second language aquisition and universal grammar

13
8/10/2019 Second Language Aquisition and Universal Grammar http://slidepdf.com/reader/full/second-language-aquisition-and-universal-grammar 1/13 SSLA 12,121-133. Printed in the United States of America. SECOND LANGUAGE ACQUISITION AND UNIVERSAL GRAMMAR Lydia White McGill University In this article, the motivation for Universal Grammar (UG), as assumed in the principles and parameters framework of generative grammar (Chomsky, 1981a, 1981b), is discussed, particular attention being paid to the  logical problem  of first language acquisition. The potential role of UG in second language (L2) acquisition is then considered. Three different positions are reviewed: (a) the claim that UG is not available to L2 learners; (b) the claim that UG is fully available; and (c) the claim that the L2 learner s access to UG is mediated by the mother tongue. This raises the issue of what kind of evidence can be used to decide between these three positions. Recent experimental research which argues for one or another of these positions by investigating the L2 status of the Subjacency Principle is reviewed, and the implications of this research are discussed. In recent years there has been a growing interest in the potential relevance of the principles and parameters approach to linguistic universals, as realized in Govern- ment and Binding (GB) Theory (Chomsky, 1981a, 1986b), for explaining certain aspects of L2 acquisition. There are several reasons for this interest. First, it is only comparatively recently that the  logic l  problem  of language acquisition has been well articulated. Focusing on the complexity of the end result of the acquisition process and the fact that many aspects of the target grammar are not at all obvious from the input, researchers have asked how language can be acquired in such cir- cumstances (e.g., Lightfoot, 1982; Pinker, 1984; Wexler & Culicover, 1980). Linguists This research was conducted with the assistance of the following research grants, for which I am grateful: Social Sciences and Humanities Research Council Canada Research Fellowship *455—87—0201, Social Sciences and Humanities Research Council of Canada research grant *410—87—1071 (with Lisa Travis), and Government of Quebec FCAR research grant *88 EQ 3630 (with Lisa Travis). ©  199 Cambridge University Press 0272-2631/90 5.00 + .00 121

Upload: kaiser

Post on 02-Jun-2018

251 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Second Language Aquisition and Universal Grammar

8/10/2019 Second Language Aquisition and Universal Grammar

http://slidepdf.com/reader/full/second-language-aquisition-and-universal-grammar 1/13

SSLA

12,121-133. Printed in the United States of America.

SECOND LANGUAGE

ACQUISITION AND UNIVERSAL

GRAMMAR

Lydia White

McGill University

In this article, the motivation for Universal Grammar (UG), as assumed

in the principles and parameters framework of generative grammar

(Chomsky, 1981a, 1981b), is discussed, particular attention being paid

to the logical problem of first language acqu isition. The potential role of

UG in second language (L2) acquisition is then considered. Three

different positions are reviewed: (a) the claim that UG is not available to

L2 learners; (b) the claim that UG is fully available; and (c) the claim that

the L2 learner s access to UG is med iated by the mother tongue. This

raises the issue of what kind of evidence can be used to decide

between these three positions. Recent experimental research which

argues for one or another of these positions by investigating the L2

status of the Subjacency Principle is reviewed, and the implications of

this research are discussed.

In recent years there has been a growing interest in the potential relevance of the

principles and parameters approach to linguistic universals, as realized in Govern-

ment and Binding (GB) Theory (Chomsky, 1981a, 1986b), for explaining certain

aspects of L2 acquisition. There are several reasons for this interest. First, it is only

comparatively recently that the  logic l problem  of language acquisition has been

well articulated. Focusing on the complexity of the end result of the acquisition

process and the fact that many aspects of the target grammar are not at all obvious

from the input, researchers have asked how language can be acquired in such cir-

cumstances (e.g., Lightfoot, 1982; Pinker, 1984; Wexler & Culicover, 1980). Linguists

This research was conducted with the assistance of the following research grants, for which I am grateful:

Social Science s and H umanities Research C ouncil Canada Research Fellowship *455—87—0201, Social S cien ces

and Humanities Research Council of Canada research grant *410—87—1071 (with Lisa Travis), and Government

of Que bec FCAR research grant *88 EQ 3630 (with Lisa Travis).

©

 199

Cambridge University Press 0272-2631/90 5.00 + .00 1 2 1

Page 2: Second Language Aquisition and Universal Grammar

8/10/2019 Second Language Aquisition and Universal Grammar

http://slidepdf.com/reader/full/second-language-aquisition-and-universal-grammar 2/13

122 Lydia White

argue that there is an innate Universal Grammar (UG), consisting of principles and

parameters which help to explain the acquisition of language. Linguistic theory aims

to characterize these principles and parameters, thereby providing an account of our

linguistic competence. This raises the question of whether L2 learners also achieve

competence which goes beyond the input, and whether L2 acquisition is constrained

by principles of UG (Bley-Vroman, 1989; Rutherford, 1988; White, 1985; Zobl, 1983).

Second, although generative grammar has always assumed an innate component,

proposals for the content of UG were surprisingly sparse in earlier versions of the

theory, so that it was difficult to come up with specific hypotheses as to what might be

guiding the language learner. In contrast, at the present time, the form and content of

UG are the subject of extensive investigation. Proposals are considerably more pre-

cise than they were, and it has become possible to test whether UG is a component of

the second language acquisition process. Third, the concept of parameters within UG

has resulted in increasingly sophisticated accounts of similarities and differences

across languages. This is particularly useful in the L2 acquisition context, where there

are potentially interesting relationships between the m other tongue and the language

being acquired which may be explained in terms of parameters of UG.

