I address three issues raised in the commentaries of Buckhalt, Jensen, and/or Murray: the heritability of intelligence, intelligence's role in economic inequality, and the time course of the debate over "The Bell Curve." In some areas, I feel these authors have advanced the scientific discourse. However, because they did not address some of the important recent articles cited in my original target article, some opportunities to advance the debate have been lost for the moment.
1. I would like to thank Joseph Buckhalt, Arthur Jensen, and "The Bell Curve" (TBC) author Charles Murray for taking the time to read and write commentaries on my target article "Revisiting the Bell Curve" (Reifman, 2000). I would also like to thank Dr. Murray for bringing to my attention the book "Meritocracy and Economic Inequality" (Arrow, Bowles, & Durlauf, 2000), one of several books I read in preparing this response.
2. In this response, I have chosen to address the three issues of the heritability of intelligence (including a discussion of the "Flynn Effect;" Flynn, 2000a), intelligence's role in economic inequality, and the time course of the debate over TBC. These are issues that were either raised by more than one of the commentators and/or are areas of vigorous, ongoing research.
3. The main thrust in my target article regarding IQ heritability was to argue that recent studies indicated it to be toward the low end of TBC's suggested range of 40-80 percent (paragraph 29).
4. Murray (2000, paragraph 7) reiterates his belief in the 40-80 percent range, stating that this and other of TBC's claims were not exceptional "at the time we wrote them, nor are they now." Murray also continues to believe "the environmental component is dominated by the nonshared environment" (paragraph 7), and he endorses a claim of "increasing heritability of IQ with age" (paragraph 6). Jensen (2000) also endorses the ideas of environment variance being predominantly nonshared (at least by mid-adolescence; paragraph 10), and of IQ heritability increasing with age (paragraph 9).
5. The problem I have is that neither Murray (2000) nor Jensen (2000) addressed the important studies I cited calling into question their claims regarding shared/nonshared environments (Stoolmiller, 1999) and changes in heritability of IQ with age (Devlin, Daniels, and Roeder, 1997). Regarding the former, as noted in paragraph 15 of my target article, Stoolmiller argues that range restriction in adoptive homes may be biasing downwards estimates of shared family environmental influence and biasing upwards the effects of genes. Regarding the latter, as described in paragraph 25 of my target article, Devlin et al. failed to find support for a model including age effects; these authors also explain how the high heritability estimates in older subjects may be an artifact. The articles by Stoolmiller and Devlin et al. do not merely report single studies; each represents a careful mathematical synthesis of large numbers of studies that should be of interest to scholars studying heritability.
6. I am not saying that Murray and/or Jensen must agree with the conclusions of Stoolmiller (1999) and Devlin et al. (1997). However, Murray and Jensen should give these studies due consideration, and then state their reasoning for accepting, rejecting, or reserving judgement on the findings.
7. Murray's (2000) Section I also raised the distinction between (a) what is the most accurate estimate (or range of estimates) of heritability per se, and (b) the degree to which societal implications depend on strict genetic heritability versus parent-to-child transmission of any form. I submit that my main focus was on the former issue, whereas Murray's was more on the latter. Perhaps paragraph 29 of my target article implied that genetic heritability was crucial to predicting societal implications. However, I was simply trying to bring my section on heritability to a close and incorporate a statement by McGue (1997) that was in my 13th paragraph, into my summary. Perhaps Murray would take issue with McGue (1997), but it is hard to know because he did not cite McGue.
8. Jensen (2000, paragraphs 4-8) spends a great deal of time arguing for the importance of the "g factor." As noted in paragraphs 5 and 6 of my target article, I am willing to accept the utility of g (or a g-like construct), with some qualifications. Heckman (1995), however, argued in one of the earliest reviews of TBC that one need not and should not rely excessively on a g-based model of intelligence in conducting the kinds of analyses relevant to TBC [NOTE 1].
9. Jensen (2000, paragraph 8) notes that I did not bring up the "Flynn effect" (increasing IQ scores over time in many countries) in my target article. Jensen appears to find the Flynn effect a useful vehicle for further discussion of g. I did not discuss it earlier as, although the empirical patterns of rising IQ scores appear well-established, in many people's minds the jury is still out over how to interpret the increases. I address the Flynn effect now.
