Linguistic theory and second language acquisition: How not to lose sight of the wood for the trees

The version of generative linguistic theory with which this paper is concerned is the Chomskyan version. The tenn "theoiy of Universal Grammar", or "UG theoty" for . short, will be used to refer to this theory. The term "UG theory of L2 acquisition" fwUl be used to refer collectively to theories which adopt UG theory as a framework ' for the study of L2 acquisition. It should always be borne in mind that the Chomskyan version of UG theory is first and foremost a theory of native speaker (LI) knowledge. Chomsky himself never intended the theory to be used for describing L2 knowledge or for explaining its acquisition.


i.
The aim of L2 acquisition research is to understand how non-mother tongues are acquired.An important part of acquiring a language is acquiring its grammar.ii.
It is impossible to understand how the grammar of a language is acquired without understanding what knowledge of grammar is and how this knowledge is represented in the mind.iii.UG theory * provides us with a hypothesis about what knowledge of grammar is and how it is mentally represented, and * is the most well-developed theory of language currently available.iv.Using UG theory as a framework for the study of L2 acquisition, therefore, will provide insight into L2 acquisition.^ Compelling as this argument may seem, tbe abstractness, complexity and frequent revision of specific proposals about syntactic stnicture made within the framework of UG theory have provoked many L2 researchers to reject the theory out of hju^ as irrelevant to their concerns.It is not difficult to see why, considering that in current versions of generative syntax the simple sentence John kisses Mary is assumed to have the following stnicture:^ kisses Mary It does not seem unreasonable to conclude that a theory requiring this amount of descriptive apparatus to represent the structure of a simple sentence can have very little to say about L2 acquisition.Such a rejection of UG theory as irrelevant to LI acquisition can be argued to be premature, however.
There are two sides to the argument.In the first place, it can be argued that the rejection of UG theory for the reason outlined above reflects a failure to make an important distinction.The distinction in question is that between the basic tenets of UG theory on the one hand, and specific proposals about the content and organisation of grammatical knowledge, i.e. the descriptive apparatus of the theory, on the other hand.
. In section 2 it will be shown that a convincing case can be made for taking the basic tenets of UG theory as a point of departure for L2 acquisition research.It will be argued in section 3 that, by using the descriptive apparatus of current versions of UG theory, L2 researchers have been able to provide (i) much more precise descriptions of the problems facing L2 learners, as well as (ii) principled explanations of problematic L2 phenomena.Also, as will be shown in section 4, jecent insights into the way in which the sentences of himian languages are structured have cast new light on some of the most persistent problems of L2 acquisition research, making it possible for researchers to suggest interesting answers to old questions.

Basic tenets of UG tlieory
It was noted above that the rejection of UG theory as irrelevant to L2 acquisition research often reflects a failure to distinguish between the basic tenets of the theory on the one hand, and its descriptive apparatus on the other hand.The aim of this section is to consider two reasons why the basic tenets of UG theory should not be rejected out of hand as a framework for L2 acquisition research.These reasons are that i. unlike the descriptive apparatus of the theory, the basic tenets of UG theory are extremely simple and have not changed since they were first articulated by Chomsky in the late 1950s, and ii.assuming these basic tenets as a point of departure for L2 research is a highly valued option in terms of considerations of conceptual coherence and theoretical simplicity.
The basic tenets of UG theory may be summarized as follows: i.
Humans have a special-purpose, species-specific genetic endowment for language, a Universal Grammar.ii.
Universal Grammar consists in unconscious linguistic knowledge which allows human children to discover the grammar of any language to which they are exposed.iii.A native speaker's attained knowledge of the grammar of his or her language is a mental construct, i.e. it is represented in the speaker's mind.iv.Knowledge of granraiar is fundamentally different from any other kind of knowledge, but interacts with other kinds of linguistic and nonlinguistic knowledge and capacities in actual language use, i.e. in the interpretation and judgment of utterances.

V.
The acquisitional mechanism(s) responsible for converting linguistic ' into fcnowiedge of grammar are specific to language as well, i.e. they ait^ used in the acquisition of any other kind of knowledge.3 These basic tenets of UG theory have not changed over the years.

