Another Take on the L2 Initial State: Evidence from Comprehension in L2 German

Another Take on the L2 Initial State:Evidence From Comprehension

in L2 German

Theres GrüterMcGill University

The nature of the initial state in second-language (L2) acquisition is a much debatedbut still unresolved issue, due in part to the empirical problem of obtaining produc-tion data from L2 learners at very early stages in development. In an attempt to cir-cumvent this problem, this article presents evidence from a comprehension-basedexperiment involving 17 adult English-speaking learners of German at the initialstate. The experiment investigates learners’ interpretations of ambiguous wh-ques-tions (e.g., Was beisst die Katze? ‘What is biting the cat?/What is the cat biting?’) ina picture interpretation task. The results present clear evidence against the MinimalTrees hypothesis (Vainikka and Young-Scholten (1994; 1996a; 1996b)), which pro-poses that early L2 grammars lack functional categories, while lending support to aFull Transfer position (e.g., Schwartz and Sprouse (1994; 1996)), which holds thatthe L1 grammar as a whole constitutes the L2 initial state.

1. INTRODUCTION

The nature of the initial state of second-language (L2) grammars has been undercontinuing—and still unresolved—discussion as it constitutes a crucial factor indetermining the course that subsequent L2 development can and must take. Thusa principled account of the initial state is a key component in any theory of L2 ac-quisition. Several accounts have been put forward over the last decade, withVainikka and Young-Scholten (henceforth VYS)’s (1994; 1996a; 1996b) Mini-mal Trees hypothesis and Schwartz and Sprouse’s (1994; 1996) Full Transfer FullAccess account being two of the most dominant. A substantial amount of empiri-cal evidence has been adduced to date in support of these different theories. Whatis problematic, however, is that this evidence consists almost exclusively of(spontaneous and elicited) production data. Many claims rely on the assumption

LANGUAGE ACQUISITION, 13(4), 287–317Copyright © 2005/2006, Lawrence Erlbaum Associates, Inc.

Correspondence should be sent to Theres Grüter, McGill University, Department of Linguistics,1085 Dr. Penfield Ave., Montréal, Québec, H3A 1A7 Canada. E-mail: [email protected]

that the absence of a property in production implies the absence of its abstract un-derlying representation. Yet this assumption has been challenged by several stud-ies showing that learners tend to observe syntactic requirements associated with aspecific functional category otherwise claimed to be absent due to its lack of mor-phological realization (e.g., correct pronoun choice with respect to case as a reflexof I0 even in the absence of inflectional morphology on the finite verb; seeGavruseva and Lardiere (1996), Lardiere (1998a; 1998b)). These studies con-clude that absence of evidence, that is, absence of overt morphological material,does not constitute evidence of absence, that is, absence of the corresponding un-derlying syntactic representation.

In view of this growing skepticism toward production data as fully reliableevidence, it seems that data from very early L2 comprehension are needed toshed further light on the nature of the initial state, as pointed out by White(2003):

Indeed, to investigate the possibility that there might be a stage prior to the emer-gence of L2 speech in which functional categories are lacking, we need methodolo-gies that do not rely on production data. Comprehension tasks where functionalproperties are manipulated are not easy to construct. (p. 75)

The aim of this article is to present evidence from precisely such a task. The ex-periment reported here takes advantage of a property of some constituent wh-questions in German, namely, their ambiguity between a subject and an objectquestion interpretation, as shown in (1).

(1) a. Was beisst die Katze?what bite-3SG the cat‘What is biting the cat?’ (= subject question)OR ‘What is the cat biting?’ (= object question)

b. Was hat die Katze gebissen?what have-3SG the cat bitten‘What has bitten the cat?’ (= subject question)OR ‘What has the cat bitten?’ (= object question)

Crucially, the representation and interpretation of these strings as either subject orobject questions involves the presence of functional categories beyond the verbphrase (VP). The picture interpretation task presented in this article is designed toevaluate initial state L2 learners’ interpretations of such questions, thus providingdirect evidence on the presence or absence of functional projections in their (ini-tial state) interlanguage grammars.

The remainder of the article is organized as follows: Section 2 gives a briefoverview of the discussion on the L2 initial state and introduces the two hypothe-

288 GRÜTER

ses under investigation (Minimal Trees and Full Transfer). The relevant syntacticbackground on wh-questions is provided in section 3, followed by the predictionsthat ensue from the Minimal Trees and the Full Transfer hypotheses for the exper-iment presented here (section 4). The experiment itself—investigating English-speaking learners of German at the beginner’s stage—is presented in section 5,concluded by a discussion of the results and their implications for L1 transfer andthe nature of initial state grammars in L2 acquisition.

2. L1 TRANSFER AND THE NATURE OF THE INITIALSTATE: MINIMAL TREES VERSUS FULL TRANSFER

In the domain of second-language learning, it has long been noted that proper-ties of the mother tongue exert some influence on the second language beinglearned (Gass (1984), Lado (1957), inter alia). Within the principles and param-eters approach to L2 acquisition, effects of transfer were captured by consider-ing L2 learners to start out with the parameter settings of their L1 (e.g., White(1985)). It is not until the mid-1990s, however, that principled accounts of thenature and domain of transfer have been proposed which link the initial state ofL2 grammars to subsequent development in the L2. As Schwartz and Eubank(1996, 3) pointed out, this recent body of work goes beyond earlier, largely de-scriptive accounts of transfer in that it seeks to provide an explanatory accountof L2 development that crucially hinges on the nature of transfer assumed at theinitial state. In this sense, every theory of the initial state is inextricably linked toclaims about the course further L2 development can take. At the same time, it iscrucial for any theory of L2 acquisition to be clear about what is posited as theinitial state.

There are several logical possibilities one might assume for the initial state ofL2 grammars: (i) the initial state is “pure UG,” that is, there is no difference be-tween the initial state in L1 and L2 acquisition; (ii) the L2 initial state is a spe-cific, pre-existing grammar, namely, that of the L1—as a whole or in part; (iii)neither Universal Grammar (UG) nor L1 constitute the L2 initial state. Position(i) is represented by Platzack (1996) and the work of Epstein, Flynn, andMartohardjono (1996; 1998), position (iii) by the earlier work of Clahsen andcolleagues (e.g., Clahsen and Muysken (1986), Meisel, Clahsen, and Pienemann(1981)).1 Neither of these two positions has a straightforward manner of ac-

ANOTHER TAKE ON THE L2 INITIAL STATE 289

11Two more recent approaches to L2 acquisition—Klein and Perdue (1997) and Pienemann

(1998)—are also representative of position (iii). However, both of these accounts admit L1 transfer insome form (cf. Klein and Perdue (1997, 314), Pienemann (1998, 82)) while leaving the nature and do-main of such transfer largely undefined. This makes it difficult, if not impossible, to derive clear pre-dictions for an experiment such as the one reported in this article. For this reason, these accounts arenot further considered here.

counting for the robustly attested L1 effects in L2 grammars. As the focus of thisarticle is on the nature and extent of L1 transfer, these positions are not furtherdiscussed here. (The reader is referred to White (2003, chap. 3) for further dis-cussion of various initial state hypotheses.) Position (ii), probably the mostwidely held, comprises a number of conflicting theories in itself. In particular,the debate is on how much of the L1 is taken to form the initial state of the L2. Idiscuss in more detail two theories that make very explicit and different claimsin this respect: the Minimal Trees account of VYS (1994; 1996a; 1996b), andthe Full Transfer Full Access hypothesis proposed by Schwartz and Sprouse(1994; 1996).

2.1. Minimal Trees

The Minimal Trees hypothesis—originally proposed by VYS (1994) and adoptedby several scholars in the field of L2 acquisition (e.g., Hawkins (2001), Myles(2005))—was put forward in analogy to Weak Continuity approaches to L1 ac-quisition. These approaches (Guilfoyle and Noonan (1992), Radford (1988), interalia) hold that children’s grammars initially contain only lexical categories,whereas functional categories and projections are added to the syntactic represen-tation in a stepwise fashion in the course of development. This account capturesthe observation that lexical material associated with functional categories (e.g.,determiners, auxiliaries, verbal inflection) as well as their syntactic consequences(e.g., verb raising) are typically missing from the speech of very young children.The same observation can be made for the speech of early L2 learners. In conse-quence, VYS (1996b) proposed that “L2 learners build up phrase structure inmuch the same way as children do” (p. 13). Under their Minimal Trees hypothe-sis, the grammars of L2 learners at the initial state will project nothing beyond abare VP, consisting exclusively of lexical categories and lacking functional ones.Moreover, the headedness of the VP projected in early L2 is assumed to be identi-cal to (i.e., transferred from) the L1.

