A centimorgan (cM) is a unit that measures how much DNA you share with a genetic match, and it is the single most useful number for estimating how closely two people are related. One centimorgan corresponds to a 1% chance that a segment of DNA will be split apart by recombination during a single generation, but the practical takeaway for anyone looking at a list of DNA matches is straightforward: the higher the centimorgan total, the closer the relationship.1ScienceDirect. Centimorgan Where things get interesting, and occasionally frustrating, is in the wide range of relationships a single centimorgan number can represent.
How Centimorgans Translate to Relationships
When a DNA testing company reports that you share a certain number of centimorgans with a match, it is summing up all the segments of your genome that appear to be inherited from a common ancestor. The total autosomal genome is roughly 3,400 cM when you account for both copies of each chromosome. A parent and child will always share close to half of that, around 3,400 cM, because inheritance from parent to child is not random in the way that more distant relationships are. Full siblings share about 2,500 cM on average, though this figure can swing considerably. Half-siblings tend to cluster around 1,750 cM. First cousins typically share somewhere in the range of 850 cM, second cousins around 200 cM, and third cousins roughly 50 to 75 cM.
Beyond third cousins, the numbers drop low enough that many matches fall below what testing companies can reliably detect. A fourth cousin might share only 20 cM or so, and fifth cousins often share nothing detectable at all, even though they do have a common ancestor. This is not a failure of the test. It reflects the fact that DNA is shuffled and halved with every generation, and by the time you reach distant cousins, luck determines whether any recognizable chunk of shared DNA survived the journey.
Why the Same Centimorgan Number Can Point to Different Relationships
One of the most common sources of confusion in genetic genealogy is discovering that a match of, say, 900 cM could be a half-sibling, a grandparent, an aunt, or even a first cousin. The overlap between relationship categories is real and sometimes substantial. This happens because the number of centimorgans shared between two people is determined by a stochastic process: the number of DNA crossovers on each chromosome during reproduction varies from one conception to the next. As a result, even fairly close relatives can land on centimorgan totals that are ambiguous about the actual number of generations separating them.2arXiv. Likelihood Models for Forensic Genealogy
In practice, testing companies handle this by comparing your centimorgan score against distributions built from self-reported relationships in their databases. Each centimorgan total gets assigned a probability for being at a certain number of meioses (cell divisions separating you and the match), with the probabilities for all possible distances adding up to one.2arXiv. Likelihood Models for Forensic Genealogy The result is a best guess, not a certainty. When your match falls in a zone where two or three relationship types overlap, the company will often list multiple possibilities and leave you to narrow things down with additional information, such as the ages of the people involved, the number of shared segments, or known family tree data.
How Shared DNA Drops With Each Generation
The basic math behind DNA sharing is appealingly simple. Because each child inherits half of each parent’s genome, the expected amount of DNA that two relatives share drops by half for every generation separating them.3PubMed Central. The rate of identical-by-descent segment sharing between close and distant relatives Your parent shares 50% of your DNA. A grandparent shares about 25%. A great-grandparent about 12.5%, and so on. This halving pattern is why a first cousin (separated by four meioses) shares roughly 12.5% of your genome, and why a third cousin (separated by eight meioses) shares so little that detecting it requires long, unbroken segments of matching DNA.
The word “expected” does real work here, though. The halving is an average, not a guarantee. Your actual sharing with a given relative can be noticeably higher or lower than the expected value, especially as you move to more distant relationships. With a parent the variance is essentially zero. With a first cousin, there is some wobble. By the time you reach third or fourth cousins, the variance becomes enormous relative to the expected total. Two third cousins might share 100 cM, while another pair shares nothing at all. Both outcomes are perfectly normal.
Why Siblings Do Not Share Exactly the Same Amount
If you and a sibling have both taken a DNA test, you may have noticed that your match lists look quite different, even though you have the same parents. This is not a mistake. Each child receives a freshly shuffled version of each parent’s genome, assembled through recombination during the formation of eggs and sperm. The specific crossover points differ every time, so each sibling ends up with a different patchwork of grandparental DNA.4PubMed Central. Sibling variation in polygenic traits and DNA recombination mapping with UK Biobank and IVF family data
This is why full siblings share about 50% of their DNA on average, but the actual number for any given pair can range from roughly 38% to 61%. It also explains why you might be a strong centimorgan match to a cousin on your mother’s side while your sibling barely registers as a match to the same person. You happened to inherit more of the relevant segments from your shared grandparent; your sibling happened to inherit more from the other grandparent. Neither of you is “more related” to the cousin in a genealogical sense, but your DNA tells a different story depending on which chunks survived the shuffle.
This variation has practical consequences for genetic genealogy. If you are trying to identify an unknown ancestor or confirm a suspected relationship, testing multiple siblings gives you a better picture than testing just one person. Each sibling carries a somewhat different sample of the family’s DNA, so collectively they cover more of the ancestral genome than any one of them does alone.