i.
UG theory is the only theory that offers the beginnings of an explanation of the knowledge that underlies one type of linguistic behaviour, namely LI behaviour.ii.
L2 behaviour, too, is a type of linguistic behaviour.iii.Given i. and ii., the assumption that UG theory can (partly) explain the knowledge underlying L2 behaviour is the more highly valued assumption in terms of considerations of theoretical simplicity.
f the knowledge underlying L2 behaviour is assumed to be UG-based, it follows hat the acquisition of this knowledge, too, must be UG-based.Schwartz's irgument, as outlined in (4), is therefore an explicit argument for adopting UG heory as a framework for the study of L2 acquisition.The crucial distinctions responsible for the word order differences between French and English on the one hand, and between different classes of verbs on the other hand, are the distinctions finite vs nonfmite and main verb vs auxiliary verb.The distinction between finiteness and nonfmiteness, whenever it is overtly expressed in a language, is typically expressed by inflectional morphemes, i.e. by functional elements (as opposed to lexical elements)®.Likewise, the crucial difference between auxiliaries and main verbs is that auxiliaries are fiinctional elements, while main verbs are lexical elements.
It is a characteristic of lexical verbs that they enter into role relationships with the noun phrases in a sentence.For example, in the sentence John kisses Mary the lexical verb kiss describes an event with two participants, John (the "kisser") and Mary (the "kissed").It is assumed that lexical verbs originate in the VP along with the NPs with which they enter into role relationships,^® The VP therefore contains the elements necessary for establishing the conceptual meaning of the sentence.
Functional elements (such as tense and agreement morphemes) do not contribute to the meaning of the sentence; that is, they do not affect the role relationships in the sentence.Moreover, they need not be affixed to the lexical verb, as sentences (6c) and ( 11)-( 13) above clearly show.It is therefore assumed that functional properties of a sentence, such as tense and agreement, are associated with functional categories, represented in sentence structures such as (5) by a separate set of nodes which occur outside the VP.However, the tense and agreement features of a sentence must ultimately be speiiej out on a verb, be it a main verb or an auxiliaiy.The simplest case, for various theoretical reasons which do not concern us here, is for the verb to move out of the VP and into the relevant functional node where it picks up the features associated with that node.This is what happens in French.The finite verb raises to TP (via AGRP about which more will be said soon) to receive tense and agreement features This explains the French facts in ( 6)-( 8) and ( 11)-( 13).This explanation does not hold for English, however.In English, main verbs remain in the VP and only auxiliaries can raise to TP, as is clear from a comparison of the English sentences in ( 6)-( 8) with those in (11)-(13).Pollock (1989: par.4) relates this difference to another difference between French and English; French has a much richer inflectional system than English.This, according to Pollock, means diat English and French have different values for a parameter which, for ease of reference, we shall call the "Agreement (AGR) parameter": French has strong AGR and English has weak AGR.Crucially, having strong AGR means that the verb need not be in the VP with its complement in order to mark that complement as bearing a certain thematic relation to it (e.g. the relation "the one bemg kissed" in the sentence John kisses Mary).That is, languages with strong AGR, are rich enough morphologically to permit transmission of the thematic roles of a verb that has moved out of the VP, so that these roles can be assigned to the relevant NPs in the VP.
In a language with weak AGR, the verb cannot transmit its theta roles from outside the VP.In English, therefore, a lexical verb cannot raise out of VP to receive tense and agreement features.Auxiliaries, by virtue of not entering into thematic role relationships with any other constituents in a sentence, are free to occur outside the VP, which explains the freedom of distribution displayed by the English auxiliary verbs in (11H15).
Havmg considered a possible explanation for the presence of at least one functional node, TP, in the structure (5), we still need an answer to the question why the structure needs to have so many additional functional nodes.There is a simple empirical reason for this.The functional nodes, apart from carrying features such as tense and agreement, which determine the functional properties of sentences, are required as landing sites for verbs which are moved out of VP.It is left to the reader to check that every one of the functional nodes in the structure (5) is required as a landing site if all the possible word orders in the sentences (6)-( 15 The second outcome of White's analysis is that a principled answer can now be given to the question why French-speaking learners of English, despite allowing finite lexical verbs to raise past adverbs in English, do not allow raising past the negative and even on to the sentence-initial position.This seemingly inexplicable phenomenon is in fact predicted by White's analysis, given that it links the movement possibilities of finite verbs in English to those of nonfinite verbs in French.As nonfinite verbs in French are not allowed to move beyond the position immediately to the left of the adverb, it is predicted that the movement of finite verbs in L2 English will be similarly restricted.
