Monday, June 5, 2017

The Bell Curve: Race and Bad Science

Science is important to me and a couple of articles have come across my desk recently that are not just bad science but bad science serving a political agenda. Bad science is bad enough. But using evidence based thought to serve an agenda usually tosses the evidence out in favor of the bias. It not only helps no one, it holds us all back.

As Stan Marsh said in the South Park episode, The Biggest Douche in the Universe:  

“The Big Questions in life are tough. Why are we here? Where are we from? Where are we going? But if people believe in… liars like you we’re never going to find the real answers to those questions. You aren’t just lying, you’re slowing down the progress of all mankind.”

From here.

Which is why I enjoy South Park.

The article that crystallized it for me wasn’t a recent one but it is representative of a lot that I’ve seen.

Back in 1995, Richard Herrnstein and Charles Murry published The Bell Curve.  This book purported to represent how intelligence influenced class structure in America. I read it the year it came out and pretty much dismissed it since it violated most of the neuroscience I knew. (My Master’s is in Neurophysiology.) The book suggested that African Americans were too dumb to educate and should therefore we shouldn’t spend any money on them or the poor. A good refutation of the book is here. Vincent Sarich’s defense of the book is here.

Sarich’s defense was reprinted recently and when I read it, it ticked me off. It was bad science. But given recent events, I’m starting to think that people are having trouble detecting bad science when it comes up. So, I’m going to analyze the defense and show at least what goes on in my mind trying to winnow out good science from bad.

The first thing to notice about Sarich’s defense is that it is not a scientific paper. Forget its publishing history. That’s not germane. A scientific paper or other publication is distinguished by two salient characteristics: a dispassionate voice and a reliance on evidence and analysis local to the paper.

This latter concept is key. What it means is that a scientific publication attempts to enclose all evidence, reasoning and conclusions within the confines of the paper. Appeals to outside material—religion, ethics, popular concepts, etc.—should not be part of the paper. This is a goal but may not be possible in practice since the scope of the argument may exceed the boundaries of the paper. Hence, the concept of citation—the citing of other, relevant materials accessible to the reader.

This is not to say a non-scientific paper is not good, relevant, exciting or important. Lincoln’s Gettysburg Address is not a scientific paper but it is good, relevant, exciting and important. But a scientific publication defines its own venue of criticism. Consequently, stepping outside that realm means that such material cannot be taken as scientific. It is, at best, digested material. Interesting but not amenable to scientific judgement.

So: Sarich’s defense is not a scientific publication. There are a number of irrelevant references. Sarich attacks commentators on The Bell Curve. This is known as the ad hominem argument where the people promoting a point of view are attacked to besmirch their opinion. This is not, in and of itself, a bad thing. For example, if one notices that Koch Industries is funding a climate denial scientist, it injures the credibility of that scientist if he comes out favoring a position supporting Koch Industries. (See here.) But it is not a scientific argument. These are arguments are not couched in dispassionate terms. Sarich quotes a number of people using fairly passionate language.

That’s all right but it means that we should not give it scientific weight whether or not it purports to be a scientific publication. It’s a polemic.

But since we want to evaluate the science in the article, let’s winnow that out.

Sarich bases his defense of The Bell Curve on the following arguments:
Races are real with actual phenotypes that are identifiable among groups. These do not constitute species or subspecies as any race can easily interbreed with any other race and produce viable offspring.
The creation of racial phenotypes derives from historical geographical separation whereupon natural selection operated differentially on those phenotypes according to environmental success.
  • Intelligence is one of those phenotypes.
  • Intelligence is accurately measured by IQ as presented by intelligence tests.
  • Intelligence is a function of brain size.
  • IQ as an accurate measurement of intelligence can substitute for brain size.
  • Brain size is genetically determined.
  • Brain size (as derived from IQ tests) is varied across races.
  • Therefore, some races have greater intelligence than others as an inherited phenotype trait determined by genetics.
These are, in fact, many of the same arguments of The Bell Curve. No surprise there. So a lot of Sarich’s defense is saying “look at the evidence. It’s obvious these guys are right.”

Hm. Not so much.

