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

 

DATE DUE DATE DUE ' DATE DUE

 

 

I ‘ MW? 9 2067
I

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1/98 WM

THE ACQUISITION OF REFLEXIVE
PRONOUNS AMONG ADULT LEARNERS OF ENGLISH

By

Margo K. Glew

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
DOCTOR OF PHILOSOPHY
Department of English

1998

ABSTRACT

THE ACQUISITION OF REFLEXIVE
PRONOUNS BY ADULT LEARNERS OF ENGLISH

By
Margo K. Glew

This study investigated the acquisition of reﬂexive pronouns by second
language learners of English. The study proposed that when individual
grammars are analyzed, the interlanguage systems of learners would reﬂect
the acquisition stages identified in Franks and Connell (1996) for English
speaking child language acquirers.

The three acquisition stages identified in this study did not, as
predicted, reﬂect the acquisition stages of child language learners. One group
of learners recognized only the referential dependency of the reﬂexive, by
allowing any pragmatically reasonable noun phrase serve as an acceptable
antecedent. A second group of learners required a local antecedent only in a
non-finite context. It is argued that these learners are respecting the structural
constraints on acceptable antecedents for the reﬂexive, yet have misanalyzed
the English reﬂexive as morphologically simplex. The third, and largest,

- group of learners exhibited target-like behavior showing that ultimate
attainment is possible for English reﬂexives.

The possible inﬂuences of the types of reﬂexives in a learner's first
language as well as the number of prior languages are considered as possible
factors slowing down the acquisition of English reﬂexives.

COPyrisht by
MARCO KATHLEEN GLEW
1998

ACKNOWLEDGMENTS

This dissertation would absolutely not have been possible if not for the
kind support of my husband Rob. The support came not only in the form of
"technical support," in terms of formatting, proofreading, and watching
children sol could work, but, most importantly, most critically, in the form of
encouragement and a conviction that I would and could do this. Thank you,
Rob, for all your patience and kind words. I depended on them both so much.

I would also like to thank Susan Gass, the chairperson of my
committee. I am so grateful to have had the Opportunity to have worked
with Sue. I have benefited incredibly from her advice and support. She was a
mentor in the truest sense of the word, helping me with my work,
encouraging me along the way, and always supporting me with whatever
direction I chose to take. As a professor, a teacher, and a friend, I learned from
the very best!

I am also so grateful to have been able to work with my committee
members. Thank you to Charlene Polio for sharing her expertise with me as
well as her humor and friendship these past few years. Thank you to David
Dwyer for his careful reading and thoughtful comments on previous drafts,
and also for his friendship and support. And thank you to Barbara Abbott,
who, as my first linguistics teacher, showed me how exciting and fascinating
the study of language really is.

iv

I would also like to thank Mimi Bentley for her friendship as well as
her advice and support on this dissertation. For all the help on my statistics
and computer "challenges" thank you to Gary Cook and Denni
Hoopingarner.

And finally, I would like to thank my children, Robbie, Tim, and
Madeleine for all of their love, laughter, and jokes throughout.’ No matter

how long the day was, they always gave me a reason to smile!

TABLE OF CONTENTS

LIST OF TABLES ................................................................................................ x
LIST OF FIGURES .............................................................................................. xii
CHAPTER 1
DESCRIPTIVE MODELS OF REFLEXIVE BINDING ................................... 1
1.1. Introduction ..................................................................................... 1
1.2. Research context ............................................................................. 3
1.3. The binding of antecedent and reﬂexive ................................... 6
1.4. Parameterized approach to reﬂexive binding .......................... 11
1.4.1. The governing catefory and the binding domain in
English ................................................................................ 11
1.4.2. The governing category and proper antecedent
parameters ......................................................................... 14
1.4.3. The acquisition of reﬂexive pronouns: the subset
principle ............................................................................ 15
1.4.4. Criticisms of the CC? and PAP ...................................... 19
1.5. Raising approaches ......................................................................... 19
1.6. Relatavized SUBIECT analysis ..................................................... 22
1.7. The argument structure perspective ........................................... 25
1.7.1. SE-anaphors ....................................................................... 25
1.7.2. Self-anaphors .................................................................... 26
1.7.3. Logophoricity .................................................................... 27
1.8. Chapter summary ........................................................................... 29
CHAPTER 2
THE SECOND LANGUAGE ACQUISITION OF ENGLISH
REFLEXIVES ....................................................................................................... 31
2.1. Introduction .................................................................................... 31
2.2. Ultimate attainment: Do adult learners acquire English
reﬂexive pronouns? ..................................................................... 32
2.2.1. Learners from various L! backgorunds acquire
English reﬂexive pronouns .......................................... 32
2.2.2. Methodological problems .............................................. 35
2.2.3. Learners form various L1 backgrounds bind English
reﬂexives on-distance ..................................................... 37

2.2.4. Learners accept more long-distance binding into

vi

non-finite subordinate clauses than inot fininte

subordinate clauses ......................................................... 40
2.3. Possible explanations for the acquisition process of English
reﬂexive pronouns .......................................................................... 46

2.3.1. Explanation #1: Learners transfer part or all of
their reﬂexive system from the L1 and apply it

to English ............................................................................ 46
2.3.1.1 Transfer XP ........................................................... 46
2.3.1.2 Transfer X ............................................................. 47
2.3.1.3 Pragmatic strategy transfer ................................ 50

2.3.2. Explanation #2: The L2 acquisition process of
English reﬂexives is a development similar to the
acquisition process of children learning English... 60
2.3.2.1. Child language acquisition of relxive

pronouns .......................................................... 60
2.3.3.3. SLA of English reﬂeives: reconsidering
the data ............................................................... 65
2.4. Chapter summary .......................................................................... 73

CHAPTER 3 .

METHODOLOGY ................................................................................................ 71
3.1. Introduction ..................................................................................... 75
3.2. Methodological considerations ................................................... 75

3.2.1. Introduction ..................................................................... 75
3.2.2. Proficiency measurement in SLA studies of
reﬂexives ........................................................................... 76
3.2.2.1. Proficiency scores as a variable to control
for in comparing learners of various L1
backgrounds ..................................................... 80
3.2.2.2. Heterogenous proficiency groups and
analysis of individual grammars ............... 83
3.2.2.3. Proficiency scores as "benchmarks" against
which changes in IL can be pegged ............... 84
3.2.3. Immersion in the L2 environment ............................. 89
3.2.4. L1 background ................................................................... 91
3.3. Research design. .............................................................................. 91
3.3.1. Research questions .......................................................... 91
3. 3. 1.1. Matching research design with research
questions ............................................................. 92
3.3.1.2. Research questions ........................................... 94
3.4 Subjects ............................................................................................... 95
3.4.1. Control group .................................................................... 95
3.4.2. Experimental group ......................................................... 95
3.4.2.1. Proficiency level ................................................. 95
3.4.2.2. Time spent in the L2 environment / years
of formal English study .................................... 96

vii

3.4.2.3. L1 background .................................................... 97

3.5. Methods and materials ................................................................... 98
3.5.1. Materials .............................................................................. 98
3.5.1.1. Task ....................................................................... 98
3.5.1.2. Sentence types ..................................................... 101
3.5.2. Response patterns ............................................................. 102
CHAPTER 4
RESULTS ............................................................................................................. 104
4.1. Introduction .................................................................................... 104
4.2. Data reporting ................................................................................. 104
4.3. Exclusion criteria ............................................................................ 106
4.3.1. Proficiency check ............................................................ 106
4.3.2. Systematicity ..................................................................... 108
4.4. ENS control group .......................................................................... 110'
4.4.1. Exclusion criterion .......................................................... 110
4.4.1.1. Proficiency check .............................................. 110
4.4.1.2. Systematicity ..................................................... 111
4.4.2. Group results ................................................................... 111
4.4.3. Individual results .......................................................... 113
4.4.3.1 Individual results: English-only speakers. 113
4.4.3.2 Individual results: bilingual speakers ......... 117
4.5. Learner results ................................................................................ 118
4.5.1. Learner groups ................................................................. 118
4.5.2. Exclusion criteria, proﬁciency check ............................ 120
4.5.2.1. Proficiency check .............................................. 120
4.5.2.2. Systematicity ...................................................... 121
4.5.3. Group results .................................................................... 122
4.5.4. SL individual results ....................................................... 127
4.5.4.1. Target-like group ............................................... 127

4.5.4.2. Local domain in the ﬁnite context/
=/ - local domain in the non-finite context. 130

4.5.4.3. "Yes" group ......................................................... 132
4.5.4.4. Unclassified ......................................................... 133
4.5.5. ML individual results ...................................................... 133
4.5.5.1. Target-like group ............................................... ~. 133

4.5.5.2. Local domain in the finite context/
=/ - local domain in the non-finite context. 135

4.5.5.3. "Yes" group ......................................................... 136

4.6. Chapter summary ............................................................................ 137
CHAPTER 5

DISCUSSION AND CONCLUSIONS ............................................................ 139

5.1. Introduction .................................................................................... 139

5.2. L2 domain patterns ........................................................................ 140

viii

5.2.1. Grammatical patterns and grouping ........................... 144

5.2.2. Pragmatic L2 strategy ....................................................... 145
5.2.2.1. L1 Acquisition of reeﬂexive pronouns:
early stages .......................................................... 145
5.2.2.2. Acquiring the referential dependency of
the L2 reﬂexive .................................................. 146
5.2.2.3. "Yes" group: non-structural binding ........... 147
5.2.3. LD group ............................................................................. 154
5.2.4. Local group ......................................................................... 164 1
5.3. Orientation and domain ............................................ 165
5.3.1. The SLA orientation of domain: predictions from
theoretical and SLA research ......................................... 165
5.3.2. SLA research context ........................................................ 167
5.3.3. Orientation and domain in the dissertation study.... 169
5.4. Conclusions and suggestions for further research ..................... 172
5.4.1. Acquisition stages .............................................................. 172
5.4.2. Suggestions for further research .................................... 175
APPENDIX A - Experimental and control task ............................................. 177
BIBLIOGRAPHY ................................................................................................... 183

ix

Table 1 -

Table 2 -

Table 3 -

Table 4 -
Table 5 -
Table 6 -
Table 7 -
Table 8 -

Table 9 -

LIST OF TABLES

Mean scores and standard deviations for Japanese speaking
learners of English

The experimental questions associated with the 13 vignettes of
the VMTT

The acquisition stages of reﬂexive pronouns for English
monolingual children

First language backgrounds ESL learners

Predicted response patterns for experimental subjects
Mean scores for English control group

ENS individual results

Individual answers for non-local tokens

Individual results

Table 9 - Number and percentage of subjects dropped due to

Table 10 -
Table 11 -
Table 12 -
Table 13 -
Table 14 -
Table 15 -

comprehension errors

L1 backgrounds of bilingual speaking learners of English
L1 and L2 backgrounds of multi-lingual learners of English
L1 backgrounds of L2 learner group

Descriptives of group performance on experimental task
Contrast tests comparing mean scores

Case summaries

Table 16 - Paired samples correlations
Table 17 - L2 group - individual results
Table 18 - ML group - individual response

Table 19 - Acquisitions stages of child language learners (both normative
and SLI)

Table 20 - Percentage of learners consistently accepting antecedents for
the reﬂexive in all domains and contexts

Table 21 - First language backgrounds of subjects exhibiting a "yes"
response pattern

Table 22 - First language backgrounds of learners exhibiting a "yes"-type
response pattern

Table 23 - First language backgrounds of learners exhibiting LD binding
pattern

Table 24 - Average years of language study for each acquisition group

xi

LIST OF FIGURES

Figure 1 - C—command domain of the NP Mick Iaggar

Figure 2 - C-command domain of the NP Mick laggar’s mother
Figure 3 - Subset relationships for the governing category parameter
Figure 4 - Subset relationships for the proper antecedent parameter
Figure 5 - Phrase structure tree of Lf raising of Chinese anaphor, ziji

xii

Chapter 1
DESCRIPTIVE MODELS OF REFLEXIVE BINDING

1.1 Introduction

This dissertation study investigates the second language acquisition of
reﬂexive pronouns. Reﬂexive pronouns have been studies extensively by
linguists specializing in a diverse range of academic disciplines including
syntax, semantics, child and second language acquisition. (SLA). Thus, not
only does this dissertation address the specific issues surrounding the study
of the SLA of reﬂexives, but it is discussed within the contexts of child
language acquisition research as well as current work in the theoretical
description of the relationship between reﬂexives and their antecedents.

The first chapter of this dissertation, then, provides explanations of the
current theoretical descriptions of reﬂexive] antecedent relationships across
languages. The chapter begins with a discussion of the assumption that
human language acquisition is an innate process, how reﬂexive pronouns
relate to this assumption, and why studying this particular linguistic
sIructure (reﬂexive pronouns) is important in helping SLA researchers
better understand the acquisition of a second language. This is followed by
descriptions of the theoretical models developed to account for the
variation across human languages in how different noun phrases may and

may not function as an antecedent for a given reﬂexive. The models I

discuss are the models which many second language acquisition researchers
appeal to in the discussion / explanation of their data on the SLA of reﬂexive
pronouns.

Chapter Two is a literature review of the SLA studies of reﬂexive
pronouns. The goal of this chapter is two-fold. The first part of the chapter
discusses the specific findings of research on the SLA of reﬂexives. It tells
what we know to be true about the behavior of second language learners
regarding English reﬂexives. From there, I discuss possible explanations for
these findings. In this section I provide the explanations that have come .
from the researchers themselves as well as my own thoughts on a possible
explanation for the data. I raise the possibility that my explanation for the
data accounts for what we currently know to be true about the behavior of
learners acquiring English reﬂexives in a second language.

Chapter Three begins with a discussion of methodological issues
which have a direct bearing on the research design of the dissertation study.
Factors addressed in this chapter include the use of proficiency
measurements (as a factor to control for in studies, as a "benchmar " against
which changes in learner grammars can be ”pegged", and as a grouping
factor which obscures systematic differences in individual learner
grammars), the importance of working with experimental subjects who
have been living in the L2 environment, and the issue of the first language
background of second language learners and how that may affect how a
second language learner acquires English reﬂexives.

The second part of the chapter provides the research questions and
background information about the methods and materials for the study as
well as the subjects (control and experimental) who participated.

Chapter Four outlines the results of the study. Included in this

chapter is a discussion of the exclusion criteria which were adopted for the
study as well as an explanation of the results and response patterns for both
the control and experimental groups.

Chapter Five discusses the results that were reported in Chapter Four.
The various response patterns identified in the learners are grouped and
evaluated. In addition to identifying three distinct acquisition stages, I
provide argumentation to suggest why the learners are operating with the
grammars they are as well as how they may proceed from one acquisition
stage to another. Finally, I give some final conclusions regarding this
research study and its place in the context of SLA research as well as
suggestions for future work on the second language acquisition of English

reﬂexives.

1.2 Research context

Linguists and language acquisition specialists have long assumed
that, barring severe mental or physical impairment or injury, all children
acquire the language of their environment to native—like competence. This
universal competence in a first language is one of several pieces of evidence
that linguists point to in supporting the argument that language acquisition
is an innate ability that human beings are born with. Another key piece of
evidence that language acquisition is an innate skill, is the observation that
the input the child receives underdetermines the grammar the child builds
based on that data. White (1989) uses the distribution of reﬂexive pronouns
as an example. In English, the reﬂexive pronoun and its antecedent
typically occur in the same sentence, but there are many apparent exceptions
to this rule. Despite the many different possible interpretations of reﬂexive
pronouns that are present in the linguistic input, children learn to follow

"the rules" of reﬂexive pronoun use. White illustrates the incredible
complexity of the input regarding reﬂexive pronouns in the following

example. The co—reference relationships are shown in italics (White 1989: 9).

(1) a. John saw himself.
b. *Himselfsaw John.
c. Looking after himselfbores John.
d. *John said that Fred liked himself.
e. John said that Fred liked himself.
f. John told Bill to wash himself.
g. John promised Bill to wash himself.
h. *John believes that himself is intelligent.
i. John believes himself to be intelligent.
j. John showed Bill a picture of himself.

White goes on to point out:

Children have to discover that the reﬂexive pronoun usually
must follow the antecedent (1a and 1b) but that this is not always
so (1c); that the reﬂexive usually must be in the same clause as
the antecedent (1a, d, e) but not always so (If); that the reﬂexive
can be in subject position of a non-finite embedded clause (1i),
but not a finite one (1h); that the closest NP is usually the
antecedent (1d, e, f), but not always so (1g); that in some cases,
there is more than one possible antecedent (lj) , but not always
so (1d, e, f, g). It would seem impossible to work out the
distribution of anaphoric pronouns on a trial and error basis
without a great many mistakes (White 1989: 9-10).

Given this example from the acquisition of reﬂexives, as well as
many other examples of cases where children quickly and easily come to

create a grammar which accounts for all of these "rules," researchers have

long assumed that human beings are born with an innate capability for
language acquisition which helps them quickly and easily create this
complex system we call human language.

The challenge, then, is to define and provide the characteristics of
what exactly the language universals are. What, exactly, constitutes the
expectations the child is born with and what, exactly, is the "system" the
child acquires and uses in speaking and understanding his/ her native
language?

Second Language Acquisition (SLA) researchers are also greatly
interested in this search for a description of linguistic universals as well as
descriptions of the language specific system children ultimately acquire.
This, of course, is because SLA researchers concern themselves with trying
to understand the process by which people learn a second language. SLA
researchers rely on the work of first language acquisition researchers and
linguists in order to compare how language acquisition by an adult is
similar to or different from language acquisition by a child.

One way to explore the possible similarities and differences between
L1 and L2 acquisition is to investigate the process of acquisition of reﬂexive
pronouns. As shown in (1), the various usage patterns of this structure are
extremely complex, yet children eventually learn to master this structure
(Franks and Connell, 1996; Bloom and Conway, 1994). Because the input, in
ways already discussed, underdetermines the system children ultimately
acquire, it has been assumed that an innate language acquisition component
in some way helps human beings acquire the reﬂexive pronoun system. If
children rely on their innate capacity for language acquisition to guide them
in acquiring the correct use of reﬂexive pronouns, how do adults acquire

reﬂexive pronouns? Do they have access to this innate system relied upon

by children or do they acquire reﬂexive pronouns by relying on what they
know about reﬂexive pronouns in their first language? These questions are
the focus of this dissertation.

First, however, this chapter describes some of the theoretical models
developed to account for the various binding patterns across human
languages. These are the models which many second language acquisition

researchers appeal to in the discussion] explanation of reﬂexive data.

1.3 The binding of antecedent and reﬂexive
In order to be interpreted correctly, nouns must convey several pieces
of information. To illustrate this point, consider the noun phrase (NP)

Mick Jaggar in sentence (2) below:

(2) Mick Jaggar sang the song.

From this sentence, we can interpret Mick Jaggar as being the actor or agent
of the sentence. We also know that the Mick Jaggar NP carries the features
[+male] and [+singular]. Furthermore, when we hear the name Mick Jaggar,
many of us can conjure up a mental picture of the real world person to
whom the NP Mick Jaggar refers.

In this short sentence, then, the NP Mick Jaggar is assigned a theta-
role (of agent) by the predicate, and, simultaneously, marked as the
grammatical subject. The NP is also marked for certain features necessary
for agreement, those being [+male] and [+singular]. Finally, the noun
phrase refers to a person in the real world (the referent), the singer and rock-
star Mick Jaggar.

Reﬂexive pronouns contain much of the same important

information as regular NPs. Consider sentence (3) below:
(3) Mick Jaggari hurt himselfi,

In sentence (3) the reﬂexive pronoun himself is assigned a theta-role
(patient), a grammatical role (direct object), and the nominal features of
gender and number [+male] [singular]. Additionally, know who hi msel f is.
Himself is linked with the NP Mick Jaggar which, in turn, refers to the real-
world rock star. In this way, reﬂexive pronouns are linked up with (are
bound by) another noun or pronoun (the antecedent). In sentence (3) the
NP Mick Jaggar binds with the reﬂexive pronoun himself. The result of
this binding process is that the two noun phrases now share the same
referent, the real-world rock—star Mick Jaggar. This relationship is indicated
by subscripts on the co-referential nouns.

In sentence (4) we see that the reﬂexive pronoun agrees in gender and
number with its antecedent. The reﬂexive, herself, refers to a single female
person. Given that the reﬂexive and the antecedent both share the same
referent, both NPs must have the same nominal features for gender and

number.
(4) Mary; gave Billj some information about herselfﬁ', j,

Reﬂexive pronouns bear a theta-role and grammatical function and
are marked with nominal features of gender and number. However, they
fail to independently assign reference. Because of this, reﬂexive pronouns
bind with a noun phrase (an overt NP or pronoun) in order to obtain a

referent. Furthermore, the reﬂexive and antecedent must agree in gender

and number.

Another elementary notion of reﬂexive-antecedent relations is c-
command. That is, in order to function as the antecedent for the reﬂexive, a
given NP must c-command the reﬂexive pronoun. C-command is defined

as follows (Haegeman 1991: 122):

(5) a node A c-commands a node B iff
(i) A does not dominate B;
(ii) B does not dominate A;
(iii) the ﬁrst branching node dominating A also

dominates B

In other words, a node A c-commands its sisters (the neighboring nodes)
and all of its sisters' descendants. The antecedent must therefore be "higher
on the tree" than the reﬂexive. Sentence (6a) is unacceptable because, as

Figure 1 shows, the NP Mick Jaggar does not c-command the reﬂexive.

(6a) Mick Jaggar'si' mother loves himself*i
(6b) Mick Jaggar's mother," loves herselfi

If we change the reﬂexive pronoun to herself, the Sentence becomes
acceptable. This is because while the NP Mick Jaggar fails to c-command the
reﬂexive pronoun, the full NP Mick Jaggar's mother , as in (6b) and
illustrated in Figure 2, does c-command the reﬂexive.

 

 

 

 

NP ‘1'
NP N ' I VP
[+TENSE] ‘
1 v-
N V NP
Mick Jaggar's ' motheri -s loves himself*[

Figure 1: C-command domain of the NP Mick Jaggar

10

./\.
NP/ \N' 1/ \VP
[+TENSE] '
/\7
l \

Mick Jaggar's mother,“ -s loves herself,‘

 

 

 

Figure 2: C-command domain of the NP Mick Jaggar's mother

11

1.4 Parameterized approach to reﬂexive binding

1.4.1 The Governing Category and the Binding Domain of English. Given
the basic assumptions that a reﬂexive pronoun must bind with another NP
that agrees in gender and number with it, and that the reﬂexive and the
antecedent must respect the c-command requirement, linguistic theory
must also provide a description of the domain in which reﬂexive pronouns
may bind with an antecedent.

Chomsky (1981) argues that anaphoric reﬂexives and their
antecedents are base-generated and co-indexed within their governing
category. Chomsky (1981) uses the concept of governing category to
characterize all referential expressions (reﬂexive pronouns, regular
pronouns, and full NPs (called R-expressions in (7)) in the following

manner:

(7) Binding Principles

a. An anaphor must be bound in its governing
category.
b. A pronoun must be free in its governing category.

c An R-expression must be free.

The governing category is the minimal domain in which the reﬂexive and
its antecedent must be contained. Conversely, pronouns, another type of
anaphor, must look outside of the governing category in order to ﬁnd an
antecedent. The governing category is defined as the minimal domain
which contains the pronoun, its governor (usually the verb which
subcategorizes for the pronoun), and an accessible subject / SUBJECT.

The governing category makes a distinction between two types of

12

subjects; subject, or "little" subject, and SUBJECT, or "big" subject. Little
subject(subject ) is characterized as the noun phrase subject of a clause or
sentence. In (8) the subject of the sentence and the higher clause is Mick
Jaggar, while the subject of the lower clause is Keith Richards. The
governing category for the reﬂexive, therefore, is the lower clause because
this clause contains the reﬂexive (himself), a governor (hurt) and a subject
(Keith Richards). Because a reﬂexive must bind with an NP inside the
governing category, Keith Richards is the only possible antecedent for the

reﬂexive in this, sentence.

(8) Mick Jaggar realized [that Keith Richardsi hurt himselfi]

A subject can also be an NP subject of a containing NP. For example,
in (9), Jane is the subject of the complex NP, Jane ’5 description of herself.
Because the noun description assigns a theta role to the noun Jane, (making
it the governor) and because the noun Jane is in the specifier position of the
complex NP, NP subjects such as Jane in (10) also serve as subjects which

deﬁne the constraints of the governing category.

(9) George appreciated [Jane'si description of herselfi]

As deﬁned above, the governing category is the minimal domain
which contains the pronoun, its governor , and an accessible
subject] SUBJECT. Subject is defined as the NP subject of clauses and the
subject of NPs. SUBJECT also serves as a necessary component of the
governing category. Consider (10a-b).

13

(10a) Yo hablo espanol.
I speak Spanish.
(10b) Hablo espaﬂol.
(I) speak Spanish.

Both of these sentences are grammatical in Spanish. However, (10b),
instantiates a null subject. The subject pronoun yo is allowed to drop in
Spanish because, in part, Spanish has a rich morphological agreement
system. The agreement morphology on the verb provides the information
that would normally be conveyed by an overt subject pronoun. The
agreement features (abstractly represented as AGR) therefore assume the
nominal features of the subject. Chomsky has proposed that AGR is
therefore like a subject. English does not have such a rich inﬂectional
system and so requires the presence of an overt subject pronoun. However,
it is assumed that AGR is nonetheless abstractly represented in English.
This means that AGR, identiﬁed as SUBJECT, functions like a subject in
deﬁning the governing category. By including SUBJECT as well as subject in
the deﬁnition of governing category, the following sentences are accounted

for:

(11a) *Mick Jaggari believes that [himself*i is famous]
(11b) Mick Jaggari believes [himself] to be famous]

In (11a) the lower clause is the governing category because this clause
contains the reﬂexive (himself), a governor (is), and a subject (AGR,
expressed as [+AGR] when the clause is ﬁnite). In (11b), however, the lower

clause, because it is non-ﬁnite, is marked [-AGR] so the governing category

14

must be enlarged to include a subject or SUBJECT. In this case the subject,
Mick Jaggar, as well as the [+AGR] or -es of the ﬁnite verb believes, fulfills
this requirement.

Reﬂexive pronouns thus have certain restrictions on which NPs are
available as potential antecedents and which are not. A reﬂexive pronoun
binds with an NP antecedent in order to identify its referent. The reﬂexive
and its antecedent agrees in gender and number and the antecedent c-
commands the reﬂexive. Furthermore, a reﬂexive ﬁnds its antecedent
within its governing category. The governing category is deﬁned as the
minimal domain which contains the pronoun, its governor and a

subject / SUBJECT.

1.4.2 The Governing Category and Proper Antecedent Parameters. In
assuming that reﬂexives and their antecedents are base-generated and co-
indexed with respect to binding principle (a) and the governing category as
deﬁned above, the theory inadequately accounts for binding phenomena in
languages such as Japanese Korean, Chinese, and Turkish. These languages
have reﬂexive pronouns which bind non-locally (long distance reﬂexives,

hereafter referred to as LDRs). Consider (12) from Glew et al. (1993).

(12) Ichiroo-waj Kenji-gai zi bu n-noi, j kuruma-o tsukau to kiita.
Ichiroo-TOP Kenji-SUB self-GEN oar-ACC ‘ use that heard.
Ichiroo,‘ heard [that Kenjii will use self‘siJ‘ car.]

In (12) the reﬂexive pronoun zibun can identify either Ichiroo or
Ke n j i as its antecedent. Assuming that the governing category only includes

the reﬂexive zibun and the subject NP Ke n j i, the theory as stated above is

15

not strong enough, in that it needs to account for the grammaticality of
I chi roo as another possible antecedent for zi bun , despite the fact that it lies
outside the governing category as deﬁned above.

Much research has focused on accounting for long—distance binding in
Japanese and other languages. One approach, which has been extensively
relied upon by SLA researchers (Thomas,1989, 1991; Finer and Broselow,
1986; Finer, 1991; Lakshmanan and Teranishi, 1994; Hirakawa, 1990) is the
parameterization approach. The parameterization approach treats both local
and long-distance anaphors within the standard binding theory. The
deﬁnition of governing category is assumed to vary parametrically. For
English, the governing category is deﬁned as the minimal category which
contains the anaphor and a subject, represented as value (a) in (13). For
speakers of languages with reﬂexives set at value (b) of the parameter (e.g.
Italian 56) the governing category is the minimal domain which contains
the anaphor and INFL (the node of the phrase structure which carries
inﬂection). For speakers of Russian, whose language contains reﬂexive
pronouns set at value (c), the governing category is deﬁned as the minimal
domain containing the anaphor and a tense. Value ((1) deﬁnes the
governing category as the minimal domain containing the anaphor and a
”referential" tense (as exhibited by Icelandic sig). And ﬁnally, certain
reﬂexive pronouns in languages such as Japanese and Korean instantiate

setting (e) which deﬁnes the governing category as the root sentence.

(13) Governing Categories (Manzini and Wexler, 1987: 419)
y is a governing category for a iffy is the minimal
category which contains or and has:

a. a subject, or

16

b. an INFL, or
a tense, or
a "referential" tense

e. a "root" tense

Under this approach, reﬂexive pronouns also exhibit systematic
structural restrictions on which NPs within a given clause can serve as an
antecedent for the reﬂexive. To address this, Manzini and Wexler (1987)
and Wexler and Manzini (1987 : 431) introduced the Proper Antecedent

Parameter, shown in (14).

(14) Proper Antecedent Parameter:
a proper antecedent for or is
a. a subject 8; or

b. any element 8

This parameter has two values. The unmarked value requires that only
subject NPs may serve as an antecedent for the reﬂexive. The second, or
marked value of the Proper Antecedent Parameter, allows NPs of any
grammatical role to serve as an antecedent for the reﬂexive pronoun.

Manzini and Wexler point out that reﬂexive pronouns which
instantiate a marked value of the Governing Category Parameter (GCP)
typically instantiate the unmarked value of the Proper Antecedent
Parameter (PAP). While noting this correlation, Manzini and Wexler
maintain that the CC? and PAP are separate parameters.

A second point made by Wexler and Manzini is that the parameter

settings can vary within languages. For example, Japanese has two reﬂexive

17

pronouns with two different parametric values.1 Reﬂexive pronouns such
as ka ré-zisin, which take a local antecedent, instantiate value (a) while the

reﬂexive pronoun zi bun instantiate value (e).

1.4.3 The acquisition of reﬂexive pronouns: the subset principle. Berwick
(1985) outlines the Subset Principle as a model of language acquisition. The
Subset Principle stipulates that where one language is a proper subset of
another, the learner assumes the smallest language compatible with the
input received so far. In this way, the learner can enlarge the linguistic
system when the positive evidence presents itself. By moving from a
smaller grammar to a larger one, a learner can acquire a particular linguistic
element by using positive evidence alone. The Subset Principle, then, is a
learning principle which guarantees that a child rely on positive evidence
alone in acquiring a linguistic element.

Manzini and Wexler (1987) interpret the Subset Principle as a model
of language acquisition which guides the learner in ﬁnding the correct value
of a particular parameter. In this case, the Governing Category and Proper
Antecedent Parameters are used as examples. Manzini and Wexler state
that the values of the Governing Category and Proper Antecedent
Parameters are subsets of one another. Figures 3 and 4 show the subset
relationships for reﬂexive pronouns as interpreted by Manzini and Wexler
(1987).

Given this relationship, Manzini and Wexler (1987) suggest that
learners follow the Subset Principle in acquiring the correct value of the

GCP and PAP. Early on in the acquisition process, the learner assumes the

 

1 For an extensive discussion of the different types of Japanese reﬂexives
from a different perspective, see Katada (1993)

18

 

Figure 3: Subset relationships for the governing category parameter

 

Figure 4: Subset relationships for the proper antecedent parameter

19

unmarked, (a) value of the GCP and PAP. As the learner begins to make
sense of the input, the grammar is enlarged to value b and then to value c,
and so on until the grammar reﬂects the parametric value of the target

language.

1.4.4 Criticisms of the GCP and PAP. Recently, the GCP and PAP have
fallen out of favor among both syntax and Second Language Acquisition
researchers. Thomas (1995) points out that researchers have criticized the
parameterization approach for basing the parametric variation on lexical
items. For example, Japanese has several types of reﬂexive pronouns.
Onesuch pronoun, the morphologically simplex reﬂexive, zi bu n, would
have value e of the GCP and value a of the PAP. On the other hand, the
morphologically comples reﬂexive pronoun karé-zi si n would have value a
of the GCP and value b of the PAP and the reﬂexive pronoun zibun-zisi n
would have value c of the GCP and value a of the PAP. Thus, three
different reﬂexive pronouns within the same language (in this case,
Japanese) instantiate three different parametric values. By tying parametric
variation to lexical items within each given language, this approach allows
for many different parameter values within one language. Consequently,
the parameter merely describes the state of affairs in a given language rather
than predicting systematic, cross linguistic variation. This is a problem for a
parameter-based model. '

Thomas (1995) also points out that the parametric values of the GCP
are arbitrary. There is no principled, theoretical motivation for the
distinctions. Because the parametric values lack theoretical motivation,
there is no constraint on the limits of parametric variation. Subsequently,
this portion of the model is also merely a description rather than a predictor

20

of cross linguistic variation.

The Subset Principle as a guiding model of language acquisition has
also been challenged (Kapur et al., 1993). The most striking problem with
the model is that child language acquisition studies show that children do
not, in fact, assume the (a) value of the GCP and subsequently enlarge their
grammars. On the contrary, studies by Read & Hare (1979), Mathei (1981),
Wexler 6r Chien (1987), Chien 8r Wexler (1990), Hyams 5r Sigurjénsddttir
(1988, 1990), Sigurjdnsdéttir 6r Hyams (1991, 1992), and McDaniel, Cairns
&Hsu (1990) provide both direct and indirect evidence that children of
various ﬁrst languages initially go through a phase of long-distance binding
before discovering the local nature of complex, XP reﬂexives. This suggests
that learners initially assume the larger grammar and then later adopt the
smaller. Additionally, Franks and Connell (1996) argue that the acquisition
models for the GCP and PAP assume that the learner is capable of parsing
complex sentences in order to set the parameter values. Lightfoot (1989,
1991) makes the important point that children do not rely on complex
sentences to set parameter values.

