You are on page 1of 15

INFANT

BEHAVIOR AND DEVELOPMENT

19,33-47

(1996)

Infant Temperament and Maternal Sensitivity


as Predictors of Attachment Security
AMYSUSMAN-STILLMAN,MARKKALKOSKE,AND

BYRONEGELAND

UniversiQ of Minnesota

IRWIN WALDMAN
Emory University
This study investigated contributions of infant irritability, sociability, and maternal sensitivity to
attachment security in a high-risk sample. Moderator, mediator, and additive models tested
hypotheses that maternal sensitivity determines security and that temperament influences type of
insecurity and subcategory placement. Composite measures of temperament and observational
ratings of maternal sensitivity at O-3 and 6 months predicted 12-month attachment classifications
and subcategory placement. Interaction of 3-month maternal sensitivity and infant irritability predicted security (moderator model). Six-month sensitivity independently predicted security (additive model) and mediated the relation between irritability and security (mediator model).
Maternal sensitivity distinguished secure and insecure infants. Three- and 6-month temperament
independently predicted type of insecurity and subcategory placement. An integrative conceptualization of attachment and temperament is supported.
temperament

attachment

maternal

A controversial
topic in the area of infant
socio-emotional
development
is the relation
between infant temperament and mother-child
attachment. Early in the debate, attachment and
temperament
theorists
represented
opposite
ends of the continuum, clashing over whether
or not infant temperament and mother-infant
attachment were orthogonal constructs (Kagan,
1984; Sroufe & Waters, 1982). Research examining direct relations between temperament and
attachment yielded inconsistent
results, with
some studies obtaining direct relations between
temperament and attachment (Calkins & Fox,
1992; Frodi,
Bridges,
& Shonk,
1989;
Goldsmith & Alansky, 1987; Miyake, Chen, &
Amy Susman-Stillman,
Institute of Child Development;
Mark Kalkoske,
now at Special Education
Services,
Morongo Unified School district, Twenty-nine Palms, CA
92270; Byron Egeland, Institute of Child Developmen$
Irwin Waldman, Department of Psychology.
Portions
of these data were presented
at the
International Conference on Infant Studies, May 1992, and
the meeting of the American Psychological
Society, June
1993. This research was partially supported by a National
Institute of Mental Health Child Development
Training
Grant to AX The authors would like to thank Megan
Gunnar and Alan Sroufe for their thoughtful comments on
earlier versions of the manuscript, John Ogawa for helpful
statistical advice, and Cheryl Gfrerer for superb editorial
assistance.
Correspondence and requests for reprints should be sent
to Byron Egeland, Institute of Child Development, 5 1 East
River Road, University of Minnesota, Minneapolis, MN
55455-0345.

sensitivity

parent-child

relations

Campos, 1985) and others failing to document


direct effects (Bates, Maslin, & Frankel, 1985;
Bohlin,
Hagekull,
Germer,
Andersson,
&
Lindberg, 1989; Gunnar, Mangelsdorf, Larson,
& Hertsgaard, 1989). Other studies found that
both infant and mother characteristics
contributed to the prediction of attachment security
(Izard, Haynes, Chisholm, & Baak, 1991). Fox,
Kimmerly, and Schafer (199 l), in their metaanalysis of 11 studies examining concordance
of mother-father
attachment, found that either
an attachment perspective (concordant parenting styles) or a temperament
interpretation
could explain why type of security, insecurity,
or subcategory classification to one parent was
dependent upon type of security, insecurity, or
subcategory classification
to the other parent.
Viewed as a whole, no firm conclusions regarding the relation between temperament
and
attachment could be drawn from the literature.
Some researchers shifted their focus towards
clarifying
the role of temperament
in the
Strange Situation. Sroufe (1985) suggested that
temperament might influence behavior in the
Strange Situation but not security of attachment.
Goldsmith,
Bradshaw,
and
Rieser-Danner
(1986) offered the hypothesis that temperament
may overtly contribute to individual differences
in attachment behaviors used in deciding attachment classifications.
Studies offer evidence to
suggest that temperament does influence certain
behaviors in the Strange Situation, such as cry33

34

Susman-Stillman,

Kalkorke,

ing. but does not necessarily influence attachment classification


(Thompson,
Connell,
&
Bridges, 1988; Vaughn, Lefever, Seifer, &
Barglow, 1989; Weber, Levitt, & Clark, 1986).
Other research supported the idea that temperament distinguishes infants within attachment
classification. Thompson and colleagues (Frodi
& Thompson, 1985; Thompson & Lamb, 1984)
discovered considerable heterogeneity in emotional expressiveness among securely attached
infants during the Strange Situation. Splitting the
securely attached infants into two subgroups
(B I + B2 vs. B3 + B4), they found differences in
emotional expressiveness
such that Bl + B2
infants behaved more like avoidant infants and
B3 + B4 infants behaved more like resistant
infants. Specifically, the Bl + B2 subgroup
showed low separation distress, long onset latency and rise time before becoming visibly distressed, and a brief recovery period, behaviors
which are also characteristic of avoidant infants
in the Strange Situation. The B3 + B4 subgroup
experienced
intense separation distress, brief
onset latency and rise time before becoming visibly distressed, and a prolonged recovery, behaviors which are also characteristic of resistant
infants in the Strange Situation.
A third line of research advances the idea
that the effects of temperament on attachment
are indirect. Van den Boom (1989) conducted
two studies investigating the roles of infant irritability and maternal sensitivity on the quality
of attachment.
The first study, which was
observational, described a pattern of interaction
between extremely irritable infants and their
mothers characterized
by maternal unresponsiveness and inconsistent
soothing strategies
and an opposite interaction pattern for nonirrtable infant-mother
dyads. The second study,
an intervention study, concentrated on increasing and improving aspects of maternal sensitivity for mothers of extremely irritable infants.
Significantly
more intervention
than control
infants were securely attached, with intervention mothers being more responsive and appropriately stimulating towards their infants. This
experiment illustrated that the relation between
infant irritability and mother-infant
attachment
was influenced by the degree to which mothers
exhibit sensitive behavior towards their irritable
infants.
Additional
evidence
comes
from
Crockenberg
( 198 1) and Crockenberg
and
McCluskey
(1986). They suggested that the

Egeland, and Waldman

relation between infant irritability and attachment is influenced by maternal attitudes about
responsiveness
towards infants and maternal
perceptions of social support.
These lines of research offer ample suggestion that the relation between infant temperament and attachment is more complex and the
effects more subtle than those assumed by earinvestigators.
lier
temperament-attachment
Theories of direct effects of infant temperament
on attachment
classification
may well be
incomplete and underestimate
the complexity
of the relation between variables. The focus of
temperament and attachment researchers, then,
shifted to formulating more complex approaches
to investigating
the nature of the relation
between temperament
and attachment.
Both
have offered alternative hypotheses for how
temperament
might influence the attachment
relationship
(Goldsmith et al., 1986; Sroufe,
1985). Yet it is still the case that the majority of
studies in the temperament-attachment
field
have examined direct relations between aspects
of infant temperament (mainly infant negative
reactivity) and attachment rather than indirect
relations, such as mediation
or moderation.
Virtually no empirical research has conceptualized complex interrelations between these variables or tested models of possible interrelations
between temperament and attachment.
A variety of alternative hypotheses regarding the temperament-attachment
relation have
been generated but not thoroughly examined.
Goldsmith et al. (1986) suggested that aspects
of temperament may mediate the social interactive processes involved in the development of
attachment between an infant and a caregiver.
Del Carmen, Pedersen, Huffman, and Bryan
(1993) offered some support for that hypothesis, testing the relative impact of maternal prenatal characteristics,
infant temperament,
and
mother-infant
dyadic patterns in predicting
They
found
the pattern
of
attachment.
mother-infant
dyadic interaction
while the
infant was distressed to be the strongest predictor of 12-month attachment security.
Sroufe ( 1985) suggested that temperament
might influence subcategory placement but not
security of attachment, or that caregiver sensitivity might influence security of attachment,
whereas temperament might influence type of
insecurity. A few studies have tested Sroufes
hypothesis that temperament affects subcategory

