Maternal Preconceptions About Parenting Predict Child Temperament, Maternal Sensitivity, and Children's Empathy

ubiquitousstrumpetMobile - Wireless

Dec 10, 2013 (3 years and 6 months ago)


Developmental Psychology

© 2004 American Psychological Association

November 2004
, Vol.
, No.
, p

text PDF

Maternal Prec
onceptions About Parenting
Predict Child Temperament, Maternal
Sensitivity, and Children's Empathy


Amanda J.

JoAnn L.


Department of Psychology, University of Denver

Department of Psychiatry, University of Colorado Health
Sciences Center

Address for Correspondence:
Lisa Kiang, the Center for Culture and Health, Neuropsychiatric
Institute, University of California, Los Angeles, 760 Westwood Plaza, Box 62, Los Angeles,
CA 90024


This material is based on
work supported by National Science Foundation Grant 0091430 to
JoAnn L. Robinson and by Colorado Trust Grant 93059 to David Olds. We thank Robert
Emde and Carolyn Zahn
Waxler for their comments on a draft of this article.

September 9, 2003.

May 17, 2004.

June 11, 2004.



Empathic Development

Maternal Preconceptions About Parenting

Temperament and Empathy

Maternal Sensitivity and Empathy

Gender, Ethnicity, and Empathy

The Present Study





Maternal preconceptions

Difficult temperament

Maternal sensitivity

Empathy indices


Means and Correlations

Group Modeling Analyses

Models Predicting Empathy

Mediating Effects of Child Difficult Temperament and Maternal



Figure 1

Figure 2

Figure 3

Figure 4

Table 1

Table 2


This study examined the influence of maternal preconceptions on child
difficult temperament at 6 months and maternal sensitivity
at 12

15 months
and whether all 3 variables predicted children's empathy at 21

24 months.
Within a low
income, ethnically diverse sample of 175 mother
child dyads,
path models were tested with 3 empathy indices (prosocial, indifference,
inquisitive) as out
comes. Results indicated that maternal preconceptions
significantly predicted child difficult temperament, maternal sensitivity, and
children's empathy. Temperament mediated the link between maternal
preconceptions and inquisitiveness, and maternal sensiti
vity mediated the
link between preconceptions and prosocial responses. Group modeling
techniques revealed no significant differences across gender or ethnicity.
Correlations suggested contextual effects based on the familiarity of the
person in distress. T
he implications and utility of developing parenting
interventions are discussed.

Empathy is central in the emotional bond between individuals. In this study,

is defined as the expression of care and concern for others in
distress with the goal

of relieving that distress (
Batson, Duncan, Ackerman,
Buckley, & Birch, 1981
Davis, 1983
Robinson, Zahn
Waxler, & Emde,
). This conceptualization views empathy as a pattern of behaviors that
convey meaning to others and the self. Although empathy is often
conceptualized as expressed concern or prosocial behavior directed toward
the person in distress (e.g.,
Robinson et al., 1994
), early in development,
empathy is variously conveyed through several channels of expression. That
is, young children may express both empathic concern

and inquisitiveness in
response to a distressed individual, or they may be initially inquisitive but
then convey indifference by re
engaging in play or ignoring the distressed
person (e.g.,
Waxler, Robinson, & Emde, 1992
Waxler, Schiro,
Robinson, Emde, & Schmitz, 2001
). In this study, we consider the
distinctiveness of children's empathic responses by examining three aspects
of empathy, namely, prosocial or concerned responses, responses conveying
indifference or lack of empathy, and degree of inquisitiveness or attempts to
make cognitive sense of the
distress situation.

Research has traditionally focused on two predictors of empathy in early
childhood: child temperament and maternal sensitivity. The possibility that a
third variable, maternal preconceptions about parenting, can serve as a
precursor not

only to empathy but also to these more proximal predictors of
empathy has not, to our knowledge, been explored.

Maternal preconceptions about parenting represent enduring features of a
child's caregiving environment, comprise attitudes that tend to be sta
ble and
more resistant to change than actual parenting behaviors, and can affect
parents' accuracy in perceiving their children's cues and responding
appropriately (
Holden, 1995
De Lisi & Sigel, 1995
Eisenberg, & Harris, 1991
). Such preconceptions embody value systems and
underlying beliefs most likely reflecting a person's own experience with
being parented (
Holden, 1995
). This, in turn, can translate into how a parent
perceives his or her infant's behavior and can also impact a parent's
subsequent sensitivity or responsiveness.

Although parenting attitudes may be susceptible to transactional
stemming from the birth and characteristics of the child (
De Lisi
& Sigel, 1995
), assessing preconceptions helps rule out the possibi
lity that
one's parenting attitudes are a function of the parent
child relationship. An
association between preconceptions and empathy, particularly one that
remains strong after the behavioral indicator of maternal sensitivity is
accounted for, would indi
cate the unique significance of the parent's
orientation and attitudes, which are likely mediated both genetically and
environmentally. Examining preconceptions adds a potentially important
piece to the existing literature on empathy development, which pri
assumes that empathic responses are learned through direct teaching and
empathic interactions with the child.

The primary goal of this study was to use longitudinal data to determine
whether maternal preconceptions (that is, prenatal maternal attitu
des and
beliefs about parenting and children) predict child difficult temperament,
maternal sensitivity, and children's empathy. We also examined previously
documented links between child temperament and empathy, and between
maternal sensitivity and empath
y, and investigated potential mediating

Empathic Development

Empathic development generally follows a normative sequence progressing
from dysregulated and “egocentric” responses in early infancy to those that
are more regulated and
oriented in early childhood (
Hoffman, 1975
Waxler & Radke
Yarrow, 1982
). Between 20 and 24
onths of age, in addition to marked changes in cognitive and linguistic
abilities, self
regulation, self
awareness, and the appreciation of standards
and morality (
Kagan & Lamb, 1987
Waxler & Radke
Yarrow, 1990
children undergo a developmental transition in which empathic behaviors
communicating concern a
nd a desire to alleviate another's distress can be
observed. This transition from infancy to early childhood reflects an
important, yet understudied, developmental period in which to examine
empathy, in that empathic and altruistic responses are just begin
ning to
emerge in a child's behavioral repertoire (
Robinson et al., 1994

Socialization and contextual factors have been found to influence empathic
ions, implicating the role of the relatedness of the “victim” (see
Eisenberg & Mussen, 1989
, for a review).
Robinson, Zahn
Waxler, and Emde

found that empathic responses in children 14

36 months were
meaningfully differentiated by relational context. Children,

on average,
exhibited higher levels of arousal when responding to an examiner's distress,
but they exhibited higher levels of approach, inquisitiveness, and prosocial
behaviors when responding to their mother's distress, suggesting that
empathy and concer
n may be more readily observed when the person in
need is familiar and loved. Given the potential for contextual effects, it is
important to investigate children's empathy as a function of familiar
(responses to mother) and unfamiliar (responses to examine
r or stranger)

Maternal Preconceptions About Parenting

Although little work has focused on the impact of expectant mothers'
preconceived notions about parenting on children's development, the
general influence of maternal attitudes on actual p
arenting behaviors and
children's adjustment has long been an area of interest (e.g.,
Benasich &
Gunn, 1996
Daggett, O'Brien, Zanolli, & Peyton, 2000
Schaefer & Bell, 1958
). Although some disagree (
Schilmoeller, & Higgins, 1990
Holden & Edwards, 1989
), literature suggests
that attitudes about children and childrearing influence the way parents
interact with and understand their childr
en (
Bavolek, 1989
Lisi & Sigel, 1995
Schaefer, 1991
Sigel, 1986

