Evolutionary psychology and behavioral genetics

https://www.parsingscience.org/wp-content/uploads/2020/09/ParsingScience082-Barbaro.mp3#t=02:19,05:04

Barbaro: They both kind of came about at the same time. Around the 70s and 80s, they began to both make major contributions and kind of develop as independent fields. And despite them having, in my opinion, very complimentary roles in understanding human behavior, they do tend to ride kind of in parallel to one another. So when we look at how traits develop over time … you know a key component of evolutionary theory is that the traits have to be heritable in order to be passed down from parents to offspring throughout generations. So despite that being a key factor in what we know about evolutionary theory at a very basic biological level, the integration of understanding heritability and the role of genetic variation in understanding any of these traits we’re interested in – as evolutionary psychologists or psychologists more generally – tends to kind of not be the primary focus of psychological research in particular. So understanding heritability and how traits are passed on and actually looking at the genetics of whatever trait we’re looking at, can actually help us understand the broader evolution of those traits. And especially now, with more advanced genome-wide association techniques that geneticists have, we can actually start to look at if there’s different types of selection on traits over human evolutionary history. Which is really important for understanding how traits evolve over time, rather than just identifying what adaptations or psychological behaviors may have evolved over time. And then on the flip side, behavioral genetics can really be informed by evolutionary psychology by having theoretically grounded hypotheses to actually test. Whereas behavioral genetics – it is a big, methodological discipline – so you can take kind of these behavioral genetic models and apply them to almost any trait. Most of them are psychology traits, and look at how heritable trait is. So a lot of behavioral genetics research … I mean, there’s been probably tens of thousands of studies at this point that just identify how heritable is trait x? How heritable is trait y? So, evolutionary psychology can kind of offer some guiding theoretical kind of framework to understand, you know, what traits are worth investigating. And what hypotheses are worth investigating. So there’s, there’s a big area of overlap, and I think, in the past few years that that overlap has just really started to get very mainstream within both of these fields.

Genetics and home environments

https://www.parsingscience.org/wp-content/uploads/2020/09/ParsingScience082-Barbaro.mp3#t=05:03,08:39

Watkins: Nicole’s study investigated not only the extent to which being spanked as a child is associated with negative behavioral outcomes later in life, but also the degree to which those outcomes could be explained by the influence of the interaction of two other factors. The first of these was genetics, namely the degree to which those behavioral outcomes differed between more genetically similar full siblings and more genetically dissimilar half siblings. The other was the home rearing experience of these siblings, specifically, the experiences of the home life environment that were shared in common between siblings versus those that were unique to each sibling, and therefore were non-shared.

Barbaro: For this paper, we’re interested in understanding the impact of spanking. And when we say spanking, we mean, open palm spanking, typically, on the back end of a child. You know, we’re not talking about physical abuse, we’re talking about spanking that’s intended to correct a problem behavior. So typically, a child did something, you know, undesirable, they get spanking as a punishment in the hopes of decreasing the child doing that behavior again.

We’re interested in how spanking impacted a wide range of developmental outcomes in children. And this is a very heavily researched topic. But one area that we were interested in that – it’s very limited – is understanding how behavioral genetics can help us understand the impact of spanking on adult outcomes. And this is an area that’s only been looked at in actually a handful of studies, despite hundreds of studies more generally looking at the effects of spanking. And all of these studies find that spanking has a negative outcome. So typically, kids have more antisocial behavior growing up, or emotional problems, any internalizing problems, poor social relationships. And we wanted to know that once we account for this process that we refer to as genetic confounding, which is akin to kind of looking at a third variable. So the genetics that underlie children’s problem behavior also underlie their propensity for getting spanked. So how this, what we call a third variable of shared genetics, can potentially explain some of this relationship that we typically see in the literature between spanking and outcomes. Because the literature so limited on this topic, we took a simulation modeling approach. So we basically took all the data that we could find from the published literature, all the different effect sizes, and put this into a simulation modeling program to estimate how much of the relationship – so that direct relationship between spanking and, say, antisocial behavior – is accounted for by this third confounding variable. And once we account for that, how much of that direct relationship is left? Which would allow us to say how large, if any, the effect of spanking on negative outcomes really is like a causal sense, in the intuitive sense that we think about spanking harming children’s development. And what we found is that there does seem to be evidence that spanking does indeed have a negative effect on child outcomes, just as we suppose. But about a half to two thirds of that relationship that we typically look at is explained for by this confounding variable. So although our research shows that there is a negative effect of spanking on child outcomes, but the effect size or how detrimental that effect is, is substantially reduced by about 50% to two thirds of what is typically reported. Which we find to be an important finding for a lot of practical reasons.

