TL;DR

  • Pama-Nyungan covers about 90% of mainland Australia; the other 27 language families are crowded into the far north (Bouckaert et al. 2018).
  • The majority view puts its homeland at the base of the Gulf of Carpentaria about 5,700 years ago. A Bayesian analysis of 306 languages found “very little support” for every rival.
  • O’Grady put the root in central-northern Australia, Wurm in the north-west, and Clendon in the eastern ranges after the last ice age.
  • Others tied the spread to Ice Age refuges or to the first settlers. Dixon denied there was a family to have a homeland at all.
  • Every named homeland but Clendon’s lies in the north.

“After their daily forays for game in their small Urheimat in central northern Australia, members of the original Pama-Nyungan speech community must regularly have been greeted with a stock phrase, ŋaana minja what meat?” — Geoffrey O’Grady, “Preliminaries to a Proto Nuclear Pama-Nyungan Stem List” (1979), p. 127

Why does Pama-Nyungan need a homeland? #

People have lived in Australia for more than 50,000 years, yet one family of languages covers nine-tenths of the mainland. Of the continent’s 28 language families, “27 are restricted to the far north, while one family—Pama–Nyungan—covers the remaining 90% of the continent” (Bouckaert, Bowern and Atkinson 2018, p. 741). Its name joins two words for “man”: pama on Cape York and nyungar around Perth.

A family that spans a continent but is only a few thousand years old must have started somewhere and spread over older languages. Its closest relatives point to the north. Nicholas Evans summarises the model that most specialists now hold: “Pama-Nyungan was a relatively recent daughter node within a larger Stammbaum” of Australian languages, and its nearest sister is the Garrwan family of the southern Gulf of Carpentaria (Evans 2003, pp. 9–10). Pama-Nyungan began as one northern branch among many and went on to cover most of the continent.

Where that branch began has been argued for fifty years. Here are the proposals, mapped.

Map / Proposed homelands

Where scholars put the root

Map of the proposed homelands of the Pama-Nyungan language familyA dot-matrix map of Australia and southern New Guinea. Grey dots in the far north, Tasmania and New Guinea mark land where the languages are not Pama-Nyungan; a dashed line marks the frontier. Copper dots mark the shallow shelf that drowned last, after about 10,000 years ago. Amber markers cluster at the base of the Gulf of Carpentaria: Evans and Jones (1997), Bouckaert, Bowern and Atkinson (2018) and, on the drowned plain, Morphy and colleagues (2020). O’Grady (1979) sits in central-northern Australia, Wurm (1972) inland from Eighty Mile Beach in the north-west, and Clendon (2006) along the Great Dividing Range. Dashed rings mark Dixon’s three northern landfalls and Williams’s three Ice Age refuges.Wurm (1972): Inland from Eighty Mile Beach, north-western Australia, Holocene, not datedWurm 1972undatedDixon (1997): Kimberley entry: Kimberley, a northern entry point, c. 40,000–55,000 years agoDixon (1997): Arnhem Land entry: Arnhem Land, a northern entry point, c. 40,000–55,000 years agoO’Grady (1979): Northern Arandic country and the land to its north-east and east, central Northern Territory, 4,000–5,000 years agoO’Grady 19794–5 kaEvans & Jones (1997); Evans & McConvell (1998); McConvell (1996): South-western Gulf of Carpentaria, c. 4,000–6,000 years agoEvans & Jones 19974–6 kaMorphy, Morphy, Faulkner & Barber (2020): The drowned Carpentarian Plain, now the floor of the Gulf, c. 8,000–6,000 years agoMorphy 20208–6 ka, now seabedBouckaert, Bowern & Atkinson (2018): Gulf Plains, at the base of the Gulf of Carpentaria, c. 5,700 years ago (95% range 4,455–6,966)Bouckaert 2018c. 5.7 kaWilliams et al. (2013, 2015): Murray–Darling refuge: Murray–Darling Depression, south-eastern Australia, c. 9,000–7,000 years agoDixon (1997): Cape York entry: Cape York Peninsula, a northern entry point, c. 40,000–55,000 years agoWilliams et al. (2013, 2015): Gulf Plains–Einasleigh refuge: Gulf Plains and Einasleigh Uplands, north Queensland, c. 9,000–7,000 years agoWilliams et al. (2013, 2015): Brigalow Belt South refuge: Brigalow Belt South, inland southern Queensland, c. 9,000–7,000 years agoClendon (2006): The Great Dividing Range and the east coast, After c. 13,000 years agoClendon 2006after 13 kaDixon 1997first landfalls, 40–55 kaWilliams 2015Ice Age refuges, 7–9 ka
  • Gulf proposals, the majority view
  • Pama-Nyungan languages
  • Other language families
  • Last land to drown, after ~10,000 years ago
  • Dry at the glacial low stand
  • Bouckaert’s most probable root
Markers are approximate centres of each proposal, placed from the authors’ own descriptions; ka is thousands of years ago. Amber marks the Gulf proposals, the majority view. Clendon’s band follows the Great Dividing Range, and the halo is the area of highest posterior probability in Bouckaert, Bowern and Atkinson (2018, fig. 1), drawn by eye. The language frontier is simplified, and small Pama-Nyungan pockets on the western Gulf coast are not shown. Seafloor from ETOPO1, banded by the southern Gulf sea-level curve of Sloss and colleagues (2018): the copper shelf, less than 30 m deep, was still dry land 10,000 years ago.

