If youve read 'health' news on the internet this week, youve seen this story:
Just put 'MHC women' into Google News. Youll get a billion hits for articles talking about this paper:
Ugh. Guys, at this point in time, research connecting MHC genotype with 'compatible' mates in humans is good old fashioned woo.
MHC Class I molecules are on the surface of all of your cells. Proteins that your cells make eventually get degraded, and MHC molecules present little bits of these proteins at the cell surface. Its a signal to cytotoxic T-cells that everything is fine.
When your cells are infected with a virus, or have become cancerous, its MHCs start presenting proteins that arent 'normal'-- they become a signal to CTLs that something is wrong, and the cell needs to be killed.
MHC Class II molecules are only on the surface of antigen presenting cells (cells that eat bacteria/viruses/etc). So while MHC I present internal proteins, MHC II present proteins that have been chewed up from the external environment.
While you and I both have MHC I and II, there are lots different genotypes. Different variations of MHCs will present different bits of chewed up proteins. If you and I are both infected with salmonella, you might present Protein Chunk A in your MHC II, and I might present Protein Chunk B.
Theoretically, this means that we would be good 'partners'. Because our MHCs are different, if we mated, our offspring might be more likely to survive a salmonella plague than either one of us individually.
A bump in the road is the fact we have no idea how I would have any idea what your MHC genotype is. We type people today for transplants/donations through a battery of assays-- how the hell could you and I, on a date, know our MHCs are different, and our offspring would be 'fit'?
The most popular explanation is that we can smell each others MHC type (other mammals might, why not humans too?). Ive read a few of the papers that explore this possibility because I know someone who works for an MHC-match dating service-- That we could 'sense' another individuals MHC genotype was first publicized in a 1995 study, 'MHC-Dependent Mate Preferences in Humans'. Well Ive read that paper, and its garbage. Its just a friggen crap shoot whether a woman picks a 'similar' or 'dissimilar' MHC type. All of their error bars overlap.
Subsequent papers havent been much better. Some not just saying that we have ways of 'smelling' someones MHC, but that we can tell just by looking at someone what their MHC type is.
Now we have this new paper, and 'science reporters' are bleating 'OMG! WOMEN ON THE PILL CHOSE THE WRONG MHC TYPE!!'
... Except in this new paper all the error bars on their graphs overlap between 'similar' and 'dissimilar' choices, for women on and off The Pill!!
(a) odour pleasantness ratings, (b) odour intensity and (c) odour desirability.
I feel like Im on Wacky Pills! Why the hell is anyone taking this 'research' seriously?? I mean, gee, just think of the money and time we could save doing organ transplants:
Transplant Surgeon: "Theres no time to use PCR to determine this motorcycle crash victims MHC type! I can tell by SMELLING HIM hes compatible with my patient! Take his kidneys!"
WTF. Stupid crappy 'science reporting' and stupid crappy research. Ugh.
Heh. I've heard about this in the past, but didn't realize exactly how bad the science was. Thanks for the update.
That said, I've been reading too many physics artcles lately, because the first thing that popped into my head when I saw "MHC" was "Medium Hadron Collider".
What people perhaps don't recognize is that the "MHC" region harbours a lot
of olfactory genes, so any correlation people find with a certain MHC ABC DR type
could simply well be specific olfactory genes, hitching a ride with a certain
At the risk of outing myself as a serious layman, can anyone point towards something that can teach me how to read those graphs? I have a feeling a certain friend of mine is going to be busting out this little tidbit as fact and I would love to be able to point her towards this post, but I need to understand it first. I know she will either want me to explain it or claim that she understands it and it proves her right if she thinks I don't understand it myself (uhm.. which I don't).
I think I heard something along these lines ages ago, so thank you for debunking it (as usual). Though, I have a vague recollection that the sniffing was supposed to be related to interrelatedness* - which would go in the other direction.
On the upside, though. At least they include the errorbars!
* roughly, we're supposedly more attracted to relatives (but not too close relatives), since if we share genes with them, there's a bigger chance that our offspring will get our genes, since they're theirs too. ... Okay ... perhaps I shouldn't have tried to write that impersonally ...
Ugh. Guys, at this point in time, research connecting MHC genotype with 'compatible' mates in humans is good old fashioned woo.
I've thought so for years but now, thanks to you, I have the data to back my position up. Bookmarked.
Also, I've noticed that most people smell differently on different days. I'd imagine that smells have more to do with hormones than anything. If someone has a paper confirming or debunking this suspicion, I'd like to see it.
Well Ive read that paper, and its garbage. Its just a friggen crap shoot whether a woman picks a 'similar' or 'dissimilar' MHC type. All of their error bars overlap.
I don't think that's right, since they (the -95 paper) used the Wilcoxon signed-rank test, which is the nonparametric equivalent of a paired t-test, and you can't get the significance of that from the error bars. The test might still be crap for other reasons but I think you're too quick to dismiss the results.
In any case I think it's better to focus on the stupid conclusions - so what if humans happen to have a preference for MHC-dissimilar mates, the hand-wringing over the pill is ridiculous! How about perfume, or makeup, or Viagra, don't they disrupt natural mate choice patterns? Aargh!
