The story of Pama-Nyungan
01
A Family Named from Its Two Ends
Europeans noticed early that words recurred across great distances in Australia. Edward Curr's The Australian Race (1886–87) gathered hundreds of wordlists sent in by settlers, missionaries and police, and readers could see the same pronouns and body-part terms turning up from Queensland to Western Australia. In 1919 the Austrian scholar Wilhelm Schmidt divided the continent's languages into a northern and a southern group, a split that roughly anticipates the modern one. In 1956 Arthur Capell drew the line on grammar instead: most of the north uses prefixes on the verb, while the south builds words almost entirely with suffixes.
The family got its name in 1966, when Geoffrey O'Grady and Carl and Florence Voegelin published a classification based on shared basic vocabulary, drawing on work with Kenneth Hale. Hale's label took the word for 'person' from each corner of the area: pama in Cape York and nyunga in the far southwest. On their count, Pama-Nyungan covered about nine-tenths of the continent, while the remaining languages of the north fell into many small families.
Counting shared words is not proof of descent, and the grouping had a prominent critic. R. M. W. Dixon argued in 1997 and 2002 that Australian languages had borrowed from one another for so long that family trees could not be drawn, and that Pama-Nyungan was an area of convergence, not a family. The reply came from the comparative method itself. Barry Alpher's reconstruction (2004) set out regular sound correspondences, several hundred cognate sets and a shared set of pronouns and case suffixes found across the whole area and not outside it. Most specialists now accept Pama-Nyungan as a genuine family, though Dixon has not changed his view.
02
From the Gulf across a Continent
People have lived in Australia for at least 50,000 years, but Pama-Nyungan is much younger. Its languages are too similar to be anywhere near that old: basic words such as ngali 'we two' are still close to identical thousands of kilometres apart. Most estimates put the ancestor at roughly 5,000–6,000 years ago, in the middle of the Holocene.
Where it was spoken is inferred from the shape of the tree. Diversity is greatest in the northeast, where the Paman languages of Cape York are unusually varied, and a homeland is usually placed where diversity is highest. Remco Bouckaert, Claire Bowern and Quentin Atkinson (2018) modelled the family's history on a map. Their analysis puts the origin in the Gulf Plains near the Gulf of Carpentaria and has the family moving south and west over the following millennia, often along coasts and rivers before crossing drier country.
What drove the spread is less clear. Nicholas Evans and Patrick McConvell (1998) linked it to changes seen in the archaeology of the mid-Holocene, including new stone-tool types and new social and ritual networks. The dingo, often mentioned in the same breath, arrived later than most estimates for the start of the spread. Genetics adds a further constraint. A genomic study led by Anna-Sapfo Malaspinas (2016) detected a Holocene population expansion in the northeast but only limited gene flow from there to the rest of the continent. On that evidence the languages spread much further than the people who first spoke them. Most researchers therefore think they travelled mainly through contact, marriage, trade and ceremony, with existing groups taking up the new language. The spread was also uneven. The Western Desert language, for example, seems to have covered its huge territory only within the last one or two thousand years.
03
Sounds without Fricatives, Words without Prefixes
An English speaker hearing a Pama-Nyungan language for the first time usually notices what is missing. There are typically no fricatives like s, f or th, and there is no contrast between voiced and voiceless stops, so spellings with p or b, t or d are arbitrary choices. Many languages have only three vowels, a i u, often with a long–short contrast. The consonants make up for it: stops and nasals come at as many as six places in the mouth, including a velar nasal ŋ that can begin a word (as in ngali 'we two'), dental t̪ n̪, palatal c ɲ and retroflex ʈ ɳ sounds. Stress nearly always falls on the first syllable, and words usually begin with a consonant. The Arandic and Northern Paman languages, which lost their initial consonants, are famous exceptions.
The grammar is built with suffixes. Nouns take case endings, and many languages mark the subject of a transitive verb with an ergative suffix. In Pitjantjatjara, for example, tjilpi-ngku 'old man-ERG' is the doer in 'the old man drank water'. Pronouns often work on a different pattern from nouns, with one form for any subject, and they distinguish singular, dual and plural: a speaker says 'we two' as readily as 'we'. Because the case endings show who did what to whom, word order is fairly free.
Speech is shaped by social rules. Many communities had 'avoidance' registers used near certain in-laws, such as the Jalnguy style of Dyirbal in north Queensland, and several had sign languages as well. None of the languages had a writing system before colonisation. Today they are written in Latin-based spellings devised by missionaries, linguists and communities, which is why the same sound can be spelt differently in neighbouring languages.
04
A Family Still Being Mapped
Pama-Nyungan is now generally accepted as a family, but many questions about it are still open.
The first concerns its internal shape. Dozens of lower-level subgroups (Paman, Yolŋu Matha, Arandic, Thura-Yura and others) are secure, but how they combine is not. Bowern and Atkinson's phylogenetic study (2012) recovered four large groups, Western, Central, Northern and South-Eastern, yet several languages move between them depending on the method. Some scholars see the top of the tree as a 'rake', with many branches splitting from the root at once. Others propose intermediate groups such as Arandic–Thura-Yura, which Glottolog adopts. The position of Yolŋu Matha, cut off from its relatives in Arnhem Land, is particularly uncertain.
The second is dating. Most computational estimates put the proto-language at 5,000–6,000 years old, but they rest on assumptions about rates of word replacement. Some linguists suspect that the widespread similarity partly reflects later borrowing between neighbours, which would make the family look younger than it is.
The third is what lies beyond the family. Nicholas Evans has proposed a 'Macro-Pama-Nyungan' grouping linking Pama-Nyungan with the northern Tangkic, Garrwan and Gunwinyguan families, on the evidence of shared pronouns and verb forms. Bowern and others consider the evidence thin. Others go further and argue that all Australian languages share a single ancestor. Mark Harvey and Robert Mailhammer (2017), for instance, reconstructed noun-class prefixes they assign to 'Proto-Australian'. Because the time depths involved are so great, few specialists regard any of these deeper links as proven.
Finally, the question of why this family, and not another, came to cover nearly the whole continent is still open.