---
title: "qpAdm models for European ancestry: the standard recipe and its regional variations"
description: "The three-source model that rebuilt European prehistory — WHG, Anatolian farmers, steppe pastoralists — as a working qpAdm recipe: exact source and right-set choices, regional adjustments, failure modes and a worked reading."
canonical: https://www.ancestrify.io/blog/qpadm-models-european-ancestry
date: 2026-09-01T09:28:00+00:00
updated: 2026-09-24
author: "Andi Thomaj"
---

# qpAdm models for European ancestry: the standard recipe and its regional variations

European ancestry is qpAdm's home fixture: the model class the method was
[literally introduced to defend](/blog/how-qpadm-changed-ancient-dna) in 2015, rerun thousands of
times since, with the best-sampled sources in the entire ancient record. That maturity makes it
the right first recipe — the version of the method where the choices are settled enough to state
as a table and the failure modes are all known by name. Here is the standard distal model, why
each piece is what it is, and how to adapt it without breaking it.

## The three streams

Present-day Europeans decompose, to first order, into three deeply divergent ancestries that met
between roughly 6000 and 2000 BCE:

- **Western Hunter-Gatherers (WHG)** — the Mesolithic foragers of post-glacial Europe.
- **Early European Farmers (EEF)** — Neolithic migrants whose ancestry traces to northwest
  Anatolia, carrying agriculture into Europe from ~6500 BCE.
- **Steppe pastoralists** — Bronze Age herders of the Pontic-Caspian steppe (themselves a roughly
  even [EHG and CHG mixture](/blog/why-qpadm-models-get-rejected#negative-weights), which matters below),
  expanding west from ~3000 BCE.

## The standard distal recipe

**Left (target + sources):** the target, plus `Turkey_N` (Barcın Neolithic, the standard EEF
proxy), `WHG` (Loschbour/Villabruna-cluster individuals), and `Russia_Samara_EBA_Yamnaya` (the
canonical steppe proxy). All three predate every living European —
[distal by construction](/blog/qpadm-distal-vs-proximal).

**Right (outgroups):** [the O9 spine](/blog/how-to-choose-qpadm-sources-and-outgroups#the-standard-right-sets), namely
`Mbuti.DG`, `Ami.DG`, `Basque` (or another O9 member per the published variant), `Onge.DG`,
`Ust_Ishim.DG`, `Mota.DG`, `MA1`, `Villabruna`*, `Vestonice16`, `ElMiron`, `Ethiopia_4500BP`,
with the era contrasts that give the model its discrimination: `Russia_EHG` and `Georgia_CHG` (or
`Iran_GanjDareh_N`) to split steppe from farmer ancestry, `Levant_N` to pin the southern edge.
(*Villabruna sits on the right only when WHG is proxied by other individuals — the
[never-cladal-with-a-source rule](/blog/how-to-choose-qpadm-sources-and-outgroups); swap it out
if your WHG label contains it.)

Run [lowest rank first](/blog/qpadm-best-practices): two-source EEF+steppe often passes for
southern targets where WHG rides inside both sources' backgrounds; admit the third source when
[the rank test calls for a third ancestry component relative to the right set](/blog/qpwave-explained).
Testing the lowest rank first is statistical parsimony: the simplest model that fits, relative to
the right set used. It is not a head count of the migrations in anyone's history.

## What to expect, calibrated

The published gradients are your sanity check: steppe ancestry is at a maximum in the Baltic and
Ireland/Scotland/Norway (~50%), EEF at a maximum in Sardinia (~70–80%) and the Mediterranean,
WHG everywhere the smallest fraction — at its peak in the Baltic — and essentially every value
between 30–50% steppe / 30–60% EEF / 5–20% WHG occurs somewhere on the map. A first result far
outside those envelopes is a [pipeline question](/blog/qpadm-data-preparation-merge-aadr) before
it is a discovery.

## The named failure modes

- **The southeast fourth stream.** Aegean, Balkan, Italian and Jewish-diaspora targets routinely
  reject the three-source model — correctly — because post-Neolithic Near Eastern ancestry
  (Anatolia_BA/Levant_BA-related) is real there. The fix is a fourth source or
  [a proximal reframe](/blog/qpadm-distal-vs-proximal), not right-set surgery. The same applies
  to Iberian targets with recent North African ancestry (add `Morocco_Iberomaurusian`-related or
  a proximal Maghreb source).
- **Steppe/farmer collinearity.** Yamnaya's CHG half overlaps Iranian-plateau ancestry: with a
  weak right set the model wanders,
  [SEs balloon](/blog/how-to-read-qpadm-p-value-z-score-standard-error), and northeast targets
  can flip between EHG-flavoured solutions. The EHG/CHG right-set members above are the cure —
  present *because* of this failure mode.
- **The Finnish/Baltic east.** Uralic-associated Siberian ancestry (Nganasan-related) enters
  late in the northeast; three-source models of Finns, Estonians and Saami-admixed targets fail
  until a fourth Siberian source is admitted.
- **Umbrella-label WHG.** `WHG` in the AADR pools individuals from Iberia to the Balkans across
  millennia; [check the .anno file](/blog/aadr-sample-labels-sg-dg-ho-explained) and prefer a
  tight cluster for source duty.

## A worked reading

A concrete pass, from [our own report format](/blog/qpadm-model-record-explained#a-worked-example): an
Irish-ancestry target, p = 0.31, Yamnaya 0.48 ± 0.021, Turkey_N 0.38 ± 0.024, WHG 0.14 ± 0.018,
with all weights [more than 2 SE from 0 and 1](/blog/qpadm-best-practices), all three nested two-source
models rejected below p = 0.01, ~710k SNPs used. Reading: compatible, well-determined,
literature-consonant (high-steppe northwest edge), and the nested rejections show that, relative to
this right set and this candidate pool, the simpler models do not fit. That is not proof that three
historical populations mixed; it is a claim the data could have rejected, which no percentage bar
chart can make, and that is [what the method is for](/qpadm).

Terms used here are defined in the [glossary](/glossary).

## References

- Haak, W. et al. (2015). Massive migration from the steppe was a source for Indo-European
  languages in Europe. *Nature*, 522, 207–211. (The recipe's debut.)
- Lazaridis, I. et al. (2016). Genomic insights into the origin of farming in the ancient Near
  East. *Nature*, 536, 419–424. (O9 and the distal framework.)
- Mathieson, I. et al. (2018). The genomic history of southeastern Europe. *Nature*, 555,
  197–203. (Regional variation and the Balkan fourth stream.)
- Patterson, N. et al. (2022). Large-scale migration into Britain during the Middle to Late
  Bronze Age. *Nature*, 601, 588–594. (Modern-era refinements to the same model class.)