J. Child Psychol. Psychial. V o l . 22, p p 1 li> 18. Pi-rgamon Press L i d . 1981. Printed in G r e a t Hritaiti. & Ass(i( iaticiti lor CliikI P.syrholi.Hy .iiid Psyc liiati y. 0012-9(1:«I/8I)/OH)1-0(K)I S02 00/(J THE REACTION OF FIRST-BORN CHILDREN TO THE BIRTH OF A SIBLING: MOTHERS' REPORTS JUDY DUNN,* CAROL KENDRICK and ROSANNEMACNAMEE Medical Research Council Unit on the Development and Integration of Behaviour, University of Cambridge INTRODUCTION clinicians have attached considerable significance to the effects of the birth of a sibling on a first-born child (Levy, 1934; 1937; Petty, 1953; Thomas and Chess, 1977; Winnicott, 1977) there has been httle systematic study of the changes in behaviour following the arrival of a sibling, or of their prognostic importance (Black and Sturge, 1979). Also, while individual differences in the response of children to potentially stressful events have been frequently noted (Rutter, 1977) little is known about their nature or about the factors associated with them. In Cambridge a sample of 40 families has been studied over the period when a second child is born, and through the infancy of the second child (Dunn and Kendrick, 1979; 1980a). Marked changes in the interaction between mother and first-born child after the sibling birth were observed, with decreases in measures of maternal attention and play, and increases in confrontation between mother and child (Dunn and Kendrick, 1980a). This paper presents data derived from mothers' reports on their first-born's behaviour. In the first section the reactions of the first-born children are described, and in the second, the association between different aspects of the children's reactions is considered. In the third section the factors associated with the different reactions are examined: age, sex, temperament, the mothers' state in the post partum weeks, and changes in the mother-first-born interaction are among the factors considered. WHILST METHODS Procedures. Forty families were studied over a period from 1 to 3 months before the birth of a second child until the second child was 14 months old. An outline of the structure of the study is given in Table 1. In this paper the interviews and observations carried out during the pregnancy are referred to as the pre-sib-birth interviews and observations, and those carried out in the second and third week after the birth of the sibling as the post-sib-birth interviews and observations. The first-born child is referred to as the child, and the second child as the baby or the sibling. The population studied. The families taking part in the study were collected through the help of General Practitioners and Health Visitors. The families were largely •Requests for reprints to: J. Dunn, MRC Unit on the Development and Integration of Behaviour, University Sub-Department of Animal Behaviour, Madingley, Cambridge CB3 8AA, U.K. Accepted manuscript received 15 February 1980 1 2 JUDY DUNN, CAROL KENDRICK AND ROSANNE MACNAMEE working class: the occupation of the father, classified according to the Registrar Ceneral (1973) was IV for 7 families. III manual for 12 families. III white collar for 13 families, and II for 8 families. Thirteen of the mothers delivered the second baby at home. There were 21 first-born boys and 19 first-born girls. The age range of the children was from 18 to 43 months at the birth of the sibling (median age 25 months). At the time of post-sib-birth interview 17 of the seeond babies were being breast-fed, and 16 were bottle-fed. A further 7 were having some combination of breast and bottle. TABLE 1. OUTLINE OF STUDY Mother pregnant with second child 2 pre-sib-birth observations with mother ± father Interview Rating of temperamental characteristics Birth of second child 2 weeks post-birth 3 weeks post-birth Post-sib-birth observation Observation Interview Child with mother + sibling (inc. feed) (± father) 2nd child 8 months old 2 observations: siblings with mother (± father) Interviews Rating of temperamental characteristics 2nd child 14 months old 2 observations: siblings with mother (± father) Interviews Collaboration. The pre-sib-birth and post-sib-birth interviews were developed in collaboration with Dr. Claire Sturge of Northwick Park Hospital. The Assessment Temperamental Characteristics Interview was developed by Dr. Sturge. Pre-sib-birth and post-sib-birth interviews. The interview conducted in the visits made before the sibling was born included questions on the child's feeding, sleeping, and toilet habits, on attention-seeking behaviour, independence and dependence, fears, worries, miserable moods and so on. The questions were concerned with detailed descriptions of the child's behaviour in specific situations, and the answers were coded on three 4-point scales. The post-sib-birth interview included the same questions; thus changes in the child's behaviour, as reported by the mother, can be analysed in terms of changes in the scores on these scales. The post-sib-birth interview also included questions concerned with the first child's behaviour when the new baby was being fed, changed and played with or cuddled. The interview, then, was concerned with the first child's reaction to the events surrounding the birth of the sibling and his/her behaviour towards the new baby, with the mother's view of these reactions, and her own state. Observation measures. Details of the method of observation, categories of behaviour used, reliabilities, etc. are given in Dunn and Kendrick (1980a). THE REACTION OF FIRST-BORN CHILDREN TO THE BIRTH OF A SIBLING 3 SECTION 1; FREQUENCY OF DIFFERENT REACTIONS IN THE FIRST-BORN CHILDREN Results (1) Disturbance and negative reactions in thefirst-born children (Table 2). (i) Increases in tearfulness, clinging, withdrawal, demanding, and negative behaviour towards the mother Table 2 (a) shows that 24 children were reported to have become more tearful, 13 of these children showing this increased tearfulness frequently or constantly. Twenty-three children were reported to have become more clinging. Eleven children were reported to have become more withdrawn, quieter and less outgoing since the sibling birth. Table 2 (a) also shows that increases in demanding behaviour and negative behaviour towards the mother, either by physical or verbal opposition or attack, were common. TABLE 2. NUMBERS OF CHILDREN REPORTED TO SHOW CHANGES IN BEHAVIOUR AFTER SIBLING BIRTH {N = 40) (a) Increases in behaviour Rare/absent Occasional Tearful Clinging Withdrawn Demanding Negative behaviour towards M. Frequent Constant 16 17 29 3 9 11 12 9 13 8 10 9 2 22 22 3 (b) Increase in incidence of problems (4-point scale) No change 1-point increase 