Growth and Intellectual Development

(1)

Growth

and Intellectual

Development

Darrell

M. Wilson,

MD, Lawrence

D. Hammer,

MD,

Paula

M. Duncan,

MD,

Sanford

M. Dornbusch,

PhD,

Philip

L. Ritter,

PhD,

Raymond

L. Hintz,

MD,

Ruth

T. Gross,

MD,

and

Ron G. Rosenfeld,

MD

From the Departments of Pediatrics and Sociology, Stanford University, Stanford, California

ABSTRACT. Data from the National Health Examina-tion Survey (cycles II and III) provided a representative sample of 13,887 US youths (6 to 17 years of age) with

which to examine the relationship between height

(nor-malized for age and sex) and measures of intellectual development (Wechsler Intelligence Scale for Children) and academic achievement (Wide Range Achievement Test). Additionally, 2,177 subjects were studied first in cycle II and 2 to 5 years later in cycle III, forming a well-selected longitudinal study group in which to examine any association between linear growth and change in IQ scores. Wechsler Intelligence Scale for Children and Wide Range Achievement Test scores were significantly correlated with height in both cycle II and cycle III. However, no significant association between change in relative height and change in IQ scores could be detected in the longitudinal group. These data suggest that ther-apies designed to increase height are unlikely to alter measures of intellectual development or academic achievement. Pediatrics 1986;78:646-650; IQ, growth, height.

The association between physical stature and

achievement has been the subject of a number of

interesting observations. Gowin’ showed that in

1916, within the same general area of employment,

those

in more prestigious positions were taller than

those

in positions of lesser prestige. Since 1900, 17 of the 21 presidential elections have been won by

Received for publication July 10, 1985; accepted March 14, 1986.

Presented, in part, at the Society for Pediatric Research, 1984,

at the Serono Symposium-The Pyschosocial Aspects of Delayed

Growth, Oct 20, 1984, Washington, DC, and at the American Federation for Clinical Research, Western Section, Feb 7, 1985, Carmel, CA.

Dr Duncan’s current address is University Pediatrics, 1 5 Pros-pect, Burlington, VT 05401.

Reprint requests to (D.M.W.) Department of Pediatrics, 5-322,

Medical Center, Stanford University, Stanford, CA 94305.

American Academy of Pediatrics.

the taller candidate. Moreover, all but two

presi-dents since George Washington have been taller

than the average height for men of their time.2

These and other studies have suggested a more than

chance association between height and

achieve-ment. Furthermore, there is evidence that children

who mature more rapidly than average have higher

IQ scores.3’4

In this study, we examined data from a national survey to determine whether an association exists

between height and intellectual development and academic achievement in children and adolescents.

We also examined a large longitudinal subset of

these subjects to determine whether there is any

association between linear growth and change in

IQ scores. We chose this recent large-scale survey

because its excellent anthropometric measures and

data on social background permitted extensive

analyses of the other factors that have been

re-ported to confound the relationship between height and intellectual development.

MATERIALS

AND

METHODS

Cycles II and III of the National Health

Exami-nation Survey (NHES) were conducted during the

1960s by the National Center for Health Statistics to obtain health-related data from a national

prob-ability sample of the noninstitutionalized US

youth.5 During cycle II (1963 to 1965), 7,119

chil-dren, 6 to 1 1 years of age, were studied, and during

cycle III (1966 to 1970), 6,768 adolescents, 12 to 17 years of age, were studied. In both samples, exten-sive data concerning the intellectual, physical, and psychosocial status of each subject were collected.

Both samples were carefully stratified to represent

the target US population with respect to age, sex,

race, geographic region, population density, and

(2)

U, Ui 0 0 U, a 108 104 100 96 92 88 <5 5-25

1 1 years of age and first evaluated during cycle II,

were subsequently reevaluated 2 to 5 years later

during cycle III, resulting in a large, well-selected longitudinal study population.

