Kinship & Family StructuresMethods & Tools

Half, step and in-law: how genealogy counts a blended family

Half siblings, half cousins, stepparents and in-laws worked through with Wright's path method, shared DNA figures, and how charts and GEDCOM files record them.

Portrait of Elspeth GilchristElspeth Gilchrist · · 10 min read
Half, step and in-law: how genealogy counts a blended family

Blended families are as old as remarriage, and remarriage is as old as early death. A family historian meets them all the time: a widower with two sets of children, a widow who brings a daughter into a second marriage. Three small words sort out who is who: half, step and in-law. Behind them sits a question the words leave open, which is how much ancestry two people actually share.

Take one hypothetical household. Walter has a son, Tom, with his first wife, Ann. After Ann dies, Walter marries Beth, a widow with a daughter, Rose. Walter and Beth then have a daughter of their own, Clara. Three children grow up under one roof, and no two of them are related in the same way.

Kin-type letters keep the relationships straight: F father, M mother, B brother, Z sister, S son, D daughter, H husband, W wife, read from Ego outward, so FW is Ego's father's wife. Our notation guide warns that the identity FW = M holds only in a model without remarriage. Walter's household is where that shortcut breaks.

Three kinds of tie under one roof

Pair Kin type, from the first named Parents in common Kind of tie r
Tom and Clara FD, but not MD Walter half siblings 1/4
Rose and Clara MD, but not FD Beth half siblings 1/4
Tom and Rose FWD none stepsiblings 0
Tom and Beth FW none stepson and stepmother 0
Rose and Walter MH none stepdaughter and stepfather 0

Half relatives share one parent, or more generally one ancestor where full relatives share a couple. Step relatives are joined only through a parent's marriage. In-laws are joined through your own marriage or a sibling's: WM, HF, ZH, BW. The English prefix has a bleak origin. Step- goes back to an Old English element meaning deprived of a relative, related to a verb meaning to bereave.

Half relationships, counted by paths

Sewall Wright's coefficient of relationship, r, adds up the paths that join two people through ancestors they share. Each path is worth one half for every parent and child link on it: r = Σ (1/2)^n. Full siblings have two paths, one through each parent, each two links long, so r = 1/4 + 1/4 = 1/2. Tom and Clara have only Walter in common. One path, Tom → Walter → Clara, two links:

r = (1/2)^2 = 1/4

The missing parent echoes down the generations. Suppose Tom has a daughter, Ivy, and Clara has a son, Ned. Ivy and Ned are half first cousins. There is one path, Ivy → Tom → Walter → Clara → Ned, with four links:

r = (1/2)^4 = 1/16

That is half the 1/8 of ordinary first cousins, who would have a second path through a shared grandmother. Clara is Ivy's half aunt: Ivy → Tom → Walter → Clara is three links, so r = (1/2)^3 = 1/8, against 1/4 for a full aunt.

The rule that makes naming easy: find the sibling pair at the top of the two lines of descent. If they are half siblings, every relationship between their descendants is a half relationship, and its r is half the full value. The cousin degree and the number of removes do not change.

Our cousin calculator handles this with its "one person only" option. It counts generations below the shared ancestor, names the relationship ("half-first cousin" and so on) and gives an expected amount of shared DNA in centimorgans (cM), assuming 3,500 cM for one copy of the genome.

Relationship Pair in the example Generations below the shared ancestor Paths r Calculator's expected cM
Full siblings for comparison 1 and 1 2 1/2 2,625
Half siblings Tom and Clara 1 and 1 1 1/4 1,750
Half aunt and niece Clara and Ivy 1 and 2 1 1/8 875
Half first cousins Ivy and Ned 2 and 2 1 1/16 438
Half first cousins once removed Ivy and Ned's child 2 and 3 1 1/32 219
Half second cousins Ivy's child and Ned's child 3 and 3 1 1/64 109
Stepsiblings Tom and Rose no shared ancestor 0 0 0

Full siblings get 2,625 cM rather than the 3,500 that doubling r would suggest, and the reason is worth knowing. At any spot on a chromosome, each parent passes the same copy to both children with probability 1/2, so with probability 1/2 × 1/2 = 1/4 the siblings match on both of their copies there. Measured as the share of the genome covered by matching stretches, full siblings expect 1/2 + 1/4 = 3/4, and 3/4 × 3,500 = 2,625. Half siblings can match on only one copy at a time, since Tom's other copy came from Ann and Clara's from Beth. Their matching stretches are expected to cover 2 × 1/4 = 1/2 of a genome copy, which is 1,750 cM.

What a DNA test can settle, and what it cannot

Those figures are expectations, and real pairs scatter around them. The Shared cM Project, which the genetic genealogist Blaine Bettinger compiles from testers who know how they are related, makes the scatter visible. Some rows from version 4.0 (March 2020):

Relationship Pairs Average cM Lowest Highest
Full siblings 2,465 2,613 1,613 3,488
Half siblings 1,266 1,759 1,160 2,436
Grandparent and grandchild 1,106 1,754 984 2,462
Aunt or uncle and niece or nephew 2,695 1,741 1,201 2,282
First cousins 3,337 866 396 1,397
Half first cousins 691 449 156 979

Three lessons follow. Half siblings, grandparents and full aunts or uncles overlap almost entirely, and Bettinger states plainly that total shared cM cannot separate half siblings from aunts and uncles. Ages and documents usually can. Next, the lowest full-sibling totals fall inside the half-sibling range, and Bettinger reads full-sibling reports between about 1,600 and 2,000 cM as likely to be unrecognized half siblings. Finally, half first cousins and full first cousins overlap from 396 to 979 cM, so one cousin match rarely settles whether a grandparent married twice. Matches with several relatives on both sides of the suspected break do much better.

