If a testing site shows you a "close family" match sharing around 1,700 or 1,800 centimorgans, that number alone cannot tell you the relationship. A grandparent, an aunt or uncle, and a half-sibling all produce similar totals; the averages for the three are within eighteen centimorgans of each other. This is not an error in your results but a known limit of the measurement. A second number, however, can help distinguish them.

Why the total cannot separate them

The Shared cM Project collects submitted results from people with documented relationships, and its fourth version, published in 2020, gives observed ranges rather than theory. The project gives the following observed ranges for these three relationships:

  • Grandparent and grandchild: 984 to 2,462, averaging 1,754.
  • Aunt or uncle with niece or nephew: 1,201 to 2,282, averaging 1,741.
  • Half-siblings: 1,160 to 2,436, averaging 1,759.

Any total from 1,201 to 2,282 is consistent with all three relationships. In this 1,081-centimorgan band, the total is ambiguous. Because the averages are so close, being near one of them is not evidence either.

The reason is that all three relationships share the same expected fraction of DNA. You get about a quarter of your genome from each grandparent, and the expected share with an aunt, an uncle or a half-sibling is also about a quarter. They are different relationships that produce the same expectation, and the spread around that expectation is wide enough to swamp the small differences between them.

The number that does help

The amount of shared DNA is similar, but its arrangement is different. DNA is inherited in blocks, and every generation those blocks get broken up and reshuffled. A relationship separated by more generational steps has had more opportunities for that shuffling, so the same total arrives split into more, shorter segments.

A grandparent and grandchild are two steps apart with relatively little reshuffling between them, and the shared DNA tends to arrive in fewer, longer pieces. Half-siblings share a parent, and their shared DNA has been through a different number of shuffles. So the segment count carries information the total does not.

Two cautions about using it. Not every service shows you segment data, and where it is available the presentation differs, so you may need to work from a site that exposes it. The segment count only shifts the probabilities. It can make one relationship more likely than another, but it does not turn an ambiguous match into a certain one.

The X chromosome sometimes decides it

A further asymmetry in inheritance can sometimes rule out a possibility outright.

The X chromosome is inherited differently from the rest. A father passes his single X to his daughters and none to his sons, so certain relationships cannot share X DNA at all. That means the presence or absence of X sharing can eliminate some candidate relationships on one side of a family while saying nothing about the other side.

This means an ambiguous match on the paternal side may be resolvable where one on the maternal side is not, or vice versa, depending on the sexes of the people involved. It is worth checking rather than assuming, and it is the one piece of evidence in this area that can be conclusive.

Numbers from different companies are not comparable

This one catches people who are doing everything else right.

Different services use different thresholds for how small a matching segment has to be before they stop counting it, and they process the raw data differently before measuring. One company applies an algorithm that discounts regions where matches are common in the general population, which reduces its totals relative to a service that does not. So the same two people, tested by two companies, produce two different centimorgan figures.

If you are comparing your number to a published chart, check which basis the chart is on. If you are comparing two of your own matches, make sure they came from the same service. A difference of a couple of hundred centimorgans between platforms is ordinary and is not evidence about the relationship.

What to do with an ambiguous match

Our shared DNA and relationship calculator will show the candidate relationships for a given total. Then look at ages, because that often resolves what the genetics will not. A grandparent relationship implies a particular generational gap; a half-sibling usually does not. Where the ages are known this is frequently decisive and costs nothing.

Then look at who else the match shares DNA with. A match's own matches place them in a branch of the tree even when your direct comparison is ambiguous, and this is usually the fastest route to an answer.

People usually ask this question because a result was unexpected. If that is why you are here, the arithmetic above is about categories of relationship and cannot tell you anything about a particular person or what any of it means for your family. There are communities and counsellors who deal specifically with unexpected DNA results, and they are a better source of help for that part than a page about centimorgans.

Where this comes from, and what will date

The ranges and averages are from version 4 of the Shared cM Project, published March 2020 and current at the time of writing. The figures are observed submissions rather than theoretical predictions, which is why the ranges are as wide as they are.

Two things carry dates. Earlier versions of the same project circulated different numbers — the 2017 edition gave a noticeably narrower overlap — so a chart found online may be out of date, and it is worth checking which version you are reading. And company processing changes: thresholds and algorithms get revised, so the comparability warning above is about a moving target.

The underlying point does not move. Three relationships share the same expected amount of DNA, so no total will ever separate them cleanly. That is a property of inheritance rather than a limitation of any particular test.