A common fallacy about a parent's room for influence
Tom Adams
Just want to point out a fallacy that I see a lot in the parenting literature. Here’s an example of it from a recent blog:
“The truth is that nature – a baby’s unique genetic make-up – accounts for somewhere between about a quarter and a half of the variation in infant nighttime sleep patterns.1,2”
“On the other hand, this leaves a lot of room for the influence of nurture – the way we care for our babies and the sleep environment that we create for them.”
So, the “room for influence of nuture” is supposed to be three-quarters and one half of the overall variation. Or, 1 to 3 times the variation due to genetics.
Now, if you look at the citations 1 and 2, you will see that these are studies that dole out the variation between genetic influences (“nature”) and environmental influences (“nurture”).
The blunder has to do with the implicit definition of “nurture” here. Nurture is defined as whatever is it that the broad population of parents just happen to be doing that has influence, as estimated from a random sample from this population.
So the story goes. Our “room for influence of nurture” is determined by whatever the population does. If the practices of a population became more varied such that the influence was evidenced in a more varied infant behavior, then our “room for influence” would become larger. If the practices of the population became more standardized, then our “room for influence” would become smaller – in principle our room for influence could shrink to nothing without any change in ourselves whatsoever if the rest of the population of parents simply started behaving exactly the way we do!
But, actually, the room for influence of nurture is not measured by this at all. It’s measured by the distance between a parent’s current practice and best practice, assuming the parent has the competence to achieve best practice. This is perhaps best measured in randomize controlled trials of practices. For instance, the randomized controlled trials used to measure the effect size of parent training courses using a wait-listed control group. And even then, you’d have to believe that the population sampled was representative of you.
I see this blunder a lot. The books The Nature Assumption and The Blank Slate are rife with it.
The simple fact is that the fraction of variation attributed to environmental influences puts absolutely no upper bound on whatsoever on the room for influence. The upper bound is limited by the best practices that have been discovered and validated as of now or may be discovered and validated in the future.
William Eden
Drilling down into more specific categories of interest, intellectual functions broadly were highly heritable at 67%, while more specific executive function metrics were 51% heritable with a high 24% shared environment contribution. Mood disorders were highly variable, from bipolar being 68% heritable to depressive episodes being 34% heritable. Height and weight showed 63% heritability, with relatively large 30% and 20% shared environment contributions respectively. The more specific values and social variables were mostly in line with the overall findings. Tendency towards religion and spirituality was 31% heritable with an even larger 35% shared environment component. Basic interpersonal interactions were similar, with 30% heritability but 36% determined by shared environment.
To summarize, basic variables in terms of intelligence, height, and weight are primarily determined by the genetic contributions. Most health and psychiatric outcomes fell somewhere in the middle, but still showed roughly half of variance explained by genetics. Variables relating to fundamental values (e.g. religion, politics) and social interactions (e.g. emotional intelligence, relationships) were by far the most malleable traits, with roughly equal contributions from genetics and shared environment. Nonetheless, even a 30% contribution from genetics is still a large effect in an absolute sense. Depending on how much you care about sharing your values with your children, relative to other features like long-run health outcomes or basic biometric markers, you may wish to elevate these criteria in your consideration.
--
You received this message because you are subscribed to the Google Groups "Less Wrong Parents" group.
To unsubscribe from this group and stop receiving emails from it, send an email to less-wrong-pare...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
Tom Adams
Tom Adams
On Wednesday, March 9, 2016 at 7:58:20 AM UTC-5, Tom Adams wrote:
Consider the Flynn Effect or something similar. The mean of a human capability has increased over time at a rate R. Population studies by behavioral geneticist have shown that a proportion (P) of variability of this capability is due to genetic influences.
Does the value of P change as the mean capability increases at rate R? Is it not the case that P could go up or down or not change at all as the mean capability increases? Is R somehow linked to P? Does the measured value of P put a limit on R? Does the measured value of P put a limit on how much the mean capability can possibly change?
Robin Lee Powell
On Tue, Mar 08, 2016 at 11:10:52AM -0800, Tom Adams wrote:
>
> So the story goes. Our “room for influence of nurture” is determined by
> whatever the population does. If the practices of a population became more
> varied such that the influence was evidenced in a more varied infant
> behavior, then our “room for influence” would become larger.
the amount of daylight in your country, are very unlikely to count
for this measurement.
I had assumed that such studies worked the *other* way, by measuring
the influence due to genetics (i.e. the statistical variances
between twins/siblings) and then calling the rest "nurture".
