencing is one such cognitive process. While the results of the current study do not
conclusively show that the use of INE is operative in the acquisition of the
ungrammaticality of SAVO word order in French by native speakers of English, they do

suggest that the use of inductive inferencing plays a role in SLA.

This dissertation is dedicated to the memory of

Lynn Ann Wehner
1957-1991

whose goodness will always be a source of strength.

iv

ACKNOWLEDGMENTS

A sports announcer/philosopher once said upon receiving an award that he was
reminded of a time when he was walking down a country road and came across a turtle
perched on top of a fence post. He could not figure out exactly how she had gotten there.
But one thing he did know for sure-«she did not get there by herself.

I am deeply grateful to many people.

The College of Arts and Letters which awarded me a dissertation fellowship that
provided the time necessary to do the data collection.

Michael Koppisch, Keith Palka, and David Wright for their cooperation and
assistance which made this study possible.

Larry Kuiper, Mary McCullough, Shawn Morrison, and Alya Mousli--the teachers
of French ISO—who willingly took part in the study, encouraged their students to volunteer,
and were extremely tolerant during the stressful times.

The students of French 150; they a_r_e the study.

Cheryl Delk, Jean Nicholas, and Catherine Tamareille who spent hours checking and
rechecking materials.

Lydia White and Martha Trahey for providing me with the AI and WOC Tasks which
were used in the study.

Stefan Hannuschke, Sabine Helling, and Amy Tickle who helped with the initial data

collection.

Gary Cook who provided hours of statistical expertise. Chapter 3 would not have
been possible without his instruction and assistance. This dissertation has benefitted from
our many conversations.

Charlene Polio for taking on the last-minute task of reading this dissertation,
discussing it with me in detail, and attending the defense.

Usha Lakshmanan who introduced me to the concept of INE and has provided
encouragement every step of the way.

Amy, Cheryl, Joanna, Tom, and Wayne for standing by me through it all.

Susan Gass, Carolyn Harford, Paul Munsell, and Dennis Preston--my committee
members to whom I am indebted for their guidance and support. Carolyn, a fantastic
teacher, clarified theoretical issues for me which prevented me from going astray early on.
Her calm understanding helped me through trying times. Paul was there at just the right
time with empathy and sympathy. Dennis contributed above and beyond the call of duty.
His sense of humor was a welcome complement to his academic acuity. Sue has been my
mentor fiom the very beginning. She instilled me with confidence, kept me focused, and
generously shared her encyclopedic knowledge of the field. I feel extremely fortunate to
have had the opportunity to work with a professional such as Sue.

My family whose support and patience has been invaluable. Faye for putting things
into perspective for me. My parents, Dallas and Charlotte Haun, for their love and the
benefit of their wisdom.

And John, where would I be without you?

vi

TABLE OF CONTENTS

LIST OF TABLES
LIST OF FIGURES
INTRODUCTION

CHAPTER 1 THE LEARN ABILITY PROBLEM
1.1. Positive Evidence
1.2. Negative Evidence in First Language Acquisition
1.2.1. Summary
1.3. Negative Evidence in Second Language Acquisition
1.3.1. Summary
1.4. Indirect Negative Evidence
1.4.1. Indirect Negative Evidence in First Language Acquisition
1.4.1.1. Summary
1.4.2. Indirect Negative Evidence in Second Language Acquisition
1.5. Chapter Summary

CHAPTER 2 INDUCTIVE INFERENCING
2.1. Inductive Inferencing
2.2. Inductive Inferencing in First Language Acquisition
2.3. Inferencing in Second Language Acquisition
2.4. Inductive Inferencing and Indirect Negative Evidence
2.5. Inductive Inferencing: A Working Definition
2.6. Summary
2.7. Learning Strategies
2.8. Inductive Inferencing as a Learning Strategy
2.9. Teaching of Learning Strategies
2.10. Chapter Summary

CHAPTER 3 METHODOLOGY
3.1. Procedures Overview
3.2. Subjects
3.3. Teaching
3.3.1. Adverbs
3.3.2. Inferencing

vii

xii

13
13
15
21
22
31
31
33
44
45
48

50
50
51
57
58
63
64
64
67
81
89

9 l
92
94
94
94
96

3.3.3. Questionnaires
3 .4. Acceptability Judgments
3.5. Testing Materials
3.5.1. Adverb Placement
3.5.1.1. Word Order Correction Task
3.5.1.1.1. Scoring Procedure
3.5.1.2. Acceptability Judgment Task
3.5. 1.2. 1. Scoring Procedure
3 . 5.2. Inductive Inferencing
3.5.2.1. Raven's Standard Progressive Matrices Test
3.5.2.1.1. Validity
3 .5 .2. 1.2. Normative Studies
3.5.2.1.3. Reliability
3.5.2.1.4. Scoring Procedure
3.6. Chapter Summary

CHAPTER 4 RESULTS
4.1. Learner Variables
4.2. Overview of All Tests
4.2.1. Acceptability Judgment Task
4.2.2. Word Order Correction Task
4.2.3. Summary
4.2.4. Word Order Correction Task Subscores
4.2.4.1. Correct Subscores
4.2.4.1.1. Correct Subscore Summary
4.2.4.2. Error Subscores
4.2.4.2.1. Error Subscore Summary
4.3. Repeated Measures Analysis of Variance
4.3.1. Acceptability Judgment Task
4.3.2. Word Order Correction Task
4.3.3. ANOVA Summary
4.4. Multiple Regression
4.4.1. Acceptability Judgment Task
4.4.1.1. Correct Score
4.4.1.2. Error Score
4.4.2. Word Order Correction Task
4.4.2.1. Correct Score
4.4.2.2. Subscores
4.4.2.3. Error Score
4.4.2.4. Subscores
4.4.3. Inferencing Group
4.4.3.1. Acceptability Judgment Task
4.4.3.1.1. Correct Score
4.4.3.1.2. Error Score

viii

100
102
106
106
106
107
108
109
110
110
112
113
113
114
115

117
117
120
125
127
129
129
130
133
133
137
137
137
139
141
142
143
143
144
144
144
145
145
146
147
147
147
147

4.4.3.2. Word Order Correction Task
4.4.3.2.1. Correct Score
4.4.3.2.2. Error Score
4.4.3.221. Subscore
4.4.4. Multiple Regression Summary
4.5. Teacher Questionnaire
4.6. Student Questionnaire
4.7. Chapter Summary

CHAPTER 5 DISCUSSION
5.1. Correlation
5.2. Acceptability Judgment Task
5.2.1. Correct Score
5.2.2. Error Score
5.3. Word Order Correction Task
5.3.1. Correct Score
5.3.2. Error Score
5.4. Observations
5.4.1. Adverb Input: A delayed effect or reanalysis
5.4.2. Unequal Variables
5.4.3. Inferencing versus Noninferencing
5.4.4. Raven's Standard Progressive Matrices
5.4.5. Summary
5.5. Caveats
5.5.1. No immediate effect of treatment
5.5.2. No significant difference between groups
5.5.3. Predictors
5.6. Implications and Suggestions for Future Research
5.7. Conclusion

APPENDIX A BIOGRAPHICAL COVER SHEET
APPENDIX B SUMMARY OF LESSONS
APPENDIX C ADVERB FLOOD EXERCISES
APPENDIX D INFERENCING EXERCISES
APPENDIX E QUESTIONNAIRE TO TEACHERS
APPENDIX F QUESTIONNAIRE TO STUDENTS
APPENDIX G WORD ORDER CORRECTION TASK

APPENDIX H ACCEPTABILITY JUDGMENT TASK

ix

148
148
148
148
148
151
151
154

155
156
158
158
160
162
162
164
166
167
169
176
178
180
180
181
182
183
184
188

190

191

193

209

219

222

223

228

APPENDIX I RAVEN'S STANDARD PROGRESSIVE MATRICES SAMPLES 233

APPENDIX J MULTIPLE REGRESSIONS 234

BIBLIOGRAPHY 246

LIST OF TABLES

Table 3.1. Retest Reliability of SPM

Table 4.1. Learner Variables by Group

Table 4.2. Learner Variables by Class Section

Table 4.3. Acceptability Judgment Task

Table 4.4. Word Order Correction Task and Raven's Standard Progressive Matrices
Table 4.5. Word Order Correction Task Correct Subscores
Table 4.6. Word Order Correction Task Error Subscores
Table 4.7. Source Table AJ Task Correct Scores

Table 4.8. Source Table AJ Task Error Scores

Table 4.9. Source Table WOC Task Correct Scores

Table 4.10. Source Table WOC Task Error Scores

Table 4.11. Multiple Regression Results

Table 5.1. Pearson Product-Moment Correlation

Table 5.2. WOC Task Change in Means of All Subjects

Table 5.3. Learner Variables by Class Section

114

117

119

121

122

131

135

138

139

140

141

150

156

167

173

Figure 2.1.
Figure 4.1.
Figure 4.2.
Figure 4.3.
Figure 4.4.
Figure 4.5.
Figure 4.6.
Figure 4.7.
Figure 5.1.
Figure 5.2.
Figure 5.3.

Figure 5.4.

LIST OF FIGURES

Inferencing and Indirect Negative Evidence
AJ and WOC Tasks Mean Correct Scores
AJ and WOC Tasks Mean Error Scores
AJ Task Mean Correct Scores

AJ Task Mean Error Scores

WOC Task Mean Correct Scores

WOC Task Mean Error Scores

WOC Task Mean C2 + C4 Subscores

AJ Task Mean Correct Scores

AJ Task Mean Error Scores

WOC Task Mean Correct Scores

WOC Task Mean Error Scores

xii

59

123

124

125

126

127

128

132

158

160

162

164

INTRODUCTION

In Lectures on government and binding, Chomsky states:

Returning to our idealized - but not unrealistic - theory of language
acquisition, we assume that the child approaches the task equipped with
U6 and an associated theory of markedness that serves two functions:
it imposes a preference structure on the parameters of U6, and it
permits the extension of core grammar to a marked periphery.
Experience is necessary to fix the values of parameters of core
grammar. In the absence of evidence to the contrary, unmarked
options are selected. Evidence to the contrary or evidence to fix
parameters may in principle be of three types: (1) positive evidence
(SVO order, fixing a parameter of core grammar: irregular verbs,
adding a marked periphery); (2) direct negative evidence (corrections
by the speech community); (3) indirect negative evidence [italics added]
- a not unreasonable acquisition system can be devised with the
operative principle that if certain structures or rules fail to be
exemplified in relatively simple expressions, where they would be
expected to be found, then a (possibly marked) option is selected
excluding them in the grammar, so that a kind of “negative evidence”
can be available even without corrections, adverse reactions, etc.
There is good reason to believe that direct negative evidence is not
necessary for language acquisition, but indirect negative evidence may
be relevant. (1981, pp. 8-9)

The potential significance of indirect negative evidence (INE) has been all but
ignored in Second Language Acquisition (SLA) research and has just recently been taken
into consideration in first language acquisition (L1) studies. In general terms, this
dissertation addresses the Learnability Problem of language Acquisition and the role that
INE may play as a partial solution to that Problem. Specifically, the purpose of this
dissertation is to investigate the relationship between the use of inductive inferencing and
the acquisition of the ungrammaticality of Subject-Adverb-Verb-Object word order in

French by native speakers of English. In order to investigate the use of INE, I argue

2

that its use is analogous to the use of inductive inferencing, and that therefore the use of
INE can be investigated via an investigation of the use of inductive inferencing. The
organization of the argument is as follows:

In Chapter 1, I begin with a brief overview of the Learnability Problem in
language acquisition in order to place the role of evidence, both positive and negative,
in its theoretical perspective. A major assumption in this research, which is supported
by the evidence gathered from previous studies, is that the Learnability Problem of L1
acquisition, with respect to incomplete evidence available to the learner, equally applies
to SLA. The Learnability Problem is often used as support for Universal Grammar (if
the output exceeds the input, then there must be a contribution from some internal or
innate ’mechanism’). However, the goal of the present research is not to argue UG into
or out of existence.

The Learnability Problem is best encapsulated in the question asked by Chomsky
in 1987 (cited in Cook, 1988, p. 55): ”How do we come to have such rich and specific
knowledge, or such intricate systems of belief and understanding, when the evidence
available to us is so meager?" There are three problems with this evidence, two of
which deal with the positive evidence available to the learner. First, the language to
which learners are exposed and which informs them of what is possible in the grammar
they are creating underdetermines their final competence; that is, the input might be
expected to mislead learners into making false generalizations. The second problem with
the positive evidence available to learners is that it is often degenerate. For example,
slips-of-the-tongue, false starts, and informal dialogue are all common input for children.

Yet, even with this degenerate and incomplete positive evidence, learners are able to

3

’work through’ the complexities of the grammar and know not only which sentences are
possible but also which sentences are impossible in their language.

I examine in relatively greater detail the last aspect of the Learnability Problem,
mmely the role of negative evidence, or the overt correction which informs the learner
of what is not possible in the grammar. I provide a brief summary of some of the major
theoretical linguistic and psychology of learning studies questioning the role of, the
existence of and what constitutes negative evidence. The impact of this work on SLA
research is shown in the review of the seminal L2 studies dealing with negative evidence.
If it is the case that learners form false generalizations, and then they are overtly
corrected, the Learnability Problem would be resolved. However, the research in both
first and second language acquisition indicates that when learners make errors they do
not consistently receive overt correction or if they do receive correction, it appears that
in many instances they ignore it.

Idiscuss the literature to date (Lasnik, 1989; Valian, 1990; Saleemi, 1990, 1992;
Archibald, 1993) which addresses the issue of INE. It should be noted that these authors
are arguing for the viability of INE from a theoretical standpoint; as previously
mentioned, no empirical studies in either L1 or L2 have been conducted to date. This
is in part due to the elusiveness and ”vagueness" (White, 1989) of the concept and the
seeming difficulty of developing a research design to test for INE. All authors are
working within a UG framework, and while Iasnik and Saleemi address in some detail
what the structural contents of UG must be, Valian focuses on the psychological reality
of an acquisitional model. Theoretical support within SLA (Gass, 1988) for a role of

[NE is discussed before turning to the first proposal within SLA (Archibald, 1993) which

4

explicitly states that the evidence available to the learner may include INE. Archibald
concentrates on the problem that ”INE is not specific as to the location of the error or
to what action to take to correct things” (p. 5), and thus turns to appropriate cues, or
triggers, combined with both inductive and deductive learning as a solution.

The proposal that INE plays a role in language acquisition crucially depends on
the assumption that a domain of expectation exists. The absence of a structure must be
noticed. Without a domain of expectation, what is not present is infinite. Additionally,
a domain of expectation would partially provide the learner with a means of determining
if nonoccurrence can be equated with ungrammaticality; that is, just because a specific
structure has not been heard may not mean that that structure is impossible in the
grammar.

The question then becomes what is it which provides this domain of expectation?
For L1 acquisition, Universal Grammar has been proposed. For SLA, the question
expands to the similarities and differences of L1 and L2 acquisition; and further, can
these be explained in terms of universal principles of grammar? Among those working
within a U0 framework, the debate continues over whether principles of UG are directly
accessible (Mazurkewich, 1984; Schwartz, 1993), partially accessible [(White, 1989)
when and to what degree (Sharwood Smith, 1990)] or not accessible (Clahsen &
Muysken, 1989; Bley-Vroman, 1989; Schachter, 1988). With respect to the issue of L1
versus L2 acquisition, Eckman (1992) has argued that the most parsimonious explanation
would be to posit a single mechanism, to treat the differences as superficial, and thus to
unify the two under a single theory. However, once the focus narrows to the role of INE

in SLA, and specifically to the assumed domain of expectation, the prudent supposition

5

would be that this domain is provided by the L1. There is no justification initially to
invoke U6 and thus assume the larger domain. The present research begins from this
position.

In Chapter 2, after cursory definitions of inductive and deductive inferencing from
the fields of philosophy and psycholinguistics, I provided examples of the use of
induction in the processing of meaning in a first language. I discuss Carton’s (1966)
Method of Inferencing in Foreign Language Learning and propose that Carton’s
inferencing and Chomsky’s ’operative principle’ are analogous processes. In order to
investigate this process, I argue that the use of INE is analogous to the use of inductive
inferencing. I propose that the second language learner can reach the correct conclusion
through the process of inductive inferencing.

A role of inferencing in SLA is not new to the research agenda. In 1966, Carton

defined inferencing

. as requiring the scanning of what is already known by the learner in the
establishment of new concepts. The process inherently requires linking of new material
with what is already known. . . . Inferencing can add salience to novel stimulus
configurations. By definition the salient features of the environment are ones we notice
and remember . . . On the other hand it is conceivable that some learning through
inference occurs completely unnoticed. . . . This might occur when an unknown stimulus
element1 is embedded in what is otherwise a completely familiar and comprehensible
context (pp. 16-17).

Carton proposed that inferencing consists of a multi-stage process. First the
learner scans what is already known in the L1, the L2, and/or ’world’ knowledge.
Second, new material is linked with what is known. Finally, new concepts and

generalisations are established based on the (mis)match between the new material and

 

‘ I would argue that an "unknown stimulus element” could be the absence of a structure in an
"otherwise completely familiar and comprehensible context. "

what is already known.

Stage 1: Scanning of what is known (either L1, L2, and/or ’world’
knowledge)

Stage 2: Linking new material with what is known

Stage 3: Establishing new concepts or generalisations

A parallel can be drawn between the ’operative principle’ of Chomsky’s
acquisition system and Carton’s multi-stage inferencing process. I argue that this
’operative principle’ relies on an inferencing process.

I offer a working definition of inductive inferencing and then review the learning
strategy research in SLA as it relates to this definition as justification for the assumption
that inductive inferencing is a learning strategy (Rubin, 1975; Stern, 1975; Naiman et al,
1978; O’Malley et al, 1985; O’Malley & Chamot, 1990). That inductive reasoning plays
a role in foreign language (FL) aptitude was suggested by Carroll as early as 1953.
Carroll proposed that FL aptitude consists of four independent abilities, one of which is
”inductive language learning ability---the ability to infer or induce the rules governing
a set of language materials, given samples of language materials that permit such
inferences.” (p. 105) While Carroll remains neutral on the issue of whether language
learning ability can be improved through teaching, I argue that inductive inferencing can
be a learning strategy which can be taught (O’Malley & Chamot, 1990). Furthermore,
it is frequently the case that a learner’s performance does not reach the level of her or
his ability; explicit instruction in the learning strategy of inductive inferencing may help
a learner to ’tap into’ this ability.

In Chapter 3, I provide justification for the instruments which were used and

describe the research design. This research, based on the premise that the use of INE

7

can be analyzed via an investigation of the use of inductive inferencing, is designed to
examine the relationship between the use of inductive inferencing and the acquisition of
the ungrammaticality of Subject-Adverb-Verb-Object (SAVO) word order in French by
native speakers of English.

I limit myself to adverbs of frequency and manner and to verbs in the present
tense. While both English and French allow SVOA and ASVO word orders, the
languages differ in that only French allows SVAO order, and only English allows SAVO
order.2 Importantly, if one assumes lack of negative feedback, nothing in the French
positive input informs learners that the SAVO order allowed in their native language is
disallowed in the target language.

Volunteers were obtained from four sections of French 150 at Michigan State
University. All volunteers (45 students) took three pretests: a Word Order Correction
(WOC) Task, an Acceptability Judgment (AJ) Task, and Raven’s Standard Progressive
Matrices Test. The WOC and AJ tasks were designed to test students knowledge of
adverb placement in French; Raven’s Matrices tests inductive ability.

The Word Order Correction Task was in the form of a cartoon story. Students
were asked to correct those sentences which are incorrect. In the Acceptability Judgment
Task, students were asked to decide whether both sentences are correct, both sentences

are incorrect, only one is correct, or they do not know. Raven’s Progressive Matrices

 

2 It must be noted that there are exceptions to this characterization of the two languages with respect
to adverb placement. In French, it is possible to place the adverb before the verb for emphasis. For
example, Marie, SOUVENI’, oublie ses devoirs. This emphasis is more common in spoken discourse than
in written discourse and will not occur in the first and second person singular. With respect to the
characterization of English, it is possible to place the adverb after the verb if the direct object is a ’heavy
NP.’ For example, I notice frequently that you are right. However, the focus of concern in this study is
basic word order and not uncommonly used constructions.

8

Test is a non-verbal test which is considered to be a valid and reliable test of inductive
reasoning. It consists of five sets of 12 problems each presented in abstract
figures/designs and combines elements of analogies, series completions, and
classifications. Students were allowed unlimited time to complete all tasks. The order
in which tests were taken was alternated both within a testing session and for each
student.

Volunteers were divided into two groups based on class section. Teachers were
asked not to correct students with respect to adverbs, nor were they to teach students the
rule of adverb placement. For a 2-week period, in group 1, teaching materials
containing adverbs were integrated into the teachers’ regular lesson plans. In group 2,
in addition to adverb ’flood,’3 students received lessons in inductive inferencing drawn
from Carton’s Method of Inference in Foreign Language Study and from Sternberg’s
(1988) The Triarchic Mind. These exercises were given in both French and English and
consisted of exercises in analogy, series completion, and classification.

Immediately following the instructional period, each group was posttested using
the Word Order Correction Task and the Acceptability Judgment Task. Presumably, as
Raven’s Test is a measure of ability, the score on this test will not change over a two-
week period. Previous studies with this test have shown that even when subjects are
coached on the diagrams specifically, scores do not change. Once again, my assumption

is that performance usually does not match ability; the goal, therefore, is to try to tap

 

3 The goal was that students would be exposed to positive evidence only of adverb placement in
French. Adverbs of frequency and manner were integrated into the regular lessons wherever possible.
Relative to the infiequency with which adverbs occur naturally in classroom discourse (Trahey & White,
1993), students are ”flooded" with adverbs with this manipulation of the exercises.

ability.

Three weeks after the first posttest, students were posttested again to determine
any long term effects of the teaching material.

I present the results of the study in Chapter 4. Two scores for each student were
computed for the Word Order Correction Task and for the Acceptability Judgment Task.
One score represents an SAVO error score and consists of the number of acceptances of
or changes to this ungrammatical order. The other score, the SVAO correct score,
represents the number of acceptances of or changes to this grammatical order. Repeated
measures ANOVAs and Multiple Regressions of these test scores were performed to
determine the change over time in addition to the main factors and interactions of the

following:

Within Groups

Word Order Correction Task - pretest and posttests performance
Acceptability Judgment Task--pretest and posttests performance
Word Order Correction Task pretest and Raven’s pretest

Word Order Correction Task posttests and Raven’s pretest
Acceptability Judgment Task pretest and Raven’s pretest
Acceptability Judgment Task posttests and Raven’s pretest

9999392"

Across Groups
Word Order Correction Pretests

Word Order Correction Posttests
Acceptability Judgment Pretests
Acceptability Judgment Posttests
Raven’s Pretests

99959:”

The following variables were taken into consideration: age, gender, previous years of
French study, adverb input, and inferencing input. These latter two variables are
described below.

An attempt was made to quantify the adverb input which the students received.

10
After the instructional period, each teacher filled out a questionnaire and described how
each lesson was taught: as a class (1), group (2), pair work (3), individually (4), or
homework (4). In this way, an adverb input score was calculated for the 2-week
instructional period for each class. If a student was absent or did not do a homework
assignment, this was deducted from her or his adverb input score. The assumption
behind this particular quantification is that if a student must do an exercise individually
(or as homework), she or he is forced to concentrate on the task at hand. Whereas, in
a teacher-fronted (or class) activity, conceivably nine out of ten students could be ’tuned

9

out. Importantly, this scale is meant to represent the potential adverb input; I am

making no claims as to the actual "apperceived input" (Gass, 1988).

Similarly, a quantified inferencing input score was given to each student in the
inference groups. While inferencing was treated as the only nominal variable, a way was
needed to take into account the effect of an absence on the ’amount’ of inference
instruction.

To summarize, the present research is based on the following
arguments/assumptions:

1. Negative evidence appears to be of little use in SLA, because 1) it is not
consistently provided to second language learners, and 2) when it is
provided, it seems to be ineffective in altering learner language in the long
term.

2. The use of Indirect Negative Evidence is similar to the process of

inductive inferencing.

Inductive inferencing is a ’reasoning’ process used in SLA.

Inductive inferencing is a learning strategy which can be taught.

Word Order Correction and Acceptability Judgment Tasks are valid

instruments for measuring students knowledge of a specific linguistic

structure.

6. Raven’s Standard Progressive Matrices Test is a valid instrument for
measuring an individual’s eductive abilities.

9P.“

11
7. In order for English speakers to learn the ungrammaticality of SAVO
word order in French, they must notice the absence of this structure; this
can be achieved through the process of inductive inferencing.

The dissertation concludes with Chapter 5, a discussion of the results and
implications of the present research. The review in this dissertation of both empirical
and theoretical studies reveals that the Learnability Problem continues to be substantiated
but we are seemingly no closer to a solution, regardless of the perspective one chooses.
The present research calls for a reevaluation of the approach to the Problem.

Specifically, Indirect Negative Evidence may prove to be a partial solution.
Arguments from L1 and L2 acquisition researchers support the view that such a role is
not theoretically out of the realm of possibility. These arguments can be corroborated
with examples from the biological sciences which show that a course of development may
indeed be determined by the absence of a specific environmental factor.4 It is not
unreasonable to assume that humans, as members of the biological community, would

also employ a mechanism which not only ’apperceives’ in general but must ’apperceive’

the absence of information. I suggest that this process is at the core of inductive

 

‘ While a defense of analogical reasoning goes beyond the scope of this dissertation, analogizing is
not without precedence among language acquisition researchers (see Lightfoot, 1989, in addition to Critical
Period studies). Examples of the use of ’indirect negative evidence’ can be found in the fields of
microbiology and ethology, among others. "The immune system must not let itself be stimulated to produce
antibodies” against its own antigens (Jerne, 1967, p. 201). That is, an organism expects’ the presence of
similar antigens and in the absence of similarity, the organism will respond. Suppose an antigen is
introducedintoananimalandthatantigenis similartotheanimal’ s, thenantibodieswill notbeproduced.
If, however, the antigen is not similar (that rs, it is absent 1n the animal’s repertoire), it will be recognized
as foreign, and antibodies will be produced. The use of indirect negative evidence can also be found in
ethology. In crocodilians, the temperature of egg incubation determines sex. If the temperature is 30°C,
100% females will hatch; at 33°C, 100% males will hatch; temperatures in between produce varying ratios
of male to female. If the crocodilian population is low, females will nest close to the water where the
temperature tends to be lower. As a result, more females will be produced and the population will rapidly
increase. Under normal conditions, females will nest both near the water and on higher ground (Deeming
& Ferguson, 1989). Thus, given the ’expectation’ of a stable population, the absence of an element
(crocodilians) has resulted in an alternative behavior pattern (preference to nest near the water).

12

reasoning.

The results of the present research may provide further theoretical and possibly
empirical support for a role of INE. Specifically, the implications of this investigation
are that I) second language learners are indwd hypothesis-testers--once INE is taken into
account, the theoretical need for negative evidence is eliminated; 2) the process of
inductive inferencing, which is not solely linguistic in nature, is used in second language
learning; and 3) SLA can be facilitated through the use (teaching/learning) of inductive
inferencing.

The study of language acquisition attempts to explain a system which produces
an output which exceeds the input but which in another sense is responsive to the input.
The system must make maximum use of the input. By extension, if we think of
ourselves as generating an ’output’ (SLA theory) based on the ’input’ (empirical
observations) plus ’principles’ of reasoning, we should consider in our hypothesis testing

all the input available which includes Indirect Negative Evidence.

CHAPTER 1
THE LEARN ABILITY PROBLEM

1.1. POSITIVE EVIDENCE

The Learnability, Logical, Projection, Plato’s and Poverty—of-the-Stimulus
Problems all refer to the question asked by Chomsky in 1987 (cited in Cook, 1988, p.
55): "How do we come to have such rich and specific knowledge, or such intricate
systems of belief and understanding, when the evidence available to us is so meager?”
Two problems with the evidence deal with the positive evidence available to the learner:
(1) it underdetermines the final competence of the learner, and (2) it is often degenerate.
White provides

a number of examples [wanna-contraction, distribution of complementizers, constraints
on wh-movement, pronouns, and parasitic gaps] where the input might be expected to
mislead the child into making false generalizations about the language being learned, or
where it might fail to allow the child to work out various subtle phenomena. . . . Despite
the problem of underdetermination, children arrive at the full complexity of adult
knowledge with comparatively little difficulty, and without the range of errors that one
might anticipate. (White, 1989, p. 11)

Here, I cite just one example, that of wh-movement:

(1—1) What did John see _?

(1-2) What did Mary believe that John saw _?

(1-3) What did Jane say that Mary believed that John saw _?
(14) *What did Mary wonder whether John had bought?

(1-5) Mary wondered whether John had bought a present.

(1-6) What did Mary hope that John had bought?

(1-7) *What does Mary believe the claim that John saw?

(1-8) Mary believes the claim that John saw a ghost.

(1-9) What does Mary believe that John claimed to have seen?

(White, 1989, pp. 8-9)

13

14

Based on the questions in 1-1, 1-2, 1-3, 1-6, and 1-9, if children were merely
generalizing from the input, they would be misled into forming the ungrammatical
questions in 1-4 and 1-7. The fact that the questions in 1—4 and 1-7 are not produced is
taken as evidence that children are not forming generalizations from the input.
Additionally, the ungrammaticality of 1-4 and 1—7 cannot be attributed to the declarative
sentences, as 1-5 and 1—8 are indeed grammatical. Furthermore, White points out that
semantics or pragmatics cannot explain the ungrammaticality of 1-4 and 1-7, since 1-6
and 1-9 are grammatical.

The second learnability issue dealing with the input is that of degeneracy. This
refers to the fact that sentences a child may hear, while acceptable, may not be
grammatical. For example:

(1-10) *Want your lunch now?

(1-11) *Going out to play?

(1-12) *Raining again.
Given such input, the child would make the false assumption that English is a null-subject
language.‘ However, the competent speaker of English knows that subjects are
obligatory in English. Even though there is some debate whether learners do in fact
receive degenerate input,2 it seems likely that they would be exposed to sentences such

as 1-10, 1-11, and 1-12 in a naturalistic setting.3 We therefore see that based on positive

evidence alone, it is theoretically impossible for learners to become fully competent in

 

‘ However, see Roeper and Weissenbom (1990, p. 159) for the proposal that ”the subordinate clause
[is] the locus of a unique trigger."

2 Snow and Ferguson (1977) and Ellis (1986) cited in White (1989).
3 As cited in Larsen-Freeman and Long (1991, p. 291): "Michael Montgomery (personal

communication) has pointed out that some varieties of English, c. g. varieties of Southern American
English, do in fact allow dropping of subject pronouns, especially in informal registers.”

15

a language because there is a mismatch between the input and the final grammar attained.
1.2. NEGATIVE EVIDENCE IN FIRST LANGUAGE ACQUISITION

Lack of negative evidence available to the learner constitutes the third problem
with the input and is the area that has attracted the most attention and remains the most
controversial. Negative evidence refers to the corrective feedback supplied to the learner
which allows the learner to arrive at intuitions regarding (un)grammaticality. Because
of its influence on SLA research, it is important to look at the evolution of the L1
studies, both theoretical and empirical, dealing with negative evidence. As cited in
Schachter (1991, p. 92), ”it was standardly assumed after Chomsky (1965) that children
constructed a grammar by using input together with innate linguistic knowledge to
formulate hypotheses about possible grammatical rules, test them out on further input,
holding, revising or abandoning them as necessary." The emerging significance of
negative evidence can be seen in 1) Braine’s (1971-written in 1967) critique of
Chomsky’s hypothesis-testing model of language acquisition and 2) Gold’s (1967)
mathematical proof showing that a wider range of languages are more learnable if the
information presented to the learner includes both positive and negative evidence.4 In

Braine’s article he argues:

Since it seems clear that a great deal of negative information must be included in the
input to a reasonable hypothesis-testing model, let us consider whether the child obtains
and uses such information. Several lines of evidence suggest that negative information
cannot be necessary for first language acquisition. First and foremost, there is the

 

‘ Gold’s paradigm has since been criticized (Gordon, 1990; Bohannon, MacWhinney, & Snow,
1990) on a mrmber of grounds: 1) Gold’s assumption that the language learner is a ‘tester’ may not be an
appropriate characterization; 2) his dissociation of learning a language from learning the lexicon of the
language; 3) his assumption that language learning involves the entire grammar simultaneously; 4) his
assumption that once a language is ’chosen,’ that choice never changes: and finally 5) Gold’s paradigm
does not guarantee language acquisition in a specified amount of time, a factor which must be taken into
account in a model of language acquisition.

16
universality with which language is acquired at a fairly rapid rate . . . despite a wide
variety of cultural conditions and child-rearing practices. . . . When one considers the
difficult circumstances under which huge numbers of children are raised, it becomes clear
that correcting their infants’ speech must be among the least of the concerns of very
many human adults. . . .

Another fact worth noting is that it would often be difficult for even the
best-intentioned parent to know precisely what correction to make in response to many
errors. . . .
Regardless of the extent to which corrections actually occur, there is little to

indicate that young children are able to use such information when it is given, even in
apparently simple cases. (Braine, 1971, p. 159-160)

Braine (1971, p. 161) argues "that children acquire grammar without benefit of the

negative information that seems required by a hypothesis-testing theory . . . An early
demarcation between those who argue either for or against the existence and necessity
of negative evidence is thus created.

The research on negative evidence has continued to center around three major
issues, the first of which is: Do children make relevant errors? Relevant in the
following sense: It has been argued that positive evidence alone (1-13, 1-14, and 1-15
below) would cause the child to formulate the incorrect generalization in 1-16 (the
underdetermination issue) and that overt correction would then be necessary to eliminate
this structure from the child’s grammar.

(1-13) Who do you think that Mary met yesterday?

(1-14) Who do you think Mary met yesterday?

(1-15) Who do you think arrived yesterday?

(1-16) *Who do you think that arrived yesterday?

(White, 1989, p. 8)

The position that the learner does not make relevant errors, as in 1-16, and
therefore negative evidence is not necessary for the learner to attain competence has been
developed by several researchers on purely theoretical grounds (with the exception of

Otsu’s 1981 empirical study, cited in White, 1989). Baker (1979, p. 535) argued for "a

grammatiml theory that is much more restrictive . . . in the range of grammatical rules

17

that are permitted” so that "children must be innately programmed in such a way that
they will only develop hypotheses that, if incorrect, can be corrected subsequently on the
basis of positive evidence alone" (Schachter, 1991, p. 93). In the same vein, Dell (1981)
proposed a learning procedure whereby the learner’s first choice will always be the
unmarked case, which will always be the subset grammar; therefore, the superset or
marked grammar will never be hypothesized unless positive evidence warrants it. Thus,
negative evidence is not necessary.

The second line of inquiry dealing with negative evidence focuses on whether
corrective feedback is provided to the learner. This research naturally becomes more
complex and convoluted because it also involves an attempt to describe the nature of the
feedback and when it occurs. Perhaps the most cited study is that of Brown and Hanlon
(1970) in which they investigated the responses of caretakers to the incorrect and correct
usage of syntactic rules by 3 children. They concluded that caretakers do not correct
grammatical mistakes but do provide overt correction if the truth value of the utterance
is incorrect. Hirsh-Pasek, Treiman and Schneiderman (1984) and Penner (1987)
concluded that mothers tended to repeat their children’s ungrammatical utterances more
frequently than their grammatical utterances and that these differential adult responses
to children’s speech decreased as children developed. A study by Demetras, Post and
Snow (1986) concentrated on the existence of and the type of negative feedback provided
by mothers. These authors concluded that explicit negative feedback (Yes, that ’5 right;
No, that’s wrong) is infrequent and when it is provided, it is directed toward the semantic
content of the child’s utterance, supporting the Brown and Hanlon study of 1970. In

contrast, however, implicit negative feedback (repetitions, clarification questions,

18

move-ons’) was provided differentially to well—formed and ill-formed child utterances.
Contracted and extended repetitions followed ill-formed utterances more often than other
forms of implicit negative feedback. Given the inconclusiveness of the results, these
studies have been used as support for both the ’no negative evidence is provided’ claim
(Gordon, 1990; Larsen-Freeman & Long, 1991; Marcus, 1993) and the ’negative
evidence is provided’ claim (Bohannon & Stanowicz, 1988; Bohannon, MacWhinney, &
Snow, 1990).

Bohannon and Stanowicz (1988) attempted to expand on the above mentioned
studies by l) investigating both parent and nonparent adult conversations with children
(previous research had been limited to parent-child dyads), and 2) examining differential
adult responses to the type of error (semantic, syntactic, or phonological).

Results indicated that adults were more likely to repeat verbatim a well-formed sentence
than an ill-formed sentence. In contrast, adults were more likely to repeat with changes,
or request clarification of, a sentence containing syntactic or phonological errors than
well-formed sentences. . . . (Bohannon & Stanowicz, 1988, p. 684)

It should be noted that the repetitious contrasts observed in our study are not
"negative evidence" in the strictest sense. . . . the information contained in adult
repetitious sequences is much greater than that in simple denials. A denial informs the
learner that the prior response was, in some way, incorrect. It neither informs the
learner about the nature of the error nor of correct alternative ways to express the same
meaning . . . The process observed in the current study offers language learners not only
simple error detection but also information as to correct (or at least acceptable)
alternatives. In this light, repetitious feedback by adults may be considered superior to
simple denials and as qualified to assume the role of ’negative evidence’. (Bohannon &
Stanowicz, 1988, pp. 688)

In response to Gordon (1990, p. 217), who argues that the Bohannon and

Stanowicz (1988) results "tell us nothing about learnability or innateness," Bohannon,

 

’ Repetitions were subdivided into exact, contracted, expanded, and extended. Contracted and
expanded repetitions provided syntactic or morphological correction; contracted repetitions were shorter
than the child’s utterance, while expanded repetitions were longer than the child’s utterance. Extended
repetitiom admd new information to the child’s utterance. Move-ons were ”responses that continued the
conversation or that began a new topic, without either questioning or repeating (parts of) the child’s
preceding utterance" (Demetras et al., 1986, p. 279). Note that these authors’ use of the term ’implicit
negative feedback’ should not be equated with recent definitions (Saleemi, 1990, 1992).

19
MacWhinney and Snow (1990, p. 221) maintain that "the rates of feedback found [in the

1988 study] are . . . sufficient to spur learning in many species, including concept
formation tasks in humans . . . [and] that using innate factors as a ’default’ explanation
is a dangerous and counterproductive scientific endeavor. "

Marcus (1993, p. 56) disputes the claims made by researchers (Bohannon &
Stanowicz 1988; Demetras, Post, & Snow 1986; Hirsh-Pasek, Treiman, & Schneiderman
1984) who maintain that children do indeed receive negative feedback. Marcus reviews
these studies and finds that "in all cases, the differences in parental replies to
grammatical versus ungrammatical sentences were statistical rather categorical . . . [He]
calls this type of feedback ’noisy feedback,’ since parents provided each type of reply
after both grammatical and ungrammatical sentences, albeit in different proportions. "

Marcus contends:

—a
C

Noisy feedback is too weak to be a plausible way of eliminating errors;

2. Noisy feedback is inconsistent across parents, declines or disappears with age, and is
probably not provided for all types of errors; and
3. Because parental reply categories are defined only with respect to the child’s utterance,

"correlations" between the two may be artifacts resulting from the definition of parental
reply categories and of constant noncontingent properties of parental and child spwch.
(Marcus, 1993, p. 57)

Marcus concludes that since there is no evidence showing that noisy feedback, if it exists,
is necessary for language acquisition, "internal mechanisms are necessary to account for
the unlearning of ungrammatical utterances" (p. 54).

The evolution of the negative evidence studies is of particular interest for the
purposes of the present research because a redefinition of negative evidence, and in fact
what could be considered INE, emerges from their results. In the Bohannon and

Stanowicz (1988) study, the fact that all adults tended to use recasts slightly more

20

frequently than other types of responses after ill-formed utterances deserves mention
because this type of response could, in theory, provide the child with evidence of what
is not possible by means of the absence of certain structures in the recast----in other
words, indirect negative evidence. Similarly, Demetras et al (1986) found contracted and
extended repetitions to frequently follow a child’s ill-formed utterance. While extended
repetitions may be considered positive evidence, contracted repetitions, which shorten the
child’s utterance, may be considered indirect negative evidence of what is not possible
in the grammar by virtue of its absence in the adult response.6

The third line of inquiry dealing with negative evidence - do children make use
of negative evidence? - has received much less attention in first language acquisition than
the first two issues and of course becomes a moot point if one answers ’no’ to the first
issue - Do children make relevant errors? ”Even when it [correction] is provided, it is
typically resisted, as many parents will readily attest. McNeill (1966:69) recorded a

celebrated illustration of this resistance" (Lightfoot, 1989, p.323):

Child: Nobody don’t like me.

Mother: No, say "nobody likes me."

Child: Nobody don’t like me.

(eight repetitions of this dialogue)

Mother: No, now listen carefully; say "nobody likes me.”
Child: Oh, nobody don’t likes me.

A similar example is provided by Braine:

Child: Want other one spoon, Daddy.

Father: You mean, you want THE OTHER SPOON.
Child: Yes, I want other one spoon, please, Daddy.
Father: Can you say ”the other spoon?"

 

6 Marcus (1993, p. 55) "collapses" indirect negative evidence with positive evidence, stating that
”INE depends on a reanalysis of positive evidence based on mechanisms internal to the child, rather than
input external to the child." Crucially, however, there are certain structures which are absent in the
positive evidence which the child reanalyzes. Thus, INE cannot, in the strict sense, be considered the same
as positive evidence.

21

Child: Other...one...spoon.

Father: Say ”other."

Child: Other.

Father: "Spoon."

Child: Spoon.

Father: "Other...spoon."

Child: Other...spoon. Now give me other one spoon?
(Braine, 1971, pp. 160-161)

Even though the data "that young children do not use negative information is indirect or
anecdotal" (Braine 1971, p. 161), these anecdotal comments are supported to some extent
in the psychology of learning literature. As cited in Schachter (1991, p. 95), Levine
(1975) found ”that younger children are more likely to exhibit a response set hypothesis,
i.e., to act as though a certain hypothesis still held immediately following disconfirmation
of that same hypothesis, indicating that they are less likely to take negative data into
account." While there are still those, such as Braine, who reject hypothesis-testing
models of language acquisition altogether, Chomskyan linguists do receive a certain
amount of support from the field of psychology.
1.2.1. Summary

Thus far I have explained the Learnability Problem of first Language Acquisition,
discussing in relative detail the issues surrounding the negative evidence available to the
learner. These issues include: 1) Do children produce utterances which theoretically
require overt correction in order to be eliminated from their grammar? 2) Is overt
correction consistently provided to children? And, 3) Do children make use of overt
correction? I have reviewed the arguments made on both sides of these issues, which are

yet to be resolved.

22
1.3. NEGATIVE EVIDENCE IN SECOND LANGUAGE ACQUISITION

The role of negative evidence becomes much more complex in SLA mainly due
to a proliferation of variables that could be influencing the effects of negative evidence
and because of the methodological difficulties in investigating those variables. Allwright
(1975) was one of the first to address this problem and attempted to categorize error
types, treatment types, and learner variables (e.g. L1, cognitive style, affective factors)
which may influence the learner’s interpretation of the negative feedback given in an
instructional setting.

Long (1977) continued this line of inquiry in a review of ”descriptive studies of
the classroom behaviours of teachers following learner error in an attempt to ascertain
what the teachers studied currently do when providing feedback" (p.278). Long
discovered a ”marked lack of clarity and consistency in teacher feedback,” leading him
to question a hypothesis-testing process of language acquisition.

A year earlier, Chaudron (1977) conducted an empirical study of French
immersion classes for English speaking students in an attempt to isolate and measure
which corrective techniques led to successful correction of errors. Chaudron concluded
that the results suggested a positive relationship between a form of correction which he
categorized as Repetition with Reduction (8: ”Le maison est jaune. " T: "La maison") and
successful correction by the student of errors mainly in phonology, syntax, and content.
(Other types of errors included morphology and lexis).

In Chaudron 1986 (a continuation of the 1977 immersion study), the goal was to
determine ”how much are learners’ L2 linguistic errors corrected in either French or

other subject classes, relative to errors of other sorts” (p.65)? Results indicated that

23

linguistic and content errors were corrected with approximately the same frequency
(100% and 95% respectively) in French class, whereas in other subjects linguistic errors
were corrected much less frequently than content errors (37% and 88%). It should be
noted that the frequency of correction did not remain constant over time. In French class
the percentage of linguistic errors which were corrected dropped from 95 % to 66% and
the percentage of content errors which were corrected dropped to 75 %; this was
attributed to an increase in student participation. Moreover, the percentage of linguistic
errors corrected in other subject classes remained at 37%, while the percentage of
content errors corrected increased to 96%.

Chun, Day, Chenoweth, and Luppescu (1982) shifted the research focus to an
examination of native speaker (NS) corrections of non-native speaker (NNS) errors in
non-instructional settings. Results indicated that NSs rarely corrected errors (8.9%) made
by their NNS friends. Additionally, ”errors of a factual nature were most often corrected,
followed by discourse and vocabulary corrections. Grammatical errors were seldom
corrected” (p.545).

Brock, Crookes, Day, and Long (1986) looked at what types of NNS errors led
to what types of NS responses and the subsequent effect of these responses on the speech
of the NNS in informal conversations. This study examined errors in lexis, phonology,
and morphosyntax (content errors were not included). Results indicated that

Phonological errors did not appear to be related to a particular NS turn. . . . Lexical
errors [were] more likely than others to trigger a side sequence in which an attempt is
made to clarify the message. Morphosyntactic errors, perhaps because of their lesser
communicative significance, [were] more likely to permit the main line of discourse to
be continued. . . . Few effects of the NS response were observed on subsequent NNS
conversation. (Brock, Crookes, Day, & Long, 1986, pp. 233-234)

The above mentioned studies are not without caveats, and while

24

overgeneralization may be dangerous, a brief comparison of the results may be of some
use. Not surprisingly, a common finding in the studies that examined errors of content
(3 of the 4 mentioned) is that teachers and NS speakers, in general, more often
responded to these error types than to other types of errors. (Interestingly, this is
consistent with the L1 studies which find parents more often react to the truth value of
a child’s utterance.) However, even teacher correction of content errors did decrease
over time (Chaudron, 1986). Additionally, the correction of errors of any type
represented an insignificant percentage of all interactions in the Chun, et al. study.
Furthermore, while only two of the studies attempted to examine the effect of correction
on subsequent NNS speech, one study (Brock, et al.) found no effect and the other
(Chaudron, 1977) found only one form of correction to be effective. Even though the
past studies mentioned here have been useful and have usually been used as support for
the nativist position that learners do not receive negative feedback (and, therefore,
correction plays a minimal role in acquisition), the results are mixed and relatively
uninforrnative with respect to providing a clearer understanding of the effects of negative
evidence in SLA. The effectiveness of corrective feedback on learner’s language has
recently become the focus of several studies.

In a series of studies, White (1991a, 1991b, 1992a), working within a
Government and Binding framework, looked at the effect over time of positive and
negative evidence on a proposed parametric difference between French and English, that

is, whether or not the language allows verb movement.’ The studies consisted of

 

7 My discussion of this work and the responses to it will focus specifically on the issue of evidence
and not on the implications on the ’contents’ of UG or the viability of Pollock’s Verb Movement
Parameter.

25

Francophones learning English as a Second Language. Students were divided into two
groups; one group received form-focused instruction, which included both positive and
negative evidence, on adverb placement; the other group received instruction only on
question formation.8 Subjects were given a pretest, two post-tests (one immediately
following instruction and another five weeks later) and a follow-up test one year after
instruction. In both studies, the group that received both positive and negative evidence
exhibited knowledge of adverb placement on the two post-tests, while the group that
received only positive evidence did not. However, the results on the follow-up test (one
year later) were the same for both groups. That is, the group that received form-focused
instruction did not retain knowledge of adverb placement. One possible explanation
suggested by White (1991a, p. 159) is "lack of suitable follow up or subsequent emphasis
on this issue.” The result that "exposure only to positive input [is] insufficient” (White,
1991a, p. 158) is consistent with studies conducted by Tomasello and Herron (1988,
1989).

From the results of their ’Garden Path’ investigations, in which learners are
guided into making errors of overgeneral ization and transfer, these authors concluded that
”students learn best when they produce a hypothesis and receive immediate feedback
because this creates maximal conditions under which they may cognitively compare their

own system to that of mature speakers" (Tomasello & Herron, 1989, p. 392).9

 

8 Assuming question formation to be one of the cluster of properties affected by the verb-raising
parameter, White considers this group to be receiving only positive evidence with respect to adverb
placement (White I992a).

9 Beck and Eubank (1991) criticize the Tomaser and Herron experiments on both theoretical and
methodological grounds. Given that it has not been shown that children consistently receive or make use
of negative evidence, Beck and Eubank object to Tomasello and Herron basing their work on a theory of
L1 learning which requires negative evidence. Beck and Eubank also argue that the study (1989) suffers

26

Trahey and White (1993) as follow-up to the White studies (and in partial
response to Schwartz and Gubala-Ryzak, 1992 [see below]) investigated "whether
preemption operates in L2 acquisition, that is, whether (naturalistic) positive L2 input
which is incompatible with the L1 parameter setting is sufficient to force parameter
resetting" (Trahey & White, 1993, p. 183). Learners were exposed to an input flood of
materials containing English adverbs during a 2-week instructional period. Subjects were
posttested immediately following instruction and again 3 weeks later. Results indicated
that ”francophone learners of English responded to properties of the L2 input flood:
without explicit instruction, learners’ use of English SAVO order (prohibited in French)
increased dramatically" (Trahey & White, 1993, p. 200). However, SVAO word order
(consistent only with the French value of the parameter) was not expunged from the
learners’ interlanguages. The authors concluded that the L1 setting was not preempted
by the L2 setting (failure of the Uniqueness Principle) and suggest that the length of time
of the input flood was insufficient.

White (1991a, b) and Trahey and White (1993) draw two conclusions (among

others) from their studies which are relevant to the present research:

 

from such flaws, among others, as: 1) Subjects’ prior linguistic knowledge was not taken into account (by
means of a pretest); 2) There was variability in the elapsed time between treatment and testing of the
various structures: and 3) Only one test - a translation task - was used; "it is plausible to assume that the
results...will not generalize to other types of performance" (Beck & Eubank, 1991, p. 75).

27

1. Negative evidence is effective in helping L2 learners to master the fact that SVAO is
ungrammatical in English while positive evidence is insufficient. '° (White, l991a, p.
158)

2. The flood might simply have been imufficient for preemption to occur. (T rahey & White,
1993. p. 201)

With respect to the first conclusion, I would argue that given the research findings
which strongly indicate that negative feedback is not consistently available to or utilized
by language learners, a conclusion which suggests that negative feedback may be
required would imply that ungrammatical structures (in this case SVAO word order in
the English of francophones) are usually never eliminated from the learners’ grammars.
This is certainly not the case, therefore one is still left with the need to explain how those
learners who are not exposed to negative feedback manage to learn the ungrammaticality
of a structure in the target language.

With respect to the second conclusion, even if research were to determine that X
amount of positive input is necessary to ’switch’ parameters, this still does not explain
how or why the parameter is reset.

