Identical twins aren't *that* identical
A lawsuit about cheating on an exam reveals some weird misconceptions about twins - perpetrated by those who should know better
If you have an interest in American crime stories—or books on free will—you’ll have heard of The Twinkie Defence.
In 1978 the California politician Harvey Milk was assassinated, along with San Francisco Mayor George Moscone, by another politician named Dan White. Despite the fact he admitted to shooting both men multiple times in the head, White was only convicted of manslaughter and jailed for a mere five years. A lot of this was down to his lawyers’ clever strategy during the trial.
Now, there’s the legendary version of the Twinkie Defence and the real version of the Twinkie Defence. The legendary version goes like this:
The lawyers argued that White had been eating a lot of junk food like Twinkies (the American mini sponge cake filled with cream) in the lead-up to the shooting. All the sugar and E-numbers caused him to lose the capacity for rational thought, lose his free will, and thus have diminished responsibility for his actions.
But as the Wikipedia article explains, that’s a common misconception. In fact, the real Twinkie Defence was something like:
The lawyers argued that White was extremely depressed, as indicated by his changing from a highly health-conscious diet to one filled with junk food like Twinkies. Clearly his whole life was going off the rails, he had less free will, and he thus had diminished responsibility for his actions.
So the reality wasn’t quite as silly as the legendary Twinkie Defence, but we could still describe both versions the same way: “an ostensibly scientific—but actually pretty flimsy—excuse to get someone off at a criminal trial”.
Well, as of a couple of days ago, we have a brand-new version of that ostensibly scientific but actually flimsy excuse. This time it isn’t the Twinkie Defence, but…
The Twin-kie Defence
The headline of a new article from Insider.com basically tells the story:
Identical college twins were accused of cheating in an exam by signaling. They won $1.5 million in damages after a jury decided they hadn't cheated because their minds were connected.
The twins, Kayla and Kellie Bingham, have spent the last six years in a defamation suit against the Medical University of South Carolina, who claimed that exam invigilators noticed the twins—who were sitting near to each other as they completed the important medical exam—seemed to be moving in peculiar ways that implied they were communicating with each other.
The university also found that the twins’ scores on the exam were nearly identical: of 307 questions, they gave the same answer 296 times, including 54 times where they gave the same wrong answer. They argued that the twins answered the questions—which were done by computer—at a near-identical pace throughout the exam. A examinations company concluded that the probability the twins took the test independently was one in 100 undecillions (a undecillion is a 1 with 36 zeros after it; 10^36).
But here’s the defence: these calculations were totally off, because they didn’t take into account the fact that the students were identical twins. Identical twins are clones of each other, with 100% of their DNA in common. Shouldn’t we expect them to be vastly more similar to one another than the average person? In fact, shouldn’t expect them to, for all intents and purposes, be exactly the same person?
To back this up, the twins’ lawyers brought in Prof. Nancy Segal, who is a well-known behaviour-geneticist who has made important contributions to the field. In her deposition, Segal didn’t quite argue that the twins’ “minds were connected”, but she made the case that twins are so similar that she would be “surprised if she didn’t see” similarities like that on the exam:
Once people are familiar with what the research shows -- and everybody knows that twins do amazing things alike, which really defies the imagination. But once they understand that, intelligent and more reasonable individuals understand that twin similarities are a function of their genetic propensities and not a function of cheating.
For this reason, Segal argued that the report by the examinations company—called Caveon—was “completely irrelevant”. Here’s part of her exchange with the university’s lawyer:
LAWYER: …and what should Caveon had done that they didn't do?
SEGAL: Well, they should have… factored in some sort of measure of genetic relatedness. I'm not sure how you do that statistically. But given the research on identical twins, both raised apart and together, in terms of similarities in test scores, this was a grievous omission. I mean, when I show these to colleagues, they hardly believe it.
L: Okay. But do you have any specific proposal for what they should have done to account for the fact that they were twins?
S: Well, there should have been some statistical correction there. There should have been a statistical correction. … Now, is there such -- do I know a statistical correction? No, I'm not a statistician.
Well, the twins’ legal team went ahead and made such a correction. They took the Caveon number—1 in 10^38—and corrected it for the similarity of the twins. If, they argued, twins are twice as similar as the average person, the number should be divided by two, giving 1 in 10^19. Still pretty big.
But if the twins are ten times more similar than the average person, you’d do 10^38 divided by 10, which is of course 1 in 10^3.8, which translates as 1 in 6309. That seems much more likely to occur by random chance! Case closed!
