Professional Documents
Culture Documents
Attachment To Mother-Attachment To Father
Attachment To Mother-Attachment To Father
REFERENCES
Linked references are available on JSTOR for this article:
https://www.jstor.org/stable/1130716?seq=1&cid=pdf-
reference#references_tab_contents
You may need to log in to JSTOR to access the linked references.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms
Society for Research in Child Development and Wiley are collaborating with JSTOR to digitize,
preserve and extend access to Child Development
Nathan A. Fox
University of Maryland
Nancy L. Kimmerly
National Institute for Child Health and Human Development
William D. Schafer
University of Maryland
Fox, NATHAN A.; KIMMERLY, NANCY L.; and SCHAFER, WILLIAM D. Attachment to M
Attachment to Father: A Meta-Analysis. CHILD DEVELOPMENT, 1991, 62, 210-225. An im
tenet of attachment theory is that classification of security or insecurity as derived from the
Situation reflects the quality of an infant's relationship with its caregiver. One piece of ev
supporting this claim is the lack of concordance of classification between an infant and its
and father. We performed a meta-analysis of the 11 studies that have examined the concor
mother/father attachment to an infant. We found that security of attachment to one par
dependent upon security to the other parent, that type of insecurity (avoidant/resistant) to on
was dependent upon type of insecurity to the other, and that subcategory classification w
secure category (B1B2/B3B4) to one parent was dependent upon subcategory classificatio
other. These data raise important questions regarding the meaning of infant attachment clas
as derived from the Strange Situation. Among the possible explanations for the pattern of
existence of concordant parenting styles and/or influence of infant temperament (possibly
dency to cry upon separation) on classification of Security/insecurity in the Strange Situati
Support for this work was provided, in part, by a grant to NAF from the National Institutes of
Health (HD 17899). We would like to thank each of the authors of the individual studies of mother/
father concordance who provided the raw data for the meta-analyses. In addition, we would like to
thank the anonymous reviewers whose suggestions and comments were helpful in revising this
paper. Requests for reprints should be sent to Nathan A. Fox, Institute for Child Study, University of
Maryland, College Park, MD 20742.
[Child Development, 1991, 62, 210-225. ? 1991 by the Society for Research in Child Development, Inc.
All rights reserved. 0009-3920/91/6201-0016$01.00]
from the Strange Situation, one would expect Researchers interested in infant tempera-
a higher degree of concordance between ment have argued that behavior in the
classification of the same infant to mother or Strange Situation and to some extent attach-
father. If temperament were orthogonal to at- ment classification reflects the infant's tem-
tachment classification, but not behavior in peramental bias (Goldsmith, Bradshaw, &
the Strange Situation (Belsky & Rovine, 1987;Rieser-Danner, 1986; Kagan, 1982; Thomp-
Sroufe, 1985; Sroufe & Waters, 1982) andson, Connell, & Bridges, 1988; Weber, Levitt,
classification reflected the individual infant- & Clark, 1986). For example, Kagan (1982)
caregiver relationship, the pattern of classifi- has claimed that threshold to respond with
cation to mother might be different than the negative affect to mild stress is a tempera-
pattern to father. This argument is based on mental characteristic that may be highlighted
the organizational view of attachment in in the Strange Situation. Infants who cry and
which security as derived from the Strange are not easily soothed in the Strange Situation
Situation reflects the ongoing relationship may more likely be classified into one cate-
that has developed between caregiver and gory as opposed to another. While attachment
child over the first year of life (Ainsworth et theorists have disputed these claims (Sroufe,
al., 1978; Sroufe & Waters, 1977). Mothers 1985), a "strong" infant temperament per-
and fathers may interact with their infants dif- spective would predict a great deal of simi-
ferently, and the degree to which they are larity of classification between caregivers
sensitive and responsive to their infants' bids based on the infant's dispositional tendency
for comfort may vary. Were that the case, a to respond to novelty or stress. To date, how-
child might not have the same attachment ever, most of the data on temperament and
classification to both. Indeed, it is possible for attachment have found only that disposition
a child to be securely attached to one parent to cry outside of the Strange Situation is re-
or caregiver and insecurely attached to a sec- lated to crying in the Strange Situation, but
ond caregiver. Attachment theorists have ar- that infant temperament does not discrimi-
gued that the data on discordance of mother/ nate among different classifications (see
father attachment support these claims. Al- Goldsmith & Alansky, 1987). Realistically,
ternatively, concordance of classification to classification of an infant and the quality of
mother and father may reflect similarity in in- attachment relationship are probably the re-
teractive style between parents. If father and sult of a complex interaction of parental
mother were both sensitive and responsive to caregiving styles, infant temperament, and
their infant's cues, one would expect a high the working model developed by the infant
degree of identity in classification between (Sroufe, 1985). However, there are only a few
parents and infant. studies that have described that interaction
between infant disposition and parental
Researchers who approach the study of caregiving style (e.g., Crockenberg, 1981).
