Genetics, Science, Science / Tech

Not Everything Is An Interaction

Albert Einstein was a brilliant man. Whether his famous equation of E=mc2 means much to you or not, I think we can all concur on the intellectual prowess—and stunning hair—of Einstein. But where did his brilliance come from? Environment? Perhaps his parents fed him lots of fish (it’s supposed to be brain food, after all). Genetics? Surely Albert hit some sort of genetic lottery—oh that we should all be so lucky. Or does the answer reside in some combination of the two? How very enlightened: both genes and environment interact and intertwine to yield everything from the genius of Einstein to the comedic talent of Lewis Black. Surely, you cannot tease their impact apart; DNA and experience are hopelessly interlocked. Except, they’re not. Believing that they are is wrong; it’s a misleading mental shortcut that has largely sown confusion in the public about human development, and thus it needs to be retired.

Despite strong genetic influences on IQ (and there are strong genetic influences on IQ), we can’t calculate the proportion of credit for Einstein’s intellect that is owed solely to his genes. He’s just one person, and this prevents us from knowing that, say, 70 per cent of his genius was scrawled in his DNA. Though true for given individuals, this reality has lead some to claim that under no circumstances can we ever talk meaningfully about genetic effects on development separately from environmental effects. Trying to split the two would be a bit like trying to calculate the area of a rectangle using only its height. Your middle school teacher would be ashamed, as everyone knows you need both height and width. As the great psychologist Robert Plomin noted, however, “[I]f we ask not about a single rectangle but about a population of rectangles, the variance in areas could be due entirely to length, entirely to width, or both.” As with rectangles, so with humans. Differences observed in a population of humans can be described as the result entirely of genetic differences, environmental differences, or some combination of the two.

The studies that can split genetic from environmental influences nearly always demonstrate that both matter to some extent. Why is it wrong, then, to say that all development is the product of interactions between genes and environment? The wrongness has to do with understanding what the word “interaction” actually means. The term is commonly used to refer to the fact that practically every complex human outcome is influenced by both genetic and environmental factors. The more technical definition of “interaction”, however, refers to instances where the magnitude of genetic effects on some trait either increases or decreases depending on the environment. Whether genes and the environment interact in this narrow sense is a mathematical issue, a different ball of wax entirely from parochial ideas about interaction. Most importantly, genes and environments might interact in the parochial sense—meaning that both might matter for development—while in the mathematical sense there is no interaction to be found.

Consider an example of a “technical interaction.” Prior research has suggested that, in the United States, the relative influence of genes on differences in general intelligence appears to depend on socioeconomic status. Genetic effects either explain more, or less, of individual differences in intelligence depending on whether one is raised in relative wealth versus poverty. Technical interactions remain a bugaboo for researchers. They are notoriously difficult to detect, prone to certain methodological problems, and like many things in science, they often do not replicate. The SES-IQ interaction gives the appearance of being isolated to the United States. A large recent review by the psychologists Elliot Tucker-Drob and Timothy Bates didn’t really find the effect in samples from other countries. This, by the way, doesn’t mean that the studies finding interactions were wrong. Not at all. But it does illustrate the fact that detecting an interaction does not guarantee that it exists everywhere, and in all places. When testing for interactions using specific genes measured with specific environments, the replication record becomes poorer still (for a variety of reasons that the behavior geneticists Laramie Duncan and Matthew Keller outline here). In short, the research landscape is not replete with confirmed interactions.

I hail from the southern part of the United States and so I speak with a lovely southern inflection (so I’ve been told!). Had I been born in Australia (or France or Britain), I would have an even lovelier accent. Accent variation is wholly environmental; it depends on where you’re born. Saying that accent is “an interaction between environment and genes”, is false. Alternatively, there are outcomes where individual differences are more fully explained solely by genetic differences. Most traits are the product of genetic and environmental influence, but the fact that both genes and environment matter does not mean that they interact with one another. Don’t be lured by the appeal of “interactions.” Important as they might be from time to time, and from trait to trait, not everything is an interaction. In fact, many things likely are not.