Thus, due to the fundamental theoretical problems with the GCP and
PAP, as well as the above mentioned problems with the Subset Principle,
the parameterized approach to reﬂexive pronoun acquisition has fallen out

of favor among both ﬁrst and second language linguists.

1.5 Raising approaches
Yet another approach, which had been developing independently,
suggests that linguistic variation in anaphoric binding relates to the
movement of anaphors at the level of Logical Form (LF) (Lebeaux, 1983;
Pica, 1987; Batistella, 1987; Cole et al., 1990; Katada, 1993). This approach

21

assumes that reﬂexive pronouns raise at the level of LF. Logical Form is the
level of the language in. which syntax and interpretation come together. It is
the place where meaning is structurally represented. It therefore makes
sense that the reﬂexive pronoun would go looking for its antecedent and,
ultimately, its referent, at the interpretive, LF, level of the language.
Lebeaux (1983) was the ﬁrst to suggest the possibility that anaphors

raised at the level of Logical Form. He observed that English reﬂexive
pronouns, but not reciprocals, could only occur in positions that could be
properly governed; that is, in positions which are either theta-governed or
could be antecedent-governed. This observation led Lebeaux to posit that
English reﬂexives raise at LF and leave behind a trace which must be
properly governed.

Pica (1987) and Batistella (1989) and, later, Cole et al. (1990) and Katada
(1993) developed this idea into an explanation of long—distance vs. local-only
binding across languages. Local vs. long-distance binding, these researchers
argue, hinges on the morphological nature of the reﬂexive pronoun.
Morphologically complex reﬂexive pronouns, such as the English h i ms e If;

are considered maximal projections (Xmax

)with the following structure:
[NPlspech i m][N' selﬂl. Because maximal projections such as the NP
reﬂexive pronoun can only adjoin to other maximal projections, the
reﬂexive pronoun is only able to raise as far as the lower VP. If the
reﬂexive pronoun does not raise at all, it is within the C-command domain
of both the subject and object NPs. Once the XP reﬂexive has adjoined to
the lower VP, however, it is within the C-command domain of the subject
NP only. In this way, complex reﬂexives can take either a subject or non-
subject NP as a possible antecedent.

Morphologically simple reﬂexive pronouns such as the Chinese zi j i,

22

however, are minimal projections, referred to as X0 pronouns. Because X0
reﬂexives are heads, they must go through head-to-head movement into
the INFL node of the phrase structure. This movement iterates, and so the
reﬂexive pronoun can move successively from the lower [NFL to the higher
INFL and therefore identify any of a number of potential NP antecedents.
This ability of the X0 reﬂexive pronoun to successively raise to higher
clauses is what allows the pronoun to ﬁnd a non-local antecedent. Once the
XO reﬂexive pronoun has raised, however, only the subject NPs C-command
the reﬂexive pronoun. It follows that once the reﬂexive pronoun has raised
out of its surface level site, the reﬂexive pronoun can only have a subject
orientation. This movement pattern is shown in Cole et al. (1991: 6) and

repeated in Figure 5.

1.6 Relatavized SUBJECT analysis

The relatavized SUBJECT analysis, proposed by Progovac and Franks
(1992) and Progovac (1992) also rests on the assumption that the
morphological complexity of the reﬂexive pronoun determines the nature
of its binding patterns.

Where in the raising approach the constraints on the binding domain
and orientation are the structural consequence of movement patterns, the
relatavized SUBJECT analysis suggests that anaphor raising is not necessary.
The different binding patterns exhibited by complex (XP) and simplex (X0)
reﬂexives are due to binding domains of different sizes.

As discussed earlier, Chomsky (1981) argues that reﬂexives and their
antecedents must both be contained in the minimal category containing the
reﬂexive and a SUBJECT. Progovac argues that only certain elements of
SUBJECT, as deﬁned by Chomsky, are relevant for each reflexive-type.

Cpl
/ \
s C
pec / \
C [P
1/ l\
NP 1'
I \
Zhangsan 11 VP1
I \
ziji V CPZ
I I \
renwi Spec C
"think"

C2
l

tlfll

23

I \

1P2
I \
NP 1'
I
12 VPz
I I \
t'” V CP3

r I \
zhidao Spec C

I\‘
C3 1P3
I I \
t" NP 1'
I I\
Wangwu
t' V NP

xihuan N
"like" I

Zhangsan ziji renwei Lisi zhiado Wangwu xihuan.

Zhangsan self thinks Lisi knows Wangwu likes.

Zhangsan thinks that Lisi knows him.

Figure 5: Phrase structure tree of LF raising of Chinese anaphor, 21' j i
(Adapted from Cole et al. (1991: 6))

24

As stated previously, SUBJECT (hereafter taken to mean

subject] SUBJECT) includes NP subjects of clauses] sentences and complex
NPs as well as ﬁnite AGR. In terms of X-bar levels, there are two types of
subjects included in this deﬁnition. NP subjects of clauses and complex
noun phrases are XP subjects, whereas ﬁnite AGR is an X0 category.
Progovac argues that NP subjects are the only relevant SUBJECT for XP
reﬂexives. The binding domain for an XP reﬂexive, then, is the minimal
category containing the reﬂexive and an XP subject (subject of a clause or
complex NP).

For X0 reﬂexives, the only relevant SUBJECT is the x0 category AGR.
This being the case, the binding domain for the X0 reﬂexive can vary,
depending on whether or not AGR is overt or null in the various clauses of
the sentence.

Russian, for example, is like English in that it has overt AGR in ﬁnite
clauses but null-AGR in non-finite complement clauses. Unlike English,
however, Russian has XO reﬂexives. Progovac argues that null-Agr is
automatically coindexed with the next higher AGR, so when an X0 reﬂexive
is used in an embedded, non-ﬁnite clause, the XO reﬂexive is bound by AGR
which, because it is null, is coindexed with the higher, overt AGR. The
result of this is that the binding domain is expanded to include the clause
containing the X0 reﬂexive as well as the higher, overt AGR containing
clause.

Languages such as Chinese, Japanese, and Korean do not have overt
AGR; therefore, the binding domain for X0 reﬂexives in these languages is
relatively free.

25

1.7 The argument structure perspective

1.7.1 SE-Anaphors. Reinhart and Reuland (1991) suggest that the different
ranges of possible antecedents for reﬂexive pronouns are reducible to certain
important differences in the syntactic and argument structure of the
pronouns.

Simplex expressions, or SE-anaphors (thus far referred to as X0 or
monomorphemic reﬂexive pronouns), they point out, are similar to XP
reﬂexives (referred to as SELF-anaphors) in that they are both "referentially
defective." That is, reﬂexive pronouns do not in and of themselves
carrylexical meaning. The lexical content is picked up when the reﬂexive
links
with an NP. The range of possible antecedents, however, differs between an
SE-anaphor and a SELF-anaphor.

The reason for this difference in possible antecedents relates to the
different structures of the two anaphors. The SE-anaphor has a structure
identical to pronouns. That is, SE-anaphors and pronouns occur in

determiner position as shown in (15) (Reinhart and Reuland, 1991: 286).

(15) [NP] DP Deti [N' . . . N(xi) . . . ]]

SE-anaphors, Reinhart and Reuland suggest, are different from pronouns in
that they lack ¢-features The SE-anaphor must get o-features from a
position which is a head (X0 category) like itself, which c-commands the
anaphor, and carries o-features. The only category to satisfy these
requirements is AGR Reinhart and Reuland argue that both ﬁnite and
non-ﬁnite INFL are associated with AGR, but only when ﬁnite INFL is

associated with AGR is AGR anaphoric. Thus, SE-anaphors associate with

26

anaphoric AGR and, because AGR is coindexed with the subject, SE-
anaphors have a subject orientation.

The domain of the SE-anaphor is the distance in which the SE-
marked verb-INFL complex can raise at LF. Reinhart and Reuland point out
that, among other language speciﬁc constraints, tensed INFL is the opacity
factor, blocking such movement. So the constraints on the movement of
SE-anaphors is what limits the domain. The predictions of this model are
roughly equivalent to those predicted in the Relatavized Subject analysis, as
the Relatavized Subject Analysis relies on "overt AGR" as the opacity factor.

1.7.2 Self-Anaphors. SELF-anaphors, on the other hand, have a different
structure and so behave in a different manner. As we know already, SELF-
anaphors are considered full noun phrases where SELF occurs in N '
position and a pronoun determiner (such as him, her, them, etc.) occupies
the speciﬁer position. Reinhart and Reuland argue that SELF operates like a
relational noun. That is, SELF subcategorizes for two arguments, one of
which is satisﬁed by the pronominal determiner. The process is shown in
(16).

(16) [NP him [N ' SELF <Y, X>l

The SELF-anaphor, therefore, is defective in that it is missing one of its
arguments. One way in which the second argument is assigned, is via a
reﬂexive predicate. A reﬂexive predicate, according to Reinhart and
Reuland, is a predicate in which two of its arguments are identical. In
English, these two identical arguments are the reﬂexive and its antecedent.
The SELF noun, when it occurs in a subcategorized position of the predicate,

27

reﬂexivizes, or reﬂexive marks, the predicate. The result of this, then, is
that the SELF-anaphor and its co—argument of the predicate are the two
identical arguments on the reﬂexive predicate's o-grid. A SELF-anaphor,
when used to reﬂexive mark a predicate, is in complementary distribution
with pronouns and SE-anaphors Furthermore, because the SELF-anaphor
reﬂexivizes the predicate, it can only share its identity with a co-argument of
the predicate, thus giving rise to the local domain of SELF-anaphors.

1.7.3 Logophoricity. Up to this point, the discussion of SE and SELF-
anaphors has related speciﬁcally to structural binding. Reinhart and
Reuland (1991) suggest that SE and SELF-pronouns can also function as
logophors, or "discourse bound" reﬂexives.

Logophoric binding is a process in which the logophoric pronoun
ﬁnds its antecedent via discoursal cues. According to Clements (1975: 141),
the logophor assigns as its antecedent the noun "whose speech, droughts,
feelings, or general state of consciousness are reported." Because logophoric
binding is an establishment of coreference rather than structural binding, it
is not subject to the binding conditions.

Reinhart and Reuland point out that although logophoric binding is
not subject to the binding conditions, there are certain constraints placed on
the context in which logophoric binding can take place. That is, SELF-
anaphors can only be bound logophorically when they do not occupy an
argument position of a predicate. The following examples from Reinhart
and Reuland (1991: 312) show the effects of this constraint.

(17) a. Luciei boasted that the chairman invited her husband
and herselfi for a drink.

28

b. * Luciei boasted that the chairman invited herselfi for
a drink.

In (17a) the non-local NP is an acceptable antecedent, whereas in 14b the
non-local NP is unacceptable. This is because in order for a SELF-anaphor to
ﬁnd a non-local antecedent, it cannot do so through structural binding, but
must bind logophorically. As mentioned previously, in order to be bound
logophorically the SELF-anaphor must occupy a non-argument position.
For if the SELF-anaphor occupies an argument position, the predicate is
reﬂexivized and the anaphor must identify another argument of the
predicate.

The examples in (17) bear out these predictions. In (17a) the SELF-
anaphor is in a non-argument position and so is free to bind logophorically
with the non-local antecedent. In (17b), however, the SELF-anaphor is in a
subcategorized position of the verb invite. Because of this, the SELF-
anaphor reﬂexive marks the predicate thus requiring that the external
argument the chairman be the other identical argument (is the
antecedent).

This approach also bears on distributional issues. As (18)
demonstrates, when a SELF-anaphor is used to reﬂexivize a predicate, it is in

complementary distribution with pronouns.
(18) Mick Jaggari hurt him* i/ himself].

When a SELF-anaphor is used logophorically, however, complementary
distribution does not obtain, as demonstrated in (19).

29

(19) Luciei boasted that the chairman invited her husband
and heri / herselfi for a drink.

The purpose of the Governing Category and Proper Antecedent
Parameters, the raising approach, the relatavized SUBJECT analysis and the
argument structure perspective is, among other things, to account for the
binding and co-reference patterns of both short distance reﬂexives (SDRs),
and long distance reﬂexives (LDRs). These approaches provide descriptions
of the structural details of languages which allow a reﬂexive pronoun
embedded in a subordinate clause to refer to any of several possible

antecedents in the local or higher clauses.

1.8 Chapter summary

This chapter discussed the current models which have attempted to
account for reﬂexive] antecedent relations across languages. After an initial
discussion of universal characteristics of the reﬂexive] antecedent
relationship such as the structural and semantic linking of reﬂexive with
antecedent and, ultimately, referent as well as the c-command requirement,
the chapter outlines different models which have attempted to describe the
behavior of reﬂexive pronouns. The parameterized approach, with its
governing category and proper antecedent parameters was summarized ﬁrst,
followed by discussions of the raising approach, relatavized SUBJECT
analysis, and argument structure perspective.

Second Language Acquisition (SLA) researchers, in their search to
explain how adults come to understand the rules and usage patterns of
reﬂexive pronouns, have appealed to each of these models at various points

in time. Thus, having outlined the major components of these analyses,

30

Chapter Two will address the results and speciﬁc explanations proposed for
the second language acquisition of reﬂexive pronouns. While the studies
discussed in the next chapter have not been designed to test the primacy of
one linguistic analysis or another, each study frames its discussion of the
results in terms of one current analysis. Therefore, the analyses outlined in
this chapter will provide a frame of reference from which the SLA studies
can be evaluated.

 

Chapter 2

THE SECOND LANGUAGE
ACQUISITION OF ENGLISH REFLEXIVES

2.1 Introduction

This Chapter considers recent work on the SLA of English
reﬂexives. First of all, I consider the question of ultimate attainment. Do
adult learners acquire English reﬂexive pronouns? The studies I review
suggest that learners do, ultimately, acquire the target-like structure for
English reﬂexives. From there, however, opinions diverge and several
explanations have been provided as to the process by which learners
eventually acquire English reﬂexives. As I summarize and evaluate each
study I suggest that despite certain limitations of much of the research
(e.g., reliance on tasks which may tap preferred intuitions on possible
reﬂexive] antecedent relationships rather than a complete range of
learner intuitions), certain important observations have been made on
the interlanguage grammars of English reﬂexives.

Several studies have shown that learners of various ﬁrst language
backgrounds (e.g., French, Spanish, Japanese, Korean) allow both local
and non-local antecedents for the reﬂexive. Additionally, there is
evidence that some learners appear to differentiate between tensed and
non-tensed embedded clauses in their grammars for reﬂexives. These
learners allow local or non-local antecedents for the reﬂexive when it is

embedded in a non-ﬁnite clause, but when the reﬂexive is embedded in a

31

32

ﬁnite clause, these learners reject long-distance antecedents and only
allow local antecedents.

Researchers have provided several possible explanations for this
behavior by L2 learners of English, most of which assume that the ﬁrst
language background of the learner somehow plays a role in the
acquisition process, either by way of structural transfer of either the
simplex or complex anaphor from the ﬁrst language or through transfer
of a pragmatic strategy from the ﬁrst language. I argue that these
explanations, while sufﬁciently accounting for the data in the particular .
study under discussion, inadequately account for the behavior] intuitions
expressed by learners in other studies.

One account of the acquisition process which is compatible with
the results of SL studies of reﬂexives comes from the child language
acquisition literature. I raise the possibility that second language learners
of English do not necessarily transfer reﬂexive structures from their ﬁrst
languages; but rather, these learners may progress along a path of
acquisition similar to that described in the child language acquisition
study by Franks and Connell (1996).

2.2 Ultimate attainment: Do adult learners acquire English reﬂexive
pronouns?
2.2.1 Learners From Various L1 Backgrounds Acquire English Reﬂexive
Pronouns. Several studies of the acquisition of reﬂexive pronouns have
produced data that suggest that learners ultimately succeed in learning
the clausemate condition of English reﬂexives.
Thomas (1991) investigates intuitions regarding reﬂexive
pronouns for 132 learners of English and 41 learners of Japanese. For this

study, Thomas relies on two tasks. The ﬁrst is an elicitation task where

33

learners repeat sentences which contain anaphors and pronouns. The
second task is an acceptability task where learners are asked about the
possible interpretations of reﬂexive pronouns. About this task, Thomas

states,

Subjects read a battery of sentences and responded to a
question following each one by identifying the referent of the
pronoun or anaphor. For example, they indicated whether
the reﬂexive in (20) refers to Mary, Sue, some other person
not mentioned in the sentence, or any combination of Mary,
Sue, or some other person (Thomas, 1991: 227).

(20) Mary heard that Sue told the doctor about herself
(Thomas, 1991: 227).

The various test sentences evaluate learners' intuitions regarding
domain (local/ non-local NPs as possible antecedents), orientation (+/ -
subject as possible antecedents) and C-command. The results show that
70-90.5% of learners across groups consistently select the local NP as the
antecedent for the English reﬂexive pronoun. These results suggest that
learners acquire target-like intuitions for English reﬂexive pronouns
quickly and easily.

Matsumura (1994) in a study of 110 Japanese native speakers
learning English, likewise reports that the English reﬂexive pronoun is
acquired quickly and easily. This study is a replication of Hirakawa (1990)
which will be discussed later in this chapter. After administering a cloze
test to break the subjects into proﬁciency levels and a pre-test to check that
the subjects' English proﬁciency was high enough, Matsumura
administered an acceptability task to evaluate judgments on binding
patterns. This task involved presenting subjects with sentences

34

containing a reﬂexive and a list of possible interpretations from which to

choose. An example is given in (21) below (Matsumura, 1994: 20).

(21) Bob told Steve not to blame himself.

. Bob
. Steve

1
2
3.
4
5

Either Bob or Steve

. someone else

. don't know

The results show that English native speakers chose the local NP as the

antecedent almost 100% of the time. The mean score for the high

proﬁciency group was slightly lower than for the native English speakers,

yet higher than the mean score for the low proﬁciency group, suggesting a

pattern where learners approach native-like intuitions as proﬁciency

levels increase.

Eckman (1994) focused on the binding patterns of individual

learners and found that many learners acquire the target-like form for

English reﬂexive pronouns. This study relied on a picture identiﬁcation

task where subjects examined two pictures with a sentence containing a

reﬂexive underneath.

The subjects were required to identify which

picture (or both pictures, or neither picture) the sentence accurately

described. The results for this study, reported in terms of individual

grammars, revealed that out of the 25 learners of English (mixed Lls and

proﬁciency levels) who participated in the study, 16 of them consistently
bound the reﬂexive to a local NP. The patterns exhibited by these 16

learners reﬂect those expressed by the majority of the ENS control group.

35

2.2.2 Methodological Problems: Tapping Preferences Rather than
Complete Intuitions. The results of the studies thus far presented suggest
that L2 learners of various ﬁrst language backgrounds acquire the local
constraints of English binding fairly quickly and easily. In this section I
discuss the possibility that the experimental tasks used in these studies
failed to tap the complete intuitions of the subjects. It is only by
examining later work which takes steps to encourage learners to consider
all possible antecedents for the reﬂexive, does it become apparent that the
acquisition of reﬂexives is more complicated than ﬁrst thought.

As I have mentioned, some of the older studies tended to elicit
preferred interpretations of the reﬂexive rather than a complete range of
interpretations. Whether a picture task (Eckman, 1994) or a sentence
interpretation task (Thomas, 1991; Matsumura, 1994), these studies asked
subjects to come up with all possible readings of the sentence themselves.
Eckman (1994) asked subjects to indicate one, both, or neither picture as
showing the meaning of the sentence above. Thomas (1991) and
Matsumura (1994) asked subjects to indicate if one NP or another, or both
NPs could be the antecedent of the reﬂexive. By relying on such a task,
the responsibility for carefully considering all possible interpretations of
the sentence fell on the person completing the task. If a person had a
strong preference for one particular interpretation of the sentence, it is
quite possible (likely, even) that the person would never consider a
second, less likely, but perfectly acceptable interpretation.

White et al. (1997) discuss this point at length, arguing that by
relying on data collected through tasks which tap only preferred rather
than complete intuitions, researchers have an incomplete picture of the
learner's system. White et al. point out that these preferences can be quite

strong, even for English native speakers when they consider sentences

36

where the reﬂexive can refer to either a subject or object NP. As I have
discussed in Chapter One, it is perfectly acceptable for an English reﬂexive
to refer to either the subject or object NP in a local domain. White et al.

write,

English NSs usually assume that the sentence's subject is the
antecedent of the reﬂexive; they will recognize the object
interpretation if it is pointed out to them or if the context favours it.
In some cases, however, their preference may be so strong that it
overrides perception of any other interpretation . . . Even so, the fact
that they choose only one interpretation does not necessarily mean
that the other is excluded from their grammar (White et al., 1997:
148).

The results of the studies we have been discussing indicate a
potential problem with preferences. Thomas (1991) reports a strong
preference among native English speakers for the subject NP. This study
ﬁnds that only 41.3% of the responses by English native speakers are for
either the subject or non—subject NP as an acceptable antecedent for the
reﬂexive pronoun, whereas 54% of the responses were for the subject NP
only.

The Eckman (1994) study also produces data suggesting a problem
with preferences. Here too, while the ENS group consistently binds the
reﬂexive in the local domain, 15 out of the 25 subjects bind the reﬂexive
to the subject NP only. Only six subjects accept either the subject or the
non-subject as possible antecedents, while four subjects were inconsistent
in their response patterns.

White et al. (1997) discuss other studies which show evidence of a
"subject strategy;" that is, in these studies, L2 learners also indicate the
subject as the only possible antecedent. Studies by Finer (1991) Finer &
Broselow (1986), and Hirakawa (1991), all report data indicating a subject

37

strategy and all rely on some form of a multiple choice or picture
identiﬁcation task where study participants are responsible for coming up
with all possible interpretations for the antecedent.

If a particular task is obviously failing to tap the complete
intuitions of native English speakers, it is equally likely, if not more so,
that the task is failing to tap complete intuitions of the learners. The key
element, according to White et al. (1997) is pragmatic context. In the
absence of pragmatic context, both native speakers and ESL learners will
rely on the preferred interpretation. Pragmatic context, however, can
help encourage speakers to consider other, non-preferred possible
antecedents. By manipulating pragmatic cues so they point to a particular
reading, speakers can consider one possibility at a time and also give due
consideration to each possible interpretation of the sentence containing

the reﬂexive pronoun.

2.2.3 Learners from Various L1 Backgrounds Bind English Reﬂexives
Long-Distance. Given this methodological problem of preferences, it is
not surprising that other studies which use pragmatic or semantic context
to help subjects consider all possible interpretations of the reﬂexive, ﬁnd
that ESL learners of various L1 backgrounds do not acquire English
reﬂexives as quickly and easily as the previously discussed studies
indicate. Studies by Thomas, (1989), Christie (1992) Demirci (1997 ), and
White et al. (1997) show that learners of English frequently accept long
distance antecedents for reﬂexive pronouns.

Thomas (1989), examines the binding patterns of Spanish and
Chinese native speakers who were learning English as a second language.
This study relies on a multiple choice task which is susceptible to tapping

preferred rather than complete intuitions. However, this study includes

38

sentences which provide pragmatic context to help the learner consider
other, possibly less preferred, possible interpretations. Interestingly, in
the neutral, bi-clausal sentences, the rate of local binding patterns for
Spanish speaking learners of English is about 60% and for Chinese
speaking learners of English, 69%. When pragmatic bias is directed
toward the non-local NP, these numbers drop to 38% and 34%,
respectively.

The "subject strategy" (preference for the subject NP over object NP
in the local domain) is also reduced with contextual cues. In neutral
sentences, close to 60% of subjects' responses (Spanish and Chinese
speaking learners of English) were for the subject NP as the only possible
antecedent for the reﬂexive. When pragmatic context is added to help the
learners consider the object NP, the percentage of "subject-only" type
responses drops to around 19% for Spanish learners of English and 29%
for Chinese learners of English. These results suggest strongly that a lack
of pragmatic context contributes to the problem of preferences overriding
complete syntactic judgments.

Christie (1992) studies reﬂexive interpretations by Spanish and
Chinese native speakers learning English as well as two groups of English
native speakers who were learning Spanish or Chinese. Learners are
presented with a picture task in which one picture and a sentence with a
reﬂexive are contained on a page. The study participant is then asked to
indicate whether or not the sentence reﬂects the message of the picture.
The results of this study indicate that both Spanish and Chinese speakers
learning English accept considerable long-distance binding.

Demirci (1997) also reports extensive LD binding among Turkish
speaking learners of English. Using an acceptability task in which subjects

are asked to read a sentence and indicate who the reﬂexive can not refer

39

to, Demirci presents subjects with both neutral and pragmatically biased
test sentences. The results show that Turkish speakers learning English
will accept a non-local antecedent more readily than English native
speaker controls when pragmatic bias is directed toward the non-local NP.

White et al. (1997) investigate the issue of methodology directly by
comparing the judgments expressed by Japanese and French speaking
learners of English who complete a picture task and a story task. The
story task involves asking subjects to read a short story, then read a
sentence underneath the story and indicate whether or not the sentence is
true given the story. An example is given below in (22) (White et al. 1997:
153):

(22) Susan wanted a job in a hospital. A nurse interviewed Susan for
the job. The nurse asked Susan about her experience, her
education and whether she got on well with people.

The nurse asked Susan about herself. True] False

For the picture task, each subject is given a booklet where one picture
with a sentence underneath it ﬁlls each page. The subject is asked to
identify whether or not the sentence matches the picture.

The results of this study show that the story task better elicits
acceptance of the object NP as a possible antecedent for the reﬂexive. Both
the ENS group and ESL learner groups are more likely to accept the non-
subject NP as a possible antecedent for the story task than for the picture
task. For the bi-clausal sentences, both the ESL groups accept signiﬁcantly
more long-distance antecedents than the ENS control groups.

It appears, then, that studies which better tap complete grammatical
intuitions suggest that L2 learners of English (whose Lls are of either type

 

40

- LDR/ SDR or SDR only) show a preference for the local NP in neutral
sentences, but will consider long distance antecedents when the pragmatic
context suggests it. Adult English native speakers, on the other hand,
have a much stronger allegiance to the local NP.

2.2.4 Learners accept more long-distance binding into non-ﬁnite
subordinate clauses than into finite subordinate clauses. Regardless of
methodology, many studies have found that some learners are more
likely to accept long-distance antecedents when the reﬂexive is embedded
in a non-ﬁnite subordinate clause than when the reﬂexive is embedded
in a ﬁnite subordinate clause (Finer and Broselow, 1986; Hirakawa, 1990;
Broselow and Finer, 1991 ; Progovac and Connell, 1991; Bennett and
Progovac, 1993; Bennett, 1994; Matsumura, 1994; Demirci, 1997; White et
al., 1997).

The ﬁrst study to discover this pattern was a pilot study conducted
by Finer and Broselow (1986). Six Korean speaking learners of English
were given a picture identiﬁcation task in which they examined two
pictures with a sentence underneath them. The learners were to indicate
whether the sentence described the ﬁrst picture, the second picture or
both pictures. The sentences tested learners' interpretations of both
reﬂexive and regular pronouns in simple and complex sentences. For the
complex sentences, two types were presented to the learners - sentences in
which the reﬂexive / regular pronoun was embedded in a non-ﬁnite
clause and complex sentences in which the reﬂexive] regular pronoun
was embedded in a ﬁnite clause.

Interestingly, the ESL learners chose a local antecedent for the
reﬂexive 91.7% of the time when the reﬂexive was embedded in a
complex sentence with a ﬁnite subordinate clause. When the subordinate

41

clause of the complex sentence was non-ﬁnite, however, that number
dropped to 58.3%. The results of this pilot study suggest that Korean
learners of English are apparently more likely to bind a reﬂexive long-
distance when the reﬂexive is embedded in a non-ﬁnite subordinate
clause than when it is embedded in a ﬁnite clause.

Hirakawa (1990) reports similar results. After a preliminary test to
screen the learners for the most basic understanding of basic grammar,
vocabulary, and elementary facts about reﬂexives and pronouns, 65
Japanese FL learners of English in four proﬁciency groups were given a
binding test. This test took the form of a acceptability task in which
learners were to read a sentence containing a reﬂexive and then circle the
chosen interpretation from a list of choices below. An example is given

in (23) (Hirakawa, 1990: 70).

(23) John said that Bill hit himself.
a. John
b. Bill
c. either John or Bill
d. someone else

e. don't know

Hirakawa ﬁnds no grade effects in this study; that is, accuracy on the task
does not improve with proﬁciency level. As a result, the four levels were
collapsed and evaluated as a group. The results of this evaluation shows
that the Japanese learners of English perform signiﬁcantly better on
complex sentences with ﬁnite subordinate clauses than on complex
sentences with non-ﬁnite subordinate clauses. These ﬁndings support

those of Finer and Broselow (1986) in that learners are more likely to bind

 

42

long-distance when the reﬂexive is embedded in a non-ﬁnite clause than
when it is embedded in a ﬁnite clause. Hirakawa points out, however,
that despite this difference, learners made errors in all sentence types.

Broselow and Finer (1991) report results of binding patterns of
Hindi, Japanese, and Korean speaking learners of English. Using a picture
task like the one used in Finer and Broselow (1986), this study ﬁnds once
again that Korean and Japanese native speakers bind the reﬂexive with a
non-local antecedent more often when the reﬂexive is embedded in a
non-ﬁnite clause than when it is embedded in a ﬁnite clause.
Interestingly, the Hindi speakers learning English did not exhibit this
pattern. Broselow and Finer suggest that these speakers grew up in a
socio-linguistic context in which English is prevalent. Perhaps, they point
out, these learners already understood this part of the English system.

Bennett and Progovac (1993) and Bennett (1994) also ﬁnd that
learners differentiate between ﬁnite and non-ﬁnite embedded clauses . In
these studies of Serbo-Croatian learners of English, learners accept long-
distance binding into object control inﬁnitive clauses more often than
tensed embedded clauses. This ﬁnding is particularly interesting because
of the fact that the Serbo-Croatian spoken by the subjects has no object
control infinitives.

Matsumura (1994) in a replication study of Hirakawa (1990) was
mainly interested in changes in accuracy across proﬁciency levels, so
while she breaks the sentence types down into those in which binding
between reﬂexive and antecedent may occur across ﬁnite embedded
clauses (Types A and B) and sentences in which binding may occur across
non-ﬁnite embedded clauses (Types C and D), she does not comment on
possible differences in accuracy. Her data suggest this possibility, but it is
not clear. The relevant results are summarized in Table 1 below. Type A

43

sentences are biclausal sentences with the reﬂexive embedded in a ﬁnite
clause (e.g., Helen knows that Alice blames herself] (Matsumura, 1994:
41)). Type C sentences are biclausal sentences with the reﬂexive
embedded in a non-ﬁnite clause (e.g., Bob told Steve not to blame
himself. (Matsumura, 1994: 41). Responses by subjects were analyzed in
two ways, those indicating the local NP as the only possible antecedent
(correct) and all other responses (incorrect). The experimental task
included four type A sentences and three type C sentence. Thus, the type
A mean scores are reported on a scale of four and the type C sentences are
reported on a scale of three. Both mean scores and their percentage
equivalents are given in Table 1.

There is a difference in mean scores between types A and C
sentences at both the low (L) and high (H) proﬁciency level. ESL learners
more accurately bind the reﬂexive locally when it is embedded in a ﬁnite
clause than when it is embedded in a non-ﬁnite clause. Thus,
Matsumura's results support previous ﬁndings that learners appear to be
more likely to consider LID binding when the reﬂexive is in a non-ﬁnite
context.

Demirci (1997 ) also reports that the Turkish speaking ESL subjects
in her study differentiate between ﬁnite and non-ﬁnite embedded clauses.
These learners were signiﬁcantly more likely to bind a reﬂexive long-
distance when it was embedded in a non-ﬁnite clause than when it was
embedded in a ﬁnite clause.

White et al. (1997) investigate learner judgments of biclausal

sentences with ﬁnite and non-ﬁnite subordinate clauses. Where the

Table 1: Mean scores and standard deviations for Japanese speaking
learners of English (Adapted from Matsumura (1994: 29)).

 

 

 

 

 

 

 

 

 

Low Proﬁciency High Proﬁciency t-score Control

Group (L) Group (H)

(n=24) (n=56) btwn (n=15)

groups

M SD M SD t M SD
Type A 2.75 0.97 3.23 0.93 2.08* 4.00 0.00
(n=4) 69% 81% 100%
Type C 1.54 0.76 2.02 0.52 3.20“ 2.93 0.25
(n=3) 52% 67% 98%

 

 

 

 

*p < .05. “p < .01.

 

45

context of the story points-to the local NP as the more likely antecedent,
the ESL groups show no difference in judgment between ﬁnite and non-
ﬁnite sentences, although the ESL subjects were less accurate than the
ENS control group. For the picture task, the Japanese group bound the
reﬂexive locally more than for the story task, yet performed less accurately
than the ENS group overall.

When the pragmatic context suggests a non-local NP as the more
likely antecedent (pragmatically speaking), differences in judgments for
finite and non-ﬁnite clauses become more evident. For the picture task,
both ESL groups bound the reﬂexive long-distance in both finite and non-
ﬁnite embedded clauses. For the story task, however, the Japanese group
was signiﬁcantly more accurate on the ﬁnite sentences than on the non-
finite.

Thus, White et al. found that only the Japanese learners of English
differentiated between finite and non-ﬁnite embedded clauses and then
only when the pragmatic context suggested the non-local NP as a more
likely antecedent.

A number of studies have therefore observed a difference in
binding patterns between reﬂexives embedded in a ﬁnite clause and
reﬂexives embedded in a non-ﬁnite clause. These ﬁndings are, however,
the result of a small pilot study (Finer and Broselow, 1986), and studies
done with methodologically problematic tasks (Matsumura, 1994;
Hirakawa, 1990; Finer, 1991 ; Bennett and Progovac, 1993; Bennett, 1994).
The White et al. (1997) study, while better methodologically, found this
difference only in very limited circumstances. These issues, along with
limitations regarding how the data have thus far been reported will be
taken up later in this Chapter.

