Twin studies stand up to the critique, again

The history of twin studies is littered with attempts to discredit them – such as this bit of rubbish. Yet every challenge has been met, with a couple of newish studies knocking off another.

The basic idea of twin studies is that by comparing the similarity of fraternal twins to the similarity of identical twins, you can tease out the influence of their genes. Twin studies tend to find that most behaviours have heritability of at least 0.2 (that is, 20 per cent of the variation is due to variation in genotype), IQ a heritability of over 0.5 and height around 0.8. However, twin studies require an assumption that identical and fraternal twins have equally similar environments, and this is where the critiques begin. If identical twins have a more similar environment, the estimates of heritability may be too high.

The responses, however, are plenty. There are studies of twins reared apart. Adoption studies find similar results. For those who believe that identical twins are treated differently to fraternal twins, there are studies of misidentified twins – where everyone thought they were identical or fraternal, but they were the other. Peter Visscher and friends took advantage of the differences in relatedness between siblings to generate estimates of heritability consistent with twin studies (You are 50% related to your siblings on average, which means you can test how similarity varies with variation in relatedness . For me, that study should have been the final nail in the coffin of any arguments that twin studies hadn’t told us anything).

One critique still floating around is that people who look more similar are treated similarly (although the misidentified twin studies deal with this to a degree). And the New York Times has reported two studies that take on that argument. In the first, Nancy Segal assessed the similarity in personality of 23 pairs of unrelated lookalikes. The similarity – effectively zero. Then in a replication, Segal got a skeptic, Ulrich Ettinger, involved in the project. They found the same result – no resemblance – unlike Ettinger’s expectation that people who looked alike would have similar personalities as people would treat them the same.

As Razib points out, these studies involves a small sample. However, they are yet another piece of evidence pointing in the same direction as all the rest.

22 comments

  1. A few nitpicks:

    The heritability of IQ is not 0.5. It’s more like 0.8-0.95. The heritability of most traits stands in the 0.4-0.6 range, which shoots up to 0.7-0.9 when you correct for measurement error.

    1. I agree the heritability of IQ is higher than 0.5 – hence the use of the word “over”. I’d say 0.8. On the other traits, the numbers I was giving are what twin studies tend to find, not the absolute heritability.

  2. Heritability, especially with respect to human behavior, is utter nonsense. The “rubbish measure” is of the same order as the notion that a p-value 95%. Twins reared apart answering criticisms? Garbage. Collins does not address what has been called the “beautiful twin/ugly twin” (BT/UT) argument. I will discuss this argument in a moment but it is important to note that the argument is illustrative of the problem with behavioral genetics, not the only issue by any means. Here is the argument framed as a syllogism:

    P1: MZ twins are more likely to resemble each other than DZ twins.
    P2: Good-looking people are treated differently than ugly people.

    Conclusion: MZ twins reared apart are treated more similarly than DZ twins reared apart since MZ twins are more likely than DZ twins to be both beautiful or both ugly.

    Similarly,the notion that people reared in the same household “have the same environment” (and, thus, that differences are “genetic”) is utter nonsense. Behavior is the result of an organism/environment interaction and TINY differences in initial conditions that ARE due to genes are amplified by the dynamics of the organism/environment interaction. The notion that heritability offers any kind of quantitative assessment of the role of genes is, as I said, as big a fallacy as the notion that a p-value offers quantitative assessment of the truth or falsity of Ho.

    Cordially,
    Glen

    1. Your website altered my post…not good…what I said (paraphrase) was that (using words rather than symbols) a p-value less than 0.05 means that the probability of Ho being false is 95%. You should fix this problem or warn people not to use certain symbols.

      Cordially,
      Glen

      1. Oops! I should have said that the notion that a p-value means that there is a 95% chance that Ho is false is incorrect. Sorry for being so careless.

        Glen

    2. @Glen M. Sizemore:

      So Mr. Troll Jorge of Leinster, you are harassing Jason Collins now?

      P1: MZ twins are more likely to resemble each other than DZ twins.
      P2: Good-looking people are treated differently than ugly people.

