Monday, February 21, 2011

Nature vs. Nurture

The question of “nature vs. nurture” has long fascinated me. Specifically, I have often tried to understand the extent to which nature was the ‘cause’ of the way a person is, and how much was left to nurture (or, more generally, the environment in which a person grows up and lives). It is only recently that I have adjusted my tentative answers to these questions quite a bit.

About 15 years ago I adopted the general tenet that "genetics provides a range and one's environment determines where one lands in the range,” which served me well. I never had any better way of describing the nature vs. nurture issue, nor did I think I needed one. Recently, though, stimulated by some stuff I learned in a lecture about neuroscience from a Cal professor named Hinshaw, I think I have come up with a new, and possibly more powerful, model for thinking about this issue.

Basically, I started with several revelations and new (new to me, not necessarily to science) pieces of info:

1) Imagine a peanut allergy that is completely determined by a single, simple gene. If you have the gene, you have the allergy. If you don't have the gene, you don't have the allergy. On the surface, the allergy seems to be 100% genetic in origin, and in some sense it is. But, in another sense, it's not. The reason I say that is that if you grow up in a world where there are no peanuts, you may think you have no peanut allergy, even if you have the gene. You'd never KNOW you had the allergy, and there wouldn't even be a test for it, because no one would know to look for it. So, in the case of someone with the allergy, it's "100% genetic" in origin (this is equivalent to saying that genes give you a 100% chance of having the allergy, with a range of +/- 0% determined by your environment). But, if you look only at outcomes, and you have a person who doesn't show the allergy, you could make a reasonable argument that in some sense, the outcome is 100% explained by environment, even though the allergy is also, in the previous sense, purely genetic in origin.

2) If you imagine a baby who is born with a healthy brain and body, but then locked in a box with feeding tubes and absolutely no sensory input, that baby's IQ after several years will probably be exceptionally low (perhaps below 35, where 100 is average). The "range" for this person, with a "normal" life, in terms of IQ, might have been 85-115, but the "range" considering these extreme conditions might be as wide as 15-100. There is nothing about this that makes the "range" invalid, but it appears to make it somewhat less useful, because it would seem there are multiple ranges, for normal vs. abnormal or rare scenarios. One might easily make a slight modification to the "range" model to try to account for this. We might call it the "additive" model. This additive model might be one where you get a "genetic IQ component" of [50 +/- 50] and an "environmental IQ component" of [50 +/- 50]. You could then add an "extreme cases" component that contributed [0 +/- 50]. In fact, scientists have sort of proposed this when they have broken down the "nurture" factors into "normal" and "abnormal" conditions. Still, though, this model seems unsatisfactory for explaining the sensory deprivation condition.

3) Many studies are done on twins to try to determine the nature vs. nurture influence. Identical twins obviously have the same DNA, but may have very different "environments" in which they live their lives. One classic study looks at the variation in height between twins and non-twins, as well as between twins raised in the "same" environment vs. twins in "different" environments. It tries to apportion the observed variation to either genes or environment. Such studies usually show something like, "90% of height variation is explained by genetics and 10% by environment." On the surface, this seems completely intuitive and reasonable. Further, you might take the extreme example of someone who had their legs amputated as a child. This "environmental" factor may cause an extreme outlier, similar to the example from #2, but other than that, the "range" model seems to be very consistent with the findings of the actual studies on twins. This is where I had a sort of epiphany.

I assert that what most people *think* the twin results show is actually only 1 of 2 possible interpretations of studies of this type. First, let me tell you what I *thought* it must mean which is what I think most people assume it means. They assume that the twin studies on height mean that, "height is 90% determined by genetics and 10% determined by environment." Indeed, the study could be showing that. It could also, however, be showing something very different. It could be showing that our environments are 9 times as homogeneous as our genes. I'll explain what I mean by this through an example. Imagine that we could, with a magic wand, make every baby born tomorrow have exactly the same DNA. If we could do this, and then we measured the heights of the same babies 30 years later, 100% of variation in height would now be attributable to environment, not to genetics. Of course, the range (or, more precisely, the standard deviation) of this variance might be much smaller. In fact, comparing the standard deviation of the variance between babies born on a day with "the same genes," to the general population might yield some sort of actual, meaningful quantitative measure. This is exactly how the twin studies are supposed to work. Twins are this "comparison group." The problem with these studies is "what does 100% environmental" mean? What if we could, tomorrow, make all babies born somehow magically grow up in "the exact same environment." First, this is difficult to even conceptualize. They can't be "the same" because they can't all live in the same bedroom in the same house, even if we could clone their parents and make exact copies of the weather patterns and whatnot. Another way of saying the same thing is:

If, today, genes account for 90% of variation in height, and then next year we do a magic experiment where we make the environments "less similar" for children growing up, we might suddenly find that now genes only account for 80% of variation in height.

In fact, studies confirm this hypothesis as well. If you break down the twin studies of height, in particular, they appear to show approximately a 95% genetic component for height amongst the higher socioeconomic classes, and only about an 88% genetic component for those in the lower classes. This might be precisely because the environments for the higher socioeconomic classes are more homogeneous (more similar food, medical care, or activities).

