ANNUAL REVIEWS

Further

Quick links to online content

Ann. Rev. Genet 1976. 10:179-207 Copyright © 1976 by Annual Reviews Inc. All rights reserved

Annu. Rev. Genet. 1976.10:179-207. Downloaded from www.annualreviews.org Access provided by University of California - Davis on 01/24/15. For personal use only.

GENETICS OF SPECIFIC

+3103

COGNITIVE ABILITIES! J. C. DeFries, S. G. Vandenberg, and G. E. MeC/earn Institute for Behavioral Genetics. University of Colorado, Boulder. Colorado

80309

CONTENTS ...................................... . ........... . ..... ..... . .... . ......... Behavioral Phenotypes .............. .............. ................. . . . . . . . . . . . . . . . . . . . . . . . . . . . ............... ....... Historical Development ............................................ . .................................... . ....... ... Parallel Between Classical Test Theory and Quantitative Genetic Theory . ...........

180 180 181 183

. . ........ .. Twin Studies ..... ..... .............. ........... ..... ............ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ................ ......... Family Studies ............ ................. ........ . . . . . . . . . . . . . . . . . . ....................... ..................... .......

185 185 188

.... ............ ....................... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ................... Differential Effects of Mental Retardation Syndromes .......................................... Spatial Ability ... .. . ...... .. .. .. . ......... . .. . ..... . . .... . .. . . .. . ...... . .. . ........ . ........... .. ... . ................... Reading ....................................................................................................................

196 196 198 201

PSYCHOMETRIC CONSIDERATIONS

EVIDENCE FOR THE HERITABILITY OF SPECIFIC COGNITIVE ABILITIES

SPECIAL CASES

PROMISSORY NOTE

.

. .

. .

............ ........... . ............................ .. ..........................................

CONCLUDING REMARKS

.

. .

.

.

.

. .

. .

.

... ..... . .......... .......... ..... ................... . .................. . ...........

202 202

Behavioral genetics was reviewed in 1960 (39), 1966 (78), 1971 (65), and 1974 (IS) for the Annual Review of Psychology and in 1970 (76) for the Annual Review of Genetics. In 1972, human behavioral genetics was reviewed by Childs (20) for the Annual Review ofMedicine. Lewontin (64) authored a critical review entitled "Ge­ netic Aspects of Intelligence" for the 1975 Annual Review of Genetics, but focused entirely upon human IQ studies. As an indication of the growing interest in this area, a quarterly journ al (Behavior Gen etics) was founded in 1970, and a professional organization (Behavior Genetics Association), which now has more than 300 members, was organized in the same year. The growth of this interdisciplinary field has been so rapid that it would be difficult to review adequately all developments between 1970 and 1976. Therefore, rather than attempting a general review of behavioral genetics, we have chosen to 'Supported in part

by a grant from the Spencer Foundation. 179

180

DEFRIES, VANDENBERG & McCLEARN

Annu. Rev. Genet. 1976.10:179-207. Downloaded from www.annualreviews.org Access provided by University of California - Davis on 01/24/15. For personal use only.

focus upon only one important behavioral domain, namely, specific cognitive abili­ ties. Even in this narrowed range, it has been necessary to be quite selective. Our primary objective is to communicate the principle that cognitive ability is far too complex to be assessed by a univariate number such as IQ and that a multivariate approach is required. Toward this end we briefly discuss the history of the develop­ ment of cognitive test instruments and the operational philosophy of test and measurement theory. This discussion is followed by a review of the evidence con­ cerning the inheritance of specific cognitive abilities. PSYCHOMETRIC CONSIDERATIONS

Behavioral Phenotypes The phenotypes studied by geneticists have a vast variety of observational require­ ments. A simple dichotomous classification requires only the ability to perceive differences between the classes. Assessment of continuously distributed characters, however, is more difficult. In addition to the purely statistical considerations, there can be very important problems of measurement. For some "quantitative" charac­ ters, measurement is relatively straightforward. In dealing with scutellar bristles, for example, it may only be necessary to count. For other characters, such as body height, there may be a fairly obvious application of some fundamental standard of measurement. On the other hand, fundamental comparison operations are not appropriate for some characters, and derived measures are required. Although the problems of derived measurement are by no means limited to the behavioral domain, they are very pronounced in this area. The difficulties have required psychologists to examine the problem of the definition of behavior with great care. As a consequence, certain aspects of operational philosophy and logical positivism have been woven into the fabric of psychological thought. Stevens (108) has succinctly summarized this philosophy as follows: A term denotes something only when there are concrete criteria for its applicability; and a proposition has empirical meaning only when the criteria of its truth or falsity consist of concrete operations which can be performed upon demand. (p. 228)

