Defining and Counting Phonological Classes in Cross-linguistic Segment Databases

Defining and Counting Phonological Classes in Cross-linguistic Segment Databases Dan Dediu, Scott R. Moisik Max Planck Institute for Psycholinguistics Wuldtlaan 1, Nijmegen, The Netherlands Dan.Dediu@mpi.nl, Scott.Moisik@mpi.nl Abstract Recently, there has been an explosion in the availability of large, good-quality cross-linguistic databases such as WALS (Dryer & Haspelmath, 2013), Glottolog (Hammarstr¨om et al., 2015) and Phoible (Moran & McCloy, 2014). Databases such as Phoible contain the actual segments used by various languages as they are given in the primary language descriptions. However, this segment-level representation cannot be used directly for analyses that require generalizations over classes of segments that share theoretically interesting features. Here we present a method and the associated R (R Core Team, 2014) code that allows the flexible definition of such meaningful classes and that can identify the sets of segments falling into such a class for any language inventory. The method and its results are important for those interested in exploring cross-linguistic patterns of phonetic and phonological diversity and their relationship to extra-linguistic factors and processes such as climate, economics, history or human genetics. Keywords: phonetics, phonology, segment classes 1. Introduction The patterns of phonetic and phonological diversity are very important to understanding the wider lin- guistic diversity, the processes shaping it and its evol- ution. For several decades now, databases such as the UCLA Phonological Segment Inventory Database (UPSID) (Maddieson, 1984) have played a major role in allowing (semi-)quantitative analyses of the pat- terns of phonetic and phonological variation around the world. Unfortunately, until relatively recently, such databases used to be relatively small (for ex- ample, UPSID has 919 different segments in 451 lan- guages; http://web.phonetik.uni-frankfurt.de/ upsid_info.html) and offer a rather sparse geograph- ical and genealogical coverage. The recent advent of the World Atlas for Language Structures (WALS) (Haspelmath et al., 2005) and its online and continu- ously updated version (Dryer and Haspelmath, 2013) available at http://wals.info, offer better coverage (currently 2,679 languages, many with lots of missing data) but at the cost of providing only a small number of variables (currently only 20 are listed under “Phon- ology”; http://wals.info/feature) that are discret- ized into a limited number of categories (e.g., feature 1A “Consonant inventories” can only have five val- ues: “small”,“moderately small”,“average”,“moerately large”, and“large”), a representation that is hard to ac- comodate in quantitative analyses and whose use has justifiably generated intense controversy. Fortunately, the last several years have seen the release of phon- etic/phonological databases that are much richer and have better coverage, such as Phoible1 (Moran et al., 2014) and the dataset accompanying a recent compar- ison of phonological and genetics patterns of diversity 1Freely available at http://phoible.org. (Creanza et al., 2015) curated by Merritt Ruhlen2. However, such segment-level databases have a major drawback intrinsic to their design in that they cannot be directly used for analyses that require generaliza- tions over classes of segments that share theoretically interesting features, such as “front rounded vowels”, “retroflex stops” or “clicks”. In this paper we3 introduce a method that builds on the notion that a …

Content truncated to 500 words