THE LOGICAL PROBLEM OF LANGUAGE ACQ UISITION

Linguists motivate UG by pointing to the end result of first language acquisition,

namely the adult grammar. The linguistic competence of adults is abstract, subtle,

and complex. All (normal) children acquire this linguistic competence in their mother

tongue, despite the fact that there is a mismatch between the primary linguistic input

(the linguistic data that they are exposed to) and the system actually attained.

In particular, the input underdetermines  linguistic competence. Children acquire

properties of language which are not immediately obvious and which are not explic-

itly taught; children and adults possess knowledge of grammaticality, ungrammatical-

ity, ambiguity, and various other subtle and complex phenomena, which go far

beyond the actual sentences that an individual learner may happen to have been

exposed to. An acquisition problem arises if we impute to the child only some general

cognitive ability to make analogies or generalizations solely on the basis of linguistic

input. If learning were of this type, one would expect the child to make many false

generalizations, to produce errors which have not been attested in child language,

and to fail to work out many properties of language.

The distribution of pronouns provides an example. A child has to learn when a

pronoun can refer to a noun phrase (NP) within the same sentence and when it

cannot. As indicated in (la ) and

 (lb),

 a noun phrase and a pronoun cannot refer to the

same person in a simple sentence, and the same is true of two pronouns, as in (lc):

(1) a. *Janej washed herj

b.  *She,

 washed Jan e

:

c. 'She, washed her.

However, these same sentences are grammatical when two different people are

involved, as in (2):

Page 3: Second Language Aquisition and Universal Grammar

8/10/2019 Second Language Aquisition and Universal Grammar

http://slidepdf.com/reader/full/second-language-aquisition-and-universal-grammar 3/13

L Acquisition

  nd

 U 23

(2)  a.  Jane ; washe d herj

b.  Shej washed Janej

c. Shej wa shed her

f

When

 the

 child hears sentences like those

 in

 (2),

 it is

 likely

 to be

 clear from context

that

  Jane

 and the  pronoun refer to two  different people. But  discovering that they

refer to two different people in particular utterances is not the same as knowing tha t

they

 may

 never refer

 to the

 same person

  in

 this kind

 of

  structure. Furthermore,

 the

situation is different  in complex sentences, as can be seen in 3):

(3)  a.  Janej wa tched television before shej had her  dinner

b.  Before Janej had her dinner she, wa tched television

c. Before

 she, had her

  dinner Jane, watched television

d.  *Shej wa tche d television before Jan ej had her  dinner

e. Shej wa tched television before she; had her  dinner

In (3a), (3b),

 and

 (3c), Jane

 and the

 pronoun she

 can be

 coreferential

  but can

 also

refer to two  different people; in  other words, the sentences are potentially ambigu-

ous).

 In

 (3d) they must

 be

 disjoint

  in

 reference.

 The

 coreference

  and

 noncoreference

possibilities are not  simply a  matter of  sentence type; it is not the  case that  coref-

erence is impossible in simple sentences and possible in multi-clause sentences, since

(lb)

 and

  3d) are both ungram matical, despite being different in this respect. Nor is

 it a

matter of  linear relationships between  the noun phrase and the pronoun; Jane pre-

cedes the pronoun  in 3a) and 3b), which are grammatical, but also precedes it in

(la),

 which is ungrammatical. Jane follows

 the

 pronoun

 in

 (3c), which is grammatical,

and in lb), which is ungrammatical. Furthermore,  the picture changes again when

the sentence contains two pronouns instead of a noun phrase and a pronoun: (3d) is

ungrammatical with intended coreference

  but

 coreference

  is

  possible between

  the

two pronouns in

 (3e).

 However, this is not the case in (lc), where coreference remains

impossible between

 the

 two pronouns.

LI acquirers

 are not

 explicitly taught

  the

  coreference possibilities between

 pro-

nouns and noun phrases. Nor is it clear how the distribution is to be learned from the

primary input alone. Although (3a), (3b),

 and

 (3c)

 may be

 heard with intended

 coref-

erence,

 it is not

 necessary

  in

  these sentences;

 but

  disjoint reference

  is

 necessary

 in

(la), (lb), and

 (3d).

 If NPs are replaced by pronouns, the possibilities change again.  If

the child comes

  to the

  acquisition task equipped solely with abilities

  to

  extract

generalizations from input data,

  it

  would seem impossible

  to

  arrive

  at the

  correct

generalizations without a great many errors, if at all. The problem is confounded by

the fact that children appear

  to get

  little

 or no

  explicit negative evidence, that

 is,

evidence about ungrammaticality, about what sentences

 are not

 possible

 in the lan-

guage they are learning (Brown & Hanlon, 1970). Yet one of the things that children

acquire is knowledge

 of

 ungrammaticality;

 in

 this case, knowledge that

 the

 sentences

in  1) and (3d) are ungrammatical if coreference  is intended.

1

It

 has

 been suggested that

  the

  logical problem

  of

  language acquisition

  is

 dimin-

ished by the fact that children get simplified input that is, input in the form of short,

simple, grammatical sentences (e.g., Ellis, 1985, p. 211; McLaughlin, 1987, pp. 105-

106). But such input will

 not

 help

 at all

 here. Suppose that

 the

 child gets input only

 in

Page 4: Second Language Aquisition and Universal Grammar

8/10/2019 Second Language Aquisition and Universal Grammar

http://slidepdf.com/reader/full/second-language-aquisition-and-universal-grammar 4/13

124 Lydia White

the form of the sentences in (2), where context indicates that

 Jane

 and the pronoun

do not refer to the same person. This provides no clue as to the different behavior of

the pronouns, in terms of their coreference possibilities, in the sentences in (3). The

simple clause structures certainly do not solve the problem of how to treat pronouns

and NPs in more complex structures; indeed, if anything, they are highly misleading

since coreference possibilities crucially differ in simple clause and multiclause struc-

tures:

  disjoint reference is always required in cases like (1), but only sometimes

required in cases like (3).

