The Genius in All of Us

What does "heritable" mean? (You won't believe it when I tell you.)

What does "heritable" mean, in the context of genetics? If you've casually followed genetics stories over the last few years, you've surely seen the term, and probably assumed that it refers to the extent to which a particular trait is genetically inherited.

For instance, here are a few recent uses in the New York Times:

"Studies of twins show that homosexuality, especially among men, is quite heritable, meaning there is a genetic component to it." (4/10/07)

"Since personality is heritable, this would be a mechanism for Yanomamo nature to evolve and become fiercer than usual." (3/12/06)

"A genetic analysis using tissue samples from 185 dolphins, 13 of them spongers, showed that it was highly unlikely that sponging was a heritable trait." (6/7/05)

So why does noted biologist and geneticist Tim Tully tell author Matt Ridley: "I can prove in a court of law that heritability has nothing to do with biology"?

And how is it that the heritability of the number-of-fingers on the human hand is close to zero?

And how can height be 90% heritable and yet subject to extraordinary influence from nutrition?

And how can I.Q. be 50-60% heritable in some well-designed studies and almost 0% in other, equally valid studies?

Welcome to the truly bizarre world of heritability and its misuses. If you're a non-scientist like me, you will be stunned to learn the word's actual definition, and its implications. In fact, a close look at how the term is misused by journalists and scientists alike goes a long way to helping us understand why the public still has such a profound misunderstanding of genetics and human development.

One thing is clear enough, to me anyway: the terms "heritable" and "heritability" should never, ever be used in a general interest publication without extensive context.

So -- here now the definition, from Merriam-Webster:

***
heritability: the proportion of observed variation in a particular trait (as height) that can be attributed to inherited genetic factors in contrast to environmental ones.
***

Let's break that down:

"the proportion"
-- a heritability estimate comes from a statistical analysis of a given population, not the results of a biological investigation (that, of course, doesn't mean it's invalid; but it does convey some important limitations); further, heritability can only be estimated, never actually known -- in fact, the "proportion" heritability seeks to estimate is ultimately unknowable, for reasons I will explain below;

"of observed variation"
-- heritability cannot actually look directly at what causes a trait (partly because every complex trait has multiple causes); it can only look at the amount of variation in that trait in a particular population. As Matt Ridley explains, heritability studies come from "measuring how similar identical twins are, how different fraternals are, and how both identicals and fraternals turn out if separately adopted into different families."

-- since the observed variation is going to differ from population to population, a heritability estimate is only relevant to the particular population studied. "It is strictly a property of a particular population," University of Arizona's Bruce Walsh explains. "Different populations, even if closely related, can have very different heritabilities."

    And, adds Matt Ridley, heritability is "meaningless for any individual person." This is critical to keep in mind.

"in a particular trait that can be attributed to inherited genetic factors in contrast to environmental ones."
-- What traits are we talking about? Not the basic Mendelian stuff like eye and skin color, obviously, but  the really complex stuff like sexuality, aspects of intelligence, personality factors, etc. How are those traits formed, to the best of our understanding? From an extremely complex combination of, and interaction between, genes and the environment. The actual equation used by behavioral geneticists is: V(p) = V(a) + V(d) + V(i) + V(e) + V(g X e) + COV(G,E). In layman's terms, there are four basic variables.

1. One or more separate genes, some of them interacting with one another (taking height, for example: hypothetically, you might inherit four different genes that help determine your height, two of whom are completely independent actors, and the other two which also have independent influence AND which will interact with one another to effectively produce a unique fifth genetic influence).

     plus

2. Environment (your in-utero and post-utero nutrition will have a direct and significant effect upon your adult height)

     plus

3. Gene-environment interactions (some of your nutritional experience will have a specific interaction with your particular genes -- with a *different* outcome than the interaction my genes might have with the exact same nutrition)

     plus

4. Gene-environment covariation (these are genetic influences on environment; for example, let's say from early on you are a particularly active baby, and as a result you get placed in a daycare group with other particularly active babies; that group happens to get a slightly different snack chosen by a different teacher, and that affects your nutrition, which effects your height. In effect, you have some non-height genes having had unexpected impact on your environment, which in turn is going to have an effect on your height. This indirect but very real phenomenon is called "covariation).

It actually gets more complex than that, but that's a fair summary. What this means, of course, is that for any such traits the nature vs. nurture paradigm simply DOES NOT EXIST. It's a false choice, like saying a meal either comes from the ingredients or the cook. Instead, the development process is continuously and inseparably nature-and-nurture.

If nature and nurture are inextricably intertwined, how can we determine what portion of a certain trait is due to genetics? We can't. We can only come up with crude averages from population studies. As a matter of biology, such a proportion does not exist. Which is why the Human Genome Project website calls heritability "a statistical construct," and warns: "heritability statements provide no basis for predictions about the expression of the trait in question in any given individual."

Heritability is not a useless measure, by any means. It has wide applications in genetics and agriculture. But in the context of conveying scientific understanding to a general audience, it is inherently misleading.

Executive function: critical and trainable

Scientific American's website reports on a new study demonstrating that IQ is far less relevant to math success than two specific brain skills -- working memory and inhibitory control, which are part of the suite of skills known as executive function.

Regular followers of this blog won't be surprised to learn that these skills are trainable. 

"It's often thought that kids don't do well because they're dumb, and there's nothing we can do about it," the site quotes Penn State professor and lead study author Clancy Blair. "But not only is executive function pivotal for academic success, it's amenable to training, and this training might make a big difference in a child's ability."

There's also a clear educational mandate here: "Preschool curricula that focus on development of these skills and self-regulation are needed in a big way," Blair says. "There is a federal push to learn our numbers, our letters and our words, but a focus on the content, without a focus on the skills required to use that content, will end up with children being left behind."

