Massive IQ gains over time showed that obsolete norms had inflated estimates of the effects of intervention, adoption, and aging; and misdiagnosis of whether individuals had met IQ cutting lines that affected everything from the administration of the death penalty to who should benefit from special education. There were also important studies cited in the literature as if they could be taken at face value – the adoption study by Skodak and Skeels, for example. In America, obsolete norms had turned the death penalty into a lottery: you survive if you took a current test and got sixty-eight; you die if you took an obsolete test and got seventy-five. Research on the causes of IQ gains showed that environmental factors had a potency hitherto unappreciated, illuminated the history of cognitive progress in the twentieth century and its social significance, and recast the debate about group differences in IQ.

At the age of 17, cognitive performance does much to determine the fate of American youth. That is the age at which they take the SAT and sort themselves out among various universities. I am going to ask you (for now) to accept two promissory notes: that we know how much family affects vocabulary at various levels of achievement; and that we have a rough estimate of the percentile gap between levels of achievement and the cognitive quality of the family typical at that level. For example, those at the 98th percentile of vocabulary come on average from homes just below the 70th percentile of cognitive quality. Assuming we have this knowledge, let us look at the consequences.

Vocabulary and family quality

When students sit the SAT, universities take scores on the SAT for reading (SAT-R) as the best measure of the viability of their students. Vocabulary is highly predictive of those scores. I will average Vocabulary results from all six of the data sets the leading IQ tests give us: Stanford-Binet tests from 1985 and 2001; Wechsler tests from 1950–55, 1975, 1992, and 2004–05 (these dates average the years when the Wechsler Intelligence Scale for Children and the Wechsler Adult Intelligence Scale, the WISC and the WAIS, were actually normed). By combining their results, I hope to eliminate the vagaries of any particular set.

Table 1 shows that family has different effects at various levels of performance. For example: students whose vocabulary puts them at +2 SD above average, which is the 97.73 percentile (they are better than almost 98 percent of 17-year-olds), suffer from a typical disadvantage thanks to their families of about 1 IQ point; those at +1 SD (or at the 84th percentile – better than 84 percent of their peers) have a typical disadvantage of 3 IQ points; typical students at −1 SD (the 16th percentile) are advantaged by just under 3 points: while those way down at the −2 SD (the 2.27 percentile) are advantaged by over 7 IQ points.

I will settle the great debate about whether family effects are still significant at the age of 17. We now know the practical significance of this question. It may seem trivial whether 3 IQ points hamper or help the typical 17-year-old thanks to his or her family environment, but this determines whether or not young people qualify for the universities to which they aspire and whether justice enters into the equation. Thus far the combination of twin studies and adoption studies have not been decisive. Fortunately, a new source of data decides the question. At least some cognitive abilities really are influenced by family environment at that crucial age.

A pause for reflection

To prepare the mind to appreciate the new method, the following is a useful exercise:

1. (1) Assume that individual differences on IQ tests and subtests are determined by two factors that are mutually exclusive: differences in genes and differences in systemic environment like family. For the moment we will set aside differences in chance environment (life's boons and ills and the effects of human autonomy) as irrelevant.

2. (2) We want to compare childhood ages with the age of 50. Assume that by that age, genes and current environment have become perfectly correlated and family background has no independent influence. While at childhood, it does have an impact independent of genes.

3. (3) Surely it would make sense to compare performance at age 50 (when the two factors are “pulling together”) with age 10 (when they are at odds) to see if we can detect a difference between the two that would allow us to measure the degree to which family environment has an independent impact.

4. (4) To follow through, it remains to isolate the proper data, to hypothesize how family might affect that data, to look for its footprint, and to find a way of measuring the size of that footprint.

Thus far, most of this book has been a contribution to the science of individual differences in intelligence. Previously, I have contributed to the science of how cognitive abilities alter over time. I will now venture into the theory of intelligence, partially for its own sake and partially to put my contributions into context.

I want to show that every science needs something I call a meta-theory just as much as it needs fertile scientific theories and, indeed, that there is a relationship between the two. The latter explain and predict phenomena, whether they pertain to the motions of the planets or to intelligent human behavior. The former consists of one or more heuristics – which is to say, concepts that offer advice to theory builders. This advice can be good or bad in the sense that it can set strictures on scientific theories that limit their explanatory potential. The quality of the meta-theory should be measured in terms of the quality of the scientific theories it engenders. In addition, there must be a body of data that allows both verification and falsification. This last is not passive. Some times new data emerge that signal either a flaw in an existing theory and, therefore, the need for a new theory, or the need for a new meta-theory because it appears that current heuristics are giving bad advice about theory building.

I will begin by demonstrating the need for meta-theory even in the most rigorous of the sciences: astronomy, as inclusive of physics and cosmology. When we review the history of astronomy, we will find that the failure to realize that every science has a meta-theory encourages three kinds of error:

1. (1) There is a failure to face the fact that new evidence shows that a hitherto successful theory needs to be revised (type-one error). Often this occurs simply because scientists, like anyone else, are reluctant to embrace new ideas. But sometimes, it is because you are under the spell of a heuristic that makes new theories seem “impossible.”

