Published in: Language Acquisition 13(1), 3-32 Copyright ©2005/2006, Lawrence Erlbaum Associates, Inc Structural and Lexical Case in Child German: Evidence from language-impaired and typically-developing children Sonja Eisenbeiss, Susanne Bartke, Harald Clahsen* *Corresponding Author: Harald Clahsen Department of Linguistics University of Essex Colchester CO4 3SQ United Kingdom Tel.: 0044-1206-87-2228 Email: [email protected] 2 Abstract This study examines the system of case marking in two groups of German-speaking children, five children with Specific Language Impairment (SLI) and five typically-developing (TD) children matched to the SLI children on a general measure of language development. The data from both groups demonstrate high accuracy scores for structural case marking and overapplications of structural cases to instances that require lexical case marking in the adult language. These results, we argue, provide evidence for the sensitivity of both TD and SLI children for abstract, structure-based regularities, and is incompatible with accounts of SLI that posit broad syntactic deficits for these children. 3 1. Introduction Syntactic accounts of case marking posit a distinction between structural and lexical cases (see e.g. Chomsky 1981:170ff) according to which structural cases are checked or assigned in particular structural positions whereas lexical case marking is based on the (case-assigment) specifications of particular lexical items (also labeled ‘quirky’ case) or associated with particular thematic roles (also labeled ‘inherent’ case). Many languages exhibit a distinction between structural and lexical case marking. German, for example, has nominative subjects and accusative objects which are both regarded as structural cases (as in English), but the direct objects of, for example, verbs such as danken 'to thank' and helfen 'to help' require datives (rather than accusatives) indicating that these lexemes have dative specifications which take precedence over any structural case marking. Likewise, there is a group of intransitive verbs that assign dative (rather than nominative) to their Experiencer argument, e.g. grauen ‘to be terrified of sth.’ These latter examples are instances of lexical case marking which cannot be predicted from the structural configuration in which an NP occurs. In the acquisition literature, children’s case marking systems have been studied for a variety of languages (see e.g. English: Schütze and Wexler 1996, Russian: Babyonyshev 1993, German: Clahsen 1984, Finnish: Clahsen et al. 1994). The question, however, of potential differences between structural and lexical case marking in children’s grammars has received very little attention. One notable exception is Babyonyshev’s (1993) study of the spontaneous speech of two Russian speaking children in the age range of 2;1 to 2;7 and 1;6 to 2;0 in which she found few mistakes in the use of structural cases and relatively many in the use of lexical cases (Babyonyshev 1993: 27). She argues that the structural case assignment mechanism of the adult grammar is already operative in two-year olds, whereas the acquisition of lexical case marking is delayed due to the idiosyncratic properties of lexical case assignment which have to be learnt on an item-by-item basis. 4 Case marking has also been studied for children with Specific Language Impairment (SLI), a (probably genetically determined) delay and/or disorder of the normal acquisition of grammar in the absence of neurological trauma, cognitive impairment, psycho-emotional disturbance, or motor-articulatory disorders (see Leonard 1998, Levy and Kavé 1999, Clahsen 1999 for review). Most previous studies on case marking in SLI dealt with English-speaking children (Loeb and Leonard 1991, Leonard 1995, Schütze 1997, Clahsen et al. 1997, Wexler et al. 1998, Radford and Ramos 2001). It was found that in the speech of SLI children, (accusative) case marking on objects is mostly correct, whereas subject pronouns often appeared in accusative case instead of the required nominative, e.g. When me go outside to play (Radford 2001:67). For German SLI children, Clahsen (1991) also found many case-marking errors, particularly overapplications of nominative forms in accusative and dative contexts. Some researchers have taken these errors as indications of a general impairment in the underlying syntactic representations of SLI individuals. Clahsen (1989, 1991), for example, argued that the grammars of SLI individuals are impaired in the grammatical mechanism, i.e. the ControlAgreement Principle (CAP, see Gazdar et al. 1985), that is responsible for matching grammatical features of different syntactic categories within a sentence. On a similar note, van der Lely and colleagues have claimed that individuals with SLI have ‘a deficit with building non-elementary complex syntactic dependencies between constituents’ (van der Lely and Stollwerck 1997: 283). Other researchers have argued that functional categories (CP, IP, DP, etc.) are particularly vulnerable in SLI (e.g. Eyer and Leonard 1995, Guilfoyle, Allen and Moss 1991, Leonard 1995, 1998). In these accounts, the case errors seen in SLI children are taken to reflect a broad syntactic deficit. An alternative proposal is that the underlying syntactic competence is basically intact and that selective areas of the grammar are not fully specified for children with SLI. The most wellknown proposal of this kind is the Extended Optional Infinitive hypothesis of Rice, Wexler 5 and their collaborators (Rice et al. 1995, Rice and Wexler 1996, Wexler, Schütze and Rice 1998) according to which tense and/or agreement features are optionally omitted from the syntactic representations of sentences produced by SLI children. On the other hand, those grammatical morphemes and elements of syntactic structure that are unrelated to tense and/or agreement are ‘not affected, at least not beyond the delays attributable to the initially slow emergence of language’ (Rice 1999: 349). A related proposal is the ‘Agreement-Deficit’ hypothesis of Clahsen and colleagues (Clahsen, Bartke, Göllner 1997, Clahsen and Dalalakis 1999) which posits an even more selective syntactic deficit for SLI. Adopting Chomsky's (1995) theory of formal features which distinguishes between interpretable features, i.e., features relevant for semantic interpretation, and non-interpretable ones that are irrelevant for interpretation, it is argued that individuals with SLI are impaired in dealing with noninterpretable features. Clahsen and colleagues pursued a narrow interpretation of the Agreement Deficit hypothesis according to which ‘just the optional non-interpretable φfeatures of verbs, i.e. the verb’s agreement features are affected’ (Clahsen et al. 1997: 153) leaving other formal features including case, tense, and categorial features intact. Empirical support for this hypothesis comes, for example, from the finding that subject-verb agreement causes difficulty for SLI subjects across different languages and different age groups, even for children for whom tense marking functions normally (see Clahsen and Dalalakis 1999). What is common to the Extended Optional Infinitive and the Agreement Deficit hypotheses is that they posit relatively narrow impairments for SLI in the domain verb finiteness marking. The case errors seen in English-speaking SLI children (i.e. non-nominative subject pronouns) receive a different interpretation from these accounts. It is proposed that these errors are a consequence of SLI children’s difficulties with finiteness marking. Given that nominative case is assigned by an agreement feature in T(ense), Schütze (1997) and Wexler et al. (1998) have argued that in root-infinitive clauses produced by SLI children, specifically in those in 6 which agreement is left unspecified, nominative case cannot be assigned, yielding sentences with null subjects or subjects in the default (accusative) case such as (him) fall down. This interpretation raises doubts on whether case marking is indeed genuinely impaired in SLI and forms part of a broad syntactic deficit. Questions also arise with respect to the case errors seen for German-speaking SLI children. As Clahsen (1991) did not distinguish between structural and lexical case, it is not clear from this study how many of the errors were in contexts that required lexical (rather than structural) case marking. Lexical case marking errors could be due to the delayed learning of lexical exceptions, similarly to what Babyonyshev (1993) argued for Russian child language, and not necessarily be indicative of a structural deficit for SLI children in the domain of case marking. The present study has two main aims. Firstly, we will determine whether children with SLI have genuine case-marking deficits. Answers to this question bear on our understanding of the nature of SLI. Case marking is indeed a core component of grammar which involves a syntactic dependency between a case assigner (e.g. a verb) and a case-marked constituent (e.g. a subject or an object). In Gazdar et al.’s (1985) terms, case marking falls under the ControlAgreement Principle, and in Chomskian theory (e.g. Chomsky 1995), it involves functional categories. Thus, if SLI involves one of the general syntactic deficits mentioned above (‘control agreement’, ‘functional categories’, ‘dependent relations’), children with SLI should exhibit impairments in the domain of case marking. However, as seen above, the