      Conclusion: MZ twins reared apart are treated more similarly than DZ twins reared apart since MZ twins are more likely than DZ twins to be both beautiful or both ugly.

      Except that this very post discusses a study that shows that the equal environment assumption holds. Not to mention that misclassified MZ/DZ twins behave as similarly as their genetics would dictate, not from social expectations. So try again.

      Or better yet, don’t.

      Similarly,the notion that people reared in the same household “have the same environment” (and, thus, that differences are “genetic”) is utter nonsense.

      They don’t have to have exactly the same environments, which no two human beings do. Their environments just need to be more similar than those who weren’t raised in the same home. And they are.

      Behavior is the result of an organism/environment interaction and TINY differences in initial conditions that ARE due to genes are amplified by the dynamics of the organism/environment interaction.

      Too bad that’s complete BS, since heritability of behavioral traits and intelligence increases with age. Even longitudinal adoption studies show this.

      1. Jayboy: So Mr. Troll Jorge of Leinster, you are harassing Jason Collins now?

        GS: There was nothing trolling about my post. WTF is wrong with you?

        P1: MZ twins are more likely to resemble each other than DZ twins.
        P2: Good-looking people are treated differently than ugly people.

        Conclusion: MZ twins reared apart are treated more similarly than DZ twins reared apart since MZ twins are more likely than DZ twins to be both beautiful or both ugly.

        Jayboy: Except that this very post discusses a study that shows that the equal environment assumption holds.

        GS: Do tell…and how exactly does it do that?

        Jayboy: Not to mention that misclassified MZ/DZ twins behave as similarly as their genetics would dictate, not from social expectations. So try again.

        GS: That finding is not at all antithetical (my that’s a big boy word, eh – when you’re grown up maybe you can use big boy words!) to what I said. Indeed, it would be exactly what one would expect. It seems you haven’t understood what I said.

        Jayboy: Or better yet, don’t.

        GS: I still don’t get what set off your childish post. Any dissent on this blog is trolling? Or is it something special about Jason. You two boys go to the same middle school or something?

        Similarly,the notion that people reared in the same household “have the same environment” (and, thus, that differences are “genetic”) is utter nonsense.

        Jayboy: They don’t have to have exactly the same environments, which no two human beings do. Their environments just need to be more similar than those who weren’t raised in the same home. And they are.

        GS: How do you figure that little feller? When twins are reared in the same household it is assumed that the variance due to the environment is zero – not just “less” than when they are reared together.

        Jayboy: Behavior is the result of an organism/environment interaction and TINY differences in initial conditions that ARE due to genes are amplified by the dynamics of the organism/environment interaction.

        Too bad that’s complete BS, since heritability of behavioral traits and intelligence increases with age. Even longitudinal adoption studies show this.

        GS: That is exactly what one would expect if my critique is correct. The behavioral trajectories of DZ twins would separate more rapidly and with greater magnitude over time than those of MZ twins. Thus, one would expect that H would increase over time.

        Cordially,
        Glen

      2. Jayboy: Except that this very post discusses a study that shows that the equal environment assumption holds.

        GS: Do tell…and how exactly does it do that?

        Here’s the longhand version of Jayman’s statment. If there is any variation in environment between fraternal and identical twins due to more similar looking people (the identical twins) being treated the same, it has no effect on their personality. Therefore, for the purposes of calculating heritability, you can safely assume equal environment. As for the specific parts of the study, they’re right at the top:

        Study one (from the abstract): A common misconception is that monozygotic co-twins’ personality resemblance results from similar treatment by others, due to their matched physical appearance. The present study brings unique evidence to this question by assessing the similarities in personality and self-esteem of 23 pairs of unrelated look-alike individuals.

        Study two (also from the abstract): Twin research critics assert that similar treatment of monozygotic (MZ) twins results from their matched physical appearance, and that their similar treatment explains their within-pair behavioral similarities. A genetic explanation of MZ twins’ resemblance is, thereby, dismissed. To address this challenge, Segal (2013) found a lack of similarity in personality and self-esteem in pairs of unrelated look-alike individuals. The present study describes a constructive replication of that work, confirming these findings.

        Continuing this point, you state:

        When twins are reared in the same household it is assumed that the variance due to the environment is zero – not just “less” than when they are reared together.

        The variance is assumed to be zero not because there is actually no variation. It is assumed to be zero because to the extent there is any variation (however marginal it might be), it has no effect on outcomes. In other words, it can be treated as though it is zero. This argument is the staple of critiques of twin studies, and the responses numerous (hence the ‘again’ in the title to the post).

      3. The “beautiful twin/ugly twin” is just a metaphor for any differences between twins. To say that such variation (which is necessarily greater in DZ twins) “…has no effect on outcomes” is mere naïve assertion – common enough for people who do not understand the role of the environment and the fact that features of the environment (I know! Let’s call them contingencies!) concatenate. Let’s say that one twin is slightly less able to pay attention (not that “paying attention” is completely understood – but I understand it better than you) At first, it is not very noticeable – but maybe some – the parents may be conscientious but the twin who is better at paying attention may read a bit better, for example. When schooling gets under way, however, the environmental contingencies begin to concatenate. For the twin who can pay attention better, school goes well. But for the other twin there are some successes but often failures, which begin to increase. At some point the very introduction of a new topic in school becomes an aversive stimulus, and the “bad twin” begins to escape the stimuli through further inattention, or maybe whatever attentional repertoire had been built simply undergoes extinction (i.e., it is seldom reinforced – err, “rewarded” is the ordinary-language term which has made a comeback in supposedly technical discourse). For the “bad twin” the entire school environment becomes aversive so that he or she begins to skip school etc. etc. As I’m sure you can see, this state of affairs has amplified the slight variability that originally existed and will impact almost every facet of his or her life. Are you saying that this scenario is unlikely or that it will not have an effect on the many measures that are looked at in human behavioral genetics? Couldn’t be…that would be silliness.

        Regards,
        Glen

      4. Lets pretend for a moment your scenario is true. How is it a critique of twin studies or the concept of heritability? The high heritability for, say, IQ suggests that this process cannot be very sensitive – changing homes or schools doesn’t change the result.

  3. Oops…forgot to respond to a line: “Similarly,the notion that people reared in the same household “have the same environment” (and, thus, that differences are “genetic”) is utter nonsense.”

    GS: I agree. That is, however, an ASSUMPTION of behavioral genetics.

      1. Did you even read the paper Trollman? I don’t know how you could have, because I just finished and I started it a while ago.

        You’re a silly nonsensical poser who’s hardly worth the time, beyond entertainment value.

      2. No I did not and I see no reason to do so. Some dipshit claims some stuff…I see it all the time. Most of what is written about behavior – broadly speaking – is just BS. If the Abstract was informative in any way I might look at it when I go into the lab tomorrow, but it (the Abstract) looks like it was written by a 14 y.o. – much the way your insipid posts look. Why don’t you critique my hypothetical treatment of a difference in the “ability to pay attention” between twins – are you saying that the scenario I proposed could not occur or that such an eventuality does not make a shambles of the naïve view that living in the same household means that the twins are exposed to the same functional environment?

  4. Having had a chance to look over the paper offered by JayBoy I see that, as I suspected, it contains nothing that could be said to counter my argument. It starts out with a straw man diatribe about those who attack behavioral genetics being afraid of linking “biology and behavior.” Aside from being an inaccurate use of the term biology (behavior acquired through ontogenic processes is just as “biological” as any unconditioned reflex), this is inapplicable to me since my argument relies on differences in genes. But, knowing something about behavior and the often powerful role of subtle environmental variables, it is clear that it is utter nonsense to claim that being reared in the same household (and going to the same school etc.) means that there is zero variation in behavior attributable to the environment is utter rubbish. There is no way to make this claim while being informed as to the nature of the organism/environment interaction that is behavior. The paper deals only with the issue of the gross phenotypic differences or similarities among MZ and DZ twins but, as I said to Jason (and could have been clear by actually paying attention to what I originally wrote), the “beautiful twin/ugly twin” argument is just a metaphor for all the possible ways that DZ twins can differ genetically. Behavior can be seen as a trajectory in a multi-dimensional space in which the current repertoire of an organism interacts dynamically with a host of variables, some of which have been elucidated by a natural science of behavior. Such a view makes it clear that differences in DZ twins, that are linked to genes, makes it so the initial conditions relevant to such a dynamic system will be further apart in the state-space than for MZ twins (who aren’t supposed to have any genetic differences but do, most likely, have epigenetic differences). Such differences in the initial conditions insures that the trajectories, for DZ twins, will diverge more rapidly and in larger magnitude than MZ twins. This is simply the way it must be given what we know about the organism/environment interaction that is behavior.

