THESlS

lllllllllllllllllllllllllIlllllllllllillllllllllllllllllllll

31293 01710 2835

This is to certify that the

thesis entitled
A CRITIQUE OF RESEARCH ON THE CRITICAL PERIOD

IN SECOND LANGUAGE ACQUISITION
presented by

Wei -Shen Tao

has been accepted towards fulfillment
of the requirements for

M. A . degree in Linguisfii Q5

 

Major professor

Date 4-27-98

0-7639 MS U is an Affirmative Action/Equal Opportunity Institution

 

 

 

LIBRARY

Michigan State
University

 

 

 

PLACE IN RETURN BOX
to remove this checkout from your record.
TO AVOID FINES return on or before date due.

 

DATE DUE DATE DUE DATE DUE

 

JAN 1 9390“

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1/98 WM“

 

A CRITIQUE OF RESEARCH ON THE CRITICAL PERIOD
IN SECOND LANGUAGE ACQUISITION

By

Wei-Shea Tao

A THESIS

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

MASTER OF LINGUISTICS

Department of Linguistics and
Germanic Slavic Asian and African Languages

1998

ABSTRACT

A CRITIQUE OF RESEARCH ON THE CRITICAL PERIOD
IN SECOND LANGUAGE ACQUISITION

By

Wei-Shen Tao

Within the framework of Chomsky's linguistic theory, the underlying assumption
of language acquisition is that a child will not be able to attain the linguistic competence
of the adult fi'om the input data alone. Because of this ‘projection problem’ in child
language acquisition, the hypothesis of innate Universal Grammar has been proposed to
account for how linguistic knowledge is acquired by the child. Furthermore, the
hypothesis of a Critical Period for language acquisition has been proposed to account for
the observed inadequacies of second language acquisition after about age 12.

In this thesis, we argue that two influential research studies on second language
acquisition which support the hypotheses, Schachter (1990) and Johnson & Newport
(1991), are inadequate to rule out the role of universal grammar in second language
acquisition. We discuss some crucial problems regarding the subjects, test sentences, and
data analyses in both studies, and accordingly argue that both studies are problematic for

their claim of the inaccessibility of Universal Grammar in second language acquisition.

Dedicated to
my father Shu-Hsiang Tao (WE?)

iii

ACKNOWLEDGMENTS

I would like to express my most sincere gratitude and appreciation to Dr. Grover
Hudson, the chairperson of my guidance committee, for his guidance, assistance, and
patience during the course of this study. Without him, I probably would not be able to
come to Michigan State University years ago and might not be able to finish my study
here.

I would like to thank Dr. Yen-Hwei Lin, Dr. Dennis Preston, and Dr. Alan
Beretta for serving as members of my guidance committee and for their time and
academic advisement. I also feel grateful to Dr. Julia Falk for her encouragement and
advice when I needed it during the past.

I also like to express my gratitude to the Department of Linguistics and Germanic
Slavic Asian and African Languages for the opportunity and financial assistance granted
to me. Being a teaching assistant not only helps to build up my self-confidence in many
ways but also makes me believe that I am able to become one of the best teachers,
although maybe not one of the best linguists.

I would like to thank my parents for what they have done for me. Words are
never enough to express my gratitude to them. Last but most, I like to express my
appreciation to my wife Kuei-Chun Liu for her devotion, patience, understanding, and
love. Without her encouragement, I would not come to study the graduate program in the
United States years ago. And, I also like to thank her for giving me two lovely children,

Kristin and Jeffrey, who case my pressure and bring me joy in life.

iv

TABLE OF CONTENTS

LIST OF TABLES ................................................................................. vii
LIST OF FIGURES ............................................................................... viii
Introduction ........................................................................................... 1
Background .................................................................................. 1
Universal grammar and language acquisition ........................................... 5
Measure of L2 competence: grammaticality judgment tasks ......................... 7
The Subjacency Principle .................................................................. 9
Purpose of this thesis ............................................................................... 12
Schachter (1990) .................................................................................... 13
A brief summary of Schachter’s study ................................................... 13
Problems in Schachter’s study ............................................................ 18
Subjects in Schachter’s study ................................................... 19
Test sentences in Schachter’s study ............................................. 21
Schachter’s theoretical conclusion: the Incompleteness Hypothesis ............... 25
What does Schachter’s Incompleteness Hypothesis tell us? ................. 25
Is her hypothesis sustained by her study? ...................................... 27
Johnson and Newport (1991) ...................................................................... 29
A brief summary of Johnson and Newport’s study .................................... 30
Problems in Johnson and Newport’s study ............................................. 35
Subjects in Johnson and Newport’s study ..................................... 36
Issues on Johnson and Newport’s analyses .................................... 37
General discussion .................................................................................. 41
References ............................................................................................ 44

LIST OF TABLES

Table 1 Means and Standard deviations of correct answer ................................... 16

Table 2 Means and Standard deviations
of correct answer on 9 wh-movement sentences ..................................... 24

Table 3 Mean number of no responses on subjacency
and control test items by subjects ...................................................... 39

Table 4 Mean number of correct responses and incorrect responses ........................ 40

LIST OF FIGURES

Figure 1 Mean number correct on S sentence types
by native English and Chinese speakers ........................................... 33

Figure 2 The relationship between age of arrival in the U. S. A. and number
correct on subj acency test items and no-inversion test items ................... 33

vii

INTRODUCTION

Mthin the framework of Chomsky’s linguistic theory, the underlying assumption
of language acquisition is that a child will not be able to attain the linguistic competence of
the adult from the input data alone. Because of this ‘projection problem’ in child language
acquisition, the innate Universal Grammar (U G) is then taken to account for how
linguistic knowledge is acquired by the child. In this paper, we argue that some research
data on second language acquisition are inadequate to rule out the role of UG. Our
criticism will mainly focus on two noticeable studies, Schachter (1990) and Johnson &
Newport (1991). We argue that both studies are problematic for their claim of UG’s
inaccessibility in SLA.

It is important to mention that we do not intend to address directly what the role
of UG is in Second Language Acquisition although arguments of both sides are reviewed
in this paper. Before we discuss these two articles, we first briefly review some relevant

theories and arguments about the age efl’ects on language acquisition.

Background

Recent studies in child language already argue that a child cannot acquire certain
complex properties of language only on the basis of input data; therefore, properties of
UG must be built in. During the past decades, one of the essential problems that linguists
argue most is the question of whether or not second language acquisition (SLA) is also
guided and controlled by UG (e.g., Birdsong 1992, Bley-Vroman 1989, Coppieters 1987,

Eubank 1991, Flynn 1993, Flynn and Manuel 1991, Long 1990, Martohardjono and Gair

1993, Newport 1990, and White 1989). The controversial discussion in SLA focuses on
the following question: do L1 and L2 learners share the same learning module to acquire
linguistic knowledge of the target language? To address this question, research refers to
the issue of Lenneberg’s critical period hypothesis.

Lenneberg (1967) proposes that there exists a critical period, between age 2 and
puberty, for primary language acquisition. His hypothesis is that this critical period
corresponds with the establishment of cerebral lateralization. He cites surveys of the
relevant clinical study and then provides support for the critical period and equipotentiality
hypothesis from evidence of the efi‘ects of early cerebral damage on language acquisition.
He claims that only children can transfer language functions completely to the right
hemisphere after an operation for an injured left hemisphere. He concludes that
interhemispheric reorganization of language functions is possible only before puberty, and
he links relevant time limitations for language learning to the lack of plasticity of the
human brain. According to Lenneberg’s proposal, first language learning turns out to be
impossible afier puberty. However, Lennebcrg’s theory has been challenged by later
research. For example, Krashen (1973) reanalyzes the data cited by Lenneberg and finds
that the timetable for language acquisition proposed by Lenneberg is irrelevant. Krashen
argues that lateralization must be completed by around age 5 rather than up to puberty.

Taking a similar line, many researchers in SLA propose a weak version of
Lenneberg’s critical period hypothesis, namely a sensitive period hypothesis (e.g., Johnson
1992a, Johnson and Newport 1989 & 1991, Oyama 1982, and Patkowski 1980 etc).
According to this hypothesis, changes in sensitivity for language acquisition show gradual

increases and declines, instead of an absolute and abrupt onset and ofl‘set. Generally

speaking, for phonology, the acquisition of native-like pronunciation is complete at age
six; and for morphology and syntax, the age limit for native-like attainment is before 15. If
a learner’s age of acquisition is before this timetable, he/she will be able to attain native-
like L2 competence. According to the notion of the sensitive period hypothesis, language
learning is still possible after the above age limits. However, the capacity for language
learning declines with age due to the maturational constraints, so older learners’ native-
like attainment is not guaranteed.

