[For reference, the text of the email exchange between Sam Harris and Ezra Klein over an article published about Harris and Charles Murray in Vox is here. References to the Vox articles in question are linked within Harris’s blog.]
Sam Harris is right that his argument with Ezra Klein suggests one side is in a delusional moral panic and the other is engaging in an intellectually honest conversation, but he is wrong about who is who. Harris is the one in a delusional moral panic, and Klein is the one attempting to have an honest conversation on a politically divisive topic. It’s a shame that Harris doesn’t see this; that he apparently can’t let go of his bone of moral victimization long enough to take a more balanced look at how The Bell Curve, and theories like it, has endured—or in this case has not endured—peer review. For as it happens, two of Herrnstein and Murray’s central claims were found unsound both before and at the time The Bell Curve was published: before by numerous scientists when the same issues came up in Jensen’s earlier visitation of the same themes, and contemporaneously by most of the scientific reviews of the book. Specifically, the two most contested claims are that 1) heritability indicates the immutability, or not, of a trait and 2) heritability as an estimate of genetic influence in some scientific sense implies that between-group genetic differences determine differences in that trait, and that those differences are therefore irremediable. Both claims are the flash point for the political dynamite that is this long-standing debate over race, genes, and IQ. They represent, as it were, the contested powder keg that both Harris and Klein ignite.
To take the issue out of its controversial context and to make clear the scientific stakes—not just the lay intuitions, but the scientific stakes—let’s consider a trait even more heritable than IQ but far less controversial, height. Consider, for instance, that a biologist wants to compare height in two ‘racial’ groups: the ‘Asians’ of South Korea and the ‘Caucasians’ of the US. Phenotypically both groups share a common trait (skin color), but they both also differ on height and other traits (such as facial features, hair color, etc.). Now, assume a biologist wants to know, in general, how much ‘genes influence height’; that is, assume he wants to know both if height is relatively fixed in each group and if genes are driving (however partially) the difference between groups. He knows, for instance, that height has a heritability ranging from 0.6 to 0.8, meaning, colloquially, that the height is 60-80% ‘due to genetic factors’ and 20-40% ‘due to environmental factors,’ but more technically and accurately that these proportions represent the partitioning of the variation in height due to genetic and environmental factors. Notably, he also knows this partitioning into genetic and environmental factors must be determined separately for each group studied, and as a separate determination, heritability is expected to vary. In other words, there is no fixed, or fixable, “heritability” for height as such outside the population being investigated, covering, as it were, the relative partitioning of genes and environment across all populations. So knowing these two things about heritability and height, the question for the biologist comparing these two groups is: using heritability estimates, what can be said about the 1) the immutability, or not, of height relative to environmental manipulation and 2) the likelihood that genetic differences are driving the between-group differences, as opposed to being driven by environmental differences alone, or some combination of the two? The entire political dynamite of the race, genes and IQ debate in psychology and social science (more on this emphasis shortly) hinges on these questions and the way they are answered. So what exactly turns on this debate? What does heritability in the technical, scientific sense say about either issue—the immutability of a trait or the relative genetic contribution to between-group differences?
Surprisingly—surprisingly, that is, given the political furor over the topic—the answer to both questions is disarming enough—nothing. Nothing whatsoever about the relative immutability, or not, of a trait can be inferred from the scientific notion of “heritability,” just as nothing whatsoever can be inferred from heritability estimates about the genetic basis, or not, of a between-group difference on any phenotypic measure (and thus the immutability, or not, of that difference). And what’s more, outside of psychology and social science this is entirely uncontroversial. Population geneticists, statisticians, and biologists are so comfortable with this limitation to heritability estimates—i.e. with their non-informative value on these two points—that they only condescend to rebut the opposite presumption when psychologists or other social scientists mistakenly make it. This means, as surprising as it might sound, that the controversy over race, genes, heritability, and IQ is almost exclusively a psychologist’s or social scientist’s recreation. And what’s more: it’s a recreation based a fundamental misunderstanding. Simply put, heritability estimates have nothing to do with whether 1) IQ is or is not a genetically immutable trait or 2) the black-white IQ difference is rooted in genes, or in the environment, or whatever mix of both, and therefore is largely or in part immutable as well. “Heritability” has nothing useful to say about either issue, and that it supposedly does is nothing more than a symptom of psychologists and social scientists borrowing from outside of their scientific expertise, then allowing their lay intuitions about genes and environments to infect the tool through conceptual creep, the scope of which they either do not acknowledge or perhaps in some cases don’t even understand.
