Intelligence

Introductory treatments of the measurement of intelligence often begin with a discussion of three pioneers in the field: the French psychologist Alfred Binet (1857–1911), the English psychologist Charles Spearman (1863–1945), and the American psychologist Lewis Terman (1877–1956). Binet initiated the applied mental measurement movement when, in 1905, he introduced the first test of general mental ability. Spearman offered support for a psychologically cohesive dimension of general intellectual ability when, in 1904, he showed that a dominant dimension (called g) appears to run through heterogeneous collections of intellectual tasks. And Terman championed the application of intelligence testing in schools and in the military. Subsequently, Terman also illustrated how tracking intellectually talented youth longitudinally (i.e., via long-term studies) affords fundamental insights about human development in general.

Binet: The Testing of Mental Ability

Binet was not the first to attempt to measure mental ability. Operating under the maxim of the fourth century B.C.E. Greek philosopher Aristotle, that the mind is informed to the extent that one's sensory systems bring in clear and reliable information, the English scientist Francis Galton (1822–1911) and others had aimed to measure intellect through fundamental psychophysical procedures that indexed the strength of various sensory systems. In contrast, Binet examined complex behaviors, such as comprehension and reasoning, directly. In doing so, his methods could not compare to psychophysical assessments in terms of reliability. But Binet more than made up for this in the validity of his assessment procedure in predicting school performance. Binet's insight was to use an external criterion to validate his measuring tool. Thus, he pioneered the empirically keyed or external validation approach to scale construction. His external criterion was chronological age, and test items were grouped such that the typical member of each age group was able to achieve 50 percent correct answers on questions of varying complexity. With Binet's procedure, individual differences in scale scores, or mental age (MA), manifested wide variation around students of similar chronological age (CA). These components were synthesized by William Stern to create a ratio of mental development: MA/CA. This was later multiplied by 100 to form what became known as the intelligence quotient ("IQ"), namely IQ = MA/CA ×100.

Spearman: The Discovery of g

While Binet was creating the first valid test of general intellectual functioning, Spearman was conducting basic research that offered tangible support for the idea that a psychologically cohesive dimension of general intelligence underlies performance on any set of items demanding mental effort. In a groundbreaking publication from 1904 called "'General Intelligence': Objectively Determined and Measured," Spearman showed that g appears to run through all heterogeneous collections of intellectual tasks and test items. Ostensibly, items aggregated to form such groupings were seen as a hodgepodge. Yet when such items are all positively correlated and they are summed, the signal received by each is successively amplified and the noise carried by each is successively attenuated. And the total score paints a clear picture of the attribute under analysis.

Spearman and William Brown formalized this property of aggregation in 1910. The Spearman-Brown Prophecy formula estimates the proportion of common or reliable variance running through a composite: r_tt= kr_xx÷ 1 + (k - 1) r_xx (where: r_tt = common or reliable variance, r_xx = average item intercorrelation, and k = number of items). This formula reveals how a collection of items with uniformly light (weak) positive intercorrelations (say, averaging r_xx = .15) can create a composite dominated by common variance. If fifty r_xx = 15 items were available, for example, their aggregation would generate an individual differences measure having 90 percent common variance (and 10 percent random error). Stated another way, aggregation amplifies signal and lessens noise. As Bert Green stated in his 1978 article "In Defense of Measurement," "given enough sow's ears you can indeed make a silk purse" (p. 666). A large number of weak positive correlations between test items is, in fact, the ideal when measuring broad psychological attributes.

Terman: The Application of IQ

Binet's approach to assessing mental ability was impressive because, unlike psychophysical assessments of sensory systems, his test forecasted teacher ratings and school performance. And Spearman's work identified the dominant dimension responsible for the validity of these forecasts. Subsequently, Terman cultivated the new enterprise of applied psychological testing. For example, he played a key role in America's military effort when he combined forces with the American psychologist Robert Yerkes (1876–1956) to facilitate personnel selection during World War I. The U.S. armed forces needed an efficient means to screen recruits, many of whom were illiterate. One of Terman's students, Arthur Otis, had devised a nonverbal test of general intelligence, and his work was heavily drawn on to build one of the two group intelligence tests used for the initial screening and the appropriate placement of recruits: the Army Alpha (for literates) and Beta (for illiterates). The role that mental measurements played in World War I and, subsequently, in World War II constitutes one of applied psychology's great success stories. Even today, an act of the U.S. Congress mandates a certain minimum score on tests of general mental ability, because training efficiency is compromised prohibitively at IQs less than or equal to 80 (the bottom 10% of those tested).