46

2.3 Possible explanations for the acquisition of English reﬂexive
pronouns

2.3.1 Explanation #1: Learners Transfer Part or all of Their Reﬂexive
System from the L1 and Apply it to English.

2.3.1.1 Transfer XP. Researchers Have Proposed Several Possible
Explanations for the Acquisition Process of English Reﬂexive Pronouns.
The most commonly used explanation is transfer. Yuan (1994) suggests
that adult learners of English simply transfer the morphologically
complex reﬂexive from their L1 and apply it to the L2. Yuan states:

It is therefore likely that when Chinese, Japanese, and Korean
learners of English encounter English reﬂexives in the input
data, they recognize them as phrasal reﬂexives through the
form of PRONOUN + self. As a result, they assume that
English reﬂexives behave like the phrasal reﬂexives in their
L1s, which is a correct assumption. On the basis of this
assumption, the learners coindex English reﬂexives only with
local antecedents and reject sentences that allow long-distance
binding (Yuan, 1994: 542).

Yuan puts it even more strongly in his 1992 paper where he states,

Quite a few SLA researchers have used English reﬂexives to
test the resetting of the governing category parameter by
native speakers of Chinese, Japanese and Korean . . . any
claims for the availability of UG and parameter resetting in
SLA on the basis of results obtained, from these tests are
illegitimate because, obviously, reﬂexives in English share the
same setting as phrasal reﬂexives in Chinese and no
parameter resetting is involved at all (Yuan, 1992).

If, indeed, learners simply transfer the value of the
morphologically complex reﬂexive from their L1 and apply it to the

English reﬂexive pronoun as Yuan suggests, certain learner behaviors

47

would result. The ESL learner would interpret the English reﬂexive
pronoun in the same manner that they would interpret the
morphologically complex reﬂexive pronoun in their L1. That is, the
learner would assume that the reﬂexive pronoun must ﬁnd its
antecedent within its own clause. As pointed out in the previous section,
numerous studies (Thomas 1989, White et al. 1997; Christie, 1992;
Demirci, 1997) have found that learners of English, no matter what their
L1 type (LDR/ SDR or SDR only) accept long distance antecedents
extensively. If learners were transferring the SDR-value from the ﬁrst
language and applying it to English, these learners would, as Yang
predicts, . . reject sentences that allow long-distance binding." This
prediction we know from the earlier discussion, is not borne out.
Learners of various L1 backgrounds consistently accept long-distance

antecedents in the English L2.

2.3.1.2 Transfer XO- Given that learners apparently do not transfer the
morphologically complex value of the reﬂexive pronoun, Lakshmanan
and Teranishi (1994) suggest that for Japanese native speakers, learners
could inappropriately transfer the value of the morphologically simple
reﬂexive pronoun (in this case, zibun) from the L1 and interpret the
English reﬂexive as such in the L2. Eventually, through a process
Lakshmanan and Teranishi refer to as inter-lingual identification, the
learners realize that English reﬂexives are not, in fact, morphologically
simple reﬂexive pronouns like zibu n, but, rather, function just like the
Japanese morphologically complex reﬂexive, ka ré-zi sin. The learner then
discards the zibun-type interpretation of the English reﬂexive and
replaces it with a karé-zisin-type interpretation.

48

While this proposal accounts for the ﬁnding that Japanese learners
of English initially accept long-distance antecedents, certain predictions of
learner intuitions given this assumption do not bear out. If learners are,
in fact, transferring the LDR value from the L1 and applying it to the L2,
there would be a difference in behavior between learners whose Lls
contain LDRs and learners whose Lls do not. Presumably, only speakers
with LDRs in their ﬁrst language can transfer the LDR value and go
through a long-distance phase of acquisition. This proposal, therefore,
predicts that one type of language learner (someone whose L1 contains
both LDRs and SDRs) will initially accept long distance antecedents, while
another type of language learner (someone whose L1 contains only SDRs)
will not accept long distance antecedents. Once again, acquisition data
(Thomas 1989; White et al. 1997; Christie, 1992) show that learners from
SDR-only Lls frequently accept long distance antecedents.

A second problem with the structural transfer proposals relates to
learner interpretation of ﬁnite and non-ﬁnite embedded clauses. Several
studies (Finer and Broselow, 1986; Broselow 8r Finer, 1991 ; Finer, 1991;
Matsumura, 1994; Demirci, 1997 and White et al., 1997) have found that
many learners of English go through a period where they are more likely
to accept a long-distance antecedent when the reﬂexive pronoun is
embedded in a non-ﬁnite clause than when it is embedded in a ﬁnite
clause. If learners are, in fact, transferring one value of a reﬂexive
pronoun from the ﬁrst language and applying it to the second language,
the learners would not differentiate between ﬁnite and non-ﬁnite
embedded clauses when no such distinction exists in the L1. The learners
who participated in the studies cited above exhibit intuitions which
suggest that they are operating from a binding system which differentiates
between reﬂexives in ﬁnite and non-ﬁnite embedded clauses. This

49

distinction is made despite the fact that no such distinction is made in
either the L1 or the L2.

Bennett and Progovac (1993) and Bennett (1994) (see also White,
1996) attempt to resolve this issue by appealing to the Relatavized
SUBJECT Analysis. They suggest the possibility that learners from
LDR] SDR-type Lls transfer the value of the X0 reﬂexive and correctly
identify the overt AGR in ﬁnite embedded clauses in English. The result
of this is that the learners, thinking that English reﬂexives are LDR, allow
both local and non-local antecedents for the reﬂexive in non-ﬁnite (i.e.,
non-AGR) contexts. In the finite context, however, learners recognize the
overt AGR and thus require local-only binding.

Once again, assuming that the learner transfers the value of the X0
reﬂexive from the ﬁrst language accounts for only half the problem. That
is, the transfer-based explanation only accounts for the long—distance
binding patterns by learners whose Lls contain LDRs. This transfer-based
explanation of LD-binding among learners of English fails to account for
the equally strong tendency of learners from SDR-only Lls to accept long-
distance antecedents.

Yoshikawa (1993) investigated the relationship between the
acquisition of AGR and LD binding in L2 learners. This study measured
the rate of errors in subject-verb agreement in a composition exercise as
well as the binding patterns of L2 English learners through a multiple
choice task. Yoshikawa found that learners whose knowledge of subject-
verb agreement (is AGR) was perfect accepted LD antecedents just as
often as learners who made many mistakes on subject-verb agreement.

Furthermore, it is possible that the ﬁnite] non-ﬁnite distinction
made by L2 learners of English can be explained in another way. I will
argue later in this chapter that the data collected so far may have been

50

misinterpreted due to the reporting methods used in these studies. It is
possible there is another, better, explanation for the ﬁnite] non-finite
distinction exhibited by L2 learners of English, which I will take up at the
end of this chapter.

Structural transfer (of either the morphologically simple reﬂexive
pronoun or the morphologically complex reﬂexive pronoun) is therefore
not a viable explanation for the acquisition of English reﬂexives. The fact
that speakers of both language types (LDR/ SDR and SDR only) accept
long-distance antecedents to a signiﬁcant degree and the fact that learners
accept signiﬁcantly more long distance antecedents when a reﬂexive is .
embedded in a non-ﬁnite clause as compared to a ﬁnite clause, suggests
that learners of English as a second language are not transferring the
value of a particular reﬂexive pronoun from their ﬁrst language and
applying it to English.

2.3.1.3 Pragmatic Strategy Transfer. In addition to the possibility of
structural transfer, Demirci (1997) suggests that Turkish learners of
English, whose L1 contains both complex and simplex reﬂexives,
innapropriately transfer a stronger sensitivity to pragmatic cues onto
English. The study used a grammaticality interpretation task that
evaluated the degree to which subjects responded to pragmatic bias.
Sentences were divided into neutral sentences, sentences biased toward
local NPs, and sentences biased toward non-local NPs. Furthermore,
sentence types included sentences with both tensed and non-tensed
embedded clauses. Demirci reports that Turkish native speakers are
highly responsive to pragmatic cues in considering possible antecedents
for Turkish reﬂexives. The English native speaking group, however, did
not respond as strongly to pragmatic bias in their intuitions on possible

51

reﬂexive] antecedent relationships. Demirci ﬁnds that Turkish speakers
learning English continue to exhibit a sensitivity toward (and willingness
to respond to) pragmatic cues. She argues that the learners have
transferred their pragmatic strategy from Turkish and applied it to their
L2, English. She further suggests the possibility that this strategy
fossilizes, as she found that even with students who had been studying
English for several years continued to apply this pragmatic strategy.

A study by Glew, Demirci, Yamagata, and Cass (1993) reports a
similar sensitivity to pragmatic cues by both Turkish and Japanese native
speakers. This study used a task similar to that in Demirci (1997),
investigated the intuitions of English native speakers learning Turkish
and Japanese. The study also included data from three control groups:
Turkish native speakers (providing intuitions on Turkish reﬂexives),
Japanese native speakers (providing intuitions on Japanese reﬂexives)
and English native speakers (providing intuitions on English reﬂexives).
Glew. et al. (1993) found that the Turkish and Japanese native speaker
control groups were most strongly inﬂuenced by pragmatic cues in
choosing possible antecedents for a reﬂexive. In some cases, the native
speakers denied the possibility of co-reference with an NP against which
pragmatic bias was directed. This behavior often occurred even if the NP
against which pragmatic bias was directed was (structurally) a perfectly
acceptable antecedent for the reﬂexive.

Similar to the ﬁndings of Demirci (1997), Glew et al. report that the
intuitions of English native speakers are much less susceptible to
pragmatic bias than the Turkish or Japanese speakers.1 The English

 

1 It should be noted, however, that both the Demirci (1997) and Glew et al.
(1994) studies report that over 20% of English native speaker responses
indicated acceptance of the long-distance NP as an antecedent for the
reﬂexive. These results show a much stronger sensitivity to pragmatic

52

native speakers who were studying Turkish and Japanese, however,
exhibited intuitions somewhere in the middle. These subjects failed to
respond as strongly to the pragmatic bias as the Turkish and Japanese
native speakers did. The English native speakers who were studying
Turkish and Japanese did, however, show more sensitivity to pragmatic
bias than the ENS control group did.

The results of these studies suggest that native speakers of
languages with both complex and simplex anaphors, such as Japanese and
Turkish, are much more sensitive to pragmatic information for
antecedent assignment of reﬂexive pronouns than native speakers of
languages such as English which only have complex anaphors.
Furthermore, Demirci (1997) suggests that these native speakers continue
to be sensitive to pragmatic cues when learning English as a foreign
language.

In order to consider the possibility of transfer, both structural
transfer and transfer of a processing cue such as pragmatic meaning, it is
necessary to consider recent research within the Competition Model, a
psycholinguistic model of language acquisition and processing.

The Competition Model contrasts sharply with many nativist
theories of language acquisition. While special nativist theories (e.g.
U.G.) are developed on the assumption that the "core grammar" is
acquired by a separate language component, the Competition Model was
developed with the assumption that language learning takes place
entirely as a result of the application of more general learning

mechanisms . In other words, special nativist researchers argue that

 

cues by English native speakers than other studies (White et. al. , 199x;
Thomas, 1989) which include "pragmatic context" to help learners
consider all possible antecedents.

53

language acquisition is driven by inborn mechanisms and those
espousing the Competition Model argue that language is learned entirely
from experience.

While these two philosophies represent two sides of the classic
nature/nurture continuum, researchers are beginning to explore possible
models which allow for both task speciﬁc and general learning
mechanisms in the language acquisition process. Researchers arguing for
a general nativist proposal for language acquisition (O'Grady, 1996;
Eckman, 1996; Wolf—Quintero; 1996) suggest that language acquisition is
the result of the application of both task speciﬁc and general learning
mechanisms.

While there continues to be much discussion and debate over
theories of L1 grammar, Eckman (1996) cautions that a theoretical
grammar of the L1, such as UG, is not necessarily the guiding system of
interlanguages. Interlanguages have grammars, but one must not
assume that IL grammars are necessarily the same as the L1 grammar.

Eckman states,

All that one can conclude from the fact that ILs have
grammars is that any theory of SLA must include a theory of
IL grammar. And, most important, it is an open, empirical
question whether a theory of IL grammar is the same as a
theory of L1 grammar. Thus, it is conceivable that L1
grammars and IL grammars are governed by different
theories. One possible conception of this situation is that L1
grammars obey the constraints of the theory of grammar,
namely, UG, while IL grammars obey the constraints
imposed by some other theory, e. g., general learning. In
other words, under the general nativist position for SLA, UG
could be the theory of L1 grammars, and general cognition
could be the theory of L2 grammars. It is an empirical matter,
which, for the time being, at least, must remain open
(Eckman, 1996: 406).

54

There are several tasks the learner faces when acquiring a second
language. While special nativist approaches account for the contribution
of the mental grammar to the acquisition task, it provides little in the way
of explanation of how the learner uses the mental grammar in
conjunction (or in competition) with the various levels of linguistic
representation (e. g., phonological, morphological, semantic, pragmatic,
social, etc.) present in the input.

Thus, while proponents of the Competition Model make speciﬁc
claims about how the L1 is acquired, those claims do not necessarily have
to be assumed in order to consider the possibility that L2 acquirers assign
various values to the different linguistic elements in the input. In other
words, regardless of how (i.e., through mathematical weightings of the
various cue strengths, through adjusting the value of the cues to
accommodate the strengths and weaknesses of the various components of
the target grammar, as a consequence of the development of the IL, etc.)
the language learner comes to balance the relative value of the various
linguistic components, the Competition Model provides a framework
whereby it is possible to explore what various cue values at different
phases of the acquisition process.

The Competition Model is a model that recognizes the fact that
learners (and language users in general) are faced with the challenge of
processing several linguistic levels simultaneously in order to interpret
the meaning of the language. The listener] speaker must coordinate
phonological, syntactic, semantic, and pragmatic information
simultaneously as the speech is being produced. Naturally, there are
limits on the attention one can devote to each of these various levels.
Because of these limitations, a language user must learn to pay close
attention to those levels that provide the best information regarding the

55

interpretation of the sentence. In so doing, the speaker can focus on the
important forms and pay less attention to the less important forms.

As a person learns a given language, different forms show
themselves to be more effective than others. This is referred to as cue
strength. Cue strength is deﬁned as "the probability or weight that the
organism attaches to a given piece of information relative to some goal or
meaning with which it is associated" (Bates a MacWhinney, 1987: 64. cf.
Heinleman Gr McDonald, 1993). In other words, cue strength refers to
how effective a given cue (function) is for a language user.

In most cases, several cues will converge on a certain function for a
given noun. In (24), for example, Joh n is clearly the "doer" of the action.
We know this because several cues, namely, word order (the ﬁrst noun is
typically the agent] subject) and animacy (animate nouns are the more
likely than inanimate nouns to be agents or "doers" of actions), converge

on J o h n as the subject] agent of the action.
(24) John built the house.

But, as (25) shows, cues do not always have to converge. In (25), word
order tells us that the sandwich is the subject] agent of the sentence.

Semantics, however, tells us that J o h n is the subject] agent because
animate beings typically eat things.

(25) The sandwich ate John for lunch.

It is through exposure to sentences in which cues conﬂict, that learners
eventually sort out the weightings of the different cues for a given
language. The cue that "wins" the conﬂict most often, or has the stronger

56

conﬂict validity, becomes stronger and more relied upon in the overall
processing of the given language.

The various weightings of the cues, however, vary from language
to language. Word order, for example, tends to be relied upon most
heavily by English speakers. This means that in sentences like (25),
English speakers would interpret the ﬁrst NP, the sandwich, as the agent
or "doer" of the action. They do this despite the fact that the semantic cue
of animacy points to the second NP, Joh n. In English, then, the syntactic
cue of word order is more strongly relied upon than the semantic cue.

Italian native speakers, however, place more weight on the
semantic cue. Therefore, if a sentence like (25) were translated to Italian
and the dominant cue, agreement marking was controlled for, John
would be the more likely choice of agent] subject because animacy has a
stronger conﬂict validity than word order.

While no Competition Model research has yet been done on the
relationship between pragmatic and syntactic cues for linking the
reﬂexive and antecedent, research has been done on the relationship
between syntactic and semantic] pragmatic cues as they relate to thematic
role assignment.

Gass (1985) conducted a study of 111 learners of English in a second
language environment. The learners completed a grammaticality
interpretation task which involved interpreting sentences which varied
according to semantic verb type (ask/promise/tell ) and also in terms of
the subject and object NP's place on the topicality hierarchy
(Human>Animate > Inanimate). By manipulating these factors, the
study examined how native speakers and ESL learners responded to
situations where semantic and syntactic information converged and

competed.

57

Gass found that when learners were presented sentences in which
semantic and syntactic cues converged, they readily provided the correct
English interpretation. When the cues competed, however, beginning
learners were more sensitive to the semantic cues (animacy in Gass'
framework) than the syntactic ones. At higher proﬁciency levels, learners
learned to accept syntax as the dominant cue for English. Furthermore,
she found that pragmatic cues dominate over both syntactic and semantic
cues when the token sentence indicates an unlikely situation such as (26).
The dog is the ﬁrst NP (indicating that it is the agent) in the sentence and
is also an animate noun (animate things are the most likely candidates
for agenthood, inanimate things are not). However, because the sentence
depicts an unlikely (pragmatically implausible) scenario, most learners
indicated confusion on this item.

(26) The dog told the chair to go (Gass 1985: 25).

When the sentence represented a scenario where pragmatic and syntactic
cues converged, learners performed similarly across proﬁciency levels.

In the case of (27), however, both semantic and syntactic cues converge.

(27) The m a 71 told the dog to go (Gass 1985: 25).

The results of this study suggest that pragmatic cues are extremely
important for learners of English. Where the pragmatic and syntactic
cues converge on one particular interpretation, learners overwhelmingly
respond to the cues and choose that interpretation. When the pragmatic
information is confusing (such as for sentences such as (26), which
indicate unlikely situations), learners do not appear to fall back on either

58

the semantic or syntactic reading, but indicate confusion. Without a
reasonable pragmatic context, therefore, learners do not have conﬁdence
to rely on the other levels of the linguistic system.

Gass ( 1987) continued investigating of the relationship between
semantic and syntactic cues in her examination of native speakers,
foreign and second language learners of Italian and English. In this study,
Gass argued that for the assignment of thematic roles, Italian native
speakers (INS) rely more heavily on the semantic cue of animacy than
English native speakers, whose dominant strategy is a syntactic one, word
order. She examined how Italian native speakers must move from
relying on the semantic cue of animacy to a language such as English,
whose native speakers rely more heavily on the syntactic cue of word
order. The results suggest that the INS learners of English initially
become aware of the importance of word order, only moving to reﬂect the
intuitions of native speakers in the case of classic noun-verb-noun
(NVN) word orders. This ﬁnding, Gass argues, suggests that learners ﬁrst
"get a feel" for a certain part of the L2 grammar before they move on to
learn the speciﬁc factors which determine the characteristics of that part of
the L2. In this study then, the INS learners of English continue to rely on
the L1 dominant one of animacy, but have become aware of the
importance of word order. The speciﬁc characteristics of this syntactic
strategy have not yet been ﬁne tuned.

English native speakers learning Italian, however, acquired the
semantic cue strategy of animacy more easily than the INS learners of
English did for word order. Gass suggests that moving from a
syntactically based system, such a English, to a semantically based one,

such as Italian, is somehow easier than vice versa.

59

Based on her ﬁndings, Gass argues that the semantic strategy is
stronger than the syntactic. This, she suggests, is supported by the fact that
the ENS learners of Italian responded to the animacy cue much more
readily and with greater similarity to INS data than the INS learners of
English did for word order. She concludes that ENS learners of Italian
acquire the native—like animacy strategy, whereas the INS learners of
English move much more slowly and with more difﬁculty, ﬁrst becoming
aware of the importance of word order and then moving on to begin to
learn the exact characteristics of English word order.

Other researchers (Harrington, 1987; Sasaki, 1991, 1994; Wulfeck et .
al., 1986) have supported Gass (1987) in ﬁnding that learners more readily
adopt L2 animacy strategies. Sasaki (1991) found that ENS learners of
Japanese acquired the animacy strategy for Japanese, whereas Harrington
( 1987) showed that Japanese native speakers learning English continued
to rely on their animacy strategy, failing to adopt the L2 (English) strategy
of word order. Wulfeck et al. (1986) report that among Spanish] English
learners and native speakers, the animacy strategy is a "strategy of last
resort." The study found that subjects relied on animacy cues only when
the "preferred cue" was missing (for group I this was word order, for
group II this was agreement marking).

The Competition Model research on the acquisition of theta-role
assignment shows that learners who come from an L1 for which semantic
cues are strong, continue to rely heavily on the semantic cue in the L2.
They are also less willing to let go of the semantic strategy in favor of the
L2 dominant syntactic strategy.

These ﬁndings bear a striking resemblance to those reported in the
Competition Model studies (Gass, 1985, 1987; Harrington, 1987; Sasaki,
1991, 1994; Wulfeck et al. ,1986). Learners whose L1 requires a heavy

60

reliance on pragmatic and semantic cues tend to "hang on" to the
pragmatic strategy and move more slowly toward a reliance on syntactic
cues than vice versa. Likewise, Glew et al. (1993) and Demirci report that
native speakers of languages such as Turkish and Japanese (languages
with both simplex and complex reﬂexives) rely heavily on

pragmatic] semantic cues and move to a stronger reliance on syntactic
cues reluctantly, if at all.

Thus, the studies by Demirci (1997) and Glew et al. (1993) suggest
that in addition to providing a structural account of the developing
grammar of an L2 learner, it is also important to consider the possibility
that a pragmatic strategy plays a role in the interlanguage system of

second language learners.

2.3.2 Explanation #2: The L2 Acquisition Process of English Reﬂexives is
a Developmental One Similar to the Acquisition Process of Children
Learning English
2.3.2.1 Child Language Acquisition of Reﬂexive Pronouns. If learners are
not transferring the L1 value and applying it to the L2, then some sort of
developmental process, independent of the learner's ﬁrst language
background, may be at the heart of the acquisition of English reﬂexives.
One possibility is that second language learners follow a process similar to
that of English monolingual children. Franks and Connell (1996) report
on a study of twenty adults, 13 normal children (age range: 3;7 to 8;6), and
11 Speciﬁc Language Impaired (SLI) children (age range 3:9 to 7;8) in order
to investigate the process by which native English speakers acquire
reﬂexive pronouns.

The study relies on a Video Multi—Interpretation Task (VMIT).
This task was developed speciﬁcally to address the problem of tapping the

61

preferred antecedent rather than the complete range of possible
antecedents. The VMTT provides study participants with the pragmatic
context essential to tapping a speaker] learner's complete intuitions. The
VMIT involves a series of 13 "video-taped vignettes" each of 30 seconds
in length. Each vignette involves a scene of puppets having a juice party
where the participants take turns pouring juice into each other's glasses.
Each vignette is followed by questions designed to elicit the subjects'
interpretation of reﬂexive pronouns. The test questions include three
questions evaluating whether subjects understand the question-answer
format of the test, ﬁve questions to test subjects' binding domain of
reﬂexives, two orientation questions to determine whether subjects accept
subject and / or non-subject antecedents, and three C-command questions
to determine if subjects accept the reﬂexive pronoun as an anaphor. (28)
provides an example question where the subjects' intuitions regarding
the domain and orientation of the reﬂexive. In this case (Franks and
Connell 1996: 443), the child sees a vignette where Mickey is pouring juice
into Ernie's glass. The child is then asked,

(28) Did Bugs ask Ernie if Mickey poured juice for himself?

The correct adult, native speaker answer would be n 0 because Mickey did
not pour juice for himself, Mickey poured the juice for Ernie (h i m). A yes
answer to this question would indicate that the child] adult is accepting a
long-distance object as an antecedent for the reﬂexive pronoun.

The subjects participating in the Franks and Connell study
completed the VMTT several times until the answers for each of the
questions were consistent. Franks and Connell then evaluated the

response patterns of each individual learner to identify the domain and

62

orientation of the reﬂexive for each subject. The response patterns were
compared across sentences to develop a complete picture of the domain
and orientation of the reﬂexive for each learner. Response patterns fell
into four distinct patterns. One learner was unable to exhibit an
understanding of C-command (through the experimental task (answering
yes to questions 11 and 12 and answering n o to question 13) as well as a
follow-up task involving an act out protocol where the subject once again
failed to demonstrate a clear awareness of the C-command requirement).
Franks and Connell conclude that this learner has not yet learned that the
reﬂexive pronoun is an anaphor in need of an antecedent. This ﬁnding
reﬂect studies by Wexler and Chien (1987), Chien and Wexler (1990),
McDaniel, Cairns, and Hsu (1990) which also identify children who fail to
respect the c-command requirement. Likewise, these researchers suggest
that learners who display this behavior are not yet treating the reﬂexive
as an anaphor. Instead, these learners interpret the reﬂexive as a lexical
NP with a meaning something equivalent to body.

A second group of learners, the "nearest NP group" (n=7), only
accepted the local NP as a possible antecedent. This was demonstrated
through their responses across sentences. Table 2 shows the experimental
questions and predicted adult responses for each item. These seven
subjects, then, gave "correct" answers for items 1;1, 2:2, 3;2, 4;3, 5;3, 6;4, 7;4,
8;4, 11:6, 12;6, and 13;6. However, for the subject orientation questions,
the learners accepted the interpretation of Bugs (the object NP) as the
antecedent for the reﬂexive in vignette test sentence 10;5, but rejected
Ernie as a possible antecedent in 9;5. This group, Franks and Connell
suggest, is at one of the earliest phases of the acquisition process. They
suggest that once the child recognizes that the reﬂexive is an anaphor
(evidenced by C-command effects), 8] he selects the nearest NP as the

63

antecedent, thus assigning the local object NP as the antecedent for the
reﬂexive.

Another response-type pattern exhibited in the Franks and Connell
study was that displayed by the "LD group." Individuals falling into this
group exhibited the following characteristics (Franks and Connell, 1996:
448):

(a) When the reﬂexive was embedded in a ﬁnite clause, individual

subjects in this group allowed both SD and LD antecedents, but

only with subject NPs.

(b) When the reﬂexive was embedded in a non-ﬁnite clause,

individual subjects in this group allowed only the LD subject as the

antecedent.
Franks and Connell suggest that members of this group have realized
(and require) that the reﬂexive raise at LF, but that these language
learners do not recognize PRO as an acceptable binder. The result of this
is that the reﬂexive (interpreted by the child as LDR) requires a subject
antecedent, but cannot accept the local subject, PRO, as an acceptable
antecedent in non-ﬁnite embedded clauses. When the reﬂexive is
embedded in a ﬁnite clause, either the local or the non-local subject NPs
is an acceptable antecedent. Franks and Connell suggest that the reason
learners go through this LD phase is that they are not yet able to lexically
parse the reﬂexive pronoun into its component parts. Learners who have
not yet reached a stage where they can lexically parse the complex
reﬂexive pronoun (a bimorphemic, XP reﬂexive such as h i m self) see the
XP reﬂexive as a simplex, monomorphemic reﬂexive such as Japanese
zi bu n. Assuming that the children mistakenly analyze the English
reﬂexive as monomorphemic, they allow long—distance antecedents for
the reﬂexive. The children also, for some unexplained reason, do not

Table 2: The experimental questions associated with the 13 vignettes of

the VMIT. Adapted from Franks and Connell (1996: 443)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

J ] JRecipient or Adult

Sentence object depicted English

Vignette number Questions on tape response

2 Format uestions '

1 1 Is Bu Ernie's friend? Yes

2 2 Did Ernie ask Bugs to get Juice N o
the lasses?

3 2 Did Ernie ask Bugs to get Glasses Yes?
the glasses?

Domain uestions: nonfinite

4 3 Did Ernie ask Bugs to pour Ernie No
juice for himself? _

F5ﬁ 3 Did Ernie ask Bugs to pour Bugs Yes
juice for himself? ﬂ
Pomain questions ﬁnite _

6 4 Did Bugs ask Ernie if Bugs No
Mickey poured Bugs juice
for himself?

7 4 Did Bugs ask Ernie if ﬁnite No
Mickey poured Ernie juice
for himself?

8 4 DTd Bugs ask Ernie if Mickey Yes
Mickey poured Mickey
juice for himself?

.. Su 'ect orientation questions

9 5 Did Ernie ask Bugs about Ernie Yes
himself?

10 5 DTd Ernie ask Bugs about Bugs Yes
himself?
Cfcommand questions

11 6 Did the mouse who said ‘Efrnie No
'Hi Ernie' pour juice for
himself?

12 7 Did Ernie's friend Bugs me N 0

ur 'uice for himself?

13 7 Did Ernie's friend Bugs Eu—gs Yes

pour juice for himself?

 

 

65

consider PRO an acceptable binder.

The ﬁnal group of learners for this study is referred to as the "Local
Group." This group exhibited intuitions in line with the English target
system; that is, individuals in this group disallowed LD binding but
allowed_either subject or object NPs as possible antecedents for the
reﬂexive in simple clauses. This response pattern, Franks and Connell
argue, is reﬂective of the ﬁnal stage in the acquisition process. At this
time, learners realize that English reﬂexive pronouns are
morphologically complex and, thus, cannot ﬁnd an acceptable antecedent
outside of the embedded clause. Because the reﬂexive is now considered .
to be morphologically complex, the learner also realizes that it can refer to

either the subject or non-subject in the local domain.

2.3.2.2 SLA of English Reﬂexives: Reconsidering the Data. Given this
model of child language acquisition, it is important to return to the
second language acquisition data and see if they provide any support for
the acquisition stages identiﬁed in Franks and Connell (1996). Table 3
outlines the developmental stages in Franks and Connell (1996).

In the existing literature, there is little evidence either for or
against the existence of a "nearest NP" stage among second language
learners of English. This may due, in part, to the task problems of earlier
studies. Many of the early studies reported results in terms of
aggregateresponse patterns and did not give results in terms of individual
grammars (e.g. Bennett 6: Progovac, 1993; Bennett, 1994; Matsumura,
1994; Thomas 1989). By reporting data in terms of "percentage of
responses directed toward NPl, NP2, etc." or the mean scores of learners
on an acceptability task where the "right" answer is the local NP
(Matsumura, 1994), it is impossible to identify individual interlanguage

66

systems. It is impossible to tell if any subjects accept the local object to the

exclusion of the local subject without examining individual grammars.

Learner intuitions which reﬂect the "nearest NP" stage are often

Table 3: The acquisition stages of reﬂexive pronouns for English

monolingual children as identiﬁed in Franks and Connell (1996:
458)

1. The reﬂexive pronoun is interpreted as a noun (independent

reference)

2. The reﬂexive pronoun is interpreted as an anaphor

a. Nearest NP phase: The learner realizes that the reﬂexive
pronoun is an anaphor in need of an antecedent, but does not yet
understand that the anaphor can raise at LF. The learner therefore
selects the nearest NP (the local object) as the antecedent for the
reﬂexive.
b. LD phase: The learner understands and requires that the
reﬂexive raise at LF, resulting in an obligatory subject orientation
for the reﬂexive pronoun. The learner does not understand,
however, that PRO is an acceptable binder. This results in the
following behavior:

(i) When the reﬂexive pronoun is embedded in a ﬁnite

clause, either the local or the non-local subject is an

acceptable antecedent.

(ii) When the reﬂexive pronoun is embedded in a non-ﬁnite

clause, only the non-local subject is an acceptable antecedent.
c. Local Phase: The learner understands that the reﬂexive can
optionally raise and that PRO is an acceptable binder. Furthermore,
the learner is linguistically mature enough to parse the reﬂexive

67

pronoun and understands that it is not a morphologically simple
pronoun but, rather, is a morphologically complex reﬂexive
pronoun which cannot raise outside of the embedded clause. This
gives rise to the English system: local domain, subject or non-

subject orientation.

lumped together with intuitions from learners who may be at the target-
like "local" phase. Aggregate data reporting like this may elicit general
trends in the acquisition process, but if one is interested in identifying
subtle changes in the learners interlanguage system, this method is too
blunt an instrument.

While recent studies have begun to report data on individual
grammars (Eckman, 1994; Thomas, 1991, 1995; White et al. 1997), the way
in which the data are evaluated and reported makes it impossible to get a
clear picture of interlanguage systems. The acquisition stages identiﬁed in
Franks and Connell (1996) have several important characteristics. It is not
enough to say, "this learner accepts local NPs as possible antecedents."
This behavior is only part of the interlanguage system, it is in fact a
characteristic of two vastly different systems. A "local-only"
interpretation could indicate a learner at the very ﬁrst stage of acquisition,
the "super-local phase," where the learner only accepts the local object as
an acceptable antecedent. Or, a "local-only" interpretation could indicate
a learner who has acquired native-like competence for English reﬂexives,
in which case this person would accept either subject or non-subject NPs,
but only in the local domain.

The problem with the studies by White et al. (1997), Thomas (1991),
and Eckman (1994) is that in reporting individual grammars, these
studies only examine one feature of the interlanguage system at a time.

68

For example, White et al. (1997) report on the percentage of subjects who
consistently accept binding to subject NPs, then they report the percentage
of subjects consistently accepting binding to object NPs. Then the subjects
are thrown back into the pool and the percentage of subjects who accept
local and] or non-local NPs are reported. With these results, we know
how many test subjects bind the reﬂexive with the object NP as well as
how many test subjects bind the reﬂexive exclusively in the local domain;
however, we do not know how many subjects bind the reﬂexive to the
object NP and simultaneously bind the reﬂexive exclusively in the local
domain.

While these studies are providing more detailed information
about learners' interlanguage systems than the aggregate percentage
reporting of previous studies, they are still falling short in that they do
not examine the entire interlanguage systems. Instead, they focus on one
feature of the system at a time, masking the possible ﬁner features of
different interlanguage systems.

Even with the recent addition of data on individual grammars, the
way in which SL data are reported and analyzed prevents us from
identifying a "nearest NP" grammar among SL learners of English.

Given the data so far provided on the acquisition of English reﬂexives, it
is impossible to tell whether or not a "nearest NP" stage exists in the
interlanguage of any ESL learner.