Results of above studies indicate no correlation between age of acquisition and
ultimate attainment of L2 acquisition. Most researchers support the likelihood that
younger learners are better learners. Accordingly, researchers of this party argue that first
language acquisition and second language acquisition may be dominated by different
learning modules. Younger learners, who are still within the critical (or sensitive) period,
are dominated by the same learning module as for their L1 acquisition, namely UG. That
is the reason why success of younger learners’ L2 acquisition is guaranteed. However,
older learners’ L2 acquisition turns out generally to be a failure, for the innate learning
faculty UG is no longer or not completely available for their L2 acquisition. Therefore, it
is assumed that older L2 learners apply some other learning modules, like general
problem-solving, which cannot fully replace the functions of UG.

However, researchers of another party propose that UG still plays an important
role in second language acquisition (e.g., Felix 1985 and 1988; and White 1992a, 1990,
and 1988). These studies on L2 learners’ interlanguage argue that learners’ immature
language performance of L2 also reveals sophisticated and complex linguistic properties.

So these features of L2 learners’ interlanguage also cannot be induced directly from the

Ll knowledge or the input data. These researchers claim that there is also a mismatch
between the input data and L2 learners’ ultimate attainment. As White (1989: 38) points
out, “we potentially have the same projection problem as in L1 acquisition, namely that
knowledge is attained which goes beyond the input and which could not be acquired on
the basis of general learning strategies or problem-solving.” As motivated for L1
acquisition, principles of UG must also mediate L2 acquisition because an L2 learner
shares the similar logical problem of language acquisition as L1 learners.

In order to examine whether or not UG is still available in SLA, recent L2 research
has examined some properties of UG, such as the principle of Subjacency, the Empty
Category Principle (ECP), or the Projection Principle, as evidenced by the L2 learner.
One of the crucial independent variables in these studies is “age of acquisition”.
Researchers usually compare the performance of difl‘erent age groups or simply study the
older learners’ L2 acquisition. The hypothesis is that older subjects’ worse performance,
compared with the native speakers and younger learners, indicates the absence of UG
(e.g., Long 1990, Newport 1990, Johnson and Newport 1989 & 1991, and Schachter
1989 & 1990 ).

Their general findings show that there are age-related efl‘ects on second language
proficiency. The younger learners usually outperform the older ones in the long run, and
this phenomenon seems to support the notion of a sensitive period for L2 acquisition.
Nevertheless, we will argue that two important studies’ claims against availability of UG in
SLA are problematic due to the failure of their methodology: Schachter 1990 and
Johnson & Newport 1991. Before any discussion on this topic, we briefly review three

theoretical aspects generally applied in the study on second language acquisition: the

general notion of Universal Grammar, the measure of L2 competence, and the Subjacency

Principle.

Universal grammar and language acquisition

Chomsky proposed the theory of Universal Grammar (U G) to account for how a
child can possibly achieve the linguistic knowledge possessed by adults. A theory of
grammars, UG is composed of the abstract principles which are believed to generally
underlie all human languages. As Chomsky (1975: 29) argues, UG provides “a system of
principles, conditions, and rules that are elements or properties of all human languages,
not merely by accident, but by necessity”. Further, he proposes that these essential
linguistic properties of UG are purely innate.

Researchers in child Ll acquisition support Chomsky’s hypothesis that some
linguistic principles have to be built in. The principles and parameters of U6 theory
provide a significant basis for a child’s L1 acquisition. One main fundamental issue that
researchers argue most is the so-called “projection (or logical) problem” in language
acquisition. That is, children cannot possibly obtain the linguistic knowledge of adults
solely based on the input which they are exposed to. White (1989) provides a detailed
discussion about three main projection (or input) problems.

According to White’s argument, the first problem is underdetermination, which is
“by far the most serious one, providing the best motivation for innate, specifically
linguistic structure” (p. 5). She mainly argues that there is not a corresponding pattern
between the environmental input and the linguistic knowledge actually attained. Instead,

the grammar of the target language is far beyond the available input. Therefore, in order

to account for how a child can arrive at the adults’ linguistic competence even though
there is no available input to trigger the overt principles of the target language, the innate
linguistic faculty “UG” must be involved.

The second problem is concerned with the degeneracy of input to a learner. The
language that a child hears may be ungrammatical or fragmentary sometimes. Logically
speaking, a child should have problems to come up with the correct grammar when
confronting both grammatical input and ungrammatical (or incomplete) input. “fithout
knowledge of UG, the child will possibly generalize either correct or incorrect rules, and
he/she has insufiicient basis for knowing which is which. Therefore, it must be innate
knowledge “UG” that guides the child to generalize the grammar of the target language,
despite the degenerate input. A child will know in advance that certain kinds of
hypotheses are impossible because of UG constraints.

The third problem is so-called “negative evidence”. When a child learns a
language, helshe may not get the negative input or make the relevant sorts of errors to
learn what are ungrammatical and impossible sentences in the target language. In other
words, a child may not have a chance to experience certain sorts of ungrammatical
sentences, or if experiencing them, to find out they are ungrammatical. However, with the
innate linguistic knowledge UG, he/she still attains the ultimate linguistic competence
which enables him/her to tell what sentences are possible and impossible in the target
language.

Theoretically speaking, UG contains a finite set of principles and parameters which
underlie possible grammars for a natural language. Because these properties of UG are

irmately programmed, language acquisition can be achieved despite the fact that there

exists a projection (or logical) problem. Guided by the constraints of UG, a child can
come up with what are possible and impossible grammars in the target language
ultimately. Following UG theory as an account of L1 acquisition, researchers on L2
acquisition then consider whether or not this innate linguistic faculty still functions. In the
latter discussion section of this paper, I will argue that UG must also be available to

second language learners.

Measure of L2 competence: grammaticality judgment tasks

In order to investigate whether or not non-primary language (L2) acquisition is
also constrained by the principles of UG, researchers apply difl‘erent methods to tap a
learner’s linguistic knowledge of L2. The most common technique is the solicitation of
grammaticality judgments. A grammaticality judgment task can involve sentences which
violate constraints of UG, so that a researcher can determine whether or not the learner
observes those constraints. In most of the research, the operation of UG is assumed only
if the learner performs accurately in both correct and incorrect test sentences. However,
there exist two main problems with this kind of tapping of linguistic competence.

The first problem is that grammaticality judgment tasks may not represent a whole
picture of the underlying competence of an L2 learner. Linguistic competence is an
abstract notion which cannot necessarily be measured directly by any task. Grarnmaticality
judgment tasks may more likely test a learner’s L2 performance than competence. If a
study involves a measure of subjects’ performance, a researcher may need to take into
consideration some variables related to testing behavior. These variables are crucial to the

results of a study. For example, as White (1989: 59) points out, the subjects may “show a

tendency to accept all sentences, regardless of their grammaticality”. Or, the subjects will
reject long sentences because of the lack of mentalinguistic skills or the parsing difficulty.
They may guess if they are not sure, instead of trying hard to access available knowledge.
They may lose patience if the test is too long, or they may expect that there is a balance
between tested sentences. Studies involving grammaticality judgment tasks should
consider such relevant variables.

The other problem is whether or not the study has validity. If some variables are
not controlled, the results of a study may reflect a false issue as in Schachter’s or Johnson
& Newport’s articles discussed later. Most recent studies in second language acquisition
are cross-sectional. Within the fi'amework of this kind of study, one variable in particular
should be controlled: subjects’ L2 proficiency level. When a researcher claims to test
certain properties of UG, he/she has to make sure that subjects have already reached the
appropriate acquisition level to recognize those tested properties of UG in an L2.

In other words, if a learner’s L2 acquisition is still one step behind some principle,
a research cannot claim that this principle of UG does not operate when the subjects’
judgments show violation of the principle. To avoid this probability, researchers usually
set up control sentences which are less advanced but structurally similar to the
ungrammatical ones (e.g., Bley-Vroman et al. 1988, Johnson & Newport 1991, and White
1988a). Before the test, a researcher first has to make sure that the subjects have no
problems with the control sentences; otherwise, the results of the study are then not
significant simply because the researcher has not tested what he/she claims to test.
Accordingly, research containing this problem is unable to argue whether or not UG is in

operation in L2 acquisition.

The grammaticality judgment task ideally is a satisfactory method to investigate a
learner’s L2 knowledge. However, if the above two discussed problems exist, it still
possibly fails to truly measure a learner’s L2 knowledge and to provide evidence as to
claim whether or not UG is still active. My later critique on Schachter’s and Johnson &
Newport’s articles will show that the results of their studies are derived from questionable
grammaticality judgment tasks in some such respects. Therefore, I argue that their claims
against UG’s role in SLA are not significant. Because both studies test Subjacency in

English wh-movement, a brief introduction to the Subjacency Principle is given below.

