Vivian Cook Writing System Topics Onlinee Writings
of paper in G. Brown, K. Malmkjaer & J. Williams (eds.), Performance and
Competence in Second Language Acquisition, 1996, CUP, 57-69
of the long-standing disputes in linguistics concerns what it is about. De
Saussure remarked, 'Other sciences work with objects that are given in advance
and that can then be considered from different viewpoints; but not linguistics'
(de Saussure, trans. 1959, p.8). The chief bone of contention has been whether
language consists of behaviour or of knowledge. Should we study large samples of
speech – performance? Should we
study the contents of individual minds – competence?
This dispute has had many battles and many battlefields; it is waged as
much in language teaching, in children's language, or in computational
linguistics, as it is in linguistics itself; for example the usual linguists'
criticisms of connectionism are, not that it is wrong, but that it is not about
language as conceived of by linguists (Pinker and Prince, 1988).
contemporary Universal Grammar (UG) theory assumes that the knowledge of
language is all; language use depends on knowledge. Hence UG theory sets an
agenda for linguistics with three main questions, as formulated by Chomsky
(i) what constitutes knowledge of language?
that is to say, what is the state of the mind of a person who knows language? What does a person who knows English actually know, what is present in their minds?
(ii) how is knowledge of language acquired?
that is to say, how does this knowledge of language come into being? What we know is constrained and facilitated by the processes of acquisition; describing what someone knows means also describing how they set about acquiring it.
(iii) how is knowledge of language put to use?
that is to say, how does the speaker of a language put this knowledge to use? Partly this means studying the psychological processes of performance, partly the pragmatic competence which enables the person to use the language for communicating, for planning, for remembering, for declaring war, for passing sentence, and the myriad functions for which language may be used.
UG model therefore starts from linguistic competence – the knowledge of
language in the speaker's mind. This 'knowledge' question is logically prior to
the questions of acquisition or of use. The first need is to establish what this
linguistic competence consists of. UG has proposed in recent years that such
knowledge consists of principles and parameters – principles that are
universal to human language, parameters that have different values in different
languages. Principles are constraints on possible structures that function
across all the rules of the language at an abstract level. A familiar example is
the principle of structure-dependency which claims that 'all known formal
operations in the grammar of English, or of any other language, are
structure-dependent' (Chomsky, 1972, p.30). For example passive sentences are
arrived at by moving constituents about in the sentence:
John drove the car.
The car was driven by John.
by, inter alia, moving the Noun Phrase the car from object to subject position. But it is the whole constituent the car in the direct object position that moves, not a particular word in the sentence. The same applies to movement in questions such as:
Will John drive the car?
where it is the auxiliary constituent in the main VP that moves, not any particular word, and to the interpretation of himself in:
John said Paul drove himself.
where the equivalence of reference between Paul and himself depends upon our knowledge of which is the subordinate clause in the sentence. All of these depend on knowledge of structure.
familiar example of a parameter is the pro-drop parameter. Some languages must
have a subject as in German:
Er spricht. (he speaks)
Il parle. (he speaks)
Other languages need not have a subject in declarative sentences, for instance Italian:
Hence one of the parameters within the knowledge of language is whether a sentence must have subjects or need not have them; any language has one or other of the values for the pro-drop parameter.
UG has answered the questions for linguistics by stating that knowledge consists
of principles and parameters; this is not to say, of course, that linguistic
competence does not have many other aspects as well as this core, in particular
vocabulary. The questions for linguistics can now be rephrased as:
(i) what are the principles and parameters that people know?
(ii) how is knowledge of principles and parameters acquired?
(iii) how is knowledge of principles and parameters put to use?
apply this to Second Language Acquisition, the starting point is question (i):
do people who know L2s also know principles and parameters?