Women and Men Shuffle DNA Differently
Recombination does not happen at the same rate in eggs and sperm. The female genetic map is substantially longer than the male map, meaning more crossover events happen during the formation of eggs than during the formation of sperm. One comprehensive mapping study found that the total female genetic map was about 44 morgans (4,400 cM) while the male map was about 27 morgans (2,700 cM), a ratio of roughly 1.6 to 1.5PubMed Central. Comprehensive human genetic maps: individual and sex-specific variation in recombination That difference is not uniform across the genome either. The ratio of female-to-male recombination peaks near the centers of chromosomes and varies from one chromosome to another.6PubMed. Using lod scores to detect sex differences in male-female recombination fractions
For people doing genetic genealogy, this means that DNA inherited through a line of women tends to be broken into smaller segments more quickly than DNA inherited through a line of men. A segment that has passed through several generations of mothers may be shorter, and harder for algorithms to detect, than a segment of the same genealogical age that has passed through fathers. It also means that two relatives connected through maternal lines might share a slightly higher total cM than theory predicts based on the sex-averaged map, because the female map is longer. This effect is modest for close relatives but can shift things enough at the second- and third-cousin level to affect which relationship category your match falls into.
The X Chromosome Follows Its Own Rules
Most centimorgan totals reported by testing companies come from the 22 pairs of autosomes, the non-sex chromosomes. The X chromosome is typically reported separately, and for good reason: it has a fundamentally different inheritance pattern. Men have one X chromosome (inherited from their mother) and one Y chromosome (from their father). Women have two X chromosomes, one from each parent. This means a father passes his X chromosome to all of his daughters but none of his sons, and a mother passes one of her two X chromosomes to each child regardless of sex.
The consequence for DNA matching is that X-chromosome sharing follows an uneven path through a family tree. Some ancestors contribute to your X chromosome and some do not, depending on the specific chain of male and female links between you and that ancestor. Your father’s father, for instance, contributed nothing to your X chromosome regardless of your sex, because your father’s X came entirely from his mother. Researchers studying X-chromosomal relationships have found that realized sharing on the X shows high deviations from what you would expect based on pedigree alone, and that individuals are frequently identical along the entire chromosome.7PubMed Central. Theoretical and empirical comparisons of expected and realized relationships for the X-chromosome In practical terms, a large X-chromosome match can help confirm or rule out certain relationship paths. If two people share a significant X match, the connection cannot run through a father-to-son link, because no X DNA passes that way.
What “Identical by Descent” Actually Means for Your Matches
When testing companies report that you share a segment of DNA with someone, they are identifying segments that appear to be identical by descent, meaning both of you inherited that stretch of DNA from the same ancestor. This is different from being identical by state, where two people happen to have matching DNA at the same spot simply because that version is common in the population. If a particular gene variant is carried by 80% of people, you will match almost everyone at that location even if your most recent common ancestor lived thousands of years ago.
The algorithms that detect matches have to distinguish between these two scenarios. They do it primarily by looking at segment length. Long, unbroken stretches of shared DNA are almost certainly inherited from a recent common ancestor, because recombination would have broken them apart over many generations. Short matching stretches are more ambiguous. Below about 7 cM, it becomes difficult to tell whether a shared segment represents a genuine recent relationship or just a common population-level pattern.8PubMed Central. Identity by descent: variation in meiosis, across genomes, and in populations Most testing companies set a minimum threshold for reporting matches, often around 7 or 8 cM for a single segment, to reduce false positives. This is also why very distant matches (fifth cousins and beyond) rarely show up: the surviving shared segments have been whittled down to sizes that fall below the detection threshold.
If you enjoy triangulation, where you check whether three or more people all share the same segment on the same chromosome, the length of shared segments in centimorgans is the key measurement. A triangulated segment above 15 cM shared among three or more people who all descend from the same ancestor is strong evidence of a genuine genealogical connection. Smaller triangulated segments can still be meaningful, but the confidence drops as the segments get shorter.
How Forensic Genealogy Puts Centimorgans to Work
The same centimorgan framework that consumer testing companies use has become a powerful tool in criminal investigations. Investigative genetic genealogy, the technique that famously helped identify the Golden State Killer in 2018, works by uploading a DNA profile from crime-scene evidence to a public database and then measuring how many centimorgans the unknown profile shares with each match. The similarity score reported is the total centimorgan length of autosomal matches, which approximates the fraction of the autosomal genome that is identical by descent.2arXiv. Likelihood Models for Forensic Genealogy
Forensic genealogists face the same ambiguity problem that consumer users do, just with higher stakes. A match of 400 cM could be a second cousin or a first cousin once removed, and the investigation has to explore both possibilities. What makes forensic work distinctive is the way analysts combine centimorgan data with traditional genealogical records to build family trees from the matches outward, eventually converging on a suspect. The accuracy of the centimorgan measurement is critical at every step, and the inherent randomness of recombination means that probability models, not hard cutoffs, drive the analysis. Analysts assign each centimorgan score a probability distribution across possible relationship distances, then use those distributions to prioritize which branches of the family tree to investigate.