available evidence for this is ambiguous and there are alternative interpretations. The second aim of the present study is to examine whether the contrast between structural and lexical case marking that Babyonyshev (1993) discovered can be replicated for German child language. We would expect that case errors in TD children are restricted to lexical case marking, e.g. to lexical datives in German, whereas structural case marking, e.g. nominative for subjects and accusative for direct objects, is largely error-free. 7 2. Linguistic background: a comparison of case marking in English and German In the domain of case marking, German differs in various ways from English. It has been argued (Emonds 1985, Schütze 1997) that in English accusative case functions as the default in that it is used in contexts in which there is no overt case assigner, e.g. for dislocated topics (Me, I like sailing), answers to subject questions, e.g. Me as an answer to Who likes sailing?, so-called mad magazine sentences (Me give up sailing, never!’). By contrast, in German (as indeed in most nominative/accusative languages), the nominative (rather than the accusative) is used in such circumstances, e.g. in mad magazine sentences (Ich das Auto putzen, kommt nicht infrage ‘Me cleaning the car, no way’). Therefore, Emonds (1985) and Schütze (1997) argued that nominative is the default case in German. Additionally, German has accusative and dative case, a prenominal possessive genitive which is marked on proper nouns or kinship terms denoting the possessor (as in Marias/Mamas auto 'Maria's/mommy's car'), and a postnominal genitive (e.g. die Blumen der Frau 'the flowers of the-Gen. woman'); the latter is rarely used in the spoken language and not attested in the data under study; see also Mills (1985:185ff.) who reported that none of the TD children studied in the early diary studies used the postnominal genitive before the age of six years. The second property of the German case marking system that is not present in English is the distinction between structural and lexical case marking. As in English, nominative subjects, accusative objects, and the prenominal genitive are regarded as instances of structural case marking in German (see e.g. Haider 1988, Olsen 1990). In addition and unlike in English, German requires lexical case marking, e.g. for transitive verbs such as danken 'to thank' and helfen 'to help' which take datives (rather than accusatives) or for intransitives such as grauen ‘to be terrified of sth.’ that assign dative (rather than nominative) to the Experiencer argument. In such instances, lexical case specifications override structural case assignment. The question, however, of whether case marking on indirect objects and complements of 8 prepositions is lexical or structural, is controversial in the syntactic literature. Some researchers have argued that dative case marking is always lexical (e.g. Haider 1985, Haegeman 1991, Heinz and Matiasek 1994). Their main point is that (as in English) accusative alternates with nominative in German passivization (1a vs. 1b), whereas the dative does not change in passives (1c vs. 1d), suggesting that (unlike accusative) dative case is not affected by structural changes such as those involved in passivization. On the other hand, there is a construction in German, the so-called recipient passive (also labeled ‘indirect passive’, see Kubo 1992), in which dative case marking does indeed alternate with nominatives, much like the accusative-nominative alternation in passivization (1e vs. 1f). This, together with the fact that dative case marking on indirect objects is fully predictable and productive suggests that indirect dative objects are instances of structural case marking (see e.g. Czepluch 1988, Wegener 1990, Gallmann 1992, Schmidt 1995, Wunderlich 1997). (1) a. Der Affe beißt den Clown. The monkey bites theACC clown. b. Der Clown wird gebissen. TheNOM clown gets bitten. c. Der Affe winkt dem Clown. The monkey waves at-theDAT clown. d. Dem Clown wird gewunken. At-theDAT clown gets waved. e. Der Affe klaut dem Clown den Hut. 'The monkey steals theDAT clown the hat.' f. Der Clown kriegt den Hut geklaut. 'TheNOM clown gets the hat stolen.' 9 Prepositions assign accusative, dative, or genitive case in German. For some researchers (e.g. Haegemann 1991, Heinz and Matiasek 1994, Müller 1999), all of these are instances of lexical case marking. Others claim that dative is the structural case assigned by prepositions and that accusative-marked complements of prepositions are lexical exceptions (see e.g. Bierwisch 1988), a view that is compatible with the fact that in corpora dative is more frequent than accusative in complements of prepositions (Folsom 1984). A third alternative is to analyze case marking in terms of categorial features. Accusative never occurs on complements of nouns or adjectives. Instead, nouns assign genitive (die Frau [meines Nachbarn]GEN 'the wife of my neighbour') and adjectives genitive or dative case to their complements (e.g. Sie fühlt sich [[ihres Mannes]GEN sicher] 'She feels secure about her husband’; Sie ist [[ihrem Mann]DAT true] 'She is faithful to her husband'). Verbs and prepositions, on the other hand, may assign accusative case to their complements. These differences in case assignment coincide with other morpho-syntactic differences. For instance, nouns and adjectives can carry case markers, whereas verbs and prepositions are never casemarked. These distributional differences can be captured in terms of the distinction between [+N] categories (= adjectives and nouns), and [-N] categories (= verbs and prepositions), and Eisenbeiss (2002) and Stiebels (2002) argued that the most economical analysis of the German case marking system is to treat accusative as the structural complement case assigned by syntactic categories containing the feature [-N], i.e. by verbs and prepositions. A consequence of this is that dative and genitive case on complements of prepositions are instances of lexical case. The third property that distinguishes case marking in German and English concerns the morphological spell-out. While English has a small number of suppletive case-marked pronouns, German has regular paradigms for case marking (together with gender and number) 10 on determiners and attributive adjectives. Nominative, accusative, and dative are marked on personal pronouns, determiners, wh-words, quantifiers, possessive pronouns, and adjectives in German. In addition, some masculine nouns carry the accusative/dative suffix –n in the singular (e.g. für den junge-nACC 'for the boy', mit dem junge-nDAT 'with the boy') and some plural forms of nouns take -n in dative plural contexts (e.g. mit den kind-erPL-nDAT 'with the children'). Personal pronouns and definite articles are suppletive and form the word-specific paradigms shown in (2) and (3). The italicized forms are non-syncretic, i.e. they unambiguously encode a particular case for a given gender/number. (2) Personal pronouns 1.ps. 2.ps. nominative accusative singular ich mich mir plural wir uns uns singular du dich dir plural ihr euch euch masculine er ihn ihm neuter es es ihm feminine sie sie ihr sie sie ihnen singular 3.ps. plural (3) dative Definite articles singular plural masculine neuter feminine nominative der das die die accusative den das die die dative dem dem der den Case marking on demonstratives like dies- 'this', wh-words like welch- 'which', quantifiers like alle 'all', and attributive adjectives (in NPs without any other so-called strong determiner) 11 involves regular suffixes which encode case, gender, and number specifications (see (4)); again, unambiguous case markings are shown in italics. Indefinite articles, possessive pronouns, and the negation element kein- 'no' all have stems ending in -ein and will be referred to as ‘-ein-determiners’. When they are used as pronouns (e.g. das ist keiner 'that is none'), they are inflected according to the paradigm in (4). In attributive use, however, the paradigm in (5) applies, e.g. das ist kein hund 'that is no dog'. (4) Strong case marking on determiners and adjectives singular (5) plural masculine neuter feminine nominative dies-er dies-es dies-e dies-e accusative dies-en dies-es dies-e dies-e dative dies-em dies-em dies-er dies-en Case marking on determiners ending in -ein singular plural masculine neuter feminine nominative ein-0 ein-0 ein-e ein-e accusative ein-en ein-0 ein-e ein-e dative ein-em ein-em ein-er ein-en 3. Case marking in German child language The acquisition of German case marking in typically-developing (TD) children has been examined in a large number of studies (Clahsen 1984, Mills 1985, Tracy 1986, Parodi 1990, Clahsen et al. 1994, 1996, Schütze 1997, Eisenbeiss 1994, 2000, 2002, Szagun 2004, Wittek & Tomasello 2005). The picture that emerges from these studies is that in the two-word stage children do not yet use case markings contrastively. Instead, most noun phrases occur in their citation form (= nominative) without any clear case contrasts and prenominal genitives lack 12 the obligatory -s marker on the possessor. An early case system that distinguishes between nominative and non-nominative forms begins to develop during the third year of life (Clahsen 1984). The –s marker of the prenominal possessive genitive is amongst the earliest case markers to appear, but is initially restricted to a few proper names or kinship terms which children frequently