Research has typically focused o
n the positive impact of sensitive, adaptive
attitudes or the negative impact of insensitive, maladaptive attitudes. For
instance, parenting attitudes that value independence, initiative, and
curiosity have been positively correlated with children's academ
performance and competence (
Schaefer & Edgerton, 1985
). Similarly,
constructionist parenting attitudes that value children's experimentation and
on have been linked with children's cognitive competence
De Lisi, 1985
Seifer & Sameroff, 1987
). Further, much of the
attachment literature focuses on the role of adaptive parenting attitudes
related to maternal warmth and sensitivity on children's later adjustment
and behavior, including empathy (
Benasich & Brooks
Gunn, 1996
& Robinson, 1991
Thompson, 1999
). In contrast, negative parenting
attitudes, such as the belief that children are “born bad,” have been
associated with negative parenting behav
iors such as excessive
authoritarianism, “breaking the will of the child,” reprimands, and physical
punishment (
Kochanska, Kuczynski, & Radke
Yarrow, 1989
Schaefer, 1991
Schaefer & Edgerton, 1985
). Research has also shown that mothers with
more negative parenting attitudes had more negative interpretations of their
s' behavior and perceived them as more irritable (
Chen &
Luster, 1999
Daggett et al., 2000

Much of the existing research has examined parenting attitudes as
concurrent outcomes differentiating risk groups, often in light of child abuse
and neglect. Several studies have shown adolescent mothers to have more
ative attitudes than older mothers (
East, Matthews, & Felice, 1994
Haskett, Johnson, & Miller, 1994
); however, results are inconsistent
Baranowski et al., 1990
). Findings are also inconsistent with regard to
differences across ethnicity and socioeconomic status. Some studies
have found ethnic minority and low
income statuses as risks regarding child
outcomes (
t et al., 1994
Lutenbacher, 2001
), and some studies have
yielded no effect (
Glass, 1983

Venturing beyond group differences and concurrent assessment of parenting
attitudes, we expected to shed light on the predictive influence of parenting
preconceptions, that is, attitudes measured prenatally, on outcomes suc
h as
children's empathy. It is reasonable to expect this association because
preconceptions, although not directly linked to children's empathy, have
been shown to impact documented predictors of empathy, namely,
temperament and maternal sensitivity.

rch supports that both observed infant temperament and subsequent
maternal responsiveness can be predicted by parenting attitudes prenatally
reported by expectant mothers (
Crockenberg & Smith, 1982
Frodi, Bridges,
and Shonk (1989)

found prenatal childrearing attitudes to predict maternal
ratings of infants' dif
ficult temperament at 4 months, and links between
maternal preconceptions and maternal warmth and sensitivity have also
been documented (
Crockenberg & McClus
key, 1986
Moss & Jones, 1977
The assumption is that an expectant mother's preconceived notions about
parenting and children persist into subsequent mother
child intera
ctions as
well as her perception of her child (see
Glass, 1983
). Indeed, research on
the intergenerational transmission of attachment has shown that expectan
mothers' working models of attachment predicted their infants' later security
in the Strange Situation (
Fonagy, Steele, & Steele, 1991
Steele, Steele, &
Fonagy, 1996
Ward & Carlson, 1995
), implicating the role o
f maternal
preconceptions and characteristics in predicting maternal sensitivity and the
quality of the subsequent mother
infant relationship.

Temperament and Empathy

Child temperament

is generally conceived as a child's typical mode of
response that is influenced by biological factors and contextually conditioned
(i.e., heritability and developing abilities to regulate emotion), as well as
influenced by ongoing socialization experience
s (
Rothbart, Ahadi, & Evans,
). Our hypothesis was that maternal preconceptions will influence
children's temperament and that temperament, in turn, will

exert an
influence on children's empathy.

Research has linked greater empathy to different aspects of temperament,
including the following: (a) behavioral inhibition and fearfulness (
Ahadi, & Hershey, 1994
Young, Fox, & Zahn
Waxler, 1999
), (b) sociability
or positive engagement (
Robinson et al., 1994
Waxler, Cole, Welsh, &
Fox, 1995
), and (c) observed
negative emotional reactivity (
Eisenberg et al.,
inson et al., 1994
Rothbart et al., 1994
Young et al., 1
however, the direction of effects depends on how and when temperament is
assessed. Two studies are noteworthy in finding contemporaneous
empathy links.
Young et al. (1999)

found a negative
association between behavioral inhibition and empathy toward an unfamiliar
adult at 2 years, and
et al. (1994)

found that greater fearfulness at
age 7 was related to less concurrently reported empathic behaviors.

However, when considering early temperament in relation to later empathy,
a more differentiated picture emerges.
Young et al. (1999)

found that
children who were high
motor as infants (a precursor of
uninhibited behavior) were later significantly more empathic toward a
stranger than

motor infants, but infants' behavior was not
related to later empathy toward mother. In a study of toddler twins,
Robinson et al. (1994)

d high levels of sociability reported by parents at
14 months to be associated with boys attaining high levels of empathy by 20
months. In addition, they found that high negative emotionality at 14
months was associated with maintaining high levels of empa
thy at 14 and 20
months. However,
Eisenberg et al. (1996)

demonstrated that preschool
aged children with low levels of negative emotionality showed high leve
ls of

(defined as an emotional response to another's emotional state
consisting of feelings of sorrow or concern) in response to a distressing film
across repeated assessments at ages 4

6 years. This finding was especially
true for boys.

These stu
dies suggest that sociable, fearful, and distress features of
temperament are associated with empathic responses, and that children with
early developing sociable or uninhibited temperaments displayed more
empathy later in development. Toddlers expressing
more fear were also
found to be less empathic to the stranger. In another study, more negative
emotional expressiveness (including anger and distress) was associated with
more empathy. However, older preschoolers with less negativity showed
more empathy la
ter in development, underscoring the role that emerging,
regulatory behaviors might play.

High negative reactivity, addressed in previous research by
senberg et al.
, can be considered one component of difficult temperament.
However, the construct of difficult temperament encompasses an additional
constellation of characteristics, including short latencies to react and minimal
ability to regulate
emotions, even with support from a caregiver (
Freeland, & Lounsbury, 1979
), and has not yet been implicated in empathy
research. We hypothesized that d
ifficulties in self
regulating emotional
experiences would be associated with less prosocial or empathic concern,
greater indifference, and less inquisitiveness. Although no research has
demonstrated links between difficult temperament and empathy, there i
substantial literature linking difficult temperament in early childhood with
disruptive behavior problems that could be partially characterized as
Rutter (1989)

Moffit and Caspi (2001)

found that children
who developed aggressive or antisocial tendencies in middle childhood and
adolescence were
rated as having difficult temperaments in early childhood.

Maternal Sensitivity and Empathy

Maternal sensitivity has also been linked to early empathic development
Robinson & Little, 1994
Robinson et al., 1994
). Although it

is not entirely
clear how maternal sensitivity translates into children's ability to empathize
with others, one might argue that children learn to be empathic through
social learning and modeling effects and exposure to empathic and sensitive
behaviors of

caregivers. It has been shown that securely attached children
display more empathy toward others than insecurely attached children (e.g.,
Kestenbaum, Farber
, & Sroufe, 1989
), presumably as a result of social
learning. Consistent with basic tenets of attachment theory, attachment
researchers argue that children with a secure relationship with their primary
caregiver regularly experience parental empathy to the
ir own distress (
der Mark, van IJzendoorn, & Bakermans
Kranenburg, 2002

The few studies that have examined parental warmth and sensitivity in
to children's empathy have documented positive associations (e.g.,
Robinson et al., 1994
Waxler & Radke
Yarrow, 1990
Yarrow, & King, 1979
), with the notable exception of
van der Mark et
al.'s (2002)

study. In that study, however, maternal behavior was measured
as “sensitive structuring” during a cognitive puzzle task, a context with
limited opportunity to reveal a mother's attention to her child that would

required sensitive or empathic behavior. We expected to find an
association in the current study given our assessment of maternal sensitivity
during a free play situation.

Gender, Ethnicity, and Empathy

Previous research has documented gender differences in the observed
expression of empathy (see
Brody, 1985
, for a review;
Eisenberg & Fabes,
Hoffman, 1977
Jensen, Peery, Adams, & Gaynard, 1981
Robinson et
al., 1994
Waxler et al., 1992
). Specifical
ly, females have been found
to be generally more emotionally expressive than males, exhibiting greater
prosocial behaviors, concern, hypothesis testing, and less indifference
toward a victim in distress. However, these differences were not dramatic,
and it

is possible that males convey caring and concern through more
instrumental actions, rather than affective displays.

