We have conducted a cross-linguistic study investigating the role of pauses in online speech processing by Finnish, Swedish, and Russian speakers. We hypothesize that pauses have better potential to explain chunking behavior in Swedish compared to Russian and Finnish.
Arguably, humans speaking different languages share at least some similarities in the way they perceive spontaneous speech. We have conducted a cross-linguistic study involving Finnish, Swedish, and Russian to investigate the similarities and differences in online speech processing across those languages. The study builds on a dynamic model of online processing that works on English (Sinclair & Mauranen 2006; Mauranen 2018). We carried out an experiment using a custom web-based application (Vetchinnikova et al., 2017) that presents the participants with a speech extract matched to its broad orthographic transcription. We have chosen 100 spontaneous speech extracts of approx. 30 seconds in each language. We invited 150 native speakers (50 per language) to listen and chunk up the extracts into segments intuitively. We have analysed the resulting behavioral data in terms of consensus and annotated it for pauses (semi-manually, using WebMaus and Praat software) as well as clauses (manually, relying on a reference grammar from each language). Many studies report that there are no major differences in pause distribution across languages (Demol, Verhelst & Verhoeve, 2006 for Dutch, English, French, Italian, Romanian, and Spanish; Yang, 2007 for English and Chinese). However, we hypothesise that when listening to speech, speakers of different languages rely on pausal cues to a different extent. We will perform regression modeling using those variables to determine the role of acoustic and structural factors across languages. We suggest that pauses will explain a smaller proportion of chunking behavior variance in Russian and Finnish compared to Swedish due to the rich morphological structure in the first two.
