Elsevier

Intelligence

Volume 32, Issue 5, September–October 2004, Pages 481-498
Intelligence

Sex differences on the progressive matrices: A meta-analysis

https://doi.org/10.1016/j.intell.2004.06.008Get rights and content

Abstract

A meta-analysis is presented of 57 studies of sex differences in general population samples on the Standard and Advanced Progressive Matrices (SPM and APM, respectively). Results showed that there is no difference among children aged 6–14 years, but that males obtain higher means from the age of 15 through to old age. Among adults, the male advantage is 0.33d equivalent to 5 IQ points. These results disconfirm the frequent assertion than there are no sex differences on the progressive matrices and support a developmental theory that a male advantage appears from the age of 15 years. A meta-analysis of 15 studies of child samples on the Colored Progressive Matrices showed that among children aged 5–11 years boys have an advantage of 0.21d equivalent to 3.2 IQ points.

Introduction

During the last two decades, the issue of sex differences in cognitive abilities has been addressed by carrying out meta-analyses of studies on verbal abilities (Hyde & Linn, 1988), spatial abilities (Linn & Peterson, 1985, Voyer et al., 1995), and mathematical abilities (Hyde, Fennema, & Lamon, 1990). No meta-analysis has yet been made of sex differences in reasoning ability. In this paper, we present the first study of this issue in the form of a meta-analysis of sex differences in nonverbal reasoning ability measured by the Progressive Matrices Test.

Raven's Progressive Matrices was constructed in the late 1930s as a test of nonverbal or abstract reasoning ability (Raven, 1939). It has become one of the leading and most frequently used tests of this ability and has been described as “the paradigm test of nonverbal, abstract reasoning ability” (Mackintosh, 1996, p. 564). It is also widely regarded as the best or one of the best tests of Spearman's g, the general factor underlying all cognitive abilities. Thus, Court (1983, p. 54) has written that it is “recognised as perhaps the best measure of g”; and Jensen (1998, p. 541) wrote, “the Raven tests, compared with many others, have the highest g loading.” There are three versions of the test: the Standard Progressive Matrices (SPM) for the ages of 6 years to adulthood; the Colored Progressive Matrices, an easier version of the test designed for children aged 5 through 12; and the Advanced Progressive Matrices (APM), a harder version of the test designed for older adolescents and adults with higher ability.

The issue of whether there are any sex differences on the progressive matrices has frequently been discussed. It has been virtually universally concluded that there is no difference in the mean scores obtained by males and females. The first statement of this conclusion was made by Raven (1939, p. 30) who wrote that in the standardization sample, “there was no sex difference, either in the mean scores or the variance of scores, between boys and girls up to the age of 14 years. There were insufficient data to investigate sex difference in ability above the age of 14.” The conclusion that there is no sex differences on the progressive matrices has been endorsed by numerous scholars. For instance, Eysenck (1981, p. 41) stated that the tests “give equal scores to boys and girls, men and women.” Jensen (1998, p. 541) writes “there is no consistent difference on the Raven's Standard Progressive Matrices (for adults) or on the Coloured Progressive Matrices (for children).” Mackintosh (1996, p. 564) writes “large scale studies of Raven's tests have yielded all possible outcomes, male superiority, female superiority and no difference.” From this, he concludes that there is no difference or, in a subsequent paper, that there is only a very small difference consisting of no more than a difference of 1–2 IQ points among adults “either way” (Mackintosh, 1998).

Both Jensen (1998) and Mackintosh (1998) rely for their conclusion that there is no sex difference in mean scores on the progressive matrices on a review by Court (1983). This review summarized 118 studies of sex differences on the progressive matrices and concluded that most showed no significant difference in mean scores, although some showed higher mean scores for males and others found higher mean scores for females. From this, he concluded that “there is no consistent difference in favour of either sex over all populations tested…the most common finding is of no sex difference. Reports which suggest otherwise can be shown to have elements of bias in sampling” (p. 62) and that “the accumulated evidence at all ability levels indicates that a biological sex difference cannot be demonstrated for performance on the Raven's Progressive Matrices” (p. 68).