Venturing beyond mean differences, gender is important to consider in
generating predictive models of empathy, given that boys and girls ha
been shown to approach social interactions and interpersonal relationships
differently (see
Benenson, 1996
). The development of empathy and
associated outc
omes can take different routes and have different
consequences for boys and girls. For instance, the exaggeration of empathy
in girls whose mothers are clinically depressed can potentially increase girls'
risk for mental health problems (
Waxler, Cole, & Barrett, 1991
whereas deficits in empathy, particularly in boys, are associated with later
disruptive behavior and conduct problems (
Cohen & Strayer, 1996

Much of the research documenting gender
differences has predominantly
used advantaged, Caucasian samples. In the current study, we sought to
replicate these gender differences in a low
income, ethnically diverse group.
The use of ethnically homogeneous samples in previous research also begs
question of whether results may be generalized to more diverse ethnic
groups. Given cultural differences in the socialization of children's emotions
and behaviors in terms of individualistic versus collectivistic values (
& Fingerson, 2003
), it is important to explore differences and similarities in
empathic development in children from a variety of ethnically diverse

Our major goal in exam
ining gender and ethnic group differences was to
determine whether there may be different predictors of empathy for boys
and girls and for ethnic minority children. Within a culturally diverse sample,
we hoped to determine whether separate models should be

used to explain
how empathy develops across gender and ethnic group.

The Present Study

The goal of this study was to examine the impact of maternal preconceptions
(negative, maladaptive attitudes about parenting and children, prenatally
assessed) on later child temperament and maternal sensitivity and to predict
children's empathy from all t
hree sources. Path analyses were used to model
the two hypotheses shown in
Figure 1
: Maternal preconceptions will predict
maternal report of child difficult temperament a
t 6 months, observed
maternal sensitivity at 12

15 months, and observed child empathy at 21

months (Hypothesis 1), and difficult temperament and maternal sensitivity
will mediate the association between preconceptions and child empathy
(Hypothesis 2).

Enlarge this image.

Conceptual model of relations. (1) = Hypothesis 1; (2) = Hypothesis 2

We examined three indices of empathy, namely, prosocial empathy or
concern, indifference or lack of empathy, and inquisitiveness. These indices
are important to consider independently because of the distinctiveness of
children's empathic responses, that is
, prosocial empathy can be seen to
convey a highly affective and other
oriented response to another's distress,
indifference indicates a lack of empathic concern, and inquisitiveness is an
early cognitive indicator of empathy. Although we examined these in
independently, we did expect them to be moderately correlated.

On the basis of previous research, we expected more negative maternal
preconceptions to lead to higher ratings of child difficult temperament, less
maternal sensitivity, less prosocial or

inquisitive responses, and more
indifferent responses. Less difficult temperament and greater maternal
sensitivity were expected to predict more positive empathic responses.
Given previous research documenting relational effects, analyses considered
hy toward both a familiar (mother) and unfamiliar (examiner) victim
of distress. Group modeling techniques were used to assess whether
pathways performed differently for boys than for girls, and across two ethnic
groups, Latinas and European Americans, rep
resented in our sample.



Women without health insurance and who were expecting their first birth
were recruited from antepartum clinics in metropolitan Denver. Following an
intake interview, participants were randomly assigned to e
ither one of two
treatment groups or a control group. This investigation focused only on
participants assigned to the control group (

= 255) in order to circumvent
potential intervention effects. Within this group, 5 participants declined
postnatal participation, 11 experienced a miscarriage or infant death, and 2
gave their child up for adoption. An additional 62 participants were excl
from final analyses because we required cases with complete data from all
postnatal assessments (6 months, 12

15 months, 21

24 months), resulting
in a final

of 175 (87 mothers of boys and 88 mothers of girls).

To ensure that the 32% of participants
not included because of missing data
did not differ significantly from our sample with complete data, we
conducted a series of independent samples

tests that revealed no
significant demographic differences between these groups. In addition, three
dummy v
ariables were created to represent excluded and included
participants for each of the model variables (6
month child temperament,

to 15
month maternal sensitivity, 21

to 24
month empathy). Data on
maternal preconceptions were available for all partici
pants, as this construct
was assessed at prenatal intake, and were used to predict each of the three
dummy variables in a series of logistic regressions. Regressions predicting
missing data for child temperament, χ

= 255) = 1.93,

maternal sen
sitivity, χ

= 255) = 2.69,

=.11, and empathy, χ

255) =.01,

=.93, were not significant, demonstrating no significant
differences between our sample of excluded participants and our final
sample with regard to maternal preconceptions, the m
ajor predictor in our

In this final sample, the mean age of the expectant mothers at intake was
19.7 years (

= 4.0 years). This was a low
income group, with 50% of
participants reporting, at intake, an annual household income of $7,500 or
less. A
verage household income was approximately $12,800 (

$12,000). The highest grade completed for 9.5% of the sample was 8th
grade; 38.8% completed some high school; 38.3% graduated from high
school; 11.2% completed some college; and 2.2% graduated from co
The sample was ethnically diverse, consisting of 49% Latinas (primarily
Mexican or Mexican American), 31% European Americans, 18% African
Americans, and 2% other.


These data were a subset of the larger HV2000 study, initiated by David
Olds and colleagues in 1994, investigating effects of an early intervention
program (
Olds et al., 2002
). Assessments consisting of maternal interviews,
child interactions, and child behavior observations were conducted in
study offices (at 6, 15, and 24 months) and in the family's home (at 12 and
21 months). This investigation focused on data collected at prenatal intake
(maternal preconceptions), 6 mon
ths (maternal ratings of child difficult
temperament), 12

15 months (average observed maternal sensitivity at
home and in the laboratory), and 21

24 months (average child empathy at
home and in the laboratory).

Following informed consent and prior to rando
m assignment to treatment or
control groups, an enrollment interview was conducted. The interview
covered numerous topics; of relevance to this study were data from the
Adolescent Parenting Inventory (AAPI;
Bavolek, 1984
). At the 6
lab visit, mothers were interviewed using the Infant Characteristics
Questionnaire (ICQ;
Bates et al., 1979
) to assess children's difficult
temperament. ICQ interviews were conducted as part of a 90
min visit,
either in person following the completion of developmental testing or later
via telephone. Maternal sensitivity data w
ere collected from mother
dyads at home (targeted for 12 months) and in the lab (targeted for 15
months) through a mother
child play episode. The play procedure was
conducted in the first half of each 90
min visit. A blanket was spread on the
and the mother was offered a standard set of toys. The episode served
as a necessary break for the child, and the mother was encouraged to use
the 10 min to relax and play. Age of children at the home assessment
averaged 12.9 months (range = 11.7

17.9). Ag
e of children at the lab
assessment averaged 16.0 months (range = 11.9


Empathy data were collected at two time points, once at home (targeted for
21 months) and once in the lab (targeted for 24 months), as part of a 11/2

hr visit. In each session,

children's empathic responses to naturalistic
simulations of distress by the examiner and the mother were videotaped.
Each simulation lasted 90 s and was similarly conducted whether at home or
in the lab. Examiners performed a simulation early in each vis
it and
instructed mothers to observe the manner in which they performed it. To
begin the simulation, the examiner pretended to pinch her finger in a
clipboard. First, she grasped her finger saying, “Ooh” and “Ouch” in a
realistic manner for 30 s, being car
eful not to have eye contact with the
child. Second, the examiner expressed that her injury “feels better now” and
talked about how the injury was better for 30 s. Third, the examiner re
engaged the child as the camera continued to record the next 30 s for

delayed responses.