The third outcome of White's analysis is that, on this analysis, L2 learners' problems with word order are related in an insightful way to a superficially unrelated difference between the LI and the L2, namely a difference in the extent to which tense and agreement features are morphologically expressed in the two languages.That is, her analysis provides "deep" insight into the problems facing the L2 learners concerned by showing that superficially unrelated properties of their L2 are related at an abstract level.We shall elaborate on this relationship in section 4 below.
These outcomes can all be attributed to a particular assumption underlying White's analysis of the relevant L2 data, namely the assumption diat sentences have functionally "rich" stnictares such as (5).In addition to structures such as (5), her analysis also assumes Pollock's AGR parameter which, in tarn, presupposes such functionally rich sentence structure.
So, the answer to the question posed at the beginning of this section must be that there is indeed much to be gained from using the descriptive apparatus of UG theory in the analysis of L2 phenomena.

New insight into old problems
Sections 2 and 3 were mainly concerned with countering the claim that UG theory is irrelevant to the study of L2 acquisition.The thrust of the argument has been to show that L2 acquisition research can indeed benefit from adopting both the basic tenets and the descriptive apparams of UG theory.In this section, I present further s^dence of the benefits to be gained from adopting specific UG-based poposaJs j^ut the content and organisation of grammatical knowledge as a point of departure f(ir L2 acquisition research.The discussion will fociw on two remarkable spin-offs ^ch application of Pollock's insights into the role of functional categories in lenience structure has had for the study of L2 acquisition. TKe first spin-off is a renewed interest in the results of the famous morpheme order ^dies conducted in the 1970s.These studies purported to show that functionaj oioiphemes are acquired in a predictable order by L2 learners of English, regardless af whether the L2 is acquired naturally or in a formal ieaming environment.^^ "Moreover, the fmdings of these studies were claimed to be remarkably siniilar to vthose of LI studies conducted earlier, as is clear from the For close on 15 years now, the findings of the morpheme order studies have been largely discounted as a result of objections to, firstly, their methodology and, secondly, their failure to provide a theoretical explanation of their findings.As the morphemes that were studied included both bound and free, and both nominal and verbal morphemes, the studies were claimed to have yielded no insight whatsoever into the reasons for the observed L2 acquisition orders, nor for the differences between the LI and L2 orders.A second spin-off that recent developments in grammatical theory have had for the jtudy of L2 acquisition is a renewed interest in an aspect of L2 acquisition which seems to be problematic for L2 learners universally, namely the acquisition of tense and agreement properties and their overt expression in the L2.Following Pollock's proposal to separate the functional and lexical properties of sentences in structures as in (5), various hypotheses have been advanced to explain L2 learners' characteristic problems with word order and concord.
A first hypothesis, advanced by Vainikka andYoung-Scholten (1994, 1996), is that beginning L2 learners (like beginning LI learners according to one school of thought^^) initially assign a veiy basic lexical structure containing only noun phrases and verb phrases to sentences in the L2.This basic lexical structure is constructed on the basis of the semantic properties of the lexical items (particularly of the verb) in the sentence and it includes no functional nodes.Such basic structures allow beginning L2 learners to express meaning, i.e. to say who does what to whom.Such rudimentary structure does not allow the expression of tense and agreement, however.Functional nodes are gradually added, as evidence for them is picked up in the input the learner receives.Similarly, the value of a parameter such as the AGR parameter is only set for the L2 once the necessary functional structure is in place.
An alternative hypothesis, advanced by Schwartz andSprouse (1994, 1996), is that L2 learners adopt the entire LI grammar as an initial hypothesis about the L2.That is, they are assumed to transfer complete structures, including lexical and functional nodes and values for parameters such as the AGR parameter, from their LI into the initial grammar of the The task facing L2 learners in this case is to pick up indications of differences between the LI and the L2 from the input and to adjust their initial hypothesis about the L2 grammar accordingly.