hear with this marker in their input (e.g. mamas 'mommy's'; Clahsen et al. 1994, Eisenbeiss 2002). As soon as –s is used in all obligatory contexts, it is overextended to common nouns, which cannot be combined with this marker in the target language (e.g. *männers wagen 'men's car'). For the other cases, suppletive forms, specifically accusative and dative pronouns, appear before regularly inflected accusative and dative suffixes such as those on determiners and adjectives. Case markers on nouns are acquired late, which might be due to the fact that they are restricted to a small class of nouns and frequently omitted by adult native speakers (Indefrey 2002, Eisenbeiss 2002). Three main types of case errors have been observed in previous studies of case marking in German child language: (i) overgeneralizations of structural cases to forms that require lexical case in the adult language; (ii) omissions of cae markers on -ein-determiners, e.g. on indefinite articles (noch ein (correct: einen) fisch malen ‘another fish paint’ (= I want to paint another fish)); (iii) substitutions of the correct dative ending -m with the incorrect (accusative) ending -n (Julia auf den (correct: dem) bein sitzen ‘J. on the leg sit’). Overapplications of structural case marking have been reported, for example, for Experiencer arguments of intransitive verbs that (exceptionally) require dative case in adult German and in which children use nominative instead (Eisenbeiss 1994). Likewise, TD children overgeneralize accusative case to the direct objects of verbs such as winken 'to wave at' and danken 'to thank' that require lexical datives in the adult language (Mills 1985, Eisenbeiss 1991). These kinds of errors can be interpreted in syntactic terms. Children (like adults) associate certain phrase- 13 structure positions, e.g. the subject and direct object positions, with particular case markings (accusative and nominative) and overapply these structural cases when they have not yet acquired the lexical exceptions The two other types of case errors mentioned above, i.e. omissions of cae markers on -eindeterminers and -m → -n substitutions, can be explained in morpho-phonological terms. Recall that in adult German, -ein-determiners exhibit strong inflection when they function as pronouns (e.g. das ist keiner 'that is none') and that they often lack overt case affixes in attributive use (das ist kein hund 'that is no dog'). Children extend the use of uninflected -ein forms to contexts in which they are ungrammatical in the adult language thereby regularizing the paradigm of -ein-determiners. Given this interpretation, affix omissions on -eindeterminers are not indicative of any specific difficulty with case marking. Along similar lines, -m → -n substitutions (…auf den (instead of : dem) Bein …’on the leg’) could be due to the phonological similarity of these two forms, rather than to a confusion of dative and accusative case (Mills 1985, Eisenbeiss 1991, 2002, Szagun 2004). Evidence for this comes, for example, from the fact that the same children who substituted -n for -m supplied the correct dative markers -r for feminine singular and -n for plural nouns and did not overapply accusative forms in such cases (Eisenbeiss 1991). Case marking in German SLI children was investigated by Clahsen (1991). Nominative forms were found to be overapplied in 64% of the 102 accusative contexts and in 42% of the 62 dative contexts. Moreover, examples were reported of overgeneralizations of dative forms in nominative and accusative contexts as well as of overgeneralizations of accusative forms in nominative and dative contexts. Note, however, that since this study did not distinguish between structural and lexical case contexts, it not clear how many of the errors were overapplications of structural case marking (parallel to what has been found for TD children). 14 Moreover, incorrect uninflected -ein-determiners, -n → -m substitutions, and elliptical utterances without overt case-assigners were included Clahsen’s analysis, even though (according to considerations mentioned above) they may not signal any genuine case-marking difficulties. Clearly, more data and analyses are required before any general conclusions about case marking in German-speaking SLI children can be drawn. 4. The present study This study examines structural and lexical marking in five German-speaking SLI children and in five typically-developing (TD) children. With respect to the latter, we expect to replicate the findings from Babyonyshev’s (1993) study of two Russian children: TD children’s production of the structural cases of German should be largely error-free; errors should be restricted to instances of lexical case marking. In particular, we expect TD children to overapply the corresponding structural cases to contexts that require lexical case. Similarly, TD children might fail to produce accusative or dative markers on nouns, as these markers only occur on a small number of nouns which have to be learnt on an item-by-item basis. For case marking in SLI children, the different accounts mentioned in the Introduction make different predictions. Models of SLI that posit a broad syntactic deficit for ‘control agreement’ (Clahsen 1991), ‘functional categories’ (Leonard 1995), or ‘dependent relations’ (van der Lely and Stollwerck 1997) predict SLI children to be impaired in the domain of case marking. Specifically, their accuracy scores for the structural cases of German (e.g. for accusatives on direct objects) should be significantly lower than those of TD children. If, on the other hand, the underlying syntactic competence is basically intact in SLI and the syntactic impairment is confined to the verb-finiteness cluster (as proposed by the Extended Optional Infinitives and the Agreement-Deficit Hypotheses), we would expect that those structural cases that do not involve verb-finiteness features (e.g. accusatives on direct and datives on 15 indirect objects) are unaffected and that SLI children perform at the same level of accuracy and produce the same kinds of errors as TD children, e.g. overapplications of structural cases to items that require lexical case in the adult language. 5. Method An overview of the data we investigated is given in Table 1. Production data from five German-speaking SLI children and five TD control children were examined. Most of the recordings of both groups of children involved dinner-time conversations or free play. For some of the recordings with both groups, we presented children with cartoon-like drawings in a semi-naturalistic setting and encouraged them to talk about the properties and actions of the people and animals depicted in the drawings. Following much previous research (see Leonard 1998: 28f. for review), we used MLU for matching SLI and control children. The MLU serves as a general descriptor of young children’s language acquisition, and high levels of test-retest reliability have been reported for MLU scores derived from speech samples of TD children between 2;7 and 3;10 (Gavin and Giles 1996), which corresponds to the age range of the group of TD children tested here. As German is a fusional language with quite a bit of inflection, determining MLU scores for morphemes may lead to ambiguous or unclear decisions about the productivity of a particular word form (see Clahsen et al. 1994: 403). Instead, MLU scores were counted based on words. Statistical comparison confirmed that there were no between-group differences in the MLU values of the SLI and the TD groups (Z = 0.731; p = 0.55). For all analyses of case marking, we present individual subject data to allow for comparisons between individual TD and SLI children with similar MLU scores. We also examine developmental trends in the data for those children for whom several recordings are available. 16 The data from four of the five SLI children (David, Peter, Sebastian, Dieter) were previously presented and studied by Bartke (1998) and Clahsen et al. (1997) with respect to other phenomena. The data from three of the unimpaired controls (Leonie, Annelie, Svenja) come from the LEXLERN project (Clahsen, Vainikka, Young-Scholten 1990), Mathias from Clahsen (1982), and Carsten from the Wagner-corpus in the CHILDES database (Wagner 1985, MacWhinney 2000). //INSERT TAB.1 ABOUT HERE// The SLI children included in this study have been independently diagnosed by speech therapists as having SLI, i.e., while they are severely impaired in language development as measured by clinical tests of language abilities, they are not reported to show any obvious non-linguistic deficits. According to the clinicians’ reports, their non-verbal cognitive abilities fall within the normal limits for their chronological age, there are no hearing losses, obvious neurological dysfunctions or motor deficits (see Bartke 1998 for more information). The data from the five SLI children are based on 11 recordings of speech of about one hour each (age range: 5;8-7;11, MLU range: 2.6 - 4.2). Participants were the child and the speech therapist or another adult who the child was familiar with. The recordings took place in the institutions and clinics where the children were treated. In these recordings, the SLI children produced a total of 3,661 grammatically analyzable utterances which were transcribed and subsequently analyzed by members of our research group; see Tab.1 for the individual subject data. The data from the five unimpaired controls (age