Plausible theories of language acquisition must assume realistic input (in other

words, that children proceed largely on the basis of positive evidence) and must

explain how children come by their unconscious knowledge of language on the basis

of this kind of input. Since input alone is insufficient to account for acquired proper-

ties of grammar, the assumption in generative grammar is that knowledge about

what is and is not possible stems in part from an innate UG, containing principles and

parameters which constrain grammars in various ways. In GB theory, the distribution

of pronouns and noun phrases (and also of reflexives and other anaphors) is deter-

mined by the Binding Theory (Chomsky, 1981a). This is part of UG and constitutes

knowledge that the child brings to bear on the acquisition task, thus reducing the

problem of acquiring the distribution of pronouns solely on the basis of input. Other

principles account for o ther subtle and complex properties.

PRINCIPLES AN D PARAMETERS O F UG

As w e have see n,  a co nsideration of the logical prob lem  of language acquisition has

led  to  proposals that certain aspects of language must  be innately present  in the first

language learner  in th e form  of  Universal Grammar, accounting for th e fact that the

child acquires

 all th e

 com plexities

  and

 subtleties

  of

  grammar although these

 are not

explicit  in th e  input data. UG consists,  in part,  of a  number  of  fixed abstract princi-

ples.

  These predispose

  th e

 child

  to

  organize language

  in

  certain ways

  and

 lead

  to

rather limited possibilities  for  grammar construction, instead  of the  full range that

would b e logically possible if  language learning proceeded only by applying general

inductive learning strategies.

An example  of a principle  of UG is Subjacency, which constrains how  far catego-

ries,

  such  as wh- ex pressions,  m ay move . Sen tence s like those  in  (4a) and (4b) are

ungrammatical, although  th e  corresponding statements  in (4c) an d (4d) ar e

grammatical:

(4)  a.  *What did Mary wonder whether John had bought?

b.  *What did Mary make the claim that John had seen?

c. Mary wondered whether John had bought a new car

d. Mary made the claim that John had seen a ghost

Furthermore, other  wh- questions  of  equivalent length, complexity,  or meaning are

grammatical, as ca n b e seen  in  (5):

(5)  a.  What did Mary believe that John had bought?

b.  What did Mary claim that John had seen?

Page 5: Second Language Aquisition and Universal Grammar

8/10/2019 Second Language Aquisition and Universal Grammar

http://slidepdf.com/reader/full/second-language-aquisition-and-universal-grammar 5/13

L Acquisition and

 UG

  125

Wh-

 movement is assumed to b e restricted by the Subjacency Principle, which stipu-

lates that movement

  may not

 take plac e over mor e than

  o n e bounding node  at a

time, where the bounding nodes

  for

  English are S and NP.

 In

 (6),

 t h e

 relevant bound -

ing nodes

  for

 (4a) and (4b) have been included. The trace (t) marks the original site of

the  wh-  word:

(6)  a.  *Whatj [3 did Mary wonder  [ • whether [3 John had bought tj]]

b.  *Whati

 [s

 did Mary make [

NP

 the claim y  that

  [5

 John had seen tj]]]

In (6a), t h e  wh-  word has crossed two S nodes, a nd in (6b) it has crossed two  S nodes

and  an NP node,  in  both cases violating Subjacency.  In th e sentences  in (5),  on the

other hand,

  what  can

  pass through

  th e

  intermediate  wh- position

  in

  COMP

 (the

position already occupied

  by

  whether

  in 6a),

 avoiding

  a

  Subjacency violation

  by

moving  in steps over one bo undin g n ode  a t a time, as shown  in (7):

(7) What, [s did Mary believe [ • t; that [g John had bought tj]]

It

  is

  unlikely that

  th e

  input will offer

  th e

 language learner clear evidence

  to

  distin-

guish between grammatical cases like (5) and ungrammatical cases like (4a) and (4b).

Much of the positive eviden ce from English will exemplify extensive  wh- movement,

especially in simple sentences, and

  it

 is not clear how th e restrictions

  on

 its operation

in more complex structures are to be learn ed, unless one assum es built-in know ledg e

of Subjacency, which provides children with advance knowledge  of  these restrictions

and prevents them from making Subjacency violations

  in the

 course

  of

  language

acquisition (Otsu, 1981).

Although linguistic principles like Subjacency  a re  universal, this does  no t  mean

that every principle  of  UG necessarily operates  in  every language. Subjacency,  for

example, does not operate

  in

 langua ges withou t syntactic move men t rules, since

 it is

a constraint  on  movement  and is of no  relevance  in th e  absence  of  movement.

Languages which  do not have  wh- movem ent, such  as  Japanese, Korean,  an d Chi-

nese, will provide

  no

 occasion

  for

 Subjacency

  to

 ope rate . One would not, however,

expect  to  find languages with  wh-  movement  bu t  without Subjacency. Although

languages vary  in terms  of  their syntactic properties, principles  a re fixed  a n d must

apply to all languages which exhibit the relevant properties.

In addition  to fixed principles, UG conta ins para me ter s, with param ete r settings

that differ from language to language. A central idea behind the concept

  of a

 parame-

ter  is that several apparently indepe ndent properties  of  language  ar e in  fact related

and cluster together. By means  of param eters, certain aspects  of  crosslinguistic varia-

tion are built into UG, aspects which could not easily

  b e

 induced from input alo ne.