One problem is going to be deciding what counts as evidence for their
knowledge. Linguistics typically depends for the description of the L1 on what
has been called single-sentence evidence
(Cook, 1993): a single sentence such as:
John likes Paul.
is clearly a sentence of English and noone would deny it. There is no need to go out and record someone actually saying it or to ask a random sample of English people whether they agree it is acceptable. For many of the obvious phenomena of the language, L1 description does not need to do more than cite such single sentences. If someone denies the sentence, one may have to think again; but one can hardly imagine an English-speaking person who could deny that:
John likes Paul.
Second Language Acquisition there is a problem with such single-sentence
evidence; L2 learners seldom get to levels of language knowledge equivalent to
the L2 native; who could say whether
John likes Paul.
is part of their knowledge in an equivalent way? One possibility is to resort to Error Analysis of actual utterances L2 learners have produced; but this is performance and needs to be compared with native performance rather than native competence as manifest in single sentences. Another possibility is to ask L2 learners directly whether something is acceptable to them – a grammaticality judgments test. It may be that a grammaticality judgment test indeed taps their underlying competence; nevertheless it has to be remembered that the grammaticality judgment test is still a kind of performance indirectly linked to competence. Again it needs to be compared with native grammaticality judgments rather than native single sentence evidence for competence. The distinction between competence and performance comes into the methodology of Second Language Acquisition research in a way that it is not involved in the statement of L1 competence as there is no clear source to which we can appeal single sentences.
is still true that grammaticality judgment tests are the most effective way of
testing what learners know within the UG model. The one that I have been
developing for a while is the MUGtest (Multi-parameter Universal Grammar test).
The aim is to test for knowledge of several aspects of UG at the same time and
to compare the results across a large range of L2 learners with different L1s.
The MUGtest has a list of 96 sentences including inter alia structure-dependency
and pro-drop; the learners have to say whether each sentence is OK,
Not OK, or Not sure. So far this
test has been carried out with university students of English who speak
Japanese, Finnish, and Chinese as well as native speakers of English.
did the L2 learners make of structure-dependency?
Take sentences that break the principle of structure-dependency such as:
Is Joe is the dog that black?
100% of Finnish learners rejected this, compared with 93.3% of Japanese, and 86.6% of Finnish; a group of native speaker scored 100%. Structure-dependency violations were highly apparent to all of these L2 learners despite the fact that they had never encountered them before and would never have been taught about them.
learners were also tested for pro-drop through sentences that miss out the
subject, ungrammatical in English, such as:
Again the results show that learners know perfectly well that null subject sentences are ungrammatical in English, ranging from 64.2% for Chinese, 73.3% Japanese, and 92.8% Finnish, with English native speakers scoring 90%. Table 1 shows the figures scored as percentage correct for the sentences with structure-dependency violations and for equivalent declaratives and questions; similarly for null subject and for equivalent sentences with subjects.
1 structure-dependency and pro-drop effects in four groups of L2 learners
established provisionally that L2 competence consists of principles and
parameters, we can now pass on to the second question: how do L2 learners
acquire principles and parameters in the L2?
The conventional answer to this has been phrased in terms of access to UG
(Cook, 1985). The three alternatives are:
i) no access
In this case L2 learners do not use UG in learning the L2. This position has been argued in similar ways by Schachter (1989), Bley-Vroman (1989), and Clahsen and Muysken (1986): L2 learning is different from L1 learning in that the learner cannot make use of the UG principles and parameters that the native child employs. But, as we can see, the MUGtest does not show that L2 learners do not know principles and parameters at all; in fact they score reasonably well on tests of structure-dependency and pro-drop. They do of course score less than native speakers. But nevertheless they clearly show that a part of their knowledge can be expressed in principles and parameters form, as can that of natives. Unless they can be shown to have learnt this by some means unavailable in L1 learning, say explanations of structure-dependency by the teacher, there seems no reason not to credit them, like the native, with access to UG.
ii) direct access
It might be that the L2 learner acquires the L2 as the child acquires the L1 through UG, or it might be that the L2 learner acquires it through the filter of the L1. Direct access is the idea that L2 learners will have UG directly available to them. How can this be tested in L2 learning? Suppose that a learner encounters something in an L2 that is not in their L1; only if UG were still available would they be able to acquire this. The testcase is an L2 learner whose L1 does not have some feature of UG present in the L2. For example Japanese does not form questions through syntactic movement: the questioned item remains in its place in the sentence:
Niwa wa doko desu ka?
