One childhood,
not many.
Fifty-four cultures. Two thousand four hundred and twenty-five practices. Nine movements that argue humanity agrees about raising children far more than folk intuition assumes — and disagrees, where it does, about things you would not have guessed.
The atlas begins with everything on the map.
Every named culture the corpus documents is placed roughly where the fieldwork was done. Region is the colour you would expect: Africa warm, Europe blue, Americas green, Asia divided. Scroll to see what this map has to say — and what it does not.
The 54 cultures with reciprocal convergence data appear here. Thirty-four thematic, pedagogical and religion cross-cuts feed the practice layers but do not sit on the graph as nodes; they surface in Explorer and search.
Then the edges. All of them, at once.
This is the hairball you were promised. Every pair of cultures that shares at least one converging practice draws a line. Held for two beats so you can see the density with your own eyes.
The web stays connected far past any threshold a sceptical reader would pick. Currently: one connected web · 54 of 54.
The map dissolves. Order comes back from another axis.
The 54 cultures leave the world layout and re-settle by how many practices they share. Region-coloured, the picture is a scatter: neighbours no longer cluster.
Switch to ecology — forager, pastoralist, agrarian, industrial, mixed — and the picture composes. The Hadza and the Ju/'hoansi (opposite ends of Africa, both foragers) sit together. Navajo and Mongolian, pastoralists across three continents, cluster. And Japan's strongest tie in the corpus, once you compute it, is not any of its neighbours — it is Yucatec Maya.
Ecology is analyst judgement, authored per cultures_ecology.csv.
Every use of it on screen says so. The pattern it reveals, though, was in the
edges before we named it.
The same practice, arrived at by unrelated peoples.
When unrelated cultures independently reach the same solution, that convergence is the strongest evidence of function a corpus like this can carry — the same logic that biologists call convergent evolution. The atlas ranks 158 practices that clear a strict bar: three or more other regions and two or more other ecologies in the convergence set.
Scrub the ranked strip below. The arcs on the stage redraw as jumps between the origin culture and every culture that arrived at this practice independently.
Convergence is evidence of function, not proof of outcome. The atlas states this once here and lets the rendering repeat it everywhere.
What can a modern family actually pick up?
Every practice in the corpus sits somewhere in this quadrant. The horizontal axis is convergence — how many independent cultures arrived at it. The vertical axis is transferability — whether a modern nuclear family could realistically adopt it. Two sliders move the lines; the dots resettle.
Village required is not a decline narrative. It says here are things
humanity broadly endorses that infrastructure, not parenting, makes possible.
Practices flagged exclude-with-note for harm are set aside entirely
and never shown as options.
Where the arguments actually are.
Consensus per domain — converging links divided by converging plus contested. All eleven domains cluster tightly around 0.85, which is itself the finding: there is no domain where cultures mostly disagree.
The three sharpest disagreements in this domain: two practices, two stated purposes, two evidence tiers. No adjudication.
Both sides of every pair get their own stated purpose at their own evidence tier. Tier 1 is direct fieldwork; tier 3 is inference from adjacent sources. Contested links appear dashed on the graph.
What the world expects of a child at every age.
Rings are age bands (0-1 at the centre, 12-18 at the rim). Sectors are the eleven domains. Each dot is a practice, weighted by evidence tier. The heavier ring between the 3-6 and 6-9 bands is the 5-to-7 shift — one of anthropology's oldest regularities, reproduced here without being told to.
Discipline, autonomy and real work all jump sharply at the 6-9 band. The clock makes it something you can point at.
Local words for things English has no name for.
The corpus carries 2,425 local terms — 1,323 of them named as concepts with no direct English equivalent. They cluster on the stage by domain and glow by culture. Hover any point for the term, its gloss, and the practice it names.
Concept ≠ translation. A term is flagged concept only when the gloss explicitly refuses a one-word English mapping, uses a non-Latin script, or expands into a phrase. Everything else files as a translation.
What the corpus has already lost, or is losing.
Every practice with a still-practised status of partially, no, or historical is a dot on the timeline. X-axis is the decade the fieldwork was done; y-axis is the practice's domain. Fill fades from partial → gone → historical.
No elegy. Some of these ended for good reasons — Sweden's 1979 ban on corporal punishment shows up as a change in "still practised" of a very different kind. Reading the timeline is the argument.
Pick a culture. The atlas rearranges around it.
The mirror is the personal manual — one culture at the centre, the other 53 rearranged around it. Its position in the web, its nearest kin, the practices where humanity converges with it, the practices only it does, its side in every debate, and a shortlist of practices from its kin that a family sharing its constraints could actually pick up.
The mirror works for any of the 54. Fifty-four cultures are worth taking seriously; no default is imposed. Choose one.