 A

particular parameter setting is triggered by input from the language being learned.

For example, Subjacency

  has a

  parameter associated with

  it .

 There

  is

  variation

across languages  in  terms  of the  bounding nodes that they adopt.  In  particular,

languages differ as to the bounding status of S; Italian and French,  for exampl e, have

NP  and S' as  bounding nodes,  bu t not S (Chomsky, 1981b; Rizzi, 1982; Sportic he,

1981).

2

  This means that  in thes e languag es certain ex tractio ns will b e possible which

are not possible  in English. Equivalents of  (4b) will still be ungrammatical because NP

and  S' are  bounding nodes,  and the wh- word  has crossed both  of  these. Sentences

Page 6: Second Language Aquisition and Universal Grammar

8/10/2019 Second Language Aquisition and Universal Grammar

http://slidepdf.com/reader/full/second-language-aquisition-and-universal-grammar 6/13

126 Lydia White

involving extractions from so-called

 wh-

 islands

3

 such as

 (4a),

 on the other hand, will

be grammatical in French and Italian, because only one bounding node, namely S', is

crossed. Evidence that S is not a bounding node in these languages is provided by

sentences involving extractions out of

  wh-

  islands. Such sentences will trigger the

appropriate parameter setting.

Thus,

 Subjacency is a principle of

 UG

 which limits movement so that at most one

bounding node can be crossed. This is fixed and applies across languages, with the

exception that it is irrelevant in the absence of syntactic movement. Associated with

it is a param eter which permits variation across languages; this parameter concerns

what counts as a bounding node for Subjacency.

ON THE AVAILABILITY OF UG IN SECOND

LANGUAGE ACQUISITION

Assuming that LI acquisition is mediated by innate universal principles, the question

arises as to whether L2 learners, particularly adults, still have access to the principles

and parameters of UG.

 A

 number of different positions on this question are represent-

ed in the literature, three of which will be discussed here (for further details, see

White, 1989b). The differences in these positions can be traced in part to whether

emphasis is on the relative lack of success of L2 learners (their knowledge, fluency,

and ultimate attainment rarely approach that of native speakers) or their success

(they do acquire many complex properties of language which are not transparent in

the input).

Many of those who focus on difficulties faced by L2 learners, and differences

between LI and L2 acquisition, have suggested that these differences and difficulties

can best be explained on the assumption that UG is no longer available to adult L2

learners (Bley-Vroman, 1989; Clahsen, 1990; Clahsen & Muysken, 1986; Schachter,

1988b). Proponents of this view argue that the learning mechanisms underlying adult

L2 acquisition are radically different from those underlying LI acquisition, and that

they are not unique to language.

In contrast is a position which assumes that

 UG

 is still available. As we have seen,

UG was motivated for LI acquisition by a consideration of the projection problem:

Despite the fact that certain properties of language are not explicit in the input,

native speakers end up with a highly complex, unconscious mental representation of

their language, suggesting that universal principles must mediate acquisition and

shape knowledge of language. It seems most unlikely that L2 input will contain

explicit information about these kinds of properties in the L2; thus, L2 input will

underdetermine the L2 gram mar in precisely the same way that LI input underdeter-

mines the LI grammar. If L2 learners attain unconscious knowledge of the L2 which

goes beyond the input and which could not be acquired on the basis of general

learning strategies or of the LI, it suggests that UG must still be involved. One

possibility, then, which represents the strongest form of the hypothesis that UG is

available, is that LI and L2 acquisition are identical with respect to the operation of

UG, and that UG explains the acquisition of complex linguistic knowledge in both

contexts.

Page 7: Second Language Aquisition and Universal Grammar

8/10/2019 Second Language Aquisition and Universal Grammar

http://slidepdf.com/reader/full/second-language-aquisition-and-universal-grammar 7/13

L Acquisition and U 127

A compromise position assumes that UG is still available but that the L2 learner's

access to it is mediated by LI knowledge. In fact, this position is consistent with two

very different assumptions about the essential availability of UG. On the one hand,

the LI might play a role and UG be unav ailable. In this view, UG is dead as an active

force in L2 acquisition, but aspects of it encoded in the LI can still be tapped. Thus,

L2 learners might give the impression of having access to complex and subtle knowl-

edge,

  but this would be entirely due to their ability to tap this knowledge as encoded

in the LI. This means that fixed principles exemplified in the LI, as well as LI

parameter settings, will be accessible to the L2 learner. However, principles not

tapped by the LI will be inaccessible, and parameters cannot be reset to their L2

values. On the oth er han d, attributin g a role to the LI is also consistent with the claim

of the essential availability of UG (e.g., Flynn, 1987; White, 1988b). White's claim, for

example, is that the L2 learner may use principles and parameter sett ings from the

L I,  at least initially, as an interim way of dealing with the L2 data. However, the

learne r is not assum ed to be necessa rily stuck with LI pa ram eter settings; instead,

parameter resetting to the L2 value is in principle possible, given appropriate L2

input interacting with a still active UG.

EVIDENCE

Theoretical and Methodological Considerations

What kind of evidence can decide among these possibilities? It is not sufficient to

point to general differences betw een LI an d L2 acquisition to argu e for non-availabili-

ty of UG, or to general similarities to argue for its availability. UG is a claim about

knowledge in a particular domain, a claim that our knowledge of language is con-

strained by certain a bstract but crucial principles. Therefore, th e poten tial availability

of UG in L2 acquisition must be investigated within this same domain. If UG is no

longer available to adults, and second language acquisition proceeds by means of

general cognitive abilities, L2 learners should not be able to work out abstract

properties of the L2 which are un derd eterm ined by the input data. W here the input is

insufficiently precise to allow L2 learners to induce the relevant properties of the

grammar, they should not be able to achieve full success. Thus, one form of evidence

for the hy pothesis that UG op erates in L2 acquisition will be ev iden ce tha t L2 lear ner s

in fact attain the kind of complex and subtle knowledge which is attributable to UG.

