Friday, January 23, 2004

Pop quiz: What is the rate of mistaken paternity?

Everyone knows that men can never really know that their children are biologically their own, short of a DNA test. But most people think that, well, cases of mistaken paternity are rare freak cases, and the people we know couldn't possibly be cases of mistaken paternity; and most men figure that when they have kids, of course the woman they choose will be trustworthy and faithful, so it's a problem for freak-show losers to worry about.


While reading Gigerenzer's Law of Indispensable Ignorance on's Edge Annual Question (this year, "What's Your Law?"), I came across the startling sentence:

The estimated 5 to 10% of children and their fathers who falsely believe that they are related might not lead a happier life by becoming less ignorant; knowledge can destroy families.

"Holy crap! 5-10%? That's got to be a mistake," I thought. Is it? Like any lazy researcher, I did a quick Google search, which turns up a thread from the sci.anthropology newsgroup archives with the rather insensitive, albeit humorous, subject "How many bastards are there, anyway?". At first, it seems, we can breathe a sigh of relief, as Lee Rudolph writes:

Yesterday I was talking to a bioethicist who studies genetic counseling, and when the conversation turned to folklore, she brought up an issue she'd recently investigated. She told me that essentially all genetic counselors believe that there is a "false paternity" rate of about 5 percent--yet few if any counselors encounter a rate nearly that high in their own practices. This discrepancy having piqued her curiosity, my friend had tried to track down just why the rate is believed to be 5 percent. One source after another said the scientific equivalent of "it was published by a friend of a friend". Eventually the FOAF-chain terminated, in a single paper whose author does not think it supports the interpretation it is popularly given! Appendix 1, immediately below, represents the present state of her research into this possible piece of genetic folklore.

... [skipping to appendix] ...

A historical search and a conversation with Dr. James Neel, one of the deans of human genetics, led us to what may be the source for the high estimate: a 1962-65 study of blood typing in a small Michigan town (1). That study found discrepancies between biological and stated parentage in 109 of 2507 nuclear familes. Many of these may have been unacknowledged adoptions, including step-parent adoptions.

...[skipping to cite]...

(1) Sing, CF et al (1971) Studies on genetic selection in a completely ascertained Caucasian population II. Family analysis of 11 blood group systems. American Journal of Human Genetics 23(2) 164-198.

Whew! OK, so it's one small study, and it may not have counted adoptions properly. But hold on --- a few messages later in the thread, Joe Quellen writes:

The statistics are covered in a very engrossing book by Jared DIamond, called *The Third Chimpanzee*, 1992, Harper Perennial, Chapter 4, The Science of Adultery. He describes a study of blood typing and genetics which had unexected results and was quashed. It was done in the 1940s at a "highly respectable" US hospital. The study found that fully 10 percent of babies were not the biological offspring of their legal fathers.

Oh. Diamond was citing a different study. A couple of messages later, "sgf" writes:

I just picked up Timothy Taylor's _The Prehistory of Sex: Four Million Years of Human Sexual Culture_ (Bantam: 1996). Here's his summarization of the subject:

(pp 77-79)
"...In tests of genetic paternity recently conducted by Robin Baker and Mark Bellis [1], they found that around 10 percent of children had been sired by someone other than their ostensible fathers -- although the fathers consciously believed these children to be their own.

...[skipping to cite]...

[1 Baker, R. and M. Bellis. 1993 "Human sperm competition: ejaculate adjustment by males and the function of masturbation." _Animal Behaviour_ 46: 861-65]

And just to pile on a bit, the next Google hit is a working paper by anthropologist Kermyt G. Anderson titled "How well does paternity confidence match actual paternity? Evidence from worldwide nonpaternity rates" (abstract; DOI) (UPDATE 2010-02-17: fixed linkrot), wherein we learn:

The median nonpaternity rate for the high paternity confidence sample is 1.9% (range: 0.4 - 11.8), while median nonpaternity for the low paternity confidence sample is 30.2% (range: 14.3 - 55.6). The median nonpaternity rates for these two groups are significantly different (Wilcoxon sign-rank test, z = -6.112, p < 0.0001), which is not surprising since the two distributions do not even overlap. Thus, men with high paternity confidence are less likely to be incorrect in their assessment of paternity than men with low paternity confidence. In other words, men with high paternity confidence are more accurate in assessing paternity.