Richard Feynman gave the 1974 Caltech commencement address, entitled Cargo Cult Science. I strongly urge anyone to read this.

The Cargo Cult was first described in Melanesia. It has come to mean circumstances where people built or rely on the attributes of what they want (runways, etc.) in hopes of attracting what they want (airplanes bearing material wealth.)

Science is hard. Figuring out what is actually going on rather than what we think is going on requires diligence, perseverance and a willingness to tear apart your own pet theories. Just pushing up straw men reflecting what you want to be true is Cargo Cult science.

He described Young’s very interesting rat experiments in 1937 trying to figure out why rats picked a particular door in his tests. While I urge you again to read the Feynman article, it’s worth quoting this bit:

“So he painted the doors very carefully, arranging the textures on the faces of the doors exactly the same.  Still the rats could tell.  Then he thought maybe the rats were smelling the food, so he used chemicals to change the smell after each run.  Still the rats could tell.  Then he realized the rats might be able to tell by seeing the lights and the arrangement in the laboratory like any commonsense person.  So he covered the corridor, and, still the rats could tell.

 He finally found that they could tell by the way the floor sounded when they ran over it.  And he could only fix that by putting his corridor in sand.  So he covered one after another of all possible clues and finally was able to fool the rats so that they had to learn to go in the third door.  If he relaxed any of his conditions, the rats could tell.”

My point with this quote is that is what is missing from Sarich’s arguments: rigor. Let’s take them one by one:

Races are real with actual phenotypes that are identifiable among groups. The phenotypes that represent the groups do not constitute species or subspecies as any race can easily interbreed with any other race and produce viable offspring.

The latter statement here is obvious: any human being can interbreed with pretty much any other. So we are all one species regardless of how much some people want to argue against that. The “race” question is much more subtle. There are group phenotypes, some with longer lasting effects than others. Any genetically isolated group will start to show them as the gene pool narrows. Similarly, these phenotypes begin to fade as soon as the gene pool widens, though some phenotypes have more staying power than others.

But the group definition of race is often defined by the qualities of those outside the group in question and it’s of particular importance here in the United States with our heritage of centuries of African American chattel slavery. 

On average, African Americans are an amalgam of 73-80% or so West and Central Africa, 18-24% European and a little less than 1% Native American. (From here) This means that the group being criticized shares 20% of the genetic material as the group doing the criticizing. So: is there a cultural phenomenon of race? Sure. Is there a genetic component to race? Probably—there are phenotypic traits we identify. Do they mean much? Probably not.

The creation of racial phenotypes derives from historical geographical separation whereupon natural selection operated differentially on those phenotypes according to environmental success.

Sure. That’s how phenotypes arise—see above. The key component here is “natural selection.” What does that mean?

It means there is differential reproduction between individuals such that particular phenotypes are preserved. That’s all it means. It does not mean “fittest.” It does not mean “successful” except with respect to that differential reproduction. And it does not indicate in any way why those traits became established.

In some cases phenotypes can be established without any natural selection whatsoever. For example, there’s such a thing as genetic drift where traits become established in a given population from pure randomness. 

There is also sexual selection where individuals are selected based on appearance that has nothing to do with any sort of physical “fitness.”

Sarich brings up the fairly tired example of the preponderance of African Americans in the National Basketball Association. He suggests that there is a genetic fitness of the individuals, based on race, that predisposes African Americans towards success in the NBA based on their statistical presence.

This is confusing correlation with causation.

There are a lot of confounding factors in his conclusion: what are the relative population sizes of of cultural groups attempting to join the NBA. Are opportunities of those groups the same? Are there differences in competing attractive opportunities? Are there differences in access to those competing attractive opportunities between groups? What are the relative cultural ambitions of the two groups? Any one or combination or additional unmentioned issues can confound a biological interpretation.

If you want to examine the hypothesis that there is a predisposition towards basketball among demographics in the US, the obvious characteristic to consider is average height and height variation. And there is some variation. (See here.) Non-Hispanic Blacks average about 12% taller than average Non-Hispanic Whites according to the CDC. This could be a salient biological factor. Does Sarich investigate it? No.

