ThreeRevision with abstract and references plus author

Establishing discourse referents Establishing discourse referents: indefinite noun phrases in German pre-­‐
schoolers’ narratives* *This work was partly supported by the German Bundesministerium für Bildung und Forschung, Grant 01UG0711. I would like to thank my colleagues at the Zentrum für Allgemeine Sprachwissenschaft in Berlin, Dagmar Bittner, Natalia Gagarina and Milena Kühnast for helpful comments. Special thanks go to Ute Bohnacker for insightful comments on an earlier draft of this paper. Insa Gülzow Zentrum für Allgemeine Sprachwissenschaft Schützenstr. 18 D-­‐10117 Berlin 1 Establishing discourse referents Abstract The use of indefinite noun phrases by German children aged 1;11-5;05 and adults was
investigated with a story-telling task in a situation of mutual knowledge. While adults
produced typical antecedent constructions, two-year-olds used indefinite noun phrases mainly
as nominal predicates in naming constructions. Children in the older age groups used less
indefinite noun phrases and produced higher proportions of definite expressions. Especially
the three-year-olds used comparatively high proportions of pronominal determiners for
referent introductions. The results suggest that children initially employ a context-bound
(naming) strategy and overestimate the role of deictic pronouns and mutual knowledge in a
narrative task. The strategy becomes more cohesion-oriented in the older children and
displays the basic requirements of referential linking, but still lacks the systematic use of
indefinite noun phrases for referent introduction.
2 Establishing discourse referents Introduction The use of an indefinite noun phrase (NP) is dependent on knowledge both on the level of morphosyntax and pragmatics. In naming structures like (1) for instance, the indefinite NP acts as a predicate identifying the class membership of a referent, while in (2) the indefinite NP is an argument introducing a discourse-­‐new referent. (1) That’s a fox. (2) And then the bird notices a fish on the table and goes for it. The current study will explore the role of these two use types in the early narrative productions of German children. Previous studies in various languages have shown that indefinite NPs as part of naming structures appear early at about two years of age (e.g. Kupisch, 2006, Rozendaal & Baker, 2008 for spontaneous production; Karmiloff-­‐Smith, 1979 for experimental data). The appearance of indefinite determiners to mark a discourse-­‐new referent on the other hand is dependent both on the text type and the children’s ability of taking the listener’s perspective into account. Preschool English children mark discourse-­‐new referents with indefinite NPs in spontaneous speech (Rozendaal & Baker, 2010) and in elicitation experiments (Schaeffer & Matthewson, 2005) before they do so in narratives (Hickmann, 2003) and French nine-­‐year-­‐olds master the use of indefinite NPs in no mutual knowledge (NMK) contexts before eleven-­‐
year-­‐olds do so in mutual knowledge (MK) contexts (Kail & Hickmann, 1992). These findings suggest that the predicative use of indefinite NPs and the use of indefinite NPs as arguments in a narrative MK context lie relatively far apart. Studies exploring German children’s narrative strategies show similar results. Bamberg (1986) found that children around age three hardly use indefinite NPs for referent introduction in a MK context at all and it is reported by Hickmann (2003) that the use of indefinite NPs for first mentions approaches adults’ distribution around age seven in a NMK context. The 3 Establishing discourse referents results of a study by Gülzow and Gagarina (2007) suggest that at phases in development when children are still learning to differentiate and express perspectives other than their own, indefinite NPs are used in functions other than the head of a referential chain. Similar proportions of indefinite NPs are documented in the narrative data of two-­‐year-­‐
olds and adults tested in a MK context. Clearly, the use of indefinite NPs by two-­‐year-­‐
olds contrasts with the near absence of this expression in older children and calls for an explanation. The present study will focus on how children who are in the process of acquiring their target system of referential expressions use indefinite NPs in a narrative MK context. As it seems that in this context a gap occurs between the children’s use of indefinite NPs as predicates and as arguments, the study will also address the children’s use of other referential expressions. It is well known that children use the definite determiner (e.g. de Cat, 2009; Hickmann, 2003; Karmiloff-­‐Smith, 1979; Maratsos, 1974, 1976; Rozendaal & Baker, 2008; Schafer & de Villiers, 2000; Wigglesworth, 1990) in places where the indefinite determiner would be (more) appropriate. Less is known about children’s use of pronouns instead of indefinite NPs, but Bamberg (1986) for instance found that German children use pronouns to introduce referents in narratives. These children generally mark the main protagonist with pronouns irrespective of whether the character is introduced, reintroduced, or reference is maintained. The overuse of the definite determiner on the other hand is explained by the children’s failure to take the listener’s perspective into account (e.g. de Cat, 2009), or an overestimation of the role of deixis (e.g. Karmiloff-­‐Smith, 1979). Deixis is a characteristic element of children’s early references and precedes referential linking. Karmiloff-­‐Smith (1979) noted in her experimental studies that children use definite determiners to signal the deictic function before they use them to signal an anaphoric relation. In 4 Establishing discourse referents German, pronominal reference is possible with two forms that mark the same distinction: while one form displays a clear deictic component, the other appears mainly as an anaphor. If at early stages of development children overuse definite forms and preferably encode deixis, it can be expected that not only definite determiners are used to introduce referents in a MK context but also pronominal forms encoding deixis. The German system of referring expressions Pronominal reference usually signals an anaphoric relationship to a previously mentioned referent, but pronouns can also be used for first mentions together with a deictic gesture. In German the personal pronoun ermasc/siefem/esneut ‘he/she/it’ (PRO) or the PRONOMINAL DETERMINER dermasc/diefem/dasneut (PRO det) can be used to indicate a referent present in the shared discourse. The term DEMONSTRATIVE is avoided here, as it may cause confusion with the proximal/distal pair of pronominal expressions in German dieserprox/jenerdist ‘this/that’. While deixis easily relates the pronominal determiners dermasc/diefem/dasneut to referents in MK contexts, deictic personal pronouns ermasc/siefem/esneut ‘he/she/it’ are the dis-­‐preferred option. (3) DERmasc/ERmasc (pointing at one student) kam zu spät. PRO det/PRO too late ‘HE was late’. came In a narrative, deictic reference is not common even if the referents are present or provided on pictures and pointing would in principle be an option. However, the different deictic potentials clearly separate the two German pronominal forms into a more exophoric (pronominal determiner) and a more endophoric (personal pronoun) form. In the present study the distribution