behavior in the Strange Situation from a tem-
perament perspective might expect a high de- In one of a number of attempts to recon-
gree of concordance. For many, temperament cile temperament and attachment interpreta-
is regarded as a constitutionally based predis- tions of behavior in the Strange Situation,
position that is stable across time and general- Belsky (Belsky & Rovine, 1987) argued that
izes across situations (Buss & Plomin, 1984; behaviors in the Strange Situation are, in part,
Rothbart & Derryberry, 1981). Thus, for ex- a function of the child's temperament, though
ample, there may be individual differences in security or insecurity is not. He demonstrated
infant irritability or threshold to respond with across a number of data sets that infants of
distress to novelty that would present them- similar temperament were more likely to be
selves in the Strange Situation in a similar classified in either a grouping of Al to B2 or
manner regardless of presence of mother or B3 to C2. Temperament, however, could not
father. Infants with a low threshold to express discriminate avoidant/insecure infants (All/
negative affect, who are highly irritable, may A2) from their "dispositionaly close neigh-
be more likely to be classified as insecure/ bors" (B1 or B2), or, on the other extreme,
resistant (Davidson & Fox, 1989; Fox, 1989). resistant/insecure infants (Cl and C2) from
Or, there may be differences in infant play B4s. In this rapprochement, Belsky argued,
behavior evident from the early months of life one could find a solution to the conflict be-
that might indicate an infant highly focused tween opposing camps. Temperament influ-
on interaction with toys and objects. Such an enced behavior in the Strange Situation, but
infant may be more likely to be classified as security was still a function of the history of
insecure/avoidant in the Strange Situation the relationship with the caregiver. This argu-
(Lewis & Feiring, 1989). ment was first proposed by Ainsworth (Ains-
'CU
0 0 0
-, 00~ - -e 0 0
-o o c 1 o 01 0
- CA CA CA~ 0 0
Q1 C4C'C')
0 CU
tC 0
C4 CU
cd 0 p 4
cn >
u Cd
0 a 0
?M o X Xo ?o o o o ?o o 4 0
w w W V) W WCU
EU o
0 0
a -01 00 0 E0 -0lo.-.
01 0
CO C IOC)
Q 00......
4) C- -
0 Cd
00 CU
4) Cl
SE m
E o
~UF-4 1l
Eg P-4 -4 0
C01 01
PO 11C 11 C)
f) CO x
>% oo E oo E
cts ts 0 0
po P-4 m Cd( C:
(M3~ 00 r( m m 0
( T
9 cdJw~O~j
P4 innSI~j~j~jr
0 U -0 0 0* *N0 0 0 '
00 0
CI c I tIc
S0
0-B
each parent. However, from a temperamentfants classified as B3 through C2. These dis-
perspective, it may confound relationship and
positional differences express themselves in
disposition since it combined infants whoinfant behavior during the Strange Situation
may present opposite temperamental patterns but do not directly influence classification.