From a theoretical perspective, Schwartz and Gubala-Ryzak (1992) have argued
against White’s claim that explicit positive evidence and negative evidence have a short—
term effect and can lead to parameter resetting in SLA. The authors explain that

within a Principles and Parameters approach to L1 acquisition of grammatical knowledge,
there are three necessary and sufficient components for attaining a steady-state grammar:

a) Universal Grammar
b) a learning procedure

 

'0 Schwartz and Gubala-Ryzak (1992) contend that "in order to maintain the claim for a need for
negative evidence to reset parameters in L2 acquisition, White must implicitly assume that a learner can
have both values of the Verb Movement Parameter operating in the interlanguage at the same time" (White,
1992b, p. 124). These authors maintain that only one setting of a parameter can be maintained at a time
(Uniqueness Principle). Citing empirical studies (White 1991a, I99Ib, I992a) and theoretical arguments
(Berwick, 1985; Cook, 1991; Valian, 1990a), White (1992b, p. 125) contends that in both first and second
language acquisition, learners may "entertain more than one value of the parameter at a time. "

28
c) (contextualized) input

They further propose that there are three types of input available to second (and first)

language learners:

1. Primary Linguistic Data (PLD), consisting of (contextualized) utterances in the ambient
language;

2. Negative Evidence (NE), consisting of information about the impossibility of a form or
utterance;

3. Explicit Positive Evidence (EPE), consisting of descriptive information about the

language. (Schwartz & Gubala—Ryzak, 1992, p. 2)
"The basic idea is that PLD alone are used by UG in combination with the learning
procedure to create a grammar" (Schwartz, 1993, p. 151).

Citing work such as Cohen and Robbins (1976), which indicates that ”in the face
of seemingly ample ND [negative data], to which L2ers have often been repeatedly
exposed, sometimes even consistently and sometimes even over years and years, L2ers’
hypotheses seem resistant to revision” (Schwartz, 1993, p. 149), Schwartz and Gubala-
Ryzak (1992, p. I) maintain that, at the level of syntax, ”the grammar-building process
cannot make use of negative evidence to restructure (lnterlanguage) grammars--
irrespective of logical need." These authors contend that only PLD (as in L1 acquisition)
can reset the Verb Movement Parameter and that the short-term effects found in White’s
studies are the result of changes in the subjects’ linguistic performance but not their
linguistic competence. Moreover, the positive evidence in the White studies was not of
the ”appropriate” kind;“ that is, "the form-focused instruction . . . was too explicit to

tap unconscious parameter setting mechanisms" (Trahey & White, 1993, p. 187).

 

“ Schwartz and Gubala-Ryzak offer examples of the positive evidence necessary to reset the
parameter such as:
Not to completely wash the soap out of my hair is a nuisance.
John believes Mary not to usually beat her brother at chess.
White (1992b, p. 127) points out that such ”data are margiml and obscure, not at all likely to occur in the
primary linguistic input to the learner, a problem Schwartz & Gubala-Ryzak recognize."

29

Furthermore, following Fodor’s (1983) modular theory of mind, Schwartz and Gubala-
Ryzak (1992) propose that the linguistic knowledge which results from NE and EPE
cannot serve as input to the language acquisition process.

Schwartz ( 1993, p. 153) offers the following analogy:

Take a bearded man who moves to a foreign place. Unbeknownst to him, all men must
be clean shaven in this culture. He soon discovers this custom, and he is the type that
tries to fit in as best he can in other cultures. Luckily for him, he has the solution: He
(coincidentally) brought an electric shaver along, so logically the shaver is what is
necessary to solve his problem. He goes to his hotel and plugs it in, but nothing
happens. The current is different. It turns out that a transformer is needed to make the
shaver work. Even though the shaver is necessary, it cannot be used.

The crucial step in this cross-cultural experience, which Schwartz ignores, is that
the bearded man notices the absence of beards in the new culture. He then compares this
with his own appearance and realizes that his facial features are in need of alteration.
In essence, the shaver is not necessary at all; he can use a straight-edge. What is
necessary is that the man initially recognizes the mismatch between the look he is
sporting and that of the men in his new culture (see Chapter 2).

Schwartz (1993) elaborates on the issue of evidence and makes three claims which

I would like to address:

PLD alone can "trigger" UG in grammar construction; (p. 158)

An approach to building linguistic knowledge that assumes general (hence, non-domain-
specific) hypothesis-testing procedures cannot explain why ND (and explicit data) do not
work more fruitfully than they do in L2A; (p. 159)

3. Negative Data [comists] of explicit and perhaps implicit [italics added] information about
the impossibility of a form, utterance, or sentence. (p. 151)

Nr—n

Regarding the first claim, as mentioned above, Schwartz (1993) and Schwartz and
Gubala-Ryzak (1992) maintain that the only ’appropriate type’ of input which can engage
the grammar-building process is PLD. One immediately wonders what it is about PLD

which makes it appropriate. To this Schwartz replies (1993, p. 158) that ”the only thing

30

we know about what ’appropriate type’ means is that PLD necessarily must be it, since
PLD are what build L1 grammars.” PLD has been defined as utterances in the ambient
language. I would argue that we are still no closer to a definition of what ’appropriate
type’ is or to a clearer characterization of PLD. Furthermore, I would suggest that the
appropriateness of the input can only be defined relative to the specific second language
learner: what may be too explicit/inappropriate for one learner may be not explicit
enough for another. In other words, attempting to find some intrinsic feature of the input
which renders it ’appropriate’ may be a futile effort. Rather, I suggest that it is features
of the input in conjunction with learner internal variables which result in
’appropriateness.’ This leads to my argument with the second claim which is closely tied
to the flaw in the third claim.

Leaving aside the issue of UG, let us assume that there are at least two necessary
components to SLA: l) a learning procedure, and 2) contextualized input. Let us further
assume that this learning procedure is responsive to the environment, or the
contextualized input. If we accept the research which suggests that negative evidence is
frequently not available to language learners, it is necessary then for this learning
procedure to adapt and function in the absence of negative evidence. If the use of
Indirect Negative Evidence (Chapter 2) is included as an integral part of the (hypothesis-
testing) learning procedure, the theoretical need for negative evidence disappears, and
it therefore becomes a moot point to "explain why ND and EPE do not work more
fruitfully." That is, they do not work, because they are not needed.

Finally, Schwartz (1993, p. 151) mistakenly conflates ”explicit and implicit

information about the impossibility of a form, utterance or sentence” as Negative Data.

31

Explicit negative data is overt correction which informs the learner of what is not
possible in a given language. It is external to the learner, a part of the contextualized
input to which the learner is exposed. However, implicit (or indirect) negative evidence
is not, strictly speaking, a part of the learner’s linguistic environment in the same way
that explicit negative evidence is. In order to qualify as Negative Data, the implicit
information, or absence of a structure, must be equated with ungrammaticality. This
necessarily involves a learning strategy or procedure and must be treated as distinct from
explicit Negative Data (Pinker 1989).
1.3.1. Summary

From this cursory review of the SLA investigations into the issue of negative
evidence, several tentative conclusions can be made. First, in both first and second
language acquisition situations, it appears that language learners are not consistently
corrected. Second, when overt negative evidence is provided, it seems to be ineffective
in altering learner language in the long term. Finally, and most importantly, these
empirical studies and theoretical discussions have either ignored and/or misconceptualized
a role for indirect negative evidence. In the following section, therefore, I return to the
original definition and elaborate on the concept.
1.4. INDIRECT NEGATIVE EVIDENCE

. . . if certain structures or rules fail to be exemplified in relatively simple

expressions, where they would be expected to be found [italics added],

then a (possibly marked) option is selected excluding them in the grammar

. . . (Chomsky, 1981, pp. 8-9)

That is, the absence of a feature or features is "noticed” by the learner. In a sense,

indirect negative evidence is a misnomer in that it is not a form of indirect correction,

32

but rather it is an indirect means of letting the learner know that a feature is not possible
because it is never present in the mtg] environment. The notion of INE assumes that
a domain of expectation exists; otherwise, the absence of a feature would be insignificant
in part due to the fact that the potentially absent features would be infinite. It is
precisely innate linguistic knowledge which establishes a framework of expectations. It,
therefore, cannot be suggested that non-occurrence of certain sentences would cause the
learner to treat them as ungrammatical or that INE replaces innate knowledge (White,
1989).

Chomsky’s statement that "indirect negative evidence may be relevant" for
language acquisition actually seems to be more applicable and testable in SLA than in L1
acquisition. Even those who are adamantly opposed to a UG approach to SLA cannot
argue against the supposition that a second language learner would expect to find certain
features in the target language based on the instantiation of features in the L1 and would
notice the absence of those features. For example, on a very simplistic level, the English
speaker learning Spanish or Italian is going to notice the absence of subjects in certain
environments. In other words, it is possible to examine the issue of INE in SLA without
first making any UG-based assumptions of markedness, principles, or parameters.
Additionally, given the inconsistent results of the negative evidence/corrective feedback
studies, INE may play a larger role in language acquisition than previously suspected and

has been ignored for too long.

33

1.4.1. Indirect Negative Evidence In First Language Acquisition

Lasnik (1989), Valian (1990), and Saleemi (1990, 1992) argue for a role of
indirect negative evidence in first language acquisition from a theoretical perspective.12
While Valian concentrates on the learner as hypothesis-tester, Lasnik assumes hypothesis
testing but focuses on the structural contents of Universal Grammar. Saleemi combines
a reanalysis of the null-subject parameter with domain-specific learning procedures.

As previously stated, Lasnik uses principles of grammar and a theory of

markedness to eliminate the need for negative evidence 1h acquisition.

If the child does not have access to negative evidence . . . then Universal Grammar
presumably does not make available choices that can only be resolved by such evidence.
. . . [and] the concern is exclusively with the situation [where] the target language is a
proper sublanguage of the hypothesized language. Every structure of T is also a structure
of H, but there are structures of H that are not in T. Thus, the learner would . . . need
the information that the ’excess’ structures are impossible. (Lasnik, 1989, p. 89)

Iasnik examines null subjects, contraction, and dative alternation. The discussion
here is limited to null subjects and dative alternation. His argument proceeds as follows:
Suppose that English (-null subject) is the target language and that markedness is
irrelevant. If the child incorrectly hypothesizes +null subject, the hypothesized grammar
will contain English (the problematic situation described in the quote above). According
to Chomsky’s definition, when the child does not hear +null subject sentences, this will
be indirect negative evidence that the hypothesized grammar is wrong, and the child will
reset to the correct -null subject parameter. Even if the parameter is preset, if +null
subject is the unmarked value, the child is forced to make an incorrect overgeneralization

for English (but correct for Spanish or Italian). But again, the absence of sentences with

 

'2 Lasnik first argues that indirect negative evidence could substitute for negative evidence and then
goes on to show that markedness could often replace the need for indirect negative evidence. Here, I will
naturally be concentrating on the initial part of his argument.

34

null subjects will indirectly indicate that the guess is wrong. As with one of White’s
criticisms of indirect negative evidence, Lasnik questions under what conditions is non-
occurrence of sentences equated with ungrammaticality. Chomsky’s Avoid Pronoun
Principle (1981) states that ”every language would prefer a null subject to an overt one.
Thus, since null subjects are to be expected, their absence is relevant evidence. This
suggests, but does not require, that the grammatical rule or rules'allowing null subjects
should constitute the unmarked case" (Lasnik, 1989, p. 91). However, Lasnik suggests
that the situation be reversed. Suppose that -null subject represents the unmarked value.
Therefore, the child’s initial guess will never contain the target language (see Dell in
section ”Negative Evidence in L1"). For the child learning English, -null subject is the
first guess and will not be disconfirmed by positive evidence. For the child learning
Spanish, the initial guess will be disconfirmed by sentences lacking overt subjects. In
other words, indirect negative evidence, the absence of overt subjects, is ’noticed’ by the
learner and aids in acquisition. It should be noted that both Lasnik and Chomsky rely
on indirect negative evidence and show how acquisition could proceed in the absence of
negative evidence; moreover, Lasnik’s proposal circumvents the problem of the target
language being a subset of the hypothesized language.

The problematic nature of dative alternation with respect to acquisition continues
to be the object of examination by researchers. Lasnik begins his argument with a brief
summary of previous work. He examines the following constructions:

(1-17) John gave the book to Fred.
(1-18) John gave Fred the book.

(1-19) John donated the book to the library.
(1-20) *John donated the library the book.

35
According to Baker (1979), 1-20 would never be produced by a child. However,

Mazurkewich and White (1984) showed that children do overgeneralize the
subcategorization rule, and argued that children pare down to the correct rule with
positive evidence alone using a lexical redundancy rule. In the following

subcategorizations, 1-21 is the target rule and 1-22 is the overgeneral initial rule:

 

 

+ V1 (monosyllabic) + V.- (monosyllabic)
1_2 .. . NP NP
( I) ._NP,(’°)NP, “ 2 '.
f0,- NP2 prospective possessor of NP1
+ Vt, V
1 fl) r 2 2 1

 

 

 

 

As can be seen, the overgeneral rule 1-22 lacks semantic and syllabic specifications.
Mazurkewich and White claim that once the child notices that the indirect object is the
prospective possessor whenever the double object construction is allowed as an alternative
to the prepositional phrase construction, this fact will be incorporated into the child's lexical
redundancy rule. Then, when the child notices that verbs in dative alternation constructions
are monosyllabic, this fact will be added to and finalize the rule. Thus, dative alternation is
acquired based on positive evidence alone.13 However, as Lasnik points out, notice that the

overgeneral rule 1-22 includes rule 1-21. This means that anything described by 1-21 is also

 

‘3 Mazurkewich and White (1984) dismiss INE as a viable solution on the grounds that 1) it would
require UG to specify which subcategorizations are possible with which verbs under which situations; and
2) "the child would have to be able to distinguish between the output of his own grammar and those of
other people, and between other children and adults” (p. 272). Lasnik’s proposal that the child notices
when alternation does not occur (that is, UG does not specify every instance when it must occur) solves
the complexities raised in the first objection. Mazurkewich and White’s second objection could be applied
to any proposal of language acquisition and therefore does not seem to be a valid argument against INE
specifically.

36
described by 1-22. So, we have returned to the problematic situation discussed above where
the target language is a subset of the hypothesized language, and in principle, negative
evidence should be required in order to retreat from the incorrect overgeneralization.
Additionally, Lasnik questions what would cause the learner to abandon a successful rule;

no positive evidence would falsify the rule, but rather the rule

fails to exclude certain classes of non-occurring alternations. What the child must be
noticing, then, is that the alternation does not occur unless the requirements are met. It is
clearly not sufficient to notice that the alternation does occur if they are met. But this is
precisely indirect negative evidence . . . and there is no obvious way around the need for [it].
(Lasnik, 1989, p.96)

In summarizing, Lasnik states:

We have considered, in a preliminary way, how to explicate the notion "where they would
be expected to be found," which is crucial to the learner's successful use as data of the
absence of certain sentences. . . . linguistic theory (i.e., the structure of the brain) simply does
not make available a grammar that is otherwise just like that of English but that fails to
generate the specific sentence The man left. Ifthis is the right answer (and it is hard to see
what an alternative might be) the learner will pay no attention to the absence of the example
in the data. There could be no rule or parameter for which that absence would constitute
relevant evidence. (Lasnik, 1989, p. 97)

Valian also works within a UG framework, but unlike Lasnik (1989), Lightfoot
(1989), and Roeper and Weissenbom (1990), her focus is not on the triggers (that is, the
linguistic environments which constitute relevant data) but rather on establishing logical and
psychological constraints which an acquisitional model must satisfy. Other researchers
assume that a parameter is preset to avoid the possibility that a learner will hypothesize a
grammar which is larger than the target language. Using the null subject parameter in her
description, Valian (1990a, l990b) argues against this assumption."

Valian's model (1990b) begins with the Input Constraint. A model of acquisition

 

“ Kim (1993) argues against Valian’s (I990a) dual-value solution for setting of the null-subject
parameter to which Valian (1993) responds. As these discussions do not deal directly with the issue of
Indirect Negative Evidence, they will not be covered here.

37
must be able to resolve the fact that children may hear ungrammatical, yet acceptable,
sentences (the degenerate problem) and that children must learn to distinguish between the
two. Otherwise, upon hearing sentences such as Want your lunch now? or Raining out, the

English speaking child's parameter would be mistakenly set to the positive value.

A parser assigns syntactic structure to a string of words. It labels each word according to its
syntactic category; it groups the individual lexical items into phrasal categories; it indicates
the hierarchical structure of the string. A parser is fed by a grammar, and cannot parse
beyond its resources. (Valian, l990b, p.122)

Therefore, the constraint on the parser is the grammar that feeds it. Valian explains how the
parsing constraint disallows a preset parametric value in the following manner. Assume that
the preset value is -null subject (English) and that the target language is +null subject
(Spanish). The child will hear +null subject sentences, but since the parser is fed by an
English grammar, these sentences will not be labeled as "sentence" by the parser, and
therefore, the child would never switch to the appropriate +null subject language. Likewise,
suppose the target is English. Upon hearing +null subject sentences, if the parser labels
these as "sentence," and switches to the +null subject setting, this would be incorrect.
Conversely, suppose the preset value is +null subject and the target language is -null subject.
When the child hears +null subject sentences, these would be labeled "sentence" which is
incorrect.” Therefore, based on the above two constraints, Valian concludes that the model,
or acquisition mechanism, must provide the child with both values of a parameter.

The Inconclusiveness Constraint states:

 

'5 Valian argues that indirect negative evidence cannot work with a preset parameter value because
of a time limit problem. How long does the child wait before concluding that she is not going to hear a
form that she expects based on the preset value. However, as will be seen, indirect negative evidence does
play a role in Valian’s model. Lightfoot (1989, 1991) and Roeper and Weissenbom (1990) avoid the time
limit problem by positing "triggering domains" which make it ”unnecessary to have any sophisticated
tabulator of input sentences" (Roeper & Weissenbom, 1990, p. 147).

38

Parser failure is not sufficient reason for the child to change her grammar, nor parser success
a sufficient reason for a child to maintain her grammar. . . . Parser failure is inconclusive
because the child is limited by her performance mechanism. As a result, she will fail to parse
sentences that are fine as far as her grammar is concerned, but are beyond her computational
ability.

Parser success is inconclusive . . . [because] by giving the child both values of the
parameter simultaneously, we eliminate some opportunities for parser failure: the child will
be able to parse strings compatible with both values . . . (Valian, l990b, p. 126-127)

As a result of this constraint the acquisition mechanism "must be able to evaluate the
importance of parser failure and parser success" (Valian, 1990b, p.127).

The Consistency Constraint "is a requirement on theory construction generally, and
is introduced as a constraint on the child's theory of the language. The child will want an
internally consistent grammar" (Valian, l990b, p.127). Given that the child is equipped with
both values of a parameter, she or he may be able to parse all input. But, since these
parsings would be inconsistent with each other, the Consistency Constraint dictates that the
child will continue to modify her or his grammar until all rules are consistent.16

Given the above constraints and the assumption that both values of a parameter are
available to the child, Valian proposes that the language acquisition mechanism must be able
to integrate and evaluate all the data and determine which value is more compatible with the
input. Her Comparison Model is such a mechanism. It allows the child to compare, both
observationally and experimentally, the input for "distributional regularities" using both
positive and indirect negative evidence. In the case of null subjects, observational
comparisons entail the child noting "differences in adult subject use as a fiinction of the
cooccurring syntactic and discourse context" (Valian, 1990b, p.131). If the target language

is +null subject, subject omission will be all-pervading; whereas, if -null subject is the target,

 

'6 The Consistency Constraint proves especially problematic for SLA. It may imply that the learner
may be entertaining multiple Interianguage grammars. No such claims are being made here, and this issue
is not addressed, as the Comparison Model is the aspect of Valian’s framework which is relevant to the
present discussion.

39

subject omission will be utterance-initial. In order for observational comparisons to be
effective, "different forms of usage in different contexts must occur with sufficient
frequency" (V alian, 1990b, p. 132). The other form of comparison, experimental, allows the
child to compensate for insufficient input. By this process, the child produces an utterance
and then compares her or his utterance with the adult response. For example, suppose "the
child is (unconsciously) trying to decide between two competing hypotheses, A and B"
(V alian, l990b, p. 132). She or he will produce both A and B-type utterances . "If the adult
replies are structurally and lexically similar to the child’s,17 they can be compared with the
child's utterance, and used to choose between hypotheses A and B" (Valian, l990b, p. 133).
Ifthe adult replies are A-type, regardless of the child's utterance type, the child will conclude
that hypothesis A is correct. Valian claims that this comparison process eliminates the time
limit problem associated with the use of indirect negative evidence. (That is, suppose a
structure is only infrequently used, how long should the child wait before concluding that
the structure does or does not exist in the grammar?) By Valian's account, the problem now
becomes an empirical one and "the child has a procedure that gives her good reason to
expect a structure in the short run, then she can integrate the lack of appearance with the
other data at her disposal" (Valian, 1990b, p. 134).

Saleemi (1990, I992), focusing on binding theory and the null-subject parameter,
first argues that the "binary-valued formulation of the parameter . . . may be descriptively
inadequate . . . [and proposes that a multi-valued formulation may] directly capture a wider

range of variation . . . [without requiring] the postulation of many additional grammatical

 

'7 Here, Valian draws on Levelt and Kelter’s (1982) "correspondence effect" - the tendency of a
listener to respond to a speaker by using a similar structure or vocabulary - as support.

40
mechanisms to the system" (Saleemi, 1990, pp. 235-36). Saleemi (1992, p. 109) suggests
the following:

The null-subject parameter
23 may optionally remain invisible in PF, where 2 represents a class of subjects

containing:

(a) 0; (English, French, Swedish)
(b) nonarguments; or (German)

(c) nonreferential arguments; or (Yiddish)

(d) any arguments whatsoever (Spanish, Italian)

This formulation erroneously predicts that pleonastic,l8 that is, (b) or (c), pro-drop
will be optional, as is referential pro-drop; while referential pro-drop is generally optional,
non-referential pro-drop is obligatory in most pro-drop languages. Saleemi, therefore,
qualifies the parameter and suggests that "since pleonastic subjects are nonreferential, they
would be superfluous and thus might not, or might have ceased to, exist in a null-subject
language . . . [lack of] fiinctional usefiilness [has resulted in] gaps in languages (though
perhaps not in their grammars) . . ." (Saleemi, 1992, p. 110).

The irregular distribution of pleonastics in languages still causes a problem, however,
both for the proposed multi-valued parameter and for learnability. Specifically, languages
do not form a subset hierarchy as would be predicted by the parameter, rather they form both
subset and intersecting relations. As a partial solution, Saleemi suggests that there may not
be a direct correlation between a parametric value and a particular language; this in turn
presents the following learnability problem: If the language learner is exposed to positive

evidence which does match her or his expectations (based on a given parametric value), how

 

‘3 Saleemi (1990, p. 237) defines pleonastics as nomeferential subjects of which there are two types.
A nonargument, which “is an expletive subject which is obligatorily construed with a postverbal NP or
clause; and a quasi-argument, which "is the expletive subject of atmospheric-temporal predicates (e. g. a
weather predicate. "

41
is this mismatch resolved?

Saleemi proposes that a "markedness [defined in terms of parametric values rather
than in terms of the respective languages] condition defines the order in which the
parametric choices expressed in the null-subject parameter are explored . . . and [a] learning
principle that can used to select the correct value of the parameter" (Saleemi, 1992, p. 116)
based only on positive evidence.

Saleemi's learning principle consists of two processes, the first of which is Positive
Identification:

A parameter value is positively identified just in case all observed positive instances
are consistent with that value.

Positive identification is a "selective mechanism" which is completely successful only if the
input perfectly corresponds with the parametric value. However, as stated above, this is not
ensured. For example, assume the target language is (b), in which argumental subjects must
be overt. When the learner is exposed to null non-arguments, the parametric value may be
identified, but nothing prevents the learner from overgeneralizing within the language and
producing overt non-arguments.

Saleemi (1992, p. 118) therefore proposes a second process, Exact Identification,
which may be considered an "observational strategy. . . . Suppose that the language defined
by the target value (a) of a parameter is L(a), and the corresponding ambient language is
L(Ia)."

A parameter value (a) is exactly identified just in case

a. it is positively identified; and

b. the difference between L(a) and L( | a) is known.

Exact Identification thus "requires that the learner should discover the exact extent of

42
difference between the language predicted by the target parameter value and the 'incomplete'
ambient language."

Exact Identification could result if the learner were to undergeneralize within the
hypothesized language, which could be achieved if the learner notices the absence of the
relevant types of overt pleonastic subjects in the ambient language. In other words, certain
structures are expected based on the learner's grammar (the positively identified parametric
value); if these structures are absent in the ambient language, "the learner is capable of the
negative inference" (Saleemi, 1992, p. 118) that these structures are impossible in the. target
language. That is, the learner makes use of indirect negative evidence.

Still, a problem remains. How does the learner know to equate absence with
ungrammaticality? The mechanism which Saleemi proposes draws on Oehrle (1985) who
proposes that the acquisition of certain features of language can be modelled after a set of
mathematical functions. Saleemi expands on this model and proposes a strength function

which distinguishes nonoccurrence from total absence.

If the attested examples do not acquire a threshold level k of strength, or if the nonoccurrence
of such examples does acquire k, by a designated time 1*, the learner, construed as ftmction
f, will then be entitled to a negative inference as to the nonexistence of [a form in the ambient
language]. On the other hand, if a derived form does acquire k by t“, the learner will decide
that form is in the ambient language. (Saleemi, 1992, pp. 1 19-120)

Importantly, the strength value is not necessarily absolute, but can be relativized. If
the frequency of occurrence of two items, x and y, is related, the strength value of each item
will be 'interpreted' in comparison with the other. In this way, the learner will not exclude

items which may just be rare in the language.19

 

'9 Related to the issue of ’rare’ structures is the criticism that the use of Indirect Negative Evidence
would result in the exclusion of more complex, and thus uncommon, structures in a learner’s language.
Saleemi appeals to Randall’s (1985) indirect positive evidence as a solution: "that certain aspects of a
grammar not figuring directly or prominently in positive data might follow indirectly from its other aspects

43
The final issue which Saleemi addresses is the psychological feasibility of the use
of indirect negative evidence. One objection to its use is that it invokes a (possibly
unwarranted) complex learning system. In conjunction with a principles and parameters
approach to language acquisition, which initially restricts the choices made available to the
learner, Saleemi offers the following constraint intended to "ensure a sufficient degree of
conservatism in the learning system" (1992, pp. 127-128):

On no account can the choice of a parameter value, or a change in the value of a
parameter, be made solely on the basis of indirect negative evidence.

In sum, according to Saleerrri's framework, the choices made available to the learner
are restricted by parametric theory; the learner is "sensitive" to the nonoccurrence of certain
data which may be expected "under the preexisting schemas;" based on positive evidence
alone, a parameter value is Positively Identified; "indirect negative evidence may then be
invoked in the event of the need to exclude some consequences of a choice" (Saleemi, 1992,
p. 128). Saleemi has offered a computational mechanism, or strength function, whereby the
learner may make use of indirect negative evidence. The strength function consists of a
frequency threshold and a time threshold. Parameters are set when a threshold is crossed.
The frequency threshold may be crossed by positive evidence; "if afier a certain amount of
time, the frequency threshold has not been crossed, (the learner) assumes it never will be;

once the time threshold is crossed a decision is made" (Archibald, 1993, p. 4).

 

that are directly and explicitly exemplified" (Saleemi, 1992, p. 124). The machine induction system of
Berwick and Pilato (1987) was able to ’learn’ fully expanded English auxiliary phrases (e. g. could have
been being given) from simple, frequently occurring strings.

1.4.1.1. Summary

The purpose of this review has been to show that within L1 acquisition, it has been
argued that INE may provide a better theoretical explanation of the acquisition of dative
alternation (Lasnik) rather than sole reliance on positive evidence (Mazurkewich & White).
Additionally, INE plays a crucial role in somewhat different accounts of the acquisition of
null-subjects. Lasnik attributes the language learner with a preset (unmarked) value of -null
subject. By this account, -null subject is the first guess and will not be disconfirmed by
positive evidence for the child learning English. For the child learning Spanish, the initial
guess will be disconfirmed by sentences lacking overt subjects. That is, Indirect Negative
Evidence, the absence of overt subjects is noticed by the learner, aids in acquisition. Saleemi
suggests a four-valued null subject parameter. Because the Subset Principle and positive
evidence will not guarantee that the correct value will always be selected, Saleemi proposes
a redefinition of markedness in terms of parameter values rather than in terms of the
respective languages along with a learning procedure which relies on both positive and
Indirect Negative Evidence. A parameter is identified based on positive evidence alone; a
parameter is then Exactly Identified through the use of INE; the learner notices the absence
of certain structures which are predicted based on the positively identified parameter.
Finally, Valian proposes a different approach to the problem and suggests a series of logical
and psychological constraints which an acquisitional model must satisfy. She then suggests
possible mechanisms which might meet the requirements imposed by the constraints.
Among these is a Comparison Model, which allows the child to observationally and
experimentally compare, using both positive and Indirect Negative evidence, the input for

distributional regularities. Thus, all three researchers incorporate the use of Indirect

45

Negative Evidence into their proposals of the acquisition of specific structures. While
Lasnik attributes the language learner with a 'preset' value, Valian and Saleemi argue against
this position and in turn focus more closely on the learning procedure. Both offer somewhat
complex computational mechanisms, each of which relies on the frequency with which an
item occurs (or does not occur) in the input. The importance of frequency has been
addressed in SLA, which is discussed in the next section.
1.4.2. Indirect Negative Evidence in Second Language Acquisition

Theoretical support within the field of SLA for the role of INE can be found in Gass'
notion of "apperceived input" (1988).20 By this account, there are certain factors which
"serve as ambient speech filters . . . and may determine why or why not some ambient
speech, or input, passes through to the leamer" (Gass, 1988, p.202, p. 204). Frequency, prior
knowledge and attention are among the ones mentioned by Gass which are particularly
relevant to the present discussion. With respect to frequency, Gass states ". . . something
which is unusual because of its infrequency may stand out to the learner" (Gass, 1988,
p.202). This concept is at the core of indirect negative evidence; that is, "infrequency" (or
absence) will be noticed by the learner. Prior knowledge is what makes the input either
meaningful or meaningless, and it in turn "activates" selective attention which "allows a
learner to notice a mismatch between what he or she produces/knows and what is produced
by speakers of the second language" (Gass, 1988, p. 203). We thus have an explanation for

Chomsky’s phrase "where they would be expected to be found."

 

2" The relationship between Gass’ entire framework and indirect negative evidence goes beyond the
scope of this proposal, but will be addressed in the dissertation. The focus here will be on the similarity
of her conceptualization of how ”ambient speech" becomes ”apperceived input" and the notion of Indirect
Negative Evidence.

46

Working within phonology, Archibald (1993) supports the view that INE must be
taken into account in language learning but argues that even if the learner does recognize a
mismatch between the input and the output (that is, notices indirect negative evidence), there
still remains the problem of knowing what needs to be changed in the output. Further,
"indirect negative evidence also does not specify what action is to be taken to remedy the
error. This is the problem of blame assignment . . . a term coined in cognitive science and
artificial intelligence [which] has to do with how a learner changes a complex system on the
basis of error detection" (Archibald, 1993, p. 5).

Archibald therefore shifis focus from the frequency of an item in the input to a
concentration on the items themselves. He argues that to date the relationship between
triggers and parameters has been neither constrained nor specified. He suggests that there
are specific cues which are appropriate for particular parameters, which he formalizes as:

If x is found where y was expected, change parameter 2.

According to Archibald (1993, p. 8), the concept of appropriate cues would

I. Restrict the hypothesis space of the learner in terms of possible actions;

2. Acknowledge that expectations (perceptions) are guided by linguistic
grammar (as well as production); and

3. Address the problem of blame assignment.

Archibald further argues that admitting the use of INE into a learning theory would
allow the inclusion of induction into language learning. Noting, however, that this process
must be constrained in some way, he suggests that UG and a principle of Lexical
Dependency would provide such constraints. Lexical Dependency was first suggested by
Newson (1990, p. 179) as "a mechanism which generalizes learned information concerning

a particular lexical item to all other relevant lexical items. . . . Let us refer to this learning

47
mechanism as a "Lexical Dependency" as the setting of lexical parameters of certain items

are, under these assumptions, dependent on those of others."

Archibald (1993, p. 10) claims that

induction constrained by possible representations of UG and a principle of Lexical
Dependency [would] account for (l) the time it takes to reset a parameter, (2) stages in
interlanguage development when the learners are apparently allowing both settings of a
parameter, and (3) the individual variation in patterns of interlanguage change over time.

There are three assumptions implicitly made by Archibald which I would like to
address. First, he assumes that the domain of expectation for the second language learner
is UG. As stated in the Introduction, I argue that the more parsimonious assumption is that
the domain of expectation is provided by the L1. Notice that limiting the domain has the
same effect as Archibald's 'appropriate cues' in that the hypothesis space of the learner is
immediately restricted. This is not to say that I disagree with his notion of appropriate cues.
Rather, a limited domain of expectation in conjunction with an appropriate cue may provide
the desired constraint on induction in addition to 'directing' the learner toward the correct
'action.‘

Second, Archibald states "that some aspects of language acquisition involve
induction while other aspects involve deduction. It . . . seems fitting that the domain of
inductive learning is the lexicon . . ." (Archibald, 1993, p. 9). If one assumes that induction
is a learning strategy, I believe it is a dangerous move to suggest that learning
strategies/procedures are limited to and vary according to (perhaps artificially separated)
linguistic subfields. A more fiuitful, but more tedious, line of inquiry may be to investigate
the variability of inductive learning as a result of individual learner characteristics. While
it may very well be that certain linguistic areas are more 'conducive to' acquisition through

induction, individual variables play a critical role and should not be eliminated fiom the

48
equation.

Finally, and most importantly, Archibald provides no justification that the use of INE
is similar to induction: "I also feel that one of the benefits of including indirect negative
evidence in our learning theory is that it allows us to incorporate the process of induction
into language learning" (Archibald, 1993, p. 8).21
1.5. CHAPTER SUMMARY

In this chapter I have provided a brief overview of the Learnability Problem in first
language acquisition. I have also summarized the theoretical and empirical arguments which
suggest that the Learnability Problem of L1 acquisition, with respect to incomplete evidence
available to the learner, applies equally to SLA. I have examined in relative detail one
aspect of the Learnability Problem, namely the role of negative evidence in both first and
second language acquisition. The research in both first and second language acquisition
indicates that when learners make errors they do not consistently receive overt correction or
if they do receive correction, it appears that in many instances they ignore it. I have
reviewed the proposals which have suggested that the use of INE in certain domains may
provide a partial solution to the Learnability Problem. These proposals come from both L1
and L2 researchers, working in somewhat differing frameworks, several of whom have
alluded to the possibility that the use of indirect negative evidence is part of the learning
procedure, a learning strategy, or induction. However, the distinctions and relationships

between 'learning strategy,’ 'induction,‘ and 'indirect negative evidence' have not been

 

2‘ Saleemi makes a similar assumption, somewhat implicitly, in his discussion of ”exact identification
(which) further consists of inferential processes..." (p.121) and his discussion of Berwick and Pilato’s
machine induction system (1992, pp.124-126). However, he later states that "the relationship between
various types of evidence and different learning strategies [italics added] certainly deserves to be probed
in considerable depth” (1992, p. 139).

49
explicitly described in the literature reviewed above. In order to assume that the use of
Indirect Negative Evidence is indeed an empirical problem which can be solved (Valian,
1990, p. 134) and then proceed to "empirically investigate the psychological plausibility of
[its] use" (Saleemi, 1992, p. 139), it is necessary to define as precisely as possible the
processes under investigation and to provide valid argumentation for any analogies made.
Therefore, the proposal that the use of indirect negative evidence is analogous to induction
which in turn is a learning strategy, while implicitly accepted by the above authors, must be

substantiated. This is the topic of Chapter 2.

CHAPTER 2
INDUCTIVE INFERENCING

In this chapter I first define inductive inferencing/reasoning. I then provide
examples of the use of induction in the processing of meaning in a first language. I
describe a process of inferencing in SLA and argue that this inferencing process is
analogous to the use of Indirect Negative Evidence (INE), both of which involve the use
of inductive inferencing. I contend that the acquisition of the ungrammaticality of a
specific linguistic structure may depend on the use of inductive inferencing or INE. I
offer a working definition of inductive inferencing and then review learning strategy
research in SLA as it relates to this definition. Finally, I summarize representative
studies in the teaching of learning strategies and discuss some of the issues left
unresolved which the present research partially addresses.

2.1. INDUCTIVE INFERENCING

In The Logic of Scientific Inference, Trusted (1979) distinguishes between
deductive inference and ampliative inference. Deductive inference is in accordance with
accepted rules of logic. If the rules are correctly followed, the conclusion cannot be
false if the premises are true, because the conclusion cannot contain more information
than is found in the premises. In ampliative inference (eduction and induction), the
conclusions contain more information than is contained in the premises. Both eduction

and induction are based on an expectation of order and consistency. Inferencing from

50

51

known to unknown particulars is a case of eduction; inferencing from known particulars
to generalizations is an instance of induction.
2.2. INDUCTIVE INFERENCING IN FIRST LANGUAGE ACQUISITION

In Psycholinguistics, Kess (1992) explains that the conclusions reached through
deductive inferencing, which is based on mathematical logic, are of the necessarily true
type. Whereas, the conclusions reached through inductive inferencing, which is based
on mental logic, are of the probably true type. Kess (1992, p. 193) cites work conducted
with syllogisms and on discourse (Johnson-Laird, 1986; Moore, 1986; Iakoff, 1987) and
concludes that "inferences derived from natural language are very often of the probably
true instead of the necessarily true type."

Moore (1986, p. 52) uses several specific examples to argue "that deductive logics
are too narrowly based to provide appropriate and revealing models of reasoning and
inference in natural language; and . . . a better understanding of inductive inference is
essential if linguists are to gain useful insight into natural language operation.” Moore
points out that deductive reasoning is based solely on the form of the argument, not on
the meaning of the premises. Whereas, form cannot be separated from meaning with
inductive reasoning. Moreover, "inductive logics accept uncertainty and indeterminacy,‘
which are primary characteristics of natural language” (Myers, Brown, & McGonigle,

1986, p. 3). Moore (1986, p. 53) offers the following examples to show the contrast

 

' Note that the term "indeterminacy” as used here varies slightly from its use in SLA, where it has
been defined as "the incomplete (or lack of) knowledge a learner has of parts of the second language
grammar" (Gass & Selinker, 1994, p. 36). The focus is on the learner. It is implied that certain aspects
of the grammar can potentially be complete or determinate; what is lacking is the learner’s knowledge.
As used here, the term refers to the inherent vagueness and indefiniteness of certain aspects of natural
language; it may be impossible for these aspects to ever be determinate.

52

between inductive and deductive reasoning:
(2-1) Deductive
Premises: Every mammal has a heart.
A cat is a mammal.
Conclusion: A cat has a heart.
(2-2) Inductive
Premise: Every cat that has been examined has a heart.
Conclusion: Every cat has a heart.
Importantly, the conclusion of the deductive argument states nothing new. Whereas, the
conclusion of the inductive argument, while uncertain, is innovative.

To support his view that inductive inference may ”provide the more appropriate
model . . . to explain how meaning emerges from the utterances of a language” (Moore,
1986, p. 65), Moore reanalyzes the work conducted by Luria in the 1930’s in
Uzbekistan. Of the 30 subjects in the study, 15 were schooled and 15 were non-
schooled, "living in a backward economy based mainly on the raising of cotton. The
kishlak (village) dwellers displayed remnants of a once-high culture together with
virtually complete illiteracy, and also showed the pronounced influence of Islamic
religion" (Luria, 1976, p. 13). Luria gave the subjects two kinds of syllogisms. The
premises of one kind of syllogism contained information which was related to the
subjects’ experiences. For example:

(2-3) Cotton grows well where it is hot and dry.

England is cold and damp.

Can cotton grow there?
The premises of the other syllogisms contained information which was new to the
subjects. For example:

(2-4) In the Far North, where there is snow, all bears are white.

Novaya Zemlya is in the Far North.
What colour are the bears there?

53

Luria concluded that the unschooled peasants refused to make deductive inferences
from any of the syllogisms, while the subjects who had some schooling completed the
deductive inference ’correctly.’ Thus, "the acquisition of new modes of thought such as
deductive inference depend on socio-historical development” (Moore, 1986, p. 58).

However, as mentioned, Moore reexamines Luria’s study from a different
perspective, arriving at a conclusion somewhat contradictory to that of Luria. Moore
cites an exchange with one subject, Abdurakhm, age 37, non-literate, from a remote
Kashyar village in Uzbekistan. When asked what color the bears in Novaya Zemlya are

(2-4 above), Abdurakhm responded:

We always speak only of what we see; we don’t talk about what we haven’t seen. . . .
If a man was sixty or eighty and had seen a white bear and had told about it, he could
be believed, but I’ve never seen one and hence I can’t say. That’s my last word. Those
who saw can tell, and those who didn’t see can’t say anything.2 (Moore, 1986, pp. 56-
57)

Moore (1986, p. 57) points out that Abdurakhm’s comment is a prime example of a
conditional argument of modus tollens:
If I could tell, I would have seen.
I did not see.
Therefore, I cannot tell.
This may take the logical form:
If p, then q.
Not q.
Therefore, not p.
In other words, Abdurakhm is quite capable of deductive argument and inference.

Moore (1986, pp. 58-60) contends ”that deductive reasoning may not have the special

’advanced’ status [Luria] accords it," and maintains that the primary difference between

 

2 Similarly, to syllogism 2-3, the non-schooled subjects responded that they had never been to
England and therefore didn’t know if cotton grew there.

54

the literate and non-literate is their ability to separate form from context, not in their
cognitive capacity for deductive reasoning. While the words of the syllogism are ”inert,
inactive" for the literate, the non-literate are involved in "active, constructive processes"
of reasoning and inference "to fill in gaps and establish links between frames of reference
representing the subject’s accumulated and categorized experience of the world."

Moore also cites the ethnographic work of Scribner (1977) to support the
conclusion that across cultures the non-literate draw on their knowledge and experience
of the world to arrive at a meaning (conclusion) of syllogisms. That is, they respond to
the content rather than the form of the syllogism. Moore offers the following example
from Scribner (1977, p. 487):

If Sumo or Saki drink palm wine, the Town Chief gets vexed. Sumo is not drinking palm wine,
Saki is. Is the Town Chief vexed?
erlle farmer: The Town Chief was not vexed on that day.

Investigator: What is the reason?
erlle farmer: The reason is that the Town Chief does not love Sumo.
Investigator: He doesn’t love Sumo? Go on with the reason.

erlle farmer: The reason is that Sumo’s drinking means a hard time. That is why when he
drinks palm wine, the Town Chief gets vexed. But when Saki drinks palm
wine, he will not give a hard time to people. He goes to lie down and sleep.
In that way people do not get vexed with him. But people who drink and go
about fighting, the Town Chief cannot love them.

While this may be the ’wrong’ logical conclusion, Moore concludes that the
erlle farmer has reasoned "clearly." The farmer rejects one of the disjuncts of the
major premise (if Saki drinks palm wine, the Town Chief gets vexed), and he adds
information regarding the different effects of palm wine on Sumo and Saki. That is, ”he
declines to restrict himself to the ’closed’ world of the deductive reasoning exercise; he
interprets the premises inductively according to their content and its plausibility, not
simply their form, and comes up with a response which, while lacking the certainty of

a valid inference, is innovative" (Moore, 1986, p. 59).

55
It may be argued that reasoning exercises with syllogisms are not comparable to
reasoning in natural language processing. However, Moore contends that the processes
involved in the interpretation of anaphoric and deictic expressions show the close relation
between inductive inferencing and natural language processing. Moore (1986, p. 61)
offers the following example:

A: An alsatian bit me.
B: They are vicious beasts.

The referent of the anaphoric expression they cannot be interpreted with 100% certainty:
they could possibly refer to the class of dogs, to the class of dogs with alsatians, or to
a particular group of alsatians. In order to understand B’s response as referring to the
class of alsatians, A must use "a form of inductive reasoning which recognises that
speaker B has linked the reported case with her general belief about the propensity of
alsatians to attack without cause" (Moore, 1986, p. 61). Even though this belief may not
be valid, A is aware of the common practice of generalizing from one instance to all
instances. Therefore, A uses information which is not present in the exchange in order

to resolve the anaphor. Moore (1986, p. 62) concludes that

interpreting anaphoric expressions may require bringing to bear information not explicit
in a text or exchange; the understander arrives at the most probable, rather than the
’correct,’ interpretation of the anaphor. Both of these factors - the introduction of new
information and the uncertainty associated with the conclusion - are . . . characteristics
of inductive inferencing. . . . anaphoric expressions do not refer to elements in an
utterance but to ’mental categories’ that the bearer constructs as the utterance, in the
course of being processed, connects with his or her existing frameworks of knowledge,
belief and experience. On this view, the fact that the antecedent to an anaphoric
expression cannot necessarily be identified with any degree of certainty does not constitute
a problem. Indeed, the relation between an anaphoric expression and its antecedent may
have to be induced. The conclusion reached is not formally valid but is the most
plausible in the context given.

Moore also argues that understanding of deictic expressions requires inductive

processes. As ”a deictic word . . . takes some element of its meaning from the situation

56

(i.e., the speaker, the addressee, the time and place) of the utterance in which it is used"
(Hurford & Heasley, 1983, p. 63), it requires the hearer/reader to bring ”previous
knowledge and experience of the situation in which the exchange takes place in order to
arrive at the most probable interpretation" (Moore, 1986, p. 63). Hurford and Heasley
(1983, pp. 64-65) illustrate "the flexibility with which deictic terms can be interpreted"
with the verb come. If a speaker says Come over there, please, while pointing to a
distant corner of a room, the hearer may "reasonably” inductively infer, but again
without 100% certainty, that the speaker intends to move to that corner also.
Additionally, come in this example is extended to include ”toward where the speaker will
be.” (In contrast to the utterance Can I come to your ofi'ice? spoken over the telephone.
Here, the speaker plans to move to the m location of the hearer). In other words,
deictic words may be considered ”as cues enabling a hearer to use both context and other
accumulated experience to construct a specific set of spatio-temporal relations" (Moore,
1986, p. 63).

Moore (1986, p. 64) concludes that "as a general rule, inductive reasoning in
anaphora, deixis and natural language processing appears to be done informally, almost
casually, on the basis of beliefs and impressions formed, with little if any reference to
evidence." Given the fact that the inductive process, and its outcome, rely on one’s
previous knowledge and experience there is no standard technique of induction. One’s
inductive inferencing ability will be dependent, in part, on the subject matter and one’s
experience/knowledge of the subject.

Moore (1986, p. 64) notes that

given [the] widespread dependence of mtural language processing on inductive reasoning
and inference, it is all the more odd that linguists should have almost invariably looked

57

for models of reasoning and inference in deductive logics. It is possible that, as in the
case of Chomsky’s earlier borrowing of grammars of formal languages to be used in the
formulation of grammars of natural languages, the reasons are in essence threefold: They
were there, they were formally precise, and they were reasonably well-understood.

In sum, Moore has provided concrete examples and convincing arguments which
illustrate the operation of inductive reasoning in natural language processing. Granted,
Moore focuses specifically on adults deriving meaning from utterances of their native
language. However, as discussed in Chapter 1, there have been suggestions for a role
of induction in both first language acquisition (Saleemi) and SLA (Archibald). There is
also additional support for the proposal that the cognitive ability of inductive reasoning
may be involved, at various linguistic levels, in the acquisition of a second language.
2.3. INFERENCING IN SECOND LANGUAGE ACQUISITION

A role of inferencing in SLA is not new to the research agenda. Carton (1966)
argued for the importance of this process in The ’Method of Inference’ in Foreign
Language Study3 in which the focus was on the acquisition of morphemes and

vocabulary.4 He defined inferencing

as requiring the scanning of what is already known by the learner in the establishment of
new concepts. The process inherently requires linking of new material with what is
already known. . . .

Inferencing can add salience to novel stimulus configurations. By definition the
salient features of the environment are ones we notice and remember. . . .

 

3 Carton does not distinguish between inductive and deductive inferencing. Later, I argue that his
method of inferencing may be considered a case of induction.

‘ The influence of the times (i.e. Contrastive Analysis) is apparent in Carton’s conceptualization of
foreign language learning in that his process not only allows for a role of the L1 but also implies a
comparing and contrasting of the L1 and L2. Importantly, however, this process is creative and not merely
a behavioristic view of language learning; thus, it represents a signal of the changing times (e. g. Corder,
I967, I983; Dulay & Burt, 1972, 1973, 1974a, 1974b; Kellerman, 1979, 1983, 1984, 1986). Canon’s
process could be integrated with (relatively) recent proposals of cross-linguistic influences in SLA. For
example, the linking stage may depend in part on the learner’s perception of the distance between the new
material/L2 and the known material/L1 (Kellerman, I979).

58

On the other hand it is conceivable that some learning through inference occurs
completely unnoticed. . . . This might occur when an unknown stimulus element’ is
embedded in what is otherwise a completely familiar and comprehensible context.
(Carton, 1966, pp. 16-17)

Carton distinguishes between three types of information which may be sources of
cues for making inferences: Intra-lingual cues, which are based on the target language;
inter-lingual cues, which are based on the native language; and extra-lingual cues, which
are based on content and/or context.

Carton proposed that inferencing consists of a multi-stage process. First, the
learner scans what is already known in the L1, the L2, and/or ’world’ knowledge.
Second, new material is linked with what is known. Finally, new concepts and
generalizations are established based on the (mis)match between the new material and
what is already known. Thus,

Stage 1: Scanning of what is known (either in L1, L2, or ’world’ knowledge)

Stage 2: Linking new materials with what is known

Stage 3: Establishing new concepts or generalizations‘5
2.4. INDUCTIVE INFERENCING AND INDIRECT NEGATIVE EVIDENCE

A parallel can be drawn between the use of indirect negative evidence and
Carton’s multi-stage inferencing process. The operative principle of Chomsky’s
acquisition system relies on an inferencing process. Chomsky’s domain of expectation
can be assumed to be UG (or that which has already been acquired of the L1); this is

analogous to Carton’s known material. New material (L1 input) and known material

(UG) must be scanned and linked. If what is known or expected fails to be instantiated

 

’ I would argue that an "unknown stimulus element" could be the absence of a structure in an
"otherwise completely familiar and comprehensible context."

‘ Carton (1971, p. 48) later refines Stage 3, noting that inferencing "applies to the acquisition of
a novel term-a novel fragment of language-and n0t to an unfamiliar concept."

59

in the new material, a selection (Carton’s concept or generalization) would be chosen

(established) which would exclude a certain structure in the grammar of the learner.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

ART N H M KY
KNOWN MATERIAL DOMAIN OF EXPECTATION
Ll UG
xyz ’0’? xy
NEW MATERIAL NEW MATERIAL
L2 L1
19’ 1?
INTER-
LANGUAGE \ OUTPUT
OUTPUT < 7 xy
xyz and/or
xzy + xy
SCAN AND LINK **********7
ESTABLISH OPTION
GENERALIZATION SELECTED
ONLY xy ONLY xy

 

 

 

 

 

 

Figure 2.1. Inferencing and Indirect Negative Evidence

 

7 Note that if xy output is produced before this process, the Learnability Problem remains. Positive
evidence alone (xy input) will not inform the learner that xyz is impossible. Assuming no direct negative
evidence, one must explain how xyz is eliminated from the grammar.