L1 transfer, under this account, is confined to lexical categories, most impor-tantly the VP. In cases where L1 and L2 differ with respect to the headedness ofthe VP, the learner is assumed to pass through a first VP-stage (“VPi-stage” inVYS (1996b)), where VP-headedness is as in the L1, followed by a second VP-stage (“VPii-stage”), where headedness is flipped to the target L2 value. Cru-cially, the switching of VP-headedness is taken to precede the acquisition of anyfunctional material (VYS (1996b, 20)). The syntactic representations availableunder the Minimal Trees account to early learners of a head-final L2 (e.g., Ger-man) whose L1 is head-initial (e.g., English, Romance) are illustrated in (2).

290 GRÜTER

(2)

(VYS (1996b, 20), (10b))

As implied by the representations in (2), the claim is that at the initial state “allclausal projections are treated as VPs since only lexical projections are trans-ferred” (VYS (1996b, 32)). To my knowledge, evidence in support of this posi-tion has come exclusively from spontaneous production data (VYS, severalarticles since 1994). These studies have shown that the speech of L2 learners ofGerman at an early stage in development complies with the authors’ five criteriafor a VP-grammar, listed in (3).

(3) At the VP-stage we find a lack of

� verb raising� auxiliaries and modals� an agreement paradigm� complementizers� WH-movement

(VYS (1996b, 16))

The exact procedures for determining presence or absence of these properties,however, are somewhat difficult and to a certain extent arbitrary, a point I returnto in the analysis of the production data obtained as part of the present experiment(see section 5.3.1). Moreover, no evidence (that I am aware of) has been presentedfrom comprehension or interpretation tasks showing that (functional) materialbeyond the VP is lacking in the grammars of early L2 learners. If it is agreedthat the responsibility of “a grammar” is for comprehension as well as production,the Minimal Trees hypothesis predicts that the comprehension of material associ-ated with functional categories beyond the VP must be compromised in early L2learners.

Comprehension-based evidence running counter to the prediction of MinimalTrees that transfer is limited to lexical categories emerges indirectly from a studyby Slabakova (2000), which investigated the functional projection for telicity(AspP) in the L2 grammars of Bulgarian and Spanish learners of English through


an interpretation task. The results showed significant differences between theSpanish and the Bulgarian learners with respect to the interpretation of telicity inEnglish. The differences are attributed to differences in the representation of as-pect, that is, the nature of AspP, in the two L1s. Thus properties of a functionalcategory in the L1 were shown to affect interpretation in the L2. This constitutesdirect counterevidence to the Minimal Trees hypothesis, under which L1 func-tional categories have no impact on L2 grammars.2

2.2. Full Transfer

An alternative to the Minimal Trees hypothesis is presented perhaps most explic-itly by Schwartz and Sprouse’s (1994; 1996) Full Transfer Full Access account,which claims that “the entirety of the L1 grammar (excluding the phonetic matri-ces of lexical/morphological items) is the L2 initial state” (Schwartz and Sprouse(1996, 41)). In other words, the representations of the L1, including both lexicaland functional categories, are fully transferred at the beginning of L2 acquisition.As a result, we expect transfer effects for both lexical and functional categories.Evidence supporting this claim has been presented in a number of studies, againbased largely on production data (e.g., Haznedar (1997) for the transfer of nega-tion phrase (NegP) from L1 Turkish into L2 English).

As does the Minimal Trees hypothesis, Full Transfer Full Access makes ex-plicit claims about the further course of L2 development, based on its character-ization of the initial state. The focus of this article being exclusively on the natureof the initial state, however, I remain silent on these further predictions. All that isof interest here is the first part of Schwartz and Sprouse’s hypothesis, namely,Full Transfer. Note that, in principle, the Full Transfer position on the initial stateis compatible with a number of very different approaches to L2 development: It isconsistent not only with the Full Transfer Full Access model but also withHawkins and Chan’s (1997) Failed Features Hypothesis, which differs from theformer only in that it does not allow for acquisition of features not present in theL1. Although differing in their accounts of subsequent development, the twomodels make precisely the same assumptions with respect to the initial state.Moreover, a Full Transfer position is not in conflict with accounts which proposethat L2 acquisition is fundamentally different from L1 acquisition (e.g., Bley-Vroman (1990)) in that none of the presumed innate mechanisms involved in thelatter are operative in the former. If the claim is that the only linguistic resourceavailable to L2 learners is L1 knowledge, we have essentially (the potential for) a

292 GRÜTER

22Several studies (e.g., Grondin and White (1996)) have presented production data from L2 learners at

an early stage containing functional elements such as determiners and wh-words. Although such data ap-pear to constitute prima facie evidence against Minimal Trees, it is not clear whether the observed func-tional categories are the result of L1 transfer or whether they had been subsequently acquired as a resultof Full Access in the L2. The latter scenario would be in full accordance with the Minimal Trees hypoth-esis, and thus such data cannot be taken as evidence against the Minimal Trees account.

Full Transfer scenario. In consequence, I refer to the position proposing transferof the entire L1 grammar for the initial state of L2 as Full Transfer only, leavingaside any discussion of subsequent development.

The aim of this article is to adduce empirical evidence relevant to deciding be-tween the Minimal Trees and the Full Transfer hypotheses of the initial state.Given the current debate on the appropriate interpretation of production data, anexperiment investigating the comprehension and interpretation of material associ-ated with L2 functional categories has been constructed. The following sectionprovides the relevant syntactic background on the L2 and the L1 under consider-ation, German and English, respectively.3

3. (AMBIGUOUS) WH-QUESTIONS IN GERMAN

Following standard assumptions, German is a V2 language with underlying SOVword order, and a requirement for overt wh-movement to the clause-initial position(den Besten (1983), Koster (1975)). As a result of these properties, all wh-ques-tions in German have the same word order: the wh-phrase (in Spec-CP), followedby the finite verb (in C0), followed by the rest of the clause. What typically disam-biguates a subject- from an object-wh-question is morphological case marking onthe wh-word, the determiner, and/or the noun, as illustrated in (4a) and (4b).

(4) a. Wer kennt den Jungen?who-NOM know-3SG the-MASC-ACC boy-ACC‘Who knows the boy?’

b. Wen kennt der Junge?who-ACC know-3SG the-MASC-NOM boy-NOM‘Who does the boy know?’

However, due to syncretism in the case-marking paradigms of German, there arescenarios where subject and object questions cannot be disambiguated by casemarking. In particular, this is the case when the gender of the determiner phrase(DP) is feminine or neuter, and the wh-word is was ‘what’, as illustrated in (5).

(5) Was beisst die Katze/ das Kamel?what bite-3SG the-FEM cat / the-NEUT camel‘What is biting the cat/ the camel?’ (= subject question)OR ‘What is the cat/ the camel biting?’ (= object question)

As was does not inflect for case, and nominative and accusative case marking isthe same (i.e., null) in the feminine and neuter paradigms, (5) is ambiguous be-


33The discussion is presented within a principles and parameters model of syntax (Chomsky

(1981)) rather than the more recent minimalist framework (Chomsky (1995)) due to the fact that thetwo initial state hypotheses under consideration were formulated within this earlier model.

tween a subject- and an object-question interpretation. The appropriate interpreta-tion in this case will depend uniquely on the context. (Note that there is nodifference in prosody.)4

The syntactic structures underlying the two interpretations of (5), however, aredifferent, as illustrated in (6). In the subject question, [die Katze] is parsed as thecomplement of V0, and the wh-word binds a trace in Spec-IP (and Spec-VP, if theVP-internal subject hypothesis (Kitagawa (1986)) is assumed). In the object ques-tion, [die Katze] is parsed in Spec-IP (or Spec-VP), and the wh-word binds a tracein the complement of V0.