10. A couple of points are in order: First, the increases in IQ must be due predominantly to environmental factors because genetic changes cannot have occurred quickly enough to account for the IQ changes (Herrnstein & Murray, 1994; Neisser, 1998), and, as Flynn (2000a) points out, "...no one argues for a trend toward better genes..." (p. 44). Second, even if as some claim, the Flynn effect does not reflect real gains in g, then as Neisser suggests, "...the tests are evidently flawed, and all arguments based on test scores become suspect" (p. 5).
11. Jensen (2000, paragraph 8) firmly comes down on the side of the Flynn effect being independent of g (at least with regard to Wechsler tests), citing research by Rushton (1999) to that effect. Flynn (2000a) notes, however, that IQ gains of the last 60 years "are largest on the tests that are supposed to be the purest measure of intelligence, Raven's and other tests of fluid g" (p. 36). Flynn (2000b) responds to some of Rushton's research.
12. Jensen (2000, paragraph 8) also claims that "...it is unwarranted to use the Flynn effect to belittle the import of the Spearman effect or the Black-White IQ difference." Flynn (2000a) provides a thoughtful analysis of how the effect named after him can be used to think about black-white IQ differences in the U.S. He starts by suggesting that blacks in 1995 should be at a similar mean IQ level to whites in 1945, and then concludes as follows:
"All we need to assume is that the average black environment of 1995 matches the quality of the average white environment of 1945. Many will find that plausible enough to keep an open mind about the causes of the black-white IQ gap" (Flynn, 2000a, p. 45).
13. Within the last year or so, two additional statistical modeling approaches relevant to heritability have been presented.
14. Dickens and Flynn (2001) focus on, among other things, the fact that genes and environment are often correlated ("If someone's genes predispose them to be good at basketball, then somewhat better play alone is likely to lead him or her into an environment supportive of better performance" [p. 349], in terms of advanced coaching, opportunities to play with and against better players, etc.). Given this and other considerations, circumstances can arise in which "...genes get credit for some of the work that is actually being done by the environment" (p. 353).
15. The upshot of this is that "Our model of the effects of environment on IQ shows that the potential impact of even very small changes in environment could be very large, even if we accept the largest estimates of the heritability of IQ" (Dickens & Flynn, 2001, p. 349).
16. A final area I'd like to discuss regarding heritability is that of cultural transmission. Although the idea of cultural transmission has been around for at least 25 years (Dawkins, 1976) and was used in some older modeling of contributions to IQ variance, it has been invoked in some very recent writings (Cavalli-Sforza, 2000; Feldman, Otto, & Christiansen, 2000; Wilson, 1999). The modeling of cultural transmission appears to be very complex; interested readers are urged to consult the aforementioned sources. Feldman et al. note that, in part, it can be thought of in terms of "a latent factor that can be transmitted culturally from generation to generation and that may include factors such as wealth, education, and residence" (p. 69) [NOTE 2].
17. Feldman et al. (2000) summarize heritability studies that have taken cultural transmission into account, as follows:
"The most detailed analyses of the heritability of IQ, those that use thousands of related individuals rather than the tens of twins studied in the papers cited by Herrnstein and Murray, are not cited in most of the psychological literature. This literature, with its focus almost exclusively on twins, produces estimates substantially higher than studies with more diverse sets of relatives that allow more detailed estimation of culturally transmitted components. In fact, cultural and genetic heritability estimates are virtually the same for IQ when more complete models are used" (Feldman et al., 2000, p.62-63).
18. I address two issues in this reply that concern IQ's role in economic inequality. The first involves reanalyses of the National Longitudinal Survey of Youth (NLSY) to probe further the role of parental socioeconomic status and IQ in predicting social outcomes. The second involves broader societal issues related to income inequality.