of different referential expressions is 5 Establishing discourse referents analyzed. If the distributional analysis shows an effect of the er/der contrast in a phase were children overuse definite determiners this will add to our understanding of how children master the shift from a predominantly context-­‐bound, deictic system of reference to a system displaying the potential of marking more textually oriented reference. Indefinite NPs that occur as arguments in anaphoric chains clearly represent a relatively advanced stage in this development. The predicative use of indefinite NPs in naming structures has so far not been related to the development of referential linking or deictic reference. If naming structures do not represent an isolated use type, but are an integral part of young children’s reference systems, it should be possible to relate them to the children’s early deictic phase as well. Naming structures and deixis In his classical study, Brown (1973: 347f) argued that in naming structures like That’s a train or That’s a bear children identify a referent by its lexical label in a MK context. This reflects the on-­‐going process of acquiring an entity’s label. Once the child has acquired the name of the category, the definite determiner can be used as in Look at the train and Look at the bear. Karmiloff-­‐Smith studied the acquisition of French determiners in an extensive number of experiments and found that the “indefinite article is initially used as part of a procedure for naming, i.e. in its appellative or nominative function” (Karmiloff-­‐Smith, 1979: 216). While Brown makes his observation on the basis of whole sentences like That’s a train, Karmiloff-­‐Smith’s data consists of indefinite NPs only as children answered questions like What’s in the bag? for instance with A doll. This is an important point to note as Karmiloff-­‐Smith discusses the role of deixis for early uses of the definite determiner, but not for the indefinite determiner although naming structures that are realized by a whole sentence involve a deictic element in preverbal 6 Establishing discourse referents position. In the German example below, the preverbal position is taken by the deictic pronominal determiner dasneut and the indefinite NP functions as a nominal predicate assigning a property to the referent of the subject NP. (4) Das ist EIN BÄR. that is a bear ‘That’s a bear’. The main function of this construction is to assign class membership to an entity that is deictically identified. The preverbal position can also be taken by the local deictic expression da. (5) Da ist EIN BÄR. there is a bear ‘There’s a bear’. German da allows both a deictic and an existential interpretation. In a MK context, da can be deictic: the location indicated by da can be associated with the referent of the indefinite NP. Together with examples like (4), these (with deictic da) will be called NAMING STRUCTURES in the present study. Indefinite NPs in pre-­‐ and postverbal position Naming structures realize the NEW-­‐LAST PRINCIPLE (cf. Hickmann, 2003: 61); the indefinite NP presents new information in post-­‐verbal position. Küntay and Koçbaş (2009) show that in the narrative data of preschool English and Turkish children “a very high percentage of presentational constructions attract the indefinite marker” and conclude that the choice of the indefinite determiner is related to the choice of construction 7 Establishing discourse referents (Küntay & Koçbaş, 2009: 90). In the study of Hickmann (2003: 231) and similar to the Küntay and Koçbaş study, most post-­‐verbal referent introductions are marked with the indefinite determiner by German and English children from seven years onwards. It is therefor likely that the preschool children in the present study prefer to use indefinite NPs in post-­‐verbal position, both in naming structures and as referring arguments. As referring arguments, indefinite NPs can represent post-­‐verbal objects and subjects; examples will be given in the results section. The current study will address German children’s early use of indefinite NPs in a MK narrative. It will focus on the changes that can be observed in the children’s use of indefinite NPs as predicates and as arguments and on the co-­‐occurring distributional changes regarding the children’s use of definite referring expressions. In a narrative, information about hearer knowledge may be in conflict with the successful use of an indefinite NP produced in a MK context and the children must be willing to violate the rule linking indefinite NPs to NMK contexts in spontaneous speech. On the basis of previous results, the following questions will be addressed in this study: (A)
How is the transition from using indefinite NPs as highly informative naming devices to using indefinite NPs as referring arguments reflected in the data? (B)
How does the distribution of definite NPs (definite determiners, pronominal determiners, personal pronouns) correlate with this transition? Method Participants The narrative data of five age groups was collected and analyzed. A total of 60 pre-­‐
schoolers between age two and five and 15 adults took part in the study. Due to 8 Establishing discourse referents technical problems only 13 of the transcripts in the age group of five-­‐year-­‐olds could be used. Table 1: Age and age range of subjects Materials The narrative was elicited using a sequence of six pictures that featured a bird, the skeleton of a fish and a fox as well as a tree and a table (see Appendix 1). The pictures were designed to offer opportunities for story telling and were mounted onto six pieces of cardboard. Choosing protagonists with male gender in German eliminated gender effects (der Fuchs ‘the fox’, der Vogel ‘the bird’, der Fisch ‘the fish’). Although little reference to the inanimate entities was expected, these were also controlled for male gender (der Tisch ‘the table, der Baum ‘the tree’). To avoid an effect of marking the protagonist by one kind of NP (cf. Bamberg, 1986), active and passive involvement of the fox and the bird in the action was distributed as evenly as possible. Procedure The children were normal hearing, monolingual speakers of German and recruited at
Kindergartens in Berlin and Potsdam, Germany. The adults were University students from
Berlin, also normal hearing, and monolingual speakers of German with no prior experience in
this kind of task. The parents of the children filled out a questionnaire documenting language
development and the languages spoken at home in order to avoid bilingual or multilingual
speakers and children with language impairment. Additionally, a reduced version of a
standardized test was used to assess typical language development in German (Kauschke &
9 Establishing discourse referents Siegmüller, 2009). The children were seen individually by one experimenter in a quiet room
with a table, two chairs and a video camera that recorded the children’s narration and filmed
their hands to capture all instances of pointing. The child and the experimenter sat opposite
each other at a low distance with no barrier between them to emphasize the situation of shared
attention. Once the children seemed comfortable at the table, the experimenter asked them to
take part in a story-telling task: Ich habe Dir ein paar Bilder mitgebracht, Du kannst gleich
eine Geschichte dazu erzählen ‘Here are some pictures for you to tell a story in a moment’.