In many ways, pointing to exemplars of a concept is the most satisfactory sort of definition. But many concepts are not concrete and require definitions which themselves need explication. Herein lie numerous traps for the unwary, because mutuality of understanding of the defined concept depends upon the degree to which the meanings of the definitional words are shared. Or, as Larrabee (61, p. 248) has warned: "talk without mutual understanding is just a means of preventing silence." Our dictionaries may state that intelligence is "the faculty of thought or reason," or that cognition is "the mental process or faculty by which knowledge is acquired." However, unless we have some way of reducing terms such as faculty, thought, reason, mental, and knowledge to operations, these magniloquent definitions are, for scientific purposes, simply gibberish. For research purposes, there are many different operational definitions of cogni­ tive functions. For example, performance level on a particular test is one definition

Annu. Rev. Genet. 1976.10:179-207. Downloaded from www.annualreviews.org Access provided by University of California - Davis on 01/24/15. For personal use only.

GENETICS OF SPECIFIC COGNITIVE ABILITIES

18 1

of intelligence; another definition involves correlational analyses. As we shall see from the following review, we should really speak of intelligences, rather than of intelligence. Were cognitive functioning of interest only to specialists, these points would not need elaboration. But intelligence is of concern to the population at large, and everyone has an understanding, however vague, of what it means. The tacit assump­ tion that there is only one meaning, that it is shared by all, and that it is the same as the meaning employed by the research specialist has created confusion and probably accounts for a considerable part of the controversy that now swirls around the subject. In a brief review such as this, it is not possible to provide all of the operational definitions for all of the cognitive functions discussed. We hope, however, that some appreciation for the diversity of definitions can be gained from the context. Detailed consideration of these matters will require reference to the original works.

Historical Development The idea of a dormant or innate attribute waiting to be developed underlies the concepts of ability, aptitude, and faculty which occur in most European languages (47). The Greek philosophers enumerated many separate attributes such as memory and imagination, and the list was further elaborated by writers of the Renaissance and those of the 18th and 19th centuries. During the latter half of the 19th century, Sir Francis Galton experimented with various measures of sensory discrimination, including bisecting a line and judging differences between small weights. From his research, Galton (40, p. 20) concluded the following: "The trials I have as yet made on the sensitivity of different persons confirms the reasonable expectation that it would on the whole be highest among the intellectually ablest." This view was a logical outcome of the British empiricist philosophy which held that all knowledge was acquired through the senses. James McKeen Cattell, who first used the term mental test in 1890, worked with Galton in England and later brought his ideas to the United States. Cattell experi­ mented with measures such as keenness of eyesight and hearing, reaction time, afterimages, color vision, perception of pitch, and judging a to-second interval ( 19). Other early mental tests were developed during the latter part of the 19th century by Oehrn (90), Munsterberg (86), Kraepelin (60), and Ebbinghaus (29). In 1904, Alfred Binet, a gifted experimental psychologist at the Sorbonne, was appointed to a commission responsible for devising a national system of education for the mentally retarded. The commission concluded that no child should be sent to a special school unless an examination revealed that "his intellectual state renders him unable to profit in the ordinary measure from the instruction given in the regular schools" (9, p. 163). Binet & Simon developed scales (9- 1 1) which were highly effective for this purpose. The test was later revised for administration in the United States by Terman ( 1 10) and by Terman & Merrill ( 1 1 1) at Stanford Univer­ sity. Success of the Binet-Simon test led to increased interest in mental testing. In 1917, the US War Department began a program of mental testing for the classification of

Annu. Rev. Genet. 1976.10:179-207. Downloaded from www.annualreviews.org Access provided by University of California - Davis on 01/24/15. For personal use only.

182

DEFRIES, VANDENBERG & McCLEARN'