(6) a. subject question: b. object question:‘What is biting the cat?’ ‘What is the cat biting?’

The experiment presented in this article relies on this structural ambiguity of wh-questions where case marking cannot disambiguate between subject- and object-question interpretations. Such scenarios are found not only with simplex verbs, asillustrated by the present-tense question in (5) and (6), but also with periphrasticverb constructions in the complex tenses. An example of an ambiguous wh-ques-tion in the perfect tense (Perfekt) is provided in (7).

294 GRÜTER

44As an anonymous reviewer pointed out, in requests to identify humans, the wh-word generally

used is wer/wen ‘who’, whereas was ‘what’ is mostly used for inanimates, an observation confirmedby any reference grammar of German (e.g., DUDEN (1998, §606)), and largely analogous to the dis-tinction between ‘who’ and ‘what’ in English. Thus, ideally, the referent of was in a was-question isinanimate. However, it is difficult to construct semantically reversible clauses with two inanimate ref-erents. For this reason, the referents in the questions used in this experiment are animals—neither hu-man nor inanimate—for which both wer/wen and was are grammatical, although some speakers mayhave a preference for the former. See section 5.4 for how such a preference and the animacy distinctionbetween wer/wen and was could have affected the responses in the present experiment.

(7) a. subject question: b. object question:‘What has bitten the cat?’ ‘What has the cat bitten?’

Classroom learners of German are typically taught that word order is not a reli-able cue to identifying the subject and the object of a German clause and that theyshould look for case marking instead. In most cases, this strategy will allow themto correctly identify the semantic roles. However, there is no strategy—taught ornot—that will help them disambiguate questions such as (6) and (7). As these am-biguous scenarios only arise under specific circumstances (see above), they arequite rare, and—to the best of my knowledge—not discussed in the classroom.Thus the chance of explicit instruction influencing learners’ interpretations here isminimal.

It is important to note that, in all cases, the syntactic representations assumedfor adult native German require (at least) the functional projections IP and CP. Insection 4, I consider the possible syntactic representations that could be assignedto strings like these by L2 learners at the initial state, under the Full Transfer andunder the Minimal Trees hypotheses.

3.1. Wh-Questions in English

As we are concerned with transfer from the learners’ L1 (English) a brief consid-eration of wh-questions in English is in order. The literature on the structure andacquisition of English wh-questions is extensive and intricate (see, e.g., Stroms-wold (1995) for discussion and further references), yet for the purposes of myexperiment, many details are irrelevant and only a few very general and uncontro-


versial assumptions are necessary: English is an SVO language with overt wh-movement to the clause-initial position; in interrogative clauses, auxiliaries (andmodals) raise to C (8a), whereas thematic verbs remain in situ and generally re-quire do-support (8b). An exception is presented by subject wh-questions (9a),where do-support is not only not required but ungrammatical (9b) (unless withemphatic stress).

(8) a. [CP Whatk hasi [IP the catj t�i [VP t�j ti [VP tj bitten tk]]]]b. [CP Whatk doesi [IP the catj ti [VP tj bite tk]]]

(9) a. What bites the cat?b. *What does bite the cat?

The representation of subject wh-questions such as (9a) is debated, as it cannot bedetermined from the surface string whether the wh-word has “vacuously” movedto Spec-CP or whether it has remained in Spec-IP (Chomsky (1986)). Becausethis distinction does not affect the predictions in my experiment, I remain agnosticon the position of the wh-word in English subject wh-questions.

4. HYPOTHESES AND PREDICTIONS

4.1. The Full Transfer Hypothesis

As previously discussed, Full Transfer holds that the entire L1 grammar consti-tutes the initial state of L2 grammars. Thus the initial phrase structure available toa native speaker of English acquiring German will look as in (10) (omitting irrele-vant detail).

(10) Initial state phrase structure for the English-speaking learner of Germanunder the Full Transfer hypothesis.

Moreover, Full Transfer will assume that the English learner transfers the prop-erty of English thematic verbs to stay in situ, that is, inside the VP. With the

296 GRÜTER

phrase structure in (10) and without verb raising, how could a learner parse a Ger-man wh-question in the present tense, such as that in (6), repeated here as (11)?

(11) Was beisst die Katze?what bite-3SG the cat‘What is biting the cat?’ OR ‘What is the cat biting?’

The phrase structure in (10)—like the grammar of English—offers no straightfor-ward possibility for parsing a postverbal noun phrase in a wh-question as asubject, thus [die Katze] here must be analyzed as an object. Consequently, thewh-phrase must be interpreted as the subject, making the only possible interpreta-tion that of a subject question, as shown in (12).

(12) Syntactic representation of (11) under the Full Transfer initial state gram-mar in (10).5

Consequently, the Full Transfer hypothesis predicts English-speaking learners ofGerman at the initial state to interpret constituent wh-questions in the presenttense as subject questions only.

The situation is exactly reversed for questions in the perfect tense (see (7), re-peated here as (13)).

(13) Was hat die Katze gebissen?what have-3SG the cat bitten‘What has bitten the cat?’ OR ‘What has the cat bitten?’

The only way for an English-like (SVO) grammar to parse a clause-final transitiveverb is by moving its object to the left. The only available parse, then, is the one inwhich the fronted wh-phrase is analyzed as the object, as shown in (14).6


55The representation in (12) assumes a vacuous movement analysis of subject wh-questions in Eng-

lish. On the alternative analysis, the learner would be expected to parse was ‘what’ under Spec-IP. Theprediction remains the same: Only a subject question should be available to these learners.

66Barring the possibility that some sort of object shift has applied in addition to wh-movement of

the subject. I can see no property of the L1, English, which would warrant such a scenario.

(14) Syntactic representation of (13) under the Full Transfer initial state gram-mar in (10).

The Full Transfer hypothesis thus predicts L1-English initial state learners of Ger-man to interpret constituent wh-questions in the perfect tense as object questionsonly.

4.2. The Minimal Trees Hypothesis

The Minimal Trees account holds that only the L1 VP will be transferred and thatFunctional Projections are absent at the initial state. Under this view, the phrasestructure at the disposal of English-speaking learners of German is that in (15).

(15) Initial state phrase structure for the English-speaking learner of Germanunder the Minimal Trees hypothesis.

How can a grammar of this shape accommodate the German questions in the pres-ent (11) and the perfect tense (13)? In the case of the present tense, the string canbe fitted into the minimal tree in (15) rather straightforwardly, if we allow for theparsing of the wh-phrase in the specifier of the VP, as shown in (16).

298 GRÜTER

(16) Syntactic representation of (11) under the Minimal Trees initial stategrammar in (15).

With the noun phrase [die Katze] in postverbal position, the representation in (16)must yield a subject question interpretation. As there is no room for movementwithin this minimal tree, it will be impossible to associate the wh-phrase with anunderlying object, thus ruling out an object question interpretation. In the case ofthe present tense, the Minimal Trees hypothesis therefore predicts a subject ques-tion interpretation only for English-speaking learners at the VP(i)-stage.

Accommodating the question in the perfect tense (13) in the minimal tree in(16) is more problematic, as illustrated in (17).

(17) Attempt to accommodate (13) within the Minimal Trees grammar in (15).