(garden where is)
Where is the garden
so far as syntactic movement is concerned, though not in other areas, Japanese
does not need structure-dependency;
the occasion never arises on which the speaker needs a knowledge of the
structure of the sentence to move parts of it about. Japanese learners of
English cannot depend on their L1; they must have access to something else. As
we have seen, they do in fact spot structure-dependency violations quite easily,
93.3% in fact. Wherever this knowledge comes from it is not Japanese; perhaps it
came from some odd form of language teaching; perhaps however it has the same
source as the native speaker, namely UG. The MUGtest (and indeed other work with
Japanese speakers and structure-dependency by Otsu and Naoi (1986)) suggests
that L2 learners have direct access, at least for this aspect of L2 knowledge.
iii) indirect access
The third possibility is that L2 acquisition depends upon the knowledge of the L1; once UG has been instantiated in one grammar, this dictates the form of subsequent grammars. So L2 learners may utilise their L1 knowledge of principles and parameters as a way-in to the L2. An example might be parameter settings. An L2 learner might start with a neutral setting, neither pro-drop nor non-pro-drop say. In this case all L2 learners would learn the L2 in the same fashion regardless of L1. Or an L2 learner might start by transferring the L2 parameter setting to the L2 and gradually have to learn that it was wrong. In this case there are going to be differences between L2 learners according to their L1s. The MUGtest results for pro-drop can test this. Japanese and Chinese are pro-drop languages without the need for subjects; Japanese learners of English score 73.3% on sentences like:
while Chinese learners score 64.2%, Finnish on the other hand is a non-pro-drop language in which subjects are necessary apart from a few set expressions. Finnish learners score 92.8% – even natives only score 90%. Of course it is not possible to equate the learners totally in knowledge of English apart from the fact that they are university level in their respective systems; nevertheless the presence of subjects in English sentences is something which is in a sense taught from very early on and hardly seems an advanced problem. So learners from pro-drop L1s have different scores in English from learners with non-pro-drop L1s. This would seem a clear case where knowledge of the L2 has been mediated through a knowledge of the L1. L2 learners are behaving as if they had indirect access to UG.
have so far simply translated the three questions for linguistics from
monolingual competence to L2 competence. Is this enough?
As we have already seen, there are problems in defining the ultimate goal
of second language acquisition: L1 acquisition leads to linguistic competence in
virtually all human beings; linguistic competence is defined as whatever it is
that native speakers of a language know, nothing more, nothing less. However, L2
acquisition leads to something that approaches native competence in a small
fraction of those who undertake it. Yet this is indeed the goal assumed by most
L2 researchers: L2 learners are aiming at native competence, as we see in
quotations such as 'adults usually fail to become native speakers' (Felix, 1987,
p.140), or 'those adults who seem to achieve native speaker
"competence"' (Selinker, 1972). It is true that since the independent
grammars assumption came into Second Language Acquisition from first language
acquisition, as described in Cook (1993), it has been an axiom that L2 learners
have grammars of their own – 'interlanguages' as this was labelled in Selinker
(1972). But the interlanguage assumption did not seem to alter the destination
that L2 learners were believed to be heading towards; the successful ones become
native-like: 'second language learner language approximates native speaker
language' (Spolsky, 1989, p.35). Native or native-like competence was still the
final goal even if noone ever reached it. Hence the vast bulk of the Second
Language Acquisition research literature talks in terms of L2 learners' failure
to reach native competence, as we see in quotations such as 'Unfortunately,
language mastery is not often the outcome of SLA' (Larsen Freeman & Long,
1991, p.153) where 'language mastery' can only be taken to refer to native
competence or 'failure to acquire the target language grammar is typical'
(Birdsong, 1992, p.706) where 'target language grammar' presumably equates to
the one thing that L2 learners cannot do by
definition is become native speakers – fully balanced bilinguals perhaps,
natives never. The goal of L2 acquisition should be seen as something other than
monolingual native competence. The term ‘multi-competence’ has been
introduced to cover knowledge of more than one language in the same mind (Cook,
1991). There is no assumption that this knowledge corresponds to a monolingual
native speaker's in either L1 or L2; this is a matter for empirical research
rather than the basis for Second Language Acquisition research. The starting
point should be what L2 learners are like in their own right rather than how
they fail to reach standards set by people that they are not by definition.