Temperament,

Sensitivity, and Attachment

classification
but not security of attachment.
Belsky and Rovine (1987), using ratings of
neonatal and 3-month temperament and attachment classifications
with each parent, conducted analyses using the traditional A-B-C
attachment groups and the subcategory placement grouping split suggested by Thompson
and Lamb (1984) (Al-B2 vs. B3-C2). Findings
revealed differences in neonatal behavior and
3-month infant temperament
when subjects
were grouped according to the Al-B2 and B3C2 subgroups and concordance
for motherfather attachment within the Al-B2 and B3-C2
subgroups but not for traditional A-B-C attachment groups. Significant differences along the
Al-B2 and B3-C2 subgroups provide evidence
that temperament plays a role in determining
subcategory placement, whereas a lack of concordance for traditional attachment groups indicates that factors other than temperament influence attachment classification.
These results
supported the idea that temperament seems to
influence infant emotional expressiveness
in
the Strange Situation rather than determine the
overall organization
of attachment behaviors
germane to classification.
Subsequent studies,
however, have been unable to replicate the AlB2 and B3-C2 subgroup split (Mangelsdorf,
Gunnar, Kestenbaum, Lang, & Andreas, 1990).
The question of whether or not temperamental
characteristics
play a role in determining
attachment subclassifications
still remains.
Earlier findings in the field as well as the
hypotheses
offered by Sroufe (1985) and
Goldsmith et al. (1986) are particularly valuable because they provide theoretical bases for
the development and validation of integrative
approaches. The aim of this study is to clarify
what integrative perspective
might be most
appropriate
to study the relation
between
attachment
and temperament.
We test the
hypotheses that caregiver sensitivity influences
security of attachment, and infant temperament
influences type of insecurity and subclassification (Sroufe, 1985). To do so, we adopted a
more complex conceptualization
of the relations between aspects of infant temperament
and attachment, testing multiple models of both
direct (additive) and indirect (moderation and
mediation) relations between the constructs.
Two dimensions of infant temperament with
particular relevance to the attachment relationship, irritability
and sociability,
as well as

35

maternal sensitivity, were selected as the main


predictors of attachment
classification.
Irritability is widely viewed as a valid dimension
of temperament (see van den Boom, 1989, for a
discussion on this topic), and a number of
investigators have found that irritability relates
to attachment
(Egeland
& Farber,
1984;
Goldsmith & Alansky, 1987). We defined irritability as the tendency of the infant to react
negatively to aspects of the environment or to
exhibit distress in affective and behavioral
ways (Goldsmith et al., 1987).
There is also support for the construct of
sociability or, as is the case with young infants,
behaviors that are thought to be indicative of
later sociable behavior. Few hypotheses have
been offered to explain the role of infant sociability on attachment security, yet it is a potentially influential factor. Lewis and Feiring (1989)
suggested that infants who are low on sociability
interact less with their mothers or respond more
avoidantly during reunion, increasing the possibility of an insecure attachment. Sociability
could also encourage positive, nonstressful interactions between the mother and infant and a
secure attachment. Alternatively, sociability may
modulate the infants reaction to the stranger,
and the effects of the stress of separation may be
influenced by interaction with the stranger. We
included sociability in order to clarify its influence on attachment. Sociability was defined as
the tendency for the infant to be involved in reciprocal interactions characteristic of later social
interaction or exhibit positive feelings affectively or behaviorally
within
social contexts
(Goldsmith et al., 1987).
The attachment
relationship
is widely
believed to exert significant influence on a
childs later development. Attachment theorists
have stressed maternal sensitivity, comprised of
awareness and accurate interpretation of and
appropriate and prompt responses to an infants
signals, as playing a vital role in the establishment of secure attachment. In this study, sensitivity is conceptualized as a global variable, reflecting a variety of maternal behaviors. Although
some research has illustrated that specific maternal behaviors, such as rejection (Isabella, 1993),
may influence attachment classifications,
we
chose to focus on broader constructs of sensitivity. This strategy takes specific behaviors into
consideration. For example, rejection is likely to
be highly negatively correlated with sensitivity

36

Susman-Stillman,

Kalkoske,

(although we are aware that low levels of rejection do not necessarily implicate high levels of
sensitivity). A number of findings indicate that
maternal behavior influences attachment formation (Ainsworth, Blehar, Waters, & Wall, 1978;
Bohlin et al., 1989; Egeland & Farber, 1984;
Weber et al., 1986).
HYPOTHESIZED
MODELS OF THE
TEMPERAMENT-ATTACHMENT
RELATIONSHIP
The fact that models of direct and indirect relations of temperament
and attachment remain
largely unexplored in the literature prompted us
to examine a number of possible models: moderator, mediator,
and additive.
Baron and
Kenny (1986) discussed the conceptual implications and utility of moderator and mediator
models for understanding
different types of
relations among variables. We selected models
at two different points (3 and 6 months) during
the infants 1st year of life, a critical time for
attachment formation.
A moderator model is evident when the
influence of the independent
variable on the
dependent variable depends upon the level of a
second independent variable (e.g., when there
is interaction
between the independent
variables). It tells one when to expect a certain
event (e.g., that high levels of irritability predict insecure attachment only when maternal
sensitivity is low). We hypothesized an interaction between irritability and sensitivity, specifically that highly irritable infants with mothers
who were low on sensitivity were likely to be
insecurely attached.
A mediator model differs from the moderator model because it attempts to discern why or
how effects may occur rather than informing
under what conditions
an effect may occur.
Thus, it gets closer to a causal explanation. A
variable is considered to be a mediator when it
accounts for all or part of the relation between
the independent
variable and the dependent
variable. Specifically,
mediation occurs when
relations exist between the independent
variable and the mediator, the mediator and the
dependent variable, and the independent variable and the dependent variable. With the addition of the mediator into the equation, the
previously significant relation between the independent variable and the dependent variable
should be significantly
weakened. Mediators

Egeland, and Waldman

represent properties of a person that transform


the predictor variable in some way (Baron &
Kenny, 1986, p. 1178). Based on evidence in
the literature (van den Boom, 1989) and expectations from attachment theory, we hypothesized that any effects of temperament
on
attachment would be mediated by maternal sensitivity. That is, infant temperament
would
influence attachment only to the extent that it
significantly influenced a mothers responsiveness to her child.
The third model we fit was the additive
model, which addresses the question, Do each
of these variables contribute independently to the
prediction of attachment? The additive model
could be complementary to either the moderator
or mediator models, because independent variables can be independent contributors as well as
moderators or mediators. An additive effect was
expected for maternal sensitivity in relation to
attachment security. Because there is support in
the literature for a relation between irritability
and anxious-resistant
attachment (Goldsmith &
Alansky,
1987), we hypothesized
that there
would be a correspondence
between irritability and type of insecurity, specifically anxiousresistant. The paucity of data regarding the
relation between sociability and attachment did
not allow us to form any specific hypotheses.
Thus, the purpose of this study was to test
models of direct and indirect effects of temperament and sensitivity on attachment security. We tested the hypotheses that maternal sensitivity determines attachment security (B vs.
non-B) and that temperamental
factors influence type of insecurity and subcategory placement using moderator, mediator, and additive
models. In addition, by testing models at O-3
and 6 months, we hoped to document developmental changes in the mother-child
relationship in the 1st half year of life.
An earlier study conducted by E&and and Farber
(I 984) used this particular data set but had as its purpose the
examination of a variety of different maternal variables.
including maternal personality, maternal behavior, and life
circumstances, and how those separately predicted attachment security and changes in attachment security from 12 to
18 months. Although we used some of the same predictor
and outcome variables as Egeland and Farber. namely maternal sensitivity at 3 and 6 months and attachment classitications at 12 months. we created a multimethod measure of
aspects of infant temperament and had as our purpose the
testing of the relative influence of temperament and sensitivity on attachment classification in different combinations.