SLA data do, however, provide vague support for stages 2b and 2c.
As discussed earlier, many studies have found that learners differentiate
between reﬂexive pronouns embedded in a ﬁnite clause as compared to a
non-ﬁnite clause (Finer and Broselow, 1986; Broselow Gr Finer, 1991 ;
Finer, 1991; Matsumura, 1994; Demirci, 1997; White et al., 1997, Bennett,
1994)]. These studies have shown that learners are more likely to accept a

69

long-distance antecedent when the English reﬂexive is embedded in a
non-ﬁnite clause than when it is embedded in a ﬁnite clause. The
ﬁndings of these studies differ slightly from those of Franks and Connell
(1996) in that they do not identify the exclusively non—local domain for
reﬂexive pronouns embedded in non-ﬁnite clauses that English
monolingual learners exhibit.

In addition to the task and data reporting problems I have just
discussed, the ﬁnite ] non-ﬁnite distinction made by second language
learners may not have been adequately measured because of problems with
proﬁciency grouping. Many SLA studies of reﬂexives (White et al. 1997;
Demirci, 1997; Matsumura, 1994; Hirakawa, 1990; Thomas 1991; Bennett 61:
Progovac, 1993; Bennett, 1994) stratify subjects into levels based on global
proﬁciency levels. In assuming that global proﬁciency levels correspond
neatly with proﬁciency in reﬂexive binding, researchers may inadvertently
be mixing subjects who are at different levels in the acquisition process of
reﬂexive pronouns. Franks and Connell point out that for monolingual
English learners, some individuals may stay at a particular level for a long
time, suggesting that ages and acquisition stages do not correlate.
Furthermore, Franks and Connell suggest that as learners move from one
phase to another, the old and new phases can exist simultaneously and

compete with one another. They state:

We propose, instead, that earlier grammars coexist in the
mind with later ones. They are, in a sense, in competition . . .
Thus, we propose that the child does not immediately
abandon G“ as soon as Gm] becomes available, but rather has
decreasing recourse to it as the subsequent grammar
strengthens. This competition may persist for a long time; . . .
(Franks and Connell, 1996; 460). '

70

For L2 learners, we can extrapolate and assume that the potential
mismatch between global proﬁciency level and acquisition stage for
reﬂexive pronouns is even stronger. Given that children progress
through acquisition stages at variable rates and also operate with two
competing systems simultaneously, it is equally likely that L2 learners
also progress at variable rates and operate with two competing
interlanguage systems. In fact, it is likely that the L2 acquisition process is
even more variable. This of course, is due to the many factors
inﬂuencing L2 acquisition such as motivation, instruction, immersion in
the L2 environment, and input.

Some studies on reﬂexive pronoun acquisition rely on data from
subjects in a foreign language setting (White, 1997; Demirci, 1997;
Progovac Gr Bennett, 1993; Bennett, 1994, Wakabayashi, 1996; Hirakawa,
1990; Matsumura, 1994). When considering intuitions on reﬂexive
binding, it is important that a distinction be drawn between second
language learners (who have exposure to the language outside the
classroom setting) and foreign language learners. It is commonly
assumed that the properties of reﬂexive binding are not taught in the
classroom. Thomas (1991: 217) states that L2 learners of English and
Japanese are not taught the rules of reﬂexive binding.

No subject in this experiment claimed to have been explicitly
taught how to determine the antecedent of a reﬂexive. An
informal survey of teachers of English as a second language
found no one who instructs students in this detail. I
inspected a dozen popular English textbooks but found at best
brief or vague treatment of the interpretation of reﬂexives.
Example sentences, if any, contain only one potential
antecedent. Four Japanese language textbooks likewise
yielded no evidence that learners are instructed in the
relevant constraints of zibu n '

(Thomas, 1991: 217).

71

Even in terms of exposure, it is not at all clear that reﬂexive
pronouns have a common place in classroom input. References to
oneself or reﬂexive actions of others are more typical of personal
interchange rather than the language typical of the classroom. A study by
Parker, Heitzman, Fjerstad, Babbs, and Cohen (1995) suggests that even in
Full Language Immersion Programs (FLlPs), language learners do not use
the L2 for personal interchange. Parker et al. investigated the language
use patterns of 32 pupils in a Spanish-language full immersion program
(Grades 3-6) and found that students used Spanish 70-88% of the time in
teacher-fronted situations, whereas for soda] and small-group situations,
students used Spanish only 22-27% of the time. These results suggest that
for situations in which personal interchange occurs, students
overwhelmingly rely on their L1. It is assumed that in a foreign language
classroom, even less time is spent involved in interpersonal, social
exchange in the L2.

Students in a foreign language setting also have little to no
exposure to the L2 outside of the classroom. White et al. (1997) point this
out in the description of their FL subjects. They state,

We gave the learners a background questionnaire to
determine their previous experience with English. All had
learned English in a classroom setting, with a small amount
of additional exposure; even the francophones reported at
most only sporadic contact with English outside of the school
context . . . Neither group reported having spent much time
in English-speaking provinces or countries: an average of 1.5
months for the French-speakers, and 2 months for the
Japanese speakers

White et al. 1997: 150).

The result of this, therefore, is that while foreign language learners

have a middle or high overall proﬁciency level, this measure may not

72

indicate one's proﬁciency for reﬂexive binding. Because classroom input
is impoverished for reﬂexives, and FL learners have little to no
opportunity for interpersonal exchange either in or out of the classroom,
these learners may not have had an opportunity to advance beyond the
most basic proﬁciency levels for reﬂexive binding.

Many proﬁciency groups, therefore, that are used by researchers to
stratify learners as they acquire reﬂexive pronouns may not, in fact, have
stratiﬁed learners according to proﬁciency levels as they apply to reﬂexive
pronouns. This problem could explain the slight difference in ﬁndings
between the L2 studies and the Franks and Connell study as they relate to
phase 2b. Franks and Connell (1996: 448) report that children at phase 2b
of the acquisition process exhibit the following behavior:

(i) When the reﬂexive pronoun is embedded in a ﬁnite
clause, either the local or the non-local subject is an
acceptable antecedent.

(ii) When the reﬂexive pronoun is embedded in a non-ﬁnite
clause, only the non-local subject is an acceptable antecedent.

Second language acquisition research has produced data suggesting
that learners are more likely to accept a long distance antecedent when
the reﬂexive is embedded in a non-ﬁnite subordinate clause than when
the reﬂexive is embedded in a ﬁnite subordinate clause. If, as I have
suggested, these studies grouped learners into proﬁciency levels that do
not accurately reﬂect proﬁciency levels for reﬂexive binding, (in addition
to the already mentioned problems of data reporting and task issues) the
learner groups could contain learners at various stages of acquisition.
This would result in a group of learners, some of whom accept non-local
antecedents out of non—finite clauses (learners at phase 2b), and others
who exhibit a local reading for the non-ﬁnite clause (learners at phase 2c).

73

Likewise, in evaluating the data on reﬂexives embedded in a ﬁnite
clause, second language acquisition researchers could be reporting on a
proﬁciency group with learners who accept both local and non-local
antecedents (learners at phase 2b) as well as learners who accept only local
antecedents for the reﬂexive (learners at phase 2c). This proﬁciency
grouping, in combination with the inadequacies of both group data and
reporting on individual grammars which I have discussed earlier, could
result in data that suggest a stronger tendency to accept the non-local
antecedent out of a non-ﬁnite clause than out of a ﬁnite. The proﬁciency
grouping of learners may, therefore, contribute to the slightly different
conclusions reached by second language researchers regarding the
finite] non-ﬁnite distinction observed in second language learners.

In order to ﬁnd out if L2 learners are actually following an
acquisition pattern like the one identiﬁed in Franks and Connell (1996)
the study needs to examine the individual interlanguage systems in their
entirety as well as investigate learners who have not been broken down
into proﬁciency levels before the experiment. These important
methodological issues will be taken up in the next chapter.

2.4 Chapter summary

By reviewing current and recent research on the SLA of English
reﬂexives, several characteristics of learner behavior become clear. First,
the research suggests that learners of various ﬁrst language backgrounds
eventually indicate a local domain for the reﬂexive (Thomas, 1991;
Matsumura, 1994; Eckman, 1994; Lakshmanan and Teranishi, 1994). For
non-targetlike learners, however, two disntinct patterns are evident. (1)
ESL learners of various ﬁrst language backgrounds allow the English
reﬂexive to identify non-local NPs as acceptable antecedents (Thomas,

74

1989, 1991; Christie, 1992; Demirci, 1997). (2) ESL learners who have both
SDR and LDRs in their ﬁrst languages allow more long distance
antecedents for an English reﬂexive when it is embedded in a non-ﬁnite
clause than when it is embedded in a finite clause (Finer and Broselow,
1986; Broselow and Finer, 1990; Bennett and Progovac, 1993; Bennett, 1994;
Matsumura, 1994; Demirci, 1997; White et al., 1997).

I have argued that LD binding in ESL learners is not likely due to
L1 transfer because studies by Thomas (1989), Christie (1992) and White et
al. (1997) have shown that learners with SDR-only ﬁrst languages (e.g.
Spanish and French) also allow LD binding.

Secondly I suggest that while SL researchers (Bennet, 1994; Bennett
and Progovac, 1993) have proposed a transfer-based explanation for the
+/ - ﬁnite distinction made by certain learners, I argue that a similar
distinction has been shown in monolingual English speaking children. I
end the chapter by raising the possibility that L2 learners do not, in fact,
transfer a reﬂexive interpretation from their ﬁrst languages, but rather,
may be going through acquisition stages identiﬁed for children by Franks
and Connell (1996).

Chapter 3
METHODOLOGY

3.1 Introduction

After a discussion of some of the methodological considerations that
affect the task design and data analysis for this dissertation study, this
chapter presents the research questions for the study as well as descriptions
of the control and experimental groups. Furthermore, a description of the
task, the sentence types used, as well as a discussion of the predicted
outcomes of the data are discussed at the end of the chapter.

3.2 Methodological considerations

3.2.1 Introduction. Proﬁciency measurements for L2 learners play a critical
role in understanding the development of L2 grammars. This is particularly
true for studies investigating interlanguage stages. In order to deduce that
changes in interlanguage systems are actually a reﬂection of a developing
interlanguage (progressing, ultimately to the steady-state grammar),
researchers ﬁnd it necessary to tie changes in interlanguage systems with
overall changes (improvement) in global proﬁciency. In this way, global
proﬁciency scores serve as a "benchmar " against which speciﬁc variations

in interlanguage systems are pegged.

75

 

76

In practice, however, studies of the L2 acquisition of reﬂexives vary
widely in how and to what degree the issue of proﬁciency level is taken into
account. Thomas (1994) investigates the different ways in which proﬁciency
level is treated in various SLA studies. She outlines the methods by which
SLA researchers have evaluated learner proﬁciency and how these
evaluations have been used in the studies. Thomas ﬁnds that the most
common method of evaluating the proﬁciency of the study subjects is
through institutional placement. That is, subjects' proﬁciency level is
inferred from their placement into one or another English class.

Reﬂexive studies also provide several examples of this sort of
proﬁciency evaluation (Finer and Broselow, 1986; Hirakawa, 1990;
Lakshmanan and Teranishi, 1994). Other reﬂexive studies rely on scores of
standardized tests such as the English language Skills Assessment (ELSA)
battery (Finer, 1990), the University of Michigan English Language
Institute's English Placement Test (MTELP) (Thomas, 1991). Still other
studies rely on scores from proﬁciency tests created speciﬁcally for the
purposes of the study (Bennett, 1993; Matsumura, 1994).

3.2.2 Proﬁciency Measurement in SLA Studies of Reﬂexives. In addition to
the variety of proﬁciency measurements used in reﬂexive studies,
researchers also vary on the range of proﬁciency levels exhibited for subjects
participating in their studies. Thomas (1989) studies a group of learners
with a wide range of proﬁciency values. She reports data on learners with
proﬁciency scores for the ELI representing a range of 56 to 100, although
scores were only available for 51 of the 97 learners in this study and no
proﬁciency measure was available for the bilingual group. Thomas (1989:

77

288) describes these learners as, . . individuals for whom English is not a
native language, but who have achieved very close to native proﬁciency."

Thomas (1989) set out to investigate the possible inﬂuence of
learners' ﬁrst language background on the SLA of reﬂexives. Subjects were
divided into two learner groups: Chinese and Spanish native speakers
learning English. This study makes an important ﬁnding; namely, that both
the Chinese and Spanish native speakers allowed non-local binding for
English reﬂexives. Thomas found the two groups behaved similarly,
regardless of the obvious differences in L1 backgrounds and likely '
differences in proﬁciency levels. Ironically, her diverse (proﬁciency-wise)
subject pool enhances her claim that ESL learners from "local-only" type
languages are just as likely to accept non-local antecedents for reﬂexive
pronouns as ESL learners whose Lls contain both simplex and complex
anaphors.l

Other researchers, such as Finer and Broselow (1986), Finer (1990),
Eckman (1994) and White et al. (1997) also study heterogeneous groups of
ESL] EFL learners exhibiting varying ranges of proﬁciency levels. Finer and
Broselow (1986) report on a small group of Korean ESL learners (n=6), ﬁve
of whom were placed in an intermediate language class. One learner was of
a low proﬁciency. Finer and Broselow report that the learners who
participated in this study (all Korean speaking learners of English) were
more likely to accept a non-local antecedent for the English reﬂexive

pronoun when the reﬂexive was embedded in a non-ﬁnite clause as

 

1 Note that this study also included test sentences which provided subjects
with contextual clues to help them consider all possible readings of the
sentence. This task design, combined with her ﬁndings for the diverse
language groups, strengthens her conclusions.

78

opposed to a ﬁnite clause. This study, then, identiﬁed a particular feature of
the learner interlanguage of English reﬂexives.

Once Finer and Broselow identiﬁed this particular interlanguage
system (i.e. that these learners were more likely to allow a non-local
antecedent in a non-ﬁnite context than in a ﬁnite), they needed to argue
when (at what point) this stage occurs in the acquisition process and also
what this interlanguage grammar tells us about how (by L1 transfer,
accessing UG, etc.) learners acquire English reﬂexives. The information on
proﬁciency levels for these subjects was somewhat vague (ﬁve students in
an intermediate-level class, one student of low level), so it would be quite
difﬁcult to speciﬁcally link this interlanguage system with any particular
point in a given learner's overall proﬁciency development.

Instead, Finer and Broselow turn to theoretical predictions in their
search to answer the "when" and "how" questions of the L2 acquisition of
English reﬂexives. Finer and Broselow recognize that the peculiar behavior
of this small group of Korean ESL learners reﬂected the grammar of a
speaker whose Governing Category Parameter (GCP) was set at value (c) (see
Figure 3 for an illustration and explanation of the various values for the
GCP). English reﬂexives exhibit the unmarked (a) value, instantiated by a
local binding domain. Reﬂexives exhibiting the (c) value of the GCP (e.g.
simplex reﬂexives in Russian) take a local or non-local antecedent for the
reﬂexive when it is embedded in a tenseless (non-ﬁnite) clause, but is
restricted to a local interpretation when the reﬂexive is embedded in a
tensed (ﬁnite) clause. Reﬂexives exhibiting the (e) value of the GCP (e.g.
simplex reﬂexives in Korean) can take local or non-local antecedents from
all clause types.

79

Finer and Broselow interpret their ﬁndings for this group of Korean
learners of English by relying on the Subset Principle, the acquisition model
discussed in Chapter One which predicts that whenever grammars fall into
a subset relationship (as the parameters of the GCP do), children will assume
the most restrictive grammar as the unmarked value (e .g. value (a) for the
GCP) and then enlarge the grammar based on positive evidence. Given that
Finer and Broselow's subjects have in their L1 reﬂexives instantiating the
(e) value of the GCP, and that English reflexives instantiate the (a) value,
Finer and Broselow suggest that their subjects began the acquisition task by
transferring the (e) value reﬂexive from Korean and applying it to English.
The interlanguage stage identiﬁed for this group of subjects represented the
"half-way" mark of the L2 learners. That is, they were in the process of
restricting their grammar from an (e)-type interpretation of the reﬂexive to
an (a)-type interpretation (the target system). The interpretation of the data
suggests that these L2 learners were behaving according to the predictions of
the Subset Principle. Instead of immediately assuming the unmarked, (a)
value of the GCP, these learners were operating with a larger grammar. AS
a consequence, they were faced with the task of constraining their grammar
rather than enlarging it.

The linguistic and acquisition theories in acceptance at the time of
this study provided the mechanism by which Finer and Broselow could put
their ﬁndings into perspective in the acquisition process (who were half—way
between their initial assumption of value (e) and the target system, value
(a)). and also provide a way of shedding light on how the learners acquired
the reﬂexive pronoun (who transferred value (e) from their ﬁrst language
and applied it to English Then learners slowly constrained their grammar
(a non-UG based mechanism) to reﬂect the target system, value (a)).

80

The obvious problem with relying on linguistic and acquisition
theories to provide explanations for learner grammars is that the theories
themselves are not always stable. When the linguistic theories move on,
the acquisition data must be reinterpreted. The Finer and Broselow study
provides an example of just such a shift. As I point out in Chapter One, the
Subset Principle has received a lot of criticism. This is due to research on
child language acquisition, which has shown that children acquiring
languages with only complex (local-only) reﬂexives go through a period of
long-distance binding. The more recent theories of reﬂexive binding and
acquisition no longer require learners to initially assume that the reﬂexive
has a local-only domain and so, ironically, the Finer and Broselow results
today are not incompatible with behaviors and intuitions identiﬁed for
monolingual children learning English.

3.2.2.1 Proﬁciency Scores as a Variable to Control for in Comparing Learners
of Various L1 Backgrounds. Finer (1990) explores the results of Finer and
Broselow (1986) in greater detail. This study examines the intuitions of a
much larger subject pool (n=79) and it directly compares the grammars of
different L1 language groups (Japanese, Korean, and Hindi speaking learners
of English). In this study, steps to control for proﬁciency differences are
more important than in Finer and Broselow (1986). This is because the
purpose of Finer (1990) is to compare L1 groups. In order to compare Hindi
speaking ESL learners with Japanese and Korean speaking ESL learners, the
groups must be comparable in all respects other than L1 background. With
large subject pools, controlling for the possibility of proﬁciency differences
across the groups is notoriously difﬁcult. This difficulty is evident in Finer
(1990). For this study, Finer was only able to provide proﬁciency scores for

81

part of the subject pool In addition, the range of proﬁciency scores exhibited
by those learners for whom scores were available was large (between 60 and
100 on part of the English Language Skills Assessment (ELSA) battery). As a
result, Finer's cenclusion that Hindi speaking ESL learners do not
differentiate between ﬁnite and non-ﬁnite embedded clauses the way that
Japanese and Korean speakers do, cannot be generalized. Possible
proﬁciency differences between the two groups make it difﬁcult to conclude
that Hindi speaking ESL learners acquire English reﬂexives any differently
than Japanese or Korean ESL learners do.

White et al. (1997) attempt to control for possible proﬁciency
differences across their learner groups. This study's purpose was to evaluate
learner grammars. Learners are divided according to L1 (French and
Japanese native speakers learning English) and compared. White et al.
report that learner responses on the story task (but not the picture task),
indicate that the Japanese speakers, but not the French speakers differentiate
between ﬁnite and non-ﬁnite embedded clauses in their binding patterns for
English. White et al. control for proﬁciency differences by comparing
proﬁciency scores across groups. The ESL learners in this study took the
University of Michigan English Language Institute Test of English as a
Foreign Language (ELI test). No signiﬁcant differences in proﬁciency scores
were found across ESL groups. By reporting proﬁciency scores and
comparing them across the groups a researcher is investigating, it is possible
to make stronger arguments for Ll-based differences when the two groups
perform differently on a certain task.

Given that there were no signiﬁcant differences in global proﬁciency
scores between the learner groups, can we then assume that each Japanese

group and the French group were at the same acquisition stage for reﬂexive

82

pronouns? Not necessarily. Reﬂexive pronouns and regular pronouns are
markers of discourse. They are used to refer back to people previously
mentioned in the sentence or in the conversation. The results of written
proﬁciency exams such as the MTELP, TOEFL, etc. may not be the best choice
in predicting proﬁciency levels for reﬂexive pronouns.

These exams are written to evaluate a learner's command of a wide
variety of linguistic elements. The scores for these exams, then, represent
the average of the leamers' abilities in many different areas of the language.
A person's command of English grammar could be very low yet their
vocabulary and inferencing/ reading skills could be high This score could
easily be comparable to a person with extremely strong grammar and low
reading skills. Given their proﬁciency scores, these two people have the
"same" proﬁciency level, can we then assume based on these scores that
these two people are at the same point in their acquisition of reﬂexive
pronouns?

To the best of my knowledge, there has been no work investigating
the relationship between leamers' proﬁciency scores and their competence
for reﬂexive pronouns. This is not surprising, because at this point we do
not have a clear picture of the actual stages L2 learners go through as they
acquire reﬂexive pronouns. This situation results in a circular problem.
Until researchers are able to characterize the interlanguage systems exhibited
by L2 learners for reflexive pronouns, we cannot develop a model of
proﬁciency levels for reﬂexive pronouns. Likewise, without a model of
proﬁciency for reﬂexive pronouns, there is no way to experimentally test the
relationship between competence in reﬂexive pronouns and global
proﬁciency level. For the time being, the correlation must be assumed.

83

3.2.2.2 Heterogeneous Proﬁciency Groups and Analysis of Individual
Grammars. Eckman (1994) also studies a group of learners of various L1
backgrounds, although he does not attempt to compare one language group
with another. Because of this, and because of the way in which Eckman
analyzes the data, he is not faced with the problem of controlling for
proﬁciency differences across groups. As a result, Eckman does not provide
much information regarding learner proﬁciency levels, except to say that
they (n=25) were enrolled in the Intensive English Program at the
University of Wisconsin, Milwaukee. While proﬁciency levels are not a .
direct concern for the interpretation of the data, they are relevant to
Eckman's experimental task. For all of the reﬂexive studies I have discussed
thus far, it is important to insure that subjects are proﬁcient enough to
understand and complete the experimental task. This is an especially
relevant concern for reﬂexive studies because the purposes of these studies
require measuring how learners interpret complex sentences.

Eckman (1994) examined individual grammars in order to identify
possible grammars that may violate UG principles ("rogue grammars).
Because these principles are considered to be universal, it is not directly
relevant for the purposes of this study when certain behaviors present
themselves. This study simply wanted to identify if rogue grammars exist
among second language learners at any proﬁciency level.

This research objective results in an interesting problem. In order to
effectively search for rogue grammars, the researcher needs to investigate
the grammars of a wide range of learners (of both L1 background and
proﬁciency level). On the other hand, the researcher can only rely on data
from learners who are proﬁcient enough to complete the experimental task.

84

We see how this problem presents itself in the results of Eckman
(1994). For the ESL learners, Eckman ﬁnds that 16 out of 25 subjects
exhibited native-like grammars.2 These ﬁndings suggest that the subjects
for this study were proﬁcient enough to compete the experimental task, but
were perhaps too proﬁcient having already acquired target-like intuitions, to
elicit data that might identify rogue grammars.

It seems, then, that all studies of the L2 acquisition of English
reﬂexives have run into methodological problems. There is the problem of
linking one's ﬁndings to current linguistic theories, there is the problem of
controlling for proﬁciency differences in studies comparing learners of
different L1 backgrounds, and there is the problem of tapping a wide range

of proﬁciency levels in research investigating individual grammars.

3.2.2.3 Proﬁciency Scores as "Benchmarks" Against Which Changes in IL
can be Pegged. Given the methodological problems encountered in the
search to identify how learners acquire reﬂexive pronouns, the research I
discuss in this section approach the problem from a different angle. They
consider the question of when certain behaviors (e.g. local binding) present
themselves in hopes of developing a model of the acquisition process. In
order to do this, researchers have had to tie changes in interlanguage to
differences in overall proﬁciency levels. These studies stratify learners into

proﬁciency levels and then examine the responses of the "low" "mid" and

 

2 Three learners indicated a local domain and a +/ - subject orientation
while 13 learners indicated a local domain and a + subject only orientation.
This study design is prone to eliciting preferred rather than complete
interpretations of the reﬂexive (see Ch. 2 discussion), a problem reﬂected in
the ENS data as well. Out of the 25 English native speakers completing the
task, 15 out of the 20 subjects who answered consistently on the orientation-
type sentences also indicated a + subject only orientation

85

"high" groups in order to better understand the acquisition process. Once
again, this sort of experimental design assumes that global proﬁciency scores
can predict proﬁciency levels for speciﬁc structures, an assumption I have
presented is only tentative at this point.

Additionally, these studies exhibit problems similar to those
identified for SLA studies in general by Thomas (1994). That is, studies of
the L2 acquisition of English reﬂexives exhibit a wide range of
measurements, cut-off scores and groupings for various proﬁciency levels.
Additionally, as Thomas, 1994 points out, many studies have problems
establishing distinct proﬁciency groups.

Thomas (1991), for example, relies on the grammar and listening
comprehension portions of the MTELP. Subjects are then stratiﬁed into
low, mid, and high proficiency groups based on the MTELP scores (low: 21-
32; mid: 33-40; high: 41 -49). Given this proﬁciency stratiﬁcation, a learner
with a score of 32 on the MTELP would be at the same acquisition stage for
reﬂexive pronouns as a learner whose score is 21. But that same learner
would be at a different acquisition stage than someone whose score is 33. So
in addition to problems of consistent proﬁciency measurements, cut-off
scores, and groupings across studies, individual studies can sometimes have
problems defending that the proﬁciency groups they use are actually distinct.

Lakshmanan and Teranishi (1994) stratify their experimental group
into three proficiency levels based on institutional placement. Group A
(n=8), the "low" group is made up of learners from levels 1 and 2 (beginning
and low-intermediate) of the intensive English program at the Center for
English as a Second Language (CESL) at Southern Illinois University.

Group B (n=15), the "mid" group, is made up of levels 3 and 4 (upper
intermediate and advanced), and the "high" group is made up of students

86

enrolled in composition courses required of international students who pass
the University’s English proﬁciency requirement. Students were placed into
levels at CESL based on their TOEFL scores.

Hirakawa (1990) also stratiﬁes subjects based on institutional
placement (grade-level in school), but examines four proﬁciency levels
instead of three: Grade 10 (n=13, ages 15—16), Grade 11 (n=14, ages 16—17),
Grade 12 (n=18, ages 17-18), and Grade 13 (n=20, ages 18-19).

Relying on institutional placement to determine proﬁciency
stratiﬁcation introduces additional variables which further complicate the
problem of creating distinct proﬁciency groups. At English language centers
across the country, administrators are under constant pressure to ﬁll
classrooms, accommodate stafﬁng requests, student appeals, and teacher
schedules and conﬂicts. All of these pressures can press administrators to
accommodate exceptions in the usual placement procedures. Students who
are of significantly lower or higher proﬁciency level can sometimes slip into
classes meant for students of a much different level of English proﬁciency.
Given this situation, it is difﬁcult to assume, for example, that Lakshmanan
and Teranishi's (1994) low group is qualitatively different from the mid and
high groups. It is quite possible that there is some overlap in abilities
between the students in each consecutive class.

For Hirakawa (1990), these problems are multiplied because instead of
three levels, she investigates four levels. Additionally, these groups were
stratified according to their grade level in school. In the case of FL English
in Japanese schools, the mechanism for advancement may be different than
at other institutions. As Thomas (1994) points out, some schools may
advance learners to the higher language class automatically rather than
according to actual increases in proﬁciency level. This may be the situation

87

for the subjects of Hirakawa's (1990) study. As students move from grade 10
to grade 11 they may move to the next year's language class. It is possible
that they did not progress as much as others in their proﬁciency levels. The
result of such a situation would be the establishment of proﬁciency groups
which were not truly distinct. There is some evidence that this problem
occurred in Hirakawa (1990), in that she found no grade effects across learner
groups.

Additionally, Matsumura (1994) attempted to stratify learners along
the same lines as Hirakawa (1990) but found that grade level insufficiently
distinguished learner groups. The results of a cloze test administered to the
subjects brought out the variation within grade levels.

Even if all students in all classes fall within the required range of
proﬁciency scores, any language teacher will quickly acknowledge that the
typical English language class consists of learners at a wide range of
proﬁciency levels. Some students can handle anything the teacher presents,
other students seem to be perpetually "lost."

Added to this issue is the problem of deﬁning proficiency levels. For
each SLA study, it is up to the researcher to deﬁne the terms of "low" "mid"
and "high" proﬁciency levels. On the basis of the studies summarized so
far, it is clear that researchers do not agree on the best way to stratify
begirming, intermediate, and advanced learners for the purposes of
idenﬁfying the stages of the L2 acquisition of reﬂexive pronouns. Because of
this, it is particularly difﬁcult to compare results across studies because the
definitions of proﬁciency levels vary so widely.

In the search to describe the L2 acquisition process of reﬂexive
pronouns, proﬁciency-based stratiﬁcation studies have in some ways raised
more problems than they have solved. We have seen that global

88

proﬁciency scores may not be the most accurate predictors of proﬁciency
levels for reﬂexive pronouns. In addition, deﬁning distinct proﬁciency
groups based on proﬁciency scores and institutional placement has been
particularly problematic. Finally, because there is no clear agreement on
standard levels and measurement instruments, proﬁciency stratifications
vary widely, making it extremely difﬁcult to put together a picture of the
acquisition of reﬂexive pronouns on the basis of the collection of research
conducted thus far.

There are several important consequences for this situation. First of
all, the wide variation in proﬁciency measurements and cut-off scores
makes it difficult to compare studies, because each time a new study is
published, the proﬁciency measures and grouping criteria have changed.
Second, most of the studies I have discussed so far stratify learners into
proﬁciency groups that may not be distinct. Qualiﬁcation for one or another
proﬁciency group can, in some cases, be determined by a one point
difference on a proﬁciency exam. In other cases, qualiﬁcation for one or
another proﬁciency group is determined by the not always so rigid
placement policies of English language programs. In still other cases, no
information on the stratification is provided at all (Bennett, 1993).

It is very possible that a particular ESL learner could ﬁnd him] herself
in a different proﬁciency group for several of the studies cited above. In
order to better share L2 research and begin mapping out the characteristics of
a "beginner/ intermediate] advanced" interlanguage grammar of reﬂexive
pronouns, the deﬁnitions of "beginner/ intermediate] advanced" learners
need to be standardized.

Thomas' (1994) methodological discussion therefore appears quite
problematic. We cannot be certain that proﬁciency scores predict proﬁciency

89

for reﬂexive pronouns, so we cannot rely exclusively on those scores to
assure that Ll-based groupings are in fact comparable. Additionally, we
cannot rely exclusively on proﬁciency scores or institutional placements
(which, assumedly, are determined by proﬁciency scores) to distinguish
proﬁciency-based groups. 50, in what way can proﬁciency measurements
help us understand the L2 acquisition process? A learner's proﬁciency score
is an aggregate marker of acquisition. Because of this, a learner's global
proﬁciency score provides only one clue to his] her competence on a
particular structure such as reﬂexive pronouns. Global proﬁciency scores
can be at best, general markers of L2 acquisition when they are considered in
conjunction with other factors inﬂuencing the development of the
structure. For reﬂexive pronouns, factors affecting the development of
discourse competence, such as time in the L2 environment and type] degree
of language use are also important factors that need to be considered in

evaluating the acquisition of reﬂexive pronouns.

3.2.3 Immersion in the L2 Environment. For acquisition studies focusing
on reﬂexive pronouns, the amount of time the learner has lived in the L2
environment is an important factor in the acquisition of reﬂexive
pronouns. Several studies report on learners with little or no experience in
the L2 environment (Hirakawa, 1990; Bennett, 1993; Matsumura, 1994;
White et al. 1997). Regardless of their global proﬁciency scores, the learners
in these studies have had little to no experience participating in extended
oral dialogue. Thomas (1991) argues that reﬂexives are not taught in the
classroom. Furthermore, Parker et al. (1995) suggest that even in L2
immersion schools, students rarely engage in natural conversation in the

L2.

90

Thus, the different learning environments (second language (SL) or
foreign language (FL)), then, provide drastically different possibilities for
oral dialogue, and, as a result, different levels of exposure to reﬂexive
pronouns. In interpreting the results of L2 studies on the acquisition of
reﬂexive pronouns, it is important to consider whether or not the subjects
are FL or SL learners. Despite the ultimate explanation of how learners
acquire reﬂexive pronouns, FL learners simply may not have had enough
input to "trigger" acquisition of the reﬂexive pronoun or to provide enough
information for the learner to develop a strategy for acquiring the structure.

Length of time in the L2 environment is, of course, only an indirect
marker of a more important factor inﬂuencing the development of reﬂexive
pronouns - how often the learner uses the L2 for communication. It is
through opportunities for oral discourse and exchange that the learner has
an opportunity to practice using reﬂexive pronouns. One must also
consider when and where the learner actually uses the L2 in order to get a
clearer picture of the quality of the learning environment. Obviously, SL
learners who do not use the L2 at home or at work, but only use it in stores,
in class, etc. will not be as proﬁcient in the use of reﬂexives as someone who
has lived in the L2 environment the same amount of time, but must rely on
the L2 all day at work.

Issues such as global proﬁciency, type of learner (FL/ SL), length of
time in the L2 environment, and language use patterns all provide clues to
a given learner's proﬁciency for reﬂexives It is therefore important to take
as many of these factors as possible into consideration in interpreting results
of studies relating to the acquisition of reﬂexive pronouns. No one factor
can actually predict where a learner will be in the acquisition process. By

considering the combination of these factors, researchers can get a much

91

clearer picture of where the learner should be in the acquisition process for

reﬂexive pronouns.

3.2.4 L1 Background. As I have shown, current research on the L2
acquisition of reﬂexives assumes that L1 background has an effect on their
acquisition among L2 learners. I have laid out various conclusions from
numerous studies suggesting why this assumption may not necessarily be
so. I discussed how the argumentation for L1 transfer does not accurately
reﬂect the evidence. In Chapter 3, I show that grouping problems such as
proﬁciency stratiﬁcation and an over-reliance on global proﬁciency scores,
make it difﬁcult to assume that differences reported in the data of different
L1 groups are, in fact, due to differences in L1 background.