Court's (1983) review is a useful starting point for tackling the question of sex differences on the progressive matrices but it cannot be accepted as a satisfactory basis for the conclusion that no sex differences exist. The review has at least five deficiencies. First, it is over 20 years old and a number of studies of this question have appeared subsequently and need to be considered. Second, it is a literary review that does not attempt to quantify the magnitude of any sex difference that may exist. Third, it includes general population samples and a variety of convenience samples including those of psychiatric patients, deaf children, retarded children, shop assistants, clerical workers, British, Indian, and French university students, Native Americans, and Inuit. Many of these samples cannot be regarded as representative of males and females, and in some cases this is stated by the authors and reported by Court. For instance, two studies showing that among British military personnel in World War II female neurotic patients obtained higher scores on the progressive matrices than males; Court reports that the authors of the studies believed that the reason for this was “probably due to the biased selection procedures in favour of bright women in the British Armed Forces” (Court, 1983, p. 60). With such diverse and unrepresentative samples in Court's review, it is not surprising that in some studies higher scores were achieved by males, in others higher scores were achieved by females and in others there were no sex differences. Fourth, Court does not provide information on the sample sizes for approximately half of the studies he lists, and where information on sample sizes is given the numbers are generally too small to give a statistically significant difference between males and females. To detect a statistically significant difference of between 2 and 5 IQ points such as may exist between males and females on the progressive matrices requires a sample size of around 500. Court's review gives only one study of adults with a sample size of this number or more. This is Heron and Chown's (1967) study (n=600) on which men obtained a significantly higher mean score than women of 0.31d, approximately 4.65 IQ points (see Table 1). Nine of the studies showing no statistically significant sex difference in Court's review have sample sizes of fewer that 100, such as those of 60 Welsh 11- to 12-year-olds and 22 American 5-year-olds. Studies with these small sample sizes that show no significant differences between males and females have no value for the resolution of the issue of whether there is a small but significant sex difference on the progressive matrices. We found that of the 118 studies covered in Court's review, only 21 met the criteria of being general population samples and having sample sizes on a minimum of 50 males and 50 females that we have adopted in the meta-analysis to be presented. Fifth, Court's review did not include all the studies; in our literature search, we found nine studies of general population samples that were not given in the review and these included two with the largest sample sizes (Hsu, 1976, Wilson et al., 1975). For all these reasons, Court's review cannot be accepted as an adequate basis for the conclusion that there are no sex differences on the progressive matrices.

The consensus that there are no sex differences on the progressive matrices and on other tests of abstract (nonverbal) reasoning ability has been challenged by Lynn, 1994, Lynn, 1998, Lynn, 1999. He has proposed a developmental theory of sex differences in intelligence that states that boys and girls mature at different rates both physically and mentally during childhood and adolescence. Boys and girls mature at about the same rate up to the age of around 7 years; from the age of 8, girls begin a growth spurt in which there is an acceleration of their physical growth in respect of height, weight, and brain size; the growth rate of girls slows at the age of 14 and 15, while the growth of boys continues. The developmental theory states that intelligence follows the same trend. Evidence supporting the theory has been provided in Lynn, 1994, Lynn, 1998, Lynn, 1999 and in Lynn, Allik, and Must (2000). In regard to abstract (nonverbal) reasoning ability, the theory as originally formulated in Lynn states that over the age range of around 9 through 12 years, girls have an advantage of approximately 1 IQ point; by the age of 16 years, this has changed to a small advantage in favor of boys and among adults the male advantage is 2.4 IQ points. These estimates were not derived from data on the progressive matrices but (in the case of adults) from the American standardization samples of the Differential Aptitude Test. In a subsequent compilation of studies, it was proposed that among adults the male advantage on abstract reasoning is approximately 5 IQ points (Lynn, 1999).

