Intelligence and Fertility in the United States: 1912-1982

By MARIAN VAN COURT
Department of Psychology
The University of Texas at Austin

By FRANK D. BEAN
Department of Sociology and Population Research Center
The University of Texas at Austin

The following article first appeared in Intelligence 9 (1985), pp. 23-32.

In the advanced industrialized nations, the rate of change in fertility and mortality - the two major forms of selection acting upon intelligence?began rapidly accelerating with the onset of the "demographic transition." In the United States, fertility began to decline in the late 1700s, moving from an average number of live births per woman of around 8.0 before the Revolutionary War to one Of less than 2.0 during the 1970s (Grabill, Kiser, & Whelpton, 1958; Rindfuss & Sweet, 1977). Although the evidence on mortality is more limited, it appears that steep declines in mortality also occurred during this period (Kitagawa & Hauser, 1973). Whereas most observers would agree that differential mortality operated prior to the demographic transition to increase intelligence, it is not certain whether differential fertility or differential mortality was the dominant selective agent in the latter stages of the transition. Nevertheless, the possibility of the former led many theorists at the beginning of this century to forecast a lowering of average levels of intelligence.

Studies conducted in the early 1900s seemed to bear out the more pessimistic prognostications. The relationship between family size and intelligence was measured by correlating the IQ scores of school children with their number of siblings. Negative correlations were consistently reported in a large number of studies conducted in the United States and England (Cattell, 1936; Lentz, 1927; Mailer, 1933; Roberts, 1939; Roberts, Norman, & Griffiths, 1937; Sutherland, 1929; Sutherland & Thomson, 1926). Naturally, these findings were met with considerable alarm. It was predicted that a loss of 1.0 to 1.5 mean IQ points would occur per decade (Cattell, 1937, 1936).

Large-scale investigations were launched in an attempt to measure predicted losses in population IQ over time. In the Scottish survey, the entire 1 l-year-old population was tested in 1932. Again in 1947, all 1 l-year-olds were given the same group verbal-intelligence test. Contrary to expectations, there was an average increase of 2.3 points (Maxwell, 1954). Cattell conducted an equally ambitious cross-decade study of 10-year-olds' performance on a nonverbal test in England, and found a 1.2-point increase in the mean IQ of children tested in 1950 over those tested in 1936 (Cattell, 1951). Intelligence test performance of US high school students showed small gains over a 20-year period (Finch, 1946), and American soldiers from World War II were reported to have significantly higher verbal-ability scores than their counterparts from World War I (Tuddenham, 1948).

Clearly, results of the family-size-IQ studies and cross-decade studies of populations yielded contradictory results. If people with lower IQ scores had larger families, why was there no discernible loss of IQ over time? Investigators adduced a variety of explanations attempting to resolve the paradox. Predicted losses might have been masked by rather substantial improvements in education, nutrition, and other facets of the environment. Tuddenham (1948) stressed the importance of better education and improved mass communication. Cattell (1951) discussed the possibility of increased test sophistication and noted that the relationship between intelligence and marriage rates had not been investigated. Reed (1965) suggested that such small gains as those reported in the cross-decade studies could easily have been caused by sampling error, testing errors, or some other unknown source of error. Osborn's (1940) "Eugenic Hypothesis", had predicted that a eugenic trend would naturally emerge in a modern democratic society, because parenthood would become wholly voluntary when there was free access to birth control (Osborn, 1940, pp. 193-198). At the opposite extreme, Cook (1951) adopted perhaps the bleakest outlook in the controversy when he characterized a decline in intelligence as "inevitable" and wrote that "(I)f this trend continues for less than a century, England and America will be well on the way to becoming nations of near half-wits" (p. 6). Still others expressed skepticism that real changes in either direction were taking place. Penrose (1950a, 1950b) believed a genetic equilibrium existed. Dobzhansky (1962, 316) concurred, suggesting a balanced polymorphism for intelligence in which both extremes failed to produce their quota of offspring.

In 1962, Higgins, Reed, and Reed provided what has come to be regarded as the definitive answer in their landmark article, "Intelligence and Family Size: A Paradox Resolved." They studied the completed fertility of a large Minnesota sample. Although they found the usual sizable negative correlation between IQ and number of siblings, they found a tiny positive correlation between IQ and completed fertility. The latter correlation was dependent upon the inclusion of individuals who had never married?apparently, their automatic exclusion in previous family-size-IQ studies using school children had biased earlier result because the unmarried were disproportionately found at the lowest IQ levels Higgins, Reed, and Reed reported that 30% of those with IQ's less than 70 were, unmarried, in contrast to 10% with IQ's between 100 and 110, and 3-4% with IQs over 110.