Kurt Vonnegut

The great Kurt Vonnegut died on Wednesday, and for a moment the earth stopped rotating. To me, he was an inspiration for his brutal honesty, intense skepticism and willingness to think and write utterly strange things. He seemed to be constantly in pursuit of a particular spectrum of truth that only he could see. When I reflect on the aspirations I have for my children, high on the list is that willingness to be brazenly your own self. If that isn't one of the key ingredients of greatness, I don't know what is.

I was fortunate to briefly visit with Vonnegut in his home almost 20 years ago, as I was just beginning my writing career. Here is the published interview that came out of it.

Is IQ actually AQ? (Mistaking Achievement for "Intelligence")

Have we been mistaking achievement for "intelligence"?

I've just gotten my hands on a copy of Andrew Elliott and Carol Dweck's mammoth Handbook of Competence and Motivation. Following the lead chapter from the editors is an utterly fascinating contribution from Yale psychologist Robert Sternberg, who, in just a few pages, seems to completely shatter the popular myth of I.Q. and intelligence testing. Being an academic text, the writing is a little dry; still, it may be the most important thing I've yet read on the subject.

According to I.Q. advocates and to popular understanding, intelligence tests are able to discern each individual's raw, natural intelligence -- which academic psychologists refer to as g for "general intelligence." So-called g is supposedly an innate, unchanging cluster of intellectual abilities that each of us simply possesses -- it is the hand we're dealt. This pure intelligence is not what we've learned, but simply how well our brains work. Furthermore, it seems to correlate so well with later job performance and life success, people have come to believe that each person has a specific amount of inherited intelligence that truly drives his/her level of success.

To reinforce the idea of pure intelligence tests, the testing community has gone to great efforts to distinguish between these so-called "ability tests," which reveal our innate intelligence, and "achievement tests," which examine the knowledge and skills we've been able to develop.

But what if those distinctions simply don't exist? What if every intelligence test measures a certain combination of skills and knowledge, revealing only what we've learned up to that point in our lives? And what if this correlation is more of a mirage than a true indication of cause-and-effect?

These are Sternberg's staggering -- and yet rational -- claims. "There is no qualitative distinction between various kinds of assessments," writes Sternberg. "The main thing that distinguishes ability tests from achievement tests is not the tests themselves, but rather how psychologists, educators, and others interpret the scores on these tests."

"Conventional tests of intelligence and related abilities," he says, "measure achievement."

Furthermore: "These skills develop as results of gene-environment covariation and interaction. If we wish to call them intelligence, that is certainly fine, so long as we recognize that what we are calling intelligence is a form of development competencies that can lead to expertise."  [We will discuss gene-environment covariation in another post].

In other words, it's not at all fine, because that is not at all how we use the word "intelligence." Intelligence is defined in the dictionary and in popular understanding as "the ability to acquire and apply knowledge" -- our natural ability. Distinct from knowledge and learned skills, it is what is built-in to our brains.

Sternberg argues that no current tests actually measure such built-in intelligence, and that intelligence testers are instead relying on a dangerous circular logic:

"Some intelligence theorists point to the stability of the alleged general (g) factor of human intelligence as evidence for the existence of some kind of stable and overriding structure of human intelligence. But . . . [w]ith different forms of schooling, g could be made either stronger or weaker. In effect, Western forms and related forms of schooling may, in part, create the g phenomenon by providing a kind of schooling that teaches in conjuction the various kinds of skills measure by tests of intellectual abilities."

In other words: we are teaching certain skills in our schools -- skills which do correlate reasonably well with Western job performance -- and then measuring how well kids learn these skill. Then we pretend that the results reveal a person's raw intelligence, when all they actually reveal is how well a child learned those skills. All we're really learning from intelligence tests is that some kids do better than others in school. We are not, as intelligence testers claim, uncovering the innate cause of these differences.

Is Sternberg saying there's no such thing as innate intelligence?
No. But he is saying that such intelligence is "not directly measurable," that it is not one general ability which can be scored, and that it is not inherently limiting. The evidence shows that skills and abilities are inextricably interwoven, and that all skills are modifiable. "The main constraint in achieving expertise is not some fixed prior level of capacity, but purposeful engagement involving direct instruction, active participation, role modeling, and reward."

What about the famous correlation between intelligence test scores on the one hand and job performance/life success on the other?
It's a mirage. The correlation does exist, says Sternberg, but not because one causes the other; rather, it's because they both measure the same abilities. Or as Sternberg puts it:
     "Such correlations represent no intrinsic relation between intelligence and other kinds of performance, but rather overlap in the kinds of competencies needed to perform well under difference kinds of circumstances. The greater the overlap in skills, in general, the higher the correlations."

Sternberg then points to a series of studies demonstrating that practical expertise does not correlate well with analytical ("intelligence") tests but do correlate very nicely with job performance and life success.
-- The Yup'ik Eskimo children of Alaska have "extremely impressive competencies and even expertise for surviving in a difficult environment, but because these skills are not ones valued by teachers" they tend to do very poorly in school. (Grigorenko et al, 2004).
-- In Brazil, street children who are extremely successful in running street businesses, and highly expert in math skills necessary for those affairs, do very poorly in abstract, pencil-and-paper math propblems. (Nunes, 1993 and 1994).
-- In Berkeley, California, there is "no correlation" between housewives' impressive abilities in comparison shopping math and scores on pencil-and-paper math tests. (Lave, 1989).

The essential point being that whatever our innate abilities -- which clearly exist but are still far from being understood and specified -- they do not limit us in a way that I.Q. scores imply. Ultimately, life success is a function not of inherent abilities, but of highly developed skills.