In conclusion, let us consider some of the implications of these con- clusions in the light of existing controversies opposing different types of models that have been proposed to account for language acquisition. I focus below (Section 11.3.1) on the need to take into account two aspects of language acquisi- tion that emerge from the findings: the multifunctionality of language, particularly the contribution of various forms to different functions at the sentence and dis- course levels of linguistic organisation; and recurrent and variable aspects of the acquisition process across languages, whereby children come to terms with this type of multifunctionality. I close (Section 11.3.2) with a few concluding remarks concerning future research that is necessary to address further questions concerning the relation among different determinants of language acquisition and the relation between language and cognitive development.
11.3.1 Relevance for available models of language acquisition
The previous review of the literature in linguistics and developmental psycholinguistics contrasted different dimensions along which models of language acquisition diverge. Among them, we saw that models focus on two different levels of linguistic organisation: some focus mostly on the sentence level (sentence-based models), while others focus mostly on the discourse level (discourse-basedmodels focusing on cohesion or on coherence). As we saw, little has been undertaken in order to combine the contributions of these models in a more complete view that takes both sentence-internal and discourse functional determinants into account. However, as
Implications and concluding remarks they learn to produce well-formed utterances to represent events, children are also learning to regulate the flow of information across utterances in discourse as a function of various contextual constraints linked to presuppositions and focus. One of the aims of this book was to show that this type of multifunctionality is a central part of children’s language acquisition.
I have argued that discourse provides children with a developmental mechanism for the acquisition of linguistic devices in all three domains of reference. The tim- ing and course of this acquisition process are partly determined by the functional properties of these devices, which result from their simultaneous contribution to the organisation of the sentence and of discourse. This contribution to two levels of organisation follows some universal principles, as well as some language-specific ones. An adequate model of acquisition in this domain, then, requires an account of how discourse and sentence-internal functions both contribute to acquisition within a cross-linguistic perspective. As discussed below (Section 11.3.2), further research is necessary to determine precisely how these two levels of organisation interact during the acquisition process. However, the research presented does show that both levels of organisation clearly have an impact on the acquisition of the same linguistic devices. In all three domains, analyses that focus only on one or the other type of factor therefore provide a partial and insufficient account of language acquisition.
Furthermore, in all domains clause-internal and interclausal factors interact, result- ing in some mutual influences of the sentence and discourse levels, so that such a partial account might be misleading.
In addition, universal and language-specific factors both affect acquisition in different domains. Several types of universal determinants have been proposed, including at least the following: formal universals, reflecting the nature of Univer- sal Grammar; semantic universals, potentially reflecting some cognitive universals;
functional pragmatic universals, such as discourse principles regulating information flow, reflecting universals of human communication; and other cognitive universals, reflecting general properties of human cognition. With respect to cognitive univer- sals, general language-independent processes presumed to apply to all domains of child development have been most frequently invoked by previous studies to ac- count for language acquisition. For example, it is likely that some general cognitive process akin to decentring might play a role in children’s ability to regulate informa- tion flow in the absence of mutual knowledge, enabling them to take the perspective of others when it differs from their own and to adapt linguistically their discourse accordingly when necessary in particular communicative situations. We also saw that some general ability to organise discourse representations might be involved in the development of discourse cohesion, for example enabling children to plan ahead the necessary initial anchoring of personal, spatial, and temporal information and the further regulation of this information in subsequent discourse. Finally, it has
Conclusions
been argued that various types of cognitive schemata for complex event sequences underlie children’s verbal performance in narrative situations after the emergence of language (scripts, story schemata), as well as their capacity for temporal sequencing at an even younger age.
However, it remains to be seen whether such general processes provide sufficient explanatory principles for two reasons: it is unclear whether all cognitive processes involved in language development are entirely domain-independent and/or whether they are all independent of specific properties of the language to be acquired. I would first argue that some of these cognitive processes might be more specific to language use, being universally specialised for the production, comprehension, processing, and construction of language. Thus, the particular type of perspective-taking, plan- ning, or more generally organisational capacity that is necessary for children’s mature participation in discourse activity might be highly constrained by the prop- erties of language, and particularly by its organisation into a system of multiple form–function relations. As put forth by Vygotsky (see Chapter 2), language in this respect provides children with a very special multifunctional tool, which mediates higher mental activity. I would further argue that the discourse-internal functions of language provide children with the potential to decontextualise their represen- tations or, more precisely, to ‘re-contextualise’ them in this semiotic medium. In this process, then, children learn to rely maximally on discourse, rather than merely on non-linguistic context, when discourse cohesion and coherence are necessary for communication. They thereby become able to use language as itsowncontext in all kinds of situations, including those that are not strictly speaking communicative (e.g. problem-solving situations). Although other uses of language, such as script- like discourse types and/or deictically anchored face-to-face conversation, also ob- viously play a crucial role throughout children’s social and cognitive development, this new semiotic context further allows them to manipulate their representations, thereby simultaneously providing a huge cognitive flexibility and constraining this potential flexibility by new organisational principles. For example, the development of discourse cohesion allows children to go beyond the principle of chronological order for the purposes of information grounding or to temporarily foreground some otherwise backgrounded information for the purpose of marking information sta- tus. I also previously argued (see Chapter 4) that narrative coherence is intricately related to the use of language as its own context and that the development of dis- course cohesion could be partially constitutive of how children construct complex forms of cognitive macrostructures during the course of development.