How can this be? Can it really be so simple that the powder keg that is heritability, race, genes and IQ only explodes because of a misappropriation, one so basic that virtually no geneticist or biologist makes it? As dismaying, it is—and to this minor scientist it is very dismaying—the fundamental misuse of heritability estimates, both with respect to what they imply or don’t imply about genetic immutability, drives the race, genes and IQ debate in psychology. If one looks elsewhere—say to the non-morally panicked reviews and papers put forth in the non-psychological literature (such as analysis of works like The Bell Curve by biologists, statisticians, and geneticists whose native domain is “heritability”)—there is simply no reason whatsoever to care if the heritability of IQ is 0.2 or 0.8, or anything in between, because heritability, paradoxically enough, says nothing about either the environmental or genetic causes of intelligence, much less anything about the genetic difference between the races that might or might not be based in them. And it says even less about immutability on either point. Only in psychology and social science is it an issue that it possibly does otherwise, meaning that both the consensus ‘on the science’ Harris sees and the controversy ‘in it’ Klein sees is (for the most part) utterly pointless.
The arguments for this no doubt controversial charge—and yes, among psychologists it would be a controversial charge, but among biologists or geneticists it would not; in any case, regarding this charge, the highly technical arguments against the psychological and the social scientific misuse of “heritability” estimates have been discussed extensively, and most of the details are difficult to convey to lay readers, but their application is simple enough. For a full account, curious and technically inclined readers are referred to Layzer’s “Science or Superstition: A Physical Scientist Looks at the IQ Controversy” (Cognition, 1, p. 265) and his “Heritability Analysis of IQ Scores: Science or Numerology?” (Science, 183, p. 1259); to Lewontin’s “The Analysis of Variance and the Analysis of Causes” (Am J Hum Genet, 26, p. 400); to Feldman and Lewontin’s “The Heritability Hang-up” (Science, 190, p. 1163; for responses and FL’s replies, see “Heritability of IQ”, Science 194, p. 6); and to Goldberger’s “Heritability” (Economica, 46, 327). All of these works are by serious scientists, and none of them exhibit anything like moral panic, yet none of them would agree with psychologists and social scientists who say there is a valid consensus on how to use heritability estimates to discern the relative immutability of traits or the relative genetic contribution to between-group differences, because, simply put, in their view there is no validity to making the attempt. The details of their arguments are often densely mathematical and conceptually technical, but fortunately the details need not be rehearsed in order to see how the conclusions affect the example proposed above. In other words, the biologist wanting to know something about genes, height, immutability and group differences between Asians and Caucasians serves well enough as an example of the scientific pointlessness of using heritability estimates to support claims like 1 or 2—i.e. the immutability (or not) of traits and the genetic contribution (or not) to between group differences, and thus the immutability or not of that difference as well.
So consider again the two groups on height: South Korean Asians and American Caucasians. Now, say two groups of biologists have heritability estimates in two different time periods, say 1914 and 2014. In both comparisons, the biologist wants to know 1) how genetically ‘fixed’ within groups relative height is and 2) how much of the height difference between groups is ‘due to genes,’ and how much is ‘due to the environment.’ To be more specific, say it is known in 1914 that the average height of 18 year old Korean men is (because it was) about 62.5 inches, and that the average height for US men is (again, because it was) about 67.5 inches—a difference of about 5 inches, or just over one standard deviation of the height for US men. The biologist comparing these two populations over such a large difference is likely to infer two things, but from lay intuitions about genetics, not the science of “heritability” as such: first, that height is somehow genetically circumscribed, meaning that the genes ‘set’ an upper and lower limit to how tall Korean and American men are; and second, that differences in these genes must be at least partially involved in such a dramatic, 1.25 standard deviation difference between the two, and that therefore this difference is in some respects immutable. For after all, genes and environment work together to create every trait, so naturally this one—height—must also be under some kind of genetic control. And what’s more, what else but genes could account for such a large difference, when for all intents and purposes both populations appear to be getting enough to eat; neither has a substantial proportion of starvation. Both lay intuitions are common, and both drive the race, heritability, genes, and IQ debate. But what, it can be asked, does heritability add to these lay intuitions? Based on heritability estimates, what inferences are likely to be made?