Genetic covariation

https://www.parsingscience.org/wp-content/uploads/2020/09/ParsingScience082-Barbaro.mp3#t=08:38,11:34

Leigh: Nicole’s study analyzed the responses to questionnaires completed by nearly 1000 children whose mothers had previously reported the frequency of their spanking of those children when they were four to nine years old. Then every two years for 28 years, these children responded to various questionnaires about their own delinquency, depression, and alcohol use. Some of these children were full siblings of one another, and some are half siblings, allowing for the determination of the extent of their shared genetics. Nicole explains next how differences in people’s genetics and home environments can explain the association between being spanked and the experience of adverse outcomes later in life.

Barbaro: Complex psychological traits such as, you know, our behaviors and personalities and all the things that we’re interested in as psychologists … we know that each of those traits is underpinned by hundreds or thousands of genetic variants. So it’s not just one gene for any particular trait. So there’s going to be hundreds to thousands of genetic variants that are underlying, you know, your personality. That are underlying your hair color. Any of these, you know, complex traits that we’re interested in. So what we mean by genetic covariation is what is the overlap in the, let’s say, 500 genes that underlie your personality? And then what overlapped those 500 genes have with the genes that underlie, for example, spanking behavior. So if you get spanked as a kid. So if we see low genetic covariation, we would expect maybe – I’m just pulling numbers here – maybe five or ten of those genes are the same. Whereas if there’s a high covariation, we would expect maybe 300 of those genes are the same. So the same 300 genes that underlie one trait are also underlying another trait. And what happens is, at the higher level that’s typical in psychological research – if you have a correlation of like, say, .3 between a personality trait and a particular behavior, but the genetic covariation is as high as a .6 – that’s going to dramatically reduce the kind of true correlational effect that we see between personality and whatever that behavior is. So understanding genetic covariation can kind of pull out this confound, and to allow us to more precisely understand the true causally consistent relationship between two what we call phenotypic traits. Which would just be, you know, your personality and a behavior, whatever those two psychological traits are. In traditional statistics, we think of this as like the “third variable problem.” So there’s a single variable that’s causing, or that is related to, the two variables of interest to the researcher. And once you control for that third variable, then you see typically, you know, the relationship goes away, or it’s significantly reduced. So that is specifically genetic covariation, but you can apply that same kind of logic to the environmental covariation as well.

Simulation

https://www.parsingscience.org/wp-content/uploads/2020/09/ParsingScience082-Barbaro.mp3#t=11:33,13:34

Watkins: Nicole’s data set was not only nationally representative in terms of demographics, but with nearly 1000 sibling pairs responding over 28 years, it was also rather sizable. Nevertheless, it wasn’t large enough to yield sufficiently powerful statistical results. So she and her team used a technique called resampling. It involves repeatedly generating multiple smaller data sets for re-analysis by randomly selecting a subset of data, as she explains next.

Barbaro: So two basic pieces to the simulation are: one we’re going to have general mathematical equation that we’re going to solve, which we have here. So we’re trying to solve … the put that we’re trying to get is what proportion of the correlational relationship that we see between spanking and each of our outcomes, is due to shared genetic covariation? And that’s for one equation, and then the other is the same thing, but for the covariation of the non-shared environment. So we have a mathematical equation. And the kind of logic behind a simulation is quite simple. There’s four values in the equation. And we are basically calculating the answer to that equation thousands and thousands and thousands and thousands of times. Each time we’re calculating it, we’re pulling a random set of values from kind of like a distribution that we have for each of those four inputs. So the logic is quite simple: we’re just solving a relatively simple math equation. But we’re doing it tens of thousands of times with slightly different values each time to come up with a general estimate. So we did this calculation, I think millions of times for this paper overall, across all these different things, and all these different variables. And our goal is to get a general estimate, bracketed by what we call confidence intervals: so kind of like a range of plausible estimates. And that’s really the whole logic to it is just solving a math equation tens and tens of thousands of times. Of course, the computer does all the solving for us, so we’re not writing this all down ourselves.

ACE phenotypes

https://www.parsingscience.org/wp-content/uploads/2020/09/ParsingScience082-Barbaro.mp3#t=13:33,16:30

Leigh: Behavior genetics research begins by assessing the variants of a behavior in a large group – which geneticists call a phenotype – and attempts to estimate how much of this is due to three factors. First heritability, or the additive shared genetic effects, and abbreviated A. Second, how much is due to the shared or common environment, and abbreviated C. Third, how much is due to the unique or non-shared environment, abbreviated E. Up next, Nicole explains how these three ACE components played out in her study

Barbaro: When we think about kind of these ACE components … so A which is the additive genetic component C, which would be defined as our shared environment, and then E is the non-shared environment. Another way to understand the distinction between C and E. C is going to be systematic family effects is kind of the broadest way to understand that, whereas E is going to be kind of unsystematic individual ways in which children interact with their parents or peers, for instance. A lot of the times C is defined as like the family environment only and then E is everything else, which isn’t quite so accurate, because C is going to be those systematic effects that are typically in the family environment when we’re looking at developmental outcomes. And so, for spanking what a significant effective of C means – which we have in this model, and typical of other spanking models – is that there’s going to be a general kind of family level effects. So if you have kind of two siblings, there’s going to be some impact of that shared systematic environment of spanking that’s going to have some sort of impact on them. Whereas E, that effect is going to be more individualized. So it can be either the individual relationship that that child has with the parent, because of course, not every interaction or parenting is not exactly the same across children. So that’s going to capture some of that – and then also kind of their individual worlds, you know, with their friends or outside environments – that significant C effect doesn’t occur for a lot of variables. So as you see, you know, the delinquency depression, alcohol use don’t have that significant C effect, which is quite common, and behavior genetic models. Some traits do have that shared kind of family shared environmental experience, like we have for spanking. So that means that there’s going to be some similar effect for both children in that family based on that particular trait that we’re thinking of. So there could be a similar effect of spanking on outcomes for those kids. But generally that C components you can see is smaller than the additive genetic or the non-shared environmental component as well. And for a lot of traits that we do see an effective C, typically it is earlier in development, and those effects start to fade away as children develop and become more autonomous and move into adolescence and adulthood. Which is why most traits when we study them in adulthood typically don’t have any effects of C at all.

Long term negative impacts of spanking

https://www.parsingscience.org/wp-content/uploads/2020/09/ParsingScience082-Barbaro.mp3#t=16:29,19:19

Watkins: Nicole and her team simulation found that genetics accounts for only about 20% of the individual differences in [the] negative long term behavioral impacts of being spanked, with variations in the home environment that are unique to each sibling explaining only about 10%. The remainder of the effect – that is to say about 70% of the differences in whether spanking may cause long term negative impacts – is explained by aspects of parents’ home environment that all siblings share. But there was one exception, as she explained after this short break.

[ad: Altmetric]

Watkins: Here again is Nicole Barbaro.