To obtain an index of height normalized for age

(to the nearest month), we calculated the number of standard deviations each subject’s height differed

from the mean height for that age and sex: the Z

score for height. An index of relative physical ma-turity was calculated by dividing each subject’s bone age by his or her chronologic age. We estimated the

socioeconomic status of the subjects by using their

family income. Family income data (dollars per

year) was obtained by interview and then divided

into ten categories. For some analyses, the subjects

were divided into three income groups: <$5,000

(cycle II, n = 2,503; cycle III, n =

1,742), $5,000 to

$10,000 (cycle II, n = 3,103; cycle III, n = 2,619),

and >$10,000 per year (cycle II, n = 1,142; cycle

III, n = 1,912). Race was described as either black

or white. Number of children was estimated from

the NHES data as the number of children younger than 20 years of age residing in the household and

the birth order of each subject was estimated by

comparing the age of the subject to the other chil-dren younger than 20 years of age residing in the

household.

The vocabulary and block design subsets of the

Wechsler Intelligence Scale for Children (WISC)

were used to estimate intellectual ability, and the reading and arithmetic subsets of the Wide Range

Achievement Test (WRAT) were used as

approxi-mate measures of academic achievement. These IQ

tests were administered during both cycle II and

cycle III.

For each subject in the longitudinal sample, we

calculated the change in height Z score, WISC, and WRAT scores by subtracting the values obtained during cycle II from the values obtained during cycle III.

Statistical analysis was performed using the Sta-tistical Analysis System (SAS Institute, Cary, NC). Pearson’s product moment correlation coefficients

were used to detect relationships between variables. Multiple linear regression models (with stepwise addition of variables) were used to determine the relative contribution of different variables to the

WISC or WRAT score. Because of occasional

miss-ing values, summation of the number of subjects across subgroups within a given analysis may differ

slightly from the total number of subjects in that

group.

RESULTS

WISC and WRAT scores significantly correlated with the height Z score in both cycles II and III

WISC WRAT

Z:

nil..

-.

:C

1

, /

I

/

25- 75- >95 <5 5- 25 75 >95

75 95 25 75 95

HEIGHT PERCENTILE GROUPS

Fig 1. Relationship between height and Wechsler

Intel-ligence Scale for Children (WISC) and Wide Range

Achievement Test (WRAT) scores for entire sample di-vided into five groups based on height (normalized for age and sex). Mean and standard error of mean of IQ

scores are shown.

(cycle II, r = .18 height with WISC, r = .17 with

WRAT; cycle III, r = .20 with WISC, r = .19 with

WRAT; P < .0001 for each r value). This relation-ship is graphically illustrated for the subjects in

cycle III

in Fig 1. Mean IQ scores increased more

than 14 points from the shortest to the tallest

group. The correlations between height Z score and

WISC and WRAT scores persisted even when the

subjects were divided into subgroups based on their

sex, socioeconomic status, or race. The correlations between height and IQ scores were less among the

subjects with family incomes >$10,000/yr in both cycles II and III (cycle II, r = .08 with WISC (P < .05), r = .06 with WRAT (P < .05); Cycle III, r = .12 with WISC (P < .0001), r = .08 with WRAT (P

< .001)).

To determine the effect of other variables on WISC and WRAT scores, we developed series of

multiple linear regression models with WISC or

WRAT scores as the dependent variables and

fam-ily income, race, relative maturity (bone age to

chronologic age ratio), height Z score, family size, birth order, and an interaction term of family size and birth order as the independent variables. This

interaction term was calculated by multiplying

birth order and family size after subtracting the

sample mean from each variable.

The results of these models for boys and girls in

both cycle II and cycle III are shown in the Table.

In cycle II, socioeconomic status, race, family size,

and height Z score contributed significantly to the

models

for WISC and WRAT. In cycle III,

socio-economic status, race, height, birth order, and

fam-ily size significantly entered the models. Relative

maturity and the interaction term of family size

and birth order did not consistently improve the

(3)

TABLE.