First, regarding “immutability,” the biologist knows that the heritability for height is relatively high, suggesting that genes are contributing to the majority of the height—in this case, 60-80%. And since presumably the genes for height don’t vary, only their relative frequency in the population does, based on the 60-80% genetic contribution, it seems safe to assume that there is an upper bound to how much environmental manipulation can influence height within the group. In other words, as this manipulation encounters the limits of genetic control, it will cease to make a difference. Based on high heritability, then, the first of two inferences will be made: height, it can be said, is relatively immutable, subject to probably small variations due to environmental changes.
Second, regarding the genetic contribution to between group differences, since the within-group heritability is so high, and since this value suggests that the lion’s share of height is under genetic control, it seems safe to assume—it even follows from the data, it will be said—that the between-group difference of 1.25 standard deviations must be under some kind of genetic control as well. In other words, since height within the group is so highly influenced by genes, and since the difference between groups is so high, this difference must in part at least be controlled by differences in these genes as well, and therefore it remains in some sense as immutable as the trait within groups. Based on high heritability, it would seem safe to assume that some genetic control is accounting for the between-group difference, especially since discontinuous traits like eye configuration and hair color are almost certainly genetically controlled. Why expect, it will be said, height to be any different?
Both of these inferences will appear to follow from the data and the standard estimates of heritability for height. High heritability suggests some degree of immutability, and high heritability suggests some degree of immutable genetic difference between the groups. This biologist will therefore likely conclude that there are intrinsic limits to remedial programs to make Koreans as tall as Americans, and more to the point, he may even conclude that the genetic differences will perpetuate over time, despite our best intentions, into a social under-class of short Koreans vis-à-vis tall Americans. They will, in short, be left behind. The policy implications will appear to follow from the science.
Now, instead of rehearsing the technical arguments why neither of these two inferences are valid from the heritability estimates (thought they remain valid as possibilities in lay intuitions), consider a similar biologist in 2014 comparing the same populations four generations later. The heritability of height is still 0.6 to 0.8, so he is likely to make the same two inferences about immutability and between-group differences as the biologist did in 1914, except now there is a problem. Now the average height of 18 year old Korean men is 68.9 inches, and the height of American men is 69.7 inches (as they are now)—a difference of 0.25 standard deviations, a full standard deviation closer. How is this possible? Given the high heritability and its presumed genetic control, i.e. its presumed immutability in 1914, how, the biologist must ask, is it possible that a full standard deviation of group difference has dissolved, even as an ‘immutable’ trait within the group has changed? Based on high heritability, the biologist in 1914 inferred with all apparent merits a high degree of ‘immutability’ to height within the groups, just as he inferred that genetic differences at least in part —and based on the “high” heritability most likely in no small part—drove the difference between the groups. Yet in four generations—far too little time for any meaningful genetic change to occur—the heights of Koreans and Americans converged from 1.25 standard deviations apart to only .25 standard deviations apart. If genetic variation can’t account for this convergence—and it can’t—what then does?
According to the partitioning by the heritability estimate into ‘genetic’ and ‘environmental’ variance, there can be only one answer: the environment. Despite the high heritability and therefore the apparently high immutability and genetic control over with the within-group trait and the between-group difference, the’ environment’ must have changed, and this change must have driven the convergence because genetic change couldn’t have done it. It is inferences like this that fuels the race and IQ debate among psychologists, with one camp arguing that heritability implies one thing (genes) and the other arguing that it leaves room for another (the environment). Parsing the relative contributions with lay intuitions like “some” or “in part” etc. only magnifies, not settles, the problem. In fact, however, the focus on heritability estimates in determining the causal ‘role of genes’ in a trait or a trait difference between groups is itself the problem, regardless of whether one prefers the ‘genetic’ or ‘environmentalist’ camp (though of course, when pressed, both sides have to agree it’s “it’s both”– yet they still argue!). Absent this focus—this misfocus, as it were—on heritability there wouldn’t even be an argument between two warring camps.