Download the data (CSV)

Where do most linguists put the Pama-Nyungan homeland? #

At the base of the Gulf of Carpentaria, in the savanna plains around today’s Burketown in north-west Queensland.

The linguistic argument came first. Evans and the archaeologist Rhys Jones found the deepest splits within Pama-Nyungan around the south-western Gulf, “suggestively close to its nearest non-Pama-Nyungan neighbours” (Evans, comment in Clendon 2006, p. 53). Under the usual rule, a family’s homeland is where its oldest branches sit closest together.1 Patrick McConvell’s “Backtracking to Babel” (1996) dated the protolanguage to 5,000–6,000 years ago, and Evans and McConvell (1998) linked its spread to the new tools and social forms then appearing in the archaeological record.

In 2018 Remco Bouckaert, Claire Bowern and Quentin Atkinson tested the question statistically. They coded 18,238 cognate sets across 306 Pama-Nyungan languages and modelled the family’s spread across the map. The root came out “at the base of the Gulf of Carpentaria, with an estimated age of 4,455–6,966 years (95% highest posterior density; mean = 5,671 years)” (p. 742). They tested the three main rival scenarios against it. “There is very little support for homelands proposed under the rival hypotheses, including the Einasleigh Uplands, Great Dividing Range, Murray–Darling basin and potential colonization routes in the north.”

The tree also dates individual branches. The Paman languages of Cape York enter Queensland “after 5076 years BP”, in step with a rise in inland occupation there (SI, p. 35).

The authors find little genetic sign of a mass migration, so they conclude that the languages “were carried as part of an expanding package of cultural innovations that probably facilitated the absorption and assimilation of existing hunter-gatherer groups” (p. 741). In the mid-Holocene, backed stone artefacts proliferated “across a range concordant with the distribution of Pama–Nyungan languages”, alongside new ways of using land and sea and new rock art (p. 745). The genomes do record something. Malaspinas and colleagues found “a population expansion in northeast Australia during the Holocene epoch… associated with limited gene flow from this region to the rest of Australia, consistent with the spread of the Pama–Nyungan languages” (Malaspinas et al. 2016, p. 207).

Is the real homeland under the sea? #

Frances and Howard Morphy and their co-authors add a variant that the map can show. When Proto-Pama-Nyungan was spoken, the sea had only just finished drowning the Carpentarian Plain, the dry land and great freshwater lake that once joined Arnhem Land to Cape York. They suggest that the protolanguage began on the plain’s last fringes, 8,000–6,000 years ago, among people pushed back by the rising water. On this reading the root “appears to be near Burketown simply because its actual point of origin is now under water” (Morphy et al. 2020, p. 155). They call the idea “hopeful speculation” and set it aside (p. 156). It would move the homeland only a short distance north of the majority site.

