Proto-Australian and the Drowned North

TL;DR

• Proto-Australian is the reconstructed ancestor of almost every Aboriginal language. It said “I” with ŋa-, and that “I” still runs through the continent, from Warlpiri ngaju in the desert to Burarra ngay-pa in Arnhem Land.
• The same “I” turns up across the water, in two small language families on the coast of New Guinea that faced Australia: Pahoturi River ŋəna, Bulaka River ŋəl.
• Until about 9,000 years ago the two shores were one land. Nobody knows how old Proto-Australian is. If it is older than the sea, its homeland may now be the floor of the Arafura, and its “I” may have spread from there both ways.

Comparative reconstruction allows us to glimpse back to a time, long ago, when these two phonological realms were less different than they are now. — Nicholas Evans, on Australia and New Guinea, ICPhS (2019)

What is Proto-Australian? #

Almost every language Aboriginal people spoke when Europeans arrived descends from one ancestor. The linguists Mark Harvey and Robert Mailhammer have reconstructed it, sounds, grammar and a core of words, and called it Proto-Australian (Proto-Australian sets out what they rebuilt).

They also say where it was spoken. A language family is usually born where its branches are most varied, because that is where it has had longest to split, and Australia’s branches are crowded into the far north. “The area with the most isolates has the greatest diversity. Within Australia, that area is the Top End” (Harvey and Mailhammer 2025). Nicholas Evans put it more starkly: the north holds “perhaps 90% of Australia’s linguo-genetic diversity in an eighth of its land area” (Evans 2003, p. 1).

How did Proto-Australian say “I”? #

With ŋa-, the velar nasal of English sing at the start of a word. Harvey and Mailhammer reconstruct the free pronoun as *ŋaj. Its descendants are everywhere. In the Pama-Nyungan languages that cover nine-tenths of the continent “I” is Warlpiri ngaju, Gamilaraay ngaya, Pitjantjatjara ngayulu. In the northern families it is Worrorra ngayu, Burarra ngay-pa, Ngandi ngaya, and in many of them the same syllable sits on the front of the verb as a prefix, nga- “I”. It is one of the steadiest things in the family. Australian languages replace words quickly, partly because words that sounded like a dead person’s name were dropped. The word for “I” has hardly moved.

Where else does that “I” turn up? #

Across the water. The south coast of New Guinea, between the Fly River and Merauke, holds three small language families that do not belong to Trans-New Guinea, the great family of the highlands, whose languages mostly say “I” with n- (Melpa na, Duna no). Two of the three say it the Australian way.

• Pahoturi River, on the Papua New Guinea coast opposite the Torres Strait island of Saibai. All six of its languages say ŋəna or ŋən (Lindsey 2019, p. 234), and Malcolm Ross reconstructs their ancestor’s “I” as *ŋa-na (Ross 2005, p. 39).
• Bulaka River, on the Arafura coast west of Merauke. Tina Gregor recorded “I” as ŋəl in two Yelmek villages, ŋər in two others, and ŋəlo in Maklew (Gregor 2021, p. 26). Yelmek also marks “me” as the object of a verb with a prefix, ŋə- (pp. 48, 216–218), much as the northern Australian languages carry nga- on the verb.

Download the data (CSV)

Bulaka River is the telling one. It does not face the Torres Strait islands, where Australian and Papuan speakers have traded and married for centuries. It faces open sea, and behind that sea the old basin of the Gulf of Carpentaria. Island contact is a poor explanation for an Australian “I” there.

The third family, Yam, keeps other things. Its languages mark the doer of a transitive verb with an ergative case, as most Australian languages do and most New Guinea languages do not (Döhler 2018, p. 99). And when Evans reconstructed the sound systems of the Yam and Pahoturi ancestors, he found them more Australian than their descendants: Proto-Pahoturi had five liquids, “highly unusual for a Papuan language, but rather normal for an Australian language”. His summary: “both proto-Yam and proto-Pahoturi look much more Australian-like than their contemporary descendants” (Evans 2019). Ryan Chon concludes that “contact with the Australian linguistic area played a role” in shaping both (Chon 2025).