The Age-Table Method provides some evidence, which, as far as I know, is novel. It measures the extent to which parents try to give their preschool children environments of equal cognitive quality, despite the fact that their children differ in giftedness.

Both the Stanford-Binet and the Wechsler tests (thanks to the Wechsler Pre-Primary Scale of Intelligence or WPPSI) allow us to partition variance among very young children who are still almost entirely conditioned by the home environment their parents provide. Given the high values for family effects at age 7, Vocabulary already at 65 percent and Similarities verging on complete dominance, we would anticipate that family would overwhelm genes among younger children. Bouchard (2013) was correct when he speculated that the variance explained by genes among preschoolers would be minimal. When the brain begins to cognize, family effects may be as high as 80 percent, with the rest split between genes and uncorrelated environment (chance factors like infant illness). Indeed, genes probably account for less than 10 percent.

This means that the direct effects of genes on individual differences in brain physiology may be slight. This fact has implications for scans of the infant brain (magnetic resonance), assuming we want to predict which brain has the most genetic potential. It would be as difficult as looking at two healthy seeds under a microscope and predicting which would grow into the taller plant. This in no way dismisses the potency of whatever genetic differences exist. However subtle they may be at birth, during the course of any normal life they are capable of matching quality of environment so as to produce a combination that differentiates all the levels of cognitive achievement we see around us.

The evidence

When we apply the method to preschool data, we find disadvantages and advantages that put family variance above 80 percent, even values so great that genetic influence would be entirely absent.

We abandon meta-theory to discuss theories in the narrow sense of scientific theories. These undertake the strictly scientific task of explaining phenomena under three headings: clarifying the nature of individual differences, or group differences (including differences between the generations), or brain physiology. I do not believe that any of them challenge either my overall heuristic for intelligence or the quasi-heuristics that give advice to scientific researchers in the three areas. They do not all fit neatly within one of the three areas. Indeed, it would be a poor theory that did not have implications for one of its neighbors. But usually, they originated in one area and I will classify them by their area of origin. The exceptions are theories that were actually designed to bridge areas – that is, were attempts to reconcile findings within two areas that seemed incompatible with one another.

Individual differences

I. g-centered theories

Jensen

I have discussed Jensen's theory at length as a counterpoint to developing my own and will be brief. Its salient feature is its focus on g, the general factor that emerges from the fact that cognitive abilities are inter-correlated (people who do better on one tend to better on all). Jensen was quite aware that factor analysis produced other cognitive abilities that were also significant: verbal factors, memory factors, and so forth. For example, he argues that black and white Americans are relatively equal for rote memory while unequal for more complex thinking.

But even so, aside from race, he did little to analyze the operational cognitive abilities tested by the various Wechsler subtests – Vocabulary, Arithmetic, Comprehension, and Information, the abilities that have so much to do with the significance of individual differences. For example, less vocabulary means less reading, less adequate performance on the SAT, and so forth. I do not mean he would deny the latter. But once he had determined the flow of g (the irreplaceable fuel) through the subtests, he seemed to lose interest in them in their own right.

What I have to say at the end is worth only a page. The science of psychology shows that those who are pre-scientific in their worldview must think again. They must all recognize that science is the best instrument to explore the real world, including the real world of human behavior, and that “common sense” cannot compete. However, I sympathize with Cardinal Bellarmine in his advice to Galileo: science should revise scripture (what ordinary people think about the significance of what they do) only when the evidence is decisive and interpretation of its consequences lucid.

Like Pinker (2002), I have no sympathy with those who believe that human nature is a blank slate that environment may do with as it will. I accept the main thrust of the twin studies. But I reject post-twin pessimism. I also have little sympathy with those who interpret the twins as a genetic veto on our sense of social justice and our efforts to improve our children, our selves, and our species. Whether this short book represents what the best science has to say, and whether it represents what most people suspect to be true, and whether I have reconciled the two, readers both expert and general can decide at leisure.

I have struck another note of optimism. The distinction between the meta-theory of psychology and scientific theories of psychology should prevent us from repeating the mistakes of the past. In addition, the meta-theory's heuristics offers good advice, good enough to allow scientific theories to get on with their job. These theories compete with one another in terms of explanation and prediction, and it will be sad if all of them are not eventually transcended. But I do not foresee the kind of failure that would require the radical step of a new heuristic.

Sources

Raven, J. C. (1941). Standardization of progressive matrices. British Journal of Medical Psychology 19: 137–50. Contains SPM 1938.

Foulds, G. A., and Raven, J. C. (1948). Normal changes in the mental abilities of adults as age advances. Journal of Mental Science 94: 133–42. Contains SPM normed on adults 1942.

