The Long Walk Out

How to read this

Ember shows modelled human presence at the selected date, with a hotter leading edge where arrival is recent. Dated sites anchor the reveal, and illustrative dispersal paths fill the space between them. This is not a census or a map of everything people knew. Dark land sits outside the current modelled presence, not necessarily outside human experience. When the model includes a retreat, the ember cools back to slate — north-west Europe and northern Siberia empty out under the ice, and Greenland is abandoned and resettled more than once. Oceans are a constant background with no explored or unexplored status.

For the first quarter of a million years this map only lights up inside Africa, and that is enforced, not hoped for: the African spread is fenced to the continent so it cannot bleed across Gibraltar or the Red Sea.

Then people leave, and mostly fail. Misliya at 185,000 years ago, Skhul and Qafzeh at 120–90,000, Al Wusta in Arabia, Grotte Mandrin in France at 54,000 — all real, all dated, all extinguished. You can watch each one fade back to slate. The dispersal that actually founded everyone alive outside Africa today comes late, after about 50,000 years ago, and the reason we can be firm about that is Neanderthal DNA: every non-African carries it, and the interbreeding is dated to 49,000–45,000 years ago. A founding population older than that would not fit.

What the map is guessing at

• Routes are drawn as smooth arcs. Real dispersals were slow, reversible, and full of dead ends; a line between two dated sites is a hypothesis, not a road.
• The 65,000-year dates in Sunda and Sahul do not fit the genetic ceiling. Lida Ajer and Madjedbebe are drawn as isolated toeholds that fade, and the surviving wave arrives around 50–45,000 years ago. That is one reading of an argument that is still live; the excavators read it differently.
• Land bridges are pinned to their own dated chronologies rather than derived purely from the sea-level curve here. Beringia opens about 37,000 years ago and drowns by 11,000; Sundaland is continuous from roughly 72,000 to 12,000. Sea level during the middle of the last glacial is itself disputed, and a coarse curve makes bridges flicker on and off, which is an artefact, not history.
• The ember is a flat fill, not terrain. It carries only whether the model places people there, and the soft edge is a drawing style rather than a measured margin of error.
• Coastlines are today's. The pale grey shelves — Sunda, Sahul, Beringia, Doggerland — are approximate reconstructions of land exposed by low sea level, not surveyed bathymetry. Most coastal archaeology of the Ice Age is underwater and undiscovered.
• Ice sheets are schematic outlines scaled by a standard sea-level curve, drawn to show the barrier, not its exact edge.
• Absence of evidence is thin in tropical, forested, and submerged places, where organic material rarely survives. Sparse regions are probably under-dated rather than empty.
• Dates on this map are calendar years before present. Where a site's age is a range, the map uses a mid-point and the log gives the range.
• A few places stay dark on purpose. Severnaya Zemlya and Franz Josef Land were only sighted in the 1870s–1910s and have never had a permanent population; Antarctica, off the bottom of this map, still has none.
• Watch the last flickers. Mauritius, the Azores, Cape Verde, Bermuda, the Falklands, the Seychelles and the Galápagos had no people on them at all until the 1500s–1800s — they are the only sizeable habitable land that waited for ships rather than canoes.