By Marian Van Court The following article originally appeared in Intelligence 32 (2004),
New evidence of dysgenic fertility for intelligence
By Richard Lynn
in the United States
University of Ulster
Marlborough, MA 01752
By Marian Van Court
The following article originally appeared in Intelligence 32 (2004), pp. 193-201.
|Data from the General Social Survey (GSS) collected in the years 19901996 are examined for the relationship between fertility and intelligence as measured by vocabulary. The results show that the relation between fertility and intelligence has been consistently negative for successive birth cohorts from to 1900 to 1979, indicating the presence of dysgenic fertility for all of the 20th century studied thus far. The most recent cohort for which fertility can be regarded as complete is that born in the years 19401949. In this cohort, the decline of genotypic intelligence arising from the negative association between intelligence and fertility is estimated at .90 IQ points per generation. The decline of genotypic intelligence of Whites is estimated at 0.75 IQ points a generation.|
For almost a century and a half there has been concern that there is a negative association between people's intelligence and their number of children. A negative association of this kind is known as dysgenic fertility. The reason it has aroused concern is that it would entail a decline in genotypic intelligence, i.e., the genetic quality of the population in respect of intelligence. In the 19th century this concern was voiced by Galton (1859) and in the earlier decades of the 20th century by Cattell (1937), Fisher (1929), and Muller (1963), among many others. Evidence for the presence of dysgenic fertility in the economically developed nations for the last 150 years or so and in most of the rest of the world during the 20th century has been reviewed in Lynn (1996). The general trend has been that fertility became strongly dysgenic in the closing decades of the 19th century, whereas in the early decades of the 20th century the dysgenic trend weakened but was still present.
The leading theory to explain the onset of dysgenic fertility in the second half of the 19th century was differential use of contraception. A variety of methods of contraception including the sponge, spermicidal chemicals, pessaries, douches, the condom made from sheep gut, and withdrawal were described in a series of books including Richard Carlile's (1826) Every Woman's Book, Robert Owen's (1832) Moral Physiology and Charles Knowlton's (1832) The Fruits of Philosophy. It is assumed that these books were read and the methods of contraception were used initially and predominantly by those with higher intelligence levels, who used this knowledge to reduce their fertility in the second half of the 19th century. By the early decades of the 20th century, knowledge and use of contraception had become widespread. This brought about a decline in fertility throughout the whole population and reduced the dysgenic effect.
In the middle decades of the 20th century, a number of those concerned with this issue believed that dysgenic fertility would be a temporary phenomenon and would disappear as contraception became used efficiently by the whole population. It was argued by Osborn (1940) that when this occurred, fertility would become eugenic because the more intelligent would tend to be higher earners, would be able to afford more children and would have more. Osborn called this the "eugenic hypothesis."
Some studies carried out in the United States in the 1960s suggested that dysgenic fertility had already disappeared and therefore that the eugenic hypothesis was right (Bajema, 1993 and Higgins et al., 1992). However, several studies in the 1980s found that dysgenic fertility was still present (Retherford & Sewell, 1988; Van Court & Bean, 1985 and Vining, 1982). This paper presents new data on this issue for the United States collected during the years 19901996.
The data for this study are drawn from the General Social Survey (GSS) carried out by the National Opinion Research Center (NORC) (Davis & Smith, 1996). These surveys are carried out annually on nationally representative samples of approximately 1500 individuals aged 18 and over drawn as national probability samples from the continental United States but excluding those who do not speak English and those in institutions. Full details of the sampling procedures are given by Davis and Smith (1996). The data from the surveys are available on disks from NORC and it is from these that the results presented in this paper have been derived.
The GSS collects a vast amount of information. The variables with which we are concerned are the vocabulary score, the number of children and the race and sex of the respondents. The vocabulary score is taken as a measure of intelligence. The vocabulary score is derived from a multiple-choice test asking the meaning of 10 words and the score is the number of words defined correctly. Vocabulary scores are highly correlated with measures of general intelligence. For example, the vocabulary subtest correlates .75 with the full-scale IQ of the Wechsler Adult Intelligence Scale, more highly than any of the other subtests (Wechsler, 1958).
The GSS data for the years 1974, 1976, 1978, and 1982 were analyzed for the relationship between vocabulary scores and fertility by Van Court and Bean (1985). The reason for using these years was that the vocabulary test is not given every year and these were the years in which it was given during the period 19741982. Van Court and Bean found negative correlations of around −.15 between vocabulary and fertility. The present study is an examination of more recent GSS data to see whether the negative association between vocabulary and fertility has continued to be present. The present study examines the GSS data collected in the years 1990, 1991, 1992, 1994, and 1996. The data from these 5 years were combined to give a single sample of 6522 respondents. The vocabulary test was not given in the 1993 or 1995 surveys.