The median nonpaternity of men whose paternity confidence is unknown is 16.7% (range: 2 - 32). This is significantly greater than the high paternity confidence sample (Wilcoxon sign-rank test, z = -4.349, p < 0.0001), and significantly lower than the low paternity confidence sample (Wilcoxon sign- rank test, z = 3.528, p = 0.0004).

When the high and unknown paternity confidence samples are combined, the median nonpaternity is 3.9% (range: 0.4 R 32). This is significantly less than median nonpaternity for men with low paternity confidence (Wilcoxon sign-rank test, z = -6.053, p < 0.0001). Thus, men with low paternity confidence are the least accurate in assessing actual paternity.

In other words, interpreting the results rather cavalierly, if you're fairly confident that the child is yours, it is probably yours (although, interestingly, nearly 1 out of 50 such men are still mistaken). If you've got suspicions, then the probability that it's not yours is nearly one out of three. Intuitively, this makes sense: men don't just doubt for no reason. If you have reason to doubt, then it's probably with some justification.

Anyway, the 5-10% number seems to be a bit of an overestimate; the actual number from the most comprehensive survey seems to be in the 2-4% range. Still disturbingly high, in my opinion.


  1. Interestingly, two groups of people latched onto the 10% figure: a.) misogynists who used it to proclaim the perfidy of the female gender, and b.) ironically, their nemeses the radical feminists who used it to counter the (equally) sexist view that women didn't enjoy sex.

    And I think even people without a political ax to grind clung to it because, as the sister of the late Tsar said about the woman who claimed to be the princess Anastasia even though she was obviously a fake, the public finds it exciting to believe.

  2. OK, wait a minute. How on earth are you concluding after all that that the true rate of non paternity is really 2-4%? This issue is so politically hot it is it not possible to look find the truth, but look what you did find:

    The median nonpaternity of men whose paternity confidence is unknown is 16.7%

    So there you go - 16.7%. That is very disturbing.

    There has been a recent campaign to discredit the previous studies showing 10% by referencing
    Anderson study. The Anderson study puts non-paternity studies of the past in 3 buckets,
    'high', 'low' and 'unknown' confidence. The high bucket, as it turns out is actually the bucket where
    couples were asked to come forth and volunteer to be in a paternity study. Those scenarios would
    tend to keep the cheating women away. The other bucket was scenarios where fathers had enough
    doubt that they wanted a test. It didn't mean that they believed the child not to be theirs, but
    just that they wanted a test to be sure. There is nothing wrong with that and that certainly does
    not put those fathers is some sort of marginalized category. They are just suspicious - as apparently they should be.

    Both the high and the low buckets are a skewed, and predictably they return skewed numbers: 1.7% and >30%.

    What does that tell us about the average rate? Practically nothing. The only thing
    that is a little revealing is the rate where 'paternity confidence' is unknown - ie there was no known
    bias: 16.7%!

    The very fact that *this metastudy* is being used to discredit the 10% number really shows you
    how high it must be. What an act of desperation.. to point to a research project that shows some
    studies to be biased in one direction and others in another to somehow dispel the other studies
    showing a 10% average, especially when the bucket of studies with no discernable or known bias
    had a rate of 16.7%! Also consider how disturbing it is that if a man is at all suspicious
    that a child *might* not be his, enough for him to get a test, then chances are 1 in 3 that in fact he is
    right! That is crazy high. The attempt to spin this that:

    "men with low paternity confidence are the least accurate in assessing actual paternity"

    is crazy nonsense. Those men didn't think they were not the fathers. They just wanted to get
    a test.

    Here's another thing that is extremely suspicious. There are many ways to take a much broader
    non-bias sample and just have the thing known once and for all. What about the overall rate
    of organ transplant donors? What about the overall rate of adoption cases, both of which have
    mandatory DNA tests? Good luck trying to find those numbers.

    I'm not saying that this is any conspiracy of course, but it sounds like a reality that is extremely
    uncomfortable and un-pc to accept. It certainly would not further the cause of feminism.
    So the internet, and particularly wikipedia, has become a battleground of clashing ideologies.

    In my opinion the jury is still out on what the real rate is.