I.  The Issue

 A.  In 1969, A. R. Jensen published a paper in the Harvard Educational Review (39: 1-123) entitled "How
       much can we boost IQ and scholastic achievement?"

  1.  Showed that IQ scores differed for blacks and whites

  2.  Showed that there is significant genetic variation for IQ within homogeneous populations
       (e.g. middle-class whites).

  3.  Concluded that the differences between whites and blacks in average IQ score are therefore
       genetically based.

  4.  This conclusion led him to conclude that it was unlikely that educational intervention would
       be successful in "bringing black intelligence up to white levels" (not his quotation)--
       essentially, if a trait is genetically determined, it is immutable and not subject to alteration
       by environmental intervention.

 B.  In 1994, Richard J. Herrnstein and Charles Murray published "The Bell Curve", which expanded on
       this argument and had 400,000 copies in print just two months after its publication.

 C.  These arguments have been attacked repeatedly on a number of fronts, including assertions that:
 

  1.  IQ tests do not measure anything meaningful about intelligence.

  2.  IQ tests are culturally biased, and therefore give false readings of intelligence differences
       among ethnic groups.

  3.  The data used for many of the analyses described by Jensen and by Herrnstein and Murray is
        riddled with ambiguity, and in some cases appears to be fabricated.  (See "Behind the Curve"
        from Scientific American--Required Reading)
 

 D.  At its core, however, the central assertion that blacks and whites differ innately in intelligence is an
       evolutionary issue: two groups supposedly differ in some trait because of underlying genetic
       differences, which by definition is evolutionary divergence.

 E.  As evolutionary biologists, then we should be able to evaluate validity of both the premises and the
       inferences in Jensen's and Herrnstein and Murray's arguments.

 F.  We shall consider the evidence, as well as the logic of these author's underlying deductions, and will
      come to three conclusions:
 

  1.  The premises claimed by the authors as true are, in fact true, i.e., the evidence is pretty clear
       that
   a.  blacks score lower on IQ tests than whites, and
   b.  IQ is heritable within groups, i.e. there is variation among genes that affects IQ.
 
  2.  The logical deduction that the truth of these two premises implies that racial differences in IQ
        are also genetically based is flawed.

  3.  Even if it were true that racial differences in IQ were due in part to underlying genetic
       differences, it is not true that environmental intervention could not eliminate this difference,
       and even raise everyone's IQ, assuming that high average IQ is a desirable goal.

 

 A.  Differences between black and white IQ scores.
 
  1.  There are many studies indicating blacks score lower on IQ (or similar) tests than an
       "equivalent" white group.

  2.  Example: The National Longitudinal Survey of Labor Market Experience of Youth (or,
       for short, the "NLSY").
 

   a.  Consisted of 12,686 youths between 14 and 22 when the study began in 1979.
   b.  Individuals were initially chosen to provide adequate sample size for various
        socio-ethnic groups, including blacks, Latinos, low-income whites, etc.
   c.  94% of the participants were administered the Armed Forces Qualification Test
        (AFQT), which is considered to be a test of "general intelligence", and thus at
        test of "IQ".
   d.  The frequency distribution of scores for blacks and whites are shown in the
         following figure:

   e.  One obvious potential explanation for this racial difference is that, on average,
        blacks and whites come from different socio-economic classes (e.g. blacks, for
        reasons of history and societal structure, are more likely to belong to the "lower"
        socio-economic classes), and that differences in educational opportunities among
        social classes may actually be the cause of racial differences in IQ.

   f.  Attempts have been made to "correct" for such biases.  For example, Herrnstein
       and Murray present a graph showing that factoring out parental socio-economic
       status (SES) still shows that blacks have a lower IQ than "equivalent" whites:

   g.  It should be kept in mind that no matter how many such analyses one performs,
        there may always be one or more unexamined environmental factors that can't be
        "controlled" in this way.  Nevertheless,

  3.  The overall data supporting the notion that blacks score lower on IQ tests than whites is fairly
       convincing.

 B.  Heritability of IQ.
 
  1.  Twin studies, despite the bad reputation they have earned in conjunction with Cyril Burt's
        fradulent work, provide the best (and practically sole) evidence regarding the heritability
        of IQ.

