The Mismeasure of Man

Book club notes

This book argues against the idea that intelligence can be meaningfully abstracted as a single number capable of ranking all people on a linear scale of intrinsic and unalterable mental worth. These types of investigations of intelligence usually found that oppressed and disadvantaged groups - races, classes, or sexes - are innately inferior and deserve their status.

There is a strong desire of people to believe that social and economic roles accurately reflect the innate construction of people, and these ideas linger today.

There were some early scientists who fought against these notions: notably Humboldt and Wallace (co-discoverer of natural selection).

Before the theory of evolution was established, there were two main styles of scientific racism: monogenism and polygenism.

Monogeny held that humans emerged from one source, Adam and Eve. From that initial perfection, humans degenerated, some more than others due to environmental factors, but that now that they’re degenerated, they can’t go back.

Polygeny held that humans emerged from different sources, but this went against scripture so wasn’t as popular. It was one of the first theories of American origin that got the attention and respect of European scientists.

Samuel George Martin gained a reputation as a great data-gatherer and objectivist. He was a highly respected scholar and gathered one of the most impressive collections of human skulls in the world. He ranked races by the cranial capacities of their skulls, which he measured by pouring various granular solids into them. These measures were reprinted repeatedly as hard data in the 19th century on the relative rank of human races. They confirmed existing biases: whites on top, Indians (Native Americans) in the middle, and black people on the bottom. Among whites, the ordering also confirmed existing biases.

Gould re-analyzed Morton’s data and found several problems, but concluded that they were unconscious mistakes because Morton was really transparent about publishing his original data. But the fact remains, all his mistakes were in the direction he wanted to believe. Gould makes the point that if the a priori biases of a scientist can affect something as basic and objective as volume measurements, they can creep into anything.

Problems with Morton’s data:

Some groups are very over-represented compared to others. His native American samples contained almost entirely small-brained Inca Peruvians, and very few large brained Iroquois for example. If this happened to be reversed, he would have found a much higher average for Native American brains. But, when faced with a similar problem with an oversampling of small Hindu brains (counted as Caucasian), he left most of them out, reasoning that they had exceptionally small brains and thus did not represent white brains generally

He did not correct for body size. He had the data to see the close association of body size and brain size, but he never even considered that brain size might represent anything other than inherent intelligence

On the same note as above, he failed to make sure sexes were equally represented in all groups. Half of his Caucasian samples were male, but only a third of his black samples were male. Thus, females have smaller brains on average, reflective of their smaller body size, and this difference contributed to the difference Morton found between black and white brains.

When Gould re-analyzed this data, the racial differences disappeared.

Robert Bennett Bean was a Virgina physician. He took special pride in his data on the corpus callosum. Following the tenet that higher mental functions reside at the front of the brain, Bean set out to rank races based on size of relative parts of the corpus callosum. So he measured length of genu (front of CC) and compared it to the splenium (back of CC). He found almost complete separation between back and white people, whites have larger genu, meaning more brains up front, and thus were smarter. Bean’s mentor became suspicious of the work and re-measured the CCs, but this time without knowing which brain belonged to which person beforehand, he found there was no difference between races at all. This was most likely fraud on Bean’s part.

Paul Broca was a professor of clinical surgery and founded the Anthropological Society of Paris in 1859. He firmly believed brain size was closely related with intelligence. He argued that the brain is larger in mature animals vs elderly and in men than women. Also, larger in superior races compared to inferior, based on the data discussed above (Morton’s and others). Though his data was gathered carefully Gould found when reading his work that he often came to a conclusion first, then used data to support that conclusion, what Gould calls the “usual route.” “By this route, the conclusions achieved not only the blessing of science, but the prestige of numbers. Broca and his school used facts as illustrations, not constraining documents.” Broca considered himself a liberal, in that he did not believe intelligence was unchangeable. Women’s brains had degenerated over time due to social underuse, but that didn’t mean they couldn’t be improved under different conditions.

One of his most famous studies concerned the steady increase in brain size as European civilizations advanced from medieval to modern times (1862). He did not find this steady increase and made all kinds of post-hoc rationalizations to still come to the conclusions he wanted to (for example, the more modern skulls were from a cemetery by a river, which meant many of those people may be drowning victims, drowning artificially lowers brain volume etc). When faced with the fact that Cro-Magnon brains seemed larger than modern ones, he temporarily abandoned brain size as a primary indicator of intelligence and focused more on when brain sutures close (later closing indicating higher intelligence). He extended this to differences between blacks and whites.

