LIRTA; OWifl EE.r LCTS
AID FLUIDr/CY SILLIZED INTELLiGENCE
Sa d .e < i..'son i Lt
A DYS:Ri A.r',i ,
E i, -L
S.,,i T: i T
L. M L OF
i'u' S F0 TI
L, R I 'v Y 0i
. 1 .
;E I n 'i" ,i
I would like to acknowledge my sincere appreciation to Dr. Walter
R. Cunrningham, the chairman of my doctoral committee and director of
this dissertation, whose valuable suggestions, constructive criticisms,
and encouragement made this study possible. Special thanks are extended
to the other members of my doctoral committee, Dr. W. Keith Berg, Dr.
Richard Griggs, Dr. James Whorton, and Dr. Nathan Perry. Their assistance
and practical s-. ]estions were greatly appreciated.
I would also like to thank Kathy Moser for her capable assistance
.ith the ration of tnis manuscript, Regis :l;,/ar for preparing thu
figures, and Patricia ,.ichardson and Tom Harbin for their assistance wi h
the t administration.
Sreover, this effort was made somewhat less difficult by the moral
sj; rt of .. husband, Bill, and the incredible motivating influence J.
my da ugters, ,.'y and Kate.
TABLE OF CONTENTS
ACKNOWLEDGMENTS . . . . . . . . . .
LIST OF TABLES . . . . . . . . . .
LIST OF FIGURES . . . . . . . . . .
ABSTRACT . . . . . . . . . . . .
C'i'PTER I INTRODUCTION . . . . . . .
Birth Order/Family Size Theories. . .
Fluid and Crystallized Intelligence
C ^- TT- II METHOD ,' 3 DATA ,-,;ALYSIS . . . .
Overview and Design . . . . .
Study 1 . . . . . . . .
St'1-v2; 2 . . . . . . . .
S 3 . . . . . . . .
CHAF'-_. III DISCUSSION . . . . . . .
University Student Sample . . . .
High School Sample . . . . .
Ii,:plications . . .
directions for Future Research ..
SUPPLr, _,;,T ., REFERENCE . . . . . . . .
E FrI, -PI S..................
BIG3iAPHICAL SKETCH . . . . . . . . .
LIST OF TABLES
1 Means and Standard Deviations for College
Student Sample . . . . . . ... .21
2 Promax Rotated Factor Loadings . . ... .23
3 Multiple Regression of Predictor Variables
on G for College Sample . . . . . 29
4 -,v, and Standard Deviations for High
School Sample . . . . . . . . 41
5 Multiple Regression of Predictor Variables
on G- for the High School Sample ...... .45
LIST OF FIGURES
Figure i Linear Structural Equation Moaels .10
Figure 2 Intercorrelations among Variables in
each model for the College Student
Sample . . . . . . . 36
Figure 3 Intercorrelations among Variables in
each Model for the High School Sample. 43
.'r tract of Dissertation Presented to the Graduate Council
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
BIRTH ORDER EFFECTS
AND FLUID/CRYSTALLIZED INTELLIGENCE
Sandra Johnson Witt
Chairman: Walter R. Cunniii.h.min
" 'or rtnent: Ps ', .-lc .'
Th- .-. ,t investi--.tion was an att,_ t to separate out some of
the co. d relationships between the family confi ,i tion correlates
of abili. and achievement performance. In the first study, a battery
of 21 tests was administered to 111 university students. After factor
anal the results for optimal marker variables of the fluid intelligence
(Gf--active problem-solving abili; ) and c tallized intelli .nce (Gc--
ac ieve ent-relate F .,'-rmance) constructs c Cattell and Horn, four
liner structural tion models relati" Gc to various tnesized
Secictor variables were evaluated in the original university sample, and
an additional :- .,le consisting of 65 high school students. The
individual models emphasized the importance of either economic, intellectuall
nviro ent," biological, or socio-psychological influences in the
iedi ion of birth order and/or family size effects on intellectual
abili y -.for rnce. t was t.,. l.;,ized that the models were not mutually
exclusive, but mi .'it be differentially applicable in different -'..,lations.
Certain unexpected correlations obtained in the university student
sample, which might be specific to such a population, rendered
straightforward interpretation of the models difficult. The finding
of a significant positive correlation between family size and
socioeconomic status was in marked contrast to the often reported
negative intercorrelation of these two variables in the general
population. This result was discussed in terms of financial resources
and limitations, which may play a decisive role in college attendance.
The only model supported in this sample was a modified version of the
The ability of the models to predict Gc performance in a more
hete-. -..i..ous high school sample was also evaluated. All of the models
were better supported in this sample. However, whereas the overall F
ratio for each of the models was significant, the theoretically
expected rc ;,-:ions for the individual predictor variables on Gc were
significant only for the economic and socio-psychological models. It
was concluded that, not only were these two models able to account for
substantial amounts of variance in the prediction of Gc performance,
but that the separate paths indicated by both models were strongly
supported by the obtained data. After comparing the individual paths
postulated by these two models, it was concluded that the socio-
psychological model was the optimal predictor of Gc performance. This
result was largely attributable to the highly significant regression
for the firstborn (versus later-born) variable on G
With regard to the very different results obtained in the college
student and high school student samples, it is apparent that one must
specify the kind of sample involved when addressing issues related to
family configuration and socioeconomic correlates of intellectual
The very positive results obtained for the heterogeneous high
school sample indicate the significance of certain social and
psychological factors as mediators of family configuration
influences--particularly firstborn superiority--on intellectual
achievement. .-,'..over, economic influences seem to exert powerful
effects, both directly and indirectly, on intellectual ability
It was s': _-ted that the utilization of innovative statistical
techniques and conceptual approaches might serve to more effectively
isolate the extremely :,- ..lex family confi ur .-tion correlates of
intellectual ititude and achievement.
,iile the relationships of birth order and family size to
achievement-related variables have been studied extensively, the
conclusions which can be drawn from these research efforts are often
contradictory and, at best, tentative. Problems involved in the study
of such relationships are complicated by the fact that (a) family size
and birth order are highly correlated, (b) personality and
ability/achievement are not independent, and (c) socioeconomic status
(SES) is correlated with both family size and the personality-ability-
achievenent complex (Eysenck & Cookson, 1970).
A consistent finding of studies which have related birth order to
eminence has been a preponderance of firstborns among eminent individuals
(J. M'. attell, 1927; Ellis, 1904; Galton, 1874; Huntington, 1928). In
Ternan's (1' 5) study of gifted children, there were more firstborns than
would nave been expected by chance. Altus (1965a) reports that the
firstborn individual is greatly overrepresented at the college level.
Nichols (1",") presents data on National Merit Scholarship finalists.
Nearly 60% of the finalists who came from families of two, three, four,
and five children were firstborns. Altus (1965b) studies Scholastic
Aptitude Test scores and concluded that firstborns in college may be more
able verbally than later borns; no differences appeared in mathematical
aptitude. Several other studies have found firstborn superiority on
various achievement-related measures (Breland, 1974; Burton, 1968;
Glass, Neulinger, & Brim, 1974; Marjoribanks & Walberg, 1975).
In addition to the greater achievement of the firstborn, there
may exist hierarchies of aptitude among the intellectually able related
to birth order and family size. Zajonc (1976) reported that
intellectual level generally declines with family size and certain age
spacin-: relationships, which are important in terms of the family's
immediate intellectual environment. The dependence of the intellectual
development of each family member on that of all other family members
is emphasized in this theory. However, the relationships between
family size and intellectual performance appear to be attenuated or
may di r in higher SES families. In a study of 11-year-old
children, Cicirelli (1967) found that family size and achievement
were unrelated. It appears that, in middle and upper-middle SES
populations, children from larger families are able to obtain adequate
attention and learning opportunities. Kennett (1973) suggests that
the at. tance of an inverse relationship between family size and
intelli ice must be modified by the influence of SES on the
relationship, and that correlations may vary from social class to
ile firstborn superiority on achievement-related variables has
been widely reported, consistent patterns relating decreasing intellectual
ability scores to successive birth order position have been less
strai ihtforward. Although several studies ,'-a*:'t hierarchical
relationships (Altus, 1965b; Belmont & Marolla, 1973; Breland, 1974;
Burton, 1968; Nichols, 1964; Wark, Swanson, & Mack, 1974), findings are
often inconsistent. For example, eminence studies (Apperly, 1939;
J. M. Cattell, 1927; Clarke, 1916; Ellis, 1904; Terman, 1925;
Yoder, 1894) report a preponderance of firstborns and lastborns ar-on-
persons of genius. Similarly, studies of birth order among older
children and young adults often demonstrate superiority of firstborn
over later-born children on IQ and achievement, with the middle child
frequently making the poorest score (Cicirelli, 1967; Corliss, 1964;
Hall, 1963; Lees & Stewart, 1957; Maxwell & Pilliner, 1960; Rosenberg
& Sutton-Smith, 1964; Schachter, 1963). Record, McKeown, and Edwards
(1969) report that the strong association between verbal reasoning
scores and birth order in the general population is mainly due to
differences between rather than within families. In upper SES
families the di..'erence between consecutive siblings (0.7 IQ points)
is sial], but in poorer families the difference (2.0 IQ points) is
The reasons for these discrepant data are unclear. Marjoribanks,
Wall and Ba .,, (1975) report that birth order effects are rarely
unitary, but often involve other familial aspects, such as the sex of
the sibli., and their differences in age. Cicirelli (1967) found
that such sibli constellation factors were related to measures of
creativity, arithmetic ability, and language achievement in a sixth-
Birth Order/Family Size Theories
There are at least four major points of view concernin- birth order
and family size influences on intelligence. The theories considered in
this investigation emphasize the importance of either economic,
"intellectual environment" (confluence), physiological, or socio-
psychological factors in the determination of birth order and
family size correlates of intellectual ability and achievement.
These models will be discussed individually.
Many investigators contend that birth order and family size
effects are largely reducible to SES and the economic factors involved.
For example, in the past the concept of primogeniture (the exclusive
right of inheritance bestowed on the eldest son) was directly related
to economics. Anastasi (1956) reports a negative correlation of IQ
with family size, which disappears or becomes positive in higher
SES levels. Fisbet and Entwistle (1967) note a negative correlation
of -. ;ly size with intellectual ability in school children, an effect
which is attenuated in upper SES groups. Kennett (1973) found a
si,..ificant inverse relationship between family size and IQ in junior
high school students, except in the upper two SES categories.
Marjoribanks et al. (1975) report that, while children with more
sibli: tend to score lower on verbal and number ability tests, the
relation is attenuated in high SES families. Schooler (1972, 1973)
conLends that birth order effects can be most parsimoniously explained
in terms of differences among SES trends in family size. These studies
sU t a changing relationship between family size and intelligence
in di' erent SES groups.
