Elsevier

Intelligence

Volume 36, Issue 3, May–June 2008, Pages 183-191
Intelligence

The Flynn effect is partly caused by changing fertility patterns

https://doi.org/10.1016/j.intell.2007.04.002Get rights and content

Abstract

The present study investigates the impact of changing fertility patterns on the Flynn effect. Intelligence test data comprised scores of army conscripts on an arithmetic, language and a Raven-similar test, and a composite score (General Ability). Family data of the conscripts enabled a decomposition of the population mean into effects of sibship size on the mean intelligence and the proportion of the persons comprising the various sibship sizes within each of 13 birth cohort groups (each comprising 3 birth years). Both the means within each sibship and the proportions of the different sibship sizes varied across cohorts. Estimated changes in means due to changing proportions of sibship sizes alone were calculated by fixing the mean intelligence test score within the different sibship sizes at the level of the oldest birth cohort (1938–1940) and letting the proportions of the different sibship sizes take their empirical values in each of the subsequent 12 three-year cohort groups. It is concluded that changing proportions of sibship sizes had a moderate effect both on General Ability and the subtest scores, and that most of the changes were connected to changing sibship means across cohorts.

Introduction

A secular increase in mean intelligence test scores – the Flynn effect – has been observed in more than 20 industrialized countries (Flynn, 1987, Flynn, 1999). Recently a substantial Flynn effect was observed in rural Kenya (Daley, Whaley, Sigman, Espinosa & Neumann, 2003). Secular increases seem to be largest in non-verbal tests, like Ravens Progressive Matrices (Flynn, 1987, Sundet et al., 2004). In Norway, testing of military male conscripts shows that most of the increases in mean intelligence have taken place from cohorts born before the Second World War to the birth cohorts shortly after the war (Sundet et al., 2004). After this period the increase rates have been substantially smaller among conscripts both in Norway and in nearby Denmark (Sundet et al., 2004, Teasdale and Owen, 2000). In these two countries the Flynn effect seems to have come to a complete stop, or even reversed in recent years (Sundet et al., 2004, Teasdale and Owen, 2005).

Both the nature of the secular trends in IQ scores and their causes has been extensively discussed. Rodgers (1998) points out that the Flynn effect could be due to changes within certain ranges of the IQ distribution. Thus, it seems that the Flynn effect in some countries is partly or mostly due to lower prevalence of low scorers in more recent cohorts (Lynn and Hampson, 1986, Sundet et al., 2004, Teasdale and Owen, 1989, Teasdale and Owen, 2000). Some studies have reported a declining g-factor in more recent cohorts (Kane and Oakland, 2000, Lynn and Cooper, 1993, Sundet et al., 2004). Wicherts, Dolan, Hessen, Oosterveld, van Baal, Boomsma and Span (2004) analyzed several data sets and found indications of lacking factorial stability across cohorts. These findings certainly complicate the interpretation of the Flynn effect, because changes in mean intelligence test scores across cohorts may partly be due to changes in different underlying constructs.

Doubts about the validity of intelligence test scores as indicators of “real” intelligence have been voiced from time to time (e.g. Flynn, 1987). This is a difficult question, but some studies indicate that the Flynn effect has been accompanied by “real world” indications of rising population intelligence (Howard, 1999, Howard, 2001).

Although a lot of potential causes of the Flynn effect, e.g. nutrition, education, and access to mass media have been widely discussed (cf. Neisser, 1998), there are still others to investigate more thoroughly. Rodgers (1998) pointed out that our knowledge about the occurrence or non-occurrence of a Flynn effect in subgroups is insufficient. Zajonc and Mullally (1997) called attention to the potential utility of studying changes of family configurations across generations. This possibility arises because there is a connection between family size and intelligence. Two factors connected to family size might be considered. One of them is the well-documented negative correlation between sibship size and intelligence test scores (Belmont & Marolla, 1973). The other is the alleged effect of birth rank order on intelligence test scores (Zajonc, 1976, Zajonc and Mullally, 1997). The empirical status of the birth order effect on intelligence is at present unclear. Recently, it has been argued that the birth order effect on intelligence is a myth created by inadequate study designs (Rodgers, Cleveland, Van den Oord, & Rowe, 2000). The birth order effect issue may be clarified by future meta-analytic studies addressing important methodological concerns like selection problems, cross-sectional versus longitudinal designs, and sample size. In the present paper we have decided to use sibship size as the main analytical unit.

Two possible factors connected to sibship size may contribute to the Flynn effect; changes of the mean intelligence within sibships across cohorts, and cohort changes of the relative number of persons comprising the various sibship sizes. This can be seen by considering a hypothetical population where the sibship sizes range from singletons to families comprising 5 siblings, with 20% of the persons belonging to each of the sibship sizes at a given time. The IQ means decrease by 2.5 IQ points across sibship sizes, ranging from 105 for singletons to 95 for persons in 5-sibships. If the mean IQ's within each sibship size and the proportions of the different sibship sizes remain constant over time, no Flynn effect will be seen in our hypothetical population (the aggregated population mean, calculated as 105  0.2 + 102.5  0.2 + 100  0.2 + 97.5  0.2 + 95  0.2, will remain at 100 IQ points). However, consider that the mean IQ's within sibship sizes remain constant over time, whereas the relative proportions of the different sibship sizes vary over time. Say that the proportion of singletons increases from 20% at time 1 to 40% at time 2, the proportion of persons comprising 2-sibships changes from 20% to 30%, 3-sibships from 20% to 15%, 4-sibships from 20% to 10%, and 5-sibships from 20% to 5%. The population mean will then change from 100 at time 1 to 102.25 at time 2 (105  0.40 + 102.5  0.30 + 100  0.15 + 97.5  0.10 + 95  0.05). In this hypothetical case, a Flynn effect has been produced by changing proportions of sibship sizes alone, without any changes within the subgroups of different sibship sizes. A similar change could be produced by changing IQ means within each sibship size without any proportion changes. Thus, if the proportions remained constant over all sibships, but the mean increased by 2.25 IQ points in all sibship sizes from time 1 to time 2 the population mean would also be 102.25, i.e. identical to the population mean in the first case.

