On the basis of these findings, it is possible to draw three general con- clusions concerning the acquisition of the linguistic devices necessary for discourse cohesion across domains: the timing and rhythm of acquisition (Section 11.2.1); the impact of sentence-internal and discourse functional determinants (Section 11.2.2);
and the impact of universal vs. language-specific factors (Section 11.2.3). I take each of these conclusions in turn, pointing out how different findings bear on them in each domain, before turning to some similarities and differences that can be observed across domains (Section 11.2.4).
11.2.1 The timing and rhythm of acquisition
The first general conclusion concerns issues revolving around the rel- atively precocious or delayed timing of first language acquisition. The findings across domains show that discourse-internal uses of linguistic devices are a rather late development, which emerges at about six to seven years of age and continues to evolve until at least ten years of age or even thereafter. This first conclusion sug- gests a delayed unfolding of the acquisition process, at least if one seriously takes into account contextual constraints affecting the uses of linguistic devices. Such a conclusion apparently goes against the conclusions of some previous developmen- tal studies, although such conclusions are often based on insufficient evidence that does not allow us to differentiate discourse-internal uses of linguistic devices from earlier deictic uses of the same forms.
11.2.1.1 Animate entities
The analyses of how children mark information status for referent in- troductions show that systematic uses of various markings of newness do not occur before about six to seven years of age in all languages. The first newness mark- ings to occur at this age are local ones (indefinite or numeral determiners, specific classifiers), including in Chinese where they are optional, whereas global markings emerge yet later on, even in Chinese where they are obligatory (see further discussion in Section 11.2.2 below). Clause structure variations serving discourse functions do occur early, mostly in French (e.g. uses of presentative existential constructions and dislocated structures from four years on). However, these structures are not used with discourse-internal functions until a later age, when children become able to vary clause structure in order fully to differentiate newness from givenness. For example, dislocations are at first used for all types of topic promotion, including with the first mention of referents (for which they are inappropriate), and they are later on reserved for reference maintenance (as required by the adult system), once appropriate markings of newness are acquired. Finally, we also observed deictic
General conclusions across domains uses of these markings. In comparison, the analyses concerning reference mainte- nance show fewer changes with age, as a result of strong discourse constraints on the uses of referring expressions across languages from the youngest age on (also see further discussion in Section 11.2.2 below).
11.2.1.2 Space
The findings in the domain of spatial reference show the early impact of the typological properties of the languages compared. Cross-linguistic differences can be observed at all ages tested in relation to predicate use for the expression of motion. I return to this point below, when I turn to language-specific factors in ac- quisition (Section 11.2.3). At this point, note that this finding shows children’s early knowledge of the structural properties of their native language (satellite-framed vs.
verb-framed orientation). In sharp contrast to this early knowledge, the findings also show their very late mastery of principles governing spatial anchoring in discourse.
In this respect, the introduction of spatial anchors presented children with some difficulties, resulting in a very late developmental progression. Roughly, only the adults in the present sample provided adequate spatial anchoring for locations and displacements in their narratives. Children only begin to do so at about seven years and they have not mastered spatial anchoring even at ten years. This developmental progression is strikingly similar in all languages, contrary to the original hypothe- sis, given the expected cross-linguistic differences that are indeed observed in other respects. I attributed it to the development of children’s ability to plan discourse in such a way as to anticipate their interlocutor’s task in reconstructing the locations and location changes of the characters in the story. In addition, anchoring in this domain shows a later developmental progression as compared to the marking of information status for animate referents (see Section 11.2.4 below).
11.2.1.3 Time
With respect to children’s uses of temporal-aspectual devices, uses of verbal inflections (Indo-European languages), of aspectual particles (Chinese), and of connectives (all languages) by young children differ from those that are observed in the older children’s and adults’ narratives. Analyses of temporal-aspectual shifts in the narratives show that it is only at about seven years in the present sample that children begin to use verbal morphology and particles for the grounding of informa- tion in discourse, and this development continues until ten years and thereafter. In addition, uses of connectives evolve with increasing age in conjunction with these morphological distinctions. In particular, regional devices marking various types of temporal relations among denoted events increase with age, particularly after seven years. However, such devices can be observed earlier in Chinese, where they
Conclusions
constitute a major means of discourse organisation. The particular contexts for the discourse-internal uses of both verbal devices and connectives from seven years on show that they serve a number of functions, such as the marking of overlaps among events and of referent introductions. Before this age, temporal-aspectual devices are mainly used for the description of successive events, with little differentiation between the discourse foreground and background.
11.2.2 Determinants in language acquisition: sentence vs. discourse factors
The evidence from the three different domains shows the impact of two main determinants: syntactic and semantic factors operating within the utterance and functional pragmatic factors operating across utterances in discourse. The particular functional factors that were considered in the present study concerned two discourse- internal principles governing the regulation of information flow across utterances:
the marking of information status and the grounding of information (see Chapter 2).
