The Austronesian tree

Where the family began was argued for a century. Hendrik Kern in 1889 used plant and animal names shared across the family to place the homeland on the coast of mainland Southeast Asia. Isidore Dyen's word-count classification of 1965 pointed instead to Melanesia, where he found the most diversity, a result later traced to the heavy vocabulary replacement of many Melanesian languages. Since Richard Shutler and Jeff Marck's proposal of 1975, most linguists have placed Proto-Austronesian in Taiwan, because that is where the deepest splits in the family are found.

The vocabulary fits a farming people on a subtropical coast. It includes rice and foxtail millet, pigs and dogs, weaving, pottery and boats, but not the outrigger canoe and sail terms that appear later in Malayo-Polynesian. The archaeologist Peter Bellwood linked the spread to farming, tracing it through red-slipped pottery from Taiwan into the Philippines around 2000 BCE and on into Indonesia. Genetics has since added support at the far end. Ancient DNA from early Lapita-period burials in Vanuatu and Tonga, published by Pontus Skoglund and colleagues in 2016, shows almost entirely East Asian ancestry, with Papuan ancestry arriving later. A Bayesian analysis of basic vocabulary by Russell Gray, Alexei Drummond and Simon Greenhill (2009) estimated the family's age at about 5,200 years and found pauses and pulses in the spread that match the archaeology.

The model has critics. Stephen Oppenheimer and, earlier, Wilhelm Solheim argued for an origin within island Southeast Asia, pointing to older maritime networks and some genetic lineages. Many archaeologists think the spread involved more trade and intermarriage than a simple farming expansion. The settlement of Madagascar, probably in the second half of the 1st millennium CE, is also debated in its route and detail.

Austronesian languages tend to have small sound systems and simple syllables. Hawaiian manages with eight consonants and five vowels, and even the larger systems of the west rarely allow more than one consonant at the start of a syllable. Most roots have two syllables: Malay rumah 'house', mata 'eye', batu 'stone'.

Words are built by adding affixes, including infixes placed inside the root. In Tagalog, sulat 'write' gives sumulat with the infix -um-, focusing on the writer, and sinulat with -in-, focusing on what was written. This is the Philippine-type voice system, found in many Formosan and Philippine languages and in Malagasy, in which the verb shows which participant the sentence is about. Reduplication is everywhere: Malay rumah-rumah 'houses', Hawaiian wikiwiki 'quick'. Almost every Austronesian language distinguishes an inclusive 'we' that takes in the listener from an exclusive one that does not, as Malay kita and kami do.

Word order varies more than the vocabulary. Formosan, Philippine and Polynesian languages usually put the verb first, and Malagasy has the rare order verb–object–subject. Malay and Indonesian are subject–verb–object, and some languages around New Guinea put the verb last under Papuan influence. Many Oceanic languages mark possession by category: Fijian uses one possessive form for things to be eaten and another for general belongings.

Writing came from outside. Old Malay inscriptions of the late 7th century use a South Indian script, and Brahmi-derived alphabets were later adapted for Javanese, Balinese, Buginese and the Philippine baybayin. Arabic script served Malay (Jawi) and Malagasy (sorabe), but today almost every Austronesian language is written in the Latin alphabet.

The top of the tree is the most disputed part. Blust divides Proto-Austronesian into ten primary branches, nine of them in Taiwan and one, Malayo-Polynesian, covering everything else. Laurent Sagart argues for a nested tree, in which the Formosan groups peel off one after another and Malayo-Polynesian emerges from within a subgroup on Taiwan's east coast; Malcolm Ross has argued from verb morphology for a 'Nuclear Austronesian' grouping that leaves out Rukai, Tsouic and Puyuma. Both challenge the idea that the Formosan languages sit on an equal footing.

Below Malayo-Polynesian the picture is also uncertain. Blust's Central-Eastern Malayo-Polynesian, which would unite the languages of eastern Indonesia with those of the Pacific, rests on a small number of shared innovations, and Mark Donohue and Charles Grimes (2008) argued that those innovations do not hold up. The large grouping of the Philippines and western Indonesia is equally hard to pin down.

Dates remain approximate. Linguists and archaeologists broadly agree on an origin in Taiwan five to six thousand years ago, but the pace of later movements, including the settlement of Madagascar and of East Polynesia, is still revised as new radiocarbon dates come in.

The family has no proven relatives. Paul Benedict proposed in 1942 that Austronesian is related to Kra-Dai (Thai, Lao and their relatives), and Weera Ostapirat and Sagart have since offered sound correspondences in support; Sagart treats Kra-Dai as a branch of Austronesian itself. Wilhelm Schmidt's 'Austric' hypothesis linking Austronesian with Austroasiatic, and Sagart's Sino-Austronesian, have found fewer supporters. None is accepted as demonstrated.