60

Chomsky does not elaborate on how expected structures would not be found; my
argument is that this operative principle would necessarily involve a scanning and linking
process. In other words, through the multi-stage process of inferencing, the learner links
what is known/expected with new material, notices the absence of an expected structure
in the new material, and therefore excludes this structure from her or his grammar. The
similarities between Carton’s method and Chomsky’s operative principle can be
schematized as in Figure 2.1.

The question remains whether the inferencing process described above is a case
of deductive inference or ampliative inference. By definition, ampliative inferences are
based on minimal evidence and the logic that connects the evidence to the generalization
is questionable. The generalization or conclusion necessarily includes more information
than what is available in the premises, or the input in the case of language acquisition.
Second language learners can never be certain that they have received all of the
information required to reach a necessarily true conclusion (see previous discussion of
Valian’s argument regarding the time limit problem), especially in light of the findings
that negative evidence is either minimal and/or ignored. It therefore follows that second
language learners are indeed not formulating generalizations based on deductive

reasoning. 8

 

3 That inductive reasoning plays a role in foreign language aptigde was suggested by Carroll as
early as 1953. Carroll (1962 cited in 1981, p. 105) proposed that foreign language aptitude consists of four
independent abilities, one of which is "inductive language learning ability—the ability to infer or induce the
rules governing a set of language materials, given samples of language materials that permit such
inferences.” While I argue that inductive inferencing is a learning strategy which can be taught, Carroll
(1981, p. 86) states: "I am in general sympathy with writers like Neufeld (1978) who want to emphasize
that foreign language aptitude, whatever it is, is not fixed or innate. . . . I am simply neutral on this
matter, since we do not have the kind of evidence that would enable us to decide it. . . . I have no hard
evidence that would impel me to disagree with the idea that foreign language aptitude, considered as the
individual’s initial state of readiness and capacity for learning a foreign language, and probably degree of

61

What follows is justification for an investigation of the possible role of ampliative
inferencing in the acquisition of the contrasts in 2-5 and 2-6 by native speakers of
English learning French. In the context of this specific linguistic example, I argue that
in order for the second language learner to reach the correct conclusion, the absence of
expected material must be noticed and the process of inductive inferencing is a means to
that end.

(2-5) Marie oublie souvent ses devoirs.

*Mary forgets often her homework.

(2-6) *Marie souvent oublie ses devoirs.

Mary often forgets her homework.

While both English and French allow SVOA, ASVO, and SAuxAde word
orders, the languages differ in that only French allows SVAO order, and only English
allows SAVO order.9 Importantly, if one assumes lack of negative feedback, nothing in
the French positive input informs learners that the SAVO order allowed in their native
language (English) is disallowed in the target language (French). Utilizing Carton’s
inferencing process, four possible courses of action can be hypothesized:

1. Assume that the learner uses intra-lingual (L2) information to reach the
generalization that only SVAO is possible. The only way to disconfirm this would be
the presence of SAVO order in the L2. Can we assume that the learner who only uses

SVAO has confirmed the generalization through noticing the absence of SAVO word

order? In order to conclude that the learner has not progressed beyond Carton’s stage

 

facility in doing so, is crucially dependent upon past learning experiences. Yet, what evidence I have
suggests that foreign language aptitude is relatively fixed over long periods . . . and relatively hard to
modify in any significant way." The present research may provide further evidence to decide this issue.

9 See Introduction, Footnote 2, for exceptions to this characterization.

62
l and has only scanned what is known in the L2 (i.e., has reached a generalization based

solely on positive input), it would be necessary to find a learner who has never used
SAVO word order in French. This is highly unlikely.

2. Assume that the learner uses inter-lingual (L1) information and only uses
SAVO word order. This learner has not advanced beyond stage 1. That is, the learner
has not related the new material with what is known in the L1, and is relying solely on
the L1.

3. Assume that the learner uses intra-lingual and inter-lingual information to form
the generalization that both SAVO and SVAO word orders are possible in French. This
generalization cannot be confirmed and can only be disconfirmed by the learner when she
or he notices the absence of SAVO word order. We can assume that the learner who
continues to use both SVAO and SAVO word order has not noticed the absence of SAVO
in French.

4. Assume the learner uses intra-lingual and inter-lingual information, notices the
absence of SAVO order in the L2 and comes to the generalization that only SVAO is
possible. Importantly, given the implausibility of a learner exhibiting the interlanguage
of #1, #4 learners would exhibit, for a period of time, the interlanguage of #3 before
progressing to #4.

In order for the learner to ’switch’ from SAVO to SVAO word order, the absence
of SAVO word order in French must be noticed by the learner. While the absence of
the structure may be noticed at Carton’s stage 2, its absence must be noticed in order to

disconfirm generalization #3 and thus progress to Carton’s stage 3.

63
2.5. INDUCTIVE INFERENCING: A WORKING DEFINITION

Within the present context, inductive inferencing refers to a multi-stage
comprehension (versus production) process which results in a probably true hypothesis.
This process may be involved in the acquisition of the syntax, morphology, phonology,
pragmatics, or semantics of a first or second language but is not reserved exclusively for
the linguistic domain. In very general terms, following Carton, the process begins with
scanning of what is known in the L1, L2, or world knowledge; in the second stage, new
material is linked with known material; and in the third stage, a probably true hypothesis
is formed.

Each of the stages may occur unconsciously, as would be the case in certain
instances in L1 acquisition if one accepts the arguments presented earlier in this chapter
and in Chapter 1 which claim that induction is the learning principle at the core of
Chomsky’s Indirect Negative Evidence. The process of inductive inferencing may also
occur unconsciously among ’good’ language learners, if one accepts the results of the
research on learning strategies. Furthermore, certain stages of the process may be
intentionally engaged in by second language learners (see discussion of ’Learning
Strategies’ below).10 Importantly, known material in this process does not refer to that
which can be explicitly articulated, although this possibility is not excluded.

It is important to note that each stage is dependent on the previous stage(s), and
therefore, in a sense, each stage consists of multiple processes. Additionally, the
processes at each stage are dependent on multiple variables. For example, as Gass

(1988, 1994) proposes, factors such as frequency of input, prior knowledge, and affective

 

‘0 Intentionality, however, may not necessarily entail specific knowledge of the process.

64

variables may play a role in determining if the input becomes ”apperceived input."
Other factors may consist of the medium of presentation (oral, written) and the wide
range of learner variables (learning style, maturational constraints, previous educational
experience). At any point the process of inductive inferencing may break down, perhaps
as a result of the aforementioned variables. Additionally, a learner may be strong in one
aspect of the process (e. g., scanning L2 knowledge) and weak in another aspect (e. g.,
linking new material with L1 knowledge).
2.6. SUMMARY

To summarize the chapter thus far: Inductive reasoning/inferencing has been
defined; specific examples of the use of induction in the processing of meaning in a first
language have been given; a process of inferencing in SLA has been explained; it has
been argued that this inferencing process is analogous to the use of INE, both of which
involve the use of inductive inferencing. In other words, the implicit assumptions of
Saleemi and Archibald (see Chapter 1) have been explicitly formulated and justified.
Furthermore, it has been argued that the acquisition of the ungrammaticality of a specific
linguistic structure may largely depend on the use of inductive inferencing, or INE.
Finally, a working definition of inductive inferencing has been provided. Support for the
proposal that this process may also be considered a strategy will now be provided.
2.7. LEARNING STRATEGIES

There is not yet complete agreement on the definitions and classification of
learning strategies nor is there an understanding of the complex relationships between
task-type, learner variables, and various strategies (Naiman, Frohlich, & Todesco, 1975;

Bialystok, 1983, 1990; Wenden & Rubin, 1987; Oxford & Cohen, 1992; Gass &

65

Selinker, 1994). Oxford and Cohen (1992, p. 3) note that "serious conceptual and
classificatory problems" exist in the domain of learning strategies. As reported by Gass

and Selinker (1994, p. 266), there are unresolved issues of

me criteria used for classifying language-learning strategies, whether such strategies are
conscious or unconscious, the relatiomhip to learning styles, and the difficulty of showing
what contributes to language learning. Furthermore, Bialystok (1990) pointed out that
it is difficult in practice to distinguish as strategic just learner behaviors that are clearly
(I) concerned with problematic tasks, (2) conscious or unconscious, and (3) intentional
or unintentional.

The lack of consensus on what should be included under the rubric of learning strategy
is evidenced in the following sample definitions.
Drawing on Rigney (1978), Oxford and Crookall (1989. p. 404) offer a general

definition of learning strategies as

steps taken by the learner to aid in the acquisition, storage, and retrieval of information.
. . . strategies can make learning more efficient and effective.

Rubin’s (1987, p. 19) definition of learning strategies
includes any set of operations, steps, plans, routines used by the learner to facilitate the
obtaining, storage, retrieval and use of information (after O’Malley et al., 1983; and
Brown et al., 1983). . . . [strategies are] the behaviors and thought processes that

learners use in the process [italics in original] of learning, not those variables which may
provide a background to learning success.

While Rubin excludes such background variables as psychological characteristics (e. g.,
learning styles), affective variables, and social style, Weinstein and Mayer (1986)
maintain that learning strategies are intentional on the part of the learner (contrary to

Seliger, 1983, 1984) and

[learning strategies] can be defined as behaviors and thoughts that a learner engages in
during learning and that are intended to influence the learner’s encoding process. Thus,
the goal of any particular learning strategy may be to affect the learner’s motivational or
affective state, or the way in which the learner selects, acquires, organizes, or integrates
new knowledge. (Weinstein & Mayer, 1986, p. 315)

Focusing on communication strategies, Blum and Levenston (1978) used the

66

criterion of temporality to distinguish between strategies and processes. They defined
strategies as ”the way a learner arrives at a certain usage at a specific point in time” (p.
402). Repeated use of a strategy can lead to internal reorganization of a second
language. This process is the ”systematic series of steps by which the learner arrives at
the same usage over time" (p. 402). According to Blum and Levenston’s
conceptualization, strategies may be either ’situation bound’ (used only once), or they
may ’initiate processes’ (repeated use of the strategy).

Seliger (1984) also uses temporality, among other criteria, in distinguishing
strategies from tactics, both of which are levels of processing in SLA.

The first level of processing, strategy, is a biological, constant, learning process used by
all human beings in all learning environments and consists of abstract unconscious
formulation of cognitive questions in the form of hypotheses attempting to relate new
material to already established cognitive structures. The second level of processing,
tactic, consists of specific responses in specific learning environments. Some tactics
(micro-tactics) provide data from which strategies may derive underlying principles or
rules of the language system while other tactics (macro-tactics) do not engage the
hypothesis-formation process at the level of strategy.” (Seliger, 1984, p. 47)

The position that temporality in addition to conscious awareness are characteristics
distinguishing tactics from strategies is supported by Oxford and Cohen (1992).

Learning tactics represent the short-term art of using specific behaviors or devices . . .

to support one or more major learning strategies during day-to-day learning situations.

Learning strategy is the long-range art of learning more easily and effectively by using
major clusters of behaviors . . . (Oxford & Cohen, 1992, p. 4) [Italics in original]

Bialystok (1978, 1990) uses the criterion of optionality to distinguish strategies
from processes.

The idea originally modvating this criterion was that there is level of performance which
is the inevitable, perhaps automatic, mental functioning of the linguistic system. These
processes are carried out by a mental executive (some control structure that oversees all

 

” Based on this definition, Seliger (1984, p. 41) maintains that ”studies of ’good’ language learners
(Rubin, 1981) or introspective research (Cohen & Hosenfeld, 1981) will not provide insights into the level
of strategy."

67

performance) in response to the demands of the problem under the constraints of the
system (including knowledge of the target language, knowledge of other languages. . .).
. . . An additional construct is necessary only when the learner/performer intercedes in
these usual processes to change the normal routine, and hence, the expected form of
response. To this end, strategies are defined as supplementary activities that the learner
can impose on the autonomous system to expedite achieving a goal. (1990, p. 19)

Faerch and Kasper (1983) propose that consciousness and problem-orientedness
are the criteria which distinguish strategies in their model of language production. Their
model consists of a planning phase and an execution phase, both of which involve
separate processing steps. Strategies, however, are involved only in the planning phase
and incorporated into the formation of a communicative plan only when a problem
occurs.

Thus, strategies have been defined as having an affective and/or conceptual basis
and may influence everything from the learning of simple tasks to more complex tasks
such as language comprehension or language production.”

2.8. INDUCTIVE INFERENCING AS A LEARNING STRATEGY

While the principal goal of the present research is not to develop a learning
strategy taxonomy or to propose a conclusive definition of strategy, results may provide
insights into the concept of a learning strategy, and more specifically, into inductive
inferencing’s status in the classification scheme. The purpose of the following selective
summary of the investigations into learning strategies is to justify the assumption that the
process of inductive inferencing, as defined in this dissertation, is a learning strategy.

The review will therefore focus on that research in which (inductive) inferencing has

 

'2 Oxford & Cohen (1992, p. 17) attribute "the lack of concordance in strategy categories and
concepts . . . [to] differences in research methodology, and research methodology is a reflection of
researchers’ goals. Some of the identifiable goals include: analyzing the causes of linguistic error,
describing a good language learner, defining learning processes, and validating teaching methods."

68

been either explicitly defined or can be located within the descriptive schemes.

Importantly, although diverse theoretical and methodological frameworks are adopted in

this research, inductive inferencing has been identified as a learning strategy in SLA

from the initial investigations in this area to the research presently being conducted.
Within SLA, research in learning strategies stemmed from attempts to identify the

characteristics of ’good language learners’ (Rubin, 1975, 1981, 1987; Stern, 1975;

Naiman, Frohlich, Stern, & Todesco, 1978; Wesche, 1979) through self-reports,

interviews, and observations.13 Expanding on the work of Stern (1975), Naiman et al.

(1978) identified five strategies, each of which was divided into sub-strategies, based on

their interviews with successful language learners. The major strategies are:

Active Task Approach

Realization of language as a system

Realization of language as a means of communication and interaction

Management of affective demands
Monitoring of L2 performance

99959?

I would like to draw attention to the authors’ descriptions of strategies 2 and 5. Learners
who engage in strategy 2 refer back to their native language and make cross-lingual
comparisons at different stages of language learning; additionally, they analyze the target
language and make inferences about it. Learners who engage in strategy 5 constantly
revise their interlanguage by monitoring their L2 and testing their inferences. In other

words, from the earliest work in learning strategies, a multi-stage process of inductive

 

'3 These characteristics of good language learners have in turn become one of the criteria by which
strategies have been identified. One of the main problems with learning strategy research—that of causality-
-has been pointed out by Skehan (1989, p. 97): “One can . . . argue that learner strategies do not determine
proficiency, but are permitted by it [italics in original]. The use of learner strategies, that is, may not lead
to higher accomplishments—instead one of the benefits of higher proficiency may be the capacity to use a
wider range of strategies.” Additionally, as Gass and Selinker (1994, p. 266) point out, ”poor learners
may be lacking the verbal skills to report what they do as readily as good learners can. If so, then
differences in reporting skills may be misinterpreted by analysts as differences in strategies used."

69
inferencing, while still nebulous, emerged: successful learners refer back to their NL
(scan their background knowledge/domain of expectation), compare it with the L2 (link
the L1 and L2), and make inferences (generalizations/select options).

Based on a compilation of research derived from self-reports and observations,
Rubin“ (1981) distinguished between cognitive strategies and processes15 that directly
affect learning and those that contribute indirectly to learning. Rubin’s classification is
as follows:

Messes mt directly affect learning

Clarification/verification
Monitoring

Memorization
Guessing/inductive inferencing
Deductive reasoning

Practice

Pflosgs that contrioote indirectly to lm’ng

9999592"

1. Creates opportunities for practice
2. Production tricks

Of particular interest to the present context is Rubin’s distinction between

guessing/inductive inferencing and deductive reasoning. Rubin (1987) states:

Guessing/inductive inferencing refers to strategies which use previously obtained
linguistic or conceptual knowledge to derive explicit hypotheses about the linguistic form,
semantic meaning or speaker’s intention. . . .

Deductive reasoning is a problem-solving strategy in which the learner looks for
and uses general rules in approaching the foreign or second language. Here the learner
uses previously acquired linguistic or conceptual knowledge to derive specific hypotheses
about the linguistic form, semantic meaning or speaker’s intention. The difference
between inductive and deductive reasoning is that in inductive reasoning the learner is
looking for a specific meaning or specific [italics in original] rule whereas in deductive
reasoning the learner is looking for and using more general rules. (pp. 23-24)

 

" Interestingly, Rubin (1987) credits the work of Carton (1966, 1971) as the impetus for her initial
research into learning strategies.

‘5 Rubin (1981, p. 118) defines "cognitive processes [as] those general category of actions which
contribute directly [emphasis in original] to the learning process. Cognitive strategies are the specific
actions which contribute directly to the learning process."

70

The contradiction between the definitions of inductive and deductive reasoning given in
the beginning of this chapter (Trusted, 1979; Kess, 1992; Moore, 1986) and those given
by Rubin are obvious. I would maintain that the distinction between the two is not the
degree of specificity of the conclusion, or rule, but rather the degree of certainty of the
conclusion. Rubin is mistaken in equating eduction (inferencing from known to unknown
particulars) with induction and in equating induction (inferencing from known particulars
to generalizations) with deduction. Unfortunately, it is precisely such variations in
definitions which have resulted in the present ”conceptual and classificatory problems"
(Oxford & Cohen, 1992, p. 3) in the area of learning strategy research. Nonetheless,
I would like to point out that all of the elements, while perhaps misclassified and
misnamed, of an inductive inferencing process are present in Rubin’s scheme.
Drawing on the work of Carton (1966, 1971), Stern (1975), and Rubin (1975),
Bialystok (1978) created a model of second language learning which included four
learning strategies; she then tested (Bialystok, 1979) the interactions proposed in the
model (see below). Bialystok (1978, p. 71) listed four strategies, which are "optional
means for exploiting available information to improve competence in a second language. "
Importantly, the strategies are defined in terms of purpose (formal or functional) and
modality (oral or written), the assumption being that "the effects of the strategies are not
generalizable; the effects, rather, are specific to the particular type of language for which

they are used" (Bialystok, 1979, p. 372). The four strategies are:

l. Inferencing—the use of available information to derive explicit linguistic hypotheses. The
information used for this purpose may be linguistic or non-linguistic, it may be taken
from the speaker or from the environment, and it may relate to the structure or the

71

meaning of the language. '5 (1979, p. 376)

2. Monitoring-noting errors.

3. Formal practicing-the specific exercise of the language code for the sake of mastering
the rule system. (1979, p. 374)

4. Functioml practicing-the language learner increases his opportunity to use the language
for communication by going to movies, reading books, or talking to native speakers.
(1979, p. 374)

According to the model, the type of learning strategy which will be employed by the
learner is dependent on the kind of knowledge that is required by the task. Three types
of knowledge are proposed: Explicit Linguistic Knowledge (conscious, articulable facts),
Implicit Linguistic Knowledge (intuitive information),17 and Other Knowledge (knowledge
of L1, world knowledge). Three sources of inferencing are possible: inferencing from
Implicit Linguistic Knowledge, from Other Knowledge, and from Context, ”the linguistic
and physical aspects of a situation" (Bialystok, 1983, p. 107).

Importantly, Bialystok restricts Carton’s original definition of inferencing to a

strategy . . . employed primarily to derive meaning from the target language rather than
to infer formal or structural features. Although formal aspects of the language, such as
form class, may be exploited in inferencing, the product of inferencing is a representation
of communicative meaning in the use of language, hence the strategy is functional. . . .
the strategy may more appropriately be called ”semantic inferencing." (1979, p. 376)

Furthermore, Bialystok proposes that the product of inferencing will always result in a

form of explicit linguistic knowledge about the second language. In other words,

 

" Bialystok (1983, pp. 105-107) offers the more general definition of inferencing as a strategy
resulting in a hypothesis about the target language based on minimal evidence; furthermore, the logic that
connects the evidence to the hypothesis is often questionable. As with Rubin, Bialystok refers to the
hypothesis as a "deduction" even though her definition is of inductive inferencing, suggesting a lack of
differentiation between induction and deduction. Again, I argue that a distinction between these two
reasoning processes must be maintained.

‘7 Implicit and Explicit Knowledge are distinguished bamd on function: ”It is important to notice that
the distinction between the two knowledge sources is defined in terms of function rather than content. Any
information may possibly be represented in either source, and certainly different second language learners
will vary greatly as a function of the nature and extent of the information found in each. A larger Implicit
Linguistic Knowledge source is associated with an ability for greater fluency; a larger Explicit Linguistic
Knowledge source is associated with extensive knowledge of formal aspects of the language but does not
necessarily imply an ability to use this information effectively" (Bialystok, 1978, p. 73).

72

according to her model, the result of the inferencing process will always be conscious,
articulable knowledge of the second language.

As stated above, Bialystok (1979) tested her model by administering strategy
questionnaires and achievement tests“3 to students learning French as a second language.
Bialystok concluded that functional practice had a stronger relationship with achievement
than any of the other strategies, regardless of the task, and despite the fact that
monitoring and inferencing were reported to have been used significantly more often than
functional or formal practicing. However, as noted by Bialystok, these findings must be
interpreted with caution given the possibility that the criterion test and/or the
questionnaire may not have been adequate measurements of inferencing.

Given that "the diversity of activities subsumed by the general strategy
’inferencing’ is large because of both the many possible types of information upon which
an inference can be based and the many logical or deductive processes which can be used
to generate inference" (Bialystok, 1983, p. 106), Bialystok’s attempt to refine and restrict
the definition of the strategy of inferencing is understandable.” However, the definition
of inferencing assumed in this dissertation contrasts with Bialystok’s in three major ways.

First, the definition assumed here does not exclude the possibility that the
inferencing process may result in intuitive (Implicit, in Bialystok’s terms) linguistic

knowledge. This assumption is supported by the theoretical arguments, presented in

 

'8 Standardized I.E.A. (International Association for the Evaluation of Educational Achievement)
Tests were adapted for three of the four tests given. The four tests were designed along the purpose and
modality parameters: formal/oral (decide the correctness of a spoken sentence), formal/written (fill in the
blank with an appropriate form on isolated sentences), functional/oral (listening for comprehension),
fimctional/written (reading for meaning).

‘9 Her choice of restricting inferencing to a comprehension, functional strategy may in part be due
to her concentration on the area of reading.

73

Chapter 1 and the beginning of this chapter, suggesting that inductive inferencing (or
INE) plays a role in the acquisition of (the intuitive knowledge of) a first and/or second
language. One of the implications of Bialystok’s claims that "the conversational use of
language proceeds perfectly well from an intuitive source" (1983, p. 108) and that
inferencing does not result in intuitive knowledge, would therefore be that inferencing
may never be used in language acquisition. This seems to be a rather drastic
consequence of her model. In light of these arguments, it would seem that an
explanation is nwded of m the result of inferencing is always in the form of conscious,
explicit linguistic knowledge; that is, what is to be gained by positing such a restriction?
Given that no substantive arguments are provided for this position and that arguments in
support of ’inferencing to intuitive knowledge’ are offered, the present definition includes
the possibility that the inferencing process results in intuitive knowledge.

Second, Bialystok’s model excludes the possibility of Explicit Linguistic
Knowledge (of either the L1 or the L2) as a possible source of inferencing. The
definition assumed here does not exclude this information as an inferencing source.
There is a slight contradiction in Bialystok’s scheme: "Other Knowledge refers to
information about languages other than the target language, especially the native
language, as well as the learner’s general knowledge of the world. . . . Part of Other
Knowledge is the learner’s metalinguistic knowledge of language in general" (1983, pp.
106-107, p. 109). Therefore, Other Knowledge may include explicit knowledge about

other languages, and yet, according to Bialystok, while Other Knowledge is a source of

74

inferencing, explicit linguistic knowledge is not.20
Additionally, Bialystok (1983, p. 107) offers the following example of inferencing

from implicit knowledge:

A learner of English as a second language can exploit the morphological information that
the ending -ly signals an adverb, the ending oor refers to an agent, and the ending -tion
is a cluster indicating a noun and is pronounced in a certain way.

Once the "regularity [is] abstracted from intuitive information [and] explicated, [it then]
is accumulated for Explicit Knowledge, the source of analyzed linguistic knowledge"
(Bialystok, 1983, p. 107). Are we to assume then that once regularities have been
abstracted and explicated they somehow lose their ability to function as sources of
inferences?

These inconsistencies and contradictions would be resolved if one assumes that
explicit linguistic knowledge may be used as a source of inferencing. Again, no
explanation is given of why explicit linguistic knowledge cannot be used to make
inferences about unknown or unfamiliar material; or, what is to be gained by excluding
it as a possible inferencing source?

Third, contrary to Bialystok, the process of inferencing assumed in this
dissertation is not restricted to semantic inferencing. According to Bialystok, syntax,
morphology, or phonology may be used only ”to derive meaning from the target
language rather than to infer formal or structural features" (1979, p. 376). It seems
incongruous that formal, and only intuitive, aspects of the L1 or the L2 are used to infer

only explicitly functional aspects of the L2. Again, we are offered no explanation of

 

2° Additionally, this description of the knowledge sources implies that explicit L1 knowledge and
explicit L2 knowledge are somehow different forms of knowledge; this seems to be an unjustified
distinction.

75

w_hy inferencing must be limited to the semantics of language or what is to be gained by
such a limitation.

In sum, while an attempt to refine the definition of the inferencing strategy is
desirable, Bialystok’s restrictions of the strategy are unmotivated and unjustified. The
claims that the strategy of inferencing I) always results in conscious, articulable facts;
2) never uses conscious, formal knowledge of a language as a source; and 3) is restricted
to a semantic, functional purpose are unsubstantiated. The present purpose is not to
develop the concept/role of inferencing within Bialystok’s framework, but rather to
explain how the present use of the term differs from her conceptualization and to
highlight the fact that inductive inferencing (see Footnote 15, Bialystok’s definition 1983,
pp. 105-106) plays an essential role in her model of second language learning.21

Based primarily on the work of Anderson (1983, 1985) and on their research
conducted within an information-processing framework, O’Malley et al. (I985a, 1985b)
and O’Malley and Chamot (1990) attempt to integrate SLA and learning strategies into
an information-processing theory of cognition. According to O’Malley, Chamot, and
Walker (1987, p. 290), ”language can best be understood as a complex cognitive skill,
and that mental processes involved in language parallel the processes used with other
cognitive skills."

Based on interviews and observations, O’Malley and Chamot (1990, p. 1)

 

2‘ In Bialystok and Ryan (1985a, 1985b) and Bialystok (1990), the model has been revised: "The
framework is generated by identifying as independent corntituents of cognition the two skills associated with
structuring knowledge and accessing knowledge. . . . and are referred to, respectively, as the dimensions
for analyzed knowledge and for cognitive control” (Bialystok & Ryan, 1985a, p. 232).
Analyzed/unanalyzed knowledge is roughly equivalent to the implicit/explicit distinction made in the model
reviewed here. Because the inferencing strategy is not explicitly dealt with in the more recent cognitive
framework, this framework will not be covered here.

76

distinguish between three types of learning strategies, which are "special ways of

processing information that enhance comprehension, learning, or retention of the

information," depending on the type or level of processing involved. Their relatively

detailed inventory, adopted from cognitive psychology (Brown & Palinscar, 1982), is as

follows:

Metaoognitive Strategies: Higher order executive skills which may apply to a variety of
receptive or productive language learning tasks.

Selective attention
Planning

Monitoring

Evaluation

Focusing on special aspects of learning tasks, as in planning
to listen for key words or phrases.

Planning for the organization of either written or spoken
discourse.

Reviewing attention to a task, comprehension of information
that should be remembered, or production while it is
occurring.

Checking comprehension after completion of a receptive
language activity, or evaluating language production
after it has taken place.

Cognitive Strategies: Operate directly on incoming information, manipulating it in ways
that enhance learning. . . . Cognitive strategies may be limited in application to the
specific type of task in the learning activity.

Rehearsal
Organization

Inferencing
Summarizing

Deducing
Imagery

Transfer

Elaboration

il/Aff ive tr

Repeating the names of items or objects to be remembered.

Grouping and classifying words, terminology, or concepts
according to their semantic or syntactic attributes.

Using information in text to guess meanings of new linguistic
items, predict outcomes, or complete missing parts.

Intermittently synthesizing what one has heard to ensure the
information has been retained.

Applying rules to the understanding of language.

Using visual images (either generated or actual) to understand
and remember new verbal information.

Using known linguistic information to facilitate a new learning
task.

Linking ideas contained in new information, integrating new
ideas with known information.

ieS° Represent a broad grouping that involves either interaction

with another person or ideational control over affect. Generally, they are considered
applicable to a wide variety of tasks.

Cooperation

Working with peers to solve a problem, pool information,
check notes, or get feedback on a learning activity.

77

Questioning for clarification Eliciting from a teacher or peer additional explanation,
rephrasing, or examples.
Self-talk Using mental redirection or thinking to assure oneself that a

learning activity will be successful or to reduce
anxiety about a task.
(O’Malley & Chamot, 1990, pp. 44-46)

While a computer-based metaphor of cognition is not being adopted in this
dissertation, inductive inferencing as defined earlier in this chapter may be utilized within
such a model. The process of inductive inferencing would involve multiple strategies
which O’Malley and Chamot classify as Cognitive: inferencing, transfer, and elaboration.
Recall Carton’s Method of Inference which provides the foundation of the present
definition:

1. Scan what is known in the L1, the L2, and/or world knowledge.

2. Link new material with what is known.

3. Establish new generalizations.

Stage 1 is roughly analogous to O’Malley and Chamot’s transfer; that is, known
linguistic information is used. Stage 2 would fall under elaboration, the linking and
integrating of new and known material.22 Stage 3 consists of, in part, O’Malley and
Chamot’s inferencing; although a slight modification in their definition is in order. The
authors seem to be limiting inferencing, as Bialystok does, to semantics ( " guess meanings
. . . predict outcomes . . ." 1990, p. 44). As argued previously, I maintain that
inferencing need not be restricted to the domain of semantics. Additionally, included in
O’Malley and Chamot’s (1990, p. 44) definition is "complete missing parts." As

discussed in the section "Inductive Inferencing: A Working Definition,” I maintain that

these missing parts (which may include syntactic structures) may be noticed during the

 

3 However, O’Malley and Chamot do not elaborate on whether the term ”ideas" includes linguistic
information. Additionally, it would seem that "organization" would be a necessary precursor to any kind
of ”elaboration.”

78

linking/elaboration stage, and rather than being completed through inferencing, they may
be used to establish new probably true generalizations regarding the syntax of the new
information. In other words, O’Malley and Chamot’s itemization of strategies does not
allow for learning through induction.

In addition to the present definition of the stages of inductive inferencing lending
themselves, in part, to application within an information-processing framework, the
possibility that this process or strategy may be both conscious and unconscious is

supported within this model of cognitive theory:

Research on learning strategies is based on the assertion that strategies begin as
declarative knowledge that can become proceduralized with practice and, like complex
cognitive skills, proceed through the cognitive, associative, and autonomous stages of
learning. At the cognitive stage, the strategy application is still based on declarative
knowledge, requires processing in short term memory, and is not performed
automatically. The student may have a firm recollection of using the strategy with a
specific task. . . . However, if the strategy application has become proceduralized and
the strategy use is performed automatically, the student may not be aware of using the
strategy. (O’Malley & Chamot, 1990, p. 85)

In an attempt to resolve some of the ”conceptual and classificatory problems”
(Oxford & Cohen, 1992, p. 3) within learning research, Oxford and Cohen (1992, p. 7)
suggest developing "a smaller number of broad strategies with their related tactics listed
underneath each strategy." Based on the research to date which utilizes interviews,
journals, observations, surveys, introspection, retrospection, and factor analytic studies,
they (1992, pp. 7-9) list seven general strategies: 1) forming concepts and hypotheses;
2) testing hypotheses; 3) personalizing mental linkages; 4) embedding new material in
long-term memory; 5) understanding one’s affective state; 6) managing one’s learning

process; and 7) producing oral or written language while lacking adequate linguistic

79
knowledge.23 Relevant to the present discussion is their first strategy.

1. Strategy = forming concepts and hypotheses (building declarative knowledge)
Tactics = simplifying through transfer, simplifying through overgeneralization;
using inferencing to figure out the meaning in the absence of adequate
vocabulary and grammar; categorizing, comparing, contrasting, hierarchically
organizing, analyzing, reasoning deductively, abstracting, creating networks,
propositions, and schemata (conceptual frameworks). (Oxford & Cohen, 1992,
p. 7)

Stage 1 of the present definition of inductive inferencing, scanning known
material, and stage 2, linking known material with new material, are not specifically
itemized as tactics within Oxford and Cohen’s organization of learning strategies and
tactics. Presumably these stages would fall under the tactics of "simplifying through
transfer," "comparing," "contrasting,” and "creating networks." That these stages can
be considered integral to a learning strategy is reflected in Oxford and Cohen’s (1992,
p. 9) ”assumption #3: learning strategies can aid learners in making initial links between
new material and stored information, strengthening existing links, and retrieving
information through declarative networks when needed (Mayer, 1988)."

According to Oxford and Cohen’s scheme, the final stage of the present working
definition of inductive inferencing (establishing new generalizations) would be a strategy
and the steps leading to this result would be classified as either tactics or assumptions.
Additionally, as with O’Malley and Chamot and Bialystok (and contrary to the present
assumption/definition), Oxford and Cohen (1992, p. 7) seem to limit inferencing (a
tactic) to semantics--"using inferencing to figure out the gleam in the absence of

adequate vocabulary and grammar." Importantly, forming concepts and hypotheses

 

2’ Vague areas stifl remain, however. Oxford and Cohen (1992, p. 8) list the tactic of "looking for
the organizing principle" under the strategy of ”managing one’s learning process." It is not clear how this
tactic is distinguished from several of those listed under the strategy of forming concepts and hypotheses:
categorizing and hierarchically organizing.

80

through induction is not explicitly referred to in their organization. However, one could
argue that the tactic of "simplifying through overgeneralization" is a necessary aspect of
induction.

Regarding the issue of conscious awareness as an essential feature of a learning
strategy, Oxford and Cohen (1992, pp. 11-12) "assert that language learning strategies
are indeed conscious behaviors undertaken to improve language learning. . . . If a
learner’s behavior is totally unconscious, then it would simply be referred to as a
process, not a strategy.” It is important to note that the authors do not rule out the
possibility that the m actions or behaviors are a strategy if performed consciously by
one learner and yet are also a process if performed unconsciously by another learner.
This characterization of a learning strategy does not oppose the present working
definition of inductive inferencing.

This cursory review of the learning strategy research supports the assumption that
inductive inferencing as defined in this dissertation may be considered a learning
strategy/process. It may be that inductive inferencing is indeed a combination of several
strategies and processes: Naiman et al.’s (1978) realization of language as a system and
monitoring of L2 performance; Rubin’s (1981) guessing/inductive inferencing and
deductive reasoning; Bialystok’s (1978) inferencing; O’Malley and Chamot’s (1990)
cognitive strategies of transfer, organization, elaboration, and inferencing, or Oxford and
Cohen’s (1992) strategy and related tactics of forming concepts and hypotheses.

While variations in terminology, classification, and essential characteristics of
learning strategies continue, the stages of inductive inferencing outlined here have been

identified within the various classifications and conceptualizations. Additionally, the

81

similarities and differences between the definitions offered by these frameworks and the
present definition have been described.
2.9. TEACHING OF LEARNING STRATEGIES

As highlighted in the previous section, "there is no complete agreement on exactly
what strategies are; how many strategies exist; how they should be defined, demarcated,
and categorized" (Oxford, 1990, p. 17). However, ”despite problems in classifying
strategies, research continues to prove that strategies help learners take control of their
learning and become more proficient . . ." (Oxford, 1990, p. 22).“ Based on this
assumption -- that language learning strategies can improve the development of language
skills -- research has been directed toward the teachability of strategies.

In this section, I summarize some of the major studies which have investigated
the effectiveness of teaching learning strategies. I then discuss some of the issues yet to
be resolved which are reflected in the summarized studies and explain how the present
research may address some of these issues.

Early research in learning strategy instruction focused on the training of just one
or two strategies. One set of studies concentrated on associative memory strategies for
acquisition of new vocabulary. Paivio and Desrochers (1979) found that the pegword,

or hook, method (a list of memorized cue words is used to learn vocabulary or

 

7" However, Gass and Selinker (1994, p. 267) maintain that one of the major obstacles still to be
surmounted is to determine if a given strategy actually facilitates second language acquisition for afl
language learners: "Finally, suppose we can show that good language learners do X, that X is strategic,
and that X in fact does contribute to their language learning. Logically, it does not follow that if X is then
taught to a poor language learner, it will necessarily [emphasis in original] lead to language improvement.
It is not impossible, of course, that the teaching of that X may in fact lead to language improvement. But
the point is that it does not logically follow that it will. What is needed is research that shows that the
strategy under comideration, if used by a so-called bad language learner, does indeed improve the latter’s
consistent performance, thus making the IL more target-like. "

82

grammatical categories) aided English-speaking subjects in the recall of new French
vocabulary. The keyword method (new words are learned through a combination of
auditory and imagery links) has been reported as successful for recalling German
vocabulary (Ott et al., 1973; Desrochers et al., 1989), Spanish vocabulary (Raugh &
Atkinson, 1975), and Russian vocabulary (Atkinson & Raugh, 1975). Cohen and Aphek
(1980, 1981) conducted a study in which English-speaking students of Hebrew were
taught how to use associations” to learn new vocabulary. The students then generated
their own paired associations. The authors found that students reported using their self-
generated associations in order to recall words and ”that their performance was better
when using this retrieval strategy than when they used a new association or used no
association at all" (Cohen & Aphek, 1980, p. 229).

Bialystok (1983) investigated the teaching of the strategy of inferencing in the
context of reading comprehension.26 The working assumption was that inferences are
"tentative propositions of meaning . . . The hypothesis [was] that providing information
that could potentially form a basis of meaningful inferences [would] improve the.reader’s
ability to understand the text” (Bialystok, 1983, p. 114).

The subjects in the study were English-speaking high school students learning
French as a second language. "Four cue conditions were created and examined for their

effects on the learners’ ability to understand the general meaning of a passage and on

 

2’ Associations included, among others, sound, meaning, image, number of letters in word, and
context.

2‘ Hosenfeld et al. (1981) taught reading strategies to high school students of French. Because no
assessment of the efiectiveness of the strategies taught, or of an increase in strategy use, was provided, this
study will not be reviewed here.

83

their ability to translate difficult lexical items in the passage" (Bialystok, 1983, p. 114).
The four conditions were as follows: 1) students were given a picture which summarized
the main point of the reading passage;27 2) students were given dictionaries which
contained all the difficult words in the passage, in addition to several distractors; 3)
students were given 15 minute lessons on how to inference by using target and native
language information (prefixes, suffixes, cognates) and contextual information; and 4)
students in this group received no cues.

After reading the passages in one of the four cue conditions, students were given
two tests in English. The first test consisted of short-answer questions on the meaning
of the passage; the second test contained difficult vocabulary items which the students
were asked to translate into English. Bialystok (1983, p. 116) found that 1) all subjects
performed better on the comprehension test, and 2) "the picture improved performance
on the comprehension test only and the dictionary and lesson improved performance on
both the comprehension and the vocabulary tests."

The study was replicated with another four groups of students and somewhat
different results were obtained. Performance on the comprehension test was better for
the picture cue group, rather than being equally effective as in the first study.28

Additionally, performance on the vocabulary test improved only in the dictionary cue

 

2’ This can be equated with Canon’s extra-lingual or contextual cue.

2‘ Bialystok notes that this may due to the fact that in the first study the picture cue group was the
lowest achieving academically. In other words, if they had been more competent initially, their
comprehension scores might have surpassed those of the other groups in the first study also.

84

group; it did not improve for the lesson cue group.29

Bialystok (1983, p. 121) concluded:

The results of the present study suggest that the deliberate provision of different kinds
of contextual information and instruction in inferencing techniques are pedagogically
expedient in second language reading comprehension. The lesson provided subjects with
a heuristic device for dealing with difficult information; stable long-term effects of this

instruction were neither measured nor expected. . . . students were able to understand
the techniques described in the presentation and to apply them to similar material on their
own.

The largest learning strategy research agenda has been put forth by O’Malley,
Chamot, and their associates.30 In addition to developing the Cognitive Academic
language learning Approach (CALLA) instructional model which is designed to
integrate learning strategy training, content topics, and academic language development
for adolescent ESL students, these researchers have set out to address the following
issues: 1) the effectiveness of strategy training on the integrated tasks of speaking and
listening; 2) strategy training performed by teachers versus researchers; 3) the training
of a combination of strategies; and 4) the transferability of strategies across tasks. One
comprehensive, representative study (O’Malley et al. , 1985b) which relates to the present
discussion will be reviewed here.

The goal of this study was to determine if combinations of learning strategies

would facilitate performance in speaking, listening, and vocabulary acquisition. Subjects

 

2’ Bialystok attributes this to the fact that the lesson, by necessity, was altered somewhat; it was
shorter and given to a larger group of students; thus, the lesson was more in the form of a lecture rather
than class participation.

3° Other learning strategy studies which are peripheral to the present discussion and therefore won’t
be summarized here include, among others: Chamot and thper (1989) investigated the effects of teacher
interest, techniques, and ability to motivate students on the success of learning strategy training. A series
of studies by Chamot and O’Malley (1986, 1987), which are part of the Foreign language Course
Development Study, are designed to 1) determine whether and how foreign language teachers incorporate
learning strategy instruction into the classroom, and 2) utilize teachers’ knowledge of student attitude and
course objectives to incorporate learning strategy instruction into a curriculum.

85

in the study were intermediate-level adolescent ESL students (one-third Hispanic, one-
third Asian, and one-third other ethnicities). Students were assigned to one of three
treatment groups. Group 1 received instruction in metacognitive, cognitive, and
social/affective strategies. Group 2 received instruction in cognitive and social/affective
strategies. Group 3, the control group, did not receive any learning strategy instruction.

50 minutes a day for eight days, students received instruction and practice in
using learning strategies. Explicit directions for strategy use were faded out over the
training period. Each day students practiced two of the three skill areas (listening,
speaking, vocabulary). The same strategies were always presented with a given skill
area, although the tasks in each skill changed. Listed below are the language activities

and accompanying learning strategies (O’Malley & Chamot, 1990, p. 173).

Wk Strategy Typ_e Mgr

Vocabulary (word lists) Metacognitive Self-evaluation
Cognitive Imagery and grouping
Social/affective None

Listening (5-minute Metacognitive Selective attention

lecture on an academic Cognitive Note taking

topic) Social/affective Cooperation

Speaking (2-minute Metacognitive Functional planning

presentation on a familiar Cognitive None

topic) Social/affective Cooperation

An example of how the strategies were integrated into the tasks is as follows:

Vocabulary: Students were presented with ten-item word lists comprised of tangible items
and encouraged to group the words based on meaningful classifications as they studied
them and to imagine themselves interacting with the object while using its name. After
a 5-mimite study period, students were tested for recall. Students were then given the
opportunity to review their tests and were encouraged to analyze which strategies worked,
which didn’t work, and why. (O’Malley & Chamot, 1990, p. 173)

Students were pretested, posttested, and given four daily tests in the three skill

86

areas. A multiple choice listening test was designed to assess students’ recall of
information from a 5-minute videotape on academic topics. ”The speaking test consisted
of ratings on a two-minute speech on one of three topics: a personal experience, their
own culture, or an academic subject. Students were given time to prepare the talks and
then tape recorded the talk before a small group of students. The tapes were scored
blind by a panel of judges who rated the talks on delivery, appropriateness, accuracy,
and organization” (O’Malley, 1987, p. 135). The vocabulary test consisted of multiple
choice items and fill-in-the-blank.

Results of the speaking task indicated that the metacognitive group scored higher
than the cognitive group, which scored higher than the control group. On the listening31
and vocabulary tests, no significant difference was found among groups. Interestingly,
on the vocabulary test, ethnic differences in strategy use were found. Asians in the
control group outperformed Asians in the training groups, while Hispanics in the training
groups outperformed Hispanics in the control group. Teachers reported that Asian
students resisted the strategy training}2

O’Malley and Chamot (1990, p. 175) conclude that

this study demonstrated that strategy training can be effective in a natural classroom
enviromnent with integrative language tasks such as speaking and listening, although it
suggests that training effectiveness depends on the difficulty of the materials or the rate
at which cues for strategy use are faded over time. The fact that the language tasks on
which the strategies proved effective were academic language skills suggests that this type
of training has promise for improving the learning ability of minority language students.

Aside from the caveats that are a part of any study (e.g., the reliability and

 

3‘ The authors suggest that this finding may be a result of the material being too difficult, decreased
interest in the material, and/or the fading of the directions for strategy use.

32 This finding agrees with ethnic differences in strategy use reported by Politzer (I983).

87

validity of testing instruments, the effects of individual learner variables), the two studies
just summarized are based on assumptions which may need to be reexamined. In both
studies the researchers taught very explicit procedures which were assumed to be
strategies/tactics, which in turn were assumed to facilitate performance in a specific skill
area. For example, the picture cue in the Bialystok (1983) study was assumed to provide
contextual information which would engage an inferencing strategy; the use of this
strategy would then be reflected in higher performance on a comprehension test. It could
be that the picture cue group did not use inferencing at all. Rather, the picture simply
supplied the students with the necessary information. Similarly, given the tenuous results
in the O’Malley et al. (1985b) study, a justifiable question would be: Were the strategies
taught actually related to or necessary for improved performance on the skill in question?
These strategies were associated with these language tasks based on previous studies in
which information was collected through interviews and observations (O’Malley et al.,
1985a) and survey questionnaires (Bialystok, 1979). Therefore, that the problems with
initially identifying strategies may transfer to and complicate subsequent research is
exemplified.

A more global approach is taken in the present research33 as a result of the
possibility that specificity of the strategy/tactic is not appropriate to or does not target
the skill area in question. In other words, the assumption of the present research is that
improvement in the general cognitive ability or strategy of inductive inferencing will be

utilized in the linguistic domain. This assumption is supported by McCawley (1983, p.

 

33 While methodology is briefly mentioned here, detailed descriptions of the research design will be
reserved for Chapter 3.

88
180):

I reject the remarkably popular non-sequitur that has taken many linguists from the quite
reasonable proposition that there are innate mechanisms specific to the acquisition of
language to the much less plausible proposition that general purpose learning mechanisms
play no role in language acquisition. If there are such things as general purpose learning
faculties, it is bizarre to suppose that they shut off when language is being acquired.
language involves not only specifically linguistic units and relations but also factors that
play a role in many cognitive domains.

The fact that some students seemed to resist the strategy training in the O’Malley
et al (1985b) study underscores the issue of whether strategy instruction should be direct
(students are informed of the purpose of the strategy) or embedded (students are not
informed of the purpose of the materials, which are designed to elicit strategy use). As
a corollary of the global approach adopted here and to avoid the possibility of student
resistance, embedded learning strategy instruction was utilized in the present research.34

As summarized above, all previous studies have focused on vocabulary
acquisition, speaking, or listening and reading comprehension. In reviewing the memory
training studies on vocabulary acquisition, Thompson (1987) points out a number of
difficulties with the techniques used in the vocabulary studies. Among those she cites
are: it has not been shown that the devices aid in the acquisition of meaningful
information; more effort is sometimes required on the part of the student to learn the
associated relationships than simply to learn the words by rote; and pronunciation
difficulties may hinder mnemonic techniques. An attempt to teach strategies for the
integrated skills (that is, meaningful information) of speaking, listening, and reading has
also proved to be problematic. In the Bialystok (1983) study, it is implied that the

reading passage was too difficult for the first picture cue group. In the O’Malley et al

 

3‘ Additionally, direct strategy instruction requires teacher training, which was impossible to
implement in the present study.

89

(1985b) study, the authors attribute a portion of their findings to the possibility that the
listening task was too difficult. In other words, extraneous variables associated with the
multiple demands of the task have entered into these studies obscuring both the use of
a strategy and the effectiveness of the training of a specific strategy. Therefore, in the
present study one specific syntactic structure is isolated for investigation, and the effect
of the teaching of a specific strategy, inductive inferencing, on the acquisition of the
ungrammaticality of that structure is examined.

Finally, in all of the studies to date the long-term effects of strategy training have
not been investigated; single post-tests have been administered immediately following
instructional periods. The present study attempts to examine the (relatively) long-term
effects of strategy instruction by posttesting subjects twice -— immediately after instruction
and again three weeks later.

In this section I have summarized some of the studies which have investigated the
effectiveness of teaching various learning strategies. While these studies leave many
questions unanswered, their results indicate that attempts at strategy instruction need not
be abandoned. I have highlighted three of the unresolved issues (the global nature of
learning strategies, the effect of strategy instruction on the acquisition of syntax, and the
long-term effects of strategy instruction) in the learning strategy research and suggested
that the present research may provide some insights into these issues.

2.10. CHAPTER SUMMARY

In this chapter, a definition of inductive inferencing from the fields of philosophy

and psycholinguistics has been given. Specific examples of the use of induction in the

processing of meaning in a first language have been described. A process of inferencing

90

in SLA and its similarity to the use of Indirect Negative Evidence have been explained.
It has been argued that the acquisition of the ungrammaticality of a specific syntactic
structure depends on the use of inductive inferencing. Learning strategy research in SLA
has been reviewed in the context of the working definition of inductive inferencing used
in this study. Finally, a selective summary of the research investigating the teaching of
learning strategies has been provided. Issues which remain problematic in this research
area have been indicated. The present research, which is designed, in part, to address

these issues, is described in Chapter 3.

CHAPTER 3
METHODOLOGY

I begin this chapter with a general overview of the research design and
procedures. I then provide biographical sketches of the volunteers who took part in the
study. The majority of the chapter is devoted to detailed descriptions of and justifications
for the testing and instructional materials which were used.

The objective of the present research, which was discussed in Chapter Two, is
to investigate the possible role of inductive inferencing in the acquisition of the contrasts
in 3-1 and 3-2 by native speakers of English learning French.

(3-1) Marie oublie souvent ses devoirs.

*Mary forgets often her homework.

(3-2) *Marie souvent oublie ses devoirs.

Mary often forgets her homework.

While both English and French allow Subject-Verb-Object-Adverb (SVOA),
Adverb-Subject-Verb-Object (ASVO), and Subject-Auxiliary-Adverb-Verb(SAuxAV)
word orders, the languages differ in that only French allows Subject-Verb-Adverb-Object
(SVAO) order, and only English allows Subject-Adverb-Verb-Object (SAVO) order.‘
Importantly, if one assumes lack of negative feedback, nothing in the French positive

input informs learners that the SAVO order allowed in their native language (English)

is disallowed in the target language (French).

 

' See Introduction, Footnote 2, for exceptions to this characterization.

91

3.1. PROCEDURES OVERVIEW

All volunteers took three pretests. One test was a Word Order Correction Task,
which was in the form of a cartoon story. Students were asked to correct those sentences
which were incorrect. The second test was an Acceptability Judgment Task,2 which
contained pairs of sentences, and students were asked to decide whether both sentences
were correct, both sentences were incorrect, only one sentence was correct, or they did
not know. The Word Order Correction and Acceptability Judgment Tasks were designed
to test students’ knowledge3 of adverb placement in French. Test sentences contained
adverbs of frequency or manner and lexical verbs (no auxiliaries) in the present tense.
The third test was Raven’s Standard Progressive Matrices, which is a non-verbal test of
inductive reasoning. Problems are presented in abstract figures/designs and combine
elements of analogies, series completions, and classifications. Students were allowed
unlimited time to complete all tasks. The order in which tests were taken was alternated
both within a testing session and for each student.