Before the onset of the actual story-telling the children were presented the sequence of six
pictures in a row. Erstmal zeige ich dir alle Bilder, damit Du weißt, was in der Geschichte
passiert ‘First, I will show you all the pictures so that you know what happens in the story’.
When the children were ready to tell the story they were presented the first picture. Hier ist
das erste Bild. Was passiert denn da? ‘This is the first picture. What is happening’? After the
children had finished with the first, the second picture was positioned to the right side of the
first and served as the new stimulus. The subsequent pictures were presented in a way that the
picture on the left (e.g. the first) was taken away so that the subsequent picture (e.g. the
second) could take its place. Then the new picture (e.g. the third) would be placed on the right
and serve as the stimulus. Apart from the very first picture children always had two pictures
in front of them: the previous picture and the one serving as the current stimulus for the
narrative. The experimenter supported the production of utterances by encouraging the
participants and praising their performance Was passiert denn da? Prima. Und hier? ‘What’s
happening? Very good. And here’? Constant feedback also served to give the children a sense
of common ground. The experimenter avoided sentences that included NPs referring to the
participants so that the children were not influenced by the experimenter’s choice. The adults
were treated to the same procedure and were encouraged in the same way as the children.
This results in an experimental situation that is somewhat unnatural for the adults and
findings cannot be interpreted taking the adults’ performance to represent the norm. However,
10 Establishing discourse referents as adults behaved in the way they did, it will be assumed that the children’s productions
approach adult distributions in the course of language development.
It may seem a daunting task to ask two-­‐year-­‐olds to produce a narrative, but we wanted to explore the narrative strategies of young children before they conform to adult standards. The procedure was chosen because very young children could be overburdened by looking at too many pictures at once. By reducing the number of pictures we hoped to elicit more references as the children concentrated on one picture at a time and could only switch to the next when the experimenter decided to let them do so. As focusing attention on one picture may discourage children from producing a coherent narrative and lead them into describing individual pictures instead (cf. de Cat, 2011), we hoped to support attempts at referential linking between pictures by making the previous picture available. Results Types of constructions with indefinite NPs The types of utterances in which indefinite NPs occur will be analyzed in this section. The examples below are taken from the adult data of the present study (expect for the first and the second) and represent the different categories used for analysis. The first category includes examples of utterances without verbs: (6) Ein Fisch. a fish ‘A fish’. child 130: 2;4,6 [without verb] Among the utterances with verbs, predicative naming structures constitute the next category [NAM]. 11 Establishing discourse referents (7) Das hier ist EIN RABE. that here is a raven ‘That’s a raven here’. child 76: 2;08,01 [Nam] Naming structures observe the new-­‐last principle. So do structures with indefinite NPs as referring arguments that appear as objects in post-­‐verbal position [svO]; a capital bold letter indicates the indefinite NP. (8) Also, der Vogel klaut EINE FISCHGRÄTE. well the bird steal a fishbone ‘Well, the bird is stealing a fishbone’. adult 13, [svO] A subject indefinite NP in preverbal position indicates newness not by position but morphological marking; both a post-­‐verbal object [Svo] or a post-­‐verbal adverbial [Svx] is possible. (9) EIN VOGEL sieht das a bird ‚This is seen by a bird and…’. see und... that and (10) EIN VOGEL fliegt über die Wiese. fly adult 16, [Svo] a bird ‘A bird is flying across the lawn’. adult 12, [Svx] across the lawn In German, subject-­‐verb inversion occurs if some other element occupies the first position [xvS]. 12 Establishing discourse referents (11) Auf dem Teller liegt EIN TOTER FISCH. on the plate lie ‘A dead fish is lying on the plate’. adult 12, [xvS] a dead fish A first step compares the number of utterances without verbs to the number of utterances with verbs (Figure 1). In this first analysis, the categories [Nam], [svO], [Svo], [Svx], and [xvS] are collapsed into the category ‘with verbs’. Figure 1: Use of indefinite NPs with and without verbs Figure 1 shows that the majority of indefinite NPs used by two-­‐year-­‐olds are utterances that contain no further constituent (81%, N=62). While structures with verbs increase proportionally across age groups, raw numbers show that their overall use decreases. In the next step, the children’s utterances with verbs will be analyzed with regard to the type of construction that contains the indefinite NP (Figure 2). Figure 2: Clause structure of utterances with indefinite NPs Figure 2 shows that two-­‐year-­‐olds exclusively produce naming structures (12) when using an indefinite NP together with a verb. Three-­‐year-­‐olds and four-­‐year-­‐olds use naming structures in a majority of cases while in the data of the five-­‐year-­‐olds indefinite NPs are mainly used as referring arguments. In the data of the adults there is only one example of a naming structure, the remainder are indefinite NPs used as referring arguments. (12) Hier ist EIN VOGEL. child 105: 2;07,17 here is a bird 13 Establishing discourse referents ‚Here’s a bird’. Of the few examples of indefinite NPs used as referring arguments that are documented in the data of the children, most occur in post-­‐verbal position. In a majority of cases, these indefinite NPs are objects rather than subjects. (13) Und da hat there has der Vogel EIN FISCH. and ‘And the bird has a fish there’. child 79: 4;05,28 the bird a fish In the adults’ data, utterances are almost evenly distributed between the three categories svO, Svo/Svx and xvS. Thus in a majority of cases indefinite NPs are subjects rather than objects, but only about half of the subjects