Much of Broca’s work concerned the differences between men and women. He found that men’s brains were much bigger, and that this difference has increased over time. His ancient samples were very limited (only 6 or 7 of each sex) so that’s not enough for a conclusion in any case. Moreover, he didn’t correct for body size, and his modern female samples were significantly older, and we know brain size declines as you age. Once these and other factors are accounted for, there seems to be no difference at all in brain size between the sexes (when corrected for body size). Gould takes issue with the fact that Broca knew taller men tended to have larger brains than shorter men, but made no move to say tallness was associated with intelligence, because he had no pre-conceived notion that that was the case.

Craniometric arguments began to lose luster in the next century, and determinists began to switch to intelligence testing as a more direct path to ranking groups by mental worth. But this too was plagued with many of the same biases.

Ernst Haeckel came up with the now famous “ontogeny recapitulates phylogeny” meaning that an individual, in its own growth, passes through a series of stages representing adult ancestral forms in their correct order. An individual climbs their own family tree, in a sense. This was seized upon by scientists who wanted to rank human groups. Black adults were compared to children of superior groups, for the child represents a primitive adult ancestor.

Cesare Lombroso was an Italian physician who came up with a theory of innate criminality. He was trying to find anatomical differences between criminals and regular men. He began to see atavistic features in the skulls of criminals recalling “an apish past rather than a human present.” He wasn’t just arguing that crime is hereditary, he was saying that criminals are evolutionary throwbacks in our midst. These people are innately driven to act as an ape would, it’s just that its no longer socially appropriate. We can spot them by their anatomical signs, which he called stigmata. These stigmata were just extreme values on a normal curve for various features in humans that are close to the average measures for those features in great apes. This is a fundamental error. For example, some people have longer arms than others, and chimps have longer arms than humans on average. However, this doesn’t mean that people with longer arms are more genetically similar to apes than shorter armed people. Normal variation within a population is different from differences in average values between populations.

This affected social policy hugely. Lombroso and his disciples argued that rehabilitation was useless against stopping crime because these efforts would just be wasted on these atavistic throwbacks and they would commit crimes again.

Alfred Binet was the director of the psychology laboratory in Sorbonne, and was interested in measuring intelligence. He was admirably introspective about his own bias when he began these studies by measuring head sizes in school children. He decided to construct a series of tests to assess reasoning ability, this had not really been done before. In 1904, he was commissioned to design tests to figure out which school kids needed extra help. He made a scale that was a hodgepodge of various activities. He hoped that if he mixed together enough tests, he could assess a child’s general potential with a single score. He assigned an age level to each child, defined as the youngest age at which a child of normal intelligence should be able to do the task. So a kid starts with the youngest age tests and proceeds in order until they can’t do the test anymore. This mental age is divided by the chronological age to yield IQ. But Binet firmly maintained that this score was just an average of many performances, not an entity unto itself. His aim was always to identify kids who needed help in order to help them improve, not impose limits. He thought that intelligence can be augmented by a good education and is not an inborn and fixed quantity.

Unfortunately, Binet’s good intentions were thoroughly dismantled by American Hereditarians in the early 1900s.

Henry H Goddard was the director of the Vineland training school for Feeble-Minded girls and boys in New Jersey. He was the first to popularize Binet’s scale in America. He believed differences in intelligence were responsible for a wide range of undesirable behaviors because he believed deficient intelligence was linked to immorality.  He thought criminals, prostitutes and ne’er do wells were feeble-minded, and thus incapable of adapting to their environments. At the upper end, intelligent men rule by right. Goddard was working during the first flourish of excitement at the re-discovery of Mendel’s work and the deciphering of heredity. Unfortunately, the idea of genes was in its infancy and was viewed really simplistically. People didn’t understand that heredity can be complicated and depend on an interaction between many genes, and also that genes and environment interact to yield a given phenotype. He assumed “morons” had a deficient gene and should not be allowed to breed.

Goddard turned his attention to immigrants. He believed that in the early 1900’s, America was getting mostly “third class” immigrants, and most of them were probably morons. His tests revealed this to be the case, but the tests were done on scared people new to the country who often spoke no English. His testers were women who were skilled at spotting potential morons on sight, and often came to conclusions without more formal testing. This all led to a severe tightening of immigration in the next decade.

Later in his life Goddard softened his approach and said that many morons could be reformed with the right environment and training.

Lewis M. Terman really popularized Binet’s scale in America, though Goddard was the first to introduce it. He was at Stanford at the time, and gave his new scale a name we recognize today, the Stanford-Binet scale, which set the standard for virtually all IQ tests that followed. Terman added many items to Binet’s original tasks, and many of his questions were very culturally biased and did not accept reasonable original answers. For the next several decades, the inference was made that if a test correlated with the Stanford-Binet, it must also measure intelligence, without any investigation into whether the Stanford-Binet actually measures intelligence.

