A paternity test looks trivial from the outside. Someone rubs a swab inside a cheek, drops it in an envelope, and a few days later a PDF arrives with a number on it. The science underneath is not complicated either, but almost nobody explains it, which leaves people unsure whether to trust the result. Here is the whole process, start to finish.
Step 1: the samples
Nearly every paternity test in the United States now uses a buccal swab - a soft-tipped stick rubbed against the inside of the cheek for about 30 seconds. The swab picks up cheek cells, and cheek cells contain the same DNA as every other nucleated cell in your body. Blood is not more accurate. It is just messier, more expensive, and needs a phlebotomist.
The standard test compares the child and the alleged father. The mother's sample is optional and usually unnecessary, though including it can sharpen the statistics in unusual cases.
Samples that labs can work with when a swab is not possible include blood, and in some cases nail clippings, hair with the root attached, or a used toothbrush. These "discreet" samples cost more, fail more often, and cannot be used for anything legal, because nobody can prove whose they are.
Step 2: extraction and quantification
At the lab, chemical reagents break the cells open and separate DNA from everything else - proteins, membranes, the cotton of the swab itself. The result is a small volume of purified DNA in solution.
The lab then measures how much human DNA it has. This matters: too little and the analysis produces incomplete results; too much and the signal saturates. If the quantity is too low, the lab requests a new sample rather than guessing. That is why a badly collected home kit sometimes comes back as "insufficient sample" instead of a wrong answer.
Step 3: amplifying the markers
DNA is copied - amplified - using the polymerase chain reaction, which makes millions of copies of specific short regions. Those regions are the interesting part. Human DNA contains stretches called short tandem repeats, or STRs, where a short sequence such as AGAT repeats a variable number of times. One person might have 11 repeats at a given location, another 14.
Those repeat counts are inherited. At each STR location you carry two values, one from your biological mother and one from your biological father. Labs test 20 or more of these locations at once. NIST maintains STRBase, the reference database describing the markers used in human identity testing, and MedlinePlus explains the underlying structure of DNA for anyone who wants the biology first.
Step 4: reading the results
The amplified fragments are separated by size using capillary electrophoresis, which produces a chart of peaks. Each peak is one repeat count at one location. The output for a person is a table like this:
| Marker | Child | Mother | Alleged father |
|---|---|---|---|
| D3S1358 | 15, 17 | 15, 16 | 17, 18 |
| vWA | 16, 19 | 14, 16 | 19, 19 |
| D8S1179 | 12, 13 | 12, 14 | 10, 13 |
Read the first row. The child has 15 and 17. The mother contributed the 15. So the 17 must have come from the biological father - the "obligate paternal allele". The alleged father has 17 and 18, so he could have supplied it. Same story in the other two rows.
Do that across 20-plus markers. If the alleged father lacks the obligate paternal allele at two or more markers, he is excluded, and the report reads 0% probability of paternity. There is no ambiguity about an exclusion; it is effectively absolute.
Step 5: the probability calculation
An inclusion is a statistical statement rather than a flat yes. If a man matches at every marker, the lab asks a follow-up question: how common is that combination of values in the general population?
For each marker the lab calculates a paternity index - roughly, how many times more likely it is that this man contributed the allele than that a random unrelated man did. A rare allele produces a large index; a common one produces a small index. Multiply the indices across all markers and you get the combined paternity index, often in the hundreds of thousands or millions.
That combined index is then converted into the number on the front page: the probability of paternity, typically reported as 99.99% or higher. A result of 99.99% means the tested man is roughly 10,000 times more likely to be the father than a random unrelated man from the same population group.
Why reports never say 100%
An exclusion is certain. An inclusion is not, and no reputable lab will pretend otherwise. The reason is simple: the test compares the man in front of it against a population, and it cannot rule out an untested close relative who shares much of the same DNA. An identical twin is indistinguishable at STR markers. A brother or a father shares far more than a stranger would.
This is why labs ask, on the intake form, whether any close relative of the alleged father could also be the father. If the answer is yes, test them too. Otherwise you can get two brothers who both come back at 99.9%.
What can go wrong
- Bad collection. A swab that is wet, mouldy, or contaminated with food or another person's saliva produces a failed or ambiguous profile. Labs usually detect this and ask for a recollection.
- Mutation. Occasionally a repeat count changes between parent and child, producing a single mismatch in an otherwise matching profile. Labs handle this by testing extra markers, not by calling an exclusion on one mismatch.
- Sample swap or substitution. This is the failure mode chain of custody exists to prevent, and the reason a home test cannot be used in court.
- Recent transfusion or transplant. A bone marrow transplant means blood carries donor DNA; labs will use a buccal swab instead and should be told beforehand.
Turnaround, in practice
The bench work takes hours, not days. What sets the calendar is shipping and queueing. A realistic timeline is one to two days in transit each way, one to three business days at the lab, plus a review step. Most people see results within a week of mailing a kit, and within two or three days of an in-person collection.
Does the type of test change the science?
No, and this is the single most useful thing to understand. A legal, court-admissible test and a $99 kit from a pharmacy shelf are often run by the same accredited laboratory on the same instruments using the same markers. The difference is entirely in how the sample got there: who watched, who checked the ID, and who signed for it. For background on the consumer end of genetic testing generally, MedlinePlus covers direct-to-consumer testing.
If you want a full comparison of the two, see our guide to legal versus at-home paternity tests, or start with Paternity Testing 101.