The Subjacency Principle

In order to investigate whether or not UG is active in L2 acquisition, researchers
employ difl‘erent properties of UG to test the learner’s L2 knowledge. One of the
principles applied in a number of L2 acquisition research is the Subjacency Principle (e.g.,
Bley-Vroman et al. 1988, Johnson & Newport 1991, Schachter 1989 & 1990, and White
1988). According to Haegeman (1991: 365), Subjacency stipulates that “movement
cannot cross more than one bounding node, where bounding nodes are [P and NP”. The
presentation below focuses on Subjacency in English wh-movement, for most studies deal
with English L2 acquisition.

Generally speaking, there may be either a short or a long movement for wh-
elements in English. The distance is not subject to grammaticality. For example, both (1),
which involves a short movement, and (2), which involves a long movement, are
grammatical.

(l) Whom did the instructor give the credits?

(2) Whom did you believe that the instructor gave the credits?

In Sentence (1), whom is moved within the main clause and crosses only one 1?; therefore
the movement is grammatical. However, in sentence (2), whom is moved from a lower
clause to a higher clause and seems to cross two IPs in question. The linguistic
assumption is that [Spec, CP] is the landing site of wh-movement. When a long wh-
movement occurs, there should be one or more intermediate [Spec, CP] positions
available for a wh-trace’s landing. Otherwise, the movement is ungrammatical because of
crossing two bounding nodes at a time. Accordingly, sentence (2) can be represented as
sentence (2)a.

(2)a [CP Whomi did [IP you believe [CP t’i that [IP the instructor gave ti the

credits]]]]?

In sentence (2)a whom is first moved to [Spec, CP] of the embedded clause and
then to the [Spec, CP] of the main clause. Both movements cross only one IP and thus
are grammatical. Sentence (3) is an example of a Subjacency violation.

(3) *[CP Whomi did [IP you believe [CP whenj [IP the instructor gave ti the

credits tj]]]]?

In sentence (3) there are two wh-elements. At first, the adjunct when has to be
moved to [Spec, CP] of the embedded clause to avoid the violation of the cyclic condition,
and this movement is okey. Because the intermediate landing site is already occupied by
when, whom then has to move to [Spec, CP] of the main clause directly and crosses two
IPs at a time. This step violates Subjacency and therefore this sentence is ungrammatical.

In a similar manner, sentence (4) shows a Subjacency violation according to which the

10

intermediate t’i crosses two bounding nodes, 1? and NP, at a time. Thus, the sentence is
ungrammatical.

(4) *[CP Whomi did [IP the instructor make [NP the claim [CP t’i

that [IP he gave ti the credits]]]]]?

In SLA, some research investigates the availability of Subjacency by studying the
acquisition of English by subjects with different L1 backgrounds. The hypothesis is that
the Subjacency Principle does not operate in the subjects’ L1 and so UG must play an
active role if the subjects observe the Subjacency violations in English.

In order to evaluate L2 learners’ performance on the Subjacency Principle, a
researcher needs to make sure that his/her subjects are highly advanced L2 learners. As
the examples that we gave above suggest, Subjacency only occurs when a wh-element
moves across those complex sentence structures with at least one or more embedded
structures within. Because of the complex sentence structures involved, we therefore
suggest that the research on Subjacency requires L2 learners at a relatively advanced
proficiency level, subjects who may reasonably be supposed to have little or no difi'rculty

with aspects of language which make up the Subjacency phenomenon.

ll

PURPOSE OF THIS THESIS

The purpose of this thesis is to propose that arguments of some extant research
results, which rule out the role of U6 by studying the Subjacency Principle, are
questionable. We will review and criticize two important articles, Schachter 1990 and
Johnson and Newport 1991, which examine the L2 acquisition of Subjacency in English.
Our critique will focus on theories of UG and research methods applied in these two
studies. In the third section of this paper, we will first give a summary of Schachter 1990
and then discuss the problems found in the study. In the fourth section, we will discuss
Johnson & Newport 1991 in a similar manner. Then, in the discussion section, we will
first discuss the similarity of the crucial problems found in these two studies and claim that
both research data of these studies are inadequate to rule out the role of UG in SLA At
last, we raise an issue that no current theory firlly demonstrates the phenomena in second
language acquisition at present, and we are looking forward to a better understanding

when the debates between UG and not-UG move on.

12

SCHACHTER (1990)

Schachter’s study investigates whether or not L2 learners acquire the principles of
Subjacency and ECP in English. She proposes that L2 learners show little or no evidence
of observing Subjacency in English if their L1 does not include the similar syntactic
phenomena. According to her results, she claims that “this experiment provides strong
evidence that the Incompleteness Hypothesis is the correct hypothesis for adult L2
acquisition as well, and therefore that completeness in second language acquisition past
puberty is likely to be an impossibility” (p. 118). However, there are some potential
problems, which are either theoretical or methodological ones, revealed in Schachter’s

study. Consequently, her claim about UG’s inaccessibility is questionable.

A brief summary of Schachter’s study

Schachter studies the acquisition of Subjacency in English by four L2 groups with
different L1 s: Dutch, Indonesian, Chinese, and Korean. The linguistic assumption is that a
language with overt wh-movement phenomena must observe constraints of the Subjacency
Principle; on the contrary, these constraints do not apply if there is no wh-movement. The
above four languages are related to difl’erent aspects of Subjacency. Schachter says that
Dutch exhibits the full range of Subjacency’s constraints as observed in English. Both
Indonesian and Chinese partially show evidence of Subjacency. As she points out, wh-
movement in Indonesian is allowed only if the wh-word is “promoted to subject in its own
clause before being moved to the beginning of the sentence” (p. 102). As for the Chinese

language, instead of overt wh-movement, the wh-word remains in situ. However, Chinese

13

allows other kinds of movements, like relative pronoun or topic extraction, which also
involves Subjacency. Different from the other four languages, Korean does not allow any
kind of extraction at the level of S-structure; therefore it does not show properties of
Subjacency.

Schachter’s experiment is designed to test knowledge of long-distance wh-
movement, knowledge of impossible wh-movement, and knowledge of Subjacency
violations in English. The research hypotheses are quoted fi'om Schachter’s study as
follows:

H1. knowledge of grammatical wh-movement will have an efi‘ect on

knowledge of Subjacency violations involving wh-movement.

H2. knowledge of the grammatical constructions that constitute islands

will have an efl‘ect on knowledge of Subjacency violations involving
extraction out of them.

H3. there will be a language-specific efi‘ect on knowledge of Subjacency.

For the Korean subjects the period of L1 development of Subjacency
will have passed before the subjects began their exposure to English
and as a result they will have no latent knowledge of Subjacency to
fall back on. The Dutch subjects, on the other hand, will have
complete knowledge of Subjacency efi‘ects, as a result of their
exposure to their L1 during the appropriate maturational stages.
(p. 105)
Schachter employs a set of sentences for written grammaticality judgments to

examine subjects’ knowledge of Subjacency in English. Subjects are asked to judge

l4

whether a sentence is clearly grammatical, probably grammatical, probably not
grammatical, or not grammatical. The test takes approximately 30 minutes. There are 66
test sentences in all, and all sentences are randomized. The test sentences include four
types of English construction: sentential subjects (SS), noun complements (NC), relative
clauses (RC), and embedded questions (EQ). There are a total of nine subtests for testing
the Subjacency Principle in this study. Five subtests constitute the control sentences (or
the syntax test): the wh-movement subtest, the SS subtest, the NC subtest, the RC
subtest, and the EQ subtest. The other four subtests are Subjacency violations: the *SS
subtest, the *NC subtest, the *RC subtest, and the *EQ subtest. Here are example
sentences of above tests which we copy fi'om Schachter’s study:

1. the wh-movement:

What does the professor expect us to know for the exam?

2. SS:

That oil prices will rise again this year is nearly certain.

3. NC:

The judge rejected the evidence that the student committed the crime.

4. RC:

The theory we discussed yesterday will be on the exam next week.

5. EQ:

The dorm manager asked me who I wanted to have as a roommate.

6. *SS:

*Which party did for Sam to join shock his parents?

7. *NC:

15

*What they have to accept the idea that they couldn’t operate by themselves?

8. *RC:

*What did Susan visit the store that had in stock?

9. *EQ:

*Who did the Senator ask the President where he would send?