Did Pama-Nyungan begin in central Australia? #

Geoffrey O’Grady thought so. In 1979 he argued that the modern map “points to an almost explosive expansion of the Pama-Nyungan speech-area 4,000 to 5,000 years ago”, which brought the family to seven-eighths of Australia (O’Grady 1979, p. 114). He placed the centre of dispersal in “the northern part of the Arandic speech-area, as well as the territory immediately to the northeast and east”, in the middle of the Northern Territory, north of Alice Springs. His evidence was again diversity. The languages to the west, Warlpiri, Walmajarri and Nyangumarda, “show every indication of quite close genetic relationship”, so their area “loses its attractiveness as a Pama-Nyungan homeland” (pp. 114–15).

From that centre, O’Grady pictured languages spreading “in all directions”. The Paman branch broke off “at around 1,000 B.C.” and moved north-east towards Cape York and the western Torres Strait (pp. 114–15). His homeland lies about 600 kilometres south-west of the Gulf site, and his dates fall at the young end of Bouckaert’s range.

Did it start in the north-west? #

Stephen Wurm also put the homeland in the north, but on the opposite side of the continent. His 1972 survey Languages of Australia and Tasmania mapped it inland from Eighty Mile Beach, between the Pilbara and the Great Sandy Desert (pp. 165–66). Wurm gave the expansion an outside cause. He compared it to the way Malayo-Polynesian voyagers had transformed life in New Guinea by bringing new skills, and proposed that “Malayo-Polynesian influence reaching the northwest coast of Australia may likewise have led to the spread of a new technology and a new linguistic element through most of the continent” (as summarised by O’Grady 1979, p. 114).2

O’Grady accepted a northern homeland but not this one: the languages of the north-west, Nyangumarda among them, are too closely related to have been diverging there for long.

Did it come from the eastern ranges? #

Mark Clendon proposed the boldest alternative in 2006. At the last glacial maximum the interior was desert, and people held on in the forested rim of the continent. Clendon argues that when “the climate became warmer and wetter after ca. 13,000 BP, the continent was reoccupied from a relatively compact demographic base consisting mainly of people from the dividing range” (Clendon 2006, p. 46). The languages of that base, already alike from long contact, “spread westward from their dividing-range Urheimat sometime after 13,000 BP” (p. 47). The model roughly doubles the usual time depth and places the origin in the east.3

The replies in the same issue of Current Anthropology were respectful but unconvinced. Evans called the model “fresh and provocative” but pointed out that the east coast is “strikingly homogeneous” linguistically: “I know of no account which regards the east coast as a serious contender” (p. 53). In 2018 Bouckaert’s team calculated that their age calibration would have to be more than 50% older to fit Clendon’s date (p. 742).

Ice Age refuges and first settlers #

Two further scenarios do not pick one homeland.

The first ties the spread to the end of the Ice Age. Alan Williams and colleagues modelled the continent’s radiocarbon record and found populations growing quickly out of a few Pleistocene refuges during the early Holocene climatic optimum. “This diffusion from isolated Pleistocene refugia provides a mechanism for the synchronous spread of pan-continental archaeological and linguistic attributes at this time (e.g. Pama-Nyungan language, Panaramitee art style, backed artefacts)” (Williams et al. 2015, p. 1). As Bouckaert’s team list them, the refuges are the Gulf Plains and Einasleigh Uplands, the Brigalow Belt in southern Queensland, and the Murray–Darling basin. The archaeologist Mike Smith reached a similar age of about 8,000 years (Bouckaert et al. 2018, SI Table 1). One of these refuges overlaps the Gulf homeland, but its date is too early for the tree.

The second scenario follows R. M. W. Dixon, who argued that Australia’s languages spread within “a few thousand years” of first settlement, 40,000–55,000 years ago. That would put the root at the first landfalls on the northern coasts: the Kimberley, Arnhem Land and Cape York (SI Table 1). To fit, Bouckaert’s calibration would have to be more than 300% older (p. 742).

Is Pama-Nyungan a family at all? #

Dixon himself did not believe there was a family to have a homeland. In his model of punctuated equilibrium, a short burst of splitting at settlement was followed by “some tens of thousands of years of an equilibrium situation, where the number of languages would have remained more-or-less static, with cultural and linguistic features diffusing across geographical regions” (Dixon 2002, ch. 14). He concluded that “‘Pama-Nyungan’ cannot be supported as a genetic group” (Dixon 2002, pp. 48, 53, quoted in Bowern and Atkinson 2012, p. 818).