Only three positions are available in the syntactic tree: the specifier of VP, V0, andthe complement of V0. I argue that the constituents competing for these three posi-tions are (i) the wh-phrase, (ii) the lexical verb, and (iii) the NP [die Katze]. Theprosodically least salient part, the auxiliary hat, will then remain unanalyzed.7 Yetthe three remaining constituents still cannot be accommodated under a head-initial VP, as the verb appears in clause-final position: Both the wh-phrase and theNP [die Katze] compete for the specifier position, while the complement positionremains empty. Strictly speaking, then, the question in the perfect tense cannot be


77See Radford (1996, 57) for a proposal that L1 learners may similarly fail to parse items that lack

acoustic salience.

parsed syntactically by a grammar of this nature, and we might expect learners toguess between the two interpretations available.8

It is possible, however, that in the absence of a syntactic representation, learn-ers might employ some other strategy for interpretation, such as following linearorder (as has been proposed for agrammatic patients; see, e.g., Grodzinsky(1995)). Studies with (normal) adult native speakers of German (and Russian)have shown that in the absence of other cues, speakers prefer to associate the firstNP in a string (declarative sentence) with the subject role and the second NP withthe object role (Kempe and MacWhinney (1999)). If this strategy is carried overto wh-questions, this would mean associating the initial wh-phrase with the sub-ject role, resulting in a subject question interpretation. Thus under the MinimalTrees account, the prediction for questions in the perfect tense is either guessing

300 GRÜTER

88An anonymous reviewer raised the possibility of VP adjunction, whereby was could be analyzed

in a position adjoined to VP. Such an analysis has been proposed for object wh-questions in child Eng-lish within a structure building approach to L1 acquisition (Radford (1996, 58)). Because VPadjunction does not involve functional structure, it appears prima facie compatible with the MinimalTrees approach to L2 acquisition. It is interesting to note, however, that the type of utterances that gaverise to the VP adjunction analysis in L1a (e.g., what kitty doing? (Radford (1996, 58)) does not seem tobe attested in the L2 corpora on which VYS’s Minimal Trees analysis is based. VYS stated explicitlythat none of their Turkish and Korean learners of German at the VP-stage produced any “wh-questionswith a fronted wh-phrase” (VYS (1994, 282; 1996b, 19)). Given that fronted wh-phrases are abundantin the input, the absence of such structures in L2 learner utterances is surprising if VP adjunction is anoption. (Their Romance learners produced “very few Wh-questions” (VYS (1996a, 155)), and the ex-amples provided suggest that object questions were not among them.) Perhaps it is for this (empirical)reason that the proponents of Minimal Trees never explicitly discuss the possibility of VP adjunctionat the VP-stage. However, there is a more serious learnability problem that arises if VP adjunction isallowed as an option at the VP-stage in L2a, which can be illustrated nicely with VYS’s Romancelearners of German. Take a German matrix clause with a periphrastic verb construction of a type abun-dant in the learner’s input (e.g., (i)), and suppose that these learners try to interpret it with their (trans-ferred) head-initial VP and the possibility of VP adjunction. Their only possible analysis could be thatin (ii), leading to interpretation B, which is marked in adult native German. Such a preference seemshighly unlikely, although I am unaware of any experimental evidence bearing on the issue. More im-portantly, the possibility of assigning an interpretation to (i) through VP adjunction would present aconsiderable obstacle to resetting the learners’ head-initial VP to head-final. Presumably, the word or-der in clauses such as (i) is an important trigger for such a resetting. If a VP adjunction analysis is pos-sible, the evidence for a head-final VP in the input will become rather unclear. At the very least, onemight expect, then, that the resetting of VP headedness would be a difficult and lengthy process. Yetthis is contrary to the observations of VYS, who found that their Romance learners reset VP headed-ness as a first step in their L2 development, that is, before acquiring any functional structure (VYS(1996b, 15)). Thus I argue on grounds of both descriptive adequacy and learnability that VPadjunction cannot be an option at the VP-stage under Minimal Trees.

(i) Die Katze hat die Maus gebissen.the-FEM cat have-3SG the-FEM mouse bittenInterpretation A: ‘The cat has bitten the mouse.’ (unmarked interpretation)Interpretation B: ‘The mouse has bitten the cat.’ (marked interpretation; typically requiresstrong emphasis on die Katze)

(ii) [VP die Katzei [VP die Maus [V� gebissen ti]]]

behavior due to the lack of a possible syntactic analysis or a potential bias towarda subject question interpretation as a result of a linear-order strategy.

In sum, the predictions of Full Transfer and Minimal Trees converge in thecase of the present tense: Both accounts predict English-speaking learners of Ger-man at the initial state to interpret ambiguous constituent wh-questions as subjectquestions only. For questions in the perfect tense, however, predictions differ.Whereas Full Transfer clearly predicts an object question interpretation only,Minimal Trees makes no principled predictions because the relevant string cannotbe accommodated within the syntactic representations available to learners at thisstage. If any bias can be expected, it would have to be based on nonsyntactic strat-egies for interpretation and would most probably lead toward a subject rather thanan object question interpretation. Table 1 summarizes the predictions of the twohypotheses under investigation.

5. THE STUDY

5.1. Participants

Two groups participated in this study: 15 native speakers of German (the L1group; M age = 37, range = 15–65), and 17 English-speaking learners of German(the L2 group; M age = 20, range = 18–30). The L1 group was tested in Switzer-land. The L2 group was recruited from German beginner classes at McGill Uni-versity in Montreal, Canada. Testing took place between Weeks 8 and 10 of thecourse, which consisted of 3 hr of contact time per week. None of the learners hadhad significant prior exposure to German.9 All learners identified English as theirfirst and dominant language.

5.2. Procedure

All participants completed a picture interpretation task (section 5.2.2), which tookapproximately 30 min. The L2 group also participated in a short (5–10 min) elic-ited production task (section 5.2.1), which was administered immediately follow-ing the picture interpretation task. I administered both tasks, as a native speaker of


TABLE 1Summary of Predictions

Present Tense Perfect Tense

Minimal Trees Subject question Guessing (potential bias toward subject question)Full Transfer Subject question Object question

99An additional L2 learner (E5) was excluded from the study due to exposure to German during

high school.

German, on a one-to-one basis with each participant. The elicited production taskwas tape-recorded and subsequently transcribed by me.

5.2.1. Elicited production. To ensure that the L2 learners studied hereare not beyond the initial state as defined by the Minimal Trees account, theVP(i)-stage (see (3) for VYS’s (1996b) criteria), that is, to ensure that their re-sponses can be considered relevant evidence for the hypotheses under consider-ation, a sample of their spoken German had to be collected. Participants weregiven an 18-panel picture story illustrating a little girl’s morning routine. Theywere asked to tell the story and/or describe the pictures to the experimenter inGerman. It was clear from the outset that this would be a very hard task for learn-ers at such an early stage. Participants were told that they were not expected to tella perfect story, that they could skip pictures if they did not have the appropriatevocabulary, and that they should just try and do their best. The experimenter inter-vened as little as possible during this task, asking only occasional questions orpointing to a picture if the learner seemed unable to proceed on his or her own.

5.2.2. Picture interpretation. In this task, participants were presented withvisual stimuli (see Figure 1 for an example), which were described to them by theexperimenter reading from a predetermined script (see (18)). They were then askedwhether the situation depicted was clear to them. The experimenter answered anyvocabulary-related questions, then proceeded with the experimental questions.10

(18) Excerpt from script describing the picture in Figure 1.Die Schlange jagt die Schildkröte. Die Schlange ist hinter der Schildkröte.Das Schwein jagt die Schlange. Das Schwein ist hinter der Schlange. (. . .)Die Katze jagt das Schaf. Das Schaf ist vor der Katze. (. . .)

302 GRÜTER

FIGURE 1 Example of visual stimulus.

1010Prior to the task, participants were made familiar with the basic vocabulary items used in the ex-

periment. Labeled pictures of the relevant animals remained within view for the entire testing session.Verbs and prepositions were translated and explained at the request of participants.

‘The snake is chasing the turtle. The snake is behind the turtle. The pig ischasing the snake. The pig is behind the snake. (. . .) The cat is chasingthe sheep. The sheep is in front of the cat. (. . .)’

In a first round of questions—the present-tense condition—each picture was fol-lowed by two (ambiguous) constituent wh-questions in the present tense (see (19)for an example and Appendix A for a complete list of the experimental stimuli)and two distractor questions.11 Participants replied nonverbally by simply check-ing the correct answer(s) on a multiple-choice answer sheet they were providedwith (see (20) for excerpt).

(19) Was jagt das Kamel?what chase-3SG the camel‘What is chasing the camel?’ / ‘What is the camel chasing?’

(20) Excerpt from answer sheet. Choice of answers for the question in (19).(Schaf (‘sheep’), Maus (‘mouse’), Hund (‘dog’); participants were in-structed to check the ? box if they were not sure how to reply.)

� Schaf � Maus � Hund � ?