Experimental subjects are 18 Dutch, 21 Indonesian, 20 Chinese, 20 Korean

subjects, and all had begun the process of English learning at age 12 or later. All are

undergraduate students and “highly proficient speakers of English” (p. 108). None of the

subjects had been exposed to the linguistic study of Subjacency. The controls are 19

native speakers. Schachter says that the backgrounds of all experimental subjects are

homogeneous and that neither age at first exposure nor residence in an English-speaking

country has an efi‘ect on subjects’ performance.

Subjects’ overall performance shows that all groups did well in judging

grammatical sentences; however, the performance differs in ungrammatical sentences. We

copy the results of her study (p. 111) as displayed in Table 1 below. The mean numbers

are the means of test sentences answered correctly.

 

 

 

 

Table 1 Mmdmmmrflanswer
Total Total

Grammatical Ungranunatical Wh-movement

(24 sentences) (24 sentences) (9 sentences)
Group Mean SD. Mn SD. Mean SD.
Native 21.6 2.7 21.2 1.9 7.1 1.07
Dutch 22.2 1.9 21.9 2.8 7.6 .98
Indonesian 21.2 2.6 15.2 4.5 5.26 1.2
Chinese 21.2 2.2 17.2 3.7 5.0 1.4
Korean 19.8 3.8 12.4 4.5 5.6 1.0

 

l6

As shown, there is no significant difference among the means of 24 grammatical sentences.
But, only the native (21.2) and Dutch (21.9) subjects performed well in the 24
ungrammatical sentences. The Indonesian (15.2), Chinese (17.2), and Korean (12.4)
groups performed significantly worse than Dutch subjects and the native group.
Moreover, the Korean subjects did worst among the experimental groups. According to
the results, Schachter claims that “native language has a significant efi‘ect on knowledge of
one principle of Universal Grammar in post-puberty-acquired second language grammars”
(p. 116).

Dutch subjects have already acquired the Subjacency effects, which are similar to
the ones in English, in their L1 acquisition; therefore they can precisely identify a
Subjacency violation. Indonesian and Chinese subjects only observed part of Subjacency
efi’ects in their Ll acquisition; therefore they may only identify some sorts of Subjacency
violations in English. That is, their knowledge of English Subjacency is incomplete and
thus they perform worse than Dutch subjects. Korean subjects did not acquire the notion
of Subjacency in their L1 acquisition; so their acquisition of English shows no constraint
of Subjacency.

Schachter further argues that the results of her study clearly rule out the role of
UG and support the Incompleteness Hypothesis in L2 acquisition. As she questions, if
UG is still available in L2 acquisition, Indonesian, Chinese, and Korean subjects will “be
able to detect Subjacency violations in English just as well as the Dutch subjects” (p. 118).
She proposes that UG is not reactivated in adult L2 acquisition. Moreover, she claims

that completeness in secOnd language acquisition past puberty can hardly be achieved.

17

However, findings of Schachter’s study may be questionable due to problems of her

subject selection and the test sentences.

Problems in Schachter’s study

There are three crucial problems in this study. The first problem concerns the
issue of subjects. Differences in the “incomplete” performance of the experimental subject
groups may be due to subjects’ insuficient L2 knowledge rather than due to their difl‘erent
Ll backgrounds, therefore results of Schachter’s study may be superficial. One may argue
that the subjects, except the Dutch group, have not acquired Subjacency in English yet.

The second problem concerns the issue of the test sentences. Although including a
syntax test, the test does not include corresponding grammatical wh-movement sentences
as the control sentences. The wh-movement tests consist only of nine grammatical
sentences which involve different wh-movement: 3 with one—clause movement, 3 with
two-clause movement, and 3 with three-clause movement. Either a wh-subject or wh-
object is moved in each group. Examples of the wh-movement test are copied fi'om
Schachter’s study as in the following:
(1) What does the professor expect us to know for the exam?

(wh = object in $2)
(2) Who did the president say he planned to appoint as ambassador?

(wh = object in S3)
(3) Who do you think Bill said Mary expected to go to the dance with Mark?

(wh = subject in S4)

18

Schachter’s wh-movement test does not include the corresponding sentence structures
that she includes in her Subjacency test. We argue that Schachter should have included
grammatical wh—questions parallel to those ungrammatical sentences on Subjacency tests.
For example,

(4) What did they know that Sarah disliked?

(5) *What did they know the fact that Sarah disliked?

Here both (4) and (5) involve a wh-movement in an NP-complement structure, but only
(5) is not grammatical because of violation of Subjacency. Without corresponding
patterns, we simply don’t know whether or not her subjects have already acquired wh-
movement in English. Further discussion of this issue will be given below.

The third problem is that Schachter’s theoretical conclusion seems to
conflict with the views derived either from the sensitive period hypothesis or from
the availability of UG. Schachter said that her subjects “had all begun the process
of acquisition of English between age 12 or later” (p. 108). That is, their age of
acquisition is obviously around puberty and still within the timetable within which
most researchers believe that native-like competence of L2 acquisition is still
highly attainable. Hence, Schachter’s Incompleteness Hypothesis seems more
likely to argue that acquisition of any non-primary language is absolutely
impossible; however, her hypothesis does not provide any significant evidence to
account for any possible age effect on UG’s accessibility. Explicit discussions of

the above three problems are given below.

19

Subjects in Schachter’s study

The subjects’ English proficiency may not be suficient. Schachter’s assumption
about her subjects’ English proficiency may be questionable. Schachter says that her
subjects, not including the 18 Dutch, “had either a) taken the USC English placement
exam and been exempted fiom any ESL requirement or b) in fact had been required to
take an ESL course but had completed all such requirements in a semester before the one
in which the Subjacency tests were administered” (p. 108). Schachter assumes on this
basis that all her subjects are “highly proficient speakers of English”. She does not apply
any other criterion to measure subjects’ accurate English proficiency level. Besides,
Schachter’s syntax test fails to confirm the L2 subjects’ qualification to some extent
regarding Subjacency in English, and we will discuss this problem later. Accordingly, we
simply don’t know whether her subjects are qualified to be selected in her study or not.

Another minor issue is that the foreign fieshmen may not be the ideal subject pool.
Schachter’s subjects were students in sections of an introductory linguistics course or
Freshman English courses. We suspect that all these “foreign” fi'eshmen still have
problems to perform as highly proficient speakers of English. Foreign undergraduate
students in many universities probably only need to pass the language test required by
their university, a test requirement is probably equivalent to 550 or less (see Custard 1998)
on the TOEFL. With this level of English proficiency, Schachter’s subjects do not
convincingly represent a highly proficient subject pool.

Besides, previous studies (e. g. Johnson & Newport 1989) “have shown no efl‘ects
of length of experience with the language for adult learners of a second language after 5

years of immersion in the target language” (Johnson & Newport 1991, p. 228). All

20

means of the 3 Asian groups’ length of residence to English are much less than 5 years.
The Indonesian and Korean groups’ length of residence are even less than 3 years. As
shown in Table 1 (p. 109), the mean of the Chinese subjects’ months in English speaking
countries is 48.5; of the Indonesians’ 27.6 months; and of Korean subjects’ 34.2 months.
If length of experience also plays a role in the second language acquisition, one may
therefore question that Schachter’s subjects’ acquisition of English is still in the immature
stage.

Because of these problems, we believe that Schachter’s Asian subjects are not
highly advanced English learners and might for this reason alone have failed to perform on
the Subjacency tests. As for the 18 Dutch subjects, one possibility is that they may
actually be the only highly proficient group. They even outperform the native speakers as
the control group. The other possibility is that they may follow a pattern of answering the

test sentences as discussed in the following section.

Test sentences in Schachter’s study

The second problem concerns the issue of the test sentences. Although Schachter
does include equal numbers of 24 grammatical sentences and 24 ungrammatical sentences
in her test, the test does not include corresponding grammatical wh-movement sentences
as the control sentences. All of her 24 grammatical Subjacency sentences are declaratives.
As we mentioned before, although Schacher’s test does include a subtest including 9
grammatical wh-movement sentences, the sentence structures of these 9 questions are not

similar to those questions of the Subjacency test.

21

Schachter’s wh-movement test only deal with the aspects of wh-words’ moving
across one clause, two clauses, or three clauses. But, most importantly, she should apply
wh-movement test sentences parallel to questions in the Subjacency test. When testing
Subjacency in English, a researcher first needs to make sure that L2 subjects have already
acquired wh-question movement in general. To clarify this, correct wh-questions should
be provided in a grammaticality judgment test. Furthermore, these grammatical wh-
questions should be structurally parallel to those ungrammatical questions which are for
the Subjacency test. Otherwise, the stnrctural difference may become another factor in
subjects’ judgments. Let’s look at the examples below.