military men, The test employed, the Army Alpha (128), was to a large extent a creation of Otis, a graduate student of Terman's. It was a "paper and pencil" test which could be administered by relatively unskilled personnel to large groups. Eventually, over 1,750,000 men were tested. A second version, the Army Beta, was developed for use with illiterates or non-English speaking recruits. After World War I, there was a rapidly increasing interest in the use of intelligence tests in industry, education, and civil service; it is likely that this was due at least in part to the fact that so many people had been exposed to mental tests during the war. In the meantime, Charles Spearman, an English army officer who later became a professor at the University of London, attempted to provide a theoretical basis for the concept of intelligence He was influenced by Galton and reasoned that even weak tests might be useful ifit could be shown that a combination of such tests could predict a criterion such as school grades. Spearman developed a statistical test to determine whether different tests were measuring something in common and some­ thing specific to each test. In this two-factor theory (104), the first factor was ascribed to general intelligence (or g), whereas the second was a specific factor unique to each test. Spearman's two-factor theory did not provide a solution to the problem of the nature of intelligence. Later investigators (58) proposed the existence of other "group factors" in addition to g. For example, in 1935, EI Koussy (30) identified a spatial group factor which he symbolized "k:" After several such group factors were recognized, a hierarchical theory of intelligence was proposed by Burt (17) and Vernon (121). In this model, illustrated in Figure I, four kinds of factors are assumed: (a) a general factor; (b) two major group factors (verbal-education and spatial-mechanical); (c) an undetermined number of minor group factors, some of which can be still further subdivided; and (d) factors specific to each test. The most prominent advocate of multiple ability factors was Thurstone (11311S). In addition to developing a shortcut method for extracting factors from a correlation matrix, Thurstone (116) advocated rotating these factors to a new posi­ tion which could provide simple structure, where: (a) each variable is identified with only one or a small proportion of the factors; (b) the number of variables loaded on (correlated with) a factor is minimized; (c) the variance accounted for by the major unrotated factors is distributed across the rotated factors; and (d) each rotated factor is more or less identified with a distinct cluster of correlated variables g

Major group factors

I

v:ed

I

I

k:m

� � I fJn rll � I r=t-,

Minor group factor

Specific factors Figure 1

1111111111111111111111111111111111111111

Hierarchical structure of human abilities. Reprinted from Reference 121.

GENETICS OF SPECIFIC COGNITIVE ABILITIES

183

(see Rummel, 98). Using these techniques, Thurstone identified seven uncorrelated factors which were found to recur across a number of large studies: verbal meaning or vocabulary (V); verbal fluency (W); numerical ability (N); spatial ability (S); inductive reasoning (I or R); memory (M); and perceptual speed (P). When later searching for simple structure by plotting test loadings on pairs of factors, Thurstone noted that more near-zero values could be obtained when factors were allowed to correlate. Such axes and the resulting factor structure are said to be "oblique," in contrast to the orig inal orthogonal" factor structure. Once factors are allowed to correlate, reSUlting factor correlation matrices may again be factor analyzed to produce higher order factor solutions. Thus, a compromise was reached between Spearman's theory of one general plus many specific factors and Thur­ stone's multiple factor theory. With the advent of electronic computers, the full characteristic equation repre­ sented by the correlations between n tests could be solved readily to yield n roots and, hence, n factors. Thurstone's concepts of rotation, simple structure, and oblique factors have survived this technical revolution. For a detailed account of factor analysis and the various methods of approximating simple structure, see Harman (45) and Rummel (98). In contrast to Thurstone, Guilford (43) insisted upon uncorrelated factors and attempted to broaden the content and format of psychological tests. Guilford recog­ nized five psychological processes which can be performed on four types of materi­ als, where the information can be packaged into six kinds of products. Because these three classification criteria are postulated to be independent, there would be 5 X 4 X 6 120 uncorrelated ability factors (see Figure 2). While Guilford's theory has stimulated considerable research, few investigators have accepted his plethora of factors. What consensus there is seems to favor the hierarchical view (49).

Annu. Rev. Genet. 1976.10:179-207. Downloaded from www.annualreviews.org Access provided by University of California - Davis on 01/24/15. For personal use only.

"

=

Parallel Between Classical Test Theory and Quantitative Genetic Theory It is interesting to note that a marked similarity exists between the mathematical models of psychological test theory and quantitative genetics, as well as between the parameters derived from these models. In classical test theory (see 67), the following linear model is assumed: X=T+E,

1.

where X is the observed test score of an individual, T is the true score, and E is the error of measurement. In quantitative genetic theory (see 32), a similar model is assumed:

P= G +E.

2.

where P is the observed phenotypic value, G is the genotypic value (or the value conferred upon the individual by its genotype), and E is the environmental devia­ tion. Two tests are said to be parallel if they yield identical true scores and if the errors are uncorrelated and have equal variances. The analogous situation in quantitative genetics is the case of identical twins reared in random environments. In such a case,

184

DEFRIES, VANDENBERG & McCLEARN

OPERATIONS Cognition Memory

Diver ent Production Convergent Production Evaluation

Annu. Rev. Genet. 1976.10:179-207. Downloaded from www.annualreviews.org Access provided by University of California - Davis on 01/24/15. For personal use only.