occur in topic position. In the children’s data, examples with subject indefinite NPs in preverbal position are not documented. NP types in first mentions Six categories were used for analysis: indefinite NPs like ein Vogel ‘a bird’ (14) and definite NPs like der Vogel ‘the bird’ (15); for the children bare nominals like Fisch ‘fish’ were also included (16). Pronominal NPs included pronominal determiners like der ‘PRO det’ (17) and personal pronouns like er ‘he’ (18). In German, it is not easy to distinguish between an utterance with the pronominal determiner dermasc and the personal pronoun ermasc if the preceding word ends in an alveolar plosive dann rennt er/der ‘the he runs’. The decision was made for the personal pronoun er if a pause was detectable before the pronominal form. The rest category includes other types of 14 Establishing discourse referents referring expressions such as the reflexive pronoun sich ‘x-­‐self’ and null subjects, or demonstrative dieser ‘this’. (14) Und das ist EIN VOGEL… and ‘And that’s a bird’. child 28: 3;0,22 that is a bird (15) DER VOGEL will the bird wants the fox ‘The bird wants to eat the fox’. (16) Da hat der den Fuchs auffressen. there has ‘There he has caught the fish’. (17) DER PRO det fish PRO det falls into water ‘He falls into the water. (18) Dann will ER den Raben fressen. then want he the raven child 20: 3;4,2 caught fällt ins Wasser eat FISCH gefangt. child 28: 3;0,22 child 48: 4;0,10 child 53: 4;5,17 eat ‘Then he wants to eat the raven’. NPs that refer to a participant for the first time are regarded as first mentions. NPs that do not refer to a participant for the first time are regarded as subsequent mentions. In the present study, subsequent mentions are not further differentiated into reference maintenances and referent reintroductions. The design of the study is an inadequate testing ground for the maintenance/reintroduction contrast as the three participants closely interact, they all have male gender and there is no main protagonist. It has been shown that the need to resolve ambiguity in this study setting leads to the use of lexical expressions even in cases where reference is maintained (cf. Bittner & Kühnast, 2012). 15 Establishing discourse referents This aspect will be returned to in the discussion where results are also related to the Bamberg (1986) study. Table 2: NPs used for first mentions Table 2 shows that references are mostly nominal (bare nominals, NPs with definite or indefinite determiners) and that none of the participants in the study use the pronouns ermasc/siefem/esneut ‘he/she/it’ for first mention of a referent. Two-­‐year-­‐olds predominantly use bare nominals and indefinites for first mentions. Across age groups, the use of bare nominals decreases with age, which is a clear sign of the children’s reference systems becoming more adult-­‐like. The proportion of indefinite NPs also decreases, which sharply contrasts with the adults’ data where indefinite NPs are used for first mentions in a majority of cases. Across age groups children increase their use of the definite determiner. While adults do not use pronominal NPs for first mentions at all, children use the pronominal determiner dermasc/diefem/dasneut. Especially the data of the three-­‐year-­‐olds displays a high proportion of pronominal determiners for referent introduction. Table 3: NPs used for subsequent mentions NP types in subsequent mentions In subsequent mentions children of all age groups produce all five referential expression types. Adults use no indefinite NPs or bare nominals, but rely on NPs with a definite determiner in a majority of cases. Quite to the contrary, two-­‐year-­‐olds 16 Establishing discourse referents predominantly use indefinite NPs and bare nominals. This use decreases to adult level across age groups for both the indefinite determiner and bare nominals. The remaining referential expressions used by adults are personal pronouns and pronominal determiners plus a rest category that mainly consist of the reflexive pronoun sich ‘x-­‐self’ (62 out of a total of 136). In contrast to the adults, all age groups of children and especially the three-­‐year-­‐olds use a high amount of pronominal determiners. They increase their use of personal pronouns from almost zero in the group of two-­‐year-­‐olds above the level of adults in the group of five year-­‐olds. When comparing first with subsequent mentions within age groups, no significant differences were found for the group of two-­‐year-­‐olds (χ2=6.55, d.f.=4, p=.161). For all other age groups on the other hand, the difference between first and subsequent mentions is significant (three-­‐year-­‐olds: χ2=26.87, d.f.=4, p<.001; four-­‐year-­‐olds: χ2=54.13, d.f.=4, p<.001; five-­‐year-­‐olds: χ2=47.15, d.f.=4, p<.001; adults, χ2=411.2, d.f.=3, p<.001). This finding suggests that there is a categorical shift between the two-­‐
year-­‐olds and the older children in that the youngest children do not differentiate their marking of first and subsequent reference by the use of different types of NPs. Fisher’s exact test was performed across ages for first and subsequent mentions. Results reveal significant differences between age groups both for first mentions (χ2=110.5, d.f.=16, p<.05) and subsequent mentions (χ2=772.9, d.f.=20, p<.05). Discussion The aim of the present study is to determine the role and ontogenesis of indefinite NPs in young German children’s narratives produced in a MK context. The results of the study suggest that preschool German children who are faced with a narrative task do not use indefinite NPs in the same way as adults do, but start with a nominative naming 17 Establishing discourse referents strategy. The oldest children in the study have not yet developed full control of the adults’ use of indefinite NPs as antecedents in MK contexts, but clearly, they have abandoned the naming strategy. In the following, the results of the study will be related to the two questions that were raised in the introduction. It will be shown that despite the near absence of indefinite NPs in the older children’s data, they have achieved important steps on their way to referential linking. The naming function in German narratives The first research question concerns the nature of the children’s transition from