19. In an attempt to address claims that the measure of parental SES used in TBC was weak and may have biased the results, Murray (2000, paragraph 10) notes, he published a post-TBC monograph (Murray, 1998) focusing on analyses of sibling pairs in the NLSY. In one type of analysis that used sibling pair as the unit of analysis, Murray (1998) assessed the correlation between sibling difference in IQ and sibling difference in earned income. Given the assumption - which I certainly consider reasonable - that two siblings will share many aspects of home environment beyond just parental SES, this analysis likely provides a more comprehensive control for home environment than was present in TBC. Yet, as Murray (1998) shows:
"The Bell Curve's method of controlling for SES and the sibling method of controlling for everything in the family background yield interpretations of the independent role of IQ on income that are substantively indistinguishable" (Murray, 1998, p. 25).
20. Murray (2000) noted that Korenman and Winship (2000) also conducted a siblings-based analysis. These latter authors note, in apparent shock, that:
"Incredible as it may seem, our sibling analyses suggest that, even though Herrnstein and Murray's parental SES index is poorly measured, it appears to be adequate for producing unbiased estimates of the effects of AFQT [Armed Forces Qualification Test, which TBC used to measure intelligence] score on socioeconomic outcomes" (Korenman & Winship, 2000, p. 138).
21. That is only half the story, however. If one is interested in the comparative magnitudes of AFQT scores and family background in predicting social outcomes, as Herrnstein and Murray (1994) were, then he or she must look at the effect of the crudeness of the TBC SES measure not just on the AFQT regression coefficient, but also on that for family background.
22. To do so, Korenman and Winship (2000, Table 7.4) added a set of family background control variables to augment the TBC SES index, in predicting a variety of social and economic outcomes. They conclude that: "The results presented in table 7.4 suggest that the combined family socioeconomic background effect was considerably larger than the effect of the index of parents' SES alone" (p. 159).
23. Others of Murray's (1998) analyses broke the sample up into five cognitive classes (slightly different from the ones in TBC): "very bright," "bright," "normal," "dull," and "very dull," and examined situations where one sibling was in the normal reference category and the other was in a different category.
24. A very interesting demonstration of the importance of non-IQ factors emerges from Murray's (1998) Table 4-3, in which the five cognitive classes are compared on the dependent variable of attainment of a Bachelor's degree. The table is broken down further according to whether or not the normal referent sibling had attained the Bachelor's degree. In sibling pairs where the referent sibling had attained the degree, 91.3% of their very bright (top 10% on IQ) siblings had also. However, in sibling pairs where the referent had not attained the degree, only 58.7% of their very bright siblings had. What is important to realize here is that, even though IQ was held absolutely constant in the two groups of very bright siblings (top 10% of IQ in both cases), something was accounting for the fact that one group of very bright individuals had a 91.3% rate of acquiring the Bachelor's degree, whereas a different, equally smart group had only a 58.7% rate of doing so.
25. Apart from the specific confines of the NLSY dataset, one can also attempt to evaluate economic inequality and its possible causes by looking at broader societal trends. The book "Created Unequal" by James K. Galbraith (1998) is very helpful in this regard. [NOTE 3]
26. Galbraith's (1998) book and TBC both have the same point of departure in discussing inequality: data from the monthly Current Population Survey in the U.S. showing increases in income disparities in the years leading up to these books' publication. The two books also note that this increasing inequality must derive from something beyond education, experience in the labour force, and gender (Galbraith mentions only the first two of these factors). The two books agree, in other words, that some unmeasured, "residual" factor(s) must have been playing a substantial role in the documented trends of increasing economic inequality.
27. Herrnstein and Murray (1994) argue that IQ makes up at least part of this residual contribution to inequality, giving as one of their reasons that, "Perhaps most obviously is that technology has increased the economic value of intelligence" (p. 98).
28. Galbraith (1998) characterizes this view, which he claims is a dominant one among economists, as follows: "...the idea appears to be that computers extend the productivity of the underlying intellectual talent. Thus, they make typists faster, writers more vivid, artists more accessible, bookkeepers more accurate, and so on" (p. 32).
