Mental representation of tonal spreading in Bemba: Evidence from elicited production and perception

Previous research has shown that listeners from tonal languages are better at processing tone compared to speakers from non-tonal languages. However, most of this research has tested Asian tone languages, particularly those which have many tonal contrasts and a dense tone-to-syllable association. In this paper we investigate the mental representation of derived tones in Bemba, a Bantu language that has a two-way tone contrast but which shows robust tone spreading patterns. Specifically, we test ternary high-tone spreading, a process that is unique from a phonological perspective. In a production task we test whether ternary spread can be extended to non-words. We complement this with an AX discrimination task comparing binary vs ternary spread, which are phonologically contrastive, on the one hand, with a tonally similarly salient but non-phonologically relevant contrast, on the other. We show that in both the production and perception of non-words, ternary spread is distinct from binary spread, suggesting that derived tone is equally mentally represented as lexical tone is in Asian tone languages. Introduction Evidence suggests that speakers of tone languages, who utilise pitch to signal lexical differences, have different representations for tonal aspects than speakers of non-tonal languages (e.g. Van Lancker and Fromkin 1973, 1978; Gandour 1983, see below for more detail; Repp and Lin 1990; Lee et al. 1996; Ye and Connine 1999; Bent et al. 2006; Zheng et al. 2007; Francis et al. 2008; Braun and Johnson 2011; Chiao et al. 2011). These studies, conducted on Mandarin Chinese, Cantonese, Taiwanese and Thai, have a number of findings of significance, in particular: (i) different lateralisation for lexical (tone) and non-lexical (intonation) pitch processing (Van Lancker and Fromkin 1973, 1978); (ii) more integral processing of segments (and in particular vowels) and tones in tone language speakers (Braun and Johnson 2011; Repp and Lin 1990); (iii) an advantage for vowel over tone information in vowel and tone monitoring tasks (Ye and Connine 1999); and (iv) more categorical processing (in particular identification) of tone by tonal listeners (Sun and Huang 2012) but no difference in tone discrimination in AX tasks with short inter-stimulus intervals (Cutler and Chen 1997). (See Francis et al. 2003 for a more critical view on the categorical perception of tones). However, tone languages are not homogenous and thus while the findings above hold of the Asian tone languages investigated, they cannot be assumed to also hold for Bantu tone languages which differ in many critical respects from Asian language tone (Yip 2002). First, Bantu tone languages have fewer tonal contrasts than seen in Mandarin, Cantonese, Taiwanese and Thai (2–3 as compared to 4–7). Second, in Bantu tone languages, only a few vowels/morphemes are lexically specified for tone, while nearly all vowels/morphemes are tonally specified in the type of Asian tone languages investigated. Third, in Bantu tone languages, the surface tonal realisation is derived from intricate tone-spreading rules (see e.g. Clements and Goldsmith 1984; Hyman and Kisseberth 1998, and below), while there are considerably fewer tonal (sandhi) processes in Mandarin, Cantonese, Taiwanese and Thai (Yip 2002: 173ff). While it is a legitimate assumption that tone is mentally represented as part of the vowel that carries it in the Asian tone languages investigated, this assumption is likely not tenable for Bantu tone languages, at least not for derived tones. If this is the case we expect tonal processing for speakers of this type of Asian language and Konstanzer Online-Publikations-System (KOPS) URL: http://nbn-resolving.de/urn:nbn:de:bsz:352-0-321211 Erschienen in: Southern African Linguistics and Applied Language Studies ; 33 (2015), 3. S. 307-323 https://dx.doi.org/10.2989/16073614.2015.1108768