The study of language origin and divergence is important for understanding the history of human populations and their cultures. The Sino-Tibetan language family is the second largest in the world after Indo-European, and there is a long-running debate about its phylogeny and the time depth of its original divergence. Here we perform a Bayesian phylogenetic analysis to examine two competing hypotheses of the origin of the Sino-Tibetan language family: the 'northern-origin hypothesis' and the 'southwestern-origin hypothesis'. The northern-origin hypothesis states that the initial expansion of Sino-Tibetan languages occurred approximately 4,000-6,000 years before present (BP; taken as AD 1950) in the Yellow River basin of northern China, and that this expansion is associated with the development of the Yangshao and/or Majiayao Neolithic cultures. The southwestern-origin hypothesis states that an early expansion of Sino-Tibetan languages occurred before 9,000 years BP from a region in southwest Sichuan province in China or in northeast India, where a high diversity of Tibeto-Burman languages exists today. Consistent with the northern-origin hypothesis, our Bayesian phylogenetic analysis of 109 languages with 949 lexical root-meanings produced an estimated time depth for the divergence of Sino-Tibetan languages of approximately 4,200-7,800 years BP, with an average value of approximately 5,900 years BP. In addition, the phylogeny supported a dichotomy between Sinitic and Tibeto-Burman languages. Our results are compatible with the archaeological records, and with the farming and language dispersal hypothesis of agricultural expansion in China. Our findings provide a linguistic foothold for further interdisciplinary studies of prehistoric human activity in East Asia.