2-point increase Extreme breakdown (1) Sleep problems (2) Feed problems (3) Toilet training problems 29 37 14 26) 2 2 1 (ii) Change in incidence ofproblems (Table 2b) (a) Sleeping problems had been rated on a 4-point scale in the pre-birth interview, on which the coding distinguished no sleep problem, occasional sleep problem (less than weekly), a marked problem, or a daily problem. For 7 children there was an increase of more than 1 point. (b) Feeding problems were rated in the pre-sib-birth interview according to a 4-point scale. For 3 children a 1-point change was noted; for 37 no change was reported. (c) Twenty-six children were toilet trained before the sibling birth. Of these, 14 showed no change, 4 had 1 or 2 accidents (1 point change), 5 now also regularly required the mother's help (2 point change), and for 3 there was an extreme breakdown. (2) Behaviour towards the new baby (Tables 3 and 4). Some children were thought by their mothers to be deliberately irritating the new baby, by jogging him/her when he/she was asleep, removing his/her dummy, taking the bottle away, or overwhelming the baby with physical attention. Table 3 shows that such behaviour was JUDY DUNN, CAROL KENDRICK AND ROSANNE MACNAMEE TABLE 3. BEHAVIOUR TOWARDS THE NEW BABY: NUMBERS OF CHILDREN SHOWING IRRITATING OR NEGATIVE BEHAVIOUR, OR IMITATING THE BABY Behaviour towards the new baby Irritates baby Negative behaviour to baby Imitates baby Absent/rare Occasional 19 31 10 Frequent 13 6 15 reported as occasional in 13 children, frequent in 8. Table 3 also shows that direct negative behaviour towards the baby, such as hitting, pinching, squeezing or poking was not common. Some children were reported to imitate their baby sibling with noises, grimaces, crying, lying on the changing mat, or sucking from the bottle. Such acts were noted occasionally in 15 children, and frequently in 15. Table 4 gives the incidence of aspects of positive interest in the baby. TABLE 4. NUMBERS OF GHILDREN SHOWING POSITIVE INTEREST IN THE BABY SIBLING (N Absent or rare Occasional Verbal references and comments Concern shown when baby cries Attempts to help M. in care of baby Entertains baby Affectionate physical interest Frequent 8 16 2 18 20 12 14 16 ,11 21 11 10 21 8 9 = 40) Constant 1 1 1 (3) Signs of regression. Twenty-eight mothers had noted signs of regression. For 15 children these were mild—occasional baby talk, wanting to be fed, or carried around, etc. For a further 13 children there was some definite regressive step over toilet training, or a regular insistence on being fed when the child previously fed himself/herself, and so on. (4) Signs of being more 'grown up'. Several children were reported by their mothers to have become more 'grown up' or independent over the three weeks since the new baby was born. The most common change was an increase in independence: this was noted in 9 children, who were reported to have become more independent about feeding themselves, about going to use the potty on their own, etc..Other children were reported to have given up the bottle, and three mothers reported that they had noticed an improvement in their child's language ability. In all, 25 children were reported to have shown at least some sign of being more grqwn up. (5) Changes in the first child's daily life. Mothers were asked to describe any changes in the child's daily life over the period when the sibling was born and the following two weeks which they thought might be important. Probe questions concerned the number of child or adult visitors the family had had, and how much the child had been taken out. Only 17 mothers thought that the child's routine had not altered. Sixteen ehildren were reported to have been taken out less. (6) Reaction of mother after the birth of the second child. The mothers were asked how they felt 2, 3 and 4 weeks after the birth of the second child. Answers were coded on THE REACTION OF FIRST-BORN CHILDREN TO THE BIRTH OF A SIBLING 5 a 4-point seale: at 2 weeks, 4 mothers reported feeling fine, 15 mothers were tired but eoping, 16 mothers were very tired, and 5 were very tired and depressed. By 4 weeks post partum the number of mothers reporting that they felt fine had increased from 4 (out of 40) to 9 (out of 35); there were still 9 out of 35 who felt very tired and three of these reported feeling depressed. Hours of sleep: 22 out of 40 mothers reported that they were getting less than 5 54 hours sleep per 24 hours at 3 weeks post partum. SECTION II. ASSOCIATION BETWEEN INDICES OF REACTION In this section, the question of how far the different aspects of disturbance, negative behaviour, and positive interest in the baby are associated is considered. Were the children who showed, for example, a very marked increase in negative behaviour towards the mother, the same ehildren who were least likely to show positive interest in the new baby? The relationships between the various aspects of disturbance, changed behaviour, and positive behaviour towards the baby sibling were examined, using a log-linear contingency table analysis (Bishop et al., 1975). This is simply an extension of the more familiar 2 x 2 contingency table analysis, in which the null hypothesis of no interaction between the variables is postulated, and the chi-square statistic is used to test this. With the log-linear method it is possible to describe the relationship between several categorical variables simultaneously, and to test for interaction effects between them. In this analysis, if the null hypothesis of no interaction between any variables is rejected, there are a variety of possible alternatives that can be tested. In Appendix 1 the procedure is outlined. The relationship between the following 'reaction indices' was examined: increase in withdrawal, in clinging, in tearfulness, in negative behaviour to the mother, in demanding behaviour, in sleeping problems; irritating behaviour towards the baby, imitation of the baby, aspects of positive behaviour towards the baby, signs of being grown up, and signs of regression. The number of variables used in each analysis is restricted by the number of cases; thus in the following analyses 4 variables were examined in each analysis. Resutts (1) Relationship between the following indices of positive behaviour towards the baby: interest in the baby (verbal references); helpful with baby; entertains the baby; affectionate physical interest in the baby. (a) Model (I): no interaction between variables. The G-statistic for goodness of fit for the model = 30.95 (11 df significant at the P< 0.005 level). This indicates a poor fit, and implies that there were interaction effects between the variables. (b) Model (II): alt 2-factor interactions fitted. The G-statistic for goodness-of-fit for this model = 3.65 (5 df). This low value implies that the model fits well. (c) Tests for interaction effects. Tests for interaction effects between each pair of variables were carried out. Table 5 shows that the interactions between the indices of 'interest (verbal)' and 'helpful', and 'interest (verbal)' and 'entertains' are highly significant, and that between 'interest (verbal)' and 'affectionate physical interest' is significant at the 10% level. (d) Modet (III): includes interactions 'interest {verbal)' 'hetpfut' and 'interest {verbal)' 'entertains'. T h e G-statistic for goodness-of-fit = 7.60 (9 df). This low value indicates a good fit. Inspection of the contingency tables showed that the associations between the indices were positive. (e) Percentage of deviance accountedfor by Modet (III). This was 87 % . Since these results indicate positive association between these different aspects of positive interest in 6 JUDY DUNN, CAROL KENDRICK AND ROSANNE MACNAMEE TABLE 5. INTERACTIONS BETWEEN INDICES OF POSITIVE BEHAVIOUR TOWARDS THE BABY Deviance Model (I): no interactions Model (I): + interaction 'interest (verbal)', 'helpful' Model (I): -l- interaction 'interest (verbal)', 'entertains' Model (I): -H interaction 'interest (verbal)', 'affectionate physical interest' Test statistic for interaction ,^ Significance level df level df 30.95 11 20.31 10 30.95-20.31 = 10.64 1 0.005 19.24 10 30.95-18.24=12.71 1 0.001 28.01 10 30.95-28.01 = 2.95 1 0.10 the baby, in the analyses which follow we have used a single measure of 'positive interest' in the baby: this is the combined score on the 4 measures. The use of a score derived from a combination of measures involves the assumptions that each type of behaviour is equally important, and that the relationships between these different types of behaviour is additive. These are of course unwarranted assumptions. However, when the analyses were carried out using each separate index of positive interest the results were substantially the same as those found using the combined index. For clarity of presentation therefore we report here just the analyses using the combined score. (2) Relationship between increases in clinging; tearfulness; withdrawal; and 'positive interest' in the baby. (a) Model (I): no interaction between variables. The G-statistic for goodness-of-fit for the model = 20.46 (11 df, P< 0.05). This indicates a poor fit, and implies that there are interactions present. (b) Modet (II): includes all 2-factor interactions. The G-statistic for goodness of fit = 7.75 (5 df). This low value implies that the model fits well, and that there are interactions present. (c) Tests for interaction effects. Tests for interaction effects between each pair of variables were carried out. Table 6 shows that the interactions between increase in clinging and withdrawal, and between increase in clinging and tearfulness were significant. TABLE 6. INTERACTIONS BETWEEN INCREASES IN CLINGING, TEARFULNESS, WITHDRAWAL Model (I): with no interaction Model (I): with interaction 'clinging', 'withdrawal' Model (I): with interaction 'clinging', 'tearfulness' Test statistic for interaction df Significance level 10 20.46-11.51=3.95 1 0.05 10 20.46-14.17 = 6.29 1 0.025 Deviance df 20.46 11 16.51 14.7 (d) Model (III): includes the two interactions above. The G-statistic for goodness-of-fit = 10.22 (9 df), indicating a good fit. The fitted values using this model were close to the observed values, confirming the finding. Inspection of the contingency tables showed that the association between the increases in clinging and tearfulness was positive. (e) Percentage of deviance accounted for by Model (III): This was 75 % . These results show that there were positive associations between the reactions of increased withdrawal and clinging, and between increased clinging and tearfulness. The relationships between withdrawal, clinging and 'positive interest' in the baby were further investigated in the next analysis which included examination of 3-factor interactions. (3) Relationship between increase in withdrawal, clinging, imitation of the baby and 'positive interest' in the baby. THE REACTION OF FIRST-BORN CHILDREN TO THE BIRTH OF A SIBLINC 7 (a) Model (I): no interactions between variables. The G-statistic of goodness-of-fit for the model = 22.59 (11 df, P< 0.05). This indicates a poor fit, and implies that there are interactions. (b) Model (II): includes all 2-factor interactions. The G-statistic of goodness-of-fit for the model = 9.833 f) (c) Tests for interaction effects. Since the previous analysis [(2), above] had shown a significant interaction between withdrawal and clinging, this interaction was included in all subsequent models; each of the remaining 2-factor interactions were added in turn to this model (III). Table 7 shows that the interaction between increase in clinging and imitation of the baby was significant. Inspection of the contingency table showed that this interaction was negative: imitation of the baby was less likely in children reported to have increased in clinging. TABLE 7. INTERACTIONS BETWEEN INGREASE IN WITHDRAWAL, GLINGING AND IMITATION OF THE BABY Deviance df Model III [model (I) + interaction increase in withdrawal, increase in clinging] 18.64 10 Model III + interaction increase in clinging, imitation of the baby 12.91 9 Test statistic df Significance level 18.64-12.91=5.73 1 0.025 (d) Model (IV): includes interactions between withdrawal and clinging and between clinging and imitation. T h e G-statistic for goodness of fit for this model = 12.91 (9 df). The possibility that there were 3-factor interaction effects was next investigated. (e) Testingfor 3-factor interaction effects. The 3-factor interactions which were tested were: (i) increases in clinging/withdrawal/imitation; (ii) increases in clinging/withdrawal/positive interest; (iii) increases in clinging/imitation/positive interest. These were fitted in turn. The G-statistic for the interaction (ii) above = 3.04. This indicates that the positive interest variable may be affecting the interaction between withdrawal and clinging. Contingency tables of the variable increase in withdrawal and clinging were set up for each level ot 'positive interest' separately. The results showed that there was a significant interaction between clinging and withdrawal when the level of positive interest in the baby was low [i.e. on or below the median; X' = 4.72 (1 df) P< 0.03]. The interaction was not significant when the level of positive interest in the baby was high. The 3-factor interaction was therefore included in the model, giving an overall goodness-of-fit statistic G = 7.51 (6 df). This low value indicates a good fit. The fitted values, under this model, were close to the observed. (f) Percentage of deviance accounted for by the model. This was 84% . These results then indicate that increased withdrawal and increased clinging were associated reactions when there was little positive interest in the baby, but there was no such association in the presence of marked positive interest in the baby. Frequent imitation of the baby was less likely in children who were reported to have increased in clinging. (4) Similar analyses were carried out to examine the relationship between the remaining indices of reactions (increase in demanding behaviour, in negative behaviour towards the mother, irritating behaviour to the baby, signs of regression and signs of being more grown up) and those already investigated. For each of these analyses, the G-statistics of goodness-of-fit for Model (I), with no interactions included, were low, ranging from 8.40 to 12.40 with 11 df These low values indicate that there were no interaction effects among the variables. The fitted values under Model (I) were in each analysis close to the observed values. As further confirmation of the lack of interaction between variables, tests for interaction effects were carried out between each pair of variables. None was significant at the 5% level. 8 JUDY DUNN, CAROL KENDRICK AND ROSANNE MACNAMEE Summary of results ^ These findings indicate that there were positive associations between aspects of positive behaviour towards the baby, and between the reactions of increased dinging and tearfulness, and increased withdrawzil and chnging where little positive interest in the baby was reported. Frequent imitation of the baby was less likely in children reported to have increased in clinging. However, there were no significant associations between the other aspects of disturbed or negative behaviour, or between these reactions and the positive interest which the first child showed in the baby. SECTION III. FACTORS ASSOCIATED WITH THE FIRST CHILD'S REACTION The contribution of factors such as temperament, sex and age to the different aspects of the first child's reaction to the birth of the sibling was analysed using a multiple regression technique. Because the dependent variables were dichotomised (see Results below), a linear logistic model was fitted. Appendix 2 gives the procedure. In order to select the independent variables which were to be used in relation to each index of reaction, the association between each 'reaction index' and a number of other variables was examined in a preliminary analysis. The following 8 reaction indices were investigated: Increases in withdrawal; clinging; demanding; negative behaviour to mother; sleep problems; irritating behaviour to baby; signs of regression; signs of positive interest in the baby. (1) Preliminary analysis for selection of independent variables to be used in regression analysis Each of the 8 'reaction indices' was examined in relation to the following variables. (a) Temperamental traits. Comparisons were made for each reaction index between children scoring either above or below the median on each of 6 temperamental traits: negative mood, intensity, withdrawal, unmalleability, intensity -i- negative mood, and unmalleability + intensity (see Dunn and Kendrick, 1980b, for description of assessment of temperamental characteristics, and for analyses of the relationship between this assessment and the observed behaviour of the child). ()^ (c) Pre-sib-birth interview items. Items selected were those concerned with feeding problems, sleeping problems, tantrums, periods of misery, worrying, demanding behaviour, affectionate behaviour towards mother and father, experiences of separation, mother's state of irritability with child. (d) Post-sib-birth interview. Items selected were those concerned with mother's state (tiredness and depression), difficulties with the new baby, location of delivery (home or hospital), first child's behaviour during hospital visit, the expression of jealousy when father or grandparents were involved with baby, the first child's behaviour when the baby was being fed, and the nature of the feeding technique (breast or bottle). (e) Observation measures. The following observation measures from the pre-sib-birth and post-sib-birth observations were included in the preliminary analysis: mother prohibit, time spent in confrontation, time in joint attention, time in joint play, time child held, mother shows, suggests, child fuss, wander, sit, verbal demand, child initiate play, hold. THE REACTION OF FIRST-BORN CHILDREN TO THE BIRTH OF A SIBLING 9 These preliminary analyses showed that for several of the 'reaction indices' there were significant associations between the particular reaction index and a number of the 'independent' variables. The logit regression analyses were next carried out to clarify the extent to which each of these associations was independent, and to examine possible interaction effects. In the logit probability analysis, then, the selection of the independent variables for the analysis of each 'reaction index' was based on the association found in the preliminary analysis: in the results section below the independent variables selected for the analysis of each reaction index are given, with the change in deviance associated with each (ANDEV table), the interaction analyses performed are described and their results given. In each case, the order in which the 'independent' variables were entered was varied, and all possible interactions were investigated. If the order of entry affected the change in deviance accounted for by the independent variables significantly, this is reported. Resutts (1) Increase in withdrawal The preliminary analysis had shown significant associations between a reported increase in withdrawal and the following. (a) Temperament: children scoring above the median on the temperamental trait of negative mood were significantly more likely to be reported to have increased in withdrawal behaviour. (b) Sex: more boys than girls showed an increase in withdrawal. (c) Mother's state in the 3rd week after the sibling birth: children whose mothers reported feeling extremely tired and/or depressed were more likely to be reported as increasing in withdrawal. The dependence of an increase in withdrawal behaviour on these variables was investigated by examining the change in deviance accounted for by each of the above variables (a-c) and by the interaction of: (d) sex and temperament trait of negative mood; (e) mother's state and temperament trait of negative mood; (f) mother's state and sex of first child. The comparison was between those children reported to show no increase in withdrawal (A^ = 29) and those reported to show increased withdrawal (A^ = 11). A number of different models were fitted, in which the order of the variables entered was varied, and the different interactions (d, e, and f above) were examined. In each of these analyses the 3 variables (a-c) accounted for significant change in the deviance. Table 8 shows the analysis of deviance TABLE 8. INCREASE IN WITHDRAWAL: ANALYSIS OF DEVIANCE Deviance Initial model constant Mother's state (post-sib-birth) Temperament trait: negative mood Sex Interaction between sex and mother's state 47.053 38.830 34.534 28.830 16.639 df 3 1 1 3 Change in deviance Significance of change 8.223 4.296 5.704 12.192 0.05 0.01 0.02 0.01 table for the model which included the interaction (f) between the mother's state and the sex of the first child. The table shows that: (1) There is a significant change in the deviance associated with the mother's state. (2) There is a significant change in the deviance associated with the temperamental trait of negative mood, after the variable 'mother's state' has been fitted. 10 JUDY DUNN, CAROL KENDRICK AND ROSANNE MACNAMEE (3) There is a significant change in the deviance associated with the sex of the first child, after the two previous variables have been fitted. (4) There is a significant change in the deviance associated with the interaction between sex and mother's state. The other interactions did not account for a significant change in deviance. (5) The percentage of deviance accounted for was 64%. Table 9 gives the regression coefficient for the variables which make a significant contribution to the deviance. It can be seen from the sign of the relevant coefficient that the probability of an increase in withdrawal after the sibling birth is significantly higher where the mother reported extreme tiredness/ depression in the third week after the baby, and where the first child is a boy, and that there was a further increase in the probability of increased withdrawal if the first child was a boy and the mother was extremely tired/depressed. It was significantly decreased where the first child scored below the median on the temperamental trait of negative mood. TABLE 9. INCREASE IN WITHDRAWAL: REGRESSION COEFFICIENTS Mother's state: extreme tiredness/depression -1-4.43 Temperament: below median on negative mood above median on negative mood fc.Male Interaction between male sex and mother's extreme tiredness -3.89 +2.83 +0.05 +5.29 (2) Increase in steepingprobtems Two measures from the preliminary analysis had shown significant associations with a reported increase in sleeping problems after the sibling birth. (a) Temperament: children scoring above the median on the temperamental trait of negative mood and intensity were more likely to develop sleeping problems. (b) Observational measures: the increase in the proportion of verbal interactions initiated by the mother prohibiting the child (increase in M. prohib.) was negatively associated with the reported increase in sleeping problems. The dependence of an increase in sleeping problems on these variables was investigated by examining the change in deviance accounted for by these two variables and by the interaction between them. The comparison was between those children reported to show no increase in sleeping problems (A'^ = 29) and those reported to show an increase of 1 point, more than 1 point, or extreme breakdown of sleeping (A^ = 11). The analysis of deviance table (Table 10) shows that: (1) There was a significant change in the deviance associated with the 'increase in M. prohibit' variable. (2) There was a significant change in the deviance associated with the temperament trait of negative mood and extreme intensity, after the 'increase in M. prohibit' variable was fitted. (3) There was a significant interaction effect between the two previous variables. (4) The percentage of deviance accounted for was 30%. TABLE 10. INCREASE IN SLEEPING PROBLEMS: ANALYSIS OF DEVIANCE Initial model constant Increase in M. prohibit Temperament trait: negative mood + intensity Interaction between: increase in M. prohibit and temperament (negative mood + intensity) Deviance df Change in deviance Significance of change 46.407 40.975 36.659 1 1 5.426 4.315 0.05 0.05 4.143 0.05 32.517 11 THE REACTION OF FIRST-BORN CHILDREN TO THE BIRTH OF A SIBLING The signs of the regression coefficients given in Table 11 show that the probability of an increase in sleeping problems was increased in the case of children who scored above the median on the temperamental trait of high intensity and mood; it decreased as the measure of 'increase in mother prohibit' increased. There was a significant interaction between these two variables: the plotted probability showed that for those children who scored high on the temperament trait of high intensity and negative mood, who also had mothers who showed no increase or a decrease in the 'mother prohibit' measure, the probability of an increase in sleeping problems was significantly increased. TABLE 11. INCREASE IN SLEEPING PROBLEMS: REGRESSION GOEFFICIENTS Increase in M. prohibit -0.10 Temperame..t • above median on negative mood + intensity below median on negative mood + intensity 0.36 -1.32 (3) Increase in clinging The preliminary analysis showed that there were significant associations between a reported increase in clinging and the following. (a) The age of the first child: younger children were more likely to be reported as showing a marked increase in clinging behaviour. (b) Temperament: children scoring above the median on the temperamental traits of negative mood and of intensity were more likely to increase in clinging. The dependence of an increase in clinging on these variables was investigated by examining the change in deviance associated with these three variables, and by the interaction between them. The comparison was between those children in which increased clinging behaviour was reported to be absent or rare (A^ = 17) and those showing occasional, frequent or constant increase (A^ = 23). There were no significant interaction effects. Table 12 shows the analysis of deviance for the three variables without the interactions fitted. The percentage of deviance accounted for was 25%. TABLE 12. INCREASE IN GLINGING: ANALYSIS OF DEVIANCE Initial model constant Age Temperament trait: negative mood Temperament trait: intensity Deviance df Change in deviance Significance of change 54.548 49.781 48.611 44.477 1 1 1 4.767 1.170 4.135 0.05 n.s. 0.05 The regression coefficients for these variables were (a) Age: -0.05; (b) Temperament traits (1) above median on negative mood: +1972; (2) above median on intensity: +3.82. The signs of these coefficients indicate that the probability of an increase in clinging is increased with the younger children and where the child scores above the median on the temperament traits of intensity. (4) Increase in negative behaviour towards the mother The preliminary analysis had shown that the following variables were associated with a marked increase in negative behaviour towards the mother. (a) Pre-sib-birth interview: mother frequently irritated by child. (b) Post-sib-birth interview: mother's state in 3rd week after sibling birth. (Children whose mothers reported feeling extremely tired and/or depressed were more likely to be reported as having shown an increase in negative behaviour to the mother.) 