Barbaro: There’s like 17 outcomes that we assessed in this particular simulation. And for almost all of those outcomes, we see that that shared genetic covariation is explaining most of the relationship. And then that non-shared environmental component is explaining the remainder of that relationship. Now there was the one key exception, which was being a victim of physical abuse. And that was explained heavily by the shared environmental component. It didn’t follow that same pattern. So the shared genetic covariation was not explaining that relationship very well at all. What we see for the effects of abuse and physical violence against children … across the board, there’s no debate in the literature that that is detrimental for child development and has negative impacts. And our results are very much in line with that. So it was encouraging from a methodological standpoint, because this simulation method that we’re using mathematically isn’t new, necessarily. It’s kind of like basic behavioral genetic math, but applying it to these developmental questions. It’s encouraging to see that it’s mapping on with that previous literature. So that shared environment – that shared environmental effect – is having a big impact in explaining the relationship between abuse and spanking, and abuse and of course, other negative developmental outcomes as well. So there’s no debate about the negative impact of physical abuse. So if you’re in a family that uses physical abuse, you know beyond traditional notions of spanking, then there’s going to be a negative impact, despite whether you or your sibling had more physical abuse or not … like that individual difference piece isn’t going to matter. Whereas just the mere fact of experiencing that physical abuse is going to have a big impact.

Non-shared environmental covariation

https://www.parsingscience.org/wp-content/uploads/2020/09/ParsingScience082-Barbaro.mp3#t=19:18,21:44

Leigh: In 2016, a meta-analysis of 50 years of research reported that spanking is consistently associated with poor developmental outcomes, appears to be largely ineffective at reducing problem behavior, and has not been shown to have any positive effects on child development. But of course, that was before Nicole’s paper, which found that genetic similarity explains a non trivial proportion of the association between spanking and psychosocial outcomes … with remaining proportional differences between people accounted for by non-shared environmental covariation. That is, by individual differences between siblings.

Barbaro: All the evidence does point to there being some degree of negative outcomes. My view based on the research would be that those negative outcomes, or the impact, is smaller than what we may have previously understood to be the case. But there have been some adoption study. So adoption studies are interesting because we’re taking a genetically unrelated child and putting them into a home with genetically unrelated parents, which allows us to kind of tease apart some of these genetic and environmental effects. What these adoption studies show is that for some, what they called “genetically at risk children,” so adopted children that may have had parents that were engaged in a lot of problem behaviors. So we all know that there’s individual differences among individuals for, you know, personality traits, and just kind of behaviors more generally. So children that have, let’s say, a family history of problem behaviors: even if you put those children into the “world’s best parents” household, they’re going to be a little bit more sensitive and have worse outcomes from harsh parenting than do children that are out what they call “genetically low risk,” or you know, children that are coming from families with no family history of kind of these problem behaviors. So we can see that the same type of parenting – so harsh spanking-type parenting – may have more negative outcomes for children that are already high risk for these problem behaviors. So it may be the case that some of the spanking effects that we are seeing may be driven by children that are already at high risk for engaging in negative behaviors, or they may not be. We don’t have you know definitively the answers on that. But there does seem to be the role of individual differences and kind of differential susceptibility to certain types of parenting practices, depending on the individual child.

Can correlation imply causation

https://www.parsingscience.org/wp-content/uploads/2020/09/ParsingScience082-Barbaro.mp3#t=21:43,24:25

Watkins: Most of us have heard the old saw that correlation does not imply causation. Nevertheless, Nicole’s study examined correlations, and yet reports that her team’s findings are, in fact, consistent with causal interpretations of the negative effects of spanking. So since it seems that correlation can imply causation, we asked Nicole to discuss the conditions under which this may be valid.