So, saying that 90% of variation is due to nature could ALSO be interpreted as "the environment is far less homogeneous than our genes." The studies that are done aren't wrong - they do precisely what they purport to do; they attribute variation to its sources. What I thought that meant, but what it does not necessarily mean, is that "90% of height is caused by one’s genes."

This brings me to the formulation of a new mental model for this topic. I think that if you imagine a coloring book, like children draw in, with black outlines on a paper but color to be added by the artist, this is perhaps a good metaphor for the nature vs. nurture question. If the lines on the paper are heavy, thick, black, and close together (perhaps even forming a solid block of black ink), there may be no way to use a crayon to ever create a pink flower on that part of the paper. The best you may be able to do is a dark-brown flower in a black box. On the other hand, if you imagine smearing heavy paint over the whole page, you might be able to nearly completely obscure thin outlines beneath and change the paper to anything except a pure white color.

The metaphorical extension of this coloring book model is then to talk about the differences between people in terms of "Do they have the same picture, with slight differences in the lines?" "Do they have different pictures outlined but colored similarly?" "Do they have some condition so extreme (red paint splashed across the page) that you can't even make out the underlying outlines (mental illness, severe trauma, etc...)?"

This model is sort of like the "range" model in the general sense that, under normal circumstances, the broad strokes of the final picture are determined by the outlines, with the details added by the color. These details, while small in some ways, can radically change the net "impact" of the picture, and they can be harmonious or discordant. I think, though, that this model/metaphor may be more useful for explaining extreme or "edge" cases and outliers. My next step would be to try to create a mathematical equivalent to this model. I don't pretend that I can quantify the actual impact of nature vs. nurture (the quantification of which may not even be a sensible concept), but I do hope that I can show the relationships between factors.

In the height example, if we imagine an overly simple world where height is ONLY determined by genes and nutrition, I could imagine a 2nd (or higher)-order polynomial that would account for the extreme range of possibilities (3 ft - 10 ft), but would also, under normal circumstances, allow the "magnitude" or "contribution" of the genetic variable to dominate (say, in a ratio of 9:1) the nutrition variable. This would require that the two variables interacted either by multiplication or by raising one to the power of the other. This would distinguish this model from the "additive" model, which I think is basically the same as the range model, which would allow the terms to interact with each other only by addition and subtraction. The difference is between a model that might be "x- squared plus 1/2 y-squared" and a model that might be more like "2 times x + 1/2 x times y." Another way to think of this mathematical difference is to imagine what happens when one factor is set to 0. In the additive model, this can't completely overwhelm the other factor. In this multiplicative model, it can. I'm not sure which one is more "right" and rightness might only be defined by which is more useful in practice, but I propose this new model as a useful one, especially for trying to unify the 'range' of possibilities in 'normal' people with the prevalence of extreme factors, mental illness, and the like.

So, how useful is this model? Personally, I would say it's moderately more useful than the "range" model. I'll give an example where it seems no more useful and one where it seems more useful:

No More Useful: Taking something more complex, like cancer for example, rather than say, "I'm genetically 2x as likely to get cancer as another person but then I influence that by living a healthy lifestyle," we might instead say something like "I have the risk factors and precursors for several types of cancer, but my lifestyle choices far outweigh those factors" or "Despite living a healthy life, I haven't been able to control the several genetic cancer risk factors I inherited." Of course, these are very subjective statements, and it would be useful to try to quantify, at least roughly, how much impact lifestyle has compared to genetics. Perhaps such a set of equations could be derived. This would require the same kinds of twin studies done now, but it would also require some sort of quantifiable measurement of "environmental homogeneity." Further, it would require comparing these things over time (since we don't know what to use as a baseline for "sameness" of environment). Such studies would be both longitudinal and latitudinal. In education and medical research, for example, they have been extremely hard to do. In this case, the two models seem to describe the situation equally well.

More Useful: Taking sexual orientation, I think the coloring book model may be more useful than the range model. I have found little use in trying to ascertain an underlying "range" of homo-vs-heterosexuality for an individual and then adding in a set of influences that occur after birth (both internal thoughts and external influences). I think the reason for this difficulty is because it tries to place homo- and heterosexuality as opposite poles on one axis (I think this has been a common attempt since Kinsey made it in "Sexual Behavior in the Human Male" around 1950. This may be the problem. A homosexual and a heterosexual may be very "similar" in some dimensions, either genetically, environmentally, or both, and very different in other dimensions. Thus, the "range" model seems unable to capture the complexity (specifically, the number of 'dimensions') whereas the coloring book model is inherently two-dimensional, better capturing this multi-dimensional possibility (although the dimensionality of the determinants of sexuality may be large).

I believe that a multiplicative model (or similar), which assumes that every factor may be both minor or conclusive, and allows for complex interactions between factors is a better mental construct for addressing such issues as “What makes a person homosexual, nature or nurture?” or “Why did one child turn out ‘smarter’ than another?”

No comments: