Bellis et al. 2005 — Measuring paternal discrepancy and its public health consequences

What the study found

Bellis and colleagues pooled results from 17 separate studies to estimate how often a child's biological father is not the man assumed to be the father — what the researchers call paternal discrepancy. Across those studies the rate varied widely, but the middle of the distribution was low: a median of about 3.7%. The much higher figures, up toward 30%, did not come from general or unselected populations; they came from samples of men who had specifically sought a disputed-paternity test, where the median was around 27%. ^[1]

vary between studies from 0.8% to 30% (median 3.7%, n = 17)

^[1]

Why it matters here

This review is the source most often mined for the manosphere's paternity-fraud statistic, because its range genuinely reaches 30%. But that top figure is a property of a self-selected group — men who already doubted their paternity — not of fathers in general, for whom the pooled median is under 4%. That distinction is central to the check on whether 10–30% of children are the product of paternity fraud, where the high end of a selection-biased sample is frequently presented as the population-wide rate.