Everywhere else these languages are Papuan. Their words for “you” begin with b- and o-, like their northern neighbours’. What looks Australian is the oldest and most conservative part of each language: the first person, the case of the doer, the sounds a child learns first. That is what an old layer looks like after thousands of years of separation, and Evans reads the trend that way: the New Guinea languages have been “becoming less Australian-like over millennia” since “they were cut off from the Australian phonological zone by the waters of the Torres Strait.”

What did the north look like before the sea rose? #

At the height of the last ice age, about 20,000 years ago, the sea stood about 120 metres below today’s level. Australia and New Guinea were one land. The Gulf of Carpentaria was dry, and in its basin lay Lake Carpentaria, “up to 250 km in width, 500 km in length”, ringed by grassland (Sloss et al. 2018). The coast of the Top End lay far out on what is now the floor of the Arafura and Timor seas.

Then the ice melted, and the north went under fast. The sea broke over the Arafura Sill into the lake about 11,700 years ago, and by 10,000 years ago the Gulf was salt water. Across Australia the coast moved “on average by 139 km”, and the greatest losses were in the north-west, “>300 km in the vicinity of Cambridge Gulf, Keep River, and Daly River”. In places the shoreline advanced “~3 km per generation” (Williams et al. 2018). People there lost their country fast enough to watch it go.

How old is Proto-Australian? #

Nobody knows. Words do not carry dates, and the comparative method has no reliable clock: Harvey and Mailhammer note that “there are no reliable examples dating back more than 10,000 years.” Their own best guess is that Proto-Australian “was most likely spoken about 6,000 years ago in the Top End” (2025). The reason is Tiwi. The Tiwi Islands were cut off about 6,000 years ago, and Tiwi is not one of Proto-Australian’s daughters, so the family seems to have spread across the mainland after that.

That dates the spread, not the language. A language can be spoken in one region for thousands of years before its descendants leave home. Latin was spoken around Rome for centuries before the Republic carried it across the Mediterranean. So Proto-Australian could be much older than its expansion, and the “I” on the far shore is a reason to wonder whether it was.

Was the homeland under the Arafura? #

Look again at where the diversity sits. The Daly River families, the Iwaidjan languages of the Cobourg Peninsula, and isolates such as Gaagudju, Umbugarla and Laragia line the coast and the lower rivers from the Daly to western Arnhem Land. They are crowded against a shoreline about 7,000 years old, on the stretch of Australia that lost the most land to the sea.

When a family’s diversity is pressed against a coast, the coast may be cutting its homeland in half. “Within Australia”, the Top End is the most diverse region, but that is the Australia of today. Before the sea rose, the land in front of the Top End ran unbroken north to New Guinea. If Proto-Australian was spoken then, the centre of its world may have been that plain, now the floor of the Arafura Sea. As the water came in, its speakers would have retreated two ways: south, up the rivers of the Top End, and north, onto what became the New Guinea coast. That would leave exactly what we see. The family’s deepest branches are bunched on the Australian shore, and the same “I” survives on the New Guinea shore opposite.

Others have looked at this coast and seen something special. Chris Gosden and Lesley Head noticed that ancestral tracks across the land are most pronounced on the two shores of the Arafura Sea, the region that changed most as the sea rose (The Gardens of the Dreaming). And Evans, who holds that Australian languages grew from the first settlers with little outside input, allowed for early contact “with Papuan speakers in northern Cape York and possibly along the plains that once lay to the west of the then Lake Carpentaria” (Evans 2005, p. 280). Those plains are the drowned shelf.

What would it take to prove it? #

Not the “I” alone. Tiwi, which is not a Proto-Australian language, says “I” as ngiya (Harvey 2003, pp. 500–501). So the ŋa- “I” may be older still, a word of the whole northern Sahul population rather than of Proto-Australian alone. That fits the picture of one people on one plain, but it means the “I” cannot by itself show that Proto-Australian crossed. Ross, who reconstructed the New Guinea forms, adds a fair caution: first-person pronouns like these “are not unambiguous markers of a genealogical relationship” (p. 39).

The proof would be something Proto-Australian invented, found on the New Guinea side: one of its noun-class markers, a sound change of its own, a quirk of a paradigm that matches cell for cell. That is the kind of evidence Harvey and Mailhammer used to build Proto-Australian in the first place, and nobody has yet looked for it across the strait. The comparison is now possible. There is a reconstructed Proto-Australian on one side and a reconstructed Proto-Pahoturi and Proto-Yam on the other.