Multiple Linear Regression Models*

Cycle and Variable Boys Girls

WISC WRAT WISC WRAT

R2 STD B R’ STD B R’ STD B R2 STD B

Cycle II

Family income .22 .342 .17 .310 .22 .340 .20 .343

Race .28 .236 .20 .168 .26 .217 .22 .167

Family size .30 -.144 .22 -.132 .29 -.214 .24 -.119

HeightZscore .32 .128 .24 .125 .30 .070 .25 .115

Birth order X .30 .060t

family size Cycle III

Family income .22 .357 .20 .337 .22 .350 .22 .353

Race .26 .207 .25 .223 .26 .226 .27 .247

HeightZscore .28 .138 .27 .127 .28 .145 .29 .127

Birthorder .29 -.084 .28 -.085 .29 -.115 .30 -.135

Family size .29 -.035t .28 -.035 .30 -.119 .30 -.147

Birth order X .30 .102t .30 .159t

family size

Bone age/chron- .30 .038t

ologic age

* R2, Cumulative R squared value; STDB,standard partial regression coefficients for each model; WISC, Wechsler Intelligence Scale for Children; WRAT, Wide Range Achievement Test. Overall, each model is highly significant (P < .0001, hypothesis F > 1). Each variable’s contribution is also significant (P < .0001, hypothesis F > 1, except where

marked otherwise).

tP.cz.05.

:1:In the model for WRAT scores, among the girls of cycle III, family size entered before

birth order.

no major differences noted between boys and girls. In all models, the height Z score contributed

signif-icantly (P < .0001) to the prediction of WISC and WRAT scores, accounting for approximately 2% of

the total variance found in IQ scores.

Whereas a correlation could be found between

height and IQ scores in subjects from both cycle II

and cycle III, no correlation existed between the

change in height Z score and change in either WISC

or WRAT scores (P > .2, Fig 2). The analysis was repeated using only those subjects whose initial

height Z score was less than the 1 SD below the

mean, and again, no correlation existed between height and IQ scores.

DISCUSSION

Several authors have related height and various

measures of intellectual functioning in normal pop-ulations. In 1892, Porter6 studied 33,500 students

and showed that taller, heavier pupils performed

better academically than did smaller students of the same age. Analysis of data from the Scottish

Mental Survey7 of 6,921 11-year old children born

in 1936 revealed a correlation between height and

IQ scores in both boys and girls (r = .24, boys; r =

.26, girls). Douglas et a18 also found a correlation

between height and IQ scores in 5,362 British

chil-dren born in 1946 (r = .14, boys; r = .12, girls). In

the Newcastle Upon Tyne study of 762 students born in 1947 and followed longitudinally until 15

years of age, Miller9 found significant correlations

between IQ scores and height. In a study by

Wein-berg et al.1#{176}of 334 children, 8 and 9 years old, in St

Louis, height correlated with IQ (r = .20). Scott”

reported a significant correlation between height

and IQ scores (r values ranged from .14 to .36) and

further noted an effect of family size on both height and IQ scores. Belmont et al’2 also reported that both family size and birth order interacted with height and IQ scores in a complex manner. Con-versely, Richards et al’3 found no association be-tween achievement scores and height in 481 seventh grade students.

There are other factors that have been associated with the results of intelligence testing and could

confound the small association between height and IQ scores found in these earlier studies. These

include socioeconomic status, race, the size of the

family, birth order, and the relative rate of physical

maturity. The extensive data obtained during the

NHES afforded a unique opportunity to analyze

the effects of these factors and, in our study of

nearly 14,000 subjects, a significant association

be-tween the height Z score and both WISC and

WRAT scores remains even after all ofthese poten-tially confounding influences are controlled for.

(4)

10.

5.

-15

- 20

-2.5 -1.5 -0.5 0.5 1.5 2.5

15-C

H

A

N

G E

N

H

C

S

C

0

R

E

CHANGE IN HEIGHT Z SCORE

Fig 2. Mean change (± 1 SD) in Wechsler Intelligence Scale for Children (WISC) scores V change in height Z score among subjects in longitudinal group. Overall mean increase in WISC scores from cycle II to cycle III of .716 has been removed. Similar plot using mean change in Wide Range Achievement Test scores was essentially identical.

effect of relative maturation on IQ scores. The

relative strengths of the variables in predicting IQ

score are slightly different in the two age groups. In particular, family size is a stronger predictor

than height in cycle II (ages 6 to 11 years), whereas

the converse is true in cycle III (ages 12 to 17

years). This finding may reflect the fact that our estimate of family size (which only includes persons younger than 20 years of age living in the subject’s

household) is a less reliable measure of family size

in the older subjects. However, when we repeated the multiple linear regression analyses using only subjects who had no siblings living out of the

house-hold, we obtained very similar results.