In addition to the theory that sex differences on the progressive matrices vary by age, there is also a theory proposed by Mackintosh (1998) that there are cohort differences such that among older generations men achieved higher means than women, but this is no longer the case among recent cohorts of young adults.

The resolution of these conflicting theories requires a meta-analysis of studies on sex differences on the progressive matrices such as we have carried out and now describe. The meta-analysis is designed to test three hypotheses. These are as follows:

  • 1.

    The Eysenck (1981), Court (1983), Mackintosh (1996), Jensen (1998) hypothesis that there is no sex difference on the progressive matrices.

  • 2.

    The Lynn, 1994, Lynn, 1998, Lynn, 1999 hypothesis stating that there is no sex difference on the progressive matrices among young children up to the age of 8 years; that girls have a slight advantage from the ages of 9 through 12 years; that boys and girls obtain about the same mean scores at the ages of 13–15 years; that at the age of 16 years boys have a higher mean than girls; that this advantage increases up to the age of 18–19 years; and that from the age of 20 onwards the male advantage lies between 2.4 (Lynn, 1994) and 5.0 IQ points (Lynn, 1999).

  • 3.

    The Mackintosh (1998, p. 538) hypothesis that the higher mean scores of men on the progressive matrices is a cohort effect such that it may have been present among older generations but that “the sex difference in general intelligence among young adults today in the USA, Britain or Israel” (and, presumably, elsewhere in the economically developed world) “ is trivially small, surely amounting to no more than 1–2 IQ points either way,” i.e., in favor of either males or females.

Section snippets

Criteria for selection of studies

The meta-analyst has to address three problems before analyzing the data. These have been identified by Sharpe (1997) as the “Apples and Oranges,” “File Drawer,” and “Garbage in-Garbage out” problems. The “Apples and Oranges” problem is that different phenomena are sometimes aggregated and averaged, where aggregation shows different effects for different phenomena. For instance, Hyde and Linn (1988) have shown in a meta-analysis of sex differences in verbal abilities that the magnitude of the

Results

The results of the studies on sex differences on the SPM and APM are shown in Table 1. The table gives data derived from 57 studies analyzed to provide effect size estimates for 195 samples, with participants numbering a total of 80,928. Samples were considered to be independent within age categories since no study provided more than one estimate per age group. The table gives the location of the study; the size of the male and female samples; the male–female difference in d scores with

Discussion

There are eight points of interest in the results of this meta-analysis. First, it was designed to test the Eysenck (1981)–Court (1983)–Mackintosh (1996)–Jensen (1998) hypothesis that there is no sex difference in mean scores on the progressive matrices. The results given in Table 1, Table 2 show that this is correct for the SPM for the age range 6–14 years. However, it is incorrect for the age of 15 years onwards. From this age into adulthood, males consistently obtain higher means than

References (102)

  • D.J. Prediger

    Dimensions underlying Holland's hexagon: Missing link between interests and occupations?

    Journal of Vocational Behavior

    (1982)
  • D. Sharpe

    Of apples and oranges, file drawers and garbage: Why validity issues in meta-analysis will not go away

    Clinical Psychology Review

    (1997)
  • A.H.G.S. Van der Ven et al.

    A Rasch analysis of Raven's standard progressive matrices

    Personality and Individual Differences

    (2000)
  • P.L. Ackerman et al.

    Determinants of individual differences and gender differences in knowledge

    Journal of Educational Psychology

    (2001)
  • Adams, E. A. (1952). Analysis of Raven's matrices scores....
  • E. Albalde Paz et al.

    El test PMS de Raven y los escolares de Galicia

    (1993)
  • O.S. Alonso

    Raven, g factor, age and school level

    Havana Hospital Psiquiatrico Revista

    (1974)
  • A.L. Angelini et al.

    Matrizes progressivas Coloridas de Raven

    (1999)
  • M.N. Baraheni

    Raven's Progressive Matrices as applied to Iranian children

    Educational and Psychological Measurement

    (1974)
  • F. Campos

    Teste das Matrizes Progresivas Escala Geral

    (1999)
  • J.B. Carroll

    Human cognitive abilities

    (1993)
  • Chaim, H. H. (1994). Is the raven progressive matrices valid for Malaysians?...
  • J. Chan et al.