Working this out by looking at variance across a general population
and calling that "nurture" is just ... 0.o Do studies really do
that? It's *obviously* stupid.
Robin Lee Powell
On Tue, Mar 08, 2016 at 11:28:09AM -0800, William Eden wrote:
> Generally the typical heritability study differentiates into (additive and
> non-additive) genetic, shared and non-shared environmental factors. Along
> most adult attributes, it's found to be roughly half genetic and half
> non-shared environmental factors explaining the variance, with very little
> shared environment. The effect of parenting is commonly equated to the
> shared environmental factors, but I think this is incorrect. I write about
> that here, including discussion of a study looking at parenting in the
> non-shared environmental component:
> http://becomingeden.com/parenting-and-the-non-shared-environment/
average of the two of us would be someone I very much want to
exist." :D, even though that's kind of off-topic.
"Intuitively, it seems crazy to me that we might not have a major
impact on our kids." -- I know, right?
"Realizing that parents do treat their kids differently, and that
this produces different results, along with the combination of a
huge number of individual life events, is my main takeaway from this
paper." -- I definitely treat my kids differently, but I think in
ways that are largely appropriate/positive/responsive to actual
differences between them, but this is something I intend to actively
check in with them about as they get older.
Tom Adams
Tom Adams
Tom Adams
Robin Lee Powell
purely theoretical populations.
Saying that "variance across a general population" is exactly the
same as "the part that's not due to genetics" implies that every
*possible* (or at least "available") environmental/parenting choice
is present in the population.
Please note that this is not an area I'm strongly expert in; I could
be misunderstanding profoundly or explaining poorly. But it seems
obvious to me that any given study population cannot possibly cover
all *possible* environmental influences, so the only way to see how
much is genetics and how much is environment is to figure out how
much is genetics and (metaphorically) subtract. Which probably
still won't really work, but at least seems marginally plausible.
Tom Adams
William Eden
Sent from my iPhone, enjoy the unusual brevity.
--
Tom Adams
On Tuesday, March 8, 2016 at 2:28:29 PM UTC-5, William Eden wrote:
Generally the typical heritability study differentiates into (additive and non-additive) genetic, shared and non-shared environmental factors. Along most adult attributes, it's found to be roughly half genetic and half non-shared environmental factors explaining the variance, with very little shared environment. The effect of parenting is commonly equated to the shared environmental factors, but I think this is incorrect. I write about that here, including discussion of a study looking at parenting in the non-shared environmental component: http://becomingeden.com/parenting-and-the-non-shared-environment/I do generally want to be cautious about rationalizing away the research literature finding minimal effects of parenting. I have a variety of inside view reasons, as well as methodological reasons, to be concerned about those results, but it is a very strongly and consistently reported finding.My biggest issue with heritability studies is that they're only accounting for explanation of variance within a given population. These studies are almost never done across country lines for example. Technically growing up in the US vs the third world should show up in "shared environment", and yet that variance is never even observed. So what can these studies really tell us? That conditional on finding yourself in a given, relatively homogenous population, your kids are mostly influenced by half genetics and half idiosyncratic factors that are difficult to measure, some of which are probably parenting. I would consider parenting effects on measured outcomes to be small, but non-negligible. I also think there are unmeasured things that matter a great deal.And of course you can always try to claim that you're doing something so radically different that your variance is never captured in existing twin studies, which may well be a fair point, depending on what you're doing.In terms of heritability estimates, some things seem much more amenable to "shared environment" factors than others, and it mostly conforms to common-sense expectations. I will excerpt from an unpublished report I wrote on the subject:"An extremely ambitious meta-analysis of all twin studies was published in May 2015, reporting heritability estimates from 2,748 studies featuring over 2 million twin pairs, encompassing virtually every published study to date.[1][2] Across very broad domains of health outcomes, almost everything falls within the 40-60% heritability range, with cancer as a representative example being 46% heritable. Similarly, neurological variables show about 50% heritability (with little shared environment involvement), while cognitive and psychiatric outcomes are similarly heritable, but also have a nearly 20% shared environment component. Social values appeared to be 31% heritable, but shared environment played nearly as big a role at 27% heritability. Similarly, social interactions were 32% heritable, with a somewhat smaller shared environment component of 18%.Drilling down into more specific categories of interest, intellectual functions broadly were highly heritable at 67%, while more specific executive function metrics were 51% heritable with a high 24% shared environment contribution. Mood disorders were highly variable, from bipolar being 68% heritable to depressive episodes being 34% heritable. Height and weight showed 63% heritability, with relatively large 30% and 20% shared environment contributions respectively. The more specific values and social variables were mostly in line with the overall findings. Tendency towards religion and spirituality was 31% heritable with an even larger 35% shared environment component. Basic interpersonal interactions were similar, with 30% heritability but 36% determined by shared environment.