PRODUCTS

Units Classes Relations Systems Transformations mplications

CONTENTS Figural Symbolic Semantic

Behavioral

Gui lford' s ability factors.

Figure 2

members of the twin pair would have identical genotypic values and the environ­ mental deviations would be uncorrelated. From the above model of test theory, the following result obtains: Vx

=

Vr + VE>

3.

where V symbolizes variance. The proportion of the variance in observed scores owing to variation in true scores is defined as the reliability of a test. It may be shown that reliability has the following equivaknt forms:

reliability = VTI Vx =

r

iT

=

bTX

=

'XX�

4.

where 'XT is the correlation between true and observed scores, bTX is the regression of true score on observed score, and 'XX' is the correlation between performances

on parallel forms of a test. From the model of quantitative genetics, 5.

when G and E are uncorrelated. The proportion of the phenotypic variance owing to variation in genotypic values is defined as "heritability in the broad sense" (68) and has the following equivalent forms: heritability = VGI Vp = r

�p

= bGP = rppl

6.

Annu. Rev. Genet. 1976.10:179-207. Downloaded from www.annualreviews.org Access provided by University of California - Davis on 01/24/15. For personal use only.

GENETICS OF SPECIFIC COGNITIVE ABILITIES

185

where P and P' are measures on members of identical twin pairs reared in random environments. In test theory, the correlation between performances on the same test adminis­ tered at two different times provides a measure of "test-retest" reliability. In quanti­ tative genetics, the correlation between successive measurements on an individual is referred to as repeatability, and it provides an upper-bound estimate of heritabil­ ity. Thus, it is seen that the basic models and concepts of psychological test theory and quantitative genetics are highly similar. EVIDENCE FOR THE HERITABILITY OF SPECIFIC COGNITIVE ABILITIES Twin Studies

In his 1975 Annual Review of Genetics article concerning the heritability of IQ, Lewontin (64) completely dismissed the results of twin studies: I have quite deliberately not discussed the large number of simple comparisons of dizy­ gotic and monozygotic twins . ..., because any such comparisons must assume that the environmental similarities for dizygotic twins are the same

as

for monozygotic [italics

added]. The environmental dissimilarities will be completely confounded with genetic differences. (p. 396)

But how much weight should be given to the scanty evidence, as contrasted to intuition, of greater environmental dissimilarities within fraternal (DZ) than within identical (MZ) twin pairs? The most recent treatments of this topic have been by Scarr (99) and Vandenberg (120). After reviewing the evidence, Scarr concluded that "genetic relatedness" of twins determines the similarity of their treatment and that this is not necessarily evidence for environmental bias. Evidence that some abilities maybe more heritable than others is presented in Tables 1 through 4. F ratios of the within-pair variances of DZ and MZ twins are tabulated for subtests of the Primary Mental Abilities Test in four studies, for the Differential Aptitude Test in two studies, for the Wechsler Adult Intelligence Scale (WAIS) in one study, and for a number of separate tests in the Swedish study. It may be seen in these tables that the rank order of F ratios (from highest to lowest) tends to be as follows: spatial, vocabulary, word fluency, speed in simple arithmetic, and reasoning. Further evidence for the differential heritabilities of specific cognitive abilities may be found in Vandenberg (119). Is there evidence for the existence of genetically independent cognitive abilities? Modest attempts to answer this question have been made by Vandenberg (118), by Loehlin & Vandenberg (66), and by Bock & Vandenberg (14). If we can assume that the environmental influences produce the same degree of covariance for pair differ­ ences in the two types of twins, one can subtract the MZ from the DZ covariances to obtain an estimate of the genetic covariances (66), or one can generalize the F test that is used in univariate studies to determine the significance of the differ­ ence between the MZ and DZ within-pair variances (14). Instead of evaluating . . 2 / 2 F = (TwDZ (T wMZ, one ascertains h ow many elgen va1ues are sign! " ficant In ' the characteristic equation, .

186

DEFR IES, VANDENBERG & McCLEARN

Annu. Rev. Genet. 1976.10:179-207. Downloaded from www.annualreviews.org Access provided by University of California - Davis on 01/24/15. For personal use only.

7.

In practice, this number is usually equal to the number of eigen values greater than or equal to unity, although with increasing sample sizes more eigen values will acquire statistical significance. Using the procedure of subtracting MZ from DZ covariances, Loehlin & Vanden­ berg (66) concluded (on the basis of rather small samples) that the same ability factors appeared in the environmental and in the genetic covariation among 15 scales of Thurstone's Primary Mental Abilities Test. However, there were three oblique higher-order factors (number, space, and verbal-educational) in the environ­ mental covariation, while only one general factor appeared in the genetic covariance. Table 1 F ratios of the within-pair variances of dizygous and monozygous twins for sub­ tests of the Primary Mental Abilities ( P MA) Testa

Name of P MA subtest

----�."