The interpretations of the increase of the population mean IQ in these two cases would obviously be very different. This reasoning indicates that it might be reasonable to distinguish two broad classes of causes of the Flynn effect: Factors that cause generational changes within subgroups (and the population mean), and on the other hand, population mean changes caused by changing proportions of these subgroups in the population. Actually, it is quite tempting to consider changes in observed means not apparent in subgroups but only due to changing proportion of these subgroups as aggregation artifacts, and that a “real” Flynn effect is the change in observed means after the effect of proportion changes has been removed. It is therefore of considerable interest and importance to clarify the relative importance of the effect of fertility changes and changes of the means within sibship sizes over time.

There are few studies directly concerned with estimating the effects of fertility changes. Zajonc (1976) argued that changing family configuration might affect the population mean scores of cognitive abilities tests. Alwin (1991) investigated the reasons for declining verbal SAT scores in the US, and did not find any evidence indicating effects of changing family configuration. The aim of the present paper is to investigate the relative contribution of fertility changes and increasing IQ means within sibship sizes to the Flynn effect in the intelligence test scores of Norwegian conscripts (mostly males) in the birth cohorts from 1938 to 1985.

Section snippets

Materials and methods

In Norway, military service is compulsory for every able young man. Before they enter the service the young men are required to meet before a draft board where their medical and psychological suitability, including intellectual ability, for military service are assessed. A great majority of the men meeting before the draft board (about 95%) are examined between their 18th and 20th birthdays. Physically or psychologically disabled are exempted from these investigations. Also, seamen and others

Results

Fig. 1 displays the observed mean scores by sibship size in13 different birth cohort groups.

Fig. 1 shows an increase in mean General Ability within all sibship sizes, and stronger in the comparatively large sibships. In sibship sizes from 1 to 3 the increase from the 1938–1940 to the 1983–1985 birth cohorts was about 5–6 IQ points. The corresponding increases in the means for 4 and 5-sibships were approximately 7–8 IQ points. In the 6+ sibships the increase was about 10–11 IQ points.

The total

Discussion

The tests used to assess the intellectual ability among Norwegian conscripts are representative of subtests regularly included in standard intelligence tests. Thus, the Arithmetic and Word Similarities subtests were similar to the Arithmetic and Vocabulary subtests in WAIS, and the Figures test was explicitly constructed to be similar to the Raven Progressive Matrices test. In Cattell's (1987) system the first two tests measure crystallized intelligence, whereas the last one measures fluid

Acknowledgements

We are indebted to the Norwegian Armed Forces and the Norwegian Social Science Data Services for kindly making the data used in this paper available to us. Psychologist Dag Barlaug at the Psychological services, Norwegian Armed Forces made subtest data available. Psychologists Tore M. Torjussen and Olav Storsve were very helpful in ascertaining data and information about the tests and testing procedures. Dag Erik Eilertsen was very helpful in organizing the data files used in the analyses.

References (31)

  • J.M. Wicherts et al.

    Are intelligence measurement invariant over time? Investigating the Flynn effect

    Intelligence

    (2004)
  • M.D.M. Al et al.

    Relation between birth order and the maternal and neonatal docosahexaenoic acid status

    European Journal of Clinical Nutrition

    (1997)
  • D.F. Alwin

    Family of origin and cohort differences in verbal ability

    American Sociological Review

    (1991)
  • L. Belmont et al.

    Birth order, family size and intelligence

    Science

    (1973)
  • J. Blake

    Family size and the quality of children

    Demography

    (1981)
  • Cited by (38)

    • The ups and downs of intelligence: The co-occurrence model and its associated research program

      2022, Intelligence
      Citation Excerpt :

      This should be contrasted with the finding of longer-term (i.e., over two centuries) declines in a social-cognitive indicator of altruism noted in Woodley of Menie, Figueredo, et al. (2017). Another finding that may appear problematic is that there are some Western populations were the data indicate that IQ is declining even in cases where it is apparently not being selected against, such as Norway (Bratsberg & Rogeberg, 2018; cf. Hegelund et al., 2021; Sundet, 2014; Sundet, Borren, & Tambs, 2008). However, there are three points that must be considered in this regard, and that illustrate why this finding is not necessarily problematic for the co-occurrence model.

    • The Flynn effect for verbal and visuospatial short-term and working memory: A cross-temporal meta-analysis

      2017, Intelligence
      Citation Excerpt :

      Furthermore, many studies have shown lower fertility rate among women in developed countries than in developing countries (Bureau, 2016; Nargund, 2006). The FE and anti-FE were also found as partly being caused by fertility patterns in many countries (Kong et al., 2017; Menie, Fernandes, José Figueredo, & Meisenberg, 2015; Sundet, Borren, & Tambs, 2008). Hence the study investigated whether or not this aspect affects memory scores of people between the two types of countries over time.

    View all citing articles on Scopus
    View full text