Why One Centimorgan Does Not Equal a Fixed Number of Base Pairs
A common misconception is that centimorgans map neatly onto a fixed physical length of DNA. The genome-wide average is roughly one centimorgan per megabase (one million base pairs), but this average hides enormous variation. Recombination is not evenly distributed across the genome. Some regions are hotspots where crossovers happen frequently, making a short physical stretch correspond to many centimorgans. Other regions, particularly near the centers of chromosomes, are recombination deserts where a long physical stretch registers as very few centimorgans.
Chromosome size matters too. Smaller chromosomes experience proportionally more recombination than larger ones. A statistical model of this relationship across multiple species found that the genetic length of a chromosome is not simply proportional to its physical length. Instead, there is an excess of recombination on smaller chromosomes, meaning each megabase on a short chromosome translates to more centimorgans than the same physical length on a long chromosome.9PubMed Central. Two-parameter characterization of chromosome-scale recombination rate One physical mapping study of a region on chromosome 3 found that 31 centimorgans corresponded to over 20 megabases in that particular stretch, a ratio of about 0.65 megabases per centimorgan, which is below the genome-wide average.10Proceedings of the National Academy of Sciences. High-resolution physical mapping by combined Alu-hybridization/PCR screening
For genetic genealogy, this means that a 20 cM segment on one chromosome might span a very different physical stretch of DNA than a 20 cM segment on another chromosome. The centimorgan measurement tells you about the probability of recombination, not about the number of genes or base pairs in the segment. Two segments of equal centimorgan length can contain very different amounts of genetic information. This is an important distinction if you are trying to understand what a shared segment might mean in terms of shared traits or medical risk. A large centimorgan match on a recombination hotspot might contain fewer genes than a smaller match in a recombination-cold region.
When Small Matches Are Worth Investigating and When They Are Not
The question of whether a small centimorgan match is meaningful comes up constantly in genetic genealogy forums, and the honest answer depends on context. A single shared segment of 10 cM between you and a match is plausibly a real connection, but it could also be a false positive, especially in populations with a history of endogamy (marriage within a community), where many people share numerous small segments without being closely related in the genealogical sense. Ashkenazi Jewish, French Canadian, and certain Polynesian communities are well-known examples where small-segment matches should be interpreted cautiously.
A reasonable rule of thumb: if your total shared cM with a match is above about 90 cM, the relationship is almost certainly real and recent enough to appear in a well-researched family tree. Between 20 and 90 cM, the match is probably genuine but could represent anything from a third cousin to a more distant connection. Below 20 cM, you are in uncertain territory. The match might be real, or it might be a coincidental overlap of common population-level DNA. If the match consists of a single small segment rather than multiple segments, be extra cautious.
Multiple small segments that add up to a modest total deserve particular skepticism. A match showing 30 cM spread across six tiny segments is less convincing than a match showing 30 cM in a single unbroken stretch. The single long segment is much harder to explain by random population-level matching and almost certainly reflects a genuine shared ancestor within the last several generations. The cluster of tiny segments is more likely to be noise, particularly in endogamous populations.
What Testing Companies Do Not Always Make Clear
Different companies use slightly different algorithms for detecting shared segments, setting minimum thresholds, and calculating total shared centimorgans. This means that if you upload your raw data to multiple platforms, you may see different cM totals for the same match. The differences are usually small for close relatives but can be meaningful at the third-cousin level and beyond, where detection is already marginal. One platform might report a match at 25 cM while another does not report it at all, simply because the second platform uses a more conservative threshold for calling a segment as shared.
The cM figure you see also depends on how the company handles phasing, the process of figuring out which alleles sit on the same copy of a chromosome versus the other copy. Without good phasing, algorithms can mistakenly stitch together segments from both copies of a chromosome, creating artificially long matches. Companies that use parent-child trios or statistical phasing against large reference panels tend to produce more accurate segment calls. If you are comparing results across platforms, keep in mind that some of the discrepancy may reflect phasing quality rather than genuine differences in shared DNA.
Companies also differ in how they handle the X chromosome. Some include it in the total cM count, while others report it separately or not at all. If you are comparing a match’s total across platforms and the numbers seem off by a few dozen centimorgans, check whether one platform is including X-chromosome sharing and the other is not. For relationship prediction purposes, it usually makes sense to evaluate autosomal and X sharing independently, since as noted earlier, the X follows inheritance rules that can make its sharing misleadingly high or low depending on the specific path through the family tree.