Other types of universal determinants have been proposed, which may or may not be reducible exclusively to language-independent processes that would apply equally to all domains of development. These factors include syntactic and semantic categories (e.g. concepts of specific vs. non-specific reference, agency, subjecthood,
Implications and concluding remarks boundedness, kinesis). They also include other factors affecting cognitive complex- ity, which might be general enough to apply to all domains, but seem to be predomi- nant for language learning and perhaps even specialised for it. For example, consider the greater functional complexity of global markings in comparison to local ones in the domain of reference to entities. Clause structure participates in the marking of distinctions at two levels: grammatical relations at the sentence level and information status at the discourse level. Although languages differ with respect to their relative reliance on global vs. local markings for the first level (e.g. morphology is richest in German, but weak in English and especially in Chinese), global markings are functionally more complex than local ones and are acquired with more difficulty than local ones in all languages. This conclusion applies to Chinese, despite the fact that global markings constitute obligatory markings of information status, while local ones are merely optional. This type of result does suggest that some cognitive factors influence the rate and course of language acquisition. However, relative cognitive complexity is in this case tied to a very domain-specific learning process, one that is imposed on the child by virtue of the very special properties of language, namely its multifunctionality. Although multifunctionality is a property that may apply to any tool engaged in human action, the particular form of multifunctionality displayed by language is not to be found in any other domain, and it requires that the child pay attention to phenomena that are specifically linguistic in the course of his or her varied activities. If universal cognitive processes are involved, some are at least highly constrained by the fact that language has to become ‘its own problem-solving space’ (see Karmiloff-Smith 1992).
Whatever the case may be, language-specific factors also affect the rate and course of development, as shown by the different developmental patterns that were found in all three domains across languages: referring expressions and clause structure are not used to mark information status at the same rate across languages; chil- dren are clearly influenced by the ways in which their language structures spatial information; and cross-linguistic variations throw some doubt on the claim that the acquisition of tense-aspect systems should be entirely determined by language- independent concepts of situations. Such language-specific factors partially result from the different systems of devices available and from the different ways in which languages map sentential and discourse functions onto these devices, accounting for the cross-linguistic differences that were observed in different domains. Thus, despite some striking similarities in the developmental patterns that were observed in all domains with respect to discourse organisation (e.g. general principles of an- choring and of grounding personal, spatial, and temporal information), some results show that language-specific factors affect either the rate or the course of develop- ment. In some cases, the course of development for particular devices is similar across different languages, while the rate of development differs from language
Conclusions
to language (e.g. the use of newness markings when denoting animate entities).
In other cases, differences in the course of development appear from the very be- ginning (e.g. the structure of utterances representing motion events, the relation between predicates and past or perfective markings), while the rate of development is similar in other related respects (e.g. spatial anchoring, temporal grounding).
As we saw in previous chapters, the more general question of cross-linguistic variability in language acquisition has led to a partially ‘relativistic’ view of de- velopment, according to which language has an impact on how children categorise the world, on which aspects of the incoming information are most salient to them, and/or on how they organise information when they engage in the activity of com- municating in discourse. Such an impact has been observed in children’s uses of language during the emergence of language, as well as during later phases of devel- opment, particularly on the basis of the implicit categorisation of spatial relations (Bowerman 1996; Bowerman and Choi 2001), the representation of motion events (Berman and Slobin 1994; Choi and Bowerman 1991), and the temporal organisa- tion of discourse (Berman and Slobin 1994). One such relativistic view focusing particularly on the type of narrative activity studied in preceding chapters (Slobin 1996) proposes that from very early on speakers learn to engage in the activity of thinking for speakingon the basis of the patterns that are provided by their specific native language.
If the impact of language could be further observed on children’s perceptual, attentional, and cognitive processes in non-verbal situations, including during the prelinguistic period and during the periods that follow the emergence of language, such results would raise fundamental questions about major assumptions in some approaches, such as assumptions of innateness and modularity (also see Section 11.3.2 below). Such results would also cast doubt on models postulating only gen- eral cognitive abilities (such as Piagetian theory) and on models (such as Karmiloff- Smith 1992) postulating that development only proceeds from general (language- independent) capacities to modular (domain-specific) ones. In particular, they would show the structuring impact of language from the earliest period onwards, for which such models cannot account so far. The results from the present study do not resolve this question, but they nonetheless show the need to relate three types of capaci- ties, all of which seem to be involved during language acquisition: (1) general all-purpose (language-independent) cognitive and/or communicative capacities;
(2) domain-specific (syntactico-semantic) capacities for linguistic organisation at the sentence level; (3) domain-specific (semantico-pragmatic) capacities for lin- guistic organisation at the discourse level. Different models have focused primarily on one or the other types of capacity, but no adequate account has been proposed so far to explain how all of them might simultaneously play a role across behavioural domains and across languages during child development.
