This study proposes the Distillation Method for the comparative studies on languages in contact. This method is applied to the Bai language, which comes under heavy contact with Chinese. The distillation method consists of the following steps: (i) Intra-comparison; (ii) Inter-comparison; and (iii) Recognition. The intra-comparison may exclude later borrowings and recent innovatios in the Bai dialects after their separation from Proto-Bai. The second step, inter-comparison is used to exclude some early borrowings before Proto-Bai. Therefore, the data on the genetic transmission of the Bai language may be made cleaner. In the recognition part, the Inexplicability Principle and the Rank Theory can be made use to recognize the inheritance status of the lexical items in the cleaner data.

Intra-comparison uses the first-hand data from nine different Bai dialects to reconstruct a Proto-Bai system, including the phonological system, some morphosyntactic structures and some basic words. In principle, Proto-Bai contains only the elements of the Bai dialects inherited from their common ancestor. Nineteen innovations, including phonological innovations, semantic innovations and lexical replacements are selected as criteria for subgrouping the nine modern Bai dialects. Two phylogenetic algorithms, PAUP and PENNY, are used to generate the tree for the nine Bai dialects based on the nineteen characters. This work proposes a subgroup structure for the Bai dialects– Western Bai vs. Eastern Bai.

Inter-comparison mainly stratifies the Sino-Bai related words thus allowing the exclusion of borrowed layers in Proto-Bai from Chinese. The rest of the Sino-Bai related words after this distillation are accumulated as the “oldest” layer, which represents the cleaner data.

Two ways, namely, the Inexplicability Principle and Chen’s (1996) Rank Theory are then used to recognize the refined Sino-Bai related words as cognates rather than borrowings.