Zajonc- rkus Confluence Model
In the confluence model (Zajonc, 1976; Zajonc & Markus, 1975),
ordinal differences in intelligence are explained in terms of family
size and age spacing between siblings. Zajonc and Markus attempt to
account for the effects of the immediate "intellectual environment"
(represented as a function of the average of the absolute intellectual
levels of the family members) on intellectual development, and to
specify how individual differences emerge in the social context of the
family. For example, if the parents' intellectual levels are 30
arbitrary units each, the birth of the first child, whose intellectual
level is zero, would cause the family intellectual environment to have
an average value of (30 + 30 + 0)/3 = 20. If a second child is born
when the first child's intellectual level reaches 4, the second born
enters an environment of (30 + 30 + 4 + 0)/4 = 16. The importance
of age spacing between siblings could be seen, for example, if the
second child were not born until the firstborn child reaches an
intellectual level of 24. The newborn would enter an environment of
(30 + 30 + 24 + 0)/4 = 21, even more favorable than the environment
of entered into by the firstborn. Negative effects of birth order
can be nullified and even reversed with large enough age gaps between
sibli., s, accordir,- to this model. In addition, it is pro. osed that
only children and last-born children suffer a handicap due to their
lack of opportunity to serve as teachers to younger siblings. Thus,
it is suggested that the intellectual growth of each family member is
dependent on that of all the other members, and that the rate of
growth depends on family configuration.
P:', biological Model
Proponents of this point of view discuss the effects of increased
chance of c~netic error and "uterine fatigue" with later-born children
(Forer, 1976). Warren (1966) indicates that the intrauterine
environment may vary with the mother's age and number of previous preg-
nancies. Such influences would favor earlier-born children. However,
a smaller placenta, longer labor, increased perinatal complications,
and increased use of forceps are often associated with pregnancy and
birth of the firstborn, influences which would favor later-born
children (Weller, 1965). Thus, it is evident that the physiological
model may si'j:'est quite complex effects of family configuration.
h.. recent research has supported the importance of socio-
ps.L' ,logical variables related to birth order and family size. Many
studies'indicate personality differences between different birth
orders and with differing family sizes (Eysenck & Cookson, 1970;
Forer, 1976). Bradley (1968) discusses personality factors which
favor firstborns in school-related behaviors:
Firstborns seem to more frequently (a) meet teachers'
expectations and (b) show more susceptibility to social
pressure than later borns. Exhibiting (c) greater
information-seeking behavior and (d) being more sensitive
to tension-producing situations, firstborns may be ju.jT- by
others as (e) serious and (f) low in ag session. These behav-
iors -, (g) strengthen firstborns' achievement motivation
and (h) help to enhance their academic performance. (p. 45)
Start and Start (1974) studied teacher ratings of first-grade children.
Firstborns received higher teacher estimates of conscientiousness and
effort. Glass et al. (1974) propose two related processes to account
for the hi- .r verbal ability and educational aspirations of firstborns.
They su,--st that better-educated parents encourage higher aspirations
and verbal skills in their children, and that this phenomenon is
especially true for earlier-born children:
The later child generates and moves into relationships
uniquely its own--"the principle of unoccupied space"--
specifically different from the role of the "achieving
child", which is already filled in the family. (p. 80)
Burton (1968) found a decreasing progression of IQ with birth
order. However, the.mean difference between firstborn and last born
was approximately 3 IQ points, a difference unlikely to have any
practical effects on the achievement of firstborn relative to that of
last born. He suggests that research on birth order effects should
focus on socio-psychological variables associated with ordinal
In an investigation of family size and sibling spacing effects
(Nuttall, Nuttall, Polit, & Hunter, 1976), there c. -r' ._d a sex-specific
pattern I relationships between family size and academic achievement.
The authors discuss the significant family size effect for boys and
the significant firstborn effect for girls in terms of socio-
psycholo-ical factors. However, because none of the proponents of the
four models emphasized sex differences as a major contributing factor
to family configurational correlates of intelligence, sex was not
included as a predictor variable in these analyses.
a although the reasons for the relationships are not at
present clear, ordinal position at birth appears to be importantly
related to si iificant social factors.
It is therefore evident that birth order and family size influences
on intelligence can be discussed in terms of economic, physiological,
age-spacing and/or socio-psychological variables. It must be emphasized
that these theories are not mutually exclusive. Socioeconomic status
is inextricably related to other variables. For example, influences of
birth order and family size are largely attenuated in higher SES groups.
Prenatal care and diet, as well as family interaction patterns, are
related to social class. In higher SES populations later-born children
and children from larger families do not appear to manifest the handi-
caps generally found in lower SES populations. Thus, it is apparent
that, as one moves up the SES scale, economic, physiological, and age-
gap variables exert less influence and socio-psychological variables
exert relatively more influence on the birth order and family size
correlates of intelligence and achievement.
Fluid and Crystallized Intelligence
In a stuJ of the relationships of Primary Mental Ability Test
scores to birth order, "'rjoribanks et al. (1975) found that verbal and
nu ber 'abilities, but not reasoning and spatial abilities, were
sensitive to ordinal position effects. The authors discuss the results
in terms of the differential influences of the fluid and crystallized
intelli l.L-c, constructs of R. B. Cattell and Horn (Cattell, 1941, 1963,
1967; Horn, 1l;' 1l:C 1970, 1975; Horn & Cattell, 1966, 1967). Horn
(19 ) contends that, during development, a number of factors interact
to ke possible the appropriation of la' : sections of the "collective
intell i ,-... of a culture (represented by crystallized intelligence,
or G ). He c.i',hj;izes the importance of the quality of both school and
home environments in promoting and shaping an individual's crystallized
intelligence. On the other hand, fluid intelliei ice (Gf), or active
problem-solvi ability, is purported to reflect the functioning of the
neurol ical structures and is thus much less influenced by accultura-
A major implication of this theory is that at some order
of analysis in factoring among ability performances two
very broad factors, each having properties of what is
putatively intelligence, should be distinguishable. One
of these factors should be defined primarily by abilities
which can be seen to be quite closely related to intensive
acculturation, whereas the other should be defined primarily
by abilities.which are less closely linked to this. (p. 444)
The greater sensitivity of verbal ability, as compared to active
problem-solving ability, to environmental factors suggests that ordinal
position influences can be best explained in terms of social rather
than physiological factors (Murray, 1971). If Horn's contention is
correct, and if environmental factors within the family are the major
determinant of ordinal position effects on achievement-related
measures, there should be a significant birth order effect on Gc
measures, whereas there should be no significant birth order effect on
Due to the important socio-psychological factors related to
ordinal position, it was predicted that firstborns would score higher
than later borns on achievement-related variables (Gc). Firstborn
superiority was not predicted for Gf measures, which purportedly reflect
neurophysiological functioning. A large battery of tests, hypotheti-
cally related to Gf, Gc, and other closely related cognitive abilities,
were administered to 111 college students and factor analyzed in order
to select optimal marker variables of fluid and crystallized intelli-
gence. Four linear structural equation models representing the economic,
Zajonc and Markus confluence, physiological, and socio-psychological
points of view are illustrated in Figure 1. In addition to comparing
these models in terms of the overall variance accounted for in the
observed results and the predictive ability of the individual
r- ... ... ] ,........ r .-
ECOji! (C /
| (7,Z J' oC ond Ui S4 R ";1
CO NFI.IU ENCE
Uirth orl //
SOCiO-, SY ECOLOGICAL
Linear Structural Equation Models
hypothesized components of each model, certain assumptions concerning
causal order were examined using path analysis procedures. Although path
analysis should not be considered a method for demonstrating causality, it
can be used to examine a set of causal assumptions imposed on a system of
relationships. Different deterministic relationships between the predictor
variables and achievement-related performance were posited by the four
models. Essential to the assumptions of the models was the order of inclu-
sion of the predictors, which implies the existence of certain relationships
among the variables, including direct and indirect influences on the
criterion variable. Moreover, potentially redundant predictors can only
be identified with later inclusion in the model. Multiple regression
procedures are used to estimate the relative strength of the individual
paths. ,These results indicate the magnitude of direct and indirect
(redundant) influences in terms of the hypothesized causal order among
the variables posited for each individual model.
Since Cattell and Horn (Cattell, 1941, 1963, 1V 7; Horn, 1966, C',,
1970, 1975; Horn & Cattell, 1966, 1967) postulate an important
deterministic relationship between Gf and Gc, the Gf variable was entered
first in the three models which included Gf as a predictor variable. The
order of inclusion of the predictor variables and the underlying assumptions
involved will be discussed separately for each model.
Predictors of G in the economic model were included in the following
order: Gf, family size, and SES. Family size was considered to be
i., C' .nt n G. since no neu-'c physiological correlates of family size
were discussed in this model. Due to the attenuation of family size
influences often reported for higher SES levels (Anastasi, 1956; Kennett,
1973; Marjoribanks et al., 1975; Nisbet & Entwistle, 1967; Schooler, 1972,
1973), SES was considered to influence Gc both directly (economic advantages
and restrictions in terms of educational and occupational opportunities)
and indirectly (through differential influences on the family size
correlates of achievement). SES was therefore entered as the third
variable in the multiple regression analysis for this model.
Very different relationships were suggested by Zajonc and Markus in
the confluence model. The proposed handicap of last-born and only children
was the first variable to enter due to its postulated direct influence on
Gc, independent of the other predictors. For example, Zajonc and Markus
(1975) treated "handicap" as a dummy variable--equal to 0 for last-born
and only children or 1 otherwise--in the regression equation used to describe
the Belmont and Marolla (1973) birth order-IQ results. Since neuro-
physiological influences were not discussed by Zajonc and Markus, Gf was
not included as a predictor variable. The second and third variables to
enter into the equation were sibling age spacing and family size, resrvc-
tively. Alt'h Ji, highly confounded, both variables were considered
important mediators of intellectual ability in this model. Since age
spaci was considered a potentially more important correlate of intellectual
abili' it was included before family size. The SES variable was entered
last due to its complex interrelationships with the other variables
postulated by the confluence model.
The first variable entered into the analysis of the physiological
model was Gf, purported to reflect neurophysiological functioning and
integriL '. Since economic factors were discussed in terms of their role
in the mediation of certain physiological influences (including diet,
i-ir.,tal care, and health), SES was entered next. Birth order--discussed
in terms of increased genetic error, "uterine fatigue," and perinatal
complications (Forer, 1976; Warren, 1966; Weller, 1967)--was postulated
to be only a potentially important influence on neurophysiological
functioning (Gf), and was thus the third variable to be considered in
this multiple regression analysis.
The first variable entered into the socio-psychological model was
Gf, as discussed above. Glass et al. (1974) contend that the higher
verbal abilities and educational aspirations of firstborn children may
be particularly prevalent at higher SES levels. The second and third
predictor variables to enter this equation were firstborn (versus later-
born) and SES, respectively. The SES variable was entered last due to its
potentially i:--ortant mediation of birth order influences and possible
direct influence on achievement-related performance through economic
resources and limitations (Forer, 1976).
Thus, the results of the present study should not only indicate the
predictive ability of these four models in terms of Gc performance, but
the evaluation of the individual components of the models may uncover
subtle interrelationships, both direct and indirect, among the predictor
variables themselves and with regard to their prediction of achievement-
related performance. Moreover, the present results should hopefully
clarify certain issues pertaining to the relationships between ordinal
position and aptitude (theoretically represented by Gf measures) versus
achievement-related measures (theoretically represented by Gc ability).
METHOD AND DATA ANALYSIS
Overview and Design
In an effort to evaluate the influence of various family size
variables, birth order variables, and fluid intelligence (Gf) on
crystallized intelligence (Gc), three separate procedures were carried
out. A preliminary factor analysis was performed on a large battery of
tests administered to 111 university students in order to identify optimal
marker variables for Gf and Gc. Following the factor analysis, the most
highly loading tests for each of these two constructs were combined and
utilized as variables in four linear structural equation models relating
G to various hy, othesized determiners, including Gf and several
sibli; I/-i ; ly relationship variables. The models were evaluated separately
for two i : (a) the original college student sample; and (b) a
s..ple consisting of 65 high school students.