The two hypotheses make different predictions about the kind of evidence about the target language which L2 learners will need.If the learner indeed starts from scratch, as claimed on the first hypothesis, then the kind of evidence needed is the same as that needed by LI learners, namely positive evidence, or evidence about what is possible in the L2.Positive evidence is gleaned from naturally occurring utterances in the language being acquired.If learners start out by adopting the LI grammar as an interim L2 grammar, positive evidence may not be sufficient to show up all the differences between their interim L2 grammar and the target grammar.In this case learners may require negative evidence, or evidence about what is not possible in the L2.Such evidence, typically, is not available in naturally occurring utterances and has to be specially provided, for example by the teacher.19 In either case the question arises as to what exactly is required to trigger the necessary changes in the learner's interim L2 grammar: does evidence about word order trigger knowledge of the tense and agreement properties of the L2, or does evidence about the tense and agreement properties of the L2 trigger knowledge of word order in the L2?The importance of the answers to these questions for those whose task it is to facilitate the acquisition of a second language in classroom conditions hardly needs to be mentioned.
Questions such as these are currently the focus of a spate of studies on the acquisition of German word order by native speakers of such diverse languages as Turkish, Spanish and Korean (cf., e.g., Vainikka andYoung-Scholten 1994, 1996;Eubank 1994;Schwartz andSprouse 1994, 1996), the acquisition of English by native speakers of French and/or German (cf., e.g., Schwartz 1993; White 1992; Eubank 1994a; Eubank 1996) and the acquisition of Bantu languages (with their rich system of noun-class prefixes and agreement markers) both by native speakers of other Bantu languages and by native speakers of languages without a noun-class prefix system (see discussion in Lardiere 1995: 553-554).
Developments in L2 research such as those discussed in this section have the potential to make a considerable contribution to our understanding of 1.2 acquisition.As such, they are clear evidence of the fruitfulness of UG theory as a framework for the study of L2 acquisition.

Conclusion
This paper has argued that a blanket rejection of UG theory as irrelevant to the study of L2 acquisition would be premature.First, it was shown, on grounds of concepmal coherence and theoretical simplicity, that a case can be made for adopting the basic tenets of UG theory as the null hypothesis for L2 acquisition research.Next, it was argued that by using the complex descriptive apparatus of UG theory, L2 researchers have been able to give more precise descriptions and more principled explanations of L2 phenomena, and to achieve greater depth of insight ^to the problems facing L2 learners.Finally, a brief review of recent L2. research jllostrated a particular advantage of keeping abreast of developments in UG theory: itjallows 12 researchers to take a fresh look at old problems and to come up with 'Mfcresting answers to questions that have plagued the; field for years.L2 •jesearchers, therefore, should not allow their view of this thriving field of L2 lesearch to be obscured by trees such as (1).commonly ascribed to, amongst others, Sportiche (1988).Cf., e.g., (Cook and Newson 1996: 146) for some discussion.Cf., e.g., (White 1992)

To summarize :