(1) What did they know that Sarah disliked?

(2) *What did they know the fact that Sarah disliked?

Here both sentence ( 1) and sentence (2) involve a wh-movement in an NP-complement
structure, but only (2) is not grammatical because of violation of Subjacency. We will
have several possible answers judged by an L2 English learner for these two wh-questions.
Ifthe learner judges either that (1) is ungrammatical and (2) is grammatical or that both
(1) and (2) are not grammatical, we will say that this subject has not acquired wh-
movement in English—more specifically, a wh-movement in an NP-complement structure.
This subject then is not a qualified candidate for the study on Subjacency.

As we suggest, before testing Subjacency, we first need to make sure whether a
subject knows wh-movement structures in English. That is, a correct answer on (1) from
the subject must be guaranteed. There are two possibilities. One is that, judging that (1)
is grammatical and (2) is also grammatical, the learner knows the structure of wh-

movement extracted from an NP-complement but does not identify the Subjacency

22

violation in (2). The other is that, judging that (1) is grammatical and (2) is not
grammatical, the learner knows the structure of wh-movement extracted from an NP-
complement and also identify the Subjacency violation in (2). We then say that this
subject performs as well as the English native speaker regarding Subjacency. Schachter’s
24 declaratives grammatical sentences obviously do not bear the function of question (1)
given above, and fail to investigate whether her subjects have acquired English wh-
movement. Accordingly, we question that her subjects are indeed qualified candidates for
a study on Subjacency.

Johnson & Newport (1991) point out that Schachter’s test sentences cannot
provide clear evidence showing that a subject does observe Subjacency. They mention
that, without including corresponding grammatical wh-movement sentences, Schachter’s
test fails to clarify the possibility that “rejection of these sentences could have happened if
subjects believed that question formation in any form is ungrammatical, or more
reasonably, that question formation in any complex sentences is ungrammatical” (p. 223).

Johnson & Newport also mention that Schachter’s test ‘firvas not designed to rule
out a simple response bias” (p. 223), namely subjects saying “yes” or “no” to all test
sentences. But, as we can see fiom Table 2 (p. 111), this is not the case. Even the
means of Korean subjects do not show this bias. However, it’s possible that a similar
response bias may be observed by some subjects. They may assume that they only need to
say “yes” to all declaratives and “no” to the questions as in the examples given below:

a. Question (say no): *Which party did for Sam to join shock his parents?
b. Declarative (say yes): That oil prices will rise again this year is nearly certain.

23

We believe that this is the case for the Dutch subjects, and that may be the reason why the
Dutch group performs even better than the native control group in both grammatical
sentences and ungrammatical ones.

In addition to the above two problems, there is another problem which can be
referred to as either a problem of the test sentences or of subject selection. Schachter
includes 9 wh-movement sentences in the test and we show the test results in Table 2 (as
her Table 2 on p. 111).

Table 2 Mg; and Standard deviau'ons of correct mer on 9 wh-mpvement sentences

Wh-movement (9 sentences)

 

Group My S.D.
Native 7. l 1.07
Dutch 7 .6 .98
Indonesian 5.26 1.2
Chinese 5.0 1.4
Korean 5.6 1.0

 

On the one hand, one may argue that these wh-movement sentences are not well designed.
The mean number of the native control group is only 7.1 (out of 9) which is below 80%
and well below the standard, 90% as in most research. One may therefore suspect that
the results of this study are questionable, for the test sentences are problematic. ,

On the other hand, if we ignore the problem of the ill-designed wh-sentences and
agree with Schachter, then the subject issue raises again as we discussed in the above
section. Subjects are supposed to answer these questions without any problem if they
know the wh—movement sentence structures in English. However, as we can see, it is not
the case and all three Asian subject groups answer the questions correctly only just above
chance level. The mean of Indonesian group is 5.26 (out of 9), of the Chinese group 5.0,

and of the Korean group 5.6. If there is nothing wrong with the test sentences themselves,

24

the means of the Asian groups should be close to the means of the Dutch group (7.6) and
the native control group (7.1). According to the means of the Asian groups, we argue
that these subjects have not completed the acquisition of the wh-movement sentence
structure in English yet, and therefore these subjects are not qualified for Schachter’s
study on Subjacency. Due to the problems we discuss above, we claim that Schachter’s

test sentences fail to examine what she planned to examine.

Schachter’s theoretical conclusion: the Incompleteness Hypothesis

Schachter’s theoretical conclusion seems to conflict with the views derived either
from the sensitive period hypothesis or from the availability of UG. All Schachter’s
subjects had begun the study of English between age 12 and 15. Their age of acquisition
is still within the time in which most researchers believe that native-like competence of L2
acquisition is still attainable. In this section, we will first review Schachter’s
Incompleteness Hypothesis, and then we will argue that her hypothesis does not provide

significant evidence to account for possible age effects on UG accessibility.

What does Schachter’s Incompleteness Hypothesis tell us?

Schachter’s Incompleteness Hypothesis is that completeness with regard to L2
acquisition will not be possible due to the inavailability of UG as a knowledge source. To
support this argument, she gives a proposal and we quote her own words as below.

If, however, it turns out that in the acquisition of the target some instantiation of

principle P is necessary and P is not incorporated into the learner’s L1, the learner

will have neither language-intcmal knowledge nor initial-state knowledge to guide

25

one in the development of P. Therefore, completeness with regard to the

acquisition of the target language will not be possible. (p. 100)

Generally speaking, her idea is that some principles of UG (e.g. Subjacency) will remain
disabled at all times if they are not activated in one’s L1 acquisition. That is, an L2 learner
will subsequently only have access to the principles of UG which are instantiated in his/her
native language. And, those inactivated principles will not be incorporated into L2
acquisition.

We argue that Schachter’s proposal is unreasonable because it rules out the
phenomena found in multilingual acquisition. As we all believe, a child will be able to
learn two or more languages while growing up even if these languages have different
properties of UG. For example, a child of a Korean woman and an American man can/will
acquire Korean and English in the same time frame, and we will say that both languages
are his/her Lls. We then have no argument against Schachter’s proposal because both
languages are his/her Lls, and there is nothing to do with L2 acquisition in this case.

Nevertheless, Let’s say that another 12-year-old Korean child moves to Michigan
and only then starts to learn English. We will say that Korean is the child’s L1 and
English his/her L2 even though he/she begins to learn English only after knowledge of
Korean is well established, and he/she should attain native English competence without
doubt. But, according to Schachter’s proposal, he/she will not be able to acquire
Subjacency in English because his/her L1 Korean does not have Subjacency. We believe
that Schachter’s proposal is in conflict with the finding of most SLA research (e.g.,

Johnson and Newport 1989 & 1991, Oyama 1982, and Patkowski 1980 etc).

26

It seems that Schachter’s proposal focuses on the availability of UG in late L2
acquisition. She does not precisely give us her own view on the time flame of the
critical/sensitive period for L2 acquisition although she does mention “the maturationally
appropriate time” (p. 99). As we discussed in the first section on the issue of the so-called
“Critical Period Hypothesis”, those who deny the full availability of UG in adult second
language acquisition agree with the general conclusion that changes in sensitivity for
language acquisition show gradual increases and declines, instead of an absolute and
abrupt onset and ofi’set. However, the Incompleteness Hypothesis seems to claim a one-
point decline which is absolute and abrupt. The properties of UG can only be triggered by
input in one’s Ll acquisition, and it seems to be a once-in-a-life-time chance. This

hypothesis is not confirmed from the results of most studies.

Is her hypothesis sustained by her own study?

We have criticized the subject issues and the test sentences in Schachter’s study
and argued that the study itself is questionable, and therefore the results that Schachter
uses to support her study are problematic. On the subject issue, we find that differences of
the “incomplete” performance of the experimental subject groups may be due to subjects’
insuficient L2 knowledge rather than due to their difl‘erent L1 backgrounds. Except for
the Dutch group, we believe that all three Asian groups have not acquired Subjacency in
English yet. In other words, Schachter fails to target the desired subject pool for her
study, and her subjects are not advanced English learners as claimed.

Moreover, because the test does not include corresponding grammatical wh-

movement sentences as control sentences, results of Schachter’s study may be superficial.

27

The test sentences fail to reveal whether her subjects have acquired wh-movement in
English, let alone Subjacency. Some subjects may not be familiar with the wh-sentence
structure in general. As I said in the last section, Schachter’s Incompleteness Hypothesis
seems more likely to indicate that acquisition of any non-primary language is absolutely
impossible. Her study does not have any convincing evidence to support her hypothesis

and therefore fails to account for any possible age efl’ect on UG accessibility.