Wow, just wow, any time error bars overlap THAT much, you can't really draw a conclusion from it. The only one that doesn't have overlapping error bars is C-Pill (probably due to error or anomalous data). This type of research makes me very, very angry.
Jared and others, please read my comment before jumping on the "error bars" bandwagon. I don't have time right now to go through the test in more detail, but it may be premature to dismiss it.
In the 1995 study, N was 49, and in the one being hyped now, N was 100. It's been too long since I've had to do nonparametric statistics in any significant depth, so I won't give an intuitive judgment on how much samples that size can tell you. . . but damn, those numbers are small.
windy-- Id love some help with the stats, cause Im totally confused as to why they are making such strong conclusions from (what I see as) really weak data! Want me to email you the paper?
Thanks, I have access to the papers online - but the 1995 study doesn't have much detail about the tests. I agree that the 2008 study is a bit weak but not for reasons of the error bars. They make about a zillion different statistical tests, should there be a correction?
Couple of interesting bits from the results:
"Our results therefore suggest that, at least in our sample, there was neither a significant general preference for MHC dissimilarity across normally cycling women, nor a significant preference for MHC similarity associated with pill use."
So this study failed to replicate the result of most previous studies of MHC preference - you don't see that hyped in the press! So what's the fuss all about?
"significant session-group interaction, whereby ratings shift in favour of MHC similarity after initiating pill use, in contrast to the control group"
So it's the change in desirability ratings in the group that started pill use, I think this is in your figure (c), the two bars on the right (white before pills, grey after pills). If you grant them the multiple comparisons, there's a significant change there.
But- compare with the "control" group on the left! The difference between the two groups of women was greater from the start than the change from pill use! They discuss possible reasons in the article, but three guesses why everyone presents this as "contraceptives DISRUPT mate choice preferences in women" when based on the data you could just as easily say that "contraceptives NORMALIZE mate choice preferences in women"! Hah!
You're right Windy, it's not just the error bars, it's the sample size, error bars, and the links they were drawing with absolutely NO means of a signal. But I was looking at the consistency of changes between the two groups. I think fMRI studies would be a good test to put a final nail in the "MHC love connection" papers.
Please e-mail me a copy of the paper in either pdf or word format. I have a background in biostatistics and would be interested in seeing how they justify their conclusions.
Interestingly, mice don't actually smell for MHC alleles, but something else to determine relatedness.
I was hoping that science would vindicate this idea because I really wanted a t-shirt with one of the following:
"I think your Major Histocompatibility Complex is hott."
"I want to have sex with your Major Histocompatibility Complex."
"I dig your MHC."
I think those would go over really well at bars, and hopefully, if all goes well, the people who find the shirt entertaining will actually be my "histo"-compatible partners :)
Hi Blake and windy,
Assuming that they used equal sample sizes for those women on and off the pill - and that they did not stratify the sample size further (e. g. to take in account for age differences such as between twenty-something and thirty-something year olds or to account for ethnic differences like white vs. black Britons), then the resulting statistical analyses and conclusions should be correct. They had more than enough of a sample size to make any assertions that they were dealing with a sample size which was normally distributed for those on and off the pill.
"What people perhaps don't recognize is that the "MHC" region harbours a lot of olfactory genes, so any correlation people find with a certain MHC ABC DR type could simply well be specific olfactory genes, hitching a ride with a certain
That's interesting. Because could that not be an argument made /for/ hypothesizing a link between MHC haplotype and mate preference? Forgive me - maybe I'm thinking of this all wrong - but if your specific MHC haplotype were linked to an olfactory receptor, this would be an important component of the mechanism for your being able to select a mate with an MHC different from yours.
I share ERV's skepticism regarding these findings, but to judge their veracity by whether "the error bars overlap" is ludicrous, especially in a post that attacks peer-reviewed scientific research as "woo." Just the tiniest amount of knowledge about statistical analysis would have been useful here.
"Strong conclusions from weak data" -- that's solid scientific criticism. "The error bars overlap" -- that's ignorant claptrap. Come on, people. This is embarrassing.
Steve-- Fine, ignore the error bars. The Pill group looks exactly like the Control group, with a few minor variations that apparently are inherent in their scoring scale (ie why does the pre-Pill group look different from the Control group? Why is there any variation in the Control group pre/post-non-Pill?).
Its the same kind of results you see with intercessory prayer studies (also peer reviewed) and fourth grade science projects on ESP. Its crap.
Ive got a post on why this is woo from a genetic/physiological aspect for later, but this was a quick post for damage control (the mass postings from pop news sources).
After rereading the paper, I want one of these naysayers to explain to me why the hell my complaints about the stats are wrong. Seriously:
However, with the core sample, we found a significant session-group interaction (F3,71=3.05, p=0.034), driven mainly by desirability ratings (F1,73=3.63, p=0.061; pleasantness F1,73=0.22, p=0.64; intensity F1,73=0.01, p=0.92).