Although the data were collected over a number of years, they are always referred to as the “1942 UK adult norms.” Raven, J. C., Court, J. H., and Raven, J. (1976). Manual for Raven's Progressive Matrices and Vocabulary Scales. London: Lewis. Contains CPM 1949 (sometimes dated as 1947).

Raven, J. C., Court, J. H., and Raven, J. (1986). Manual for Raven's Progressive Matrices and Vocabulary Scales. London: H. K. Lewis. Contains 1979 SPM and 1982 CPM.

Raven, J., Raven, J. C., and Court, J. H. (2003, updated 2004). Manual for Raven's Progressive Matrices and Vocabulary Scales. San Antonio, TX: Harcourt. Table APM XIV links ages 15.5 with ages 20–70 for APM 1992 raw scores.

Raven, J., Rust, J., and Squire, A. (2008a). Manual: Coloured Progressive Matrices and Crichton Vocabulary Scales. London: Pearson. Contains CPM 2007.

Raven, J., Rust, J., and Squire, A. (2008b). Raven's Standard Progressive Matrices (SPM) and Raven's Standard Progressive Matrices Plus (SPM Plus). London: Pearson. Contains SPM Plus 2008.

The most recent data (linking 2007 CPM with 2008 SPM with 1992 APM)

First, I will do the most recent data for all results throughout.

1. Step I. Data: raw scores by age at various performance levels.

2. Step II. How to get the values by SD level: given the great interest of this data, I will detail my calculations throughout.

3. Step III. Subtract the difference at the median from the difference at all other levels.

Standard Progressive Matrices Plus (2008)

First, I norm age 7.5 on 9.5 and then, age 9.5 on 15.5 to get a cumulative total (for 7.5 on 15.5). This version of the Progressive Matrices added new items of greater difficulty, which means that the raw scores by percentile are not analogous to previous standardizations of the SPM. But the test is similar in kind. The values in bold are relevant to the first norming (7.5/9.5), the values in brackets relevant to the second (9.5/15.5).

My estimates of the percentage of IQ variance family accounts for will prove sufficient to accomplish their purpose. I aim at supplementing kinship studies (with fresh estimates of family effects by performance level and subtest), not at replacing them. This chapter will show that my estimates are comparable with the twin studies, taking the twin estimates as a given. It will also show that my estimates are in accord with adoption studies, with particular reference to identifying the age at which family effects cease.

Twin studies

We can use my results for the percentage of variance due to family effects to partition cognitive variance into its three main components. These are: genes (including environment matched to genes), family, and current environment uncorrelated with genes. Fortunately, kinship studies show that “uncorrelated with genes” environment (or chance environment or uncommon environment) is steady between ages 6 and adulthood, which is something we would expect from a set of “random” factors. If we add that percentage on to the family percentage and deduct the sum from 100 percent, we get my own estimate of the influence of genes.

Dutch kinship study

Using Stanford-Binet (2001) Vocabulary, Table 9 performs that service. The Dutch values are from Holland (McGue et al., 1993) and their estimates are typical of kinship studies.

Their estimate of 18 percent accounted for by chance (“uncommon environment”) is a bit lower than the usual 20 percent (Haworth et al., 2010; McGue et al., 1993), but it is close enough. The values in bold show that we have achieved our objective of a good match with kinship data: the six comparisons show on average a difference of only 7.45 percent; the match is almost perfect at age 18.

The termination of family effects

However, there is one result of the twin studies we have yet to confirm: that family effects disappear entirely sometime during adulthood.

The Age-Table Method of measuring family effects can also be used on Raven's Progressive Matrices. In 1935, when John C. Raven (then 35 years old) sat at his kitchen table and began to experiment with snowflake designs, he was merely a M.Sc. candidate at the University of London. And yet, what he invented became the most important test in the history of intelligence. It had as its companion the Mill Hill Vocabulary Scale. Spearman, the great inventor of factor analysis, had distinguished two main components of intelligence: eduction, or the ability to make sense out of complexity – that is, perceive logical sequences despite distracters that can become very confusing indeed; and reproduction, or the ability to store and retrieve what is learned – for example, vocabulary. The Progressive Matrices were designed to measure the former; the Vocabulary Scale, the latter.

Particularly after twin studies began to partition individual IQ differences into variance explained by unequal genetic endowment and by unequal environments, Jensen and his followers deemed Raven's the best measure of intelligence existent; and described what it measured to be an attribute deeply embedded in the human brain and therefore largely impervious to any kind of sudden change.

The presumed stability of intelligence

Intelligence was supposed to be stable precisely because it was genetically influenced. Therefore, it was unlike learning, which is at the mercy of what kind of education your environment provides. Just as, under normal conditions, your genes determine your height relative to others, so your genes determine the neural framework of your brain and its capacity to solve complex cognitive problems. As for the average intelligence of the human race, it had evolved by natural selection and while that process will of course continue, it will be slow.

There is a corollary: some of the behaviors peculiar to the last two centuries can accelerate the rate of change but only because they impinge on genes.