There are two problems with this. First, the “ten times” figure is just pulled from thin air - but how similar twins are compared to the average person is at least a debate we could have. The second, much bigger problem is that the calculation is complete, utter nonsense. That’s not how you divide numbers - you don’t divide the exponent, but the number itself! So, 10^38 divided by 10 is not 10^3.8 - it’s 10^37 - still an unbelievably enormous number. Perhaps the legal team weren’t statisticians either.
In a post-trial motion, the university argued that the twins’ lawyers misled the jury with this “astronomically incorrect”, “astonishingly inaccurate” calculation. I don’t know how post-trial motions work in the South Carolina legal system (or, if I’m honest, in any legal system), so I can’t tell you whether this will make any difference to anything.
The university also put forward some counter-arguments to the idea that the performance would be expected to be extremely similar:
The twins’ similarity on other exams across the year varied dramatically - sometimes it was as low as 60% and sometimes as high as 96% - so it wasn’t always near-identical;
The specific pattern of strengths and weaknesses differed between the twins: Kayla did better in exams on the nervous system, for example, whereas Kellie was much better at exams on the endocrine and renal systems. Again, not what you’d expect if they were fundamentally the same person. There was an exception to this pattern, and it was the exam in question.
Segal, for her part, argued that she took into account—along with her general knowledge of the twin-study literature—another important piece of evidence: the twins’ testimony. And they were adamant that they hadn’t cheated. Here’s another section:
LAWYER: Okay. So as I understand it, when you get one of these cases where twins have been accused of cheating, your methodology is to consult -- and you're familiar with the literature. And then if they said they weren't cheating, then your opinion is that they weren't cheating; is that correct?
SEGAL: Yes. But there's more to it than that… I need to know a little bit more about the circumstances. I mean, Kayla, in her initial e-mail to me, went on and on about her relationship with her sister, what the circumstances were. You know, I need to know more about it, because I need to work in an informed fashion.
L: Okay. So do you accept at face value what Kayla told you?
S: I do.
L: Okay. Did you do any investigation to determine whether what she was telling you was true?
S: No.
L: Okay.
Just to be clear: I haven’t come near to reading through all of the voluminous documents relating to this years-long case. I’m not saying that I know these twins cheated on the exam, or even that I suspect it - I have no idea. There could be all sorts of reasons for the similar scores. But I think it’s fair to say we should be highly sceptical that “they’re identical twins” is a scientifically-sound explanation of the similarities in this case.
Jim^2
I wonder if it’s easier to believe that identical twins are literally identical because of some of the famous stories about astounding twin similarities. Some of these come from the kinds of cases that behavioural geneticists themselves often talk about.
The most famous is “The Jim Twins” - identical twins who were separated at 4 weeks old, adopted, and upon being reunited at age 39, found they’d led uncannily similar lives. Their bizarre story is on websites like Ripley’s Believe it or Not, but also mentioned in more serious science reporting, and I’ve seen the Jim Twins referenced many times in textbooks, papers, and lectures on behaviour genetics.
They’re also on a page from the University of Minnesota, which was the site of the original twins-raised-apart study the Jim twins were involved in (Nancy Segal was also heavily involved with this study). Aside from both being called Jim, the site notes that:
“Both twins are married to women named Betty and divorced from women named Linda.
One has named his first son James Alan while the other named his first son James Allan.
Both twins have an adopted brother whose name is Larry.
Both named their pet dog ‘Toy.’
Both had some law-enforcement training and had been a part-time deputy sheriff in Ohio.
Each did poorly in spelling and well in math.
Each did carpentry, mechanical drawing, and block lettering.
Each vacation in Florida in the same three-block-long beach area.
Both twins began suffering from tension headaches at eighteen, gained ten pounds at the same time, and are six feet tall and 180 pounds.”
I don’t think it’s ever explicitly stated that all these similarities are due to their being genetically identical. But the implication is definitely there.
How many of these coincidences do you think are really ultimately caused by genetics? The two most plausible ones, which don’t seem like that much of a coincidence, are in italics above. Physical, health-related, and psychological similarities make perfect sense.
But to be honest, some of the others are such incredible levels of coincidence that they would still be surprising if the twins were 100% genetically identical and had been raised in the same home from birth and hadn’t been allowed to do anything without the other twin being present and had been subject to a fairly authoritarian parenting regime that instilled values about what names you should choose in a partner and for your dog.
In other words: I don’t really think these coincidences are evidence for, as one article put it, “the staggering influence of genetics”. I don’t know what they’re evidence for, though: are these cherry-picked from a much longer list of comparisons, many of which showed no similarities? Did someone just lie about some of them? Has the story gotten a bit out of control over the years? Did the twins find out about each other earlier than is recorded, and then deliberately do things like giving their dog the same name? But surely that kind of “leak” story wouldn’t be an explanation for the coincidence about their wives’ names being the same twice in a row, or the name of the adopted brother.