Subjects. A sample, consisting of 111 university students, including
seven graduate students, was recruited for test administration by classified
advertisement in the campus newspaper. Since this sample was to be
included as part of a larger grant-supported investigation of cognitive
abilities and aging, each subject was paid $10 for participation. The
sample consisted of 57 male and 54 female white, middle and upper-middle
SES students. The age range of the subjects was 17 to 30 years, with a
mean age of 20 years.
Instruments. A battery of 21 tests which were hypothesized to
represent Gf, Gc, and other closely related cognitive factors were
administered to the subjects by the author. The group test
administration was approximately four hours in length, with two 10-
minute rest intervals. The tests, in their order of presentation,
included the Gf-Gc "Sampler," an unpublished battery of tests constructed
by Horn which consists of three tests designed to represent Gf and two
tests designed to represent Gc the Army Alpha Examination (First
Nebraska Edition, Guilford, 1938), four subtests of the Primary
Mental Abilities (PMA) battery (Thurstone & Thurstone, 1949), three
Educational Testing Service (ETS) measures of French, Ekstrom, and
Price (1963), and the Omelet Test from the Structure-of-Intellect
measures ('"cpfner & Guilford, 1969). A description of each measure
1. Vocabulary (Horn "Sampler"). Subjects are asked to select the word
which is closest in meaning to a given word from five test words. There
are 20 items in this subtest. This test, like all of Horn's test in
the present investigation, was untimed.
2. Analogies (Horn "Sampler"). This test consists of fairly simple
verbal analogies in the form: "A is to B as C is to ." In each of
15 items .the subject is given five test words from which to choose the
appropriate answer. Since the words are relatively simple, perceiving
the relationship between them is theoretically what is evaluated, which
should reflect Gf ability.
3. High-Level Vocabulary Analogies (Horn "Sampler"). This test involves
exactly the same format as Horn's Analogies Test. However, this test
consists of more difficult verbal analogies in the sense that it
contains higher-level vocabulary words designed to involve more advanced
verbal ability, thus reflecting Gc ability.
4. Letter Series (Horn "Sampler"). After examining a sequence of from
5 to 15 letters, the subject is required to write the next letter in the
series in a blank space which follows the sequence (15 items).
5. Fi .:..-l Relations (Horn "Sampler"). Subjects are presented with a
square form with geometric, simple figural, and letter patterns in 3 out
of 4 grants, and, on the basis of the interrelationships of these
fi 'u, must select from several choices the appropriate figure for the
fourth quadrant (20 items).
6. Following Directions (Army Alpha). Subjects must carry out verbally
presented instructions on items consisting of geoi.etric and letter
patterns. Short-term memory ability is required for the rapid execution
of these 12 items.
7. Arithmetic Problems (Army Alpha). This test consists of verbal prob-
lems which require arithmetic manipulations.
8. Common Sense (Army Alpha). This test involves questions designed to
evaluate practical judgment and reading comprehension.
9. .. -Ant.. ., (Army Alpha). Subjects must designate whether 40
sets of word pairs are the same or opposite in meaning.
10. Disarranged Sentences (Army Alpha). After mentally re-arranging
a set of words in order to make a sentence, subjects are instructed to
indicate whether each of the resulting 24 sentences is true or false.
11. Number Series Completion (Army Alpha). Subjects must examine the
relationship which exists within a series of numbers and indicate what
two numbers should logically follow (for each of 20 number series).
12. Analogies (Army Alpha). After determining the relationship between
a pair of words, subjects must select one of a group of four words which
is related to a second word in an analogous manner.
13. Information (Army Alpha). Subjects must answer 40 multiple-choice
items which involve general information.
14. Word Grouping Test (PMA). In this test the ability to understand
and group words of the same type is involved. Subjects must select
which of five presented words does not belong with the other words.
Each of.the PMA subtests is timed.
15. Letter Series Test (PMA). This test consists of 20 letter series,
each of which has a specific pattern. The task is to select which of
five letters should logically follow for each individual letter sequence.
16. Spatial Relations Test (PMA). Subjects are presented with 30 geo-
metric forms, to the right of which are five similar or identical forms
which have been rotated to different positions. The task is to identify
which of these forms are identical to the original form.
17. Number Series Test (PMA). This test consists of 20 multiple-choice
items in which the task is to select the number which should come next
in a distinctive pattern of numbers.
18. Addition Test, N-l (ETS). This two-part test assesses how quickly
and accurately subjects can add series of numbers. The time limit is
two minutes for each 60-problem part.
19. ;:.,n;er Comparison Test, P-2 (ETS). This is a highly speeded test
consisting of two parts. The task is to compare two series of numbers
(the pairs of number sequences are of different lengths) and indicate
whether or not they are the same by inserting an X between the number
series that are not identical.
20. Identical Pictures Test, P-3 (ETS). This is a two-part test in
which subjects must select from five alternative pictures the one that
matches a target picture.
21. Omelet Test (Hoepfner & Guilford, 1969). In this test the task
is to rearranr,_ sets of four letters each to make familiar words. This
test was constructed to measure Guilford's Structure-of-Intellect factor
of convergent production, defined as the "generation of logical conclusions
from given information where emphasis is upon achieving unique or
conventionally best outcomes" (Guilford & Hoepfner, 1971, p. 20).
Variables. The variables included for analysis were determined by
the assumptions of the four models, which are illustrated in Figure 1.
The G construct was the criterion variable for each model. The predictor
variables included Gf, SES, birth order, family size, firstborn (as opposed
to later-born siblings), age spacing between siblings, and the "handicap"
of the only child and last-born child (Zajonc & Markus, 1975). The SES
scale utilized in this study was a composite of three separate standardized
variables: (a) prestirn value of father's occupation (Treiman, 1977);
(b) father's educational level; and (c) mother's educational level.
Educational level was rated along a 6-point scale, from "not a high
school graduate" to "doctorate." The number of children in the
families in this sample ranged front 1 to 13, with a mean family size of
3.432. There were 40 firstborns (including four only children).
Birtn order ranged from 1 to 11 (mean = 2.234). In order to normalize
the extremely skewed birth order distribution, birth order was divided
into three categories before any further analysis took place. These
categories were: (a) firstborn (n = 40); (b) second- or third-born
(n = 56); and (c) fourth- or later-born (n = 15). The adjusted percent-
ages of the total sample for these three categories were 36%, 50"', and
14%, respectively. The sample included 36 middle children and 35
lastborn children. Due to certain assumptions specific to the different
models, two dichotomizations of birth order frequencies were calculated.
The firstborn (versus later-born) construct was a predictor variable
in the socio-psychological model. The "last-born/only-child handicap"
was included as a variable in the Zajonc and Markus confluence model.
This category included 39 subjects. In addition, this model proposed
that spacii.' between siblings would be an important correlate of
intellectual development. Age spacing was calculated as the sum of two
separate indices, the number of years to the next oldest sibling and the
number of years to the next youngest sibling. Since this model proposed
that an arbitrary maximum intellectual level of 100 would be attained at
approximately 19 *errs of age, 19 years was used as the maximum age
separation index. For example, the age spacing index for an only child
would be (19 + 19) = 38. This somewhat crude index of age spacing
between siblings was utilized die to the extreme difficulty encountered
when attempting to calculate each individual subject's intellectual
level according to the sigmoid function presented by Zajonc and Markus
(1975). This equation was analyzed in terms of average age gaps, not
individual age separation data. Other approaches to this age gap index
have included mean age separation between siblings (McCutcheon, 1977)
and median number of months separating the siblings in the entire
family (Nuttall et al., 1976).
The means and standard deviations of the variables for the college
student sample are listed in Table 1.
Factor Analysis. A 21-by-21 variable correlation matrix was
computed from the scores of the 111 subjects on the 21 measures. The
factor analysis was carried out with the Exploratory Factor Analysis
Program (EFAP), devised by Joreskog and Sorbom (1976a). Rotation was by
the Promax method, utilizing the Varimax orthogonal rotation procedure
as the basis for the subsequent oblique solution.
Kimum likelihood procedures (Joreskog & Sorbom, 1976b) of factor
extraction were utilized to obtain 4- through 10-factor solutions. The
most easily interpretable solution appeared to be the 4-factor solution.
The factor loadings and factor correlation matrix are found in Table 2.
The first factor seemed to be a numerical facility factor which had
substantial loadings on the two number series tests, the Addition Test,
Arithmetic Problems, and also the Omelet Test. Inclusion of the Omelet
Test su--e.,ts the possibility that convergent production may be
involved in the first factor. Factor II appeared to be a verbal ability
factor, hypothetically representing Gc. Tests which had interpretable
loadi : on this factor included Vocabulary, High-Level Vocabulary
.r.,,l ies, Syn.' .- -Antonym, Information, Common Sense, Disarranged
Sentences, Com on ,.~,logies, and the 'Word Grouping Test. The loading
of Horn's Common Analogies subtest on this factor was somewhat
surprising since this measure was constructed as a marker of Gf ability.
However, the low correlations of this test with other nonverbal measures
Neans and Standard Deviations for College Student Sample
Mean Standard Deviation
Age 20.2160 1.9970
Family Size 3.4324 1.6713
Birth Order 2.2340 1.4270
Mother's Education 2.8200 1.1380
Father's Education 3.3578 1.3846
Occupational Prestige 50.4364 13.1961
Age-Spacing Index 17.9550 8.8360
Vocabularya 10.8830 2.0960
Vocabulary .ilogiesa 10.6220 2.2360
Common Analogiesa 11.6577 1.8360
Letter Seriesa 12.4230 1.8660
Figural -lationsa 14.6130 2.9300
Follow, Directionsb 9.4050 1.7960
Arithmetic Problemsb 11.9910 2.8170
Common Senseb 11.5500 2.4600
Synonym-Antony.,b. 30.8010 5.1610
Disarr.in ed Sentencesb 20.2700 2.6250
Number Seriesb 14.6670 3.3070
Analogiesb 34.5860 4.6720
Information 29.3330 3.2680
Table 1 continued
Mean Standard Deviation
Word Groupingc 23.3330 3.1400
Letter Seriesc 15.1530 2.7640
Spatial Relationsc 120.4230 13.1670
Number Seriesc 11.5860 2.8870
Addition 43.1980 13.2110
Number Comparisond 26.1350 5.9980
Identical Picturesd 79.4180 11.4440
Omelet Teste 21.5590 4.8570
uArmy Alpha Examination
cPrimary Mental Abilities Test
dEducational Testing Services
eSheridan Psychological Services
Promax Rotated Factor Loadings
Numerical G Speed G
Facility c f
1 2 3 4
Common Analogiesa 0.006 0.465 -0.132 0.178
Analogiesb -0.085 0.693 0.133 0.119
Word Groupingc 0.021 0.701 -0.179 0.175
Letter Seriesa 0.160 0.115 -0.046 0.565
Letter Seriesc 0.157 0.043 0.118 0.630
Figural Relationsa 0.063 0.008 -0.019 0.530
Spatial Relationsc -0.031 -0.135 0.058 0.716
Number Seriesb 0.730 -0.109 -0.002 0.357
Number Seriesc 0.741 -0.013 0.001 0.138
Following Directionsb 0.074 0.230 0.067 0.312
Vocabularya -0.263 0.792 0.026 0.076
Vocabulary Analogiesa 0.071 0.663 -0.091 0.022
Synonym-Antonymb -0.053 0.734 0.268 -0.239
Informationb 0.150 0.779 -0.213 -0.068
Arithmetic Problemsb 0.387 0.209 -0.018 0.300
Additiond 0.498 -0.092 0.504 0.055
Number Comparisond 0.065 -0.057 0.756 0.088
Facility c Speed Gf
1 2 3 4
Identical Pictures -0.122 0.094 0.287 0.290
Common Senseb 0.020 0.648 0.225 -0.076
Disarranged Sentences 0.137 0.470 0.074 -0.022
Omelet Teste 0.448 0.132 0.227 -0.055
A,-i. Alpha Examination
CPrimary Mental Abilities Test
educational Testing Services
hheridan ;ychological Services
Table 2 -
may have been a result of the restriction of range in this particular
sample. It would be informative to analyze the intercorrelations of these
tests in a more heterogeneous sample. Factor III was a highly speeded
factor, which had its highest loading on the Number Comparison Test.