28

JOHNSON AND NEWPORT (1991)

Johnson and Newport (1991) also examined the L2 acquisition of Subjacency in
English. The experimental groups are Chinese speakers, whose L1 does not include the
principles of Subjacency according to Johnson and Newport’s interpretation. There are
two parts of this study. In the first one, they investigated adult learners whose age of
acquisition was after the age of 17 . According to these results, they concluded that:

This maturational effect was shown not merely to be the result of an inability

to set the parameters of this principle. Indeed, the entire integrity of the principle

appears to be afi‘ected by maturation. (p. 245)

In order to look at the maturational issue in detail, they compared the adult
learners’ performance in the first part of their study with the performance of three other
groups-ages 4-7, ages 8-13, and ages 14-16. The same materials and procedure were
used. Based on the results of the second part, Johnson and Newport claim that there is a
continuous decline in performance which correlates with age of acquisition (age of arrival
in the States): earlier starters perform better.

Compared with Schachter’s study, Johnson and Newport’s study is much better.
Nevertheless, there are still some problems which we would like to raise in this paper, and
accordingly we suggest that the results of their study remain ambiguous regarding nrling
out the availability of UG in adult second language acquisition. Before discussing those

problems, we first give a brief summary of their study in the following section.

29

A brief summary of Johnson and Newport’s study (1991)

Johnson and Newport’s study examines whether or not there is a critical period
effect for the application of UG to second language acquisition. As in Schachter’s study,
they also select the principle of Subjacency to be their primary test category. They select
Chinese subjects for their study, assuming that “subjects cannot get the correct answers
for English Subjacency structures by strict transfer from their native language” (p. 227).
That is, if the Chinese subjects observe the Subjacency principle and have no problems
with English wh-questions, we have evidence of the operation of UG. On the contrary, if
subjects fail to observe Subjacency, we have evidence to question the availability of UG to
these L2 subjects. Considering these alternatives, Johnson and Newport conduct two
experiments in order to determine any age efi‘ects on the acquisition of Subjacency by
difi‘erent age groups.

In the first experiment, there are 23 adult learners, 14 males and 9 females, who
were first immersed in English in the U. S. A. “between the ages of 18 and 38, with a
mean age of arrival of 25.8” (p. 228). Their average stay in the States before the time of
test is 6.2 years, with a range of 5-12 years. At the time of test, subjects are between the
ages of 23 and 44. All subjects are selected from graduate student, post-doctoral, and
faculty subject pools at University of Illinois; therefore Johnson and Newport claim that all
subjects’ social background is homogeneous. There are 11 undergraduate native speakers
selected as a control group.

Subjects were asked to judge the grammaticality of a series of taped-recorded
sentences and say yes/no after hearing them. And, subjects were told to guess if they were

unsure. All sentences were read twice, with 1-2-second pause separating the repetitions.

3O

There was again 6-9-second pause between sentences for subjects to make their
judgements. There was a break after half of the test, and subjects were told that they
could request one anytime if they felt tired or if the tape was too fast for them. After the
grammaticality judgement task, there was an interview to gather subjects’ background
information. Besides, as Johnson and Newport mentioned, “a subset of the subjects, in
addition, took a short written comprehension test to test whether or not they were
correctly interpreting the questions with subjacency violations” (p. 230).

The tape-recorded test sentences are 180 sentences in all, representing 3 English
structures: noun phrase complements (NP-comp), relative clauses (RC) and WH-
complements (WH-comp). Johnson and Newport provide 12 sentences as basic sentences
representing each of the 3 structures, and each of these basic sentences is presented in four
syntactic forms: declarative (grammatical sentences), subjacency violation (ungrammatical
sentences), control (grammatical sentences), and no subject-auxiliary inversion
(ungrammatical sentences). Here, we copy the examples of the four sentence types for
each of the three sentence structures on Johnson and Newport’s Table 1 (p. 232).

NP-complements

(1a) The teacher knew the fact that Janet liked math. (delcarative)

(lb) *What did the teacher know the fact that Janet liked? (subjacency)
(1c) what did the teacher know that Janet liked? (control)

(1d) *What the teacher did know that Janet liked? (no inversion)

SS relative clause

(2a) The policeman who found Cathy should get a reward. (declarative)

(2b) *Who should the policeman who found get a reward? (subjacency)

31

(2c) What should the policeman who found Cathy get? (control)

(2d) *What the policeman who found Cathy should get? (no inversion)

Wh-complement structures

(3 a) Sally watched how Mrs. Gomez makes her cookies. (declarative)

(3b) *What did Sally watch how Mrs. Gomez makes? (subjacency)

(3 c) Who did Sally show how Mrs. Gomez makes her cookies? (control)

(3d) *Who Sally did show how Mrs. Gomez makes her cookies? (no inversion)
There are another 12 simple grammatical wh-questions. Thus, there are 156 (12x3x4+12)
sentences as the core test sentences. In addition, there are 24 filler sentences served to
increase the variability but not included in analyses.

Johnson and Newport give the overall results in Figure 1 (their Figure 3, p. 235)
which shows the numbers of correct answers over the five sentences types for both the
control group and the subject group. As shown on Figure 1 (see p. 33), the Chinese
group does worst on the subjacency violations. Based on the results, Johnson and
Newport claim that “the subjacency principle is not fully accessible to the mature learner
for the learning of a second language” (p. 243). Besides, due to subjects’ better
performance for relative clauses structures in comparison to the other sentence types, they
also suggest that “language learners show only weak tendencies to observe linguistic
universals, and may therefore implicitly form hypotheses about their late-learned language

that violate these universals” (p. 245).

32

Figure 1 Mean number correct on 5 sentence types by native English

 

 

 

 

 

 

and Chinese speakers
36 T
34 t - t
32 r .
80 r ..
§ 28 r .
3
0 26 " r
3 24 - -
E
a
z 22 ’ —0— Native . ‘
—"°— Chinese
20 * r '
1 L l 1 l A
Simple Declarative Control Nc-lnverslon Subjacency

Sentence Type

Figure 2 The relationship between age of arrival in the U. S. A. and number

correct on subjacency test items and no-inversion test items.

 

 

 

 

 

36 T V I r r
‘6 34 ~
2
a .
o 32 r
8 soL J
U
5
E 28 ‘
0
m
26 r 4
>-
S
o 24 ' r
0
.1
g 22 ' r
m —°— Subjacency
-c-o-o-
20 L No—lnverslon ,
18 L L A A A
Native 4 lo 7 a to 13 14 to 16 Adult

Age ol Arrival

33

Johnson and Newport then do a second experiment to observe efl‘ects of
maturation on subjacency. In order to see the difl‘erent patterns over a range of ages of
acquisition, they find another 21 Chinese subjects, 9 males and 12 females, who learned
English before adulthood. This group of subjects is selected to represent a group of
younger learners whose age of arrival in the States was from 4 to 16, and their years in the
States ranged from 5 to 15, with a mean of 9.6 years. All these younger subjects are
undergraduate students, with the exception of two graduate students. Another difi‘erence
with the first study is that none of the subjects in the second study had education in
English before arriving in the States. The test procedure and materials were identical with
those used in the previous study.

Johnson and Newport compare the results of the first experiment with the results
of the second experiment. The younger learners of the second experiment were grouped
by age of arrival into three groups: 4-7 (6 subjects), 8-13 (9 subjects), and 14-16-year-old
arrival (6 subjects). Performance on subjacency violations and no-inversion sentences
were displayed on Figure 2 (their Figure 7, p. 248).

As shown (see p. 33), only the youngest group (4-7-year-old arrivals) performed
well on both types and was not different fi'om the control group on Subjacency. Based on
the results, Johnson and Newport claim that “there is a fairly continuous decline in the
observance of Subjacency as age of exposure to the language increases” (p. 247). They
say that “linguistic universals such as Subjacency become less accessible to the language

learner with increasing maturation” (page 254).

34

Problems in Johnson and Newport’s study

Johnson and Newport’s study is a better and more convincing study than
Schachter’s study (1990) First, they narrow down their subject group to those with one
language background. This step enables them to focus on the issue of Subjacency alone
without possible interference that may involve other issues relative to linguistic aspects or
language acquisition to some extent. Secondly, they provide enough test tokens per
sentence type, and that increases the validity of their data. Furthermore, Johnson and
Newport apply many statistical analyses to display their test results and to show the
difi’erence among subject groups, and we suggest that their argument is therefore more
persuasive than Schachter’s.