I havent had a stats course in six years, but their own analysis says their results are insignificant except the 'desirability' rating, kinda.
Why am I wrong?
Your skepticism is well taken. I wouldn't even say that the desirability ratings is statistically significant (The P value would have to be at least 0.05 or lower for significance.).
Abbie-- There was only one problem with your diatribe: you critiqued published results based (in part) on "overlap of error bars." That's silly. Neither statistical significance nor scientific significance can be meaningfully judged by "overlap of error bars." That's all. I look forward to your comments on the physiology; I'm sure you and I have the same grounds for skepticism.
But let's keep our skepticism under scientific control. Hard-nosed skepticism is a scientific virtue, but is easily confused with wild-eyed incredulity, which is a virtue in certain Seattle neighborhoods but is the antithesis of sound scientific reasoning. To criticize overinterpretation of equivocal data is scientifically praiseworthy; to label as "woo" a conclusion that just seems implausible or is thus far unexplained is scientific malpractice. A word to the wise.
Now back to work.
I'll suggest another grounds for skepticism that doesn't draw so much on statistics (I'm a neophyte in that arena). There is zero conceptual work done to connect MHC to normative sociocultural categories like "better" or "mate" and some ridiculous assumptions about what might count as successful mating in Homo sapiens. Are we really that stupid as to believe that MHC determines successful mating practices because of a t-shirt test that says nothing about the actual, real-world preferences, desires and outcomes of the people involved?
Perhaps the single dumbest claim in the article is this sentence from the abstract: "Women using oral hormonal contraceptives have been reported to have the opposite preference, raising the possibility that oral contraceptives alter female preference towards MHC similarity, with possible fertility costs." OH NOES?!? For real? Women taking the birth control pill suffer from a possible fertility cost? Maybe that's because they don't want a friggin baby. Get these dudes a Nobel, pronto.
The only thing that this study proves is that Evolutionary Psychologists continue to suffer from their profound anxiety about any form of female sexuality that doesn't match their imagination of a Neolithic patriarchy. You don't need a statistics degree to figure that out. Take a sociology of gender or philosophy of biology course once in awhile and maybe we wouldn't have to argue about p-values of experiments that have zero conceptual value in the first place.
"Its the same kind of results you see with intercessory prayer studies (also peer reviewed) and fourth grade science projects on ESP. "
The publishers of the study know who their target audience is.
And there's a non-zero, but extremely small, probability that any members of that target audience are going to log onto SB to read the rebuttals.
And they know that too.
One thing I didn't like were the wild extrapolations from this. 'The pill makes users pick bad mates' makes no sense when you consider that most women on the pill are already in a relationship and want to stay in that relationship without having babies.
*sigh* so crazy. And who gives a damn if the pill changes a womans partner preference anyway? If shes happy and hes happy (or shes happy and shes happy, I just assume woman/woman couples wouldn't need to use the pill, but may be wrong) what does it matter?
I give a damn! I started dating my girlfriend two years ago when she was on the pill. Now I have to decide whether to marry her and take her with me wherever I get stationed for the Army (which is already difficult enough for certain reasons I won't go into) and just blindly guess as to whether these studies actually mean anything. I mean, even if it's only a chance that it's true, I'm going to have to go overseas on deployment in about a year with the knowledge right now that she may lose interest in me and write me a Dear John letter all because she was on a stupid pill when she met me. I mean I could have dated lots of other girls but I liked her attitude and personality the best. I trusted her more than any other girls and I felt like she was a girl I could possibly leave to go into combat without worrying about her love and fidelity (something that can really affect job performance and the lives of my subordinates). But I finally got her to go off the pill about a month ago and already she picks fights more often with me and she is less loving. Yeah sure, you guys could say its some kind of psycological placebo phenomenon but I really am not so sure. And I have to decide whether to marry her soon. So yeah, it is important to the male gender which gets ignorantly accused in this post as hell-bent on reinstating patriarchal beliefs. I don't know why some of the women in this post got so offended by a little bit of criticism about the pill. The criticism isn't set to bring down women, it smartly points a skeptical finger at pills that radically alter a human being's body chemistry. It's not as dangerous as anabolic steroids as far as we know but going against nature always carries a lot of risks.
There is evidence in numerous species that genes involved in immunity influence mate choice. Factors like body odor may subconsciously favor partners with different immunity alleles, to avoid inbreeding and/or endow offspring with broad resistance to pathogens. A previous study, based on HapMap genotypes, reported that European-American mates were extremely dissimilar from each other in immunity alleles compared to non-mates. Upon re-examining the results and methods, and visually comparing mates and non-mates, we found that this effect was weak, strongly dependent on extreme pairs and on arbitrary choices in methodology, and not significant after correcting for the multiple hypotheses tested. More importantly, examination of new couples from the same population did not support this hypothesis.
Derti A, Cenik C, Kraft P, Roth FP (2010) Absence of Evidence for MHCâDependent Mate Selection within HapMap Populations. PLoS Genet 6(4): e1000925. doi:10.1371/journal.pgen.1000925