Factor IV, hypothesized to represent Gf, had interpretable loadings on
the two letter series tests, Figural Relations, the Spatial Relations Test,
,Number Series Completion, and the Following Directions Test. Factor
interrelationships were moderately oblique ranging from .263 to .453.
After interpreting the factor loading matrix, it was decided that
the five best marker variables of the Gf construct were: (a) the PMA
Letter Series Test; (b) the PMA Spatial Relations Test; (c) the Army Alpha
Number Series Completion Test; (d) Horn's Letter Series subtest; and (e)
Horn's Fi'gural Relations subtest. Since the two letter series tests were
measures of the same ability, it was decided to combine them as a single
measure in the Gf composite. The measures selected to represent Gc were:
(a) the PMA Word Grouping Test; (b) the Army Alpha Synonym-Antonym subtest;
(c) the .nrmn Alpha Information subtest; (d) Horn's Vocabulary subtest; and
(e) Horn's High-Level Vocabulary Analogies subtest. The marker variables
for both composites were individually standardized before being added
together. As has been found in the past (Horn, 1970, 1975) and is expected
theoretically, the Gf and Gc composite scores were substantially positively
intercorrelated, r = + .3674, p <.01.
Subjects. The sample consisted of the same 111 university students
examined in the previous study.
Method. A multiple regression procedure was employed to examine
the results in terms of both the overall variance accounted for in the
observed data and the hypothesized causal orderings for the individual
components for each of the models described in Chapter 1. Multiple
regression was the method of choice since it is a statistical technique
which permits analysis of the relationship between a criterion variable
and a set of predictor variables. A hierarchical specified inclusion
method was used. The criterion variable for each model was Gc. This
procedure permits the addition of variables to the regression equation
in a predetermined order. The estimation of the relative strengths of
the individual paths indicates the magnitude of direct and indirect
influences in terms of the hypothesized causal order posited for each
model. The order of inclusion of the predictor variables was determined
a priori by the author with regard to the hypothesized influences of
family configuration variables and SES on intellectual aptitude and
achievement as specified by the individual models. In the economic
model, such influences are discussed in terms of SES and the related
economic advantages and restrictions. Physiological mediators of birth
order differences in cognitive ability are considered in the physio-
logical model. In the confluence model, ordinal differences in intel-
lectual ability are discussed in terms of family size and age spacing
between siblings. The relationship of social and psychological influences
and birth order differences are emphasized in the socio-psychological
model. These models, which are illustrated in Figure 1, are presented
in more detail in Chapter 1.
In the hierarchical decomposition procedure, each predictor variable
is added to the equation in a single step, and the increment in R2, or the
explained variance, at each step is considered to be the component of
variation attributable to that particular variable, given the contri-
bution of the other variables to the equation. The first predictor is
evaluated by the F ratio:
(1 R2-12,....,k)/(N k 1)
.where r 1 is the incremental sum of squares due to the independent
variable X., 1 R2-12,....,k is the sum of squares of the residuals,
N is the sample size, and k is the number of independent or predictor
variables. Each successive predictor variable is tested by an F ratio
with 1 and (N k 1) degrees of freedom. This procedure indicates
the total influence of each successive variable since it refers to
adjust 6nts only for those variables that precede each given variable
in a predetermined order of inclusion.
The F ratios were calculated for each model and for the inclusion of
each predictor variable within each model. In addition, the models were
compared to each other for overall variance accounted for in terms of
tne obtained data. The computer methodology used to evaluate these
results was outlined in the Statistical Package for the Social Sciences
(SPSS), edited by Nie, Hull, Jenkins, Steinbrenner, and Bent (1975).
Results. Since Gc ability may be much more sensitive to accultura-
tional and environmental influences than Gf ability, which purportedly
reflects neu:-rp', biological functioning, it was hypothesized that there
would be a si nificant birth order influence on the Gc construct, but not
on the Gf construct. As predicted, while the correlation between birth
order and Gc was significant (r = -.1633, p_<.05), the correlation
between birth order and Gf was not significant (r = .0710). Since no
significant birth order patterns were found for the Gf construct, it
appears that birth order differences are not mediated by neurophysio-
logical factors in the present sample.
However, the prediction that, due to certain socio-psychological
influences related to ordinal position, particularly in the restricted
SES range of the present sample, that firstborns would obtain higher
scores than later borns on Gc but not on Gf measures was not upheld.
While the correlations of the firstborn (versus later-born) variable
on the Gc and Gf constructs were in the predicted direction
(r = .1398 and r = .0393, respectively), neither correlation was
.iults of the multiple regression analyses are listed in Table 3.
Althiu .j three of the models--economic, physiological, and socio-
psychological--obtained significant overall F ratios (p<.01), it is
evident that these results were attributable to the highly significant
F ratio obtained in the first step in the equations, i.e., the regression
of Gf on Gc. lone of the other predicted paths between the predictor
variables and Gc were significant for any of the models. Regarding the
Zajonc and M.arkus confluence model, neither the overall F ratio nor any
of the predicted paths were significant. The overall variance explained
by the economic, confluence, physiological, and socio-psychological
Sodels was i roximately 1 C, C 17', and 17 respectively. Thus,
alti the economic model appeared to be the most predictive model in
terms of variance accounted for, the explained variance was not large
in absolute terms.
E o n
4- ro L W
C Ll. V)
- CMj C)
S co >,
c" c) E
4- LU *
CL0 VC Co
C-- J M0
_0 U0 0
4- C-J C'
L- - n
> 0 *
Ll ) *r-
SC LO -
Therefore, upon initial investigation, it appeared that none of
the four models could effectively account for Gc in the present sample.
The possibility was considered that these negative results were largely
attributable to the restriction in range of the present university
student population. Upon closer analysis, however, certain unexpected
relationships within this particular sample were uncovered. It became
apparent that, besides restriction in range, certain relationships
related to SES and family size which might be specific to a college
sample were statistically suppressing certain predicted family configur-
ational influences. Specifically, these relationships appear to have
suppressed the correlations between several of the predictor variables
and the criterion variable G It was decided to use partial correla-
tion procedures to separate out some of these relationships. Since
partial correlation statistically controls for the effects of one or
more specified variables on the relationship between other variables,
this technique can be utilized to uncover suppressor relationships. The
general form of suppressor influences can be described as: A shows no
relationship to B because A is negatively related to C which is positively
related to B. Thus, when the effects of C are controlled, A may be
positively related to B. Further assumptions and issues concerning such
suppressor relationships will be discussed in further detail after dis-
cussion of the unexpected SES-family size relationships found in the
college student sample.
SES-family size relationships which may be specific to university
student samples. Several unexpected relationships among the predictor
variables in the present sample may have suppressed their relationship
to G These results rendered straightforward interpretation of
the models difficult. It is conceivable that certain possibly artifact-
ual relationships may be specific to college student samples. Such
relationships may complicate attempts to generalize from more hetero-
geneous populations regarding family configurational influences. Schooler's
(1973) contention that birth order effects do not appear to persist into
adulthood may be particularly apparent in a college sample. Firstborn
overrepresentation appears to be the stablest finding reported in
studies of university students (Altus, 1966; Capra & Dittes, 1962;
Schachter, 1963). The finding that 36% of the current sample were
firstborns was slightly, although not significantly, higher
(Z = 1.''".3, p(<.10) than the expected 30 firstborn representation
for this particular age group (Dunn, 1956).
r -- ..... i- effect Kennett (1973) discusses the changing rela-
tionship between family size influences on intellectual ability. He
suggests that the acceptance of a negative correlation between family
size and intelligence must be qualified by the possibility that SES may
influence the relationship. Thus, the correlations may vary from social
class to social class. In this r-c 'rd, perhaps the most surprising find-
ing in the -e- nt college sample was the significant positive corre-
lation between family size and SES (r = +.1972, p <.02). This finding
was in marked contrast to the widely reported negative correlation
between these two variables in the general population (Belmont & Marolla,
1973; Eysenck & Cookson, 1970; Murray, 1971; Nisbet & Entwistle, 1967;
Record et al., 1`69). The observation that students in this sample who
belc,., to larger families also tend to be of a higher SES level suggests
that financial resources and related issues may play an important and
perhaps decisive role in the opportunity to attend college. This posi-
tive relationship between family size and SES might be expected to be
even more substantial in a private, more expensive institution.
Birth order effects. Family size and birth order are highly con-
founded variables, as evidenced by their substantial intercorrelation
of +.4775 (p_, .001). While the economic model and the confluence model
discuss family configurational influences primarily in terms of family
size effects, the physiological model posits such effects in terms of
sibling birth order while the socio-psychological model emphasizes the
influence of firstborn versus later-born on G c
Family size and, therefore, birth order are negatively correlated
with SES in the general population. However, in the college student
sample who participated in this study, there existed a positive although
nonsignificant correlation between these two variables (r = +.1200, p< .10).
Despite the possibility that this positive intercorrelation may have
suppressed to some extent the negative correlation between birth order
and Gc, as was predicted by the physiological model, the obtained corre-
lation between birth order and G was nevertheless significant
(r = -.1633, p <.05).
Statistical Su ,pression
Cohen and Cohen (1975) discuss the effects of three types of
suppressor relationship--classical, net, and cooperative suppression--
on the patterns of association between the predictor variables. In the
present sample, the positive correlation between family size and SES
along with the resulting effects on the criterion variable, Gc is an
example of cooperative suppression. Cohen and Cohen describe cooperative
suppression as the case in which the intercorrelation of two independent
variables involves a portion of their variance which is irrelevant to
the criterion variable. The partialling of these predictor variables
from the others enhances all indices of relationship with the criterion
variable. It is a completely symmetrical phenonemon; whenever it is
determined that X2 suppresses X it is also evident that X suppresses
X2. One indication of cooperative suppression is a situation in which
a beta coefficient exceeds its simple r with the dependent variable and
is of the same sign. With reference to Table 3, it can be seen that
beta coefficients for both SES and family size in the economic model
meet this criterion.