Temperament,

Sensitivity, and Attachment

METHOD
Sample
Data for this study were obtained from the Mother-Child
Interaction Project, an ongoing longitudinal study being conducted at the University of Minnesota (Egeland & Farber,
1984). The original sample consisted of 267 poor, primiparous women who received prenatal health care through public assistance at the Maternal and Infant Care Clinics,
Minneapolis Health Department. At the time of birth, mothers ranged in age from 12 to 37 years (M = 20.52, SD =
3.65). Mothers educational level ranged from junior high
school to postcollege graduate level, and 40% had not graduated from high school by the time of their infants birth.
Most mothers were European American (80%); the others
were African American (14%). Native American,
and
Chicano (6%). Approximately
15% of the infants were of
mixed racial backgrounds.
At the 12.month attachment
assessment, 212 infants were tested. Due to the longitudinal
nature of the study, the number of subjects available for
these analyses vary (range = 170-206). The maximum number of subjects available for each analysis was used.

Procedure
Mothers and their infants were visited by trained interviewers and observers during the neonatal period and at 3, 6,
and 12 months. Data were collected through interviews and
observations in the hospital and University laboratory. Data
for this study were drawn from available measures of infant
temperament, mother-child
interaction, and mother-infant
attachment.

Measures
O-3 and b-Month Temperament Scores. Measures of infant
irritability and infant sociability were constructed by drawing on multiple data sources collected neonatally and at 3
and 6 months of life. Nurses conducted naturalistic observations of the infants in the hospital nursery during the
infants hospital stay. Items were rated on a 3-point scale.
For reliability purposes, a sample of 30 infants was rated by
two nurses on the same shift on the same day. Interobserver
reliability could not be calculated because there was little
or no within-nurse variability. Reliability was conducted on
all 381 item ratings, with 67% perfect agreement and 38%
having a l-point discrepancy. Most of the correlations were
significant, and directionality of the ratings appears to be
consistent (Vaughn, Crichton, & Egeland,
1982). Five
items relevant to either irritability or sociability
were
selected from the total items rated (see Table 1). Scores
used were the average of all nurses ratings for each item.
At 3 and 6 months, temperament data were collected
from maternal report via the Carey Infant Temperament
Questionnaire
(ITQ; Carey, 1970), a structured questionnaire measure of infant temperament assessing infant behavior in a variety of situations and in-home observations of
mother-child interactions by independent trained observers
during a feeding session. In-home observers were trained to
an average 85% agreement for the entire set of items using
videotaped feedings prior to observing the mothers and
infants. Continual checks throughout data collection were
conducted. Interrater agreement on the 3- and 6-month inhome observations using the Lawlis-Lu ~2 (for recommendations on its use, see Tinsley & Weiss, 1975) was calculat-

37

ed with agreement defined as a discrepancy of 2 points or


less for a 9-point scale with two raters and 1 point for scales
less than 9 points with two raters. The Lawlis-Lu ~2 index
allows researchers to define a criterion for agreement and
thus distinguish between serious and unimportant disagreements. It corrects for the number of agreements expected by
chance and is appropriate only when interrater agreement is
greater than the agreement expected on the basis of chance.
Reliability for the items used in this study is exhibited by
the moderate-to-high
T value of agreement, ranging from
.7 1 to .90 at 3 months and 64 to .77 at 6 months, and a significant x2 for all items, p < .05. which indicates that the
observed agreement was greater than the agreement expected on the basis of chance.
Composite temperament scores using observational and
maternal report data were created. This strategy was used to
follow good practice of current temperament researchers
(Goldsmith & Rieser-Danner,
1990). In addition, methodological problems exist with this early version of the ITQ
(Vaughn, Taraldson, Crichton, & Egeland, 1980), and this
strategy allowed utilization of all available data on infant
temperament.
The procedure for developing each temperament measure was as follows. Items relevant to irritability and sociability were selected from the nurses ratings, the ITQ, and
the feeding observation.
Scores for these items were
reversed when necessary, such that high scores on all items
pertaining to a given dimension contributed positively to
that dimension (i.e., an irritability item for which a high
score represented only whimpers if hungry was reversed
so that the previously low score for cries loudly when hungry became the high score, as it represented a more irritable characteristic).
Following necessary item reversals, all
item scores were standardized. Cronbach alphas were then
computed for each scale at both O-3 and 6 months. A few
items which significantly depressed the alpha (i.e., were not
consistent with others on the scale) at either time period
were removed from each scale, until a common item set
remained at both 3 and 6 months (see Table 1). Once sets
were finalized, temperament scores were computed by taking the mean z score for all cases with at least two thirds of
the available items. This method allowed the computation
of scores for more subjects than would have been permitted
had it been required that data on each individual item exist
for all cases, while at the same time not producing low
scores for subjects with some missing data.
Maternal Sensifivify. Scores for maternal sensitivity were
based on observer ratings during the feeding situation (one
observation at 3 months, the mean of the two observations
at 6 months). At 3 and 6 months, a 9.point sensitivity scale
was used. Mothers were rated on varying degrees of sensitivity, attention, and accommodation
to their infants. This
scale is based on the Ainsworth et al. (1978) conceptualization of sensitivity, which includes awareness and accurate
interpretation of and appropriate and prompt response to
the infants signals. Mothers who were inattentive or unresponsive to obvious cues received very low sensitivity
scores; mothers who paid variable attention and intermittently accommodated
their infants received average sensitivity scores; and mothers who were consistently accommo-

The nurses ratings are an exception and were only


included in the 3-month scores because they were collected
neonatally.

Susman-Stillman,

38

Items Comprising

Kalkoske,

Egeland, and Waldman

TABLE 1
Temperament Scales at 3 and 6 Months
3- and 6-Month

Measure

Temperament

Items

lrritabilityo
Carey

Feeding
Nurses Ratingsc

Sleep routine bothered considerably by changes in time, place, health


Cries with interruption of feeding
Cries loudly when hungry
Nothing stops hunger cry
Unpleasant res onse to new foods
Does not read1.Py accept changes in foods
Cries until receivin type of food desired
Cries when having % owe1 movement
Fussy when diaper soiled
Fussy when diaper wet
Cannot be distracted when fussing about diaper
Fussy during diapering and dressing
Objects to bathing after more than 2 weeks
Fusses or cries in response to new procedures
Continues to object to new procedures after 2-3 times
Generally fussy or crying once procedures established
Not distracted when fussing about procedures
Fussy during doctor exam
Cries for several minutes following shots
Cries and fusses with any illness
Reacts negatively to repeated exposure to sensory stimulation
Does not tolerate new situations well
Rejects new toys when first presented
Generally fussy during pla
Babys temperament (irrito i: le vs. pleasant and cheerful)
More fussy than other babies
Cries more than other babies
Less easy to soothe or quiet than others
Less content after feeding than others

Sociabilityb
Carey

Feeding

Nurses

Ratingsc

a3-month ITI= .70; &month a


b3-month cc= .72; &month a
CBecouse nurses ratings were
in the computation of the 3-month

= 56
= .49
completed in the neonatal nursery, they were only included
temperament measures.

dating and responsive to their infants cues and needs


received high sensitivity scores. As was done for all inhome observations,
interrater agreement
was calculated
using the Lawlis-Lu index, with agreement defined as a
2-point discrepancy among raters on the 9-point scale. At
both 3 and 6 months, the Lawlis-Lu x2 was significant at p
< .05, with the T value of agreement in the moderate-tohigh range (3-month T = .75; 6-month T = .X7).
Attachment Classication. Attachment
classification
was
assessed at 12 months using Ainsworth
and Wittigs
Strange Situation (Ainsworth
et al., 1978). Videotaped
Strange Situations
were coded by two expert coders
(Everett Waters and Brian Vaughn), who independently
classified each of the infants. Rater agreement on major
classifications
(A-B-C) was X9%, and disagreements
were

resolved through conferencing.


Mains disorganized
(D)
category was not used. Twenty-one percent of infants were
classified as insecure-avoidant
(11 = 46); 567~ were claasified as secure (Bl + B2 = 66; B3 + B4 = Sl); and 23%
were classified as insecure-resistant
(n = 48).

Construct Independence
Pearson correlation coefficients were computed to examine
the relative independence of the irritability, sociability, and
sensitivity measures (see Table 2). Irritability and sociability were negatively correlated at both 3 and 6 months (r =
-.29, p = .OO; I = -.12, p = .04, respectively). Irritability
also correlated negatively with maternal sensitivity at both
3 and 6 months (I- = -.25, p = .CQ r = -.l2, p = .02, respectively). As these correlations are low to moderate in magni-

Tempemment,
Intercorrelations

Sensitivity, and Attachment

TABLE 2
of Tempemment and Maternal

Variables
1. infant Irritability
(3 months)
2. Infant Irritability
(6 months)
3. Infant Sociability
(3 months)
4. Infant Sociability
(6 months)
5. Maternal Sensitiviiy
(3 months)
6. Maternal Sensitivity
(6 months)
p -C.05, two-tailed test.

Sensitivity Variables
4

39

.49

-.29

-.25

-.13

-.13

.35

-.25**

-.14

.31 *

.06

-.14

-.12

.25**

.37

l*p c

.50

.Ol,two-tailed test.

tude, they suggest that the measures have a large degree of


unshared variance. Sociability
and maternal sensitivity
were positively correlated at 3 and 6 months (r = .31,
p = .OO; Y = .36, p = .OO, respectively). Again, these correlations are significant, but it is clear that a large proportion of
the variance is unshared.

Analyses
Three sets of logistic regression analyses were conducted to
predict the likelihood
that infant temperament
and/or
maternal sensitivity accounted for (a) attachment security
(B vs. A + C), (b) type of insecurity (A vs. C), and (c) subcategory placement (A + Bl + B2 vs. B3 + B4 + C).
Follow-up t tests comparing levels of maternal sensitivity
were conducted within subcategory placement groups (A
vs. Bl + B2; B3 + B4 vs. C).
Logistic regression analysis is a procedure used when
modeling the relationship between a dichotomous outcome
variable and a set of predictor variables (Hosmer &
Lemeshow, 1989). This method was preferred over a discriminant function analysis because the assumptions are
less restrictive. Because logistic regression directly predicts
the likelihood of an occurrence, it was also better suited to
the hypotheses in this study.
Logistic regressions were run separately for each model
(additive, moderator, mediator) at each measurement time
(3 and 6 months). Results for each set of analyses will be
reported by measurement period in the following order:
security of attachment, type of insecurity, and subcategory
placement groups. Table 3 illustrates group means for each

Statistics reported for logistic regression are as follows: -2 log likelihood ~2, model ~2, and the Wuld statistic. -2 log likelihood answers the question, Is the actual
model significantly different from a perfect model (one in
which the predictors are perfectly related to the outcome; a
saturated model)? The model ~2 answers the question,
Is the proposed model better than chance? (Do the covariates increase ones ability to predict the outcome over and
above chance?). The Wuld statistic tests the significance
of each predictor or term in the logistic regression equation.
As in linear regression, the sign indicates directionality.

infant and mother variable, and Tables 4 through 6 summarize the significant and nonsignificant models for the sets of
analyses at 3 and 6 months.

RESULTS
Predicting

Security/Insecurity

At 3 months, the only significant model was the moderator


model, with an interaction between infant irritability and
maternal sensitivity (see Table 4 and Figure 1; R = -.08).
This interaction improved prediction over a nonsignificant
additive model which included independent effects of irritability and sensitivity (model x2(2, N, = 195) = 3.03,
p < .22; -2 log likelihood x2( 193) = 267.03, p < ,001;
Wald(1) = 0.38 and 2.09, p < .54 and .15; B = -0.29 and
0.15; R = .oO and .02, respectively). There was no evidence
for a mediator effect of maternal sensitivity on either temperament variable. Follow-up tests on the interaction performed by splitting the temperament variables into highand low-irritability groups so that each included one half of
the sample indicated that, contrary to our a priori hypothesis, high sensitivity increased the likelihood of security for
the low-irritability group only (model x2( 1, N = 99) = 4.36,
p < .04; -2 log likelihood x2(97) = 132.39, p < ,001;
Wa/d(l) = 4.07, p < .05; R = -.12; B = -0.30).
At 6 months, when tested in additive models including
either irritability or sociability, respectively, an additive
effect was found only for maternal sensitivity, indicating a
positive relation between sensitivity and security of attachment (see Table 4 and Figure 2; R = .17). There was no
moderator effect, contrary to what was found at at 3 months.
Maternal sensitivity also mediated the effects of irritability
on attachment by accounting for the direct relationship
between irritability and attachment. When entered alone in a
model predicting security, 6.month irritability contributed
negatively at a marginally significant level (see Figure 3).
An independent regression demonstrated that irritability at 6
months marginally contributed to the prediction of 6-month
sensitivity, r2 = .13, F(1, 168) = 3.24, p < .09. In a model
including both 6-month sensitivity and 6-month irritability,
however, the overall model remained significant (see Table
4 and Figure 3; R = -. 17). but the contribution of irritability
was no longer significant (Wald( 1) = 1.24, p < .23; B = 0.55;

Susman-Stillman,

40

Kalkorke,

Egeland, and Waldman

TABLE 3
Group Means for Infant Temperament and Maternal
&CUM2

Variable

(SD1

lrr~itdili~li
6 Months

Avoidant

Insecure
(SD1

-.03

(.33)

.02 (.29)

-.05

(.33)

.05 (.33)

Sensitivity Variables

Resistant

(SD1
.02 (.29)
-.02

ABlB2

(SD)

(.34)

B3B4C

(SD1

(SD1

.02 (30)

-.02

(.33)

.Ol (.30)

.12 (.32)

-.08

(.32)

.08 (.33)

Sji;abft\T
6 Months

Ma3~oy~~Sensitivity
6 Months

.05 (53)

-.03

(.58)

-.OO (.62)

-.04

(.53)

6.26
6.25

5.94
5.72

(I .4)
(1.3)

(1.4)
(1.2)

.08 (.51)
.Ol (.55)

5.89
5.66

-. 13 (.64)

(1.4)
(1.3)

-.09

(.53)

6.00
5.78

(1.5)
(1.3)

.08 (.48)
.Ol (.62)
6.18
5.99

(1.5)
(1.3)

-.07

(.62)

-.06

(.55)

6.04

(1.4)
(1.2)

6.02

TABLE 4
Summa
of Logistic Regression Models Testing Direct, Mediated,
an 7 Moderated Prediction of Securitv Versus lnsecuritv
D-3 Months
Question
Does sensitivity significantly predict
security vs. insecurity?
Does temperament significantly predict
security vs. insecurity?
IS the effect of temperament significantly
mediated by sensitivity?
Is the prediction of security moderoted?