using indefinite NPs as highly informative naming devices to using indefinite NPs as referring arguments. I will first comment on the distribution of the two use types across age groups and then briefly discuss their relation in terms of the syntactic position of the indefinite NP. Similar to findings in spontaneous production (e.g. Brown, 1970; Kupisch, 2006) and experimental data (cf. Karmiloff-­‐Smith, 1979) the children in the study initially use the indefinite determiner to signal the naming function. As long as children fail to realize that the task in the present study demands the production of a cohesive narrative, or the task verges on their linguistic and cognitive abilities, it is likely that they rely on the conventions of spontaneous speech. The two-­‐year-­‐olds in the study exclusively produce naming structures, three-­‐ and four-­‐year-­‐olds use indefinite NPs both in naming structures and as referring arguments, five-­‐year-­‐olds use indefinite NPs predominantly as referring arguments, and adults use indefinite NPs only as referring arguments. Clearly, the change in the distribution of the two use types marks a developmental change towards a more adult like usage of indefinite NPs. 18 Establishing discourse referents At the onset, the children’s use of indefinite NPs mirrors labeling routines that can frequently be found in spontaneous speech (19). (19) adult: What’s that? child: A fox. adult: Yes, and what’s that? child: A fox. The example shows that the multiple appearances of indefinite NPs in places other than first mentions are licensed in labeling. Unlike adults who use indefinite NPs no more than once to introduce a new referent in a MK narrative context, children up to the age of four produce indefinite determiners both in first and subsequent mentions. The children’s willingness to use indefinite NPs in subsequent mentions (with and without verbs) correlates with a general predominance of naming structures over indefinite NPs used as referring arguments. That is, children who give up naming structures in MK narratives more clearly mark subsequent mentions by the use of forms other than indefinite NPs. This relation is especially evident in the two-­‐year-­‐olds: these children exclusively produce naming structures when using indefinite NPs together with verbs and do not differentiate between first and subsequent mentions in their use of referring expressions at all. All other age groups use indefinite NPs (also) as referring arguments and show a distributional difference between the referential expressions used for first and subsequent mentions. It can therefore be concluded that not only the first appearance of the use type itself (indefinite NP as referring argument) marks an important step in referential linking, but its appearance in the group of three-­‐year-­‐olds 19 Establishing discourse referents coincidences with the realization of a basic requirement of referential linking: the general differentiation of first and subsequent mentions. Parallel to the increase in the proportion of indefinite NPs used as referring arguments, the use of indefinite NPs generally decreases to 10% (N=4) in the oldest age group (cf. Bamberg, 1986; Hickmann, 2003). The somewhat contradictory finding that children increase their skills regarding the formation of anaphoric chains and expand the variety of structures with indefinite NPs (see below) while at the same time they seem to abandon their use will be returned to when discussing the second research question. The new-­‐last principle Although naming structures can in principle serve as antecedents in referential linking (That’s a fox. The fox is chasing the bird.), the adults in this study introduce referents with indefinite NPs used as referring arguments. In these structures, subject indefinite NPs are most typical with 30% occurring in preverbal (Svo/Svx) and 32% occurring in post-­‐verbal position (xvS). These results align with those of Hickmann (2003: 202) who found that the German adults in her narrative study use a majority of post-­‐verbal subjects in subject-­‐verb inversions (about 50%) and a high number of preverbal subjects (about 30%) when introducing referents with an indefinite NP (Hickmann, 2003: 203f). Regarding position, post-­‐verbal indefinite NPs realize the NEW-­‐LAST PRINCIPLE (Hickmann, 2003: 231) and are typical in the adult data of the present study. In addition to 32% post-­‐verbal subjects (xvS, see above) adults produce another 24% post-­‐
verbal objects (svO). When using indefinite NPs as referring arguments, children do not produce examples of pre-­‐verbal Svo and Svx structures, but only produce examples of 20 Establishing discourse referents xvS (20), (21) and svO (22). These share the post-­‐verbal position of the indefinite NP with naming structures (23). (20) Da taucht ein Fuchs auf. there appears a fox verb.part ‘A fox appears there’. (21) Auf dem Teller liegt ein toter Fisch. On the plate a dead fish. ‘A dead fish lies on the plate’. (22) Der Vogel hat einen Fisch auf dem Tisch gefunden. child 189: 5;5,13, [svO] the bird has a fish on the table ‘The bird found a fish on the table’. (23) Da ist ein Fuchs. there is a fox ‘There is a fox’. lies adult 2, [xvS] adult 12, [xvS] found child 20: 3;4,2, [Nam] Besides the post-­‐verbal position of the indefinite NP, examples of xvS often exhibit a sentence-­‐initial locative expression that they share with naming structures. The sentence-­‐initial position can either be filled with the locative da ‘there’ (21) typical for naming structures, or some other locative expression such as auf dem Teller ‘on the plate’ (22). In the present study, children older than two produce utterances representing both the xvS and the svO category, but no examples of Svo/Svx. It can therefore be argued that the children are sensitive to input frequencies with regard to the post-­‐verbal position of indefinite NPs and that they observe the new-­‐last principle. The children gradually expand the naming structure by using verbs other than the copula and a wider variety of locative expressions in xvS structures and preserve the post-­‐verbal position both in their use of xvS and svO. 21 Establishing discourse referents Definite referring expressions The second research question concerns the use