Other specialists answered sharply. O’Grady and Hale called the claim “extravagant and spectacularly erroneous”. Evans gave the decisive test: Pama-Nyungan words such as its pronouns turn up across the family, but “none of these forms is ever found in a non–Pama-Nyungan language” (Evans 2005, pp. 262, 269). Borrowing does not stop at a family boundary in that way, but inheritance from a single ancestor does. Since then, computational studies have built trees for the family and gone on to look for its root.

What about Cape York? #

Cape York is full of Paman languages, many of them sharply distinct, and it faces New Guinea across Torres Strait. However, no published proposal puts the origin of Pama-Nyungan there. The Paman languages form one branch of the family, and they reached the peninsula from the south-west. Kenneth Hale’s work on the Wik languages shows that the branch has been settled there a long time: “there is no plausible location outside the area that can be identified as a homeland from which an intrusive Wik-speaking people could have come” (Hale 2021, p. 147). That establishes a deep Wik homeland on Cape York, not a Pama-Nyungan one.

What do the proposals have in common? #

ProposalYearWhereWhenArgument
Wurm1972Inland from Eighty Mile Beach, north-westHolocene, undatedNew technology from Malayo-Polynesian contact
O’Grady1979Northern Arandic country, central north4–5 kaWestern languages too close-knit to be the root
McConvell; Evans & Jones; Evans & McConvell1996–98South-western Gulf of Carpentaria4–6 kaDeepest splits, and the Garrwan sister, beside the Gulf
Clendon2006Great Dividing Range, eastAfter 13 kaIce Age bottleneck in the eastern ranges, then westward spread
Williams et al.; Smith2013–15Gulf Plains–Einasleigh, Brigalow, Murray–Darling refuges7–9 kaEarly Holocene growth out of refuges
After Dixon1997Northern landfalls: Kimberley, Arnhem Land, Cape York40–55 kaSpread with the first settlers
Bouckaert, Bowern & Atkinson2018Gulf Plains, base of the Gulf4.5–7.0 kaBayesian phylogeography of 306 languages
Morphy et al.2020Drowned Carpentarian Plain8–6 kaOrigin on the shrinking plain, now under the Gulf
Dixon1997, 2002None—No family, only long diffusion

Laid out on the map, the proposals show a clear pattern. Every named homeland except Clendon’s sits in the north, close to the non-Pama-Nyungan languages that hold the family’s nearest relatives. Wurm’s is the only one with an outside trigger. The rest look inside Australia, to new tools, rites and kin systems or a changing climate, for whatever carried one northern language across the continent. Most of those inquiries end at the base of the Gulf, where the diversity, the sister languages and the statistical tree agree.

For the wider linguistic setting, see how Australian and New Guinea languages are connected and proposed cognates across Australian languages. For the mid-Holocene changes that the Gulf expansion is usually tied to, see Australia and the Sapient Paradox.