Irritability

Sociability

Irritability

Sociability

No

No

Yes

Yes

No

No

No
Yes

No
No

R = .OO). This supports our hypothesis that the effects of irritability on attachment security are indirect and operate
through maternal sensitivity.
Although maternal sensitivity remained a significant
predictor in relation to sociability in the 6-month mediator
model, the mediator model should not be considered significant because there was no additive effect of sociability on
attachment (mode1 x2( I, N = 198) = 0.23. p < .63; -2 log
likelihood ~(196) = 272.23, p < .0003; Wafd(l) = 0.23,
p < .63; B = -0.12; R = ,oO). Irrespective of aspect of temperament, maternal sensitivity was a significant predictor of
attachment security at 6 months.

Predicting

6 Months

Type of Insecurity

Within the insecure group, the only significant models were


additive. At 3 months, infant sociability predicted anxiousavoidant attachment, with avoidantly attached infants scoring higher on our measure of sociability
(model ~2
(2 N = 91) = 5.24, p < .07; -2 log likelihood x2(87)
= 119.48, p < .Ol; Wuld(1) = 4.45, p < .04; B = -0.91; R =
-, 14; see Table 5). At 6 months, there was a trend for high
levels of infant irritability
to predict anxious-resistant
attachment (model x2(2 N, = 78) = 3.82, p c .15; -2 log
likelihood x2(74) = 102.91, p < .02; Wald(l) = 2.56.
p < .I 1; B = 1.17; R = .07; see Table 5). We view this trend
as noteworthy because this p value represents a two-tailed
test, and the results fit with our a priori hypothesis. It
should be noted that, given this statistical procedure, these
variables were successful in improving the probability of

Marginally

No
No
No

Yes
No

predicting resistance versus avoidance with prior knowledge of membership in the insecure group.

Predicting

Subcategory

Placement

At 3 months, the only model approaching significance was


the additive model in which sociability increased the likelihood of membership in the A + B 1 + B2 subgroup as compared to the B3 + B4 + C subgroup (model x2(2, N = 198)
= 4.42, p < ,004; -2 log likelihood x2(194) = 267.21,
p < .OOl; Wald( 1) = 4.20, p < .OS; B = -0.55; R = -.09; see
Table 6). At 6 months, only irritability had an additive
effect, increasing the likelihood of membership in the B3 +
B4 + C subgroup as compared to the A + BI + B2 subgroup (model x2(2, N = 171) = 7.96, p < .04; -2 log likelihood x2( 167) = 226.86. p < .OOl; Wald( 1) = 7.49, p < ,005;
B = I .44; R = IS; see Table 6).
A follow-up I test comparing the B I + B2 versus B3 +
B4 subgroups on irritability and sociability was conducted
to insure that these temperamental differences were not due
solely to differences among the A or C subgroups. The B3
+ B4 subgroup was significantly higher on 6-month irritability (Bl + B2 M = -.12; B3 + B4 M = .05), t(97) =
-2.55, p < .O I, There were no differences in sociability or
maternal sensitivity at either point.
Follow-up r-test comparisons within the A + Bl + 82
subgroup (A vs. Bl + B2) of irritability, sociability. and
maternal sensitivity revealed marginal differences in maternal sensitivity at 3 months which became significant by 6
months (3.month A group M = 5.8X; Bl + B2 group M =

CN
Model Chi Square
-2 Log Ukelihood
2-Pathwaydgnifkance:

df
1

.06

192

.OOl

3.98
263.07

ss

Bo-.70.~~3.n1911,Q<.05

Figure 1. Significant predictors of attachment security at 3 months.

ChlSauare
Model CM

square

-2 lag Likdihood

12.45
219.7

1 - Pathway significance:

I2s
2
166

B = .43, M

.002
.003
=9.04, df = 1, p e .003.

Figure 2. Significant predictors of attachment security at 6 months.


41

42

Susman-Stillman,

Kalkoske, Egeland, and Waldman

6-month
M8ternal
Sensitivity
L

Model Chl Square


-2 Log ukelihood
1 - Pathway signlfbncez

12.46
221.71

2
167

.002
.003

B = -.61,~=9.04.df=l,p<.OO3.~=-.17

2 - When preliminary model induded only this pathway:


Model Chi Square = 3.30, df = 1, Q < .07
Pathway significence: B = -.87, W
= 3.21 df = 1,

p < .07.

Figure 3. Significant predictors of attachment security at 6 months.


6.3X: 3-month t(105) = -1.75, p < .08; h-month A group
M = 5.66, Bl + B2 group M = 6.23; 6-month t(104) =
-2.27, p < ,025). Difference, in maternal sensitivity within
the B3 + B4 + C subgroup (B3 + B4 vs. C) were found only
at 6 months (B3 + B4 group M = 6.26, C group M = 5.77).
t(93) = -2.03, p < .05. As expected, there were no temperament differences within either of those two subgroups (A
vs.Bl+B2orB.?+B4va.C).

DISCUSSION
Based on models which included aspects of
infant temperament and a global measure of
maternal sensitivity, the 3-month results partially
confirmed and the 6-month results more fully
supported our original hypotheses: Maternal sensitivity predicted attachment security, and temperament predicted type of insecurity and subclassification.
Findings
from the
category
subcategory placement groups strengthened those
conclusions, because maternal sensitivity distinguished
between
securely
and
insecurely
attached infants, and aspects of infant temperament corresponded to type of insecurity and subcategory placement. Significance
of different
models from 3- to 6-months indicated a possible developmental change in the temperamentsensitivity component of the mother-child relationship during that period.

