Part III of this series examines perhaps the two most controversial and misunderstood aspects of intelligence: heritability and test bias. Full books could, and have, been written on both of these topics, so obviously the information presented here is not all encompassing. I’ll try to give a balanced view of both issues while making it clear what the research shows, even if it slightly contradicts the views of current pop-science.
Heritability has a very precise definition in the world of genetics – it is defined as the proportion of the variance of a particular phenotype that can be attributed to variance in genotype. Although often misrepresented as such, heritability is not the percentage of a phenotypic trait that can be explained by genetics. To clarify the difference, think of feet. The number of feet on a human is a trait that has negligible variance from person to person, and therefore a near zero heritability, but it is a trait that is almost totally controlled by one’s genetic makeup. Foot size however, varies from person to person and that variance is largely due to genetics, so it has a high heritability.
Investigating the heritability of intelligence is basically asking if the difference in intelligence from person to person is more a product of genetics or environment. It’s asking if math genius is the result of genes, or a product of Sesame Street.
Actually studying heritability is facilitated by the existence of twins, both identical (monozygotic or MZ), and fraternal (dizygotic or DZ). As the name suggests, identical (MZ) twins are genetically indistinguishable, and fraternal (DZ) twins are no more genetically alike than normal siblings. Comparing the correlations of IQ scores between twins gives some indication of the heritability of intelligence. A high heritability means that the more genetically similar the pair, the higher the IQ correlation should be.
Numerous studies indicate that the IQs of MZ twins are more highly correlated than those of DZ twins, with one comprehensive meta-analysis putting the correlation at .81 and .59 respectively (McCartney, et al., 1990). Though suggestive, looking at MZ versus DZ twins only falls short of conclusiveness because environmental factors must be taken into account. For example, it is known that MZ twins share more similar environments than DZ twins (Loehlin and Nichols, 1976), so it is conceivable that the difference in IQ correlation could result solely from environmental factors.
In an attempt to take environment out of the picture, scientists look at siblings who have been raised apart from each other due to an early childhood adoption. In theory, MZ twins who have been raised separately would share totally uncorrelated environments (see note 1), so the correlation between their IQ scores should be a direct estimate of heritability. Figure 1 sums up the results of one twin study conducted by Bouchard and McGue in 1981.
|Relationship||Average IQ Correlation|
|MZ twins raised together||.86|
|MZ twins raised apart||.72|
|DZ twins raised together||.50|
|Siblings raised together||.47|
|Siblings raised apart||.24|
The data from this study shows exactly what one would suspect if intelligence has a high heritability – the more genetically similar the people, the higher the correlation.
Another point of interest is that the heritability of IQ tends to increase with age. The data from Bouchard and McGue displayed above shows that the IQ correlation for siblings raised apart was around .24, but further studies indicate that is increased to about .47 in adulthood. Likewise, IQs of unrelated people who were raised together correlate around .25 in childhood, but decrease to about zero when they grow up (McGue et al., 1995; Paul, 1980; Scar, 1989; Segal, 1997; Teasdale & Owen, 1984). This data indicates that intelligence, a highly heritable trait to begin with, increases in heritability as one ages, probably due to the fact that adults are much more in control of their environments than children.
Intelligence itself is a phenotypic trait – it’s the manifestation of certain behavior from the complex interaction of genetics and environment. The above data indicates that the majority of the variance in IQ is caused by genetics, not the social class of your parents, not your early schooling, not your self-esteem. A common analogy is that intelligence is the total area of a lawn, with genetic components being its width and environmental influences being its length.
Are IQ tests culturally biased? I posed this question to a number of people I encounter on a daily basis, and the answers ranged from “a little” to “of course.” It’s really not surprising that the majority of people think IQ tests are biased – the hoopla surrounding the Bell Curve and the resultant public backlash has all but branded IQ tests instruments of neo-Nazi propaganda. Couple that with the idea of “multiple intelligences” and the politically correct reluctance to give anyone a score without a caveat, and you’ve got a sizeable population that thinks IQ tests are just a two-and-half hour, individually administered pat on the back for the white upper middle class.
Let’s put on our nerd pants and try to examine the idea of test bias from a scientific and mathematical standpoint. First, a definition: bias is the systematic over- or under-estimation of whatever true value is being measured. Supposed two groups of people, group A and group B, are given an IQ test called T. Now imagine group A consistently scores far below group B. In and of itself, that is not an indication of test bias; however, if it is found (using some hypothetical perfect measure of intelligence) that both groups are actually equally intelligent, it would then be fair to say text T is biased.
Since bias is concerned with group differences in IQ scores, and the biggest systematic group difference in test scores are between the sub-Saharan African population (often referred to as blacks in the literature) and the European Caucasoid population (called whites), I’ll spend a little time talking about this controversial gap. Most studies agree that if the mean IQ of the American white population is set to be 100 with a standard deviation of 15, the black population has a mean IQ of about 85 with a standard deviation of 12 (Jensen, 1998; shuey, 1958, 1966; Loehlin et al., 1975; Kaufman and Doppelt 1976; Reynolds et al. 1987). As above, with group A and group B, this score differentiation doesn’t necessarily indicate test bias but it certainly requires further investigation.