Subjects were divided into two groups based on class section. Teachers were
asked not to correct students with respect to adverbs, nor were they to teach students the
rule of adverb placement. For a two week period (8 days) in Group 1, which I will refer

to as the noninferencing group, the teachers’ regular lessons were altered so that,

 

2 The terms grammaticality judgment, acceptability judgment, and preference are often used
interchangeably. However, as Birdsong (1989, p. 73) highlights in citing Newmeyer (1983, p. 51),
”grammaticality . . . is a theoretical construct, it is not directly accessible to the intuitions of the speaker
of the language . . . The feelings speakers have about the well-formedness of sentences in their language
are referred to by the term acceptability." Thus, as Gass (1994, p. 303) points out, grammaticality is
inferred through the use of acceptability judgments. In this dissertation, the terminology is used in its
strictest sense.

3 The subject of "knowledge" is discussed later in the chapter in the sections ”Acceptability
Judgments" and "Testing Materials."

93

wherever possible, the exercises contained adverbs. I refer to this as adverb input flood.
Fifteen students in the noninferencing group participated in the pretest and the first
posttest (see below); thirteen of these students took the second posttest. In Group 2,
which I will refer to as the inferencing group, in addition to the adverb flood, students
received verbal inductive inferencing exercises in both French and English. These
consisted of exercises in analogies, series completions, and classifications. Thirty
students in the inferencing group participated in the pretest and first posttest; twenty-nine
of these students took the second posttest.

Immediately following the two week instructional period, each group was
posttested using the Word Order Correction Task and the Acceptability Judgment Task.
Three weeks after the first posttest, students were posttested again to determine any long
term effects of the teaching material.

All tests were administered at the regular class times on a Wednesday, the day
of the week the classes did not meet (see "Teaching").

In an attempt to take into account the possible confounding variable of teaching
style, a questionnaire was administered to the teachers after the two week instructional
period. Even though the teachers were working from the same syllabus, I could not be
assured that all of the students were receiving the same type of input. An attempt was
therefore made to quantify the input based on the responses to the questionnaires.

Following the two week instructional period, a questionnaire was also
administered to the students. The goal of this questionnaire was to obtain a descriptive
analysis of the students’ perceptions of the input and to ensure that no corrective

feedback or explicit rule formation with respect to adverb placement had been provided

94

to the students.
3.2. SUBJECTS

Forty-five volunteers were obtained from four sections of intensive first year
French (French 150) at Michigan State University. Biographical information was
obtained from a cover sheet (Appendix A) attached to the tests. The database consists
of 4 males and 41 females. Subjects ranged in age from 17 to 22 years old (average age
18 years 4 months). Previous years of studying French ranged from 2 to 5.5 (average
years of previous French instruction 3 years 7 months). Four subjects reported that they
grew up speaking a language in addition to English: Latvian, Croatian, Amharic and
French, and Vietnamese. Of these, Iatvian, Croatian, and Amharic are used in addition
to English in the respective subjects’ homes. All of the subjects were American with the
exception of two Canadians. Seven subjects reported living in French-speaking
countries; of these three had lived in a French-speaking country for a duration of one
month, one for a duration of three months, one for a duration of one year and one
month, one for a duration of eight years, and one for a duration of nine years.“
3.3. TEACHING
3.3.1. Adverbs

The text used in French 150 is Bonjour, ca va? (Rochester, Muyskens, Omaggio
Hadley, & Convert-Chalmers, 1991). The material scheduled for the two week period
of the study included: passé compose, imparfait, question formation, reflexive

pronominal verbs, reciprocal reflexive verbs, and prepositions with geographical names.

 

“ The test scores of the students who had lived in French-speaking countries did not differ from the
other students’ scores; therefore, these students were not eliminated from the database.

95

A summary of the material covered in the text during the instructional period is provided
in Appendix B. This semester-long course is designed for students who I) do not quite
place into second year French; 2) do not feel confident enough to take second year
French; and/or 3) have not been exposed to French for a period of time. The four
sections were taught by four different teachers. The course meets for fifty minutes, four
times per week (Monday, Tuesday, Thursday, Friday) and can be classified as
communicative in approach. Students had not been explicitly taught adverb placement
in French 150 prior to the study.

Following Trahey and White (1993), the goal was that students would be exposed
to only positive evidence of adverb placement in French. As adverbs are not frequently
present in naturally occurring classroom discourse (Trahey & White, 1993), adverbs of
frequency and manner were integrated into the regular lessons, if the exercise was
conducive to such manipulation. To the extent possible, sentences were used in which
the verb was in the present tense and with main verbs rather than auxiliaries, because
French and English do not differ with respect to adverb placement with auxiliaries.
Materials contained adverbs in all possible positions, but were predominantly in the
SVAO word order. Activities included games, stories, dictées, and exercises which were
done individually, in groups or pairs, and/or as a class. Some of the exercises in the text
already contained adverbs, and the teachers were asked to be sure to incorporate these
into their daily lessons. As mentioned previously, teachers were asked not to correct any
errors of adverb placement which the students might produce. Importantly, students did
not receive form-focused instruction in adverb placement but rather were flooded with

examples of this structure. The altered lessons used during the two week instructional

96
period are given in Appendix C.5
3.3.2. Inferencing

The lessons taught to this group included both adverb flood (Appendix C) and
inferencing exercises as shown in Appendix D.

The inferencing lessons were drawn from four sources: Carton (1966), the
students’ text (Bonjour, ca va .7), Sternberg (1988), and those which I created. The
exercise from Carton (1966, pp. 83-86) and those from the students’ text (e.g., ”Mise
en pratique: Anticipating content” pp. 165-167) can be classified as explicit skill
instruction (O’Malley & Chamot, 1990). The inductive reasoning exercises adapted from
Sternberg and those which I developed can be considered embedded (see Chapter Two,
"Learning Strategies" for definition) in that the strategy being taught was not made
explicit to the students.“ Based on the assumption, argued in Chapter Two, that the
inductive inferencing process is a general cognitive process which may be utilized in
various domains, exercises were presented in both French and English. For those
exercises which were conducted in French, and which did not focus on acquisition of
new vocabulary, an attempt was made to use vocabulary items which the students would
know so that students would not be hindered in practicing the inferencing process as a
result of not knowing the vocabulary.

Example 3—3 below (Carton, 1966) illustrates a portion of an explicit lesson taught

to students to help them guess the meaning of a vocabulary item unknown to them:

 

5 All exercises were checked by a native speaker of French.

“ This does not mean, however, that the reason for the amwer to an analogy, for example, was not
explained to the students. Rather, it means that the targeted strategy was not explained to the students.

97

(3-3) Questions to Facilitate Inferencing

A. Is it a NOUN?

1. Grammatical questions

a. Is it singular or plural?

b. What adjective modifies it, if any?
What gender is it?
. Is it the subject or object of a verb?

as

2. Pragmatic questions

a. Does it refer to person, things, conditions, events, qualities?
b. How many?

c. What kind?

B. Is it a VERB?

1. Grammatical questions

a. Does it show future, present, or past?

b. What is its subject?
Exercises were used during the two week instructional period which were intended to
allow students to practice Carton’s method. For example, an exercise used on October
7 is shown in 3-4.
(3-4) Write the following sentences on the board and ask the questions listed below.
s’est levé a six heures.
s’est regardé dans le miroir.

. Je me leve lentement a six heures.
se leve rapidement a neuf heures.

 

 

war-e

 

Are the subjects masculine or feminine? How do you know?

In this example, students must know to look to the verb for a possible answer to
the question. After identifying the verb as reflexive pronominal, they can then ask
themselves what information, if any, the verb provides them with respect to the subject.
Once they identify the verbs in sentences 3 and 4 as present tense, they can then
conclude that the gender of the subject cannot be determined. After identifying the verbs

in sentences 1 and 2 as passé compose, they then can ask themselves if agreement

98

features are masculine or feminine, and thus conclude that the subjects must be
masculine.

Listed in 3-5 below is an example of an exercise drawn from the students’ text
which explicitly explains how students can use clues to help them increase their reading
comprehension.

(3-5) Avant de lire.
Anticipating content. Text illustrations, titles, and lead lines or highlighted
material are valuable aids in reading. Before starting to read a passage, you can
examine them for clues about the content of the accompanying text. Clues serve
as advance organizers to the reading. In-text illustrations and captions can also
clarify meaning during the reading process itself, as visual representations can
convey specific and general, explicit and implicit information. Readers who start
with a working hypothesis when confronted with an unfamiliar text seem to read

more quickly and with greater understanding. (Rochester et al., 1991, pp. 165-

166)

While eductive ability may be unchanging, it has been suggested that it can be
developed, refined, and displayed in activities which people value (Raven, Raven, &
Court, 1991, p. G5). Sternberg (1988) is one of the strongest proponents of the
possibility for individuals to improve their performang in activities which require
inductive reasoning and has set out to develop exercises which will achieve this goal.
The majority of the inferencing exercises were therefore drawn from Sternberg (1988).
He enumerates four processes which can be subsumed under inductive reasoning:
encoding, inference (”the discovery of one or more relationships between objects or
events,” p. 117), mapping ("the recognition of the relationship between two
relationships," p. 130) and application ("applying a relation that has been previously

inferred,” p. 132). In addition to exercises in these individual processes (see below),

students were given exercises in analogy, series completions, and classifications which

99

require the use of all four processes. Examples of each are provided in 3-6 through 3-8.

(3-6) Aoalogy
Parent is to father as child is to
a. uncle
b. adult
c. son
d. brother (p. 135)

(3-7) Series Completion

Seed, seedling, sapling: teenager,
a. adolescent
b. child
c. adult
(1. infant (p.139)

(3-8) ngsifigtion

Book,

chapter, page

magazine, letter
publisher, printer

shelf, library (p. 147)

9.057!”

In the instance of analogy, for example, first the various terms must be encoded
(identified); obviously, the inductive process may be interrupted here if the terms of the
analogy cannot be identified. Next, a relationship between the attributes of the first two
terms, a father is a parent, is inferred. Then, the relationship that links the first half of
the analogy to the second half must be mapped or connected. Finally, the relationship
inferred between the first two terms (from the third) is applied to each of the possible
answers. Thus, a father is a parent and a son is a child.

As mentioned above, students were also given exercises in the individual
processes which are components of inductive reasoning. For example, an exercise was

created which combined vocabulary and conjugation review (from the regular lesson)

100

with inference. A portion of an exercise used on October 4 is listed in 3-9.

(3-9) Which word in each column does not belong? Why?

beau partir
chaud monter
froid rentrer
vent donner
pleut tester

In this exercise, students must infer the relationship between the words in the
column by recognizing the relationship each has to the expression "11 fait..." (column 1)
and "étre" (column 2). This inference is then applied in order to eliminate 'pleut" and
"dormer. "

Appendix D provides a complete list of the inferencing exercises and the process
which the exercise is intended to target.

3.3.3. Questionnaires

As stated in the "Procedures Overview” section, a questionnaire (Appendix E)
was administered to the teachers after the two week instructional period. The goal was
to be able to quantitatively compare across class sections and determine the different
types of input, if any, provided to the students. Teachers were asked to indicate whether
the activity was conducted individually (4 points), as homework (4 points), as pairs (3
points), as a group (2 points), as a class (1 point), or was not done (0 points).7 For those
exercises which were assigned as homework, teachers informed me if a student did not
do the homework. The reasoning behind this quantification is that if an exercise is done

individually, presumably the student is attending to the task and receiving the input.

 

7 It was possible for a student to receive more than 4 points for a particular exercise. For example,
if the exercise was done as homework (4) and then gone over in class (1), the total points would equal 5.

101

Whereas, the possibility exists with class activities that students are not paying attention
and thus not receiving the input. If one assumes that the maximum amount of input is
received if exercises are conducted either as pairs or individually and then reviewed as
a class, the highest adverb input score possible was 151.

An inferencing input score was also assigned for each student. The total points
possible for a student in the inferencing group was 8 (1 point for each day). Those
students in the noninferencing group all received an inferencing input score of 0.

Teachers informed me of the students who were absent and which day(s) they
missed. The adverb input score and inferencing input score of those students were
adjusted accordingly.

As stated previously, a questionnaire (Appendix F) was also administered to the
students following the two week instructional period. As I was not permitted to observe
or record the classes, this questionnaire, coupled with the teachers’ questionnaires, was
a means to obtain a descriptive analysis of the classroom procedures and methodology.
The primary goal of the questionnaire was to determine if corrective feedback or explicit
instruction with respect to adverb placement had been given to the students. In addition
to the questions dealing with this topic were similar questions regarding the passé
composé, a grammatical point which I knew had been explicitly taught; the purpose was
to be able to compare students’ responses to whether or not the two structures were
explicitly taught. Secondary goals of the questionnaire were to gain information
regarding the students’ perceptions of the similarities and differences between French and
English, the ease or difficulty of learning French, in addition to their reasons for taking

French.

102
3.4. ACCEPTABILITY JUDGMENTS

Valid and reliable measures of second language learners’ grammatical knowledge
continue to be the topic of debates among SLA researchers (Schachter, Tyson, & Diffley,
1976; Chaudron, 1983; Sorace, 1988; Birdsong, 1989; Eubank, 1989; Bley-Vroman &
Chaudron, 1990; Ellis, 1990, 1991; Schachter & Yip, 1990; Christie & Lantolf, 1991;
Tarone, Gass, & Cohen, 1994). "Perhaps one of the most controversial (and yet most
commonly used) methods of doing second language research is through the use of
intuitional data. In broad terms, intuitional data refer to a type of performance in which
one is asked about the language rather than being asked to use the language" (Gass &
Selinker, 1994, p. 35). Although lost sight of at times, it is now accepted knowledge
that these intuitional data, as obtained by current methodology, are not a direct reflection
of a learner’s competence, which is an abstraction or theoretical construct (Chomsky,
1986; Sorace, 1988; Birdsong, 1989; White, 1989; Ellis, 1990; Gass, 1994).8

In an attempt to create more valid measures of grammatical competence,
researchers have focused on identifying what Sorace (1988, p. 172) refers to as ”extra-
grammatical factors" which may influence or be involved in the intuitional process.
Among these factors are: processing constraints, which may result from vocabulary or
sentence length and/or complexity (Schachter, 1989; Schachter & Yip, 1990; Cowan &
Hatasa, 1994); presentation of the material, which includes order, position, and
contextualization of the items in addition to mode of presentation (Cowan & Hatasa,

1994); personality or the ’yea-saying,’ ’nay-saying,’ or ’maybe-saying’ factors (Felix,

 

3 However, Sorace (1993) reports that the technique of magnitude estimation, given as a timed
procedure, may tap tacit, not metalinguistic, knowledge.

103
1988; Bley-Vroman, Felix, & Ioup, 1988); linguistic training of the informants; and
judgments by the informants of something other than the grammatical structure under
investigation (Kellerman, 1985). Concern for these factors has thus resulted in a certain
amount of variation in the design of acceptability judgment tasks.
Gass (1994, p. 310) enumerates

three dimensions in which [acceptability] judgments vary. First researchers differ in
whether or not they ask learners to correct those sentences that are judged
ungrammatical. . . . Second, differences occur in what is being asked: In some cases
learners are asked for judgments on single sentences; in others they are asked for
preference judgements. . . . Within [acceptability] judgment tasks, another difference
arises: Some sentences are contextualized; some are not. A third dimension has to do
with the number of possible responses that one can give. In some cases, responses are
dichotomous; a sentence can be either grammatical or ungrammatical. In others, there
is a range of possibilities that include the degree of confidence a learner has in making
responses. Additiomlly, researchers vary widely in the number of sentences subjects are
asked to give judgments about, ranging from 30 or 40 to more than 200.

In discussing the methodological considerations of studies employing acceptability
judgments, Chaudron (1983,p. 369) emphasizes that ”it seems mandatory that, though
they are indicators of subjects’ knowledge or values, judgments should be validated by
other measures on the same or comparable items and subjects.” Among those measures
he lists are: 1) productive transformations, imitations, or completions; 2) ability to
identify and correct errors; and 3) other judgments or performance.

The present study attempts to respond to these issues of validity9 by obtaining
judgments from two somewhat different procedures, a Word Order Correction (WOC)
Task and an Acceptability Judgment (AJ) Task, which incorporate various aspects of

9

Gass three dimensions. The WOC Task consisted of contextualized sentences

(dimension 2) and subjects were asked to correct those sentences which were incorrect

 

9 Other validity issues are implicitly addressed later in the chapter in the description of the tasks.

104

(dimension 1). The AI Task consisted of pairs of sentences (dimension 2) and subjects
were asked to decide if both sentences were correct, both were incorrect, only was
correct, or they did not know (dimension 3).

In discussing the issue of the reliability of acceptability judgments, it is necessary
to maintain a clear distinction between the reliability of informants’ responses and the
reliability of researchers’ interpretation of the data. For example, Bley-Vroman, Felix,
and Ioup (1988) allowed for a Not Sure response and counted it as an Incorrect response;
whereas, Coppieters (1987) counted the Not Sure responses as Correct. Similarly, on
4 point scales of Clearly Grammatical, Probably Grammatical, Probably Ungrammatical,
and Clearly Ungrammatical, the middle two categories have been at times kept distinct
(Schachter & Yip, I990; Gass & Ard, 1984) and at times combined with the responses
at the top and bottom of the scale (Schachter, 1990). Furthermore, Gass (1994, pp. 308-
309) argues that recent studies attesting to the unreliability of acceptability judgments
may be more a result of methodological flaws (Ellis, 1990) or an ’anti-acceptability
judgment’ bias on the part of the researchers in interpreting the results (Christie &
Lantolf, 1991).

Sorace (1988) suggests that there are two solutions to the problem of inconsistent
learner judgments. The first solution is that ”sentences about which there is no clear
judgment are not sentences that should be included in our descriptions of second-language
grammars" (Gass, 1994, p. 308). For this reason, the AI Task in the present study

included a Don’t Know category, and if this response was circled, the item was not

105

counted as either Correct or Incorrect. 1°

The second solution is to acknowledge and accept the fact that learner’s grammars
are inherently indeterminate. That is, learners’ grammars are incomplete, unstable, and
constantly evolving (Adjemian, I976; Sorace, 1988; Gass, 1994).‘1 Therefore, while on
the surface the acceptability judgment instrument may appear unreliable, the possibility
exists that inconsistent responses are actually representative of an indeterminate grammar.
The issue then becomes to distinguish determinate from indeterminate knowledge.

In an investigation of the reliability of acceptability judgment data, in which
reliability was considered a function of syntactic constraints,” Gass (1994, p. 320)

concludes:

The data do not provide evidence for the view that judgment data are unreliable. Nor
do they provide support for the view that individuals behave in an inconsistent manner.
On the other hand, there is evidence to suggest that low reliability occurs in just those
areas where greater indeterminacy is predicted. . . . By comparing judgment data with
data based more directly on linguistic principles, it has been shown that judgment data
can, when used properly and appropriately, be useful in second-language acquisition
research. In sum, we find that grammaticality judgments are indeed reflective of patterns
of second-language use.

 

‘9 Additionally, on pairs of SAVO-SVAO sentences on the A] Task, respondents could only get a
correct score if they chose the SVAO word order, not if they chose the Both Right response. The number
of Both Right responses is also tabulated. In this way, one is able to obtain a clearer picture of which
structures are and are not being accepted by the subjects. This is discussed in detail in "Scoring
Procedures. "

” While Birdsong (1989, p. 73) states that "acceptability is in principle a hierarchical or scalar
concept, while grammaticality is not,” he acknowledges Mohan’s (1977) notion of ’fuzzy grammar’ which
Sorace (1988) in part bases her work on.

‘2 The acceptability judgment test comisted of 30 individual sentences (6 distractors and 4 sentences,
2 grannnatical and 2 ungrammatical, of each of the 6 relative clause types. Subjects were asked to judge
each sentence as Correct or Inconect and then to indicate their degree of confidence on a scale from +3
(Definitely Correct) to -3 (Definitely Incorrect).

106
3.5. TESTING MATERIALS

3.5.1. Adverb Placement

The Word Order Correction Task and the Acceptability Judgment Taskl3 were
those used by White (l991a, 1991b; 1992b) and Trahey and White (1993). These tests
were translated from English into French and checked by a native speaker of French.

As implied in the previous section, "Acceptability Judgments," many ". . .
variables, along with dimensions of skill, mental representation of linguistic knowledge,
and retrieval and application of that knowledge, add up to a distinctly Heisenbergian
picture of the language competence that is available through metalinguistic performance"
(Birdsong, 1989, p. 69). For this reason, along with the fact that students were allowed
unlimited time on all tasks (see Footnote 7), I assume that the tests used here represent
metalinguistic performance and not competence.
3.5.1.1. Word Order Correction Task

This task (Appendix G) was a cartoon story. Instructions were in English.
Students were told that in some of the sentences, some of the words were in the wrong
place. They were asked to correct those sentences which were incorrect by circling the
word which was misplaced and drawing an arrow to the location where the word should
be. An example was provided. A single version of the task was administered at the

three testing sessions. The story contains 33 sentences, 16 of which involve adverb

 

'3 The cartoon task was originally referred to as a Grammaticality Judgment/Correction Task and
the multiple choice task was referred to as a Preference Task. Following Footnote 1, I have chosen to use
different terminology. In the strictest sense, a Preference Task asks students to decide which sentence they
prefer. I refer to this task as the Acceptability Judgment Task because this task asks students to decide the
correctneg of the sentences. For the sake of clarity, in order to clearly distinguish the cartoon task, which
is another form of an acceptability judgment task, from the multiple choice task, I refer to this task as a
Word Order Correction Task.

107

placement. Of those 16, 4 sentences were in the ungrammatical SAVO order, 3 were
in the ungrammatical SAVPrepositional Phrase (SAVPP) order,“ 2 were in the
grammatical SVAO order, 2 were in the grammatical SVAPP order, and 5 were in the
grammatical ASVO or SVOA order. Ten sentences were grammatical and did not
contain adverbs. Seven ungrammatical distractor sentences were also included in the
cartoon. As an indication that students were attending to the task, at least three sentences
had to be altered in order for the test to be included in the results.
3.5.1.1.1. Scoring Procedure

Two scores for each student were computed, an SAVO error score and an SVAO
correct score. The SAVO error score consists of the number of acceptances of or
changes to this ungrammatical order. The SVAO correct score consists of the number
of acceptances of or changes to this grammatical order. In order to take into account
those sentences, for example, which were changed from SAVO to ASVO or SVOA
which clearly is not showing an acceptance or usage of the French word order, the
correct and error scores were further classified according to the guidelines shown in 3-10
and 3-11.

(3-10) Correct Score

C1: Unchanged from SVAO
C2: Changed SAVO to SVAO

C3: Changed SAVO to ASVO or SVOA
C4: Changed ASVO or SVOA to SVAO

 

‘4 Following Trahey & White (1993), SAVPP sentences were counted as SAVO and SVAPP
sentences were counted as SVAO, because none of the sentences with prepositional phrases were
problematic such as:

John quickly went to his car.

John went quickly to his car.
In other words, all the sentences with indirect objects clearly followed the same constraints as sentences
with direct objects with respect to adverb placement.

108

(3-11) Error Score
E1: Unchanged from SAVO
E2: Changed ASVO, SVOA, or SVAO to SAVO
E3: Changed SVAO to ASVO or SVOA
Sixteen of the sentences contained adverbs, therefore, the maximum possible error is 16
and the maximum possible correct score is 16.
3.5.1.2. Acceptability Judgment Task
This task (Appendix H) consisted of 32 pairs of sentences. Instructions were in
English. Students were told that in some of the sentences, some of the words were in
the wrong place. They were asked to decide if both sentences were correct, both were
incorrect, only one was correct, or they did not know. Four examples were provided.
A sample item from the test is provided in 3-12.
(3-12)
a. Les enfants détestent les devoirs ordinairement.
b. Les enfants ordinairement détestent les devoirs.
only a is right only b is right both right both wrong don’t know
Of the 32 pairs of sentences, there were 4 distractors and 28 contained adverbs. Of these
28 pairs, 10 pairs contained adverbs in sentences with prepositional phrases,l5 and 18
pairs contained sentences with adverbs in SVAO, SAVO, ASVO, or SVOA order. Of
these 18 pairs, 6 pairs consisted of sentences in the ungrammatical SAVO order paired
with sentences in the ASVO or SVOA order, 7 pairs consisted of sentences in the

grammatical SVAO order paired with sentences in ASVO or SVOA order, and 5 pairs

consisted of sentences in the SAVO and SVAO order. In other words, 11 sentences were

 

'5 Because these sentences proved questionable to a native speaker of French with respect to
acceptable adverb placement (see Footnote 13), these sentences were not included in the Error or Correct
scores, unlike Trahey & White’s (1993) procedure and unlike the scoring procedure for the WOC Task
in the present study.

109

in SAVO order and 12 sentences were in SVAO order.

White (1991a, 1991b) originally created two versions of this test, containing
different pairs of sentences with the items in each version in different orders. No effects
for version or order were found. Therefore, in the present study the same pairs of
sentences were used at the three testing sessions. The order of the pairs, however, was
changed for the first posttest so that students took the same test for the Pretest and for
the second Posttest.
3.5.1.2.1. Scoring Procedure”

Two scores for each student were computed, an Error score and a Correct score.
The Error score was computed as follows. For the 6 pairs of sentences containing a
sentence in SAVO order and a sentence in ASVO/SVOA order, an error was counted if
a student chose SAVO as correct or the Both Right response. For the 7 pairs of
sentences containing a sentence in SVAO order and a sentence in ASVO/SVOA order,
an error was counted if the student chose only ASVO/SVOA or the Both Wrong
response. For the 5 pairs of sentences containing a sentence in SAVO order and a
sentence in SVAO order, an error was counted if the student chose only SAVO. The
maximum Error score is 18.

The Correct score was computed as follows. For the 6 pairs of sentences

 

'“ These scoring procedures differ from Trahey & White (1993) whose correct scores actually include
acceptance of the incorrect word order (Trahey & White, 1993, p. 202, Foomote 10). The present method
of scoring the 5 pairs of SAVO-SVAO sentences is intended to distinguish those students who clearly know
the correct word order fiom those who are still accepting both structures. Thus, a response of Both Right
or Both Wrong is counted as neither an error nor as correct, contrary to Trahey & White who counted a
Both Right response as correct. While Trahey & White seem to be focusing more on the inclusion of the
correct word order (correct scores were not tallied for rejection of incorrect word orders), the present study
is equally interested in the exclusion of the incorrect word order. Therefore, a response of Both Wrong
to an SAVO-ASVO/SVOA pair is counted as correct.

110

containing a sentence in SAVO order and a sentence in ASVO/SVOA order, a correct
score was counted if the student chose only ASVO/SVOA or the Both Wrong response. ‘7
For the 7 pairs of sentences containing a sentence in SVAO order and a sentence in
ASVO/SVOA order, a correct score was counted if the student chose only SVAO or the
Both Right response. For the 5 pairs of sentences containing a sentence in SAVO order
and a sentence in SVAO order, a correct score was counted if the student chose only
SVAO. The maximum Correct score is 18.

The number of Both Right responses to the 5 pairs of SAVO-SVAO sentences
were also tabulated in order to obtain further information regarding those students who
are still accepting both structures.

3.5.2. Inductive Inferencing
3.5.2.1. Raven’s Standard Progressive Matrices Test

Raven’s Standard Progressive Matrices (SPM) Test was developed in order to
assess ”eductive ability, that is, the ability to educe correlates, the ability to generate high
level schemata, which make it easy to handle complex events” (Raven, 1989, p. 1).18
The test consists of five sets (A-E) of 12 problems each which are presented in abstract
figures and designs. A problem consists of a 3 x 3 matrix with the lower right hand
element blank. Given 8 options, students must choose the one which will complete the
matrix by inferring the rule(s) of the columns and rows. The sets in the test become

increasingly more complex and difficult. Students were allowed unlimited time to

 

‘7 While the ASVO/SVOA sentence is correct, the response of Both Wrong reveals that the student
is rejecting the SAVO order, which is the focus of the investigation.

'3 "The objective of Raven’s research was to identify the environmental [italics in original], as well
as genetic, origins of mental defect" (Raven, Raven, & Court, 1991, p. 010).

111

complete the test, which was administered at the Pretest session.

Raven’s SPM (Appendix I) "was never intended to be used on its own as measure
of (Spearman’s) g, and still less as a measure of ’General Intelligence" (Raven, Raven,
& Court, 1991, p. G6). However, the test has been used and interpreted as a measure
of nonverbal IQ. There are only three (of which I am aware) SLA studies which have
used Raven’s SPM (Flahive, 1980; d’Anglejan & Renaud, 1985; Abraham & Vann,
1987), and even though the authors have misrepresented what the test is measuring, the
findings of one of these studies are particularly interesting.19

Flahive (1980, p. 34) conducted a study with 20 students, representing seven
different first languages, to study "the relationship between scores on a nonverbal IQ test
[that is, Raven’s SPM], three reading tests, and the Test of English as a Foreign
language.” The three reading comprehension tests were "the paragraph comprehension
portion of the McGraw-Hill Basic Skills System Reading Test (Raygor, 1970) . . . a
paraphrase recognition test . . . and a cloze test” (Flahive, 1980, p. 35). While both
the McGraw-Hill test and the paraphrase recognition test were multiple-choice, the author

classifies only the former as a traditional multiple-choice reading test. The author found

 

'9 d’Anglejan & Renaud (1985), investigated the relationship between learner characteristics -
nonverbal IQ, years of schooling, cognitive style, age, classroom anxiety, competence in English, contacts
with Francophones, use of French, and literacy/illiteracy - and achievement in French as a Second
Language of 391 adult immigrants from 50 different countries in Montreal. The ages of the students
ranged from 17-63. Predictor variables included: Raven’s SPM, The Group Embedded Figures Test
(Witkin, Otrnan, Raskin, & Karp, 1971), and a Sociolinguistic and Attitudinal Questionnaire developed by
the authors. Criterion variables included teachers’ evaluations and The Test de Rendement en Francais-
Ministere de l’Immigration du Quebec (MIQ). This latter test contains four subtests in listening
comprehension, reading comprehension, speaking, and writing. The authors found that ”nonverbal IQ was
the most powerful predictor of second language achievement in (the) study" (1985, p. 12). Two comments
are in order. First, the authors have mistakenly interpreted Raven’s SPM as a measure of IQ. Second, it
is unfortimate that no detailed description of the MIQ is provided. Given the results of Flahive (1980),
it would be interesting to determine if the MIQ is in fact mainly a test of inferencing skills.

1 12
that

all the measures are highly correlated [which] is not surprising, since three of the tests
are purportedly testing the same skill, reading. Nor are the correlations between the
[paraphrase recognition] test and the TOEFL and the cloze and TOEFL surprising. They
are all measures of language proficiency. What is surprising is the high correlation
between Raven’s and the McGraw-Hill test. (Flahive, 1980, p. 37)

The author (1980, p. 38) concluded that "it is reasonable to conclude that reading
subtests . . . are not simply tests of reading ability. They are also tests of intelligence."

As stated above, the author is mistaken in assuming that Raven’s SPM is a test
of intelligence. What is interesting is that the author describes the McGraw-Hill test .
(which highly correlated with Raven’s SPM) as a "measure [of] a subject’s ability to
make inferences, to pick out main thoughts and supporting ideas, and to discover
organizational patterns in paragraphs and essays” (Flahive, 1980, p. 35). Given that this
description could very easily be a characterization of Raven’s SPM, with slight
modifications, it is not surprising that there was a high correlation between the two tests.
3.5.2.1.1. Validity

Raven’s SPM Test is "consensually accepted as the quintessential test of inductive
reasoning" (Alderton & Larson, 1990, p. 888). Raven, Court, and Raven (1992, p.

SPM6) claim that the SPM is

a valid index of the construct [of eductive ability and] that the test measures a range of
abilities that are built one on top of the other: it is not generally possible to solve the
more difficult problems if one does not possess the abilities required to solve the easier
ones.

This has been supported by studies of item difficulties, the order of which is invariant
across ability levels, conducted during numerous standardizations of the test. Item

Characteristic Curves also indicate

a) that the items are all measuring a common factor (if they were not,
there would be no relationship between the percentage able to solve a

113

given problem and the total scores);

b) that the abilities that are required to solve the more difficult problems
form part of a continuum which starts with the easiest items; and

c) that, although the abilities that are required to solve the more difficult
items may appear to be qualitatively different from those required to
solve the easier ones, these apparently different abilities shade
imperceptibly one into the other. (Raven, Court, & Raven, 1992, p.
SPM6)

3.5.2.1.2. Normative Studios

Since its development in 1936, the test is continually standardized based on
norming studies conducted approximately every five years. These studies have taken
place in Great Britain, Germany, Australia, China, and the United States with all ages.
"The 1986 US standardization was based on a series of local norming studies conducted
(with) over 22,000 students" (Raven, Court, & Raven, 1992, p. SPM9). Based on these

studies, the authors (Raven & Others, 1990, p. 73) concluded:

a) that, at any point, there is considerable similarity in the norms obtained in
different countries with a tradition of literacy;

b) that the tests scale in very much the same way in different cultures, socio-
economic groups, and ethnic groups;

c) that the differences between the mean scores of the various ethnic groups within
the U.S. parallels differences in their height, birth weight, and rates of infant
mortality; and

d) that the bulk of the variance in test scores is within, not between, ethnic and

socio-economic groups and is not accounted for by the major and obvious social
and economic inequalities within our society.

3.5.2.1.3. Reliability

Internal Consistency. The internal consistency of the SPM is estimated in a different way
than the factor-analytic tradition, because, "it does not make sense to inter-correlate the
items because one cannot use the fact that someone gets the easiest item right to predict
whether he or she will get the most difficult item right" (Raven, Court, & Raven, 1992,
p. SPMII). Internal consistency of the SPM is determined by 1) "whether the items

scale in the same way for different populations" (Raven, Court, & Raven, 1992, p.

114
SPMll), and 2) if the Item Characteristic Curves have few cross-overs, are equally
spaced, and cover the whole range of ability.

In the US standardisation the correlations between the item difficulties established
separately for different ethnic groups (Black, Anglo, Hispanic, Asian, and Navajo)
ranged from .97 to 1.0. [Additionally], the correlations between the item difficulties
established separately in the UK, US, Germany, New Zealand, and Chimr range from .98
to 1.0. (Raven, Court, & Raven, 1992, p. SPMII)

The test therefore is extremely robust and works in the same way - measures the same

thing - in a wide variety of ethnic, socio-economic, and age groups. (Raven, Court, &
Others, 1989, p. RS4.13)

T est-Retest. Listed in Table 3.1 are the retest reliabilities from the original studies
(1948) of the SPM. The time period between test and retest is not provided; however,
Raven, Court, and Raven (1992, p. SPM 13) report that "well-conducted studies . . .
indicate a satisfactory re-test reliability for SPM for periods up to one year." Note that

the retest reliability is the best for the age group of the subjects used in the present study.

 

 

 

 

 

 

 

Table 3.1
Retest Reliability of SPM
Age range (years) Mean Score Retest reliability %|
13 +/- 1 41 0.88
Under 3020 48 0.93
30-39 37 0.88
40—49 35 0.87 ||
50 and over 29 0.83

 

 

 

 

3.5.2.1.4. Scoring Procedure
A student’s total score on the SPM is the number of problems solved correctly.

The maximum score is 60. If an item was not answered or if two (or more) options

 

9° The authors do not state the lower age range of this group.

115

were selected that item was not counted as correct. The student’s total score is
composed of the subscores for each of the five Sets of problems. The total score is then
compared with the expected score composition for that score. These expected score
compositions are based on the 1979 standardization of the SPM. The difference between
the student’s score on each Set and that normally expected for that total score is then

listed as "discrepancies." An example of the computation is given in 3-13.

(3-13) SET Total
A B C D E
Student Score 11 10 10 9 5 45
Expected Score
Composition 12 1 1 9 9 4 45
Discrepancy -1 -1 + 1 0 + 1

If a person’s score on one of the Sets deviates by more than 2, the total score on the
scale cannot be accepted at its face value as a consistent estimate of general capacity for
intellectual activity. For general purposes, the total score appears to be relatively valid
even when discrepancies of more than 2 points occur. (Raven, Court, & Raven, 1992,
p. SPM30)

The time that each student took to complete the test was recorded in order to
determine the average test time for this population and eliminate any who deviated
significantly from the norm.

3.6. CHAPTER SUMMARY

In this chapter I have described the research design of the present study. I
provided biographical sketches of the volunteers who participated in the study and
explained the instructional materials which were used. I reviewed the justifications for
the use of Acceptability Judgment Tasks in SLA and then described the tasks and the

scoring procedures used in the present research. I provided a description of Raven’s

116

Standard Progressive Matrices Test and a summary of the validity, reliability, and
normative studies which support its use as a measure of inductive inferencing. The

results of the research are presented in Chapter 4.

CHAPTER 4
RESULTS

In this chapter I first present descriptive results of learner variables of age, previous
years of French, and gender by treatment group in addition to adverb input score1 by
treatment group. These variables are also described by class section. I also provide a
descriptive overview of the mean scores on all tests by treatment group and note observed
trends. I then present the results of the repeated measures ANOVA. Next, I report the
results of the multiple regressions, focusing on those results which are significant. I
conclude with a descriptive summary of the Teaching Questionnaire (Appendix E) and
Student Questionnaire (Appendix F).

4.1. LEARNER VARIABLES

Table 4.1. Learner Variables By Group

 

 

 

 

 

VARIABLE GROUP N it MDN SD MIN MAX
AGE INF 30 18.53 18.0 1.04 17.0 22.0
NON 15 18.27 18.0 .88 17.0 20.0
YEARS OF INF 30 3.46 3.8 .86 2.0 5.0
FRENCH NON 15 4.13 4.2 1.05 2.0 5.6
ADVERB INF 30 84.17 90.0 12.87 37.0 92.0
INPUT NON 15 69.6 75.0 15.24 17.0 75.0
GENDER INF 30 NON IS
-MALE 3 -MALE l
-FEM 27 -FEM 14

 

 

 

 

 

 

 

 

 

 

 

' Recall from Chapter 3, the adverb input score was derived from the Teaching Questionnaires as
a means to take into account student absences and the variation in teaching style.

117

118

As shown in Table 4.1, students ranged in age from 17 to 22 years old, with the
oldest student being in the inferencing group. Students' previous years of French study
ranged fi'om 2 years to 5.6 years; on the average, students in the noninferencing group had
approximately 6 months more of French study than students in the inferencing group. Out
of the 45 students participating in the study, 41 were female; 3 males were in the inferencing
group and 1 male was in the noninferencing group.

Assuming a total possible adverb input score of 151 (see Chapter 3 for an explanation
of this score), Table 4.1 shows that the inferencing group, on the average, received 56% of
the total possible adverb input, while the noninferencing group received, on the average,
46% of the total possible adverb input. The minimum score under the adverb input score
reflects student absences. In the noninferencing group, three students were absent during
the instnictional period of the study: one student was absent 6 times; one student was absent
two times; and one student was absent one time. In the inferencing group, a total of 15
students were absent during the instructional period: one student was absent 4 times; one
student was absent 3 times; and 8 students were absent one time each.

Table 4.2 displays the same data by class section and pooled as a single group.
Some differences between class sections which are not readily apparent by describing the
data based on treatment group are revealed in this table. Of note is the fact that the n-size
of class #4 is considerably less than the other sections. With respect to age, all sections are
roughly similar. However, with respect to previous years of French, class #3 stands apart
from the other three sections. The minimum number of years of French study was twice (4
years) that of the minimum number of years of French study for the other sections (2 years).

Therefore, the mean number of years of French study for class #3 is over one standard

119

deviation2 greater than the means for classes #1 and #2, and it is almost 2 standard deviations

greater than that of class #4.

Table 4.2. Learner Variables By Class Section

 

 

 

 

 

 

 

 

 

 

 

 

VARIABLE Classl ClassZ Class3 Class4 ALL
INF INF NON NON
n=l3 n=l7 n=ll n=4 n=45
MEAN 18.77 18.35 18.27 18.25 18.44
AGE SD 1.30 .79 .79 1.29 .99
MIN 17.0 17.0 17.0 17.0 17.0
MAX 22.0 20.0 20.0 20.0 22.0
MEAN 3.40 3.51 4.6 2.83 3.68
YEARS OF SD 1.04 .72 .56 .99 .97
FRENCH MIN 2.0 2.0 4.0 2.0 2.0
MAX 5.0 4.3 5.6 4.1 5.6
MEAN 79.85 87.47 75.0 54.75 79.31
ADVERB SD 17.14 7.29 0.0 26.13 17.0
INPUT MIN 37.0 73.0 75.0 17.0 92.0
MAX 90.0 92.0 75.0 75.0 15.21
MEAN 7.23 7.59 7.43
INFERENCING SD 1.25 .49 .92
INPUT MIN 4 7 4
MAX 8 8 8

 

 

With respect to the adverb input score, class #4 stands apart from the other three
sections. The mean adverb input score for class #4 is approximately 1.5 standard deviations
less than those of classes #1 and #3, and it is over 2 standard deviations less than the mean
adverb input score of class #2. As mentioned in the discussion of Table 4.1, one student in
the noninferencing group was absent six times; this student was in class #4, as reflected in
the minimum adverb input score of 17. Of note, also, is that no students were absent in class

#3 during the instructional period.

 

2 The standard deviation referred to here is from Table 4.2, Years of French, all classes pooled as
a single group.

120

Looking just at the inferencing classes, more students were absent in class #1 than
in class #2. This is reflected in the lower minimum adverb input score and the minimum
inferencing input score of class #1.

4.2. OVERVIEW OF ALL TESTS

Table 4.3 lists the means, medians, standard deviations, and minimum and maximum
scores on the Acceptability Judgment Task for both groups. Table 4.3 shows that the
minimum correct score of the inferencing group on each the AJ tests was lower than that of
the noninferencing group. The maximum correct score of the inferencing group was higher
than that of the noninferencing group at the Pretest, and each group obtained the same
maximum correct score at Posttest I and Posttest 2. The minimum error score of the
inferencing group was lower at the Pretest and the same as that of the noninferencing at
Posttest 1 and Posttest 2. The maximum error score of the inferencing group was higher than
that of the noninferencing group on all occasions. Furthermore, at all occasions the standard

deviation of both the error and the correct scores was greater for the inferencing group.

121

Table 4.3. Acceptability Judgment Task

 

 

 

 

 

 

 

 

TEST GROUP N it MDN so MIN fiMAX
PRE AJ INF 30 9.20 9.0 4.38 2.0 18.0
CORRECT NON 15 9.27 10.0 2.84 4.0 14.0
K=18
PRE AJ INF 30 7.13 7.0 3.98 0.0 15.0
ERROR NON 15 6.20 6.0 2.37 3.0 13.0
K=l8 _
POST] AJ INF 30 11.03 11.0 5.14 0.0 18.0
CORRECT NON 15 11.33 14.0 4.43 5.0 18.0
K=l8
POST] AJ INF 30
ERROR NON 15
K=18

,
POST2 AJ INF 29
CORRECT NON 13
K=18
POST2 AJ INF 29 4.66 3.0 4.28 0.0 16.0
ERROR NON 13 3.39 3.0 3.02 0.0 10.0
K=18

 

 

 

 

 

 

 

 

 

 

Table 4.4 lists the means, medians, standard deviations, and minimum and maximum
scores on the Word Order Correction Task and Raven's Standard Progressive Matrices for
both groups. Table 4.4 shows that the minimum correct score of the inferencing group on
all occasions of the WOC task was lower than that of the noninferencing group. The
maximum correct score of the inferencing group was higher than that of the noninferencing
group on all occasions. The standard deviation of the correct score for the inferencing group
was less than that of the noninferencing group at the Pretest and greater at Posttest l and
Posttest 2. The minimum error score of both groups was the same. The maximum error
score of the inferencing group was higher on all occasions. The standard deviation of the
error score for the inferencing group was less than that of the noninferencing group at the

Pretest and Posttest 2 and greater at Posttest 1.

122

Table 4.4. Word Order Correction Task And
Raven's Standard Progressive Matrices

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

ll TEST GROUP N St MDN SD MIN MAX
PREWOC INF 30 7.93 8.0 2.39 2.0 14.0
CORRECT NON 15 7.47 8.0 2.85 3.0 13.0
K=l6
PREWOC INF 30 3.97 4.0 2.16 0.0 11.0
ERROR NON 15 3.93 3.0 2.55 0.0 9.0

ll K=16
PSTIWOC INF 30 8.97 9.5 3.26 2.0 14.0
CORRECT NON 15 7.93 9.0 2.55 4.0 11.0
K=16
PST 1 WOC INF 30 3.20 2.0 2.99 0.0 13.0
ERROR NON 15 3.53 3.0 2.53 0.0 8.0
K=l6

hr
PSTZWOC INF 29 9.69 10.0 3.71 1.0 16.0
CORRECT NON 13 9.31 10.0 2.93 4.0 12.0
K=16
PSTZWOC INF 29 2.79 2.0 2.70 0.0 12.0
ERROR NON 13 2.62 1.0 2.84 0.0 9.0
K=l6
RAVEN'S INF 30 49.40 51.0 5.56 34.0 56.0
K=60 NON 15 49.27 51.0 5.05 38.0 57.0
TIME INF 30 20.07 19.0 4.65 1 1.0 30.0

NON 15 18.20 18.0 2.51 14.0 25.0

 

 

 

The mean scores on Raven's SPM of both groups are approximately the same. The
minimum and maximum scores of the inferencing group are slightly lower than those of the
noninferencing group. With respect to the time taken to complete the test, the mean and
median times of the two groups are approximately equivalent. The inferencing group shows
a greater standard deviation. Additionally, the inferencing group's minimum time to
complete the test is slightly less than the noninferencing group's minimum time, and their
maximum time is greater than that of the noninferencing group. A t-test reveals a significant

difference (t=7.48; p<.001) between the two groups with respect to time taken to complete

123
Raven's SPM.

Figures 4.1 and 4.2 show the mean correct and mean error scores on the Pretest,
Posttest 1, and Posttest 2 for the Acceptability Judgment and Word Order Correction Tasks
for each group.

Figure 4.1 shows that the general trend was improvement over time on the mean

correct scores for both groups on both tasks.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

   

AJ PRETEST AJ POST 1 AJ POST 2 W00 PRETEST WOC POST 1 W00 POST 2

 

1 1:] INFERENCING - NONINFERENCING

 

Figure 4.1. AJ and WOC Tasks Mean Correct Scores

Similarly, Figure 4.2 shows that the mean error scores of both groups decreased on

both tasks over time.

124

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

AJPRETEST “POST1 AJPOSTZ WOCPRETEST WOOPOST‘I WOOPOSTZ

 

L [j INFERENCING I uomnsenencmo ]

 

Figure 4.2. AJ and WOC Tasks Mean Error Scores

The same data are displayed somewhat differently in Figures 4.3 to 4.6.

125

4.2.1. Acceptability Judgment Task

 

 

 

MEANS

 

 

 

 

 

 

l
POSTTEST 1 POSTTEST 2

 

l - -.- - INFERENCING —.— uomnmsncmo

 

Figure 4.3. AJ Task Mean Correct Scores

Figure 4.3 shows the mean correct scores of both groups on the AJ Task for each
testing session. Both the inferencing and noninferencing groups showed improvement from
Pretest to Posttest 1 to Posttest 2, with the noninferencing group performing slightly better
than the inferencing group at each testing session. Additionally, the change in means across
testing sessions was slightly greater for the noninferencing group. From Pretest to Posttest
l, the change in means was 2.06 for the noninferencing group versus 1.83 for the inferencing
group; from Posttest 1 to Posttest 2, the change in means was 1.98 versus 1.42; and from

Pretest to Posttest 2, the change in means was 4.04 versus 3.25. Thus, the difference

126

between the two groups' mean correct scores steadily increased from .07 at the Pretest to .3

at Posttest 1 to .86 at Posttest 2.

 

 

 

 

 

 

 

 

 

 

7 .~~§'~.‘
0n.
a “a
2 “~.
‘ ~.\
Ill “~.
3 ““““
5 "‘~-.
4
3 1
PRETEST POSTTEST 1 POSTTEST 2
- - . - - INFERENCING —'—— NONINFERENCING

 

 

 

Figure 4.4. A] Task Mean Error Scores

Similar trends can be seen in Figure 4.4 of the mean error scores of both groups on
the A] Task for each testing session. The mean error scores for both groups decreased at
each testing session, with the noninferencing group again performing slightly better (that is,
fewer errors) at each testing session. Additionally, the change in means across testing
sessions was slightly greater for the noninferencing group. From Pretest to Posttest l, the
change in means was - 1.3 for the noninferencing group versus - 1.23 for the inferencing

group; from Posttest 1 to Posttest 2, the change in means was - 1.51 versus - 1.24; and from

127

Pretest to Posttest 2, the change in means was -2.81 versus -2.47. Thus, the difference
between the two groups' mean error scores steadily increased from .93 at the Pretest to .97
at Posttest l to 1.27 at Posttest 2.

4.2.2. Word Order Correction Task

 

10

O
'
O
O
O
'O
O

 

9.5 .v
0 C

 

 

 

 

 

 

 

7 l
PRETEST POSTTEST 1 POSTTEST 2

- m. INFERENCING —.— NONINFERENCING

 

 

 

 

Figure 4.5. WOC Task Mean Correct Scores

Trends are somewhat reversed on the Word Order Correction Task. Figure 4.5 shows
the mean correct scores of both groups at each testing session for this task. Both the
inferencing and noninferencing groups showed improvement from Pretest to Posttest 1 to
Posttest 2, with the inferencing group performing slightly better on each test than the

noninferencing group. The change in means was slightly greater for the inferencing group

128

from Pretest to Posttest I (1.04 versus .46 for the noninferencing group). However, the
noninferencing group showed a greater change in means from Posttest 1 to Posttest 2 (1.38
versus .72 for the inferencing group) and from Pretest to Posttest 2 (1.84 versus 1.76 for the
inferencing group). Thus, the difference between the two groups' mean correct scores
increased from .46 at the Pretest to 1.04 at Posttest l and then decreased to .38 at Posttest

2.

 

 

\
‘
3.8 \
\
\
\
\
S
\
\
\
\
3 6 ‘
n \
S
\
\
S
\
\
\
§
\
3.4 ‘
i
\
\

3.2 ~ s

 

 

 

MEANS
O

 

 

 

 

 

 

3 ‘

2.8

2.6 1

maresr POSTTEST 1 POSTTEST 2
--.-. INFERENCING —'— NONINFERENCING

 

 

 

Figure 4.6. WOC Task Mean Error Scores

Figure 4.6 displays the mean error scores of both groups on the Word Order
Correction Task at each testing session. Again, both groups showed improvement from

Pretest to Posttest I to Posttest 2. While the change in means fi'om Pretest to Posttest 1 was

129

slightly greater for the inferencing group than for the noninferencing group (- .77 versus -.4),
the noninferencing group showed a slightly greater change in means from Posttest l to
Posttest 2 (- .91 versus — .41 for the inferencing group) and from Pretest to Posttest 2 (- 1.31
versus - 1.18 for the inferencing group). Somewhat different trends between the two groups
can be seen on the mean error scores of this task. While the mean error score of the
noninferencing group was slightly lower than the inferencing group at the Pretest (a
difference of .04), the noninferencing group's mean error score was .33 higher than the
inferencing group's mean error score at Pretest 1. At Pretest 2, the noninferencing group's
mean error score was again slightly below (.17) that of the inferencing group's mean error
score.
4.2.3. Summary

An examination of these descriptive results indicates the following trends: 1) Both
groups showed improvement over time on both the correct scores and the error scores on
both tasks; 2) The noninferencing group's mean correct scores and mean error scores on
each of the AJ tasks were slightly higher than those of the inferencing group; 3) The
noninferencing group showed slightly greater improvement on the correct and error scores
of the A] task over time; 4) The inferencing group's mean correct scores on each of the WOC
tasks were slightly higher than those of the noninferencing group; and 5) The inferencing
group showed slightly greater improvement between the Pretest and Posttest 1 on both the
correct score and the error score of the WOC task.
4.2.4. Word Order Correction Task Subscores

As described in Chapter 3, the correct and error scores on the Word Order Correction

Task were subdivided in order to differentiate 1) active usage of the (in)correct word order

130

from passive acceptance of the (in)correct word order, and 2) acceptance and usage of
French word order (SVAO) from acceptance and usage of English and French word orders
(SVOA or ASVO).