Because the present study is a follow-up of the investigation carried out by Van Court and Bean in the early 1980s, it is useful to start by considering the results of the two studies together so that they can be considered as a whole. This is done in Table 1, which divides the subjects into eight birth cohorts. The first two of these are 20-year birth cohorts consisting of those born between 18801899 and 19001919, followed by 10-year birth cohorts of those born 19201929, 19301939, etc. to 19701979 (the first two cohorts are 20-year cohorts because of small numbers). Table 1 shows the numbers in each birth cohort and the correlations between vocabulary and number of children as found by Van Court and Bean and as found in the present study. There are three interesting features in this table. First, all the correlations in both data sets are negative, indicating consistent and prolonged dysgenic fertility. Secondly, there is a close similarity between the correlations obtained by Van Court and Bean and those in the present data. Thirdly, there is no tendency for the magnitude of the negative correlations to decline in more recent birth cohorts. On the contrary, it increases.
In evaluating these negative correlations, it is important to consider whether the fertility of the cohorts is complete. The reason for this is that the more intelligent tend to have their children later (see, e.g., Vining, 1995), so young cohorts can show a negative association between intelligence and fertility, which may be reduced or disappear as the cohort grows older and the more intelligent start to have children. For practical purposes, fertility can be considered to be largely complete for those who have reached the age of 40. In the Van Court and Bean series, fertility is complete up to and including the 19201929 cohort and can be assumed to be more or less complete for the 19301939 cohort, which was aged 35 to 52 at the time the data were collected between 1974 and 1982. In the present series, the latest cohort that can be considered to have completed its fertility is that born between 1940 and 1949, which was aged 41 to 56 at the time the data were collected between 1990 and 1996. The next cohort born 19501959 was aged 31 to 46 at the time the data were collected between 1990 and 1996. Probably its fertility was largely but not entirely complete.
We now analyze the 19901996 data in more detail by breaking down the association between vocabulary scores and fertility by sex and race. In regard to race, the GSS categorizes respondents as White, Black and other. The numbers in our sample are 5450 Whites, 806 Blacks and 286 other. The "other" category is considered to be too few for analysis, so the analysis is confined to Whites and Blacks. Table 2 shows the numbers and correlations for Whites and Blacks, broken down by males and females and by cohorts. The correlations vary somewhat, probably because of small numbers, particularly for the Blacks. To provide a clearer overall picture, the four first age cohorts, 19001919 through 19401949 of those whose fertility can be regarded as complete, have been aggregated and the results are shown in Table 3. There are two interesting features of the data. First, the negative correlations between vocabulary and fertility are present within the two racial groups and in males and females. Secondly, the negative correlation is approximately twice as great for Blacks as for Whites.
Because the 19401949 cohort is the most recent for which fertility can be regarded as complete, it provides the most recent data on which to examine the magnitude of the deterioration of genotypic intelligence per generation arising from the negative association between intelligence and fertility. The formula for calculating the change in a trait as a result of differential fertility (the response to selection) is given by Plomin, DeFries and McClearn. (1990, p. 281) as the product of the narrow heritability of the trait multiplied by the selection differential (narrow heritability is the additive heritability, i.e., the heritability attributable to the effect of additive genes, while total heritability includes the effects of dominant and recessive genes). The formula is derived from Fisher (1929) whose work on the problem is summarized by Plomin et al. (1990, pp. 284285). These authors also provide an extensive discussion of selective breeding studies (Plomin et al., 1990, pp. 278295).
For the present problem of calculating the magnitude of the deterioration of genotypic intelligence, the figure adopted for the narrow heritability of intelligence is .71 given by Jinks and Fulker (1970). The selection differential is the correlation between IQ and fertility and is −.17. Thus, we obtain a decline in genotypic intelligence of .12. This is in the metric of vocabulary scores. To express this in conventional IQs, we need to express it in S.D. units. The S.D. is 2.08, so the decline is .06 S.D. units and this is the equivalent of .90 IQ points. For Whites, the correlation between IQ and fertility is lower than for the total sample at −.15 as compared with −.17. Hence, for Whites the decline of genotypic intelligence is also less and is −.15 multiplied by .71=.11. The S.D. for whites is 2.02, so the decline is .05 S.D. units and is the equivalent of .75 IQ points.