Clearly for something like “height”, the point being made here is hardly controversial, except perhaps to psychologists and social scientists who want to insist that one example, however real, doesn’t undermine the use of heritability estimates to determine the relative contributions of genes to immutability and between-group differences on phenotypic traits. And to some degree, they would be right: one example is hardly enough to discredit heritability estimates for this purpose, unless it illustrates a broader point, as this one does. Specifically, this routine and non-controversial example shows that conceptual creep from lay intuitions informs a technical concept like “heritability,” even as the technical expertise of heritability computation gives sanction to purely speculative or tautologically obvious lay intuitions. From a generic standpoint, it is meaningless to say that genes play “some” role in a trait being immutable or not, or that because of them the trait falls within a limited range, and it’s meaningless because genes are involved in all traits that fall within a limited range, as all traits invariably do. And some of those traits are ‘immutable.’ Similarly, it is equally meaningless to say that genes play ‘some’ role in the between-group differences in a phenotypic trait, and it is meaningless because, again, genes are involved in all traits, those both similar and different, both between and within groups. What misused heritability estimates adds to these platitudes is the illusion of a numerical estimate of relative genetic causation—an illusion that is tantamount to saying that the length of a rectangle ’causes’ this or that much “proportion” of area opposed to the width, when only the multiple of the length and width produces area in the first place. Like with length and width in area, so with genes and environments in traits. Both always play a role—both acting together are the trait—and when it comes to immutability and between group differences, heritability estimates only offer the foolish illusion that relative causes of a trait are being partitioned out. As this example nicely illustrates, the inability of [sic] “perfectly valid” inferences about genetic influence from heritability estimates to account for how height actually gets expressed in and between two populations should be warning enough against using them to estimate ‘immutability’ and ‘genetic control’ both within and between groups for intelligence. For as it stands with causes of a trait and its immutability, heritability estimates are meaningless.
Now, it might be asked: how well does this non-controversial example of height map onto the highly controversial example of race, genes, and heritability with respect to intelligence? How instructive is this one example for examining the contours and limits of the other?
Given the reality of the Flynn effect in IQ scores despite its relatively high heritability (HM source it between 0.4 and 0.8, settling on 0.6), and given the convergence of black and white averages between 1972 and 2002 (0.33 SD, or 5.5 points), the parallel between “height” and “IQ” is instructive, if not exact. First, like with height over the generations in question, there has been a secular rise in IQ scores that cannot be accounted for by genetic changes, the time period being too short for meaningful genetic change to have taken place. Similarly, while not as pronounced, the average IQ scores between blacks and whites are converging, and they are converging as the relative equality and civic participation of blacks in general improves, not as the genetic makeup of the populations changes. Like for height, then, in 1914, it is foolish now to look at the high heritability of IQ and infer anything about the immutability of the trait, or the genetic basis and therefore immutability of the between-group difference—an application that in any case is theoretically invalid for any heritability estimate (see the above cited technical literature). Just as height can change as environments change, so can intelligence, and whatever the upper limits of those changes relative to environmental manipulation, heritability estimates contain no information on those limits, or on the causes governing them. Simply put, using heritability estimates to estimate relative fixity of traits or the genetic immutability of between group differences is as foolish now as using heritability estimates to do the same for height would have been in 1914. Secular changes can and have occurred in both height and IQ through causal mechanisms that are as yet unknown—mechanisms entirely unknowable through heritability estimates.
It is crucial to point out at this point that nothing in this argument says that there are not (in a loose sense) ‘immutable’ traits within groups, or ‘fixed’ and ‘unbridgeable’ differences between groups on traits of interest, much less that these traits and differences don’t have genetic markers. It is not even to say that in the right environments, traits like height and intelligence will converge in all compared groups, and that remaining differences won’t have genetic markers as well. Finally, it is not even to say that the Asians and the Caucasians compared in this example will converge completely, and that all genetic markers of differences in height will disappear, much less that all groups of Asians (say the Chinese or Japanese) will completely converge in the ‘right’ (whatever that means) environment, leaving, again, no genetic markers for differences (though it does bear pointing out that both Japanese and Chinese men are converging, and even that Chinese men are converging at different rates relative to rural, urban, and capital residence, meaning Chinese men in Beijing are taller than their rural confederates and are as tall as America men, though this was not the case just two generations ago). In any case, nothing in this comparison between height and intelligence suggests that ‘suitable’ (whatever that means) environments will absolve all differences between races on any given trait, much less on intelligence or height, leaving, as it were, no genetic markers for differences between them. It is only to say that in the quest to understand how ‘immutable’ the traits are within a group and how genetically related the differences are between groups, heritability estimates are meaningless. And to this point, note that in the recent New York Times opinion piece on genetics and race that Harris cites as supporting evidence, the geneticist writing it never refers to the “heritability” of traits, not even once. Instead, unlike the race, genes, and IQ psychologists and social scientists, he focuses exclusively on known genetic markers for persistent traits and the causal mechanisms—such as we yet know them—for their expression. This is a far cry from inferences about immutability and group difference based on heritability estimates because those inferences are as unscientifically uninteresting as they are invalid. In other words, those invalid inferences have nothing to do with the important scientific questions of what causes traits within groups, or the differences between them. It is only through conceptual creep from lay intuitions that heritability estimates become the focal point for questions on genes, intelligence, and race.