Approximately 30 min or more later, the mother sat on the floor with her
child and played with a toy. After engaging her child with the toy, she stood
up and pretended to hurt her knee, rubbing it as she said, “Ooh” and
“Ouch.” Moth
ers were instructed not to engage their child through eye
contact or call their child's name. The examiner signaled the mother after 30
s by asking, “[Mother's name], are you okay?” At this point, the mother
expressed that the injury was feeling better and

the examiner and mother
discussed the injury for 30 s. The examiner signaled to the mother that the
procedure was complete by stating, “I'm glad it's feeling better now.” The
camera continued to record the child's behavior for an additional 30 s.
age of children at the 21
month assessment was 21.7 months
(range = 20.8

24.6), at the 24
month assessment children averaged 25.5
months (range = 22.0

30.0), and the average length of time between home
and lab visits was 96 days (

= 33 days).


Maternal preconceptions

The AAPI (
Bavolek, 1984
) is a 32
item measure designed to assess
parenting attitudes and beliefs. Items were scored on a 5
point Likert
scale (
strongly disagree

strongly agree
). Items assess a range of attitudes
about parenting and child rearing, including empathy toward children's
needs, developmentally appropriate expectations about children's abilities,
endorsement of the use of physical punishment, and parent
child role
reversal (i.e., that children should help their parents to feel better). Although
the measure consists of four subscales (Lack of Empathy, Unrealistic
Expectations, Physical Punishment, and Role Reversal), correlations among
subscales were high


.60). Because we were interested in how
negative attitudes, as a whole, influence outcomes, responses were
averaged to form an overall construct of maternal preconceptions, with
higher scores indicating higher agreement with maladaptive parenting
ttitudes. The internal consistency of the measure was good (

=.91) and
did not vary considerably across ethnic groups (range =.88


Difficult temperament

The ICQ (
Bates et al., 1979
) was used to assess maternal report of
children's difficult temperament. The ICQ is a psychometrically sound means
of measuring difficult temperament in infants, focusing on dimensions such
as fussiness and low soothability,

slow adaptation to changes in the
environment, irregularity and unpredictability of needs, and low sociability.
The full measure consists of 24 items; however, we used a 9
item short
form of the measure. Items were scored on a 7
point Likert
type scale, w
higher scores reflecting more difficult temperament. The internal consistency
obtained from our data was good (

=.73; range across ethnic groups =.72


Maternal sensitivity

child play episodes at 12 and 15 months were rated using the
tional Availability Scales (
Biringen & Robinson, 1991
Robinson, & Emde, 1994
), global scales that assess features of the mother
child interaction that reflect the dyad's closeness, rhythmicity, and both
partners' motivation to be engaged in interaction. The Sensitivity subscale
was the focus of analyses, assessing maternal behavior such as accuracy
and responsiveness to infant signals, warm and genuine affect, variety and
creativity in play, and flexibili
ty in negotiating distress or conflict situations.

Four coders were trained by JoAnn L. Robinson to a criterion of 80%
agreement in observing children within a specific age range. Ratings were
made on a 9
point scale, with higher numbers reflecting greater

Interobserver reliability was assessed for the larger sample across 62 cases
following initial criterion training and was high (intraclass


made at home (12 months) and at the lab (15 months) were significantly
correlated (


<.01) and thus were averaged to form one overall
rating of 12

15 month maternal sensitivity.

Empathy indices

Children's empathic reactions to simulated distress events were scored by
three coders trained in a recent version of a coding system developed by
Robinson and Zahn
Waxler (2002)

that was adapted from a system
originally developed at the Developmental Psychology Laboratory of the
National Institute for Me
ntal Health (
Waxler, Radke
Yarrow, Wagner, &
Chapman, 1992
), which has been reported in previous work (
Waxler, Usher, Robinson, & Bridges, 2000
Waxler et al., 1992
Ratings from 21

and 24
month simulations were aggregated to form 21

month constructs, one for the mother and one for the examiner.
Approximately 25% of the 175 cases were double
coded for reliability
purposes. Children's responses were scored on a variety of subscales
aracterized by behavioral dimensions and discrete behaviors. Subscales
were standardized and then aggregated to form the following composites:

1. The
Prosocial Index

was created to reflect positive, prosocial concern. This
index comprised the following rat
ings: global empathy (1

7 scale, higher
scores reflecting more global empathy), concern for victim (1

5 scale,
higher scores reflecting more concern), and prosocial responsiveness
(determined from the raw number of prosocial acts toward the victim,
ng helping, sharing, distracting, and verbal sympathetic behavior).
Final scores on each of these ratings were standardized and averaged to
create an overall index of prosocial empathy. Intercoder reliability for each
of these ratings ranged from.70 to.80
(intraclass correlations), and the
average reliability across all scales of the constructed index was.76.

2. The
Indifference Index

reflects negative, rejecting responses. Included
are children's callousness and hostility (1

3, higher scores reflecting mor
callousness and hostility), anger (1 =
no anger
, 2 =
presence of anger
), and
whether children engaged in ignoring or active play behavior during the
simulation (each scored 0 if behavior did not occur, 1 if behavior did occur).
These scores were standard
ized and aggregated to form an overall index of
indifference. Intercoder reliability for subscales ranged from.74 to.80, and
average reliability across scales was.78.

3. The
Inquisitiveness Index

was created by standardizing and aggregating
scores from the

hypothesis testing (1

5 scale, higher numbers reflecting
more frequent and complex hypothesis testing) and social referencing (0 =
no occurrence of behavior
, 1 =
social referencing look
, 2 =
referencing look
plus verbalization or action directed toward no
nvictim adult
) scales,
reflecting attempts by the child to gain more information about the situation.
Intercoder reliability ranged from.78 to.88, resulting in an average reliability


Means and Correlations

Table 1

displays means and standard deviations of all study variables.
Results from independent samples

tests to determine whether variable
means differed by gender are also shown in
Table 1
. We did not expect
significant gender differences with respect to maternal preconceptions,
difficult temperam
ent, or maternal sensitivity, and indeed, no significant
gender differences among these variables were found. In terms of empathy
indices, however, we did expect girls to convey greater empathy than boys,
and small, sometimes significant, differences were
found. In the sample as a
whole, inquisitiveness toward the mother differed,
(173) = 2.50,

with girls exhibiting significantly more inquisitiveness compared with boys.
Prosocial responses toward the examiner also differed,
(173) = 2.05,

<.05, w
ith girls displaying more prosocial responses than boys. Gender
differences within ethnic groups exhibited similar effects, with two
exceptions. European American girls expressed less inquisitiveness toward
the examiner compared with boys (though this diff
erence was not
significant), and Latina girls exhibited significantly more indifference toward
the examiner than did boys. In general, however, children's empathy across
the entire sample and within each ethnic group showed patterns consistent
with previou
s research, such that girls were more empathic than boys.

Enlarge this image.

Variable Means and Standard Deviations by

Correlations by child gender and for the entire sample are shown in
Table 2
Although there did not appear to be consistently different patterns, several

relations are noteworthy. Maternal preconceptions were significantly
correlated with difficult temperament and children's prosocial responses
toward mother, but only for boys. Also for boys, difficult temperament was
significantly related to indifference
toward the mother; however, this relation
was not significant for girls. In contrast, correlations among preconceptions,
maternal sensitivity, and indifference toward the mother were significant,
but only for girls. These differences support the use of gro
up modeling
techniques to determine whether the overall pattern of associations differed
significantly by gender.

Enlarge this image.

Correlations of Study Variables by Child Gender and Victim of Distress

According to correlations using the entire sample,
Table 2

shows that, in
general, empathy toward the mother was significantly associated with
maternal preconceptions, child difficult temperament, and maternal
sensitivity, although the patterns and strengths of association differed
slightly by empath
y index. It is interesting that maternal preconceptions
were more strongly correlated with maternal sensitivity than with difficult
temperament, despite shared method variance for the self
report measures
of preconceptions and temperament.

responses were correlated across mother and examiner contexts,
as were inquisitive responses. However, it is important to note that maternal
preconceptions, difficult temperament, and maternal sensitivity were only
weakly associated with empathy toward the

examiner, with the exception of
maternal sensitivity relating to indifference toward the examiner. The
generally weak associations between predictor variables and empathy
toward the examiner do not permit modeling; thus, our Results and
Discussion focus o
n children's empathy toward the mother.

