An accurate reconstruction of Sino-Tibetan language evolution would greatly advance our understanding of East Asian population history. Two recent phylogenetic studies attempted to do so but several of their conclusions are different from each other. Here we reconstruct the phylogeny of the Sino-Tibetan language family, using Bayesian computational methods applied to a larger and linguistically more diverse sample. Our results confirm previous work in finding that the ancestral Sino-Tibetans first split into Sinitic and Tibeto-Burman clades, and support the existence of key internal relationships. But we find that the initial divergence of this group occurred earlier than previously suggested, at approximately 8000 years before the present, coinciding with the onset of millet-based agriculture and significant environmental changes in the Yellow River region. Our findings illustrate that key aspects of phylogenetic history can be replicated in this complex language family, and calls for a more nuanced understanding of the first Sino-Tibetan speakers in relation to the “early farming dispersal” theory of language evolution.

准确重建汉藏语言的演变将极大地增进我们对东亚人口历史的理解。最近的两项系统发育研究试图这样做，但他们的一些结论彼此不同。在这里，我们使用贝叶斯计算方法应用于更大、语言上更多样化的样本，重建了汉藏语系的系统发育。我们的结果证实了之前的研究结果，即汉藏人的祖先首先分裂为汉族和藏缅族，并支持关键内部关系的存在。但我们发现，这个群体的最初分化发生得比之前认为的要早，大约在距今8000年前，与以小米为基础的农业的兴起和黄河流域的重大环境变化同时发生。我们的研究结果表明，系统发育历史的关键方面可以在这个复杂的语系中复制，并要求对第一批汉藏语使用者与语言演化的“早期农业传播”理论的关系有更细致的了解。