Participants were instructed at the beginning of the task that there may be caseswhere more than one answer is possible and that they should check every answerthat constitutes a possible reply to the question asked. Moreover, half of thedistractor questions were constructed such that two answers would be possible.12

There were five visual stimuli, making for 10 constituent wh-questions in thepresent tense, plus 10 distractors.


1111The construction of appropriate distractors was not trivial. Ideally, distractors should not yield

any information about the property under investigation, which in this case consists of the relative orderof the lexical verb and its arguments—features hard to eliminate from any clause. To at least minimize

12

their potential basis for analogy, distractor questions were constructed as copular constructions, as in(i).

(i) Was ist hinter dem Schaf?‘What is behind the sheep?’

Even though these questions are subject questions, it is deemed that the absence of a lexical verb willmake the analogy with the ambiguous constituent wh-questions less salient and thus not lead to a biasfor a subject-question interpretation in the experimental items. As no such bias was found in the con-trol group, it is assumed that the distractor questions have no influence on the results.

12For example, (i) is a potentially ambiguous distractor question for the visual stimulus in Figure 1.

(i) Was ist vor dem Schaf?‘What is in front of the sheep?’

Depending on the interpretation of the preposition vor (‘in front of ’), it is possible to answer just‘the camel’ or both ‘the camel’ and ‘the mouse’. (For learners’ replies to these questions, see footnote21.)

In a second round of questions—the perfect-tense condition—the same fivestimuli were used.13 This time, participants were asked to memorize the picture,which was then removed before two constituent wh-questions in the perfect tense(21) were asked.

(21) Was hat das Kamel gejagt?what have-3SG the camel chased‘What has chased the camel?’ / ‘What has the camel chased?’

After trying to answer the two questions from memory, participants were shownthe picture again, and the two questions were repeated. They then had the oppor-tunity to correct their answers if they found that they had not memorized the pic-tures correctly. Only the corrected answers were considered in computing theresults, thus eliminating memory as a factor. Administering the perfect-tense con-dition as a disguised memory task served two purposes: to divert participants’ at-tention away from language and to make the use of the perfect tense pragmaticallymore appropriate (as the picture being referred to is no longer within view whenthe question is asked). Ten constituent wh-questions in the perfect tense were in-cluded in the task. Two versions of the task were created by varying the order ofpresentation of the five visual stimuli.

5.3. Results

5.3.1. Elicited production. Not surprisingly, L2 learners at this earlystage in development had great difficulty producing anything in German. Never-theless, between 4 and 16 verbal utterances could be elicited from each participant(M = 8.3), making for a corpus of 141 utterances. Following VYS’s (1996a) pro-cedure for analysis, this corpus excludes “utterances either lacking a verb or con-taining only a verb and nothing else as well as imitations or clear idiomaticphrases” (p. 151). In keeping with VYS’s (1996a; 1996b) diagnostics for the VP-stage, the corpus was then analyzed for the following five properties: verb raising,auxiliaries and modals, complementizers, wh-movement, and acquisition of anagreement paradigm. Table 2 presents a summary of verbal elements produced bythe L2 learners.

As illustrated in Table 2, none of the learners produced any auxiliaries ormodals, and this corpus does not contain overt complementizers or evidence of

304 GRÜTER

1313The two conditions are called present tense and perfect tense for ease of reference. As an anony-

mous reviewer correctly points out, tense itself is not directly relevant here. The distinguishing charac-teristic between the two conditions is that one contains a simplex verb (present tense), the other aperiphrastic verb construction (perfect tense).

wh-movement.14 With regard to verb raising in German, VYS (1996a) pointed outthat this is difficult to determine in the case of learners whose L1 has a head-initialVP: “We cannot distinguish verbs located in the head-initial VP from verbs poten-tially raised to a head-initial functional projection in a typical sentence” (p. 157).Evidence could come from the position of temporal adverbs and negation relativeto the main verb or the occurrence of postverbal subjects. As no such elements arepresent in this corpus, we can only conclude that based on the data available, thereis no evidence of verb raising in the L2 German of these learners.15


TABLE 2Distribution of Verbal Elements

Learner TotalMainVerbs

ist (‘is’)(Copula 3SG)

OtherCopula Aux Modal

E1 11 3 8 0 0 0E2 10 4 6 0 0 0E3 9 2 6 1 0 0E4 16 11 5 0 0 0E6 13 3 10 0 0 0E7 4 4 0 0 0 0E8 6 1 5 0 0 0E9 7 6 1 0 0 0E10 5 4 1 0 0 0E11 7 3 4 0 0 0E12 6 2 2 2 0 0E13 5 4 1 0 0 0E14 11 8 3 0 0 0E15 7 3 4 0 0 0E16 12 9 3 0 0 0E17 4 1 3 0 0 0E18 8 5 2 1 0 0Total 141 73 64 4 0 0

Note. Compare VYS (1996a, 152, Table 4).

1414It must be pointed out, however, that the elicitation context—a narrative—was not conducive to

the production of interrogatives, which would constitute the main evidence for wh-movement. Thisdisclaimer applies to the present study as well as to VYS’s own data. In consequence, it seems to methat not too much emphasis can be placed on the absence of such structures from these productionsamples.

1515One learner utterance is problematic in this respect, as it contains a particle verb in a construction

with the main verb in targetlike second and the particle in sentence-final position.

(i) Sie bereite Essen vor. (E11)She prepare food (particle)‘She is preparing food.’

The last property to be considered is the acquisition of an agreement paradigm.This is a notoriously difficult property to establish because criteria for “acquisi-tion” vary greatly (see White (2003, 77) for a critique). Again following VYS’s(1996a) procedure, only main verbs were considered for the purpose of analyzingagreement.16 Table 3 shows the proportion of correct agreement on main verbs.

Overall, only 40% of all verbs were correctly marked for agreement. This isclearly below the 60% cutoff point employed by VYS. Yet due to the small sam-ple sizes, percentages for individual learners should be viewed with caution. Fourof the 17 L2 learners (E7, E10, E13, E16) used agreement correctly in more than60% of cases. All of these correct utterances, however, involve the third-personsingular suffix -t. Based on these data alone, I believe it cannot be concluded thatthese four learners have fully acquired the agreement paradigm in German. Thusalthough the present data to some extent underdetermine the facts, I feel that it isnevertheless safe to assume that, overall, the L2 learners studied here fulfill thecriteria for being at the VP-stage as defined by VYS.

306 GRÜTER

TABLE 3Agreement on Main Verbs

LearnerMain Verbs

(Excluding Suppletive Forms)Correct

Agreement

E1 3 1 (33%)E2 3 1 (33%)E3 2 0 (0%)E4 9 2 (22%)E6 3 1 (33%)E7 4 3 (75%)E8 1 0 (0%)E9 5 0 (0%)E10 3 2 (67%)E11 3 1 (33%)E12 2 0 (0%)E13 4 4 (100%)E14 8 3 (38%)E15 3 1 (33%)E16 9 6 (67%)E17 1 0 (0%)E18 4 2 (50%)Total 67 27 (40%)

Note. Compare VYS (1996a, 154, Table 5).

16A syntactic analysis of this string would require a head-final VP, at least one functional category, aswell as verb raising to that functional category. This is clearly inconsistent with the VPi-stage. Yet thisbeing the only utterance of this kind in the present corpus, it may represent a memorized lexical chunkand has therefore been excluded from the analysis.

16As in VYS’s analyses, verbs requiring suppletive forms such as haben (‘to have’) wereexcluded.

Finally, it has to be ensured that the learners are at the VP(i) stage, that is, theheadedness of their VP is still that of their L1. Table 4 shows the headedness ofthe VP for those instances where this could be clearly determined.17

Not a single instance of a clear head-final VP was found. This suggests thatthese learners retain an English-like, head-initial VP and have not yet acquired thetarget head-final setting for German. Consequently, it seems safe to conclude thatthese learners are at the VP(i)-stage as defined by VYS (1996b).

In sum, the results of the elicited production task indicate that the L2 learnersstudied here can be clearly situated at the first stage of development within theMinimal Trees account. Thus they constitute the right population for testing thehypotheses at stake.