We turn now to the issue of the fertility of those with very low vocabulary scores. The interest of this question is that the method used early in the century to investigate the problem of whether fertility is dysgenic consisted of examining the correlation between intelligence and numbers of siblings. It was found that these correlations were invariably negative. It was inferred that there must be a negative correlation between the intelligence of parents and their number of children (see, e.g., Lentz, 1927, for the United States, and Cattell, 1937, for Britain). An objection made to this method was that it failed to sample those in the parental generation who were childless. If these had low IQs, their lack of children would counterbalance the dysgenic fertility inferred from the negative association between intelligence and numbers of siblings. Studies by Bajema (1993) and Higgins et al. (1992) reported that childlessness was most prevalent among those with very low IQs. These results have been widely considered to invalidate the methodology of inferring that fertility was dysgenic from the negative associations between intelligence and numbers of siblings (e.g., Ehrlman & Parsons, 1976). However, several subsequent studies reviewed in Lynn (1996) have found that those with low IQs do not have a high rate of childlessness. To throw further light on this problem we have analyzed vocabulary scores in relation to numbers of children. All those born up to 1949 have been analyzed, those born from 1950 onwards being excluded because they may not have completed their fertility. The results are shown for Blacks and Whites and for males and females in Table 4. The results do not confirm the theory that the childless tend to have low IQs. On the contrary, their vocabulary scores are higher than average.
Stating the same claim slightly differently, it has been argued that those with very low IQs tend to have relatively few children (e.g., Erhman & Parsons, 1976). To examine this claim the mean numbers of children have been calculated for Black and White males and females, for those born 19001949. The results are shown in Table 5. They show no tendency for those with the lowest vocabulary scores to have small numbers of children. The mean vocabulary score of the entire sample is 6.1 and the standard deviation 2.1. Hence, those with vocabulary scores of 01 score 2 standard deviations below the mean, equivalent to conventional IQs in the range 5570. Inspection of the data set out in Table 5 will show that if those in this range are aggregated they have about the same numbers of children as the total sample.
This study contains five principal points of interest. First, it goes some way towards resolving the problem of the differences between the Higgins et al. (1992) and the Bajema (1993) studies, showing a positive relationship between intelligence and fertility, and the Van Court and Bean (1985), Vining, 1982 and Vining, 1995, and the Retherford and Sewell (1988) studies, showing a negative relationship. The results of the present study confirm and extend the second set of studies in that they show that the association between intelligence and fertility has been consistently negative for all birth cohorts from 19001919 up to 19701979. This negative association holds for the American population as a whole and within White and Black and male and female subpopulations.
When Vining (1982) found a negative association between intelligence and fertility he proposed that this could be reconciled with the positive association reported earlier by Higgins et al. (1992) and by Bajema (1993) if fertility had been dysgenic in the early decades of the century, subsequently turned eugenic (as found by Higgins et al. and by Bajema), and then had turned dysgenic again. This interpretation of the evidence is not supported by the present results showing that fertility has been consistently dysgenic from the 18801899 cohorts onwards. These results are consistent with the negative associations between educational level and fertility that were present in the cohort born in the last decade of the 19th century and has continued throughout the 20th century, as shown in Lynn (1996, p. 114). Because of the association between educational level and intelligence, it is improbable that educational level could be negatively associated with fertility, while in the same cohorts intelligence was positively associated with fertility. Since the negative associations between educational level and fertility are derived from census data they have to be regarded as stronger evidence than the positive associations between intelligence and fertility found by Higgins et al. and by Bajema in rather small samples whose representativeness is doubtful. In fact in the Higgins et al. study the initial sample showed a negative association between intelligence and fertility (r=−.08 for men and −.11 for women). It was only when the sample was reconstructed by including the siblings of the sample that the association appears to have turned positive, although the correlations were not reported. As regards the Bajema result, it was obtained on an urban sample from a school in Kalamazoo, Michigan. The positive association between intelligence and fertility may have arisen because of the omission of rural subjects since rural populations typically have lower mean IQs and higher mean fertility, so their inclusion might have turned the association negative.