  2.   Example:  comparison of monozygotic and dizygotic twins.
 

   a.  Measure of similarity in IQ between twins: intraclass correlation--measures
        how closely twins resemble each other for a particular character.
 
    i.  A value of  0 indicates that twins are no more similar than randomly chosen
        strangers
    ii.  A value of 1 indicates that twins are exactly alike.
 
   b.  Correlation between twins is due to two causes:
 
    i.  inheriting the same alleles affecting the trait of interest
    ii. growing up in the same environment


   c.  When monozygotic and dizygotic twins are compared, the contribution of
        environment to the correlation should be similar.

   d.  By contrast, when the genetic contribution to the correlation is considered, expect
        this contribution to be greater for monozygotic twins because they are genetically
        identical.

   e.  Hence, the difference between the correlations for monozygotic and dizygotic
        twins should reflect the degree of genetic control over the trait:



   f.  Data summarized by Cavalli Sforza and Bodmer indicate that Monozygous twins
        (MZ) tend to have higher intraclass correlations for IQ than dizygous twins (DZ):

                             Author                  MZ                 DZ
                            Newman et al.      0.886             0.6 31
                            Shields                 0.76               0.51

   g.  This data indicates a fairly strong genetic component to differences among
         (non-twin) individuals in IQ.
 

  3.  Example:  comparison of monozygotic twins separated and reared together
 
   a.  Using the same framework as in the previous example,  the contribution of
        genetic variation to the intraclass correlation coefficient for IQ should be
        the same for twins reared together as for twins reared apart, since in both
        cases, the twins are genetically identical.

   b.  By contrast, the contribution of the environment to the correlation should be
        smaller for twins raised apart, since the the environment experienced by separated
        twins is probably more different than the environment experienced by twins
        reared together.

   c.  This means that we expect the intraclass correlation coefficient to be sustantially
         smaller for twins reared apart if home environment has a large effect in IQ, relative
         to the effects of genetics.

   d.   There is such an effect, as revealed by several studies (Data from the Minnesota
          Twin Project; Bouchard, T. J.  et al. 1990.  Science 250: 223-228 below):

                                                                        intraclass correlation, r
                                         Test                       MZ apart       MZ together

                                 WAIS IQ--full                      0.69            0.88
                                 WAIS IQ--verbal                 0.64            0.88
                                 WAIS IQ--performance       0.71             0.79
                                 Raven, Mill-Hill Comp.         0.78             0.76

   e.  However, the effect is rather small, suggesting that only a small portion of the
        variation among individuals (non-twins) in IQ is explainable by differences in
        home environment;  this result is consistent with the notion that genetic variation
        contributes importantly to IQ variation.

 

 A.  The real question: do blacks and whites differ inherently (genetically) in intelligence (IQ)
 
  1.  This is a question that evolutionary biologists deal with all the time: do two groups
       differ genetically with respect to some character.
 
  2.  And evolutionary biologists normally answer this question using a very simple tool: the
       common garden experiment (Remember Clausen, Keck and Hiesey)

  3.  Unfortunately, experimentation on humans, even relatively benign common garden
       experiments, are prohibited for moral reasons, which means that we do not have this
       tool available to answer the basic question.

  4.  Consequently, some people are led to using logical reasoning to deduce whether groups
       differe genetically for a character.

  5.   In most cases this approach will not work.
 

 B.  The faulty logic of Jensen and Herrnstein and Murray
 
   1.  Demonstrating that this logic is faulty requires demonstrating that the premises, which
       we have conceded to be true, do not imply the conclusion that racial differences in IQ
       are genetically based.  In particular, such a demonstration requires showing two things:
 
   a.  Heritability of IQ within groups does not imply genetic differences between groups
   b.  Genetic differences between groups does not imply heritability within groups.
 
  2.  This is equivalent to showing that populations may exhibit any of the four possible
       combinations of genetic variation, or lack thereof, within groups and genetic variation, or
       lack thereof, between groups.
 
 C.  To demonstrate that individual populations may exhibit significant heritable variation for a trait and
       average phenotypic differences between populations, without there being any
       genetic differences underlying those average differences in phenotype between populations,
       consider the following thought experiment employing maize (corn), with the character of interest
       being plant height on the 4th of July:
 
  1.  Suppose that we start out with a bag of genetically variable corn seeds (we are assuming
       this genetic variation affects plant height).

  2.  Next, let us choose two samples of 1000 seeds randomly from the same bag.

  3.  Let us plant one of the samples out in a field that is plowed but not fertilized, and let
       us plant the other sample in a field that is plowed and fertilized.