      1. “So what you’re saying is that genetic differences can result in big differences in outcomes.”

        GS: That would be a misleading characterization of what I said, especially given the context. [Your sycophantic sidekick seemed to enjoy it, though! But, then, he is very…well…simple.] Anyway, the point I made is that small differences in behavior (and large differences, of course), tied to genetic differences, are amplified by the organism’s ongoing interaction with the environment. This assertion is consistent with the increase in H across time often found in longitudinal studies.The issue is not “Are genes important?” (they obviously are) – the issue is: “Can the variance due to genes be estimated?” That answer is, obviously, “No, and/or the question, itself, is literal nonsense.” I have already given an explanation of why this is, but you either: A) misunderstood it or B.) deliberately misrepresented it. I have repeatedly given hypothetical examples of how the behavior of DZ twins can diverge, UNDER THE IMPETUS OF THE ENVIRONMENT, faster and more completely than MZ twins. If that perspective has merit, and I’m pretty sure it does, the notion that “being reared together” is somehow a control for “environmentally-caused variance” is utter nonsense. But I have said that before.

      2. @Trollman:

        Too bad there’s no evidence for gene-environment correlation effects (which is what you are describing), and indeed, evidence against them from extended twin studies.

  5. “Too bad there’s no evidence for gene-environment correlation effects…”

    Of course there is. But what’s published barely scratches the surface. Most of the behavior of humans that we are interested in is shaped and maintained by its consequences (you know, reinforcement, punishment…). And the person’s behavioral repertoire at any given time influences further interactions and so on. Now, if genes influence behavior (and I’m sure you would agree that they do) then by necessity genes will influence the types of environments in which people find themselves. The problem for analysis is that we do not know what sorts of processes are affected by genes, but it is safe to assume that almost any aspect of basic behavioral processes may be so influenced. So, for example, one DZ twin may have an allele that makes it so only richer schedules of reinforcement will maintain behavior (do you think that this is impossible?). Such a person would quit interacting with certain environments (because the reinforcement provided is inadequate; i.e., his or her behavior undergoes extinction) in which the other twin persists. Examples like that could be multiplied endlessly. Genetic susceptibilities to social reinforcement, conditioned reinforcement (like that which maintains “paying attention”), sexual reinforcement etc. all would influence the type of environments with which the individual engages. There may be different sensitivities to delays to reinforcement that are tied to genes. Such genes would influence the characteristics of the hyperbolic delay functions (“delay discounting”) that describe whether or not the person tends to choose immediate reinforcement (“impulsivity”) over delayed reinforcement. Laboratory studies with non-humans could identify the existence of such genes through artificial selection. I’m guessing one could breed rats that have extreme indifference points on choices between small, immediate reinforcement, and large delayed reinforcement (“degree of impulsivity”), and that goes for many aspects of basic behavioral processes. And, of course, DZ twins are the ones who might have different alleles relevant to such processes. Thus, the notion that being reared in the same household means little or no environmental variability is hogwash – simply from a logical point of view.

    1. Once again, what is the evidence that a dynamic of this type is confounding twin study results? We have studies of misidentified twins, which generate the same results. We have twins reared apart – same results. And we have Visscher and friends methodology using actual relatedness, which does not rely on the same environment assumption. If the dynamic you describe was messing up the analysis using traditional twin studies, these results would not line up so nicely.

Comments welcome

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s