Eight generations back, a man has 34 ancestors who could have passed him part of his X chromosome. In 40,000 simulated family trees at each X map length we assumed, from 170 to 200 cM, about eight or nine of the 34 passed on any of it on average; at our middle setting the count fell between five and twelve in nine trees out of ten. Taken alone, each of the 34 was likelier to have passed on nothing than something. Two rules keep the candidate list short: a man's single X comes from his mother, while a woman has one from each parent, and her father hands his on whole.
A man's X ancestry is narrower than his sister's
FamilyTreeDNA's help page on X-Matches (updated 8 January 2026) says fathers pass their lone X to daughters "exactly as they received it from their mothers", and never to sons. The 23andMe DNA Relatives article (updated 13 October 2026) says women inherit an X from each parent and men from the mother only.
Every X ancestor a man has, his sister has too. She also carries their father's X, which came whole from his mother, so that grandmother's lines are open to her and closed to her brother. The father's father is closed to both. For a woman that shuts one quarter of the tree; for a man, the whole paternal half. This assumes the usual XX and XY pattern and ignores the X's short pseudoautosomal tips, which recombine in men.
The Fibonacci sequence predicts the X ancestor count better than doubling
Each man in a tree adds one X ancestor above him, his mother, and each woman adds two. So the count at any generation is the sum of the two counts before it, the Fibonacci rule. We checked it by brute force to twelve generations, for both sexes. Blaine Bettinger charted the pattern on The Genetic Genealogist in 2008.
Ten generations back, 89 of a man's 1,024 ancestral slots can have carried his X, about 8.7% on our arithmetic. For a woman it is 144, about 14.1%. Ahnentafel numbers name the slots: a father is twice his child's number, a mother twice plus one. Francis Galton described the same numbering in Nature on 6 September 1883.
Most real shares come out lumpier than the expected ones
An expected share is an average over many families. A mother passes her son one X spliced from her two, so on average half is her mother's and half her father's. Across our assumed range, about one man in ten got his whole X from a single maternal grandparent, because his mother's meiosis, the cell division that makes an egg, put no crossover on it. The two parents of his mother's mother each have an expected 25%. At 180 cM each gave nothing in about 19% of 40,000 simulated trees (16% to 21% across the range), and the middle 90% of shares ran from 0 to 67%. One share is as tidy as its average: a woman's paternal grandmother supplies 50% of her X in every family.
Our X-DNA inheritance calculator lists the expected shares, which our simulation reproduces within 0.3 percentage points. Until 14 October 2026 it headed that column "Expected X" with nothing on the spread, its generation ladder stopped at ten while its slider reached twelve, and it gave a one-in-ten eligibility figure without saying that holds only for a man. The revision committed that day relabels it "Average X share" with a note that a listed ancestor may have given much more or nothing, extends the ladder to the chosen depth, and gives a woman's figure, about one in seven.
Distant X ancestors of a man left nothing more often than not
Louis Kessler estimated this in a post on his Behold blog in December 2016. Averaging men and women and assuming about two crossovers on a mother's X, he found that on the X "it only takes 6 generations" for an ancestor's chance of passing on a segment to fall below half. Our simulation is a cross-check that runs each sex separately and models interference, the tendency of one crossover to discourage another nearby. Because his figures blend the sexes, we do not set them beside ours.
For a man eight generations back, all 34 X ancestors gave nothing in more than half of the 40,000 simulated trees, at every map length we tried. Yet eight or nine gave something on average: the X arrives in a few large pieces. Three generations nearer, ancestor 63, at the top of his direct maternal line, and her husband, ancestor 62, each carry an expected 6.25%; each gave nothing in 54% to 59% of trees, and the median share was zero. At the seventh generation one of a man's 21 X ancestors fell below half, and at the eighth one of a woman's 55 did: ancestor 341, on the line alternating father and mother, where the X recombines least.
X segments run longer for their age than autosomal ones
The autosomes, chromosomes 1 to 22, recombine in every parent's meiosis; the X only in a woman's. A centiMorgan (cM) measures genetic length by how often crossovers fall, so on our reading an X segment meets fewer cuts per cM each generation. FamilyTreeDNA's X-Matches page puts a working figure on it: "a 16 cM block on the X chromosome" carries about as many shuffles "as a 7 cM block on any other chromosome". By that rule of thumb, a long X segment can be older than it looks.
FamilyTreeDNA treats an X match more cautiously than an autosomal one
FamilyTreeDNA keeps X matches out of its headline figure: centiMorgans on the X "do not contribute to overall shared cM". An X segment alone puts nobody on a match list either; a match must first meet the minimum "not counting X DNA". 23andMe says it compares "autosomal DNA (chromosomes 1-22) and X chromosome(s)", with a general threshold of 7 cM and 700 SNPs (single-letter DNA markers) for a matching region, and does not say whether that applies to the X separately. Neither page offers a way to prove a line. Relationships are covered in our guide to shared DNA with grandparents and half-siblings.
Pedigree collapse makes real trees messier than the model
Every figure here comes from an idealised pedigree, with a different person in every slot and every parent a biological one. Real trees fold back on themselves, as our guide to ancestors a thousand years ago explains. "Contributed" means any length of X; with a floor we chose of 7 cM per segment, the man's eight-generation average at 180 cM falls to about six. Crossovers follow the Housworth-Stahl interference model in the ped-sim simulator of Caballero and colleagues (2019), with the X values in its parameter file. Without interference the man keeps about seven or eight of his 34. The female X map length is our assumption: we could not open a published figure, so we ran 170 to 200 cM, and at every setting all 34 of the man's eighth-generation X ancestors stayed more likely than not to have given nothing.
View author profile → · Editorial policy