Correlations of the temperament and sensitivity variables suggest that the variables share
low-to-moderate
proportions of the variance.
Not surprisingly, they also suggest that there is
a negative relation between irritability and sensitivity and a positive relation between sociability and sensitivity. This fits with findings in the
field (Crockenberg & McCluskey,
1986) suggesting that maternal behavior is influenced by
infant characteristics
(e.g., characterized
by
lower levels of maternal responsiveness towards
irritable infants).
The correlation table also illustrates low-tomoderate stability in irritability and sociability.
These findings are consistent with other documentation of moderate stability in infant characteristics during infancy. Rothbart ( 1986), using
both home observations and the Infant Behavior
Questionnaire
(IBQ) to assess infant temperament, found convergence and stability in the
global measures of positive and negative reactivity from 3 to 6 months (positive reactivity was
significantly stable only from the IBQ). GarciaCo11 Zenah, Walk. Lester. and Vohr (as cited in
Belsky, Fish, & Isabella, 199 1) also found modest stability in positive and negative emotionality
between 3 and 7 months. Belsky et al. ( 1991 )

Temperament,

Sensitivity, and Attachment

43

TABLE 5
Summary of logistic Regression Models Testing Direct, Mediated,
and Moderated Prediction of Type of Insecurity (A vs. C)
6 Months

D-3 Months
Question
Does sensitivity significantly predict
security vs. insecurity?
Does temperament significantly predict
security vs. insecurity?
Is the effect of temperament significantly
mediated by sensitivity?
Is the prediction of security moderated?

irritability

Sociability

irritability

Sociability

No

No

No

No

No

Yes

Marginally

No

No
No

No
No

No
No

No
No

TABLE 6
Summary of logistic Regression Models Testing Direct, Mediated,
and Moderated Predictton of Subcategory Placement (A + Bl + 82 vs. 8384

&3
Question

Months

+ C)

6 Months

Irritability

Sociability

Irritability

Sociability

No

No

No

No

No

Yes

Yes

No

No
No

No
No

No
No

No
No

Doessensitivity significantly predict


security vs. insecurity?
Does temperament significantly predict
security vs. insecurity?
Is the effect of temperament significantly
mediated by sensitivity?
IS the prediction of security moderated?

reported more success in obtaining stability for


negative emotionality than for positive emotionality. Furthermore, low-to-moderate
stability is
what temperament theorists and behavior geneticists would predict given the conceptualization
of temperament as emerging and changing with
development (Thompson, 1986).
The fact that, in our sample, both our global
measure of maternal sensitivity and our constructed measures of infant temperament provided information about security of attachment
supports the use of an integrative perspective
for studying temperament-attachment
relations.
Our data show that maternal sensitivity and
infant temperament
exerted both direct and
indirect effects on attachment
classification
which were time and situation
dependent.
Employing
moderator and mediator models
yielded information
suggesting
that, when
mother and infant variables are considered in
conjunction, there seems to be a more complex
relationship
between the two than previous
research designs have assessed. For example,
when the effects of infant temperament
on
attachment classification were direct, they con-

tributed to the type of insecurity manifested for


insecure infants and contributed to the subcategory classification for securely attached infants.
Maternal sensitivity exerted a direct effect on
attachment security when tested independently
but also mediated the effects of infant irritability. Our data suggest that although the maternal
role is important, the importance of child characteristics in relation to the attachment relationship should not be discounted when maternal
sensitivity is low. By examining the combination of maternal and child factors, we were able
to account for situations where there are both
direct and indirect child or maternal effects on
the mother-child relationship.
These results concur with the findings of
Vaughn and colleagues (Vaughn et al. (1992)
who suggested
that the common
ground
between personality variables, such as attachment and temperament, lies in their contributions to emotional
development
and affect
modulation. To test this hypothesis, they correlated temperament and attachment measures for
six different samples ranging in age from 5 to
42 months. Low-to-moderate
correlations were

44

Susman-Stillman,

Kalkoske,

found between temperament


and attachment
measures for five of the six samples, and the
correlations increased in magnitude with the
age of the child. Although the correlations were
modest, the consistency and increase in magnitude are noteworthy because they offer empirical support for the idea that there is overlap
between these two behavioral domains. Furthermore, the hypothesis of overlap between
temperament and attachment may explain why,
in our sample, temperament tended to have a
direct effect on the insecurely attached infants.
From an attachment theory perspective, sensitive mothers are responsive to their childrens
needs, which helps infants modulate affective
expression and arousal. Mothers who are low in
sensitivity, then, are likely to be less effective at
helping their children regulate their arousal.
Without that support, the infants predisposition
to variations or extremes in emotional expressiveness
plays
a more obvious
role in
mother-child
interactions.
Calkins and Fox
( 1992) made a similar argument to explain the
relation
between
behavioral
inhibition
and
attachment, suggesting that children who have
difficulty modulating fear and who cannot rely
on their attachment figures for help may develop a response to fear that is exacerbated by the
lack of sensitivity on the part of their caregivers.
Detecting some of the nuances that are part
of the mother-child relationship, such as conditions under which the contributions
of infant
characteristics
are recognized,
may depend
upon integrative research designs. Other studies
using research strategies tapping both mother
and infant contributions
have yielded similar
results. For example, van IJzendoorn, Goldberg,
Kroonenberg, and Frenkel (1992) tested the relative effects of maternal and child contributions
to the mother-child
relationship. Conducting a
meta-analysis
of attachment
studies using a
variety of clinical populations, they found that
atypical distributions of attachment classifications were more likely in samples with maternal problems than in samples with child problems. However, they did not disregard the role
of the child in the formation of the attachment
relationship.
They concluded
that, although
maternal problems are stronger determinants of
attachment security than child problems, child
the way in which
problems may influence
insecurity is expressed (pp. X53-854). Data
from Del Carmen et al. (1993), showing a

Egeland, and Waldman

mother-infant
dyadic variable (distress management) to be a stronger predictor of attachment than either maternal or infant variables,
reiterated the need to consider mother and child
variables in combination.
Although the measurement of temperament
may change in important ways over time, the
differential
model significance
from 3 to 6
months could also be suggesting that the joint
effects of temperament and sensitivity on the
attachment relationship may undergo a developmental change between 3 and 6 months. At
3 months, there was an interaction between irritability and sensitivity. Contrary to our a priori
hypothesis, follow-up tests on the interaction
indicated that the effects of maternal sensitivity were moderated by the level of irritability
such that high levels of sensitivity increased
the likelihood of secure attachment but only
among less irritable infants. Crockenberg and
McCluskey ( 1986) examined patterns of motherinfant interaction for irritable and less irritable
infants, finding a negative relation between 3month maternal responsiveness, prenatal maternal responsiveness,
and separation crying for
the less irritable infants. Their result fits with
our interaction finding which suggests that high
sensitivity predicts security of attachment for
the low-irritable infants only.
However, this interaction was not evident at
6 months. By 6 months, maternal sensitivity
independently contributed to attachment security and served as a mediator of the effects of
irritability. A transformation
of the interaction
effect into a mediation effect in the context of
an increase in the importance
of maternal
sensitivity may be occurring. In other words,
maternal sensitivity emerges as a pivotal mechanism through which irritability influences the
attachment
relationship.
The
association
between
irritability
and insensitivity
may
evolve as the mother-child
relationship
progresses during the infants 1st year of life
(Crockenberg & McCluskey, 1986).
These disparate models may also suggest
that predictions to attachment from measures of
the infant-mother
relationship
during early
infancy may be less successful simply because
the relationship is not stable. Stabilization
of
the attachment relationship in the latter half of
the 1st year may be enhanced by the contribution of sensitivity from 3 to 6 months and by
increasing maturation of the infant. The media-