Terman relentlessly emphasized limits and their inevitability. Based on a 40 minute test, he would condemn a child to never being “an efficient worker or a responsible citizen.” He recognized the argument that IQ may be influenced by environment, but dismissed it by saying that while environment can have a role, it is far outstripped by the role of heredity. Also, if parents provide a good home, it’s because they possessed the superior heredity to get rich and provide a good home in the first place. It all comes down to heredity.

Robert M. Yerkes was a psychologist at Harvard, and he yearned for his field to shed its soft science reputation. Mental testing yielded quantified results, so he was attracted to it. Yerkes brought together all the major American hereditarians in his field (including Terman and Goddard) to write the army mental tests in 1917. Literate recruits were given the alpha, illiterates and men who failed the alpha would get the beta. Men would be graded, and the grades would be used for proper military placement. These were the first mass produced written tests of intelligence. The army did not use these tests much to determine rank, but they were used to reinforce existing biases. Black people scored worst, though the influence of slavery and oppression was not considered. Worryingly, the average score of a white male American put him at a mental age of about 13, just above “moron.” The tests also had a ton of problems. They were administered very inconsistently, with some officers not being able to hear the commands at all, many had not held a pencil before and had never taken a test and didn’t understand what they were supposed to do at all. They were heavily culturally biased as well. The summary statistics are very telling Yerkes expected that there would be a normal distribution, but for most tests, it is bimodal, with one peak at a middle value, and the other around the minimum, strongly implying that about half of recruits had no idea what to do on the test. His statistician, a psychologist named “Boring” actually amplified this flaw by correcting for zero scores by looking at other tests scores of the same individual. If the individual had scored poorly on other tests, the zero value was given a negative value. Interestingly, Yerkes actually found several correlations between environmental factors and scores on his tests, but instead of considering that the test might not be measuring inherent intelligence, he excused the association away in each case individually. For example, Yerkes found strong correlations between average scores and hookworm infestation but concluded that “low native ability may induce such conditions of living as to result in hookworm infection.”

Of note, Yerkes found a strong correlation between years of schooling and intelligence. But again, he argued that men with more intelligence spent more time in school. He had data showing that half the black recruits in Southern states had not attended school beyond the 3rd grade, but in northern states, this number was 2 grades higher. But he didn’t think about those regional differences, because he had such a strong notion already that blacks are inherently less intelligent than whites.

C.C Birgham was a disciple of Yerkes, and an assistant professor of psychology at Princeton. He became a primary vehicle of translating the army test results on group differences into social action. He admitted that alpha was subject to the effects of education, since you had to be literate to do it, but beta had no such limitation. He wrote a book defending many criticisms of the army tests. For example on the regional differences between Southern and Northern blacks, he said northern ones were still inherently smarter because they mixed more with white blood, and because the better conditions up there drew the smarter black people north. But eventually, he retracted all this because he recognized that the data were incredibly internally inconsistent. Nations had taken alpha and beta in different proportions, and the two tests couldn’t be compared. Thus, there was no way to make a comparison between nations. Also, he agreed that the tests were really culturally biased. But it was too late to undo the damage. Quotas had been established to slow immigration from southern and eastern Europe to a trickle. Throughout the 1930’s, many Jewish people tried to come to America, but were not admitted, condemning them to the Holocaust.

Cyril Burt was the official psychologist of the London County Council, and was responsible for administering and interpreting mental tests results from London schools. He stressed the heredity of IQ by showing strong correlations in IQ scores between twins raised apart. His twins data was shown to be fraudulent. He also tried to declare himself the inventor of factor analysis, even though it was really well documented that his mentor Charles Spearman had invented it. But most of the damage Burt did was by advocating for the reification of intelligence. Factor analysis is a mathematical technique to reduce a complex system of correlations into fewer dimensions. Spearman calculated the first principle component in his analyses and called it “g” for general intelligence, but then he made the error of interpreting it as a an entity, and gave it an unambiguous causal interpretation. For that, we need independent information beyond the fact of the correlation itself. As it stands, it could be interpreted in two completely different ways.

It reflects inherited mental ability, some people do well on most tests because they are generally smarter

It reflects environmental conditions that tend to cluster together, some people do well on most tests because they had good educations, had enough to eat, had stable families etc.