Analysis of suppression effects was performed to ascertain what,
if any, relationships might exist in the present sample which could have
resulted in the suppression of predicted family confirJration effects on
Gc. In order to identify the nature of the specific suppressor relation-
ships between the predictor and criterion variables, the partial corre-
lation coefficients were calculated for several predictor variables and
Gc, with the potential suppressor effects of the predictor variables
statistically controlled. The obtained multiple regression and partial
correlation coefficients are found in Figure 2. Results for the partial
correlation procedure will be discussed separately for each individual
Economic model. With reference to the economic model, it was evi-
dent that the unexpected positive correlation between family size and
SES i -ressed the predicted significant correlations of both family
size and SES with the Gc construct. When the effect of family size was
partialled from the SES-Gc relationship, the correlation between these
tw.o variables was significant (+.1683, p <.05). Similarly, when the
Intercorrelations among Variables in each Model
for the College Student Sample.
Numbers in parentheses indicate partial correlations between
the independent and criterion variables.
Significance levels: <.05, **p<.01.
Sj S.!3)0 >- [.,
[-. C ", I C; 91 -. 7-;8 .3674*;
/ L I. !
.3972 .13. 0
CO( FLULrC : I C"
PIiSOlGQ YCHOLAL -. 071OGCL .1 ) .7
--- 0.- --- 1 i ( rs o
.1340 ..] L
suppressor effects of the significant positive correlation between
family size and SES were controlled, a modified version of the economic
model was supported by the present results. The hypothesized relation-
ships between the predictor variables and the criterion variable were
significant when the suppressor effects were controlled. This model
was termed "modified" because of the significant positive, not nega-
tive, correlation between family size and SES. In a more heterogen-
eous sample, it would appear likely that the unmodified version of the
economic model would be supported due to the generally reported nega-
tive correlation between family size and SES (Belmont & Marolla, 1973;
Eysenck & Cookson, 1970; Murray, 1971; Nisbet & Entwistle, 1967;
Record et al., 1969).
Zajonc-Markus confluence model. Since the overall F ratio was
non-significant, it was considered inappropriate to further analyze
specific predicted relationships of this model. It is quite possible
that the unimpressive predictive power of this model may have been
attributable to the restricted range of the present sample and/or the
unexpected relationships which may be specific to a college student
population. Such unexpected trends might suppress the relationships
between SES and age spacing between siblings and, in turn, between age
spacing and G Similarly, in a study of ability of 291 community
college students, as measured by Factor B of the 16 PF, McCutcheon
(1977) found little support for the confluence model. She concluded
that this model may have limited application. The overall validity of
this model might be expected to be substantially higher in a more heter-
Physiological model. The generally observed negative correlation
between birth order and SES, an important component of this model, was
not obtained in the college student sample. However, due to this
model's predictions that birth order and SES influences would be mediated
through Gf, which reflects neurophysiological functioning, further
analyses for suppressor relationships were precluded. Since Gf ability
was for all practical purposes unrelated to both birth order
(r = .0710) and SES (r = .0404), any potential suppression effects
among these variables would have been statistically trivial in this par-
Socio-psychological model. The possibility of cooperative suppres-
sion between the firstborn construct and SES was considered. However,
when t1e relationship between the firstborn predictor variable and Gc was
par-ialled for SES effects, the resulting firstborn-Gc intercorrelation
(r = +.1423, L1.05) failed to reach significance. Similarly, although
the partial lig of the firstborn construct from the SES-Gc correlation
sli i.tly i --roved the predictability of G from SES, the correlation
(r = +.1366, p>.05) was still nonsignificant.
Although the overall F ratios for three of the four models were
significant, when the individual models were analyzed in terms of their
separate paths, they were not supported. However, when the existence
of certain statistical suppressor effects was recognized and these effects
were controlled by partial correlational techniques, a modified version
of the economic model was supported.
Subjects. Due to certain relationships found in the previous
analyses which appeared to be specific to a university student popula-
tion, a more heterogeneous sample was utilized in the attempt to evaluate
family configurational and SES correlates of the Gc construct. This
sample consisted of 65 twelfth grade students from the P. K. Yonge
Laboratory School of the University of Florida. The student population
of this school is stratified so as to be demographically representative
of the general population of the state of Florida in terms of economic
level and race. This tested sample consisted of 55 white and 10 black
students, whose age ranged from 17 to 18 years (mean age = 17.37 years).
There were 35 females, 30 males, including 14 firstborns, 2 only child-
ren, and 25 last-born children.
Instruments. Six tests were administered to these subjects. The
group administration was approximately one hour in length. The tests
included three markers of Gf (PMA Spatial Relations, PMA Letter Series,
and the Army Alpha Number Series subtests) and three markers of Gc
(Horn's Vocabulary subtest, Horn's High-Level Vocabulary Analogies sub-
test, and the ETS Wide-Range Vocabulary Test). Although a somewhat
abbreviated battery was utilized in the third study, the tests were
optimal markers of Gf and Gc ability in the initial factor analysis.
They were thus considered equivalent theoretically and practically to
the longer battery administered to the college student sample. The ETS
Wide-Range Vocabulary Test was included because of its stabilized relia-
bility and its particular appropriateness for a high school level
Variables. As in Procedure 2, the variables analyzed were
determined a priori by the assumptions of the individual models found
in Figure 1 and described in Chapter 1. The variables were fairly
normally distributed, with the exception of father's educational level
(fathers of 25 subjects had obtained doctorates) and the occupational
prestige index (Treiman, 1977). Both variables were somewhat skewed,
with relatively more subjects at the higher levels. These two variables,
along with mother's educational level, were standardized before they
added together for the SES index. Similarly, the individual tests were
standardized before they were summed for the Gf and Gc composites.
Family size in this sample ranged from 1 to 10, with a mean family
size of 3.6 children. Birth order also ranged from 1 to 10 (mean =
2.4?1): The age spacing index for the confluence model was calculated
in the same manner as in Study 2, that is, the sum of the number of years
to the next oldest sibling and the number of years to the next youngest
sibling. Also, the two birth order dichotomizations, firstborn (versus
later-born) and the "last-born/only child handicap," were included as
predictor variables exactly as they were for the college student sample.
The means and standard deviations of the variables studied in the
high school sample are listed in Table 4. The intercorrelations between
the variables for the individual models are found in Figure 3.
method. As in Procedure 2, multiple regression procedures were used
to examine the amount of variance accounted for in terms of the observed
data and the direct and indirect influences of the hypothesized paths
in terms of Gc with regard to the economic, physiological, Zajonc and
Markus confluence, and socio-psychological models. Since the
Means and Standard Deviations for High School Sample
Mean Standard Deviation
Age 17.3692 0.4862
Family Size 3.6000 1.6845
Birth Order 2.4963 1.5220
Mother's Education 3.6615 1.3610
Father's Education 4.2308 1.7657
Occupational Prestige 58.0000 17.5944
Age-Spacing Index 15.9219 8.5027
Vocabularya 10.3077 2.1645
Vocabulary Analogiesa 10.3231 2.3326
Wide-F.Pie, Vocabularyb 14.1231 3.3237
Letter Seriesc 13.6462 3.4705
Spatial Relationsc 121.0615 13.6516
Number Seriesd 12.3538 2.9709
Educational Testing Services
cPrimary Mental Abilities Test
dArmy Alpha Examination
Intercorrelations among Variables in each Model
for the High School Student Sample.
Significance levels: *p <.05, ** <.01.
I ..'': ... . rV !:,C ..... . C .
. ,0C -.413 -.210 .5 9
( Z"JO .C ,c: ..,A,.i,KUS) /
*,r i Cap <*
P- 1-!Y 10 -0 GICAL525
0 1 S.27CIC, ) 1C ., 13 C|3
-v. *y \ 1 ---"- -
-5C92 .1 75
PH YSI LOGICAL .O~,6 ,)
[nirth orc.cL .3380- G:, ]
S 0 C I O- S Y C" L O --, 1 ( .L ,136 30 o
..S .2cl --- I -orn
assumptions of the individual models remained the same, the hierarchical
specified inclusion method with G as the criterion variable which was
described in Study 2 was used. F ratios were calculated for each
model and for the successive inclusion of each predictor variable within
each model. The individual models were also compared to each other for
the overall variance accounted for in terms of the obtained data.
Results. The results of the multiple regression analyses are
listed in Table 5. It is apparent that all of these models yielded
better predictions of Gc performance in the more heterogeneous high
school sample examined in this procedure. Not only were the overall
F ratios for all of the models significant (p <.01), but the overall
variance explained by the economic, physiological, Zajonc-Markus, and
soci( ychol- ical models was Epr: -,.imately 51", 52';, 40%, and 54",
res-_. ively. However, while the variance accounted for by each model
was substantial, the regressions of all of the individual predictor
variables on Gc were significant only for the economic and socio-
psychological models. Results for the multiple regression analyses
will be discussed separately for each individual model.
Economic model. In contrast to the unexpected positive correlation
between family size and SES (r = +.1972, p_<.02) found in the more homo-
geneous colle-- sample, the predicted negative correlation between these
two variables was found for the high school population (r = -.2104, p_<.05).
This .. ,tive correlation has been widely reported in other studies
(Beli-;ont & Marolla, 1973; Eysenck & Cookson, 1970; Murray, 1971; Nisbet
& Entwistle, 1-.7; Record et al., 1969) and is an essential component of
the economic model, in which it is proposed that birth order and family
size influences are largely reducible to SES and the economic factors
S LMn mL .D 0 m c\j .D h- LO
Q 0 CO I- Lr ) 't oz o
S[ C m)D L) Cj m C D G
U -.- C0 O CD (\ c 'O
C/ C) O d) 1 1 O O,
c- C 0 0m L- L C 0n D
C LO LLO C C -
D O ,-
D C CJ Or r- CD CC OM
S. o 04 00 r- r O0 o co ( -. -
c-- i- M 0 LO Oh LO LO
CD LO C CD LO r-\- O'-
C 4-- m 0) LO LfCD O ) C m
C- CD CD- CD CD0
C) 0 (A C/ .
4- o o C oo -) O o "ow"
S. -- C0 i 0 0 0 .
I( I -m l l l ^1 r- r
0 0 0
O U 0 >,
-- I .-
- c\j n
4- r LU
S.-c : )
- e C)
involved (Anastasi, 1956; Kennett, 1973; Schooler, 1972, 1973).
Moreover, the predicted significant increments to the explained sum
squares attributed to each successive predictor variable (Gf, family
size, and SES) in the hierarchical decomposition procedure were statis-
tically supported. It was concluded that, in addition to the ability
to account for a very substantial amount of variance in the obtained
data, an unmodified version of the economic model was fully supported
in the more heterogeneous high school sample.
Zajonc-Markus confluence model. In contrast to the findings with
regard to this model (Zajonc, 1976; Zajonc & Markus, 1975) in the college
student sample, the overall significant F ratio for the high school sample
permitted the further analysis of the individual predicted relationships
of the confluence model. After examining these results, however, it
became evident that, while the age-spacing index, family size, and SES
were adequate predictors of the criterion variable, the "handicap" of
lastborn and only children (ostensibly due to their lack of opportunity to
serve as teachers to younger siblings) was not an adequate predictor of
Gc performance. Therefore, the confluence model was not fully supported
by the obtained data.
Physiological model. In the physiological model (Warren, 1966; Weller,
1965), emphasis is placed on physiological mediators of birth order
differences in cognitive ability, such as mother's age, length of labor,
and perinatal complications. Again, an overall F ratio and a substantial
amount of the variance in Gc performance accounted for permitted
further analysis of the individual components of this model. The first
two predictor variables in the multiple regression equation, Gf and SES,
were significantly predictive of G scores. However, birth order, whose
correlation of -.1752 (p >.05) with Gc failed to attain significance in
this sample, also failed to add significantly to the explained sum of
squares for the physiological model. Thus, it was concluded that this
model was not supported by the data.