Tempramant,

Sensitivity, and Attachment

tion finding implies that, when irritability does


influence security of attachment, it may lead
some mothers to respond to their infants in less
sensitive ways. Other evidence regarding the
relation between maternal behavior and infant
characteristics
(Crockenberg
& McCluskey,
1986; van den Boom, 1989) supports this finding, suggesting that changes in maternal behavior during the 1st year of life towards unresponsivity are more easily detected in mother-infant
pairs that include an irritable infant.
We did find support for direct relations
between temperament, type of insecurity, and
subcategory
placement.
Within the insecure
group at 3 months,
sociability
predicted
avoidant attachment and differentiated the A +
Bl + B2 and B3 + B4 + C subgroups. Within
the insecure group at 6 months, irritability tended to predict resistance and differentiated the
B3 + B4 + C and A + Bl + B2 subgroups.
These results offer strong support for temperament playing a role in determining the type of
insecurity manifested and replicate the work of
Belsky and Rovine (1987), suggesting that temperament contributes to subcategory placement
and not to security of attachment. In addition,
these findings may help to explain Braungart
and Stifters (1991) observation that the subcategory groups show a variety of reactivity and
regulatory patterns.
Goldsmith et al. (1986) noted that a weak
theme present in temperament-attachment
literature is the idea that others tend to perceive
avoidant infants as possessing positive temperamental characteristics
early in life. Both
Bradshaw, Goldsmith, and Campos (1987) and
Rieser-Danner,
Roggman,
and Langlois (as
cited in Goldsmith et al., 1986) found support
for a positive relation between positive mood
and avoidant attachment. In addition, Lewis
and Feiring (1989) found a relation between
3-month sociability and avoidant attachment.
Sagi, Lamb, and Gardner (1986) found stranger
sociability to be a correlate of both secure
attachment and avoidant attachment, and that
levels of distress-proneness
were independently
correlated with variations in stranger sociability
for resistantly attached infants. Calkins and Fox
(1992) found avoidant infants to be less inhibited than resistant infants. Our findings are
consonant with the literature, but it is unclear
how early sociability may relate to avoidant
attachment. Under certain caregiving circum-

45

stances, such as low levels of maternal sensitivity, avoidant infants may look sociable early in
life in their quest to elicit sensitive care from
their attachment figures. If their efforts at garnering care are unsuccessful,
the infants may
orient towards other sources, such as objects, to
help them modulate their arousal. Alternatively,
Belsky et al. (1991) offered the suggestion that
attachment security is associated with developmental processes of affect regulation with
regard to both positive and negative emotionality (p. 430). More research is needed to understand how positive temperamental
characteristics influence
the developing
mother-child
relationship.
The finding of a direct relation between irritability
and
anxious-resistant
attachment,
although weak, is consonant with low-to-moderate empirical findings that the more irritable
infants tend to be resistantly attached (Egeland
& Farber, 1984; Goldsmith & Alansky, 1987).
The lack of clear statistical significance in our
data could be due to lower reliability in the
measure of irritability at 6 months.
Other factors in conjunction with infant temperament, such as current mental representations of past and present attachment relationships (primarily
with parents), and specific
insensitive maternal behaviors, such as rejection, intrusiveness,
or inconsistent caregiving,
may contribute to type of insecurity (Bohlin
et al., 1989; Fonagy, Steele, & Steele, 1991;
Isabella, 1993; Lewis & Feiring, 1989). Our
use of a global measure of sensitivity precluded
us from examining specific maternal behaviors
that may have, along with irritability or sociability, influenced the type of insecurity manifested. Our measure of maternal sensitivity,
however, did not significantly contribute to the
type of insecurity, although aspects of infant
temperament did. Future research should examine parents current mental representations
of
their past and present attachment relationships,
specific maternal behaviors, and infant characteristics in combination
to better understand
how insecurity is manifested.
It is important to note a few caveats when
examining our models. Although the reported
model x* suggested that the models were significantly better than chance, the -2 log likelihood ~2 statistic suggested that the models
were also significantly
different from perfect
models. Therefore, we consider these models to

46

Susman-Stillman,

Kalkoske,

be a good start towards understanding how both


infant and maternal variables influence
the
attachment relationship. Future research should
incorporate other variables which have been
shown to account for change in attachment
classification,
such as life stress or maternal
social support and infant reactivity and regulation, which can help explain individual
differences
(Crockenberg,
1981; Rothbart
&
Derryberry, 1981; Vaughn, Egeland, Sroufe, &
Waters, 1979). Researchers should also consider running analyses by sex, because there is
suggestion in the literature that gender influences maternal behavior toward infants. In
addition, we did not have a multimethod measure of sensitivity. A more refined measure of
maternal sensitivity would enable researchers
to better understand which aspects of this construct influence security and type of insecurity.
Another caution in interpreting our findings is
the decrease in internal consistency in our measures of irritability and sociability from 3 to 6
months. This may be due partly to lower reliabilities for the 6-month observational
measures.
Although we were able to obtain significant
results for irritability at 6 months, it is possible
that the lower alpha for sociability made it difficult to obtain significant results for sociability at
6 months. Alternatively, our measurement of the
sociability construct may not be tapping a clear,
unitary construct because our measure of sociability did not distinguish between sociability in
different contexts (e.g., with family members vs.
strangers). Decreases in mean levels of sociability from 3 to 6 months could also have influenced
the ability of the 6-month sociability measure to
predict attachment.
In conclusion, the finding of both direct and
indirect effects of infant temperament and maternal sensitivity on attachment security validates
our adoption of a more complex conceptualization of the relations of these variables and
encourages the use of these types of models.
Furthermore, they point to the necessity and utility of an integrative approach to examining the
relation between temperament and attachment.
REFERENCES
Ainsworth,

M., Blehar. M., Waters, E., & Wall. S. (1978).


Parremsofattachment. Hillsdale. NJ: Erlbaum.
Baron, R.M., & Kenny. D.A. (1986). The moderatormediator variable distinction in social psychological
research: Conceptual, strategic. and statistical con-

Egeland, and Waldman


siderations.
Psychology,

Bates,

Journal
of Personality
51, 1173-l 182.

and

Sock1

J.E., Maslin, C.A., & Frankel,


K.A. (1985).
Attachment security, mother-child
interaction, and
temperament as predictors of behavior problems at
age three years. In I. Bretherton & E. Waters (Eds.),
Grou?ng points in attachment theory and research.
Monographs
of the Society for Research in Child
Development, 50, (l-2, Serial No. 209).

Belsky, J., Fish, M., & Isabella, R.A. (1991). Continuity and
discontinuity in infant negative and positive emotionality: Family antecedents and attachment consequences. Developmental Psychology, 27.42143
1.
Belsky, J., & Ravine, M. (1987). Temperament and attachment security in the strange situation: An empirical
rapprochement. Child Development, 58,787-795.
Bohlin, G., Hagekull, B., Germer, M., Andersson, K., &
Lindberg, L. (1989). Avoidant and resistant reunion
behaviors
as predicted
by maternal
interactive
behavior and infant temperament. Infant Behaviorand Development,

12, 105-l 17.

Bradshaw, D.. Goldsmith, H.H.. & Campos, J.J. (1987).


Attachment, temperament,
and social referencing:
Interrelations
among three domains of affective
behavior. InfantBeharxior and Development.
IO,
223-231.

Braungart, J., & Stifter, C. (1991). Regulation of negative


reactivity
during
the strange
situation:
Ten.perament and attachment in l2-month-old
infants.
Infant

Calkins.

Behavior

and Del,elopment,

14, 349-364.

S.D., & Fox, N.A. (1992). The relation among


infant temperament, security of attachment, and behaviorial inhibition at 24 months. Child Develop
ment, 63, 1456-1472.

Carey,

W.B. (1970). A simplified method for measuring


infant temperament.
Journal
of Pediutrics,
77,
188-194.
Crockenberg,
S.B. (1981).
Iniant irritability,
mother
responsiveness, and social support influences on the
security of infant-mother
attachment. Child Dewlopment, 52, X57-865.

