German children's comprehension of word order and case marking

In this paper we explored factors that affect how children learning German understand which participant is the agent and which the patient in simple causative sentences (i.e. who does what to whom). Most languages, including German, have multiple cues to mark these roles. In two comprehension experiments we examined whether German children are able to use the grammatical cues of word order and word endings (case marking) to correctly identify agents and patients in causative sentences and whether they weigh these two cues differently across development. Older two-year-olds correctly understood only sentences with both cues supporting each other the prototypical form but not sentences with either cue on its own. Five-year-olds were able to use word order by itself, but not case marking. Only seven-year-olds behaved like adults by comprehending both cues on their own, and also, importantly, by relying on case marking over word order when the two cues conflicted. These findings suggest that prototypical instances of linguistic constructions with redundant grammatical marking play a special role in early acquisition, and only somewhat later do children isolate and weigh individual grammatical cues appropriately in terms of their reliabilities for signalling specific functions. Comprehension of Case and Word Order 3 One of the important tasks of early childhood is mastering a conventional language. Languages differ not only in their words, but also in the grammatical constructions they employ for assembling words into meaningful utterances. Grammatical constructions are composed of multiple words, or word categories, structured into patterns in particular ways by such things as word order and grammatical markers (e.g., a different ending on a word when it is the subject rather than the direct object in a sentence socalled case marking). Thus, in English, the sentence "The dax mibbed the gazzer a toma" (the ditransitive construction) implies a transfer of some kind, even though all of the contentful words are meaningless (Goldberg, 1995). One construction of particular importance in early development is the basic transitive construction, prototypically used to indicate an agent causally acting on an object, as in simply "The dax mibbed the gazzer". The importance of this construction stems from the fact that it is one of the ontogenetically earliest in which it is critical to distinguish the different roles of the participants in some event. Thus, "The toma mibbed" creates no problems for deciding who was doing the action because there is only one participant. But if we hear "The toma the gazzer mibbed" we must decide who is mibbing whom, and to do this we need to understand the grammatical conventions of the particular language being learned. Interestingly, in most languages there are multiple, redundant cues for helping the listener do this in many utterances although in other utterances there can be just a single cue. For example, in the English sentence "He mibs pencils", we identify the agent of the action as he based on the facts that: (i) it is said before rather than after the action word or verb [word order]; (ii) it is the subject pronoun he (and not the object pronoun him) [case marking]; (iii) it agrees in number with the verb (we say "He mibs" but "Pencils mib, without an -s) [subject-verb agreement]; and (iv) it is a statistical fact that animate beings, such as male persons, are more likely to act on inanimate things, such as pencils, than the other way around [animacy]. A child acquiring the English transitive construction, therefore, could on a particular occasion be using any or all of these cues to determine who is mibbing whom in the utterance. Comprehension of Case and Word Order 4 The Competition Model of Bates and MacWhinney (1987, 1989) represents an attempt to assess how children acquire the different cues of different languages especially in the transitive construction and how they weigh these cues relative to one another when they conflict (see also Slobin & Bever, 1982). For example, in a comprehension task in which they are asked to identify who is doing what to whom, young children hear a sentence such as "Him kissed she". In this sentence the casemarked pronouns indicate that the female kissed the male (she > him), but the word order indicates that the male kissed the female (him > she). The finding is that from early in the preschool period English-speaking children privilege word order over all other tested cues (e.g., animacy and subject-verb agreement) in interpreting transitive sentences. Other researchers have tested English-speaking children's comprehension of word order when it is the only cue available (so not conflicting with any other cues) and found that even young two-year-olds already distinguish between such things as 'X is tickling Y' and 'Y is tickling X' (Hirsh-Pasek & Golinkoff, 1996, with familiar verbs; Gertner, Fisher, & Eisengart, 2006, with novel verbs). In many other languages, the grammatical cues in transitive sentences are much more evenly weighted than in word-order dominant English. For