6677
Mike7896
Zahn-Waxler1992.pdf
Developmental Psychology 1992, Vol. 28, No. 1,126-136
In the public domain
Development of Concern for Others
Carolyn Zahn-Waxler, Marian Radke^rrow, and Elizabeth Wagner
National Institute of Mental Health Bethesda, Maryland
Michael Chapman University of British Columbia
Vancouver, British Columbia, Canada
The development of prosocial and reparative behaviors was investigated by examining children's responses to distresses they caused and those they witnessed in others during the 2nd year of life. Prosocial behaviors (help, sharing, provision of comfort) emerged between the ages of 1 and 2, increasing in frequency and variety over this time period. These behaviors were linked to expres- sions of concern as well as efforts to understand and experience the other's plight. Children's reparative behaviors after they had caused distress also increased with age. Age changes in these early signs of moral development were accompanied by social-cognitive changes in self-recogni- tion. In assessments at age 2, children were most responsive to distress in their mothers but also showed some sensitivity toward unfamiliar persons.
Children are frequently exposed to others in distress, both as bystanders and as causes. Hence, opportunities to alleviate the distress commonly arise. When children intervene as by- standers, constructs such as altruism, prosocial behavior, em- pathy, sympathy, and compassion are used to describe their emotions and behaviors. When children show similar concern toward someone whom they have harmed, concepts of con- science, guilt, reparation, and remorse also become relevant. For many years these aspects of moral development were con- sidered to be outside the cognitive, social, and emotional reper- toires of young children. In psychoanalytic (Freud, 1930/1958) and social-cognitive (Piaget, 1965) theories young children were portrayed as primarily egocentric, demanding, depen- dent, and socially inept. Responsiveness to others' needs as well as responsibility for one's own actions were hypothesized to evolve during middle childhood. More recent work, however, suggests much earlier origins.
According to Kagan, children universally begin to display a moral sense during the second year of life as adherence to stan- dards becomes important (Kagan & Lamb, 1987). This period is also viewed as the time during which higher order, self- conscious, moral emotions reflecting some capacity for role- taking emerge (e.g., empathy, guilt, shame; Campos, Barrett,
This research was supported by the National Institute of Mental Health (NIMH) Intramural Research Program and the John D. and Catherine T. MacArthur Foundation, Network on Early Childhood Transitions.
We would like to express our appreciation to Geri Cooperman, Christine Grewell, Jean Mayo, Claudia Pyle, and Jean Welsh for their help on this project and to the mothers and children who participated in the research.
Correspondence concerning this article should be addressed to Car- ofyii Zahn-Waxler, Laboratory of Developmental Psychology, NIMH, 9000 Rockville Pike, Bethesda, Maryland 20892.
Lamb, Goldsmith, & Stenberg, 1983; Darwin, 1872; Izard, 1977). Some socialization approaches focus as well on the early years and highlight the role of the environment in shaping indi- vidual differences in moral orientations (Emde, Johnson, & Easterbrooks, 1987; Yarrow, Scott, & Waxier, 1973; Zahn- Waxler, Radke-Yarrow, & King, 1979).
Hoffman (1975) has postulated a biological preparedness for empathy. It is first in evidence in the reflexive crying of infants in response to the crying of other infants. As children begin to differentiate self and other during the second year of life and hence to develop understanding of others as separate beings, their emotional involvement in another's distress begins to be transformed from personal, self-distress to sympathetic con- cern for the victim. Comforting interventions are hypothesized to emerge during this period. Both empathy and guilt develop and show similar developmental courses (Hoffman, 1982). Em- pathy occurs in response to distress in a victim, whether this distress is caused or witnessed by the child. Interpersonal guilt results from the conjunction of an empathic response to some- one's interpersonal distress and awareness of having caused that distress.
Although Hoffman's theory is not data based, supportive em- pirical evidence has begun to accumulate in recent years. From birth onward, infants are responsive to emotions in others. Contagion (Sagi & Hoffman, 1976; Simner, 1971) and imitation of others' emotions (Field, Woodson, Greenberg, & Cohen, 1982; Meltzoff& Moore, 1977) are present even in the first days of life, suggesting a biological predisposition. This is consistent with recent views of the adaptive nature of empathy from evolu- tionary (MacLean, 1985), neuroscience (Brothers, 1989), and functionalist perspectives. As early as the first months of life, infants discriminate between mothers' presentations of differ- ent emotions and match these emotions under some conditions (Haviland & Lelivica, 1987; Termine & Izard, \9%%). Other in- vestigators have demonstrated infants' responsiveness to the
126
DEVELOPMENT OF CONCERN 127
affective states of the caregiver, emphasizing the negative arousal induced by maternal depression (Cohn, Campbell, Ma- tias, & Hopkins, 1990; Cohn & Tronick, 1983).
The beginnings of representational thought during the sec- ond year of life (Bruner, 1972; Piaget, 1962) and the use of symbols (e.g, McCall, 1979; Nicolich, 1977) have implications for children's abilities to infer others' perspectives and feelings (Butterworth, 1980; Rheingold & Emery, 1986). During this period, self-recognition and self-other differentiation develop (eg., Amsterdam, 1972; Bertenthal & Fischer, 1978; Lewis, Sul- livan, Stanger, & Weiss, 1989; Watson & Fischer, 1977). Chil- dren begin to use emotion language (including moral terms) to describe others' internal states as well as their own (Bloom, Lightbown, & Hood, 1975; also see review by Bretherton, Fritz, Zahn-Waxler, & Ridgeway, 1986). Stipek, Gralinski and Kopp (1990) and Lewis et al. (1989) identify aspects of self-develop- ment that may be necessary preconditions for the expression of self-conscious and moral emotions.
Research also provides some evidence for the emergence of helping, sharing, and cooperation in the second year of life (Eckerman, Whatley, & Kutz, 1975; Rheingold, Hay, & West, 1976; Ross & Goldman, 1977). These behavior patterns may have precursors in earlier attachments to the caregiver. These bonding patterns, in turn, prepare the child for later empathic development through the sharing and exchange of emotions as well as cooperation and turn-taking in social interactions be- tween parent and infant (Stern, 1985; Trevarthen, 1989). Proso- cial behavior does occur in preschool-age children (e.g, Eisen- berg-Berg & Lennon, 1980; RadkeYarrow & Zahn-Waxler, 1976; Yarrow, Scott, & Waxier, 1973), although base rates are low. Work by Caplan and Hay (1989) on nursery school chil- dren's responses to distressed peers links low rates of prosocial responding to children's expressed beliefs that they are not sup- posed to help when adults are present, rather than to a lack of capacity.
In summary, research already documents a broad repertoire of competencies that young children bring to their relation- ships with others. Even children as young as 2 years old have (a) the cognitive capacity to interpret the physical and psychologi- cal states of others, (b) the emotional capacity to affectively experience the other's state, and (c) the behavioral repertoire that permits the possibility of trying to alleviate discomfort in others. These are the capabilities that, we believe, underlie chil- dren's caring behavior in the presence of another person's dis- tress. Our interest here is in how and when these competencies become integrated to form organized response patterns.
To our knowledge, only one study has attempted to explore this question. Mothers were trained to observe and report on their 1- to 2V>year-olds' responses to naturally occurring dis- tress in others, distress that they witnessed as bystanders and distress that they caused (RadkeYarrow & Zahn-Waxler, 1984; Zahn-Waxler & RadkeYarrow, 1982). Mothers also simulated distress emotions and recorded their children's reactions. Pro- social behaviors appeared at the beginning of the second year and increased with age, taking more varied forms as children grew older. Similar behavioral interventions occurred in re- sponse to distresses caused and witnessed, illustrating chil- dren's tendencies to repair harms created as well as to provide aid as bystanders.