Crockenberg,
S.B., & McCluskey, K. (1986). Change in
maternal behavior during the babys first year of
life. Child Development, 57. 746-7.53.
Del Carmen, R.. Pedersen, F.A., Huffman, L.C., & Bryan,
Y.E. (1993). Dyadic distress management predicts
subsequent security of attachment. Infant BrhnlYor
und Development, 16, I3 I-147.
Egeland, B.. & Farber, E. ( 1984). Infant-mother attachment:
Factors related to its development and changes over
time. Child Dewlopment. 55, 7.53-77 I.
Fonagy, P.. Steele, H., & Steele, M. (199 I ). Maternal reprcsentations of attachment during pregnancy predict
the organization of infant-mother
attachment at one
year of age. Child Development, 62, 891-905.
Fox, N., Kimmerly, N., & Schafer, W. ( I99 1). Attachment
to mother/attachment
to father: A meta-analysis.
Child De\,elopment, 62, 2 I O-225.
Frodi, A., Bridges, L., & Shonk. S. (1989). Maternal correlater of infant
temperament
ratings
and 01
infant-mother
attachment:
A longitudinal
study.
Injant

Frodi,

Mental Health Jortrnol.

IO. 273-289.

A., & Thompson,


R. (1985). Infants affective
responses in the strange situation: Effects of prcma-

Temperament,
turity

and quality

of attachment.

Senrith rity, and Attachment

Child Develop-

ment, 56, 1280-1291.

Goldsmith, H.H., & Alansky, J.A. (1987). Maternal and


infant temperamental
predictors of attachment: A
meta-analytic
review. Journal
of Clinical
and
Consulting Psychology,

Temperament and social interaction in infants and


children (Vol. 31). San Francisco: Jossey-Bass.

Goldsmith, H.H., Buss, A., Plomin, R., Rothbart, M.K.,


Thomas, A., Chess, S., Hinde, R.A., & McCall,
R.B. (1987). Roundtable:
What is temperament?
Four approaches. Child Development, 58,505-529.
Goldsmith, H.H., & Rieser-Danner,
L. (1990). Assessing
early temperament.
In CR. Reynolds & R.W.
Kamphaus (Eds.), Handbook of psychological and
educational
assessment of children:
Personality.
behavior, & context. New.York: Guilford.

Gunnar, M.R., Mangelsdorf, S., Larson, M., & Hertsgaard,


L. (1989). Attachment, temperament, and adrenocortical activity in infancy. Developmental Psychology, 25.355-363.

Hosmer, D.W., Jr., & Lemesnow, S. (1989). Applied logistic regression. New York: Wiley.
Isabella, R.A. (1993). Origins of attachment:
Maternal
interactive behavior across the first year. Child
Development, 64, 60542 1.
Izard, C.E., Haynes, M.O., Chisholm, G., & Baak, K.
(1991). Emotional determinants
of infant-mother
attachment. Child Development, 62,90&917.
Kagan, J. (1984). The nature of the child. New York: Basic
Books.
Lewis, M., & Feiring, C. (1989). Infant, mother, and
mother-infant
interaction behavior and subsequent
attachment. Child Development, 60,831-837.
Mangelsdorf, S., Gunnar, M., Kestenbaum, R., Lang, S., &
Andreas, D. (1990). Infant proneness-to-distress,
temperament,
maternal personality,
and motherinfant attachment: Associations and goodness of fit.
Child Development, 61, 820-83 1.
Miyake, K., Chen, S., & Campos, J.J. (1985). Infant temperament, mothers mode of interaction, and attachment in Japan: An interim report. In I. Bretherton &
E. Waters (Eds.), Growing points of attachment theory and research. Monographs
Research in Child Development,

of the Society for


50 (No. l-2, Serial

No. 209).
Rothbart, M.K. (1986). Longitudinal observation of infant
temperament.
Developmental
Psychology,
22,
356-365.

Rothbart, M.K., & Derrybetry, D. (1981). The development


of individual differences in temperament. In M.E.
Lamb & A.L. Brown (Eds.), Advances in developmental psychology. Hillsdale, NJ: Erlbaum.
Sagi, A., Lamb, M.E., & Gardner, W. (1986). Relations
between Strange Situation behavior and stranger
sociability
among infants on Israeli kibbutzim.
Infant Behavior

Sroufe,

L.A. (1985). Attachment classification


from the
perspective of infant-caregiver
relationships
and
infant temperament. Child Development, 56, I-14.
Sroufe, L.A., & Waters, E. (1982). Issues of temperament
and attachment. American Journal of Orthopsychiatry, 52, 743-746.

55, 805-8 16.

Goldsmith, H.H., Bradshaw, D.L., & Rieser-Danner, L.A.


(1986). Temperament as a potential developmental
influence on attachment. In J. Lemer & R. Lemer
(Eds), New directions for child development.

and Development,

9, 27 1-282.

47

Thompson, R.A. (1986). Temperament, emotionality, and


infant social cognition. In J. Lemer & R. Lemer
(Eds.), New directions for child development:
Temperament and social interaction in infants and
children (Vol. 31). San Francisco: Jossey-Bass.

Thompson, R.A., Connell, J.P., & Bridges, L.J. (1988).


Temperament,
emotion,
and social interactive
behavior in the Strange Situation: A component
process analysis of attachment system functioning.
Child Development, 59, 1102-l 110.
Thompson, R.A., & Lamb, M. (1984). Assessing qualitative dimensions
of emotional responsiveness
in
infants: Separation reactions in the Strange Situation. Infant Behavior and Development, 7,423+45.
Tinsley, H.E.A., & Weiss, D. (1975). Interrater reliability
and agreement of subjective judgments. Journal of
Counseling Psychology, 22.358-376.
van den Boom, D.C. (1989). Neonatal irritability

and the
development of attachment. In G. Kohnstamm, J.E.
Bates, & M.K. Rothbart (Eds.), Temperament in
childhood. New York: Wiley.
van IJzendoom, M.H., Goldberg, S., Kroonenberg, P.M., &
Frenkel, O.J. (1992). The relative effects of maternal and child problems on the quality of attachment:
A meta-analysis
of attachment in clinical samples,
Vaughn,

Child Development,
B.E., Crichton,

63, 840-858.

L., & Egeland,


B. (1982).
Individual differences in caregiving during the first
6 months of life: Antecedents in maternal and infant
behavior during the newborn period. Infant Behavior and Development,

Vaughn,

5.77-95.

B.E., Egeland, E., Sroufe, L.A., & Waters, E.


(1979). Individual
differences
in infant-mother
attachment at twelve and eighteen months: Stability
and change in families under stress. Child Development, SO, 971-975.

Vaughn,

B.E., Lefever, G.B., Seifer, R., & Barglow, P.


(1989). Attachment behavior, attachment security,
and temperament during infancy. Child Development, 60,728-737.

Vaughn, B.E., Stevenson-Hinde, J., Waters, E., Kotsaftis, A.,


Lefever, G.B., Shouldice, A., Trudel, M., & Belsky, J.
(1992). Attachment security and temperament in
infancy and early childhood: Some conceptual clarifications. Developmental Psychology, 28,463473.
Vaughn, B.E., Taraldson, B., Crichton, L., & Egeland, B.
(1980). Relationships between neonatal behavioral
organization and infant behavior during the first year
of life. Infant Behavior and Development, 3.47-66.
Weber, R.A., Levitt, M.J., & Clark, M.C. (1986). Individual
variation
in attachment
security
and Strange
Situation behavior: The role of maternal and infant
temperament. Child Development, 57, 5M5.
05 July 1994;

Revised 27 February

1995

You might also like