5.3.2. Picture interpretation. Replies to the ambiguous constituent wh-questions in both test conditions (present tense, perfect tense) were coded as fourtypes: (i) subject-question interpretation checked only (“subject only”), (ii) ob-ject-question interpretation checked only (“object only”), (iii) both subject- and


TABLE 4Headedness of Verb Phrase

Learner Total VO VXTotal

V-Initial VP OV XVTotal

V-Final VP

E1 2 2 0 2 0 0 0E2 3 1 2 3 0 0 0E3 2 2 0 2 0 0 0E4 10 8 2 10 0 0 0E6 3 3 0 3 0 0 0E7 3 2 1 3 0 0 0E8 1 1 0 1 0 0 0E9 6 2 4 6 0 0 0E10 2 2 0 2 0 0 0E11 1 1 0 1 0 0 0E12 2 2 0 2 0 0 0E13 3 2 1 3 0 0 0E14 8 5 3 8 0 0 0E15 2 2 0 2 0 0 0E16 7 6 1 7 0 0 0E17 1 1 0 1 0 0 0E18 3 2 1 3 0 0 0Total 59 44 15 59 0 0 0

Note. Compare VYS (1996a, 157, Table 7). VO = main verb followed by object; VX = main verbfollowed by constituent other than object; OV = main verb preceded by object; XV = main verb pre-ceded by constituent other than object.

1717As in VYS (1996a), only main verbs were considered, and utterances consisting of a subject and

a verb only were excluded.

object-question interpretations checked (“both”), and (iv) neither.18 Figures 2 and3 show the results for the L1 control group and the L2 group, respectively.

The native speakers supplied both possible answers at a rate of 40.7% in thepresent-tense condition, and 45.3% in the perfect-tense condition. In the caseswhere only one answer was provided, the object-question interpretation was pro-vided significantly more often than the subject-question one, in both tense condi-

308 GRÜTER

FIGURE 2 Results for the L1 control group (n = 15).

FIGURE 3 Results for the L2 group (n = 17).

1818Type (iv) theoretically comprises both incorrect and “not sure” answers. As no incorrect and only

a single “not sure” reply was recorded, it was decided to collapse the two.

tions (present tense: 43.3% object questions vs. 16.0% subject questions; perfecttense: 47.3% object questions vs. 7.3% subject questions). The L2 learners neversupplied both possible answers. In the present-tense condition, their preferred in-terpretation was a subject question (71.2% vs. 28.8% object questions), whereasin the perfect tense, they supplied almost exclusively object-question interpreta-tions (97.1% vs. 2.4% subject questions).

Frequencies of answer types were analyzed in a three-way Learner (L1, L2) ×Tense Condition (present, perfect) × Answer Type (subject only, object only,both, neither) ANOVA. The main effect for answer type was significant, F(3, 90)= 23.32, p � .0001, as were all interaction effects.19 In particular, the three-way in-teraction between the variables was significant, F(3, 90) = 34.24, p � .0001. Sub-sequent Tukey pairwise comparisons (� = .05) revealed significant differences foranswer type according to tense condition for the L2 group only: Whereas the sub-ject-question interpretation was their most frequent answer type in the present-tense condition, an object-question interpretation was by far their most favoredanswer type in the perfect-tense condition. For the L1 control group, on the otherhand, no significant differences between the two tense conditions were found.20

As grammars are properties of individuals rather than groups, it needs to beconfirmed that the distribution of response types within groups is mirrored by thedistribution of response types within individual learners (see Appendix B for indi-vidual results). This is clearly the case for the L2 group. With the exception of twolearners who consistently gave object-only replies in both conditions (see belowfor a bias in this direction found in the L1 group), individual results reflect thegroup result: 15 of the 17 L2 learners gave subject-only replies in the present-tense condition on at least 5 of 10 items (mode: 10/10), whereas no L2 learnergave a subject-only reply in the perfect-tense condition more than twice (mode: 0/10).

The distribution of responses within the L1 group, on the other hand, is bi-modal. Six (out of 15) learners checked both possible answers on 16 or more (outof 20) items, suggesting that they were aware of the questions’ ambiguity. An-other 6 learners never gave both possible answers, indicating that they wereunaware of the ambiguity. Of these 6, 4 had a clear overall preference for the ob-ject-question interpretation (16 or more out of 20), whereas the other 2 seem tohave chosen between subject- and object-question interpretations randomly (inboth tense conditions). The remaining 3 learners gave both possible answers occa-sionally (4/20, 4/20, 9/20). One of these learners checked both possible answerson the first 4 items, then gave object-only replies on the remaining 16 items. For


1919Due to the design of the study, no main effects for learner and tense condition could be calcu-

lated.20

20An additional three-way Tense × Answer Type × Order of Presentation (Version A, Version B)ANOVA revealed marginal effects for order of presentation. However, separate Learner × Tense ×Answer Type ANOVAs for each order of presentation mirror the overall effects in both versions of thetask, indicating that the overall results are not due to one order of presentation only.

the other two, the response pattern appears to be random. Thus three subgroupscan be identified within the L1 group: (i) those aware of the ambiguity overall (n =6), (ii) those with a clear preference for object-question interpretations overall(16+/20, n = 5), and (iii) those whose performance appears random (n = 4). It isimportant to note here that this within-group variation is independent of tensecondition. In other words, the crucial aspect of the group result—no difference be-tween the two tense conditions—is mirrored in the individual results.

5.4. Discussion

Having established through the elicited production task that the L2 learners stud-ied here represent the right population for testing the hypotheses at stake—that is,they can be placed at the first VP-stage within the Minimal Trees account—the re-sults from the picture identification task bear directly on the predictions of the twoinitial state hypotheses previously outlined (see Table 1).

For the present-tense condition, both Full Transfer and Minimal Trees predictL2 learners to choose the subject-question interpretation. This prediction wasborne out: As shown in Figure 3, L2 learners chose a subject-question interpreta-tion for questions in the present tense in 71.2% of cases. In only 28.8% of casesdid they choose an object-question interpretation. No L2 learner—unlike nativespeakers (a further discussion is next)—ever checked both possible answers forthese questions, indicating that they were not aware of their ambiguity.21

For the perfect-tense condition, the two hypotheses make different predictions.Whereas Full Transfer clearly expects an object-question interpretation, MinimalTrees predicts that questions in this condition cannot be parsed syntactically. Con-sequently, Minimal Trees might expect guessing behavior, or, if in the absence ofsyntactic representations a nonsyntactic, linear-order-based strategy is employed,a potential bias toward a subject-question interpretation. As illustrated in Figure3, L2 learners overwhelmingly (97.1%) chose an object-question interpretation inthis condition. This result lends clear and strong support to the Full Transfer hy-pothesis of the initial state. The Minimal Trees hypothesis, on the other hand, can-not account straightforwardly for these learners’ performance in the perfect-tensecondition.

Having discussed the rather clear evidence presented by the results from the L2learners, the performance of the L1 control group requires some comment. Asshown in Figure 2, the crucial aspect of their performance is the fact that—unlikethe L2 group—they performed identically in the two tense conditions. This resultindicates that the split between the present- and the perfect-tense condition ob-

310 GRÜTER

2121This is in contrast to their performance on potentially ambiguous distractor questions (see foot-

note 12). For 42.5% of these, L2 learners checked both possible answers, indicating that they did con-sider this answer type as an option.

served in the L2 group is uniquely attributable to learner group and does not resultfrom factors inherent to the task itself.

One might wonder, however, why although we assume all test questions to beambiguous in German, native speakers checked both possible answers in only40% to 45% of cases (see Figure 2). Moreover, in the cases where they did notcheck both answers, object-question interpretations were favored over subject-question interpretations in both tense conditions (present tense: 43.3% objectquestions vs. 16.0% subject questions; perfect tense: 47.3% object questions vs.7.3% subject questions). This performance suggests that there is an overall biastoward an object-question interpretation in this experiment, which was not ex-pected. A possible explanation for this phenomenon might be found in theanimacy of the NPs involved. It has been shown that in the absence of other cues,animate nouns are more readily interpreted as agents than nonanimate nouns(Kempe and MacWhinney (1999)). The two NP constituents in the wh-questionsof the present experiment—the animal term and the wh-phrase was (‘what’)—clearly differ with regard to animacy: Whereas the animal term is clearly [+ani-mate], the wh-phrase was can have both animate and inanimate referents (see alsofootnote 4). If speakers indeed use animacy to resolve ambiguity in these ques-tions, they will prefer to associate the [+animate] animal term with the agent, andthus the subject, which leaves the wh-phrase for the patient, resulting in an object-question interpretation.