Because distress in others is infrequent and unpredictable, mothers may be the only regular witnesses of their young chil- dren's reactions to distress incidents. Hence, there is merit in obtaining data from caregivers. A tradition of research using parents as observers of their young children supports such an approach, (e.g, Goodenough, 1931, on emotional development [anger]; J. E. Bates & Bayles, 1984, on temperament; E. Bates, Bretherton, & Snyder, 1988, on language development; Stipek, Gralinski, & Kopp, 1990, on self-development; and Grusec, 1991, on prosocial behavior). There is evidence that caregivers can be trained to be reliable observers and that their observa- tions predict other independently assessed aspects of parent and child functioning. Maternal reports can also contain dis- tortion and bias; hence there is a tradition, as well, for question- ing the reliability and validity of such data (Radke-Yarrow, Campbell, & Burton, 1970). Here we attempted to determine whether our earlier research findings on children's prosocial development would be replicated with a second sample of children, when a broader range of methodologies was em- ployed.
Research Problem
Children's responses to distresses that they witness and cause in others were examined longitudinally between the ages of 1 and 2 years. This is the period in development during which conscience and concern for the welfare of others are hypothe- sized to emerge. The affective, behavioral, and cognitive compo- nents of children's responses to others' distress were explored. Children's responses to witnessed and caused distress were compared, as were their responses to natural and simulated distresses. Data obtained from maternal reports in situations not observable by others also were compared with indepen- dently videotaped observations. The development of self-recog- nition was examined and analyzed in relation to children's re- sponses to distress. This provided an opportunity to explore linkages between social-cognitive abilities and prosocial behav- ior patterns early in the development of both processes. To study families likely to provide a range of naturally occurring distress situations, children of both psychiatricalry normal and depressed mothers were studied. Depression in a caregiver can create an environment in which distress is chronic and perva- sive, and young children's sensitivities are known to be in- fluenced by familial distress and discord (Cummings, Zahn- Waxler, & RadkeYarrow, 1981; Zahn-Waxler, Kochanska, Krupnick, & McKnew, 1990). Hence, it becomes useful to in- troduce into research designs such potential contributors to in- dividual differences (RadkeYarrow & Zahn-Waxler, 1990).
Method
Sample
Thirty mothers of 1-year-old children were recruited through news- paper notices and community bulletin boards (n = 23) and through referrals from a National Institute of Mental Health (NIMH) adult affective disorders research laboratory (n = 7) to participate in a longi- tudinal study of young children's social-emotional development. Three mothers left the study, either because the family relocated to another city or because the mother was unable to commit sufficient
128 ZAHN-WAXLER, RADKE-YARROW, WAGNER, AND CHAPMAN
time to data collection. Twenty-four children were White, 2 were Black, and 1 was Eurasian. Families were mainly middle class (mean years of education for the mothers was 16.2, and for fathers, 17.0). There were 17 boys and 10 girls, an imbalance that was exaggerated by the fact that the 3 mothers who left the study all had daughters. The Schedule for Affective Disorders and Schizophrenia-Lifetime Version (SADS-L) interview (Spitzer & Endicott, 1979), which diagnoses affec- tive disorders and schizophrenia, was administered. Twelve mothers had major depression, and 15 mothers were diagnosis free.1 Mothers received financial remuneration through the National Institute of Health volunteer program.
Overview of Procedures
Mothers of 12- to 13-month-old children were trained (during the month before the study) to observe their children's responses to the emotions of others. Mothers reported incidents throughout the child's second year; however, the focal points of measurement (of which mothers were uninformed in order not to sensitize them) were three 3-month periods: 13-15 months, 18-20 months, and 23-25 months. On the basis of earlier conceptual (Hoffman, 1975) and empirical (Zahn-Waxler & Radke-Yarrow, 1982) work, these periods were hy- pothesized to be significant transition stages from self-distress to em- pathic, active engagement with distress victims. Mothers observed and reported emotion events both caused and witnessed by their children. Mothers also simulated emotions and recorded their children's re- sponses to these events. Once a month, a home visitor came to the family residence to review the mother's observation reports and to interview her about other aspects of the child's development. During home visits at 18 and 21 months, the mother simulated distresses, and the home visitor videotaped responses. On three occasions corre- sponding to the focal time points of measurement, a self-recognition task was administered in the home. Simulations were performed in a research laboratory when children reached the age of 2, and responses were videotaped. At this time, playmates of the target children also visited the laboratory, and their responses to simulations were assessed as well.
Observation Training
Mothers identified emotion episodes and described the events in a detailed, sequential account. They dictated their observations into a tape recorder. The event was to be recorded as soon as possible. Re- ports included the (a) date and time of incident and dictation, (b) events and atmosphere before incident and cause of incident, (c) location of child, (d) description of feelings and emotions expressed in the inci- dent and manner of expression, (e) the child's exact response (words, actions, emotions) or lack of response, (f) the responses of others in the situation. Two classes of events were reported: situations in which the child caused distress (e.g., pulling the cat's tail, taking the baby's bottle away, teasing a sibling, biting the mother while nursing, dropping a toy on the father's foot) and in which the child was a bystander (e.g., a hungry baby cries, mother and father argue, mother burns self on stove, father stubs toe). Mothers were asked to report all salient emo- tion events. It was not assumed that they would report every event, but by asking them to report on salient events, we hoped to obtain a reason- able sampling of children's responses to the more overt emotional events encountered in their young lives. Mothers also simulated emo- tions on a specified schedule (see below) to introduce controlled sam- pling procedures that would provide a basis for comparison with the naturalistic observations. Mothers were asked to observe and smwlate positive as well as negative emotions. Positive emotions were included (a) to provide an affectively balanced observing framework and (b) to keep the mother from focusing unduly on the child's responses to nega-
tive emotions. In keeping with the research questions, however, only the responses to negative incidents were transcribed, coded, and ana- lyzed.
Details of observing were elaborated in a training manual, and exam- ples of events were provided. These illustrated a range of types of distress and also a variety of responses to avoid guiding the mother into one particular response pattern. Mothers practiced the observa- tion procedures for several weeks, then sent in audiotapes that were transcribed and reviewed during the next home visit. Confusions were resolved, missing information was requested, and questions were clari- fied. After this session, the mother began to provide data-based obser- vations for the study. Two mothers had difficulty with the narrative procedure, and an experimenter regularly visited the home and called to obtain more detailed accounts of the child's reactions.
Simulation Training
The home visitor instructed the mother on how to simulate speci- fied emotions and modeled each simulation for the mother. Scripts and directions were also provided in the training manual. The simulation situations were the following: (a) respiratory distress, which was cough- ing or choking for 5-10 s, blowing one's nose, or patting one's chest; (b) pain, which was bumping one's foot or head, saying "ouch," and rub- bing an injured part for 10-15 s; (c) listnessness, which was 10-15 min of apathetic, fatigued behavior, sighing, and sitting; (d) sadness, which was audible sobbing for 5-10 s. Mothers were to practice performing each simulation until they felt comfortable expressing these emotions in the child's presence. They performed one simulation per week for the duration of the study. The importance of inserting these simula- tions naturally into the existing flow of activities was emphasized.