Implications and concluding remarks 11.3.2 Open questions and future research
This book has hopefully convinced the reader that adequate models of acquisition must take into account sentence and discourse determinants of acqui- sition, as well as universal and variable aspects of the acquisition process across languages. It has been argued that the exclusive focus on only one level of linguis- tic organisation clearly leads to an incomplete and perhaps misleading view of acquisition, while controlled cross-linguistic comparisons are indispensable tools to validate claims in the light of recurrent and diverging patterns in acquisition across languages. However, among other areas to be explored in the future, at least three further lines of cross-linguistic research would be essential to pursue some of the questions raised by this book: research relating more precisely the sentence and discourse determinants of language acquisition; research linking linguistic and non- linguistic behaviour throughout development; and research allowing us to articulate early capacities during the prelinguistic period and later capacities emerging with language during child development.
First, more research needs to be carried out in order to show the precise ways in which the sentence and discourse levels of organisation are related during develop- ment. Although our results and other studies show the impact of both sentence and discourse determinants, there are at least two different and complementary ways in which these levels might be related. On the one hand, syntactic and semantic factors determine which aspects of situations children focus on, thereby influencing how they organise information across utterances in discourse (see also Berman and Slobin 1994). On the other hand, discourse provides children with a developmental mechanism for the acquisition of linguistic devices, which influences how they or- ganise their sentences. The present study suggests that these influences presumably involve a two-way process, such that discourse factors influence some aspects of sentence organisation, while semantic and syntactic ones influence some aspects of discourse organisation. Thus, the structure of the spatial semantics in a given language, as well as the nature of its temporal-aspectual morphology, have a strong impact on the selection and organisation of spatio-temporal information across clauses in discourse. Similarly, the degree to which a language invites speakers to rely more or less heavily on word-order for the marking of grammatical relations may influence their relative reliance on clause structure for the purpose of mark- ing information status in discourse. However, we also saw that functional pragmatic factors linked to presupposition and focus in discourse strongly determine local and global form variations in all languages, for example affecting how speakers denote characters and spatial anchors in discourse (forms of referring expressions, clause structures) and partially determining various aspects of the spatio-temporal infor- mation represented (choices of predicates and of temporal-aspectual markings).
Although different strands of evidence show the impact of each of these levels of
Conclusions
linguistic organisation on the other, little is known about how to relate these two possible types of influences during the course of development. Some of the results indicate that the direction of these two types of influences may differ across domains and perhaps across languages, but that both types of influences can occur within the same domain and within the same language.
Second, within a cross-linguistic perspective, it is likely that variable aspects of language development have implications on all types of behavioural domains. If one takes the Whorfian view seriously, the prediction is that the properties that are specific to each linguistic system should affect not only how speakers organise their thoughts for the purposes of communication in interpersonal discourse situations (thinking for speakingin Slobin 1996), but also their perceptual, attentional, and cognitive processes more generally, including those that have been previously as- sumed to be independent of language structure and of language use. As previously noted (Chapter 2), such a prediction is supported by adults’ performance in various classification tasks, even though such tasks have been typically presumed to be non-linguistic in much previous research, showing that language may have a more pervasive structuring role on cognition than has been assumed so far (e.g. Lucy 1992a, 1992b). However, it is still unknown whether this impact can be observed in similar tasks with children, although some studies have begun to show differences in young children’s behaviours (from infancy onwards, see below) that may be re- lated to language-specific factors. For example, Gopnik and Choi (1990) compare children’s linguistic productions and their performance in non-verbal cognitive tasks across several languages (English, Korean, French), showing a correlation between cognitive differences and a heavier reliance on verbs in Korean and in French as compared to English. Such correlations are interesting, but must be explored fur- ther in order to determine how precisely verbal and non-verbal behaviours might be related.
For example, it would be revealing to test the hypothesis that the strikingly heavy reliance on verbs that characterises some child languages from the earliest phases of development onwards (e.g. see Bassano 2000 for French; Choi and Bowerman 1991 for Korean) is related to the verb-framed orientation of these languages and that the impact of this typological property may have general implications that go beyond the spatial domain examined in preceding chapters. More generally, the variable ways in which languages grammaticalise or lexicalise information leads to some predictions concerning the impact of language on cognitive processes. In this respect, Talmy (2000) makes a set of specific proposals concerning the implications of languages’ relative reliance on different devices for the relative salience of the coded information in cognitive functioning. For example, he proposes that different open-class and closed-class items can be ranked along a hierarchy that characterises the degree to which the encoding of information within the sentence is automatised