If we were to investigate it, the hypothesis might be phrased:
  1. Height is important in basketball.
  2. African Americans are statistically taller than other ethnic groups
  3. Therefore, can African Americans be statistical overrepresentation in NBA basketball be related to a function of height?

That’s as far as the scientific analysis. It’s not, “African Americans are better at basketball.” That’s a conclusion that cannot be demonstrated. All that can be demonstrated is the correlation. But even that’s not a scientific approach. One would want to bracket each statement:
  • Is height a determining factor for player admission to the NBA?
  • Is the tall African American population the same population drawn on by player admission?
  • Are there countervailing factors that are more important than height in the admission process?
  • Do these factors outweigh the influence of height?

This is where the “fitness” argument begins to break down. People use the word “fitness” as a substitute for being “better”—that is not evolution. Evolution is all about statistical reproductive success. If that individual asthmatic neighbor with blue eyes has one more child than the NBA neighbor, the asthmatic is more reproductively successful.

It is common to confuse societal success with evolutionary success. They are completely different. 

Sarich’s example with the NBA is a classic result of this. He’s suggesting that success in the NBA is a result from genetic heritage and demonstrates some sort of evolutionary fitness. The only component amenable to scientific analysis on a racial basis is something demonstrable like height and that’s not a clear relationships. Yet, Sarich presents this as a scientific predisposition when it is not.

Intelligence is one of those phenotypes.

This is one of those “obvious” statements that’s anything but. In order to understand this statement we have to define intelligence and relate it to the concept of phenotype. Certainly, intelligence is demonstrable and it is a phenotype. However, in order to determine its genetic heritage, we have to be able to define it quantifiably and reproduce that quantity on demand. This brings the next three statements up:

Intelligence is accurately measured by IQ as presented by intelligence tests.
Intelligence is a function of brain size.
IQ as an accurate measurement can substitute for brain size.

In this context, “intelligence” is quantified as the score on an IQ test. This means that the definition of “intelligence” must be the ability to take IQ tests—we’ve redefined an important quality of a human being and narrowed it to the ability to take a standardized test.

Think what this means. Let us say, for example, we pull a Neanderthal from the past and plunk him down in a classroom with a pencil, booklet and test sheet. This would be a person that does not know what a pencil, booklet and test sheet is or how to read. By the above definition, the Neanderthal is not intelligent.

IQ tests are notoriously biased towards the culture that originates them. (See here.) As the cultures narrow towards each other, the IQ scores begin to approach one another. A wonderful example of this is the The Black Intelligence Test of Cultural Homogeneity (B.I.T.C.H.) invented by Robert Williams in 1972. Williams deliberately developed a test with a cultural bias and, sure enough, African Americans did better than their white counterparts.

Then, the IQ score is linked to brain size, i.e., the ability to take standardized tests is a function of brain size.

Again, this is not a scientific proposition. Certainly, brain size is a variable. However, elephants have bigger brains than humans, men have bigger brains than women, big people have bigger brains than smaller people. Does this mean elephants would be better at taking IQ tests?

One has to define what is meant by “brain size” and justify the proposition that brain size, as it is defined, reflects what is defined to be intelligence. Sarich does neither of these. (There’s a good analysis of what does constitute intelligence vs brain component relative sizes here and here.) Note from those links that there is only a marginal relationship between head size and brain volume.

Studies of Cro Magnon Man (28k years ago) and NeandertalsClose analysis of brain size/intelligence correlation suggests that it’s not size alone that drives the relationship but the size of differing components. So size alone doesn’t matter so much. In both cases, their brain size is greater than modern man's. Would Sarich think we are, therefore, less intelligent than past members of genus Homo? I suspect not.

Brain size is genetically determined.

This is the point where Sarich’s argument begins to narrow towards his conclusion. Up to now the different propositions have been more or less separate from one another. But now, he is linking brain size to IQ and genetics.

It is, in part, the “genetics is destiny” idea. Essentially, it says your physical inheritance determines who you are. Historically, it’s been the basis for as diverse political ideas as eugenics, class breeding and the divine right of kings.

