Is tone phonologically atomic? The significance of syllable shape in tone diachrony
The significant role that syllable shape plays in tonogenesis is not fully utilized in synchronic studies of segment-tone interaction. Some theories take tonemes to be phonologically atomic and composable with every syllable shape, and thus accounts in those frameworks must explain restrictions on tone distribution without recourse to diachronic origin. For example, of the five tones of Standard Thai, only three—low, high, and falling—occur on ‘checked’ syllables (those with stop codas). This restriction has been referred to as “puzzling” (Yip 2002:23) and “previously unexplained” (Morén & Zsiga 2006:116). Morén & Zsiga attribute it to a relationship between a glottal feature and low tones. In fact the origins of the distribution of Thai tones have been known for many decades, and syllable shape is fundamental to their origin (Gedney 1972). Diachrony implies that tone is not synchronically atomic, because syllable shape is partially deterministic of toneme distribution (as opposed to tone shape).
In my corpus of tonal inventories from 278 Tai doculects, all 25 possible combinations of two numerals under Chao (1930) tone notation are attested, as summarized in Table 1. Morén & Zsiga’s (2006) connection of particular phonetic tone shapes to an underlying glottal feature in the synchronic phonology is not likely sustainable, or must at least be considered too ad hoc to accept unless a complete factorial typology is considered.
It is now well understood that lexical tone compensates for the loss of segmental contrasts (Krauss 1973:963). However, syllables with stop codas behave differently from other syllable shapes in tonogenesis in important ways. First, they can remain atonal after other shapes have acquired tone, as in Hmong-Mien (Ratliff 2015). Second, the conditioning environment is not fully destroyed in the process, unlike for most other tonal categories, where codas are lost or laryngeal contrasts on onsets are neutralized (Haudricourt 1954, Ratliff 2015). As a result, checked syllables may continue to change as a coherent category, long after the disappearance of tonogenetic conditioning environments of ‘smooth’ syllables (those ending in sonorants).
Complicating the matter is the fact that synchronic accounts inherit assumptions about toneme identity that may be artifacts of differing conventions for documenting tone languages. There are two general conventions for checked-syllable tones in Asia: (1) assuming they are allotones of the most phonetically similar smooth-syllable tone, and (2) assuming they are fully disjoint from the smooth-syllable toneme inventory.
The firewall between synchrony and diachrony leads to puzzlement over straightforward phenomena, and results in synchronic accounts that may fail to generalize to closely related languages, let alone the entirety of the Asian tonbund or tone as a global phenomenon. By appreciating the fundamental connection between syllable shape and tonal development, we avoid taxing synchronic accounts when consonant-tone interaction does not occur.