    The intelligence of six year olds in Hong Kong

    Journal of Biosocial Science

    (1989)
  • J. Cohen

    Statistical power analysis for the behavioral sciences

    (1977)
  • R. Conrad

    The deaf schoolchild: Language and cognitive function

    (1979)
  • N. Cortada de Kohan

    Logos en educacion secundaria y su relacion con intelligencia

    Revista Latinoamericana de Psicologia

    (1998)
  • V. Costenbader et al.

    A Kenya standardisation of the Coloured Progressive Matrices

    School Psychology International

    (2000)
  • J.C. Court

    Researchers' bibliography for Raven's Progressive Matrices and Mill Hill Vocabulary Scales

    (1980)
  • J.C. Court et al.

    Normative, reliability and validity studies

    (1995)
  • J.H. Court

    Sex differences in performance on Raven's Progressive Matrices: A review

    Alberta Journal of Educational Research

    (1983)
  • M. De Lemos

    Standard progressive matrices: Australian manual

    (1989)
  • J.J. Deltour

    Echelle de Vocabulaire Mill Hill de J.C. Raven. Braine de Chateau

    (1993)
  • M.V. Despande

    Sex differences on Raven's Matrices Test

    Journal of Psychological Researches

    (1971)
  • C.D. Dufouil et al.

    Sex differences in the association between alcohol consumption and cognitive performance

    American Journal of Epidemiology

    (1997)
  • Dupont, J. B. (1970). Sex differences on the PM 47 in Switzerland....
  • K.A. Ericsson et al.

    Long-term working memory

    Psychological Review

    (1995)
  • H.J. Eysenck

    Intelligence: The battle for the mind

    (1981)
  • A. Feingold

    Gender differences in personality: A meta-analysis

    Psychological Bulletin

    (1994)
  • P.S. Freyberg

    The efficacy of the Coloured Progressive Matrices as a group test with young children

    British Journal of Educational Psychology

    (1966)
  • F. Florquin

    Les PM 1947 de J.C.Raven au niveau des classes terminals du cycle secondaire

    Revure Belge de Psychologies et de Pedagogie

    (1964)
  • L.I. Garrity et al.

    Preschool children's performance on the Raven's Coloured Progressive Matrices and Peabody Picture Vocabulary Test

    Educational and Psychological Measurement

    (1976)
  • G. Goosens

    Etalonnage du Matrix 1947 de J.C. Raven

    Revue Belge de Psychologie et de Pedagogie

    (1952)
  • R. Guttman

    Genetic analysis of analytic spatial ability

    Behavior Genetics

    (1974)
  • L.V. Hedges et al.

    Statistical methods in the meta-analysis of research on gender differences

  • A. Heron et al.

    Age and function

    (1967)
  • C. Higgins et al.

    A comparison of Stanford-Binet and colored Raven progressive matrices IQs for children with low socioeconomic status

    Journal of Consulting Psychology

    (1958)
  • Higher Education Statistics Agency Online Information Service. Student Table 14a-HE Qualifications Obtained in the UK...
  • C. Hsu

    The learning potential of first graders in Taipei city as measured by Raven's Coloured Progressive Matrices

    Acta Pediatrica Sinica

    (1976)
  • C.C. Hsu

    Chinese children's responses on the Coloured Progressive Matrices

    Journal of the Formosan Medical Association

    (1971)
  • J.S. Hunter et al.

    Methods of meta-analysis

    (1990)
  • Cited by (241)

    • The general factor of personality as a female-typical trait

      2024, Personality and Individual Differences
    • Relationship between neuropsychological test scores and hippocampal atrophy in non-demented Japanese older adults

      2022, Interdisciplinary Neurosurgery: Advanced Techniques and Case Management
    View all citing articles on Scopus
    View full text