Children's responses to additional simulations of distress were video- taped (a) in the home at 18 months and 21 months by home visitors and (b) in the laboratory at 2 years by research assistants for later coding by our research staff. This provided a separate database on children's reactions to distress at time points in development similar to those that were based on mothers' observations. Therefore, when the child was 18 months old, the mother simulated three distress incidents during a home visit (pain, sadness, and withdrawal-unresponsiveness). At 21 months, respiratory distress, sadness and withdrawal-unresponsive- ness were simulated by the mother in the home. And when children were 2 years old, they visited the laboratory three times, typically spaced 1 month apart. On each occasion, pain was simulated, once by the mother and once by each of two female experimenters. During one of these sessions, a tape recording of an infant cry was played in an adjacent room. Again, all incidents were videotaped for later coding. Because children came to these laboratory sessions with a same-age peer (Zahn-Waxler, Iannotti, Cummings, & Denham, 1990), this pro- vided the opportunity to assess independently (i.e., at different points in the session for each child) for a replication sample of children, re- sponses to the same simulations.
Coding and Reliability
Naturally occurring caused and witnessed distress incidents took a variety of forms. These included arguments and conflicts, injuries and pain, psychological distress (sadness, crying, withdrawal), illness (fa-
1 Of the 12 mothers with major depression, 7 were, as indicated, the result of referrals from the research laboratory specializing in adult affective disorders. The remaining 5 cases entered the study through community recruitment efforts. The proportion of depressed to well mothers obtained through community recruitment is consistent with the larger population ratio, as identified in epidemiological studies.
DEVELOPMENT OF CONCERN 129
tigue, colds), and the like. Victims could include anyone with whom the child came in contact (family, relatives, strangers; adult or child). Simulated distresses were performed by mothers and testers as de- scribed earlier. Coders listened to audiotapes or worked from typed transcripts of mothers' reports to categorize the children's reactions to caused and witnessed distresses. Two types of reliabilities were com- puted. For mothers' naturalistic observations, intercoder reliability was estimated by having two research assistants code half of the mater- nal observation records across different time periods. Percentages of agreement for intercoder reliability are listed first in the parentheses that follow each code description below. Videotaped simulation situa- tions provided an opportunity to assess observer-coder reliability. Per- centages of agreement between two research assistants' independent observations of responses to these simulations at 18, 21, and 24-28 months of age are listed second in parentheses below.2
The following dimensions were coded for both naturally occurring (caused and witnessed) and simulated distresses. The first four catego- ries reflect patterns suggested by previous research and theory as rele- vant to prosocial development (RadkeYkrrow, Zahn-Waxler, & Chap- man, 1983). In addition, three nonprosocial behavior patterns (5, 6, and 7 below) were present and hence were coded as well.
1. Prosocial behavior (89%, 94%), which was children's spontaneous behavioral efforts to intervene on behalf of the victim, to change the situation, or to alleviate the distress. Specific forms included (a) physi- cal comfort (e.g., hugs, pats, kisses), (b) verbal comfort (eg., "You be okay," "Are you okay?"), (c) verbal advice (e.g., "Be careful"), (d) helping (e.g., puts on a bandage, gives bottle to crying baby, brings shawl to shivering person), (e) indirect helping (e.g., gets mother to retrieve in- fant's rattle), (f) sharing (e.g., gives food to sister), (g) distraction (e.g., closes picture book that makes mother sad), (h) protection-defense (e.g., tries to prevent another from being injured, distressed, or at- tacked).
2. Empathic concern (86%, 78%), which was emotional arousal that appears to reflect sympathetic concern for the victim. The arousal is manifested in facial or vocal expressions (e.g., sad looks, sympathetic statements such as "I'm sorry" [in a soothing or reassuring tone of voice], or gestures [such as rushing to the victim while looking worried]).
3. Hypothesis testing (74%, 94%), which was attempts to label or understand the problem, ranging from simple verbal statements or inquiries ("What happened?" or "hurt foot") to more complex infer- ences such as "Baby cry, mommy all gone" or "Baby cry—locked door —open" in response to an infant cry in the next room. Nonverbal indicators could be seen in visual search patterns, such as repeated looks from the person's injured foot to the chair that had been stum- bled on.
4. Self-referential behaviors (81%, 90%), which were imitations or enactments of others' distress (e.g, rubs own toe, blows nose, winces, says "I hurt my hand"); these are a "trying on" of the other's experi- ence.
5. Self-distress (81%, 100%), which were emotions evoked by the other's distress that were more intense, negative, and reflective of per- sonal distress. The child cries, sobs, frets, whimpers.
6. Aggressive behavior (62%, 94%), which were aggressive or angry reactions to victims (e.g., hits, pushes, pinches, shoves, yells, "I hit you" [and does], "bad girl.")
7. Positive affect (88%, 98%), which was the child showing amuse- ment, smiling, or laughing during the distress.
Assessment of Self-Recognition
At 12-14, 18, and 24 months, a visual self-recognition task (Ber- tenthal & Fischer, 1978) was administered in the home. The task mea- sures the reaction of an infant to his or her own mirror image and the
progressive ability to identify self and objects in the environment in relation to the mirror image of self on tasks graded in difficulty. The mother was requested to put a special vest on the child (to the back of which objects visible to the child only in the mirror were later at- tached). The mother also was asked to put a spot of rouge or lipstick on the child's nose as unobtrusively as possible. The mother seated the child in front of a full-length mirror and stood to one side so that her image was not visible. Then children were administered five tasks that are described below.
1. Tactual exploration task. The child passed this task if he or she attempted to touch his or her mirror image.
2. Hat task. A woollen cap was suspended approximately 15.24 cm (6 in.) above the child's head by stiff wires fixed to the back of the special vest mentioned above. The child passed this task if he or she found the hat above his or her head by looking into the mirror.
3. Toy task. The investigator dangled a toy on a string behind the child sitting in front of the mirror and asked the child to find it. The child passed this task if he or she turned around and found the toy after seeing its reflection.
4. Rouge task. The child passed if he or she looked in the mirror and then touched his or her nose or otherwise indicated that something was different about his or her nose.
5. Name task. The mother pointed to the child's reflection in the mirror and asked, "Who's that?" The child passed if he or she stated his or her own name or said "me." As described by Bertenthal and Fischer (1978), these tasks form a developmental sequence. Therefore, children were assigned a level score ranging from 0 to S, based on the highest level score that was passed (i.e., 2 was assigned if the hat task was the highest task passed, 4 if the rouge task was the highest passed, etc}. Mothers were told to respond if the child initiated interaction during testing but not to initi- ate interactions herself. Two testers were present at each visit: One administered and scored the assessments for all children, and the sec-
2 The data were not retained in a form that would permit computa- tion of Kappas, which correct for chance agreements. For intercoder reliability, the percentage estimates are already very conservative, as they do not include instances where coders agreed on the absence of a particular behavior. For observer-coder reliability, such agreements (i.e., of zero scores) were included, hence reliability figures may reflect overestimates of accuracy to some degree. To estimate observer reli- ability between mothers and outside observers, on one occasion, at 18 months, the home visitor simulated pain and both mother and home visitor recorded the child's response. Percentages of agreement be- tween mother and home visitor on categories of interest averaged in the .80s, as in our earlier study when the mother simulated pain for an observer-reliability estimate (Zahn-Waxler et al., 1979). Because chil- dren are less responsive to unfamiliar persons than to mothers (Zahn- Waxler & RadkeYarrow, 1982), we were unable to sample as many behaviors in the present study. However, we also wished to avoid plac- ing the mother in a position of feeling her performance was being evaluated any more than it already was. Although in the abstract, it might seem more optimal to obtain a broader sampling of incidents (based on simulations of caregiver and home visitor), it is necessary to balance reliability concerns with dual problems of children's overexpo- sure to simulations of distress and of maintaining rapport with the mothers. Thus, our efforts to assess the reliability and validity of ma- ternal observation reports rested mainly with the comparison of find- ings produced by these procedures with data obtained from indepen- dent sources. This latter approach is more consistent with the way self-report data obtained from parents (e.g., attachment Q sorts, child- rearing Q sorts, temperament questionnaires, behavior problem checklists) are typically used in the research literature.