Verbal Space Number Reasoning Word fluency Memory

Blewett 1954 3.13b 2.04c 1.07 2.7Sb 2.7Sb not used 26 DZ

26 MZ a Reprinted from Reference 119. bp < 0.01. cp < 0.05.

Thurstone 1955 2. S1b 4.19b

Vandenberg ( Michigan) 1962 2.65b

Vandenberg (Louisville) 1966

1.52 1.35 2.47b 1.62 53 DZ

1.77c 2.5Sb 1.40 2.57b 1.26 37 DZ

1.74c 3.51b 2.26b 1. 10 2.24b not used 36 DZ

45 MZ

45 MZ

76 MZ

Table 2 F ratios of the within-pair variances of dizygous and monozygous twins for the Differential Aptitude Testa

Verbal reasoning Numerical ability Abstract reasoning Space relations Mechanical reasoning Clerical speed and accuracy Language useI: Spelling Langu age use II: Sentences

a Reprinted from Reference 119. b p < 0.01.

1961 2.29b

1965 2.3Sb

1.39 1.47 1.67 1.36 2.54b 3.64b 3. 06b 25 DZ

1.37 1.23 2.19b 1.463.13b 2.58b 2.00b 86 DZ

47 MZ

109MZ

GENE TICS OF S PEC IF IC C OGN IT IVE A BIL IT IES

187

Table 3 F ratios of the within-pair variances of 60 dizygous and 60 monozygous twins for the scaled scores of II subtests of the Wechsler Adult Intelligence Scalea F

Subtest

3.88b 2.2Sc

Annu. Rev. Genet. 1976.10:179-207. Downloaded from www.annualreviews.org Access provided by University of California - Davis on 01/24/15. For personal use only.

Information Comprehension

2.78b

'Arithmetic Similarities Digit span Vo cabu lary Digit symbol

3.14b 2.06c

Picture completion Block d esign Picture arrangement

l.74d

1.8Id

1.S3d

LSD 2.3Sc

Object assembly

1.36

Verbal score Performance score Tot al s cor e

3.3Sb

3.41b 3.47b

aReprinted from Reference 119. bp < 0.001. c p < om. dp < 0.05.

Table 4 F ratios of the within-pair variances of like-sexed dizygous and monozygous twins for 14 psychological tests. administered to Swedish twins of elementary and high school agea

Simplex C-test Verbal analysis Form perception Picture perception Number perception Number series Number analysis Numerical classification Numerical reasoning Routine simple arithmetic Memory for 2-digit numbers (recall) Memory for 3-dig it numbers (recognition) Paired associates word-numb er memory

Boys

Girls

1.98b

2.84b

1.14

0.96

2.41b I.Slc

1.17

4.3Sb

1.23 I.S4c

All

cases 2.38b

3.37b

1.12

1.34c 1.36c

1.58c

1.91b

1.59b

1.61c

1.68c

1.63b

2.83b

1.96b

2.18b

1.15

1.39

1.24

2.37b

1.47

1.87b 0.94 1.32

1.7Dc

1.64c 1.Sle 1.34

2.01b l.57b

1.68b 1.17

66 DZ

1.00 75 DZ

141 DZ

66MZ

62MZ

128MZ

a Reprinted from Reference 119, ased on Reference 122. b bp < 0.01. cp < 0.05.

1.16

Annu. Rev. Genet. 1976.10:179-207. Downloaded from www.annualreviews.org Access provided by University of California - Davis on 01/24/15. For personal use only.

188

DEFR IES, VANDEN BERG & McCLEARN

Applying the second method to twin-pair differences on the Differential Aptitude Test, three dimensions were significant in the genetic covariance for boys and two for girls ( 1 4). In a large Finnish study of adult twins (91), a set of tests measuring six cognitive abilities (verbal, word fluency, space, number, memory, and reasoning) was adminis­ tered. A multivariate generalization of the F test led to the conclusion that the genetic covariance in the six abilities had six significant roots, that is, there were six independent genetic factors. On the other hand, when Eaves & Gale (28) reanalyzed the data of Loehlin & Vandenberg (66) with a method by which the additive genetic and dominance components of covariance could be estimated, they conclUded that a single genetic component fit well enough to suggest an absence of strong major gene effects for the abilities tested by the Primary Mental Abilities Test. It is clear that future studies will have to utilize larger samples, and more tests, to answer conclusively the question concerning the existence of genetically indepen­ dent abilities.

