The Evidence

For cross-checking and for systematic searches across large numbers of individuals, this research also drew on the AADR, a curated compendium maintained by the Reich Lab at Harvard that reprocesses published ancient genomes through a consistent analytical pipeline. This is particularly useful because different original studies sometimes use different methods or levels of resolution when assigning haplogroups. The AADR therefore provided an additional way of checking whether a proposed placement on one of the four family lines was supported by the available ancient-DNA data.

Ancient DNA is fragmentary and degraded. Ancient-genome studies commonly report a "coverage" figure, indicating how many times positions in the genome were independently read, alongside estimates of possible modern contamination. Low coverage can make a very precise haplogroup assignment less secure because the relevant downstream mutations may not have been recovered at all. These factors were considered when assessing how much confidence to place in a genetic result, particularly where a highly specific branch assignment depended on limited ancient DNA.

Mitochondrial and Y-chromosome haplogroups form branching genetic trees. Some ancient individuals can be resolved only to a broad, early branch because their DNA is incomplete or was not sequenced deeply enough, while others can be placed much farther downstream. When assessing whether an ancient individual genuinely belonged on one of the four family lines, the published level of resolution was checked rather than assuming that a broad haplogroup call automatically identified the family's more specific branch. Different studies and databases can also use different names for equivalent or closely related positions on the tree, so those relationships were cross-checked where necessary.

This research also used genetic-genealogy databases, haplotree browsers and community-maintained tools that can sometimes provide more detailed branch assignments than a published paper itself. These tools are valuable, but they are not peer-reviewed and their automated calls can occasionally overstate precision, particularly for low-coverage ancient genomes. Where a claim from one of these sources could be checked against the original study, its supplementary data, the underlying sequencing evidence or a curated resource such as the AADR, that check was made before deciding whether and how the claim should be used on this page.

Not every search produces a result. Where a particular genetic marker has been searched for across the available ancient-DNA evidence and no matching individual has been identified, that is treated as a genuine limit of the current evidence rather than something to fill with assumption.

A negative search therefore means only that no published ancient genome examined for this project has yet been found carrying that marker. It does not mean that the lineage itself was absent from the past.

Ancient DNA survives and is recovered unevenly. Whether someone's DNA survives at all depends on burial conditions, soil chemistry and chance. Whether those remains are excavated and sequenced depends on where archaeologists have worked, which individuals were sampled and what questions particular research projects were designed to answer.

The four family lines illustrate that limitation in different ways. On Xenia's line, C13538 shows that the downstream X2c1a branch was present in early medieval Britain, but the modern family result resolves only to X2c, so the particular downstream maternal route leading to the Wilde family remains unknown. On Oisin's line, no ancient genome examined for this project has yet been identified carrying the family's tested R-FGC10125 marker.

These are limits of the ancient DNA so far recovered, sequenced, published and examined for this project, not limits on what actually happened in the past. A marker not found in the datasets searched here should therefore be read as “not yet found”, never as “was never there”.

Ancient DNA research is advancing rapidly, and the picture presented here can only reflect the evidence available when this page was researched and updated. New excavations continually add individuals to the ancient genetic record, while improvements in DNA recovery, sequencing and analysis can produce more detailed results from remains that have already been studied. Haplogroup trees and ancient-DNA databases are also revised as new samples and mutations are identified, so the placement or interpretation of an individual can change even when the archaeological discovery itself has not. The reconstruction presented here is therefore a snapshot of an evolving body of evidence, based on the published data, databases and analytical tools available to me, and on my own interpretation and cross-checking of them. Where the evidence changes, or better evidence becomes available, the conclusions on this page may need to change with it. That is a necessary consequence of following a scientific record that is itself still being built. I have tried throughout to represent the evidence accurately and to distinguish clearly between what the data demonstrates and what I have inferred from it; any errors of interpretation, transcription or synthesis are mine alone.

Genetic evidence can trace branches through time, but it cannot by itself reconstruct the landscapes in which those populations lived. The Ice Age chapters therefore combine published palaeoclimate and palaeoenvironmental datasets with spatial reconstructions of glacial refugia to build a broader picture of conditions across Eurasia.

All of the climate graphs are built from published numerical data. The underlying records come from established repositories including PANGAEA, the NOAA/NCEI World Data Service for Paleoclimatology, and the Neotoma Paleoecology Database. Regional temperature comparisons use CHELSA-TraCE21k, a transient palaeoclimate model reconstruction covering the last 21,000 years; its published scale, offset and units were confirmed directly from the archive's own file metadata, and its chronological convention was independently checked against the methodology described in the original paper, rather than assumed from a secondary description.