  4.  Then, we go out on the 4th of July and measure the height of every plant in both fields.

  5.  First, consider what we will see in the unfertilized field, if we plot a frequency histogram of
       number of individuals in a given height category:
 

   a.  We will see a more-or less bell-shaped distribution of heights.
   b.  Moreover, we suspect some of that variation in height is genetically based.
 
  6.  We can test this assumption by using standard quantitative genetic analyses  to estimate the
       heritability of height in this population. (See Lecture 11)
 
   a.  One way--mate each of a number of randomly chosen plants acting as pollen parents
        (males) to a randomly chosen plant acting as a seed parent (female), collect, say, 20
        offspring seeds from each mating, and plant them out in the same field the next year.
   b.  A parent-offspring regression will then indicate the heritability, and thus how much
         of the overall variation in height seen in the first year is due to underlying genetic
         variation.
  7.&ginal seed was genetically variable, that this analysis will
       confirm that the sample planted in the unfertilized field was genetically variable.

  8.  Next, consider what we will see in the fertilized field:
 

   a.  Again, we will see a bell-shaped distribution of heights, but
   b.  The mean height is greater than in the unfertilized field.
   c.   This difference in mean is due to an environmental effect on the character, in this
          case an effect of fertilization causing an increase in plant height.
 
  9.  We could again explicitly confirm the presence of genetic variation for height in this sample
       the same kind of parent-offspring analysis described for the unfertilized field, and we would.

  10.  We have thus set up a situation in which the following statements are true:
 

   a.  There is genetic variation for height within each population (sample), and
   b.  The two populations differ in mean height, yet
   c.   The two populations do not differ, on average, in genetic constitution.
 
  11.  We know proposition c. is true because the two samples of corn seed used to set up the
         experiment were drawn from the same bag of seeds, and the number of seeds in each
         sample (1000) is too large for there to be substantial differences in genetic composition
         between the samples.

  12.  In other words, we have just shown that it is possible to have genetic variation for a trait (e.g
         IQ) within populations, differences in the mean of the trait between populations (races),
         without there being differences in genetic composition between the populations (races).

  13.  This demonstration proves that Jensen's and Herrnstein and Murray's conclusion is not
          implied by their premises.
 

 D.  To complete the argument, though, we would like also to be able to show that absence of genetic
        variation within groups does not imply absence of genetic variation between groups.
 
  1.  Again, we can do this with a very simple thought experiment.

  2.  Suppose we start with two different bags of seeds.
 

   a.  One bag is produced by crossing two highly inbred lines (the standard way
        of producing the hybrid corn seed that most farmers grow), which yields F
        individuals that are genetically identical.
   b.  The other bag is produced by crossing two different highly inbred lines, lines
         that produce plants shorter than either of the lines used to produce the seeds
         in the first bag.  All the seeds in this bag will also be genetically identical to
         each other.
 
  3.  However, the seeds from one bag will be genetically different from each other because
       they were formed from different inbred lines that contributed different alleles affecting
       height.  So we start out with two "populations" that we know are genetically different
       from each other with respect to the trait of interest.

  4.  Now, we take the seeds from both bags and plant them in the same field, with their
       positions randomized (why is this important?), keeping track of which bag each
       seed, and the resulting plant, is from.

  5.  On the 4th of July we measure the height of every plant in the field and construct a
       frequency histogram of plant height for the two populations:
 

   a.  The two populations will have different mean heights because they are
        genetically different.
   b.  However, we set the experiment up so that there is no genetic variation within
        either population. (This could, of course, be confirmed by parent-offpsring analysis.)
 
  6.  We have thus just shown that it is possible to have two populations that differ genetically for
       a trait without either population exhibiting any heritability for that trait.
 
 E.  Conclusion:  Knowledge of how much genetic variation affecting a trait exists  within
       populations tells us nothing about how much genetic varation contributes to between-population
       differences in that trait, and vice versa.

 F.  Implication with respect to IQ:  There is no good reason to believe that racial differences in mean IQ
      score are caused by underlying racial differences in intelligence.

G.  General Conclusion:  Anyone versed in these simple principles of evolutionary genetics could have
      recognized the invalidity of Jensen's and Herrnstein and Murray's logic.