Added to these grouping problems is the issue of data are analysis and
reporting. This issue will be discussed within the context of the results of
this dissertation study. Even in the absence of a discussion of these two
items (methods of data analysis and reporting), it is difficult to conclude
with any certainty what, if any, inﬂuence a learner's L1 background has on

the acquisition of reﬂexive pronouns.

3.3 Research design

3.3.1 Research Questions. Language acquisition research concerns itself
with questions of "what," "when," and 'how." That is, researchers are
interested in understanding what the characteristics of interlanguage
systems are. Additionally, researchers attempt to outline when changes in

92

interlanguage systems occur, and if these changes are present in certain.
learners, but not others. That is, a completely developed theory of L2
acquisition should be able to provide the following information:

' What

 

Learners begin with a certain grammar, G1, which has the
characteristics A, B, C . . .

11911
the learner then acquires more information, recognizes certain
features of the input, has a chance to try and use the language,
transfers value X from the L1, etc. . . .

When
and then progresses to the next stage, G2, which has the characteristics
X, Y, Z.

This process, theoretically, continues until the learner reaches a steady state,
Gs, for that particular linguistic element. Gscan be the target system or some
sort of fossilized interlanguage stage.

3.3.1.1 Matching Research Design with Research Questions. What are the
characteristics of learner grammars, when do they occur (in what order), and
how do learners progress from one acquisition stage to another? These are
the relevant questions for researchers in their search for a theory of the L2
acquisition of reﬂexive pronouns. How well have these questions been

answered so far? I believe that research on the acquisition of reﬂexive

93

pronouns has focused itself so much on answering the "when" and 'how"
questions, that the "what" questions have been inadequately answered.

The studies I have discussed in this chapter, with the exception of
Eckman (1994)3, have looked at the behavior of groups of learners. Some
studies compared learner groups based on the ﬁrst language of the learners
(Thomas, 1989; Finer, 1990; White et al., 1997). Other studies compared
groups of learners based on a pre-established scale of proﬁciency groups (e.g.
Hirakawa, 1990; Thomas, 1991; Bennett, 1993; Lakshmanan and Teranishi,
1994; Matsumura, 1994). Because of the problems linking proﬁciency level
to proﬁciency based grouping, it is difﬁcult to conclude with certainty
whether the group based differences (L1 or proﬁciency based) are actually
due to the deﬁning characteristics of the group or some intervening variable
associated with the grouping. '

Despite these grouping problems, it is tempting to continue with this
type of research design. The reason for this is because in establishing group-
based differences, studies can generalize to the larger population and,
additionally, consider the "meatier" questions of how and why acquisition
proceeds the way it does. Group—based research design has a lot of power for
generalizability and, thus, for considering how L2 acquisition in general
proceeds. It is possible, however, that these grouping problems will not
always be with us. These problems can be addressed eventually if we take a
different look at the question and the way we pose it in research design.

It is impossible to create a coherent argument of how a learner moves

from 61 to G2, or that G1 comes before G2, without at least a minimal

 

3 White et al. (1997) include in their results a section titled "individual
grammars." I will show later that in spite of the title "individual
grammars," what the authors do with the data in this section is really
another form of group reporting.

94

description of the deﬁning features of G1 and G2. Until recently, L2
acquisition research on reﬂexive pronouns has focused primarily on how
groups of learners interpret reﬂexive pronouns. While this research design
can be very effective in answering the when (through proﬁciency grouping)
and the how (through L1-based grouping) type questions, the data these
studies produce are not designed to develop an accurate picture of the details
of the interlanguage systems of individual learners. The designs of the
studies I have discussed thus far in this chapter were not powerful enough
to answer the "what" question, and, because of the serious grouping
problems I have outlined, not able to produce a completely satisfying answer
to the "when" or "how" questions. Without some observational data on
the interlanguage systems expressed by learners of va rious proﬁciency levels
and various L1 backgrounds, we will not only fail to answer the "what"
questions of SLA, but also fail to adequately address the "when" and 'how"
questions.

In that vein, this study has been designed to ask the "what" questions.
Below are the research questions for this study.

3.3.1.2 Research Questions

1. What are the characteristics of the emerging system of English reﬂexive
pronouns?

2. Are these characteristics present in learners of various different L1
backgrounds?

3. Does this emerging system reﬂect the emerging system identiﬁed in
Franks and Connell (1996) for child learners of English?

95

The research design for this study has sacriﬁced some power to
directly address the "when" and "how" questions in order to give the

"what" questions more primacy.

3.4 Subjects

This study collected data from 81 learners of English as a second
language, two bilingual speakers and 43 English native speakers (ENS). In
order to gather intuitions from learners of diverse L1 backgrounds and
proﬁciency levels, subjects were selected from existing student populations
at Michigan State University and two campuses of Oakland Community .
College, a large, urban community college system in the Detroit, Michigan
area.
3.4.1 Control Group. The ENS (n=43) control group was drawn from the
student population at Michigan State University. The subjects ranged in age
from 18—28. In addition to this group of subjects, two bilingual speakers also
participated in the study. One subject was Greek-English bilingual and the
other was ASL-English bilingual. Both subjects grew up speaking a language
other than English in the home.

3.4.2 Experimental Group

3.3.2.1 Proﬁciency Level. The subject pool for this study was deliberately
drawn from a diverse population. Because the purpose of this study is to
identify as close to the complete range of possible L2 grammars as possible,
subjects with a wide range of proﬁciency levels, years in the United States,
and language use patters were chosen. The ESL group was drawn from the
campuses of Michigan State University and Oakland Community College.
Eighty-one students, ranging in age from 18-44, participated in this study.

96

Subjects also exhibited a wide range of proﬁciency levels. The Oakland
Community College subjects all took the MTELP proﬁciency exam and
scored in the range of 43-79 (maximum score: 100). For comparison
purposes, this group of subjects falls into the intermediate-advanced
proﬁciency range. The Michigan State University students completed the
English Language Center's placement exam and scored between 65 and 80 (a
score of 75 is roughly comparable to a 520 TOEFL score and a score of 80 is
comparable to a 550), putting them in the high-intermediate to advanced
range.
3.4.2.2 Time Spent in the L2 Environment/Years of Formal English Study.
All subjects were living in the L2 environment at the time they participated
in the study. SL learners were specifically selected because they are more
likely to have had extensive opportunities for communication and, thus,
have adequate exposure to reﬂexive pronouns. The amount of time living
in the United States, however, varied widely. The most recent arrivals had
only been in the United States for two months, while some subjects were
born in the US. and had been living in the US. for twenty or more years.
Amount of formal study also varied widely. In many cases, subjects who
had been living in the United States the longest, had the least formal
education of English. One subject, for example, had been living in the
United States for 1 1 years, but had only been in an English class for two
months. She states, "I have not really learned English until I came to the
United States. I learned to converse in English but this is the formal class I
started to study the English at."

On the other end of the spectrum, there were also many subjects who
had many years of formal education in English, but very little in the way of

experience in the L2 environment. One subject, a woman from Thailand,

97

had only been in the United States for two months, but she had studied
English extensively in Thailand. When asked how many years she had
been studying English, she replied, "15 years, but it mostly was grammar,
writing, and listening skills."

There are also examples of subjects representing other combinations
of variables, some with little formal education and recent arrival to the US,
others with more formal education and experience living in the United
States. These variables all have an effect on a person's competence and
should be considered in addition to overall proﬁciency level when
analyzing the results. I will return to these factors later in the dissertation.

3.4.2.3 L1 Background. Subjects were chosen to represent a diverse range of
ﬁrst language backgrounds. The primary language background of the

learners are given in Table 4.

Table 4: First language backgrounds of ESL learners

 

 

 

 

 

 

 

 

 

 

 

 

Language N Language N
Arabic 6 French 1
Thai 2 HmonL 1
Russian 3 Korean 3
Japanese 4 Chinese 2
Italian 1 German 1
Chaldean 3 Assyrian 1
Serbian 1 Lithuanian 1
Albanian 1

 

 

 

3.5 _ Method and materials
3.5.1 Materials.
3.5.1.1 Task. There has also been much thoughtful discussion on the issue
of the type of task which best elicits the competence of both L2 learners and
native speakers regarding reﬂexive pronouns. ' In Chapter Two I discuss at
length the problem of speaker preferences when attempting to evaluate
learner intuitions for reﬂexive pronouns. Studies by Finer and Broselow
(1986), Finer (1991), Thomas (1991) and Eckman (1994) report results
indicating a subject strategy for English native speakers. That is, the ENS
control groups exhibit a reluctance to acknowledge an object NP in a
monoclausal or embedded sentence as a possible antecedent for a reﬂexive.
Likewise, these English native speakers show overwhelming acceptance of
the subject NP as a possible antecedent for the reﬂexive. Of course,
theoretically, the subject and object NPs (in the local domain) are equally
viable candidates to be an antecedent for the reﬂexive pronoun. In actual
use, however, it appears that English native speakers prefer the subject over
the object NP as an antecedent for a reﬂexive pronoun‘.

In designing a task to investigate the various interlanguage systems of
L2 learners of reﬂexive pronouns, it is important to design the task to elicit
native speaker and learner com petence, rather than preferred
interpretations. As I suggest in Chapter Two, studies which indicate subject

 

4 For a discussion of the subject strategy of English native speakers, see
Thomas, 1991 and White et al. 1997.

99

binding but fail to elicit object binding by English native speakers have very
likely relied on tasks which tapped the preferred interpretations of subjects,
rather than a full range of interpretations.

It is therefore important to discuss how the task used for this study
attempts to overcome the problems of learner preference faced by earlier
studies. I have suggested that it is essential to provide learners with a
pragmatic context from which they can consider all possible interpretations.
It is especially important that the subjects in this study have "help"
considering all possible interpretations of the reﬂexive.

A second consideration relates to individual grammars. Earlier I
pointed out the importance of reporting individual grammars. This, I
argued, is because reporting on aggregate response patterns fails to
differentiate learners into the various levels identiﬁed by Franks and
Connell (1996) for monolingual English learners. In reporting individual
grammars, this study will speciﬁcally investigate the possibility of the
learner grammars as identiﬁed by Franks and Connell (1996), but will also
identify any other systematic grammars that may emerge among the L2
learners participating in the study.

Given these assumptions, the method for this study will rely on a
story task developed by White et al. (1997). This task consists of a short story
(2-5 sentences in length) followed by a one-sentence statement relating to
the story. The subject is asked to read the story and then read the sentence
and decide if the sentence is true based on the meaning of the story. An

example is given in (29).

(29) Bill was sick and in the hospital. Nobody knew what
was wrong with Bill. The hospital did a lot of tests on

100

Bill to ﬁnd out what was wrong. Bill had to wait a long
time in his hospital room. Finally, a doctor came in to
tell Bill why he was sick.

After the medical tests, the doctor informed Bill about
himself.

True False

This task gives learners the contextual information necessary to
consider all possible antecedents. In the case of (29), the context helps the
learner consider the local, object NP as a possible antecedent. White et al.
(1997) conducted a study comparing the story task with a traditional picture
task in which subjects were asked to interpret a picture and indicate if the
sentence below the picture reﬂected the meaning of the picture. Her study
found that the story task more effectively encouraged learners and English
native speakers to consider non-preferred readings of the sentence. White
et al. (1997) found English native speakers to be signiﬁcantly more likely to
accept non-subject NPs in monoclausal sentences than in the picture task.
Furthermore, the story task also identiﬁed the ﬁnite] non-ﬁnite distinction
established in previous studies (Finer and Broselow, 1986; Broselow and
Finer, 1991; Finer, 1991; Matsumura, 1994; Demirci, 1997; White et al. 1997;
Bennett, 1994).

The task for this dissertation study involved 27 stories. The sentences
were kept as simple as possible. Pronouns were used as little as possible, but
where they were absolutely necessary, they were used in sentences where
there was only one reasonable antecedent. The stories were roughly the
same length. The test included three items designed as a proﬁciency check

on the learners. Each of the stories resembled the reﬂexive stories with the

101

exception of the true or false statement that followed. For these items, the
true or false sentence was written to check that the subject comprehended

the passage. An example of this item-type is given below in (30).

(30) Julie and Mary went to a restaurant for lunch. For lunch
they each had a cheeseburger, french fries and a Coke.
When Julie finished her lunch, she said, "I am ﬁnished
with my lunch, I will throw our garbage out." 'That is
so nice, thank you!" Mary replied.

Julie was happy that Mary threw out the garbage.

True False

3.5.1.2 Sentence Types. The experimental portion of the task included four
tokens for each sentence type. A portion of the task was piloted with 27 ESL
learners prior to ﬁnal preparation of the task. Sentences] stories were
constructed to test orientation (one type with context directed toward the
subject, one type with context directed toward a non-subject NP) and
domain (one type with context directed toward a non-local NP and one type
with context directed toward a local NP) for both ﬁnite and non-ﬁnite
embedded clauses. The sentence types are listed below in (31).

(31) a. Monoclausal sentence, context suggests the subject.
- Tomi gave Bill a picture of himselfi
b. Monoclausal sentence, context suggests the object.
- Tom sent Billi a letter about himselfi
c. Biclausal, ﬁnite embedded clause, context suggests local
NP

102

- Tom noticed that Billi was looking at himself; in the
mirror.

(1. Biclausal, finite embedded clause, context suggests
non-local NP
- Tomi heard that the doctor will operate on himselfi.

e. Biclausal, non-ﬁnite embedded clause, context suggests
local NP
- Tom told Bill, to trust himself; more.

f. Biclausal, non-ﬁnite embedded clause, context suggests
non-local NP

- Tomi asked Bill to return the book to himself].

3.5.2 Response Patterns. Individual response patterns for each subject are
examined for systematicity. That is, in order to conclude that a given
learner exhibited a particular orientiaton or domain, that learner was
required to provide the same answer for 3 out of the 4 tokens for that
sentence type. The resultant "grammars" were then evaluated within the
context of the Franks and Connell (1996) study. If L2 learners follow a
pattern similar to monolingual English speaking children, the responses for
the subjects participating in this study would follow the patterns outlined in
Table 5.

Each of the response patterns outlined in Table 5 reﬂect the different
response patterns identiﬁed in Franks and Connell (1996). If a given learner
is at the "superlocal" phase of acquisition, they will accept the local object in
mono-clausal sentences and reject the local subject. In the biclausal

103

Table 5: Predicted response patterns for experimental subjects, assuming a
response pattern that reﬂects the acquisition stages identiﬁed by
Franks and Connell (1996) for child language acquisition

sentence-
a

C

e

 

sentences, the local subject is the nearest NP so acceptance of the local
subject in this case is predicted. For learners who accept long-distance
antecedents, sentence—types "d," "e," and "f" will elicit response patterns to
show the speciﬁc nature of how individual learners differentiate between
ﬁnite and non-finite embedded clauses. If learners are, indeed, progressing
through an LD-phase similar to that identiﬁed by Franks and Connell (1996),
these learners will accept local and non-local subject NPs when the reﬂexive
is embedded in a ﬁnite clause, yet will simultaneously reject the local subject
when the reﬂexive is embedded in a non-ﬁnite clause.

By focusing on the grammars of individual learners from a diverse
subject pool, this study intends to identify a wide range of interlanguage
systems relied upon by L2 learners.

Chapter 4
RESULTS

4.1 Introduction

This chapter begins with an explanation of the exclusion criteria
which were included in this study. Subjects were required to pass a
proﬁciency check as well as a systematicity requirement in order to be
included in the study. The control group results are addressed ﬁrst. Both
group and individual results are reported and discussed. This is followed by
a discussion of the group and individual results of the learner groups. The
learner groups exhibited three language patterns. (1) a target-er behavior,
reﬂected in a local-only domain for the reﬂexive. (2) a +/ - ﬁnite behavior,
reﬂected in a local-only domain for the reﬂexive in a ﬁnite context with a
local] non-local domain for the reﬂexive in a non-ﬁnite context, and (3) a
"yes" - type behavior, reﬂected in an acceptance of antecedents in all

domains and contexts.

4.2 Data reporting

One of the fundamental characteristics of our current understanding
of interlanguage is that L2 grammars are not group phenomena, but rather
are the systems within which individual learners operate. Recently, SLA
researchers have criticized earlier research whose purposes related to the

104

105

identiﬁcation and analysis of interlanguage grammars, but whose data
analysis focused on reporting group data (Eckman, 1994; Thomas, 1991, 1995;
Wakabayashi, 1996)1. By evaluating the general tendencies of learner
groups, this type of data analysis can easily obscure the deﬁning
characteristics of certain interlanguage grammars. Standard grouping
measures (such as proficiency level, institutional placement, etc.) can easily
contain learners at different stages of the language acquisition process and,
as a consequence, contain subjects with distinctly different interlanguage
grammars. By reporting group scores, then, the distinguishing features of
interlanguage systems can easily be obscured.

Eckman (1994) argues that this problem in research design is
prevalent among studies of the SLA of reﬂexives. The goal of many of these
studies is to identify the interlanguage stages learners progress through in
the acquisition process. However, the data analysis reports only on the
degree to which. groups of learners accept or reject certain NPs (local/ non-
local, subject] nonsubject) as acceptable antecedents for the reﬂexive. He

states,

.if a study reports that one of the L2 groups gave local
responses on English reﬂexives m 80% of the cases. .this
result does not tell us whether the principles of UG are being
obeyed unless we know how those responses are distributed
across the subjects. Clearly, the case for UG can be made if all
the subjects scored 80%. If, on the other hand, the 80% group
result is a composite of subjects compiling many different
scores, some of which indicate systematic adherence to UG
principles and others which either are unsystematic or indicate
non-adherence to UG, then the case for UG governing SLA is
not at all clear (Eckman, 1994: 221).

 

1 These studies and their grouping mechanisms] data analysis are discussed
in Chapters Two and Three.

106

It is therefore important to examine individual interlanguage systems as a
whole, rather than measuring group averages for each separate component
of the target grammar (e.g. “percentage of responses accepting the local NP,"
”percentage of responses for the non-local NP," etc.).

The data analysis for this study will focus speciﬁcally on identifying
the characteristics of individual grammars. I also report group results in
order to provide additional context to the individual results.

4.3 Exclusion criteria

In order to develop a complete description of an individual’s
grammar, the experimental task was designed to tap intuitions on a range of
factors relevant to the grammar of reﬂexives. The task included sentences
pertaining to the orientation of the subject’s grammar (subject-only or +/ -
subject) as well as the domain of the reﬂexive in both ﬁnite and non-ﬁnite
contexts . In order to develop as detailed a picture as possible of the
interlanguage systems of L2 learners, this study focuses on the various
components of the grammar (orientation and domain in both finite and
non-ﬁnite contexts) as they relate to each other for each subject, rather than

in isolation from each other for groups of subjects.

4.3.1 Proﬁciency Check. The experimental design includes two
exclusionary measures which were designed to control for learners who
were 1) unable to either understand and follow the task, or 2) provide a
systematic picture of their interlanguage grammar. The ﬁrst measure is a
proﬁciency check. The task was designed to identify subjects whose low
English proﬁciency level prevented them from understanding and
completing the task.

107

The task required that subjects be able to:
1) Read and understand a paragraph of 4-5 sentences.
2) Read and understand a bi-clausal sentence about the passage.
3) Decide if the sentence was true or false according to the passage.
Three such passages were integrated among the test items. Two required a
"true" response. One item required subjects to answer "false." This item
was included in order to have tokens requiring both true and false answers
in order to check for possible response biases. The three comprehension g
tokens are provided in (32-34) below. ‘

 

(32) Mario went to the library to do some research for a school project. He
asked the librarian for some help. ”Excuse me," Mario said, "I am looking
for a book about computers." The librarian showed Mario a large stack of
books all about computers.

The librarian showed Mario where to ﬁnd books about computers.

True False

(33) Julie and Mary went to a restaurant for lunch. For lunch they each had
a cheeseburger, french fries and a Coke. When Julie ﬁnished her lunch, she
said, ”I am ﬁnished with my lunch, I will throw our garbage out." “That is
so nice, thank you!" Mary replied.

Julie was happy that Mary threw out the garbage.

True False

(34) Barry and George went to the movies together every week. This week,
Barry arrived at the theater before George. Barry sat down in a seat. Soon,

108

George arrived, carrying popcorn and soda. ”Look Barry, I bought popcorn

and soda for us!”
Barry knew that George bought some popcorn.
1.112 was

The ENS and learner groups completed identical tasks. That is, the
ENS subjects answered three questions designed to test whether or not they
understood the content of the test paragraphs and the complex sentence
which followed. After the test administration was completed, it became
apparent that one of the items, number (33) above, could be answered either
True or False depending on how one interpreted the paragraph and the
sentence. Both the learner groups and the ENS group provided great
variation in their responses to this item. This item, therefore, was
discarded. Left with two true] false questions, the L2 learners and the
English native speakers were then required to answer both remaining

comprehension questions correctly in order to be included in the study.

4.3.2 Systematicity. The second exclusionary measure relied upon for this
study was a systematicity requirement. That is, for each sentence type in this
study, four tokens were included. In order to ”qualify" for a particular
interpretation, the subject had to answer three out of the four tokens in the
same way. For example, in order to conclude that a particular learner's
interlanguage requires a local domain (i.e., only the NP in the embedded
clause is an acceptable antecedent), that learner must accept three or four of
the four test sentences which require a local interpretation of the reﬂexive.
As I have mentioned, it is necessary to consider the various
characteristics of the interlanguage (IL) for reﬂexives (e.g., domain in ﬁnite

 

109

and non-ﬁnite contexts and orientation) as they relate to each learner, rather
than consider each characteristic separately as they relate to groups of
learners. This study includes six different sentence types to identify the
various characteristics of interlanguage systems for reﬂexive pronouns. In
order to make inferences about each particular component of the grammar,
each subject was required to answer three or four of the four tokens (per
sentence type) systematically. This systematic response pattern only
provides information about how each subject interprets one particular
component of the grammar. It is important to understand how all of the
components of the grammar are patterned (i.e., to develop a description of
the entire interlanguage grammar for the subject). Thus, the subject needs
to answer consistently across all sentence types.

Because the experimental group consists cf second language learners,
it is difﬁcult to obtain systematicity across all six sentence types for the
experimental subjects. Because these subjects are learners, we assume that
for many of them their interlanguage grammars are not constant or stable.
The interlanguages are developing and changing and may simultaneously
contain multiple and competing grammars. As the learners move from one
interlanguage system to another, their grammars may be in a state of ﬂux.
Some learners could be operating with two competing grammars (e.g., G1
and G2) until the new grammar ﬁnally takes over (Franks and Connell,
1996: 459), or the learners could be restructuring their grammar as they
move to the next level . In light of these unstable grammars, it is likely that
as learners move from one interlanguage system to another, the grammar
may show certain inconsistencies.

Thus, due to the nature of the acquisition process, non-systematic
patterns across sentence types were expected. A systematicity requirement of

 

110

ﬁve or six out of the six sentence types was therefore established. That is, in
order to be included in the ﬁnal subject pool, a given subject must have
provided systematic responses (to either accept or reject the particular

antecedent) for at least ﬁve of the six sentence types.

4.4 ENS control group

4.4.1 Exclusion Criterion

4.4.1.1 Proﬁciency Check Forty-four ENS subjects took the test. One subject
did not complete the task, and so that test was discarded. Of the remaining
43 subjects, 12 subjects (one bilingual and 1 1 English-only native speakersz)
failed to answer one or both of the two comprehension questions correctly.
Six subjects gave an incorrect response for number (33), above; 5 subjects
gave an incorrect response for number (32), above, and one subject gave
incorrect responses for both comprehension items.

Thus, despite the fact that a large number of subjects were lost (largely
due to the fact that only two test items were left to discriminate those who
gave thoughtful consideration to the test items (and could understand the
reading passages) from those who did] could not), the comprehension errors
were evenly distributed across test items. Six subjects answered incorrectly
for number (33), ﬁve were incorrect for (32) and one subject was wrong for
both items. This distribution of errors implies that neither of the remaining
two comprehension items were unnecessarily drawing people into a

"wrong" interpretation.

 

2 Most of the English speakers who participated in the study indicated that
they have had some foreign language study. Usually, this involved no
more than the equivalent of a semester or two of a college language course.
Because these subjects did not grow up speaking anything other than
English until their exposure to a foreign language class in high-school
and] or college, they are referred to as English-only native speakers.

111

The subjects, who were enrolled in an undergraduate college course,
were either poor readers and unable to comprehend the paragraph, or were
distracted and] or not interested in completing the task. In any event, given
the errors on the comprehension questions, it is impossible to assume that
performance on the experimental items would be any better and so these
tests were discarded.

4.4.1.2 Systematicity. As discussed earlier, each sentence type consisted of
four tokens. When a subject answered true to three or four out of four
tokens, the subject systematically accepted (A) the reﬂexive-antecedent
relationship being evaluated for that sentence type. Likewise, if the subject
answered false to three or four of the four tokens, the subject systematically
rejected (R) the reﬂexive-antecedent relationship. When the subject
answered true to two tokens and false to two tokens, the subject exhibited an
inconsistent (1) pattern for that sentence type. For the six sentence types
investigated in this study, subjects were allowed to exhibit no more than
one inconsistent (1) response pattern.

Two ENS subjects exhibited inconsistent response patterns for more
than one sentence type. These leamers' response patterns are highlighted
inTable 7.

4.4.2 Group Results. The ﬁnal subject pool, then, consists of 29 English-
only native speakers and 2 bilingual speakers. The test results for the
English-only native speakers (hereafter referred to as the ENS group) are
considered separately from the bilingual subjects. First, I consider the group
data for the ENS group.

112

The results are reported in terms of mean scores. There were 24
experimental items (4 tokens per sentence type). For certain sentence types
(types X, Y, Z) "true" was the correct answer and for other sentence types (A,
B, C) "false" was correct. The mean score for the ENS group is 21.7931 with a
standard deviation of 1.6340. These results indicate that most ENS subjects
had high scores on the test as the mean score is high and the standard
deviation is low (minimum score: 18, maximum score: 24).

The mean scores and their percentage equivalents are given inTable
6, below. The scores for the different sentence types are reported on a scale

of 4 as there were 4 tokens per sentence type.

Table 6: Mean scores for English control group (individual sentence types:
maximum score 4; overall mean: maximum score 24)

 

Orientation Finite Domain Non-Finite Domain

 

Subject I Object Local J Non-local Local INon-local

 

 

 

3.90 (98%) I 2.90 (73%) 3.72 (93%) J 3.72 (93%) 3.93 (98%) I 3.62 (91%)

 

 

Overall Mean: 21.79 (90%)

 

 

The group results suggest that English native speakers exhibit local
binding patterns when the reﬂerdve is embedded in either a ﬁnite or non-
ﬁnite clause. The subjects for this study were over 90% accurate across
sentence types, overwhelmingly accepting the local antecedent (93% accurate
for ﬁnite, 98% accurate for non-ﬁnite) and rejecting the non-local antecedent
(93% accurate for ﬁnite, 91% accurate for non-ﬁnite) in both ﬁnite and non-

ﬁnite contexts. The mean scores also suggest that English native speakers

113

are less likely to accept the object NP as a possible antecedent for the
reﬂexive. The ENS group was only 73% accurate in accepting the object NP
as a possible antecedent for the reﬂexive. Conversely, they were 98%
accurate in accepting the subject NP as an antecedent for the reﬂexive.

The mean scores therefore suggest that this ENS group exhibits a local
binding domain in both ﬁnite and non-ﬁnite contexts, yet the orientation is
not strongly established here. The ENS group appears to accept, although
less strongly, the possibility of object binding with a reﬂexive pronoun.
Thomas (1989) refers to this phenomenon as the ”subject strategy.” That is,
English native speakers exhibit a strong preference for the subject NP. Often
this preference goes so far as to exclude the object NP. This preference is
apparent in the group data above.3

4.4.3 Individual Results
4.4.3.1 English-Only Speakers. The argumentation for focusing on the
individual grammars of L2 learners does not necessarily hold for the native
English speaking control group. Assumedly, there is little substantive
variation among native speakers. However, individual grammars were
analyzed for the control group in this study. This was done primarily to
identify the exact nature of the subject preference exhibited by many native
English speakers.

While the group data suggests a homogeneous control group
(exhibiting a high mean score and low standard deviation), a close look at
individual grammars suggests that this group actually exlu'bits three distinct

response patterns. : Table 7 summarizes the consistency with which each

 

3 For an extensive discussion of the possible task effects on the subject
strategy, see White et al. (1997)

114

subject interpreted each sentence type. There were four tokens per sentence
type. In order to be considered consistent, the subject had to answer three or
four out of four tokens in the same way.

The individual results shown in Table 7 show more detail than the
group results provide. While the group means reveal that the ENS group
was only 73% accurate at accepting the object NP in monoclausal sentences,
the individual results show how the “incorrect” responses were spread
across learners. We ﬁnd here that only four out of the 27 ENS subjects are
responsible for most of the rejections of the object NP. In fact, 23 of the 27
subjects (79% of the ENS subjects) consistently accepted both the subject and
object as an antecedent for the pronoun.

By examining how responses are spread across learners, a different
picture emerges regarding the grammars of English native speakers. The
group mean score implies that all of the subjects who took the test were only
76% accurate on the subject] non-subject sentence types. However, upon
examination of the individual response patters we see that nearly all of the
subjects (27] 29 or 93%) were either systematically correct or systematically
incorrect on the subject] nonsubject sentence types. The distribution of
incorrect responses, then, is not spread across all the learners, but is
concentrated in a few individuals. The group as a whole is therefore not
made up of individuals who are 76% accurate, but consists of subjects who
have very strong opinions in both directions. There are a few (four) who
consistently reject the object NP as a possible antecedent, and there are many
(23) who consistently accept subjects and nonsubjects as acceptable
antecedents for the reﬂexive.

In considering both domain and orientation for the ENS group, the
data show that most English native speakers exhibited ”target-like”

115

Table 7: ENS individual results (A = subject consistently (3 or 4 out of 4)
accepted this NP as a possible antecedent; R = subject consistently
rejected (3 or 4 out of 4) this NP as a possrble antecedent; I = subject
accepted this NP as a possible antecedent for two tokens and
rejected this NP for two tokens (inconsistent))

ta

te

 

116

behavior. There was one smaller group, of ﬁve subjects who bound the
reﬂexive locally in both ﬁnite and non-ﬁnite contexts, but exhibited a subject
orientation. As I have suggested, it appears that these subjects are
employing the “subject strategy” by rejecting the grammatically possible, but
non-preferred object antecedent.

Two speakers whose responses were not totally consistent were
included in this group of “subject oriented, local domain" speakers. Subject
E25 accepted two of the tokens which required binding the reﬂexive with
the object NP and rejected the other two. This subject appears to have a _
preference for the subject NP, but when the context is strong enough, will
apparently allow co-reference between a reﬂexive and an object NP. This
speaker, then, does not belong entirely to either the ”target-like” group or
the ”subject-strategy" group. This person’s language system appears to fall
somewhere between the two groups. While this individual shows signs of a
subject strategy, it does not appear to be as strong as it is for those subjects
who more consistently rejected object binding.

Subject E26 consistently accepted the subject NP and rejected the
object NP, qualifying for the ”subject oriented" group. This person, while
consistently rejecting the non-local NP in a ﬁnite context, was inconsistent
in accepting the local NP. Because this person consistently rejected the non-
local NP in both ﬁnite and non-ﬁnite contexts, and consistently accepted the
local NP in the non-ﬁnite context, I have tentatively included this subject in
the ”subject-oriented, local domain" group.

This leaves the last three subjects. These subjects exhibited grammars
that did not reﬂect the ”target-like" group, or the ”subject-strategy" group.
Two of the three subjects (E27, and E29) exhibited inconsistent response
patterns for more than one sentence type, so interpreting the response

117

patterns for these subjects is difficult given the lack of systematicity for these
sentence types.

Subject E28 appears to allow local and non—local binding in both ﬁnite
and non-ﬁnite contexts. Additionally, this subject allows both subject and
object NPs as antecedents for the reﬂexive. Given that this response pattern
was exhibited by a native, monolingual English speaker, it is not likely that
the person's mental grammar allows such unconstrained relationships
between reﬂexive and antecedent. However, it is possible that this
individual subject recognized the grammatical "violations" of the sentence
types where non-local binding makes pragmatic sense, yet nonetheless
accepted the relationship between the reﬂexive and non-local NP because it
made pragmatic sense. A second possibility is that the subject was simply
exhibiting a "yes" response bias.

4.4.3.2 Bilingual Speakers. The two bilingual speakers who participated in
the study and passed the comprehension check exhibit grammars similar to
the subjects in the ENS control group. Table 8 shows the results of their
tests. B1 is Greek] English bilingual and subject 82 is Spanish] English
bilingual.

Subject B1, a Greek-English bilinguaL exhibits a target-er grammar,
accepting either the subject or object NP as an antecedent for the reﬂexive
and allows only local NPs as acceptable antecedents. Subject B2 likewise
binds the reﬂexive exclusively in the local domain, yet is inconsistent in
accepting the object NP as a possible antecedent. This pattern is reﬂective of
ENS subjects E25 and E27. The inconsistent orientation of subject B2 does

118

not appear. to be due to L1 inﬂuence, as this person is a Spanish-English
bilingual. Spanish reﬂexives function similarly to English, in that they
allow either subject or non-subject NPs as acceptable antecedents in the local

Table 8: Individual Results - bilingual group (A: speaker consistently
accepted this NP as an antecedent for the reﬂexive; R: speaker
consistently rejected this NP as an antecedent for the reﬂexive; 1:
subject accepted this NP as a possible antecedent for two tokens and
rejected this NP for two tokens (inconsistent))

 

domain. Likewise, the Greek] English bilingual does not appear to be
operating with any L1 inﬂuence at this point as Greek is a language with
both LDRs and SDRs but exhibits a target-like grammar here.

4.5 Learner results
4.5.1 Learner groups. The experimental task was administered to 80
learners of English as a second language, studying English in the United
States. Five tests were discarded because subjectsfailed to complete the
entire task.

Seventy-ﬁve subjects therefore remained. Many of these 75 subjects
indicated that they were multi-lingual. That is, these subjects speak at least
two languages other than English, or speak several other languages in

119

addition to English. The experimental group was therefore divided into
two: a multi-lingual group and an SL group.