4.2.4.1. Correct Subscores

Table 4.5 displays the pooled and group mean correct subscores and standard
deviations at each testing session. In the discussion of the correct subscores, it must be
pointed out that the subscores do not represent equal percentages of the total correct score.
In other words, if all sentences are correct, the Cl (unchanged from SVAO) subscore is 25%
of the total correct score; C2 (changed SAVO to SVAO) and C3 (changed SAVO to ASVO
or SVOA) sentences are the same sentences and these subscores are 44% of the total correct
score; the C4 (changed ASVO or SVOA to SVAO) subscore is 31% of the total correct
score; the (C2 + C4) subscore could potentially be 75% of the total correct score.

Table 4.5 shows that both groups remained fairly consistent over testing sessions in
their acceptance of the correct SVAO word order (subscore C1). Ifthe mean subscores are
examined as a percentage of a group's total mean correct score, the C1 subscore of the
inferencing group represented 34% of their total mean at the Pretest, 33% at Posttest I, and
31% at Posttest 2. The C1 subscore of the noninferencing group was slightly higher at each
testing session and represented 37% of their total mean correct score at the Pretest, 40% at

Posttest I, and 34% at Posttest 2.

131

.2395 E “gamma 8? a £023 820 v.83 a .«o Oman: 95% no can condo v.53

82.80 £5 .«0 oozefioooa 25mg macaw 5 uoEo BOB eunuch O<>m 05 .«o owum: 25% Beam 8.503% 805 we 53 mm nonhuman 355 98 v0 can N0

3% a <O>m no 93. confine H5
3% .5 3?. 2 03% Basso H8
O<>m 3 03% e855 ”8
O<>m ace penance: cu

MBO=£ ma 33,693 06>» $.58 Ochoa 35 :33—

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

med 3d mmd mes mwd :8 202
Sam mod omfi 3.x and mag. m7:
9am nmd Nod New omN mug. Buo— den—.09.
R..— woN mm; On; 2 .N ow._ 202
mm; mm; e9. 0.: mm; CON m2“
Z; EL mm; mm; 3; EN 5.1.x mu
bod we;4 OWN 9am ohm med 202
end bow mm.m 09v NmN 3N m7:
Sam one wmfi Ne wmd ced Suv— vU+NU
mm. m_.m mo. oN.m mm; med 202
mo._ 86 mm. SN 2: SN #2.—
3. SM mm. vo.m eq— eoN vnx .0
Gm M Om M Om M Om M Om M Om m
:0 mac A? A: Ado Ago A: 412 ABC ago 1:404 442
N HmBHmOm _ HmmHHmOm Hmmhmdm

 

 

 

 

88035 8380 xmah cocootoo SEO «:03 .3» 035.

 

132

The groups show somewhat different trends on the second (C2 + C4) and third (C3)
subscores. The inferencing group showed a slightly greater increase in their active usage of
SVAO word order (C2 + C4), while the noninferencing group's usage remained somewhat
consistent. Again, if the subscores are examined as a percentage of the total mean correct
score, the (C2 + C4) subscore of the inferencing group was 37% of their total mean correct
score at the Pretest, 51% at Posttest 1, and 53% at Posttest 2. While the noninferencing
group's (C2 + C4) subscore percentage at the Pretest was slightly higher (39%) than the
inferencing group's subscore percentage, the noninferencing group showed a smaller increase
at Posttest 1 (43%) and again at Posttest 2 (44%). The (C2+C4) mean subscores of both

groups at each testing session are shown in Figure 4.7.

 

 

 

 

5.5
Lt"
5 "ai
ooo
0".
o"
-"‘
a"'
0""
I
4
4.5
’
19
z
< 4 ‘
Ill

 

 

 

 

 

 

2.6 l
PRETEST POSTTEST 1 POSTTEST 2

"Q- - INFERENCING —-'— NONINFERENCING

 

 

 

 

Figure 4.7. WOC Task Mean C2 + C4 Subscores

133

With respect to subscore C3, which reflects changes from the incorrect SAVO word
order to a word order which is possible in both French and English (ASVO or SVOA), the
inferencing group's mean C3 subscore represented 29% of their total mean correct score at
the Pretest, dropped to 16% at Posttest 1, and remained at 16% at Posttest 2. The
noninferencing group's C3 subscore represented slightly less of their total mean correct score
at the Pretest (24%) than the inferencing group's percentage; it dropped to 17% at Posttest
I, then rose to 22% at Posttest 2.
4.2.4.1.1. Correct Subscore Summary

An examination of the mean correct subscores indicates the following trends: 1)
While both groups remained fairly consistent in their acceptance of the correct word order
(C1), this subscore represented a slightly higher percentage of the noninferencing group's
“ total mean correct score than the inferencing group's subscore did of their total correct score;
2) The C1 subscores of both groups represented a higher percentage of their total mean
correct score than that category does of the total correct score; 3) The inferencing group
showed a slightly higher increase over testing sessions than the noninferencing group in their
changes to the correct word order (C2 + C4); and 4) the inferencing group decreased over
time (and maintained that decrease) in their usage of word order which is possible in both
French and English (C3), while the noninferencing group decreased in their usage at Posttest
1 but then increased at Posttest 2.
4.2.4.2. Error Subscores

Error scores for the Word Order Correction Task were also subdivided in order to
distinguish between passive acceptance of the incorrect word order and active changes to the

incorrect word order. Table 4.6 displays the pooled and group mean error subscores and

134

standard deviations at each testing session. In the discussion of the error subscores, it must
be pointed out that the subsCores do not represent equal percentages of the total error score.
In other words, if all sentences are incorrect, the E1 subscore is 44% of the total error score;
the E2 subscore is 56% of the total error score;3 and, the E3 subscore is 25% of the total
error score.

Table 4.6 shows that the majority of errors made by both groups on the WOC task
were of the same kind --- E1 errors, or passive acceptance of the incorrect SAVO word
order. If the mean error scores are examined as a percentage of a group's total mean error
score, the E1 subscores represent a higher percentage of each group's total mean error score
than the E1 category does of the total error score. The El subscore of the inferencing group
represented 60% of their total mean error score at the Pretest, 59% at Posttest 1, and 61%
at Posttest 2. Thus, the E1 subscore of the inferencing group represented approximately the
same percentage of their total mean error score across all testing sessions. The E1 subscore
of the noninferencing group represented 66% of their total mean error score at the Pretest,
rose to 76% at Posttest l, and fell to 65% at Posttest 2. Therefore, at each testing session,
the El subscore of the noninferencing group represented a slightly higher percentage of their
total mean error score than the inferencing group's subscore did of their total mean error

SCOI'C.

 

2 The E2 subscore consists of any changes to the incorrect SAVO word order. Therefore, this
subscore includes those 5 sentences in SVOA/ASVO word order and those 4 sentences in SVAO word
order. If SVAO sentences are changed to SVOA/ASVO word order, this constitutes subscore E3.

135

 

 

<O>m .5 O>m< 9 O<>m cowSEU ”mm

O><m 2 o<>m co .<o>m .o>m< Basso em
03% 52.. Basses cm

”950:8 ma covSGenm BOB 388 She 35 Am 5328 =aoom

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

441111111111. Idle.” one 1 m2 m2 112%-.
OEN 2N meN ON.m 3 .N 35. m7:
woN EN owN SUM VNN mew 2.4x HOB
Nv. in n. no. 34 NA 202
5. Nb. mm. cm. 2: so; 2
on. 31. on. on. 8.— m: eno— mm
mm. a. vm. ON. N. no. 202
mm. mm. m: ow. em. cm. 2
2.. In. em. ov. be. 0m. and. Nm
mmN e9— emN $.N mmN 8N 202
me; mo; VON 0e; 2: EN 2
coN mo; EN EN woN SN nuc— _m
Om M Om M Om M Om M Om M Om M
:0 Ago at? 42 ago a0 ‘12 4‘2 Ago :0 4‘2 412
N HmmHHmOm L HmmHHmOm Hmmhmmm
8.5896 Sum x93. nozoobou Sea 203 64‘ 035.

 

136

The E2 subscores (changes of SVOA, ASVO, or SVAO to SAVO) of both groups
represented the smallest percentage of their total mean error scores, even though this
category does not represent the smallest percentage of the total error score. The inferencing
group's E2 subscores remained fairly consistent over testing sessions. It represented 13%
of their total mean error score at the Pretest, 16% at Posttest l, and 14% at Posttest 2. The
E2 subscore of the noninferencing group represented 2% of their total mean error score at
the Pretest, 5% at Posttest l, and 6% at Posttest 2. Thus, at each testing session, the E2
subscore of the noninferencing group represented a smaller percentage of their total mean
error score than the inferencing group's subscore did of their total mean error score.
Furthermore, there is a slight increase in the E2 subscores of both groups from Pretest to
Posttest 1; this increase is maintained at Posttest 2.

The E3 subscore (changes of SVAO to SVOA or ASVO) of the inferencing group,
ofcourse, remained consistent over testing sessions. It represented 27% of their total mean
error score at the Pretest, 25% at Posttest 1, and 25% at Posttest 2. The E3 subscore of the
noninferencing group represented 32% of their total mean error score at the Pretest, fell to
19% at Posttest 1, and rose to 29% at Posttest 2. Thus, at all testing sessions, the E3
subscore of the inferencing group represented approximately the same percentage of their
total mean error score as that category does of the total error score. The E3 subscore of the
noninferencing group was slightly higher than that of the inferencing group at the Pretest and
Posttest 2; additionally, the E3 subscore of the noninferencing group at the Pretest and
Posttest 2 represented a slightly higher percentage of their total mean error score than the E3

category does of the total error score.

137

4.2.4.2.]. Error Subscore Summary

An examination of the mean error subscores indicates the following trends: 1) The
inferencing group remained consistent across testing sessions in the type of errors they were
making; 2) For both groups, El type errors, or acceptance of the incorrect SAVO word
order, represented a higher percentage of their total mean error score than the E1 category
does of the total error score; 3) The E1 subscore of the noninferencing group represented a
slightly higher percentage of their total mean error score than the inferencing group's El
subscore did of their total mean error score; and 4) For both groups, the smallest percentage
of errors consisted of E2 type errors, or active usage of the incorrect SAVO word order, even
though the E2 category does not represent the smallest percentage of the total error score.
4.3. REPEATED MEASURES ANALYSIS OF VARIANCE

Repeated measures Analysis of Variance of the correct scores and the error scores
on the Acceptability Judgment task and the Word Order Correction Task for both groups
were run to determine any significant differences between and within groups across testing
sessions.“ Where significant differences between mean scores were found, post hoc Scheffé
procedures were used to determine the source of the significant differences. A significance
level of .05 is used.
4.3.1. Acceptability Judgment Task

The source table of the repeated measures AN OVA of the correct scores on the AJ

Task is provided in Table 4.7 .

 

“ In order to perform the repeated measures ANOVA, the data from the three subjects who were
absent from Posttest 2 were eliminated from the analysis.

138

Table 4.7 . Source Table A] Task Correct Scores

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Between Groups df SS MS F
INF 1 7.7 7.7 .33
ID(INF) 82 1885.0 22.99
Within Groups
OCC 2 246.92 123.46 *32.32
OCC x INF 2 1.43 .72 .19
ID(INF) x OCC 164 626.99 3.82
Total 251 2768.04

(*p<.01)

Differences between the inferencing and noninferencing groups on the correct scores
of the AI Task are not significant (F(1,82) = .33; Fed, = 3.96). There are significant
differences between testing sessions (OCCasion) for both groups (F (2, 164) = 32.32, p< .01;
Fed = 3.89). Scheffé tests show that the changes from Pretest to Posttest 1 and hem Posttest
1 to Posttest 2 are not significant. The interaction between group and testing session is not
significant (F(2,164) = .19).

The source table for the repeated measures ANOVA of the error scores on the A]

Task is provided in Table 4.8.

Table 4.8. Source Table AJ Task Error Scores

139

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Between Groups df SS MS F
INF 1 40.62 40.6 2.48
ID(INF) 82 1344.11 16.39
Within Groups
OCC 2 124.87 62.44 *2021
OCC x INF 2 1.06 .53 .17
ID(INF) x OCC 164 506.74 3.09
Total 251 2017.30

*p<.01

Differences between the inferencing and noninferencing groups on the error scores
of the AJ Task are not significant (F (1,82) = 2.48). There are significant differences between
testing sessions for both groups (F(2,164) = 20.21, p< .01). Schefi‘é tests show that the

changes from Pretest to Posttest l and from Posttest 1 to Posttest 2 are not significant. The

interaction between group and testing session is not significant (F(2,164) = .17).

4.3.2. Word Order Correction Task

The source table for the repeated measures ANOVA of the correct scores on the

WOC Task is provided in Table 4.9.

 

140

Table 4.9. Source Table WOC Task Correct Scores

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Between Groups df SS MS F
INF 1 5.63 5.63 .61
ID(INF) 82 763.37 9.29
Within Groups
OCC 2 62.11 31.06 *14.31
OCC x INF 2 2.49 1.25 .57
ID(INF) x OCC 164 356.06 2.17
Total 251 1189.66

*p<.01

Difi’erences between the inferencing and noninferencing groups on the correct scores
of the WOC Task are not significant (F(1,82) = .61). There are significant differences
between testing sessions for both groups (F(2,164) = 14.31, p< .01). Scheffé tests show that
the changes from Pretest to Posttest 1 and from Posttest 1 to Posttest 2 are not significant.
The interaction between group and testing session is not significant (F (2, 164) = .57).

The source table for the repeated measures AN OVA of the error scores on the WOC

Task is provided in Table 4.10.

Table 4.10. Source Table WOC Task Error Scores

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Between Groups df SS MS F
INF 1 .07 .07 .00
ID(INF) 82 763.37 7.43
Within Groups
OCC 2 27.44 13.72 *9.66
OCC x INF 2 1.68 .84 .59
ID(INF) x OCC 164 232.88 1.42
Total 251 871.30

*p<.01

Differences between the inferencing and noninferencing groups on the error scores
of the WOC Task are not significant (F(1,82) = .00). There are significant differences
between testing sessions for both groups (F(2,164) = 9.66, p< .01). Scheffé tests show that
the changes from Pretest to Posttest 1 and from Posttest 1 to Posttest 2 are not significant.
The interaction between group and testing session is not significant (F (2, 164) = .59).
4.3.3. ANOVA Summary

Results of the repeated measures AN OVA reveal that there is no significant
difference between the two groups on either the correct scores or the error scores of the A]
Task and the WOC Task. While both groups showed significant improvement fiom the
Pretest to Posttest 2 in both their acceptance/usage of the correct SVAO word order and in
their rejection/usage of the incorrect SAVO word order on both Tasks, the interaction
between group and testing session (OCCasion) is not significant. In other words, there is not

a significant difference between the groups in how they changed over time.

4.4. MULTIPLE REGRESSION

Multiple regressions were performed to determine the significant predictors of each

of the outputs listed under A and B.5

A. 5 Task: Correct Score
1. Pretest

2. APretest-Posttest 1
3. APosttest I-Posttest 2
4. APretest-Posttest 2

B. WQC Task: Corrog Sooro
9. Pretest

10. APretest-Posttest 1
ll. APosttest l-Posttest 2
12. APretest-Posttest 2
13. Pretest Subscores

14. AC1, A(C2+C4), AC3

AJ Tosk: Error Score
5. Pretest

6. APretest-Post 1
7. APost l-Post 2

8. APretest-Post 2

WOC Task; Error Sooro

15. Pretest

16. APretest-Post 1
17. APost l-Post 2
18. APretest-Post 2
19. Pretest Subscores
20. AEl, AE2, AE3

The following variables were considered as predictors:
C. Age

Previous years of French

Gender

Inferencing group“/Inferencing input score7
Adverb input score8

Raven's score

99:592.“?

 

5

Assumptions of normality, homoscedasticity, and independence were not violated.

“ The interval variable of inferencing input score and the nomiml variable of inferencing group were
found to be multicollinear; therefore, the inferencing input score was removed from the multiple
regressions which considered the subjects as a single group.

7 The interval variable of inferencing input score, which incorporated student absences, was used
in the regressions on the inferencing group. This variable was not included in the regressions of the
pretest.

9 To emure that class/teacher effect was being captured by the adverb input score, class sections
were ordered on a nominal scale and a multiple regression was performed with classes as the predictor
variables and adverb input score as the output. The Omnibus test proved significant (p=.000) and each
predictor was significant (class I p=.000; class 2 p=.001; class 3 p=.000; class 4 p=.008), accounting
for 36.5% of the variance in the adverb input score. It was concluded that class/teacher effect was
captured in the adverb input score. The variable of adverb input score was not included in the regressions
of the pretests.

143

Each of the outputs listed under A and B was run with all of the predictors listed
under C, producing combined equations for each of the values. For each equation an
Omnibus Test was performed to determine if the variability between predictors and outcome
could be accounted for by chance alone. Ifthe Omnibus Test proved significant, subsequent
specific t-tests were performed on the predictors.

Regressions which combined all subjects (as a single output) were run first.
Regressions on the inferencing group alone were then run to determine any significant
predictors within this group which were not revealed by placing the groups together.9 The
primary concern is whether or not inferencing group/inferencing input, Raven's score (as a
measure of inductive reasoning), and/or adverb input is/are significant predictors of the
change in the correct or error scores on the AJ Tasks and WOC Tasks. I therefore report
only the fitted equations in which any of these variables is a significant predictor of initial
score or of change in score. The MINITAB printout of the fitted equations and their
corresponding combined equations (that is, the trimming stages have been removed) is
provided in Appendix J.

4.4.1. Acceptability Judgment Task
4.4.1.1. Correct Score

Raven's score is a significant predictor (p=.04) of the Pretest correct score on the AI
Task and accounts for 9.4% of the variance in the Pretest correct score. As Raven's score
increases, the correct score is predicted to increase.

No significant predictors of the change in the correct score from Pretest to Posttest

 

9 Regressions could not be performed on the noninferencing group alone because of the small sample
size (n=15).

144

l or fi'om Posttest 1 to Posttest 2 were found.

When the variables of age and Raven's score are considered together, the Omnibus
test approaches significance (p=.056), and this model accounts for 13.7% of the variance in
the change in the correct score fiom Pretest to Posttest 2. While the t-test on Raven's score
is significant (p=.05, 9.0% of the variance), age is not a significant predictor (p=.083, 4.7%
of the variance). When these predictors are considered separately, neither one is significant.
As age increases, the change in the correct score is predicted to increase; as Raven's score
increases, the change is predicted to decrease.
4.4.1.2. Error Score

No significant predictors of the Pretest error score or of the change in the error score
on any of the tests were found.

4.4.2. Word Order Correction Task
4.4.2.1. Correct Score

No significant predictors of the Pretest correct score or of the change fi'om Pretest
to Posttest 1 on the Word Order Correction Task were found.

When the variables of adverb input and previous years of French are considered
together, the Omnibus test proves significant (p=.028), and this model accounts for 16.8%
of the variance in the change in the correct score fi'om Posttest 1 to Posttest 2. However,
while the t-test on previous years of French is significant (p=.03 7; 9.9% of the variance),
adverb input is not (p=.064; 6.8% of the variance). As previous years of French and adverb
input increase, the change in the correct score is predicted to increase. When each of these
predictors is considered separately, only previous years of French approaches significance

(p=.053) and accounts for 9.0% of the variance.

145

No significant predictors of the change in the correct score from Pretest to Posttest
2 were found.
4.4.2.2. Subscores

No significant predictors of the Pretest or of the change in the correct subscores on
the WOC Task were found.
4.4.2.3. Error Score

Gender (p=.046) and Raven's score (p=.034) were significant predictors for the
Pretest error score on the WOC Task. They account for 14.4% of the variance in the error
score; gender accounts for 4.6% of the variance, and Raven's score accounts for 9.8% of the
variance. An error score is predicted to be less for females than for males.10 As Raven's
score increases, the number of errors is predicted to decrease.

No significant predictors of the change in the error score from the Pretest to Posttest
1 were found.

When the variables of previous years of French and adverb input are considered
together as predictors of the change in error score from Posttest l to Posttest 2, the Omnibus
test proves significant (p=.001). However, only the variable of previous years of French is
significant (p=.000), and the variable of adverb input approaches significance (p=.051).
These predictors account for 31.3% of the variance in the change in error score, with
previous years of French accounting for 24.2% and adverb input accounting for 7.1%. As

adverb input and previous years of French increase, the change in the error score is predicted

 

‘0 However, recall that the database consisted of 45 subjects, 41 females and only 4 males.

146

to move fi'om positive to negative.‘1 Previous years of French alone is a significant predictor
(p=.001) and accounts for 24.2% of the variance. However, adverb input alone is not a
significant predictor (p=.129).

No significant predictors of the change in the error score from Pretest to Posttest 2
were found.
4.4.2.4. Subscores

The variables of gender (p=.035) and Raven's score (p=.001) are significant
predictors of the variance in the Pretest E2 subscore (changes to SAVO) and account for
26.2% of the variance. Gender accounts for 2.5% of the variance in the E2 subscore, which
is predicted to be less for females than for males. Raven's score accounts for 23.7% of the
variance, and as Raven's score increases, the E2 subscore is predicted to decrease.

There was only one significant predictor of the change in the error subscores on the
WOC task. Raven's score (p=.048) accounts for 8.8% of the variance in the change in the
E2 subscore fi'om the Pretest to Posttest 1. As Raven's score increases, the change in the E2

subscore is predicted to become positive.

 

“ Note that this is a ohggo in the error score; therefore, a negative change from Posttest l to
Posttest 2, for example, indicates fewer errors on Posttest 2 than on Posttest 1.

147

4.4.3. Inferencing Group12
4.4.3.1. Acceptability Judgment Task
4.4.3.1.1. Correct Score

No significant predictors of the Pretest correct score on the AI Task were found.

The only significant predictor of the change in the correct score of the AI Task was
inferencing input (p=.048). This predictor accounted for 13 .7% of the variance in the change
from Posttest 1 to Posttest 2. As inferencing input increases, the change in the correct score
is predicted to increase.”
4.4.3.1.2. Error Score

No significant predictors of the Pretest error score or of the change in error score
from Pretest to Posttest 1 of the A] Task were found.

Inferencing input is a significant predictor (p=.027) of the change in the error score
from Posttest 1 to Posttest 2. This predictor accounts for 16.9% of the variance in the
change. As inferencing input increases, the change in error score is predicted to become
negative.

Inferencing input is also a significant predictor (p=.042) of the change from the
Pretest to Posttest 2 and accounts for 14.5% of the variance in the change. As inferencing

input increases, the change in the error score is predicted to become negative.

 

'2 Given the small sample size (n=30), the strength of these results is questionable. The fitted
equations were examined, and assumptions of normality, homoscedasticity, and independence were not
violated.

'2 However, note that for subjects with an inferencing input score between 4 and 6, this is a $831112
change. That is, these subjects did worse on Posttest 2 than on Posttest 1.

148

4.4.3.2. Word Order Correction Task
4.4.3.2.1. Correct Score

No significant predictors of either the correct score or the correct subscores on the
Pretest of the WOC Task were found. Nor were any significant predictors of the changes
in these scores found.
4.4.3.2.2. Error Score

NO significant predictors of the Pretest error score or of the change in the error score
from Pretest to Posttest 1 and from Posttest l to Posttest 2 were found.

Gender (p=.029) and inferencing input (p=.016) considered together are significant
predictors of the change in the error score from Pretest to Posttest 2 and account for 24.8%
of the variance in the change. Gender accounts for 5.5% of the variance in the change of the
error score, which is predicted to be less for females. Inferencing input accounts for 19.3%
of the variance in change. As inferencing input increases, the change in the error score is
predicted to become negative.
4.4.3.2.2.]. Subscore

The variables of gender (p=.049) and Raven's score (p=.001) are significant
predictors of the Pretest E2 subscore and account for 38% Of the variance in the subscore.
Gender accounts for 3.6% of the variance in the subscore, which is predicted to be less for
females. Raven's score accounts for 34.4% of the variance. As Raven's score increases, the
Pretest E2 subscore is predicted to decrease.

4.4.4. Multiple Regression Summary
The results of the multiple regressions are summarized in Table 4.11. All significant

predictors have been included. The predictors which are the focus of this investigation and

149

their relationship to the output are in boldface. Parentheses indicate that the Omnibus test
and one predictor are significant, and one predictor approaches significance. Brackets
indicate that the Omnibus test and one predictor approach significance, and one predictor is
significant. Variables separated by a slash mark indicate that separately the variables are
significant predictors, while together they are not. Variables connected with a plus sign
indicate that both variables are needed for significance.

With respect to the variables which are the focus of investigation, Table 4.11 shows
that adverb input, when combined with previous years of French, approaches significance
as a predictor of the change in the correct and error scores of the WOC Task fi'om Posttest
1 to Posttest 2 of both groups together.

Raven's score alone is a significant predictor of the change in the E2 subscore of the
WOC Task from the Pretest to Posttest l of both groups together. This variable is also a
significant predictor of the Pretest correct score of the A] Task of both groups. When
Raven's score is combined with gender, these variables are significant predictors of the
Pretest E2 subscore of both groups and of the inferencing group alone. These variables are
also significant predictors of the Pretest error score of the WOC Task of both groups. When
Raven's score is combined with age, these variables approach significance as predictors of
the change in the correct score of the A] Task from the Pretest to Posttest 2 of both groups.

Inferencing input alone is a significant predictor of the change in correct score of the
AJ Task fi'om Posttest l to Posttest 2, as well as a predictor of the change in the error score
of this task from the Pretest to Posttest 2 and fi'om Posttest l to Posttest 2. When combined
with gender, these variables are significant predictors of the change in the error score of the

WOC Task from the Pretest to Posttest 2.

150

Table 4.1 1. Multiple Regression Results

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

WOC PRETEST PTl-PRE PT2-PT1 PT2-PRE
TASK
Cl INF YRS/AGE
BOTH AGE
C3 INF MF
BOTH YRS
C2+C4 INF YRS
BOTH
TOT INF YRS/AGE AGE
COR
BOTH YRS YRS/AGE
(YRS+ADV)! AC!
El INF MF/YRS YRS
BOTH YRS YRS YRS
E2 INF F+RAV! E! MF
BOTH F+RAV!E1 RAV! +A MF
E3 INF
BOTH
TOT INF MF MF/YRS AGE/YRS M+INF1 -A
ERR
BOTH F+RAVI El MF/YRS AGE/YRS
(YRS+ADV)! -A
AJ
TASK
CORR INF AGE INF! A!
BOTH RAV! C! AGE [AGE+RAV]
AGE! A!
RAV! Al
ERR INF INF! -A INF! -A
B=OTH AGE
MF=Gender INF =Inferencing Input

YRS=Previous Years of French RAV=Raven's Score

ADV=Adverb Input

151

4.5. TEACHER QUESTIONNAIRE

As discussed in Chapter 3, if one assumes that the maximum amount of input is
received if exercises are conducted either as pairs or individually and then reviewed as a
class, the highest adverb input score possible from the 8 day syllabus given to the teachers
was 151. The results of the teacher questionnaire are as follows. Given how the lessons
were taught by each of the teachers, the maximum possible adverb input score for a student
in the noninferencing sections was 75; the maximum possible adverb input score for a
student in one section of the inferencing group was 92 and the maximum possible score for
a student in the other section was 90.
4.6. STUDENT QUESTIONNAIRE

The results of the student questionnaire are listed below.
1. Compare the first four weeks of French class with the last two weeks of class. Was

there any difference in the way the class was conducted?

yes no
If yes, what was different? (lessons, amount of work...)

Class 1 Class 2 Class 3 Class 4
INF INF NONINF NONINF
n= 1 3 n=17 n=1 1 n=4
Yes-13 Yes-l4 Yes-6 Yes-4
NO-O No-3 No-5 No-O

Of the students who reported that the classes were different, students in the
noninferencing classes reported that they had more work to do and were working at a faster
pace. Students in the inferencing classes reported that they were participating in class
more,” they had more work to do and that the work was a little more difficult, because

"some of it was brain-teasers."

 

" This is supported by the higher adverb input score which was computed for these classes based
on the teachers’ responses to the Teacher Questionnaire.

152

In the last two weeks, were you taught the passe' composé? yes no
Ifyes, when is the passé composé used?

Class 1 Class 2 Class 3 Class 4
INF INF NONINF NONINF
n=1 3 n=l7 n=1 1 n=4
Yes-10 Yes-l7 Yes-11 Yes-4
No-3 No-O No-O No-O

All students reported the correct rule for use of the passé compose'.

In the last two weeks, were you taught where adverbs are placed in a French
sentence? yes no
If yes, where are the adverbs placed?

Class 1 Class 2 Class 3 Class 4
INF INF NONINF NONINF
n=1 3 n=1 7 n=1 1 n=4
Yes-5 Yes-8 Yes—5 Yes-3
No-6 No-9 No-6 No-l

"Indirectly"-2

Approximately 50% of the students in the study reported that they had been taught

the rule of adverb placement. The 5 students in class #3 who reported having been taught

the rule of adverb placement gave the correct rule (SVAO). The 3 students in class #4 who

reported having been taught the rule for adverb placement stated the rule was SAVO word

order. Of the 5 students in class #1, 2 reported that adverbs are in ASVO or SVOA position,

1 reported that SAVO was the correct order, and 2 reported the correct SVAO order. The

2 students who reported having been taught the rule "indirectly" gave the correct SVAO

order. Of the 8 students in class #2 who reported having been taught the rule, 5 gave the

correct SVAO word order and 3 gave the incorrect SAVO word order.

Are your mistakes corrected in class? yes no
If yes, can you think of an example of when you were corrected?

153

Class 1 Class 2 Class 3 Class 4
INF INF NONINF NONINF
n=13 n=17 n=11 n=4
Yes-12 Yes-16 Yes-11 Yes—4
No-l No-O No-O No-O

" Sometimes"- 1
Students in all classes reported that they are corrected on their pronunciation and use
of improper tense. No students reported being corrected on adverb placement.
5. Why are you taking French?
Approximately 50% of the students reported that French is a requirement for

graduation; the other 50% reported that they are taking French because "it is important to

know a foreign language."

6. Do you think French is easy or difficult to learn?
Class 1 Class 2 Class 3 Class 4
INF INF NONINF NONINF
n=13 n=17 n=11 n=4
Easy-4 Easy-3 Easy-5 Easy-1
Dif-9 Dif-8 Dif-6 Dif-l

Medium-6 Medium-2

7. Do you think French and English are similar or dissimilar? Can you think of an
example of how they are similar? Can you think of an example of how they are

dissimilar?

Class 1 Class 2 Class 3 Class 4
INF INF NONINF NONINF
n=13 n=17 n=1 1 n=4
Sim-O Sim-0 Sim-0 Sim-0
Dissim-l Dissim-O Dissim-2 Dissim-2
Both-12 Both-17 Both-9 Both-2

The majority of students provided examples of what they found easy and difficult in
French (question #6), even though they were not asked for examples. Interestingly, the

examples given of what is easy in French corresponded with the examples in question #7 of

154

similarities between French and English; and the examples of what is difficult in French
corresponded with the examples of the differences between French and English. However,
the similarities do not outweigh or equal the differences in making French easy to learn as
is evidenced in the minority of "easy" responses to question #6. Students find that cognates
make French easy to learn and similar to English. Students find that gender, agreement, the
use of definite articles, the verb tenses, and the "exceptions to rules" make French difficult
to learn and dissimilar from English.
4.7. CHAPTER SUMMARY

In this chapter I have described the treatment groups in the study based on age,
gender, previous years of French, adverb input, inferencing input, and Teacher and Student
Questionnaires. I have provided a descriptive summary of the groups' performances on the
AI Task, the WOC Task, and Raven's SPM Test. I presented the results of the repeated
measures ANOVA which was performed to determine any significant differences between
and within groups across testing sessions on both tasks. Finally, I presented the results of
the multiple regressions which were performed to determine any significant predictors of the

group's performance on the tasks. These results are analyzed and discussed in Chapter 5.

CHAPTER 5
DISCUSSION

In this chapter I analyze the performance of the treatment groups, as revealed in the
repeated measures ANOVA, in terms of the significant predictors,‘ as revealed in the
multiple regressions. This analysis involves a comparison within and across groups with
respect to performance on the two tasks and with respect to significant predictors.2 I also
analyze the performance of all subjects considered as a single group. I highlight four
observations that can be made from the present study and discuss them based on their
relevance to the goal of the present study-«an investigation of the relationship between the
use of indirect negative evidence and Second Language Acquisition. I discuss the caveats
of the study and conclude the chapter with suggestions of future research that may help
elucidate the topic of this dissertation.

Several preliminary observations with respect to the variables in this study are
necessary. These observations are an attempt to tease apart the interaction, and at times

conflict, between the variables. They also provide a partial framework within which the

 

‘ As stated in Chapter 4, the variables of adverb input, inferencing input, and Raven’s score are the
main focus. However, other variables are discussed as they relate to the variables under question. Those
regressions which indicate gender as a significant predictor will be regarded as spurious and not discussed,
given that there were only 4 males out of a total of 45 subjects.

2 It must be emphasized that the repeated measures ANOVA revealed no significant difference
between groups on either the correct scores or the error scores of either task nor any significant interaction
between group and testing session; finally, both groups showed significant improvement from the Pretest
to Posttest 2. It may still be informative to analyze the trends of each group with respect to the significant
predictors of performance.

155

156

analyses are placed.
5.1. CORRELATION

Recall from Chapter 4 that there is not a wide range between the ages of the subjects
in the two groups.3 It is therefore somewhat surprising that the variable of age was found
so ofien as a significant predictor of performance. To determine if another variable was
perhaps 'hidden' in the age variable, Pearson product-moment correlation coefficients were
calculated for age, previous years of French, adverb input, inferencing input, and Raven's

score.“ The results are listed in Table 5.1.

Table 5.1. Pearson Product-Moment Correlation

 

 

 

 

 

Age Years Frn AdvInput InferInput
Years Fm -.356*
AdvInput -.1 17 .089
InferInput .063 -.318* .581"
Raven's .159 -.009 -.199 -.0305

 

 

 

 

 

 

 

(*p< .01; *"‘p< .001)

Based on these correlations, in addition to the information shown in Tables 4.1 and

4.2, "Learner Variables," several generalizations can be tentatively made. First, note that

 

5 Subjects ranged from 17 to 20 years old, with one 22 year old.

“ MINITAB’s multiple regression analysis will determine if two variables are strongly correlated
(see Footnote 6, Chapter 4); that is, the correlation must be .80 or greater. To uncover weak correlations,
Pearson product-moment correlation was applied.

5 Recall from Table 4 in Chapter 4 that the inferencing group did perform slightly better on the
Raven’s test. Additionally, this group received a higher adverb input score. The extremely weak negative
correlations between Raven’s score and inferencing input and adverb input shown here would indicate that
those students who were got absent (and thus obtained higher inferencing input and adverb input scores)
received lower Raven’s scores.

157

there is a significant negative correlation between age and previous years of French study.
In other words, younger subjects have had more years of French. Second, there is a
significant negative correlation between inferencing input and previous years of French
study. In other words, subjects in the inferencing group had fewer years of French than
subjects in the noninferencing group. Finally, there is a significant positive correlation
between adverb input and inferencing input. That is, subjects in the inferencing group
received higher adverb input scores than subjects in the noninferencing group.

In effect, therefore, the variable of previous years of French is reflected in the
variable of age. However, there must be an aspect to the age variable other than merely
previous years of French study, because this latter variable was not found as a significant
predictor in all the instances in which age was a predictor. It may be that the variable of age
is also indicative of years of continuous French study. For example, it is not unreasonable
to assume that the 17 year old subjects were taking French in high school hmediately prior
to enrolling in French 150 at MSU; however, the 20 year old subjects have been away fi'om
French for several years.“ If this is in fact the case, two somewhat contradictory variables --
contradictory with respect to their relationship to a subject's improvement/performance --
are contained within a single variable. That is, an Older learner who has had fewer years of
French has more room, so to speak, for improvement; yet at the same time, this older learner

has been away from French for a period of time, which may hinder performance.

 

“ Recall that French 150 is an intensive course in which first year French is covered in one
semester. Therefore, students would not have enrolled in any other MSU French courses prior to taking
French 150.

158

5.2. ACCEPTABILITY JUDGMENT TASK7

5.2.1. Correct Score

 

 

 

 

 

 

 

 

 

l
POSTTEST 1 roar-rest 2
l - -.—- msenencmo —-.— noumrensncmo

 

 

Figure 5.1. A1 Task Mean Correct Scores

Figure 5.1 shows that the groups performed similarly on the AI Pretest with respect
to their acceptance of SVAO word order and their rejection of SAVO word order. Of note
is the relatively high mean correct score (+9 out of 18) of both groups. In other words, all
subjects began with a relatively high knowledge of both the correct and the incorrect word
orders. Raven's score was found to be a predictor of the Pretest scores for both groups. This

finding supports the proposal that inductive ability somehow facilitates or is positively

 

7

For convenience, the line graphs from Chapter 4 will be repeated here.

159

related to second language acquisition. Recall that the inferencing group performed slightly
better on the Raven's test, which may explain why this group, even though they had fewer
years of previous French study, obtained approximately the same Pretest scores as the
noninferencing group.

From Pretest to Posttest 2, age and Raven's score taken together approached
significance as predictors of the change in score; the Older subjects showed greater change,
and the subjects with higher Raven's scores showed less change. These results follow from
the facts that 1) the older students had fewer years of French study, and therefore, in a sense,
had more to learn; and 2) those students with higher Raven's scores obtained higher Pretest
scores, and therefore, their change in scores was less; that is, they were initially closer to a
perfect score of 18 than those subjects who received lower Raven's scores.8

During the treatment period, that is, by Posttest 1, the inferencing group performed
roughly similarly to the noninferencing group. Age was a significant predictor of change
during this time period. If one assumes that the age variable is representative, in part, of
fewer years of French study, then again, it follows that those older students with fewer years
of French showed greater change because they had more to learn. It is possible that the age
predictor is also indicative of those Older students who received higher Raven's scores
(r=.159), suggesting that inductive ability (as measured by Raven's) is indirectly revealed
as a predictor of change in correct score.

After the treatment period, the difference between the groups increases, with the

 

5 Here again, however, the variables seem to be in conflict. There is a weak positive correlation
between Raven’s score and age; that is, older subjects received higher Raven’s scores. Older subjects are
predicted to show greater change, while subjects with higher Raven’s scores are predicted to show less
change.

1

noninferencing group performing better.

60

It may be that the more years of French study of

the noninferencing group allowed them to continue on their path of improvement while the

inferencing group fell behind after an initial improvement. Note that from Posttest l to

Posttest 2, inferencing input was a significant predictor of improvement in the correct score.

This may indicate that the treatment had a somewhat delayed effect for certain subjects in

the inferencing group.

5.2.2. Error Score

 

 

 

 

 

 

 

 

7 "~“
8 -\ -.“
0..~
to ’~~,
2 ~‘~‘
<
III ‘-.~
, .‘~~‘
5 "t:..
4 \
3 l
PRETEST POSTTEST 1 POSTTEST 2

 

u. -- INFERENCING

 

—.— NONINFERENCING

 

 

Figure 5.2. AJ Task Mean Error Scores

Figure 5.2 shows the mean error scores for both groups at each testing session.

Consistent with the results of the Al Pretest correct score, all subjects made relatively few

161

errors at the Pretest; at the beginning of the study, the students were neither rejecting SVAO
word order nor accepting SAVO word order.

Among those variables of concern, no significant predictors of performance of both
groups combined were found.9 However, it is possible that the predictor of age is a
reflection of not just older subjects but those older subjects who received higher Raven's
scores, as mentioned in the discussion of the A] Correct Score. Thus, Raven's score is
indirectly revealed as a predictor of a lower A] pretest error score.

Within the inferencing group, inferencing input was again (as with the A] Correct
Score) a significant predictor of this group’s change from Posttest 1 to Posttest 2, as well as
from Pretest to Posttest 2. As mentioned in the discussion of the AI Correct score, a delayed
effect of the treatment on this task as a whole may be evidenced. Additionally, if one
assumes that previous years of French, though not found as a significant predictor, account
for both the higher correct scores and the lower error scores of the noninferencing group,
then it must be explained why the change in error score of the inferencing group, who has
fewer years of French study, is roughly equivalent to that of the noninferencing group. It
may be that while the apparent effect of inferencing input does not result in the inferencing
group making as few errors as the noninferencing group, this input does contribute to the

former group keeping pace with the latter.

 

9 Age was a significant predictor of the Pretest; older subjects are predicted to make fewer errors.
This finding, coupled with the finding on the correct score from the Pretest to Posttest 1, indicates that,
in general, older subjects were predicted to perform better on the A] Task than they did on the WOC Task.

162

5.3. WORD ORDER CORRECTION TASK

5.3.1. Correct Score

10

 

 

 

 

 

 

 

 

 

 

9.5
a t" /
a I
i a 5 .
m I
s ‘ /
3 x4
O
7.5 *7
7 I
PRETEST POSTI’EST 1 Posrresr 2
-..... INFERENCING —.— NONINFERENCING

 

 

 

Figure 5.3. WOC Task Mean Correct Scores

Figure 5.3 shows the mean correct scores of both groups on the WOC Task at each
testing session. Of note is the relatively high mean correct scores of both groups at the
Pretest, supporting the finding fi'om the A1 Task that the subjects were already familiar with
the rules of adverb placement in French when the study began. This is the only score on
which the inferencing group consistently scored higher than the noninferencing group, and
as summarized in Chapter 4, a greater percentage of the inferencing group's mean correct

scores on all tests consisted of changes to the correct SVAO word order rather than passive

163

acceptance of this word order. However, the difference between the groups is not significant
and results of the multiple regression analysis indicate that this finding cannot be attributed
to the inferencing input. The noninferencing group did exhibit slightly greater improvement
fiom Pretest to Posttest 2, although again, the difference between the groups' performances
is not significant.

As with inferencing input on the AI Task, there seems to be a delayed effect of the
treatment on this score. That is, adverb input and previous years of French study, taken
together, are significant predictors of change from Posttest l to Posttest 2 for both groups
combined. Interestingly, even though the inferencing group received a higher adverb input
score, their change in means during this time period was almost half that of the
noninferencing group. Additionally, adverb input was not found as a predictor of change for
the inferencing group during this time period; rather, older subjects (or, as has been argued,
subjects with fewer years of French) in the inferencing group are predicted to show less
change in their correct scores. As with the A] Task, this suggests that the noninferencing
group was perhaps at an advantage as a result of more years of French study.

It should be noted that from the Pretest to Posttest l for the groups combined, not
only is adverb input pot a predictor of greater change but also le_ss_ change is predicted during
this time period for students with more years of French (thus, the lower change in mean of
the noninferencing group). It was suggested in the discussion of the correct score on the AI
Task that the change was less for certain students because they were initially closer to a
perfect score. This cannot explain the lack of change on the correct score of the WOC task,
because no student received a perfect score on this task at Posttest 1. Perhaps an

understanding of this finding can partially be found in the error scores of this task during this

l 64
time period.

5.3.2. Error Score

 

38 M
O \

306 ‘s
\
\
S
\
s
\
\
\
\
14 ‘
\

31

 

 

 

MEANS

d
I
I
o
I
I
I
I
6’
I

 

 

 

2.8 T ;?

 

 

 

23
lfluflEST

I
POSUEST1

 

--.- - INFERENCING

—'— NONINFERENCING

POSHESTZ

 

 

 

Figure 5.4. WOC Task Mean Error Scores

Figure 5.4 shows the mean error scores of both groups on the WOC Task at each
testing session. At the Pretest, both groups are making a minimal number of errors (4 out
of 16). Raven's score was found as a significant predictor of the initial error score; those
subjects who obtained higher Raven's scores made fewer errors. Interestingly, Raven's score
was also a predictor of a lower E2 Pretest subscore (changes to SAVO), which indicates
elimination of active use of this structure fi'om the grammars of the subjects. As with the AI

Pretest correct score, both groups received almost identical Pretest error scores on the WOC

165

task. Given that the noninferencing group had more years of French study, one would expect
their Pretest error score to be lower than that of the inferencing group. However, as
suggested in the discussion of the Al Task, the inferencing group performed slightly better
on the Raven's test, which may explain why they obtained approximately the same Pretest
error score as the noninferencing group. This suggests that greater inductive ability may to
some extent facilitate second language acquisition.

From the Pretest to Posttest 1, previous years of French was found as a significant
predictor of change as it was on the correct score of this task during this time period. Recall
the somewhat counterintuitive finding that those subjects with more years of French were
predicted to show less improvement in their correct score. The prediction is the opposite on
the error score; that is, those subjects with more years of French are predicted to show a
decrease in error score from Pretest to Posttest I. It may be then that the subjects are more
'focused' on the rejection of SAVO word order at the expense of using the correct SVAO
word order. However, this still does not explain why the noninferencing group, who has
more years of French, did not perform as well as the inferencing group on the error score
from the Pretest to Posttest 1.‘0

From Posttest 1 to Posttest 2 adverb input and previous years of French are
significant predictors of a decrease in error score for both groups combined. These variables
were also significant predictors of a change in correct score during this time period; there,
therefore, seems to be a delayed effect of the treatment, when combined with years of

French, on this task as a whole. As with the correct score, there again is a greater influence

 

‘° From the Pretest to Posttest 1, a higher Raven’s score was a predictor of a positive change in error
score. This finding is contradictory to what is found in other instances in which Raven’s score is a
predictor, suggesting that this is a spurious result.

166

of previous years of French than of adverb input, as revealed in the multiple regression;
additionally, the noninferencing group, who has had more years of French study, had a
change in means during this time period of more than twice that of the inferencing group,
who had more adverb input. Furthermore, previous years of French and not adverb input
was found as a significant predictor of change for the inferencing group alone, indicating a
greater effect of the former variable.

Inferencing input was a significant predictor of overall change (Pretest to Posttest 2)
for this group. It may be that their lack of French study, relative to the noninferencing
group, was somehow compensated for by the inferencing input since this group's final error
score was virtually the same as that of the noninferencing group's final score. Additionally,
recall that inferencing input was also a significant predictor of the change in error score on
the AI Task fi'om Posttest 1 to Posttest 2 and from the Pretest to Posttest 2. Recall also that
a low error score indicates that the subject 'knows' what is got possible. That inferencing
input was found as a predictor of the change in error score on both tasks lends support to the
argument that the acquisition of the ungrammaticality of a structure may depend on the use
of inductive inferencing.

5.4. OBSERVATIONS

In this section I highlight four observations that have been made regarding the
influence of the variables on the performance of the two groups and suggest possible
explanations for these observations, which are as follows: 1) There appears to be a delayed
effect of the adverb input and inferencing input treatments on certain scores; 2) Previous
years of French study have a stronger influence on performance than adverb input; 3) There

is no significant difference between the performance or improvement of the two groups; and

167

4) Inductive ability, as measured by Raven's Standard Progressive Matrices, is to some
extent an indicator of Second Language Acquisition ability. These observations, which are
integrally and complicatedly related to each other and to the findings of this study, will be
discussed in turn.

5.4.1. Adverb Input: A delayed effect or reanalysis“

The multiple regression analysis indicates that there is a delayed effect of the adverb
input combined with previous years of French on the performance on both scores from
Posttest l to Posttest 2 of the WOC Task by all subjects considered as a single group. For
the purpose of this discussion, it may be useful to examine the change in means of both

groups combined. This is listed in Table 5.2.

Table 5.2. WOC Task Change In Means Of All Subjects

 

 

 

Post 1 - Pretest Post 2 - Post 1 Post 2 - Pretest
Correct Score .75 1.05 1.8
Error Score -.585 - .66 -1.245

 

 

 

 

 

 

It seems somewhat counterintuitive that a greater change in means was obtained after

Posttest 2 rather than immediately following the treatment period at Posttest 1.‘2 I would

 

” The apparent delayed effect of the inferencing input on the performance of this group on the A1
Task and on the error score of the WOC Task is discussed in the next section, "Unequal Variables." I
suggest that the finding of inferencing input as a significant predictor may be better explained in terms of
the interaction of the variables than in terms of reanalysis.

‘2 Particularly in light of the fact that these results contradict those of Trahey and White (1993), who
reported significant difference fiom the Pretest to Posttest 1 (but not from Posttest I to Posttest 2) on both
scores of the WOC Task and on the correct score of the AJ Task. These differences were attributed to
the adverb input flood. It may be that the authors’ tallying procedures (we Chapter 4) resulted in ’inflated’
correct scores. One is unable to determine what percentage of the correct score consisted of acceptance
of the incorrect word order. It is possible that the ’true’ correct score did not change significantly. Other
factors which could account for the difference in findings include 1) Subjects in the Trahey and White study

168

like to suggest that a process of "reanalysis" (Gass, 1988) or "restructuring" (McLaughlin,
1990) is taking place.

Gass (1988) proposes that

evidence for integrated knowledge can be seen in one of two ways. First there can be
changes in the rule system that surface in the output. This is in fact what is typically
thought of when one considers developmental changes. Second, there may be changes
in the underlying system although there is no output change. Changes in the underlying
systems with no surface manifestation are typically subsumed under the category of
reanalysis. Within a second language context, we can think of reanalysis in two ways.
First, a reanalysis of the underlying system may affect the potential for output . . . .
Second, on a syntactic level, prefabricated patterns may be analyzed with, initially, little
output change. (Gass & Selinker, 1994, p. 305)

On the WOC Task, then, it is possible that the adverb input and previous years
of French facilitate a reanalysis of the underlying system from the Pretest to Posttest 1,
or affect the potential for output change, which is not realized until Posttest 2.

Further support for reanalysis is provided by examining the predicted outcomes
of the two scores from Pretest to Posttest 1 on the WOC Task with respect to each other.
Recall that those subjects with more years of French were predicted to show less change
on the WOC correct score from the Pretest to Posttest 1 and to show more (negative)
change on the error score during this same time period. In other words, improvement
in one area coincides with ’stagnation’ in another area. Lightbown (1985, p. 177)

suggests that

[Restructuring] occurs because language is a complex hierarchical system whose
components interact in non-linear ways. Seen in these terms, an increase in error rate
in one area may reflect an increase in complexity or accuracy in another, followed by
overgeneralization of a newly acquired structure, or simply by a sort of overload of
complexity which forces a restructuring, or at least a simplification, in another part of
the system.

