R You Sure That Your /r/ Is Trilled? A Methodological Caveat - Archive ouverte HAL Access content directly
Conference Papers Year :

R You Sure That Your /r/ Is Trilled? A Methodological Caveat

François Pellegrino
Dan Dediu
  • Function : Author


Databases are great tools for typological studies and for understanding the evolution of language, but they can rarely be taken at face value. Many (if not all) databases have various limitations and usage constraints (see, for example, for WALS (Cysouw, Dediu, & Moran, 2012; Hunley, Bowern, & Healy, 2012) and UPSID (Simpson, 1999)), which can "trickle down" to (or even be amplified by) the studies based on them. We focus here on phonemic databases, which can be used, among others, to compare phonemic inventories across languages and to make claims about language evolution through perceptual or articulatory mechanisms for instance. However, such databases often suffer from an intrinsic ambiguity between phoneme/allophone/phone, which often percolates down to research that aims to study the phonetic systems of languages from (supposedly) phoneme inventories (for a recent example, see (Winter, Sóskuthy, Perlman, & Dingemanse, 2022)). One common hurdle is that sometimes the phoneme is seen as a monolithic object without variation in its phonetic realizations, and without contextual dependencies. Moreover, written symbols can be misleading because the same symbol may represent different phonetic realities (Anderson et al., 2018). To drive these points home, we will focus here on r (Barry, 1997). Not all segments are directly comparable, and the rhotics class is a perfect example because of their great phonetic variability (Chabot, 2019; Lindau, 1985). In particular, by default the rhotic r is often considered as an alveolar trill: from a phonemic point of view, UPSID (Maddieson, 1984) reports that among the languages that have a rhotic, at least half have one alveolar trill or more. But to say that a language has an alveolar trill phoneme does not inform about its potential phonetic realizations. Several studies show that the trill alveolar phonemes have few of their allophones that are actually trilled (Sebregts, 2014; Blecua, 2002; Rafat, 2010), potentially leading to a distorted picture. As an example, we consider here a recent article (Winter et al., 2022) which correlates the presence of the trilled /r/ sound with "roughness" cross-modally. Their result is supported by several approaches, one being a cross-linguistic statistical study of 332 languages,
JCoLE2022_paper_202_Anselme.pdf (65.56 Ko) Télécharger le fichier
Origin : Files produced by the author(s)

Dates and versions

hal-03900020 , version 1 (15-12-2022)


  • HAL Id : hal-03900020 , version 1


Rémi Anselme, François Pellegrino, Dan Dediu. R You Sure That Your /r/ Is Trilled? A Methodological Caveat. Joint Conference on Language Evolution, Sep 2022, Kanazawa, Japan. ⟨hal-03900020⟩
3 View
0 Download


Gmail Facebook Twitter LinkedIn More