The motivation for dividing the learner groups comes from
differences between multi-lingual and unilingual FSL learners, identiﬁed by
Zobl (1995). Zobl reports that multi-lingual (ML) ESL learners4 appear to
function with wider grammars than unilingual (UL) ESL learners. Zobl
(1995) studied the linguistic judgments of UL and ML learners of English
across several English structures, including the c-command requirement for
reﬂexive binding as well as backward anaphora. He found that in most of
the structures studied, the multi-lingual learners exhibited larger, less
conservative grammars than the monolingual subjects. In the case of
reﬂexive binding, the ML learners allowed binding between reﬂexive and a
non-c-commanding NP more often than the UL learners. Conversely, the
ML learners allowed backward anaphora (whose acceptance requires a
knowledge of c-command) less often. Because of these differences in
interpretation of reﬂexives by ML and UL learners of English, it was
necessary to separate the ML from the UL (hereafter referred to as SL) group
for this dissertation study.

The SL group consists of learners for whom English is the only
second language spoken by the subject. In some cases, subjects indicated that
they spoke or studied a third language, but were only beginners at that
language. These people were included in the SL group. The SL group
consists of 46 subjects.

The multi-lingual group consists of two types of learners. First, this
group contains several learners who are bilingual in two languages and are

 

4 Multilingual learners are learners who know languages beyond their
native language. Unilingual learners (UL) are learners who speak one
language and are learning another as a second language.

120

studying English as a third language. All of the bilingual subjects with the
exception of one indicated that they were "advanced" "high-advanced" or
"bilingual" in their second language (not English). One subject indicated an
intermediate proﬁciency in the second language (not English). The second
type of learner in the multi-lingual group is a learner who is studying
English as at least a fourth language. The proﬁciency in the other L23 vary,
but because there are multiple L2s, these learners were also included in the
multi-lingual group. The multi—lingual group consists of 29 subjects.

4.5.2 Exclusion criteria .
4.5.2.1 Proﬁciency check. As with the ENS group, a large percentage of the
ESL groups failed to answer both of the comprehension questions correctly.
Number (34) was not considered with the learner groups because it could be
answered either true or false and be correct). Twenty-three subjects from the
L2 group answered one or more of the comprehension questions incorrectly
and 10 subjects from the multi-lingual group (ML group) missed one of the
comprehension questions. The distribution of errors and percentage of
subjects dropped per learner group is given inTable 9.

As Table 9 shows, half the L2 group was lost due to problems with
one or more of the comprehension questions. The multi-lingual group
fared better, losing 35% of its subjects. Compared with a loss of nearly 1] 3 of
the ENS speakers, it is not surprising that such a large proportion of the
learner groups failed the comprehension criteria of the task. The effects of
this loss of subjects will be taken up at length in the next chapter. As a
result, the subject groups were reduced to 23 subjects in the L2 group and 19
subjects in the multi-lingual group.

121

Table 9: Number and percentage of subjects dropped due to comprehension

 

 

 

 

 

 

 

errors
L2 Group ML Group
N 46 29
No. of subjects w] errors 23 10
on compreh. items
rcentge of group lost 50% 35%

 

4.5.2.2 Systematicity. As discussed previously, this study requires that each
subject give the same answer (True or False) for 3 or 4 of the tokens in each
sentence type in order for that person's response to be included in any
particular category. If a subject answered true to two items and false to the
other two items for the sentence type, they were labeled "inconsistent" for
that sentence type. There were six sentence types included in the
experimental task, so in order to be included in the experimental group, the
learner must have answered consistently on 5 or 6 of the 6 sentence types
(i.e. no more than one inconsistent response pattern).

The L2 group lost ﬁve subjects and the ML group, lost ﬁve subjects
due to inconsistent response patterns. The ﬁnal subject pool is therefore 18
for the L2 group and 14 for the multi-lingual group.

Tables 10-12 show the language backgrounds of the subjects
participating in the study. Table 10 gives the Lls of the bilingual learners,
Table 11 gives the L1 and L23 of the multi-lingual subjects, and Table 12
gives the L1 backgrounds of the L2 subjects.

 

122

4.5.3 Group results. As I discussed in Chapter Three, the goal of this study
is to idenb'fy as many different acquisition stages as possible. As a result, I
intentionally drew upon a diversity of learners to achieve more
heterogeneous groups than in previous studies. Therefore, the objectives of
the group data analysis are limited. While the learner and control groups

are compared for differences in overall mean scores, the analysis of

w.

performance on the different sentence types within the test are considered
for each group separately. The results provide evidence of certain behavior

tendencies for the learners and, as outlined in the next section, these

tendencies are reinforced and made clearer upon examination of individual
grammars.

Table 13 provides the descriptives for the performance of the
experimental and control groups on the task. The SL and ML groups have
lower mean scores (around 18.5) than the control group, whose mean score
is 21.7931. In looking at the standard deviations (SD) for the groups, the
ENS group had very little variance (SD: 1.6340). For the learner groups, the
SDs were larger. Of the two groups, the L2 group showed the greatest
variance with a SD of 2.8901. The ML group's SD was lower, at 2.1381.5

In order to compare the learner groups with the ENS control group, a
one-way ANOVA with planned contrasts was run. The contrast test, shown
inTable 14, revealed that while there was no signiﬁcant difference between

the mean scores of the learner groups (ML vs. SL mean scores) the mean

 

5 In the next section I analyze the individual grammars of the L2 and ML
groups and the results I report in this section support these ﬁndings. The L2
group exhibited a larger range of distinct grammatical patterns. While the
ML group exhibited fewer distinct grammatical patterns, the largest
represented pattern for the ML group was a non-target-like pattern, which is
probably responsible for the slightly lower overall mean score represented by
the ML group.

 

123

Table 10: L1 backgrounds of bilingual speaking learners of English

 

 

 

 

 

Lla le n = 10
Arabic Chaldean 4
Russian French 1
Russian Ukranian 1
Assyrian Arabic 1
Albanian Italian (intermed) 1
Mandarin Taiwanese 1
Lithuanian Russian 1

 

‘

 

Table 11: L1 and L2 backgrounds of multi-lingual learners of English

 

 

 

 

 

 

 

Slovenian, German

 

L1 L25 n = 4

Korean German, Japanese 1

Albanian Italianj‘urkish, Serbian 1

'Ffench GermanﬁSwiss German, 1
Italian

Serbian Macedonian, Croatian, 1

 

 

Table 12: L1 backgrounds of L2 learner Group

 

 

124

Table 13: Descriptives of group performance on experimental task (SL
=Unilingual learners of English group; ML = Multi-lingual
learners of English group; ENS = English control group)

 

95% Conﬁdence ‘
Interval for Mean

 

 

 

 

N Mean St. Dev. St. Error Lower Upper
I—r r Bpund Bound
SL 18 18.6667 2.8901 .6812 17.2294 20.1039
T41. 14 18.4286 2.1381 .5""_'714 "—17.1941 19.6631
TENS 29 217931 1.6340 .3034 21.171? 22.4146

 

 

 

 

 

 

 

 

 

 

Table 14: Contrast tests comparing mean scores of learner groups with ENS
control (contrast 1) and mean scores of SL group with ML group

 

 

 

 

 

(contrast 2)

Contrast Value of StcfEnor t df Sig. (2-
* Contrast 1 tailed)
Does not 1 3.2455 .5382 6.030 52.229 .000
assume
equal
variances 2 .2381 .8891 .268 29.950 .791

 

 

 

 

 

 

 

 

 

scores for the English native speakers were signiﬁcantly (p < 0.003) higher
than the learner scores

In addition to comparing overall performance, one of the purposes of
this study was to examine the extent to which learners differentiate between
finite and non-ﬁnite embedded clauses in the domain they set for English
reﬂexives. Table 13 shows the mean, median, minimum and maximum
scores as well as the standard deviation for the ﬁnite and non-finite
sentences in the test. The scores are given on a scale of 8 because there
were two sentence types (local and non-local) for the finite and non-ﬁnite

items.

125

The results inTable 15 show that the ENS cr'mtrol group does not
appear to treat ﬁnite and non~ﬁnite embedded clauses differently in its -
assignment of an antecedent to the reﬂexive. The mean score for ﬁnite
sentences is 7.4483, while the score for non-ﬁnite sentences is 7.5517.

Likewise, the ML group exhibits similar mean scores for both ﬁnite

Table 15: Case Summaries - ﬁnite and non-finite sentences (Maximum
score = 8;Minimum score = 0; SL = Unilingual learners of English
group ML = Multilingual learners of English group; ENS = English

speaking control group)

 

126

and non-ﬁnite sentences; although the overall means, at 6.2143 and 6.0714
for the ﬁnite and non-ﬁnite sentences respectively, are lower than those for
the ENS group .

The SL group, however, does show differences in accuracy for the
ﬁnite and non—ﬁnite sentences. The overall mean for ﬁnite sentences is
6.611 1, while the mean score for non-ﬁnite sentences is lower at 5.833.

A paired t-test was run on the mean scores on the ﬁnite and non-
ﬁnite sentences for the SL group. The results, shown inTable 16 below,
indicate that the difference in score, while not statistically signiﬁcant,
indicate that the SL learners appear to be treating the reﬂexive differently
when it is embedded in a non-ﬁnite clause as compared to a ﬁnite clause,
but more examination of the data is necessary. As I show in the next
section, analysis of individual learner grammars reveals that while a large
group of SL learners do not differentiate between ﬁnite and non-ﬁnite
clauses in the domain of the reﬂexive, a group of learners require a local
context in the ﬁnite domain but allow both local and non-local binding of

the reﬂexive in a non-ﬁnite context.

Table 16: Paired samples correlations for SL group

 

 

I LN [Correlation ll - i.g J
Pair Finite & Non 18 .421.208
1 Finite Sentences -

In the next section, the individual learner grammars are examined and the
results of this examination add some important context to the initial group

data reported above.

127

4.5.4. SL Individual Results. The individual results for the L2 group reveal
some interesting patterns. In these patterns, the subjects are loosely grouped
into similar grammar systems. Each of the groups includes members whose
intuitions exactly match the deﬁning characteristics of the system, but some
of the group members exhibited inconsistencies in their response patterns
and so inferences were made to ﬁnd the most logical place for those
particular learners. The response patterns for the subjects suggest two
different orientations: a subject-only orientation and a +] - subject
orientation. Three different domain patterns appear as well: a local-only
domain, a +] - local domain in both ﬁnite and non-ﬁnite contexts, and a
local-only domain in ﬁnite contexts with a +/ - local domain in non-ﬁnite
contexts. Furthermore, the intuitions represented by these subjects do not
provide evidence to support a correlation between domain and orientation,
as for each domain grammar represented, there are examples of subjects
with +] - subject orientation as well as subjects who exhibit a subjedcnly

orientation.

4.5.4.1 Target-Like Group. Subjects SL1, SL2, and SL3 appear to have
acquired the English system in that they consistently accept both the subject
and object NPs as antecedents for the reﬂexive. Additionally, these subjects
accept the local NP and reject the non-local NP as a possible antecedent in
both ﬁnite and
non-ﬁnite contexts.

Subjects SL4 and SL5 also require a local domain for the antecedent,
but show inconsistencies in their responses to the orientation sentences.

Subject SL4 is inconsistent in accepting the object as a possible antecedent, a

128

Table 17: L2 group - individual results (A = subject consistently (3 or 4 out
of 4) accepted this NP as a possible antecedent; R = subject
consistently rejected (3 or 4 out of 4) this NP as a possible
antecedent; I = subject accepted this NP as a possible antecedent for
two tokens and rejected this NP for two tokens (inconsistent))

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Orientation Domain Response
r finite non-ﬁnite pattern

_‘Su bL" 7L' '1 "'s"'u"bjec"t"' Object Local N. Local 'Local NTiocal
SL1
French A A A R A R Target
“SE2
Arabic A A A R A R Target
”81.3
Arabic A A A R A R Target
STA
Arabic A I A R A R Target
SL5
Thai I A A R A R Target
'sﬁ
Arabic A A A R I R Target
5177'
Hmong A A I R A R Target
"51.8
ﬂbic A A I R A R Targt
SL9 Target
Russian A R A R A R (+subj.)
SLTO Target
Korean A R A R A R (+subj.)
I§L11
Korean A A A R A A +/ - ﬁnite
81:12
_J_a_panese A A A R A A +/ - ﬁnite
SL13
Romanian A I A R A A +/ - ﬁnite
”51714
Japanese A A I R A A +/- ﬁnite
SL15 +] - ﬁﬁte
lhji A R A R I A (+subj) ,
SL16 +]- ﬁnite
Chinese A R A R I (+subj)
SL17
Japanese A A A A A "Yes"-type

 

 

 

 

 

 

129

 

 

 

 

 

 

SL18
Japanese A I A f R R R Unclassif.

 

grammar exhibited by a number of the ENS subjects. Subject SL5, however,
only accepted two of the four sentenCes which require binding with the
subject NP. This subject goes on, however, to consistently accept the object
NP as a possible antecedent. SL6 consistently accepts both the subject and
object NPs as possible antecedents, accepts the local NP and rejects the non-
local NP in the ﬁnite domain, yet is inconsistent in accepting the local NP as
a possible antecedent in the non-ﬁnite sentences. This learner nonetheless
consistently rejects the non-local NP in non-ﬁnite sentences. Because
subject SL6 consistently rejects the non-local NP in both ﬁnite and non-
ﬁnite clauses, and accepts both subject and object NPs as possible
antecedents, I also include her in the "target-like" group.

Subjects SL7 and SL8 consistently accept both subjects and objects as
antecedents for the reﬂexive. Likewise, they are consistent in accepting the
local NP and rejecting the non-local NP in non-ﬁnite contexts. The
inconsistency with these subjects lies in their acceptance of the local NP in
the ﬁnite context. This is unusual because these subjects consistently reject
the non-local NP in the ﬁnite context; both of these subjects rejected item
(35).

(35) Billy delivered newspapers every morning for his job. For three
months, Billy had been saving his money for a new bicycle. Billy ﬁnally
bought the bicycle and told Charlie about it, "Hey Charlie, look at the new
bicycle I bought!" _

Charlie heard that Billy had bought himself a bicycle. True] False

130

The target-like group is therefore made up of three subjects who
behave consistently in all aspects of English reﬂexives tested with this task,
and four subjects who appear to have acquired the English system, but due
to inconsistencies in one of the six sentence types tested, are only tentatively
included in the group.

In addition to the "target-like" group, are two learners (SL9 and SL10)
who require local binding in both ﬁnite and non-ﬁnite contexts, yet
maintain a subject-only orientation. These subjects consistently accepted the
subject NP and consistently rejected the object NP as a possrble antecedent.
It is difﬁcult to conclude with any certainty what this response pattern
reveals about the mental grammar of these learners. Looking back at the
ENS results, we see four native-speaker subjects who likewise consistently
reject the object NP as a possible antecedent. We know, however, from
earlier discussions that English allows both subject and object NPs as
possible antecedents. For the ENS speakers, then, we interpret the rejection
of the object NP as a symptom of a strong preference for the subject NP.
With this group of learners, however, it is equally possible that the mental
grammar of these learners instantiates a subject orientation. This is possible
due to the fact that these subjects, native speakers of Russian and Korean,
respectively, have morphologically simple (i.e., subject oriented) reﬂexives
in their L1 8. It is therefore impossible to say whether the mental grammars
of these subjects instantiate a subject orientation or a + / - subject orientation
with a strong preference for the subject NP.

4.5.4.2 Local Domain in the Finite Context] +/- Local Domain in the Non-

Finite Context. The second domain grammar exhibited by the learner group

 

131

is that in which the domain of the reﬂexive is determined by the ﬁnite] non-
finite nature of the embedded clause. When the reﬂexive is embedded in a
ﬁnite clause, it is restricted to the local domain. When the reﬂexive is
embedded in a non-ﬁnite clause, however, it can take either a local or non-
local NP as an acceptable antecedent.

Subjects SL11 and SL12 provided systematic responses on all six
sentence types. These subjects consistently accepted both subject and object
NPs as acceptable antecedents. In the ﬁnite context, these learners
consistently accepted the local NP and rejected the non-local NP. In the
non-ﬁnite context, however, subjects SL11 and SL12 consistently accept both
the local and non-local NPs as possible antecedents.

Two other subjects exhibited interpretations similar to subjects SL1 1
and SL12, but because these subjects answered one of the sentence types
inconsistently, it is difﬁcult to say with certainty that these learners belong
in the "ﬁnite/non—ﬁnite" group. Subject SL13 consistently differentiates
between finite and non-ﬁnite contexts. This person requires a local domain
in a finite context (accepting the local NP and rejecting the non-local NP)
and consistently allows both local and non-local antecedents in a non-ﬁnite
context. This subject, however, provided an inconsistent response pattern
on the sentences requiring binding between reﬂexive and object. As I have
already stated, it is difﬁcult to say whether the individual responses to the
object oriented items are due to competence or performance factors.

Subject SL14 exhibits a +/ - subject orientation, yet is inconsistent in
accepting the local NP in a ﬁnite context. Other than the inconsistent results
on the ﬁnite] local sentences, subject SL14 consistently rejects the non-local
NP as a possible antecedent in the ﬁnite sentences. In addition, he

132

consistently accepts both the local and non-local NPs as possible antecedents
in the non-ﬁnite sentences.

There is one example of a person who differentiates between ﬁnite
and non-ﬁnite clauses in her binding patterns, yet exhibits a subject
orientation. Subject SL15 accepts subjects and rejects objects as possible
antecedents, she also instantiates a local—only domain for the reﬂexive in the
finite sentences. In the non-ﬁnite sentences, this subject is inconsistent in
accepting the local NP as a possible antecedent. She does, however,
consistently accept the non-local NP as a possible antecedent.

Subject SL16 also exhibits a subject orientation and a local-only
domain in the ﬁnite sentences. This person, however, is inconsistent in
accepting non-local NPs in the non-ﬁnite sentence, while consistently
accepting the local NP. Given that this person is inconsistent in his
interpretation of the non-local NPs in the non-ﬁnite sentences, it is not at
all clear which group he best ﬁts into. He could ﬁt into the "target-like"
group, assuming that the two items in which he accepted long-distance
binding were due to factors other than his competence. Likewise, he could
ﬁt into the "ﬁnite / non-ﬁnite" group, assuming that his rejections of the two
long-distance sentences were due to factors other than his competence.
And, ﬁnally, all of his responses could be accurate reﬂections of his
competence. The inconsistent response pattern could therefore be due to
the fact that he is moving from one grammar (the ﬁnite] non-ﬁnite
grammar) to another (the target-like grammar).

4.5.4.3 "Yes" Group. Subject SL17 appears to allow binding in all contexts.
This subject consistently allows both subjects and objects as possible
antecedents for the reﬂexive. She also consistently allows local and non-

133

local NPs in both ﬁnite and non—ﬁnite contexts. It should be noted that
although this person consistently accepted the binding patterns in all
contexts, she did not answer "true" to all the items in the task. In order to
qualify for the experimental group, this subject needed to answer "true" to
one comprehension question and "false" to another. Furthermore, subject
SL17 also gave two "false" answers, although these "false" answers did not
affect the overall systematicity of the responses.

4.5.4.4 Unclassified. Subject SL18 does not appear to ﬁt into any of the
groups outlined above. This subject accepts subject NPs but is inconsistent
in allowing object NPs as possible antecedents. This is a pattern we have
seen before. Her answers to the domain questions are consistent, but do not
reﬂect any of the domain grammars above. Subject SL18 exhibits a local-
only domain in ﬁnite clauses, but consistently rejects both local and non-
local antecedents for the reﬂexive in the non-ﬁnite sentences.

4.5.5 ML Individual Results. The ML group exhibited patterns similar to
the L2 group, but with different representations across groups. The
individual results are given inTable 18, below.

4.5.5.1Target-Like Group. As with the L2 group, there are several members
of the multilingual (ML) group who exhibit native-like intuitions. Subjects
ML1 and ML 2 provide consistent responses across all sentence types. These
subjects accept both subjects and objects as antecedents for the reﬂexive. In
addition, they require a local domain for the reﬂexive. That is, they
consistently accept local NPs (in both ﬁnite and non-ﬁnite contexts) and
reject non-local NPs as possible antecedents.

134

Table 18: ML group - Individual responses (A = subject consistently (3 or 4
out of 4) accepted this NP as a possible antecedent; R = subject

consistently rejected (3 or 4 out of 4) this NP as a possible
antecedent; I = subject accepted this NP as a possible antecedent for
two tokens and rejected this NP for two tokens (inconsistent))

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Orientation Domain
Finite Non-finite 1
Subj Native languages] subj. obj. loc. n.loc loc. n.loc resp.
other L28 NP NP NP . NP NP . NP pttrn
”Kl—Li Italian] Turkish, A A A R A R target
__ Serbian _ _ ~ _
ML2 Arabic] Chaldean A A A R A R target
”Ki-ls German] Swiss A i A it A R target
_G_erman, Italian _ _
ML4 Chaldean] Arabic A I A R A R target
I———.: —— f
ML5 Assyrian] Arabic A A I IR A R target
_ ! l +subj
ML6 Serbian] A R A R A R target
Macedonian, +subj
Croatian, Slovenian,
German - '
ML7 Lithuanian] A R A R A R target
Russian _ +subj
ML8 ”ﬁeldean7 Arabic A it A "'_‘R _I R target
+subj
TIE Mandannf A A A ‘RT A 1 target
Taiwanese / yes
ML 10 Russian] French A I A A A A yes
MLll Russian] Ukranian A A I A A A yes
ML12 Albanian] Italian A A I A A A yes
ML13 Korean] German, A A A I A A yes
Japanese
ML 14 Chaldean I Arabic A A A A A A yes

 

 

135

Subjects ML3 and ML4 also require a local domain in both ﬁnite and
non-ﬁnite contexts and accept the subject NP as a possible antecedent for the
reﬂexive. These subjects, however, are inconsistent in accepting the object
NP as a possible antecedent. Once again, it is difﬁcult to say whether or not
this inconsistency is a reﬂection of the leamers' mental grammars. There
are examples of this response pattern (accept subject, inconsistent on object)
in both the L2 group and the ENS group.

Subject ML5 also appears to have a target-like grammar, in that he
accepts both subject and object NPs as antecedents for the reﬂexive.
Furthermore, he requires a local domain in the non-ﬁnite context and
rejects non-local NPs in the ﬁnite context. This person, however, was
inconsistent in accepting the local NP in the ﬁnite domain.

ML6 and ML7 require a local domain across ﬁnite / non-ﬁnite
contexts, yet consistently reject the object NP as a possible antecedent. These
subjects, then, exhibit a target-like grammar but with a subject orientation.

ML8 also appears to exhibit a target-like grammar with subject
orientation in that he accepts subjects and rejects objects as possible
antecedents for the reﬂexive. This person also requires a local domain in a
ﬁnite context, and consistently rejects non-local NPs in a non-ﬁnite context.
The response pattern for the local NP as a possible antecedent in a non-ﬁnite

context is inconsistent, however.

4.5.5.2 Local Domain in the Finite Context] +/- Local Domain in the Non-
Finite Context. Among the ML group, only subject ML9 ﬁts the
requirements of the ﬁnite] non-finite grammar pattern, but his response
pattern makes his membership in this group tenuous, at best. ML9 allows
both subjects and objects as possible antecedents for the reﬂexive. This

136

subject also requires a local domain by consistently accepting local NPs and
rejecting non-local NPs in a ﬁnite context. ML9 also consistently accepts the
local NP in a non-ﬁnite context, yet is inconsistent in accepting the non-
local NP in a non-ﬁnite context. As with subject SL17, who also exhibited
this pattern, it is difficult to say whether or not ML9 really allows non-local
NPs, or does not allow non-local NPs as possible antecedents for English
reﬂexives. Furthermore, it is also possrble that this person is in the process
of restructuring his grammar and so his acceptance of non-local NPs is
variable. All of these possibilities will be taken up again and discussed in
the next chapter.

4.5.5.3 "Yes" Group. Subjects ML10, ML11, ML12, ML13, and ML14 appear to
accept NPs in all grammatical roles and contexts as possible antecedents for
reﬂexive pronouns. Assuming that the inconsistent response patterns
(object - subject ML10, local NP/ ﬁnite context ML11, local NP] ﬁnite context
ML12, non-local NP] ﬁnite context ML13) indicate acceptance of that
particular NP, these learners accept subjects and objects as antecedents, and
they also accept local and non-local NPs (inﬁnite and non-ﬁnite contexts) as
antecedents for the reﬂexive.

As with the L2 group, this behavior does not appear to be the result of
a "true" response bias, as all of these subjects did provide "false" answers on
the test. Nonetheless, their answers suggest that any NP in any of the
grammatical roles tested are acceptable antecedents for the reﬂexive
pronoun.

While the group data analysis did not reveal any significant
differences between the ML and the SL groups, the individual grammar
analysis does. The ML group is most strongly represented by "target-like"

137

grammars and "yes" grammars, whereas the SL group is most strongly

represented by "target-like" grammars and "+] - ﬁnite" type grammars. -

4.6 Chapter summary

The analysis of individual grammars in this study reveal several
consistent response patterns. The ENS control group overwhelmingly
rejected non-local binding (in 26 out of the 27 subjects with one inconsistent
pattern or fewer). While some (six) either equivocal or consistently rejected
local object binding, twenty subjects responded in perfect accordance with
theoretical predictions; that is, these subjects consistently accepted both
subject and object antecedents in the local domain and required local
antecedents (also rejecting non-local antecedents) for the reﬂexive.

The learner groups also contain learners who exhibit local-only
binding pattern, both with and without a subject orientation. Slightly more
than half of the SL and ML groups were made up of learners who
consistently rejected non-local antecedents for the reﬂexive across contexts.

Two interlanguage patterns were also found. These patterns were
disproportionately represented by one or the other learner groups. The SL
group showed a large percentage (33.33%) of learners who distinguished
between ﬁnite and non-ﬁnite contexts. These learners consistently rejected
non-local binding in a ﬁnite context, yet allowed non-local antecedents for
the reﬂexive in a non-ﬁnite context (for all but one learner, who accepted
two and rejected two sentences in which the reﬂexive referred to a non-local
NP in a non-finite context).

Finally, a large percentage (36%) of the ML group learners (with one
SL learner) exhibited a "yes"-type grammar which allowed the reﬂexive to
ﬁnd an antecedent in any domain (local / non-local) and context (ﬁnite/ non-

138

ﬁnite). The possible reasons for these different interlanguage response
patterns, as well as the implications these results have for second language
acquisition theory in general are the topic of the next chapter.

 

Chapter 5
DISCUSSION AND CONCLUSIONS

5.1 Introduction

This dissertation study has resulted in the identiﬁcation of three
distinct learner response patterns. Approximately a third of the ML learners
exhibited a "yes"-type response pattern. This response pattern is
characterized by the acceptance of antecedents for the reﬂexive in both
domains (local and non-local) and contexts (ﬁnite/ non-ﬁnite embedded
clause). I argue in this chapter that this behavior represents an early stage of
SLA whereby the ML learner (who, is more likely to entertain "wider"
grammars) is more likely to remain at the "yes" stage of acquisition.
Furthermore, the ML learners in this dissertation study also had simplex
reﬂexives (monomorphemic, LDRs) in one of the background languages. I
suggest that learners who have LDRs in their ﬁrst language are also more
likely to remain at the "yes" stage of acquisition.

The SL group, however, was predominated with learners who
exhibited a +/ - ﬁnite grammar. This grammar is characterized by requiring
a local antecedent for the reﬂexive in a ﬁnite context, but allowing both local
and non-local antecedents for the reﬂexive in a non-ﬁnite context.

I suggest that these learners, whose ﬁrst languages include LDRs,
have come to recognize that English reﬂexives require both a referent and a

139

140

structural antecedent. Thus, assuming that the learners in this category
have mistakenly analyzed the English reﬂexive as a simplex, LDR, these
learners restrict binding to the lower clause in the ﬁnite context (because
AGR is present in the ﬁnite context), but allow both local and non-local
antecedents for the reﬂexive in the non-ﬁnite context (because AGR is not
present, thus expanding the governing category to include the higher clause
(Progovac and Connell, 1991; White, 1995; Bennett, 1993, 1994).

Both the SL and ML groups, however, were predominantly made up
of learners who exhibit target-ﬁke (local-only) grammars. This chapter also
addresses the issue of orientation and domain. Because the design of the '
experimental task does not elicit data that may disprove the assumption that
orientation and domain are not linked, I argue that the data from this
dissertation study does suggest that orientation and domain may be acquired
separately.

Finally, this chapter provides some ﬁnal conclusions of the study as

well as suggestions for future research.

5.2 L2 Domain patterns

The research objective of this dissertation study was to examine the
intuitions of L2 learners of English as these intuitions relate to the
acquisition of reﬂexive pronouns. In Chapter Two I reviewed the results of
recent studies of the acquisition of reﬂexives. The results of many of these
studies are analyzed in terms of a transfer-based account of the SLA of
reﬂexives. Some authors suggest that some learners transfer the complex
reﬂexive from the native language and apply it to the L2, English (Yuan,
1994). Other studies suggest that some learners may initially transfer the
simplex reﬂexive from the native language and apply it to the L2

141

(Lakshmanan and Teranishi, 1995). Still others argue that learners initially
misanalyze the English (complex) reﬂexive as a simplex reﬂexive (Franks
and Connell, 1996 for child language acquisition). For the latter two
explanations, the learner eventually discovers that English reﬂexives are
complex and so acquire the target form. Finally, researchers such as
Progovac and Connell (1996) and White et al. (1997) propose that learners
from certain language backgrounds (e.g., Japanese and Korean) transfer the
simplex reﬂexive from the L1 and apply it to the L2; yet, these learners
simultaneously recognize the agreement features on ﬁnite embedded
clauses. Assuming a relativized SUBJECT analysis, the result of these
assumptions would be a learner whose interlanguage allows long-distance
binding in a non-ﬁnite context, but requires local binding in a ﬁnite context.
I propose that while these explanations of the acquisition process
adequately account for the data of the individual studies, these explanations
do not account for the data across studies. That is, while it is reasonable to
assume that because all languages contain complex reﬂexives, L2 learners
should logically transfer the complex reﬂexive) from the L1 and apply it to
the L2. The data, however, suggest that many learners allow reﬂexive
binding to local and non-local antecedents in English. That is, these studies
suggest that learners are not, in fact, transferring the complex reﬂexive from
the L1 because they allow agreement with non-local antecedents. Secondly,
the SL research also provides evidence suggesting that learners are not
transferring the simplex reﬂexive from the L1 and applying it to the L2.
This research reports ﬁndings that learners from both types of ﬁrst language
backgrounds (those whose Lls contain simplex and complex reﬂexives
(Japanese, Korean, Chinese) as well as those whose Lls contain only
complex reﬂexives (French), allow local and non-local antecedents for

142

English reﬂexives. Thirdly, I provide arguments by Hamilton (1996) which
call into question the interpretation of learner data by White (1995) and
Progovac and Connell (1991).

Given the problems inherent with current explanations of the
interlanguage stages of the L2 acquisition of reﬂexive pronouns, the
objective of this dissertation is to focus on the acquisition stages of children
acquiring English reﬂexives. The acquisition stages proposed by Franks and
Connell (1996), provide one possible interpretation of the acquisition process
which is compatible with current and recent L2 data. This dissertation study
sets out to investigate this possibility (that adult learners of English might
follow acquisition stages similar to those identiﬁed in Franks and Connell
(1996) for child language acquisition). Speciﬁcally, the experimental task was
designed to consider the possibility that second language learners of English
go through the same acquisition stages as those identiﬁed in Franks and
Connell (1996).

Franks and Connell identify four distinct acquisition stages for
monolingual learners of English. These are summarized in Table 19.

I point out in Chapter Three that these ﬁndings, while much more speciﬁc
in their description of the characteristics of each acquisition stage, are
consistent with the results of SLA studies of reﬂexives. In fact, the
characteristics of the LD-phase as Franks and Connell have described it,
provide a possible explanation for the peculiar behavior of some L2 learners
of English where they bind the reﬂexive to a long distance antecedent more
often in a nonﬁnite context than ﬁnite.This dissertation study, then, was

 

143

Table 19: Acquisition stages of child language learners (both normal and

SL1) as identiﬁed by Franks and Connell (1996) for English
reﬂexives

 

Stage

Characteristics

 

Child does not yet identify the reﬂexive as an anaphor, treats the

reﬂexive as an independently referring noun meaning "bod "

 

Child realizes that the reﬂexive is an anaphor in need of an
antecedent, selects the nearest NP (the local object)

 

Child is unable to parse the reﬂexive and so does not understand
that the reﬂexive is morphologically complex. Child interprets the
reﬂexive as a morphologically simple LDR, resulting in LB binding
patterns as well as a subject orientation. PRO appears to be
unavailable as a binder to these learners, resulting in local/ non-
local binding patterns in a ﬁnite context, but non-local only binding

in a non-ﬁnite context.

 

 

IV

 

Child's language facility matures and s / he learns to analyze the
reﬂexive as morphologically complex. This triggers a move to a

local domain with subject/ nonsubject orientation

 

 

144

designed to speciﬁcally consider the possibility that L2 learners of English
exhibit interlanguage patterns similar to those identiﬁed in Franks and
Connell. In the event that L2 learners exhibit similar interlanguage
patterns, it would be possible to suggest the possibility that the L2 acquisition
of English reﬂexives is similar to the L1 acquisition process. In the event
that different interlanguage patterns are identiﬁed, those patterns could be
described and considered as to how and possibly why they differ from those
patterns identiﬁed by Franks and Connell. In the following section I show
that while this study was able to identify three distinct domain patterns,
some of these response patterns are different from those identified and
described in Franks and Connell (1996).

Three distinct response patterns were identified in this study. One
pattern, referred to as the "yes"-type response pattern, was characterized by
responses which allowed NPs in all domains (local/ non-local) and contexts
(ﬁnite/ non-ﬁnite). The second pattern, referred to as the LD or +/ - ﬁnite
pattern, was characterized by an exclusively local domain in the ﬁnite
context, but a non-local domain in the non-ﬁnite context. The third
response pattern, referred to as the target-like pattern, was characterized by a
local domain in both ﬁnite and non-ﬁnite contexts.