Family

Studies

Another methodology which has been employed to study the genetics of specific cognitive abilities is the family study. Although relatively few such studies have been reported to date, a major report was published by Willoughby (125) in the I 920s. He administered a battery of 11 individual tests of verbal and nonverbal abilities to 14 1 children between 12.5 and 13.5 years of age and to members of their families (90 fathers, 100 mothers, and 280 sibs). It may be seen from Table 5 that the average correlation for the verbal tests exceeded that for the nonverbal tests for all parent­ offspring combinations and for two of the three sib combinations. Since there is no difference between the spouse correlations, the greater parent-offspring and sib resemblance for verbal ability may be due either to its higher heritability and/or to a larger environmental correlation between relatives. Five years later, a family study of arithmetic and vocabulary abilities was reported by Carter ( 1 8). Four arithmetic subtests and two vocabulary tests were administered to members of 108 families in which both parents and one or more children over 12 years of age were available for testing and to 3 1 incomplete family groups. From Table 5

Familial correlations for verbal and nonverbal abilitiesa Verbal average

Nonverbal average

0.41

0.29

Mother-daughter

0.41

0.31

Mother-son

0.38

0.35

0.34

0.26

Father-son

Father-daughter Brother-sister

0.38

0.35

Brother-brother

0.50

0.37

0.40

0.52

0.44

0.44

Sister-sister Husband-wife aReprinted from Reference 125.

Annu. Rev. Genet. 1976.10:179-207. Downloaded from www.annualreviews.org Access provided by University of California - Davis on 01/24/15. For personal use only.

GENETICS OF SPECIFIC COGNITIVE ABILITIES

189

the resulting familial correlations presented in Table 6, it may be seen that the vocabulary tests yielded higher correlations for three of the four parent-offspring combinations and for all sib combinations. The greater spouse correlations for performance on the vocabulary tests may account at least in part for the higher parent-offspring and sib correlations. However, as in the Willoughby ( 1 25) study, a higher heritability and/or greater environmental correlation between relatives for verbal ability may also be involved. Three family studies of specific cognitive abilities have only recently been re­ ported. Utilizing a western Canadian sample, Williams (123) compared WISC subscale scores of 55 sons, 10 years of age, to WAIS scores of their parents. Although the same tests were not administered to both age groups, Williams (124) has shown that the two test batteries measure similar abilities. Spouse correlations and regressions of offspring on midparent value for the 11 subscales, for the verbal and performance IQ aggregates obtained from these subscales, and for total IQ are presented in Table 7. In agreement with previous evidence, the parent-son regres­ sions and spouse correlations tend to be higher on the verbal tests (subscales 1-6) than on the performance tests (subscales 7-11). This contrast is especially clear between verbal IQ and performance IQ. Note that the regression of offspring on midparent, which provides an upper-bound estimate for heritability, is only 0.46 for total IQ. Two other studies, the Hawaii Family Study of Cognition and the Boulder Family Study, are currently in progress. In both of these studies, data are being collected on 15 cognitive tests (see Table 8), various environmental indices, and blood group and enzyme systems. There are two major ethnic groups included in the Hawaii Study, Americans of Japanese ancestry (AJA) and Americans of European ancestry (AEA). Results of multivariate analyses have revealed that the cognitive structures of these two groups are nearly identical (26). Phenotypic correlations among the 15 cognitive variables were obtained for both AJA and AEA groups (N = 262 and 782, respectively), and principal component analyses with varimax rotations were performed. Four readily interpretable factors emerged from these analyses: spatial visualization, verbal, perceptual speed and accuracy, and visual memory. Common Table 6

Familial correlations for vocabulary and arithmetic abilitya Vocabulary

Arithmetic

n Father-son

n

0.21

116

0.10

116

Mother-daughter

0.34

136

0.24

136

Mother-son

0.07

126

0.12

127

Father-daugh ter

0.26

114

114

Brother-sister

0.31

III

0.04

Brother-brother

0.43

46

Sister-sister Husband-wife a Reprinted from Reference 18.

0.17 0.21

III 48

0.32

53

0.26

53

0.21

108

-0.04

108

190

DEFRIES, VANDENBERG & McCLEARN

Table 7 Spouse correlations and regressions of offspring on midparent for the Wechsler Intelligence ScalesB

Regression of offsp ring on midparent Spouse correlations

WISC/WAfS subscales 1.

