Liquid consonants and onset sonority in Keywords: Dogon languages, phonology, lexical statistics, liquids

Introduction Liquid consonants in Dogon languages (, Burkina Faso) are subject to strict phono- tactic restrictions and certain idiosyncratic alternations. In particular, several languages evidence a conspiracy such that any combination of liquid onsets across a derivational mor- pheme boundary is conflated to [l. . . r]. This presentation lays the empirical foundation for a comprehensive analysis of these phenomena by performing a comparative, lexical-statistical study of onset consonants in 12 Dogon languages.

Background Liquid alternations may affect either suffix consonants (e.g., Ben Tey /pile-le/ −→ [pile-re] ‘white-inchoative’; Heath 2015a) or root consonants (e.g., Nanga /kOri-ri/ −→ [kOlli-ri] ‘hook-reversive’; Heath 2016), or may involve a liquid mutation (e.g., Toro Tegu /bEru/ −→ [bela] ‘near-inch’; Heath 2015b) and a ‘flip-frop’ (e.g., Ben Tey /Oru-li/ −→ [Oli-ri] ‘moist-inch’; Author et al., 2016). Existing explanations appeal to the trochaic nature of feet in Dogon languages. Namely, (Heath, 2015b) appeals to unstressed deletion in Toro Tegu, and Author et al. (2016) appeal to embedded feet in suffixed trisyllabic words. It is unclear, though, how, if at at all, these proposals fit in with monomorphs and the lexicon as a whole.

Methodology A corpus of 27,603 monomorphs was constructed from the Dogon comparative wordlist spreadsheet (Heath et al., 2015). Table 1 provides a breakdown of number of unique forms within each language, as well as that language’s proposed internal genetic classification.

Table 1. Makeup of corpus

Classification Language Forms Classification Language Forms Western Mombo 2773 Eastern Yorno So 1923 Eastern Jamsay 2185 Northwestern Bankan Tey 2069 Eastern Perge Tegu 1940 Northwestern Ben Tey 2563 Eastern Togo Kan 1712 Northwestern Nanga 2330 Eastern Tommo So 2424 Northeastern Najamba 2815 Eastern Toro Tegu 2815 Northeastern Yanda Dom 1923

IPA transcriptions were simplified, and segments were categorized as either (P), (F), nasals (N), laterals (L), rhotics (R) and semivowels (J). Consonant clusters, though rare, were automatically syllabified according to their sonority profile, and the onsets of first, second and third were extracted. Results & discussion Representative results from Bankan Tey are presented in Figures 1a and 1b. Unsurprisingly (e.g., Smith (2004)), low-sonority onsets are overwhelmingly preferred in σ1 and σ2. Onsets in σ3 stand in stark contrast, however, in that /r/ is preferred after lower sonority onsets in σ2, while low-sonority onsets are again preferred in σ3 after σ2 liquid onsets.

(a) First and second syllables (b) Second and third syllables

Fig. 1: Correlations between onsets (Bankan Tey). Frequency colour is relative to onset type in syllable closer to left edge (= x-axis). Token counts indicated in tiles.

These results suggest not only a lexical avoidance of /l, r/ combinations but also a preference for third-syllable rhotics outside of this constraint. This information, along with other phenomena (e.g., vowel reduction, tonal shifts), may elucidate Dogon prosody and its potential role in the behaviour of liquid consonants.

References Heath, Jeffrey. 2015a. A grammar of Ben Tey (Dogon of Beni). Unpublished manuscript. Heath, Jeffrey. 2015b. A grammar of Toro Tegu (Dogon): Tabi Mountain dialect. Unpublished manuscript. Heath, Jeffrey. 2016. A grammar of Nanga: Dogon , Mali. Unpublished manuscript. Heath, Jeffrey, Laura McPherson, Kirill Prokhorov, and Steven Moran. 2015. Dogon comparative wordlist. Data retrieved from Dogon and Bangime Linguistics, https://dogonlanguages. org/sources/heathetal2015. Smith, Jennifer L. 2004. Making constraints positional: toward a compositional model of con. Lingua 114(12): 1433–1464.