There are many stories about families where the baby of an unwed teen was passed off as her own mother’s child. A recent query in a FaceBook group asked whether DNA testing could tell whether her aunt, her mother’s much younger presumed sister, was one of those cases, so actually the asker’s half sister. The problem is that the amount of cM shared by those two relationships completely overlaps, roughly 1600-2000 with outliers (click here for the charts at DNApainter).

Known paternal half siblings at DNA-sci

Previously I collected data and studied segment sizes to see how to tell those relationships apart (click here), finding that paternal half siblings will share more larger segments than aunt/uncles. However in this particular case, maternal half siblings, the difference is less clear. Click here for the calculator at DNA-SCI that takes segment sizes into account by using the number of segments. When I tried using it with the numbers of a pair of known paternal half siblings, that calculator predicted that aunt or uncle was more likely (see image). So this approach will not give a clear answer.

A half sister will get half their DNA from their other parent. This will usually result in some close matches that are not shared.  On the other hand, an aunt’s close matches should all be shared with her niece, since they share the same ancestors, so looking at their “not in common” matches might work. The niece, however, is expected to have close matches from her other parent’s side. Here are some half sibling examples from the tool “Match Both or 1 of 2” on GEDmatch:

The kits that match only half sibling A0 and not A5

The kits that match only half sibling A5 and not A0

Notice that each half sibling has a match larger than 200 cM that is not shared with the other. If you have pro tools and access to both accounts on Ancestry, you can also look at this by comparing each person’s list and the matches in common. There is no automated way to do that there yet. Family Tree DNA has a “not in common with” feature, but no other company has it.

What can interfere with this finding of large “not in common” matches? If the parents are related or from an endogamous population. Also if each other parent is from a population that is sparsely tested like Germans, Chinese, French, and others, so that there are no close matches. In that case, the difference in ethnicity estimates can help.

There is a tool on GEDmatch to test if parents are related (click here for that post). That relatedness will often show for a very endogamous population as well.

I have some large matches that my niece does not have

The other problem I found with the niece/aunt comparison is the generation difference leading to some ancestors falling off. My own niece did not get much DNA from my Munson great grandfather resulting in my having good matches to some people descended from him that she does not! These people are my second cousins once removed so her third cousins. The ISOGG wiki suggests that about 10% of 3rd cousins will not match. However, she does match a few others on that line at about half what I share.

other than to her father, Shipley, my niece’s (NB) matches are significantly smaller than mine (PB)

 So consistently having smaller matches than I do to the same person is another indicator of the niece relationship. A half sibling would share similar amounts or larger in some cases and smaller in others. As in the example below.

Half siblings share similar cM with close matches, varying randomly as to who has more.

Conclusion: There is no easy way to determine an aunt from a half sister. It takes careful analysis, looking at shared matches as well as the matches not shared. The exception being half sisters sharing a father who will always share a full X (click here for that post)