Annu. Rev. Genet. 1976.10:179-207. Downloaded from www.annualreviews.org Access provided by University of California - Davis on 01/24/15. For personal use only.

2. 3. 4. 5. 6. 7. 8. 9. 10. 11.

Information Comprehension Arithmetic Similarities Digit span Vocabulary Digit symbol (coding) Picture completion Block design Picture arrangement Object assembly Verbal fQ PerformanceIQ TotalIQ

BReprinted from Reference

Estimate

Standard error

0.65

0.25

0.13

0.48

0.26

0.13

0.19

0.45

0.12

0.41

0.36

0.15

0.06

0.13

0.15

0.58

0.53

0.10 0.15

0.20

0.56

0.37

-0.12

0.20

0.19

0.34

0.19

0.30

0.07

0.18

-0.07

0.02

0.24

0.62

0.55

0.35 0.61

0.22

0.09 0.17

0.46

0.10

123.

factor loadings of the 15 cognitive tests on these four varimax-rotated principal components are presented for both ethnic groups in Figure 3. The marked similarity of the factor-loading profiles for AJA and AEA subjects is obvious from this illustration. Coefficients of congruence between the two groups were 0.96 or larger for each of the four factors. When these analyses were subsequently repeated with a larger data set (N for AJA 648, N for AEA 1971), similar results were obtained (126). In fact, congruence coefficients calculated from the larger data set were all 0.99 or larger. Zonderman and associates (131) have subjected the data from the Boulder Family Study to principal components analysis and have obtained the same four varimax­ rotated factors. Coefficients of congruence between this mainland AEA sample and the Hawaiian AEA sample were all greater than 0.94. Taken together, these data demonstrate that the 15 cognitive variables provide measures ofthe same underlying abilities in Hawaiian AJA, Hawaiian AEA, and mainland AEA samples. Assortative marriage for specific cognitive abilities has been assessed in both the Hawaii and Boulder family studies (57, 131). Previous studies of assortative mar­ riage for cognitive ability were recently reviewed by Jensen (53), who reported that the unweighted mean of 43 spouse correlations for various tests of mental ability was +0.45. However, at least some of these correlations may have been inflated. When cross-sectional data concerning mental ability are examined, some change occurs as a function of age (see 126). Thus, when both older and younger couples =

=

GENETICS OF SPECIFIC COGNITIVE ABILITIES

191

Table 8 Administration time and reliabilities for tests utilized in the Hawaii Family Studyof Cognition and the Boulder Family Studya Cognitive Test

factor

Test time

Reliability

V

3min

0.96

Primary Mental Abi lities (PM A) Vocabulary

Annu. Rev. Genet. 1976.10:179-207. Downloaded from www.annualreviews.org Access provided by University of California - Davis on 01/24/15. For personal use only.

Visual Memory (immediate)

M

Things (a fluency test)

V

I-min exposure/

0.58

2parts/3min

0.74

I-min recall each

Sheppard-Metzler Mental Rotations

S

10min

0.88

(modified for group testing by Vandenberg) Subtraction and Multiplication

P

2parts/2min

0.96

each Elithorn Mazes ("lines and dots"),

S

5 min

0.89

V

2 parts/3min

0.71

shortened form Educational Testing Service (ETS)

each

Word Beginnings and Endings ETS Card Rotations

S

2 parts/3 min

0.88

each Visual Memory (delayed recall)

M

1min

0.62

PM A Pedigrees (a reasoning test)

V,S,P

4min

0.72

S

2parts/2min

0.92

ETS Hidden Patterns

eaeh S

3 min

0.84

ETS Number Comparisons

P

2parts/1.5 min

0.81

Whiteman Testof SocialPerception

V

10min

0.69

Raven's Progressive Matrices,

S

20min

0.86

Paper Form Board

each

modified form a Reprinted

from Reference 126. b Varimax rotated factor loading;;' 04 . 0

ceptual speed and accuracy, M

visual memory).

=

are included in a sample, some of the observed spouse similarity in mental ability is due to their similarity in age. For this reason, it is essential that age adjustment be employed in studies of assortative marriage for cognitive ability. As discussed below, method of test administration may also influence spouse correlations. Age-adjusted spouse correlations for couples included in the Hawaii and Boulder family studies are presented in Table 9. With regard to the Hawaiian samples, AEA spouse correlations are higher than those for AJA subjects on 13 of the 15 individual tests. However, the magnitude of the differences is not large (median AEA and AlA spouse correlations were +0.13 and +0.06, respectively). In addition, spouse corre­ lations for factor scores were highest for the verbal factor in both ethnic groups.