range: 2;6 - 3;6, MLU range: 2.5 - 4.2) are based on 31 recordings of speech. Participants were the child and the parents or another adult who the child was familiar with. The recordings took place in the children’s homes or day care centers. In these recordings, the unimpaired children produced a total of 9,461 analyzable 17 utterances which were transcribed and subsequently analyzed by members of our research group; see Tab.1 for the individual subject data. The data were analyzed as follows. For prenominal genitives, we only included unambiguous obligatory contexts: (i) noun phrases with two nouns that could be identified as possessor and possessum (e,.g. das ist mama(s) auto 'this is mommy('s) car'), (ii) utterances without a possessum, but with an overtly marked possessor (e.g. das ist mamas 'this is mommy's'). The analysis of nominative, accusative and dative markers only included sentences in which the child produced both an argument that required one of the above case markings as well as the corresponding case assigner, i.e. the case-assigning verb or preposition. Elliptical utterances without a verb or (in the case of PPs) with a preposition were excluded. Moreover, we carried out a separate analysis for case markers on nouns as the studies mentioned above have shown that these markers are acquired later than other case forms. We determined how frequently the accusative/dative singular marker –n appeared on nouns. We did not include –n in dative plural contexts as –n is also one of the five nominative plural suffixes in German, and it is not possible to determine whether –n in plural contexts is a plural or a dative marker. For similar reasons we did not analyze –n overgeneralizations on nouns in plural contexts, because in such circumstances it is not possible to decide whether the -n form is a (dative) case or a plural marker. The analysis of nominative, accusative and dative forms on determiners and pronouns only included unambiguous non-syncretic case markings, i.e. the forms shown in italics in the paradigms (2) to (5) above. Ambiguous forms, e.g. the 3rd fem.sg. and 3rd pl. forms of personal pronouns (= sie), were excluded from the analysis of direct accusative objects, because they are identical in nominative and accusative case. Our analysis focuses on case marking in (pronominally and attributively used) definite articles, demonstratives, quantifiers with strong inflection, wh-words, and personal pronouns. Case marking on -ein-determiners, 18 however, was not included. As mentioned above, TD children often omit case markings on -ein-determiners, even at a stage at which, in the same syntactic environments, case marking on other kinds of determiners and pronouns is correct. Case marking omissions on -eindeterminers are likely to be due to additional morphological complications specific to the paradigms of these determiners, rather than indicative of incorrect case marking. For subject forms, we determined accuracy scores for the suffix -r in nom.masc.sg. contexts and for unambiguously correct nominative pronoun forms. Subjects with the accusative suffix –n or the dative -m as well as non-nominative subject pronouns were counted as errors. In contexts in which accusative case was required in adult German, accuracy scores for the suffix -n in acc.masc.sg. contexts and for unambiguously correct accusative pronoun forms were determined; instances in which a determiner appeared with the nominative ending -r or the dative marker -m (instead of the required ending –n) as well as non-accusative pronouns in accusative-requiring contexts were treated as case errors. In contexts in which adult German requires dative case marking, we determined accuracy scores for the suffix -r in dat.fem.sg. contexts and for unambiguously correct dative pronoun forms. Corresponding nominative or accusative forms as well as non-dative pronouns in dative-requiring contexts were treated as errors; note, however, that due to the ambiguity of these errors, -n → -m substitutions were not included in these counts. 6. Results We first analyzed the data of the two groups of children with respect to prenominal genitives. Three of the five TD children and none of the SLI children had unambiguous obligatory contexts for prenominal genitives in their data. The three TD children produced 41 correctly inflected forms (Leonie: 16, Mathias: 5, Svenja: 20; see e.g. (6)), three overgeneralizations of 19 –s to common nouns (Leonie: 2, Svenja: 1; see e.g. (7)), and no omissions of –s. These results indicate that prenominal genitives are error-free in obligatory contexts. (6) a. Daniels kopf is härter (Mathias) ‘Daniel's head is tougher’ b. mamas kissen war hier (Svenja) ‘mommy's cushion was here’ (7) a. clowns hut (Leonie) ‘clown’s hat’ (uttered while packing the belongings of the clown) b. das is junges gürtel (Svenja) ‘this is boy’s belt’ The next analysis concerns accusative and dative case markings on nouns. In both groups of children, obligatory contexts for case suffixes on nouns were rare, 14 for the TD and 8 for the SLI group. In the TD group, there was just one correct nominal case suffix (from Leonie), and in the SLI group there were two, both from Julius. Accusative or dative noun suffixes were not overapplied to nominative (singular) contexts. All errors were due to the omission of the accusative and dative case suffix, and these omissions were not restricted to particular case contexts. Instead, the children’s use of nominal case suffixes seems to be lexically restricted. In Julius’ data, for example, the case suffix –n only occurs with the noun elefant ‘elephant’ (8), while other nouns are left unmarked in obligatory contexts. Moreover, there were two instances in which Svenja omitted the nominal case suffix, but in both instances she correctly marked accusative case on the determiner (see (9)). These data are consistent with the results from previous studies reported above and indicate that omissions of accusative and dative 20 case markings on nouns are due to the delayed acquisition of the idiosyncratic properties of the few masculine nouns that require nominal case suffixes in German. (8) wink den (corr: dem) kleinen elefant-en (Julius) ‘wave (to) the small elefant’ (9) a. wir b(r)auchen den eisbär (corr: eisbär-en) (Svenja) ‘we need the polar bear’ b. also das kind kriegt den elefant (corr: elefant-en) (Svenja ‘so, the child gets the elefant’ While obligatory contexts for prenominal genitives and nominal case suffixes were rare in our data, there were enough contexts for case-marked determiners and pronouns to permit statistical analyses. We first calculated a general measure of correctness in the two groups of children by collapsing the different correctly used case forms (nominatives, accusatives, datives) into one overall score. In the SLI group, there were 1038 obligatory contexts for case markings, 945 of which were correctly marked (= 91%), and in the control group the overall correctness score was 98% (3137/3201). Between-group comparison showed that these scores were not significantly different (Mann-Whitney; Z=.313; p=.841) indicating that overall case marking is largely accurate in both TD and SLI children and that errors were rare. A further analysis was performed to determine differences between structural and lexical case marking. Note that (for the reasons given in section 2), we take nominatives (for subjects), accusatives (for direct objects and complements of prepositions), and datives (for indirect objects) as instances of structural case marking, and datives required for arguments of certain intransitive and transitive verbs and for prepositions as instances of lexical case marking. Table 2 presents 21 overall correctness scores for structural and lexical case marking in the two groups of children as well as the corresponding individual subject data. //INSERT TABLE 2 ABOUT HERE// In both groups of children, structural case marking is mostly correct whereas lexical case marking is more error-prone. The differences in the correctness rates for lexical and structural case are significant both for the SLI group (Wilcoxon: Z=2.023, p<.05; two-tailed) and for the control group (Wilcoxon: Z=2.023, p<.05; two-tailed). Further between-group comparisons did not reveal any significant differences between the SLI and the control groups – either for structural (Mann-Whitney: Z=.529, p=.69; two-tailed) or for lexical case marking (MannWhitney: Z=-.731, p=.55; two-tailed). The contrast between lexical and structural case marking also holds at an individual subject level in that each of the ten children achieved better accuracy scores for structural than for lexical case marking. Moreover, each child produced errors in lexical case marking, whereas for structural case marking there are two SLI children (Peter and Dieter) and one control child (Leonie) who did not produce a single structural case marking error. Finally, Tab.2 shows that in both groups there is more individual variation for lexical than for structural case marking, a finding which is expected for phenomena such as lexical exceptions which have to be learnt on an item-by-item basis. In the following, we will present more detailed analyses, separately for structural and lexical case marking. Structural case marking Tab.3 presents overall correctness scores for the four types of structural case marking we examined in the two groups of children as well as the corresponding individual