Several more studies of IQ and completed fertility reported similar result (Bajema, 1963, 1971; Olneck & Wolfe, 1980; Spuhler, 1962; Waller, 1970). In Bajema's Michigan sample, the higher rate of childlessness among those with very low IQs was due more to their childless marriages than to their lower marriage rates, but the net result was the same. With direct evidence such as this the dire predictions of the early 1900s were rejected as totally unfounded (Falek 1971; Osborn & Bajema, 1972). In a 1971 review article, Falek wrote

There is no evidence of a decrease in intelligence from generation to generation .... (B)ehavioral scientists concerned with the problem have resolved, in approximately a quarter of a century, all the contradictions which plagued the understanding of the direction of human intelligence. In doing so, most investigators have turned around 180 degrees, and are now confident that, with regard to intelligence, evolution is on a positive track. (p. 14)

Despite the wide acceptance these studies received, they contained two potentially serious sources of bias. First, the samples were not random, because they were composed principally of white, native-born Americans living in either the Great Lakes states or New England (Cattell, 1974; Jensen, 1969; Osborne, 1975; Weyl, 1973). Second, they were largely restricted to a narrow range of birth cohorts (Vining, 1982). In a pathbreaking recent analysis, Vining (1982) cast doubt on conclusions derived from previous fertilit
y-IQ studies, suggesting that the absence of a negative correlation may well have been peculiar to the cohorts studied. Previous samples were largely confined to cohorts which had their main reproductive years during the "baby boom," a period of rising fertility, unprecedented; since such records began to be kept. Vining hypothesized that during periods of rising fertility, there will be a zero or slightly positive relationship between fertility and intelligence, but during periods of falling fertility which characterize the entire modern era, with the one exception of the baby boom years of the late 1940s and the 1950s—there will be a negative relationship. He correlated intelligence test scores with number of children, using a large, national probability sample of men and women aged 25 to 34 as of the late 1970s. For each category of age, sex and race examined, correlations were negative, ranging from -.104 to -.221.

One acknowledged limitation of Vining's sample is that many of the respondents had not yet completed their fertility. In addition, the information it provides is confined to a restricted age cohort. The purpose of this paper is to report the results of research on the relationship between IQ and completed fertility (as well as partly completed fertility) which extends the range of cohorts to encompass those born between 1894 and 1964, whose major reproductive years span 1912 to 1982. In so doing, we may also be able to reconcile results of previous research on this issue and to discern whether a positive or negative relationship in fact emerges during periods of rising or declining fertility.

DATA SET: THE NORC GENERAL SOCIAL SURVEY

The National Opinion Research Center (NORC), a nonprofit research organization affiliated with the University of Chicago, conducted the General Social Survey (GSS) in the United States each year from 1972 to 1982, except for 1979 (Davis, 1982). A combination of block quota and full probability sampling was employed. Hour-long interviews were completed with a total of 12,120 respondents, who were English-speaking, noninstitutionalized adults (18 years or older) living within the continental United States. Such questions as age, place of birth, income, and occupation were asked in each interview. Other questions about attitudes on various social, political, and moral issues were rotated in different years.

Variables of relevance to the present investigation include total number of liveborn children, number of siblings, and scores on a steeply graded, untimed vocabulary test given in 1974, 1976, 1978, and 1982 (total N = 6,021). The vocabulary test is made up of 10 questions selected from a test originally devised by Thorndike for use in large demographic surveys in which a full-scale IQ test would not be feasible (Thorndike, 1942; Thorndike & Gallup, 1944). The two forms of the original version were standardized against the Otis Test (Miner, 1957) and included 20 multiple-choice questions each (one item taken from every level of the I.E.R. Intelligence Scale CAVD).