The stretches matched on both copies are the best tell between full and half siblings: on average a quarter of the genome for full siblings, and none for half siblings whose other parents are unrelated to each other.

Step and in-law ties: real, and r = 0

Tom and Rose share a home and a half sister, but no ancestor. Their r is 0, and they are expected to share no DNA by descent. The same holds for Tom and his stepmother, and for anyone and their in-laws, unless the two families were related before the marriage. A count of shared ancestry says nothing about how much a tie matters; households, inheritance law and kin terms can all treat step and in-law relatives as close kin.

Two consequences are easy to miss. A step relation can sit right next to a genetic one without passing anything along. Clara is a half sister to both Tom and Rose, so Tom's daughter Ivy and any child of Rose's are both Clara's half nieces or nephews, yet they share no ancestor with each other. The reverse happens too. When two families already joined by one marriage are joined by a second, in-laws become blood relatives of the next generation, which is how double cousins come about.

Drawing a remarriage

In a kinship diagram each marriage gets its own equals sign and its own line of descent, and each child hangs from the marriage it was born into. Put Walter in the middle, Ann on one side of him and Beth on the other, and Beth's first husband on her far side. Tom hangs from the line under Walter and Ann, Clara from the line under Walter and Beth, Rose from the line under Beth's first marriage. Mark Ann as deceased.

Read back, the half siblings appear as children of different marriage lines that share one parent. The step relations appear as nothing at all, which is correct: they are not descent, and a descent diagram should not invent a line for them. Our guide to drawing a kinship diagram works through a similar family in which a divorce and a second marriage produce half siblings.

How genealogy files store it

Genealogy programs exchange data in GEDCOM, a format maintained by FamilySearch, the family history service of The Church of Jesus Christ of Latter-day Saints. GEDCOM has no record for a half sibling or a stepmother. It has individuals and families. A family record names at most one partner tagged HUSB, at most one tagged WIFE, and any number of children tagged CHIL. Each individual points to the family in which they are a child (FAMC) and to every family in which they are a partner (FAMS). Walter's household becomes three family records:

0 @F1@ FAM
1 HUSB @WALTER@
1 WIFE @ANN@
1 CHIL @TOM@
0 @F2@ FAM
1 HUSB @BETHS_FIRST@
1 WIFE @BETH@
1 CHIL @ROSE@
0 @F3@ FAM
1 HUSB @WALTER@
1 WIFE @BETH@
1 CHIL @CLARA@

Everything else is inferred. A program sees that Tom and Clara are children of F1 and F3, which share Walter, and calls them half siblings. It sees that Rose's mother is a partner in F3 alongside Tom's father and calls them stepsiblings. The specification explains that the family record was first designed for a husband, a wife and their children, and allows it to stand for other unions, including cohabitation and adoption, whatever the partners' genders.

The link from a child to a family can carry a pedigree value, PEDI: BIRTH, ADOPTED, SEALING (a ceremony of that church), OTHER, and one more for a child placed in a guardian family. The specification warns that BIRTH is ambiguous. Some users have taken it to mean the genetic parent, others the social parent at the time of birth, so programs are told not to assume either meaning in data they import. A separate STAT value can mark a child's link to a family as PROVEN, CHALLENGED or DISPROVEN, and the authors say openly that the structures for family relationships have known limitations and are likely to change.

The commonest mistake, and what it costs

Enter Walter, Beth, Tom and Clara as one family and the software will call Tom and Clara full siblings. The error doubles every r beneath it. Ivy and Ned would show as first cousins, r = 1/8 with about 875 cM expected, when they are half first cousins at r = 1/16 and about 438 cM. If they then share 450 cM, the match looks low for first cousins, though it sits almost exactly on the half-first-cousin average of 449. The tree is right and the data entry is wrong. Keeping each person's father and mother in separate fields, as in the spreadsheet method from teaching kinship with a spreadsheet, makes the slip easy to catch.

One to try

Rose later has a son, Sam. What is Sam to Ivy, and what is Sam to Ned?

Sam and Ivy have no ancestor in common. Rose and Tom are stepsiblings, so r = 0. Families sometimes call such a pair step-cousins; genealogically they are unrelated.

Sam and Ned are another matter. Their mothers, Rose and Clara, are half sisters through Beth. One path, Sam → Rose → Beth → Clara → Ned, has four links, so r = (1/2)^4 = 1/16. Enter 2 and 2 with "one person only" in the calculator and it reports half-first cousins, 1/16, about 438 cM. Ivy and Ned are both Walter's grandchildren, yet to Sam one is no relation at all and the other is a cousin by descent.

Bettinger's guide takes the DNA side of a blended family further, from choosing a test to sorting matches whose shared totals fit more than one relationship.

Sources and further reading

More in Kinship & Family Structures