However, LI knowledge is a confounding factor. If a particular principle of UG

operates in both the LI and L2, and if L2 learners show evidence of observing this

principle, this could be attribute d to transfer of LI kno wled ge. Similarly, if L2 learn ers

show evidence of applying LI parameter settings to the L2, this is actually neutral

con cernin g the availability or non-availability of UG. Thu s, th e strongest arg um en ts

in favor of the operation of UG (complete or partial) in L2 acquisition will be made in

cases where effects of the LI can be minimized.

In order to eliminate the LI as a source of UG-like knowledge, two situations can

be isolated, one relevant to the operation of principles and the other relevant to

parameters. In the case of principles, if UG is not available, then L2 learners should

not

  be able to sort out aspects of the L2 where both of the following hold:

Page 8: Second Language Aquisition and Universal Grammar

8/10/2019 Second Language Aquisition and Universal Grammar

http://slidepdf.com/reader/full/second-language-aquisition-and-universal-grammar 8/13

128 Lydia White

(8)  a.  some principle operates in the L2 but not the LI, and

b.  the input underdetermines the L2 grammar.

Similarly,

 in th e

 case

 of

 parame ters, L2 learners should

  not be

 able

  to

 acquire

  the

  L2

value of a parameter where:

(9)  a.  the LI and

 L2

 have different values for some parameter, and

b.

  the input underdetermines the L2 grammar.

If  L2  lea rne rs successfully arriv e  at the  relevant properties  of the L2  under such

conditions, then there  is support  for th e claim that UG is still truly accessible, rather

than inaccessible or weakly accessible only via the L I .

How  can on e discover whether  L2 learn ers have indeed arrived  at the relevant

abstract properties

 of

 the L2? This is not so met hing that can

 be

 dete rmined simply by

observing  the productions  of L2 learners.  For example, absence  of UG violations in

spontaneous production does  not  allow  o n e automatically  to  assume that  UG is

operating, since  a  learner might fail  by chance  to  produce structures which violate

principles of UG. Many of the relevant struc tures are fairly complex and simply might

not arise  in the  normal course  of conversation  or  during observation by an experi-

menter. An L2 learner might never have occasion to utter a sen ten ce like (4b), and yet

might consider

  it to be

 grammatical. Conversely,

  the

 presence

  of

 violations does

 not

necessarily mean that  UG is un available;  o ne needs  to be  sure that violations are

systematic rather than random.

  If a

 learner utters

 a

 sentence like (4b)

 on

 one particu-

lar occasion, this could just  be a  performance error,  or it  might  be  indicative of

linguistic com peten ce w hich  is no t cons trained by Subjacency.

Thus,  it is necessary  for the experimenter  to be able  to manipulate  the sentence

types  to be  investigated, rather than relying  on  their chance occurrence  in produc-

tion data, and to investigate various aspects of L2 knowledge where  the operation of

UG might have effects, looking  at L2 comprehe nsion  and production  in a  controlled

way. Since linguistic competence includes knowledge  of ungrammaticality,  L2 learn-

ers must somehow  be made  to  reveal, directly  or  indirectly, whether they have this

knowledge,  by means  of  tasks where sentences which violate universal constraints

are deliberately included

  for

 investigation.

  If

  learners show that they

  do

  indeed find

violations of UG to be ungrammatical, this lends support  to the UG hypothesis. If, on

the other hand, they systematically accept UG violations,  it  supports claims  for the

unavailability of UG.

Experimental Evidence for and Against UG Principles in SLA

In this section, some experimental evidence will be considered for and against the

operation of principles of UG. Subjacency will be used as an example, since a number

of experimental L2 acquisition studies have recently investigated the status of this

principle.

In order to demonstrate the operation of Subjacency most effectively, both of the

conditions outlined in (8) must hold. That is, L2 learners must not be able to draw on

LI knowledge to work out the restrictions on

  wh-

  movement in the L2, and these

restrictions must not be inducible from the L2 input alone. The first condition is met

by investigating learners whose Lls do not make use of  wh-  movement, who are

Page 9: Second Language Aquisition and Universal Grammar

8/10/2019 Second Language Aquisition and Universal Grammar

http://slidepdf.com/reader/full/second-language-aquisition-and-universal-grammar 9/13

L Acquisition and

 U

129

learning an L2 which does. Such is the case with Chinese, Japanese, or Korean

learners of English. In these languages, wh- questions are formed without wh- move-

ment. The

 wh-

 word stays in situ in its deep structure position, rather than moving to

the front of the sentence as it would in English. In the Japanese sentences in (10), the

wh- word d re stays in the direct object  DO) position:

4

(10) a.  John-wa dare-o korositaka

John-TOP who-DO killed Q particle

'Who did John kill?'

b.  John-wa Mary -ga dare-o kiratte-iru to sinzite-ita ka?

John-TOP Mary-particle who-DO hating-is that believing-was Q particle

'Who did John believe that Mary hated?'

As can be seen in (10b), wh-  words are not extracted from embedded clauses to

form questions. It is precisely in the case of extraction that Subjacency becomes

relevant in languages like English, preventing certain kinds of extractions and allow-

ing others. If Japanese-speaking learners of English successfully acquire knowledge

of restrictions on wh- movement out of embedded structures in the L2, this will not be

attributable to their LI.