of definite referring expressions and the relation of their distribution to the children’s transition regarding the use of indefinite NPs as described above. I will discuss the general distribution of referring expressions in each age group and comment on the main changes between adjacent age groups. Two-­‐year-­‐olds: indefinite NPs and naming structures The exclusive use of indefinite NPs in naming structures is typical for the two-­‐year-­‐olds whose referential systems consist mainly of bare NPs, NPs with definite determiners and especially NPs with indefinite determiners (cf. Table 2 and Table 3). These children do not differentiate between first and subsequent mentions in their choice of referential expressions and anchor reference in the situational context. The referential systems of two-­‐year-­‐olds are mainly nominal and incorporate deixis when using naming structures for the identification of referents. Three-­‐year-­‐olds: pronominal determiners and deictic reference As already mentioned, the appearance of indefinite NPs as referring arguments in the data of the three-­‐year-­‐olds contributes to the shift towards referential linking which is also documented in the finding that different distributions of forms occur in first and subsequent mentions. In comparison to the two-­‐year-­‐olds, the referential systems of the three-­‐year olds become more definite and less nominal; uses of bare nominals and the indefinite determiner decrease and uses of the definite determiner and the pronominal determiner increase both in first and subsequent mentions. Children are gradually abandoning the naming strategy and it is likely that they begin to relate referents across 22 Establishing discourse referents pictures in order to construct a story line. The definite determiner is the most frequent expression in first mentions and the pronominal determiner is the most frequent expression in subsequent mentions (cf. Table 2 and Table 3). Clearly, this distribution of forms displays some basic knowledge of adequately ranking expressions to encode different degrees of hearer knowledge (cf. Gundel, Hedberg & Zacharski, 1993) and thus represents another step towards referential linking. In comparison to the adults, the use of the pronominal determiner for first mentions in narratives is atypical (for spontaneous production see Rozendaal & Baker, 2010) and will be interpreted here as a symptom of a referential system that is still firmly anchored in the situational context. Bamberg (1986) also found that preschool German children introduce referents with pronominal forms in a narrative and that they use proportionally more pronouns than adults performing the same task. He argues that the children in his study use pronominal forms for the main character irrespective of whether the character is introduced or reintroduced or reference is maintained. Bamberg collapsed the der/er forms into one category, which makes it impossible to assign different functions to the forms and show developmental tendencies. In the present study however there is a clear contrast: pronominal determiners are much more frequent than personal pronouns and it is only the pronominal determiners that are used for referent introductions. The deixis-­‐first principle Contrary to the Bamberg study, the setting of the present study does not feature a main protagonist, but active and passive involvement of the characters ranking high on the animacy scale (the fox and the bird) is balanced (see appendix 1). This study setting is therefore an inadequate testing ground for Bamberg’s finding as children may find it difficult to identify a main character. The fact that the results resemble Bamberg’s 23 Establishing discourse referents findings nonetheless in addition to the fact that only pronominal determiners are used in first mentions suggests a strong influence of factors in addition to or other than differences between the marking of main and secondary characters. It is argued here that one influence on the referential system of preschool children (among others) is what will be called the DEIXIS-­‐FIRST PRINCIPLE. The deixis-­‐first principle is not restricted to deitic expressions such as the pronominal determiner, but can affect different expressions at different stages in development. In the case of naming structures it applies to a whole structure and not to an individual expression, but the mechanism is the same: indefinite NPs occur in structures that can be used for deictic reference before they occur in structures that are typical for referential linking. Lyons (1975) argued that deixis is the source of reference and that the anaphoric use of pronouns is derivative of their function as deictics. If children’s understanding of referring expressions is based on their prior understanding of deixis, it comes to no surprise that an overuse of pronominal determiners with a strong deictic potential is documented in the data of the younger children in the study. It can be argued that the children use pronominal determiners to mark a salient referent in the common ground available to both speaker and hearer, but (still) fail to realize that in a story-­‐telling narrative the use of pronominal NPs for referent introduction is non-­‐standard. As the progressive decline of naming structures that can be observed in the group of three-­‐year-­‐olds reduces the children’s opportunities to deictically realize reference in the situational context, the use of pronominal determiners represents an alternative for this kind of use. In comparison to the two-­‐year-­‐olds, the referential system of three-­‐