130 ZAHN-WAXLER, RADKE-YARROVV? WAGNER, AND CHAPMAN
ond also scored assessments for reliability purposes. Observer reliabil- ity for coding the different stages of visual self-recognition was r = .84.
Results
A brief description of children's response repertoires is pre- sented, followed by analyses of developmental changes. Chil- dren's reactions to problems that they caused are compared both with naturally occurring and simulated problems of others that they witnessed as bystanders. (The terms bystander and witnessed are used interchangeably here.) Interrelations of mea- sures and their stability over time are presented to examine coherences of response patterns. Age changes in self-recogni- tion then are considered, followed by analyses of relations be- tween self-recognition and prosocial behavior at similar points in time. Finally, replicability and validity of results are explored by (a) comparing findings from data obtained from maternal reports with findings based on independently sampled, video- taped records and (b) comparing responses of the target sample and a peer sample at age 2.
Developmental changes in children's responses to distress are examined in multivariate analyses of variance (MANOVAs) for each of the dependent variables, with (a) age (13-15 months [Time 1 ], 18-20 months [Time 2 ], and 23-25 months [Time 3 ]) and (b) context or situation (caused [C], witnessed natural [WN], and witnessed simulated [WS] distresses) as within-sub- ject factors and with sex as a between-subject factor. (Analyses indicated no effects of mother's depression, either as a main effect or in interaction with other variables; hence this factor was excluded from all subsequent analyses). Proportion scores are used for each dependent variable, with the number of re- sponses divided by the number of distress episodes. The means and standard deviations for number of natural distress inci- dents (i£., on which proportion scores were based) at each of the time periods are as follows: For natural distresses witnessed as bystanders, M = 8.96, SD = 7.42 at Tl; M= 7.52, SD= 6.15 at T2; M = 6.15, SD = 4.51 at T3. For caused distresses, M = 4.07, SD= 3.29atTl;M= 5.96,£D= 7.38atT2;M= 3.40,£D =3.30 at T3. Analysis of variance (ANO\A) is used to examine age changes in self-recognition. All significance levels reported are based on two-tailed tests.
Descriptive and Developmental Patterns
The means and standard deviations for children's responses to distress on the basis of maternal observations are indicated in Table 1. The data are presented separately for caused, wit- nessed natural, and witnessed simulated distress situations at each of the three time periods. Both as causes and bystanders, young children show significant involvement in the emotional lives of others. They express concern, attempt to comprehend or experience the situations, and engage in behaviors directed toward alleviating distress in others. Shortly after their first birthday, more than half of the children have made at least one prosocial response, and by 23-25 months, all but 1 of the chil- dren had prosocial behavior in their repertoire of social actions. The first acts at 13-15 months were predominantly physical (e.g., hugging, patting), but already by 18-20 months the entire range of eight different types of prosocial actions (described in
the Method section) could be seen across the sample of chil- dren. However, children also sometimes found others' distress amusing and under some circumstances themselves become distressed or aggressive.
Children's responses to others' distress undergo marked transformations during the second year of life. MANOVAs of main effects of age indicate that children showed more proso- cial behaviors as they grew older, F(2, 24) = 18.45, p < .001, both between Tl (13-15 months) and T2 (18-20 months) (p < .001) and between T2 (18-20 months) and T3 (23-25 months) (p < .001). Follow-up univariate ANO\As indicated that proso- cial interventions increased significantly with age, regardless of whether children caused or witnessed the other's distress (Table 1, row 1). For naturally occurring distresses children witnessed as bystanders, F(2,22) = 10.53, p < .001; for increase from Tl to T2, p = .057; for increase from T2 to T3, p < .05). For simulated distresses, F(2, 22) = 7.82, p < .005; from Tl to T2, p < .05; from T2 to T3, p = .054). For caused distresses, F(2,22) = 7.37, p < .005; from Tl to T2, p = ns; from T2 to T3, p < .005.
MANOVAs also revealed other developmental patterns. Em- pathic concern increased with age, F(2, 24) = 6.98, p < .005, only between Tl and T3. There were significant increases with age in hypothesis testing, F(2,24) = 7.97, p < .005, between Tl and T2, p < .001, and between T2 and T3, p < .01. There were age changes in self-referential behaviors, F(2, 24) = 19.46, p < .001, the pattern being curvilinear (i.e., an increase from Tl to T2, p < .0001, and a decrease from T2 to T3, p < .05). Of the nonprosocial response patterns (i.e., self-distress, aggression, and positive emotion in response to another's distress), only 1 showed age changes. Aggression increased with age, F(2,24) = 3.41, p< .05, between T2 and T3, p < .05.
Children's Responses to Caused and Witnessed Distresses
The findings that were based on univariate ANOV\s suggest that prosocial and reparative behaviors follow similar develop- mental trajectories. However, there also were differences across caused and witnessed situations and interactions with age that qualify this interpretation (see Table 1). Children were less likely to show empathic concern, F(2, 24) = 19.42, p < .001, when they caused distress than when they witnessed it, espe- cially as they grew older, Age X Situation interaction, F(4,22) = 2.83, p < .05 (Table 1, row 2). Follow-up tests indicated that the interaction resulted from differences at T2 in responses to caused versus witnessed (simulated) distresses, 7(26) = 3.89, p < .005.3 Children also were less likely to engage in hypothesis testing when they caused distress than when they were by- standers, F(2,24)= 8.8K/> < . 001 (Table 1, row 3). Self-referen- tial responses were less frequent in response to caused than witnessed distresses, F(2,24) = 3.85, p< .05, and there was also a significant interaction of situation with age, F(4, 22) = 4.18, p < .05 (Table 1, row 4). Follow-up tests revealed that the differ- ences were significant only at T2; here, self-referential re- sponses to caused distresses were lower than responses to wit- nessed (simulated) distresses, t(26) = 3.59, p < .05, and to wit- nessed natural distresses, t(26) = 3.52, p < .01.
3 7" tests were used for follow-up tests of interactions, and Bonferroni corrections were employed.
DEVELOPMENT OF CONCERN 131
Table 1 Children's Responses to Distress During the Second Year of Life (Maternal Observations)
Prosocial behavior M SD
Empathic concern M SD
Hypothesis testing M SD
Self-reference M SD
Self-distress M SD
Aggressive behavior M SD
Positive affect M SD
Bystander distress natural (WN)
Tl
.09
.16
.09
.21
.05
.12
.02
.07
.15
.19
.01
.001
.11
.15
T2
.21
.24
.10
.16
.06
.11
.16
.18
.12
.16
.01
.006
.09
.13
T3
.49
.49
.25
.36
.24
.28
.10
.21
.07
.13
.03
.08
.05
.11
Situations
Bystander distress simulated
Tl
.10
.27
.08
.19
.04
.19
.03
.11
.17
.30
.01
.003
.19
.29
(WS)
T2
.29
.28
.23
.26
.16
.19
.19
.19
.03
.09
.01
.04
.23
.20
T3
.49
.54
.27
.35
.29
.36
.10
.21
.02
.07
.02
.05
.15
.24
Tl
.07
.13
.03
.10
.01
.04
.03
.08
.34
.33
.01
.04
.15
.24
Caused distress (C)
T2
.10
.17
.03
.10
.06
.14
.04
.11
.41
.38
.04
.10
.11
.25
T3
.52
.66
.14
.25
.07
.21
.08
.27
.33
.38
.19
.35
.20
.32
Note. Scores represent mean proportion of distresses to which child responded. Tl = Time 1 (13-15 months); T2 = Time 2 (18-20 months); T3 = Time 3 (23-25 months); N= 27.