Multi-competence is then a necessary basis for Second Language Acquisition
research. L2 learners are not failed monolinguals but people in their own right.
This is already the accepted approach in bilingualism studies: to take two
quotations, 'it is clear that a reasonable account of bilingualism cannot be
based on a theory which assumes monolingual competence as its frame of
reference' (Romaine, 1989); 'For the vast majority of bilinguals,
"bilingual competence" is not measurable in terms of monolingual
standards' (Hoffmann, 1991). It is only in the discipline of Second Language
Acquisition research, and perhaps language teaching as well, that it has been
assumed that L1 monolingualism is an acceptable standard for L2 learners, as
implied in such statements as in 'The lack of general guaranteed success is the
most striking characteristic of adult foreign language learning. Normal children
inevitably achieve perfect mastery of the language; adult foreign language
learners do not' (Bley-Vroman, 1989, p.42).
us rephrase the three questions in terms of multi-competence:
(i) what is multi-competence?
(ii) how is multi-competence acquired?
(iii) how is multi-competence put to use?
Let us sketch some answers to the first two questions in turn. These are based on work in progress, partly already published, namely Cook (1991; 1992; 1993; 1994). Question (iii) involves issues of performanc and sociolinguistics that go outside the scope of this contribution.
(i) what is multi-competence?
starting point for Second Language Acquisition is the knowledge of more than one
language in the same mind. It cannot be taken for granted that this is the same
as two L1s added together, even if this is a possibility; rather it will be
presumed that our task is to look at the whole system of both L1 and L2 in the
same mind. The first question is whether the two language systems in the mind
are the same as those of monolinguals. What little evidence there is for this is
reviewed in Cook (1992). It suggests that some of the L1 knowledge is different
from monolinguals, for example in the case of Voice Onset Time; additional
evidence from the MUGtest contrasts English-speaking people who know French with
those who do not, as we see in Table 2.
2 pro-drop in native speakers of English with and without French
from these results English people who know French are much more tolerant of null
subject-sentences in English than English people who do not know French. This
result is very surprising in that French is a non-pro-drop language like
English! Perhaps later data will
show it is some kind of artefact. Nevertheless in general the point is true:
people who know L2s do not necessarily have the identical knowledge to
monolinguals. The knowledge of the L2 may also be different. Research by
Coppetiers (1987) showed that American/French bilinguals had different
perceptions of grammaticality in the L2 than monolinguals; other research
however suggests that the knowledge may be more similar.
further question is whether multi-competence should be described as two separate
systems or as a single system, again surveyed in Cook (1992). The lexicon in
particular appears to be a merged system, according to a variety of
psycholinguistic research (Caramazza & Brones, 1979; Beauvillain &
Grainger, 1987); codeswitching requires, if not a single system, tight
interconnections for switching from one language to another; L1 knowledge cannot
be cut off while the L2 is being processed (Altenberg & Cairns, 1983; Blair
& Harris, 1981); the two languages appear to be stored in the same areas of
the brain (Paradis, 1989).
ii) how is multi-competence acquired
acquisition question above was phrased chiefly in terms of access. It was argued
that the no-access position was invalidated by showing L2 learners know
principles and parameters; direct access was substantiated by use of UG
principles in an L2 that are not found in the L1 and that indirect access was
revealed in the transfer of parameter settings were transferred from L1 to L2.