example, in many languages in which all nouns are case marked for their role in the sentence (not just pronouns, as in English), word order is much more flexible because if a word is locally marked with a case marker indicating its role in the sentence, then word order may be used for pragmatic functions such as emphasis and perspective (as English does awkwardly in such sentences as "Him I like"). Thus, if German adults are presented with a sentence parallel to the English sentence above ("Him kissed she"), they interpret it in the opposite way to English adults, that is, they insist that the subject-marked pronoun she indicates the one doing the kissing even though it comes after the verb (whereas it most often comes before the verb) (MacWhinney, Bates, & Kliegl, 1984). Importantly, in the Competition Model there are methods for quantifying the strength of various cues in a particular language, for example, in the transitive construction. To do this one looks at the general dimensions of: the frequency of a cue Comprehension of Case and Word Order 5 (cue availability), the consistency of a cue in indicating a function (cue reliability), and the complexity of a cue (cue cost). Thus, in English, case-marked pronouns (e.g., I-me, hehim) are highly reliable in transitive sentences (when they are present they indicate accurately agent and patient), but they are not always available (often there are only full nouns, which are not case marked). In German, word order is almost always available, but it is often not reliable (because sometimes transitive sentences have the agent after the verb and the patient before it which works because they are both case-marked for role). These two dimensions of cues availability and reliability can be combined to give an overall measure of cue validity (Kempe & MacWhinney, 1998). In terms of acquisition, Bates and MacWhinney (1987) predicted that children should acquire first those cues with highest cue validity. In addition, because sometimes several cues may indicate the same function redundantly and this provides extra information children should find especially easy to comprehend prototypical transitive sentences with both word order and case marking (and perhaps other cues) working in coalition: the coalitions-as-prototypes model (Bates & MacWhinney, 1987). This should be true especially if, as is often the case, the prototype occurs especially frequently. Thus, an agent of a transitive action, for instance, should be identified most easily by a German child if it is not only marked by its position before the verb but also by the relevant case marker. In a study of English and Italian speaking children, Bates, MacWhinney, Caselli, Devescovi, Natale and Venza (1984) provided evidence for this approach by comparing the use of word order and animacy cues (agents tend to be animate, patients inanimate) in transitive sentences. They found that the high cue validity of word order in English led English two-year-olds to rely on word order and ignore animacy when these two cues conflicted (i.e., when they heard "The pencil is kicking the cow" they tried to make the pencil kick), whereas the low cue validity of word order in Italian led Italian two-year-olds to rely on animacy and ignore word order (thus making the cow kick the pencil). Some researchers have proposed that the particular aspects of cue validity that children follow change over development. In a study with Hebrew-speaking children and adults, Sokolov (1988) found that cue availability – how often a particular cue occurs – Comprehension of Case and Word Order 6 played a stronger role in sentence interpretation for younger children, whereas cue reliability – the proportion of relevant sentences for which a particular cue correctly indicated agent or patient played a stronger role for older children and adults. Of special importance, to establish unequivocally which cue is most reliable in their language, children have to notice which cue adults follow when two cues conflict [conflict validity]. In many cases this may be a quite drawn-out process, as the relevant conflict situations are sometimes fairly rare in the language children experience (McDonald, 1986). Supporting this general view, Matessa and Anderson (2000) found that in adult artificial language learning cue validity predicted which cues are used early in the learning process, and conflict validity predicted which cues are used in later learning. Cue cost (essentially, complexity) has been much less studied. Taking off from Slobin’s (1982) Local Cues Hypothesis, one claim is that ‘local cues’ such as animacy or case marking can be processed on the spot without taking the entire sentence into account, whereas ‘distributed cues’ such as word order or subject-verb agreement impose a greater burden on short term processing capacity (because sentential fragments need to be held in memory until the next relevant component is processed). Support for this hypothesis was provided by Lindner (2003), who found that early in development German children tended to rely on ‘local cues’ such as animacy (two-and three-year-olds) and case marking (fouryear-olds) and