(A and C). Indeed, one would expect the For example, B4 infants may be as likely to
weakest relations of concordance. On the cry as infants classified as C. We examined, in
other hand, since most infants in each indi-analyses 5 and 6, concordance within each
vidual study were classified as secure, one half of this dichotomy. In analysis 5 we exam-
might expect high concordance in this analy-ined the relation between classification as A
sis. versus classification as either B1 or B2. If
The second comparison examined the re-
Belsky and Rovine (1987) are correct, there
should not be significant dependence found
lation of classifications of insecurity to either
in this table. Infants within this continuum
parent by comparing the type of insecurity
should as likely be classified as A (avoidant)
(avoidant or resistant). In this analysis, only
as
those infants who fell into one of the four cells B1 or B2. Similarly, for analysis 6, we ex-
amined the relation between classification as
of the 2 x 2 (A or C to mother/A or C to
C versus classification as either B3 or B4.
father) were included. Lack of independence
here would indicate that infants classified as Here again, if the dichotomy that Belsky pro-
avoidant and insecure to mother are more posed (one based on infant predisposition to
cry in response to mild stress) is correct, there
likely to also be classified as avoidant and in-
secure to father, or that infants classifiedshould
as not be significant dependence found
resistant and insecure to mother are more in this latter analysis. Infants should as likely
be classified as B3 or B4 as C.
likely to be classified as resistant and insecure
to father. Independence would signify that These six separate meta-analyses were
the two types of insecurity are, indeed, performed
as among the 11 studies, each of
many have argued, orthogonal and perhaps which was based on a series of 2 x 2 tables of
represent either different interactional his-
frequencies. The numbers of studies in the
six analyses varied from nine to 11 according
tories, different temperamental styles, or their
interaction. to whether data for the particular comparison
of interest were available. The chi-square
The third analysis compared infants in
significance level for each of the six meta-
the subcategories of the B classification. The
analyses was set at .0083 in order to control
underlying hypothesis is that the four sub-
the family-wise type I error rate at .05 over
categories of security may be dichotomized
the six analyses using overlapping series of
along a temperamental dimension (not unlike
studies.
Belsky's analysis; Belsky & Rovine, 1987).
Significant dependence in this analysis would The approach used for each meta-analy-
further support that perspective and may ar- sis is described by Fleiss (1981, pp. 165-168).
gue that within the category of security tem-The natural log of the odds ratio, after adding
peramental differences are evident in the ex-.5 to each cell frequency, was used as the ef-
pression of behavior in the Strange Situation.
fect size estimate for each study and weighted
by the reciprocal of its sampling variance. An
The second and third analyses describedoverall effect size was estimated as the
above are in a sense subsets of a grouping first
weighted average of the individual effect
proposed by Belsky (Al through B2 vs. B3
sizes. Two significance tests of interest are
through C2; Belsky & Rovine, 1987). Belsky
available: a test that the weighted average ef-
had argued that temperament played a role in
fect size is zero, indicating independence in a
the overt behaviors of the child in the Strange
Situation but did not influence classification common 2 x 2 table, and a test of homogene-
per se. Analysis 4 examined differences in ity, indicating that the studies share a com-
mon effect size. In addition to these results,
classification based on the division proposed
by Belsky. Lack of independence in this anal- we present, in a series of tables, the average
observed and average expected percentages
ysis would support Belsky's dichotomy and
for each 2 x 2 contingency table. These were
argue that styles of behavior in the Strange
computed by summing the observed marginal
Situation, rather than classification, may have
a common basis. frequencies across studies to obtain overall
marginal percentages and allocating these
Analyses 5 and 6 were a further attemptpercentages among the cells by the log odds
to examine the "Belsky dichotomy." Belsky
ratio. Cohen's kappa (Fleiss, 1981, p. 217) was
and others have argued that infants classifiedcomputed for each of these tables. Finally, we
Al through B2 are less likely to cry than in-present the 95% confidence interval for the
OBSERVED CELL SIZES SUMMED ACROSS STUDIES AND EXPECTED CELL SIZES
FOR A MODEL WITH No MOTHER-BY-FATHER ASSOCIATION
MOTHER
A B C
A:
N ............. 62 42.5 88 95.5 7 19.0 157
% ............. 9.2 6.3 13.1 14.2 1.0 2.8 23.4
B:
N ............. 65 83.3 308 292.2 67 64.5 440
% ........ ..... 9.7 12.4 45.8 43.5 10.0 9.6 65.5
C:
N.............. 10 11.2 42 50.3 23 13.5 75
% ........ ..... 1.5 1.7 6.3 7.5 3.4 2.0 11.2
Total:
N ............. 137 438 97 672
% ............. 20.4 65.2 14.4
1 Interested researchers may contact the first author for copies of the raw data from each of the
11 studies.