The main reason people have advanced for no-access is the L2 failure argument
attacked above. This clearly accepts the monolingual as the standard, not the
multicompetent speaker. The measure for access is taken to be whether something
is learnable from the environment or not – the poverty-of-the-stimulus
argument; if L2 users know aspects of language that they could not have learnt
from the input these must be contributed by the UG built-in to their minds; the
argument applies regardless of whether monolingual competence or
multi-competence is involved. Indeed it might well be that some aspects of UG
are revealed in an L2 that are not found in an L1; L1 native competence has no
privileged status for the poverty-of-the-stimulus argument.
main issue in the access argument that multi-competence raises is however:
access to what? The argument in
the L1 is that input is processed by the internal UG in the mind, erstwhile
known as the Language Acquisition Device (LAD); this produces a linguistic
competence couched in terms of principles and parameters. In the L2 acquisition
input either goes through this module (direct access) or it does not (no
access), or it relates to it via the L1 linguistic competence (indirect access).
All these possibilities distinguish the language product (linguistic competence)
from the acquisition module (UG). But this is an inadequate formulation of UG in
principles and parameters terms. Universal Grammar is the language faculty
itself, not something that the language faculty draws on: 'The language faculty
has an initial state, genetically determined; in the normal course of
development it passes through a series of states in early childhood, reaching a
relatively stable steady state that undergoes little subsequent change, apart
from the lexicon ... we call the theory of the state attained its grammar
and the theory of the initial state universal
grammar' (Chomsky & Lasnik, 1991, in press). UG is one state of the
mind; L1 linguistic competence is another. Or, as sometimes put in the
literature, UG + input = linguistic competence (Atkinson, 1992). Again to come
back to basics, linguistic competence is knowledge of language – a state of
the language faculty in a human mind. It is not that UG clones itself and
produces an L1 grammar complete with all the principles and parameters of UG but
with the form of one language; rather it is UG itself, alias the language
faculty, that transforms itself. This becomes particularly crucial to
multi-competence. If UG is cloned every time a language is learnt there will be
as many copies as there are languages; polyglots will have large numbers of
grammars knocking round in their minds. If, however, we see UG as the language
faculty, the question is how a single language faculty contains more than one
setting for each parameter etc. This argument is developed in Cook (1994).
have reached a powerful and controversial punch-line: if multi-competence
involves knowing principles and parameters simultaneously in different forms and
with different settings, does this not feed back in to our description of L1
linguistic competence? Any human
being has the potential to know more than one language; it has often been
claimed that monolinguals are a minority in global terms. If our description of
L1 acquisition commits us to a single rigid form of UG in which all the
principles have been instantiated and the parameters set with no possibility for
change, then we are ignoring this. The description of L1 linguistic competence
is not explanatorily adequate if it fails to recognise the normal capability of
the human mind to know two languages (not like monolingual native speakers of
course but in their own unique multi-competent way). As Stenson (1990, p.194)
puts it, 'Any grammatical theory that purports to account for human linguistic
competence must also be able to account for bilingual competence and the
associated performance.' The study
of first language acquisition is never going to be enough; in a sense it can be
considered a deprived form of acquisition with impoverished input. First
language acquisition may indeed lead to a language system that differs from
multi-competence; but that is its loss
not its gain. The argument is that the description of linguistic competence has
been misleadingly based on monolinguals, like a description of juggling based on
a person who can throw one ball in the air and catch it, rather than on a
description of a person who can handle two or more balls at the same time.