Maybe all these coincidences really did happen as described… once. But if that’s the case, the Jim Twins anecdote is relegated to the status of a fun story to share online, and shouldn’t really be relied on for any kind of scientific—or, indeed, legal—purpose.
Don’t get me wrong - I’m not saying this to criticise twin studies, or any particular researcher, in general. I’m not one of those weirdos who spends their life shouting at behaviour geneticists on Twitter. I’ve done twin research; I’ve defended behaviour genetics; I’m on board, here. But looking back, it is a bit mad that these kinds of anecdotes and other implausible-sounding stories are rolled out so often when talking about the importance of genetics.
The annoying thing is that the evidence for twin similarities—and for the major influence of genetics—is impressive enough without these coincidence-mongering anecdotes. Let’s just take one example from near the top of a Google Scholar search: As discussed here, in one sample of identical twins, the correlation of scores on a reading test was r = 0.84 (non-identical twins had a correlation of 0.46). Remember that correlations can only go as high as 1.00, or as low as -1.00; a correlation as high as .84 is usually about as high as you’d get if you gave the same person the same test twice. That’s incredibly impressive - and it’s based on hard data rather than anecdotes.
Against anecdotes in behaviour genetics
Ah! But now you might be thinking: doesn’t the 0.84 correlation you just mentioned validate the Twin-kie Defence? Doesn’t this mean that the lawyers in the South Carolina case were right to say that the students were essentially identical, and that we shouldn’t be surprised by their essentially identical results?
No. First off, the correlation is .84, and doesn’t usually go higher than that - to get scores as close as the twin students got on the exam, we’d be talking about even higher correlations. The reason the correlation isn’t 1.00 is that there’s always going to be some degree of error on any test of this nature: even if you’re correlating the results with your own results from a month earlier, you’d never expect it to be an absolutely perfect similarity (or even close to that). Maybe you slept better before one of the occasions and were a bit more on-the-ball; maybe you were distracted by a loud motorbike outside the second time you took the test so got a slightly lower score. It’s all measurement error, and it’s totally to be expected. That is, even if it was the exact same person doing the exam, it would still be pretty astonishing if they got such a high level of similarity across the two tests.
0.84 is also an average correlation across all the questions on the reading test: it’s saying that across all the questions on this particular test, the twins’ total scores were strongly related. It’s not saying they got exactly the same questions right and wrong, let alone at exactly the same rate, which is what the university was arguing in the South Carolina case. In other words: twins will be similar, but not that similar.
As I said above: cheating is not the only other explanation. Maybe, like the Jim Twins, it really was just a massively-unlikely coincidence. But I don’t think it’s right to use all our hard-fought knowledge about how genes influence behaviour to produce hand-wavey arguments which essentially say that any coincidence can be explained by genetic similarity - especially without even attempting a serious analysis to back them up.
Which brings us back to the debates around free will. Among the many reasons people feel uncomfortable about behaviour genetics is that it seems to imply that we’re determined to act in a certain way; that we’re stuck on genetic tram-lines into the future, unable to change course. Telling and re-telling stories like the Jim Twins, and arguing that identical twins are essentially the exact same person in every respect, just plays into this stereotype of the field, and must be confusing to readers who also hear behaviour geneticists talking about how genes aren’t deterministic, how the environment plays a big role, how it’s all about probabilities, and so on.
(And by the way, even the most hardcore anti-free-will determinist should acknowledge that our genes have little to do with whether we’re determined or not - for them, we’re determined by the laws of physics—all the atoms bumping into each other that were predictable right from the Big Bang—and our particular genetic endowment is way down the list of ultimate causes for our behaviour).
Maybe it’s time to stop bringing up these stories of coincidences, as intriguing as they are, and focus on interpreting the huge pile of data we’ve gathered over the decades in behaviour genetics. Debating how those data apply to cases like that of the South Carolina students—and to how we should understand free will—is too interesting to be left to vagueness.
Image credits: Top photo: Getty. Spider-Man photo: internet meme.
> Both twins have an adopted brother whose name is Larry.
I mean I don't see any plausible way this could be genetic. A child doesn't have any meaningful level of control over who their parents choose to adopt or what they name them.
I was wondering if you'd cover this! I hope you had a nice honeymoon. The other complicating factor here is how the twins studying together for the duration of their training program would influence their exam answers. You'd expect two people who've studied intensively together to pick the same wrong answers more frequently than two students who studied independently. (Make those two students identical twins on top of that and I assume their propensity to get tripped up by the same wrong answers is a little higher.)
I don't actually know how much of the training program they did together, but I don't hear anyone considering it as a factor. I see it from time to time (e.g., a group of students who studied together tending to misunderstand operant conditioning terms the same way).