If this explanation is on the right track, the observed object-question bias ispredicted to disappear if the target questions are reformulated as in (22), whereboth DPs are equally [+animate].

(22) Welches Tier jagt das Kamel?which animal chase-3SG the camel‘Which animal is chasing the camel?’ / ‘Which animal is the camel chas-ing?’

A replication of the present experiment using questions of this type should helpclarify—and eliminate—the influence of animacy-related properties on the experi-mental results presented here. This would be a desirable task for future research.22

Whether or not this explanation of the observed object-interpretation bias inthe L1 group is on the right track, it is important to emphasize that it extends overboth tense conditions and thus does not affect the relevance of these data for thehypotheses at stake. The split in the L2 group’s performance between the present-


2222An anonymous reviewer raised the possibility that the L2 learners’ preference for the object-

question interpretation may be the result of the same bias as that observed with the native speakers. Itwould be somewhat surprising, however, that this effect should only show in the perfect-tense condi-tion for the L2 group. Furthermore, if this is the case, the prediction would be that in a replication ofthe experiment using stimuli such as (22), L2 learners should no longer have a preference for the ob-ject-question interpretation. This appears unlikely but remains for future research to investigate.

and the perfect-tense conditions is entirely independent of this bias and thus con-tinues to present strong evidence against Minimal Trees and support for the FullTransfer hypothesis of the initial state in L2 acquisition.

6. CONCLUSION

The aim of this article was to construct an experiment that could provide compre-hension-based evidence bearing on the long-standing debate on the nature of theinitial state in L2 acquisition. Such evidence is scarce, as comprehension and in-terpretation tasks involving functional categories are not easy to construct. Thepicture interpretation experiment presented here, however, assessing the compre-hension of ambiguous constituent wh-questions in German, is argued to consti-tute precisely such a task. The performance of 17 English-speaking initial-statelearners of German on this experiment strongly suggests that functional structuretransferred from the L1 is implicated in their interpretation of these ambiguouswh-questions in their L2. Even without being more precise about which functionalcategories are involved exactly, these results present counterevidence to the Mini-mal Trees hypothesis of the initial state, which claims that no functional architec-ture beyond VP is in place in initial-state interlanguage grammars. On the otherhand, the results are in accordance with the Full Transfer hypothesis, which holdsthat the entire L1 grammar constitutes the L2 initial state. The results obtainedhere are explained with ease if the learners are using the full syntactic representa-tions of their L1.

The results presented in this article show clearly that some functional structuremust be transferred from the L1, thus they constitute evidence against the Mini-mal Trees hypothesis of the L2 initial state. However, based on these results, itcannot be argued conclusively that all functional structure is transferred—as pro-posed by the Full Transfer hypothesis. A relatively recent proposal by Bhatt andHancin-Bhatt (2002) maintains that the L2 initial state is characterized by L1-transfer of only the IP-system (and below). The present study was not set up todistinguish between this proposal and Full Transfer. Further evidence bearing onthe amount of functional structure transferred could be obtained by varying the L1of the learners involved in the experiment presented in this article. For this pur-pose, the experiment has recently been conducted with Afrikaans-speaking learn-ers of German (see Grüter and Conradie (2006) for further details). Although thedata presented here have established clearly that some functional structure mustbe transferred, it remains for such future work to determine if all of it is trans-ferred into initial-state L2 grammars.

ACKNOWLEDGMENTS

This research was supported by FQRSC (2001-ER-66973) and SSHRCC (410-2001-0719) grants (to Lydia White et al.) and McGill graduate fellowships to the

312 GRÜTER

author. Earlier versions of this article were presented at CLA 2003 and GASLA2004 and appeared in the McGill Working Papers in Linguistics 18(2).

I thank Lydia White for her support and guidance at all stages of this project. Iam grateful to Robert Bracewell for advice on statistics; to the editors and threeanonymous reviewers of Language Acquisition, whose comments have signifi-cantly improved this article; and to all participants who took part in the experi-ments. Any remaining errors are entirely my own.

REFERENCES

Bhatt, R. and B. Hancin-Bhatt (2002) “Structural Minimality, CP and the Initial State in Second Lan-guage Acquisition,” Second Language Research 18, 348–392.

Bley-Vroman, R. (1990) “The Logical Problem of Foreign Language Learning,” Linguistic Analysis20, 3–49.

Chomsky, N. (1981) Lectures on Government and Binding: The Pisa Lectures (Studies in GenerativeGrammar, 9), Foris, Dordrecht, The Netherlands.

Chomsky, N. (1986) Barriers, MIT Press, Cambridge, Massachusetts.Chomsky, N. (1995) The Minimalist Program, MIT Press, Cambridge, Massachusetts.Clahsen, H. and P. Muysken (1986) “The Availability of Universal Grammar to Adult and Child

Learners: A Study of the Acquisition of German Word Order,” Second Language Research 2,93–119.

den Besten, H. (1983) “On the Interaction of Root Transformations and Lexical Deletive Rules,” in W.Abraham, ed., On the Formal Syntax of the Westgermania: Papers from the “3rd Groningen Gram-mar Talks” Groningen, January 1981 (Linguistik Aktuell 3), John Benjamins, Amsterdam, TheNetherlands.

DUDEN (1998) Grammatik der deutschen Gegenwartssprache, Dudenverlag, Mannheim, Germany.Epstein, S., S. Flynn, and G. Martohardjono (1996) “Second Language Acquisition: Theoretical and

Experimental Issues in Contemporary Research,” Brain and Behavioral Sciences 19, 677–758.Epstein, S., S. Flynn, and G. Martohardjono (1998) “The Strong Continuity Hypothesis: Some Evi-

dence Concerning Functional Categories in Adult L2 Acquisition,” in S. Flynn, G. Martohardjono,and W. O’Neil, eds., The Generative Study of Second Language Acquisition, Lawrence ErlbaumAssociates, Inc., Mahwah, New Jersey.

Gass, S. (1984) “A Review of Interlanguage Syntax: Language Transfer and Language Universals,”Language Learning 34, 115–132.

Gavruseva, L. and D. Lardiere (1996) “The Emergence of Extended Phrase Structure in Child L1 Ac-quisition,” in A. Stringfellow, D. Cahana-Amitay, E. Hughes, and A. Zukowski, eds., Proceedingsof the 20th Annual Boston University Conference on Language Development, Cascadilla Press,Somerville, Massachusetts.

Grodzinsky, Y. (1995) “Trace Deletion, �-roles, and Cognitive Strategies,” Brain and Language 51,469–497.

Grondin, N. and L. White (1996) “Functional Categories in Child L2 Acquisition of French,” Lan-guage Acquisition 5, 1–34.

Grüter, T. and S. Conradie (2006) “Investigating the L2 Initial State: Additional Evidence From theProduction and Comprehension of Afrikaans-Speaking Learners of German,” in R. Slabakova,S. A. Montrul, and P. Prévost, eds., Inquiries in Linguistic Development: In Honor of Lydia White,John Benjamins, Amsterdam, The Netherlands.

Guilfoyle, E. and M. Noonan (1992) “Functional Categories and Language Acquisition,” CanadianJournal of Linguistics 37, 241–272.


Hawkins, R. (2001) Second Language Syntax: A Generative Introduction, Blackwell, Oxford, Eng-land.

Hawkins, R. and Y.-H. Chan (1997) “The Partial Availability of Universal Grammar in Second Lan-guage Acquisition: The ‘Failed Functional Features Hypothesis’,” Second Language Research 13,187–226.

Haznedar, B. (1997) “L2 Acquisition by a Turkish-Speaking Child: Evidence for L1 Influence,” in E.Hughes, M. Hughes, and A. Greenhill, eds., Proceedings of the 21st Annual Boston University Con-ference on Language Development, Cascadilla Press, Somerville, Massachusetts.