5.2.1 Grammatical Patterns and Grouping. The subjects for this study have
been grouped according to the grammatical patterns they exhibited in
completing the experimental task. This methodological choice creates a
clearer picture of the characteristics of various possible interlanguage
grammars in which learners operate. In grouping the learners according to
their grammatical patterns, this methodology is able to identify and describe
the various characteristics of learner grammars in a way that traditional

145

grouping (e. g., by language background or proﬁciency level) or data analysis
(e.g., by comparing group mean scores) are unable to do. Conversely, this
methodological approach has sacriﬁced some explanatory power because the
grouping takes place post-hoc. Arguments and possible explanations can be
made, however, especially with regard to ﬁrst language inﬂuences on the
various IL grammars for the learners as well as some important differences

found between the SL and ML learners in this project.

5.2.2 Pragmatic L2 Strategy: "Yes"-type Response Pattern. The individual
grammatical analysis reveals (that all but one of the subjects from the ﬁnal
subject pool exhibited one of three distinct binding domains. The widest
domain grammar was that exhibited by the "yes" group who consistently
accepted both local and non-local NPs in both ﬁnite and non-ﬁnite contexts.

5.2.2.1 L1 Acquisition of Reﬂexive Pronouns: Early Stages. Franks and
Connell (1996) discuss a study by McDaniel, Cairns, 8r Hsu, (1990) which
relied on the analysis of individual grammars of children and identiﬁed
that at ﬁrst, children do not recognize the reﬂexive pronoun as an anaphor.
Instead, these children interpret the reﬂexive as an independently referring
noun phrase, not subject to the rules of structural binding. Franks and
Connell included two components in their experimental design to identify
children at this stage of acquisition. First of all, the experimental task
included several sentences designed to test the child's adherence to c-
cornmand. If a child allowed non-c-commanding antecedents for the
reﬂexive, a second test was administered where the child was given two
male dolls and asked to show the doll hitting himself and hitting him (the
other doll). A contrast in behavior between the tWo actions suggests an

146

awareness of the reﬂexive as an anaphor; no such contrast would suggest
that the child does not interpret the reﬂexive as an anaphor. This method
resulted in identifying one SLI child (age 4;11) who did not recognize the
reﬂexive as an anaphor. This child exhibited a "yes"-type response pattern
in that the child agreed with every binding possibility presented in the
experimental task (Franks and Connell, 1996: 449).

Previous SLA studies have not employed this approach to
investigating whether or not c-command violations reﬂect a lack of
awareness of the anaphoric nature of the reﬂexive. However, several SLA
studies have examined learners for evidence of their awareness of the c- .
command rule (Thomas, 1991, 1995; Christie, 1992). These studies report the
interesting finding that some L2 learners allow reﬂexive pronouns to
identify non-c-commanding NPs as acceptable antecedents for the reﬂexive,
despite the fact that the c-command requirement, a requirement instantiated
by UG, is assuredly in place and active in the mental grammars of the

leamers' native languages.

5.2.2.2 Acquiring the Referential Dependency of the L2 Reﬂexive. It is
difﬁcult to say exactly why L2 learners may allow non-c-commanding
antecedents for the reﬂexive. As I have mentioned, this behavior among
children is interpreted as evidence that the child is at an early stage of the
acquisition process and does not interpret the reﬂexive as an anaphor
subject to the rules of structural binding. Given that L2 acquirers are adults
who have presumably come to interpret reﬂexives in their native language
as anaphors, they should not have to re-learn the concept of anaphor.
However, L2 learners, while linguistically mature enough to be familiar
with the concept of anaphor, still need to identify which English words are

147

independently referring nouns, which nouns are reﬂexive anaphors and
which nouns are regular pronouns. In other words, the learners need to
sort out the different referential dependencies of each noun in addition to
acquiring the target-like syntactic constraints on binding.

Gass (1987), in a study I discuss at length in Chapter Two and later in
this chapter, showed that second language learners initially become aware of
or notice salient features of the L2 before they are able to apply the full range
of linguistic knowledge applicable to that feature. In this sense, L2 learners
may initially recognize the referential dependency of the reﬂexive, and the
importance of linking the reﬂexive with a referring NP, but may not yet be
able to call upon the syntactic constraints of structural binding such as c-
command, the clause-mate condition, etc. This initial phase of noticing the
referential dependency would result in the learner behavior reported in the
L2 studies above. That is, learners would understand the need to assign a
referent for the reﬂexive but would not be able to differentiate between
structurally available (i.e. c-commanding) and unavailable antecedents. The
result would be that the learner would, in effect, be relying on pragmatic
cues to satisfy the referential dependency introduced by the reﬂexive.

5.2.2.3 Yes - Group: Non-Structural Binding. Some supporting evidence for
this possibility can be found in the individual grammars of the subjects for
this dissertation study. One group of learners exhibited a "yes" type domain
pattern. That is, these learners allowed the reﬂexive to identify NPs in any
domain or context as acceptable antecedents for the reﬂexive. Both local and
non-local NPs in both ﬁnite and non-ﬁnite contexts (as well as both subject
and non-subject NPs in most cases) were acceptable binders for the reﬂexive

among these learners. There were learners of this type in both the L2 and

148

the ML group, although there was a much higher percentage of "yes"-type
learners in the ML group than in the L2 group (shown in Table 20).

Table 20: Percentage of learners consistently accepting antecedents for the
reﬂexive in all domains and contexts1

Ps1. '1m |

[5.6% l35.71% l

 

 

 

While these learners accepted NPs in both ﬁnite and non—ﬁnite contexts as
well as local and non-local domains, they did not accept every item on the
task. All of these subjects accurately rejected the comprehension item which
required a "false" answer. Furthermore, in many cases these learners were
consistent with 3 out of the 4 tokens for some of the sentence types,
suggesting that members of this group were not exhibiting a "yes" response
bias. Finally, there are some cases where one of the six sentence types
resulted in an inconsistent response pattern, also entailing some "no"
answers.

Table 21 outlines the ﬁrst language background of those learners
exhibiting a "yes"-type response pattern. This response pattern, as I have
stated, is characterized by the acceptance of local and non-local NPs in ﬁnite
and non-ﬁnite contexts as well as (in most cases) subject and object NPs in
the local domain. This binding pattern does not violate UG constraints on
binding. Icelandic sig and Serbo-Croatian sebe reﬂexives exhibit just such a
binding pattern (nonsubject oriented (NSO), LDR); thus, it is possible that

 

1 Some of these learners provided inconsistent responses for one of the six
sentence types. I suggest in Ch. 4 that because these learners consistently
accepted the other ﬁve sentence types, they belong in the "yes"-group.

149

Table 21: First language backgrounds of subjects exhibiting a "yes" respons

 

 

 

 

 

 

 

 

 

 

 

pattern . '
Subject L1 Other Lls
ML10 Russian French
MLll Russian Ukranian
ML12 Albanian Italian
ML13 Korean German, Japanese
ML14 Chaldean Arabic
SL 17 Japanese n/ a

 

these learners are interpreting the English reﬂexive as a morphologically
simple LDR which instantiates a nonsubject orientation in the local
domain.

While there is no direct evidence to contradict this possibility, the
data from other studies (Thomas-1991, 1995; Christie, 1992) that have
identiﬁed learners who allow non-c-commanding NPs as a possible
antecedent, raise the possibility that some L2 learners bind the reﬂexive
discoursally (exempt from the structural constraints on reﬂexive] antecedent
relationships) at some stage of the acquisition process. Furthermore,
Hamilton (1996) argues that one cannot automatically equate LD binding
with an X0 analysis of the reﬂexive. He points to a study by White (1995) in
which White tested L2 learners on their awareness of gender and agreement
properties of English reﬂexives. She reported that learners were aware of
the agreement properties of reﬂexives, suggesting that these learners

150

understood the morphological complexity of the reﬂexive. Additionally,
these same learners exhibited a high degree of LD binding, raising the
possibility that these learners allowed LD binding of XP reﬂexives.
Hamilton points to this ﬁnding to suggest the possibility that learners who
exhibit LD binding may not be interpreting (or transferring) the reﬂexive as
an LDR, but rather, are transferring a logophoric interpretation of the
reﬂexive.

As I discuss in section 1 .7.3 of Chapter One, logophoric binding is a
non-structural, discoursal form of binding where the reﬂexive is .
semantically linked with its antecedent. Thus, Hamilton (1996) suggests that
one cannot rule out the possibility of non-structural binding. Hamilton
(1996: 433) admits, however, that logophoric binding is infrequent in the L2
data and so the only way this possibility could occur would be for learners to
transfer the logophoric strategy from the L1. Given the overall infrequency
of logophoric binding in both the L1 and the L2, it is difﬁcult to account for
all LD binding based on logophoric transfer}

However, the possibility that learners bind the reﬂexive non-
structurally remains strong. Demirci (1997) and Glew et a1. (1993) have
suggested that learners from some language backgrounds may rely more
heavily on pragmatic strategies than on other language groups.
Furthermore, I suggest in Chapter Two that studies by Gass (1985, 1987),
Harrington (1987), Sasaki (1991, 1994), Wulfeck etal., (1986) have made a

 

2 One subject in this study who allowed LD binding in both ﬁnite and
nonfinite contexts. This subject was a bilingual speaker of Arabic and
Chaldean. Hamilton (1996;433) points out that LD binding of an Arabic
learner of English could not be explained through logophoric transfer, as
Arabic does not have logophoric binding. This subject's LD binding could
suggest against logophoric binding; however, this person also speaks
Chaldean and it is not clear whether or not logophoric binding is present in
Chaldean.

151

strong case that in other components of the grammar (in this case, the
assignment of thematic roles to the NPs in the sentence), learners who
come from an L1 for which semantic cues are strong, continue to rely
heavily on the semantic cue in the L2. They are also less willing to let go of
the semantic strategy in favor of the L2 dominant syntactic strategy.

The "yes" type learners in this dissertation study may represent a
similar situation. With the exception of one subject, all the "yes" type
learners come from languages which include both SD and LD reﬂexives.
The use of a pragmatic strategy for reﬂexive binding may vary among
learners, depending on their ﬁrst language. Additionally, the results of the
studies by Demirci (1997) and Glew et al. (1993) suggest that native speakers
of languages such as Japanese and Turkish (which have both complex and
simplex reﬂexives) are more sensitive to pragmatic information for
antecedent assignment of reﬂexive pronouns than native speakers of
languages with only complex reﬂexives such as English. Furthermore,
Demirci (1997) ﬁnds that native speakers of Turkish continue to be sensitive
to pragmatic cues when learning English as a foreign language.

These ﬁndings bear a striking resemblance to those reported in the
Competition Model studies (Gass, 1985, 1987; Harrington, 1987; Sasaki, 1991,
1994; Wulfeck et al. ,1986). Learners whose L1 requires a heavy reliance on
pragmatic and semantic cues tend to 'hang on" to the pragmatic strategy and
move more slowly toward a reliance on syntactic cues than vice versa.
Likewise, Glew et al. (1993) and Demirci report that native speakers of
languages such as Turkish and Japanese (SDR/ LDR languages) rely heavily
on pragmatic / semantic cues and move to a stronger reliance on syntactic
cues reluctantly, if at all.

152

As I discussed earlier, SLA studies of reﬂexive pronoun acquisition
have shown that some learners initially allow reﬂexives to identify non-c-
commanding antecedents. I have suggested that this is a reﬂection of an
early stage of acquisition where the learner recognizes the referential
dependency of the reﬂexive, but is unable to call upon and employ the
syntactic constraints on structural binding. The Competition Model
research discussed above has found that pragmatics and semantics are relied
upon strongly and early on in the acquisition process. An understanding of
and reliance on syntactic cues develops more slowly and is characterized by
an initial awareness of the importance of the cue before the actual syntactic
rules can be called upon.

The IL grammars identiﬁed in this dissertation study support these
ﬁndings. The "yes"-type learners appear to recognize the referential
dependency of the reﬂexive and select a referent for the reﬂexive. However,
their behavior suggests that these learners are employing a non-syntactic
strategy in assigning reference for the reﬂexive by allowing any
pragmatically reasonable NP to serve as the referent.

. Furthermore, the preponderance of learners in the "yes" group with
native languages which contain both LD and SD reﬂexives (shown in Table
22), suggests that learners with this type of language background remain at
this stage of acquisition longer than learners whose Lls contain only SDRs.
This ﬁnding is in line with the studies by Glew et al. (1993) and Demirci
(1997) who report that learners from such SDR/ LDR language backgrounds
as Japanese and Turkish rely more strongly on pragmatic cues in their
native languages than English native speakers do for English. Furthermore,
Demirci (1997) reports that Turkish EFL learners not only rely heavily on

153

pragmatic cues in their native language, they continue to assign great
importance to pragmatic cues in their L2 English.

Thus, the learners in the "yes" group of this dissertation study
represent a group of learners from SDR/ LDR ﬁrst languages who are
exhibiting a reluctance to let go of their pragmatic strategy and begin
focusing on the syntactic element of reﬂexive binding.

Table 22: First language backgrounds of learners exhibiting a "yes"-type
response pattern

 

 

 

 

 

 

 

 

Subject L1 Other languages
SL17 Japanese na

ML10 Russian French

MLll Russian Ukranian

ML12 Albanian Italian

ML13 Korean German, Japanese
ML14 ‘ Chaldean Arabic

 

 

 

 

A second communality across the "yes"-type learners (with the
exception of SL17) is that they are multi-lingual learners of English. This
factor may also contribute to slowing learners down from moving to
consider the syntactic elements of reﬂexive binding. As discussed in section
Zobl (1995) reports that multi-lingual (ML) ESL learners3 appear to function
with wider grammars (allowing a wider range of structures and

 

3 Multilingual learners are learners who know languages beyond their
native language. Unilingual learners (UL) are learners who speak one
language and are learning another as a second language.

 

154

relationships) than unilingual (UL) ESL learners. Zobl (1995) found that ML
learners were less likely to respect the c-command requirement than their
UL (SL) counterparts. This ﬁnding can be interpreted as indirect evidence
that ML learners may stall in the stage of acquisition where learners can
only recognize and call upon pragmatic strategies to satisfy the referential
dependence of the reﬂexive.

It therefore appears that the "yes"-type response pattern may be the
result of a combination of two factors. First of all, the majority of learners
come from ﬁrst language backgrounds which contain both LDRs and SDRs
and, thus, are more reluctant to let go of the pragmatic strategy and focus on
the syntactic cues so important for English binding. Secondly, these learners
are largely, ML learners who also appear to allow non-structural binding to a

greater ex tent than UL learners.

5.2.3 LD Group (+/ - Finite Group). As I suggest above, ESL learners appear
to begin the acquisition task by allowing non-structural binding in all
contexts for the reﬂexive. Some learners stay at this phase longer than
others. Previous research suggests that ML learners allow more non-
syntactically constrained binding than UL learners. Also, SLA research
(Gass, 1985, 1987; Harrington, 1987; Sasaki, 1991, 1994; Wulfeck et al. ,1986)
has suggested that beginning learners tend to rely heavily on semantic and
pragmatic cues, adding "weight" to syntactic cues develops more slowly.

These two factors may have combined to prolong the non-syntactic
phase of acquisition for the ML learners with LDR/ SDR Lls. Sooner or later,
however, learners come to recognize the importance of the syntax in
determining which NPs can serve as antecedents for the reﬂexive and
which NPs are excluded.

155

A second domain pattern identiﬁed in this dissertation study reﬂects
a growing focus on the syntactic elements of the L2 grammar. In Chapter
Three I suggest the possibility that L2 acquirers go through a phase of
acquisition similar to the LD phase identiﬁed in Franks and Connell (1996).
Franks and Connell argue that previous research on L1 acquisition has
provided both direct and indirect evidence that children learning either
SDR or SDR/ LDR languages go through a stage of LD binding on their way
to acquiring the target system (Read G: Hare, 1979; Matthei, 1981; Wexler 8r
Chien, 1987; Chien & Wexler, 1990; Hyams 6r Sigurjénsdéttir; 1988, 1990;
Sigurjénsdéttir & Hyams, 1991, 1992). Franks and Connell also identify an
LD stage for child acquirers of English. Through their emphasis on
individual grammars and their VMTT experimental task, Franks and
Connell are able to provide a detailed picture of the characteristics of this LD
phase. Franks and Connell identify one group of learners whose response
pattern allowed both local and non-local antecedents for reﬂexives in a
ﬁnite context, but required a non-local antecedent for the reﬂexive in a non-
ﬁnite context. In addition, these learners exhibited a subject orientation for
the reﬂexive in the ﬁnite context.

In Chapter Three I suggest the possibility that L2 learners also pass
through an LD phase in their acquisition of reﬂexive pronouns.
Furthermore, I raise the possibility that the LD phase of acquisition for L2
learners is also characterized by the behavior identiﬁed in Franks and
Connell but because of the way the data have thus far been reported, the
specific characteristics of this stage have been obscured.

This dissertation study identiﬁed a group of learners whose binding
domains are sensitive to the nature of the clause in which the reﬂexive is

embedded. When the reﬂexive is embedded in a non-ﬁnite clause, it can

156

take either a local or non-local antecedent. In a ﬁnite context, however, the
reﬂexive instantiates a strictly local domain. This response pattern reﬂects
the ﬁndings of previous SLA research and contrasts with the ﬁndings in
Franks and Connell (1996).

The response pattern for this LD group is characterized by a local-only
interpretation of the reﬂexive in a ﬁnite context and a local/ non-local
binding pattern in a non-ﬁnite context. Subjects SL11, SL12, SL13, SL14, SL15
and ML9 appear to be applying a different binding domain for the reﬂexive
in ﬁnite and non-ﬁnite contexts.

As Table 23 indicates, this response pattern is strongly represented by
the SL learners and nearly absent in the ML group. Five of the 18 SL
learners (28%) exhibit this LD pattern yet only one ML learner could ﬁt this
pattern. In fact, this ML learner gave an inconsistent response pattern for
the critical non-local test items in a non-ﬁnite context, making it uncertain
whether or not this individual best belongs in the LD or target-like group.

Table 23: First language backgrounds of learners exhibiting LD binding

 

 

 

 

 

 

 

 

pattern

Subject L1 Other Languages
SL11 Korean na

SL12 Japanese na

SL13 Romanian na

SL14 Japanese na

SL15 Thai na

ML9 Mandarin Taiwanese

 

 

 

 

157

Several researchers have provided explanations for the distinction
some learners make between ﬁnite and non-ﬁnite embedded clauses in
binding patterns for English reﬂexives.

Hamilton (1996) provides a thorough critique of the researchers who
have sought to account for this LD phase through the Relatavized SUBJECT
Analysis. .

Progovac and Connell (1991) reconsider the data from Thomas (1989).
In this study, Thomas studies Spanish and Chinese native speakers learning
English. While the study reported LD binding for both groups, the Spanish
speakers allowed more non-local antecedents (although not signiﬁcantly so).
Progovac and Connell (1991) proposed that while both the Spanish and
Chinese groups incorrectly analyzed the reﬂexive as X0, the Spanish
speakers were not able to recognize that English has overt AGR due to how
little overt agreement marking exists in English as compared to Spanish.
Conversely, because Chinese has null AGR, these learners were able to
recognize and respect AGR. Hamilton (1996) challenges this interpretation
of the data and argues that while Thomas ﬁnds the Spanish learners bound
LD more than the Chinese learners, that difference was not signiﬁcant.
Thus, Progovac and Connell's claims that AGR is somehow more or less
recognizable by learners depending on their L1 background are
compromised. Furthermore, Hamilton discusses a study by Christie (1992)
who studied English speakers learning Spanish and Spanish speakers
learning English. She reported LD binding for both groups, the ENS
learners of Spanish actually showed higher rates of LD binding than the
SNS learners of English. This ﬁnding also contradicts Progovac and
Connell's prediction that differences in the prominence of agreement

158

marking in the L2 (as compared with the L1) somehow affects the LD
binding patterns of learners.

White ( 1995) also adopts an explanation of her data similar to the one
by Progovac and Connell (1991). White found that French speaking ESL
learners also exhibited considerable LD binding. In this case, White suggests
that the learners misanalyzed the reﬂexive as X0 and, because French has a
more obvious agreement system, failed to recognize that English has overt
AGR. This resulted in the LD binding by the French speaking ESL learners.

Bennett (1993, 1994) also suggests this explanation for her ﬁnding that
Serbo-Croatian speaking learners of English allowed LD binding into ‘
embedded non-ﬁnite clauses but not ﬁnite. The learners exhibited this
behavior despite the fact that the dialect of Serbo Croatian spoken by the
subjects does not have object control inﬁnitives. Bennett suggests that these
learners transferred the X0 reﬂexive from their native language and also
recognized overt AGR in the L2, English. This resulted in the leamers'
different binding domains in the ﬁnite and nonﬁnite contexts.

Hamilton suggests that tense, rather than AGR is the opacity factor for
movement of the reﬂexive. However, another possibility is that the
Relatavized SUBJECT Analysis can still account for the data if one re-
analyzes the role of the learner's L1 in the acquisition of +[AGR].

Hamilton's main criticism of Progovac and Connell (1991), White
(1995), and Bennett (1993; 1994) is that English AGR does not appear to be
any more "recognizable" for one language group or another. It appears that
learners of all L1 backgrounds LD bind. The Thomas (1989) data show that
learners from -[AGR] Lls such as Chinese LD bind. The data show that
Spanish learners of English and English learners of Spanish LD bind. The

159

data also show that Serbo-Croatian learners of English LD bind in a non-
ﬁnite context.

It appears that these studies (and other SLA studies which report
similar data) are attempting to account for two different types of data: data
that L2 learners LD bind (e.g., Thomas 1989, 1991; Demirci, 1997;White et al.
1995) as well as data that show L2 learners LD bind only in a non-ﬁnite
context (e.g., Finer 6r Broselow, 1986; Hirakawa, 1990; Bennett, 1993; White et
al., 1997). Much of the data showing overall LD binding have come from
studies which examined group means and] or did not include test sentences
investigating binding patterns in a non-ﬁnite context so it is difficult to tell
if the LD percentages reported are the result of LD binding across contexts or
if these learners disﬁnguished between ﬁnite and non-ﬁnite contexts. As I
discussed in Chapter Two, these studies have found LD binding in learners
whose L1 backgrounds instantiate SD—only binding or SDR/ LDR binding.

The studies showing LD binding in a non-ﬁnite context have only
been able to show that learners whose ﬁrst languages contain both SDRs and
LDRs differentiate between ﬁnite and non-ﬁnite contexts in their binding
pattern.

Despite this limitation, as well as the results of this dissertation study
which only include individuals from LDR/ SDR-type Lls in the +/ - ﬁnite
group‘, It is premature to assume that this stage of acquisition phase is
limited to learners whose Lls contain both SDRs and LDRs.

 

4 There is one example of a subject who was removed from the subject pool
due to an error on one of the comprehension items. Nonetheless, this
individual answered consistently on all six sentence types. This subject, a
French native speaker, exhibited a local only domain in the ﬁnite context
and a local/ non-local domain in the non-ﬁnite context, suggesting that the
ﬁnite/ non-ﬁnite distinction is not necessarily a function of the learner's L1.

160

First of all, SLA research, as well as the behavior of the "yes"-type
learners in this dissertation study, has shown that learners of all ﬁrst
language backgrounds LD bind. I have argued that LD binding across all
contexts represents a stage of acquisition where the learner is binding the
reﬂexive non-structurally. If research has shown that learners of all L1
backgrounds LD bind, why must we then assume that the LD binding
associated with the +/ - ﬁnite IL is necessarily the result of transfer? Simply
because an example has not yet turned up, does not mean that there does
not exist an example of a +/- ﬁnite IL by a learner whose L1 only has SDRs.

Furthermore, there is some evidence, although small, that we cannot
rule out the possibility of learners from SDR-only Lls LD binding
exclusively in a non-ﬁnite context. White et al. (1997) report on a study
which sought to compare the effectiveness of story and picture tasks in
measuring binding domain and orientation. The study examined the
intuitions of a group of French speakers and Japanese speakers learning
English as a foreign language. The mean scores of the biclausal sentences,
particularly those sentences in which context was directed toward the non-
local NP as the likely antecedent, showed signiﬁcant differences in accuracy
for the Japanese speakers between sentences with ﬁnite and non-ﬁnite
embedded clauses. That is, the Japanese speakers were signiﬁcantly more
accurate at requiring a local-only domain for the English reﬂexive in a ﬁnite
context than in a non-ﬁnite. This ﬁnding reﬂects the results reported in
Finer 8r Broselow (1986), Hirakawa (1990), and others which identiﬁed a
pattern where Ieamers appear to allow more LD binding in a non-ﬁnite
context than ﬁnite. White et al. (1997) do not report a signiﬁcant difference
in accuracy between ﬁnite and non-ﬁnite contexts for the French speakers.
However, the results, reproduced in Table 21, show a small difference in

161

accuracy for the French speakers. Even if this difference is not signiﬁcant, it
is quite possible that contained within the group of individual learners,
there could be several learners who exhibit this + / - ﬁnite distinction in their
binding domains. The group mean scores reported in this study hint at the
possibility that French speakers also pass through a stage where they LD bind
exclusively in a non-ﬁnite context. Thus, while it is not possible to argue
deﬁnitively that learners from all language backgrounds go through a stage
where they LD bind exclusively in a non-ﬁnite context, it is likewise
premature to assume the opposite (that only those learners coming from an
LDR/ SDR L1 LD bind exclusively in a non-ﬁnite context) is true. The debate
is still open on this matter.

If it is true that learners of various L1 backgrounds progress from the
non-structural stage to a +/ - stage, it is possible that for learners from
SDR/ LDR-type Lls, this stage is prolonged. Thus, the heavy representation
of learners from certain Lls in the data showing this response pattern.

Progovac and Connell (1991), White (1995), and Bennett (1993; 1994)
argue that learners are faced with two distinct tasks in acquiring the target
like system for reﬂexive binding. For English, learners need to recognize
that English as a +[AGR] parameter setting. That is, English has overt AGR
which, when present, serves as SUBJECT for a monomorphemic, x0
reﬂexive (not present in the L1 grammar of English, but could be present in
a learner's IL). Secondly, the learner needs to recOgnize that English
reﬂexives are complex, XP, reﬂexives.

The behavior of the +/ - ﬁnite learners in this dissertation (and other
studies such as Finer Gr Broselow, 1986; Bennett, 1993; White et al., 1997)
suggests that learners come to recognize +[AGR] in English before they learn
to analyze the reﬂexive as an XP reﬂexive. This dissertation study includes a

162

group of learners, listed in Table 23, whose grammars are characterized by a
local only domain when the reﬂexive is embedded in a ﬁnite clause and a
local/ non-local domain when the reﬂexive is embedded in a nonﬁnite
clause. This response pattern is indicative of a learner who has come to
recognize AGR in English yet interprets the reﬂexive pronoun as if it is
monomorphemic. In assuming the reﬂexive is monomorphemic, the
learner allows the reﬂexive to bind outside of the embedded clause only
when AGR is not present, as is the case with a nonﬁnite embedded clause.

Hamilton (1996) argues against crediting an X0 analysis to learners
who bind long distance. It is equally possible, he argues, that the learners are
interpreting the reﬂexive as a logophor and binding it non-structurally. In
my discussion of the "yes"-type group, I suggest that non-structural binding
is a likely possibility, given that the learners show no evidence of a
syntactically constrained system. Additionally, I predict that this
nonsyntactic strategy is an initial strategy used by the learner which is
predicted to be followed by a focus on the important syntactic component of
reﬂexive binding.

The response pattern of the +/ - ﬁnite group of learners suggests a
systematic difference in domain based on a syntactic element - the nonﬁnite
nature of the embedded clause. Thus, because the deﬁning characteristic of
this response pattern is a syntactic one, the motivation for the domain
patterns is more likely a syntactic one.

The questions which arise are, 1) Are learners misanalyzing the
reﬂexive as an LDR? 2) Are they transferring the LDR from their ﬁrst
language? Or, 3) does their ﬁrst language somehow guide them into a
misanalysis, representing a combination of factors? As I discuss in the
previous section, the evidence collected from SLA research so far only

163

includes examples of learners whose ﬁrst languages contain both SDRs and
LDRs. However, we should not assume that just because we have not seen
any counterexamples yet, these counterexamples do not exist. There are
examples of learners from SDR-only Lls who bind LD. Many of these
studies reported group mean scores and / or did not speciﬁcally investigate
the +/ - ﬁnite issue, so the debate is still open and is an issue for further
research.

Thus, upon closer evaluation I have shown that a common
interlanguage pattern among L2 learners is to require a local domain when
the reﬂexive is embedded in a ﬁnite clause, yet allow both local and non-
local antecedents for reﬂexives embedded in non-ﬁnite clauses. This
response pattern differs from that identiﬁed by Franks and Connell (1996)
for child language acquisition. Franks and Connell identiﬁed a response
pattern where a group of children required non-local binding in a non-ﬁnite
context and allowed local and non-local binding in a ﬁnite context. There is
an important communality here, however. Both groups of learners appear
to recognize and assign importance to the ﬁniteness of the embedded clause.
Ironically, while the ﬁnite status of the embedded clause is of critical
importance to child and adult learners of English, it has no effect on the
binding domain of the target grammar.

Franks and Connell (1996) suggest that the reason for the LD
children's binding domain described above is that these children misanalyze
the reﬂexive as monomorphemic and do not allow PRO, the non-lexical
subject of the clause, to be an acceptable binder. This is a curious possibility,
according to Franks and Connell, given that Wexler (1992) suggests that PRO
becomes available to children through maturation of the case ﬁlter to

164

children at a much earlier age than the subjects were at in the Franks and
Connell study.

The L2 learners in this dissertation study, as well as the learners
identiﬁed in studies such as those by Finer & Broselow (1986), Hirakawa
(1990), Bennett (1993; 1994), and White et al. (1997) appear to have an
adequate grasp of PRO and allow it to serve as a binder for the reﬂexive
because the local NP in object control inﬁnitivals is consistently accepted by
learners as a possible antecedent. This makes sense, as Franks and Connell
(1996; 456) argue that the reason the children do not allow PRO as a binder is
due to maturational issues which, assumedly, the adult L2 learners bring to
the learning task.

5.2.4 Local Group. One place where the child and second language
acquisition research on reﬂexives ﬁnd common ground is on the issue of
ultimate attainment. Both child and adult acquirers show evidence that
they ultimately acquire the local-only interpretation of English reﬂexives.
Franks and Connell (1996) identify a group of 6 subjects (3 normal, 3 SLI)
who exhibit a target-like interpretation of the reﬂexive.

The results of this dissertation study also include a group of learners
who interpret the reﬂexive as a local-only anaphor. Seventeen subjects
(SL1, SL2, SL3, SL4, SL5, SL6, SL7, SL8, SL9, SLIO, ML1, ML2, ML3, ML4, MLS,
ML6, and ML7) consistently rejected non—local binding in both ﬁnite and
non-ﬁnite contexts. These learners come from a variety of ﬁrst language
backgrounds and contain learners who have SDR—only Lls as well as Lls
which contain both SDRs and LDRs. Both the SL and the ML groups
contained learners who exhibited a target-like response pattern and were

165

comparably represented. Ten of the 18 SL subjects (56%) and 8-of the 14 ML
subjects (57%) required a local domain for the reﬂexive.

While it is quite possible that factors such as ﬁrst language and
mono/ multi-lingual backgrounds have an effect on the rate and means of
L2 acquisition of English reﬂexives, it does not appear that they prevent
learners from ultimately adopting a local domain for the reﬂexive.

5.3 Orientation and domain

5.3.1 The SLA Orientation of Domain: predictions from theoretical and
CLA research. As I discussed at length in Chapter Two, theoretical accounts
of reﬂexive binding such as the LF-movement analysis and the Relatevized
SUBJECT analysis predict that the natural consequence of LD-binding is that
only subject NPs can serve as antecedents for the reﬂexive. Thus, LD-
binding entails a subject orientation for the reﬂexive. It is important to
note, however, that the opposite scenario does not hold. Subject orientation
does not entail LD-binding, as it is very possible for a reﬂexive pronoun, the
Japanese ka re-zi si n for example, to be a short distance reﬂexive yet require a
subject orientation. Non-subject binding, however, entails local binding.
Because structural LD-binding is subject oriented, object binding must be
restricted to the local domain.

These predictions for domain and orientation relate speciﬁcally to the
adult target grammar, yet also lay out the boundaries within which child
learners are assumed to operate. That is, the LF-movement analysis and
Relatavized SUBJECT analysis assume that the mental grammar of child
language learners (i.e., learners who have access to UG) disallows LD object
binding throughout the acquisition process. An important question for SLA
researchers is whether or not L2 learners also unconditionally disallow LD

166

object binding. Through this line of inquiry, researchers investigating this
link hope to ﬁnd evidence either for or against access to UG.

Franks and Connell (1996) investigate the acquisition of English
reﬂexives by normal and speciﬁc language impaired (SL1) children with the
intention of identifying the various characteristics 'of learner grammars as
they relate to both domain and orientation of reﬂexives. In their study,
Franks and Connell (1996) argue that their data show that domain and
orientation are linked for monolingual learners of English. As learners
come to discover that the reﬂexive has a speciﬁc domain (in the LD and,
later, local phases of acquisition), so too do they show evidence of a speciﬁc
orientation for the reﬂexive. The learners who allowed long distance
binding also required a subject orientation for the reﬂexive. Likewise, all
but one of the learners who required local binding also allowed both subjects
and non-subjects as possible antecedents.

While these initial ﬁndings suggest that the link between domain
and orientation has been established, the Franks and Connell (1996; 449) data
show that the task problems prevalent in recent SLA reﬂexive studies,
detailed in Chapter Three, may have similarly affected this study. Despite
the fact that English reﬂexives theoretically allow both subject and non-
subject NPs as possible antecedents in the local domain, ﬁve control subjects
(adult English native speakers) exhibited a subject orientation in their
response patterns. Likewise, Franks and Connell indicate that one
experimental subject (an SLI child) also exhibited a subject orientation with
a local domain.