With regard to nonprosocial patterns of responding, chil- dren showed more self-distress when they had caused rather than witnessed another's distress, F(2, 24) = 12.96, p < .001 (Table 1, row 5). They also showed more aggression toward victims when they had initially caused rather than witnessed distress, F(2,24) = 3.84, p< .05 (Table 1, row 6). There was an interaction of Sex X Situation, F(2,24) = 3.35, p < .052, indicat- ing that boys aggress more than girls for caused distresses, t(26) = 3.17, p < .05, but not for natural or simulated witnessed distresses where rates of aggression were near zero for both boys and girls. Finally, positive affect was more frequent for caused than for witnessed distresses (natural bystander events only), F(2, 24) = 6.50, p< .01 (Table 1, row 7). In many ways, then, children react differently to distress situations they create from those they witness, despite the similar attempts to inter- vene and repair harm that are common to both circumstances.
Table 1 indicates that when children witnessed distresses as bystanders, there were virtually identical patterns of respond- ing to naturally occurring and simulated distress, with one ex- ception. Children showed more positive affect in response to simulated than naturally occurring distress. This may have re- sulted from occasional exaggerations of distress expressions by mothers during their simulations, that sometimes seemed to have a comic quality.
Sex Differences in Response to Others' Distress
Whereas prosocial behaviors appeared with similar frequen- cies for boys and girls, some gender differences in other behav-
ior patterns were observed. Girls expressed more empathic concern than boys, F(9,17) = 2.79, p < .05, but only for wit- nessed and not for caused distresses, as seen in a significant Sex X Situation interaction, F(2,24) = 5.59, p < .01. Follow-up tests indicated that this sex difference was significant only for naturally occurring bystander distress incidents, t(26) = 3.35, p < .05. Self-referential behaviors also were more frequent in girls than boys, F(9, 17) = 4.82, p < .005. This main effect was qualified by an interaction with age, F(2, 24) = 5.15, p < .05, with girls showing more self-referential behaviors than boys, at 18-20 months, f(26) = 3.43, p < .05, but not at the other two age periods.
Interrelations Among Measures of Responses to Distress
To explore further the components of responding that may motivate children's attempts to alleviate others' distress, associa- tions between measures were examined. This was done sepa- rately for caused and witnessed (natural plus simulated)4 dis- tresses and separately for T2 (18-20 months) and T3 (23-25 months) when each of the prosocial patterns becomes more prominent. For bystander distresses at T2, prosocial behavior was associated marginally with empathic concern, r(25) = .37,
4 Because overall frequencies and developmental patterns of re- sponse in natural and simulated bystander situations were so similar for most variables, scores for these two conditions were combined for this and subsequent analyses to reduce the number of analyses con- ducted with a relatively small N.
132 ZAHN-WAXLER, RADKE-YARROV^ WAGNER, AND CHAPMAN
Table 2 Correlations of Stability Across Time
Time 3
Time 2 PB EC HT SR
Bystander distress
Prosocial behavior (PB) .56*** Empathic concern (EC) .26 Hypothesis testing (HT) .74**** Self-reference (SR) .54***
.40
.28
.33*
.57****
.49**
.10
.38**
.04
.18
.38*
.22
.18
Caused distress
Prosocial behavior (PB) . 11 Empathic concern (EC) Hypothesis testing (HT) Self-reference (SR)
.18
.29
.28
.04
.42**
.09 -.07
.07 -.02
.05
.26
-.23 .10
-.20 -.29
* p < . 1 0 . **p<.05. ***/>< .01. ****p<.005.
p = .058, and significantly with hypothesis testing, r(25) = .45, p < .05. At T3, prosocial behavior was associated with em- pathic concern, r{25) = .64, p < .001, and hypothesis testing, r(25) = .38, p < .05. For caused distresses at T2, children who showed more prosocial behavior also expressed more empathic concern, r(25) = .48, p < .01. For caused distresses at T3, chil- dren who showed more prosocial behavior expressed more em- pathic concern, r(25) = .58, p < .001, engaged in more hypothe- sis testing, r(25) = .40, p < .05, and made more self-referential responses, r(25) = .43, p < .01. Examination of correlations between prosocial and nonprosocial patterns and among non- prosocial measures yielded few associations: For bystander dis- tresses at T2, self-distress is negatively related to prosocial be- havior, r{25) = -.40, p < .05. For caused distresses at T2, self- distress is positively related to prosocial behavior, r{25) = .44, p < .05, and to empathic concern, r(25) = .38, p < .05.
Stability was examined separately for witnessed and caused distresses, for the same measures (again, considering only T2 and T3). Correlations for prosocial measures are presented in Table 2. There was moderate stability across time, mainly in bystander situations. Nonprosocial measures did not show sta- bility over time.
Self-Recognition and Responses to Distress
Level of self-recognition was examined as a function of age and gender in an ANOVA. There was a main effect of age, with self-recognition increasing over time, F(2,22) = 22.18, p< .001, Mean of Time 1 (MT1) = 1.32, Mean of Time 2 (MT2) = 3.12, and Mean of Time 3 (MT3) = 3.72. Differences were significant between Tl and T2 (p < .001) and marginally between T2 and T3 (p = .061). In addition to exploring developmental changes, it was also possible to examine whether at a given time point children's self-recognition predicts responses to distress accord- ing to mothers' observations. At T2 there was one marginally significant association; self-recognition was correlated with ob- served self-referential behavior, r(25) = .36, p < .10. At T3, self-recognition was correlated with prosocial behavior, r(25) = .41, p < .05, with self-referential behavior, r(25) = .40, p < .05,
and marginally with empathic concern, r(25) = .34, p < .10. Correlations between self-recognition and nonprosocial ac- tions did not approach significance. These findings provide modest evidence for links between the emergence of self-aware- ness and prosocial actions.
Validity Assessments
The validity of findings based on maternal observations was examined by comparing mothers' reports of children's re- sponses to simulations at 18-20 months with observer-coders' reports of children's responses to the different set of (six video- taped) simulations during the 18-21-month period. The dis- tress simulations were similar but not identical under the two conditions, hence providing a conservative test of similarity. Paired comparison t tests were used to examine differences between each of the four prosocial variables across the two settings (i.e., prosocial behavior, empathic concern, hypothesis testing, and self-reference). Average frequencies for prosocial behaviors and empathic concern were similar, showing no sig- nificant differences across the two settings (see Table 3). Mothers reported more self-referential behaviors than were ob- served on videotapes, t{26) = 2.54, p < .05, whereas data coded from videotapes revealed somewhat more hypothesis testing than was reported by mothers, f(26) = -2.01, p = .055. Inter- correlations of measures based on videotaped data showed pat- terns of association like those reported by mothers (e.g., proso- cial behavior was correlated with hypothesis-testing), r(25) = .49, p < .01, and with empathic concern, r{25) = .47, p < .05. There were fewer correlations of these measures with self-re- cognition, though empathic concern was related to self-recogni- tion, r{25) = .42, p < .05.
Replication of findings across methods also was explored by examining age changes between T2 and T3 in responses to mothers' simulated pain (the one similar episode videotaped in the home at 18 months and in the laboratory at 2 years). Chil- dren showed more prosocial behavior, /(26) = 2.27, p < .05,
Table 3 Childrens Reactions to Mothers' Simulations of Distress: Comparisons Across Two Settings at Time 2
Mothers' reports of
responses to distress
simulations (18-20
months)
Observations based on
videotapes of responses to
distress simulations (18 and 21
months)
Children's reactions M SD M SD
Prosocial behavior (PB) Empathic concern (EC) Hypothesis testing (HT) Self-reference (SR)
.29
.23
.16
.19
.28
.26
.19
.19
.35
.31
.25
.07
.22
.23
.22
.11
Note. Simulations for these two data sets were performed on different occasions. PB = Prosocial behavior; EC = Empathic concern; HT = Hypothesis testing; SR = Self-reference.
