The story of Kra-Dai
01
Out of Sino-Tibetan
For most of the 19th and early 20th centuries Thai, Lao and Zhuang were filed with Chinese. The case looked strong on the surface: all of them have tones, short words and no inflection, and Thai shares with Chinese a great deal of vocabulary, including most numerals above two. Chinese linguists still call the family Zhuang-Dong and, for much of the 20th century, kept it inside Sino-Tibetan.
Two lines of argument broke the link. The first concerned tones. In 1954 André-Georges Haudricourt showed that Vietnamese had acquired its tones from lost final consonants, so tones could develop independently and were no proof of kinship. The second concerned vocabulary. In 1942 Paul Benedict argued that the Chinese-like words in Tai were loans, concentrated in trade, culture and counting, while the basic vocabulary (body parts, pronouns, 'eye', 'die', 'fire') was shared instead with Hlai on Hainan and a handful of little-known languages of Guizhou and northern Vietnam, such as Gelao and Lachi. He called the grouping Kadai and linked it to Austronesian.
The internal evidence grew as the smaller branches were described. Li Fang-Kuei's Handbook of Comparative Tai (1977) set out regular sound correspondences within Tai; Chinese fieldworkers published grammars and word lists of Kam, Sui, Hlai and the Kra languages from the 1950s to the 1990s; and Weera Ostapirat's reconstruction of Proto-Kra (2000) showed that even the most divergent branch fitted the same system. Ostapirat proposed the name Kra-Dai, built from native words for 'person'. The Chinese loans in Tai can now be sorted into layers of different ages, which is what one expects of borrowing, not inheritance. Today almost all specialists outside China, and many within it, treat Kra-Dai as an independent family.
02
From the South China Coast
The branches are most diverse in a compact area: Guizhou, Guangxi and western Guangdong, Hainan Island and the Vietnamese border. Kra, Hlai and Kam-Sui are all still spoken there, and so are the Northern and Central groups of Tai. By the usual rule that a family's homeland lies where its deepest splits are found, Proto-Kra-Dai was spoken somewhere in this region, and most estimates put it around 2000 BCE, though no date rests on firm evidence.
How it got there is disputed. If Kra-Dai is a branch or sister of Austronesian, its ancestors must once have lived near the Austronesian homeland on and opposite Taiwan. Laurent Sagart argues that pre-Kra-Dai speakers left Taiwan and returned to the mainland in the third millennium BCE. Ostapirat prefers two sister groups that parted on the Chinese coast. James Chamberlain has placed early Kra-Dai speakers further north, in the middle Yangtze basin, around the end of the second millennium BCE. Chinese records of the many 'Yue' peoples of the south are often linked with Kra-Dai speakers, but ethnic labels in Chinese sources cannot be matched with languages with any certainty.
The later history is clearer. Chinese states pushed south from the 3rd century BCE, absorbing much of the old Kra-Dai area; the Kra languages survive only in scattered upland pockets. The Southwestern Tai languages, by contrast, spread out of Guangxi and northern Vietnam into Laos, Thailand, Myanmar and Assam. Pittayawat Pittayaporn used the Chinese loanwords shared by these languages to date that spread to roughly the 8th to 11th centuries CE, which explains why Thai, Lao and Shan are still so close to one another.
03
Tones, Classifiers and Borrowed Scripts
A reader meeting Thai, Zhuang or Kam will notice first that pitch distinguishes words. The family's tones share a history. Proto-Tai had three tones on syllables ending in a vowel or nasal, conventionally A, B and C, while syllables ending in -p, -t or -k formed a fourth class, D. Each of these later split according to the initial consonant, and many languages split again, so modern Thai has five tones and Kam, by some counts, fifteen, yet the categories still line up from language to language. Many linguists think the three original tones themselves go back to lost final consonants, as in Vietnamese.
Words are mostly one syllable long and do not change their form: there is no case, no agreement and no tense ending. Grammar is carried by word order and by small particles. The basic order is subject, verb, object, and, unlike Chinese, modifiers follow the noun: Thai bâan yài is 'house big', a big house. Counting needs a classifier, a word chosen by the kind of thing counted, placed after the numeral: mǎa sǎam tua is 'dog three body', three dogs. Several verbs can be strung together without linking words to express one event, and sentence-final particles mark politeness, questions and attitude.
Writing came from outside, in two directions. The Southwestern Tai languages adopted scripts of Indian origin through Khmer, Mon and Burmese: the Thai and Lao alphabets, the Tai Tham script of northern Thailand and the Shan and Ahom scripts. In China, Zhuang and Sui speakers built characters on Chinese models, and in the 1950s the government introduced Latin alphabets for Zhuang, Bouyei, Kam and Hlai.
04
Open Questions
The largest question is the family's place in the wider world. Few scholars now doubt that Kra-Dai and Austronesian share basic vocabulary: Ostapirat's matches include 'eye', 'die', 'fire', 'louse', 'hand' and the pronouns 'I' and 'you', with regular sound correspondences. What that sharing means is debated. Benedict's 'Austro-Tai' and Ostapirat's view make the two families sisters; Sagart makes Kra-Dai a branch inside Austronesian, close to certain languages of Taiwan; others think contact or a deeper, older relationship cannot yet be ruled out. Benedict also once added Japanese to Austro-Tai, a proposal now largely set aside. A minority of scholars in China have continued to argue for kinship with Chinese.
The internal tree is also unsettled. Kam-Sui, Tai and Be (a language of northern Hainan) are usually grouped together as Kam-Tai, but where Hlai and Kra attach, and whether they form a group of their own, is not agreed. Small languages such as Lakkja, Jiamao and Biao are hard to place, partly because of heavy borrowing between neighbours. Within Tai, Li Fang-Kuei's three groups, Northern, Central and Southwestern, are standard, but Pittayaporn argues that the groups are better seen as a series of successive splits, with Southwestern Tai one late offshoot.
Dating is the third problem. Estimates for Proto-Kra-Dai rely on comparisons with archaeology, since written records of Kra-Dai languages begin only in the medieval period. Finally, the reconstruction itself is incomplete. Proto-Tai is well established, but Proto-Kra-Dai is known from a few hundred proposed roots, with many sounds still written as unspecified capitals in the reconstructions, so every claim about the family's deepest history remains provisional.