Group Modeling Analyses

Group modeling techniques described by
Jöreskog and Sörbom (1996)

used to assess whether our conceptual path model performed differently for
boys and girls. The logic underlying this technique
is to compare the global
square of one model in which male and female paths are free to vary
with the global chi
square of another model in which male and female paths
are constrained to be equal. If the difference between chi
square values
indicates t
hat the free
vary model performs significantly better than the
constrained model, we can assume the model performs differently across
genders, indicating the need to model boys' and girls' data separately. If
global chi
square values are not significant
ly different, it can be concluded
that the model performs similarly across genders, allowing for the
combination of both groups into a single model.

Potential gender group differences were assessed for each of the three
empathy outcomes. Differences in chi
square values (

= 5) between the
vary and constrained models for prosocial responses, indifference,
and inquisitiveness were 3.03, 5.21, and 3.14, respectively. None of these
values met the chi
square significance value of 11.07 for

=.05, allo
wing us
to conclude that boys and girls did not exhibit significantly different
pathways. Because group differences were not found, data were combined
for further analyses.

Similar group modeling techniques were used to test potential differences
across et
hnicity. Because African Americans comprised a subset of our
sample too small to permit modeling, data from Latinas (49% of our sample)
and Caucasians (31%) were compared using the group modeling procedure
described above. Differences in chi
square values

= 5) between the free
vary and constrained models for prosocial responses, indifference, and
inquisitiveness were 6.40, 0.99, and 3.25, respectively. Again, comparisons
between these values did not meet the chi
square significance value (

= 5)
11.07 for

=.05. Thus, data across all ethnic groups were combined for
the final analyses.

Models Predicting Empathy

As shown in
Figure 2
, the path model predicting the Prosocial Index fit the
data well, χ

= 175) = 1.42,

=.23, comparative fit index (CFI) =.99,
square error of approximation (RMSEA) =.05. Examination of
standardized path coefficients reveals that maternal preconceptions
predicted difficult temperament and maternal sensitivity but only
approached s
ignificance in predicting prosocial empathy. Maternal sensitivity
was significant in predicting prosocial responses.

Enlarge this image.

Model predicting Prosocial Index. χ2(1, N = 175) = 1.42, p =.23, comparative

fit index
=.99, root
square error of approximation =.05. †p <.10. *p <.05. **p <.01

Figure 3

shows that the path model predicting indifference also provided a
good fit, as evidenced by the fit statistics: χ

= 175) = 1.42,

CFI =.99, RMSEA =.05
. Examination of path coefficients reveals that
maternal preconceptions predicted all three variables and was the only
significant predictor of indifference.

Enlarge this image.

Model predicting Indifference Index. χ2(1, N = 175) = 1.42, p
=.23, comparative fit index
=.99, root
square error of approximation =.05. **p <.01

The model of inquisitiveness (see
Figure 4
) produced similar fit statistics,

= 175) = 1.42,

=.23, CFI =.98, RMSEA =.05. Path coefficients
show that maternal preconceptions predicted maternal sensitivity and
difficult temperament but did not predict inquisitiveness. Difficult
temperament, however, did significantly predict inq

Enlarge this image.

Model predicting Inquisitiveness Index. χ2(1, N = 175) = 1.42, p =.23, comparative fit
index =.98, root
square error of approximation =.05. *p <.05. **p <.01

All relationships were in the hypothesized

directions. Further analysis of


values revealed that approximately 4% of the variance in difficult
temperament and 12% of the variance in maternal sensitivity were
accounted for by maternal preconceptions. In terms of empathy outcomes,
7% of prosocial

empathy, 11% of indifference, and 2% of inquisitiveness
were explained by model variables.

Mediating Effects of Child Difficult Temperament and Maternal

As evidenced by significant path coefficients, difficult temperament may
mediate the relation between maternal preconceptions and inquisitiveness,
and maternal sensitivity may mediate the relation between maternal
preconceptions and prosocial r
esponses. Although the direct path from
maternal preconceptions to inquisitiveness was not significant, and the direct
path from preconceptions to prosocial empathy only approached
significance, zero
order correlations fit the criteria set by
Baron and Kenny
, that is, that there should be associations between two variables and
a potential mediator in order to test for mediation.

The logic underlying test
s for mediation is to compare how the direct path
between two variables changes once the hypothesized mediated path is
removed and also to compare any changes between the global chi
values in a full (with mediator) and alternative (mediator removed)

Evidence for mediation would be found if the direct path between two
variables becomes stronger without the mediator in the model and if the full
model provides a better fit to the data compared with the model with the
mediator removed (
Baron & Kenny, 1986

Thus, to test the mediational effect of difficult temperament, we set the path
from maternal preconceptions to difficult temperament to zero and re
the path from difficult temperament to inquisitiveness. Maternal sensitivity
was allowed to remain in the model, as this variable did not qualify as a
mediator for this index of empathy. With difficult temperament removed, the
path from maternal prec
onceptions and inquisitiveness became stronger (β
= −.12) but remained nonsignificant. Comparison of global chi
values for the full model presented in
Figure 4

and the model with the
temperament path removed revealed a significant difference, Δχ
(2) =

<.05, suggesting a better fit for the model including difficult
temperament. These results indicate that child difficult temperament
partially mediated th
e relation between maternal preconceptions and

Enlarge this image.

Model predicting Inquisitiveness Index. χ2(1, N = 175) = 1.42, p =.23, comparative fit
index =.98, root
square error of approximation

=.05. *p <.05. **p <.01

Similar procedures were carried out to test the mediational effect of
maternal sensitivity on prosocial empathy. We set the path from maternal
preconceptions to maternal sensitivity to zero and removed the path from
maternal sensitivity to prosocial respon
ses. Temperament was allowed to
remain in the model, as this variable did not qualify as a mediator for this
index of empathy. With the removal of maternal sensitivity, the previously
nonsignificant path from maternal preconceptions to prosocial responses
emerged as significant (β = −.19,

<.05). Further, comparison of global
square values for the full model and the model with the maternal
sensitivity path removed revealed a significant difference, Δχ
(2) = 24.58,

<.05, with better fit statistics ass
ociated with the model including maternal
sensitivity. These results indicate that maternal sensitivity partially mediated
the relation between maternal preconceptions and prosocial empathy.


The goals of this study were to examine the effect
of maternal
preconceptions on child difficult temperament and maternal sensitivity and
to also examine how all three variables influence children's empathic
development. Direct effects of maternal preconceptions were found, such
that expectant mothers with

more negative attitudes toward parenting rated
their children higher on difficult temperament at 6 months and were less
sensitive toward their children at 12

15 months, compared with expectant
mothers with more positive preconceptions. Direct effects of m
preconceptions on children's empathy were also found such that more
negative preconceptions related to more indifference toward mothers'

Indirect effects of maternal preconceptions on children's empathy were also
found. Maternal
sensitivity mediated the relation between maternal
preconceptions and prosocial responses to mothers' distress; more sensitive
mothers had children who were more prosocial. Difficult temperament
mediated the effect of maternal preconceptions on inquisitive
ness; children
rated by mothers as having more difficult temperaments were less

Our models demonstrate the importance of conceptualizing empathy not as
a global construct but as a collection of psychological and behavioral
expressions. For exa
mple, empathic responses that convey caring and
concern for an individual may be expressed affectively, through facial
expressions or verbal communication of concern, through helping behavior
directed toward the person in need, or more cognitively, through

attempts to
gain additional information and an understanding of distress event. Our
results support a nuanced model of empathy in which different features of
children's empathic repertoire have different sources when considering a
range of individual (i.e
., temperament) and relational (i.e., maternal
preconceptions and sensitivity) factors.

Although specific expressions of empathy (e.g., prosocial vs. inquisitiveness)
were ultimately influenced by different variables, parenting preconceptions
played a
particularly influential role, directly and indirectly affecting all three
indices of empathy. In fact, although previous research has linked child
temperament and maternal sensitivity to children's empathy, our data
suggest that the impact of maternal pre
conceptions may largely account for
these associations, particularly with respect to empathic indifference. For
example, zero
order correlations between difficult temperament and
indifference and between sensitivity and indifference were significant but di
not emerge as significant predictors when entered into a model with
maternal preconceptions, which subsumed these zero
order effects.