Father
Secure Insecure
?-
Kappa = . 309
N = 11
FIG. 1.-Percentages for the average observed and average expected for secure versus i
judgments allocated for the margins according to total marginal frequencies for the studies and for
according to the average odds ratio.
Father
A C
Kappa = .734
N = 10
FIG. 2.-Percentages for the average observed and average expected A versus C judgmen
for the margins according to total marginal frequencies for the studies and for the cells acc
average odds ratio.
Father
B1-B2 B3-B4
O4
Kappa = .578
N=9
Father
Al -B2 B3-C2
- -
Kappa = .514
N=9
rN
Kappa = .290
N=-9
FIG. 5.-Percentages for the average observed and average expected A versus B1B2 judgments
allocated for the margins according to total marginal frequencies for the studies and for the cells according
to the average odds ratio.
the average odds ratio is 1.125-3.819. The hy- pothesis of independence was not rejected,
pothesis of independence was not rejected, X2(1) = 3.921, p = .0477, at the alpha level set
X'(1) = 5.481, p = .0192, at the alpha level set (.0083), and so no significant relation between
(.0083), and so no significant relation between ratings of C or B3 and B4 to mothers and
ratings of A or Bi and B2 to mothers and fathers was found. The hypothesis of homoge-
fathers was found. The hypothesis of homo- neity was not rejected, X2(8) = 7.729, p =
geneity was not rejected, X2(8) = 10.844, .4604, suggesting again that there was no evi-
p = .2107, suggesting again that there was no dence that there was variability in the mag-
evidence that there was variability in the nitude of the odds ratio across studies.
magnitude of the odds ratio across studies.
Figure 6 contains the average observed
Figure 5 contains the average observed and average expected percentages across all
and average expected percentages across all studies showing an odds ratio of 1.970. The
studies showing an odds ratio of 2.073. The percentages were computed in an identical
percentages were computed in an identical manner to Figure 1. They present the relative
manner to Figure 1. They present the relative proportion of subjects most likely to appear in
proportion of subjects most likely to appear in any new sample. For these percentages, Co-
any new sample. For these percentages, Co- hen's kappa is .540; according to the test for
hen's kappa is .290; according to the test for independence from the meta-analysis, this is
independence from the meta-analysis, this is not significantly different from zero at the
not significantly different from zero at the .0083 level.
.0083 level.
Discussion
The sixth analysis compared classifica-
tion between mother and father of C versus The results of the meta-analyses clearly
B3 and B4. There were nine studies with data support a position that argues for dependenc
available for this analysis. The average logof attachment classification to mother and
odds ratio was .678; the corresponding oddsfather. They suggest that the assignment of
ratio is 1.970. The 95% confidence interval forsecurity/insecurity that the child receives as a
the average odds ratio is 1.006-3.857. The hy- result of behavior in the Strange Situation is
Father
1 1
C B3-B4
C'-
Kappa = .540
N=9
and B4The
similar to two different parents. to both father
data and mother. The fourth
also
analysis replicated
suggest that the type of insecurity that is findings
ob- of Belsky and
Rovine
served (avoidant or resistant) is (1987). Along
similar a continuum of individ-
to both
parents. And they indicate ual differences
that groups in reactivity,
of infants were
the subcategories of the securemoreclassification
likely to be categorized as Al through
B2 orclassification
may reflect distinct types since B3 through C2 to both parents.
within the secure infants was similar across
The fifth 2infants
mother and father. In addition, within x 2 comparison did not find
who are considered to exhibitthat if a childbehav-
similar was classified as A or as B1/B2
iors in the Strange Situation, to one parent reflect-
perhaps he or she was likely to be simi-
larly classified
ing common disposition (Belsky & Rovine, to the second parent. Similarly,
1987), there is some evidence the sixth
for meta-analytic
concor- 2 x 2 comparison did
dance based on classification. not find that if a child were classified as C or
B3/B4 he or she was likely to be similarly
The first comparison of the meta-analysis classified to the second parent. Within either
demonstrated that infants classified as secure continuum (Al through B2 or B3 through C)
to one parent were unlikely to be classified asthere was no evidence of dependence of
insecure to the other parent. The second com-classification between mother and father.
parison further clarified these relationships,
finding that there is dependence for the We performed the latter two comparisons
specific type of insecurity that is exhibited inas a check against the hypothesis that our first
the Strange Situation. Infants categorized as four analyses were illustrating differences be-
avoidant or resistant to one parent were likelytween those infants more or less likely to cry
to be similarly categorized to the other parent.in the Strange Situation (Al through B2 vs.