DEVELOPMENT OF CONCERN 133
Table 4 Longitudinal (L) and Control (C) Childrens Responses to Videotaped Laboratory Simulations of Distress at Two Years (Percent of Children)
Children's reactions
Pain (Mother)
L" C
Simulations
Pain (Experi- menter
1)
L C
Pain (Experi- menter
2)
L C
Cry (Infant tape)
L C
Prosocial behavior (PB) 78 85 4 10 15 15 30 30
Empathic concern (EC) 50 60 50 63 48 55 41 40
Hypothesis testing (HT) 65 65 31 32 33 40 74 75
Self-reference (SR) 19 15 4 0 4 25 0 10
Note. PB = Prosocial behavior; EC = Empathic concern; HT = hy- pothesis testing; SR = self-reference. * = longitudinal sample. b = sample seen only at 2 years (n = 20 be- cause of staffing constraints).
MT2 = .30 and MT3 = .58, and hypothesis testing, f(26) = 3.95, p < .001, MT2 =. 30 and MT3 = .65, as they grew older, corrobo- rating in part the developmental patterns described earlier based on maternal reports. Concern, MT2 = .37 & MT3 = .46, and self-reference, MT2 = .04 and MT3 = .12, did not change with age.
Further Replication Analyses
At 2 years of age, the target sample's responses to the four videotaped, simulated distresses in the laboratory were com- pared with responses of the new peer sample. The percentages of children responding in each group for each of the prosocial measures are presented in Table 4. Scores for the two groups are very similar, indicating both (a) that the patterns are replicated in a second sample and (b) that exposure to distress simulations throughout the second year did not appear to alter children's later responses to distress. Although prosocial behaviors are mainly directed toward their own mothers, children are not insensitive to the distress of the experimenter and infant. They show high levels of empathic concern for both the experimenter and infant and high levels of hypothesis testing with regard to the infant.
Discussion
During the second year of life, children develop a capacity to intervene on behalf of others. The prosocial interventions take a variety of forms, including helping, sharing, and comforting victims in distress. These action-oriented patterns begin to re- place the more intense emotional arousal, identified in pre- vious research as characteristic of infants (e.g., Sagi & Hoffman, 1976) as well as in our earlier work (Zahn-Waxler & Radke- Yarrow, 1982).5 Personal or self-distress is increasingly replaced
by more modulated, constructive, action-oriented patterns. The transition is not fully accomplished during this period as individuals clearly struggle between concern for the needs of self and the welfare of others throughout the life cycle. The evidence for early moral internalization, however, highlights the need to reformulate theories emphasizing the egocentrism and narcissism of young children. This longitudinal study sup- ports our earlier empirical work on prosocial and reparative behaviors, in demonstrating similar developmental patterns with a second, independent sample of mothers and children. It supports as well the integrative, conceptual work of Hoffman, who was the first to articulate a comprehensive theory of early moral development. His theory emphasizes the transformation during the second year of life from self-concern to empathic concern for others, grounded in social-cognitive changes re- flecting the development of self-awareness and self-other dif- ferentiation.
Behavioral expressions of concern for others do appear to emerge in conjunction with the development of self-awareness and role-taking abilities. Empathic, other-oriented patterns were linked in development to self-recognition in this sample of children. Self-referential behaviors (e.ĝ pointing out one's own injury when another is injured) may also reflect some capacity for role-taking. Such responses are sometimes interpreted as egocentrism and self-concern. Rather, at this age, they may serve to connect another's experience to the child's own and hence increase comprehension of the other's experience. Our data support this interpretation: Self-referential responses were correlated with self-recognition; they predicted later prosocial behaviors and empathic concern, and they were unrelated to self-distress.
It is not clear from the present research design whether so- cial-cognitive capacities are prerequisites for more mature em- pathic involvements. It is probably not coincidental that in- creases in prosocial development occur at time periods (around IV2 and 2 years) in which there are corresponding increases in the use of symbols and the capacity to imagine and pretend. The development of representational ability permitting aware- ness and understanding of others' needs may facilitate the ca- pacity to act on behalf of another individual. It is plausible as well that children's emotions and behaviors when others are in need may also guide the ways in which children come to under- stand these experiences. These issues merit further exploration in research designs with larger samples, more controlled assess- ments, and sequential analyses of the dynamic interplay of cog- nitive, affective, and behavioral components of empathic pat- terns.
The similarity of children's constructive interventions when they cause and witness distress suggests common developmen- tal pathways and conceptual underpinnings of prosocial and reparative behaviors. This is consistent with Hoffman's hypoth-
s In our first developmental study, self-distress showed a significant decrease with age during the second year of life. In the present re- search, the decreases in self-distress with age to both witnessed natural and witnessed standard distresses (Table 1) were not significant. This may reflect the fact that the children in the first research sample were somewhat younger at the beginning of the study (9 months for the first cohort) and initially showed more self-distress than the children here.
134 ZAHN-WAXLER, RADKE-YARROW! WAGNER, AND CHAPMAN
esis that empathic arousal similarly mediates internalization behaviors reflecting both altruism and conscience. In our re- search we have refrained from a priori use of such constructs, preferring first to describe the emotions, behaviors, and verbal- izations that characterize children's response patterns and only then to make inferences about their potential relevance to con- cepts such as altruism and conscience. It seems reasonable to infer (but not possible to demonstrate fully) that the patterns identified here reflect early forms of caring and remorseful behaviors that underlie more mature forms of compassion and conscience in children. This interpretation is strengthened by consistently observed linkages between prosocial actions, ex- pressions of concern, and verbal attempts to comprehend the nature of distress events that children cause and witness.
Although there is evidence for common developmental tra- jectories for altruism and conscience, there are also indications of differences. Children's responses to distresses they caused and witnessed differed in a number of important respects in our study. Children's prosocial behaviors relate quite differently to patterns of self-distress in these two situations, suggesting possible differences in mediating mechanisms. Moreover, when children caused physical or psychological harm, they showed more enjoyment, more aggression, more personal dis- tress, less concern for the other and were less likely to explore the reasons for distress than when they witnessed distress as bystanders. Thus, reparative behaviors potentially indicative of early conscience are more likely to develop within contexts of turbulence, conflict, and ambivalence. This is evidenced not only in children's responses, but also in caregivers' reactions, as they begin to differentiate these contexts for children (e.g., using more forceful discipline practices when children cause distress, (Zahn-Waxler et al., 1979). This period in development (the sec- ond half of the second year of life) may mark the evolution of two different forms of harmdoing that are important, both for caregivers and for researchers, to distinguish, namely, acciden- tal versus intentional harms. Further conceptual work is needed to address relations between prosocial patterns and in- tentional aggression versus accidental harmdoing and the im- plications for theories about empathy.
Additional work is also needed to assess better the range of motives and reasons that underlie the development of overt prosocial acts. For example, actions based on fear and anxiety may be motivated by personal distress, whereas actions based on empathic concern or sadness may reflect sympathetic dis- tress. Both may be distinguishable from more matter-of-fact, perfunctory, or sometimes self-serving actions that also benefit others. More sophisticated systems for assessing affect (e.g., psy- chophysiological measures, facial affect coding) could aid in making these kinds of distinctions (Eisenberg et al., 1989).