Are the restrictions inducible from the L2 input? It was argued earlier that the

input to LI acquirers underdetermines properties of  wh- movement. This same un-

derdetermination holds true in the L2 acquisition context, unless L2 learners are

explicitly taught these rather obscure aspects of

  wh-

  movement in the classroom,

which seems implausible. While L2 learners will presumably both hear wh- structures

in naturalistic input and be taught certain aspects of question formation (especially

relating to simple clauses), it seems highly unlikely that they will be taught the

difference between grammatical cases of extraction from embedded clauses such as

(5),  and ungrammatical cases such as (4). Thus, the L2 input (both naturalistic and

classroom input) underdetermines restrictions on

 wh-

 movement.

Two recent studies have made use of the differences between English and lan-

guages such as Korean, Japanese, and Chinese to investigate the availability of

Subjacency to L2 learners. Schachter (1988a) studied 21 adult native speakers of

Korean and 18 native speakers of Dutch; both groups were advanced ESL learners.

5

Korean, like Japanese, forms wh- questions in situ, as in (10). In consequence, Korean

speakers will not have had any exposure to Subjacency in the LI. Dutch, on the other

hand, is very similar to English with respect to wh- movement and Subjacency. Thus,

if Dutch speakers show evidence of observing Subjacency in English, this might be

due either to the operation of UG or to knowledge of Subjacency gleaned from the

LI. Korean speakers, then, provide the test case. If UG is available, they should

unconsciously know that Subjacency applies as soon as they discover that English has

wh-

  movement; they should not make errors involving Subjacency violations when

they form questions, and they should be able to recognize Subjacency violations as

ungrammatical. In contrast, if UG is actually dead or only accessible in limited form

via the LI, such learners should not be able to work out the restrictions on

 wh-

movement in the L2. In that case, only the Dutch learners should be able to recognize

Subjacency violations, since their LI provides them with knowledge of the relevant

principle to fall back on.

Page 10: Second Language Aquisition and Universal Grammar

8/10/2019 Second Language Aquisition and Universal Grammar

http://slidepdf.com/reader/full/second-language-aquisition-and-universal-grammar 10/13

130 Lydia White

Schachter tested subjects with a grammaticality judgment task, including ungram-

matical Subjacency violations and grammatical sentences of equivalent complexity.

She found that the native speakers of Korean were accurate in judging the grammati-

cal sentences, but many of them failed to reject Subjacency violations and appeared

to be operating at chance. In contrast, the Dutch-speaking subjects performed with

considerable accuracy, accepting the grammatical sentences and rejecting the un-

grammatical Subjacency violations, just like her native-speaker control group. These

results suggest that where a principle operates in a similar fashion in both the LI and

L2,  it can still be accessed by the L2 learner. Where the LI does not exemplify a

principle, it does not appear to be accessible in the L2, explaining the poor perfor-

mance of the Korean speakers. In other words, these results support claims of the

essential unavailability of UG or its limited access only.

6

  However, Grimshaw and

Rosen (in press), commenting on results from experiments on the Binding Theory in

LI acquisition, point out that if learners are not constrained by some principle of UG,

they should treat grammatical and ungrammatical sentences alike. If they treat UG

violations and nonviolations differently, this suggests that they must know the princi-

ple in question even though they do not always obey it in an experimental task. Thus,

Schachter's finding of accurate performance on grammatical sentences and chance

performance on ungramm atical ones in fact suggests that the two sentence types are

distinguished by L2 learners, and hence that Subjacency may be part of their uncon-

scious knowledge, although it may not be consistently obeyed.

In contrast to Schachter's results are those obtained by Bley-Vroman, Felix, and

Ioup (1988). Once again, Subjacency was investigated, with English as the L2 and

Korean as the LI. Subjects were 92 adult Korean speakers, at an advanced level of

English proficiency. The test was a grammaticality judgment task, using

  wh-

 move-

ment sentences, approximately half of which were ungrammatical. Although subjects

did not achieve the accuracy of native speakers, they correctly rejected Subjacency

violations at a considerably higher than chance level. This finding contrasts with

Schachter's results and suggests that

 UG

 must after all be active. If language learners

fail to perform perfectly but nevertheless show greater-than-chance knowledge of a

principle of UG, and if there is no other way they could have arrived at this knowl-

edge, one must assume that the principle is in fact available and that other factors

sometimes lead learners to override the principles in performance (Grimshaw &

Rosen, in press).

On the basis of these two studies, the L2 acquisition status of Subjacency in

particular and UG in general remains unresolved. Schachter's results suggest that

Subjacency is inaccessible to learners whose mother tongue does not make use of

wh-

  movement. The results of Bley-Vroman et al. suggest that learners do attain

knowledge which is not attributable to the LI. Clearly, we need to isolate the source

of the differences in such results, and the question of UG availability must be pursued

with rigorous investigation of additional principles of UG.

Experimental Evidence for Parameter Resetting

Neither of these studies was concerned with the param eterization of bounding nodes

for Subjacency. Looking at parameters provides another way of investigating the

Page 11: Second Language Aquisition and Universal Grammar

8/10/2019 Second Language Aquisition and Universal Grammar

http://slidepdf.com/reader/full/second-language-aquisition-and-universal-grammar 11/13

L Acquisition

  nd

 UG  131

accessibility of UG to L2 learners. Again, in order to eliminate the effects of the LI,

the conditions in (9), repea ted here as (11), must hold.

(11) a. The LI and L2 have different values for some parameter, and

b.

  the input underdetermines the L2 grammar.