Nevertheless, we also find two crucial problems in their study and then suggest
that their study also fails to rule out the role of UG in second language acquisition. We
will first take a closer look at the subjects, especially those selected in the first study. We
suspect that those adult L2 learners are not qualified to serve as advanced English
learners, so they may not be the ideal subject pool for this study. Secondly, we point out
that Johnson and Newport’s analysis of the adult learners’ test results is misleading to
some extent; and, as we find out, the adult learners’ performance on the Subjacency
violations may not be much difi‘erent from their performance on the control sentences.
These two problems will certainly challenge the validity of Johnson and Newport’s
conclusion fi'om their study. We discuss these two problems in detail in the following

sections.

35

Subjects in Johnson and Newport’s study

We believe that the adult subjects in their first experiment may fail to represent a
proper subject pool of advanced English learners. Johnson and Newport did not provide
any information, such as subject’s TOEFL scores or scores of any other English placement
test, about their subjects’ language proficiency level. As we saw in the previous review
section, the primary criteria of Johnson and Newport’s subject selection are age of arrival
in the States, 5-year minimum stay, and homogeneity of social background. Their average
years of residence are 6.2 and their mean age of arrival is 25.8. It seems that these adult
subjects all meet the criteria. Nevertheless, the problem is that these subjects are not
necessarily qualified as advanced language learners even though they do meet all these
criteria.

Although all 23 Chinese adult learners have a graduate level of education, we still
do not know whether they are really qualified advanced English learners at the time of the
test. Johnson and Newport say that their study “requires that subjects have sufficient
competence in English to be tested on a principle which involves the use of complex
sentences” (p. 229). And, as we mentioned above, they did not apply any standard
language test to measure their subjects’ English proficiency level prior to their study.
Regarding their subject’s English proficiency, the only statement that we can find from
their article is quoted below.

Any subject who did not perform above chance on the simplest structure of our

test, wh-movement in simple sentences, was eliminated. One subject was

eliminated for this reason. (p. 229)

36

In other words, one subject was eliminated because he/she did not respond to at
least half of the 12 simple grammatical wh-questions correctly. Nevertheless, Johnson and
Newport did not provide us any information about the other 23 subjects who perform
“above chance” on these 12 simple questions. We reasonably should see the mean number
and the standard deviation of subjects’ responses on these 12 simple questions.
Otherwise, we suspect that there probably exist significant differences among subjects’
English proficiency level, between above-chance performers and high-scoring performers.

Like Schachter, Johnson and Newport do not give us any information about
possible variation among their subjects. These “above-chance” subjects may be indeed
among the better speakers of English, and they actually had at least “above-chance” scores
on all the tests as we can see in Figure 1. But they may not be good enough to be
qualified as highly advanced subjects who are supposed to be ideal subjects for the study
on Subjacency. Without further information provided in Johnson and Newport’s article,

we simply do not know how good the subjects’ English really is.

Issues on Johnson and Newport’s analyses
Johnson and Newport apply many statistical analyses to strengthen their arguments
in the study. But, there are some ambiguous results, presented in Figure 3 and Figure 4
which show crucial data to support their claim, and may be interpreted difi‘erently. If this
is the case, then we suspect that the results are significant enough to support the
conclusion that UG is not available to the adult learners.
In Figure 3, Johnson and Newport say that “adults learners find the subjacency

sentences more unacceptable than the control sentences” (p. 23 6). Nevertheless, we can

37

clearly see that these Chinese have 24, or 67%, correct responses on the control type; and
22, or 61%, on subjacency violations. Is 6%, 2 test responses, difference between two
sentence types significant enough to make such a claim?

There are two possible interpretations to address the subjects’ performance here.
One is that scoring merely 67% correctly in the control test indicates subjects’ inadequate
English proficiency. Therefore, one may anticipate that they will not do well on
subjacency tests anyway. Ifthis is true, as we discussed in the section on subjects, above,
these adult learners may not be advanced English learners and are not proper candidates
for the study on Subjacency.

The other possible interpretation is that these Chinese adult subjects simply do not
like or are not good enough to handle long and complex wh-questions. As we know, the
test sentences were recorded on tape and the subjects were asked to give yes/no answers
after the questions were read. One may suppose that some (if not all) of these subjects
judged short wh-questions as grammatical ones and long wh-questions as ungrammatical.
If this is the case, one may not be surprised to see that the difl‘erence between the mean
number of correct responses on control sentences and on the Subjacency sentences is
little.

It is normal to predict that researchers will balance the length of sentence of their
test sentences. Because Johnson and Newport do not mention any information about the
length of their test sentence and do not include the hill list of their test sentences in the
paper, we simply do not know and therefore can only suspect this possibility.

Another possible misleading figure is Figure 4 (p. 23 7). The way that Johnson and

Newport interpret Figure 4, which reports the mean number of “No” responses on

38

Subjacency test and control test items by native English and Chinese speakers, is more
complicated than for Figure 3. Here, Johnson and Newport’s analysis is to examine
whether or not there might be a response bias to reject the wh-question sentences in
general. To respond the test sentences correctly, the subjects are supposed to say “No” to
the subjacency violation type of sentences and say “ es” to those control sentences. Here
it is not clear to us why Johnson and Newport compare the number correct in the
subjacency violation type of sentences with the number incorrect in the control type as

summarized in Table 3 below.

 

 

 

Table 3 M number of no res cases on ' cen d control te i b
NO Responses in Subjaceng NO Responses in Control

Mes: 35/ 97% 4/ 1 1%

Ngnnatiyes 22/ 61% 12/ 33%

 

Accordingly, Johnson and Newport claim that “adult learners reject sentences with
Subjacency violation more often than they reject the control sentences”-22: 12 or 61%:
33%. They further suggest that “there is some remaining tendency in the adult learner to
obey Subjacency, and this tendency is not an artifact or either a response bias or a general
belief that wh-movement cannot occur in complex sentences” (p. 237).

Johnson and Newport’s statement here may lead one to come up with a false
interpretation that adult subjects seem to do a better job on the Subjacency test. And, this
false interpretation is certainly not consistent with what they have argued for in their
article. Here we revise the comparison also by using the data of Figure 4. First, we
compare the number of correct responses on Subjacency violations with the number of

correct responses on control sentences. Then, we compare the number of incorrect

39

responses of Subjacency violations with the number of incorrect responses on control

sentences, in Table 4.

Table 4 Mean number of correct responses and incorrect responses

Correct responses on Sub. vs. Con. Incorrect responses on Sub. vs. Con.

 

 

flfives 35 (97%) : 32 (89%) 1 93%) : 4 (11%)
Nonpap'ves 22 (61%) : 24 (67%) 14 (39%).- 12 (33%)

 

The results here give us another view which is somewhat difl‘erent fiom their
report. As revised in Table 4, the results become 22 (61%) : 24 (67%) and 14 (39%) : 12
(33%). The difl‘erence of adult nonnative subjects’ performance between judgment of
Subjacency violations and control sentences, 6% (or 2 tokens precisely), is not significant
at all. White (1996: 104) gives a possible explanation for the non-difference and suggests
that “these L2 speakers might have grammars without long-distance movement, using pro
as their empty category.” Another explanation is that these subjects’ English proficiency
is actually not advanced enough although they may be better learners. They are able to
perform only above chance on both the control sentences and the Subjacency violations.

In short, although Johnson and Newport’s study is a much better study compared
to Schachter’s, there are still some problems as we discussed here. Accordingly we
suggest that the results of their study remain ambiguous regarding ruling out the

availability of UG in adult second language acquisition.

4O

GENERAL DISCUSSION

Both Schachter’s and Johnson and Newport’s studies are in favor of the Critical
Period Hypothesis to some extent. Schachter’s conclusion gives a stronger version of the
issue and her Incompleteness Hypothesis seems to propose that L2 acquisition is not
attainable because the principles of UG remain idle if not triggered in L1 acquisition and
the values for those triggered cannot be revised. Therefore, the acquisition of L2 is
deemed to be incomplete. As we discussed in the previous section (3.3.1. on p. 29),
Schachter seems to suggest that UG is a one-time chance and can only happen in L1
acquisition. That is, the decline of UG is abrupt.

For example, a Chinese child completes the acquisition of wh-questions in
Mandarin at age of 3 or 4 presumably. According to Schachter’s hypothesis (see our
critique in 3.3.1. on p. 29), his/her language system does not include the principle of
Subjacency to produce wh-questions, and therefore he will not be able to become an
English native speaker even if he/she moves to the States since then. In another words,
even for a child, to learn a second language is impossible. Of course, we do not believe
and do not like to believe this is true.