L.L Thurstone was a professor of Psychology at the University of Chicago, he was interested in the underlying structure of data sets. Came up with the varimax rotation, still in common use, which searches for the maximum variance upon each rotated axis. Information is neither gained or lost by rotating axes, it is merely redistributed. Preferences depend upon the meaning assigned to factor axes. In the 1950’s and 1960’s, he identified seven primary mental abilities (PMAs): verbal comprehension, word fluency, number (computation), spatial visualization, associative memory, perceptual speed, reasoning. He used the same data as Spearman and Burt, but found that seven independent PMAs was a better fit for the data, with no hierarchy and no dominant general factor. He found no common factor in the battery of 56 tests.

Gould concludes that if intelligence sets us apart from other organisms, it is probably what we have is a greater ability to adapt to changing circumstances, we have very flexible behavior.

He includes a final section critiquing The Bell Curve, which was a book that was published in 1994 and makes the case that IQ leads to an individual’s success in society, both economic and personal, and that the US was quickly becoming stratified by it. It also makes the case that there are large IQ differences between different races and that these differences are at least partially genetic. Gould takes objection to this common fallacy. For example, in an Indian village, average male height may be 5’6” and height is quite heritable. But in America, the average may be 5’9”. But heritability of height doesn’t mean that better nutrition may not raise the Indian average to 5’10”. Gould is especially mad that they accept Spearman’s g factor as a real entity without investigating if it represents a true entity.

Gould also brings up the interesting point that it doesn’t make sense to lump all black people together into a race as though they have a lot of common genetic background. The genetic variability among Africans exceeds the sum total of genetic diversity for everyone else in the world combined, because Homo sapiens spent more time in Africa than out of it.

Thought questions:

Is Gould saying science can’t be objective? It seems like there are specific biases in the bad science he presents that don’t adhere to good scientific principles. But is Gould arguing that in principle science can’t be objective?

After reading this book, do you think the idea of g is valid?

One alternative theory that describes the data as well as the idea of g is sampling theory. Sampling theory contends that the positive associations between different cognitive tasks arises because tests draw upon similar abilities, not because of unitary underlying capacity. According to this theory, there are a number of uncorrelated mental processes, and all tests draw upon different samples of these processes. The intercorrelations between tests are caused by an overlap between processes tapped by the tests. Thus, the positive association arises due to a measurement problem, an inability to measure more fine-grained, presumably uncorrelated mental processes. Importantly, this is still consistent with g predicting life success.

In a more general sense, what we include in the measure of g is subjective, and doesn’t include all types of intelligence. IQ tests typically don’t include athleticism, artistic ability, or emotional intelligence. The more broadly we define intelligence, the less variability between individuals we’ll find.

In what ways do racial biases exist in science today?

Dorothy Roberts argues compellingly in Fatal Invention that race is used widely in medicine today, and that it is not evidence based and does real harm. For example, when assessing dementia, a correction is added in some studies for black people, to make their baseline cognition lower. This results in missed diagnoses in many black patients. Roberts argues that race is a social construct and has no underlying genetic validity. If something appears to cluster with race, it’s likely a societal factor due to generations of oppression. BiDil, the first race-based pharmaceutical was marketed to black people for hypertension, but really it was just a combination of 2 generics packaged together and sold this way to be granted a patent. It was originally made for everyone. Roberts finds that investigations of race-based medicine mostly conclude that the interventions work no better in the specified self-identified race group than other groups.

Roberts argues that it is more profitable to do race-based medicine than find that system inequality leads to poorer health in racial minorities, because that is a problem that would actually cost money to fix.

Some thoughts on the state of g today:

First, we have to remember that the idea of even the most "reified" version of g is not racist in and of itself. It can be useful for predicting performance, but I think the consensus is that it is very obviously conditioned mainly by education and the complex interplay involved in learning.

The idea of g has definitely been refined due to some observations. And this has helped determine what  g is not.

Cattell's idea of fluid and crystallized intelligence has more weight now. The g enthusiasts think that g is synonymous with fluid intelligence.

Gardner had a theory of multiple intelligences and popularized the idea that the brain is quite remarkable in all sorts of ways. For instance, we can learn how to shoot basketballs. People's performance on physical tests is not always as highly correlated as between separate pen and paper tasks (although it is correlated). The g enthusiasts basically say that this is not what they mean by g. They have clarified that it is more related to reasoning and thinking abstractly.

Training cognitive tasks increases performance, but does not transfer to g. For instance, working memory ability is quite strongly correlated with g. If you train working memory, you will get better at working memory. However, your g will likely not change.

Very specific abilities or deficits arise some of the time with specific lesions or conditions, indicating that intelligence is at least, in part, modular.

Use of cognitive tests that are meant to specify a particular cognitive domain are much more common because we find modular or specific associations with performance on those particular tasks.

Use of g is more common for people who are interested in transfer and cognitive training because it is a good predictor of future achievement, or more common for its original intended use as diagnostic for people who need help.