To summarize, basic variables in terms of intelligence, height, and weight are primarily determined by the genetic contributions. Most health and psychiatric outcomes fell somewhere in the middle, but still showed roughly half of variance explained by genetics. Variables relating to fundamental values (e.g. religion, politics) and social interactions (e.g. emotional intelligence, relationships) were by far the most malleable traits, with roughly equal contributions from genetics and shared environment. Nonetheless, even a 30% contribution from genetics is still a large effect in an absolute sense. Depending on how much you care about sharing your values with your children, relative to other features like long-run health outcomes or basic biometric markers, you may wish to elevate these criteria in your consideration.
[2] They published a data viewer to experiment with: http://match.ctglab.nl/#/multiple/reported_ace"
On Tue, Mar 8, 2016 at 11:10 AM, Tom Adams <tada...@gmail.com> wrote:
Just want to point out a fallacy that I see a lot in the parenting literature. Here’s an example of it from a recent blog:
“The truth is that nature – a baby’s unique genetic make-up – accounts for somewhere between about a quarter and a half of the variation in infant nighttime sleep patterns.1,2”
“On the other hand, this leaves a lot of room for the influence of nurture – the way we care for our babies and the sleep environment that we create for them.”
So, the “room for influence of nuture” is supposed to be three-quarters and one half of the overall variation. Or, 1 to 3 times the variation due to genetics.
Now, if you look at the citations 1 and 2, you will see that these are studies that dole out the variation between genetic influences (“nature”) and environmental influences (“nurture”).
The blunder has to do with the implicit definition of “nurture” here. Nurture is defined as whatever is it that the broad population of parents just happen to be doing that has influence, as estimated from a random sample from this population.
So the story goes. Our “room for influence of nurture” is determined by whatever the population does. If the practices of a population became more varied such that the influence was evidenced in a more varied infant behavior, then our “room for influence” would become larger. If the practices of the population became more standardized, then our “room for influence” would become smaller – in principle our room for influence could shrink to nothing without any change in ourselves whatsoever if the rest of the population of parents simply started behaving exactly the way we do!
But, actually, the room for influence of nurture is not measured by this at all. It’s measured by the distance between a parent’s current practice and best practice, assuming the parent has the competence to achieve best practice. This is perhaps best measured in randomize controlled trials of practices. For instance, the randomized controlled trials used to measure the effect size of parent training courses using a wait-listed control group. And even then, you’d have to believe that the population sampled was representative of you.
I see this blunder a lot. The books The Nature Assumption and The Blank Slate are rife with it.
The simple fact is that the fraction of variation attributed to environmental influences puts absolutely no upper bound on whatsoever on the room for influence. The upper bound is limited by the best practices that have been discovered and validated as of now or may be discovered and validated in the future.
--
You received this message because you are subscribed to the Google Groups "Less Wrong Parents" group.
To unsubscribe from this group and stop receiving emails from it, send an email to less-wrong-parents+unsub...@googlegroups.com.
Gunnar Zarncke
- researching interventions at the level of their competence
- changing country or place of residence
In a way if you accept that you can't achieve a change and refuse to use the best avialble methods then that might mean that you underperform.