subject data. 22 //INSERT TABLE 3 ABOUT HERE// The overall scores shown in Tab.3 indicate that structural case marking in the four environments tested was highly accurate for both groups of children. Within-group comparisons of the overall correctness scores did not reveal any significant differences between the four structural case conditions, either for the SLI or for the control group i . Furthermore, between-group comparisons showed that the SLI group did not significantly differ from the control group in any of the four structural case marking conditions (MannWhitney: nominatives: Z=.386, p=.84; accusatives/direct objects: Z=1.230., p=.31; accusatives/PP: Z=.707, p=.80; datives/indirect objects: Z=.805, p=.56; all two-tailed). These results show that the SLI children taken as a group were unimpaired in the domain of structural case marking and achieved the same high accuracy scores as MLU-matched TD children. The individual subject data shown in Tab.3 indicates that structural case marking errors were rare. While four of the five SLI children did not produce any non-nominative subject, one child (Julius) had 192 correct nominative subjects and 61 case errors. However, all but one of these errors are substitutions of the nominative pronoun ich ‘I’ by the corresponding dative form mir ‘me’ (e.g. mir weiss diesmal ‘me know this time’ (= I know (the answer) this time)) which occurred in all three recordings. This together with the otherwise correct case marking on subjects suggests that the mir substitutions are due to an incorrect feature specification of this particular word form, rather than due to a more general difficulty with nominative case marking. In the control group data, subject case marking was also largely error-free. There were 2619 correct nominative subjects against three errors in which the accusative determiner form den was used (instead of the correct nominative der) in sentences with the verb sein 'to 23 be' (e.g. is den koffer von die von de(r) sascha ‘is the suitcase of the Sascha’ (= This is S.’s suitcase). Non-nominative subjects such as this one from Svenja have been reported in other studies of TD German-speaking children (Parodi 1990: 183, Eisenbeiss 2002). Interestingly, such errors only occur in sentences with the verb sein ‘to be’ indicating that children when they produce such errors have not yet acquired the idiosyncratic case requirements of this particular verb, namely that it takes two nominative arguments. Accusative case marking on direct objects and complements of prepositions was also largely correct in both groups of children. There were just three errors in the SLI and four in the control group data, in which the children used (six) nominatives (10a-10d) and one dative form (10e): (10) a. der da muss wir doch erst wieder anziehe(n) [corr.: den] (Svenja) ‘this one must we first of all again dress’ (= We have to dress up this one again.) b. für jeder eine [corr.: jeden] (Carsten) ‘for everybody one’ c. sich zieh der rechte strumpf [corr.: den] (David) ‘oneself put on the right sock’ (= to put on the right sock) d. jetz du der böse mann verloren [corr.: den bösen] (Sebastian) ‘now you the bad man lost’ (= Now you’ve lost the bad man.) e. mir hat der der süsse [corr.: den süssen] ‘me have the sweet’ (= I’ve got the sweet (one).) (Julius) 24 Case markings errors on indirect objects were also rare. The SLI children produced just two non-target like forms in indirect object position and the control children six; see examples in (11). There were no (unambiguous) overapplications of nominative forms; instead, all of these errors can be regarded as accusative overgeneralizations. Four of the six errors of the TD children came from one child (Mathias), three of which occurred in the same sentence frame (see (11c) which appeared twice in identical form in the same recording, and (11d)). (11) a. wenn ich mich eins faengt (Sebastian) ‘if I me one catch’ (= If I catch one for myself) b. da must du hier dich was aussuchen (Sebastian) ‘then must you here you something choose’ (= Then you must choose something for yourself here) c. das sag ich die mama. (Mathias) ‘this tell I the mommy’ (= I will tell this to mommy) d. das habe jetz die mama zählt. (Mathias) ‘this have I now the mommy told’ (= I’ve told this to mommy now) e. da fress ich dich die zuckerkluntje auf ‘then eat I you the sugar cubes up’ (= Then I’m going to eat your sugar cubes) (Mathias) 25 An interesting related finding is that in cases in which the target language would have an indirect dative object, an (obligatory) Goal or Recipient argument of a ditransitive predicate was sometimes omitted (see (12a)) or encoded as a PP containing a preposition, e.g. zu ‘to’, (12b, 12c). There were 31 such cases in the data from the TD children and 11 in the SLI data, most of which were parallel to (12b) and (12c), with a PP instead of an indirect dative object. Unlike in English, this way of expressing the third argument of a ditransitive verb sounds odd in adult German. (12) a. wir müssen das sente (=schenken) (Svenja) ‘we must give this’ b. ich schenk’s zu das baby (Julius) ‘I give-it to the baby’ (corr: ich schenk’s demDAT baby c. zu ente geben (Leonie) ‘to duck give’ (corr: derDAT ente geben) (= (I want to) give (that) to the duck) Examples such as those in (12) indicate that the encoding of indirect objects is not always target-like in SLI and TD children. Obligatory (third) arguments are not always present, and in other instances children introduce a case-marking preposition to license the third argument. Interestingly, however, these non-target-like ways of dealing with the third argument do not yield any structural case marking errors, either in SLI or in TD children. In the PPs, the lexical datives required by the preposition zu ‘to’ are substituted with structural accusative case (12b, 12c), an error that is familiar from other prepositions (and verbs) that require lexical datives. At the same time, however, case marking errors on indirect object NPs are 26 extremely rare. Thus, while some children experience difficulties expressing obligatory (third) arguments in target-like ways, this does not affect dative case marking in indirect object NPs. Summarizing, we found that structural case marking is largely error-free in both SLI and TD children. Moreover, we did not observe any developmental trends with respect to structural case marking in our data. Except for Julius, all children produced either no or very few isolated errors; and the errors we found in the data of Julius (discussed above) occurred in all three recordings for this child. Lexical case marking We examined two types of lexically case-marked noun phrases, datives exceptionally required on arguments of intransitive and transitive verbs and on the complements of prepositions. Recall that both the SLI and the TD children performed significantly worse on lexical than on structural case marking (see Tab.2). Note that this cannot be attributed to specific problems with dative case marking. The overall correctness rate for dative marking was significantly higher for indirect objects than for lexical datives assigned by verbs or prepositions (Z=2.192, p=.028). Tab.4 presents a breakdown of the two types of lexical case marking. //INSERT TABLE 4 ABOUT HERE// The overall scores shown in Tab.4 indicate that lexical case marking is far from accurate in both groups of children. Within-group comparisons of the overall correctness scores (using the Wilcoxon test) did not reveal any significant differences between the two lexical case marking conditions, either for the SLI (Z=1.826, p=.07) or for the control group (Z=1.604, p=.11). Furthermore, between-group comparisons (using Mann-Whitney) showed that the SLI group did not significantly differ from the control group in either of the two lexical case 27 marking conditions (verbal arguments: Z=1.468, p=.23, prepositional arguments: Z=-.313, p=.84). These results show that the SLI children taken as a group performed at the same level as the MLU-matched control group in the domain of lexical case marking. The individual subject data shown in Tab.4 demonstrates a considerable degree of individual variation for lexical case marking in both the SLI and the TD children. The scores range from 9% to 100% within the SLI group and from 20% to almost 90% in the TD children. The individual differences are at least in part due to the fact that lexical datives (like other lexical exceptions) have to be learnt on an item-by-item basis. In addition, it should be noted that for some children the percentages in the column ‘verbal’ shown in Tab.4 are only based on a small number of cases and are therefore hard to interpret. In particular, the SLI children Peter and Sebastian produced just one sentence each with a verb that requires a lexical dative. From these data, it is not clear whether lexical dative marking on arguments of verbs is indeed adult-like in these three SLI children. Among the TD children, the following types of case-making error were found in contexts in which lexical datives were required on verbal arguments: (i) overgeneralizations of unambiguous accusative forms or of forms that could either be nominatives or accusatives (n=4) to the internal argument of a transitive verb (see example (13a)); (ii) overgeneralizations of nominatives (n=16) to the Possessor argument of the verb gehören ‘belong’ (13b) or the Theme argument of the intransitive verbs