Having disposed of the utility of heritability estimates for estimating the relative immutability of traits or the genetic contribution (and therefore fixity) to between-group differences, it remains to be seen if Herrnstein and Murray fall victim to its illusions. In other words, in The Bell Curve do HM draw invalid inferences from heritability estimates, and has the invalidity of these inferences been called out, either directly or indirectly, in the scientific treatments of their book?
Unfortunately yes. In The Bell Curve HM draw both of the invalid inferences about heritability discussed so far, and scientific reviewers have called them out on it in their reviews.
With respect to the first inference—the immutability of traits—after elaborating on a “syllogism” that links the inheritance of IQ to success in life through the notion of “heritability” (as though “heritability” is a measure of “inheritance”), HM specifically refer to the “limits that heritability puts on the ability to manipulate intelligence” (p. 109). Taken together, there are two problems this syllogism. First, heritability is not a measure of inheritance—a fallacy even greater than the two discussed in this essay. And second, with respect to inferences about “immutability,” “heritability” says nothing about the ‘genetic fixity’ of a trait, or its ‘immutability,’ as saying ‘puts a limit on manipulability’ implies. Goldberger and Manski (above cite) point this out in their review, even as they note that social scientists appear alone in making the invalid inference. To date, Murray has not taken this correction to heart, as he still writes and speaks as though because it is highly heritable, IQ is somehow ‘fixed’ or ‘immutable’ and otherwise not amendable to environmental manipulation. On this, however, he is wrong—or alternatively put, he is using “heritability” infected with lay interpretations that belay its proper scientific sense. Heritability properly understood implies nothing about “immutability” or the limits of environmental manipulation.
With respect to the second inference—the genetic basis of between-group differences, and therefore the relative immutability of that difference—HM judiciously enough note the technical limit of heritability estimates on this point, i.e. they note that simply because “a trait is genetically transmitted in individuals,” this “does not mean that group differences in that trait are also genetic in origin” (p. 298). But in parsing their supporting example in the very next paragraph, HM “use the within-group heritability estimate of 0.6 as if it were a between-group measure.” Specifically, as an illustrative example HM note that using a within-group estimate of heritability to measure between-group differences is like assuming that handfuls of genetically identical corn will produce the same yield in two different environments like the Mojave Desert and Iowa. As they note, whatever the heritability of any trait, “the seeds will grow in Iowa, not the Mojave, and the result will have nothing to do with the genetic differences” (ibid). True enough, of course, but when parsing the example’s application to IQ, race, and environmental conditions between blacks and whites in the US, HM ask, invalidly: given that IQ is 60% heritable, “how different would the environments […] have to be [between blacks and whites and Asians] in order to explain the observed differences” in their average IQ scores? (p. 298). The very question, however, is misposed, and the calculations they make to answer it are as invalid as the conclusion they draw from them, to wit, that “environmental differences of this magnitude and pattern are implausible” (p. 299). The error here? Nothing in a within-group heritability estimate permits one to calculate the relative effects of a between-group difference in either genes or environments because heritability estimates say nothing about the relative genetic or environmental differences between groups. All HM do is commit the fallacy of between-groups inference with respect to the environmental proportion of the variance, not the genetic one, even as they assert that the implausibility of the computed environmental differences suggests the genetic immutability of the difference. This is not only committing the second invalid inference from heritability estimates on between-group contributions; it is committing it with a vengeance. Again, Goldberger and Manski pointed out this error 28 years ago, yet Murray still writes and speaks as though heritability estimates contain some kind of important information about the genetic basis, and therefore the relative immutability, of between-group racial differences in IQ.