Secondly, our results give no support to the eugenic hypothesis advanced by Osborn (1940) that dysgenic fertility would prove to be a temporary phenomenon of the demographic transition and would soon be replaced by eugenic fertility. On the contrary, the magnitude of the dysgenic fertility has increased from the cohorts of 19001919 to that of 19401949, whose fertility can be regarded as complete, and to that of 19501959, whose fertility can probably be regarded as approaching completion. These results are inconsistent with the secular trend of fertility in relation to educational levels, which show reduced dysgenic fertility in more recent cohorts (Lynn, 1996). The reason for this inconsistency is not clear.
Third, our results show that dysgenic fertility among Blacks is about twice as great as among Whites. This confirms the results obtained by Vining, 1982 and Vining, 1995. It is also consistent with census data on the relationship between educational level and fertility, which shows a stronger negative relationship among Blacks than among Whites (Lynn, 1996).
Fourth, our results show that there is no tendency for the childless to have low IQs or for those with low IQs to be childless. This suggests that the studies finding negative associations between intelligence and numbers of siblings were correctly interpreted as indicating the presence of dysgenic fertility, and makes these studies consistent with the results of the Retherford and Sewell (1988) and Vining, 1982 and Vining, 1995 studies and the present data.
Fifth, it is useful to compare the present results with those obtained by Retherford and Sewell (1988). In the present data, the decline of genotypic intelligence for the 19401949 birth cohort is calculated at .9 IQ points per generation for the overall population, and .75 IQ points per generation for the White population. Retherford and Sewell calculated a genotypic decline of .81 IQ points from their data set consisting almost entirely of Whites and born around the same time. The present results are therefore very close to those obtained by Retherford and Sewell results in showing that fertility is slightly dysgenic.
We now consider a limitation of the study that the sample excludes institutionalized individuals of whom the majority will have below average IQs. If these have fewer than average children, the effect of their exclusion from the sample would be to reduce the magnitude of the negative correlation between intelligence and numbers of children. Those in institutions and excluded from the sample are the severely mentally retarded, psychotics in psychiatric hospitals, and criminals in prisons. The severely mentally retarded in institutions most of whom have IQs below 50 have lower than average fertility, so their exclusion reduces the magnitude of dysgenic fertility, but these constitute only about 0.3% of the population and the effect of this will be negligible. Psychotics in institutions also have below average fertility but these are fewer than 1% of the population and the effect of their exclusion will also be negligible. We do not know of any data on the numbers of children of criminals in the United States, but in Britain criminals tend to have above average numbers of children (Lynn, 1995). If this is also true for the United States it would provide some counterbalance to the below average fertility of the mentally retarded and mentally ill. In any case the numbers excluded from the sample because they are in institutions are considered to be too few to have any appreciable effect on the results.
We consider finally the significance of the decline of genotypic intelligence. A decline of .9 IQ points of genotypic intelligence for one generation cannot be regarded as of great practical consequence. However, the consistently negative association between intelligence and fertility from the birth cohort of 18801899 onwards shows that dysgenic fertility has been present for three generations and, therefore, that over this period genotypic IQ has declined by approximately 2.7 IQ points. This is an appreciable decline but it has been counteracted by the much greater increase in phenotypic intelligence that has increased by approximately 3 IQ points per decade from the 1930s up to 1978 (Flynn, 1984). The fact that phenotypic intelligence has increased while genotypic intelligence has declined is not a problem. The increase of phenotypic intelligence is a result of improvements in the environment such as better nutrition and possibly other factors such as the greater availability of cognitively stimulating toys, computer games, television, and radio discussed by a number of contributors to Neisser's (1998) book. These have brought about an increase in phenotypic intelligence that has greatly outweighed the deterioration in genotypic intelligence arising from dysgenic fertility. It seems probable that the increase of phenotypic intelligence will not continue indefinitely but is likely to peter out with diminishing returns from environmental improvements. These is some evidence that this has already begun insofar as the mean IQ in the United States tested with Wechsler and Binet tests increased by approximately 3 IQ points per decade over the period 19321978 (Flynn, 1984), but increased by only 1.7 IQ points over the years 19781995 (Flynn, 1998). If this trend of declining secular gains is projected into the future, and if dysgenic fertility continues, the secular increase in phenotypic IQ would be expected to fall to zero and then be replaced by a decline. As first argued by Galton (1859) and later by Cattell (1937) and Fisher (1929), this would have an adverse impact on the nation's economic and military strength, its intellectual and cultural achievements and of the efficiency with which work is performed at all levels of society.
Bajema, C.J., 1993. Estimation of the direction and intensity of natural selection in relation to human intelligence by means of the intrinsic rate of natural increase. Eugenics Quarterly 10, pp. 175187.