Brian Boutwell is an Associate Professor of Criminology and Criminal Justice at Saint Louis University. Follow him on Twitter @fsnole1

Filed under: Genetics, Science, Science / Tech


Brian Boutwell is an Associate Professor of Criminology and Criminal Justice at Saint Louis University. His research interests include the biological evolution of human traits, genetic and environmental underpinnings of human violence, and general intelligence. His published articles have appeared in PLOS One, Behavior Genetics, Developmental Psychology, Journal of Psychiatric Research, Criminology, and Social Science and Medicine as well as others. He was also a coeditor of The Nurture versus Biosocial Debate in Criminology: On the Origins of Criminal Behavior and Criminality (Sage).


  1. Good post!

    A couple of things:

    The studies that can split genetic from environmental influences nearly always demonstrate that both matter to some extent.

    While this is technically true, the “environment” in question is almost never in the sense of what most people mean when they say “environment.”

    IQ-SES interactions might not be real in the U.S. even they to be a result of publication bias.

  2. Genes interact with the environment to form traits the way the contents of your pantry interact with your kitchen tools to create food. You would not have an accent if you didn’t have the genetic material allowing you to properly develop language, for example (you probably take it for granted, but some people would have neither southern nor Australian accent regardless of where they grew up).
    From reading this essay it seems like focusing on the mathematical definition of “interaction” (as measured by general linear regression analysis) is completely obfuscating the mechanistic, biological level, where gene expression is dictated by various signals, many of which are influenced by the environment, while development of various traits are contingent on availability of the necessary genes and the system’s ability to generate the necessary proteins from them. For an example of this principle, check out this paper:

    This isn’t to say that studying behaviors at the gene level is necessarily useful. Since the brain is a learning machine (learning from experience and interactions with the environment, i.e. producing actions and perceiving outcomes), as long as we have the necessary genes to learn whatever it is that we’re working on, there’s no reason to give too much attention to the genetic level, it’s many steps removed from most behavioral outcomes.

    • “Genes interact with the environment to form traits the way the contents of your pantry interact with your kitchen tools to create food.”

      Can you show me the biological studies for this similarity? All you provided was some allele association with tiny sample…from 2003.

      Show me the actual biological studies you got your claim from. Thanks.

  3. My take on the SES-H is more pessimistic, and I think you should be too. Small initial study was widely publicized as showing very large effect of supposedly significant social important — near heritability among low SES groups. Fast forward 10 years of mixed replications. Now comes a meta-analysis. This meta-analysis finds no good evidence of an overall effect: .029 [CI95: -.015, .073]. So, the authors try a division of studies into two parts: USA and non-USA. And in the USA subsample there is an effect with a confidence interval that doesn’t have zero: .074. The effect estimate for the other subsample is slightly negative at -.027. This kind of method is ripe for disaster as was pointed out by Schmidt and Hunter decades ago. In a delicious meta-logic, this finding represents an interaction… about an interaction effect.

    By the way, for the USA subsample, the meta-analysis effect size is 2.7 times smaller than the original study everybody was all up about and which has 1033 citations on GScholar as of now. I shall not be particularly surprised if larger studies further reduce the effect size of this.

    It is also my general view that statistical interactions are pretty rare, yet everybody keeps testing for them. Interactions are inherently interesting it seems, main effects not so much. When one does a lot of tests, one finds a lot of stuff, p<alpha. But if the base rate of these interactions is very low, these will mostly be false positives. (The continuous version of this argument is if the distribution of such effects is tightly around 0, most reported estimates with p < alpha in generally small studies will be very inflated.)

    Some day I will be annoyed enough to do a large scale study to determine the base rate of interactions. Maybe using the OKCupid dataset.

    • aljones909 says

      I think you’re probably correct regarding SES. It’s likely that advanced countries (good education system, good life chances) will stratify on the basis of intelligence. Do you have a link to the studies you talk about here “Now comes a meta-analysis. This meta-analysis finds no good evidence of an overall effect: .029 [CI95: -.015, .073].”

  4. John says

    I defy you to find the Brummy (in the UK) accent lovely

  5. Santoculto says

    Repetition of Einstein master genius by (((media))) is equivalent to holocaust repetition?? In other words, highly dubious?