As has been explained, a low error score is indicative of knowledge of what is

 

were approximately ten years younger than subjects in the present study; and 2) Trahey and White had
greater control over the classroom lessons (see the section 'Caveats").

169

mt possible. The correct score is indicative of knowledge of what is possible.‘3 In the

present study, a high correct score implies that SVAO word order has been incorporated

into the learner’s grammar. A low error score implies that SAVO word order has been

eliminated from the learner’s grammar, which, it has been argued necessarily results

from noticing the absence of this structure in the target language. There is a qualitative

difference between these aspects of grammatical development. Thus, there is a

qualitative change (or restructuring) taking place on this task from the Pretest to Posttest

l which may be attributed to previous years of French study.

5.4.2. Unequal Variables

To summarize thus far, we have seen that on the WOC Task

1)

2)

3)

4)

previous years of French study is a significant predictor of change on both
scores from the Pretest to Posttest 1;

previous years of French study have a greater influence than adverb input
on the change in both scores from Posttest 1 to Posttest 2;

inferencing input is a significant predictor of overall change in error
score; and

the noninferencing group, who had more years of French, showed a
greater overall change in means of both scores, though not statistically
significant, than the inferencing group.

We have seen that on the Al Task

5)

previous years of French (as ’hidden’ within the age variable) is a
significant predictor of change in the correct score from the Pretest to
Posttest 1;

 

‘5 The correct score, to some extent, also indicates knowledge of what is not possible. For example,
if SAVO word order is changed to ASVO, one may assume that the student knows that SAVO is incorrect.
To conclude, however, that SAVO has been eliminated from the grammar, the correct and error scores
must be considered together.

170

6) inferencing input is a significant predictor of change in both scores from
Posttest l to Posttest 2;

7) inferencing input is a significant predictor of overall change in the error
score; and
8) the noninferencing group, who had more years of French, showed a

greater overall change in means of both scores, though not statistically
significant, than the inferencing group.

Observations 1, 2, 4, 5, and 8 suggest that the variable of previous years of
French has a stronger influence on performance than adverb input. ‘“ Somewhat
contradictorily, observations 3, 6, and 7 suggest that m the inferencing group,
inferencing input has a stronger influence than either previous years of French or adverb
input. These observations raise the following three questions. First, why is it that the
adverb input did not have more of an influence? Second, if one assumes a strong effect
of previous years of French, then why is it that this variable was not a ’clear"5
significant predictor of performance on the AI Task? Finally, if one assumes a strong
effect of previous years of French, then why is it that this variable was not a significant
predictor of change for the inferencing group? Or, the corollary to this, why is it that
inferencing input was a significant predictor?

To answer the first question, the adverb input score must be reexamined. A two-
tailed t-test shows that there is a significant difference (t=3. 18; p< .01) between the two

groups with respect to adverb input. However, the inferencing group received more

 

‘“ Recall also that there is a significant negative correlation (r= -.318; p< .01) between previous
years of French and inferencing input. However, a two-tailed t-test reveals that there is not a significant
difference (t=2.14; p< .05) between the two groups with respect to previous years of French. Given that
the noninferencing group had approximately 6 months more of French study, which in some cases may
make a substantial difference, might explain there slightly better performance.

‘5 Recall that it was argued that the variable of previous years of French was ’hidden’ in the age
variable.

171

adverb input than the noninferencing group, which would imply, assuming a positive
effect of the input, that the inferencing group should have outperformed the
noninferencing group. This finding was not obtained. Perhaps the error is in the
assumption that more adverb input, as quantified in the present study, is necessarily
better.

As stated in Chapter 4, the adverb input score was designed as an attempt to
quantitatively compare across class sections and to determine the different types of input,
if any, provided to the students. It was assumed that if an exercise was done
individually, or as pair work, students were attending to the task and receiving the input;
whereas, with class activities, the possibility exists that students were not paying attention
and thus not receiving the input. Therefore, students received a higher adverb input
score if the activity was performed individually or in pairs.

The finding that adverb input did not have a stronger influence on the
performance of the subjects lends support to Gass’ (1988, p. 206) proposal that
"pressures of conversational interaction may preclude sufficient analysis for the purposes
of intake. In this case, the input (even though comprehended) may have no further role
in acquisition.” Thus, those students in the inferencing group may actually have been
at a disadvantage as a result of the demands placed on them to perform. They were not
given the time necessary to assimilate the new linguistic material; thus, the input did not
become intake, which is a prerequisite to grammar formation.‘“

This time factor provides only a partial explanation, however, because if it were

 

‘“ Gass (1988) also proposes that prior linguistic knowledge is one of the factors mediating what
becomes comprehended input and what becomes intake; the finding in the present study that previous years
of French is a significant predictor of improvement on the WOC Task supports this proposal.

172

the case that a lower adverb input score (or less pair work) in fact facilitated the
assimilation of new linguistic material, this would have been revealed in the multiple
regression analyses. That is, a lower adverb input score would have been a significant
predictor of change in scores. This was not the case. I would like to suggest that
another variable, which is somewhat elusive and not easily quantified, is obscuring the
findings.

Recall that it was concluded based on a multiple regression analysis that variations
in teaching style, as measured by the way lessons were taught, were being captured in
the adverb input score. Multiple regressions using test scores as the outputs and class
sections as the predictor variables provide us with information of which class is predicted
to perform better. Significant predictors are found for both scores on Posttest 2 of the
WOC Task and on the correct score of Posttest 2 of the AI Task. In each case,
performance, from best to worst, is ordered as follows: class #3, class #1, class #2,

class #4. For convenience, Table 4.2 is repeated here as Table 5.3.

173

Table 5.3. Learner Variables By Class Section

 

 

 

 

 

 

VARIABLE Classl Class2 Class3 Class4 ALL
INF INF NON NON
n=13 n=17 n=11 n=4 n=45
MEAN 18.77 18.35 18.27 18.25 18.44
AGE SD 1.30 .79 .79 1.29 .99
MIN 17.0 17.0 17.0 17.0 17.0
MAX 22.0 20.0 20.0 20.0 22.0
MEAN 3.40 3.51 4.6 2.83 3.68
YEARS OF SD 1.04 .72 .56 .99 .97
FRENCH MIN 2.0 2.0 4.0 2.0 2.0
MAX 5.0 4.3 5.6 4.1 5.6
MEAN 79.85 87.47 75.0 54.75 79.31
ADVERB SD 17.14 7.29 0.0 26.13 17.0
INPUT MIN 37.0 73.0 75.0 17.0 92.0
MAX 90.0 92.0 75.0 75.0 15.21
MEAN 7.23 7.59 7.43
INFERENCING SD 1.25 .49 .92
INPUT MIN 4 7 4
MAX 8 8 8

 

 

 

 

 

 

 

 

 

If it were the case that previous years of French is the 'strongest' variable,
performance by class section would be as follows: 3, 2, l, 4; if it were the case that loss
adverb input is the 'strongest' variable, performance by class section would be as follows:
4, 3, 1, 2; if it were the case that me adverb input is the 'strongest' variable, performance
by class section would be as follows: 2, I, 3, 4. None of these results is obtained. I believe
that the rank order obtained reflects not only the complex interaction of the variables in this
study but also indicates the influence of afl‘ective variables on students' reception of the
input, regardless of input type. This in turn affects their performance. These variables may
include student/teacher rapport, motivation, or anxiety.

Thus, in answer to the first question, "why is it that the adverb input did not have

more of an influence?" I have suggested that the findings support the theoretical proposals

174

that 1) individual or pair work, which may ensure that students receive the input, does not
necessarily lead to intake; and 2) other (affective) variables mediate between the type of
input received and performance.17

I now turn to the second question, "if one assumes a strong effect of previous years
of French, then why is it that this variable was not a 'clear' significant predictor of
performance on the A] Task?" This finding may appear somewhat surprising given that the
variable of previous years of French is a significant predictor of performance on the WOC
Task and, fiirthermore, that the WOC Task and AJ Task are intended to measure the same
thing -- knowledge of adverb placement in French. However, notice that each task poses
quite different demands on the students. The WOC Task requires that students manipulate
the language; the AI Task requires that students choose the correct sentence(s). In other
words, different 'levels' of facility with the language are required to perform each of the
tasks. Given the difference in the tasks, it therefore follows that different factors may be
involved in the performance on each task. While more years of French study (or prior
linguistic knowledge) may be required for performance/improvement on the WOC Task,
other variables can, in a sense, compensate for a relative lack of prior linguistic knowledge
and can facilitate performance on the AJ Task.18

What are these other variables? I believe the answer lies, in part, in the third

 

‘7 One can also speculate that those students who had more years of French study had received
(more) negative evidence, or overt correction, on adverb placement while those students with fewer years
of French study had not (or had received less negative evidence). Therefore, negative evidence (previous
years of French) has more of an effect than only positive evidence (adverb input). However, one would
still have to explain why this negative evidence was not of use on the A] Task; that is, why previous years
of French study was not a clear predictor of performance on this task.

‘9 The implications of this finding, that is, the possibility that different abilities or skills may be
operating in successful completion of the two tasks, will discussed in the section "Future Research. "

175

question, "why is it that inferencing input was a significant predictor of performance?" A
partial answer to this question has been alluded to in the responses to the first and second
questions above. That is, I have suggested that pressures to perform may hinder input from
becoming intake. NO such pressures existed with the inferencing exercises; students worked
alone and then the exercises were reviewed as a class. In other words, the yyay in which
these exercises were taught contributed to their effectiveness. I have also suggested that the
demands of each of the tasks are quite different and may require the use of different abilities
or skills. The relationship of these abilities to each other, to task type, and to the exercises
which were intended to utilize these abilities, while still nebulous, is suggested in the
multiple regression analyses. While previous years of French are advantageous for
improvement on the WOC Task, they are not a significant factor in improvement on the AI
Task. Conversely, while inferencing input is not a significant factor in improvement on the
WOC correct score, it is beneficial for overall improvement on the error scores of both tasks
and on the A] Task correct score from Posttest l to Posttest 2. Ifone assumes that inductive
reasoning was indeed employed in the inferencing exercises, these findings then suggest that
these exercises facilitated the use of inductive reasoning on later linguistic tasks.‘9
Moreover, it facilitates performance on tasks and during time periods when previous years
of French appear to have little effect. Interestingly, inferencing input is a significant
predictor of overall change in error score on both tasks. As has been argued, low error
scores suggest that learners have noticed the absence of SAVO word order in the target

language and, therefore, have eliminated this structure from their grammars. It may follow

 

‘9 Note, however, that this explanation does not account for the improvement of the noninferencing
group.

176

then that inductive reasoning plays a role in noticing the absence of SAVO word order.20

Thus, in answer to the second question, "why is it that the variable of previous years
of French was not a clear predictor of performance on the A] Task?" and to the closely
related third question, "why is it that inferencing input was a significant predictor?" I have
suggested that 1) the WOC and AJ Tasks pose different linguistic demands, 2) the way in
which the inferencing exercises were presented to the students facilitated this input
becoming intake, and 3) the type of inferencing exercise facilitated performance on linguistic
tasks in which previous years of French seem to have little effect.

In this section I have concentrated on the second observation that the variables in this
study do not have equal effects on performance nor does a single variable have consistent
or uniform effect on performance across tasks. I have explained these observations in terms
of the complex relationships between type of input, teaching methodology, and task
demands. These complex relationships contribute, in part, to the third observation -- there
is no significant difference between the performance or improvement of the noninferencing
and inferencing groups.

5.4.3. Inferencing versus Noninferencing

Results of the repeated measures ANOVA reveal that there is no significant
difference between the performance of each of the groups on either task at each testing
session, nor is there a significant difference between how the groups changed over time.
However, results of the multiple regression analyses reveal that the factors or variables

contributing to the change of each group on each task are not necessarily the same. That is

 

2“ Additionally, previous years of French study do not guarantee the elimination of SAVO word
order, given that this previous study probably cornisted of positive and negative evidence but not exercises
in inductive reasoning.

177

to say, one cannot necessarily conclude that because there is no quantitative difference
between the improvement exhibited by each group that this improvement is characteristically
similar.

While somewhat of an oversimplification, we have seen that at times (AJ Correct
Score) fewer years of French study predict improvement; at other times (WOC Task) more
years of French study predict improvement; at still other times (AJ Task and WOC Error
Score) inferencing input predicts improvement; and finally, at times (WOC Task and AJ
Correct Score) affective variables seem to be influencing improvement. I have suggested
that the fluctuating influence of a variable is dependent on both the type of task and the
relative presence of other variables. For example, it may be that the effect of inferencing
input, only on the AJ Task, varies with previous years of French study.

We have seen that the groups differ with respect to adverb input, previous years of
French, possibly affective variables such as student/teacher rapport, and, of course,
inferencing input. Yet, in spite of these differences, the groups performed similarly with
respect to their quantitative change over time. Additionally, the change of one group relative
to the other cannot be adequately described in terms of their differences with respect to these
variables. For example, while fewer years of French study predict change on the AI Correct
Score, the inferencing group, who had fewer years of French study, did not show greater
change than the noninferencing group.

That the groups did differ with respect to certain variables is one of the major
shortcomings of the study, which is discussed in the section "Caveats." As a result of this
shortcoming, explanations of the performance of one group relative to the other become

extremely speculative, and these speculations have been provided in the previous sections

178
of this chapter.

However, I would like to point out that it may still be informative to consider the
performance of the inferencing group alone in terms of one of the variables which is the
focus of this study, inferencing input. Even though this input did not result in this group's
performance differing significantly from the noninferencing group, it did have a significant
effect on the performance of the inferencing group on both scores of the A] Task and on the
WOC Error Score.” As has been discussed, if one assumes that the inferencing exercises
utilized inductive reasoning, this finding lends support to the proposals that 1) inductive
reasoning is a learning strategy which can be taught, and 2) inductive reasoning can facilitate
performance on certain linguistic tasks.22
5.4.4. Raven's Standard Progressive Matrices

Further support for this second proposal is provided by the fourth observation -
inductive ability, as measured by Raven's Standard Progressive Matrices, is to some extent
an indicator of Second Language Acquisition ability.

We have seen that Raven's score is a predictor of the pretest correct score of the A]
Task, the pretest error score of the WOC Task, and the pretest E2 subscore (changes to
SAVO) of the WOC Task. I have suggested that perhaps Raven's score is reflected in the

age predictor of the pretest error score of the A] Task and of the change in the A] correct

 

2‘ It could be argued that the variable of inferencing input is merely an indicator of student
attendance. However, if this were the case, one would need to explain why adverb input, in which
absences were also incorporated, was not a predictor of performance.

22 If one assumes that the noninferencing group was at a slight advantage as a result of the fact that
they had more years of French, and one assumes no effect of adverb input, it must be explained why the
inferencing group exhibited progress similar to the noninferencing group. One can speculate that the
inferencing input, coupled with the fact that the inferencing group performed slightly better on the Raven’s
test, facilitated their improvement. However, this is pure speculation at this point.

179

score from the Pretest to Posttest l. I would like to point out two possible implications Of
these findings.

First, that Raven's score was found as a predictor on these particular tests suggests
that a high Raven's score indicates that SAVO word order has been eliminated from the
learner's grammar. Recall that the correct score of the AJ Task reflects error-free usage.
That is, in order for a correct score to be counted on an SAVO-SVAO pair, for example,
subjects had to choose only SVAO. In other words, the correct score indicates both
acceptance of SVAO and rejection of SAVO. Additionally, a high Raven's score is not only
a predictor of a lower error score on the WOC Task, but also of those errors (E2) which
consist of changes to SAVO word order. This finding appears to support the proposal that
there is a positive relationship between the use of inductive reasoning and the acquisition of
the ungrammaticality of a structure, which, I have argued, depends on noticing the absence
of this structure in the target language.

Second, unlike the variable of previous years of French study, the predictor of
Raven's score crosses tasks.23 Recall that in the section "Unequal Variables" I suggested that
the efi‘ect of certain variables depends, in part, on the demands of the task. This appears not
to be the case with Raven's score, which is a clear predictor of the error score on the WOC
Task and of the correct score on the A] Task. This finding suggests that rolotivo to the othor
voriobles in the study inductive reasoning has a somewhat global and 'steadfast'2“ effect on

Second Language Acquisition.

 

25 This is also the case with the variable of inferencing input.

2“ Steadfast in the sense that this variable is not as easily affected by other variables as, for example,
adverb input seems to be.

180

5.4.5. Summary

In this section I have focused on four observations gleaned from the present study:
I) There appears to be a delayed effect of the adverb input and inferencing input treatments
on certain scores which may be explained in terms of reanalysis; 2) There is an unequal
effect of the variables in this study on a subjects's performance which may be explained in
terms of task demands, teaching methodology, and type of input; 3) There is no significant
difference between the performance or improvement of the two groups which may be
explained partially in terms of (2) above and partially in terms of the differences between
groups with respect to these variables; and 4) Inductive ability, as measured by Raven's
Standard Progressive Matrices, is to some extent an indicator of Second Language
Acquisition ability which is suggested by this variable predicting the scores on certain tasks.

The strength of these observations must be judged in light of the weaknesses of this
study, some of which have been alluded to in previous sections and are more fully discussed
in the following section.
5.5. CAVEATS

In this section I discuss the methodological shortcomings of this study in terms of
their possible roles in the following results:25 1) There was not a significant change in
scores on either task for either the inferencing group or the noninferencing group
immediately following the treatment period, suggesting no effect of treatment; 2) There was
not a significant difi‘erence in the performance or improvement on either task between the

two groups; and 3) There was no clear and consistent predictor of performance across tasks

 

25 Note that a single factor, for example small n-size, probably plays a role in all three findings.
For ease of discussion, however, the flaws are categorized according to the result on which they may have
the greatest effect.

181

and time.
5.5.1. No Immediate Effect of Treatment

One of the factors contributing to this result may be the length of time of the study.
Students received the adverb input and the inferencing input over a two week period, or eight
class sessions. Additionally, these sessions did not consist exclusively of adverb exercises.
Furthermore, the inferencing exercises consisted of approximately 15 minutes of each class
period. Recall that Valian (1990a, 1990b), Saleemi (1990, 1992), Archibald (1993), and
Gass (1988) all suggest a learning procedure which relies on the frequency with which an
item occurs or does not occur in the input. It may be that as a result of the short duration of
the treatment period, neither the "frequency threshold" nor the "time threshold" (Saleemi,
1992, p. 128) was crossed.

Related to the time factor is the fact that the subjects already know tho mlo of odverb
placement at the beginning of the study, as revealed in the high pretest correct scores and the
low pretest error scores. This is related to the time variable because frequently, especially
with subjects of this age, one finds marked improvement initially and then slower progress
in the later stages of acquisition. Perhaps more concentrated work is required for these
students. And, of course, the input would have no effect on the improvement of those
students who initially received perfect scores.”

Related to the fact that the students were relatively proficient in adverb placement
is the issue of sample-size, and more specifically the rapoomness of the sarnplo. While all

the students received the input, students volunteered to take the tests. That is, the test scores

 

2“ Additionally, a more concentrated and longer treatment period may result in more pronounced
differences between groups.

182

are of students who volunteered to come on their 'off-day' to participate in the study. In
other words, the results are representative of motivated students and not necessarily of all
students at this proficiency level.

Finally, given that the students were already familiar with adverb placement in
French, one can suppose that they were equally familiar with the fact that odverbs are osed
infrequently in French, relative to their use in English.27 To a certain extent, therefore, the
adverb input was unnatural. On the one hand, one might expect this fact to make the input
more salient to the students, assuming they recognize the awkwardness of it. On the other
hand, however, this overuse may result in students immediately 'tuning out' the input.
5.5.2. N 0 significant difference between groups

As discussed in the section "Unequal Variables," perhaps the two most confounding
variables in this study are 1) the groups differed with respect to previous years of French,
and 2) four different teachers participated in the study resulting in, among other things,
nonuniform adverb input across class sections.

Because the prior linguistic knowledge of both groups was not equivalent, and
specifically because the noninferenoing grottp possessotl more yoass of Fronoh, it is difficult
to separate the effect of treatment from the efi'ect of prior linguistic knowledge. In other
words, it cannot be asserted that there was no effect of inferencing input and that therefore
the groups performed similarly, however likely this assertion may be.

Any comparisons made between the performance of each group are fiirther

complicated by the fact that fo r diff ren h r i i ed in he . The most

 

27 Adverbs are not formed as easily in French as they are in English; thus, many English adverbs
must be expressed as adverbial phrases in French (Darbelnet, 1977). For example, urgently would be
expressed as de toute urgence; contentedly would be expressed as d ’un air satisfait.

183

obvious result of the fact that the variable of teacher was not held constant is the inconsistent
adverb input provided to students in each of the four sections. As discussed previously,
there is also the issue of the effect on students' performances of affective variables stemming
from such sources as teacher\student rapport, teaching style, and/or teaching methodology.

Finally, it must be pointed out that afier the treatment period, between Posttest 1 and
Posttest 2, all students were taught the rule of adverb placement. While the teachers reported
that they merely presented the rule and did not provide any follow-up exercises, it is still
difficult to determine the effects of this explicit instruction on Posttest 2. Even though it is
possible that learning the rule may account for the fact that there was no significant
difference between groups from Posttest I to Posttest 2, one is still lefi with the question of
why there was no significant difference between groups from the Pretest to Posttest 1. The
methodological shortcomings of the study mentioned previously may, in part, explain this
finding.
5.5.3. Predictors

As discussed above, the fact that the groups differed both initially and during the
treatment period (with respect to variables which should have been held constant), possibly
confounded the effects of the treatment which otherwise would have resulted in changes
within groups and/or differences between groups. The effects of these confounding variables
are clearly evident in the multiple regression analyses in which no clear and consistent
predictor was found across time and tasks.” One of the most problematic issues is the fact

that there was a significant negative correlation between age and previous years of French,

 

2“ Of course, as has been discussed, it is possible that different variables had different effects at
different times on different tasks.

184

thus making it difficult to separate these two variables and definitively attribute
change/improvement to one or the other variable. This naturally leads one to question if
other factors were 'hidden' in the defined variables.

Multiple regression analyses proved to be both an advantage and a disadvantage in
the present study. The power of this stotisticfl onolysis allowed me to discover predictor
variables in an extremely small sample size. Yet, it is just this power which is also a
disadvantage. For example, recall that gender was found as a significant predictor of change
on several tasks, and yet, there were only four males out of 45 subjects. Therefore, while it
may be tempting to do so, the findings obtained here cannot be generalized beyond this
study.

5.6. IMPLICATIONS AND SUGGESTIONS FOR FUTURE RESEARCH

In this section I first suggest alterations in methodology which may help to remedy
the shortcomings discussed in the above section and which would be beneficial to future
studies. I then note SLA issues, implicated in the findings of the present study, which
deserve fiirther attention and which future research may help to explicate.

This study would have benefitted (at most, in the findings themselves and at least,
in the strength of assertion of the current findings) had two preliminary procedures been
taken. First, it is absolutely essential to pretest subjects on the structure under investigation.
Had the subjects been pretested on adverb placement, I would have avoided the pitfall that
subjects were already relatively knowledgeable of the rule by choosing a lower level French
class. Even though French 150 is the level at which this grammatical structure is taught, the
students' had obviously covered this structure in their previous years of French study.

Second, background questionnaires should be administered to subjects before the onset of

185

the study. If this had been done, I would have discovered that 1) subjects possessed a
relatively large number of previous years of French; 2) the groups were not identical with
respect to previous years of French; and 3) there was a significant negative correlation
between age and previous years of French. Knowing these facts may not have altered the
present study given that 1) it was not possible for me to manipulate the groups in any way
(e.g., move a student fi'om one class section to another); 2) the n-size was so small that I was
hesitant to eliminate subjects (e.g., those with more years of French); and 3) there were only
four sections of French 150, and therefore, it was not possible to add more subjects to the
study. However, future studies which are not limited by such constraints would benefit from
the incorporation of a pre-study background questionnaire.

Five modifications to the study proper would have made for 'cleaner' research and
perhaps have resulted in clearer results.29 First, the sample size should be larger, which
would allow for greater generalizability of results and perhaps reveal differences both
between groups and between predictors that are not immediately evident in the present data
base. Second, the treatment period should be both longer and more concentrated.
Particularly with respect to inferencing input, it is unreasonable to suppose that eight class
sessions will have a significant efi‘ect on student performance. Third, all class sections
should be taught by a single teacher, which would control for the confounding variable(s)
associated with teaching style. Fourth, the tests should be administered to all students during
the scheduled class period, which would partially increase the randomness of the sample.

Finally, a class which receives no adverb input should be included in the study. This would

 

29 These modifications are a natural consequence of the caveats which have already been discussed.
Therefore, the implications of these modifications will not be elaborated on as they have already been
implicitly discussed.

186

perhaps allow one to determine that there indeed is no effect of adverb input treatment, as
the findings in the present study generally suggest.

I believe that the four observations made fiom this study (see "Observations") are of
use in that they highlight the need for the methodological recommendations mentioned
above and they have specific implications for firture research.

First, that there was no significant difference between groups and that there was an
apparent delayed effect of both adverb input and inferencing input emphasize the need for
more longitudinal studies in SLA. A long term study, with periodic testing, would perhaps
reveal the cumulative efiect of treatment in addition to allowing time for differences between
treatment and control groups to emerge. Such studies may provide a more detailed
understanding of the relationship between learner variables, type of input, and even the
length of time between input and effect.

Second, the fact that adverb input, as quantified here, did not have more of an
influence brings into question the effectiveness of the communicative approach, which is
dominant in many current foreign language textbooks. That those students who performed
more pair work were not predicted to show greater improvement lends support to recent
investigations (VanPatten & Cadiemo, 1993; Harley, 1993; Spada & Lightbown, 1993)
which are reexamining the important role of explicit instruction in SLA.

Third, the fact that different predictors were found at different times on different
tasks suggests that firrther investigations of task type are in order. While much recent work
(Crookes & Gass, 1993) has looked at the effects of task type on performance, and in turn
how that output affects acquisition, the findings of this study suggest that more research is

needed which investigates the relationship between the specific demands of the task and

187

specific learner variables such as prior linguistic knowledge and inductive ability which
respond to those demands.

Finally, the fact that Raven's Standard Progressive Matrices was a significant
predictor of the WOC pretest error score and the A] pretest correct score and that inferencing
input was a significant predictor of the overall improvement on both scores of the AJ Task
and on the error score of the WOC Task indicate that fiirther investigations into the role of
inductive reasoning in SLA is justified. The former finding supports, to some extent,
proposals made by Carroll (see Chapter 2, Footnote 8) that inductive reasoning plays a role
in foreign language aptitude.3o The latter finding supports, to some extent, proposals made
by Neufeld (see Chapter 2, Footnote 8) and those working within a learning strategy
paradigm (see Chapter 2) that inductive inferencing is a strategy which can be partially
developed through the instructional use of embedded and explicit exercises.” Further
research into the question of the roles of inductive reasoning and learning strategies in SLA
would benefit from studies which control for the variable of Raven's score (if one assumes
this to be a valid and reliable measure of inductive ability) and keep the positive and lack of
negative evidence constant across groups; learning strategy instruction would then be
provided to groups with exclusively low or high Raven's scores, and then the performance
of these various groups would be compared. For example, would a group with high Raven's
scores show improvement similar to a group with high Raven's scores which had received

learning strategy instruction? Would a group with low Raven's scores which had received

 

5“ Assuming all things being equal, those students with higher Raven’s scores have achieved a higher
level of proficiency with respect to adverb placement as measured by the pretest AJ and WOC Tasks.

5‘ Assuming that the embedded inferencing exercises indeed require the use of inductive reasoning
which is then utilized in the acquisition of adverb placement.

188

learning strategy instruction show improvement similar to a group with high Raven's scores
which had received the same instruction? The observations made with respect to Raven's
SPM and inferencing input indicate that such future research warrants attention.

The tentative findings of this study, taken together, underscore the conundrum of
what type of instruction at what point in study is most effective for a specific grammatical
structure. In other words, more issues have been raised than have been resolved. Future
research which draws on the merits of this study and corrects its defects may help to clarify
some of these issues.

5.7. CONCLUSION

The purpose of this dissertation has been to investigate the relationship between the
use of Indirect Negative Evidence (INE) and the acquisition of the ungrammaticality of
Subject-Adverb-Verb-Object (SAVO) word order in French by native speakers of English.
In order to investigate the use of INE, I argued that its use is analogous to the use of
inductive inferencing, and therefore, the use of INE could be investigated via an
investigation Of the use of inductive inferencing.

Manifested in the results of this study is the now well-known fact of the necessary
interrelatedness of the variables involved in SLA: input, tasks, prior linguistic knowledge,
grammatical structure, teaching style... Just as one of these variables cannot be examined
in isolation from the others, so too, the cognitive processes which both affect and are
affected by these variables must be considered. I have provided theoretical support for the
proposal that the use of INE or inductive inferencing is one such cognitive process. While
the results of the current study do not conclusively show that the use of INE is operative in

or facilitates the acquisition of the ungrammaticality of SAVO word order in French by

189

native speakers of English, they do suggest that the use of inductive inferencing plays a role
in SLA. Future research in this area is needed to determine more precisely the nature of that

role.

APPENDICES

APPENDIX A

BIOGRAPHICAL COVER SHEET

Number

 

Date

 

Teacher’s Name

 

Age
Gender
Nationality

What is your major?

 

How long have you been studying French?
years months

Have you ever lived in a French-speaking country?
yes no If yes, how long? years months

 

Do you know any languages other than English?
yes no If yes, which language(s)?

 

 

What language(s) are spoken in your home?

 

When you were growing up, did you speak any languages other than English?

Do you consider yourself bilingual? yes no

190

yes

 

DO

APPENDIX B

SUMMARY OF LESSONS FROM BONJOUR, CA VA?

October l-October 4:
"Le passage des saisons"

Vocabulaire:

Grammaire:

Functional:

Misc en
pratique:
Skills:

October 5-October 7:

"C’est la vie!"
Vocabulaire:

Grammaire:
Functional:
Misc en

pratique:

Skills:

(Rochester et al., 1991, pp. vi-ix)

Les saisons et le temps: Quel temps fait-il?
Qu’est-ce que je vais mettre aujourd’hui?
The groups mettre, partir, and venir
Summary of Information Questions

The passe compose with ctre

Getting information

Talking About the Past

Lecture: A has la mode et vive le look!
Prononcez-bienl

Mots-clés: Expressing when you did something
Face a face: Role play

Avant de lire: Anticipating content

Par ecrit

Functional minidialogue

Trois generations d’une meme famille
Lundi matin

Some Reflexive Pronominal Verbs
Reciprocal Reflexive Verbs

The passe compose of Pronominal Verbs
Expressing Some Everyday Actions
Describing Mutual Relations

Talking About the Past

Commentaire culture]: Les Francais et le travail

Situation: Visite a domicile

Functional minidialogue

Face 8 face: Role play

A propos: Pour consoler un(e) ami(e) malade ou malheureux(-uese); comment
féliciter quelqu’un

191

October 8-Octobcr 11:
"Faisons un voyage! "
Vocabulaire:

Grammaire:

Functional:

Misc en
pratique:
Skills:

October 12-October 14:

'Vivc les vacanccs!"
Vocabulaire:

Grammaire:

Functional:

Misc en
pratique:

Skills:

192

1c pars en voyage...

Or‘r vas-tu? Destinations

Attendre and Other Regular -rc Verbs
The Verbs voir and croirc
Prepositions with Geographical Names
Summary of passe compose
Expressing Location

Talking About the Past

Lecture: Les U.S.A. en jeans

Functional minidialogue

Prononccz-bicn!

Mots-c168: Putting events in chronological order
Face a face: Role play

Avant dc lire: Anticipating content

Par écrit

Que pcut-on fairc en vacanccs?

En route

The Group conduire

The imparfait

Use of the passe compose and the imparfait
Describing Past Circumstances

Narrating Past Events

Commentaire culturcl: Histoire et architecture

Situation: Une nuit a l’auberge de jeuncssc

Functional minidialogue

Mots-clés: Time expressions with the imth and the passe compose
Face a face: Role play

A propos: Comment chercher uric chambrc d’hOtel

APPENDIX C

ADVERB FLOOD EXERCISES‘

30 September: Give students the following homework: (c0pies enclosed)
Lisez le passage:

Nous venous dc Saint-Malo. Nous voulons vraiment visiter les Brats-Unis. En juin, nous obtenons
rapidement un visa au consultat américain ct nous partons joycusement pour les Etats-Unis. Nous tenons
absolumcnt a visiter New York, Washington, ct La Nouvelle-Orléans.

Maintenam, imagincz que c’cst vous qui partez en vacanccs en France. Reprencz 1c passage ct faitcs tous
les changements néccssaires. Commencez par "Je viens dc..."

R6pondez les questions suivantes aux phrases completes. Préparez-vous a cngager avec un(e) camaradc
dc classe une conversation baséc sur les questions.

1. Pars-tn souvent en voyage? Oh vas-tu d’habitude? Viens-tu d’achctcr dcs vetements ou d’autrcs
objets necessaircs pour tcs vacanccs? Qu’cst-cc que tu viens d’achctcr?

2. Sors-tu souvent pendant (during) 1c week-end ou restes-tu toujours a la maison? Sors-tu souvent
pendant 1a scmaine? Qu’est-cc que tu portes quand tu sors?

3. Fumes-tit? (Do you smoke?) Tes amis fument-ils? Est-cc que ton appaitement ou ta chambrc sent
(smells of) parfois la fumee? Deviens-tu completcment désagréablc si un ami (unc amie) fume
chez toi? Sens-tu vraimcm une difference si un ami (une amie) arrétc dc fumer? Devient-il (elle)
calme ou nerveux?

OCTOBER 1
1. Introduce seasons and weather, then»-
Page 152: Altered.
C. Climats. Posez les questions suivantes a plusieurs camarades.

1. Qucls sont les mois dc l’été? dc l’automne? dc l’hiver? du printcmps?

 

' While a native speaker of French reviewed all of these materials before they were
administered, those sentences containing soigneusement and rapidement in SVAO word order have
subsequently been judged as unacceptable. This certainly brings into question the ’authenticity’ of the
input. Future research would benefit from obtaining as many native speaker judgments as practicably
possible.

193

194

2. Quel temps fait-i1 d’habitude ici en (:15? en automne? en hiver? au printemps?
Est-cc que le climat est different dans ta ville d’origine? (hometown)
3. Qu’est-ce que tu aimcs toujours manger et boire quand...il fait tres chaud? il fait froid

et qu’il neigc? il fait beau ct frais? il pleut?

II. Introduce clothing, then---
Page 154: Altered.

B. Et vous. Posez les questions suivantes a un(e) camaradc.

l. Quels vétements trouves-tu d’habitudc confortablcs err 616? cu hiver? en automne? au
printemps?

2. Quels vetcmcnts mets-tu souvent pour sortir (go out) 1c samedi apres-midi? le samedi
soir? (Je mets....)

3. Qui dans la classe porte souvent les vétcmcnts les plus origirnrux? Decrivcz-lcs.

III. Introduce verbs with enclosed pictures.
Page 155: Altered.

f‘ _
11$
0 .1
\\\ “111“”
Then--

Pictures:

Ellcs sortent dans la rue.

Le bus part exactement a huit heures.
Il sert soigncusement 1e cafe.

Elle sent joycusement lc cafe.

ll dort encore.

Elle merit vigourcuscment a sa mere.

 

 

 

 

 

 

 

 

 

 

Page 156: Altered.

Exercises.
A. Bon voyage! Completez l’histoire avec une forme (au present) d’un des verbes de la colnne

au-dessous.

devenir dormir promcttre quittcr revenir sortir
Chaque matin nous 1 tranquillement jusqu’a huit heures. Apres un petit dejeuner
américain copieux, nous 2 rapidement l’hOtel pour Mcouvrir la ville. Tous les soirs,

nous 3 joycusement diner dans un nouveau restaurant. En un mois nous

 

195

4 completcmcnt américains. Mais hélas, c’est la fin du voyage ct nous
5 malhcureusement en France. Jc me 6 toujours dc
rctourner bicntOt en Amériquc.

 

Page 157: Altered.
B. Curiosité. Imaginez ce que ces personnes vicnnent de faire. Donnes trois possibilités pour
Chaque phrase.

Modele: Pierre rcntrc d’Afriquc.
Il vicnt dc visiter 1e Sénégal. Il vicnt dc fairc un safari. II vicnt d’achctcr une belle
sculpture.

Jennifer quittc rapidement Paris.

Jc sors soigncusement du magasin de sport.

Nous revenons toujours dc la montagne.

Jean-Jacques et Yvon revienncnt quelquefois dc la compagnc.

ewp~

IV. Go over homework from 30 Sept. (”Lisez 1c passage")

OCTOBER 4
1. Introduce information questions, then---
Page 160: Altered.
"A votre tour"
Maintenant posez uric question avec que ou qu’est-ce que.

1. 11 y a toujours des vestes chic ici.
2. Je regardc souvent les vetements d’automne.
3. Jacques ct Pierre cherchcnt attentivcment des Chemises dc Cardin.

Page 161: Altered.
B. En Hautc-Savoie. Vous interviewez une championne dc ski francaisc pour un magazine dc
sport. Voici ses réponscs. Posez les questions correspondantes.

Mots utiles: quand, pourquoi, de quoi, d’ou.

? Jc suis dc Megeve, un petit village dans les Alpes francaises.

? J’ai commence a fairc du ski a six ans.

? Eh bien, peut-etre parce que c’cst une tradition dc la region. Et bien
, parce que j’aimc ca!

? Pour m’cntrainer (to train), j’ai surtout besoin de calme—ct dc temps!

 

 

 

Pgwl"?

 

Mots utiles: a quoi, qui, qu’est-cc que, combien dc

 

5. ? Chaque jour, en hiver, je passe d’habitude trois ou quatre heurcs
sur les pistes (slopes).
6 ? A l’université, j’étudic la geologic. Un jour jc veux absolumcnt

 

devenir spécialiste dcs glaciers.
7. ? En etc, jc joue souvent au tennis et au squash.

8. ? C’cst Jean-Claude Killy qui a le plus (the most) influence mon style.
C’cst un ami de la famillc.

 

 

196

II. Dictée
Page 164: Before students do the written exercise!
Dictéc: B. Week-end en Suisse.

Bernard vicnt joycusement chercher Brigitte dc bonnc hcurc (early) pour aller prendrc 1c train.
11s y montent soigncusement ct ils cherchent rapidement lcurs places. Le train part quelques
minutes plus tard. Il entre lentement en garc (station) a Geneve a midi. Bernard ct Brigitte
descendent joycusement du train ct vont immediatcment a l’HOtel du Lac. L’apres-midi, ils
sortent visiter la ville. Le soir ils dinent dans un restaurant rustique, lc Relais dc Chambésy.
Dimanche Brigitte va au Musée d’Art ct d’Histoire. Elle prend aussi dcs photos dc la ville.
Bermrd rue tranmiillement a l’hOtel. Brigitte et Bernard quittent Geneve en fin d’apres-midi.
Ils arrivent a Paris fatigués mais contents de leur week—end.

III. Activity

Questions—Pair work. Have the students work in pairs asking and answering the following questions. As
a class, the teacher asks students about the student’s partner’s answers. (Review of question
formation/vocabulary)

Questions: (Adapted from Trahey & White, 1993)
1. Prepares-tn 1c dejeuner quelquefois?

Que prepares-tn d’habitudc?
Je prepare d’habitudc

 

2. Fais-tu souvent la vaissellc chez toi?
Qui fait d’habitudc la vaissellc chez toi?
fait(s) d’habitudc la vaissellc chez moi.

 

3. As-tu un animal domestique?
Qui 1c nourrit d’habitude?
le nourrit d’habitudc.

4. Fais-tu toujours tes devoirs?
Qui t’aidc a fairc tes devoirs quelquefois?
m’aide quelquefois a fairc mes devoirs.

 

5. Laves-tu jamais 1c planchcr?

De quoi tc sers-tu d’habitude pour laver 1e planchcr?

Je me 8ch d’habitude dc pour laver le plancher.
6. Ta mere te dit-elle souvent dc nettoyer ta chambrc?

Pourquoi oublies-tu quelquefois dc nettoyer ta chambrc?
Quelquefois, j’oublic de nettoyer ma chambrc parce que

7. Manges-tu jamais dc la glacc en etc?
Quel parfum de la glacc achetes-tu d’habitude?
J’achete d’habitudc dc la glacc

 

8. Manges-tu toujours tout ce qu’il y a dans ton assiettc?
Quellc nourriture nc manges-tu jamais?
Je ne mange jamais dc

197

9. Joues-tu jamais au tennis?
En hiver, a que! sport joues-tu?
En hiver, je joue d’habitude a

 

10. Ecoutes-tu jamais dc la musique?
Quel genre de musique n’ecoutes-tu jamais?
Jc n’écoutc jamais dc

 

OCTOBER 5
1. Introduce family names, then---
Page 173: Altered.

Pictures.

lundi matin

 

 

 

  

 

 

 

 

 

  

 

 

 

  
 

H. at Home Dalston-Oath Tun-Niche!

A. En famillc. Completes les phrases suivarnes avec le nom d’un membre dc ccttc famillc. Qui

 

 

 

fait quoi?

1. sc reveille lentement et son mari se leve.

2. Leur fils se douche rapidemem et se lave les cheveux, pendant que son petit
fiere se brosse soigncusement les dents.

3. Dans le couloir cormnence a se maquiller attentivement; son grand-pare
s’impatiente: il a besoin de se raser.

4. se peigne tranquillement.

5. s’habille tout seul.

6. Et et se couchent gentiment. Ils vont bientbt s’endormir. Ils n’ont

 

pasbesoindesortircematin.

II. Introduce reflexive pronominal verbs, then--
Pagc 176: Do ”A votre tour”-paits A and B as they are in the text.

A. Habitude matinales. Qui dans votre famille a les habitudes suivantes? Créez des phrases
completes avec les Elements donnes.

mon pere se rcgarder longtemps dans le miroir
ma mere sc lever du pied gauche

ma soeur se revcillcr toujours tres tOt

mon fiére s’habiller rapidement/lentement

mes parents sc maquiller/sc raser tres vite

198

se preparer a la dernicre minute

se laver les cheveux tous les jours

sc brosser les cheveux longucment

sc facher quand i1 y a quelqu’un dans la salle dc bains

'Q '0 ND °Q

Comparcz vos habitudes avec celles des membres de votre famillc ("Moi aussi, je..." or "Mais
moi, jc...").

Puis, comparez vos habitudes avec cefles dc vos camarades. Trouvcz quelqu’un qui...se levc dix
minutes avant dc partir; sc revcillc avam six heurcs; se leve souvent du pied gauche; se maquille,
sc rasc ou prend le petit déjcuner dans la voiturc...

B. Le dimanche. La famillc lament sc comporte un peu différemmcnt pendant 1e week-end.
Créez des phrases completes.

jc/pouvoir/se révciller/tard

Sylvie/8e regarder/longtcmps/dans/la glacc
nous/8e lever/dc bonnc humcur

Bernard ct Martin/s’habillcr/sans sc dépécher
Maman/ne pas se maquiller

Papa/ne pas se raser

je/s’cnnuyer

$9999.“?

Racontcz un pcu votre samedi ou votre dimanche. En quoi diffcre-t-il dcs autres jours de la
scmaine?

Page 177: Do "Entrc vous" A as it is in the text.
A. Interview. Interiogez un(e) camaradc sur une journee typique de sa vie a I’universite. Posez-
lui des questions avec les verbes se 'révelller, s’habiller, se dépécher, s’amuser, s’ennuyer, se
reposer, se promener et se couches. Ensuite, expliqucz a la classe les differences et les
ressemblances entre votre journee ct celle dc votre camaradc.

Page 178: Be sure to cover 17, reciprocal reflexive verbs, to the bottom of the page.

Vous quittez-vous bientOt? Oui...nous devons nous quitter juste avant Noel.
Mireille ct sa socur sc téléphonent régulieremcnt.

199

III. Cartoon. Can be done in pairs or individually. (Adapted from Trahey & White, 1993)

Organisez les images dans le bon ordre pour raconter une
histoire.

 

 

   

 

 

ll ouvre rapidement la “‘igre. ll resolute joycue-ent. Frank to reveal: tout I coup.

 

 

-- I-
e .1.
E . - F r]

\

 

 

 

 

 

 

 

[1 ac gauche. n enléve a robe dc clan-bra. n ouvrc undo-en! Is porn.

 

.1

.0
.0.

H

 

 

 

 

II entend [rapper
viola-neat.

[1 wit u chat m (labors. [I ouvrc soigncusement Ia poi-re

tT

 

 

 

 

11 net rapidement a robe dc chambrc.

 

   

 

 

 

 

n Geode “dense-cut I! u live lento-eat.

ll 43""! “hm“ “ll IL pet. the In pone.

200
OCTOBER 7
I. Verbs continued.
Page 179: ”A votre tour" Altered.

Une amitié sincere. Madame Chabot raconte l’amitic qui unit sa famillc a la famille Marnier.
Complétez son histoirc au present.

s’écrirc se rcncontrcr se telephoner

Gisele Marnier et moi, nous sommcs amies depuis plus de quinzc ans. Nous

 

 

1 tous les jours et nous parlons longtemps. Nous 2 souvent
en ville. Quand nous partons en voyage, nous 3 toujours dcs cartes posrales.
s’aimer sc disputer se parler se reconcilier sc rctrouver
Nos maris 4 aussi trés souvent ct les enfants 5 souvent surtout pendant
les vacances pour faire du sport. Les jeuncs 6 parfois, mais comme ils

7 bien, ils 8 vite.

 

 

11. Introduction of passe compose of pronominal verbs, then---

Page 181: Do "A votre tour" A and B as they are in the text.
A. le coup de foudie. Voici l’histoire d’amour de Pierre et de Sophie. Completez les phrases
suivantes avec un des verbes suggérés au passe compose.

se lever sc parler sc promener sc rcncontrcr se dire

s’cmbrasscr sc prendre sc regarder s’aimer se disputer

se marier sc quitter

Pierre et Sophie 1 chez des amis l’annec derniere. Le lendcmain matin ils 2
tres tOt pour sc téléphoncr. 11$ 3 longtemps. L’apres-midi ils 4 dans le parc.
D’abord ils 5 du coin dc l’oeil, puis ils 6 par la main. 118 7 dcs mots d’amour
ct ils 8 timidcment.

Apres, ils (ne...plus) 9. 11s 10 deux mois plus tard. Aujourd’hui, ils formcnt
lc couple parfait. 118 11 (present) toujours ct depuis qu’ils sont maries, ils
(nc. . .pas) (present). 12

Maintenant, avec un(e) camaradc, formulez uric ou deux autres conclusions possibles.

B. Souvenirs. Aline rctrouvc un vieil album dc photos. Racontez son histoirc au passe compose.
Aline s’installe pour regarder son album dc photos. Elle s’arrcte a la premiere page. Elle se
rappefle son premier amour...mais ellc ne se rappelle pas son nom. Elle oublie aussi 1e jour de
leur premiere rencontre. Elle sc demande 00 il est aujourd’hui. Elle se sent soudain trcs fatiguee
ct ellc s’endort sur la page ouvcrte.

A votre avis, pourquoi Aline a-t-cllc un souvenir si imprecis de son premier amour?

Page 182: Do ”Entrc vous" A and B as they are in the text.

A. Trouvcz quelqu’un qui...lnterviewcz quelques camarades dc classe ct trouvez quelqu’un qui...

201

s’est couche(e) tard hier soir

s’est reveille(e) pendant la nuit

s’est disput6(e) avec un ami recemment
s’cst amuse(e) la semainc derniere
s’est endormi(e) en classe cc semestrc

9‘95”!”7‘

B. Mon emploi du temps. Diviscz la classe en petits groupcs pour decrire une journee typique
de votre vie. la vie dc votre camaradc 11 ’est pas commc la v0tre. IllElle vous interrompt pour
vous cxpliquer comment il/clle a past sa journee.
Modele: A: Moi, je me leve toujours 11 sept heures moins le quart.
B: Cc matin, jc me suis leve(e) a dix heures et demie.

Page 183: Do ”Commentaire Culture!"
Les Francais et le travail

La mentalite des Francais vis-a-vis du travail a beaucoup change depuis quelques annees.
Beaucoup d’entre eux considerent la qualité de la vie plus importante que la reussite matérielle. Ils
préférent travailler moins meme s’ils doivent gagner moins.

De nombreuses reformes sociales out etc mises en place pour rcpondre 8 cc changement dc
mentalite: les salaries francais beneficient dc cinq semaincs dc vacances par an, ct beaucoup travaillent
seulcment acute-cinq heures par semainc. De plus, 1c travail a mi-temps et les horaires flexibles sont ties
populaires, surtout parmi les femmes qui pcuvent ainsi se consacrer un peu plus a leur famillc. C’est 1e
cas de cette jeune mere qui préfbre renoncer a une partic de son salairc pour passer les mercredis, jours
dc congé scolairc, avec son enfant.

Ce refus dc l’aspect alienant du travail revele l’importance qu’on donne, en France, a la qualité
de la vie. lie bonheur pour beaucoup de Francais, c’est la r6alisation dc soi. 0n aime prendre le temps
de vivre. Mais ccttc nouvellc conception du travail n’est pas obligatoirement synonyme d’improductivite.
Eliminer le stress, c’est améliorer la qualité du travail. On retrouve alors un rythme plus nature! et efficacc
qui pennet d’avoir un mcilleur equilibre personnel dans son travail. C’est 1e rejet de la routine. Ainsi 1e
Francais part a la reconquete du temps!

III. Dictee (Adapted from Muyskens, Omaggio Hadley, & Convert-Chalmers, 1990, p. 348).
Marie veut absolumcnt devenir medecin. Elle travaille constamment. Le matin, ellc arrive
d’habitude a l’hOpital a six heures et ellc y reste south jusqu’a neuf heures du soir. Dans la
journee ellc travaille beaucoup ct prend seulement quinze minutes pour déjeuner. Souvent, elle
est fatiguee le soir. Mais, je pensc qu’elle va reussir parce qu’elle est tres travailleuse ct
ambitieuse.

OCTOBER 8
1. Introduce "Travel" then-
Page 191: Pictures. Altered.

A. Dans la rue: attendre and Other Regular 45 Verbs

 

 

 

 

 

202

[Is attendcnt paticmment lc bus.

Elle entend soudainement un bruit.

Il perd toujours ses clés.

Elle vend facilcment des fleurs.

Elle rend joycusement la monnaie au client.

11. Activity: After introducing verbs and geographical prepositions,
"Memory Tricks" (Adapted from Rinvolucri, 1984)

After covering geographical prepositions. (pp. 194-197): (practice in Listening comprehension, speaking,
with simple past, prepositions of geography)

1. Send four of your better students out of the room with an exercise they can do outside in about
15 minutes.

2. Dictate a story to the others. They should not write anything down!

3. Make sure they understand all the words in the story.

4. Explain that one student in the group is to prepare to retell the story to one of the four who have

gone out. This one student can ask two questions about the story after it has been told to him/her.
The second ’outsidc’ student will be brought back in, and the first ’outside’ student will tell
him/her the story. During this telling the class is to note down how the story gets changed using
these three categories:

Things added Things left out Thin s chan
5. Each ’outside’ student is brought into the class, and the previous ’outside’ student tells him/her

the story and he/shc can ask two questions. During this time, the class is taking notes regarding
the above categories.