12 JUDY DUNN, CAROL KENDRICK AND ROSANNE MACNAMEE (c) Observation measures: increase in confrontation between mother and first child from pre-sib-birth observation to post-sib-birth observation. The dependence of the increase in negative behaviour to the mother on each of these variables was investigated by examining the change in deviance associated with them, and by the interactions between them. The comparison was between those children reported to show no or infrequent increase in negative behaviour to the mother (A'^ = 16), and those reported to show frequent or marked naughtiness (A'^ = 24). The only variable to account for a change in the deviance significant at the 5% level was the observational measure of a change in confrontation. Table 13 shows the analysis of deviance for the three variables fitted, without interaction effects. (None of the interaction effects accounted for significant changes in the deviance.) The percentage deviance accounted for by the change in confrontation was 19%. The regression coefficient was: change in confrontation: +0.36. That is, a greater increase in confrontation after the sibling birth was associated with an increased probability of increased negative behaviour towards the mother being reported after the sibling birth. TABLE 13. INCREASE IN NEGATIVE BEHAVIOUR TOWARDS MOTHER: ANALYSIS OF DEVIANCE Deviance Initial model constant Increase in confrontation Pre-birth interview: mother's irritability Mother's state of tiredness (post-sib-birth) 48.114 39.164 30.735 29.141 df 1 4 3 Change of deviance Significance of change 8.950 8.429 1.594 0.01 0.10 n.s. (5) Irritating behaviour towards the baby The preliminary analysis had shown that there were significant associations between reported irritating behaviour shown by the first child to the baby, and the following. (a) The type of feeding the new baby was receiving (breast or bottle). (b) The observed increase in frequency of maternal prohibition. The comparison was between those children reported to show no or rare irritating behaviour to the baby (A' = 18) and those reported to show such behaviour frequently or constantly (A^ = 22). Table 14 shows the analysis of deviance for these two variables. No significant interaction effects were found. The regression coefficients were: bottle-feeding: + 2.078; increase in maternal prohibition: -1-0.09. That is, the probability of irritating behaviour to the baby being reported was increased where the baby was being bottle-fed, and where a marked increase in the frequency of maternal prohibitions had been observed. TABLE 14. IRRITATING BEHAVIOUR SHOWN TOWARDS BABY: ANALYSIS OF DEVIANCE Initial model constant Breast/bottle feeding Increase in M. prohibit. Deviance df Change in deviance Significance of change 53.834 46.379 42.438 1 1 7.455 3.941 0.01 0.05 (6) Positive interest in the baby To examine the factors contributing to the index of positive interest in the baby, a standard regression model was used, as the index could be treated as continuous (Draper et al., 1966). The preliminary analysis had indicated that age, temperamental trait of withdrawal, sex, location of delivery of the 2nd child, and the reported presence of sleep problems in the pre-sib-birth interview were associated (at the 10% level of significance) with positive interest in the baby. These variables were fitted in the regression analysis, in the above order. Table 15 gives the ANOVA for the overall regressic, which THE REACTION OF FIRST-BORN CHILDREN TO THE BIRTH OF A SIBLING 13 TABLE 15. POSITIVE INTEREST IN THE BABY: ANOVA TABLE FOR AGE, SEX, SLEEPING PROBLEMS AND TEMPERAMENTAL TRAIT OF WITHDRAWAL Regression Residual Total df SS MS Fratio 6 6 31 129.7 129. 7 214. 9 344. 6 21.614 21. 614 3.12 37 6. 931 was significant (2 !4 % point of /^^ .;, is 2.86). The contribution of each variable was assessed by comparing the sum of squares for a model with all variables included, and for a model with all variables except that under consideration. The results showed that two variables contributed significantly to the variance (1) Temperamental trait of withdrawal [Fratio 5.29 {F,, ,,)]. (2) The presence of sleep problems before birth of sibling [F ratio 3.34 (/^3,:ii)]. The regression coefficients in Table 16 show that children who scored on or below the median on the trait of withdrawal showed more positive interest in the baby: those who were reported either to have no sleep problems, or to have extreme sleep problems showed more positive interest in the baby than those who were reported to have occasional sleep problems. There was n^ significant interaction between these two variables [Fratio 0.73 {F.^ j,,)]. TABLE 16. POSITIVE INTEREST IN THE SIBLING: REGRESSION GOEFFICIENTS Temperament: on or below median for withdrawal Pre-sib-birth sleep problems: no sleep problem Occasional (less than weekly) Marked (more than weekly) Very marked (daily) + 2.136 0.000 -2.31266 -0.15159 +2.4204 Signs of regression: increase in demanding behaviour No clear pattern of association was found for the reaction indices of increase in demanding behaviour, or 'signs of regression'. Summary of findings on factors associated with the reaction of thefirst-born The following variables were found to be associated with the different aspects of reaction of first-born children to the birth of the sibling. (1) Temperament. Children who scored above the median on negative mood were more likely to be reported to have increased in withdrawal and in sleeping problems, while those who scored above the median in intensity and negative mood were more likely to show increase in clinging. Those who scored above the median on withdrawal were less likely to be reported to show positive interest in the baby. (2) Sex. Boys were more likely to be reported to have increased in withdrawal. (3) Mother's state. Mothers who felt