29. Galbraith (1998) then offers several arguments against the notion that some intersection of technology, skill, and intelligence is a major contributor to inequality. He first asks, "But does the use of computers for typing, graphical draftsmanship, calculation, and bookkeeping really require more 'skill' in the abstract than these occupations in their noncomputerized forms? Does it really raise the value of such skills where they already exist?" (p. 32). He also notes that, "In competitive theory, a worker's wage depends not on the amount of raw product that he or she can loosely supervise, but on the degree of difficulty in replacing that worker for that purpose" (p. 33).
30. During the time income inequality was increasing, Galbraith (1998) also notes, productivity in the U.S. economy was stagnant. This led him to the following observations:
"The idea of skill-biased change that does not raise productivity on average leaves some big questions unanswered. Why would business firms adopt new technologies if they actually took this form? ...Even more troubling, why should (and why does) society as a whole tolerate technological change that merely creates inequality without raising even the average standard of living? I have not seen any effort by partisans of this approach to come to grips with either of these intriguing questions" (Galbraith, 1998, p. 30).
31. Galbraith's (1998) book presents a large number of statistical analyses, of which I shall share just one (see his Chapter 8). Using multiple regression, he attempts to explain variation in annual levels of U.S. inequality (using different measures of inequality, such as the "Gini" index) with various macroeconomic predictors (unemployment rate, economic growth, inflation, real exchange rate, and the level of the minimum wage). R-square values ranged from .72 to .87, depending on the particular inequality measure used as the criterion. One of the conclusions Galbraith draws is the following:
"...the fact that almost 90 percent of the variation in manufacturing wage inequality as measured by this series can be accounted for by these variables leaves little else to discuss. There is little left over for other forces, such as changes in education or the supply of skill, to explain" (Galbraith, 1998, p. 145).
32. Economic outcomes result from what is clearly a complex mixture of societal and individual factors. Although individuals' intelligence levels certainly play some role in economic inequality, Galbraith's (1998) arguments call into question whether that role is as large as TBC and other writings by Murray suggest.
33. Buckhalt (2000) and Murray (2000) appear to suggest that my target article may have represented a premature attempt to impose closure on the debate over TBC (one cannot help notice how Buckhalt's commentary, written concurrently with the protracted dispute over the Florida voting in the U.S. presidential election, is laden with election imagery such as "referendum" and "count the ballots").
34. Buckhalt (2000) states that, "...I have serious doubts that the evidence reviewed constitutes the last word on the topics." Let me reassure Buckhalt and all other readers that so do I! There is no question in my mind that research into the issues raised by TBC will - and should - continue. However, given Murray's specific challenge and the rapidity with which the relevant literature continues to expand, I also feel that a taking of stock five years post-TBC was useful.
35. Murray (2000, paragraph 14) invokes his three-stage framework, originally specified in the Afterword to the paperback version of TBC, of how he envisioned the debate over TBC would unfold. He writes in his recent commentary that, "...I am puzzled that Reifman could think the three-stage sequence... has gotten past the first stage."
36. With this last characterization I must respectfully, but firmly, dissent. According to Murray's description in TBC's Afterword, hallmarks of first-stage criticism of TBC include "invective and dismissiveness," with critics feeling "no need to be judicious" (p. 557).
37. For whatever reason, Murray does not credit the many sophisticated, thorough, and careful studies cited in my target article (studies that were published in such prominent journals as Nature, Psychological Bulletin, and American Psychologist) as taking the debate beyond the first stage. I believe we are well beyond the first stage in evaluating TBC, and I look forward to reading future contributions in this area.
[1] As an aside, some Bell Curve aficionados may find it of interest that Heckman received the 2000 Nobel Prize in economics (see report at: http://www.nobel.se/economics/laureates/2000/ ).
[2] Although some readers of Feldman et al.'s description of cultural transmission may find it to appear similar to the notion of shared environment as used in the psychological literature, they are not the same thing. Instead, cultural transmission contributes to shared environment, as individuals within a family share cultural influences. I would like to thank Sarah Otto for attempting to clarify cultural transmission for me.
[3] Again as an aside, some may find it of interest that Galbraith is the son of John Kenneth Galbraith.
Arrow, K., Bowles, S. & Durlauf, S. (Eds.) (2000). Meritocracy and economic inequality. Princeton: Princeton University Press.
