The role of segments and prosody in the identification of a speaker's dialect

The objective of this study is to investigate the role of segments, rhythm, and rhythm combined with intonation in the identification of a speaker's dialect. In a between-subjects design using three conditions, we tested 62 listeners (Zurich Swiss German) in a two-alternative-forced choice dialect identification experiment: in condition one, 21 listeners were asked to identify two dialects (Valais and Bern Swiss German) in unmorphed form. In condition two, 20 different listeners had to identify the same two dialects but with swapped speech rhythm, and in condition three, 21 different listeners had to identify the same dialects with swapped speech rhythm and intonation. The experiment showed that exchanging speech rhythm alone or speech rhythm combined with intonation had very little effect on the listeners’ dialect identification performance: listeners appear to use primarily segmental information in the identification process. Further results revealed that (a) superimposing the prosodic structure of one dialect (Bern Swiss German) onto another (Valais Swiss German) caused greater variability across some listeners than the other way around and that (b) identification performance varies as a function of sentence material used, i.e. how the sentences differ in segmental and prosodic make-up. We discuss implications for forensic phonetics, language and cognition, and automatic speech recognition.This research is funded by the Swiss National Science Foundation, grant Nr. P300P1_151210, http://p3.snf.ch/project-151210