6. After the four students have returned, one student reads the original story.

7 Categories are put on the board and class lists information in each category.

Story:

Sabine est étudiante a Rouen. Elle veut absolumcnt faire un voyage en Asie. Elle y pensc
comtamment. Alors, ellc a decide tout a coup dc partir en vacances. Elle prend joycusement son billet
d’avion. Elle fait rapidement ses valises. Elle attend son depart avec impatience. Et puis, le jour arrive.
Son vol part exactement a six heures. Elle ne manque pas lc vol. Mais....malheureusement, elle est
montee dans le mauvais avion! la destimtion dc cet avion est en Amérique du Nord. Aimi, ellc est restee
cn transit au Mexiquc quelques heures. Elle va arriver en Californie lc soir. Elle vcut vraiment s’amuser.
Donc, ellc va passer la scmaine a las Vegas dans le Nevada et puis ellc va probablcment aller a La
Nouvelle-Orléans.

OCTOBER 11
1. After covering ’croiie’ (p. 192), as a pre-imparfait exercise, give the students the following exercise:
(enclosed)

"En cours de socilogie" (Rochester et al., 1990, p. 345).

Sabine: Je pense qu’en France, l’idéal du succes c’est toujours la ’performance culturelle.’

Le Professcur: Amlysez rapidement pour nous les rapports entre cet ideal dc 1a reussite et le systemc
scolaire francais.

Sabine: On peut dire que c’est a cause de cet ideal que l’enseignement fiancais est principalement
centre sur le diplOme.

JérOme: Le dipldme, en France, est un moyen d’ascension socialc; les Francais croient beaucoup

a unc elite dipIOmee.

203

Le Professeur: Dites-moi maintenant si vous pensez que cette attitude diflEre sensiblement dans les autres
pays. Aux Etats-Unis, par exemple...

1. Pour Sabine, est-ce que l’ideal du succes en France est parfois la ‘performance culturelle’?

2. Le professeur demande-t-il a Sabine d’analyser en detail les rapports entre cet ideal et le systeme
scolaire?

3. Est-ce que JérOme pense que les Francais croient peu a une elite dipldmée?

II. "Vente aux encheres” (Adapted from Rinvolucri, 1984)

Review of vggabulg, gmgraphicgg prepositions, passe compose.

1. Ask students if any of them have been to an auction. Ask questions of those who have. Introduce
necessary words like to bid, auctioneer, a bid, What am I bid? Going, going, gone! (une fois, deux fois,
adjugé, vendu!)

2. Pair the students off and give each pair an auction sheet. Tell them that some of the sentences on the
sheet are correct and some incorrect. They are to read through and decide which sentences are correct and
which incorrect. In the auction that is to follow they are going to have to bide for sentences, the aim being
to buy only correct sentences. Tell them each pair has F5,000 for buying sentences and ask them to note
down in the budget column of their auction sheets how much they are willing to bid for a given sentence.
They may not spend more than F5,000 in the auction. The winners of the auction are the pair with the
most correct sentences an_d the most money left.

As the students work on the sentences and discuss their budgeting in pairs deny them all language help-it
is up to them to take responsibility for what they think they know. They may NOT use their books.

3. Before starting the auction tell them you will not accept bids of less than F200.

START THE AUCTION:

a. Read out the first sentence in a lively, persuasive way, even if it happens to be wrong, and ask for bids.
b. Keep the bidding moving fast, keep up a fast pace.

c. When you come to the ’going, going, gone’ stage, be ready to accept last minute bids. When a
sentence has been auctioned off make sure students keep a note of the buyer and the amount in their bought
columns.

(I. Start the auction with the first sentence on the auction sheet but then auction the rest of the sentences
in random ordernthis heightens the feeling of expectancy.

4. After each sale tell the group if the sentence is correct or not and in the latter case give the correct
version. Do this fastudon’t spend time during the ’auction’ going over grammar points.

5. When the auction is over, go over the sentences the students did not understand and the incorrect
sentences which the students thought were correct.

*1. Les Yamamoto retoument joycusement en Japan.

*2. Pour voyager en TGV hier, je reserves une place a l’avance.
3. Quand revoyez-vous vos amis? Nous nous voyons souvent.
*4. La vol n° 2 part immediatement a destination a Bruxelles.

5. Je veux vraiment aller a cet endroit parce qu’il y fait du soleil.
*6. Elle n’est pas descendu.

*7. Elle croit absolumcnt a cette route est bonne.

204

8. A dix heures precises, 1e train s’est mis a rouler.
9. Nous avons rendu visite a notre tante.
*10. Elle s’est souvenu de quelque chose.
11. D’abord, elle a voyage en vélo. Puis, elle a pris la voiture.
12. Le train entre lentement en gare.
13. Sabine vient d’Allemagne.
14. Geneve est en Suisse.
15. J ’habite tranquillement a Montreal.

111. Homework--”Le cas de l’auto-stoppeur affamé" (Adapted from Trahey & White, 1993)

To the teacher:
Give each student a copy of the Encyclopedia Brown story ”The Case of the Hungry Hitchhiker. " Tell
the students who Encyclopedia Brown is according to the following description.

Encyclopedic Brun habite a Idaville. II a dix ans et i1 aime resoudre les mysteres. Le pere d’Encyclopedie
est le Commissaire de police d’ldaville. Tous les gens pensent que la police d’ldaville est la meilleure du
monde. Mais, le Commissaire Brun sait que c’est Encyclopedic qui l’aide a resoudre les problemes les
plus difficiles. Ils discutent des problemes a table, et souvent Encyclopedic les résoud immediatement.
Mais, personne d’autre ne 1e sait pas. Qui va le croirc?

"Le cas de l’auto-stoppeur affamé"

L’apres-midi, Encyclopedic Brun allait souvent a la peche avec ses amis. Les soirs 00 ii faisait tres chaud,
le Commissaire Brun se rendait quelquefois en voiture au fleuve et il a reconduisait Emyclopédie chez eux.
"C’est magnifique,” a dit joycusement Encyclopedie, un soir comme i1 montait dans la voiture climatisee.
"Il fait tres chaud.” ‘

”II fait trente-cinq degrés,” a dit son pere.

Tout a coup, une voix s’est fait bruyamment entendre a la radio de police.

11 y avait eu un hold-up a la Banque Royale i] y a dix minutes. Les quatre voleurs se sont échappés dans
une voiture bleue. Ils se sont rapidement dirigés vers le nord.

Le pére d’Encyclopedie a radiotélephoné rapidement au poste de police.

”C’est de la part du Commissaire Brun," a-t-il dit soigncusement. 'J’irai au grand chemin. Faites sortir
les voitures de reconnaissance et appelez immediatement les autres postes. " Le Commissaire Brun est
retoumé rapidement a la voiture.

Encyclopédie a regarde son pere silencieusement. Il souriait joycusement. Il n’avait jamais donné la
chasse aux voleurs.

"Nous ne les venous pas,” a dit lentement 1e Commissaire Brun. "Ils vont trop vite. J’éspere que quelqu’un
verra le chemin qu’ils ont pris.”

be Commissaire Brun a mule prudemment sur le grand chemin. Encyclopedic a vu un auto-stoppeur qui
attendait avec patience que quelqu’un le prenne. C’était un jeune homme qui portait un grand sac.

”Si la voiture passait par ici, cet auto-stoppeur la verrait,” a crié Encyclopédie.

"Peut—étre," a repondu 1e Commissaire Brun. "Cela depend combien du temps il est la.”

Le Commissaire Brun s’est arrete a cOté de l’auto-stoppeur.

”Depuis combien de temps attendez-vous?"

"Depuis une heure,” a répondu soigneusement l’homme.

”Une voiture bleue est-ellc passée rapidement ici?”

'Oui,” a repondu l’auto-stoppeur a haut voix. "Elle a pris cette route la. Ils conduisaient rapidement.
'Montez," a dit rapidement le Commissaire Brun.

L’homme a regardé silencieusement l’uniforme du Commissaire Brun. ”Pourquoi?” l’homme a demande
timidcment. "Voyager en auto-stop, c’est contre le loi?"

"Non, soyez tranquille," a dit le Commissaire Brun. ”Si vous voyiez la voiture, la reconnaitriez-vous?”
"Oui," a dit l’homme et il est rapidement monté dans la voiture. II a ouvert son sac lentement.

205

'Voudriez-vous une orange?" a-t-il demande avec bonte a Encyclopedie. "Ou du chocolat?"

"Du chocolat, s’il vous plait,” a repondu Encyc10pédie.

Encyclopédie a pris 1e chocolat. 1] en a pris deux morceaux. Le chocolat etait tres dur. II a rendu le
reste.

L’auto-stoppeur a mis le chocolat dans son sac. Lentement, l’homme a commence a manger son orange.
11 a soigncusement mis la pelure dans un sac en papier.

Lentement Encyclopedic a mange le chocolat. Tout a coup, il a eu tres peur. II a pris un crayon dans sa
poche, mais il n’avait pas de papier.

"Est-ce que je peux avoir encore du chocolat?" a demande Encyclopedic timidement.

L’auto-stoppeur riait doucement et a donné son chocolat a Encyclopédie.

Encyclopedic a rapidement mange le chocolat. Lentement et soigncusement, il a écrit sur le papier:
”l’auto-stoppeur est le voleur.” Puis, i] a silencieusement mis le papier a cdte de son pere. Le
Commissaire Brun a rapidement regardé le papier eta continue a conduire. Enfin, ils sont arrives au poste
de police. Lentement, le Commissaire Brun a ouvert la porte de la voiture. Tout a coup, i1 s’est mis a
c0té de la porte de l’auto-stoppeur. 11 a soigncusement braqué son canon sur l’homme.

"Nous n’avom pas retrouvé vos amis, mais nous vous avons attrapé,” a-t-il dit avec colere. "Vous etes
en état d’arrestation. "

Comment est-ce que E. a su que l’auto-stoppeur etait un voleur?

Solution:

L’auto-stoppeur a fait une faute quand il a donne le chocolat a Encyclopedic. Le chocolat etait ferme.
Mais l’homme a dit qu’il attendait prés du chemin depuis une heure par une chaleur de trente-cinq degrés.
Le chocolat aurait du completement fondre a cause de cette chaleur. ll aurait du étre tres mou. L’auto-
stoppeur a menti parce que c’était le voleur qui etait censé attirer l’attention de la police et la mener sur
une fausse piste. L’homme a explique qu’il avait gardé le sac dans la fraicheur de la voiture pour filer
jusqu’a ce que les autres soient partis.

Les autres voleurs ont été arrétés rapidement.

(Adapted from: Sobol, 1978).

OCTOBER 12
In groups have students try to find solution to ”Le cas de l’auto-stoppeur affamé"

1. Introduce "Les vacances” with--
Page 205:
Do B "Mes vacances" with following alterations:

Mes vacances. Posez les questions suivantes a un(e) camarade et dites-lui vos preferences.

1. Oh préféres-tu d’habitude passer tes vacanccs? a la mer? a la ontagne? a la campagne? dans
le desert? en ville? a la maison? a l’étranger? Pourquoi?

2. Qu’est-ce que tu aimes souvent faire pendant les vacanccs? des voyages? du sport? un stage
(short course) de yoga, de photographic...? autre chose? Qu’as-tu fait pendant tes dernieres
vacanccs?

H. Introduce ’conduire’ verbs, then-

Page 207: Do exercise at bottom of page in text.

A. Séjour a Geneve. Complétez chaque phrase avec la forme de conduire, produire, réduire
ou traduire qui convient.

206

Dans le cours de langue pour etrangers, nous ne jamais.

Si vous prenez le TGV, vous radicalement le temps du voyage.
En ville, les automobilistes ne pas toujours avec prudence.

Au Palais des Nations, les interprétes tous les discours des diplomates.
En Suisse, on aussi de tres bons vins.

9:59P!"

Page 208: Do 'Nostalgie"
Nostalgic

C’est le premier aout. Beaucoup de Francais par-tent en vacances. Pierre et Louise Dupoirier vont a la

montagne comme chaque ete. Leur voiture, chargée de valises, est prise dans un embouteillage.

Pierre: Quand j’étais enfant. tu sais, les vacances, c’était vraiment le paradis.

Louise: Tu allais toujours en province, n’est-ce pas?

Pierre: Oui, nous passions nos vacances sur la ccte normande, chez ma grand-tame. Aujourd’hui, il me
semble qu’il y faisait toujours un temps splendide, que les journees etaient longues...

Louise: Et que faisais-tu la-bas?

Pierre: En fait, mes peu de chm. On jouait, on se baigmit dans la mer, on s’occupait un peu du jardin.

Louise: Mais regardc un peu cet embouteillage! Nous allons passer nos vacances sur l’autoroute.

Repondez aux questions suivantes.

1. Oh est-cc que Pierre passait autrefois ses vacances?

2. Qu’est-ce qu’on y faisait?

3. Comment les Dupoirier passent-ils leurs vacances maintenant?

III. Introduce imparfait, then—-

Page 210: ”A votre tour” do A and B as they are in the text.

A. Une ville de Bourgogne. Quand Herve habitait a Dijon, il sortait regulierement avec ses amis.
Decrivez ses activites en mettant tous les verbes du paragraphe suivant a l’imparfait.

A Dijon, nous sommes une bande de sept amis. Nous nous voyons presque tous les jours. Le
samedi, nous allons a bicyclette pique-niquer a la campagne. Quelquefois on fait des achats ou on se
promene dans la Vieille Ville. Nous aimons aussi nous promener sur le quai.

Le soir, quelqu’un choisit un film. On va aussi souvent a la disco, mais moi je préfbre écouter
de la musique classique. On rentre se coucher tard.

Nommez trois ou quatre des activités d’Herve et de ses amis.

B. Creez une ambiance. Imaginez que vous etes romancier (-iere) et que vous commence; un nouveau
livre. Voici la scene que vous imagine: ct... votre premier paragraphe:

   

 
  
 

noun [BEE
nunu all

[I [I E

B
B

   

 
        
  
 

  
 

  
        

207

'Il est huit heures du matin. De la fenétre de mon hOtel, je vois la place Daudet. Quelques personnes
attendent l’autobus. Sur la terrasse du cafe, les garcons servent du café et des croissants. Les trottoirs
commencent a se remplir de touristes. I] fait beau et pas trop chaud. Je me sens tranquille. "

Mais non! Vous n’étes pas satisfait(e). Recommencez. D’abord, mettez le paragraphe a l’imparfait: "ll
était...”

Mais vous n’étes toujours pas satisfait(e). Essayez encore une fois. Toujours a l’imparfait, creez une
atmosphere sombre et mystérieuse. Commencez par: '11 etait onze heures du soir...” Finalement,
presentez votre debut de roman a vos camarades. Demandez-leur d’en faire la critique.

IV. Introduce imparfait vs. passe compose, then--
Page 214:

"A votre tour" Altered.

Do A with following revisions:

Une soirée mouvementée (eventful). Vous étiez Chez vous hier soir. Vous vouliez faire certaines
choses...mais il y a eu toutes sortes d’interruptions! Suivez le modele.

étudier silencieusement / 1e francais...teléphone / sonner
J ’étudiais silencieusement le francais quand (lorsque) le télephone a sonné.

1. preparer soigneusement/ lecon de maths....un ami/ telephoner
2. attendre poliment/ copain....mes parents/arriver
3. aller sortir...quelqu’un (someone)/ sonner a la porte
4. lire silencieusement/ joumal...la propriétaire/ venir demander/ argent du loyer (rent)
5. ctre en train de regarder/ tele...ma camarade/ mettre/ radio
V. Vocabulary review/recognition and listening comprehension-

For this activity, the class should be divided into two teams. You have a set of numbered index cards.
On each card is a scrambled word. You also have a list of numbered sentences which will serve as clues
that only you may see. Show each team one card. Then read out the sentence clue that corresponds to
that word. The team will have one minute (or less) to unscramble it, the other team gets a try and the
point if they succeed. Then the other team gets shown a word, and so on. The team that reaches 30 first
wins. (Adapted from Trahey & White, 1993)

Quel est le mot????

Quand il fait du soleil. nous allons souvent a la plage.

J ’envoie toujours un cadeau :1 mon ami pour son anniversaire.
Quand je suis fatigue, je vais d’habitude au m.

Nous achetons souvent du l’essence pour notre voiture.

En hiver, je joue quelquefois W avec mes amis.
Quand je suis presse, j’oublie quelquefois mes clés.

Quand j’entends une farce, je IiS toujours.

Je mange d’habitude trois LCM par jour.

Au printemps, mon pere plante toujours des fleurs dans le 1m.
J’étude toujours beaucoup quand j’ai un examen.

Je prends souvent une orange avec le déjeuner.

""".‘°.°°.".°‘!~":“S*’.Nt"‘

I-‘O

12.
13.
14.
15.
l6.
l7.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.

208

J’écris d’habitude avec une 2111.91:-

A Noel, ma famille va toujours rendre visite a ma m.

Quand je suis malade, je vais d’habitude chez le mgggin.

Le matin, je me lave toujours les chevegg.

Quand il n’y a pas de chaises, nous nous asseyons quelquefois par terre.
A la ferme, je monte d’habitude a cheval.

Je mange toujours des céréales avec du 1a_it_.

Les gens se rendent quelquefois au travail en m,

Quand je vais a la plage, je prends d‘habitude une serviette.

Je me brosse toujours les m apres avoir mange.

Nous nous asseyons d’habitude sur une chaise.

Quand il fait beau, nous ouvrons toujours la fgnége.

Quand je vais a la plage, la gm entre toujours dans mes chaussures.
Je prends souvent du pain 1e matin.

Le professeur écrit toujours avec de la me au tableau noir.

Quand nous faisons du camping, nous cuisinons souvent au ffl de bois.
J ’aide quelquefois ma mere a nettoyer la maison.

Une i_lg est toujours entourée d’eau.

 

 

APPENDIX D

INFERENCING EXERCISES

OCTOBER 1
I. Teacher-centered instruction/explanation. Explicit inferencing exercise (Carton, 1966, pp. 82-86).

angxt Cues-meaning of words can frequently be found in the context of a reading passage.
Write on the board the nonsense sentence: "Le barratineur a pavoisé la baraque avec des liquettes. "
Discuss questions to facilitate ’sensible guessing’ of unknown words:

A. Is it a noun?
1. Grammatical questions
a. Is it singular or plural?
b. What adjective modifies it, if any?
c. What gender is it?
(I. Is it the subject or object of a verb?
2. Pragmatic questions
a. Does it refer to person, things, conditions, events, qualities?
b. How many?
c. What kind?

B. Is it a verb?
1. Grammatical questions
a. Does it show future, present, or past?
b. is it subjunctive?
c. What is its subject?
2.Pragmatic questions
a. Does it refer to an action, change, condition, relationship?
b. When?
c. Do people do it?
(I. If not, what kind of thing does it?

C. Is it an adjective or adverb?
1. Grammatical questions
a. Which word in the sentence does it modify, or apply to?
2. Pragmatic questions
a. What is it describing in this sentence?
b. Does it describe some fact, or does it indicate somebody’s attitude?
c. What kind of fact or attitude? (physical attributes, manner, judgment)

209

210

”Le barratineur a pavoise la baraque avec des liquettes."
Request information shown by sentence on parts of speech, ending, role, and relationships.

a. barratineur: -eur, noun marker. ’le’: masculine agent

b. a pavoisé: verb, passe compose, transitive

c. la baraque: object of verb, marked by ’la’, therefore feminine noun

d. liquettes: marked by ’des’, therefore plural, -ette, feminine ending, something used as object of
prepositional phrase

The Operation of Context

Definition of a word may be signaled by the use of a form of ’étre’.
Definitions may also be signaled by the use of EXAMPLES which may be signaled by words such as:

comme autre

surtout ca, ce, cet(te)

ou ..... ou ces
Examples:

étre: Mona Lisa est up tableau de da Vinci. If you don’t know the meaning of ’un tableau’, you can
infer it from knowing what the Mona Lisa is.

ca: Mais ce qui est arrive hier, devant mes yeux, je dois dire que ca dgpgse tout. If you don’t know
the meaning of ’dépasser’, you can infer it by refering back to ’ce qui est arrivé’.

autre: Racontez un peu votre samedi ou votre dimanche. En quoi différe-t-il des autres 1% de la
semaine. If you don’t know the meaning of ’samedi’ or ’dimanche’, you can infer the meanings
from ’jour’.

Notice that by knowing the meaning of one term allows you to infer the meaning of the other term.

Meaning may also be signaled by RESTATEMENT in the text. This may be signaled by words such as
on or by introductory phrases such as:

en d’autres mots ce que ceci veut dire
c’est-a-dire
ce qui signifie
Example:
En gros, la France a la forme d’gm hexagone, c’est-a-dire d’gg figure gmmgtrigue a six gg‘tés.

Where relationships in the text are not explicit but must be INFERRED, connecting words such as the
following will help guide you in determining the relationship:

cependant pourtant
donc alors
de la méme facon

Examples:
Je n’aime pas danser, donc je ne danse pas ce week-end.
Ma cousine étudie l’espagnol, mais j’étudie le francais.

211

OCTOBER 4
Exercise Added: .
Review of verbs. Embedded inference exercise: an application of Carton’s method.

Write the following sentences on the board and ask the questions.

Tu es parti. Is ’tu’ masculine or feminine? How do you know? (No ’e’)
Vous etes partis. masculine or fem? How do you know?

Nous sommes parties. masc or fem? How do you know?

Ils sont partis. Is this correct? How do you know?

Exercise Added.
Page 165-167: Mise en pratique. Explicit Inference Exercise from text.

Lecture: A bas la mode et vive le look!

“Avant de lire"

Anticipating content. Test illustrations, titles, and lead lines or highlighted material are valuable aids in
reading. Before starting to read a passage, you can examine them for clues about the content of the
accompanying text. Clues serve as advance organizers to the reading. In-text illustrations and captions
can also clarify meaning during the reading process itself, as visual representations can convey specific and
general, explicit and implicit information. Readers who start with a working hypothesis when confronted
with an unfamiliar text seem to read more quickly and with greater understanding.

Le concept tradionnel de la mode n’intéresse pas beaucoup les jeuncs Francais d’aujourd’hui:
l’important pour eux, c’est "le look." Quellc est la difference? Le look exprime des idées personnelles,
c’est un commentaire existentiel, politique, une prise de position. Les ancetres du look, ce sont les
Hippies. Dans les années 60, leurs vétements (chemise indienne et cheveux longs) et leurs styles de vie
(communes) communiquaient un message de paix et de retour a la nature.

Comme eux, aujourd'hui, les jeunes s’habillent selon leurs opinions: extréme-droite, hyper-
gauchiste, BCBG, Punks, New-wave, Babas, Rockers, Silicon Valley, Smurfs, Nouveaux-riches...Chacune
de ces ”tribus" de jeuncs a aussi son propre territoire, comme par exemple a Paris, les BCBG pres de
l’Ecole militaire, les Mods dans le Marais, les Punks dans le vingtieme arrondissement et les Rockers dans
la banlieue nord. Ils ont aussi adopte des idoles, un langage et des symboles particuliers. En voici
quelques exemples. (Pictures provided in text).

Les cools

Le look: jeans, tee-shirt Fruit of the Loom, les badges

Les idées: americanophiles

Les passions: les Etats-Unis, le Coca-Cola, les jeeps, le rock américain

Les BCBG (Bon Chic, Bon Genre)

Le look: britannique, le loden vert, les mocassins, un foulard Hermes
Les idées: bourgeoises, droite libérale

Les passions: Lady Di, les grands couturiers, le polo, le golf, le tennis

Les punks

Le look: blouson de cuir noir, cheveux en créte, lunettes noires
Les idees: nihilistes, anticonformistes

Les passions: la laideur calculee, le noir, le rock, les skates

Le genre "boyfriend"

Le look: vétements d’homme, manches et pantalons retroussés, bretelles, cravate, accessoires mode
Les idees: vivre pour le moment, 1e standing

Les passions: la bande et les lieux de rencontre "in”

212

Page 167: Exercise Added.

Avez-vous bien compris? Corrigez les phrase inexactes.

Les jeunes Francais adorent vraiment la mode tradionnelle.

La mode chic exprime souvent des idees personnelles.

Les ancétres du look sont Yves Saint-Laurent et Pierre Cardin.

La jeunesse francaise est unifiée.

Le look des BCBG est américain.

Les partisans du genre ”boyfriend" portent toujours des tee-shirts Fruit of the Loom.
Les cools sont nihilistes.

$999959?"

Exercise Added. Vocabulary Review. Inference exercise.

Which words do not belong? Why?

beau partir

chaud monter

froid rentrer

vent dormer

pleut rester
pleut; all others ’il fait...’ donner; all others conjugated with étre

la robe

1e chandail

la boutique

le tailleur

le costume

la boutique; all others articles of clothing

dormir quitter
partir mourir
sortir mettre
quitter prendre
dormer
dormir; all others mean ’to leave‘ ’mourir’; all others must take d.o.
OCTOBER 5

Do Inferencing exercise I. For all of these exercises (I - V), ask students to try to answer the French
exercises in French. If they can’t, it’s ok to use English. Be sure to go over the answers with them.

Exercise in Inference.
I. Ci-dessous voici une liste de mots arranges deux par deux. Pour chaque paire, trouvez la relation qui
existe entre les deux mots. (Adapted from Sternberg 1988: 119).

Mn entre les deg mots

manche: chemise

bleu: peu

crayon: écrire

fourchette: couteau

nombre naturel: nombre compte
gigantesque: énorme

les: sel

femme: homme

livre: page

 

 

 

 

 

 

 

 

99°35’99pr

 

213

10. resident: non-resident

11. mal: pis

12. chaise: meubles

13. canadien: francais

14. véhicule: autobus

15. professeur: enseignement

 

 

 

 

 

 

Exercise Added. Vocabulary review. Exercise in Inference.
Write following lists on board. Ask students which words go together. Why?

se reveiller le bureau

se laver tard

se preparer les cheveux

travailler a
OCTOBER 7

Do Inference exercise H. Exercise in Mapping.

II. Ci-dessous voici quatorze ensembles de mots arranges deux par deux. Essayez de trouver la relation
qui existe entre les deux listes. Six ensembles sont en francais et huit sont en anglais. (Adapted from
Sternberg 1988: 131-132).

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Premier Rapport Deuxieme Rapport Ra les rts
1. midi: douze minuit: douze

2. le plus court: mai le plus long: septembre
3. violet: bleu rose: rouge

4. Afrique: continent Zaire: pays

5. balle: jeter gant: porter

6. cafe: tasse vin: verre

7. French: poodle German: shepherd

8. Mom: Dad 383:121

9. Seagull: flock wolf: pack

10. warts: straw lived: devil

11. socks: pair eggs: dozen

12. freshman: youngest senior: oldest

l3. cabbage: cole slaw potatoes: french fries
14. trip: strip lime: slime

Exercise Added. Embedded Inference.
Which word does not belong? Why?
l’oeil
1’oreille
la téte
l’épaule
l’oeil is the only masculine term. However, l’épaule is the only body part below the neck. Both
generalisations are correct.

Exercise Added. Application of Carton’s Method of Inference.
Write the following sentences on the board and ask the questions.

214

1. _ s’est levé a six heures.

2. _ s’est regardé dans le miroir.

3. Je me leve lentement a six heures.

4. _ se leve rapidement a neuf heures.

Are the subjects masculine or feminine? l and 2 have to be masculine-mo ’e’ on the past participle. 3
and 4, it’s impossible to know.

OCTOBER 8
Do Inference exercise III. Analogies.

III. Complétez l’analogie. Elles sont en anglais et en fiancais. Donnez la raison de votre choix.
(Adapted from Sternberg 1988: 135).

l. journal: papier comme chaussure:

a. pair b. chaussette c. cuir d. pied
2. parent: pere comme enfant:

a. oncle b. tante c. fils d. frere
3. 7/4: 175% comme 3/25:

a. 3% b. 12% c. 25/3 d. 4/7
4. Rodin: Le Penseur comme Da Vinci:

a. l’art b. l’Italie c. Mona Lisa (I. LaRenaissance
5. farine: pain comme laine:

a. chandail b. mouton c. coton d. tricot
6. meubles: charpentier comme roman:

a. livre b. auteur c. rayon d. écrit
7. son: silence comme mouvement:

a. vitesse b. degré c. decoration d. calme
8. breeze: gale as shower:

a. raindrop b. wind c. weather d. cloudburst
9. cowardice: yellow as envy:

a. green b. hatred c. fear d. jealousy
10. touch: numb as sight:

a. bright b. optical c. clear d. blind
11. eggs: beat as cream:

a. pie b. cheese c. whip d. pour
12. water: steam as fire:

a. chimney b. wood c. smoke d. arson
13. compass: direction as watch:

a. strap. b. clock c. sight (1. time

APPENDIX H

ACCEPTABILITY JUDGMENT TASK

Number Date

Here are some sentences. In some of the sentences, some of the words are in the wrong place. Look at
each pair of sentences and circle the answer that you think is best.

EXAMPLE 1:

a. Les jeuncs filles vont au parc.

b. Les jeuncs filles vont le a parc.
only a is right only b is right both right both wrong don‘t know
EXAMPLE 2:

a. Jean pres de l’école habite.

b. Jean habite pres de l’école.
only a is right only b is right both right both wrong don’t know
EXAMPLE 3:

a. Hier Marie a été en retard.

b. Marie a été en retard hier.
only a is right only b is right both right both wrong don’t know
EXAMPLE 4:

a. Ce chien est grand tres.

b. Chien ce est tres grand.
only a is right only b is right both right both wrong don’t know

228

1. a.
b.

only a is right

2. a.
b.
only a is right

3. a.
b.

only a is right

4. a.
b.
only a is right

5. a.
b.

only a is right

6. a.
b.
only a is right

7. a.
b.

only a is right

8. a.
b.
only a is right

9. a.
b.

only a is right

229

Tom au travail va en motocyclette.
Tom en motocyclette va au travail.
both right

only b is right both wrong

Gabrielle mange toujours a six heures.
Gabrielle toujours mange a six heures.
both right

only b is right both wrong

Le bébé d’habitude sourit a sa mere.
Le bébé sourit a sa mere d’habitude.
both right

only b is right both wrong

Frank souvent travaille la nuit.
Frank travaille souvent la nuit.
both right

only b is right both wrong

Carole deteste l’odeur des cigarettes.
Carole l’odeur des cigarettes deteste.
both right

only b is right both wrong

Souvent Thomas oublie ses devoirs.
Thomas souvent oublie ses devoirs.
both right

only b is right both wrong

Louise joue tranquillement avec sa poupee.
Louise tranquillement joue avec sa poupee.
both right

only b is right both wrong

Les jeunes filles lentement finissent leur travail.
Les jeuncs filles finissent leur travail lentement.

only b is right both right both wrong
Linda toujours prend le metro.
Toujours Linda prend le metro.

only b is right both right both wrong

don’t know

don’t know

don’t know

don’t know

don’t know

don’t know

don’t know

don’t know

don’t know

14.

15.

215

chairman: meeting as judge:

a. trial b.

repel: leper
a. banker

as remit:
b. envelope

lawyer

c. criminal

C. CRDCCI'

d.

SCIItCIlCC

d. timer

Page 196: "A votre tour” B Added. Embedded Inference Exercise.

B. Retour de voyage. Vous rencontrez des touristes qui rentrent de vacances.

d’apres les objets declares a la douane.

Ce/Cette touriste arrive de/du/d’ldes...

Souvenirs

owflawewwr

10.
11.

une montre

du parfum

une bouteille de Tequila
un masque d’initiation
des chaussures en cuir fin
un pull en Shetland

du chocolat

un couscoussier

du cafe

des tulipes

une camera video

OCTOBER 11
Do Inference exercise IV. Series Completion.

Pays visités
1e Cameroun
les Pays-Bas
l’Italie

le Mexique
le Japon
l’Ecosse

le Maroc

la Colombie
la Belgique
la Suisse

la France

Dites d’ou ils arrivent

IV. Ci-dessous voici quinze series de termes. Complétez chaque série. Elles sont en anglais et en

francais. Donnez la raison de votre choix.

l.

toujours, d’habitude, quelquefois:

a. peu probable,

Cleveland, Ohio, aux Etats-Unis:

a. l’Europe,

grand-pere, pere, frere:
a. parent,

8060!,

b. fille,

éclair, pAtisserie, dessert: fruit,

a. pomme,

elementary school, junior high, high school: 18,
b. 10,

a. 0,

1e nord du Minnesota, l’est du Connecticut, le sud du Nebraska:
b. minuit,

a. neuf heures,

b. nourriture,

seed, seedling, sapling: teenager,

a. adolescent,

b. child,

possible,
b. probable,

France,
b. region,

C.

C.

C.

C.

C.

certain,

Paris,

mere,

legume,

65,

c. reveil,

C.

adult,

six heures,

(Adapted from Sternberg 1988: 139-140).

(1.

chance,

l’Allemagne,

enfant,

banane,

22,

heure,

infant,

216

8. knuckle, wrist, elbow: knee,

a. foot, b. leg, c. toe, d. hip,
9. bee, sea, dee: pew,

a. pod, b. owe, c. cue, d. bean,
10. broomstick, pencil, toothpick: nail,

a. hammer, b. metal, c. bolt, d. tack,
11. fog, drizzle, shower: flurry,

a. blizzard, b. snowflake, c. drift, d. ice,
12. luxury car, family car, jalopy: cottage,

a. house, b. cheese C. but, (1. mansion,
13. anthem, kickoff, first quarter: last act,

a. intermission, b. overture, c. actors, (I. curtain call,

14. reprimand, suspension, expulsion: robbery,

a. criminal, b. murder, c. theft, d. misdemeanor,
15. animal, vertebrate, reptile: tree,
a. plant, b. forest, c. willow, d. branch,

Exercise Added.
Page 201: "Misc en pratique" Explicit Inference Exercise from Text.

Lecture: Les U.S.A. en jeans
”Avant de lire"
Anticipating content. When you begin any reading, you already have certain expectations. They may be
bawd on your opinion of the publication in which the passage appears, the titles and subheads of the piece,
the name of the author, a review you have read, recommendations by others, the illustrations and
photographs provided, or prior experience with the subject. These expectatiom may or may not play a role
in your understanding.

For example, you expect the readings in Borgiaur, ca va? to contain information about some aspect
of the life of French speakers. You may have had certain preconceptions about the subject. By now, you
are prepared to discover similarities and differences between Francophone culture and your own.

Before reading ”Les U.S.A. en jeans," you may wish to articulate your own expectations by doing
the following exercise.

1. First, look at the title of the passage, and at the illustration(s) that accompany it. Write a sentence
explaining what you think the passage is about.

2. Read the first sentence or two of each paragraph and describe some facts you expect to find in
those paragraphs.

3. Finally, write a third sentence describing a personal experience or an experience of an

acquaintance that reminds you of the content of the reading.

After reading ”Les U.S.A. en jeans," look back at the sentences you have written. Did they anticipate
correctly the content of the passage? You will probably find that your expectations shaped and facilitated
your reading comprehension.

OCTOBER 12
Do Inference exercise V. Classification.

217

V. Below are fourteen verbal classification problems. Decide which option pairs fit with the word given.
Explain why you have made the choice you have. (Adapted from Sternberg 1988: 147-148).

1. book,
a. chapter, page b. magazine, letter c. publisher, printer
d. shelf, library

2. burgundy,
a. scarlet, crimson b. wine, beer c. Paris, France
(1. bottle, cork

3.’ thread,
a. needle, pin b. strand, spool c. sewing, mending
d. rope, string

4. stomach,

digestion, circulation
esophagus, intestines
ulcer, operation
neck, lungs

apes-p:

5. ring,
a. chime, bell b. circle, sphere c. bracelet, necklace
(1. jewelry, money

6. hair,
a. blond, brunette b. fur, coat c. shampoo, haircut
d. teeth, nails

7. chimpanzee,
a. baboon, gorilla b. tree, zoo c. ostrich, antelope

d. jungle, Afiica

8. jack,
a. tire, repair b. card, deck c. ace, seven
(1. rabbit, pot

9. compass,
a. magnet, nwdle b. stencil, ruler c. direction, bearing
(1. north, south

10. pancake,
a. french toast, waffle
b. breakfast, lunch
c. butter, syrup
(1. bread, bagel

11. relish,
a. enjoy, sentiment b. cucumber, pickle c. salt, pepper
d. ketchup, mustard

12.

13.

14.

earth,

a.
(1.

soil, dirt
moon, eclipse

difficult,

a.
b. test, exam
c.
(1. easy, boring

unachievable, impossible

challenging, hard

doctor,

a.
d.

nurse, hospital
professor, lawyer

218

b. star, sun

b. surgeon, artist

C.

C.

VCDUS, comet

illness, disease

APPENDIX E

QUESTIONNAIRE TO TEACHERS

So that I can compare across classes, could you please circle how the lessons were taught? I need to know
if the students did the exercises individually, in pairs, in groups, as a whole class (teacher directed), or as
homework. Please return this to me at your earliest convenience via campus mail:

India Plough

English Language Center

1 International Center
I hope this is the last time I have to bother you!!!! Many thanks again.

 

September 30:
”Lisez le passage” and questions given as homework. Did you correct as a class?

Yes No
October 1:
Page 152: Climats questions
individually pairs group class homework couldn’t do
Page 154: clothing questions
individually pairs group class homework couldn’t do
Page 156: "Bon voyage" fill-in-the-blank exercise
individually pairs group class homework couldn’t do
Page 157: ”Curiosité" exercise.
individually pairs group class homework couldn’t do
October 4:
Page 160: "A Votre Tour" .
individually pairs group class homework couldn’t do
Page 161: ”En Haute-Savoie"
individually pairs group class homework couldn’t do
Question activity
individually pairs group class homework couldn’t do
Dictée-5 points

219

October 5:
Page 173: "En famillc" fill-in-the-blank.

individually pairs group
Page 176: ”A votre tour” A
individually pairs group
Page 176: "A votre tour” B
individually pairs group
Page 177: "Entre vous" A

individually pairs group
Cartoon Activity

individually pairs group

October 7:

Page 179: "A votre tour: Une amitié sincere"

individually pairs group
Page 181: "A votre tour" A
individually pairs group
Page 181: "A votre tour” B
individually pairs group
Page 182: ”Entre vous" A

individually pairs group
Page 182: "Entre vous" B

individually pairs group

Page 183: ”Commentaire culturel"

individually pairs group
Dictee--5 points

October 8:

Memory tricks activity

individually pairs group
October 11:

”En cours de sociologie"

individually pairs group
Vente aux encheres

individually pairs group

October 12:
”Le cas de l’auto-stoppeur affamé" solution
individually pairs group

220

class

class

class

class

class

class

class

class

class

class

class

class

class

class

class

homework

homework

homework

homework

homework

homework

homework

homework

homework

homework

homework

homework

homework

homework

homework

couldn’t do

couldn’t do

couldn’t do

couldn’t do

couldn’t do

couldn’t do

couldn’t do

couldn’t do

couldn’t do

couldn’t do

couldn’t do

couldn’t do

couldn’t do

couldn‘t do

couldn’t do

Page 205: B "Mes vacances"
individually pairs

Page 207: Exercise at the bottom of the page

individually pairs

Page 208: "Nostalgic"
individually pairs

Page 210: "A votre tour" A
individually pairs

Page 210: ”A votre tour" B
individually pairs

Page 214: "A votre tour"
individually pairs

Scrambled word activity
individually pairs

group

group

group

group

group

group

group

class

class

class

class

class

class

class

homework

homework

homework

homework

homework

homework

homework

couldn’t do

couldn’t do

couldn‘t do

couldn’t do

couldn’t do

couldn’t do

couldn’t do

APPENDIX F

QUESTIONNAIRE TO STUDENTS

Please answer the following questions with as much detail as you can. Thank you!

1.

Compare the first four weeks of French class with the last two weeks of class. Was there any
difference in the way the class was conducted? yes no
If yes, what was different? (lessons, amount of work...)

In the last two weeks, were you taught the passe compose? yes no
If yes, when is the passe compose used?

In the last two weeks, were you taught where adverbs are placed in a French sentence? yes
no
If yes, where are the adverbs placed?

 

Are your mistakes corrected in class? yes no
If yes, can you think of an example of when you were corrected?

 

Why are you taking French?

Do you think French is easy or difficult to learn?

Do you think French and English are similar or dissimilar? Can you think of an example of how
they are similar?

Can you think of an example of how they are dissimilar?

222

APPENDIX G

WORD ORDER CORRECTION TASK

Read all of the sentences in the cartoon carefully. In some of the sentences, some of the words are
in the wrong place. Use arrows to put them in the right place.

EXAMPLE:

 

 

 

GARFIELD EST GROS.

@MANGEA

223

224

 

 

 

 

 

 

 

 

 

 

GARFIELD ET JEAN vrsrrENT QUAND 1L3 ARRIVENT. JEAN
D’HABITUDE CAMP GRIZZLY , SOIGNEUSEMENTGARE
ENJUILLET. LAVOI'I'URE. '

 

 

 

GRRR! JE srns UN CHAT SAUVAGE! ,
11! News DORS sous UNE mm;
romomtsuzs crrA'rs SAUVAGES

. 0 mm DANS LES mar-s. /

 

 

 

 

 

 

 

 

JEAN DISCHARGE LA VOITURE
RAPIDEMENT.

 

225

V

TRAVAIL, TRAVAIL, TRAVAIL...
on EST LA T.V.?

JEAN TOUJOURS OUBLIB CE QUI
EST IMPORTANT.

 

 

 

 

 

 

QUEIQUEFQIS JEAN PREPARE
UNinoranPAs.
crasr UNn BONNE caosn Truss
QUE JE sors rcr pour: mm
m

 

 

 

 

 

 

 

 

 
      

MATS n. S‘EST' SOUVENU DE
MON DEJEUNER.

 

 

 

GARFIELD OUVRE LE PANIER DE
PIQUE-NIQUE LENTEMENT.

 

 

 

<‘
j. .

O

A
er:

 

 

 

 

 

 

. , A \\
\\_—% 4: 6
§ (/ '- § v:
. (”a '1)
(I. 5'
@ \
§S§§s C5
CHERCHEGARFIELDIEAN.

IL MONTE SOIGNEUSEMENT
DANS LE PANIER.

226

 

 

  

GARFIELD, 8'“. TE PLAIT, I.
CHERCHE DES VERS PLUS.

   

JEAN, TU N‘AS PAs APPORTE
ASSEZ DE nrscurrsr

  

 

 

 

 

 

 

PLUS TARD, GARFIELD LENTEMENT RAPIDEMENT, JEAN DONNE

 

 

SORT DE LA VOITURE. LA PELLE A GARFIELD.
Dasvnrtsrcmmnomman. I'JED'nAnrrUDEPrtENDSLPsvms.

  

MAIS, D'ABORD IE DOIS ME REPOSER.

 

 

 

 

 

 

 

GARFIELD SOIGNEUSEMENT SE CACHE
DANS UN TRONC D’ARBRE.

227

 

 
   
 

BIEN 30R, PATRON.
JE VEUX vous AmER.

GARFIELD,
vauxsru DU POISSON?

  

 

 

 

 

 

 

PLUS TARD CE’I‘ APRES-MIDI, JEAN
LENTEMENT RENTRE AU CAMP.

 

 

TU SAIS. GARFIELD, DES OURS AUSSI
VIENNENT- QUELQUEFOIS AU CAM?

 

 

 

 

 

 

 

 

SOUPER APRES, JEAN RACONTE UNE
GARFIELD TROUVE VITE DU BOIS moms.

DE CHAUFFAGE.

 

10. a.
b.

only a is right
11. a.

b.
only a is right
12. a.

b.
only a is right
13. a.

b.
only a is right
14. a.

b.
only a is right
15. a.

b.
only a is right
16. a.

b.
only a is right
17. a.

b.
only a is right

18. a.
b.

only a is right

230

Quelquefois Alexandra nettoie sa chambrc.
Alexandra nettoie quelquefois sa chambrc.

only b is right both right both wrong
Charles coupe attentivement le papier.
Attentivement Charles coupe le papier.

only b is right both right both wrong

Superman sauve toujours la vie des personnes.
Superman sauve la vie des personnes toujours.

only b is right both right both wrong
Pierre silencieusement ferme la porte.
Pierre ferme silencieusement la porte.

only b is right both right both wrong

Rapidement la jeune fille mange 1e Big Mac.
La jeune fille mange rapidement le Big Mac.
only b is right

both right both wrong

Pierre parle habituellement francais.
Habituellement Pierre parle francais.
only b is right

both right both wrong

Visiter New York Jean veut.
Jean veut visiter New York.
only b is right

both right both wrong

Les enfants détestent les devoirs ordinairement.
Les enfants ordinairement détestent les devoirs.
only b is right both right both wrong
Les enfants sortent de l’école rapidement.
Les enfants sortent rapidement de l’ecole.
only b is right

both right both wrong

don’t know

don’t know

don’t know

don’t know

don’t know

don’t know

don’t know

don’t know

don’t know

19. a.
b.

only a is right

20. a.
b.

only a is right

21. a.
b.

only a is right

22. a.
b.
only a is right

23. a.
b.

only a is right

24. a.

b.
only a is right
25. a.

b.
only a is right

26. a.
b.

only a is right

27 . a.
b.

only a is right

231

Henri court chez lui rapidement.
Henri court rapidement chez lui.
both right

only b is right both wrong

Les étudiants silencieusement passent l’examen.
Silencieusement les étudiants passent l’examen.
both right

only b is right both wrong

Jeanne va au cinema quelquefois.
Jeanne quelquefois va au cinema.
both right

only b is right both wrong

Susan quelquefois joue du piano.
Susan joue du piano quelquefois.
both right

only b is right both wrong

Robert écrit attentivement avec sa nouvelle plume.
Robert eerit avec sa nouvelle plume attentivement.
only b is right

both right both wrong

Anne conduit attentivement sa nouvelle voiture.
Anne conduit sa nouvelle voiture attentivement.
only b is right

both right both wrong

Marie perd ses livres souvent.
Marie perd souvent ses livres.
both right

only b is right both wrong

Lizette a une grande tres voiture.
Lizette a une grande voiture tres.
both right

only b is right both wrong

Lentement le train part de la gare.
Le train lentement part de la gare.
both right

only b is right both wrong

don’t know

don’t know

don’t know

don’t know

don’t know

don’t know

don’t know

don’t know

don’t know

28. a.
b.

only a is right

29. a.
b.
only a is right

30. a.
b.

only a is right

31. a.

b.
only a is right
32. a.

b.

only a is right

232

Le vieil homme raconte lentement l’histoire.
Le vieil homme lentement raconte l’histoire.

only b is right both right both wrong
David regardc quelquefois la T.V.
David quelquefois regardc la T.V.

only b is right both right both wrong

Les jeuncs filles lisent des livres tranquillement.
les jeuncs filles tranquillement lisent des livres.
both right

only b is right both wrong

Les garcons lentement vont a l’école.
Les garcons vont lentement a l’école.
both right

only b is right both wrong

Jacque d’habitude boit du Coca.
Jacque boit d’habitude du Coca.
both right

only b is right both wrong

don’t know

don’t know

don’t know

don’t know

don’t know

APPENDIX I

RAVEN’S STANDARD PROGRESSIVE MATRICES TEST
SAMPLE ITEMS

 

 

 

 

he
«GO
1363

 

om

 

 

D-ID
IIID

233

APPENDIX J

MULTIPLE REGRESSIONS

(HROKHEBCCMHBHNEDD
ACCEPTABILITY JUDGMENT TASK
s cors

Combined equation:
MTB > Regress 'precorr’ 5 ’SX' ’AGE’ 'YR—F’ ’INF’ ’rav’.
The regressitniequation is

precorr = - 10.9 - 0.79 SX + 0.431 AGE + 0.752 YR—F + 0.26 INF + 0.201 rav
Predictor Coef Stdev t-ratio p
Constant —10.90 15.26 —0.71 0.479

SX - 0.785 2.299 -0.34 0.734

AGE 0.4312 0.6828 0.63 0.531

YR-F 0.7519 0.6806 1.10 0.276

INF 0.264 1.295 0.20 0.840

rav 0.2014 0.1141 1.76 0.085

s = 3.869 R-sq = 12.8%

Analysis of Variance

SOURCE DF SS MS F p
Regression 5 85.83 17.17 1.15 0.353
Error 39 583.95 14.97

Total 44 669.78

SOURCE DF SEQ SS

SX 1 13.88

AGE 1 2.46

YR-F 1 22.12

INF 1 0.75

rav 1 46.62

Fitted model:

Regress ’AJprecorr' 1 ’rav’.

The regressitnlequation is AJprecorr = - 1.87 + 0.225 rav
Predictor Coef Stdev -ratio p
Constant -1.870 5.268 -0.35 0.724

rav 0.2247 0.1061 2.12 0.040

s = 3.756 R-sq = 9.4%

Analysis of Variance

SOURCE DF SS MS F p
Regression 1 63.25 63.25 4.48 0.040
Error 43 606.53 14.11

Total 44 669.78

234

235

Correct Scores Posttest 2 - Pretest

Combined equation:

Regress 'AJC2-0' 6 ’SX’ ’AGB' 'YR-F’ ’INF' ’ADVIN’ ’rav'.

The regressituiequation is

AJC2-0 = - 23.0 + 2.85 SX + 1.58 AGE - 0.206 YReF - 1.51 INF + 0.0727
ADVIN - 0.188 rav

42 cases used 3 cases containrnissing values

Predictor Coef Stdev t-ratio p
Constant ~23.03 17.14 -1.34 0.188

SX 2.853 2.384 1.20 0.239

AGE 1.5787 0.6973 2.26 0.030

YR-F -0.2059 0.7708 -0.27 0.791

INF -1.510 1.519 —0.99 0.327
ADVIN 0.07269 0.05899 1.23 0.226

rav -O.1875 0.1344 -l.40 0.172

s = 3.824 R—sq = 23.0%

Analysis of Variance

SOURCE DF SS MS F p
Regression 6 152.55 25.43 1.74 0.141
Error 35 511.73 14.62

Total 41 664.29

SOURCE DF SEQ SS

SX 1 20.98

AGE 1 68.31

YR-F 1 0.70

INF 1 4.49

ADVIN 1 29.60

rav 1 28.46

Ffimdnmfld:

MTB > Regress ’DPC2—0' 2 ’AGE’ ’rav’.

The regressituiequation is DPC2-0 = - 3.6 + 1.09 AGE - 0.262 rav
42 cases used 3 cases containtnissing values

Predictor Coef Stdev t-ratio p
Constant -3.61 12.02 -0.30 0.766

AGE 1.0880 0.6115 1.78 0.083

rav -0.2620 0.1293 —2.03 0.050

s = 3.833 R-sq = 13.7%

Analysis of Variance

SOURCE DF SS MS F p
Regression 2 91.33 45.66 3.11 0.056
Error 39 572.96 14.69

Total 41 664.29

SOURCE DF SEQ SS

AGE 1 31.01

rav 1 60.32

WORD ORDER CORRECTION TASK
Posttest 2 - Posttest 1 Correct Scores

Combined equation:

MTB > Regress 'Dng2-1’ 6 'sx' 'AGE’ ’YR-F' ’INF' ’ADVIN’ 'rav’.