sein ‘to be’ and werden ‘to become’ (13c). In the SLI data, all errors (n=7) were of type (i) in which a lexical dative on a direct object was replaced by the corresponding structural (accusative) case form; see (13d). In PPs in which adult German requires lexical datives on the complements of prepositions, all of the errors (n=61) in both groups of children were overgeneralizations of accusative forms or of forms that could either be nominatives or accusatives (13e). While the particular morphological form of the determiner in examples such as (13e) is ambiguous, the fact that 28 among the 61 case-marking errors in this condition there was not a single case of an unambiguous nominative form, e.g. ich ‘I’ instead of mir ‘me’, indicates that errors such as those in (13e) can be interpreted as accusative (rather than nominative) overgeneralizations, because otherwise there should have been instances of nominative overgeneralizations in this condition. (13) a. ich hab michACC da wehgetan (Leonie) ‘I have me there hurt’ (corr: …mirDAT…) (= I hurt myself there) b. werNOM dasNOM/ACC gehört (Svenja) ‘who that belongs’ (corr: wemDAT…) (= Who does this belong to?) c. ichNOM bin kalt (Leonie) 'I am cold'(corr: mirDAT ist kalt) d. helf denACC dieNOM/ACC frau (Julius) ‘help the woman' (corr: …derDAT…) e. mit dieNOM/ACC birne (Peter) 'with the pear'(corr.: ..derDAT…) These data show that lexical case marking yields many errors in both SLI and TD children in which an exceptional lexical case was replaced by the corresponding structural case. To determine developmental changes in this domain, we examined lexical dative assignment by prepositions; for lexical datives assigned by verbs, there were not enough data. We included children who had more than one recording and produced relevant contexts as well as errors (TD children: Mathias and Svenja; SLI children: Sebastian and Julius). While Svenja 29 produced errors in contexts requiring lexical datives after prepositions throughout the entire recording time, Mathias and Sebastian made errors only in the first three recordings and produced correct lexical datives in the remainder of the recording period. Julius started to produce correct lexical datives only in his final recording. These findings indicate that accuracy rates for lexical case marking improve over time. 7. Discussion Our most important results are that structural case marking is (largely) error-free in TD children and unimpaired in SLI children. In the following, we will discuss how these findings can be explained both in grammatical terms (e.g. with respect to the distinction between structural and lexical case) and in terms of non-grammatical factors, specifically the role of frequency and productivity. Input frequencies, productivity, and children’s case marking systems In two recent treatments of case marking in German child language (Szagun 2004, Wittek and Tomasello 2005), it has been claimed that children’s case systems are not a ‘mirror of syntactic categorization’ (Szagun 2004: 26), but are instead influenced by non-grammatical factors. Wittek and Tomasello (2005: 122) argue, for example, that nominative and accusative case are easier to learn for German children than the dative, because they are relatively more frequent and consistent local cues in the input than the dative. Along similar lines, one might speculate that English-speaking children produce non-nominative (= accusative) subjects, because accusative pronouns are highly frequent in a child’s input. Moreover, Tomasello, Lieven and their collaborators (see also Lieven et al. 1997) have argued that what looks like abstract syntactically structured elements in young children’s utterances are often unproductive, internally unstructured, lexeme-specific forms. This could apply to instances of 30 correct subject case marking in finite clauses reported for early English child language (as well as for SLI) in which the child may combine a small number of nominative pronouns with particular finite verb forms that frequently co-occur in the input. Our findings provide clear evidence against any such explanation of the data. Note first that according to the frequency counts in Meier (1967), dative is more frequent than accusative case in German usage, in contrast to what Wittek & Tomasello (2005) claim. Moreover, complements of prepositions more frequently occur in dative than in accusative case (Folsom 1984) ii . From a frequency-based perspective, one would expect to find dative overapplications in such environments. This was not the case, however. The TD and the SLI children did not produce a single dative overgeneralization in a prepositional context and overapplied the less frequent accusative case instead. Secondly, nominative is more frequent in German usage than accusative, in particular for verbal arguments (Meier 1967), and if frequency was a relevant predictor for children’s case-marking errors, we should find nominative overgeneralizations. In the child data, however, (both from the SLI and the TD children) nominative case was not overgeneralized to direct or indirect objects, and for lexical exceptions (e.g. direct objects that require lexical datives in German), the children overgeneralized the structural case for direct objects (= accusative), rather than the most frequent case (=nominative). Thirdly, dative case is less frequent in the input to a child than nominative. We would therefore expect lower accuracy scores for datives on indirect objects than for nominatives (on subjects). This prediction is also not borne out. Instead, we found the same high accuracy scores both groups of children in these environments. Finally, case marking in the German child data appears to be productive, and not restricted to a small number of (possibly fixed) word forms. Correct (structural) case marking is not only found on high-frequency pronouns (as in English), but also on different types of determiners and in different sentential positions in combination with a range of different lexical elements. 31 We conclude that case marking in German child language is not due to formulaic or unproductive word combinations and that attempts to explain the German case-marking data in non-syntactic terms (e.g. input frequencies) are not particularly promising. Lexical and structural case marking in German child language Our results indicate a contrast between structural and lexical case marking in TD German children (as well as in children with SLI). We found that structural case-marking (i.e. prenominal genitives, the nominative for subjects, the accusative for direct objects of verbs and complements of prepositions, and the dative for indirect objects) is essentially error-free once the relevant forms are used. On the other hand, dative marking of arguments of intransitive and transitive verbs and dative marking on complements of prepositions is subject to case-marking errors in which the dative is replaced by the corresponding structural cases. These findings fit in with observations made in previous studies of unimpaired German children. For grammatical subjects, for example, Clahsen (1984), Tracy (1986), Clahsen et al. (1994) and Schütze (1997) did not find any overgeneralizations of accusatives or datives in the speech of 2-to-3 year-old TD children, and 90% of the accusative objects as well as more than 90% of the indirect (dative) objects were correctly case-marked (Clahsen et al. 1996, Eisenbeiss 2002). On the other hand, for complements of prepositions Mills (1985), Clahsen (1984), Tracy (1986), and Eisenbeiss (2002) reported frequent overgeneralizations of accusative forms after prepositions that require datives. As in our data, however, unambiguous nominative overgeneralizations were not documented at all, and examples of substitutions of datives for accusatives after prepositions were reported to be extremely rare (see Stenzel 1994, Preyer 1882, and Stern and Stern 1927). Taken together, these findings show that the only unambiguous case-marking errors that are found in unimpaired German children are overapplications of structural cases to forms that 32 require lexical case in the adult language and that substitutions of one structural case for another are extremely rare. The results of the present study show that the same is true for SLI children. The contrast seen between structural and lexical case marking in German child language replicates Babyonyshev’s (1993) results for child Russian, i.e. few mistakes in the use of structural cases and relatively many in the use of lexical cases. These findings indicate that the structural case assignment mechanism of the adult grammar is already operative in twoyear-old unimpaired children (as well as in children with SLI) and that children (like adults) associate certain phrase-structure positions, e.g. the subject and direct object positions, with particular case markings (accusative and nominative) overappling these structural cases when they have not yet acquired the lexical exceptions; see also Clahsen et al. (1994). Children’s lexical case marking, on the other hand, is developmentally delayed and more error-prone than structural case marking because of the idiosyncratic properties of lexical case assignment which have to be learnt on an item-by-item basis. Similarly, the fact that some nouns require accusative or dative case markers has to be learnt on an item-by-item basis, and both groups of children produced omission errors