year-­‐olds is reorganized with respect to the use of nominal and pronominal expressions and the marking of first and subsequent mentions, but reference is in large parts still anchored in the situational context. 24 Establishing discourse referents Four-­‐year-­‐olds: definite determiners in first and subsequent mentions In the group of four-­‐year-­‐olds the use of the pronominal determiner generally decreases and the willingness of children to introduce referents with the pronominal determiner drops down to 5 percent (N=2). This finding is a clear indicator that the children refrain from following the conventions of spontaneous speech in their use of pronouns for referent introduction (cf. Rozendaal & Baker, 2010). With regard to nominal expressions on the other hand, children behave similar to adults in spontaneous speech when introducing a new referent to a MK context and prefer to use the definite determiner (cf. Rozendaal & Baker, 2008: 789). The indefinite determiner is also used for first mentions but still as part of a naming structure in the majority of cases. It can therefor be concluded that the children’s choice of a nominal expression for referent introduction is strongly influenced by the deictic or ostensive potential of an expression. While naming structures feature a deictic expression in sentence-­‐initial position, definite determiners used in a MK context are not deictic expressions per se but can be regarded as ostensive in that they refer to entities present in the situational context (cf. Lyons, 1999: 160f). Karmiloff-­‐Smith (1979) has argued that the definite article functions as a deictic (pointing) device in early phases of language acquisition. She separates the function of articles into DESCRIPTORS, which link the visual and social context to language, and DETERMINORS, which link linguistic contexts. According to Karmiloff-­‐Smith, children acquire the descriptor function of articles prior to their determinor function. Following this line of reasoning, the deixis-­‐first principle can be interpreted as an extension of Karmiloff-­‐Smith’s (1979: 215) model. The deixis-­‐first principle affects the whole system of referring expressions in preschool children which develop from deictic systems that 25 Establishing discourse referents are anchored in the situational context to more textually oriented systems that allow the expression of referential linking. In the subsequent mentions, the four-­‐year-­‐olds reach an adult level of definite determiners (cf. Table 2 and Table 3). In both the adult’s and the children’s data, the high amount of definite determiners in subsequent mentions can be interpreted as an effect of frequent topic change since the story was controlled for an even distribution of agency among the characters. Children are still unable to exploit the potentials of the er/der contrast (see below) for marking different participants (cf. Bittner & Kuehnast, 2012; Bosch Rozario & Zhao, 2003), but a definite NP can be used to identify the intended referent with a lexical label. The increasing amount of definite determiners in subsequent mentions can therefore be interpreted as a consequence of understanding the need to unambiguously identify a referent. In German adult L2-­‐data for instance, over-­‐explication with nouns is attributed to a concern for communicative success (Hendriks, 2003). In the present study, equal gender among the characters adds to the same effect and may also explain the adults’ use of a comparatively high amount of unambiguous reflexive pronouns. As already mentioned, the proportion of the pronominal determiner in subsequent mentions is much lower than in the three-­‐year-­‐olds but it is still well above the proportion that adults produce. When using pronominal forms in subsequent mentions adults rely on personal pronouns, which clearly signal referential linking and the need to find an antecedent. Four-­‐year-­‐olds on the other hand rely on the potentially deictic pronominal determiners when using pronominal forms in subsequent mentions. It can be concluded that the four-­‐year-­‐olds systems of reference are similar to those of the three-­‐year-­‐olds in that the anchoring of reference in the situational context is still visible in their use of the indefinite determiner in naming structures, the predominance of the 26 Establishing discourse referents pronominal determiner (in subsequent mentions) and the children’s use of the definite determiner for referent introduction. The four-­‐year-­‐olds systems of reference are similar to the adults on the other hand in that lexical expressions are used for referent introduction and the most frequent expressions in subsequent mentions are NPs with the definite determiner. Five-­‐year-­‐olds: personal pronouns versus pronominal determiners, the er/der contrast The indefinite determiner represents referring arguments in a majority of cases in the group of five-­‐year-­‐olds. In subsequent mentions, the ratio of personal pronouns increases and proportionally equals the children’s use of pronominal determiners (cf. Table 2 and Table 3). Clearly, both findings must be interpreted as a further step towards adult realizations of referential linking. Adults show a clear preference for personal pronouns and thus demonstrate a textually oriented strategy of reference. It is likely that the adults’ use of the pronominal determiner does not exploit the expression’s deictic potential, but is motivated by the er/der contrast in German that allows a differentiation of topic maintenance and topic shift. According to Bosch et al. (2003) the German personal pronouns are more restrictive than the pronominal determiners in that the former signal topic maintenance and the latter prefer less salient referents. Along similar lines it is argued in CENTERING THEORY that the process of anaphor resolution connects the shortest expression (ermasc/siefem/esneut ‘he/she/it’) to the most salient discourse referent (Grosz, Joshi & Weinstein, 1995). (24) Peri möchte mit Emilj ins Kino gehen. Eri / derj Per wants with Emil to.the cinema go ‘Per wants to go to the cinema with Emil. But he isn’t allowed to’. darf PRO / PRO det allow aber nicht. but not 27 Establishing discourse referents In the example given, the personal pronoun ermasc ’he’ more easily relates to the topic of the preceding sentence Per while the pronominal determiner dermasc more easily relates to Emil and thus signals a topic change. This distinction is unavailable to preschool children (cf. Bittner & Kuehnast, 2012), but may motivate the adult’s choices. Across age groups personal pronouns are not used for first mentions even in the youngest age group. This can be interpreted as a sign of an early awareness of the expression’s different conditions of use in comparison to the pronominal determiner. The comparatively late increase of personal pronouns can be interpreted as an effect of the gradually unfolding competence of the children with regard to