The records come from very different environmental archives, including Greenland ice cores, marine and lake sediment cores, fossil pollen sequences, cave speleothems, biogenic silica and plant-wax biomarkers. Each captures a different part of the environmental story. Some record temperature directly or indirectly; others reflect vegetation, hydroclimate, atmospheric dust or ecological change. They are treated as complementary lines of evidence rather than as interchangeable measurements.

The graphs are plotted from the published source data rather than reconstructed visually from existing figures. Values are retained as reported, and gaps remain visible where the underlying datasets contain no observations, rather than being smoothed or interpolated across. Where a record contains a genuine break, such as a dated hiatus in a speleothem's growth, that break is shown as a gap in the line rather than joined across. Where new series are derived from published data, such as combining individual pollen taxa into broader woodland, steppe or tree/shrub categories, those groupings are calculated directly from the source counts rather than estimated from a chart, and follow groupings the original researchers themselves proposed wherever a paper defined them. Where a paper also reported its own summary statistic for a derived value, that figure was recalculated independently from the raw data as a check before use.

Published dataset descriptions were not always taken at face value. Where possible, the chronological range and content of a dataset were checked against the original study or its supplementary data rather than relied upon from a summary title or abstract alone. In at least one case, this check showed that a dataset's own title described a narrower time span than the data it actually contained; the graph reflects the verified range confirmed in the data itself, not the title.

A common chronological framework, and a consistent set of shaded bands marking major climatic episodes such as Heinrich Stadials, the Bølling-Allerød and the Younger Dryas, is used across every figure so that events can be compared directly between regions. Greenland ice-core records provide the principal reference chronology, while regional records from Iberia, the Balkans, the Levant and Siberia show how those wider climatic shifts were expressed in very different local environments.

The refugia maps were built as geospatial syntheses rather than copied from a single published map, combining several separate, real geographic and climate datasets. Modern coastlines come from Natural Earth, via geojson-maps.kyd.au. The lower Ice Age coastline, showing the extra land exposed by the drop in sea level, including the ground between Britain and mainland Europe, is drawn from Assis et al. (2017, Bio-ORACLE v2.0) bathymetry data. Ice-sheet extent comes from Batchelor et al. (2019, Nature Communications). The shaded relief showing the shape of the land itself is a Natural Earth dataset (Tom Patterson/NACIS), derived from NASA SRTM elevation data; its coverage extends east into Siberia, so the same relief layer forms the base of the Baikal map as well as the European one.

The four European refuge zones, Franco-Cantabrian, Iberian, Italian and Balkan, were drawn using a single, consistent test applied evenly across the map: was this ground's coldest month warmer than −5°C during the Ice Age, using the CHELSA-TraCE21k climate reconstruction? Wherever the answer was yes, within the general area each refuge is known from the literature to have occupied, that land was included. The same threshold was tested against all candidate regions, including areas ultimately excluded, before the final zones were drawn, so no zone was drawn more generously than another.

The Siberian map around Lake Baikal was built differently, since the question there was not where conditions stayed mild, nowhere in this landscape did they, but where the ground remained open, ice-free and passable despite extreme cold. Rather than a temperature threshold, this zone was drawn using the region's own geography: the outline of Lake Baikal itself, and the boundary of major glaciated mountain areas in the Altai, Sayan and Baikal ranges, drawn from GLACIMONTIS (Lima et al., 2025), a published global database of mountain glacier extents at the Last Glacial Maximum, which were excluded from the zone. What remained, open steppe and tundra-steppe ground capable of supporting large grazing animals and human movement, centred on the broader southern Siberian region around and east of the lake rather than its shoreline, forms the mapped range.

The refuge zones are deliberately broad. They mark areas identified in the literature as having remained viable, occupied or ecologically stable through the coldest part of the Ice Age, not precise borders in the way a modern country has borders.

Neither the graphs nor the maps are meant to place a particular maternal or paternal lineage at an exact site or within a precise shaded zone. Their role is to set out the wider environments those lines were living in, so the genetic and archaeological evidence has a landscape to sit within.

As with the genetic evidence, palaeoclimate and palaeoenvironmental datasets are periodically revised and extended; the reconstruction presented here reflects the data available and verified at the time of writing.