A more complete account of functional determinants should of course also consider other functional aspects of linguistic devices. The data presented, however, do show that children’s acquisition of linguistic devices in all three domains and across several languages results from an interplay among some factors that affect how they learn to represent events in well-formed utterances (who did what to whom when and where) and others that affect how they learn to organise these utterances in a well-formed stretch of discourse (what is in focus or presupposed across utterances).
11.2.2.1 Animate entities
Both clause-internal and interclausal factors have an impact on how children referred to the story characters. In this domain, clause-internal factors con- cern the expression of syntactic and semantic roles (subject vs. object, agent vs.
patient), while interclausal factors concern the marking of information status in dis- course (new vs. given information, thematic status). These two determinants jointly affect children’s uses of referring expressions (nominal determiners, pronouns) and of clause structure (word order, clause types). Thus, syntactic roles are the main de- terminants of NP position in reference maintenance (preverbal subjects, postverbal objects). They also determine to some extent the position of referent introductions before the point where global markings of newness emerge, but after this point introductions tend to be postverbal (although to various extents across languages, see Section 11.2.3 below). Syntactic roles are also seemingly related to form vari- ations (pronominal subjects, nominal objects), although this relation was shown to be merely the by-product of discourse factors, particularly the presence/absence of immediate coreference and the particular types of coreference relations involved.
General conclusions across domains Discourse factors are in fact the main determinants of form variations, overriding most other factors. These factors determine the forms of first mentions, at least after the emergence of newness markings, as well as the forms of subsequent mentions in all languages, particularly the choice of lean vs. full forms. Finally, uses of global markings of information status are less likely in languages that heavily rely on these markings for the expression of grammatical relations. As discussed below (Section 11.2.3), this finding not only suggests the impact of language-specific factors in lan- guage acquisition, but it also shows that the availability of some linguistic devices for the organisation of discourse is partly the result of the functions served by the same devices for the organisation of grammatical relations at the sentence level.
11.2.2.2 Space
With respect to space, clause-internal factors include mainly the struc- tural properties of the languages (satellite-framed vs. verb-framed) and therefore the extent to which information is grammaticalised or lexicalised in different com- ponents of the clause. These typological properties clearly affect the information that is selected and the ways in which this information is organised within or across clauses. This effect is most evident in the analyses of the different predicates that are used in the narratives, particularly in relation to the representation of motion events (see Section 11.2.3 below). Interclausal factors in this domain include the distribution of some information across clauses (e.g. the compact vs. distributed representation of manner vs. path information), as well as the information status of spatial grounds at different points in discourse (first vs. subsequent mentions, beginning vs. end of story). Given the semantic properties of various types of static vs. dynamic predicates, children’s choices of one or the other predicate type par- tially depend on whether and how spatial reference points have been set in discourse.
Children’s uses of different forms to set these reference points and to maintain ref- erence to them also partly depend on the semantic role of the noun phrases used in relation to predicates for first and subsequent mentions. However, the findings also show a general reliance on world knowledge in this domain and the role of grounding processes in relation to the story characters (see Section 11.2.4 below).
11.2.2.3 Time
Finally, in all languages both verb semantics and interclausal dis- course factors affect children’s uses of temporal-aspectual devices The bounded vs. unbounded nature of predicates partly determines uses of past and/or perfective verbal devices. Shifts in verbal devices also serve to mark relations among events (overlaps), information status (referent introductions), and more generally the dis- tinction between the foreground vs. background of discourse. Furthermore, uses of
Conclusions
connectives (conjunctions and adverbials) were shown to co-occur with particular verbal inflections or aspect particles as a function of these two factors. In gen- eral, uses of bounded predicates and of perfective markings are more likely in the foreground, while uses of unbounded predicates and imperfective (or aspectually unmarked) markings are more likely in the background.
11.2.3 Universal vs. language-specific aspects of acquisition
The results also show similarities and differences across languages in all domains, suggesting two facets of acquisition: some aspects of development can be generalised to all languages, presumably reflecting universal aspects of language acquisition, while others are clearly language-specific, showing the impact of the particular properties of children’s language on acquisition. I illustrate these two facets of development, before turning to their implications below (Section 11.3).
11.2.3.1 Animate entities
With respect to how children denoted the story characters, some im- portant differences were found across seemingly related languages, as well as some similarities across typologically distinct languages. For example, referring expres- sions and clause structure are not used to mark information status at the same rate across languages. In English and Chinese, where local markings are less complex than in French and German, young children do not rely on clause structure at all to mark referent introductions, ‘reserving’ this type of device for the marking of grammatical relations. This finding complements the results of studies focusing on sentence comprehension showing that English speakers (including adults and children as young as two years) heavily rely on word order to interpret grammatical relations, as compared to other languages such as Italian or French (MacWhinney and Bates 1989; Charvillat and Kail 1991). Although the Chinese four/five-year- olds use postverbal position more than the English group, they use it less than expected, given that NP position constitutes an obligatory marking of information status. Thus, regardless of the optional vs. obligatory nature of different newness markings, local markings are cognitively less complex than global ones, as is also the case in sentence comprehension (Ammon and Slobin 1979; Slobin and Bever 1982; Slobin 1982, 1985a). In this study the greater functional complexity of global markings results from their simultaneous contribution to different levels of organi- sation: the sentence and discourse. Indeed, the main function of Chinese determiners is to mark information status (in addition to their numerical function), whereas po- sition marks both information status in discourse and grammatical relations in the clause. Again, these results complement other studies of sentence comprehension in Chinese (MacWhinney and Bates 1989; Chang 1992).