More disconcerting, however, is the ﬁnding that one control subject
and one experimental subject consistently accepted a sentence in which the
long-distance object served as the antecedent for the reﬂexive. LD object

167

binding is an unattested binding pattern. Because this response pattern
clearly violates the rules of English binding, and one control subject
exhibited this pattern, it is possible that the task did not always accurately tap

learner/ speaker grammars.

5.3.2 SLA Research Context. SLA studies of reﬂexives have also faced
considerable obstacles in controlling for task effects. Thomas (1995)
speciﬁcally investigates the relationship between domain and orientation
among L2 learners of Japanese. In a summary of the individual grammars
of zibu n binding for 34 learners from two of her previous studies, Thomas
reports that while only seven subjects recognized zi bu n as an anaphor and
also bound zi bun long distance, ﬁve out of these seven learners (71%)
required a subject orientation in a local domain (2 i bu n requires a subject
orientation, even in the local domain). While the majority of learners who
bound zibu n long distance required a subject orientation, some learners
inappropriately allowed zi bu n to bind with nonusubject NPs in the local
domain. Thomas (1995: 220) argues that this behavior is not a violation of
UG, as it is theoretically possible for an X0 anaphor to bind with nonsubject
NPs in the local domain (e.g. Icelandic sig or Serbo-Croatian sebe). While
LDRs instantiating nonsubject orientation in the local domain is not a UG
unattested grammar, Thomas does not address the issue of motivation of
this particular grammar. It is not clear why these particular learners, most
of whom do not have LDRs which allow object binding in their Lls
(English, French, Spanish, German, Korean, Thai), might interpret zibu n in
this way. Nor is there any theoretical motivation for why learners would

naturally assign an LDR with a nonsubject orientation in the local domain.

168

Thomas then turns to learner interpretations of zi bu n in the non-
local domain. Less than half of the original subject pool (23 out of 58)
allowed 21' bu n to bind long-distance. Of the 23 L2 learners of Japanese who
bound zi bun long—distance, 16 (70% of the LD binding subjects) required a
subject antecedent in the non-local domain. The remaining seven subjects
(30% of the LD binding subjects) did not require a subject antecedent for the
reﬂexive. Of these seven subjects, six of them were low proﬁciency learners
of Japanese, suggesting that the subject orientation of zi bu n does not co-
occur with the acquisition of LD-binding, but rather proceeds at a slower

pace. Thomas concludes the following:

Thus if longitudinal evidence were to confirm that L2 learners
pass through a stage in which they allow long-distance,
nonsubject antecedents for zi bu n on their way to acquiring the
native-speaker grammar, this fact would be problematic for
both Manzini and Wexler‘s parameterized binding principles
and for the proposal that anaphors move in LF (Thomas 1996:
228). ,

The other possibility, Thomas suggests, is that learners misanalyze
zibun as a pronoun with the features [+pro, -ana], resulting in a grammar
which allows binding with LD subjects and nonsubjects as well as local
subjects. The research design of this part of the study, however, made
answering this question impossible.

Other SLA studies which address the domain] orientation
relationship such as White et al. (1997) and Christie (1992), do not address
the possibility of LD object binding, but report that L2 learners of English
from various ﬁrst language backgrounds as well as English native speaking
adults accept nonsubject antecedents in the local domain to varying degrees.

169

5.3.3 Orientation and Domain in the Dissertation Study. The results of this
dissertation study support these results. While this study did not directly
investigate the possibility of LD nonsubject binding, the experimental task
included four stories] sentences in which context was directed toward the
local subject as the antecedent and four stories / sentences in which context
was directed toward the local object as the likely antecedent for the English
reﬂexive. The results indicate that learners from various ﬁrst language
backgrounds (French, Arabic, Thai, Hmong, Korean, Japanese, Italian,
Asssyrian, Mandarin, Russian, Albanian, Chaldean) consistently accept both
subject and nonsubject NPs as possible antecedents in the local domain.

What is most interesting is that for each domain grammar pattern
identiﬁed in this study there are examples of learners whose response
patterns indicate each orientation type (subject only as well as nonsubject
orientation (NSO) patterns).

These results suggest that learners appear to acquire the rules of
orientation at varying rates and also that these rates differ from the
acquisition rate for the domain of the reﬂexive. However, it is also possible
that the story task relied upon for this dissertation study may have fallen
short in its goal to minimize the subject strategy so common among both
English native speakers and ESL learners. There are several cases (SL9, SL1 0,
ML6, ML7, ML8) where learners exhibit a local domain but reject the object
NP, implying a local domain with a subject orientation. However, as my
previous discussion suggests, it is impossible to say whether or not the
rejection of object binding by the ESL subjects is a reﬂection of a subject
orientation or a "subject strategy by learners who may not, in fact, instantiate
a subject orientation in their mental grammars. This uncertainty is made

stronger in light of the fact that three control group subjects (E22, E23, and

170

E24) also consistently rejected local object binding, despite the fact that these
people were adult monolingual speakers of English who should
theoretically allow local object binding.

A second possible explanation for these results, then, would be that
the L2 learners in this study, like some of the English control group subjects,
actually instantiate a subject] nonsubject orientation in their mental
grammars, but some learners are not sufﬁciently convinced by pragmatic
cues that the object NP is indeed allowable. Assuming this second
explanation, then, it is still difﬁcult to ﬁnd evidence to support the .
assumption that domain and orientation are linked. In this case we assume
no differences in orientation across subjects (i.e., they are all probably
subject] nonsubject oriented) yet these same subjects exhibit deﬁnite
differences in domain patterns (there are those who strictly require local
binding across contexts, those who only allow LD binding in a non-ﬁnite
context, and those who allow LD binding in all contexts).

A third explanation is the more likely; yet, unfortunately, this
explanation is more difficult to establish: L2 learners of English who
consistently reject local object antecedents represent a mix of learners who
instantiate a subject] nonsubject orientation mentally, yet exhibit a strong
preference for the subject as well as learners who actually instantiate a
subject orientation in their mental grammars. Given that experimental
tasks have thus far been unable to overcome the strong preference for the
subject NP exhibited by some English native speakers (e.g., Finer 6r
Broselow, 1986; Hirakawa, 1990; Finer, 1991; Thomas 1991), it is difﬁcult to
differentiate those learners who reject object binding as a result of
competence factors from those who rdect object binding as a result of
performance factors.

171

This dissertation study has produced data on the relationship between
domain and orientation that can be interpreted in three ways. In the ﬁrst
case, it is possible that the behavior of the learners accurately reﬂects their
mental grammars. In this case, then, the data do not support the
assumption that domain and orientation are linked for ESL learners, as
there are examples of subject oriented learners as well as subject / nonsubject
oriented learners for each domain group identiﬁed.

Secondly, the data can be interpreted as if learners who consistently
reject object binding actually allow object binding in their mental grammars.
These learners are merely exhibiting a strong preference for the subject
antecedent that cannot be overcome by the experimental task. In this case,
the data reported here would imply that the orientation of the learner
grammars remains constant (i.e., subject] nonsubject orientation), while
several domain grammars can be identiﬁed. This interpretation of the data
also suggests that the acquisition of domain and orientation proceed at
separate rates.

Finally, the data can be interpreted as reﬂecting a mixture of learners
with different orientation grammars (both subject-only and
subject] nonsubject orientation), but whose behavior cannot be relied upon
as an accurate predictor of the underlying mental grammar. Speciﬁcally, it is
impossible to infer the orientation of the reﬂexive in the minds of the
learners who consistently reject the object antecedent. Given this state of
affairs, the data do not conclusively support or falsify the assumption that
domain and orientation are linked in the SLA of English reﬂexives.

172

5.4 Conclusion and suggestions for further research

5.4.1 Acquisition Stages. In the previous sections I have argued that the
"yes"-type response pattern represents the earliest phase of the acquisition
process. I make these arguments based on the ﬁndings of previous research
in both CLA and SLA. Previous studies have identiﬁed that both child and
adult acquirers of English go through a period where non-c-commanding
antecedents are allowed for the reﬂexive pronoun. The child language
researchers interpret this behavior as indication that the child is interpreting
the reﬂexive as a non-anaphor. Thus, the reﬂexive is seen as a regular noun
meaning something roughly equivalent to body. I argue that this behavior
among L2 learners represents a period in the acquisition process where the
learner has not yet been able to put into action a syntactic system for
reﬂexive binding. The learner, while understanding that the reﬂexive is an
anaphor in need of an antecedent, does not impose structural constraints on
the binding of antecedent and reﬂexive.

While this dissertation study did not measure learners for their
acceptance or rejection of c-command, I argue that the "yes"—type response
pattern also implies a non-structural binding pattern, given the unusual,
though not unattested, pattern of NSO in simple clauses with nonlocal
binding across contexts.

Using years of English study as a source of rough proﬁciency
comparison across groups, differences appear between the two interlanguage
systems I identify and the target like response pattern. The average years of
study for each group is given in Table 22 below.

I argue that on a theoretical basis, the "yes”-type IL should precede the
+] - ﬁnite IL. The "yes"-type response pattern shows no indication of
syntactic constraints on the binding of reﬂexive and antecedent. The + / -

173

ﬁnite IL, however, represents a ﬁrst step toward the application of syntactic
constraints, although non-target like, of reﬂexive binding. In examining the
years of study for learners in Table 24, there is almost no difference between
the two groups. On the basis of years of English study alone, then, there
appears to be no evidence indicating that one IL precedes another.

In addition to the theoretical reasons for positing that the "yes"-type
IL precedes the +] - ﬁnite, there is the issue of ML and SL learners. The "yes"
group is almost totally made up of ML learners. Zoble (1995) reported that
his ML subjects were not as accurate at respecting the c—command rule as the
subjects. Zoble (1995: ) suggests that for ML learners, the c-command rule is

Table 24: Average years of language study for each acquisition group

 

Group Average Years of Study

 

Yes-allbuthL 6

 

+]- ﬁnite - all SL 6.8

 

 

 

 

Target - SL 9.6
Target - ML 8.9
Target - SL/ ML 9.3

 

 

 

only weakly instantiated. I argue that rather than having a "weakly
instantiated c-command," these learners, because they are ML learners who
have more L1 data to sort through and accommodate to the new L2, stall in
the ”yes"-type stage for a longer time than their SL counterparts.

174

The SL learners, then, are able to recognize the importance of syntax
for English reﬂexive binding more quickly and begin applying their
knowledge. Initially, it appears that the learners become aware of AGR and
differentiate between finite and non-ﬁnite embedded clauses in their
binding domains. This is likely due to a correct awareness of the clause
structure of English (particularly the ﬁnite] nonﬁnite distinction) combined
with a misanalysis of the English reﬂexive as an x0 anaphor. I argue for a
"misanalysis" of the reﬂexive rather than a transfer-based argument for two
reasons. First of all, I suggest that we cannot rule out the possibility that
learners from SDR-only Lls do not make this distinction simply because we
have not yet produced an example of this evidence.

Secondly, there are studies which show that learners from SDR-only
Lls bind LD (e.g. Thomas, 1989; Christie, 1992; White et al. 1997). These
studies did not examine the possibility that the subjects distinguished
between ﬁnite and non-ﬁnite contexts so it is unclear what their individual
grammars reveal. According to the argumentation developed in this
dissertation study, the subjects could be operating with either a yes-type IL or
a +/ - ﬁnite type IL.

AssumingthatallLl typesprogressthroughthetonLdomainsI
have identiﬁed in this dissertation study, why, then, is the + ] - ﬁnite group
composed entirely of learners whose Lls contain SD and LD reﬂexives?
These learners, while not directly transferring their L1 intuitions, still can be
subject to L1 inﬂuences. In languages which contain both SD and LD
reﬂexives, the SD reﬂexives are used much more infrequently than the
LDRs and so it is possible that learners from SDR] LDR language
backgrounds take longer to recognize the complex nature of the English

reﬂexive pronoun.

175

Equally possible, of course, is that while learners of all language
backgrounds bind non-structurally once they recognize the reﬂexive as an
anaphor, only those learners from SDR] LDR Lls apply an LDR analysis to
the English reﬂexive once they move into the syntactic phase of acquisition.

5.4.2 Suggestions for Future Research. The ILs identiﬁed here and their
proposedsequencesaretheresultofcrosssectionaldatathatneed tobe
conﬁrmed through longitudinal data. The idea that learners pass through a
stage of non-structural binding needs to be experimentally tested. Particular
attention to linking c-command violations and non-local object binding
with this "yes"-type response pattern would help establish the lack of
structural binding I have suggested to be in the minds of these learners.

Another important question to be addressed is whether or not the +/ -
ﬁnite distinction established for learners from SDR] LDR language
backgrounds is also present in learners from SDR-only Lls. It is unclear
whether or not learners from SDR-only Lls pass quickly through this stage
or skip it altogether, recognizing the complex nature of English reﬂexives as
similar to those in the L1.

None of these questions can be sufficiently answered without serious
attention to the development of a reliable experimental task. Despite the
best efforts of child and second language acquisition researchers, the issue of
preferred interpretations for the reﬂexives continues to plague research on
reﬂexives. Despite all of the positive qualiﬁes of the story task which I
outline in Chapter Three, the results of this dissertation study also produced
data which suggests limitations in tapping complete grammatical intuitions
rather than preferred interpretations of the reﬂexive.

176

This problem of preferences and complete grammatical possibilities
may relate to the issue of the lexicon in reﬂexive binding. There were
several particular test items where learners and English native speakers
were more likely to "stray" into allowing LD binding when the rules of
structural binding do not allow it. This appeared to be particularly strong in
the cases of three-place predicates in the experimental task, such as d rive
and give. While the reﬂexive was in a subcategorized position in each of
these cases, where it was in the position of the third argument, some
learners allowed the reﬂexive to LD bind. The idea that reﬂexives in none
subcategorized positions are not subject to the rules of structural binding,
may not be a cut and dried distinction in the minds of some speaker and
learners. It is possible that some predicate arguments are ”stronger" than
others, meaning that some argument positions, particularly those in three-
place predicates, are freer to bind semantically than others. This lexical
factor could also play a role in the challenging task problems in reﬂexive
research. The highly inconsistent response patterns identiﬁed in much
research on reﬂexives, and the resultant reluctance to let go of group
reporting (due to the fact that it is so much "neater" than investigating
individual grammars), may be helped by taking into consideration these
possible lexical effects. There is no doubt that these and many other
interesting questions relating to the acquisition of reﬂexive pronouns will

be the subject of debate for a long time to come. ‘

APPENDIX A

APPENDIX A

Experimental and control task.

Backggound Information
Directions: Please answer the following questions.

1. Please circle one:
Male Female

2. What is your native language?

3. What other languages do you know and how well do you know them
(beginner, intermediate, advanced)?

4. When did you come to the United States?

5. How many years have you been studying English (including English

classes in your country)?

6. Do you speak English outside of class? When d 0 you speak English?

7. What is your age?

177

178

Directions: Read each story below. After each story is a sentence. Read the
sentence and decide if the sentence is true or fa lse according to the story. you
read. Circle your answer. If you read a word that you do not understand,

circle the word.

Example: George and Tim went to the park to play. Tim brought a soccer
ball. George and Tim played soccer together all afternoon. Finally, George
said, 'T have to go home for dinner. My mom is waiting for me."

George brought a soccer ball to the park

True False

1. John really wanted to work at the video store. He went to the store to talk
to the manager. John explained that he was a hard worker. John also said
that he knew all about video movies and that he could start work right
away.

John told the manager about himself.

True False

2. Bill was sick and in the hospital. Nobody knew what was wrong with
Bill. The hospital did a lot of tests on Bill to ﬁnd out what was wrong. Bill
had to wait a long time in his hospital room. Finally a doctor came in to tell
Bill why he was sick.

After the medical tests, the doctor informed Bill about himself.

True False

179

3. John was sick in the hospital. He was nervous because he needed an
operation. A doctor came into John's hospital room and said, "Hello John, I
am the doctor. I will operate on you tomorrow."

John heard that the doctor will operate on himself.

True False

4. Teresa and Madeleine went to a party one night. Teresa drank too much
beer at the party. When it was time to go home, Teresa was worried because
she didn't want to drive her car. 'TTl drive you home, " said Madeleine.
"Oh, thank you Madeleine. I really appreciate it" said Teresa.

Teresa was happy that Madeleine drove herself home.

True False

5. Barry and George went to the movies together every week. This week,
Barry arrived at the theater before George. Barry sat down in a seat. Soon
George arrived, carrying popcorn and soda. "Look Barry, I bought popcorn
and soda for us!"

Barry knew that George bought some popcorn."

6. Helen was reading a great book. The book belonged to Carrie. One day
Carrie called Helen on the telephone. Helen was not home so Carrie left a
message on the answering machine: "Helen, this is Carrie. Can you give
me my book? I need it right away!"

Carrie told Helen to give the book to herself right away.

True False

180

7. Jeff wanted to buy a new house. He applied for a loan at the bank. The
banker asked Jeff for some important information. "I am sorry," Jeff said, "I
left that information at home." The banker replied, "Oh, that's OK. you can
give me that information later.

The banker told Jeff to give himself the information later.

True False

8. Sharon was studying hard for a math test. Sharon knew the material
well, but would always become afraid during the test. Before the exam, the
teacher said, "You should not be afraid, Sharon, you understand the
material. You just need to believe that you can do it!"

The teacher told Sharon to believe in herself more.

True False

9. A little girl and her mother went to the store. The little girl saw a toy she
really liked and said, "Mother, I really want that toy! Would you buy it for
me?: The mother said yes, and the little girl said, "I am so happy!"

The little girl was happy that her mother bought herself a toy.

True False

10. Molly and Cheryl were going to the store together. When Molly arrived
at Cheryl's apartment, she could not ﬁnd Cheryl. Molly looked all over the
apartment but could not ﬁnd Cheryl. Finally, Molly saw Cheryl looking in
the mirror in the bedroom.

Molly noticed that Cheryl was looking at herself in the mirror.

True False

181

11. Jerry could not sleep at night. This was a big problem, Jerry tried
sleeping pills, but they didn't work. Finally, Jerry went to see a psychologist
about his problem. The psychologist asked Jerry a lot of questions about
Jerry's life and work.

During the session, the psychologist asked Jerry about himself.

True False

12. Patrick decided to have a party for his birthday. Patrick invited many of
his friends. At the party, Patrick walked up to a man and said hello. The
man responded, "Hello, my name is David. I came to me party with my ‘
brother who is a friend of yours."

David gave Patrick some information about himself.

True False

13. Tom was driving home on a very snowy day. He drove over a patch of
ice and the car slid off the road and into a tree. Wen the policeman arrived,
Tom gave his name to me policeman. Tom also explained that his arm
hurt and that his head was bleeding.

After the accident, Tom told the policeman about himself.

True False

14. For a school project, Sally decided to interview her grandma and write a
short story. "Grandma," Sally asked, "Please tell me, where were you born?
What was your school like? Also, please describe your family for me."

Sally questioned grandma about herself.

True False

182

15. Ken bought a new gun to use for hunting. Ken wanted to show the gun
to Stan. "Hey Stan, look at my new gun," Ken said. Ken took the gun out
and showed it to Stan. While Ken was holding the gun, Ken accidentally
pulled the trigger and the gun ﬁred, shooting Ken in the foot.

Stan saw that Ken shot himself.

16. Harold and James live together. They share a small apartment. Usually
Harold cooks dinner and James washes the dishes. Tonight, however,
Harold has to work. Harold said, "James, I will not be home tonight, you
will have to cook dinner alone."

Harold told James to cook dinner for himself.

True False

17. Julie and Mary went to a restaurant for lunch. For lunch they each had a
cheeseburger, french fries and a Coke. When Julie ﬁnished her lunch, she
said, "I am ﬁnished with my lunch, I will through our garbage out." "That
is so nice, thank you!" Mary replied.

Julie was happy that Mary threw out the garbage.

True False

18. Dan moved to Michigan for a new job. Michigan was so cold compared
to his old home in Florida. Dan's friend, Greg, said, "Dan, I know why you
are so cold, you do not have a warm coat! Go to the store and buy a warm
coat, then you will feel better!"

Greg told Dan to buy himself a warm coat.

True . False

183

19. Paul and Mark were brothers. Paul lived in Idaho and Mark lived in
New York. Mark called Paul every week to talk on the phone. Paul and
Mark were talking on the phone one day and Paul said, "I know these long-
distance phone calls are expensive, Mark, why don’t you write me a letter?"
Paul asked Mark to write himself a letter.

True False

20. Ellen cut her hand and it would not get better. She went to the hospital.
At the hospital a nurse came into the exam room and looked at the cut. The
nurse said, "Wash well with this special soap. The doctor will come in and
look at it."

The nurse asked Ellen to wash herself well before the exam.

True False

21. A boy and his father went on a bike ride together. The boy went down a
hill very fast. "Don't go so fast!" shouted the father. It was too late, the boy
fell off his bike and started crying. The father gave the boy a hug. Then the
boy was happy again.

The boy was happy that the father hugged himself.

True False

22. Billy delivered newspapers every morning for his job. For three
months, Billy had been saving his money for a new bicycle. Billy ﬁnally
bought the bicycle and told Charlie about it, "Hey Charlie, look at the new
bicycle I bought!"

Charlie heard that Billy had bought himself a bicycle.

True False

184

23. Mario went to the library to do some research for a school project. He
asked the librarian for some help. "Excuse me," Mario said, "I am looking
for a book about computers." The librarian showed Mario a large stack of
books all about computers.

The librarian showed Mario where to ﬁnd books about computers.

True False

24. Harry and George loved their baseball team, the "Pittsburgh Pirates."
One day Harry went to George's house to drive to the game. When Harry
saw George he screamed "George! You look great! How did you paint that
pirate on your face? I love it!"

Harry was that George had painted himself.

True False

25. Mike is a respected man in his community. Mike collects warm clothes
to give to the poor and he also helps cook them meals at the local church. A
reporter from the newspaper came to write a story about Mike. The reporter
asked Mike a lot of questions about Mike's life and'his work with the poor.
During the interview, the reporter asked Mike about himself.

True False

26. Monica got a new job in another state. Before beginning the new job,
Monica wrote a letter to Nicole, the manager of the company. In the letter,
Monica explained what a hard worker she was. Also, Monica described how
happy she was to have this new job.

Monica wrote a letter to Nicole about herself.

185

True False

27. At the beginning of class, the teacher collected homework from all of the
students. One boy did not hand in his assignment. After class, the teacher
called the boy to his desk and said, "I see you forgot to bring your homework
today. Please give me your assignment tomorrow morning."

The teacher told the boy to give himself the assignment tomorrow.

True False

BIBLIOGRAPHY

BIBLIOGRAPHY

Bates, E. at B. MacWhinney. 1987. Competition, variation and lagnguage
learning. In B. MacWhinney (Ed.), Mechanisms of language
acquisition. Hillsdale, NJ: Erlbaum.

Batistella, E. 1989. Chinese reﬂexivization: a movement to INFL approach.
Linguistics 27, 987-1012.

Bennett, S. 1994. Interpretation of English reﬂexives by adolescent speakers
of Serbo-Croatian. Second Language Research 10, 125-56.

Berwick, RC. 1985. The Acquisition of Syntactic Knowledge. Cambridge,
MA: MIT Press.

Broselow, E. and ﬁner, D. 1991. Parameter setting in second language
phonology and syntax. Second Language Research 7, 35—59.

Chien, Y-C. 6r Wexler, K. 1990. Children's knowledge of locality conditions
in binding as evidence for the modularity of syntax and pragmatics.
Language Acquisition I, 225-295.

Chomsky, N. 1981. Lectures on government and binding. Dordrecht: Foris.

Christie, K. 1992. Universal Grammar in the second language: an ‘
experimental study of the cross-linguistic properties of reﬂexives in
English, Chinese and Spanish. Unpublished Ph.D. dissertation,
University of Delaware.

Christie, K. 6: Lantolf, J. 1992. Bind me up bind me down: A UG study of
reﬂexives in the second language. Unpublished manuscript.

Clements, G. 1975: The logophoric pronoun in Ewe: its role in discourse.
Journal of West African Languages 2, 141-177.

Cole, P., Hermon, G 8: Sung, L.-M. 1990. Principles and parameters of long-
distance reﬂexives. Linguistic Inquiry, 21, 1-22.

186

187

Crain, S. 6: McKee, C. 1986. Acquisition of structural restrictions on
anaphora. Proceedings of the North Eastern Linguistics Society, 16,
94-110.

Culy, C. 1994. Aspects of logophoric marking. Linguistics, 32, 1055—1094.

Demirci, M. 1997. The role of pragmatics in the acquisition of reﬂexive
binding in L2. Unpublished Ph.D. dissertation, Michigan State
University.

Eckman, F. 1994. Local and long—distance anaphora in second-language
acquisition. In E. Tarone, S. Gass 6: A. Cohen (Eds), Research
methodology in second-language acquisition (pp 207-225). Hillsdale,
NJ: Erlbaum.

. 1996. On evaluating arguments for special nativism in second
language acquisition theory. Second Language Research 12, 4 398-419.

Finer, D. 1991. Binding parameters in second language acquisition. In
Eubank, L., editor. Point counterpoint. Amsterdam] Philadelphia,
PA: John Benjamins, 351-74.

Finer, D. 8: Broselow, E. 1986. Second language acquisition of reﬂexive
binding. In Proceedings of the North Eastern Linguistics Society, 16,
154-168.

Franks, S. & Connell, P. 1996. Knowledge of binding in normal and SLI
children. Journal of Child language, 23, 431-64.

Gass, S. 1985. An interactionist approach to L2 sentence interpretation.
Studies in second language acquisition 8, 19-37.

. 1987. The resolution of conﬂicts among competing systems: A
bidirectional perspective. Applied Psycholinguistics 8, 329—350.

Glew, M., Demirci, M., Yamagata, A., Gass, S. 1993. Semantic bias as it
applies to the acquisition of anaphoric reﬂexives by English native
speakers learning Turkish and Japanese. Paper presented at the 1993
Linguistics Symposium, University of Wisconsin, Milwaukee.

Haegeman, L. 1991. Introduction to Government and Binding Theory.
Oxford, U.K: Blackwell.

Hamilton, R. 1996. Against undetermined reﬂexive binding. Second
Language Research, 12:4, 420-446.

188

Heinleman, L.K. 6: J. MacDonald. 1993. Processing strategies in L2 learners
of French: The role of transfer. Language Learning, 507-557.

Hirakawa, M. 1990. A study of the L2 acquisition of English reﬂexives.
Second Language Research, 6, 60-85.

Hyams, N. 6: Sigurjénsdéttir, S. 1990a. A crosslinguistic comparison of the
development of referentially dependent elements. Paper presented at
the 15th Boston University Conference on Child Language
Development, Boston, MA.

. 1990b. The development of 'long-distance anaphors': a
cross-linguistic comparison with special reference to Icelandic.
Language Acquisition , 57-93.

#

 

Kapur, S., Lust, B., Harbert, W. & Martohardjono, G. 1993. Universal
Grammar and learnability theory: The case of binding domains and
the 'Subset Principle'. In Reuland, E. and Abraham, W., editors,
Knowledge and language: From Orwell '5 problem to Plato '5 problem.
Dordrecht: Kluwer Academic, 185-216.

Katada, F. 1993. The LF Representation of Anaphors. Linguistic Inquiry,
22:2, 287-343.

Lakshmanan, U. 6: Teranishi, K. 1994. Preferences versus grammaticality
judgments: some methodologica issues concerning the governing
category parameter in second language acquisition. In S. Gass, A.
Cohen Gr E. Tarone (Eds), Research methodology in second language
acquisition (pp. 185-206). Hillsdale, NJ: Lawrence Erlbaum.

Lebeaux, D. 1983. A distributional difference between reciprocals and
reﬂexives. Linguistic Inquiry, 14, 723-30.

Lightfoot, D. 1989. The child's trigger experience: Degre-O learnability.
Brain and Behavioral Sciences 12, 321-334.

MacLaughlin, D. 1992. Second language acquisition and the subset
principle. Paper presented at the Boston University conference on
language development, Boston, MA.

Manzini, MR. 8: Wexler, K. 1987 . Parameters, binding theory, and
learnability. Linguistic Inquiry, 18, 413-44.

Mathei, E. 1981. Children's interpretation of sentences containing
reciprocals. In S. Tavakolian (ed.), language Acquisition and
linguistic theory. Cambridge, MA: MIT Press.

189

Matsumura, M. 1994. Japanese learners' acquisition of the locality
requirement of English reﬂexives: Evidence for retreat from
overgeneralization. Studies in Second Language Acquisition,16,19-42.

McDaniel, D., Cairns, H. St Hsu, R 1990. Binding principles in the
grammars of young children. Language Acquisition I, 121-139.

O'Grady, W. 1996. language acquisition without Universal Grammar: a
general nativist proposal for L2 learning. Second language research
12,4 374-397.

Parker, J; Heitzman, S; Fjerstad, A; Cohen, A. 1995 Exploring the role of
foreign language immersion education. In F. Eckman, D. Highland,
F. Lee, J. Mileham, & R. Rutkowski Weber (Eds), Second Language
Acquisition Theory and Pedagogy (pp. 235-253). Mahwah, NJ:
Lawrence Erlbaum.

Pica, P. 1987 . On the nature of the reﬂexivization cycle. Proceedings of
NELS 17, 483-99.

Progovac, L. 1992. Relatavized SUBJECT, long-distance reﬂexives, and
accessibility, Linguistic Inquiry, 23, 671-80.

Progovac, L. 8: S. Franks. 1992. Relatavized SUBJECTS for reﬂexives.
Proceedings of the NELS. University of Massechusetts, Amherst.

Read, C. 6: Hare, V. 1979. Children's interpretation of reﬂexive pronouns in
English. In F. Eckman 6: A. Hastings (eds.), Studies in first and second
language acquisition. Rowley, MA: Newbury House.

Reinhart, T. and Reuland, E. 1991. Anaphors and logophors: an argument
structure perspective. In Koster, J. and Reuland, E., editors, Lo n g
-distance anaphora. Cambridge: Cambridge University Press, 283-
321.

, . 1993. Reﬂexivity. Linguistic Inquiry 24, 657-720.

Sasaki, Y. 1991. English and Japanese interlanguage comprehension
strategies: An analysis based on the Competition Model. A pl ied
Psycholinguistics, 12:1, 47-73.

. 1994. Paths of processing strategy transfers in learning Japanese and
English as foreign languages: A Competition Model Approach.
Studies in Second Language Acquisition, 16:1, 4372.

190

Saleemi, AP. 1992. Universal Grammar and language learnability.
Cambridge: Cambridge University Press.

Sigurjénsdéttir, S. 5: Hyams, N. 1991. The acquisition of binding in
Icelandic. Paper presented at the 16th Boston University conference
on Child Language Development. Boston, MA.

. 1992. Some issues in the growth of control. In R. Larson et
al. (eds), Control and grammar. Dordrecht: Kluwer.

#

 

Tarone, E.; S. Gass 8: A. Cohen (Eds) 1994. Research methodology in second-
language acquisition. Hillsdale, NJ: Lawrence Erlbaum.

Thomas, M. 1989. The interpretation of English reﬂexive pronouns by non-
native speakers. Studies in Second Language acquisition, 11,281-303.

. 1991. Universal Grammar and the interpretation of reﬂexives in a
second language. Language, 67, 211-239.

. 1993. Knowledge of reﬂexives in a second language.
Amsterdam] Philadelphia, PA: John Benjamins, 351-74.

. 1994. Assessment of L2 proﬁciency in second language acquisition
research. Language Learning. 44:2, 307-336.

. 1995. Acquisition of the Japanese reﬂexive zi bu n and movement of
anaphors in Logical Form. Second Language Research, 11, 206-234.

Wakabayashi, S. 1996. The nature of interlanguage: SLA of reﬂexives.
Second language Research, 12, 266-303.

Wexler, K. Gt Chien, Y-C. 1987. children's acquisition of the locality
condition for reﬂexives and pronouns. Papers and Reports on Child
Language Development 24,30-39.

Wexler, K. 6: Manzini, R. 1987. Parameters and learnability in binding
theory. In T. Roeper & E. Williams (Eds), Parameter Setting (pp. 41-
76). Dordrecht: Reidel.

White, L. 1989. Universal Grammar and second language acquisition.
Amsterdam] Philadelphia, PA: John Benjamins.

. 1995. Input, triggers, and second language acquisition: can binding
be taught? In Eckman, F., Highland, D., Lee, R, Mileham, J. and
Rutkowski Weber, R., eds., Second language acquisition theory and
pedagogy, Hillsdale, NJ: Erlbaum, 63-78.

191

1996. Universal grammar and second language acquisition:
current trends and new directions. In Ritchie, W. and Bhatia, T., eds.,
Handbook of language acquisition, New York: Academic Press.

White, L.; Bruhn-Garavito, J.; Kawasaki, T.; Pater, J. 6: Prevost, P. 1997. The
researcher gave the subject a test about himself: Problems of
ambiguity and preference in the investigation of reﬂexive binding.
Language learning, 47:1, 145-172.

Wolfe-Quintero, K. 1996. Nativism does not equal Universal Grammar.
Second language research 12,4 335-373.

Wulfeck, B.; L. Juarez; E. Bates; K. Kilborn. 1986. Sentence interpretation
strategies in healthy and aphasic bilingual adults. In J. Vaid (Ed.),
Language processing in bilinguals: Psycholinguistic and
neuropsychological perspectives (pp. 199-219). Hillsdale, NJ:
Erlbaum.

Yuan, B. 1994. Second language acquisition of reﬂexives revisited.
Discussion note. Language, 70:3, 539-45.

1992. Long distance and short distance reﬂexives in second
language acquisition. Unpublished manuscript.

Zobl, H. 1995. Prior linguistic knowledge and the conservatism of the
learning procedure: Grammaticality judgments of unilingual and
multilingual learners. In Gass, S. 5: L. Selinker. 1992. Language
transfer in language learning. Amsterdam] Philadelphia, PA: John
Benjamins, 176-196.

HICHIGAN smTE UNIV. LIBRRRIES
IIHI Will W llll “NW“ Ill 1| l" l"! "ll W m ill WI
31293017721527