It has been shown in this section that a blanket rejection of the basic enets of UG theory as a framework for the smdy of L2 acquisition is unwarranted, rhe basic tenets of UG were claimed to be quite simple and to have remained essentially unchanged since they were first proposed.Moreover, the assumption that the basic tenets of UG hold for L2 acquisition was shown to be highly valued in terms of considerations of theoretical simplicity, whereas the alternative assumption was shown to be based on a conceptually incoherent claim.casefor adopting the basic tenets of UG theory as a framework for the study of L2 acquisition, let us turn now to the descriptive apparatus of the theory.Is anything to be gained from using the complex descriptive apparams of UG theory in the study of L2 knowledge and its acquisition?The aim of this section is to show, on the basis of a case study from the L2 literature, that the descriptive apparatus of UG theory makes it possible (i) to give much more precise descriptions of the problems facing L2 learners and (ii) to give principled explanations of problematic L2 phenomena.Let us first consider, briefly, why a simple sentence such as John kisses Mary is assigned a complex structure such as the one shown in (1) above on current versions of UG theory.For purposes of the discussion, the somewhat simplified version of (a) "John loves not Mary (c) John does not love Mary (7) (a) *Likes he Man?(8) (a) *John kisses often Mary (c) John often kisses Mary Nonnnite (9) (a) To not own a car... (c) "To own not a car... (b) Jean n'aims.pasMarie (d) Veanne pas gime Marie (b) Aime-t-il Marie?(b) Jean embrasse souvent Marie (d) Vean souvent embrasse Marie (b) Ne pas oosseder de voiture...(d) *Ne vosseder pas de voiture... (10) (a) To hardly understand French... (b) ^ / (c) "To understand hardly French... John has not kissed Mary (12) (a) Hss he kissed Mary? (13) (a) John has often kissed Mary Nonnnite (14) (a) To not bs.returning early... (c) To ^ not returning early... (b) Jean n'a pas embrasse Marie (b) A-t-il embrasse Marie?(b) Jean a souvent embrasse Marie (b) Ne pas etre retoume tdt...(d) N'etre pas retoume tot... (e) To not have had a childhood... (f) Ne pas avQireu d'enfance... (g) To have not had a childhood... (h) N'avoir pas eu d 'enfance... (15) (a) To often te returning early... (b) ...de souvent etrg retoume idt (c) To is.often returning early...(d) ...d'etre souyent retoume tot (e) To olten have kissed Y... (f) ...de souvent avoir embrassi y (g) To have often kissed Y... (h) ...d'avoir souvent embrasse Y Looking first at the sentences in (6)-(8), the following generalisations can be made (16) i.In English, finite main verbs cannot appear to the left of the negative, whereas in French they must appear to the left of the negative.(See (6)) ii.In English, finite main verbs cannot appear sentence-initially, whereas in French they can, (See (7)) iii.In English, finite main verbs cannot appear to the left of a time adverb, whereas io French they must appear to the left of the adverb, (See (8))What about nonfinite main verbs?Here the following generalisations can be made;(17) i.In both English and French, nonfinite main verbs can appear only to the right of the negative.(See (9)) ii.In English, nonfinite main verbs can appear only to the right of an adverb, whereas in French they can appear either to the right or to the left of an adverb.(See (10)) And, finally, the behaviour of main verbs has to be compared with that of auxiliaries.Here English seems to pattern like French, as a quick scrutiny of the sentences in (11)-(15) will show, i.e. (18) i.In both English and French, finite auxiliaries can appear sentenceinitially, and to the left of the negative and of adverbs.(See (11)-(13)) ii.In both English and French, nonfinite auxiliaries can appear to either the left or the right of the negative and of adverbs.(See (14) and (15)) We seem, then, to have the following situation: http://spilplus.journals.ac.za .^19) i-Atixiliaries (both finite and nonfinite) have the same distribution in English and in French, ii.As far as main verbs are concerned, there is a difference-.* In finite clauses, French requires main verbs to appear to the left of the negative and adverbs, and optionally allows them to appear sentence-initially.Finite verbs in English, by contrast, cannot appear to the left of the negative or an adverb, or sentence-initially.* In nonfinite clauses, both French and English disallow verbs from appearing to the left of the negative.French verbs can optionally appear to the left of adverbs, however, while English verbs cannot.