The regressitniequation is

Dng2-l = 5.7 - 0.19 SX - 0.476 AGE + 0.631 YR-F - 0.607 INF + 0.0558
ADVIN - 0.0443 rav

42 cases used 3 cases containtnissing values

236

Predictor Coef Stdev t-ratio p
Constant 5.72 10.97 0.52 0.605

SX —0.193 1.525 —0.13 0.900

AGE -0.4757 0.4460 -1.07 0.293

YR-F 0.6305 0.4931 1.28 0.209

INF —0.6067 0.9716 -0.62 0.536
ADVIN 0.05579 0.03773 1.48 0.148

rav -0.04432 0.08598 —0.52 0.609

s = 2.446 R-sq = 21.6%

Analysis of Variance

SOURCE DF SS MS F p
Regression 6 57.608 9.601 1.61 0.175
Error 35 209.368 5.982

Total 41 266.976

SOURCE DF SEQ 88

SK 1 13.174

AGE 1 18.472

YR-F 1 9.861

INF 1 0.000

ADVIN 1 14.511

rav 1 1.590

Ffimflrmxbk

MTB > Regress ’Dng2-1’ 2 'ADVIN' 'YR-F'.

The regressitwiequation is Dng2-1 = - 7.12 + 0.0582 ADVIN + 0.914 YR-F
42 cases used 3 cases contain missing values

Predictor Coef Stdev t-ratio p
Constant -7.117 3.015 —2.36 0.023

ADVIN 0.05816 0.03047 1.91 0.064

YR~F 0.9138 0.4222 2.16 0.037

s = 2.386 R-sq = 16.8%

Analysis of Variance

SOURCE DF SS MS F p
Regression 2 44.889 22.445 3.94 0.028
Error 39 222.087 5.695

Total 41 266.976

SOURCE DF SEQ SS

ADVIN 1 18.205

YR-F 1 26.684

Pretest Error Scores

Combined equation:

MTB > Regress 'POGJE' 5 ’SX' 'AGE' ’YR-F' 'INF’ 'rav’.

The regressitnrequation is POGJE = 30.6 — 2.95 SX - 0.791 AGE - 0.811 YR-F
- 0.376 INF - 0.124 rav

Predictor Coef Stdev t-ratio p
Constant 30.596 8.018 3.82 0.000

SX -2.953 1.208 -2.44 0.019

AGE -0.7913 0.3588 -2.21 0.033

YR-F -0.8111 0.3577 -2.27 0.029

INF -0.3756 0.6807 —0.55 0.584

rav -0.12393 0.05998 -2.07 0.046

s = 2.034 R-sq = 28.6%

Analysis of Variance

SOURCE DF SS MS F p
Regression 5 64.630 12.926 3.13 0.018

Error 39 161.281 4.135

237

Total 44 225.911
SOURCE DF SEQ SS
SK 1 10.472
AGE 1 12.335
YR—F 1 22.802
INF 1 1.370
rav 1 17.651
kadnmfld:

MTB > Regress 'POGJE’ 2 'SX' 'rav’.
The regressitniequation is POGJE = 13.0 - 2.40 SX - 0.138 rav

Predictor Coef Stdev t-ratio p
Constant 12.953 3.571 3.63 0.001

SX —2.399 1.169 -2.05 0.046

rav —0.13803 0.06307 -2.19 0.034

s = 2.146 R-sq = 14.4%

Analysis of Variance

SOURCE DF SS MS F p
Regression 2 32.526 16.263 3.53 0.038
Error 42 193.385 4.604

Total 44 225.911

SOURCE DF SEQ SS

SK 1 10.472

rav 1 22.054

Posttest 2 - Posttest 1 Error Score

Combined equation:

MTB > Regress ’DEgj2-1’ 6 ’SX’ ’AGE' ’YR-F’ 'INF’ 'ADVIN’ ’rav'.

The regressitwiequation is

DEgj2-1 = - 4.47 + 0.52 SX + 0.443 AGE — 0.969 YR-F + 0.211 INF - 0.0390
ADVIN + 0.0364 rav

42 cases used 3 cases containinissing values

Predictor Coef Stdev t-ratio p
Constant ~4.467 7.711 -0.58 0.566
SX 0.522 1.072 0.49 0.630
AGE 0.4429 0.3136 1.41 0.167
YR-F -0.9688 0.3467 ~2.79 0.008
INF 0.2110 0.6832 0.31 0.759
ADVIN —0.03895 0.02653 -1.47 0.151
rav 0.03643 0.06046 0.60 0.551
s = 1.720 R-sq = 36.1%

Analysis of Variance

SOURCE DF SS MS F p
Regression 6 58.588 9.765 3.30 0.011
Error 35 103.531 2.958

Total 41 162.119

SOURCE DF SEQ SS

SX 1 8.001

AGE 1 19.264

YR-F 1 22.705

INF 1 0.361

ADVIN 1 7.182

rav 1 1.074

Ffimdnxflbk

MTB > Regress ’DEng-l’ 2 ’YR-F' ’ADVIN’.
The regressitntequation is DEgj2-1 = 7.15 - 1.14 YR—F - 0.0434 ADVIN

238

42 cases used 3 cases containinissing values

Predictor Coef Stdev t-ratio p
Constant 7.149 2.135 3.35 0.002

YR—F -1.1412 0.2989 —3.82 0.000
ADVIN -0.04337 0.02157 -2.01 0.051

s = 1.689 R-sq = 31.3%

Analysis of Variance

SOURCE OF SS MS F p
Regression 2 50.816 25.408 8.90 0.001
Error 39 111.304 2.854

Total 41 162.119

SOURCE DF SEQ SS

YR-F 1 39.274

ADVIN 1 11.541

Pretest Error Subscore

Combined equation:

MTB > Regress 'pgj-e2’ 5 ’SX' 'AGE’ ’YR-F' 'INF' 'rav’.

The regression equation is pgj-e2 = 6.66 - 0.950 SX — 0.123 AGE + 0.022
YR-F + 0.459 INF - 0.0722 rav

Predictor Coef Stdev t—ratio p
Constant 6.663 2.610 2.55 0.015

SX -0.9498 0.3933 -2.41 0.021

AGE -0.1230 0.1168 -1.05 0.299

YR-F 0.0221 0.1164 0.19 0.850

INF 0.4587 0.2216 2.07 0.045

rav -0.07216 0.01952 -3.70 0.001

s = 0.6619 R—sq = 35.1%

Analysis of Variance

SOURCE DF SS MS F p
Regression 5 9.2254 1.8451 4.21 0.004
Error 39 17.0857 0.4381

Total 44 26.3111

SOURCE DF SEQ 88

SK 1 0.6831

AGE 1 0.5365

YR-F 1 0.2201

INF 1 1.8018

rav 1 5.9840

Pkwdnmdd:

MTB > Regress 'pgj-e2' 2 ’SX' 'rav’.
The regressitniequation is pgj—e2 = 4.70 - 0.806 SX - 0.0732 rav

Predictor Coef Stdev t-ratio p
Constant 4.703 1.132 4.16 0.000

SX -0.8060 0.3705 -2.18 0.035

rav -0.07320 0.01999 -3.66 0.001

s = 0.6801 R-sq = 26.2%

Analysis of Variance

SOURCE DF SS MS F p
Regression 2 6.8850 3.4425 7.44 0.002
Error 42 19.4261 0.4625

Total 44 26.3111

SOURCE DF SEQ SS

SK 1 0.6831

rav 1 6.2020

239

Posttest l - Pretest Error Subscore

Combined equation:

MTB > Regress ’WEl’ 6 'SX’ ’AGE’ 'YR—F’ ’INF’ 'ADVIN’ ’rav'.

The regressitniequation is

WEI = - 3.97 + 0.771 SX - 0.023 AGE - 0.009 YR-F - 0.251 INF + 0.00927
ADVIN + 0.0650 rav

Predictor Coef Stdev t-ratio p
Constant -3.971 3.317 -1.20 0.239
SX 0.7708 0.4941 1.56 0.127
AGE -0.0233 0.1448 —0.16 0.873
YR-F —0.0086 0.1487 -0.06 0.954
INF —0.2507 0.3246 -0.77 0.445
ADVIN 0.009265 0.009977 0.93 0.359
rav 0.06499 0.02461 2.64 0.012
s = 0.8203 R-sq = 19.9%
Analysis of Variance
SOURCE DF SS MS F p
Regression 6 6.3433 1.0572 1.57 0.182
Error 38 25.5678 0.6728
Total 44 31.9111
SOURCE DF SEQ SS
SX 1 1.3014
AGE 1 0.0166
YR-F 1 0.1386
INF 1 0.0600
ADVIN 1 0.1370
rav 1 4.6898
Fitted model:
MTB > Regress ’WEl’ 1 'rav’.
The regressituiequation is WEI = - 2.29 + 0.0473 rav
Predictor Coef, Stdev t-ratio p
Constant -2.292 1.154 -1.99 0.053
rav 0.04734 0.02325 2.04 0.048
s = 0.8227 R-sq = 8.8%
Analysis of Variance
SOURCE OF SS MS F p
Regression 1 2.8069 2.8069 4.15 0.048
Error 43 29.1042 0.6768
Total 44 31.9111
MULTIPLE REGRESSIONS
HNFEREFKHFKIGHUDUP

ACCEPTABILITY JUDGMENT TASK
Posttest 2 - Posttest Correct Scores

Combined equation:

MTB > Regress ’DPC2—1’ 6 'SX’ 'AGE’ ’YR-F' ’ADVIN' ’ININF’ ’rav'.

The regressitnlequation is

DPC2-1 = - 17.3 + 1.70 SX - 0.086 AGE + 0.56 YR-F - 0.299 ADVIN + 5.26
ININF + 0.059 rav

29 cases used 1 cases containinissing values

Predictor Coef Stdev t—ratio p

Constant -17.29 21.70 -0.80 0.434

240

SX 1.701 2.803 0.61 0.550

AGE -0.0862 0.8090 —0.11 0.916

YR-F 0.563 1.015 0.55 0.585
ADVIN -0.2994 0.1673 -1.79 0.087

ININF 5.259 2.279 2.31 0.031

rav 0.0585 0.1643 0.36 0.725

s = 3.779 R-sq = 26.1%

Analysis of Variance

SOURCE DF SS MS F p
Regression 6 111.11 18.52 1.30 0.300
Error 22 314.13 14.28

Total 28 425.24

SOURCE DF SEQ SS

SX 1 21.20

AGE 1 3.79

YR-F l 0.03

ADVIN 1 10.04

ININF 1 74.24

rav 1 1.81

Ffimdnmdd:

MTB > Regress 'DPC2-1’ 1 ’ININF’.

The regressitnlequation is DPC2-1 = - 9.81 + 1.52 ININF

29 cases used 1 cases containinissing values

Predictor Coef Stdev t-ratio p
Constant -9.808 5.516 -1.78 0.087

ININF 1.5205 0.7349 2.07 0.048

s = 3.687 R-sq = 13.7%

Analysis of Variance

SOURCE DF SS MS F p
Regression 1 58.20 58.20 4.28 0.048
Error 27 367.04 13.59

Total 28 425.24

Posttest 2 - Posttest 1 Error Scores

Combined equation:
MTB > Regress 'DPE2-l' 6 ’SX' 'AGE’ 'YR-F' ’ADVIN’ ’ININF’ ’rav’.
The regressicurequation is

DPE2-1 = 3.1 - 0.87 SX + 0.566 AGE + 0.395 YR—F + 0.291 ADVIN —
- 0.036 rav

29 cases used 1 cases containrnissing values

Predictor Coef Stdev t-ratio p
Constant 3.11 20.42 0.15 0.880

SX -0.866 2.638 -0.33 0.746

AGE 0.5659 0.7614 0.74 0.465

YR-F 0.3955 0.9549 0.41 0.683
ADVIN 0.2912 0.1574 1.85 0.078

ININF -5.133 2.145 -2.39 0.026

rav -0.0357 0.1547 -0.23 0.820

s = 3.556 R-sq = 29.4%

Analysis of Variance

SOURCE DF SS MS F p
Regression 6 115.98 19.33 1.53 0.215
Error 22 278.23 12.65

Total 28 394.21

SOURCE DF SEQ SS

SX l 9.04

5.13 ININF

241

AGE 1 11.20
YR-F 1 12.28
ADVIN 1 10.68
ININF 1 72.11
rav 1 0.67
Fitted model:

MTB > Regress 'DPE2-1' 1 ’ININF'.
The regressitniequation is DPE2-1 = 10.8 - 1.63 ININF
29 cases used 1 cases contain missing values

Predictor Coef Stdev t-ratio p
Constant 10.811 5.211 2.07 0.048

ININF -1.6274 0.6942 -2.34 0.027

s = 3.483 R-sq = 16.9%

Analysis of Variance

SOURCE DF 88 MS F p
Regression 1 66.67 66.67 5.50 0.027
Error 27 327.54 12.13

Total 28 394.21

Posttest 2 - Pretest Error Scores

Combined equation:

MTB > Regress ’DPE2-O’ 6 ’SX’ 'AGE' ’YR-F’ ’ADVIN’ ’ININF' 'rav’.

The regressitniequation is

DPE2-0 = 8.2 - 2.20 SX - 0.086 AGE + 0.59 YR-F + 0.270 ADVIN - 4.71 ININF
+ 0.065 rav

29 cases used 1 cases containrnissing values

Predictor Coef Stdev t-ratio p
Constant 8.15 22.01 0.37 0.715
SX —2.198 2.844 -0.77 0.448
AGE -0.0860 0.8208 -0.10 0.918
YR-F 0.586 1.029 0.57 0.575
ADVIN 0.2703 0.1697 1.59 0.126
ININF ~4.715 2.312 —2.04 0.054
rav 0.0650 0.1667 0.39 0.700
s = 3.834 R-sq = 25.4%

Analysis of Variance

SOURCE DF SS MS F p
Regression 6 109.88 18.31 1.25 0.322
Error 22 323.36 14.70

Total 28 433.24

SOURCE DF SEQ SS

SX 1 15.46

AGE 1 0.04

YR-F 1 13.97

ADVIN 1 11.44

ININF 1 66.74

rav 1 2.24

Ffimdnxfld:

MTB > Regress ’DPE2-0' 1 ’ININF’.
The regressitulequation is DPE2-0 = 9.27 - 1.58 ININF
29 cases used 1 cases containtnissing values

Predictor Coef Stdev t-ratio p
Constant 9.271 5.543 1.67 0.106
ININF -1.5781 0.7384 -2.14 0.042

s = 3.705 R-sq = 14.5%

242

Analysis of Variance

SOURCE DF SS MS F p
Regression 1 62.69 62.69 4.57 0.042
Error 27 370.55 13.72

Total 28 433.24

WORD ORDER MANIPULATION TASK
Posttest 2 - Pretest Error Scores

Combined equation:

MTB > Regress ’DEng-O’ 6 ’SX’ ’AGE’ ’YR—F’ ’ADVIN' 'ININF’ ’rav’.

The regressitniequation is

DEgj2-0 = 2.2 + 3.05 SX — 0.057 AGE + 0.430 YR-F + 0.066 ADVIN - 2.04
ININF + 0.061 rav

29 cases used 1 cases containtnissing values

Predictor Coef Stdev t-ratio p
Constant 2.22 14.75 0.15 0.882
SX 3.052 1.906 1.60 0.124
AGE -0.0572 0.5501 -0.10 0.918
YR—F 0.4295 0.6899 0.62 0.540
ADVIN 0.0663 0.1137 0.58 0.566
ININF —2.044 1.549 -1.32 0.201
rav 0.0615 0.1117 0.55 0.588
s = 2.569 R-sq = 28.2%

Analysis of Variance

SOURCE DF SS MS F p
Regression 6 56.914 9.486 1.44 0.246
Error 22 145.223 6.601

Total 28 202.138

SOURCE DF SEQ SS

SX 1 11.176

AGE 1 0.487

YR—F 1 8.608

ADVIN 1 21.188

ININF 1 13.455

rav 1 1.999

Fitted model:

MTB > Regress 'DEgj2-0’ 2 'sx' 'ININF’.
The regressiturequation is DEgj2-0 = 5.61 + 3.73 SX - 1.36 ININF
29 cases used 1 cases contain missing values

Predictor Coef Stdev t-ratio p
Constant 5.613 3.622 1.55 0.133

SX 3.730 1.614 2.31 0.029
ININF -1.3600 0.5276 -2.58 0.016

s = 2.419 R-sq = 24.8%

Analysis of Variance

SOURCE DF SS MS F p
Regression 2 50.040 25.020 4.28 0.025
Error 26 152.098 5.850

Total 28 202.138

SOURCE DF SEQ SS

SX 1 11.176

ININF 1 38.864

243

Pretest Error Subscores

Combined equation:

MTB > Regress 'pgj-e2’ 4 'SX’ ’AGE' ’YR-F’ ’rav'.

The regression equation is pgj—e2 = 7.97 — 1.08 SX — 0.101 AGE + 0.020
YR-F - 0.0950 rav

Predictor Coef Stdev t-ratio p
Constant 7.970 3.350 2.38 0.025

SX -1.0788 0.5091 ~2.12 0.044

AGE -0.1011 0.1550 -0.65 0.520

YR-F 0.0196 0.1808 0.11 0.915

rav -0.09502 0.02622 -3.62 0.001

s = 0.7555 R-sq = 39.3%

Analysis of Variance

SOURCE DF SS MS F p
Regression 4 9.2288 2.3072 4.04 0.012
Error 25 14.2712 0.5708

Total 29 23.5000

SOURCE DF SEQ 88

SK 1 0.8333

AGE 1 0.8792

YR~F 1 0.0188

rav 1 7.4975

Fdenmdd:

MTB > Regress 'pgj-e2' 2 ’SX’ 'rav’.
The regressituiequation is pgj-e2 = 6.16 - 0.945 SX - 0.0974 rav

Predictor Coef Stdev t-ratio p
Constant 6.163 1.400 4.40 0.000

SX -0.9452 0.4584 -2.06 0.049

rav -0.09742 0.02518 —3.87 0.001

s = 0.7349 R-sq = 38.0%

Analysis of Variance

SOURCE DF SS MS F p
Regression 2 8.9190 4.4595 8.26 0.002
Error 27 14.5810 0.5400

Total 29 23.5000

SOURCE DF SEQ SS

SK 1 0.8333

rav 1 8.0857

PERFORMANCE BY CLASS SECTION

Tcl=Class Section 1
Tc2=Class Section 2
Tc3=Class Section 3
y intercept=Class Section 4

Acceptability Judgment Task
Posttest 2 Correct Score

MTB > Regress ’pst2corr' 3 ’Tcl' ’Tc2' 'Tc3’.

The regressitniequation is

pst2corr = 9.00 + 5.31 Tel + 1.94 Tc2 + 5.60 Tc3

42 cases used 3 cases containtnissing values
Predictor Coef Stdev t-ratio p

244

Constant 9.000 2.499 3.60 0.001

Tcl 5.308 2.773 1.91 0.063

Tc2 1.938 2.724 0.71 0.481

TC3 5.600 2.850 1.97 0.057

s = 4.329 R-sq = 18.4%

Analysis of Variance

SOURCE DF SS MS F p
Regression 3 160.46 53.49 2.85 0.050
Error 38 712.11 18.74

Total 41 872.57

SOURCE DF SEQ SS

Tcl 1 47.80

Tc2 1 40.29

TC3 1 72.37

MTB > Regress ’pst2corr' 2 ’Tcl’ ’TC3’.

The regressitniequation is

pst2corr = 10.6 + 3.68 Tcl + 3.97 TC3

42 cases used 3 cases containinissing values

Predictor Coef Stdev t-ratio p
Constant 10.6316 0.9868 10.77 0.000

Tcl 3.676 1.548 2.37 0.023

TC3 3.968 1.680 2.36 0.023

s = 4.301 R-sq = 17.3%

Analysis of Variance

SOURCE DF SS MS F p
Regression 2 150.98 75.49 4.08 0.025
Error 39 721.59 18.50

Total 41 872.57

SOURCE DF SEQ SS

TCl 1 47.80

TC3 1 103.18

Word Order Correction Task
Posttest 2 Correct Score

MTB > Regress 'PT2GJC’ 3 ’Tcl’ 'Tc2' 'Tc3'.
The regression equation is

PT2GJC = 4.33 + 6.36 Tel + 4.54 Tc2 + 6.47 TC3
42 cases used 3 cases containinissing values

Predictor Coef Stdev t-ratio p
Constant 4.333 1.799 2.41 0.021

Tcl 6.359 1.995 3.19 0.003

Tc2 4.542 1.960 2.32 0.026

TC3 6.467 2.051 3.15 0.003

s = 3.115 R-sq = 24.8%

Analysis of Variance

SOURCE DF SS MS F p
Regression 3 121.500 40.500 4.17 0.012
Error 38 368.786 9.705

Total 41 490.286

SOURCE DF SEQ SS

Tcl 1 23.654

Tc2 1 1.343

TC3 1 96.503

245

Posttest 2 Error Score

MTB > Regress ’PT2GJE’ 3 'Tc1’ 'Tc2’ ’Tc3’.
The regressitnlequation is

PT2GJE = 7.33 - 5.41 Tcl - 3.83 Tc2 - 6.13 Tc3
42 cases used 3 cases containinissing values

Predictor Coef Stdev t-ratio

P
Constant 7.333 1.315 5.58 0.000
Tcl -5.410 1.459 -3.71 0.001
Tc2 -3.833 1.433 -2.67 0.011
TC3 -6.133 1.500 -4.09 0.000
s = 2.278 R-sq = 34.7%
Analysis of Variance
SOURCE DF SS MS F p
Regression 3 104.929 34.976 6.74 0.001
Error 38 197.190 5.189
Total 41 302.119
SOURCE DF SEQ SS
Tcl 1 12.506
Tc2 1 5.613
Tc3 1 86.810

BIBLIOGRAPHY

BIBLIOGRAPHY

Abraham, R. G., & Vann, R. J. (1987). Strategies of two language learners: A case study. In A. Wenden &
J. Rubin (Eds), Learner strategies in language learning (pp. 85-102). Englewood Cliffs, NJ:
Prentice/Hall lntemational.

Adjemian, C. (1976). On the nature of interlanguage systems. Language Learning, 26, 297-320.

Alderton, D., & Larson, G. (1990). Dirnensionality of Raven's Advanced Progressive Matrices Items.
Educational and Pyschological Measurement, 50, 887-900.

Allwright, R. (l 975). Problems in the study of the language teacher's treatment of learner error. In M. Burt
and H. Dulay (Eds), 0n TESOL ’75 (pp. 96-l09). Washington, DC: TESOL.

Anderson, J .R. (1983). The architecture of cognition. Cambridge, MA: Harvard University Press.
Anderson, J .R. (1985). Cognitive psycho/0g! and its implications. 2nd. Edition. NY: Freeman.

d'Angeljan, A., & Renaud, C. (1985). Learner characterisitics and second language acquisition: A multivariate
study of adult immigrants and some thoughts on methodology. Language Learning, 35, 1-19.

Archibald, J. (1993). Indirect negative evidence and blame assignment in L2 parameter setting. Paper
presented at the 22nd University of Wisconsin Milwaukee Linguistics Symposium. October, 1993.

Atkinson, R., & Raugh, M. (1975). An application of the nmemonic keyword to the acquisition of a Russian
vocabulary. Journal of Experimental Psycholog: Human Learning and Memory, 104, 126- l 33.

Baker, CL. (1979). Syntactic theory and the projection problem. Linguistic Inquiry, 10, 533-581.

Baker, C.L., & McCarthy, J. (Eds). (1981). The logical problem of language acquisition. Cambridge, MA:
MIT Press.

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

Berwick, R. (1985). The acquisition of syntactic knowledge. Cambridge, MA: MIT Press.
Bialystok, E. (1978). A theoretical model of second language learning. Language Learning, 28, 69-83.

Bialystok, E. (1979). The role of conscious strategies in second language proficiency. Canadian Modern
Language Review, 35, 372-394.

Bialystok, E. (1983). Inferencing: Testing the "Hypothesis-Testing" hypothesis. In H. Seliger & M. Long

(Eds), C lassroom-oriented research in second language acquisition (pp. 104-123). Rowley, MA:
Newbury House.

246

247

Bialystok, E. (1990). Communication strategies. Cambridge, MA: Basil Blackwell.

Bialystok, E, & Ryan, E. (1985a). Toward a definition of metalinguistic skill. Merrill-Palmer Quarterly, 31,
229-251.

Bialystok, E., & Ryan, E. (1985b). A metacognitive framework for the development of first and second
language skills. In D.L. Forrest-Pressley, G.E. MacKinnon, & T.G. Waller (Eds), Metacognition,
cognition, and human performance (pp. 207-252). Orlando, FL: Academic Press.

Birdsong, D. (1989). Metalinguistic performance and interlinguistic competence. Berlin: Springer-Verlag.

Bley-Vroman, R. (I 989). What is the logical problem for foreign language learning? In S. Gass & J. Schachter
(Eds), Linguistic perspectives on second language acquisition (pp. 41-68). Carnbrige: Cambridge
University Press.

Bley-Vroman, R., & Chaudron, C. (1990). Second language procesing of subordinate clauses and anaphora-
first language and universal influences. Language Learning, 40, 245-285.

Bley-Vroman, R., Felix, S., & Ioup, G. (1988). The accessibility of universal grammar in adult language
learning. Second Language Research, 4, 1-32.

Blum, S., & Levenston, EA. (1978). Universals in lexical simplification. Language Learning, 28, 399-416.

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

Bomstein, M. (Ed). (1987). Sensitive periods in development: Interdisciplinary perspectives. Hillsdale, NJ:
Lawrence Erlbaum.

Braine, M. (1971). On two types of models of the internalization of grammars. In D. Slobin (Ed), The
ontogenesis of grammar (pp. 153-186). NY: Academic Press.

Brock, C., Crookes, G., Day, R, & Long, M. (1986). Differential effects of corrective feedback in native
speaker-nonnative speaker conversation. In R. Day (Ed), Talking to learn: Conversation in second
language acquisition (pp. 229-236). Rowley, Mass: Newbury House.

Brown, A.L., & Palinscar, AS. (1982). Inducing strategic learning from texts by means of informed self-
control training. Topics in Learning and Learning Disabilities, 2, l-l7. Special issue on
metacognition and leamign disabilities.

Brown, R., Cazden, C., & Bellugi, U. (1969). The child's grammar from I to III. In JP. Hill (Ed), Minnesota
symposium on child psychology, Volume 2. Minneapolis, MN: University of Minnesota Press.

Brown, R., & Hanlon, C. (1970). Derivational complexity and the order of acquisition in child speech. In J .R.
Hayes (Ed), Cognition and the development of language (pp. 155-207). New York: John Wiley.

Carroll, J.B. (198l). Twenty-five years of research on foreign language aptitude. In KC. Diller (Ed),
Individual differences and universals in language learning aptitude (pp. 83-118). Rowley, MA:
Newbury House.

Carroll, 8., & Swain, M. (1991). Explicit and implicit negative feedback: An empirical study of the learning
of linguistic generalizations. Studies in Second Language Acquisition, I 5, 357-386.

Carton, A (1966). The 'Method of Inference' in foreign language study. New York: The Research Foundation
of The City University of New York.

248

Carton, A. (1971). Inferencing: A process in using and learning language. In P. Pirnsleur & T. Quinn (Eds),
The psychology of second language learning (pp. 45-58). Cambridge: Cambridge University Press.

Chamot, A., & Kapper, L. (1989). Learning strategies in foreign language instruction. Foreign Language
Annals, 22, l3-24.

Chamot, A, & O'Malley, J. (1986). A cognitive academic language learning approach: An ESL content-based
curriculum. Wheaton, MD: National Clearinghouse for Bilingual Education.

Chamot, A., & O'Malley, J. (1987). The cognitive academic language learning approach: A bridge to the
mainstream. TESOL Quarterly, 21, 227-249.

Chaudron, C. (1977). A descriptive model of discourse in the corrective treatment of leamers‘ errors.
Language Learning, 27, 29-46.

Chaudron, C. (1983). Research on metalinguistic judgments: A review of theory, methods and results.
Language Learning, 33, 343-377.

Chaudron, C. (1986). Teachers' priorities in correcting leamers' errors in French immersion classes. In R. Day
(Ed), T alla'ng to Learn: Conversation in second language acquisition (pp. 64-84). Rowley, Mass:
Newbury House.

Chaudron, C. (1988). Second language classrooms: Research on teaching and learning. Cambridge:
Cambridge University Press.

Chiszar, D. (1981). Learning theory, ethological theory, and developmental plasticity. In ES. Gollin (Ed),
Developmental plasticity: Behavioral and biological aspects of variations in development (pp. 71-99).
New York: Academic Press.

Chomsky, N. (1965). Aspects of the theory of syntax. Cambridge, Mass: MIT Press.

Chomsky, N. (1981). Lectures on government and binding. Dordrecht: F oris.

Chomsky, N. (1986). Knowledge of language: Its nature, origin, and use. NY: Praeger.

Chomsky, N. (1987). Kyoto Lectures. Unpublished MS.

Christie, K., & Lantolf, J. (1991). The ontological status of learner grammaticality judgments in UG
approaches to L2 Acquisition. Paper presented at Second Language Research Forum, Los Angeles.

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

Clahsen, H., & Muysken, P. (1989). The UG paradox in L2 acquisition. Second Language Research, 5, 1-29.

Cohen, A., & Aphek, E. (1980). Retention of second language vocabulary over time: Investigating the role
of mnemonic associations. System, 8, 221-235.

Cohen, A., & Aphek, E. (1981). Easifying second language learning. Studies in Second Language
Acquisition, 3, 221-236.

Cohen, A., & Hosenfeld, C. (1981). Some uses of mentalistic data in second language acquisition. Language
Learning, 31, 285-313.

Cook, V. (1988). Chomsky's universal grammar. Cambridge, Mass: Basil Blackwell.

249

Cook, V. (1991). The poverty-of-the-stimulus argument and multicompetence. Second Language Research,
7, 103-117.

Coppieters, R. (1987). Competence differences between native and non-native speakers. Language, 63, 544-
573.

Corder, SP. (1967). The significance of leamers' errors. International Review of Applied Linguistics, 5, 161-
170.

Corder, SP. (1983). A role for the Mother Tongue. In S. Gass, & L. Selinker (Eds), Language transfer in
language learning (pp. 85-97). Rowley, MA: Newbury House.

Cowan, R., & Hatasa, Y.A. (1994). Investigating the validity and reliability of native speaker and second-
language learner judgments about sentences. In E. Tarone, S. Gass, & A. Cohen (Eds), Research
methodolog/ in second-language acquisition (pp. 287-302). Hillsdale, NJ: Lawrence Erlbaum.

Crookes, G., & Gass, S. (Eds). (1993). Tasks and language learning: Integrating theory & practice.
Clevedon: Multilingual Matters.

Crookes, G., & Gass, S. (Eds). (1993). Tasks in a pedagogical context: Integrating theory & practice.
Clevedon: Multilingual Matters.

Darbelnet, J. L. (1977). Pensée et structure. NY: Charles Scribner's Sons.

Deeming, D., & Ferguson, M. (1989). The mechanism of temperature dependent sex determination in
crocodilians: A hypothesis. American Zoologist, 29, 973-985.

Dell, F. (1981 ). On the learnability of optional linguistic rules. Linguistic Inquiry, 12, 31-37.

Demetras, M., Post, K., & Snow, C. (1986). Feedback to first language learners: The role of repetitions and
clarification questions. Journal of Child Language, 13, 275-292.

Desrochers, A., Gelinas, C., & Wieland, L. (1989). An application of the mnemonic keyword method to the
acquisition of German nouns and their grammatical gender. Journal of Educational Psychology, 81,
25-32.

Dulay, H. C., & Burt, M. K. (1972). Goofing, an indicator of children's second language strategies. Language
Learning, 22, 234-252.

Dulay, H. C., & Burt, M. K. (1973). Should we teach children syntax? Language Learning, 23, 245-258.

Dulay, H. C., & Burt, M. K. (1974a). Natural sequences in child second language acquisition. Language
Learning, 24, 37-53.

Dulay, H. C., & Burt, M. K. ( 1974b). A new perspective on the creative construction process in child second
language acquisition. Language Learning, 24, 253-278.

Eckman, F. (1992). Second-language acquisition or secondary languages? A linguistic perspective. Plenary
address at the Second Language Research Forum. Michigan State University.

Ellis, R. (1986). Understanding second language acquisition. Oxford: Oxford University Press.
Ellis, R. (1990). Grammaticality judgments and learner variability. In H. Burmeister & P. Rounds (Eds),

Variability in second language acquisition: Proceedings of the Tenth Meeting of the Second Language
Research Forum. Eugene, OR: University of Oregon, Department of Linguistics.

250

Ellis, R. (1991). Grammaticality judgments and second language acquisition. Studies in Second Language
Acquisition, I3, l6l-186.

Eubank, L. (1989). Parameters in L2 acquisition: Flynn revisited. Second Language Research, 5, 129-153.

Faerch, C., & Kasper, G. (1983). Plans and strategies in foreign language communication. In Faerch, C. & G.
Kasper (Eds), Strategies in interlanguage communication (pp. 20-60). London: Longnan.

Felix, S. (1988). UG-generated knowledge in adult second language acquisition. In S. Flynn & W. O'Neil
(Eds), Linguistic theory in second language acquisition (pp.277-294). Dordrecht: Kluwer.

F lahive, DE (1980). Separating the g factor from reading comprehension. In J.W. Oller & K. Perkins (Eds),
Research in language testing (pp. 34-46). Rowley, Mass: Newbury House.

Gass, S. (1988). Integrating research areas: A framework for second language studies. Applied Linguistics,
9, 198-217.

Gass, S. (1994). The reliability of second-language grammaticality judgments. In E. Tarone, S. Gass, & A.
Cohen (Eds), Research methodology in second-language acquisition (pp. 303-322). Hillsdale, NJ:
Lawrence Erlbaum.

Gass, S., & Ard, J. (1984). L2 acquisition and the ontology of language universals. In W. Rutherford (Ed),
Second language acquisition and language universals (pp. 33-68). Amsterdam: John Benjamins.

Gass, S., & Selinker, L. (1994). Second language acquisition: An introductory course. Hillsdale, NJ:
Lawrence Erlbaum.

Gold, E. M. (1967). Language identification in the limit. Information and Control, 10, 467-474.

Harley, B. (1993). Instructional strategies and SLA in early French immersion. Studies in Second Language
Acquisition, 15, 245-259.

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

Homstein, N., & Lightfoot, D. (1981). Explanations in linguistics: The logical problem in language
acquisition. London: Longman.

Hosenfeld, C., Arnold, V., Kirchofer, J. Laciura, J., & Wilson, L. (1981). Second language reading: A
curricular sequence for teaching reading strategies. Foreign Language Annals, 14, 415-422.

Hurford, J .R., & Heasley, B. (1983). Semantics: A coursebook. NY: Cambridge University Press.

Jerne, N.K. (1967). Antibodies and learning: Selection versus instruction. In G. Quarton, T. Melnechuk, &
F. Schrnitt (Eds), The neurosciences: A study program. Rockefeller University.

Johnson-Laird, P.N. ( 1986). Reasoning without logic. In T. Myers, K. Brown, & B. McGonigle (Eds),
Reasoning and discourse processes (pp. 13-49). London: Academic Press.

Kellerman, E. (1979). Transfer and non-transfer: Where we are now. Studies in Second Language
Acquisition, 2, 37-57.

Kellerman, E. (1983). Now you see it, now you don't. In S. Gass & L. Selinker (Eds), Language transfer in
language learning (pp. 1 12-134). Rowley, MA: Newbury House.

251

Kellerman, E. (1984). The empirical evidence for the influence of the L1 in Interlanguage. In A. Davies, C.
Criper, & A. P. R. Howatt (Eds), Interlanguage (pp. 98-122). Edinburgh: Edinburgy University
Press.

Kellerman, E. (1985). Dative alternation and the analysis of data: A reply to Mazurkewich. Language
Learning, 35, 91-101.

Kellerman, E. (1986). An eye for an eye: Crosslinguistic constraints on the development of the L2 lexicon.
In E. Kellerman & M. Sharwood Smith (Eds), Crosslinguistic influence in second language
acquisition (pp. 35—48). NY: Pergamon Press.

Kess, J. (1992). Psycholinguistics: Psychology, linguistics, and the stuoy of natural language. Amsterdam:
John Benjamins

Kim, J. (1993). Null subjects: Comments on Valian (1990). Cognition, 46, 183-193.

Lakoff, G. (1987). Women, fire, and dangerous things: What categories reveal about the mind. Chicago:
University of Chicago Press.

Larsen—Freeman, D., & Long, M. (1991). An introduction to second language acquisition research. New
York: Longman.

Lasnik, H. (1989). On certain substitutes for negative evidence. In R.J. Matthews & W. Demopoulos (Eds),
Learnability and linguistic theory (pp. 89-105). Dordrecht: Kluwer.

Levelt, W.J.M., & Kelter, S. (1982). Surface form and memory in question answering. Cognitive Psychology,
1 4, 78-106.

Levine, M. (1975). A cognitive theory of learning: Research in hypothesis testing. Hillsdale, NJ: Lawrence
Erlbaum.

Lightbown, P. (1985). Great expectations: Second language acquisition research and classroom teaching.
Applied Linguistics, 6, 173-189.

Lightfoot, D. (1989). The child's trigger experience: Degree-0 learnability. Behavioral and Brain Sciences,
12, 321-375.

Lightfoot, D. (1 991). How to set parameters. Cambridge, Mass: NHT Press.

Long, M. (1977). Teacher feedback on learner error: Mapping cognitions. In H. Brown, C. Yorio, & R.
Crymes (Eds), 0n TESOL '77 (pp. 278-293). WashingtonDC: TESOL.

Luria, A. R. (1976). Cognitive development. Cambridge, MA: Harvard University Press.

Mayer, R. (1988). Learning strategies: An overview. In C. Weinstein, E. Goetz, & P. Alexander (Eds),
Learning and study strategies: Issues in assessment, instruction, and evaluation (pp. 1 1-22). NY:
Academic Press.

Mazurkewich, I. (1984a). The acquisition of the dative alternation by second language learners and linguistic
theory. Language Learning, 34, 91-109.

Mazurkewich, I. (1984b). Dative questions and markedness. In F.Eckman, L. Bell, & D. Nelson (Eds),
Universals of second language acquisition (pp. . Rowley, Mass: Newbury House.

252

Mazurkewich, I., & White, L. (1984). The acquisition of the dative alternation: Unleaming
overgeneralizations Cognition, 16, 261-283.

McCawley, J. (1983). Towards plausibility in theories of language acquisition. Communication and
Cognition, 16, 169-183.

McLaughlin, B. (1990). Restructuring. Applied Linguistics, 11, 113-128.

McNeill, D. (1966). Developmental linguistics. In F. Smith & G.A. Miller (Eds), The genesis of language:
A psycholinguistic approach (pp. 15-84). Cambridge, Mass: MIT Press.

Mohan, BA. (1977). Acceptability testing and fuzzy grammar. In S. Greenbaum (Ed), Acceptability in
language (pp. 133-148). The Hague: Mouton.

Moore, T. (1986). Reasoning and inference in logic and language. In T. Myers, K. Brown, & B. McGonigle
(Eds), Reasoning and discourse processes (pp. 51-66). London: Academic Press.

Muyskens, J ., Omaggio Hadley, A., & Convert-Chalmers, C. (1990). Rendez-vous: An invitation to French
3rd ed. New York: McGraw-Hill Publishing Company.

Myers, T., Brown, K., & McGonigle, B. (1986). (Eds). Reasoning and discourse processes. London:
Academic Press.

Naiman, N., Frohlich, M., Stern, H., & Todesco, A. (1978). The good language learner. Toronto: Ontario
Institute for Studies in Education.

Naiman, N., Frohlich, M. & Todesco, A. (1975). The good language learner. TESL Talk, 6, 58-75.

Neufeld, G.G. (1978). A theoretical perspective on the nature of linguistic aptitude. International Review of
Applied Linguistics, 16, 15-25.

Newmeyer, F. (1983). Grammatical theory: Its limits and its possibilities. Chicago: University of Chicago
Press.

O'Malley, J .M. (1987). The efl‘ects of training in the use of learning strategies on learning English as a second
language. In A. Wenden & J. Rubin (Eds), Learner strategies in language learning (pp. 133-144).
Englewood Clifls, NJ: Prentice/Hall International.

O'Malley, J.M., & Chamot, AU. (1990). Learning strategies in second language acquisition. New York:
Cambridge University Press.

O'Malley, J.M., Chamot, A.U., Stewner-Manzanares, G., Kupper, L., & Russo, RP. (1983). Learning
strategies utilized in developing listening and speaking skills in English as a second language.
Rosslyn, VA: InterAmerica Research Associates.

O'Malley, J .M., Chamot, A., Stewner-Manzanares, G., Kupper, L., & Russo, R. (1985a). Learning strategies
used by beginning and intermediate ESL students. Language Learning, 35, 21-46.

O'Malley, J .M., Chamot, A.U., Stewner-Manzanares, G., Russo, R., & Kupper, L. (1985b). Learning strategy
applications with students of English as a second language. TESOL Quarterly, 19, 285-296.

O'Malley, J.M., Chamot, A.U., & Walker, C. (1987). Some applications of cognitive theory to second language
acquisition. Studies in Second Language Acquisition, 9, 287-306.

253

O'Malley, J .M., Russo, R.P., & Chamot, AU. (1983). A review of the literature on learning strategies in the
acquisition of English as a second language: The potential for research applications. Rosslyn, VA:
InterAmerica Research Associates.

O’Malley, J.M., Russo, R.P., Chamot, A.U., Stewner-Manzanares, G. & Kupper, G. (1983). A study of
learning strategies for acquiring skills in speaking and understanding English as a second language:
Uses of learning strategies for different language activities by students of different language
proficiency levels. Rosslyn, VA: InterAmerica Research Associates.

Otsu, Y. (1981). Universal grammar and syntactic development in children: toward a theory of syntactic
development. Unpublished doctoral dissertation. MIT, Cambridge, Mass.

Ott, C., Butler, D. , Blake, R., & Ball, J. (1973). The effects of inter-active image elaboration on the
acquisition of foreign language vocabulary. Language Learning, 23, 197-206.

Oxford, R. (1990). Language learning strategies: What every teacher should know. New York: Newbury
House.

Oxford, R., & Cohen, A. (1992). Language learning strategies: Crucial issues of concept and classification.
Applied Language Learning, 3, 1-35.

Oxford, R., & Crookall, D. (1989). Research on language learning strategies: Methods, findings, and
instructional issues. The Modern Language Journal, 73, 404-419.

Paivio. A., & Desrochers, A. (1979). Effects of an imagery mnemonic on second language recall and
comprehension. Canadian Journal of Psychology, 73, 780—795.

Penner, S. (1987). Parental responses to grammatical and ungrammatical child utterances. Child
Development, 58, 376-384.

Phinney, M. (1987). The pro-drop parameter in second language acquisition. In T. Roeper & E. Williams
(Eds), Parameter setting (pp. 221-238). Dordrecht: Reidel.

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

Politzer, R. (1983). An exploratory study of self-reported language learning behaviors and their relation to
achievement. Studies in Second Language Acquisition, 6, 54-65.

Pollock, J.-Y. (1989). Verb movement, Universal Grammar, and the structure of IP. Linguistic Inquiry, 20,
365-424.

Raugh, M., & Atkinson, R. (1975). A mnemonic method for learning a second-language vocabulary. Journal
of Educational Psychology, 67, l-l6.

Raven, J. C. (1976). Raven's Standard Progressive Matrices, Sets A, B, C, D & E. Oxford: Oxford
Psychologists Press.

Raven, J. (1989). The Raven Progressive Matrices: A Review of National Norrning Studies and Ethnic and
Socioeconomic Variation Within the United States. Journal of Educational Measurement, 26, l-l6.

Raven, J., Court, J .H., & Others. (1989). Raven Manual Research Supplement 4: Normative, reliability and
validity studies. Oxford: Oxford Psychologists Press.

254

Raven, J.C., Court, J.H., & Raven, J. (1992). Raven Manual: Section 3, Standard Progressive Matrices.
Oxford: Oxford Pyschologists Press.

Raven, J ., & Others. (1990). Raven Manual Research Supplement 3: American and international Norms,
neuropsychological uses. Oxford: Oxford Psychologists Press.

Raven, J ., Raven, J.C., & Court, J.H. (1991). Raven Manual: Section I General Overview. Oxford: Oxford
Psychologists Press.

Raygor, A. (1970). McGraw-Hill Basic Skills System Test Manual. Del Monte Research Park, Monterey:
McGraw-Hill.

Rigney, J. (1978). Learning strategies: A theoretical perspective. In HF. O'Neill (Ed), Learning strategies
(pp. 165-205). New York: Academic Press.

Rinvolucri, M. (1984). Grammar games. Cambridge: Cambridge University Press.

Rochester, M.B., Muyskens, J.A., Hadley, A.O., & Convert-Chalmers, C. (1983). Bonjour, ca va? New York:
McGraw-Hill.

Roeper, T., & Weissenbom, J. (1990). How to make parameters work: Comments on Valian. In L. Frazier
& J. de Villiers (Eds), Language processing and language acquisition (pp. 147-162). Dordrecht:
Kluwer.

Rubin, J. (1975). What the 'good language learner' can teach us. TESOL Quarterly, 9, 41-51.

Rubin, J. (1981). Study of cognitive processes in second language learning. Applied Linguistics, 1 I, l 17- l 3 l.

Rubin, J. (1987). Learner strategies: Theoretical assumptions, research history and typology. In A. Wenden
& J. Rubin (Eds), Learning strategies in language learning (pp. 15-30). Englewood Cliffs, NJ:

Prentice-Hall International.

Saleemi, A. (1990). Null subjects, markedness, and implicit negative evidence. In I.M. Roca (Ed), Logical
issues in language acquisition (pp. 235-258). Dordrecht: Foris.

Saleemi, A. (1992). Universal grammar and language leamability. Cambridge: Cambridge University Press.

Scribner, S. (1977). Modes of thinking and ways of speaking. In P.N. Johnson-Laird & P.C. Wason (Eds),
Thinking. Cambridge: Cambridge University Press.

Sharwood Smith, M. (1990). Second language learnability. In I.M. Roca (Ed), Logical issues in language
acquisition (pp. 259-273). Dordrecht: Foris.

Schachter, J. (1984). A universal input condition. In W. Rutherford (Ed), Language universals in second
language acquisition (pp. 167-183). Amsterdam: John Benjamins.

Schachter, J. (1986). Three approaches to the study of input. Language Learning, 36, 21 1-225.

Schachter, J. (1988). Second language acquisition and its relationship to Universal Grammar. Applied
Linguistics, 9, 219-235.

Schachter, J. (1989). Testing a proposed universal. In S. Gass & J. Schachter (Eds), Linguistic perspectives
on second language acquisition (pp. 73-88). Cambridge: Cambridge University Press.

Schachter, J. (1991). Corrective feedback in historical perspective. Second Language Research, 7, 89-102.

255

Schachter, J ., Tyson, AF ., & Diflley, F .J. (1976). Learner intuitions of grammaticality. Language Learning,
26, 67-76.

Schachter, J., & Yip, V. (1990). Grammaticality judgments: Why does anyone object to subject extraction?
Studies in Second Language Acquisition, 12, 379-392.

Schwartz, B. (1993). On explicit and negative data effecting and affecting competence and linguistic behavior.
Studies in Second Language Acquisition, 15, 147-163.

Schwartz, 8., & Gubala-Ryzak, M. Learnability and grammar re-organization in L2A: Against negative
evidence causing the unlearning of verb movement. Second Language Research, 8, 1-38.

Scribner, S. (1977). Modes of thinking and ways of speaking: Culture and logic reconsidered. In P.N.
Johnson-Laird & P.C. Wason (Eds), Thinking: Readings in cognitive science (pp. 483-500).

Seliger, HW. (1983). The language learner as linguist: Of metaphors and realities. Applied Linguistics, 4,
179-191.

Seliger, HW. (1984). Processing universals in second language acquisition. In F. Eckman, L. Bell, & D.
Nelson (Eds), Universals in second language acquisition (pp. 36—47). Rowley, MA: Newbury House.

Snow, 0, & Ferguson, C. (Eds). (1977). Talking to children: Language input and acquisition. Cambridge:
Cambridge University Press.

Sobol, DJ. (1978). The case of the hungry hitchhiker. In Encyclopedia Brown and the case of the secret
pitch. NY: Bantam Books.

Sorace, A. (1988). Linguistic intuitions in interlanguage development: the problem of indeterminacy. In J.
Pankhurst, M. Sharwood Smith, & P. van Buren (Eds), Learnability and second languages (pp. 167-
I89). Foris.

Sorace, A. (1993). What do speakers know about their languages: Investigating linguistic acceptability
judgments. Paper presented at Michigan State University, English Language Center.

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

Stern, HH (197 5). What can we learn from the good language learner? Canadian Modern Language Review,
31, 304-318.

Sternberg, R.J. (1988). The triarchic mind. New York: Viking Press.

Tarone, E, Gass, S., & Cohen, A. (Eds). (1994). Research methods in second language acquisition. Hillsdale,
NJ: Lawrence Erlbaum.

Thompson, 1. (1987). Memory in language learning. In A. Wenden & J. Rubin (Eds), Learning strategies in
language learning (pp. 43-56). Englewood Clifls, NJ: Prentice/Hall International.

Tomasello, M., & Herron, C. (1988). Down the garden path: Inducing and correcting overgeneralization errors
in the foreign language classroom. Applied Psycholinguistics, 9, 237-246.

Tomasello, M., & Herron, C. (1989). Feedback for language transfer errors. Studies in Second Language
Acquisition, I I, 385-395.

256

Trahey, M., & White, L. (1993). Positive evidence and preemption in the second language classroom. Studies
in Second Language Acquisition, 15, 181-204.

Trusted, J. (1979). The logic of scientific inference. London: Macmillan Press.

Valian, V. (1990a). Null subjects: A problem for parameter-setting models of language acquisition.
Cognition, 35, 105-122.

Valian, V. (1990b). Logical and psychological constraints on the acquisition of syntax. In L. Frazier & J. de
Villiers (Eds), Language processing and language acquisition (pp. 1 19-145). Dordrecht: Kluwer.

Valian, V. (1993). Parser failure and grammar change. Cognition, 46, 195-202.

VanPatten, B., & Cadiemo, T. (1993). Explicit instruction and input processing. Studies in Second Language
Acquisition, 15, 225-243.

Weinstein, CE, & Mayer, RE (1986). The teaching of learning strategies. In M.C. Wittrock (Ed),
Handbook on research on teaching, 3rd ed. (pp. 315-327). New York: Macmillan

Wenden, A., & Rubin, J. (1987). Learning strategies in language learning. Englewood Cliffs, NJ: Prentice-
Hall International.

Wesche, MB. (1979). Learning behaviors of successful adult students on intensive language training.
Canadian Modern Language Journal, 35, 415-427.

White, L. (1989). Universal grammar and second language acquisition. Amsterdam: John Benjarnins.

White, L. (l991a). Adverb placement in SLA: Some effects of positive and negative evidence in the
classroom. Second Language Research, 7, 133-161.

White, L. ( l 99lb). The Verb-Movement Parameter in Second Language Acquisition. Language Acquisition,
I, 337-360.

White, L. (1992a). Long and short verb movement in second language acquisition. Canadian Journal of
Linguistics, 37, 273-286.

White, L. (1992b). On triggering data in L2 acquisition: A reply to Schwartz & Gubala-Ryzak. Second
Language Research, 8, 120-137.

Witkin, H.A., Oltrnan, P.K., Raskin, E, & Karp, SA. (1971). A manual for the Embedded Figures Test. Palo
Alto, CA: Consulting Psychologists Press.

N STRTE UN V L

r . IBRARIES
lllllllllllllllllllllllll
154221

nICHroA
lllllllll 11 ll ll
:31

293010