The idea has been tried once before. In 2006 Mark Clendon proposed that Australia’s language families were born on the land bridge about 13,000 years ago, and that some of its people “would have moved north into what is now New Guinea as the inundation proceeded” (Clendon 2006, pp. 44–45). His critics turned it down, rightly: he compared broad grammatical types with the wrong New Guinea families. Evans, replying, pointed to the right ones, “the nonTNGP languages spoken along the southern New Guinea coast”. Those are the families where the Australian “I” turned up.

So the question is open, and it is a good one. A word for “I” sits on both shores of a drowned plain. Perhaps the people who first said it lived on the plain.

Sources #

1. Brockwell, Sally, et al. “Holocene settlement of the northern coastal plains, Northern Territory, Australia.” The Beagle 21 (2011): 1–22.
2. Chivas, Allan R., et al. “Sea-level and environmental changes since the last interglacial in the Gulf of Carpentaria, Australia: an overview.” Quaternary International 83–85 (2001): 19–46.
3. Chon, Ryan. “Reconstructing the Proto-Pahoturi Consonantal Inventory: Historical Phonology in Southern New Guinea and Beyond.” Oceanic Linguistics 64(1) (2025): 215–245.
4. Clendon, Mark. “Reassessing Australia’s linguistic prehistory.” Current Anthropology 47(1) (2006): 39–61, with comments by Alpher, Bowern, Clarkson, Evans, McConvell and Nash.
5. Döhler, Christian. A Grammar of Komnzo. Berlin: Language Science Press, 2018.
6. Evans, Nicholas. “Introduction: Comparative non-Pama-Nyungan and Australian historical linguistics.” In Evans (ed.), The Non-Pama-Nyungan Languages of Northern Australia, 3–25. Pacific Linguistics 552, 2003.
7. Evans, Nicholas. “Australian languages reconsidered: a review of Dixon (2002).” Oceanic Linguistics 44(1) (2005): 242–286.
8. Evans, Nicholas. “Australia and New Guinea: Sundered hemi-continents of sound.” Proceedings of the 19th International Congress of Phonetic Sciences, Melbourne, 2019.
9. Gregor, Tina. A Documentation and Description of Yelmek. PhD thesis, Australian National University, 2021.
10. Harvey, Mark. “Reconstruction of pronominals among the non-Pama-Nyungan languages.” In Evans (ed.), The Non-Pama-Nyungan Languages of Northern Australia, 475–513. Pacific Linguistics 552, 2003.
11. Harvey, Mark, and Robert Mailhammer. Proto-Australian: Reconstruction of a Common Ancestor Language. Berlin: De Gruyter Mouton, 2024.
12. Harvey, Mark, and Robert Mailhammer. “New evidence confirms our Indigenous languages have a common source, but how they spread remains a mystery.” The Conversation, 29 January 2025.
13. Lindsey, Kate L. Ghost Elements in Ende Phonology. PhD dissertation, Stanford University, 2019. Appendix A.
14. Reesink, Ger, Ruth Singer, and Michael Dunn. “Explaining the linguistic diversity of Sahul using population models.” PLoS Biology 7(11) (2009): e1000241.
15. Reeves, Jessica M., et al. “The sedimentary record of palaeoenvironments and sea-level change in the Gulf of Carpentaria, Australia, through the last glacial cycle.” Quaternary International 183 (2008): 3–22.
16. Ross, Malcolm. “Pronouns as a preliminary diagnostic for grouping Papuan languages.” In Pawley, Attenborough, Golson and Hide (eds.), Papuan Pasts, 15–66. Pacific Linguistics 572, 2005.
17. Sloss, Craig R., et al. “Holocene sea-level change and coastal landscape evolution in the southern Gulf of Carpentaria, Australia.” The Holocene 28(9) (2018): 1411–1430.
18. Williams, Alan N., Sean Ulm, Tom Sapienza, Stephen Lewis, and Chris S. M. Turney. “Sea-level change and demography during the last glacial termination and early Holocene across the Australian continent.” Quaternary Science Reviews 182 (2018): 144–154.