indicating that the acquisition of these lexical idiosyncracies is delayed relative to the development of structural case marking on determiners and pronouns. Case marking and the nature of the syntactic deficit in SLI Recall from the Introduction that the nature of the syntactic deficit in SLI is controversial. Some accounts posit broad syntactic deficits, e.g. for ‘dependent relations’ (van der Lely and Stollwerck 1997), while others claim that the underlying syntactic competence is basically intact and that selective areas of the grammar are not fully specified for children with SLI (see 33 e.g. Rice and colleagues’ Extended Optional Infinitives’ hypothesis). These accounts make different predictions for case marking in children with SLI. Models that posit general impairments in the underlying syntactic representations of SLI individuals, e.g. difficulties in matching grammatical features of different syntactic categories within a sentence (Clahsen 1989, 1991), impairments with a wide range of functional categories (Leonard 1995), or a deficit with building syntactic dependencies (van der Lely and Stollwerck 1997) predict case marking deficits in SLI children. German-speaking SLI children should either produce random case errors or fall back on the default nominative for arguments that require other structural cases (= accusatives or datives) in the adult language. Our results clearly disconfirm these predictions. Accuracy scores for structural case marking in SLI children were high, error patterns were not random, and the default (nominative) case was not overapplied in environments that required a different structural case. Our conclusion is that the case-marking data reported here are incompatible with the idea of a general impairment in the underlying syntactic representations of SLI individuals. Other models of SLI claim that the syntactic impairment in SLI is confined to the verbfiniteness cluster and that grammatical morphemes and elements of syntactic structure that are unrelated to verb finiteness are not affected (see Rice et al. 1995, Rice and Wexler 1996, Rice 1999, Wexler et al. 1998, Clahsen, Bartke, Göllner 1997, Clahsen and Dalalakis 1999). These accounts predict that SLI children do not produce overapplications of the default case (= nominative) in circumstances in which other structural cases (= accusative and dative) are required in the adult language. This is because case marking of direct and indirect objects does not form part of the verb-finiteness cluster. Structural case marking should be productive in SLI children in the same way as in TD children. We would therefore expect to find overapplications of structural cases to forms that require lexical case in adult German, specifically, overgeneralizations of accusative to the direct objects of verbs and the 34 complements of prepositions that require lexical datives. Our results confirm these predictions. The SLI children produced correct accusative case marking on direct objects and on complements of prepositions as well as dative case marking on indirect objects, and they overapplied structural case marking in the same way as TD children, i.e. in circumstances in which adult German requires lexical case. Models of SLI that posit selective impairments in the domain verb finiteness marking also provide an account of the case-marking errors seen in English-speaking SLI children (see Wexler et al. 1998). Recall that nominative case in English is said to be assigned by an agreement feature in T(ense), and that if agreement is left unspecified, nominative case cannot be assigned; consequently, English-speaking SLI children either drop the subject or fall back on the default (accusative) case. For German-speaking SLI children, one would expect to find the same contingency, i.e. when subject case cannot be assigned (due to unspecified verbfiniteness features), SLI children should overapply the default case. This prediction, however, cannot be tested for German, since the nominative is both the default case and the structural case required for subjects. Thus (unlike in English) it is impossible to decide whether a nominative subject produced by a German SLI child is an instance of correct structural case marking or an overapplication of the default case. Nevertheless, we can conclude from the present set of findings that the patterns of case marking seen in German-speaking SLI children are compatible with the idea of positing a selective syntactic deficit that does not affect (structural) case marking. 8. Conclusion We found that both TD and SLI children achieved high accuracy scores for nominative subjects, accusatives on direct objects and complements of prepositions, and for datives on indirect objects, i.e. for structural case marking. Moreover, structural case markers are 35 overgeneralized to exceptions in which lexical datives are required in the adult language. The SLI children performed at the same high accuracy level as MLU-matched TD children in the domain of structural case marking and produced the same kinds of overgeneralization errors on lexical datives indicating that case marking is unaffected by SLI. Our results, we believe, provide evidence against models of SLI that posit a general impairment in the underlying syntactic representations of SLI individuals. Finally, our acquisition data are consistent with the view that the dative on indirect objects is a structural case and the dative on complements of prepositions a lexical case in German. We found that dative marking on indirect objects patterns with two (uncontroversial) instances of structural case marking, accusatives on direct objects and nominatives on subjects, in that case marking in these domains was largely error-free. We also observed an asymmetry in the cases assigned by prepositions: accusatives were overapplied to complements of prepositions that require datives in the adult language, but not vice versa, yielding errors for datives after prepositions, but not for accusatives. Together with the linguistic arguments mentioned in section 2, these findings indicate that the dative on indirect objects is a structurally assigned case, whereas the dative on prepositions is an instance of lexical case. ACKNOWLEDGEMENTS Supported by a German Science Foundation grant to HC (grant No. Cl 97/5.1-3). We are grateful to Jürgen Bohnemeyer, Bettina Landgraf, Keiko Murasugi, Carson Schütze, Andrew Radford, and three anonymous LA reviewers for helpful comments and suggestions. 36 Notes i ii The Wilcoxon scores were as follows: Pair-wise comparison Control group SLI group acc./dir. obj. - nom. Z=1.069; p=.285 Z=1.089; p=.276 acc./dir. obj. - dat.ind. obj. Z=-.730; p=.465 Z=1.342; p=.180 acc./dir. obj. - acc./PP Z=.447; p=.655 Z=1.000; p=.317 acc./PP - nom. Z=.447; p=.655 Z=1.089; p=1.000 acc./PP - dat.ind. obj. Z=1.604; p=.109 Z=1.000; p=.317 dat.ind. obj. - nom. Z=1.604; p=.109 Z=.447; p=.655 In her acquisition study, Szagun (2004) claims the opposite for adult German (‘dative markers are infrequent’), but this is based on a mis-interpretation and not backed up by any frequency data. 37 References Babyonyshev, M. (1993). The acquisition of the Russian case system. MIT Working Papers in Linguistics 19: 1-43. Bartke, S. (1998). Experimentelle Studien zur Flexion und Wortbildung. Pluralbildung und lexikalische Komposition im unauffälligen Spracherwerb und im Dysgrammatismus. Niemeyer: Tübingen. Bierwisch, M. (1988). On the grammar of local prepositions. In M. Bierwisch, W. Motsch and I. Zimmermann (eds.), Syntax, Semantik und das Lexikon. Akademie-Verlag Berlin, pp. 1-65. Chomsky, N. (1981). Lectures on government and binding. Foris: Dordrecht. Chomsky, N. (1995). The Minimalist Program. MIT Press: Cambridge, MA. Clahsen, H. (1982). Spracherwerb in der Kindheit. Eine Untersuchung zur Entwicklung der Syntax bei Kleinkindern. Tübingen: Narr. Clahsen, H. (1984). Der Erwerb von Kasusmarkierungen in der deutschen Kindersprache. Linguistische Berichte 89: 1-31. Clahsen, H. (1989). The grammatical characterization of developmental dysphasia. Linguistics 27: 897-920. Clahsen, H. (1991). Child language and developmental dysphasia. Linguistic studies of the acquisition of German. Benjamins: Amsterdam. Clahsen, H. (1999). Linguistic perspectives on specific language impairment. In: W. C. Ritchie and E. K. Bhatia (Eds.), Handbook of child language acquisition, Academic Press: London, pp. 675704. Clahsen, H., Vainikka, A. and M. Young-Scholten (1990). Lernbarkeitstheorie und Lexikalisches Lernen. Eine kurze Darstellung des LEXLERN-Projekts. Linguistische Berichte 130: 466-477. Clahsen, H., Bartke, S. and S. Göllner (1997). Formal features in impaired grammars: a comparison of English and German SLI children. Journal of Neurolinguistics 10: 151-171. 