turn to some L2 data.French speakers learning English as an L2 have considerable difficulty with adverb placement in English, producing sentences such ^^ as (20a) instead of the correct (20b).^ ^ (20) (a) *Afarie rates always the mHro (b) Marie always takes the metro According to a study conducted by "White (1992-.285), the leamets producing sentences such as (20a) do so in spite of the fact that they appear to have reset the AGR parameter to its English value (i.e. from strong to weak), as evidenced by the fad that they do not produce sentences such as those in (21), in which a finite lexical verb has been moved out of the VP.* (21) (a) *Likes Jean the girls?(b) *Jean likes not the girls According to White (1992 : 285), facts such as those in (20) suggest that Frenchspeaking learners of English treat finite verbs in English like nonfmite verbs in French in sentences such as (20a).Nonfmite verbs in French can occur to either the left or the right of an adverb, as is clear from a comparison of (10b) and (lOd) above.In terms of Pollock's analysis, nonfmite verbs should not be able to occur to the left of adverbs.In order to occur to the left of an adverb, a verb would have to move out of the VP.This should not be possible in the case of nonfmite lexical verbs, as their lack of tense and agreement would not permit them to transmit their theta roles.Our concern is not so much with the question of why French has the exceptional property referred to above.Rather, our concern is with White's explanation of the L2 data and what it tells us about the advantages or disadvantages of using the descriptive apparatus of UG theoiy in the study of L2 acquisition.Three outcomes of White's analysis of the L2 English data are relevant to our discussion.The first outcome is that White's analysis links a problematic property of French-speaking learners' English to an exceptional property of their LI granunar.That is, her analysis has made it possible to give a much more precise description of a problematic aspect of these learners' English and to pinpoint exactly what it is that needs to be explained.
Recent theoretical developments relating to the distinction between lexical and functional categories outlined in section 3 above, have prompted L2 researchers Helmut Zobl and Juana Liceras to reanalyze the findings of the morpheme order studies.In a nutshell, they take the particular clustering of morphemes at the top middle and bottom of the hierarchy in the L2 data to indicate that L2 acquisition is driven by the distinction between bound and free morphemes, rather than by the functional-lexical distinction.According toZobl (1995:41-42)  this is evidenced by the fact that i. in L2 acquisition, free morphemes are acquired early and in cross-categorial fashion (i.e.regardless of whether they are nominal or verbal), as indicated by the clustering of copula be and auxiliary be (which are verbal) and articles (which are nominal) at the top of the hierarchy, whereas ii. the acquistion of bound morphemes is delayed, but also occurs crosscategorially, as indicated by the clustering of past tense morphemes (verbal) along with 3rd person singular -s and possessive -s (both nominal) at the middle and bottom of the hierarchy.The LI data, by contrast, indicate that category type, rather than the distinction between bound and free morphemes, drives LI acquisition, witness the fact that i. in LI acquisition, the nominal morphemes (possessive -s and articles) cluster together towards the middle of the hierarchy, whereas the majority of verbal morphemes (past tense -ed, 3rd person singular -s and auxiliary be) occur at the bottom of the hierarchy; whereas ii.there appears to be no clear separation in acquisition order between bound and free morphemes in the LI data.^"^On the basis of their reanalysis, Zobl and Liceras (1994:162-163) conclude that in LI acquisition, functional categories are acquired gradually according to a maturational schedule, with the emergence of a panicular functional category allowing the learner to be sensitive to both its free and its bound exponents.In L2 acquisition, by contrast, functional categories are already available (as a result of the fact that the learner already possesses an LI grammar) and learners need only acquire the language-specific exponents of these categories.^^Briefly then, what Zobl and Liceras's (1994) reanalysis of the findings of the morpheme order studies illustrates, is how interesting L2 data which have been lying around largely unused for dose on two decades, have acquired a new significance ±anks to developments in the UG theory of grammar.
that this account appears not to explain the presence tight at the top of the hierarchy (for both LI and 12 acquisition) of progressive -ing and plural -s.These two moiphemes are claimed to be lexical rather that functional on the grounds that they differ from morphemes such as the tense, agreement and possessive morphemes in not being related to case-marking, for example.Cf.(Zobl 1995: 51-52 n. 7).15 The development of the free exponents of fimctional categories before theii inflectional exponents is explained either by iheii greater perceptual salience or by the different ways in which free morphemes and bound morphemes move.Cf. (Zobl and Liceras 1994: 173) for discussion.16 This is the position adopted by Radford (1990) amongst others.17 Note the convergence between Schwartz and Sprouse's hypothesis and Zobl and Liceras's finding that functional categories are available to L2 learners right from the outset.See discussion above.18 But see (Eubank 1994), (Eubank 1994a) and (Eubank 1996) for the slightly weaker claim that both lexical and functional projections tramfer, but that parametric values of morphology-driven features, such as the strength of agreement, do not.19 Cf., e.g., (White 1987: 100-107) for a discussion of the circumstances in which L2 learners need negative evidence.