Socio-psychological model. Proponents of this model (Bradley,
1968; Burton, 1968; Glass et al., 1974; Murray, 1971; Nuttall et al.,
1976; Start & Start, 1974) discuss family configuration influences in
terms of significant social and psychological factors. The variance
accounted for in terms of the data from the high school sample was 54
for this model, the most substantial of all the models. Also, after
analyzing the individual paths, it was evident that the predictor
variables, Gf, firstborn (versus later-born), and SES, were each sig-
nificant predictors of the criterion variable, in terms of both simple
correlations and individual paths.
Summary. After analyzing the economic, confluence, physiological,
and socio-psy,'. .logical models in terms of both total variance accounted
for and in terms of the individual paths, the models which were fully
supported were the economic and socio-psychological models, which
explained 51. and 54 of the variance, respectively. Although both
models were supported, the question as to which model was the optimal
predictor of Gc performance in the heterogeneous high school sample had
to be considered. The overall F ratios for both models were highly
significant (- <.01). Also, highly significant F ratios (p <.01) were
obtained for both Gf and SES, predictor variables in both models. How-
ever, when family size (the second variable to enter the regression
equation of the economic model) was compared to firstborn (the second
variable of the socio-psychological model), it was found that the F
ratio for the firstborn variable was 12.2787 (df = 1/62, p<.01), more
significant than the F ratio for family size (F = 6.4335, df = 1/62,
p_<.05). Thus, when the F ratios for the individual hypothesized paths
were evaluated, it was evident that the predictor variables of the socio-
psychological model were more strongly related to Gc than were the
individual components of the economic model. It was concluded that--
although both the economic and socio-psychological models were fully
supported by the data--the socio-psychological model was the optimal
predictor of achievement-related performance in this sample, in terms of
both overall variance accounted for and with regard to the significance
levels of the individual hypothesized paths.
Although sibling constellation correlates of intellectual
abilities have been studied extensively, the results have often
been difficult to interpret. The contradictory findings and tentative
conclusions are partially attributable to the highly confounded
nature of birth order and family size variables. Interpretations
are further complicated by the relationships between ability,
achievement, and personality and the intercorrelations of SES with
both family size and the personality-ability-achievement complex
(Eysenck & Cookson, 1970).
In the present study, an attempt was made to separate out some
of the confounded relationships between family configuration, SES,
and the performance on ability and achievement measures. In the
initial procedure, a battery of 21 tests was administered to 111
university students. After factor analyzing the resulting correlation
matrix for optimal marker variables of the fluid intelligence (Gf or
active problem-solving ability) and crystallized intelligence (Gc or
achievement-related performance) constructs of Cattell and Horn
(Cattell, 1941, 1963, 1967; Horn, 1966, 1968, 1970, 1975; Horn &
Cattell, 1966, 1967), four linear structural equation models
relating Gc to various hypothesized predictor variables were
evaluated for two separate samples, the original university student
sample and a more heterogeneous sample of high school students.
The four models emphasized the importance of either economic,
"intellectual environment" (confluence), physiological, or socio-
psychological influences in the mediation of birth order and/or
family size effects on intellectual ability. Proponents of the
economic model (Anastasi, 1956; Kennett, 1973; Nisbet & Entwistle,
1967; Schooler, 1972, 1973) contend that family size effects are
largely reducible to SES and related economic advantages and
restrictions. The confluence model (Zajonc, 1976; Zajonc & Markus,
1975) considers ordinal differences in intelligence in terms of
family size and age spacing between siblings and the resultant
influences of the family's immediate "intellectual environment" on
intellectual development. In the physiological model (Warren, 1966;
Weller, 1965), it is suggested that physiological factors are
important mediators of birth order differences in cognitive ability.
Advocates of the socio-psychological model (Bradley, 1968; Burton,
1968; Glass et al., 1974; Nuttall et al., 1976; Start & Start, 1974)
contend that birth order, particularly firstborn academic superiority,
appears to be importantly related to significant social and psychological
factors. Such influences are often discussed in terms of personality
and motivational differences which may be manifested in individuals
of differing family sizes and ordinal positions. Issues and
assumptions relevant to these models were discussed in the first
chapter. It was emphasized that these theories are not mutually
exclusive. It was suggested that, as SES level increases, physiological,
"intellectual environment," and economic variables might conceivably
exert less influence, and that socio-psychological variables might
exert relatively more influence on family configurational effects.
University Student Sample
It was predicted that the socio-psychological model would be the
best predictor of achievement-related performance, or Gc, in this
sample. In the second procedure, multiple regression linear
structural equation techniques were utilized to evaluate the amount
of variance accounted for in terms of the observed data and for the
individual hypothesized paths for each of the models in a college
student'sample. The overall significant F ratios for the economic,
physiological, and socio-psychological models were found to be
attributable to the significant positive regression of Gf on Gc
(r = +.3674, p <.01), the variable entered at the first step in
each of these models. The overall F ratio of 2.2989 for the confluence
model was not significant. The results of this multiple regression
analysis are listed in Table 3. Therefore, while the prediction
that ordinal position would significantly influence Gc performance
(r = .1633, p-<.05) but not Gf performance (r = .0710) was supported,
none of the models appeared to be adequate predictors of the G
construct in this sample. However, upon closer examination of the
results, certain unexpected relationships among the predictor variables
were observed. It seemed reasonable that such relationships might
be specific to a college sample, as they have not been generally
reported in studies of more heterogeneous samples. These relationships
acted to statistically suppress the validity of several of the
predictor variables to explain performance on the Gc composite.
It was concluded that the contention that ordinal position
effects might not persist into adulthood (Schooler, 1972, 1973) may
be particularly evident in a university student sample. While the
attenuation of birth order and family size effects in higher SES
samples has been widely reported, the results of the present procedure
suggested that such effects might not only be attenuated, but might
be reversed, in a college population. These reversals appear to be
largely attributable to certain relationships which may be specific
to such a sample. Specifically, the unexpected significant positive
correlation between family size and SES (r = +.1972, p <.02) was in
marked contrast to the often reported negative intercorrelation of
these two variables in more heterogeneous samples (Belmont & Marolla,
1973; Eysenck & Cookson, 1970; Murray, 1971; Nisbet & Entwistle, 1967;
Record et al., 1969). This unexpected finding was interpreted in terms
of financial resources and limitations, which play an important and
perhaps decisive role in college attendance. Given this positive
relationship between family size and SES in addition to the highly
confounded nature of the relationship between family size and birth
order, it was not surprising to find a positive, although nonsignificant,
correlation between birth order and SES (r = +.1200, p <.10). The
effect of these relationships was to alter the hypothesized inter-
relationships among certain predictor variables, which in turn rendered
straightforward interpretation of the individual models difficult.
After further analysis of the data, it was determined that
cooperative suppression, as described by Cohen and Cohen (1975),
existed among several of the predictor variables. When partial
correlation procedures were utilized to statistically control for
these suppressor effects, a modified version of the economic model
was supported, in terms of both overall predictive significance and
with regard to the hypothesized relationships of the individual paths
and the achievement-related construct. The economic model was termed
"modified" due to the significant positive, not the predicted negative,
correlation between family size and SES.
Even with suppressor effects controlled, the prediction that the
socio-psychological model would be the optimal predictor of G
performance in this sample was not supported. Reasons for this
result were not obvious, although the unexpected positive relationships
between SES level and the family size/birth order variables appear
to have been an important contributing factor to these negative
results. Economic restrictions and other issues related to SES
might conceivably exert less influence on the opportunity of firstborn--
as opposed to later-born--siblings to receive a higher education.
The widely reported finding of firstborn overrepresentation at the
college level (Altus, 1966; Capra & Dittes, 1962; Schachter, 1963)
has led some investigators to contend that firstborns may be more
able (Nichols, 1964; Terman, 1925) or more achievement-oriented
(Breland, 1974; Burton, 1968; Forer, 1976; Glass et al., 1974; Nuttall
et al., 1976). However, in light of the present findings, the
possibility that economic factors importantly mediate firstborn
college attendance should also be considered.
High School Sample
Due to certain unexpected relationships in the previous procedure
which might conceivably be specific to a college student sample, it
was decided to evaluate the ability of the four models to predict Gc
performance in a more heterogeneous sample. Six tests which were
assumed to represent Gf or Gc ability were administered to 65 high
school seniors. The results for the hierarchical specified
inclusion methods for this sample are listed in Table 5. It was
found that all of the models were better predictors of G
performance in terms of overall variance and with regard to the
individual paths. In contrast to the college student sample results,
the predicted negative correlation between family size and SES
(r = -.2104, p <.05) was obtained for this more heterogeneous sample.
However, while the overall F ratios for all of the models were
significant, the regressions of the individual predictor variables
on Gc were significant only for the economic and socio-psychological
models, which explained approximately 51% and 54% of the variance,
respectively. It was concluded that, not only were the economic
and socio-psychological models fully supported in this sample, but
that t;,,y were both able to account for very substantial amounts of
the variance in the prediction of Gc performance. In an effort to
determine which of these two models was the most predictive of Gc,
the individual paths postulated by these models were examined. Due
to the observation that the firstborn variable of the socio-
psychological model was a more highly significant predictor of G
performance than was the family size variable of the economic model,
it was concluded that the hypothesis that the socio-psychological
model would most adequately explain achievement-related performance
It was evident that the current versions of the physiological
and Zajonc and Markus confluence models were less adequate predictors
of Gc performance, particularly for the more heterogeneous high school
sample. This observation might have important implications with regard
to future consideration of these models. It is conceivable that
modified versions may have to be considered. For example, the present
results indicate that the Zajonc and Markus "last born/only child
handicap" variable may have to be more carefully formulated. This
predictor variable was not significantly related to G for either
3nI11le. Moreover, contrary to the predictions, there was a trend for
only children to perform somewhat better than the other subjects on
Gc measures. For the college sample the correlation between only
child and Gc was +.1438 (p .10), in contrast to the significant
negative correlation between lastborn and Gc (r = -.1831, pR<.05).
Similarly, the correlations of Gc with only child and lastborn for the
high school sample were +.1299 (p) .10) and -.1236 (p >.10),
respectively. Thus, it is conceivable that only children should not
be included in this "handicap" category. However, due to the small
number of only children in this investigation and the very different
relative representations in both samples (4 only children as opposed
to 35 lastborn children in the college sample; 2 only children as
opposed to 25 lastborn children in the high school sample), any
statistical comparison with regard to these constructs would have been
With regard to the physiological model, the contention that birth
order correlates of intellectual performance might be largely physiological
in origin was seriously challenged by the statistically trivial inter-
correlations between birth order and Gf of .0710 for the college sample
and .0526 for the high school sample. This very negative result
concerning the basic component of this model--the influence of birth
order on the physiological mediation of intellectual ability performance--
indicates the need for considerable reformulation in terms of the
theoretical predictions of the physiological point of view, particularly
in terms of birth order effects. A serious shortcoming appears to be the
contradictory predictions associated with ordinal position. More
specifically,whereas the discussion of mother's age, increased chance of
genetic error, and "uterine fatigue" (Warren, 1966) indicates deleterious
birth order influences for later-born children, the problems often
associated with the pregnancy and birth of the firstborn child, including
a smaller placenta, longer labor, and increased perinatal complications
(Weller, 1965) suggest relative physiological advantages for later-born
children. Before further consideration of even a modified version of
this model is indicated, it is apparent that such complex predictions
concerning the physiological influences of birth order will have to be
more carefully isolated.