In reality, there is no exact measure of intelligence to which we can compare IQ tests (or else we wouldn’t have IQ tests in the first place). However, since IQ tests were originally designed to see how children did in school, one possible way to determine if a test is biased is to look at scholastic achievement. If a Caucasian and an African-American student both have IQs of 100 they should end up doing about equally as well in school. However, if one group’s IQ score underestimates their scholastic performance, then it gives at least some reason to believe IQ is a biased metric. As it turns out, IQ tests do accurately predict the scholastic achievement for just about everyone, including whites, blacks, and other ethnic minorities (Herrnstien and Murray, 1994; Jensen, 1980; Mackintoch and Mascie-Tayloe, 1986).
Another way to validate the results of IQ tests is by looking at performance on the job versus IQ score. The results are the same – IQ tests do not underestimate minority job performance (Jensen, 1980; Hunter et al., 1984; Hartigan and Wigdor, 1989).
Looking at the data, some may protest that the entire structure of American society is set up to favor the white majority so IQ and achievement are both insidious ways of predicting intelligence. The same people who raise this objection probably don’t realize that Asian-Americans tend to outscore Caucasian-Americans by a few points on IQ tests, and also have the higher scholastic and occupational achievement that this would suggest.
Some people suggest that if IQ tests aren’t biased with regard to ethnicity they are at least biased with regard to socio-economic class. This is another case where the evidence contradicts popular opinion. It’s true that the correlation between an adult’s social class and IQ is around .5-.6, but the correlation between a child’s IQ and the social class of his parent’s is only around .3 (Waller, 1971). Another study showed that children of working-class mothers who, through adoption, were raised by middle-class families scored an average of 8 points lower than their adoptive siblings (Dumaret and Stewart, 1985). These studies imply that intelligence predicts social class, not the other way around. If that’s the case, then it means IQ tests don’t discriminate against the poor, but capitalism discriminates against the unintelligent.
A Quick Aside
If the difference in IQ scores between African-Americans and Caucasians is not due to test bias then what is it due to? Is it due to an actual difference in intelligence? That question isn’t easy to answer and I don’t feel right giving it just a cursory examination in this article (though discussion in Avant’s forums would be a good idea). For further information I suggest consulting The G Factor: The Science of Mental Ability, and Bias in Mental Testing, both by Arthur Jensen, Bell Curve by Richard Herrnstein, and The Black-White Test Score Gap by Christopher Jencks and Meredith Phillips.
When there is a Bias
Even if test bias is overstated, it would be exceedingly myopic to dismiss it altogether. The data shows that in the current testing population there is little reason to believe that bias exists, but it says nothing about other cultures that are extremely different than those of Western, industrial nations. N. J. Mackintosh presents a few examples of IQ tests failing to give meaningful results in his book IQ and Human Intelligence. In one case the Soviet psychologist Alexander Lura presented this syllogism to a group of illiterate Central Asian peasants:
In the far north, where it snows, the bears are white.
Nova Zemblya is in the far north, and it is always snowy there.
What colour are the bears in Nova Zemblya?
He would get replies such like:
How should I know? I have never been to the north.
Why are you asking me? You have traveled and I have not.
So-and-so said the bears were white, but he is always lying
Another example presented by Mackintosh is the case of an African peasant who was asked to sort a knife, fork, spoon, orange, apple, and banana into two groups. He replied that a wise man would sort the knife with the orange because it is the tool needed to cut it. When asked how a fool would sort the items, he unhesitatingly separated the fruit and the utensils into two different groups.
Although the peasants would have scored poorly on this test, it is obviously not indicative of a people incapable of reason. There is reasoning in the responses, even if it is not being applied in the way the psychologists would like. It’s like giving a baseball player a golf club and being amazed by the fact that he can’t hit a 300 yard drive…. After all, he knows how to swing.
More Explanation on Test Bias
Notice that the above discussion of test bias didn’t touch on individual test items at all. This is important because some people will look at an IQ test and decide for themselves whether or not it is biased based on the appearance of its items rather than empirical evidence. For example, the Information subtest of the WAIS-R which contains questions like “Who was Napoleon Bonaparte?” might seem to be a very biased and poor indicator of intelligence, but it is actually the second best predictor of IQ of the entire test (Gignac & Vernon, 2002).
Those sort of questions work because intelligence is phenotypic in nature, meaning it is a product of both environment and genetics (though we’ve already established that genetics cause the majority of the variance in IQ). Even if someone lives in a poor neighborhood he most likely has the opportunity to attend school, and even if the schooling he receives is second-rate he’ll still at least be exposed to the most basic information. The intelligent person is likely to remember this information (even without trying) and, years later, regurgitate that information to some psychologist. I guess third-grade geography really is useful after all.
All this talk about intelligence and IQ theory is interesting (well, for me), but what fun is theory without practice? To round off this series of articles I’ve made an intelligence test specifically for the Avant readership. I decided to host the test on my site and require registration to help preserve the integrity of the test. I’ll start a thread in the forum and update it as norms and averages become available. For more information, and for the actual test, please visit [Test has been removed].
Note 1: The environment or separated MZ twins are almost never totally uncorrelated because of placement bias. It’s more likely, for example, for a twin with middle-class parents to be placed in a middle-class household. In The G Factor Jensen draws on data from the Minnesota twin study to show that this bias can only account for an IQ correlation of about .1.