John Cowan sent me a link to Roger Blench’s paper (draft circulated for comment) “Suppose we are wrong about the Austronesian settlement of Taiwan?,” a fascinating attempt to upend the usual narrative. Here’s the abstract:
The current model of the prehistory of Taiwan assumes that it was first settled some 25,000 years ago by a population of unknown affinities, who reached what is now an island via a landbridge, at a time of much lower sea-levels. Some 5500 years ago, the Ta Pen Keng (TPK) culture, attested on the Peng Hu islands in the Taiwan Strait, apparently represents an incoming Neolithic population. Similar TPK sites are recorded around the shores of Taiwan in the centuries immediately following this. The pervasive assumption has been that these early settlers were the bearers of the Austronesian languages, which then diversified. If so, related Austronesian languages were formerly spoken on the Chinese mainland and these subsequently disappeared as a consequence of the Sinitic expansions. The indigenous Austronesian languages of Taiwan are claimed to reconstruct to a single proto-language, PAN, and from these reconstructions we can derive hypotheses about the lifestyle and subsistence of the earliest settlers.
This paper will argue that the single migration model is mistaken, and that it is not consistent with either the archaeology or the lexicon. If Formosan languages appear to reconstruct to a proto-language it is because they have been interacting over a long period, but they actually represent a continuing flow of pre-Austronesian languages from the mainland. Part of the evidence for this is the exceptional diversity of lexical items which are supposedly part of basic subsistence vocabulary.
Three phases of migration are distinguished, the TPK, the Longshan type culture and the Yuanshan, all of which originate on different places on the Chinese mainland. A further back migration from the Philippines may be responsible for the primary settlement of Green island and parts of the east coast, resulting in the present-day Amis population.
From the conclusion:
From this it follows that a single PAN cannot be reconstructed, in the sense of an apical ancestor, merely a Common Formosan (CF). The Formosanisms identified by Dyen and Blust are not PAN but rather local innovations. This explains why reconstructions of PAN phonology and grammar have always tended to be inconclusive. The flat arrays proposed by Blust and Ross would thus be a reflection of prehistory, although not in the sense originally intended.
It’s clearly written, and there are very useful maps, tables, and illustrations; I’ll be interested to see if the theory becomes accepted.