DEFRIES, VANDENBERG & M cCLEARN

192 1.0 .8

SPATIAL

VERBAL

.6 .4 .2





Annu. Rev. Genet. 1976.10:179-207. Downloaded from www.annualreviews.org Access provided by University of California - Davis on 01/24/15. For personal use only.

c:::i

a -.2 -.4

� ...,J

-.5

� "

1.0

(",)



.8

I I I PERCEPTUAL

--AEA --------

SPEED

AJA

VISUAL

MEMORY

.6 .4 .2 0 -2

4

-.

-!j

I I

3

Figure 3

7

9

13

15

I I

3

5

7

9

"

13

I I

15

Loadings of 15 cognitive tests on fou r varimax-rotated principal components in

Americans of Japanese ancestry (AJA) and Americans of European ancestry (AEA). Re­ printed from Reference 26.

Thus, it does not appear that ethnicity is a major influence on the extent and nature of assortative marriage in Hawaii. When the mainland AEA sample is compared with the Hawaiian samples, the mainland spouse correlations are somewhat higher (median value of +0.22). This difference, however, may be due in part to a difference in method of test administration. In the Hawaii study, tests are administered to large groups under highly standardized conditions (automated tape and slide presentation of instructions, etc). In contrast, tests in the Boulder study are usually administered to individual nuclear family units by various testers, so that differences among testers could cause increased spouse correlations in this study. Nevertheless, results of both studies suggest relatively less assortative marriage for cognitive abilities than previously reported. When effects of educational attainment were statistically re­ moved, resulting partial correlations became even smaller. Thus, there is very little assortative marriage for that part of cognitive ability which is independent of educa­ tional level (57, 131). Parent-offspring resemblance for specific cognitive abilities has also been exam­ ined in both the Hawaii and Boulder family studies (25, 105). When environmental effects are randomized among subjects, the regression of offspring's score on midpar­ ent value provides a direct estimate of heritability (32). However, because environ­ mental influences on at least some measures of cognitive ability may be common to parents and their offspring, regressions of offspring on midparent value for such characters should be regarded as measures of phenotypic similarity (genetic and/or environmental) and only as upper-bound estimates of heritability (77).

1 93

GENE TICS OF SPECIFIC C OGNI TI VE ABILI TIES Table

9

Age-adjusted spouse correlations Hawaii a AEA (N 555 couples)

Annu. Rev. Genet. 1976.10:179-207. Downloaded from www.annualreviews.org Access provided by University of California - Davis on 01/24/15. For personal use only.

=

Tests Vocabulary Visual Memory (immediate) Things MentalRotations Subtraction and Multiplication Elithorn Mazes ("lines & dots") Word Beginnings and Endings Card Rotations Visual Memory (delayed) Pedigrees Hidden Patterns Paper Form Board Number Comparisons Social Perception Progressive Matrices. Factors Spatial Verbal Perceptual speed and accuracy Visual memory First principal component (unrotated) a Reprinted from Reference 57. AEA icans of Japanese ancestry . b Reprinted from Reference 131.

0.27

(N

=

148

couples)

Mainland b AEA (N 1 23 couples)

0.16

=

0.36

0.07

0.15

0.09

0.22

0.08

-0.12

0.11

0.04

0.03

0.10

0.18

0.07

0.00 0.12 0.10

0.13

0.02

0.12

0.18

0.04

0.03

0.10

0.41

0.06

0.26

0.06

0.28

0.16

om

0.10

0.21

0.22

0.Q3

0.20

0.00

0.23

-0.05

0.13

0.19

0.20 0.13

0.22

0.13

=

Hawaii a AJA

0.20

0.00

0.00

0.13

0.28

0.13

Americans of European ancestry; AJA

0.36 0.35

0.09

0.35

0.41 0.05

0.35

=

Amer­

Regressions of midchild on midparent obtained in the Hawaii study are presented in Table 10. These regressions have been adjusted for lack of perfect test reliability (an adjustment which is essential when estimates from tests with different reliabili­ ties are to be compared). Standard errors for the regression coefficients are suffi­ ciently small to suggest that at least some of the variability among tests is real. Another indication of this differential parent-offspring resemblance is the marked congruence between ethnic groups. The Spearman rank correlation between the 1 5 AEA and AlA regressions for the individual tests was +0.77 (p

Genetics of specific cognitive abilities.

ANNUAL REVIEWS Further Quick links to online content Ann. Rev. Genet 1976. 10:179-207 Copyright © 1976 by Annual Reviews Inc. All rights reserved...
1MB Sizes 0 Downloads 0 Views