Although no attempt has been made (to our knowledge) to standardize the GSS version against another test of mental ability, there is good evidence that brief vocabulary tests such as this perform quite well as measures of general intelligence. Miner (1957, pp. 28-29) found a median correlation of .83 between scores on several dozen similar short-vocabulary tests and scores on standard IQ tests. Vocabulary correlates more highly (r = .75) than any other subtest with total score on the Wechsler Adult Intelligence Scale (WAIS) (Wechsler, 1958, p. 98). Furthermore, preliminary analyses of the GSS test showed internal characteristics and relationships with other variables which accord well with those reported for traditional, full-scale IQ tests. Scores are normally distributed with a Mof 6.0 and a SD of 2.2. The internal reliability (Cronbach's Alpha) is .79. Test scores correlate 0.5 with highest educational level obtained. As with other measures of crystallized intelligence, there is a very gradual improvement in performance until old age and then a gradual decline. Blacks average 0.70 SDs below the mean for whites, and there is a negligible sex difference (0.06 SDs) favoring women. A previous study found the GSS vocabulary test to be the most powerful predictor of adult white men's income (r = 0.29), better than both educational level and family background (Peterson & Karplus, 1981). Another study reported it to be strongly negatively correlated with "anomie" (r = .42) (Segilman, 1981).

The test was designed to provide only a rough grading of mental ability. In the "ideal world," the test might have been nonverbal and longer, though unreliability could only vitiate relationships between intelligence and other variables, and correlations to be presented could hardly result from random errors. The unique opportunity this data set affords is an overview of the relationship between intelligence and fertility for a nationally representative sample of Americans whose major reproductive years fell between 1912 and 1982.

COHORT ANALYSIS

Data were consolidated from the four surveys in which the vocabulary test was given (1974, 1976, 1978, and 1982). Respondents were divided into 15 birth cohorts of 5-year intervals ranging from before 1894 to 1964. Cohorts 1-9 can be considered to have completed their fertility (because the youngest would be 40 years old), whereas cohorts 10-15 would have completed their fertility to varying degrees. Correlations between vocabulary test scores and total number of children ever born for all 15 cohorts are presented in Table 1. Correlations

TABLE 1
Number of Offspring and Vocabulary Scores, Zero-Order Correlations by Cohort

corrected for attentuation (divided by the square root of 0.79, the coefficient of test reliability) are also presented in parentheses. It is clear that the relationship is predominantly negative, with 12 of 15 correlations statistically significant.
It is of particular interest to see whether the correlations of cohorts 8 and 9 are typical or atypical. Cohorts 8 and 9, whose fertility occurred squarely within the baby boom years, are the cohorts which largely comprised the samples of previous studies which reported small positive correlations. Although the correlations for both are negative, they are less negative than the other correlations. Vining's hypothesis of zero or slightly positive correlations during periods of rising fertility might thus be considered partly substantiated, and his hypothesis of a negative relationship during declines in fertility is more fully substantiated. But it appears that other factors, in addition to cohort effects, will be required to account fully for the differences between Vining's results and those of previous studies. To determine whether exclusion of nonwhites may have also constituted a source of bias in previous studies, a separate analysis of whites was performed for all 15 cohorts (see Table 2). Comparing the correlations in Tables I and 2, it can be seen that overall, the effect of exclusion of nonwhites is negligible.

TABLE 2
Number of Offspring and Vocabulary Scores for Whites, Zero-Order Correlations by Cohort

Although nonwhites average more children and lower test scores, they comprise only 11% of the sample. However, with nonwhites excluded, cohorts 8 and 9 exhibit a more positive relationship than the other cohorts (t = 2.04, p < 0.025, one-tailed test. It should be added that nonwhites were not over-sampled in the General Social Survey (as are blacks in some large surveys), so the total number of nonwhites (N = 622) does not permit separate analysis by cohort.

Correlations between vocabulary scores and number of siblings are presented in Table 3. They are markedly negative across all 15 cohorts, in agreement with the numerous family-size IQ studies of the early 1900s. Vocabulary-sibling correlations are more negative in every cohort than vocabulary-offspring correlations. If the childless had disproportionately low scores in this sample as in previous ones, this would weaken the negative relationship between vocabulary and offspring, but it would not affect the correlations between vocabulary and siblings, and thus it might reconcile the two sets of correlations. In actuality, the opposite turned out to be the case&#151;the childless were found to score higher than those with one or more children in nearly every cohort. It should be noted, with regard to the difference in magnitude between the two sets of correlations, that the variability is considerably greater for number of siblings (M = 4.3, SD = 3.3) than it is for number of offspring (M = 2.1, SD = 1.9).

TABLE 3
Number of Siblings and Vocabulary Scores, Zero-Order Correlations by Cohort