extremely tired or depressed were more likely to have children reported to have increased in withdrawal. (4) Age. Younger children were more likely to be reported to show increased clinging. (5) Changes in the interaction between mother and first-born. With greater increases in confrontation, children were more likely to be reported to have increased in negative behaviour towards the mother. With greater increases in maternal prohibition, 14 JUDY DUNN, CAROL KENDRICK AND ROSANNE MACNAMEE children were more likely to be reported to show behaviour irritating to the new baby, and less likely to have increased sleeping problems. (6) Feeding of new baby. If the new baby was bottle fed, the first child was more likely to be described as irritating the new baby. (7) Interactive effects. These were seen between the factors of temperament, sex and mother's state, and affected the likelihood of increase in withdrawal, and sleeping problems. DISCUSSION It should be emphasised that the findings reported in this paper are based on a relatively small non-random sample of families. Any generalisation from our results must be made with great caution. For this group of families the results demonstrate how difficult the early weeks after the birth were for both mother and elder child. The majority of children in this sample were reported to show signs of disturbance or negative behaviour towards the mother. Similar findings are reported by Trause et al. (1979) in a study of hospital visiting: 92% of the 31 children in their sample showed an increase of at least one problem. The lack of close association between the different aspects of disturbance and negative behaviour, other than the links between clinging, tearfulness and withdrawal, suggests that a single index of 'disturbance' which combined all these different reactions would be inappropriate. The follow-up data from the later visits to the families supports this conclusion: the prognostic implications of the different aspects of reaction are quite different. We find for instance that there is an association between the reaction of increased withdrawal after the sibling birth and a poor relationship developing between the siblings, but no such association between increased negative behaviour or demanding behaviour and the later sibling relationship. The categorisation of an elder child's behaviour as deliberately interfering or irritating the baby is obviously a problematic one. It could be that similar behaviour by a 2 J/2-year-old might be described by one mother as 'showing affectionate interest', and by another as 'unnecessarily interfering with the baby'. However there were some children who were described by their mothers both as showing frequent physical affection, and as often irritating the baby. Our own observations suggest that some children do individually show a range of behaviour towards the baby sibling that includes both affectionate gentle caressing, and behaviour that is much more clearly categorised as irritating—pulling the dummy out of the sleeping baby's mouth and throwing it away, and so on. Such ambivalence is not surprising; it is also clearly reflected in the comments which the children make about the new babv. Reports of deliberately aggressive behaviour towards the baby were much less common, and the few children described as showing this behaviour were also reported to show little positive interest in the baby. This was the only clear negative relationship between aspects of positive interest in the baby and the more negative features of the fir.st child's reactions. The finding that 'signs of being more grown up' were reported for more than half the sample supports the observation of Anna Freud (1965) that children sometimes THE REACTION OF FIRST-BORN CHILDREN TO THE BIRTH OF A SIBLING 15 show a rapid increase in developmental maturity during or after periods of stress. Trause et al. (1979) also noted 'improvements' in behaviour after the birth of a sibling. The analysis reported in Section III showed that for this group of children temperamental differences, the quality of the relationship with the mother and age were all linked to the ways in which the first-born reacted, but that there were different patterns of association for the different aspects of reaction. The absence of clear associations with the 'signs of regression' score suggests, again, that it is not useful to combine different aspects of 'regressive' behaviour. Follow-up data shows that for instance different kinds of 'imitative' behiiviour can have very different prognostic implications (Dunn and Kendrick, 1980c). One possible interpretation of some of the patterns of association reported here is that they are the result of a consistently biased perception of the first child by the mother. Thus the link between a 'difficult' temperament (that is, negative in mood and intense in expression), and an increase in reported sleeping problems after the sibling birth could be interpreted as a biased description from a mother who continued to view her child in a critical fashion. Against this interpretation, however, there are a number of independent lines of evidence. First, evidence that the temperamental differences between the children are not simply differences 'in the eye of the mother' is found in the agreement between the description of a child's temperament by the mother, and by the observer, for traits other than activity. There are also some suggestions from the observations that direct measures of behaviour support the mothers' temperamental assessment (Dunn and Kendrick, 1980b). Second, there is evidence that children of difficult temperament have mothers with a wide range of attitudes towards, and practices in, prohibiting and punishing. In the pre-birth observations a clustering of prohibiting and punitive measures was described (Dunn and Kendrick, 1980a). This group of measures was negatively correlated with measures of affectionate playful mothering. However the group of prohibiting and punitive measures from the obsexvations was not significantly associated with the description of children as difficult in temperament. And mothers' replies to the pre-birth interview questions about the frequency of smacking and punishment showed that children rated as more extreme in negative mood, unmalleability or intensity of response were not more likely to have mothers who smacked frequently. Third, there is evidence that children who increase in demanding behaviour, other aspects of 'negative' behaviour, and children who are described as extreme in temperament, may also be described as interested in the new baby, and are observed to be active in attending to and