Gunnar
To unsubscribe from this group and stop receiving emails from it, send an email to less-wrong-pare...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
Tom Adams
On Wednesday, April 6, 2016 at 9:52:41 AM UTC-4, Tom Adams wrote:
Consider the case of PKU. The recessive gene is clearly heritable. But the intellectual disability phenotype can be prevented via diet. It was not practical to prevent it until after 1960 when a newborn screening test and a low-PHE baby formula were developed. There was probably little or no shared environment effect on the phenotype before 1960. The shared environment component probably peaked in the US during the spotty introduction of screening over a decade or so. The shared environment component probably went down to near zero again after the screening was mandated in the entire US. Note that these shared environment effect measurements are not like measuring some physical constant like the mass of an electron at rest (assuming that is a constant).I don't think it's rationalizing to simply state that the behavioral genetics research literature strong and consistent findings of minimal effects of parenting completely ignore the potential development of interventions and even it's lack of power to detect existing interventions in some cases.This interesting thing about IQ is that it's been fairly resistant to interventions so far. One has to look at the interventions or lack of interventions, the behavior genetics findings of minimal effects are not enough to discount the potential for influence. And even if there is a large shared environmental effect, one would still need turn that into an intervention for it to be of any value for a parent's goal of having influence.When people take the behavior genetics measurements too seriously, I think they are forgetting about all sorts of implicit assumptions about interventions and the lack of interventions. The behavioral genetics don't imply interventions are impossible. It just seems that way because they are interpreted against this implicit background information about interventions that is taken for granted.What do you mean by "common sense expectations"?Some of the interesting interventions like the low-PHE diet and the Lovaas method operate in a critical developmental window. Kids needs something within the window and there is no catching up later so the childhood intervention has a big long-term effect. We are learning more about how developmental windows relate to IQ:(not to say that this necessarily implies an intervention to increase IQ)
On Tuesday, March 8, 2016 at 2:28:29 PM UTC-5, William Eden wrote:
Generally the typical heritability study differentiates into (additive and non-additive) genetic, shared and non-shared environmental factors. Along most adult attributes, it's found to be roughly half genetic and half non-shared environmental factors explaining the variance, with very little shared environment. The effect of parenting is commonly equated to the shared environmental factors, but I think this is incorrect. I write about that here, including discussion of a study looking at parenting in the non-shared environmental component: http://becomingeden.com/parenting-and-the-non-shared-environment/I do generally want to be cautious about rationalizing away the research literature finding minimal effects of parenting. I have a variety of inside view reasons, as well as methodological reasons, to be concerned about those results, but it is a very strongly and consistently reported finding.My biggest issue with heritability studies is that they're only accounting for explanation of variance within a given population. These studies are almost never done across country lines for example. Technically growing up in the US vs the third world should show up in "shared environment", and yet that variance is never even observed. So what can these studies really tell us? That conditional on finding yourself in a given, relatively homogenous population, your kids are mostly influenced by half genetics and half idiosyncratic factors that are difficult to measure, some of which are probably parenting. I would consider parenting effects on measured outcomes to be small, but non-negligible. I also think there are unmeasured things that matter a great deal.And of course you can always try to claim that you're doing something so radically different that your variance is never captured in existing twin studies, which may well be a fair point, depending on what you're doing.In terms of heritability estimates, some things seem much more amenable to "shared environment" factors than others, and it mostly conforms to common-sense expectations. I will excerpt from an unpublished report I wrote on the subject:"An extremely ambitious meta-analysis of all twin studies was published in May 2015, reporting heritability estimates from 2,748 studies featuring over 2 million twin pairs, encompassing virtually every published study to date.[1][2] Across very broad domains of health outcomes, almost everything falls within the 40-60% heritability range, with cancer as a representative example being 46% heritable. Similarly, neurological variables show about 50% heritability (with little shared environment involvement), while cognitive and psychiatric outcomes are similarly heritable, but also have a nearly 20% shared environment component. Social values appeared to be 31% heritable, but shared environment played nearly as big a role at 27% heritability. Similarly, social interactions were 32% heritable, with a somewhat smaller shared environment component of 18%.Drilling down into more specific categories of interest, intellectual functions broadly were highly heritable at 67%, while more specific executive function metrics were 51% heritable with a high 24% shared environment contribution. Mood disorders were highly variable, from bipolar being 68% heritable to depressive episodes being 34% heritable. Height and weight showed 63% heritability, with relatively large 30% and 20% shared environment contributions respectively. The more specific values and social variables were mostly in line with the overall findings. Tendency towards religion and spirituality was 31% heritable with an even larger 35% shared environment component. Basic interpersonal interactions were similar, with 30% heritability but 36% determined by shared environment.
To summarize, basic variables in terms of intelligence, height, and weight are primarily determined by the genetic contributions. Most health and psychiatric outcomes fell somewhere in the middle, but still showed roughly half of variance explained by genetics. Variables relating to fundamental values (e.g. religion, politics) and social interactions (e.g. emotional intelligence, relationships) were by far the most malleable traits, with roughly equal contributions from genetics and shared environment. Nonetheless, even a 30% contribution from genetics is still a large effect in an absolute sense. Depending on how much you care about sharing your values with your children, relative to other features like long-run health outcomes or basic biometric markers, you may wish to elevate these criteria in your consideration.
[2] They published a data viewer to experiment with: http://match.ctglab.nl/#/multiple/reported_ace"