General conclusions across domains Furthermore, children’s heavy reliance on word order to mark grammatical rela- tions in English might explain why local newness markings are used least frequently by the four/five-year-olds in comparison to the same age groups in all other lan- guages. In particular, I found a strong association between postverbal position and indefinite forms in the referent introductions of all languages. I also noted an asso- ciation between preverbal position and definite forms, but mostly among the young English speakers. That is, these children used frequent preverbal definite forms to denote referents in subject role within the clause, in comparison to other children of the same age who relied more on clause structure to introduce referents. At four/five years the French children turned out to be the ones who used clause structure the most for referent introductions (e.g. frequent existentials). Young French children who do not introduce referents appropriately also frequently use definite left-dislocated forms, whereas older children use these constructions in conjunction with appropri- ate introductions, thereby promoting NPs to topic status before pronominalisation.
This and other results suggest that French might be more akin to ‘topic-oriented’
languages than to ‘subject-oriented’ ones in some respects, as has been pointed out by other researchers (see Lambrecht 1981; Bates and Devescovi 1989; Dasinger and Toupin 1994). Finally, some similarities in reference-maintenance were attributed to universal discourse pragmatic factors that override the cross-linguistic differences observed with both referent introductions and reference maintenance (e.g. uses of zero elements and of other referring expressions as a function of coreference).
11.2.3.2 Space
As shown above, spatial anchoring follows a very similar develop- mental progression, which was attributed to universal cognitive factors. In sharp contrast, the results also show that children’s organisation of spatial information in discourse from four/five years on is influenced by the particular ways in which their language grammaticalises or lexicalises spatial information. Children’s productions clearly reflect the typological distinction between satellite-framed and verb-framed languages. In English, German, and Chinese dynamic predicates are extremely var- ied, frequently and productively combining multiple pieces of information that are compactly packaged within the clause by means of verb roots and satellites. Such predicates provide information concerning manner, deixis, direction, and causati- vity, thereby showing a concern for representing motion in great detail. In contrast, French dynamic predicates are much less varied, they rarely encode multiple types of information within the clause, and the few distinctions encoded are marked in the verb roots. As a result, less information is provided about motion and some infor- mation is frequently left to be inferred by the interlocutor (e.g. particularly manner and causativity in the case of changes of location). Finally, when multiple types of
Conclusions
information are provided, they are typically distributed across clauses in discourse, rather than compactly packaged within the clause as in the other languages. Such differences clearly follow the pattern that is found in the adult languages, reflecting the satellite-framed vs. verb-framed nature of the system to be acquired by the child.
Similar differences have been reported in other studies. For example, the patterns observed in the present study are in line with those reported by Berman and Slobin (1994), who contrast three verb-framed languages (Spanish, Hebrew, and Turkish) and two of the satellite-framed languages represented in our study (English and German). I return below (Section 11.3) to the relativistic views of development that have emerged from such results.
11.2.3.3 Time
With respect to temporal-aspectual markings, language-specific pat- terns suggest the need to qualify the defective tense hypothesis (see Weist 1986;
Weistet al. 1984). In the present study predicate types have a different impact on tense/aspect markings across languages. Children’s use of the past seems to depend on the relative complexity and transparency of the markings available in the partic- ular system they are acquiring. It is likely that children’s reliance on the present in French and German is related to the fact that this tense neutralises aspect, while the past tenses are more complex, since they require a choice between perfective and imperfective forms. In contrast, no such difference in cognitive complexity occurs in English, where the aspectual opposition progressive vs. non-progressive is inde- pendent of tense. As a result, thehistorical presentis functionally more marked in English than in German or French and it is therefore more frequent in the English narratives of the adults than in those of the children. The greater effect of verb semantics in English is also partly due to this factor, since children use a greater range of inflections and therefore have more ‘degrees of freedom’ in comparison to the French or German children.
In addition, although discourse context is a major determinant for the use and non- use of perfective markers in all languages, the relative impact of this factor varies among the young children across languages. From seven years on these markings are used to relate events in various ways, thereby differentiating the background and foreground in discourse, whereas the four-year-olds’ uses of these markings coincide mainly with descriptions of events and/or of results. This pattern at four/five years, however, is most salient in French and Chinese. In Chinese this pattern is presumably linked to the tight association between bounded predicates and perfective markers, resulting in frequent perfective uses with the resultative predicates that are typically involved in such description contexts. In French the high frequency of temporal- aspectual shifts in event and/or result descriptions may be related to the fact that French sharply opposes perfective and imperfective aspect in the past (and only in