If the L2 value of a param eter is acquired under these conditions, the accessibility of

UG is favored. In contrast, if the LI value is adopted, the situation is not so clear, since

this may be due to the unavailability of UG, to the fact that learners are in a stage

prior to resetting, or to a complex interplay of other factors, including the lack of

positive input in the L2 to lead to param eter resetting (cf. W hite, 1989a).

White (1988a) investigated whether native speakers of French acquire knowledge

of the bounding status of S in English. As discussed earlier, French has different

bounding nodes from English. The LI, therefore, cannot give precise evidence as to

what kinds of  wh-  extractions are possible in the L2. The positive L2 input is not

transparent on this point either, since English allows a wide range of

 wh-

 movement

structures and the input does not give sufficient cues as to what extractions are not

possible. Once again, then, we have a situation where the L2 input underdetermines

the L2 grammar, and the LI grammar cannot provide the required abstract knowl-

edge in the relevant form.

White's study included two groups of adult learners of English (one low intermedi-

ate and one high intermediate), who were asked to perform a number of tasks,

including a grammaticality judgment task containing ungrammatical Subjacency vio-

lations. Some sentences were ungrammatical in both French and English, involving

extractions from complex noun phrases, as in (4b), whereas others involved extrac-

tions from wh- islands, which were ungrammatical in English but not in French. Both

groups showed considerable accuracy on complex noun phrase violations sentences.

In these sentences, parameter resetting is not at issue; they are ungrammatical in

both languages, and the subjects' treatment of these sentences might have been due

to LI knowledge or to UG. However, in the case of

  wh-

 island violations, where the

bounding status of S is at issue, there is a significant difference between the two

experimental groups; the low intermediate group failed to reject structures where a

wh-

 word had been moved out of a

  wh-

 island. These results suggest that English is

being treated like French in not having S as a bounding node; namely, the LI value of

the parameter for bounding nodes for Subjacency has been adopted. Subjects in the

high intermediate group, on the other hand, accurately rejected these sentences,

suggesting that they had reset this parameter and were treating S as a bounding node

in the L2. If so, the accessibility of UG is supported, given that appropriate informa-

tion about the bounding status of S is not available in the LI, and not easily inducible

from the L2 input alone.

IMPLICATIONS O F UG BASED RESEARCH

The question of

 UG

 accessibility in L2 acquisition is still unresolved. However, at the

very least, the studies discussed in this article indicate that there is accessibility via

the LI. While much promising recent research suggests that principles and parame-

Page 12: Second Language Aquisition and Universal Grammar

8/10/2019 Second Language Aquisition and Universal Grammar

http://slidepdf.com/reader/full/second-language-aquisition-and-universal-grammar 12/13

132 Lydia White

ters of UG are available (e.g., Flynn & O'Neil, 1988; Pankhurst, Sharwood Smith, &

Van Buren, 1988; and see W hite, 1989b, for general discussion), others argue that this

is not the case (for example, Clahsen, 1990). Although the main issue is still being

researched, UG-based approaches to L2 acquisition have a number of important

implications which hold true regardless of the final outcome. Generative grammar

holds that linguistic competence is essentially modular (Fodor, 1983). The UG per-

spective assumes the correctness of this claim for L2 acquisition as well. Second

language acquisition is not a unitary phenomenon, and it is unrealistic to expect

there to be one paradigm that will be able to em brace the whole field. Claims must be

made and tested within subdomains, of which the theory of UG is one . What the UG

perspective offers is the means to identify abstract properties of language, to study

certain aspects of L2 competence in depth, and to reach a greater understanding of

precisely what formal properties L2 learners unconsciously internalize (or fail to

internalize). Furthermore, this perspective suggests the need for sophisticated experi-

mental techniques, since L2 learners' use of language often conceals phenomena that

are not available to simple observation.

 UG

  certainly cannot provide an explanation

of everything that goes on in L2 acquisition and does not claim to do so. It does not

even account for all universal phenomena. For example, universal processing strate-

gies are different from, but certainly not incompatible with, principles of Universal

Grammar. The relevance of UG is strictly limited to providing a potential explanation

of the acquisition of rather formal aspects of language structure. Within this domain,

UG provides a suitable framework or paradigm from which to address issues of

importance within second language acquisition, a framework which gives new in-

sights and suggests new lines of research.

NOTES

1. Th e unavailability of negative evidenc e in LI acquisition has recen tly been questioned (e.g., Bohannon

& Stanowicz, 1988; Hirsh-Pasek, Treiman, & Schneiderman, 1984). However, the error types that parents

supposedly respond to do not involve the kinds of phe nom ena that UG seeks to explain. See White (1989b) for

a more detailed discussion of the negative evidence issue.

2. S' differs from S in that it includes a position for com ple m ent ize rs suc h as

  that

  an d

  whether.

  As

illustrated by the bracketing in (6a) and (6b), an S is usually contained within an S'. In current theory

(Chomsky, 1986a), S' is known as CP (complem entizer phrase) and S as IP (inflection phrase).

3.

 A

 wh-

  island has a

  wh-

  phrase in the lower COMP, rathe r tha n a

  non-wh-

  complem entizer l ike

 thai.

  In

(4a), whether  occupies the lower COMP position.

4.  Japanese is an SOV language. Direct objects and clauses which are complements to the verb precede

the verb.

5. In addition, Scha chter looks at native speak ers of othe r L ls. Since the Subjacency issue is more

com plex in thes e cases, I will not discuss them here . Her results are also reported in Schach ter (1989), but

without those of the Dutch speakers.

6. See Martoha rdjono and Gair (1989) for an alte rnative analysis which explains Sc hachter's results

without having to assume the nonavailability of UG.

