Variation of /r/: a challenge for speakers and listeners

Every person produces speech sounds differently. One speech sound that is especially prone to variation is the sound that, in Latin script, is represented as "r". In Austrian German, the sound /r/ can be produced as differently as by trilling the tip of the tongue against the gums behind the teeth, like in Italian, or by producing frication noise with the back of the tongue, like in French, with many other variants occurring alongside those. This vast amount of potential variation poses a challenge as well as an opportunity to understand the language system and the cognitive underpinnings of language use. The goal of the project is to understand how this variation in the sound /r/ affects speakers and listeners. If there is indeed no constraint on the production of /r/ then, in each word, speakers could randomly pick whatever variant they feel like. However, listening to Austrian German, the variation does not sound quite as random as this may suggest. In this project, a large-scale speech production study using audio recordings and ultrasound tongue imaging will assess how speakers produce the different variants of /r/ in different word and sentence positions. It will test whether the produced variation is at least in part systematic, for instance, with regard to stress or emphasis in words and sentences. For example, imagine someone saying "He bought a red rose" in reply to "What did he buy?" as opposed to a reply to "Did he buy a yellow tulip?". How would the /r/s in "red rose" sound in each case? In addition to linguistic phonetic analyses, machine learning techniques will be used to model acoustic variation and build an automated classification system of /r/ variants in different contexts. A complementary series of speech perception experiments will then test how listeners are affected by variation in /r/. It will assess whether listeners expect a given speaker to only use one variety of /r/, and whether and to what extent the word or sentence position affects their expectations and processing of the /r/ sounds. Would different variants of /r/ help or hinder the understanding of the abovementioned sentence about a red rose? Taken together, results of the project will speak to questions about how language works as a system by answering how different variants of sounds may interact with sentence information. It will also map out cognitive processes underlying speech production and perception by illustrating the interaction between speakers` freedom in variant choice and listeners` flexibility in tuning into this variation.