  6. Santoculto says

    People have personal optimum to sub-optimum environment. When we are reached in personal optimum environment we will be more prone to become receptive, with the higher expression of “positive” or relaxed traits while when we are in our environmental nightmare we will be likely to express a defensive or aggressive approach or higher expression of “negative” behavioral traits like anxiety, mood states.
    The levels of perfectionism and exotism explain why some people are highly prone to express defensive and or aggressive traits.

    Perfectionism tend to reflect insular or general attention to the details and intrinsic motivation to reach it at personal way and often expect that other people will reach this level soon something that rarely happen.

    Perfectionistic temperament generally at insular or specific nature tend to be closely related to hyper emotional sensitivity and specially in far-to-be perfectionistic environments.

    There are, as always, a spectrum of intrinsicability (how intrinsic or intense certain behavior expression is). More intense is certain behavior more instinctive, more difficult to change via interactions. In the same ways less intense behavior or whatever traits is, less prone to change. A verbally gifted versus a verbally deficited, for example. A verbally gifted easily will develop or reach your potential because it’s practically hardwire for him/her. A verbally deficited whatever the quality of pedagogical intervention difficultly will surpass your Lower levels in this cognitive facet. Someone who are in the middle of this spectrum will be, or more prone to reach your avg potential or will be capable to surpass but not so much, still a possibility. Some can be more plastic than others.

  7. sonali S says

    Depends on the threshold level of what is considered a technical interaction or not. May be a 50% or higher positive or negative correlation between two outcomes is interaction while a lesser correlation is not.

  8. DiscoveredJoys says

    I’ve read recently that Albert Einstein suffered from Uncombable Hair Syndrome, a rare genetic mutation and not an interaction with the environment.

  9. Chester says

    Accent (Dialect) variation is not wholly environmental. Genetic variation in the population produces individuals with craniofacial anomalies that make speech production difficult or impossible, thus changing the dialect of those speakers. Genetic variation can also alter the ability to control the motor neurons of the muscles involved in speech from birth or as a result of a neurological disorder, as in a seizure disorder or Parkinson’s.

  10. Santoculto says

    Some people have more neutral accent than others. Also we have ”accent self-consciousness”. Some people is so submerged from certain culture that they simply can’t emulate other accents.

    Other thing, some people can’t talk some sounds of words, for example, the american [specially deep south] accent to the R word.

  11. Santoculto says

    Literally speaking everything is a interaction, 😉

    Without interactions of molecules, or whatever the nano-particules you named, every-thing would be none-thing.

    or níííet

  12. He was brilliant but so were many others and some, like Tesla much more brilliant, but Einstein was ‘fashionable’ and supported in many ways and his supposed outstanding ‘brilliance’ was sold on the open market. He was not as brilliant as many believe and like any scientist, he got things wrong. He also resisted Quantum Mechanics, which, if he had been as brilliant as claimed, one presumes he would have embraced.

  13. this article has a point.

    when calculating the effect our environments have on the heritability of some genetic trait, the mathematical strength of the interaction is measured in terms of how significantly the environment affects genetic expression, where the latter is postulated to have the definitive role in deciding whether and also the way in which the environment can produce such effects. and so in these interactionist theories, where researchers are more or less trying to quantify what part nature and what part nurture we are, it is not the importance of genes that is up for debate- for it is their characteristics that set the very parameters of the theories- but the extent to which the environment has any role at all

    but i think we can simplify this lesson a bit, and put an even finer point on it

    for my money, the short adage suffices to illustrate this idea that no environmental factor can influence an organism without that organism’s genes’ first having allowed the environment to have such an influence

    a similar reining in of the giddy excesses of empiricism can be found in continental philosophy and in leibniz’ pithy but profound response to the originally aristotelian, intermediately aquinian, and finally lockean notion of the blank slate, where leibniz adds to empiricism’s textbook formulation of ‘there being nothing in the mind that was not first in the senses’, that this is true of everything except for the mind itself

    in one deft phrase leibniz helped philosophy remember that the structure of the mind has to be pre-specified to make learning possible and that this innate potential for structure is the very thing that makes experience possible, a thing which empiricist theories of knowledge are not licensed to take for granted as they do

    a simple sentence that is pound for pound one of the most arresting things i think anyone has ever said:

    “there is nothing in the mind that was not in the senses, except the mind itself”
    g v. w leibniz, new essays on human understanding (1704)

Comments are closed.