referential linking. Expressions like personal pronouns that are closely associated with the need to find an antecedent seem to attract the children’s attention relatively late. The children realize that referent introduction on the basis of deixis or ostension is inadequate in a narrative before they realize that German not only provides a pronoun that can index a referent in the situational context but also a pronoun that can index a referent in the preceding text. Definite determiners remain on a similar level both in first and subsequent mentions in the group of five-­‐year-­‐olds; children still have to realize that indefinite determiners are not unnecessarily unspecific, but can be used for referent introduction in MK contexts. Conclusion The children’s use of the indefinite determiner in narratives follows a similar direction as their use of other referring expressions in development. Their early systems of reference develop from context-­‐bound to more textually oriented in that (potentially) deictic realizations decrease and (potentially) anaphoric realizations increase. It has been argued here that children initially follow a deixis-­‐first principle that affects all expressions; it could e.g. be shown to manifest in the children’s early use of indefinite 28 Establishing discourse referents determiners in naming structures and in the distribution of the two German pronouns er/der. When the use of expressions with a deictic potential becomes less frequent, the decrease does not parallel a random increase of other expressions, but it could be shown that the distribution of forms changes systematically towards more textual orientation. The somewhat paradoxical findings that the use of indefinite NPs marks both the onset and the end of the development of narratives in a MK context could be resolved as it is different use types that dominate in the two phases. The use of indefinite NPs in naming structures and as referring arguments is nevertheless not unrelated and it could be shown that naming structures are structurally related to the children’s early use of indefinite NPs as referring arguments. References Bamberg, M. (1986). A functional approach to the acquisition of anaphoric relationships. Linguistics 24, 227-­‐284. Bittner, D. & Kuehnast, M. (2012). Comprehension of intersentential pronouns in child German and child Bulgarian. First Language 32/1-­‐2, 176-­‐204. Bosch, P., Rozario, T., & Zhao, Y. (2003). Demonstrative pronouns and personal pronouns. German der vs. er. Proceedings of the EACL 2003: Workshop on The Computational Treatment of Anaphora. Budapest. Brown, R. (1973). A first language: the early stages. Cambridge, MA: Harvard University Press. De Cat, C. (2009). Experimental evidence for preschoolers' mastery of topics. Language Acquisition 16, 224-­‐239. De Cat, C. (2011). Information tracking and encoding in early L1: linguistic competence vs. cognitive limitations. Journal of Child Language 38, 828-­‐860. Grosz, B. J., Joshi, A. & Weinstein, S. (1995). Centering: A framework for modeling the local coherence of discourse. Computational Linguistics 21, 203-­‐225. Gülzow, I. & Gagarina, N. (2007). Noun phrases, pronouns and anaphoric reference in young children narratives. ZAS Papers in Linguistics 48, 203-­‐223. Gundel, J., Hedberg, N. & Zacharski, R. (1993). Cognitive status and the form of referring expressions in discourse. Language 69, 274-­‐307. Hendriks, H. (2003). The use of nouns in reference maintenance: the seeming contradiction in adult second language acquisition. In A. Giacolone Ramat (ed.), Typology and Second Language Acquisition, Berlin, New York: Mouton de Gruyter, 291-­‐326. Hickmann, M. (2003). Children’s discourse: Person, space and time across languages. Cambridge, MA: Cambridge University Press. 29 Establishing discourse referents Kail, M. & Hickmann, M. (1992). French children's ability to introduce referents in narratives as a function of mutual knowledge. First Language 12, 73‑94. Karmiloff-­‐Smith, A. (1979). A Functional Approach to Child Language. Cambridge, MA: Cambridge University Press. Kauschke, C. & Siegmüller, J. (2009). PDSS -­‐ Patholinguistische Diagnostik bei Sprachentwicklungsstörungen (2. standardisierte, vollständig überarbeitete Auflage). München: Elsevier. Küntay, A. C., & Koçbaş, D. (2009). Effects of lexical items and construction types in English and Turkish character introductions in elicited narrative. In J. Guo, E. Lieven, N. Budwig, S. Ervin-­‐Tripp, K. Nakamura & Ş. Özçalışkan (eds.), Crosslinguistic approaches to the psychology of language: Research in the tradition of Dan Isaac Slobin, New York, London: Psychology Press, Taylor & Francis Group, 81-­‐92. Kupisch, T. (2006). The acquisition of determiners in bilingual German-­‐Italian and German-­‐French children. München: Lincom Europa. Lyons, J. (1975). Deixis as the source of reference. In E. L. Keenan (ed.), Formal Semantics of Natural Language, 61-­‐83. Cambridge: Cambridge University Press. Lyons, C. (1999). Definiteness. Cambridge, MA: Cambridge University Press. MacWhinney, B. (2000). The CHILDES Project: Tools for analyzing talk. 3rd Edition. Mahwah, NJ: Lawrence Erlbaum Associates Maratsos, M. (1974). Preschool children's use of definite and indefinite articles. Child Development 45, 446-­‐455. Maratsos, M. (1976). The Use of Definite and Indefinite Reference in Young Children. Cambridge, MA: Cambridge University Press. Rozendaal, M. I. & Baker, A. E. (2008). A cross-­‐linguistic investigation of the acquisition of the pragmatics of indefinite and definite reference in two-­‐year-­‐olds. Journal of Child Language 35, 773–807. Rozendaal, M.I. & Baker, A.E. (2010). The acquisition of reference: Pragmatic aspects and the influence of language input. Journal of Pragmatics 42, 1866-­‐1879. Schaeffer, J. & Matthewson, L. (2005). Grammar and Pragmatics in the Acquisition of Article Systems. Natural Language and Linguistic Theory 23: 53-­‐101. Schafer, R. & de Villiers, J. (2000). Imagining Articles: What a and the can tell us about the emergence of DP. In Proceedings of the Boston University Language Development conference. Somerville: Cascadilla Press. Wigglesworth, G. (1990). Children’s narrative acquisition: a study of some aspects of reference and anaphora. First Language 10: 105-­‐125. 30