Instead, Johnson and Newport propose a weaker version of the age effects and
suggest that UG is under the maturational constraint and UG gradually declines with age.
As we can tell from their Figure 7 (p. 248), subjects’ performance on subjacency shows no
significant difference, compared to the scores of the control group, up to the age group of
8 to 13. Therefore, according to Johnson and Newport’s hypothesis, that Chinese child

can still become an English native speaker, although most principles of UG were already

41

set when he/she acquired Mandarin as his/her L1. Perhaps he/she can achieve near
complete English competence if he/she moves to the States before the age of 13.

We have discussed some crucial problems found in these two studies and then
argue that both studies fail to claim that their adult subjects have no access to UG in
English acquisition. What’s in common to both studies is that their adult subjects are not
the proper subject pool for the study on Subjacency. Subjects’ English proficiency level is
questionable in both studies; therefore, the validity of the test results of both studies is
problematic. Nevertheless, as we mentioned in the introduction section, the purpose of
this paper is not to argue for the availability of UG in SLA. We have to admit that the
debate between the two opposite views on availability of UG in SLA will continue.

We also have to admit (or to accept) the common phenomenon that, unlike in L1
acquisition, most adult subjects have less (or little) chance to be successful in their L2
acquisition. To account for the difference between L1 acquisition and L2 acquisition,
Bley-Vroman (1990) proposes the Fundamental Difl‘erence Hypothesis which suggests
that UG is no longer accessible in L2 acquisition. He claims that the general failure of
adult L2 learners indicates that UG is not applicable in L2 acquisition. Furthermore, the
learner can only rely on those principles of UG triggered in L1 acquisition, and they have
to adopt other general learning mechanisms such as analogy, reasoning, or hypothesis
testing. Unfortunately, all these mechanisms are insuficient for learning a language, so
the learner is no longer guaranteed to succeed in L2 acquisition.

Felix (1985, 1986) proposes the Competition Module Hypothesis that provides
another explanation for the less successful L2 acquisition. He claims that UG is still

available to the L2 learner but it has to compete with general learning mechanisms.

42

Unfortunately, the competition between UG and those learning mechanisms may result in
some sort of interference in the domain of UG and then causes the failure of L2
acquisition.

Because of the general failure of L2 acquisition, superficially it seems easier to
come up with the hypothesis that the difference between LIA and L2A entails
unavailability of UG in SLA We all may easily make such a claim because we will not
have problems to find a lot of adult learners to support our argument. But, the truth is
that there ARE highly advanced L2 learners who have near-native performance, and we
then owe ourselves an explanation for that. Greg (1996: 66) points out the fact that
researchers who favor UG’s unavailability in L2 acquisition need to show that “all
successful examples of L2 acquisition can be accounted for either by ‘transfer’ of L1
grammatical properties to the IL grammar, or else by the application of learning
mechanisms not intended specifically for acquisition tasks.”

Most research in the field of SLA is inspired by Chomsky’s “Government and
Binding” (1981). Until this decade, researchers have tried to come up with all sorts of
theories to account for L2 acquisition, whether for the success or for the failure. So far,
as Long (1993: 225) claims, there are at least 40 to 60 theories in the field of SLA The
study of second language acquisition is still in a developing process. There is still plenty
of work left to be done. Hopefully, in the near future we will have a better understanding
and come up with a unique theory of second language acquisition, a linguistic theory

which will benefit the field of second language education.

43

REFERENCES
Birdsong, D. 1992. Ultimate attainment in second language acquisition. Langu_age 68:
706-755.

Bley-Vroman, R. 1990. The logical problem of foreign language learning. Lingaistic
Analysis 20: 3-49.

Bley-Vroman, R. 1989. What is the logical problem of foreign language learning. In S.
Gass and J. Schachter (Eds), Lingm'stic perspectives on second langaage
aaquisition, pp. 41-48. Cambridge: CUP.

Chomsky, N. 1981. Lectures on government and binding. Dordrecht: Foris.

Chomsky, N. 1975. Reflections on lam_gu_aga. New York: Pantheon Books.

Coppieters, R 1987. Competence difi‘erences between native and near-native speakers.
Langaage 63: 544-573.

Custard, E. 1998. The complete book of college_s. New York: Random House.

Eubank, L. 1991. Introduction. In L. Eubank (Ed), Point-counte_rpoint; 11de
war in the second language, pp. 1-48. Amsterdam: John Benjamins.

Felix, S. 1988. UG-generated knowledge in adult second language acquisition. In S.
Flynn and W. O’Neil (Eds), Linguistic theog in second Image aaguisition, pp.
277-294. Dordrecht: Kluwer.

Felix, S. 1986. Cogm‘tion and langaage growth. Dordrecht: Foris.

Felix, S. 1985. More evidence on competing cognitive systems. econd Lan e
Research 1: 47-72.

 

Flynn, S. 1993. Interactions between L2 acquisition and linguistic theory. In F. Eckrnan
(Ed), Confluen : Lin ' ' L2 isii n h . PP. 15-36.
Amsterdam: John Benjamins.

Flynn, S. and Manuel, S. 1991. Age-dependent effects in language acquisition: an
evaluation of “critical period” hypothesis. In L. Eubank (Ed), Point-countgpoint:
universal ggammar in the second language, pp. 117-146. Amsterdam: John
Benjamins.

 

Gregg, K. 1996. The logical and developmental problems of second language
acquisition. In Ritchie W. and Bhatia T. (Ed), Handbook of second language
muisition, pp. 49-81. San Diego, Calif: Academic Press.

Haegeman, L. 1991. Introduction to government and binding theog. Cambridge, MA:
Blackwell. Ch. 7, Wh-movement, pp. 335-392. Ch. 8, An inventory of empty
categories, pp. 393-438. Ch. 10, Barriers: an introduction, pp. 477-512.

Johnson, J. 1992a. Critical period effects in second language acquisition: the efi‘ect of
written versus auditory materials on the assessment of grammatical competence.
Language Learning 42: 217-248.

Johnson, J. and Newport, E. 1991. Critical period efl‘ects on universal properties of 1
language: the status of subjacency in the acquisition of a second language. 1
Cogm'tion 39: 215-258.

Johnson, J. and Newport, E. 1989. Critical period efi’ects in second language learning:
the influence of maturational state on the acquisition of English as a second
language. Cogm'tive Psychology 21: 60-99.

Krashen, S. 1973. Lateralization, language learning, and the critical period. Lanmge
Learning 23: 63-74.

Lenneberg, E. H. 1967. Biological foundations of language. New York: John Wiley &
Sons. Ch. 4., Language in the context of growth and maturation, pp. 125-187.
Ch.9, Toward a biological theory of language development, pp. 371-395.

Long, M. 1993. Assessment strategies for second language acquisition theories. Applied
Linguistics 14: 225-249.

Long, M. 1990. Maturational constraints on language development. Studies in Second
Language Aguisition 12: 251-285.

Martohardjono, G. and Gair, J. 1993. Apparent UG inaccessibility in second language
acquisition: misapplied principles or principled misapplications? In F. Eckrnan
(Ed), Confluence: Linguistics, L2 amuisition and spgaah pathulogy, pp. 79-103.
Amsterdam: John Benjamins.

 

Newport, E. 1990. Maturational constraints on language learning. Cogm'tive Science 14:
1 1-28.

Oyama, S. 1982. The sensitive period and comprehension of speech. In S. Krashen, R.

Scarcella, and M. Long (Eds), Child-adult differences in second Laugflge
amuisition, pp. 39-51. Rowley, MA: Newbury House.

45

Patkowski, M. 1980. The sensitive period for the acquisition of syntax in a second
language. Language Learning 30: 440-472.

Schachter, J. 1990. On the issue of completeness in second language acquisition.
Second Language Research 6: 93-124.

Schachter, J. 1989. Testing a proposed universal. In S. Gass and J. Schachter (Eds),

Linguistic perspectives on seggnd languaga Muisition, pp. 73-88. Cambridge:
CUP.

White, L. 1996. Universal grammar and second language acquisition: current trends and
new directions. In Ritchie W. and Bhatia T. (Ed), Hand ok of second lan e
amuisition, pp. 85-120. San Diego, Calif: Academic Press.

White, L. 1992a. Subjacency violations and empty categories in L2 acquisition. In H.
Goodluck and M. Rochemont (Eds), Island constraints, pp. 445-464. Dordrecht:
Kluwer.

White, L. 1990. Second language acquisition and universal grammar. Studies in Secon_r_l
Language Aguisition 12: 121-133.

White, L. 1989. Universal ggammar and second language aaquisition. Amsterdam:
John Benjamins.

White, L. 1988. Island efi‘ects in second language acquisition. In S. Flynn and W.

O’Neil (Eds), Linguistic thggg in mnd lauguage amuisition, pp. 144-172.
Dordrecht: Kluwer.

46

“Ilillillilllllllilif