We’ve already suggested the relationship between brain size and brain function is a complex relationship—as we should expect. Leg length and stride have a relationship but it’s not completely straightforward. Certainly, hip to ankle length makes possible a certain stride. But this is mitigated by the hip to knee and knee to ankle lengths. In addition, the human hip and the hips of other animals are quite different as human beings are adapted to upright walking. 

Consequently, while we can say the relationship between leg length is straightforward, even a cursory analysis shows that’s not the case.

In addition, we also know that nutrition, birth traumas and other issues can dramatically change the leg length. We can argue something similar for “brain size.” In addition, while “brain size” might be genetically determined it must be corrected for body size, etc., to be meaningful.

By simplifying the statement to just saying brain size (stand in for IQ test performance stand in for intelligence) is genetically determined, the inference is that intelligence is genetically determined.

Which brings us to the next statements:

Brain size (as derived from IQ tests) is varied across races.
Therefore, some races have greater intelligence than others as an inherited phenotype trait determined by genetics.

Sarich tries to couch this in evolutionary terms saying this is analogous to leg length and other such items. The issue makes two very profound assumptions:

  • That the variation within brain size (stand in for intelligence) serves as an opportunity for natural selection.
  • That the natural selection on brain size (stand in for brain function, i.e., intelligence) must necessarily cause differences between the races.

That’s a lot to unpack.

There is variation in cognition. Everybody knows somebody smarter than they are or not as smart. All cognition tests, IQ or otherwise, shows a curve of ability. So we know there is variation. The implied shift between Brain Size and Race is couched as a derivative from the known evidence (which we have questioned but for this part of the discussion let stand) to a solid conclusion.

Instead, there are a lot of implied assumptions in that leap that are not obvious.

For one, there’s the presumption that intelligence variation between races is meaningful. For such a variation to occur, much less be meaningful, would indicate that natural selection within a racial group is operating differently when compared to a second group. The brain is expensive—it takes on the order of 20% of the entire energy budget of the body to keep it going. If we could do without it or could manage a reduction in its size and demand, natural selection would immediately push us in that direction.

There are quite minor variations between racial groups in brain size—much less than the width of variation within the same group. One would expect any group that could reduce the energy load would very quickly naturally select to do so. The fact that no human group has reduced that much when compared to one another suggests that the advantage of a large brain overcomes the disadvantage of the expense.

Given that, the presumption of Sarich’s argument then is that the mechanism that created racial differences is the same mechanism that creates differences in intelligence (brain size.)

This is a huge assumption. It’s comparable to saying that the same selection mechanism that gave us long legs got rid of body hair. Or the selection mechanism for stubby toes is the same one that gave us the taste of sugar.

No doubt there is selection for increasing intelligence in humans—that’s our heritage. And no doubt variation in intelligence and selection against that variation caused it to increase. But to presume that selection for one phenotype is equivalent to another is madness.

In point of fact, I argue right here that since the variation within groups is greater than between them, that argues that the selection mechanism for human beings evolving larger brains is independent of racial groups. It acts on all human groups equally—suggesting that whatever the mechanism is, it is intrinsic to human beings independent of their environment. I think it’s likely sexual and social selection. The complexity of human groups is much greater than any environmental complexity we’ve faced.

Regardless, there is no evidence that there is differential selection between "racial" groups which suggests that Sarich's conclusions are spurious.

The Bell Curve and its criticism and defense happened in 1995. Vincent Sarich died in 2012. You could say the discussion is more than twenty years old but, as I said, I'm started to see it rear its ugly head again. I’ve been fairly careful not to introduce subsequent neurological research. It turns out that the relationship between IQ and brain size is somewhat nebulous. (See here.) It also turns out that (what a surprise) when corrected for body size, brain size is astonishingly similar for everybody.

But my point here is that by just thinking critically about the material one can come away with pretty much knowing what is bad science and what is good science. The problem is that thinking critically is hard. Thinking critically about ourselves (human beings) is much, much harder.

It does seem in this post-reality society, we’re making our lives harder for ourselves.

No comments:

Post a Comment