It is apparent from the present results that certain social and
psychological factors may be significant mediators of family configuration
influences--particularly firstborn superiority--on intellectual
achievement. Moreover, economic influences seem to exert powerful
effects, both directly and indirectly, on intellectual ability
With regard to the very different results obtained for the college
student and high school student samples, it is apparent that one must
specify the kind of sample involved when addressing issues related to
family configuration and SES correlates of intellectual ability.
Kennett',s (1973) discussion of the importance of SES influences on the
relationships between family size and intelligence seems particularly
appropriate to the results of the present investigation. The failure
of many studies to separate out the confounded relationships between
family configuration, SES, and other related variables, which may differ
widely depending on sampling techniques, may be a major contributor to
the contradictory findings and tentative conclusions so prevalent in
An issue which should be considered in the study of relatively
homogeneous samples, such as the university students in the present
investigation, is the possible restriction in range in terms of certain
family configuration variables, SES, and ability. Differences in
restriction in range across samples may result in difficulties in
comparison and interpretation. Moreover, such restrictions may be
differentially manifested for different variables. For example, in
the college sample, SES appeared to be the predictor variable which
manifested the least restriction in range. In contrast, while the
SES construct had relatively more variance in the heterogeneous high
school sample, the age range of from 17 to 18 years was extremely
narrow. This age constant aspect of the high school sample may have
been an important contributing factor to the increased correlations
obtained among the predictor variables and Gc.
Thus, while age was not included as a predictor variable in any
of the four models evaluated in this study, it is conceivable that a
potentially important predictor of G performance of the university
students in this study was the age of the subjects. While it might
generally be assumed that a college student population is homogeneous
in many ways, this assumption may be decreasingly valid with respect
to age. The age range of 17 to 30 years in this sample reflects the
increasing tendency for older students to attend universities, at both
undergraduate and graduate levels. Future studies of family
configuration correlates of intelligence and achievement among college
students may have to take this trend into account. Considerable
differences in age may present problems in interpretation in such
studies. For example, there existed in the present college sample a
highly significant positive correlation between age and Gc performance
(r = +.2753, p <.01) but a nonsignificant negative relationship between
age and Gf ability (r = -.0999). It is interesting to note that this
age trend for these constructs is in agreement with Horn's (1970, 1975)
postulated age-range functions of Gf and Gc abilities. As described
by Cunningham, Clayton, and Overton (1975), this theory predicts that:
Fluid intelligence, which is purported to reflect the
functioning of neurological structures, increases until
the cessation of neural maturation, generally during
adolescence, and then declines thereafter. In contrast,
crystallized intelligence is believed to reflect cultural
assimilation. In particular, it seems to be highly
influenced by formal and informal educational factors
throughout the life span. Assuming adequate health,
crystallized intelligence is presumed to increase steadily
across the adult age span. (p. 53)
However, although of theoretical interest, this cross-sectional age-
related increase on Gc or achievement-related measures may be
attributable to factors unrelated to Horn's contentions. For example,
age and college level (freshman, junior, graduate student, etc.) are
highly confounded. It is very difficult to separate the effects of
age and ability in a cross-sectional design such as that employed in
the present study. However, the influence of age in such samples may
be an important factor which should be taken into consideration in
Another potentially important issue is the possible existence
of different family configuration mediators of intellectual ability
performance with regard to racial background. Such race-related
differences might be expected to be particularly evident in terms of
SES and family size influences on intellectual ability measures.
Zajonc and Markus (1975) discuss the possible relationships between
family configuration variations among different national, regional,
ethnic, and racial groups and differences in intellectual test
performance. They suggest that differences in family configuration
patterns, including larger families and closer age spacing between
siblings, in the black population may be a contributing factor to
racial differences in intellectual performance scores. The high
school sample in the present investigation included ten black subjects.
The student body of this particular school is stratified so as to be
demographically representative of the state of Florida in terms of
economic level and race. Therefore, this sample was considered to
be a representative one, in contrast to the relatively homogeneous
and somewhat artificial university student sample. Although it would
have been interesting to examine specific patterns among the predictor
variables and Gc in addition to the predictive ability of the
individual models, across the two racial groups of the high school
sample, the small absolute number of black students and the very
different relative sample sizes (10 black students and 55 white
students) precluded further statistical analysis.
The present investigation introduced several new approaches to
the study of family configuration and intellectual achievement
which may possibly have important application for this research area.
Specifically, the use of the Gf and Gc constructs of Cattell and Horn
(Cattell, 1941, 1963, 1967; Horn, 1966, 1968, 1970, 1975; Horn &
Cattell, 1966, 1967) as measures of aptitude- as opposed to achievement-
related ability, the use of multiple regression techniques to
evaluate the predictive ability of the different path models, and
the analysis of suppressor effects may be potentially effective tools
for future studies of birth order and family size relationships to
Fluid and Crystallized Intelligence
Significant ordinal position influences on achievement-related
variables but not on certain aptitude or ability measures have been
reported (Burton, 1968; Forer, 1976; Glass et al., 1974; Nuttall et al.,
1976). These findings suggested to the author the possible
application of the Gf and Gc constructs of Cattell and Horn as
measures of aptitude and achievement-related performance. Horn (1966,
1968, 1970, 1975) hypothesizes that Gf reflects neurophysiological
functioning and active problem-solving ability. Gc is defined by
abilities appropriated through education and acculturation. Thus,
it was decided to utilize Gf measures to represent active problem-
solving ability or aptitude and G measures to represent achievement-
Marjoribanks (1976a, 1976b) contends that the equivocal results
reported in the birth order literature are partially due to the use of
restricted statistical techniques which have precluded examination of
the full character of the sibling constellation correlates of
intellectual abilities. The majority of studies have used product-
moment correlations, which reveal only bivariate relationships, and
analysis of variance, which requires the grouping of variables into
levels. Marjoribanks et al. (1975) suggest that the use of multiple
regression techniques--which provide a goodness-of-fit for any
functional form of the variables and do not require grouping--should
overcome some of the statistical shortcomings of many previous
investigations. In the present study, hierarchical specified inclusion
techniques were used to test the assumptions of four individual models.
This statistical technique was chosen because it permits the
analysis of the relationship between a criterion variable and a
set of predictor variables without the artificial grouping of
variables into levels. Moreover, the estimation of the relative
strengths of the individual paths in the models indicates the
magnitude of both direct and indirect influences in terms of the
hypothesized causal ordering for each model. In the hierarchical
inclusion procedure, each predictor variable is added in a single
step, and the increment in R2 at each successive step is considered
to be the component of variation attributable to that particular
Although statistical suppressor effects are rather uncommon in
behavioral sciences research, their existence can be devastating to
data interpretation. It was determined that certain unexpected
relationships which were observed in the college sample, particularly
the significant positive correlation between family size and SES, were
acting as cooperative suppressors (Cohen & Cohen, 1975) in at least
two of the proposed models. These suppressor effects resulted in a
decreased relationship between the family configuration and SES
variables and the achievement-related criterion variable. When
partial correlation techniques were utilized to control for these
statistical suppressor effects, a modified version of the economic
model was supported in terms of both overall variance accounted for
and the predictive ability of the individual paths.
The discovery of the existence of suppressor effects in the
college student sample may have potentially important implications
for the application of multiple regression techniques in general, and
for the use of these procedures in family configuration research in
particular. It is evident that birth order and family size variables
may exert very different effects on intellectual ability and achievement,
depending on different sampling procedures and the specific sample
investigated. The use of partial correlation techniques to control
for statistical suppression may be an important tool in the isolation
of the many complex relationships and issues in this literature.
Directions for Future Research
In the present study, an attempt was made to separate out some of
the confounded relationships of certain family configuration and SES
level correlates of intellectual ability and achievement. The present
results--along with the contradictory and tentative conclusions
prevalent in this research area--indicate the need for the generation
of new theories and hypotheses, which may be specific to particular
samples. The four models evaluated in this investigation should not
be considered to be mutually exclusive. Moreover, it is conceivable
that alternative models may be equally or even more predictive of
certain family configuration influences. For example, in a sample
consisting of 185 11-year-old boys, Marjoribanks et al. (1975)
found that a multiple regression equation containing father's
occupational status, inverse of the number of children in the family,
and the product of the two variables accounted for as much significant
variance as complex equations with many terms.
With regard to the present results, it is evident that, besides
the variables examined in this investigation, certain additional
variables should possibly be considered in future family configuration
studies. Such general influences include subject sex differences,
sex of siblings, age range of subjects, and the possible mediation of
racial and ethnic influences on intellectual ability performance.
Moreover, the supported predictions concerning firstborn
superiority on achievement-related variables might be more analytically
evaluated in a variety of ways. Such analytical paths might include
separate analyses for specific mechanisms related to the performance
of firstborn children on achievement-related measures. For example,
SES mediators of such firstborn superiority might include parental
attitudes toward achievement and educational aspirations which may
differ across SES levels, especially regarding earlier-born children
(Glass et al., 1974), certain concrete measures such as the number of
books, periodicals, and other literature in the home, and parental role
models--which might be particularly interesting for the daughters of
achieving women. Another area of investigation might involve certain
personality attributes which might be related to firstborn performance
on achievement-related variables. Such attributes might involve
increased susceptibility to social pressure and sensitivity to tension-
producing situations (Bradley, 196S), conscientiousness and effort
(Start & Start, 1974), social conformity and achievement motivation
(Forer, 1976). The examination of these hypothesized contributors to
firstborn superiority on achievement-related performance might further
clarify certain family configuration correlates of intellectual
ability performance. Of particular relevance to these considerations
is the importance of differential learning environments created by
families for their children (Marjoribanks, 1976a). These learning
environments were described in terms of five process variables:
(a) press for achievement; (b) press for activeness; (c) press for
intellectuality; (d) press for independence; and (e) press for English
(related to the reinforcement of the correct use of English).
Marjoribanks indicates that the relationships between birth order and
cognitive performance are complex:
The findings support the proposition that, if parents
provide different environmental experiences for their
children, then the environmental differences will be
related to birth-order variation in the cognitive
performance of children. (p. 764)
It is thus apparent that the often contradictory findings and
tentative conclusions in the birth order and family size literature
may be largely attributable to the failure to adequately control for
the many confounded variables and possible artifactual relationships
which may exist in certain samples. The results of the present
investigation indicate the need for the generation of new, more
analytical hypotheses with regard to family constellation influences
on aptitude and achievement. Moreover, the utilization of innovative
statistical techniques and conceptual approaches may serve to
effectively isolate the extremely complex family configuration correlates
of intellectual ability performance.
St' i Li'/F[ ARY F'. ::' F [
" 1-t z I
, "S -.:r." Unpuh!is icd
Li I. i., fc Delnver,
tr L .
Altus, W. D. Birth order and academic primogeniture. Journal of
Personality and Social Psychology, 1965, 2, 872-876. (a)
Altus, W. D. Birth order and scholastic aptitude. Journal of Consulting
Psychology, 1965, 29, 202-205. (b)
Altus, W. D. Birth order and its sequelae. Science, 1966, 151, 44-49.
Anastasi, A. Intelligence and family size. Psychological Bulletin,
1956, 53, 187-209.
Apperly, F. L. Characteristics of birth order position. Journal of
Heredity, 1939, 30, 493.