The association between height and these

imper-fect estimates of intelligence does not appear to be

limited to children and adolescents. Significant

cor-relations between height and intelligence have also

been detected in studies of adults (as reviewed by

Tanner4). Although these studies each have

meth-odologic flaws, they suggest that the association

between height and IQ in childhood persists into

adulthood and does not reflect transient differences

in relative physical maturity or growth.

Reasons for the association between height and

IQ scores remain unclear. Both of these parameters

may covary with another factor, such as subtle

(5)

alternative explanation is that, regardless of the biologic factors that contribute to stature, children

and young adults of varying height are treated

differently by others. Adults and peers may interact

differently with short children than they do with

tall children. Thus, the differences in WISC and

WRAT scores between short and tall children may

be, in part, the consequence of years of cumulative

adult height-biased expectations.

We have demonstrated a small, but statistically

significant, association between height and IQ in

both cycle II and cycle III of the NHES. However,

within the longitudinal subgroup, numbering more

than 2,000 subjects, no association between change in height and change in IQ score could be detected.

Taken together, these data imply that the processes that contribute to the association between height and IQ must occur relatively early in childhood, at

least before the age of the children studied in the

longitudinal sample.

This study also demonstrates that changes in

relative height between the ages of 8 and 13 years

do not significantly affect IQ scores. With the

avail-ability of virtually unlimited amounts of

recombi-nant DNA-derived human growth hormone for

clinical use, the possible effect of this therapy,

designed to alter height, on IQ scores becomes

particularly interesting. The results of this study

clearly indicate caution in any expectations of such

treatment. Until careful clinical studies examining

this issue have been completed, any effect of such

growth-promoting therapies upon intellectual

de-velopment or academic achievement must be

con-sidered unlikely.

ACKNOWLEDGMENT

This work was supported, in part, by funds from the

Stanford Center for Study of Youth Development and Father Flanagan’s Boys’ Home.

REFERENCES

1. Gowin EB: The Executive and His Control of Men. New

York, Macmillan Co, 1916, pp 22-33

2. Gillis JS: Too Tall, Too Small. Champaign, IL, Institute for Personality and Ability Testing, Inc, 1982, p 16

3. Duke PM, Carlsmith JM, Jennings D, et at: Educational correlates of early and late sexual maturation. J Pediatr

1982;100:633

4. Tanner JM: Relation of body size, intelligence test scores, and social circumstances, in Mussen PH, Largen J, Coving-ton M (eds): Trends and Issues in Developmental Psychology.

New York, Holt, Rinehart & Winston, Inc., 1969, pp

183-209

5. Plan and Operation of a Health Examination Survey of U.S.

Youths 12-1 7 Years of Age, Vital and Health Statistics,

National Center for Health Statistics, series 1, No. 8. Gov-ernment Printing Office, 1974

6. Porter WT: The physical basis of precocity and dullness.

Trans Acad Sci 1893;6:161

7. Scottish Council for Research in Education. Social

Implica-tions ofthe 1947 Scottish Mental Survey. London, University

Press, 1953, p 91

8. Douglas JWB, Ross JM, Simpson HR: The relationship between height and measured educational ability in school children of the same social class, family size, and stage of

sexual development. Hum Biol 1965;37:178

9. Miller FJ: The School Years in Newcastle Upon Tyne

1952-62: Being a Further Contribution to the Study of a Thousand

Families. New York; Oxford University Press, 1974, p 2238 10. Weinberg WA, Dietz SG, Penick EC, et at: Intelligence,

reading achievement, physical size and social class. J Pediatr

1974;85:482

1 1. Scott JA: Intelligence, physique, and family size. Br J Prey

Soc Med 1962;16:165

12. Belmont L, Stein ZA, Susser MW: Comparison of

associa-tions of birth order with intelligence test score and height.

Nature 1975;255:54

(6)

1986;78;646

Pediatrics

L. Ritter, Raymond L. Hintz, Ruth T. Gross and Ron G. Rosenfeld

Darrell M. Wilson, Lawrence D. Hammer, Paula M. Duncan, Sanford M. Dornbusch, Philip