Sources #

  1. Bouckaert, Remco R., Claire Bowern and Quentin D. Atkinson. “The origin and expansion of Pama–Nyungan languages across Australia.” Nature Ecology & Evolution 2 (2018): 741–749. Open copy; Supplementary Information.
  2. Bowern, Claire. “Historical linguistics in Australia: trees, networks and their implications.” Philosophical Transactions of the Royal Society B 365 (2010): 3845–3854.
  3. Bowern, Claire and Quentin Atkinson. “Computational phylogenetics and the internal structure of Pama-Nyungan.” Language 88(4) (2012): 817–845. Open copy.
  4. Clendon, Mark. “Reassessing Australia’s linguistic prehistory.” Current Anthropology 47(1) (2006): 39–61, with comments by N. Evans, P. McConvell, D. Nash and others. Open copy.
  5. Dixon, R. M. W. The Rise and Fall of Languages. Cambridge University Press, 1997.
  6. Dixon, R. M. W. Australian Languages: Their Nature and Development. Cambridge University Press, 2002.
  7. Evans, Nicholas. “Introduction: Comparative non-Pama-Nyungan and Australian historical linguistics.” In N. Evans (ed.), The Non-Pama-Nyungan Languages of Northern Australia. Pacific Linguistics 552. Canberra: ANU, 2003.
  8. Evans, Nicholas. “Australian languages reconsidered: a review of Dixon (2002).” Oceanic Linguistics 44(1) (2005): 242–286.
  9. Evans, Nicholas and Rhys Jones. “The cradle of the Pama-Nyungans: archaeological and linguistic speculations.” In P. McConvell and N. Evans (eds.), Archaeology and Linguistics: Aboriginal Australia in Global Perspective, 385–417. Oxford University Press, 1997.
  10. Evans, Nicholas and Patrick McConvell. “The enigma of Pama-Nyungan expansion in Australia.” In R. Blench and M. Spriggs (eds.), Archaeology and Language II, 174–192. Routledge, 1998.
  11. Hale, Kenneth. “Wik subgroup lexical history.” In P. Sutton and K. Hale, Linguistic Organisation and Native Title. ANU Press, 2021 (written 1997).
  12. Malaspinas, Anna-Sapfo, et al. “A genomic history of Aboriginal Australia.” Nature 538 (2016): 207–214.
  13. McConvell, Patrick. “Backtracking to Babel: the chronology of Pama-Nyungan expansion in Australia.” Archaeology in Oceania 31 (1996): 125–144.
  14. Morphy, Frances, Howard Morphy, Patrick Faulkner and Marcus Barber. “Toponyms from 3000 years ago? Implications for the history and structure of the Yolŋu social formation in north-east Arnhem Land.” Archaeology in Oceania 55 (2020): 153–167. Open copy.
  15. O’Grady, Geoffrey N. “Preliminaries to a Proto Nuclear Pama-Nyungan stem list.” In S. A. Wurm (ed.), Australian Linguistic Studies, Pacific Linguistics C-54 (1979): 107–139.
  16. O’Grady, Geoffrey N. and Kenneth L. Hale. “The coherence and distinctiveness of the Pama-Nyungan language family within the Australian linguistic phylum.” In C. Bowern and H. Koch (eds.), Australian Languages: Classification and the Comparative Method, 69–92. John Benjamins, 2004.
  17. Smith, Mike. The Archaeology of Australia’s Deserts. Cambridge University Press, 2013.
  18. Williams, Alan N., Sean Ulm, Andrew R. Cook, Michelle C. Langley and Mark Collard. “Human refugia in Australia during the Last Glacial Maximum and terminal Pleistocene.” Journal of Archaeological Science 40 (2013): 4612–4625.
  19. Williams, Alan N., Sean Ulm, Chris S. M. Turney, David Rohde and Gentry White. “Holocene demographic changes and the emergence of complex societies in prehistoric Australia.” PLoS ONE 10 (2015): e0128661.
  20. Wurm, Stephen A. Languages of Australia and Tasmania. Mouton, 1972.

  1. David Nash, commenting on Clendon in the same issue, reads Evans and Jones’s summary of the subgroups as placing the point of greatest diversity at “roughly the Barkly Tableland” (Nash 2006, p. 54), just south-west of the Gulf. Pama-Nyungan also has two outlying blocs that both touch the western Gulf: Yolŋu in north-east Arnhem Land and Yanyuwa on the south-western Carpentaria coast (Evans 2005, p. 262). The map simplifies the frontier and leaves out these pockets. ↩︎

  2. Wurm’s pages are cited here through O’Grady’s summary, which also gives the location: “the area inland from the Eighty-Mile Beach… (as Wurm does in his map, p. 166)” (O’Grady 1979, p. 114). Wurm’s own text could not be read online. ↩︎

  3. Clendon’s location is described differently in different places. Bowern (2010) summarises his model as a “bottleneck linguistic area which expanded from the South”, and his glacial “peripheral zone” did include “the south-eastern corner of the continent centred on southern Victoria and Tasmania, and the south-west of Western Australia” (Clendon 2006, p. 42). The homeland he names, however, is the dividing range (pp. 46–47), and Bouckaert’s team coded his hypothesis as “Dividing range”. David Nash notes that this region runs as far north as the Dyirbal rainforest of north-eastern Queensland (Nash 2006, p. 55). The map draws Clendon’s homeland as a band along the range. ↩︎