So Murray commits two quite fundamental errors about race, genes and IQ by drawing two invalid inferences from heritability estimates, namely: 1) that heritability estimates indicate something about the immutability of a trait, and therefore its resistance to environmental manipulation, and 2) that heritability estimates indicate something important about the genetic basis of a between-group difference on any given trait, and therefore, again, something about the immutability of that trait difference. Neither inference is valid when heritability estimates are properly understood, so in both respects, however “mainstream” Murray is within the “scientific” consensus of psychology and social science, he is flat wrong in the consensus of geneticists and biologists, arguably the only consensus that matters on understanding “heritability” estimates. In any case, psychological or social science consensus aside—and the very fact that prominent detractors deny a consensus is evidence enough that one does not exist—relative to the proper consensus on what heritability does and does not mean, Murray is a muddle-headed outlier proposing, perhaps not “junk science” but at least wrong or inaccurate science. In this respect, Harris is simply wrong to support Murray as a victim of nothing but moral panic and unfair political lynching (though he has been to a minor degree a victim of both). On the merits of peer-review in multiple sources, two critical elements of the so-called “science” of The Bell Curve were rejected both before it was published and when it was reviewed once published. There is simply no universe in which it makes sense to say that detraction from and attacks on Murray can’t be based in serious science—once, at least, one leaves the vitriol of the popular press, the agitation of which the marketing of book in no small part brought on itself.
With these conclusions in mind, how does a more balanced look at the science behind heritability estimates, race, and IQ bear on the argument between Harris and Klein? Is the original Vox article, as Harris says, “a disingenuous hit piece” that unfairly makes him look like Murray’s dupe, simply because Harris admittedly intended—indeed felt morally obligated—“to provide him [Murray] some cover” from unfair political attacks and character assassinations (Harris, blog)? Does this article that started the whole argument, for all its rhetorical flaws, represent the worthless mutterings of an ideological fringe that polices the margins of honest scientific debate by vilifying opponents with character attacks and libel—partisans who deny “uncontroversial” science that is, “in fact, uncontroversial” (Harris to Klein, 5/21)? Are Klein’s repeated attempts to clarify his position and assess the positions of Turkheimer, Harden, and Nisbett relative to those of Murray and Harris gas lighting attempts to “wish away” the hard truths Murray has the courage to say, truths that Harris wishes weren’t true but are nevertheless as scientifically irrefutable as anything gets in psychology? Finally, is Harris right, as he obviously thinks he is, to feel wronged in this whole process; to feel like he is the victim of nothing but “intellectual dishonesty” and “libelous articles” (Harris, 5/28) bent on discrediting him (never mind, for God’s sake, the actual motives he imputes to Klein)?
Hardly. While the Vox article is unnecessarily inflammatory, highly unflattering, and in some respects conceptually sloppy, the science it proposes remains a valid antidote to a position that is without any scientific merit whatsoever—to wit, Murray’s assertion that because of high heritability, IQ is for the most part immutable, and that the differences between blacks and white is both some percentage genetic and therefore to that degree similarly fixed, meaning that because of the ‘genetic contribution,’ there is nothing that can be done about it. Serious scientists who have no stake in the psychological debate between the ‘environmentalists’ and the ‘hereditarians’ on either side of the “heritability” debacle have pointed out, time and time again, that the very terms in which the debate gets posed are scientifically meaningless. In this respect, then, the authors of the Vox article are at worst skirmishing in what is at bottom, to everyone else, a pointless ideological science war among psychologists, but at best they are providing a much needed antidote to one of that war’s most toxic doctrines. And that doctrine is simple. Ever since Arthur Jensen in 1969, a faction of psychologists who claim to speak for “the scientific consensus” have argued that blacks are dumber than whites; that their genes make them this way; that all the evidence in the world points to this fact and that nothing can be done about it; therefore attempts to remedy their deficiencies are bound to fail. To be sure, when pressed there is a lot of hedging, hawing, and prevarication with qualifiers like “some”, “in part,” “most” and “reasonable to assume” liberally sprinkled in; but the essence of the matter stays the same. Given the policy recommendations and the ultimate worthlessness of the science behind them, it is hardly surprising that partisans against this position—even otherwise intellectually honest scientists—will lean toward inflammatory rhetoric and unflattering imputations (besides, whoever said public intellectual life wasn’t a contact sport?). In any case, the science in the Vox article and the follow up reply to critics is as valid as anything Murry proposes, and since Murray’s two main inferences from the heritability of IQ have been shown to be scientifically worthless—and shown by his peers, not the morally-panicked, libelous popular press—for all its own reliance on “heritability” estimates, that alternative science is at least still better than worthless, opening as it does a look into avoidable environmental causes of the racial IQ difference, causes which surely exist. Either way, the Vox authors have the benefit of proposing a scientific error that at least doesn’t lead to morally toxic indifference or reckless policy assumptions. The same cannot be said for Murray’s [sic] “science.”
