How Much
Italian
Are You?

You send off a tube of saliva. A few weeks later a website greets you with a pie chart and a confident sentence: you are 34% Italian. Maybe there's a little animation, a map with glowing regions, a suggestion that some part of you belongs to a place you've never been.

These tests are a real feat of engineering, and they are useful for some things: finding relatives, tracing broad migrations, satisfying plain curiosity. But that headline percentage carries more meaning than it can actually bear. Let's take it apart.

01 What the number actually means

Your genome isn't stamped with country labels. There is no stretch of DNA that says "Italy", and no gene is "an Italian gene". So what is a company doing when it decides you're 34% Italian?

Start with what your DNA is: a very long sequence of letters, and at scattered positions along it people differ. Those positions are the raw material. Almost no single variant belongs to one country. The letter that is common in Naples is usually just as common in Athens and Marseille. What separates one population from another is not which variants exist, but how often each one turns up, and which variants tend to sit together along a chromosome.

So a company lines your DNA up against a reference panel: a few thousand living volunteers who each said all four grandparents came from the same place. It slides along your genome in small windows and, for every window, asks one comparative question: of the reference groups I hold, whose members carry this combination of variants most often? Each window is handed to its best-matching group. Count the windows that landed on the Italian pile, divide by the total, and there is your 34%.

Read literally, then, "34% Italian" says something quite narrow: 34% of the small windows of my genome looked more like that company's present-day Italian volunteers than like any other group on their list. It is a statement about resemblance to a group of living strangers. It is not a count of genes that "came from" Italy, and it is not the share of your ancestors who lived there. Many of those windows would sit just as happily in Greece, Iberia or the Balkans; they were filed under "Italian" only because it was the closest label on offer, and every window has to be filed somewhere.

Notice that three quiet substitutions have already slipped in:

• "Italian"= a modern nation-state
• the yardstick= today's volunteers, not ancestors
• "you are"= a sample of your genome

Italy has only existed as a country since 1861. The reference "Italians" are alive now; they are your contemporaries, not your forebears. And the panels are built and rebuilt by each company from whoever they can recruit. Change the volunteers and you change everyone's answer, which is what you can watch happen next.

02 Same spit, different verdict

Here is one imaginary person, call her Alex, with one fixed genome. Nothing about her DNA changes. All that changes below is which company reads it, and which version of their panel they use. Flip between them.

Alex's "Italian-ness" lurches from 12% to 34% to 21% without a single base of her DNA changing. Categories appear and vanish ("Broadly Southern European", "Anatolian", "Aegean") because they are administrative bins the company invented, not natural kinds written into her cells. Companies really do revise customers' results when they update their panels. So the number is not really about Alex. It is about how she compares to a reference set that someone chose.

None of this is fraud. Each percentage is a defensible statistical estimate, with error bars that the pie chart quietly hides. The problem is only that it looks like a fact about Alex when it is really a fact about a comparison.

03 Your family tree explodes

Now for the deeper reason these labels can't mean what we want. Go up your family tree: 2 parents, 4 grandparents, 8 great-grandparents. Each step doubles the number of slots to fill, and that doubling outruns the entire human race almost immediately.

Drag the slider. It counts the ancestor slots your tree demands at each generation back, against the number of people who were actually alive on the whole planet at the time.

By around 40 generations your tree calls for over a trillion slots, but there were only a few hundred million humans alive, most of them nowhere near your ancestors' villages. The arithmetic is impossible unless the same people sit in your tree thousands of times over, reachable by many different paths. Genealogists call this pedigree collapse. Your family tree stops being a tree and becomes a densely woven net, one that knots together with everyone else's.

That same net makes room for something that sounds impossible: that everyone alive can trace their line back to a single shared ancestor. Watch it happen below. A small village is followed down the generations, each child joined to its two parents. In every generation some lines simply stop, because a person has no children, or none who go on to have their own. The pale lines are those dead ends. The survivors spread and cross, and before long a few of the original founders sit above the entire population. The gold founder is an ancestor of every person in the bottom row.

04 We are all recent cousins

Follow that net far enough and a striking result appears, backed by real numbers from real studies.

In 2013 Peter Ralph and Graham Coop scanned the genomes of over 2,000 Europeans and found that everyone of broadly European descent shares essentially the same set of ancestors from about 1,000 years ago. A person in Portugal and a person in Poland, seemingly unrelated, carry shared segments of DNA tracing to common forebears in the Middle Ages. In their words, nearly everyone alive in Europe a millennium ago who left any descendants at all is an ancestor of every European living today.

Push back a little further, a few thousand years, and Rohde, Olson and Chang showed that the whole of humanity converges on a single most-recent common ancestor: one individual who is a direct forebear of every person alive today, from Patagonia to Siberia. Keep going to roughly 5,000 to 7,000 years ago and you reach the "identical ancestors point", where every human then living is either an ancestor of all of us or of none of us.

So a boundary like "Italian versus Greek versus Slav" is not a wall between separate lineages. On the timescale your ancestor-slots demand, those populations mix together many times over. You don't have a sliver of "Italian" tucked in beside a sliver of "Greek". You descend from medieval Italy and medieval Greece and almost everywhere else in the neighbourhood, by countless overlapping routes.

05 The gene you got, not the tree you have

One last point undercuts the number completely. You have far more genealogical ancestors than genetic ones. DNA is passed down in chunks, and each generation you inherit a random half. Go back just ten generations or so and you carry no DNA at all from a large fraction of your actual ancestors. They sit on your family tree, but not in your genome.

So when a test reports "34% Italian", it isn't even reading 34% of your family tree. It is reading the fraction of the particular genetic chunks you happened to inherit that best match today's Italian volunteers. Change which half you had inherited (a different sibling, a coin-flip in meiosis) and the number would move again, before any panel was even chosen.

06 So what?

Enjoy the pie chart, and message the fourth cousin it finds you. That a tube of saliva can trace the broad sweep of human migration at all is real science, and worth appreciating.

Just don't read the percentages as a passport. "34% Italian" is a statement about a comparison to a chosen reference set, with hidden error bars, sampling one random half of a family tree so vast that it folded in on itself long ago. That tree is one you share, more recently than seems possible, with almost everyone.

The tests are right about one thing they never print on the box: reach back far enough and there is no "them", only a very large and very recent "us".