Self-interest may partially motivate some of the prosocial behaviors seen here. Many of the children's responses were to their mothers on whom they are clearly dependent for their physical and psychological well-being. Yet many times the sig- nificant components of actions reflecting caring and concern were observed to be directed toward strangers (in the labora- tory) even at this early age. And based on the mothers' naturalis- tic observations, it was also evident that children directed pro- social actions toward others both within and outside the family setting. Because one's well-being is inevitably tied to the receipt
as well as the provision of benevolent actions, the issue of self- less versus selfish altruism (Batson & Shaw, 1991) is not likely to be readily or fully resolved. The behaviors shown by children here, however, are consistent with Trevarthen's view that early forms of "cooperative awareness" between mother and child begin to create a world of shared meaning, empathic under- standing, and appropriate linking of one's own emotions with those of others, that then generalize beyond the mother-child dyad.
Boys and girls did not differ in their interventions on behalf of others but did differ on some of the correlates or possible mediators of prosocial behavior. Girls showed more concern and reproduced or imitated the affective experience of the other more often than did boys. Typically, these are seen as indicators of joining in to the emotional experiences of others, and there is a long history of discussion regarding why females are some- times more prone than males to participate in others' affective states (e.g, Brody, 1985; Eisenberg & Lennon, 1983). Although evidence is somewhat contradictory, there are indications that females are more empathic than males. Sex differences are strongest when self-report measures are used, but recent obser- vational research (Zahn-Waxler, Robinson, & Emde, 1991) with children as young as 14 months confirms greater empathic and prosocial involvement in other people's distress by girls than boys. This may reflect a biologically based predisposition that begins to prepare females from early on for the caregiving role. At the same time, as soon as children are born, there are strong socialization pressures, both direct and indirect, for females to be responsive to the physical and emotional needs of others.
Individual differences in young children's responses to dis- tress can be predicted from different socialization experiences. In research by Zahn-Waxler et al. (1979), maternal sensitivity and reasoning were positively related to children's empathic, prosocial behaviors during the second year of life. A caregiver's emphasis on the child's accountability and responsibility for others' problems may foster guilt as well as empathy. An overem- phasis on the child's causal role could lead some children to blur the distinction between problems that they observe and those that they cause in others and to mistakenly assume that they have created problems for which they are not responsible. Emotionally vulnerable, depressed caregivers sometimes may foster these feelings (Zahn-Waxler, Kochanska, Krupnick, & McKnew, 1990). This could lead initially to frequent prosocial behavior, especially toward the caregiver.
Other investigators have found that children of depressed caregivers do initially attempt to repair interactions by affec- tively engaging the parent, even in the first year of life (e.g., Cohn et al., 1990). In our study, children were old enough to engage their mothers behaviorally as well as affectively. We did not find differences, however, between children of depressed and well mothers with this relatively small sample. It is possible that depressed mothers were more or less selective than nonde- pressed mothers in what they recorded, and they could have differed on accuracy as well. That is, systematic differences in manner of recording of the two groups of mothers may have negated differences. The two mothers who consistently had dif- ficulty tape recording incidents as indicated earlier both experi- enced depression. It is worth noting, however, that at different age points, data obtained by our own investigators in the home
DEVELOPMENT OF CONCERN 135
and in the laboratory, rather than from maternal reports (Ta- bles 3 and 4) also failed to yield significant differences between depressed and well mothers. Because the psychiatric diagnostic interview was administered only once, it is not possible to de- termine whether unmeasured changes in depressive symptoms over time could have contributed to unreliable findings. Thus, it still remains a viable hypothesis that the two groups differ and that more adequate methodologies will detect these differ- ences.
Methodological issues pertaining to the use of caregivers as observers merit further consideration. Characteristics of chil- dren's developing reactions to distress in others could have been influenced by sampling problems (e.g, mothers' changing focus on what they consider to be salient emotional events). The simi- lar patterns of findings based on comparisons of children's reac- tions to simulated events (both those reported by mother and those videotaped by us in the home and in the laboratory) with their reactions to those events that mothers chose to report helps to alleviate this concern. The independent assessments also help to assure that developmental patterns were not the result of mothers' own theories of development.
In summary, behaviors reflecting concern for others emerge during the second year of life. The developmental transitions replicate those obtained in earlier research (Zahn-Waxler & Radke^arrow, 1982) and mothers' observation reports are cor- roborated by videotaped observations of children's reactions. The data suggest early origins for altruism and conscience in children and identify both commonalities and differences in the developmental pathways. Young children seem to show pat- terns of moral internalization that are not simply fear based or solely responsive to parental commands. Rather, there are signs that children feel responsible for (as well as connected to and dependent on) others at a very young age. There are early indi- vidual differences in these capacities as well, and further re- search might be directed to familial and constitutional patterns that contribute to these differences.
References
Amsterdam, B. (1972). Mirror self-image reactions before age two. De- velopmental Psychobiology, 5(4), 297-305.
Bates, E., Bretherton, I.,&Snyder, L. (1988). From first words to gram- mar: Individual differences and dissociable mechanisms. Cambridge, England: Cambridge University Press.
Bates, J. E., & Bayles, K. (1984). Objective and subjective components in mothers' perceptions of their children from age 6 months to 3 years. Merrill-Palmer Quarterly, 30,111-130.
Batson, C. D., & Shaw, L. L. (1991). Evidence for altruism: Toward a pluralism of prosocial motives. Psychological Inquiry, 2(2), 107-122.
Bertenthal, B. I , & Fischer, K. W (1978). The development of self-re- cognition in the infant. Developmental Psychology, 14, 44-50.
Bloom, L , Lightbown, P., & Hood, L. (1975). Structure and variation in child language. Monographs of the Society for Research in Child Development, 40 (2, Serial No. 160).
Bretherton, I., Fritz, I , Zahn-Waxler, C, & Ridgeway, D. (1986). The acquisition and development of emotion language: A functionalist perspective. Child Development, 57, 529-548.
Brody, L. R. (1985). Gender differences in emotional development: A review of theories and research. Journal of Personality, 55(2), 102- 149.
Brothers, L. (1989). A biological perspective on empathy. American Journal of Psychiatry. 146(1), 10-19.
Bruner, X S. (1972). The nature and uses of immaturity. American Psy- chologist, 27,1-22.
Butterworth, G. (1980). A discussion of some issues raised by Piaget's concept of childhood egocentrism. In M. V Cox (Ed.), Are young children egocentric? (•pp. 17-40). York, England: Batsford Academic and Educational, Ltd.
Campos, J. J., Barrett, K. C, Lamb, M. E., Goldsmith, H. H., & Sten- berg, C. (1983). Socio-emotional development. In P. H. Mussen (Ed.), Handbook of child psychology (4th ed., pp. 783-915). New York: Wiley.
Caplan, M. Z , & Hay, D. F. (1989). Preschoolers' responses to peers' distress and beliefs about bystander intervention. Journal of Child Psychology and Psychiatry, 30(2), 231-242.
Cohn, J. F, & Tronick, E. Z. (1983). Three-month-old infants' reaction to simulated maternal depression. Child Development, 54,185-193.
Cohn, J. F, Campbell, S. B., Matias, R., & Hopkins, J. (1990). Face-to- face interactions of postpartum depressed and nondepressed mother-infant pairs at 2 months. Developmental Psychology, 26,15- 23.
Cummings, E. M , Zahn-Waxler, G, & Radke-Yarrow, M. (1981). Young children's responses to expressions of anger and affection by others in the family. Child Development, 52,1274-1282.
Darwin, C. (1872). The expression of the emotions in man and animal. London: John Murray.
