Final Report Summary - DISCOM (Discourse connectives and the mind: a cross-linguistic analysis of processing and acquisition) The DISCOM project investigated pragmatic aspects of language acquisition by normally-developing children and children suffering from autism, language processing by adult speakers in their native and non-native languages, and cross-linguistic differences between closely related languages as well as their effects on translation. These dimensions of study have been tackled in a unified manner through a number of empirical investigations focusing on the differences between various domains of use of discourse connectives. This multi-faceted and cross-linguistic exploration of discourse connectives provided answers to important questions for the relation between language and cognition, namely: the role of children’s mother tongue on cognitive development, the development of coherence in textual productions and the nature of the linguistic impairment in autism.Discourse connectives like ‘because’ or ‘so’ in English can be used to relate various kinds of propositional contents, such as two events unfolding in the world (The temperature rises because the sun is shining) or a premise and a conclusion in the mind of a speaker (Max is ill because he didn’t come to work). In the first example, the connective is used in what is called the content domain while in the second it is part of the epistemic domain. DISCOM investigated the cognitive differences between these different uses of connectives along the complementary axes listed above. In the domain of language acquisition, it has often been observed that languages vary in the number of connectives they possess for a given rhetorical relation. In Dutch, a causal relation is expressed by two different connectives depending on the domain of use (omdat and want) while in spoken French there is only one (parce que). However, studies conducted within DISCOM revealed that despite this difference, children acquiring the two languages have a similar ability to handle objective and subjective causal relations, indicating that cognitive development sets the pace for their acquisition.DISCOM also brought new answers concerning the role of connectives for text processing by adult speakers and school-age children, with important practical applications for pedagogical purposes. Indeed, the use of connectives is often also considered as an indication of global text quality. The studies conducted within DISCOM indicated that evolutions off usage in a language have direct consequences for processing and that children master the subtle processing cues of causal connectives very late in their development.In the field of autism, DISCOM contributed to highlight the existence of subtle impairments in complex syntax even at the higher end of the spectrum. These results led to the conclusion that some communicative impairments found in this population might reflect problems in the structural domains of language.Finally, the cross-linguistic comparisons made between French and Dutch in all these dimensions of study were complemented with corpus analyses of parallel corpora for other major Indo-European languages. These cross-linguistic comparisons also led to address the question of how non-native speakers understand and process the meaning of connectives in a second language when there is a discrepancy of usage with the learners’ first language. Results from these experiments confirm that transfer from the first language plays an important role for learners’ difficulties to understand the meaning of connectives in a second language. However, important differences are also found depending on the task. While learners scored lower than native speakers in an offline grammaticality judgment task, they demonstrated a native-like ability to process the meaning of connectives in an online reading task. The existence of cross-linguistic differences also open the possibility that connectives are used differently in translated texts compared to original ones due to source language interferences. Corpus-studies of multilingual parallel corpora revealed that source language interferences specifically occur when a connective does not have a one-to-one translation equivalent in the target language and is also dependent on the type of discourse relation conveyed.