Parenting preconceptions exert a wide
reaching impact on children's
socialization influences, including mothers' percept
ion of their infants'
temperament and behaviors, the subsequent sensitivity mothers display to
their children, as well as children's later empathy. These findings are
striking, given the tendency for parenting attitudes to influence steady and
enduring asp
ects of a child's environment, suggesting potentially stable and
term effects.

Basic principles of attachment theory can be drawn upon to speculate why
preconceptions had this impact on children's empathy. What has been found
regarding the intergenera
tional transmission of attachment is that parents'
sensitivity in parenting, stemming from their own working models, shapes
children's attachment strategies (
van IJzendoorn, 1992
). Although we do not
have data regarding mothers' working models of attachment, per se, we
have found that mothers' negative attitudes toward child rearing result in
parenting patterns characterized by lower sensitivity to infants' cu
es. As a
result, much like infants with avoidant attachments exhibiting overt
indifference toward being reunited with their attachment figures after a brief
separation, children of mothers with negative preconceptions may be more
likely to show greater ind
ifference toward their mothers' distress.

In addition to attachment theory, perspectives from social learning theory
Bandura, 1977
) can be used to account fo
r some of the variation in
children's empathy. Mothers with more positive and adaptive parenting
attitudes may express greater empathy toward their children, which, in turn,
could lead to the development of children's own empathic responses.
Indeed, explic
it parental socialization of children's empathic responding may
be an important variable to consider in future research. The influence of
interventions and direct instructions on the part of the caregiver during
distress events has been studied previously
Waxler et al., 1979
), and
further research should address the direct impact of social modeling effects
and reinforcement for empathic behavior (or puni
shment for lack of
empathy). Moreover, our results also suggest that an intervention using
direct maternal instruction of empathic behavior may be more effective if it
includes a component addressing mothers' own attitudes about parenting
and instructing t
heir children.

Additional variables contributing to children's empathic development also
deserve further research. One factor not accounted for by our models is the
influence of caregivers other than the mother. It is possible that other
caregivers in chil
dren's lives, for example, a father figure or grandparent,
who share similar values of parenting with a child's mother, might further
mediate the relation between mothers' parenting attitudes and
developmental outcomes. Other adults, siblings, and relation
ships in a child's
life may also have a direct impact on development (e.g.,
Cochran & Niego,
Stocker, Burwell, & Briggs, 2002
). For instance,
Joye et al. (2004)

found that fathers' levels of child
centered par
enting beliefs were predictive
of children's empathy, even after accounting for mothers' child

More thoroughly examining the differential effect of a familiar versus
unfamiliar person in distress would also be important for better
anding the development of empathy. Eisenberg and colleagues'
research on sympathy in children is typically measured in relation to
unfamiliar others (e.g.,
hou et al., 2002
). Our data indicated that maternal
preconceptions, child difficult temperament, and maternal sensitivity were
significantly related to empathy directed toward the mother but not to
empathic responses toward an unfamiliar examiner (with the

one exception
of maternal sensitivity being negatively associated with indifference toward
the examiner in distress). Although zero
order correlations between model
variables and empathy toward the examiner were not as strong, this is not
to say that empa
thic responses toward the mother and the examiner were
unrelated. In fact, correlations between mother and examiner empathy
indices were significantly correlated; prosocial and inquisitive responses
toward mothers' distress were significantly correlated wi
th all three indices
of empathy directed toward examiners' distress. However, results suggest
that empathic responses toward the mother versus an unfamiliar person are
predicted by different variables. These findings corroborate previous
research also show
ing contextual effects on children's empathic responses
Robinson et al., 2001
Young et al., 1999
), and further research could
better disentangle such differential effects. For instance, the types of
variables that might predict empathy toward a stranger could be those less
specifically related to the mother
ld relationship and more related to
constructs such as general social competence or sociability.

There were several strengths to this study that deserve some mention. Our
results provide important information as to how empathy develops in an
ethnically div
erse, low
income sample. In particular, there may be
advantages to studying the importance of maternal preconceptions about
parenting in a sample that, because of assumed greater variability in both
predictors and outcomes, can include greater indices of r
isk for child
outcomes (e.g., poverty, adolescent and/or single parents, low education).
Indeed, one limitation in prior research has been reliance on middle
Caucasian samples. What was also unique to this study was the use of
longitudinal data, rat
her than contemporaneous measures such as those that
have often been used in previous research, to understand the links between
maternal preconceptions before birth and outcomes at 6, 12

15, and 21

months. In addition, our data comprised both observed a
nd self

We should note that one limitation to the present study was that we focused
only on empathy observed at 2 years. Without further replication, it is yet
unclear whether these results might generalize to older children. An

limitation was the use of maternal reports of both preconceptions
and child difficult temperament. It is possible that the strong relation
between these two variables was an artifact of shared method variance
reflecting mothers' negative perceptions of ch
ildren in general. Although this
link would be more compelling with the use of an independent measure of
temperament, this study demonstrated that maternal preconceptions were
predictive, not only of difficult temperament but also of observed parenting
sitivity and child empathy. In fact, as evidenced by bivariate correlations,
preconceptions were actually more strongly associated with observed
sensitivity than temperament, providing support for the strong impact of
preconceptions on maternal behavior an
d child outcomes, above and beyond
shared method variance.

Inadequate statistical power may have resulted in the lack of gender or
ethnic group differences found using modeling techniques. Although we had
a large combined sample to work with, we had only 8
7 boys and 88 girls and
only 87 Latinas and 54 European Americans in each group model with which
to test differences. Future studies with a similarly diverse but even larger
sample would be needed to confirm our findings.

Despite these few limitations, inf
ormation gained from this study has
potentially broad societal implications. Lack of empathy has been implicated
in numerous maladaptive behaviors including aggression, antisocial, and
disordered behaviors (e.g.,
Cohen & Strayer, 1996
). Empathic
responsiveness has been deemed critical to all social processes including
morality, caring, and general positive growth and development (
Waxler, & McShane, in press
). Given the importance of empathy in
everyday interactions and in providing a foundation for social relationships,
he goal of our larger program of research is to continue to study individual
differences in empathic development, focusing on children from ethnically
diverse, low
income families. Our ongoing work examines potential
variations and additional predictors of

empathic development using
longitudinal data from large, understudied populations.

Information from the current study can be used in the development of
effective parenting intervention programs. One of the first steps might be to
educate parents about the

relation between parenting attitudes and
children's development (
Schaefer, 1991
). It may be important to
demonstrate to parents that their preconceptions, t
hat is, parenting
attitudes they have before their child is even born, can be associated with
their perceptions of their child's behaviors, their own future parenting
behaviors, and actual long
term child outcomes. Intervention strategies such
as cognitive

reframing could be aimed at directly questioning and improving
negative preconceptions. Furthermore, parenting preconceptions are most
likely the result of a long personal history, and attention to the emotional as
well as cognitive sources of those attit
udes would be necessary for change.
Our data support the utility of implementing such interventions prenatally,
and indeed,
Daggett et al. (2000)

argued that

negative parenting attitudes
might best be addressed through preventive interventions carried out before
the birth of the child.

Intervention strategies that have been shown effective in improving child
outcomes could be drawn upon to help in developing p
renatal programs
Daggett et al., 2000
Izard, 2002
Lutenbacher, 2001
Olds, Henderson,
Kitzman, et al., 1998
Schaefer, 1991
). For instance, Olds and colleagues
found that pre

and postnatal home visitation services geared toward the
prevention of child maltreatment have

significant benefits both in the short
term reduction of negative parenting attitudes (
Olds, Henderson, Tatelbaum,
& Chamberlin, 1986
) as well as in long
rm reductions of child
maltreatment (
Olds, Henderson, Kitzman, et al., 1998
) and of later juvenile
delinquency (
Olds, Henderson, Cole, et al., 1998
). Given the strength of
negative maternal preconceptions in longitudinally predicting child difficult
temperament, less maternal sensitivity, and children's empathy,
it seems
imperative to develop strategies to ameliorate these preconceptions to
provide more optimal opportunities and outcomes for children.