Calling the knowledge of a person who knows one language linguistic competence
may be as misleading as calling throwing one ball in the air juggling. To sum
up, the description of the competence of people who know more than one language
has repercussions for the study of people who know only one. It seems an
inadequate excuse to say that this is simply normal idealisation of competence
when the form of language knowledge itself may have to be couched in a different
way. Rather than knowledge of two languages being a side-issue, we should agree
with Jakobson (1953), 'Bilingualism is for me the fundamental problem of
E.P., & Cairns, H.S. (1983), 'The effects of phonotactic constraints on
lexical processing in bilingual and monolingual
subjects', JVLVB, 22, 174-188
M. (1992), Children's Syntax, Blackwell, Oxford
C., & Grainger, J. (1987), 'Accessing interlexical homographs: some
limitations of a language-selective access', J. Mem. & Lang.,
D. (1992), 'Ultimate attainment in Second Language Acquisition', Language,
68, 4, 706-753
D., & Harris, R.J. (1981), 'A test of interlingual interaction in
comprehension by bilinguals', J. Psycholing. Res, 10, 4, 457-467
R.W. (1989), 'The logical problem of second language learning', in S. Gass &
J. Schachter (eds.), Linguistic perspectives on second language acquisition,
A, & Brones, I. (1980), 'Semantic classification by bilinguals', Canad.
J. Psychol., 34, 1, 77-81
N. (1972), Problems of Knowledge and Freedom, London, Fontana
N. (1986), Knowledge of Language: Its Nature, Origin and Use, New York,
N., & Lasnik, H. (in press), 'Principles and parameters theory', in J.
Jacobs, A. von Stechow, W. Sternefeld, W & T. Vennemann (eds.), Syntax:
An International Handbook of Contemporary Research, Berlin, de Gruyter.
H., & Muysken, P. (1986), 'The availability of universal grammar to adult
and child learners – a study of the acquisition of German word order', Second
Language Research, 2, 2, 93-119
V.J. (1985), 'Chomsky's Universal Grammar and second language learning', Applied
Linguistics, 6, 1-18
V.J. (1991), 'The poverty-of-the-stimulus argument and multi-competence', Second
Language Research, 7, 2, 103-117, 1991
V.J. (1992), 'Evidence for multi-competence', Language Learning, 42, 4,
V.J. (1993), Linguistics and Second Language Acquisition, Macmillan
V.J. (1994), 'The metaphor of access to Universal Grammar', in Ellis, N. (ed.), Implicit
Learning of Language, Academic Press
R. (1987), 'Competence differences between native and near-native speakers', Language,
63, 3, 545-573
Saussure, F. (1916), Cours de Linguistique Generale, ed by C. Bally, A.,
Sechehaye, & A. Reidlinger, Paris, Payot (trans. 1959), Course in General
Linguistics, Peter Owen, London
S.W. (1987), Cognition and Language Growth, Foris
C. (1991), An Introduction to Bilingualism, Longman, London
R. (1953), 'Results of the conference of anthropologists and linguists', IJAL
Supplement, Memoir 8, 19-22
D., & Long, M.H. (1991), An Introduction to Second Language Acquisition
Y., and Naoi, K. (1986), 'Structure-dependence in L2 acquisition', paper
presented at JACET, Keio University, Tokyo, September. Cited in White, L.
(1989), Universal Grammar and Second Language Acquisition, John Benjamins
M. (1989), 'La lateralizacion cerebral diferencial en los bilingues: !Basta, por
favor!', Investigaciones psicologicas, 7, 95-105
S., & Prince, A. (1988), 'On language and connectionism: analysis of a
parallel distributed processing model of language acquisition', Cognition,
S. (1989), Bilingualism, Blackwell, Oxford
J. (1989), 'Testing a proposed universal', in S. Gass & J. Schachter (eds.),
Linguistic perspectives on second language acquisition, CUP
L. (1972), 'Interlanguage', IRAL, X/3
B. (1989), Conditions for Second Language Learning, OUP
N. (1990), 'Phrase structure congruence, government, and Irish-English
code-switching', in Hendrick, R. (ed.), Syntax and Semantics Vol. 23 The
Syntax of the Modern Celtic Languages', San Diego, Academic, 169-199