Eckerman, C. Q, Whatley, J. L., & Kutz, S. L. (1975). Growth of social play with peers during the second year of life. Developmental Psy- chology, 11, 42-49.
Eisenberg, N., Fabes, R. A., Miller, P. A., Fultz, J., Shell, R., Mathy, R. M., & Reno, R. R. (1989). Relation of sympathy and personal distress to prosocial behavior: A multi-method study. Journal of Per- sonality and Social Psychiatry, 57, 55-66.
Eisenberg, N., & Lennon, R. (1983). Sex differences in empathy and related capacities. Psychological Bulletin, 94,100-131.
Eisenberg-Berg, N., & Lennon, R. (1980). Altruism and the assessment of empathy in the preschool years. Child Development, 51, 552-557.
Emde, R., Johnson, W, & Easterbrooks, M. A. (1987). The do's and don'ts of early moral development: Psychoanalytic tradition and current research. In J. Kagan & S. Lamb (Eds.), The emergence of morality in young children (pp. 245-276). Chicago: University of Chicago Press.
Field, T., Woodson, R., Greenberg, R., &Cohen, D. (1982). Discrimina- tion and imitation of facial expressions by neonates. Science, 218, 179-181.
Freud, S. (1958). Civilization and its discontents. New York: Doubleday. (Original work published 1930).
Goodenough, F. (1931). Anger in young children. Minneapolis: Univer- sity of Minnesota Press.
Grusec, J. E. (1991). Socializing concern for others in the home. Devel- opmental Psychology, 27, 338-342.
Haviland, J. M , & Lelwica, M. (1987). The induced affect response: 10-week-old infants' responses to three emotion expressions. Devel- opmental Psychology, 23, 97-104.
Hoffman, M. L. (1975). Developmental synthesis of affect and cogni- tion and its interplay for altruistic motivation. Developmental Psy- chology, 11, 607-622.
Hoffman, M. L. (1982). Development of prosocial motivation: Em- pathy and guilt. In N. Eisenberg (Ed.), The development of prosocial behavior (pp. 281-313). San Diego, CA: Academic Press.
Izard, C. E. (1977). Human emotion. New York: Plenum Press. Kagan, I , & Lamb, S. (1987). The emergence of morality in young chil-
dren. Chicago: University of Chicago Press.
136 ZAHN-WAXLER, RADKE-YARROW, WAGNER, AND CHAPMAN
Lewis, M., Sullivan, M., Stanger, G, & Weiss, M. (1989). Self develop- ment and self-conscious emotions. Child Development, 60,146-156.
MacLean, P. D. (1985). Brain evolution relating to family, play, and the separation call. Archives of General Psychiatry, 42, 405-417.
McCall, R. B. (1979). Qualitative transitions in behavioral develop- ment in the first two years. In M. H. Bornstein & W Kessen (Eds.), Psychological development in infancy (pp. 183-224). Hillsdale, NJ: Erlbaum.
Meltzoff, A. N , & Moore, M. K. (1977). Imitation of facial and manual gestures by human neonates. Science, 198, 75-78.
Nicolich, L. (1977). Beyond sensorimotor intelligence: Assessment of symbolic maturity through analyses of pretend play. Merrill-Palmer Quarterly, 23, 89-99.
Piaget, J. (1962). Play, dreams and imitation in childhood. New \brk: Norton.
Piaget, J. (1965). The moral judgment of the child. New York: Harcourt Brace & World. (Original work published in 1932.)
Radke-Yarrow, M., Campbell, J. D., & Burton, R. V (1970). Recollec- tions of childhood: A study of the retrospective method. Mono- graphs of the Society for Research in Child Development, 35 (5, Serial No. 138).
Radke-Yarrow, M., & Zahn-Waxler, C. (1976). Dimensions and corre- lates of prosocial behavior in young children. Child Development, 47, 118-125.
Radke-Yarrow, M., & Zahn-Waxler, C. (1984). Roots, motives, and pat- terning in children's prosocial behavior. In E. Staub, D. Bar-Tal, J. Karylowski, & J. Raykowski (Eds.), The development and mainte- nance of prosocial behavior: International perspectives on positive mo- rality. New York: Plenum Press.
Radke-Yarrow, M., & Zahn-Waxler, C. (1990). Research on children of affectively ill parents: Some consideration for theory and research on normal development. Development and Psychopathology, 2, 349- 366.
Radke-Yarrow, M., Zahn-Waxler, C, & Chapman, M. (1983). Children's prosocial dispositions and behavior. In P. H. Mussen & E. M. Heth- erington (Eds.), Handbook o]"child psychology: Socialization, person- ality and social development (Vol. 4, pp. 469-545). New York: Wiley.
Rheingold, H. L., & Emery, G. N. (1986). The nurturant acts of very young children. In D. Olweus, J. Block, & M. Radke-Yarrow (Eds.), The development ofanti- and prosocial behavior (pp. 75-94). San Diego, CA: Academic Press.
Rheingold, H., Hay, D., & West, M. (1976). Sharing in the second year of life. Child Development, 47,1148-1158.
Ross, H. S., & Goldman, B. D. (1977). Establishing new social relations
in infancy. In T. Alloway, P. Pliner, & L. Krames (Eds.), Advances in the study of communication and affect: Vol. 3. Attachment behavior (pp. 61-79). New York: Plenum Press.
Sagi, A., & Hoffman, M. L. (1976). Empathic distress in the newborn. Developmental Psychology, 12,175-176.
Simner, M. L. (1971). Newborn's responses to the cry of another infant. Developmental Psychology, 5,136-150.
Spitzer, R. L., & Endicott, J. (1979). Schedule for affective disorders and schizophrenia - Lifetime version (3rd ed.). New York: New York State Psychiatric Institute, Biometrics Research.
Stern, D. (1985). The interpersonal world of the infant. New York: Basic Books.
Stipek, D. J., Gralinski, J. H., & Kopp, C. B. (1990). Self-concept devel- opment in the toddler years. Developmental Psychology, 26, 972- 977.
Termine, N. X, & Izard, C. E. (1988). Infants' responses to their mothers' expressions of joy and sadness. Developmental Psychology, 24, 223-229.
Trevarthen, C. (1989). Origins and directions for the concept of infant intersubjectivity. Society for Research in Child Development, news- letter, pp. 1-4.
Watson, M., & Fischer, K. (1977). A developmental sequence of agent use in late infancy. Child Development, 48, 828-836.
Yarrow, M., Scott, P., & Waxier, C. (1973). Learning concern for others. Developmental Psychology, 8, 240-260.
Zahn-Waxler, C, Iannotti, R. J., Cummings, E. M., & Denham, S. (1990). Antecedents of problem behaviors of children of depressed mothers. Development and Psychopathology, 2, 271-291.
Zahn-Waxler, C, Kochanska, G., Krupnick, J., & McKnew, D. (1990). Patterns of guilt in children of depressed and well mothers. Develop- mental Psychology, 26, 51-59.
Zahn-Waxler, C, & Radke-Yarrow, M. (1982). The development of altru- ism: Alternative research strategies. In N. Eisenberg(Ed.), The devel- opment of prosocial behavior (pp. 109-137). San Diego, CA: Aca- demic Press.
Zahn-Waxler, C, Radke-Yarrow, M., & King, R. A. (1979). Child-rear- ing and children's prosocial initiations toward victims of distress. Child Development, 50, 319-330.
Zahn-Waxler, C, Robinson, 1, & Emde, R. (1991, April). The develop- ment and heritability of empathy. Poster presented at the meeting of the Society for Research in Child Development, Seattle, WA.
Received September 5,1990 Revision received June 13,1991
Accepted July 22,1991 •
