This dissertation investigates the phenomenon of tonogenesis in Vietnamese by building and testing theoretical models using a predictive algorithm. The investigation is divided into two studies: Study I examines correlations between Shorto's (2006) reconstructions of Proto-Mon-Khmer and Modern Vietnamese reflexes. Study II deals with relationships between Middle Chinese words and their Sino-Vietnamese pronunciations.

Results from Study I suggest that Haudricourt's (1954) classic model can be enhanced by distinguishing between singleton sonorant codas and sonorant-stop coda clusters in Proto-Mon-Khmer. Words with the former coda type correlate with ngang or huyen tone as Haudricourt observed, while words with the latter coda type correlate with sac or naˇng tones as do words with singleton stop codas.

Findings from Study II confirm Maspero's (1912) observation that sonorant-onset words in Middle Chinese píng tone pattern tonally with voiceless obstruent-onset words in Sino-Vietnamese, not with voiced obstruent-onset words. Maspero also claims that sonorant-onset words in other Middle Chinese tone categories pattern differently from voiced obstruent-onset words, but Maspero's proposed correlations for the other categories are not consistent with the Sino-Vietnamese data in Study II. Maspero further claims that the sonorant-onset patterning in the píng category is the result of Chinese influence, whereas I believe the patterning to be endogenous to Mon-Khmer languages, based on correlations in the protolanguage data from Study I.

This dissertation builds upon traditional theoretical models of Vietnamese tonogenesis and of Sino-Vietnamese tone, and it demonstrates the value of predictive modeling to the formulation and evaluation of hypotheses in historical linguistics.