American Psychologist

© 2004 American Psychological Association

April 2004
, Vol.
, No.
, p

text PDF

The Heroism of

Women and Men

Selwyn W.

Alice H.


Graduate School of Business, University of Chicago

Department of Psychology, Northwestern University

Address for Correspondence:
Selwyn W. Becker, Graduate School of Business, University of
Chicago, IL 60637


Preparation of this article was supported by National Science Foundation Grant SBR
9729449 to Alice H. Eagly. Authorship was determined alphabetically and does not reflect
relative contributions.

thank Claartje Vinkenburg, Marloes Van Engen, Pieter Inia, Maaike Ligthart, and Kim
Kasten for classifying the names of the Dutch Righteous Among the Nations; Maria Zagatsky
for classifying the names of the French Righteous; Abigail Masory for assisting in

the names of the Dutch Righteous; Tim Baker for helping to obtain records from the United
States Holocaust Memorial Museum; Zbigniew Kominek for helping to locate and translate
Polish documents; June P. Farris for helping to locate library materi
als in the Regenstein
Library of the University of Chicago; Carol Maraist, whose stimulating questions and
literature searches led Selwyn W. Becker to the database of the Righteous; Peter
Dembowski, who shared information on conditions in Poland during the

Nazi occupation;
and John Archer, Daniel Batson, Janine Bosak, Anne Campbell, Amanda Diekman, Paul
Eastwick, Douglas Kenrick, Anne Koenig, Melvin Lerner, Abigail Mitchell, Carmen Tanner,
and Wendy Wood for comments on a draft of this article.



Definition of Heroism

Psychological Theory and Research Relevant to Heroism in Women and Men

The Relation Between Gender and Risk Taking

The Relation Between Gender and Ma
nifesting Empathic Concern
With Others

Heroism and Gender

Extremely Dangerous
Heroic Acts

Carnegie Hero Medal Recipients

Righteous Among the Nations

Classification of the Righteous Among the Nations by sex

Determination of popul
ation baselines

Percentages of female and male Righteous Among the Nations in
relation to population baselines

Characteristics of the Righteous Among the Nations

Other Heroic Acts

Living Kidney Donors

Peace Corps Volunteers

Doctors of the World


Distinctive Characteristics of Carnegie Heroism

Sources of Heroism in Empathic Concern, Relational Self
Construal, Ethics,

and Male Protectiveness

Cultural Association of Men With Heroism



Table 1


Heroism consists of actions undertaken to

help others, despite the possibility
that they may result in the helper's death or injury. The authors examine
heroism by women and men in 2 extremely dangerous settings: the
emergency situations in which Carnegie medalists rescued others and the
t in which some non
Jews risked their lives to rescue Jews. The
authors also consider 3 risky but less dangerous prosocial actions: living
kidney donations, volunteering for the Peace Corps, and volunteering for
Doctors of the World. Although the Carnegie
medalists were
disproportionately men, the other actions yielded representations of women
that were at least equal to and in most cases higher than those of men.
These findings have important implications for the psychology of heroism
and of gender.

The human propensity to commemorate heroes appears to be a universal
feature of human culture. Heroes are honored in ancient cave paintings and
in folklore and myth. Societies transmitted stories of heroism in oral
traditions and molded legends, folktales,

and myths into poems, epics, and
eddas (
Carlyle, 1891
Hook, 1943
Klapp, 1948
). Contemporary societies
maintain the tradition of honoring heroes not only in literary works but also
in film, television, and journalism. One striking fea
ture of the heroes who
have achieved public recognition is that they are almost exclusively male.
The phenomenon that we evaluate in this article is the resulting cultural
consensus that “The hero is undeniably
, the male of the human species”
Lash, 1995
, p. 5). To explore this prominence of heroic men and the
apparent infrequency of heroic women, we define heroism and evaluate
contexts in which heroic behavior oc
curs more often in one sex

than the
other. The empirical evidence that we present consists of behavior that
arises from real
life decisions of men and women who have faced different
degrees of danger in a variety of situations. Finally, we evaluate whether the
position of men and w
omen in society has created differential access to
achieving recognition and commemoration for one's heroic acts.

In Western culture, the linking of heroism and masculinity can be traced in
myth and religion. Myths of the creation of humans featured heroes

succeeded in bringing forth humans or endowing them with the wisdom to
cope with their environments. Given the association of women with
procreation, it is not surprising that the heroes of early creation myths of
Western cultures included deities of
both sexes. Many goddesses such as
Isis, Ishtar, Inanna, Demeter, Cybele, and Cerridwen were portrayed as the
equals of gods and as possessing powerful natural forces of fertility and
creation (
Monaghan, 1990
Stone, 1978
). Nonetheless, male deities became
more prevalent over time, and goddesses, to the extent that they conti
to exist, came to play subordinate roles. When monotheism developed, there
remained no possibility of intertwined pantheons of male and female deities.
Fueled by medieval chivalric codes as well as shifts in religion, conceptions
of ideal male behavio
r in Western culture came to feature courageous
behavior in the service of others (
Hearnshaw, 1928
Keen, 1984
). In this
article, we consider the extent to which this cultural association of heroism
with men and masculinity is congruent with evidence of heroic behavior in
natural settings.

Definition of Heroism

To begin this analysis, we first define
. Whereas heroes and heroism
are generally defined in terms of courage and risk of one's life as well as
nobility of purpose (e.g.,
American Heritage Dictionary, 2003
Oxford English
Dictionary, 2003
), these definitions do not clearly indicate that it is the
conjunction of risk tak
ing and service to a socially valued goal that yields
heroic status. Yet, actions that have both of these attributes are far more
likely to yield heroic status than actions that have only one attribute. Thus,
people who take risks merely for pleasure or to

attract attention, as in
extreme sports, are not deemed heroic, nor are people who serve valued
social goals without risk to their own life or health, as in community
volunteering. Consistent with our definition of heroism, the Carnegie Hero
Fund Commissi
on, which was established by
Andrew Carnegie (1907)

honor heroes, recognizes the necessity of both risk and service to others in
identifying heroes as individual
s who voluntarily risk or sacrifice their life for
others' benefit (
Carnegie Hero Fund Commission, 2002
). Consistent with this
definition, actions recognized as her
oic are ordinarily performed voluntarily
in the sense that they are not coerced by external pressures or at least go
beyond the bounds of the behavior ordinarily induced by external pressures,
as in the case of military heroism. In this article, we therefo
re consider as
heroes only individuals who choose to take risks on behalf of one or more
other people, despite the possibility of dying or suffering serious physical
consequences from these actions.

Heroism can be identified within the broader category of
prosocial or helping

most of which do not involve much risk to the helper. It is the
acceptance of risk to one's life that calls for valor or courage and thus
transforms prosocial behavior into heroism. These actions, like other
prosocial behavior, need not be motivated by pu
re altruism in

sense of “a motivational state with the ultimate goal of increasing
another's welfare” (p. 6). As we illustrate in this article, vari
ous motives
could underlie helping another person by means of acts dangerous to
oneself. Before examining heroic actions of women and men, we consider
some insights about heroism that emerge from psychological theory and

Psychological Theory a
nd Research Relevant to
Heroism in Women and Men

Our requirement that heroism involves an unusual amount of risk to one's
life or health in helping one or more other people frames our discussion of
two themes: the psychology of risk taking and the
psychology of manifesting
empathic concern with others' welfare. We take into account several bodies
of theory and research that consider gender in relation to each of these