Buckhalt, J.A. (2000). A call for a new referendum on conclusions of the bell curve. PSYCOLOQUY 11(128) ftp://ftp.princeton.edu/pub/harnad/Psycoloquy/2000.volume.11/ psyc.00.11.128.bell-curve.4.buckhalt http://www.cogsci.soton.ac.uk/cgi/psyc/newpsy?11.128
Cavalli-Sforza, L.L. (2000). Genes, peoples, and languages. New York: North Point Press.
Dawkins, R. (1976). The selfish gene. Oxford: Oxford University Press.
Devlin, B., Daniels, M., & Roeder, K. (1997). The heritability of IQ. Nature, 388, 468-471.
Dickens, W.T., & Flynn, J.R. (2001). Heritability estimates versus large environmental effects: The IQ paradox resolved. Psychological Review, 108, 346-369. http://www.apa.org/journals/rev/rev1082346.html
Feldman, M.W., Otto, S.P., & Christiansen, F.B. (2000). Genes, culture, and inequality (pp. 61-85). In K. Arrow, S. Bowles & S. Durlauf (Eds.), Meritocracy and economic inequality. Princeton: Princeton University Press.
Flynn, J.R. (2000a). IQ trends over time: Intelligence, race, and meritocracy (pp. 35-60). In K. Arrow, S. Bowles & S. Durlauf (Eds.), Meritocracy and economic inequality. Princeton: Princeton University Press.
Flynn, J.R. (2000b). IQ gains and fluid g (comment). American Psychologist, 55, 543.
Galbraith, J.K. (1998). Created unequal. New York: Free Press.
Heckman, J.J. (1995). Cracked bell. Reason Magazine. (http://www.reason.com/9503/dept.bk.HECKMAN.text.html ).
Herrnstein, R.J., & Murray, C. (1994). The bell curve: Intelligence and class structure in American life. New York: Free Press.
Jensen, A.R. (2000). Some recent overlooked research on the scientific basis of "The bell curve". PSYCOLOQUY 11(106) ftp://ftp.princeton.edu/pub/harnad/Psycoloquy/2000.volume.11/ psyc.00.11.106.bell-curve.3.jensen http://www.cogsci.soton.ac.uk/cgi/psyc/newpsy?11.106
Korenman, S., & Winship, C. (2000). A reanalysis of The Bell Curve: Intelligence, family background, and schooling (pp. 137-178). In K. Arrow, S. Bowles & S. Durlauf (Eds.), Meritocracy and economic inequality. Princeton: Princeton University Press.
McGue, M. (1997). The democracy of the genes. Nature, 388, 417-418.
Murray, C. (1998). Income inequality and IQ. Washington, DC: American Enterprise Institute.
Murray, C. (2000). Heritability and the independent causal role of IQ in "the bell curve". PSYCOLOQUY 11(105) ftp://ftp.princeton.edu/pub/harnad/Psycoloquy/2000.volume.11/ psyc.00.11.105.bell-curve.2.murray http://www.cogsci.soton.ac.uk/cgi/psyc/newpsy?11.105
Neisser, U. (1998). Introduction: Rising test scores and what they mean. In U. Neisser (Ed.), The rising curve: Long-term gains in IQ and related measures (pp. 3-22). Washington, DC: American Psychological Association.
Reifman, A. (2000). Revisiting The Bell Curve. PSYCOLOQUY 11(099) ftp://ftp.princeton.edu/pub/harnad/Psycoloquy/2000.volume.11/ psyc.00.11.099.bell-curve.1.reifman http://www.cogsci.soton.ac.uk/cgi/psyc/newpsy?11.099
Rushton, J.P. (1999) Secular gains in IQ not related to the g factor and inbreeding depression unlike Black-White differences: A reply to Flynn. Personality and Individual Differences, 26, 381-389.
Stoolmiller, M. (1999). Implications of the restricted range of family environments for estimates of heritability and nonshared environment in behavior-genetic adoption studies. Psychological Bulletin, 125, 392-409.
Wilson, E.O. (1999). Consilience: The unity of knowledge. New York: Vintage.