Barbaro: So correlation doesn’t necessarily imply causation, but things that are causal are likely going to have a correlation underlying them. This is kind of where in the paper we discuss very carefully that, “based on x, y and z evidence, this evidence is consistent with a causal explanation.” So that’s why we’re very careful. And I think it’s italicized on page five of the paper that this association between spanking adverse psychological outcomes is generally inferred to be consistent with causal explanations. And in this literature, there’s kind of this social science kind of requirements to say that “this line of evidence is consistent with a causal explanation” for research topics such as this where you can’t run an RCT and randomly assign spanking to children and then wait 10 years and see what happens … for obvious ethical reasons. So when we’re looking at these things, you want to see that 1) there is a correlation, a reliable, significant correlation between, you know, in this case spanking and these adverse psychosocial outcomes that we’re measuring. 2) you obviously need the inferred causal variable to happen before the outcomes. So spanking has to occur before those negative adverse outcomes, which developmentally speaking, if you get spanked when you’re five, and you’re measuring something when you’re 15, you know, we have that covered. And then point 3) which is a really important one, because you can define causality – theoretically speaking, you know, practically, it’s much harder, but theoretically speaking – as long as you control for all relevant confounds, you can make causal claims. Now, this is way easier said than done: to get rid of all the potentially confounding variables. And that’s really that point 3 is really where our research efforts were focused, kind of conceptually speaking. Because we know that there’s a correlation between spanking and negative outcomes. We know spanking happens first. But our argument was that “the genetic confounds have not been adequately explored across these different outcomes. Here’s why it’s important. And here’s our demonstration of why these need to be taken seriously because it does have a significant impact on the overall effect size of how spanking impacts negative outcomes.”

Is spanking ever worth it?

https://www.parsingscience.org/wp-content/uploads/2020/09/ParsingScience082-Barbaro.mp3#t=24:24,26:33

Leigh: Setting aside the long term psychosocial impacts of spanking, a large body of literature has shown that it’s also just not very effective in the short term in curbing children from acting out. And based on a research letter published this summer in the medical journal JAMA Pediatrics – which found that spanking has declined in the United States and 50% from 1993 to 35% in 2017 – it seems like Americans might be catching on to this fact. So, while it wasn’t central to our paper, we asked Nicole is spanking ever worth it?

Barbaro: So the goal of punishment, of course, is to reduce the likelihood that a child is going to engage whatever that behavior was, that problem behavior. Stealing a toy from another kid, for example: we want them to not do that. The effectiveness of spanking can be connected to a very basic psychological learning concept of positive and negative punishment. So when we’re spanking a child, we are adding a punishment. So this would be a positive form of punishment, rather than other forms of punishment, such as negative punishment, where we take something away. Timeouts are a really good example of this. So a child does something bad and you take away their ability to play or have, you know, the freedom to continue playing with whoever they’re playing with. So timeouts – when you take something away in this form of negative punishment, – seems to be a more effective way of disciplining a child than positive punishment. And some of the reasons I suspect that spanking – and this form of positive punishment – is not effective is typically “user error” is the problem Because it’s not administered “correctly.” That word sounds kind of weird to say that way. But it’s not administered in the way that would make it more effective. So if a child does something bad, then the positive punishment needs to come immediately after that poor behavior. But if there’s kind of intermittent, you know, discussion or other things that happen – or the punishment is delayed – it’s going to become less and less effective, because now it’s difficult for the child to associate the spanking with the particular problem behavior or the punishment with whatever that problem behavior is.

Including genetic covariation in meta-analyses

https://www.parsingscience.org/wp-content/uploads/2020/09/ParsingScience082-Barbaro.mp3#t=26:32,28:50

Watkins: One of Nicole’s recommendations was that meta-analyses could benefit from the inclusion of genetic covariation as a moderator variable … but that the genetically informed literature is not yet mature enough to support a meta-analysis. Doug and I asked Nicole to elaborate on why the literature might not yet be up to the task.

Barbaro: The whole goal of our paper was to understand how much of that association between spanking and any negative developmental outcomes is accounted for by this shared third variable, genetic covariation. Because if a substantial portion of it is, then the impact that we presume that’s spanking has on these negative outcomes is much smaller than what has typically been assumed and what policies are based on. So our goal was to do this simulation because there wasn’t enough literature to run a formal meta-analysis. So there’s approximately, I think, four maybe studies that kind of show this relationship using a genetically informed sample: so sibling sample or twin sample. So there’s only a few studies that have that available information. And four studies isn’t enough to run a meaningful meta-analysis. For meta-analysis, you want to have a lot of studies. Such as you know, like 100 studies that have looked at similar relationships with similar enough variables, so you can kind of lump them together and get an average effect. So this is kind of a little bit of like a chicken and egg problem that we’re encountering. Because our goal is to say, “these are our results, and this is showing that genetic covariation is a serious issue. It has an impact on the knowledge that we have now in terms of how much spanking impacts negative outcomes. And it’s important, and we need to research it.” But then on the other hand, there’s not enough research to demonstrate that it’s a problem. So therefore, why are you doing this work? So we’re kind of in like this chicken and egg problem of, you know, we’re trying to demonstrate there’s a problem. But it’s hard to demonstrate there’s a problem, because there’s not enough data to demonstrate that there’s a problem. So that’s where our simulation came in. And our simulation maps on quite closely with the few available studies that there are. Which gives us confidence that these numbers and these estimates that our simulation models have produced are useful: are a rough, useful average.