REFERENCES

Bley-Vroman, R. (1989). What is the logical problem of second language learning? In S. Gass & J. Schachter

(Eds.), Linguistic perspectives on second language acquisition  (pp. 41-68). Cambridge: Cambridge Univer-

sity Press.

Bley-Vroman, R., Felix, S., & loup, G. (1988). The accessibility of universal grammar in adult language

learning.  Second Language Research, 4,  1-32.

Page 13: Second Language Aquisition and Universal Grammar

8/10/2019 Second Language Aquisition and Universal Grammar

http://slidepdf.com/reader/full/second-language-aquisition-and-universal-grammar 13/13

L Acquisition and U 133

Bohannon, J. N., & Stanowicz, L. (1988). The issue of negative evidence: Adult responses to children's

language errors.

 Developmental Psychology 24

684-689.

Brown, R., & Hanlon, C. (1970). Derivational complexity and the order of acquisition in child speech. In J. R.

Hayes (Ed.),

 Cognition and the development of language

  (pp.  11-53 ). New York: Wiley.

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

Chomsky, N. (1981b). Principles and parameters in syntactic theory. In N. Hornstein & D. Lightfoot (Eds.),

Explanation in linguistics: The logical problem of language acquisition

 (pp. 32-75 ). London: Longman.

Chomsky, N . (1986a). Barriers.  Cam bridge, MA: MIT Press.

Chomsky, N. (1986b).

 Knowledge of language: Its nature origin and use.

 New York: Praeger.

Clahsen, H. (1990). Th e com parative study of first an d secon d la nguag e de velopm ent.

  Studies in Second

Language Acquisition 12

135-153.

Clahsen, H., & Muysken, P. (1986). The availability of universal gram ma r to adult and child learn ers: A study of

the acquisition of German word order.  Second Language Research, 2,  93-119.

Ellis,

 R. (1985). Understanding second language acqu isition.  Oxford: Oxford University Press.

Flynn, S. (1987).

 A parameter-setting model o fL2 acquisition.

 Dordrecht: Reidel.

Flynn, S., & O'Neil, W. (Eds.) (1988). Linguistic theory in second language acquisition.  Dordrecht: Kluwer.

Fodor, J. A. (1983).

  The modularity of mind.

 C ambridge , MA: MIT Press.

Grimshaw, J., & Rosen, S. T. (in press). Knowledge and obedience: The developmental status of the binding

theory. Linguistic Inquiry.

Hirsh-Pasek, K., Treiman, R., & Schne iderm an, M. (1984). Brown and Hanlon revisited: M others' sensitivity to

ungrammatical forms.

 Journal of  hild Language 11

81-88.

Lightfoot, D. (1982). The language lottery: Tow ard a biology of grammars.  Cambridge, MA: MIT Press.

Martohardjono, G., & Gair, J. (1989).

 Apparent UG inaccessibility in  SLA: Misapplied principles or principled

misapplications?  Paper presented at the Conference on the Interaction of Linguistics, Second Language

Acquisition and Speech Pathology, University of Wisconsin, Milwaukee.

McLaughlin, B. (1987).

  Theories of second language learning.

  London: Edward Arnold.

Otsu, Y. (1981).

  Universal grammar and syntactic development in children: Touiard a theory of syntactic

development.  Unpublished doctoral dissertation, MIT.

Pankhurst, J., Sharwood Smith, M., & Van Buren, P. (Eds.). (1988).  Learnability and second languages: A book

of readings. Dordrecht: Foris.

Pinker, S. (1984).  Language learnability and language development.  Cam bridge, MA: Harvard U niversity

Press.

Rizzi, L. (1982). Issues in Italian syntax.  Dordrech t: Foris.

Rutherford, W. (1988).

  Questions of learnability in second language acquisition.

  Paper presented at the 13th

Annual Boston University Conference on Language Development.

Schachter,

 J

(1988a).

 On the issue of completeness in second language acquisition.

 Paper presented at the 13th

Annual Boston University Conference on Language Development.

Schachter, J. (1988b). Second langu age acqu isition and its relationship to U niversal Gram mar.  Applied Linguis-

tics, 9,

 219-235.

Schachter, J. (1989). Testing a proposed universal. In S. Gass & J. Schachter (Eds.), Linguistic perspectives on

second language acquisition  (pp. 73-88). Cambridge: Cambridge University Press.

Sportiche, D. (1981). Bounding n odes in French.

 Linguistic Review, 1,

 219-246.

Wexler, K., & Culicover, P. (1980). Formal principles of language acquisition.  Cam bridge, MA: MIT Press.

White, L. (1985). Is the re a logical problem of second langu age acquisition?  TESL Canada, 2(2),

 29- 41 .

White, L. (1988a). Island effects in second language acquisition. In S. Flynn &  W. O'Neil (Eds.), Linguistic theory

in second language acquisition  (pp. 144-172). Dordre cht: Kluwer.

White, L. (1988b). Universal grammar and language transfer. In J. Pankhurst, M. Sharwood Smith, & P. Van

Buren (Eds.), Learnability and second languages: A book of readings (pp. 36-61). Dordrecht: Foris.

White, L. (1989a). The principle of adjacency in second langua ge acquisition: Do L2 learners o bserve the

subset principle? In S. Gass & J. Schachter (Eds.),

 Linguistic perspectives on second language acquisition

(pp. 134-158). Cambridge: Cambridge University Press.

White, L. (1989b).

  Universal grammar and second language acquisition.

  Amsterdam: John Benjamins.

Zobl, H. (1983). Marke dness and the projection problem .  Language Learning, 33, 293 -313 .