 A.  A second inference drawn by J and by H&M is that, given that the racial difference in IQ scores is
       genetically based, environmental intervention (e.g. education) would not be effective in narrowing
       the racial gap in intelligence.
 
  1.  This inference clearly has tremendous social implications, and it therefore behooves one
        to consider whether this is also a legitimate inference, independent of whether the premise
        is true or false.

  2.  In particular, one would like to know whether it is hopeless to elevate the average intelligence
       of a racial group were racial differences in IQ to prove genetically based.

  3.  Again, J's and H&M's conclusion that it is hopeless is based on a fundamental misunderstanding
       of what it means for differences to be genetically based, and in particular on failure to
       understand or appreciate the common occurrance of genotype x environment interactions.
 

 B.  What are genotype x environment interactions?
 
  1.  The phenotype produced by an individual of a given genotype often depends on the
        environment in which the individual finds itself.

  2.  The simplest manifestation of this principle is that the environment affects the phenotype:
       e.g. fertilizing a plant will, on average, increase height, so that a given genotype can
       produces different height phenotypes, depending on the fertility of the soil in which an
       individual grows.

  3.  In general, most genotypes respond to fertilizer in a similar way, by increasing vegetative
         size.

  4.  However, this pattern is not always the case for other characters.

  5.  Example:  metal tolerance in plants.
 

   a.  Mine tailings in general are contaminated with heavy metals (lead, copper, silver),
        which are toxic most plants and greatly reduce growth and survival.
   b.  In some plant species, however, natural mutants tolerant to heavy metals occur,
        and plants carrying these mutant alleles are able to grow on mine tailings.
   c.  The ability to grow in soils contaminated with heavy metals is achieved at a price,
        however, and that price is reduced competitive ability, so that genotypes that
        tolerate heavy metals on mine tailings do not grow well in adjacent pastures and
        fields.
   d.  So, if we plot a character such as height for tolerant and non-tolerant genotypes
        when grown in two environments (e.g. tailings and pastures), we would get
        something like this:
   e.  What this shows is a genotype x environment interaction for height:  the difference
        in height between the two environments is not the same for the two genotypes.  In
        particular, in this case the rankings of the two genotypes for height change from
        one environment to the next.

  6.  More generally, there are many different patterns a g x e interaction may take:

  7.  In all cases, note the difference of a difference:  the difference between the character in
       to environments for one genotype differs from that difference for the other genotype.

  8.  A concrete example:  Bristle number in Drosophila (Gupta and Lewontin. 1982.  Evolution
       36: 934-948)
 

   a.  Grew fly larvae from different strains (genotypes) under three different temperatures
        (14, 21 and 26°C) and counted the number of bristles on the fourth and fifth
        abdominal sternite.
   b.  Separate plots for both males and females indicate the presence of a G x E
        interaction:

 C.  The potential importance of G x E interactions for IQ
 
  1.  The virtual ubiquity of differences in relative performance of different genotypes in
       different environments for a wide range of characters suggests that they may also
        occur for traits like IQ.

  2.  The importance of this is that just because genetic differences causing differences in IQ are
        expressed in one environment, doesn't mean that they will be expressed in other
        other environments, or even that the relative ranking of IQ scores won't be reversed in
        other environments.

  3.  It is thus quite possible that educational, or other environmental, intervention might boost
       IQ scores of blacks up to or beyond that of whites.

  4.  In no way, then, can evidence for a genetic basis to racial differences in IQ under current
       environmental conditions be taken to imply that intervention can not eliminate those
       differences.

  5.  In fact, a story that appeared quite a few years ago suggests one possible way that
       environmental intervention might boost IQ's of both blacks and whites up to more or
       less equal levels.
 

   a.  A study reported that the physiological susceptibility of black children to adverse
        effects on mental development of lead in their environment (primarily in old paint)
        is greater than the physiological susceptibility of whites.
   b.  This report suggests first that a genetically based physiological difference in
         susceptibility to an environmental factor could contribute to an apparent
         genetically based racial difference in IQ.
   c.  It also suggests that by that by ridding the environment of lead, mental ability (IQ) of
         both black and white children might be enhanced, but that that of blacks would
         be enhanced more.
   d.   In terms of norms of reaction, the following situation is consistent with this
         scenario:
 

 
   e.  In other words, genetic differences in IQ present in one type of environment should
        not been taken to indicate that changes in the environment could not eliminate those
        differences.