Kempe, V. and B. MacWhinney (1999) “Processing of Morphological and Semantic Cues in Russianand German,” Language and Cognitive Processes 14, 129–171.

Kitagawa, Y. (1986) Subjects in Japanese and English, Doctoral dissertation, University of Massa-chusetts, Amherst.

Klein, W. and C. Perdue (1997) “The Basic Variety (or: Couldn’t Natural Languages Be Much Sim-pler?),” Second Language Research 13, 301–347.

Koster, J. (1975) “Dutch as an SOV Language,” Linguistic Analysis 1, 111–136.Lado, R. (1957) Linguistics Across Cultures, University of Michigan Press, Ann Arbor.Lardiere, D. (1998a) “Case and Tense in the ‘Fossilized’ Steady State,” Second Language Research

14, 1–26.Lardiere, D. (1998b) “Dissociating Syntax From Morphology in a Divergent End-State Grammar,”

Second Language Research 14, 359–375.Meisel, J., H. Clahsen, and M. Pienemann (1981) “On Determining Developmental Stages in Natural

Second Language Acquisition,” Studies in Second Language Acquisition 3, 109–135.Myles, F. (2005) “The Emergence of Morpho-Syntactic Structure in French L2,” in J.-M. Dewaele,

ed., Focus on French as a Foreign Language: Multidisciplinary Approaches, Multilingual Matters,Clevedon, England.

Pienemann, M. (1998) Language Processing and Second Language Development: Processability The-ory, John Benjamins, Amsterdam, The Netherlands.

Platzack, C. (1996) “The Initial Hypothesis of Syntax: A Minimalist Perspective on Language Ac-quisition and Attrition,” in H. Clahsen, ed., Generative Perspectives on Language Acquisition:Empirical Findings, Theoretical Considerations, Crosslinguistic Comparisons, John Benjamins,Amsterdam, The Netherlands.

Radford, A. (1988) “Small Children’s Small Clauses,” Transactions of the Philological Society 86,1–46.

Radford, A. (1996) “Towards a Structure-Building Model of Acquisition,” in H. Clahsen, ed., Genera-tive Perspectives on Language Acquisition: Empirical Findings, Theoretical Considerations,Crosslinguistic Comparisons, John Benjamins, Amsterdam, The Netherlands.

Schwartz, B. D. and L. Eubank (1996) “What Is the ‘L2 Initial State’?,” Second Language Research12, 1–5.

Schwartz, B. D. and R. Sprouse (1994) “Word Order and Nominative Case in Nonnative LanguageAcquisition: A Longitudinal Study of (L1 Turkish) German Interlanguage,” in T. Hoekstra andB. D. Schwartz, eds., Language Acquisition Studies in Generative Grammar, John Benjamins, Am-sterdam, The Netherlands.

Schwartz, B. D. and R. Sprouse (1996) “L2 Cognitive States and the Full Transfer/Full AccessModel,” Second Language Research 12, 40–72.

Slabakova, R. (2000) “L1 Transfer Revisited: The L2 Acquisition of Telicity Marking in English bySpanish and Bulgarian Native Speakers,” Linguistics 38, 739–770.

Stromswold, K. (1995) “The Acquisition of Subject and Object Wh-Questions,” Language Acquisition4, 5–48.

Vainikka, A. and M. Young-Scholten (1994) “Direct Access to X�-theory: Evidence From Korean andTurkish Adults Learning German,” in T. Hoekstra and B. D. Schwartz, eds., Language AcquisitionStudies in Generative Grammar, John Benjamins, Amsterdam, The Netherlands.

Vainikka, A. and M. Young-Scholten (1996a) “The Early Stages in Adult L2 Syntax: Additional Evi-dence From Romance Speakers,” Second Language Research 12, 140–176.

314 GRÜTER

Vainikka, A. and M. Young-Scholten (1996b) “Gradual Development of L2 Phrase Structure,” Sec-ond Language Research 12, 7–39.

White, L. (1985) “The Pro-Drop Parameter in Adult Second Language Acquisition,” LanguageLearning 35, 47–62.

White, L. (2003) Second Language Acquisition and Universal Grammar, Cambridge UniversityPress, Cambridge, England.

Submitted 13 October 2004Final version accepted 25 October 2005

APPENDIX A

Experimental Stimuli

Present-tense condition

Picture 1 (verb: jagen ‘to chase’)1. Was ist vor dem Schaf? (distractor)2. Was jagt die Schlange?3. Was jagt das Kamel?4. Was ist hinter dem Schwein? (distractor)

Picture 2 (verb: tragen ‘to carry’)5. Was trägt das Schwein?6. Was ist unter der Schlange? (distractor)7. Was ist auf der Schildkröte? (distractor)8. Was trägt das Schaf?

Picture 3 (verb: schlecken ‘to lick’)9. Was ist neben der Katze? (distractor)

10. Was schleckt die Kuh?11. Was schleckt das Schaf?12. Was ist neben dem Pferd? (distractor)

Picture 4 (verb: stossen ‘to push’)13. Was stösst die Ente?14. Was ist vor dem Schwein? (distractor)15. Was ist hinter dem Schaf? (distractor)16. Was stösst die Katze?

Picture 5 (verb: beissen ‘to bite’)17. Was ist hinter der Kuh? (distractor)18. Was beisst das Kamel?19. Was beisst die Katze?20. Was ist vor der Schlange? (distractor)


Perfect-tense condition

Picture 1 (verb: jagen ‘to chase’)21. Was hat das Schwein gejagt?22. Was hat das Schaf gejagt?

Picture 2 (verb: tragen ‘to carry’)23. Was hat die Schlange getragen?24. Was hat die Schildkröte getragen?

Picture 3 (verb: schlecken ‘to lick’)25. Was hat die Katze geschleckt?26. Was hat das Schwein geschleckt?

Picture 4 (verb: stossen ‘to push’)27. Was hat das Schaf gestossen?28. Was hat das Schwein gestossen?

Picture 5 (verb: beissen ‘to bite’)29. Was hat die Kuh gebissen?30. Was hat die Schlange gebissen?

APPENDIX B

Picture Interpretation: Individual Results

TABLE B1L1 Group


S Only O Only Both Neither S Only O Only Both Neither

D1 0 1 9 0 0 0 10 0D2 4 4 2 0 0 8 2 0D3 0 10 0 0 0 10 0 0D4 0 0 10 0 2 0 8 0D5 2 8 0 0 0 10 0 0D6 5 4 1 0 0 2 8 0D7 0 1 9 0 0 0 10 0D8 4 0 6 0 0 0 10 0D9 0 10 0 0 0 10 0 0D10 0 0 10 0 0 0 10 0D11 2 8 0 0 2 8 0 0D12 0 6 4 0 0 10 0 0D13 3 7 0 0 3 7 0 0D14 0 0 10 0 0 0 10 0D15 4 6 0 0 4 6 0 0Total 24 65 61 0 11 71 68 0% 16.0 43.3 40.7 0.0 7.3 47.3 45.3 0.0

Note. n = 15. Values are number of response types in each condition.

316 GRÜTER

TABLE B2L2 Group


S Only O Only Both Neither S Only O Only Both Neither

E1 7 3 0 0 0 10 0 0E2 10 0 0 0 0 10 0 0E3 0 10 0 0 0 10 0 0E4 10 0 0 0 0 10 0 0E6 7 3 0 0 0 9 0 1E7 1 9 0 0 2 8 0 0E8 10 0 0 0 0 10 0 0E9 6 4 0 0 0 10 0 0E10 10 0 0 0 0 10 0 0E11 8 2 0 0 0 10 0 0E12 10 0 0 0 0 10 0 0E13 5 5 0 0 0 10 0 0E14 8 2 0 0 0 10 0 0E15 10 0 0 0 0 10 0 0E16 5 5 0 0 0 10 0 0E17 9 1 0 0 2 8 0 0E18 5 5 0 0 0 10 0 0Total 121 49 0 0 4 165 0 1% 71.2 28.8 0.0 0.0 2.4 97.1 0.0 0.6

Note. n = 17. Values are number of response types in each condition.


Another Take on the L2 Initial State: Evidence from Comprehension in L2 German

Documents