The third meta-analytic comparison found B3 through C2). The lack of significant depen-
that groups of the subcategories of the B dence of classification to mother and father
classification were distinct for individual in- found within either the "less likely to cry"
fants. Within the secure classification, the group (analysis 5) or the "more likely to cry"
style of behavior exhibited in the Strange group (analysis 6) does not rule out an inter-
Situation will be similar for BI and B2 or B3 pretation that a general tendency to cry to
the toddlers' relationship to mother and toships and infant temperament. Child Develop-
father: Related to conflict behavior and the ment, 48, 1-14.
readiness to establish new relationships. Child Sroufe, L. A., & Waters, E. (1977). Attachment as an
Development, 52, 932-940. organizational construct. Child Development,
Owen, M. T., & Chase-Lansdale, L. (1982, April). 48, 1184-1199.
Similarity between infant-mother and infant- Sroufe, L. A., & Waters, E. (1982). Issues of temper-
father attachments. Paper presented at the ament and attachment. American Journal of
biennial meetings of the Southwestern Society Orthopsychiatry, 52, 743-746.
for Research in Human Development, Galves- Stifter, C. A., & Fox, N. A. (1990). Infant reactivity:
ton, TX. Physiological correlates of newborn and five
Parke, R. D. (1978). Parent-infant interaction: Prog- month temperament. Developmental Psychol-
ress, paradigms, problems. In G. P. Sackett ogy, 26, 582-588.
(Ed.), Observing behavior: Vol. 1. Theory and Thomas, A., & Chess, S. (1977). Temperament and
application in mental retardation (pp. 69-94). development. New York: Brunner/Mazel.
Baltimore: University Park Press. Thomas, A., & Chess, S. (1980). The dynamics of
Parke, R. D., & Sawin, D. B. (1980). The family in psychological development. New York: Brun-
early infancy: Social-interactional and attitu- ner/Mazel.
dinal analyses. In F. A. Pedersen (Ed.), The Thompson, R. A., Connell, J. P., & Bridges, L. J.
father-infant relationship (pp. 44-70). New (1988). Temperament, emotion, and social in-
York: Praeger Special Studies. teractive behavior in the Strange Situation: An
Rothbart, M. M., & Derryberry, D. (1981). Develop- analysis of attachment system functioning.
ment of individual differences in temperament. Child Development, 59, 1102-1110.
In M. E. Lamb & A. L. Brown (Eds.), Ad- Thompson, R. A., & Lamb, M. E. (1982). Stranger
vances in developmental psychology (Vol. 1, sociability and its relationship to temperament
pp. 37-86). Hillsdale, NJ: Erlbaum. and social experience during the second year.
Sagi, A., Lamb, M. E., Lewkowicz, K. S., Shoham, Infant Behavior and Development, 5, 277-287.
R., Dvir, R., & Estes, D. (1985). Security of Vaughn, B. E., Lefever, G. B., Seifer, R., & Barg-
infant-mother, father, metapelet attachments low, P. (1989). Attachment behavior, attach-
among kibbutz-reared Israeli children. In I. ment security, and temperament during in-
Bretherton & E. Waters (Eds.), Growing points fancy. Child Development, 60, 728-737.
of attachment theory and research (pp. 257- Weber, R. A., Levitt, M. J., & Clark, M. C. (1986).
275). Monographs of the Society for Research Individual variation in attachment security and
in Child Development, 50(1-2, Serial No. 209). Strange Situation behavior: The role of mater-
Sroufe, L. A. (1985). Attachment classification from nal and infant temperament. Child Develop-
the perspective of infant-caregiver relation- ment, 57, 56-65.