only later on ‘distributed cues’ such as subject-verb agreement. However, Lindner’s analyses did not involve a direct comparison between conditions in which the different cues supported or conflicted with one another. Studies that have made such a direct comparison have found that German pre-school children comprehend sentences in which case marking and word order conflict more poorly than sentences in which case marking and word order collaborate (e.g., Mills, 1977; Primus & Lindner, 1994; SchanerWolles, 1989). However, these studies differ as to the age at which German children accurately comprehend sentences with conflicting cues, most likely because the different studies used only sentences with highly familiar verbs, and exactly which familiar verbs were used varied between studies. The use of familiar verbs opens up the possibility that children could respond on the basis of only verb-specific knowledge (e.g., knowing only Comprehension of Case and Word Order 7 that 'the hitter comes before hit) whereas mature grammatical knowledge is based on verbgeneral, abstract knowledge of grammatical constructions. In the current study, therefore, we investigated German children's understanding of word order and case marking cues in transitive sentences, and unlike previous studies in the Competition Model framework we did this using novel verbs. Our specific question was when German children come to understand that in their language case marking is a 100% reliable cue (even if it is not always available), whereas word order is not (even though it is quite often available). In two experiments, we gave children test sentences that contained various combinations of word order and case marking cues all grammatically correct. In one condition both cues supported one another: case marking and word order both indicated the first noun as the agent. In a second condition these two cues were in conflict: word order indicated the first noun as the agent whereas case marking indicated the second noun as agent. Finally, in a third condition agent and patient were case-marked ambiguously and therefore the only cue children could rely on was word order. Following Bates and MacWhinney’s (1987) concept of coalitions-as-prototypes, we predicted that sentences containing multiple, redundant cues (as in the first condition) should be easiest to acquire. From McDonald’s (1986) findings we predicted that sentences containing conflicting cues (as in the second condition) should be the most difficult because robust knowledge of relative cue reliabilities from relatively rare conflict situations is needed for adult-like comprehension. The findings from this study should be relevant not only for elucidating basic processes of language development, but also to for elucidating processes of children's learning more generally, as it addresses such domain-general issues as the role of prototypes, the individuation of particular cues from prototypes, and children's sensitivity to more local versus more distributed cues in sequential learning in general. Study 1 As a preliminary to our two comprehension experiments, we first looked at how German adults use word order and case marking in transitive sentences addressed to young children. Since the Competition Model predicts that the cue validity of word order and Comprehension of Case and Word Order 8 case marking should play a key role in children's comprehension and that cue availability and reliability might play different roles at different points in development we computed all of these values for these two cues for a corpus of child-directed speech. The German grammar relevant to the current studies is as follows. In active transitive sentences, the agent of the action is subject and is marked with nominative case marking, and the patient is direct object and is marked with accusative case marking. For both of these, the case marking is either a special form of pronoun or a noun with a special form of determiner (e.g., a or the). For example, if a dog is agent the form is der Hund (the+nominative dog) or er (he), whereas if a dog is patient the form is den Hund (the+accusative dog) or ihn (him). Additional complexity comes from the fact that nominative and accusative marking take different forms when applied to nouns of different genders, and in some cases they are not distinct. For example, unlike the example of dog above (which is masculine), if a cat is the agent the form is die Katze (the+nominative cat), but if a cat is the patient the form is exactly the same die Katze (the+accusative cat). This means that in some instances case marking is not an available cue in the sense that it does not identify case role unambiguously. Finally, although in German transitive sentences agents typically come before the verb and patients after the verb, as in English, to highlight the patient pragmatically the reverse order may be used with the case roles marked by case marking and unaffected by the reverse order. Thus, "Den Hund beisst der Mann" has the first noun, Hund, marked as accusative and the second noun, Mann, marked as nominative and so, despite word order, it is the man biting the dog.