Following World War I, Terman was one of the first to draw a generalization between the utility of military intellectual assessments and problems in America's schools. In the early 1920s, Terman developed one of the most famous longitudinal studies in all of psychology, exclusively devoted to the intellectually gifted (the top 1%). Terman, a former teacher himself, was aware of the ability range found in homogeneous groupings based on chronological age and became an advocate of homogeneous grouping based on mental age. Drawing on solid empirical findings from his study of 1,528 intellectually precocious youth (a study that continued after his death in 1956 and into the twenty-first century), he proposed that, at the extremes (say, two standard deviations beyond either side of IQ's normative mean), the likelihood of encountering special student needs increases exponentially. Terman noted that structuring educational settings around chronological age often results in classes of students with markedly different rates of learning (because of markedly different mental ages). Optimal rates of curriculum presentation and complexity vary in gradation throughout the range of individual differences in general intelligence. With IQ centered on 100 and a standard deviation of 16, IQs extending from the bottom 1 percent to the top 1 percent in ability cover an IQ range of approximately 63 to 137. But because IQs are known to go beyond 200, this span covers less than half of the possible range. Leta Hollingworth's classic 1942 study, Children above 180 IQ, provided empirical support for the unique educational needs of this special population. These needs have been empirically supported in every decade since.

The Modern Hierarchical Structure of Mental Abilities

Modern versions of intelligence tests index essentially the same construct that was uncovered at the turn of the twentieth century in Spearman's 1904 work, "'General Intelligence': Objectively Determined and Measured"–albeit with much more efficiency and precision. For example, g is a statistical distillate that represents approximately half of what is common among the thirteen subtests comprising the Wechsler Adult Intelligence Scale. As noted by intelligence researcher Ian J. Deary in "Intelligence: A Very Short Introduction," the attribute g represents the research finding that "there is something shared by all the tests in terms of people's tendencies to do well, modestly, or poorly on all of them" (p. 10). In 2001 Deary's team published the longest temporal stability assessment of general intelligence to date (covering a span of sixty-six years, from age eleven to age seventy-seven); they observed a correlation of .62, which rose to over .70 when statistical artifacts were controlled.

John B. Carroll and other modern psychometricians have come to a consensus that mental abilities follow a hierarchical structure, with g at the top of the hierarchy and other broad groups of mental abilities offering psychological import beyond g. Specifically, mathematical, spatial-mechanical, and verbal reasoning abilities all have demonstrated incremental (value-added) validity beyond g in forecasting educational and vocational criteria. Although mathematical, spatial, and verbal reasoning abilities do not have the breadth or depth of external correlates that g does, the incremental validity they offer makes them especially important for educational and vocational planning.

Psychological and Social Correlates of g

Psychologists at poles of the applied educational—industrial spectrum, such as Richard Snow and John Campbell, respectively, have underscored the real-world significance of general intelligence by incorporating it in lawlike empirical generalizations, as in the following two passages:

Given new evidence and reconsideration of old evidence, [g] can indeed be interpreted as "ability to learn" as long as it is clear that these terms refer to complex processes and skills and that a somewhat different mix of these constituents may be required in different learning tasks and settings. The old view that mental tests and learning tasks measure distinctly different abilities should be discarded. (Snow, p. 22)

General mental ability is a substantively significant determinant of individual differences in job performance for any job that includes information-processing tasks. If the measure of performance reflects the information processing components of the job and any of several well-developed standardized measures used to assess general mental ability, then the relationship will be found unless the sample restricts the variances in performance or mental ability to near zero. The exact size of the relationship will be a function of the range of talent in the sample and the degree to which the job requires information processing and verbal cognitive skills. (Campbell, p. 56)

Modern research on general intelligence has sharpened validity generalizations aimed at forecasting educational outcomes, occupational training, and work performance. But empiricism also has escalated in domains at the periphery of general intelligence's network of external relationships, such as aggression, delinquency and crime, and income and poverty. For some benchmarks, general intellectual ability covaries .70–.80 with academic achievement measures, .40–.70 with military training assignments, .20–.60 with work performance (higher correlations reflect greater job complexity), .30–.40 with income, and around .20 with law-abidingness.