Belmont, L., & Marolla, F. A. Birth order, family size, and intelligence.
Science, 1973, 182, 1096-1101.
Bradln-, R. '. Birth order and school-related behavior: A heuristic
review. Psychological Bulletin, 1968, 70, 45-51.
Breland, H. M. Birth order, family configuration, and verbal achievement.
Child Development, 1974, 45, 1011-1019.
Burton, D. Birth order and intelligence. Journal of Social Psychology,
l.J, 76, 199-206.
Capra, P. C., & Dittes, J. E. Birth order as a selective factor among
volunteer subjects. Journal of Abnormal and Social Psychology,
1 52, 64, 602-608.
Cattell, J. M. Men of science. (4th ed.) Garrison, New York: Science
Cattell, R. B. Some theoretical issues in adult intelligence testing.
Psychological Bulletin, 1941, 38, 592.
Cattell, R. B. Theory of fluid and crystallized intelligence: A critical
( ..'.iment. Journal of Educational Psychology, 1963, 54, 122.
Cattell, R. B. The theory of fluid and crystallized intelligence checked
at 5-6 year-old level. British Journal of Educational Psychology,
137, 37, 209-224.
Cicirelli, V. G. Sibling constellation, creativity, IQ, and academic
achievement. Child Development, 1967, 38, 481-490.
Clarke, E. L. American men of letters, their nature and nurture. New
York: Columbia University Press, 1916.
Cohen, J., & Cohen, P. Applied multiple regression/correlation analysis
for the behavioral sciences. New York: Wiley & Sons, 1975.
Corliss, W. S. Relationships in achievement by sex and number of
children in the family. Michigan Education Review, 1964, 42,
Cunningham, W. R., Clayton, V., & Overton, W. Fluid and crystallized
intelligence in young adulthood and old age. Journal of Gerontology,
1975, 30, 53-55.
Dunn, H. L. (Ed.), Vital statistics of the United States, 1952 (Vol. 1).
Washington, D.C.: U.S. Government Printing Office, 1956.
Ellis, H. A study of British genius. London: Hurst & Blackett, 1904.
Eysenck, H.J., & Cookson, D. Personality in primary school children:
Family background. British Journal of Educational Psychology, 1970,
Forer, L. K. The birth order factor: How your personality is influenced
by your place in the family. New York: McKay, 1976.
French, J. W., Ekstrom, R. B., & Price, L. A. Manual for kit of reference
tests for cognitive factors. Princeton: Educational Testing Service,
Galton, F. English men of science. London: MacMillan, 1874.
Glass, D. C., Neulinger, J., & Brim, C. G. Birth order, verbal
intelli ence, and educational aspirations. Child Development,
1974, 45, 807-811.
Guilford, J. P. A new revision of the Army Alpha Examination and a
weighted scoring for three primary factors. Journal of Applied
Psychology, 1938, 5, 239-246.
Guilford, J. P., & Hoepfner, R. The analysis of intelligence. New
York: McGraw-Hill, 1971.
Hall, E. E. Psychological correlates of ordinal position in the two-child
family. Unpublished doctoral dissertation, University of Florida,
Hoepfner, R., & Guilford, J. P. Structure of Intellect Tests. Orange,
California: Sheridan Psychological Services, 1969.
Horn, J. L. Integration of structural and developmental concepts in the
theory of fluid and crystallized intelligence. In R. B. Cattell
(Ed.), Handbook of multivariate experimental psychology. Chicago:
Rand McNally, 1966.
Horn, J. L. Organization of abilities and the development of
intelligence. Psychological Review, 1968, 75, 242-259.
Horn, J. L. Organization of data on lifespan development of human abil-
ities. In L. R. Goulet & P. B. Baltes (Eds.), Lifespan develop-
mental psychology. New York: Academic Press, 1970.
Horn, J. L. Psychometric studies of aging and intelligence. In S.
Gershon & A. Ruskin (Eds.), Aging: Genesis and treatment of psycho-
logic disorders in the elderly. New York: Raven Press, 1975.
Horn, J. L., & Cattell, R. B. Refinement and test of the theory of
fluid and crystallized intelligence. Journal of Educational
Psychology, 1966, 57, 252-270.
Horn, J. L., & Cattell, R. B. Age differences in fluid and crystallized
intelligence. Acta Psychologica, 1967, 26, 107-129.
Huntington, E. Season of birth. New York: Wiley & Sons, 1928.
Jureskog, K. G., & Sorbom, D. EFAP: Exploratory factor analysis
program. Chicago: National Educational Resources, 1976. (a)
Joresk--, K. G., & Sorbom, D. LISRELL III: Estimation of linear
structural equation systems by maximum likelihood methods.
Chicago: National Educational Resources, 1976. (b)
Kennett, K. F. Measured intelligence, family size, and socio-economic
status. Alberta Journal of Educational Research, 1973, 19, 314-320.
Lees, J. P., & Stewart, A. H. Family or sibship position and scholastic
ability. Sociological Review (British), 1957 (July), 85-106.
Marjoribanks, K. Birth order, family environment, and mental abilities:
A rVct tssion surface analysis. Psychological Reports, 1976, 39,
759-7 3. (a)
Marj -ibanks, K. Sibsize, family environment, cognitive performance,
and a,. active characteristics. Journal of Psychology, 1976, 94,
Marjoribanks, K., & Walberg, H. J. Ordinal position, family environment,
and mental abilities, Journal of Social Psychology, 1975, 95, 77-84.
Marjoribanks, K., Walberg, H. J., & Bargen, M. Mental abilities:
Sibling constellations and social class correlations. British
Journal of Social and Clinical Psychology, 1975, 14, 109-116.
Maxwell, J., & Pilliner, A. E. The intellectual resemblance between
sibs. Annals of Human Genetics, 1960, 24, 23-32.
McCutcheon, L. E. Dumber by the dozen? Not necessarily Psychological
Reports, 1977, 40, 109-110.
Murray, C. The effects of ordinal position on measured intelligence
and peer acceptance in adolescence. British Journal of Social
Psychology, 1971, 10, 221-227.
Nichols, R. G. Birth order and intelligence. Unpublished manuscript,
1964. (Available from the National Merit Scholarship Corporation.)
Nie, N. H., Hull, C. H., Jenkins, J. G., Steinbrenner, K., & Bent, D. H.
(Eds.) SPSS: Statistical package for the social sciences (2nd ed.).
New York: McGraw-Hill, 1975.
Nisbet, J. D., & Entwistle, N. J. Intelligence and family size, 1949-1965.
British Journal of Educational Psychology, 1967, 37, 188-193.
Nuttall E. V., Nuttall, R. L., Polit, D., & Hunter, J. B. The effects
of family size, birth order, sibling separation and crowding on
the academic achievement of boys and girls. American Educational
Research Journal, 1976, 13, 217-223.
Record, R. G., McKeown, T., & Edwards, J. H. The relation of measured
intelligence to birth order and maternal age. Annals of Human
Genetics, 1969, 33, 61-69.
Rosenberg, B. G., & Sutton-Smith, B. The relationship of ordinal position
and sibling sex status to cognitive abilities. Psychonomic Science,
1964, 1, 81-82.
Schachter, S. Birth order, eminence, and higher education. American
Sociological Review, 1963, 28, 757-768.
Schooler, C. Birth order effects: Not here, not now: Psychological
Bulletin, 1972, 78, 161-175.
Schooler, C. Birth order effects: A reply to Breland. Psychological
Bulletin, 1973, 80, 213-214.
Start, A., & Start, K. B. The relation between birth order and effort
or conscientiousness among primary school children. Research in
Education, 1974, 9, 1-8.
Terman, L. M. Genetic studies of genius. The mental and physical traits
of a thousand gifted children. Stanford, California: Stanford
University Press, 1925.
Thurstone, L. L., & Thurstone, T. G. SRA primary mental abilities.
Chicago: Science Research Associates, 1949.
Treiman, M. Occupational prestige in comparative perspective. New
York: Academic Press, 1977.
Wark, D. M., Swanson, E. D., & Mack, J. More on birth order:
Intelligence and college plans. Journal of Individual Psychology,
1974, 30, 221-226.
Warren, J. R. Birth order and social behavior. Psychological Bulletin,
1_:3, 65, 38-49.
Weller, G. M. Arousal effects of the newborn infant of being first or
later born. American Journal of Orthopsychiatry, 1965, 35, 341-342.
Yoder, A. H. Birth order. Pedagogical Seminary, 1894, 3, 146.
Zajonc, R. B. Family configuration and intelligence. Science, 1976,
Zajonc, R. B., & Markus, G. B. Birth order and intellectual development.
Psychological Review, 1975, 82, 74-88.
Sandra Johnson Witt was born in Sanford, Florida, on July 17, 1946,
the daughter of Betty Beeler Johnson and Elmer Hunter Johnson. She
graduated from Hialeah Senior High School in Hialeah, Florida. A German
major at the University of Florida, she graduated in 1968 with high honors
and was elected to Phi Beta Kappa. She received her master's degree in
German and French from the University of Florida in 1969. She worked for
Trans World Airlines in New York until 1971 as an international airline
hostess and in-flight instructor. She married William Morris Witt in 1971.
In 1973 she received a master's degree in psychology from Emory
University under the direction of Dr. Philip Dreyer and Dr. Boyd
McCandless. She worked as staff psychologist and director of adult
remedial education at the Key Training Center for Mentally Retarded and
Physically Handicapped Adults in Lecanto, Florida, from 1973 until 1974.
In 1975 she entered the Ph.D. program in developmental psychology at the
University of Florida. While in this program, she received a pre-doctoral
fellowship from the Center for Gerontological Studies and worked as a
research assistant for Dr. Walter R. Cunningham, who is also the chairman
of her doctoral committee. She has two daughters, Amanda and Katherine.
I certify that I have read this study and that in my opinion
it conforms to acceptable standards of scholarly presentation and
is fully .dequ te, in scope and quality, as a dissertation for the
degree of Doctor of Philosophy.
Walter R. Cunningham, Chaiir *maI
Assistant Professor of Psych.ology
I certify that 1 hdve read this study and that in my cpinion
it conforms to acceptable standards of scholarly presentation and
is fully adequate, in scope and quality, as a dissertation for the
decree of Doctor of Philosophy.
William Keith ero /
Associate Prof.ssor of Psychology
I certify that I have read this study znd that in my opinion
it con*aor;,m to acceptable standards of scholarly presentation and
is f;lly adecuate, in scope and quality, as a dissertations for the
dea--e of Dcctc" of Philosophy.
Richard A. CGriggs '
Assistant Professor of Psychology
I certify that I have read this study and that in my opinion
it conforms to acceptable standards of scholarly oresentatior and
is fu iy dn-quate, i1 scope and -' ;-lity, as a dissertation for the
de;i ee 0on Duo"i- of Philosophy.
NaPt'ian ; Perry
Prfc essor of C inici1 Psvchrclogy
I certify that I have read this study and that in my opinion
it conforms to acceptable standards of scholarly presentation and
is fully adequate, in scope and quality, as a dissertation for the
degree of Doctor of Philosophy.
JAes E. Wh'rton
Us ociate Professor of Special Education
This dissertation was submitted to the Graduate Faculty of the
Department of Psychology in the College of Arts and Sciences and
the Graduate Council, and was accepted as partial fulfillment of
the requirements for the degree of Doctor of Philosophy.
Dean, Graduate School