38 Clahsen, H., Eisenbeiss, S. and A. Vainikka (1994). The seeds of structure. A syntactic analysis of the acquisition of case marking. In: T. Hoekstra and B. Schwartz (Eds.), Language acquisition studies in generative grammar. Amsterdam: Benjamins, pp. 85-118 Clahsen, H., Eisenbeiss, S. and M. Penke (1996). Lexical learning in early syntactic development. In: H. Clahsen (ed.), Generative Perspectives on Language Acquisition. Amsterdam: Benjamins, pp.129-159. Clahsen, H. and J. Dalalakis (1999). Tense and agreement in Greek SLI: A case study. Essex Research Reports in Linguistics 24: 1-25. Czepluch, H. (1988). Kasusmorphologie und Kasusrelationen: Überlegungen zur Kasustheorie des Deutschen. Linguistische Berichte 116: 275–310. Eisenbeiss, S. (1991). Passiv und Kasus. Unpublished ms.: University of Cologne. Eisenbeiss, S. (1994). Kasus und Wortstellungsvariation im deutschen Mittelfeld. Theoretische Überlegungen und Untersuchungen zum Erstspracherwerb. In: B. Haftka (Ed.), Was determiniert Wortstellungsvariation? Opladen: Westdeutscher Verlag, pp. 277-298. Eisenbeiss, S. (2000). The acquisition of the Determiner Phrase in German child language. In: M.A. Friedemann and L. Rizzi (Eds.), The Acquisition of Syntax: Studies in Comparative Developmental Linguistics. London: Longman, pp. 26-62. Eisenbeiss, S. (2002) Merkmalsgesteuerter Grammatikerwerb. Unpublished Dissertation: University of Düsseldorf. Emonds, J. (1985). A unified theory of syntactic categories. Foris: Dordrecht. Eyer, J. and L. Leonard (1995). Functional categories and specific language impairment: A case study. Child Language Teaching and Therapy 10: 127-138. Folsom, M. H. (1984). Prepositions with the dative or accusative in written and spoken German. In: J. A. Pfeffer (Ed.), Studies in Descriptive German Grammar. Heidelberg: Groos, pp. 19-32. Gallmann, P. (1992). Dativanhebung? GAGL 35: 92-121 (Groninger Arbeiten zur germanistischen Linguistik). 39 Gavin, W.J. and L. Giles (1996). Sample size effects on temoral reliability of language sample measures of preschool children. Journal of Speech and Hearing Research 39: 1258-1262. Gazdar, G., Klein, E., Pullum, G. and I. Sag. (1985). Generalized phrase structure grammar. Harvard University Press: Cambridge, MA. Guilfoyle, E., Allen, S. and S. Moss (1991). Specific language impairment and the maturation of functional categories. paper presented at the Boston University Conference on Language Development, Boston, MA. Haider, H. (1985). The Case of German. In: J. Toman (ed.), Studies in German Grammar, Foris: Dordrecht, pp. 23–64. Haider, H. (1988). Die Struktur der deutschen Nominalphrase. Zeitschrift für Sprachwissenschaft 7: 32-59. Haegeman, L. (1991). Introduction to Government and Binding. Oxford: Blackwell. Heinz, W. and J. Matiasek (1994). Argument structure and case assignment in German. In: J. Nerbonne, K. Netter and C. Pollard (eds.). German in Head-driven Phrase Structure Grammar. Stanford/CA, pp. 199-236. Indefrey, P. (2002). Listen und Regeln. Erwerb und Repräsentation der schwachen Substantivdeklination des Deutschen. Unpublished Disss.: University of Düsseldorf. Kubo, M. (1992). Japanese syntactic structures and their constructional meanings. Unpublsihed PhD dissertation. MIT: Cambridge, MA. Leonard, L. (1995). Functional categories in the grammars of children with specific language impairment. Journal of Speech and Hearing Research 38: 1270-1283. Leonard, L. (1998). Children with specific language impairment. MIT Press: Cambridge, MA. Levy, Y. and G. Kahé (1999). Language breakdown and linguistic theory: a tutorial overview. Lingua 107: 95-143. Lieven, E., Pine, J. and G. Baldwin (1997). Lexically-based learning and early grammatical development. Journal of Child Language 24: 187-219. 40 Loeb, D. and L. Leonard (1991). Subject case marking and verb morphology in normallydeveloping and specifically-language-impaired children. Journal of Speech and Hearing Research 34: 340-346. MacWhinney, B. (2000). The CHILDES project: Tools for analyzing talk. Third Edition. Mahwah, NJ: Lawrence Erlbaum Associates. Meier, H. (1967). Deutsche Sprachstatistik. Vol.1 and 2. Hildesheim: Olms. Mills, A. (1985). The acquisition of German. In: D. Slobin (ed). The cross-linguistic study of language acquisition. Vol. 1. Hillsdale, NJ: Lawrence Earlbaum, pp. 141-254. Müller, S. (1999). Deutsche Syntax deklarativ. Head-Driven Phrase Structure Grammar für das Deutsche. Tübingen: Max Niemeyer Verlag. Olsen, S. (1990). AGR in the German noun phrase. In: C. Bhatt, E. Löbel and C. Schmitt (eds.). Syntactic phrase structure phenomena in noun phrases and sentences. Amsterdam: Benjamins. Parodi, T. (1990). The acquisition of word order regularities and case morphology. In: J. M. Meisel (Ed.), Two First Languages: Early Grammatical Development in Bilingual Children. Foris: Dordrecht, pp. 157-192. Preyer, W. T. (1882). Die Seele des Kindes. Leipzig: Schaeffer. Radford, A. and E. Ramos (2001). Case, agreement and EPP: Evidence from an English-speaking child with SLI. Essex Research Reports in Linguistics 36: 42-81. Rice, M. (1999). Specific grammatical limitations in children with Specifric Language Impairment. In: H. Tager-Flusberg (ed.), Neurodevelopmental disorders. MIT Press: Cambridge, MA, pp.331-359. Rice, M., Wexler, K. and P. Cleave (1995). Specific language impairment as a period of extended optional infinitive. Journal of Speech and Hearing Research 38: 850-863. Rice, M. and K. Wexler (1996). Toward tense as a clinical marker of specific language impairment in English-speaking children. Journal of Speech and Hearing Research 39: 1239-1257. 41 Schmidt, C. M. (1995). Satzstruktur und Verbbewegung. eine minimalistische Analyse zur internen Syntax der IP (Inflection-Phrase) im Deutschen. Tübingen: Niemeyer. Schütze, C. (1997). Infl in child and adult language: Agreement, case and licensing. Ph D. dissertation: MIT: Cambridge, Mass. Schütze, C. and K. Wexler (1996). Subject case licensing and English root infinitives. In: A. Stringfellow et al. (Eds.), Proceedings of the 20th Boston University Conference on Language Development. Cascadilla Press: Boston. Stiebels, B. (2002). Typologie des Argumentlinkings: Ökonomie und Expressivität. Berlin: Akademie Verlag. Stenzel, A. (1994). Case assignment and functional categories in bilingual children: Routes of development and implications for linguistic theory. In: J. M. Meisel (Ed.), Bilingual First Language Acquisition: French and German Grammatical Development. Amsterdam: Benjamins, pp. 161-208. Stern, C. and W. Stern (19071, 19284). Die Kindersprache. Leipzig: Barth. Szagun, G. (2004) Learning by ear: on the acquisition of case and gender marking by Germanspeaking children with normal hearing and with cochlear implants. Journal of Child Language 31: 1-30. Tracy, R. (1986). The acquisition of case morphology in German. Linguistics 24: 47-78. van der Lely, H. and L. Stollwerck (1997). Binding theory and specifically language impaired children. Cognition 62: 245-290. Vainikka, A. (1994). Case in the development of English syntax. Language Acquisition 3: 257-325. Wagner, K. (1985). How much do children say in a day? Journal of Child Language 12: 475-487. Wegener, H. (1990). Der Dativ – ein struktureller Kasus? In G. Fanselow and S. Felix (eds.), Strukturen und Merkmale syntaktischer Kategorien. Tübingen: Narr, pp. 70–103. Wexler, K., Schütze, C. and M. Rice (1998). Subject case in children with SLI and unaffected controls: Evidence for the Agr/Tns omission model. Language Acquisition7: 317-344. 42 Wittek, A. and M. Tomasello (2005). German-speaking children’s productivity with syntactic constructions and case morphology: local cues act locally. First Language 25: 103-125. Wunderlich, D (1997). Cause and the structure of verbs. Linguistic Inquiry 28: 27-68. Tab. 1 : Overview of German SLI and control group data group control SLI child age MLU number of number of number of in words recordings analysable analysable utterances case markers Mathias 3;1-3;6 2.91-3.51 6 719 261 Carsten 3;6 4.2 1 1795 775 Svenja 2;9-3;3 3.3-4.1 15 3811 1644 Annelie 2;8-2;9 2.5-3.1 2 674 199 Leonie 2;6-2;11 2.1-3.1 7 2462 322 Julius 5; 9 - 6;0 2.8 - 4.2 3 1015 284 David 7;11 4.0 1 411 69 Peter 6;6 – 7;6 3.5 - 3.7 2 737 221 Sebastian 5;8 – 6;6 2.6 - 3.4 4 1006 228 3.9 1 492 236 Dieter 7;2 Tab. 2 : Percentages of correct case markings for determiners and pronouns (raw numbers shown in parantheses) group control SLI child structural lexical Mathias 98.39 (244/248) 69.23 (9/13) Carsten 99.60 (747/750) 40.00 (10/25) Svenja 99.74 (1539/1543) 80.20 (81/101) Annelie 99.47 (188/189) 20.00 (2/10) Leonie 100.00 (308/308) 64.29 (9/14) Overall 99.61 68.1 Julius 76.52 (202/264) 15.00 (3/20) David 98.46 (64/65) 50.00 (2/4) Peter 100.00 (214/214) 85.71 (6/7) Sebastian 98.16 (213/217) 81.82 (9/11) Dieter 100.00 (222/222) 71.43 (10/14) Overall 93.18 53.57 Tab. 3: Percentages of correct structural case markings for determiners and pronouns (raw numbers shown in parantheses) nominative accusative dative direct object preposition 100.00 100.00 100.00 42.86 (231/231) (8/8) (2/2) (3/7) 611/612 44/44 50/51 42/43 99.84 100.00 98.04 97.67 99.85 99.27 100.00 98.00 (1345/1347) (136/137) (9/9) (49/50) 100.00 90.91 X 100.00 (173/173) (10/11) 100.00 100.00 100.00 100.00 (256/256) (35/35) (4/4) (13/13) Overall 99.89 99.15 98.48 94.92 Julius 75.89 88.89 X 100.00 (192/253) (8/9) 100.00 75.00 (58/58) (3/4) 100.00 100.00 (201/201) (13/13) 100.00 Mathias control Carsten Svenja Annelie Leonie SLI David Peter Sebastian Dieter Overall (5/5) (2/2) X 100.00 (3/3) X X 75.00 100.00 75.00 (194/194) (6/8) (7/7) (6/8) 100.00 100.00 100.00 100.00 (203/203) (10/10) (4/4) (5/5) 93.29 90.91 100.00 88.89 Tab. 4 : Percentages of correct lexical case markings for determiners and pronouns dative control SLI verbal preposition Mathias 50.00 (1/2) 72.73 (8/11) Carsten X 40 (10/25) Svenja 76.19 (16/21) 81.25 (65/80) Annelie X 20 (2/10) Leonie 20.00 (1/5) 88.89 (8/9) Overall 64.29 68.89 Julius 22.22 (2/9) 9.09 (1/11) David X 50.00 (2/4) Peter 100.00 (1/1) 83.33 (5/6) Sebastian 100.00 (1/1) 80.00 (8/10) Dieter 100.00 (4/4) 60.00 (6/10) Overall 53.33 53.66