The Relation Between Gender and Risk Taking

Consideration of risk
taking as sex typed emerges mainly in two theoretical
traditions: analyses of the male gender role and of the possible evolutionary
origins of sex differences in the propensity to take risks. Role theorists
identify societal influences that produce a socia
lly constructed male gender
role, defined as shared expectations about how men do and should behave
Eagly, Wood, & Diekman, 2000
). Many analyses of the male gender

have included the element of risk taking. For example,
David and Brannon

maintained that the essential themes of masculinity encompass the

of “reckless adventure, daring exploits, and bold excesses of all
kinds” (p. 30), and
Levant and Kopecky (1995)

similarly included risk taking
and the ability to r
emain calm in the face of danger as aspects of the male
Thompson and Pleck's (1986)

questionnaire measure of masculine
ideology included items expressing bra
vado and a taste for danger. Also,
Mosher and Sirkin's (1984)

effort to assess an extreme version of the male
gender role yielded a Hypermasculinity Inventory that
included a Danger As
Exciting subscale. In support of such analyses, numerous studies of gender
stereotypes have shown that the male gender role

or cultural stereotype

includes attributes such as daring, adventurous, calm in a crisis, willing to
take risks
, and stands up under pressure (e.g.,
Broverman, Vogel,
Broverman, Clarkson, & Rosenkrantz, 1972
Diekman & Eagly, 2000
). In a
cultural study of gender stereotypes,
J. E. Williams and Best (1990)

found that the traits of daring, adventurous, and courageous were
associated more with men than with women in all 25 nations that they

Role theorists ascribe these qualities and other aspects of gender roles to
distribution of women and men into different specific roles in societies
and to the tendency for social perceivers to assume correspondence between
the behaviors demanded by roles and the dispositions of role occupants
Eagly et al., 2000
). Because men are more likely than women to
occupy social roles that require taking actions entailing risk to one's life
(e.g., warrior, fire fighter), men are imputed to have

the role
characteristic of the propensity to take risks. Men's occupancy of these roles
derives primarily from their greater physical prowess and the restrictions
that women's reproductive activities place on women's activities in many
ents (
Wood & Eagly, 2002
). To prepare boys for these adult role
occupancies, they are socialized to take risks (e.g.,
Goldstein, 2001
). For
example, mothers reported slower and less frequent interventions and
cautions directed to sons than to daughters in situations of potential physical
injury (
Morrongiello & Dawber, 2000
). Moreover, men's positioning in society
as eligible for rewards of power and high status that have generally been
inaccessible to women may produce more striving amo
ng men for the public
recognition that allows them to rise in social hierarchies (
Baumeister &
Sommer, 1997
). For men but not women, risks taken may yield large gai
of status and power.

Evolutionary psychology offers a different perspective on risk taking by
assuming that humans evolved in a context in which ancestral men
competed with one another for sexual access to women. As a consequence,
men, more than women,
learned to take risks, especially in the form of
dangerous competitive interactions with other men, presumably because
these qualities increased ancestral men's access to the resources and status
that enhanced their mating opportunities (
Daly & Wilson, 1988
). From this
perspective, risk taking is an “evolved aspect of masculine psychology as a
result of sexual selection” (
Wilson & Daly, 1985
, p. 66). Men thus “engage in
dangerous confrontations and other forms of risky behaviour where the
reward is an elevation of status in the local community” (
Campbell, 1999
, p.
204). Although the emphasis in such discussions is often on men's violence
toward one another, the reasoning of evolutionary psychologists also
suggests that a portion of men's

risk taking could be deployed in the service
of other members of their group. Such behavior could yield rewards of
increased power, status, and mating opportunities.

Some evolutionary psychologists have argued not only that men evolved a
tendency to take
risks but also that women evolved a tendency to avoid
them (
Campbell, 1999
). If mothers' care was more crucial than fathers' care
to the survival of offspring, wome
n may have evolved a strong concern with
merely remaining alive because their nurturing was crucial to increasing
their children's chances to survive. By


argument, women
are more avoidant of danger than men and generally more fearful and

The quite diverse paradigms used in psychological research designed to
study risk taking have provided modest support for theories predicting that
men take more risks than women. A large longitudinal study of adolescents'
tendencies to engage in risky or p
roblem behaviors found weak associations
between gender and varied risky behaviors,
(1976) =.18, and thrill
(1976) =.22 (
Cooper, Wood, Orcutt,
& Albino, 2003
). In
Miller, and Schafer's (1999)

analysis of research on risk taking, men
appeared to be slightly more likely than women to engage in r
isky behavior

= 0.13), although this sex difference showed a secular trend of
decreasing magnitude over the years in which this research was conducted.
However, given ethical restraints on psychological research, behavioral
studies of risk taking c
onducted in psychological laboratories do not
encompass situations in which loss of life or serious injury are genuine
possibilities. Nonetheless, using self
report or observational methods,
investigators of risk taking have examined some relatively danger

for example, drug use, unprotected sex, risky driving behavior,
and hazardous recreational pursuits. Studies of some of these types of risk
taking produced larger effect sizes than studies in laboratory paradigms
Byrnes et al., 1999
). As evidence for men's risk taking, evolutionary
psychologists commonly have pointed to the greater prevalence of violent
behavior, especially homicides, in men than in wo
men as well as men's
higher accident rates, especially among younger men (
Wilson & Daly, 1985
In the United States, men currently account for 82.6% of arrests fo
r violent
crime and 89.2% of arrests for murder and nonnegligent manslaughter
Federal Bureau of Investigation, 2002
, Table 42). Also, men's death rate
from acciden
ts is 2.2 times that of women, and their death rate from
homicide is 3.3 times that of women (
National Center for Health Statistics,
, Table C).

It is notable t
hat studies of risk taking rarely examined behaviors that are
prosocial in the sense that they benefit others. Also, perhaps reflecting
infrequent observation of situations involving extreme risk, most
psychologists have not taken into account the influenc
e of physical prowess
on risk taking. Some dangerous actions are more safely executed by people
who are physically strong and trained in the relevant physical skills (e.g.,
rescuing potential drowning victims). Men's greater size and upper
strength wo
uld contribute to their risk taking when physical demands are
high, although most risky actions would not be selective in this manner.

The Relation Between Gender and Manifesting Empathic Concern
With Others

In contrast to psychologists' attention to risk taking in understanding male
psychology, their consideration of empathic concern with others has more
often arisen in theories of female psychology. For example,
Gilligan (1982)

maintained that the moral reasoning of women and men differs, with
women's reasoning tending to display a logic based on caring and
responsibility to others, and men's displaying a logic based o
n rights and
abstract principles. In addition, many feminist scholars have emphasized
that women, especially mothers, are expected to place the needs of others
before their own (e.g.,
Chodorow, 1978
Miller, 1976
). In concert with these
analyses, the female gender role features norms fostering nurturing and
caring behavior. Nu
merous studies of gender stereotypes (e.g.,
Diekman &
Eagly, 2000
ce & Helmreich, 1978
) have shown that social perceivers
ascribe to women, more than to men, prosocial qualities such as helpful,
kind, compassionate, and devoted to others.
J. E. Williams and Best (1990)

found that the traits of helpful and kind were associated more with women
than men in 23 of the 25 nations that they examined. From a social
perspective, these traits are expected to be stereotypic of women. Be
women are more likely to occupy social roles that require caring for others
(e.g., mother, teacher), women are thought to have role
characteristics (
Eagly et al., 2000
) and are socialized to embody them (e.g.,
Barry, Bacon, & Child, 1957
Whiting & Whiting, 1975

Although evolutionary theory offers varied accounts of human prosocial
behavior (i.e., altruism; see
G. C. Williams, 1999
), a recent analysis,
supported by physiological and behavioral evidence, focused on women's
tendency to behave supportively toward others in sit
uations of stress or
danger. Thus,
Taylor et al. (2000)

maintained, like most evolutionary
psychologists, that women evolved to maximize not only their own surviva
but also that of their children, consistent with the greater parental
investment of women than men (
Trivers, 1972
). Taylor and her colleagues
argued that women's

evolutionary history of responsibility for children's
welfare led not only to the avoidance of risk emphasized by
Campbell (1999)

but also to the development by wo
men of a “tend and befriend” response to
stress (p. 411). By this argument, women's actions in stressful situations,
facilitated by oxytocin and female reproductive hormones, focus on
protection of their offspring from harm and affiliation with others that

reduces risk. Women's prosocial concern with others could thereby derive
from an evolved disposition fostering supportive behavior toward them.