Learning from behavioral genetics

https://www.parsingscience.org/wp-content/uploads/2020/09/ParsingScience082-Barbaro.mp3#t=28:49,31:41

Leigh: To close out our conversation. Ryan and I were interested in learning Nicole’s thoughts on how other fields beyond behavioral genetics may benefit from considering genetics as a moderator of outcomes that are of interest to their disciplines.

Barbaro: I would say that pretty much any social science field that is interested in human behavior at a very broad level would have various uses for incorporating behavioral genetic frameworks, thinking, and models into their work. Because we know from decades of twin studies – there is a excellent meta-analysis in 2015 by Holderman – and it showed that – I think it was on average 49% of every trait; it was like 18,000 different human traits and behaviors – showed, on average a 49% heritability estimate. And that’s to say that genes matter, right? So when we’re thinking of human psychology and human behavior: genes matter, and they’re going to matter for a whole host of different reasons. And if we’ve learned anything in the last five years, I would say, is that the impact of genes on human behavior is way more complex and nuanced than what we thought, maybe 20 years ago when behavioral genetics was first, you know, coming up with these big replicable findings of heritability estimates and twin studies. And the kind of old school behavioral genetics – I call it an old school kind of behavioral genetics views – you know, genes matter, family environment doesn’t matter. And, you know, all these unsystematic chance effects are, you know, also driving individual differences. And it’s true to a very simplistic extent. But from what we’ve learned … there’s really great work in developmental behavioral genetics at showing really complex interplays between genes and other genes. And genes and environment. And evocative effects. And parenting effects. And there’s so much more nuance and complexity in the field of behavioral genetics over the last five years. It’s really become apparent that it seems like it was such a straightforward answer of you know, genes matter which they do. But how they matter – and the processes, and what we can do with that information – is much more complex than what we thought. So I think that pretty much any field that involves human psychology or human behavior, in general, would benefit from strongly incorporating behavioral genetic thinking and principles into their work.

[ Back to topics ]

Links to manuscript, bonus audio and other materials

Leigh: That was Nicole Barbaro discussing her study “The effects of spanking on psychosocial outcomes: revisiting genetic and environmental covariation,” which she developed with four co-authors and which they uploaded to the preprint server PsyArXiv on March 8, 2020. You’ll find a link to their paper at parsingscience.org/e82 along with transcripts, bonus audio clips, and other materials that we discussed during the episode.

Watkins: If you like what you’ve been hearing, then head over to Apple podcasts, Google podcasts, or wherever else you might get your podcasts, and subscribe to Parsing Science. And if you’ve already subscribed, consider leaving a review on iTunes. It’s a great way not only for others to learn about their show, but also a great way to help spread the word of the scientist on it.

Preview of episode 83

Leigh: Next time, in episode 83 of Parsing Science, we’ll talk once again with Manvir Singh from the Department of Human Evolutionary Biology at Harvard University about his research into why shaman healers among a group of people off the coast of Malaysia called the Mentawai, observe costly prohibitions – such as abstinence or food restrictions – especially given that they could exploit their position to devise self serving rules instead.

Manvir Singh: The seem to have this capacity to manipulate these taboos and self-serving ways. And yet they were really strikingly – apparently – disadvantaged by them. They’re prohibited from having sex all the time. They’re prohibited from eating these tasty foods. And so that was something very puzzling for me.

Leigh: We hope that you’ll join us again.

[ Back to topics ]