The Austroasiatic tree

Three kinds of evidence bear on where Proto-Austroasiatic was spoken. The first is its vocabulary. Reconstructed words include terms for rice at different stages of processing, such as *rəŋkoːˀ 'husked rice' (Vietnamese gạo, Khasi khaw), which point to a community of farmers rather than foragers. Sidwell and Roger Blench also argued that the vocabulary fits people who lived and travelled along rivers.

The second is geography. The greatest variety of branches is found in the middle Mekong, where Katuic, Bahnaric, Vietic and Khmeric meet, and in 2011 Sidwell and Blench placed the homeland there, with a rapid spread along the river systems around 2000 BCE. Sidwell has since moved the likely centre north-east, to the Red River delta of northern Vietnam. A third proposal, favoured by archaeologists such as Charles Higham and Peter Bellwood, ties the family to rice farmers who moved south from the Yangtze region of southern China in the third millennium BCE.

The third is genetics. Ancient DNA from Neolithic sites in Vietnam, such as Man Bac, published by Mark Lipson and colleagues in 2018, shows newcomers with ancestry from southern China mixing with local hunter-gatherers, and present-day Austroasiatic speakers carry much of that ancestry. This fits a farming spread, though it cannot say exactly where the language began. In India, Munda-speaking groups share paternal lineages with Southeast Asia, while their maternal lineages are mostly South Asian. Gyaneshwer Chaubey and colleagues (2011) read this as a movement of men who married into local communities.

The precise starting point, the route taken into India and the date at which Munda arrived there all remain open.

On the mainland, the typical Austroasiatic word has one stressed syllable, often preceded by a light 'minor' syllable, as in Khmer crɑmoh 'nose'. Stress falls at the end of the word, and over time many languages have worn the minor syllables away. Vowel systems are large: Khmer distinguishes around thirty vowels and diphthongs. Rather than tones, many languages contrast two voice qualities, a clear 'head' register and a breathy 'chest' register, which grew out of old voiced and voiceless initial consonants. Vietnamese went further and turned such distinctions into six tones.

Mainland grammar is analytic. Words do not inflect for tense or case, the order is subject–verb–object, nouns come before their adjectives and possessors, and counting uses classifiers, as in Vietnamese hai con chó 'two dogs', literally 'two animal dog'. The inherited word-building survives mainly as fossils. Khmer daə 'walk' has the noun dɑmnaə 'journey', and slap 'die' has the causative sɑmlap 'kill', both formed with infixes. Many languages also have a large class of 'expressives', vivid sound-symbolic words that describe sights, sounds and sensations.

Munda turned this profile inside out. Santali puts the verb last and builds long verb forms with suffixes for tense, subject and object. Its pronouns distinguish singular, dual and plural, and have separate forms for 'we' including and excluding the listener: alaŋ 'we two, you and I' but aliɲ 'we two, not you'.

The oldest writing, in Mon and Khmer, uses scripts derived from those of southern India. Old Mon writing was the model for Burmese, and Khmer writing for Thai and Lao. Vietnamese and Khasi are now written in Latin letters, and Santali also in its own Ol Chiki script.

The biggest open question is the shape of the tree. For most of the 20th century the family was divided into Munda and Mon-Khmer. Gérard Diffloth later proposed a more detailed tree, with Munda, a Khasi-Khmuic group and a 'Nuclear Mon-Khmer' group. Sidwell's comparative work found that few of these groupings rest on shared innovations, and he described the family as a 'rake' of about thirteen branches that separated almost at once. More recently he has tentatively recognised a few regional groupings, such as Khasian with Palaungic and Nicobarese with Aslian. The rake itself is sometimes read as evidence for a very rapid dispersal.

Dating is tied to the homeland. Those who connect the family with the arrival of rice farming favour the late third millennium BCE; others allow a greater time depth, and computational dating studies have not settled it.

Munda's history in India is a live topic of its own. F. B. J. Kuiper argued in 1948 that Sanskrit had borrowed a number of words from a Munda-like language, and Michael Witzel has proposed a lost 'Para-Munda' language behind some words in the Rigveda. Critics reply that many of the proposed etymologies are uncertain and that other sources, including Dravidian and unknown languages, are equally possible.

Austroasiatic may also have left traces further north. Jerry Norman and Tsu-lin Mei argued in 1976 that some Old Chinese words, such as 江 jiāng 'river', were borrowed from Austroasiatic languages then spoken in southern China. Beyond that, Schmidt's Austric hypothesis, linking Austroasiatic with Austronesian, still has defenders, notably Lawrence Reid, but most specialists regard it as unproven. No external relative of Austroasiatic is generally accepted.