An excellent compilation of positive and negative correlates of g can be found in a 1987 work by Christopher Brand that documents a variety of weak correlations between general intelligence and diverse phenomena. For example, g is positively correlated with altruism, sense of humor, practical knowledge, responsiveness to psychotherapy, social skills, and supermarket shopping ability, and negatively correlated with impulsivity, accident-proneness, delinquency, smoking, racial prejudice, and obesity. This diverse family of correlates is especially thought-provoking because it reveals how individual differences in general intelligence "pull" with them cascades of direct and indirect effects.

Charles Murray's 1998 longitudinal analysis of educational and income differences between siblings is also illuminating. Murray studied biologically related siblings who shared the same home of rearing and socioeconomic class yet differed on average by 12 IQ points. He found that the differences in IQ predicted differences in educational achievement and income over the course of 15 years. His findings corroborate those of other studies that use a similar control for family environment, while not confounding socioeconomic status with biological relatedness.

Experts' definitions of general intelligence appear to fit with g's nexus of empirical relationships. Most measurement experts agree that measures of general intelligence assess individual differences pertaining to "abstract thinking or reasoning," "the capacity to acquire knowledge," and "problem-solving ability." Naturally, individual differences in these attributes carry over to human behavior in facets of life outside of academic and vocational arenas. Abstract reasoning, problem solving, and rate of learning touch many aspects of life in general, especially in the computer-driven, information-dense society of the United States in the early twenty-first century.

Biological Correlates of g

General intelligence may be studied at different levels of analysis, and, as documented by Arthur Jensen in "The g Factor," modern measures of g have been linked to a variety of biological phenomena. By pooling studies of a variety of kinship correlates of g (e.g., identical and fraternal twins reared together and apart, and a variety of adoption designs), the heritability of general intelligence in industrialized nations has been estimated to be between 60 and 80 percent. These estimates reflect genetic factors responsible for individual differences between people, not overall level of g. In addition, research teams in molecular genetics, led by Robert Plomin, are working to uncover DNA markers associated with g. Using magnetic resonance imaging technology, total brain volume covaries in the high .30s with g after removing the variance associated with body size. Glucose metabolism is related to problem-solving behavior, and the highly gifted appear to engage in more efficient problem-solving behavior that is less energy expensive. Also, highly intellectually gifted individuals show enhanced right hemispheric functioning, and electroencephdographic (EEG) phenomena have been linked to individual differences in g. Finally, some investigators have suggested that dendritic arborization (the amount of branching of dendrites in neurons) is correlated with g.

A Continuing Field of Debate

The above empiricism is widely accepted among experts in the measurement/individual differences field. Yet, it has been common for empiricism pertaining to general intelligence (and interpretative extrapolations emanating from it) to stimulate contentious debate. Indeed, psychologists can be found on all sides of the complex set of issues engendered by assessing individual differences in general intelligence. But this is not new, and it is likely to continue. Because psychological assessments are frequently used for allocating educational and vocational opportunities, and because different demographic groups differ in test score and criterion performance, social concerns have followed the practice of intellectual assessment since its beginning in the early 1900s. In the context of these social concerns, alternative conceptualizations of intelligence, such as Howard Gardner's theory of multiple intelligences, Daniel Goleman's theory of emotional intelligence, and Robert Sternberg's triarchic theory of intelligence have generally been positively received by the public. Yet, measures of these alternative formulations of intelligence have not demonstrated incremental validity beyond what is already gained by conventional measures of intelligence. That is, they have not yet demonstrated incremental validity beyond conventional psychometric tests in the prediction of important life outcomes such as educational achievement, occupational level, and job performance. This is not to say that there is no room for improvement in the prediction process. Innovative measures of mental abilities, however, need to be evaluated against existing measures before one can claim that they capture something new.