Spontaneous chunking while writing: semantic and constructional highlights

Abstract Summary

Abstract :

This paper takes the point of view of the performance in order to i) question the nature and the mechanisms of chunks and chunking while writing; ii) describe what we call here 'performance units', i.e. text sequences spontaneously produced between two pauses during the textualization process, recorded in real time with Inputlog (Leijten & van Waes 2006) and Scriptlog (Wengelin et al. 2009). Performance units like [pause] the direction of [pause] may be compared to the data analyzed online "in reception" (Chafe 1992; Sinclair and Mauranen 2006). In this, our study is situated at the junction of the two approaches proposed by the symposium, offline and online.







Chunking is a general cognitive mechanism that reflects the ability to group different pieces of information into a single cohesive unit (Gobet et al., 2001; Christiansen and Chater 2016; see also Sinclair's 1993 "encapsulation"). A chunk can therefore be defined from the point of view of its idiomatic status, as well as from the point of view of its semantic consistency and its capacity to underpin new textual strings. The present work will take the second option as a reference frame, within which will be discussed the potentially idiomatic status of performance units produced during composition of texts, in terms of constructional regularities (see Hunston & Francis 1999 on pattern flows; Blumenthal-Dramé 2012 on entrenchment; Legallois & Koch 2019 on motifs). Indeed, the analysis of 12 000 bursts produced in French shows less than 50% of bursts are shaped into syntactically valid categories such as phrases or sentences. We then switch to the evaluation of the semantic relevance of performance units and advocate for the necessity of preserving a relational dimension in chunks, due to priming and projection.

Pre-recorded video :