What, then, can be said to conclude this essay on science, heritability, race, and IQ? What positive lesson can be drawn from the conceptual clarification (hopefully) that the natural experiment of height provides, i.e. that heritability estimates are not informative for the kinds of questions we need to ask and answer about genetics, intelligence, and race? What should “intellectually honest” scientists be discussing instead? What should the debate look like?
Simply put, given the best ignorance we have about ‘genes,’ ‘traits’ and ‘environments,’ how these three causally interact to produce “IQ” and the racial disparities on it is as yet entirely unknown—and that’s what should be discussed: the nature of this ignorance, and the nature of the empirical tools we’ll use to remedy it. Furthermore, what causes the racial differences in IQ averages, and what causally, if anything, can be done about improving them, is almost equally unknown. That too should be on the discussion table. And what about our prospects? What about the stakes in the here and now? Again, simply put, despite our prevailing ignorance, if the “natural experiment” with height is any indication—i.e. if relevant factors in intelligence are analogous to that highly heritable trait of height—then there is more reason for optimism than the ‘hereditarians’ like Murray suggest. But in either case, heritability estimates themselves are useless in the endeavor to make these determinations because, as David Deutsche noted in his TED Talk on explanations, they “settle nothing”; they are merely illusions conjured by “wizards.” When it comes to an issue as important as racial differences within a context of racial equality, why should magical thinking lead honest intellectual inquiry? When it comes to an issue with such moral, political, and even personal implications, why not start the discussion with an honest admission of ignorance? If Harris wants honest intellectual conversation, why not start by discarding his own moral panic in defense of the foolish consensus that is Murray’s magical thinking on “heritability and IQ”? Why not discard the wizardry and start with plain, simple ignorance? After all, despite Murray’s and his own protests otherwise, that is where we stand right now when it comes to understanding racial differences, genes, and intelligence. We are just ignorant, and there should be no acrimony in that.
 Among scientists, see Andrews and Nelkin (1996), “The Bell Curve: A Statement”, Science, 271, p.13; Goldberger and Manski (1995) “Review article: The Bell Curve by Herrnstein and Murray, Journal of Economic Literature, 33, p. 762; Heckman (1995), “Lessons from The Bell Curve,” The Journal of Political Economy, 103, 1091. For an indirect response see Neisser et. al. (1996), “Intelligence: Knowns and Unknowns,” Report of a Task Force by the American Psychological Association, American Psychologist, 51, p. 77. All of these reviews rejected the central theses in HM that heritability implies immutability of traits and therefore un-amenability to environmental differences, and that heritability implies that between-group differences in IQ has a genetic basis, partial or otherwise, and is also therefore irremediable. For a compilation of other scientific responses on broader topics than the two focused on here (most of them critical), see Intelligence, Genes and Success: Scientists Respond to The Bell Curve, eds. Devlin et. al, (1997). Suffice it to say that there has been ample detraction from Murray’s conclusions in the non-morally panicked peer-review process. His work is hardly ‘incontrovertible scientific fact.’
 Dickens and Flynn (2006), “Black Americans reduce the racial IQ gap: Evidence from standardization samples,” Psychological Science, 17, p. 913.
 Goldberger and Manski, 1995 p. 770-1.
 Just to name two. Scientific reviewers have rejected many others, particularly the more economically oriented claims of the book (see the convergence of Heckman and Goldberger, for instance). For a book length treatment complete with a bibliography of reviews, see the above cited Genes, Intelligence, and Success. All these reviews represent a ‘fair hearing’ from intellectually honest scientific critics– Murray’s peers, not a morally panicked press.