holding the new baby. The description of a child's behaviour towards the new sibling as positive;ly interested was supported by observation measures (Dunn and Kendrick, 1980b). It would be difficult to reconcile this evidence with the interpretation that the temperament assessment and the description of the child's behaviour in the postbirth period are linked only through a consistently negative maternal view of the child. The finding that for this sample the mothers who were bottle-feeding their second child reported more irritating behaviour towards the baby by the first born also 16 JUDY DUNN, CAROL KENDRICK AND ROSANNE MACNAMEE presents problems of interpretation. First, it could be that mothers who breast-feed make more effective attempts to provide distraction for the first child before they start feeding the baby. There is some evidence that this is the case. Second, it could be that mothers who breast-feed do view their children in a more positive way, and are prepared to interpret their behaviour less critically. Third, it could be that first-born children are indeed less liable to interfere and demand when their mothers are breast-feeding. These alternative possibilities are discussed in Kendrick and Dunn (1980) where the observational data on the mother-child interaction during feeds is reported. In considering the comparatively small contribution of age differences and the home/hospital delivery differences, the nature of the particular sample of families studied here must be borne in mind. The age range of the children was rather narrow, and the majority of families had very close contact with the grandparents, who were frequent visitors. It is possible that in some families the close relationship between grandparents and first-born may have lessened the trauma of separation from the mother which the hospital delivery involved. In a group of families without these close ties and with longer separations the effects of separation on the first-born might well be more marked as in the Robertsons' (1971) study. It should be noted however that in their study of the effects of hospital visiting on the first-born child's response to the arrival of a sibling, in which children visiting several times were compared with those who did not visit at all, Trause et al. (1979) found that differences in the degree of separation experienced did not affect the incidence of disturbed behaviour, although those first children who visited frequently were less likely to react to the actual arrival of the mother home by indifferent or hostile behaviour. What must be emphasised is that even when the mother remained at home to deliver her second child the first-born was likely to show signs of disturbance and upset. The relationship between a mother and a 2-3-year-old is often one that involves much stress and difficulty, and the constellation of changes involved when a sibling is born, even without a separation of mother and first-born, was apparently enough to increase the difficulties considerably for both mother and child. SUMMARY The reactions of first-born children to the birth of a sibling, as reported by their mothers, were studied in 40 families. Increased clinging, withdrawal and tearfulness were negatively related to positive interest in the baby, but other aspects of disturbed behaviour were not significantly associated with each other, or with positive interest. Differences in temperament, in the quality of the mother-first-born relationship, in age and in the mothers' state in the post partum weeks were found to be associated with differences in the reaction of the first-born children. Acknowtedgement—This work is supported by the Medical Research Council. We are very grateful to the families in the sample for their generous help, to the Health Visitors of the Cambridge Area, and to Robert Hinde for his helpful comments on the manuscript. We would like to thank Claire Sturge for permission to use her Temperamental Characteristics Assessment, and for collaboration with us over interview design. THE REACTION OF FIRST-BORN CHILDREN TO THE BIRTH OF A SIBLING 17 REFERENCES BISHOP, Y . M . M . , FIENBERG, S. E . and HOLLAND, F . W . (1975) Discrete Multivariate Anatysis. Massachusetts Institute of Technology Press, Boston. BLACK, D. and STURGE, C . (1979) The young child and his siblings. In Perspective in Infant Psychiatry (Edited by HOWELLS, J. G.). Oliver & Boyd, Edinburgh. DRAPER, N . R . and SMITH, H . {1966) Applied Regression Analysis. Wiley, New York. DUNN, J . and KENDRIGK, C . (1979) Interaction between young siblings in the context of family relationships. In The Child in its Family (Edited by LEWIS, M. and ROSENBLUM, L. A.). Plenum Press, New York. DUNN, J. and KENDRICK, C . (1980a) The arrival of a sibling: changes in patterns of interaction between mother and first-born child./ Chitd Psychol. Psychiat. 21, 119-132. DUNN, J. and KENDRIGK, C . (1980b) Studying temperament and parent-child interaction: a comparison of information from direct observation and from parental interview. DevlMed. Child Neurot. (in press). DUNN, J. and KENDRICK, C . (1980C) Usurping Companions: a Study of Young Siblings. Harvard University Press, Cambridge (fonhcoming). FREUD, A. (1965) Normality and Pathology in Chitdhood. International Universties Press, New York KENDRICK, C . and DUNN, J . (1980) Caring for a second child: effects on the interaction between mother and first-born. Devt Psychot. 16,303-311. LEVY, D . M . (1934) Rivalry between children in the same family. Chitd Study 11, 233-261. LEVY, D . M . (1937) Studies in sibling rivalry. Research Monograph No. 2; American Orthopsychiatric Association . PETTY, T. A. (1953) The tragedy of Humpty Dumpty. Psychoanal. Study Child 8, W'i-^12. REGISTRAR CENERAL (1973) Census 1971 Great britain Summary Tables 1% sample, H.M.S.O. ROBERTSON, J . and ROBERTSON, J . (1971) Young children in brief separation: a fresh look Psychoanal. Study Child 2 6 , 264-315. RUTTER, M . (1977) Individual differences. In Child Psychiatry: Modem Approaches (Edited by RUTTER, M. and HERSOV, L . A.). Blackwell, London. THOMAS, A. and CHESS, S. (1977) Temperament and Devetopment. Brunner-Mazel, New York. TRAUSE, M . A., BOSLETT, M . , VOOS, D . , RUDD, C , KLAUS, M . and KENNELL, J . (1979) A birth in the hospital: the effect on the sibling. (Submitted for publication.) WINNICOTT, D . W . (1977) ThePiggle. Hogarth Press and Institute of Psychoanalysis, London. APPENDIX 1. LOC-LINEAR CONTINGENCY ANALYSIS If, for example, the data are classified in a 3-dimensional taWe according to 3 categorical variables Xi, X2, X}, with n^ji, cases in the {i, j , A:)th cell, then various log-linear models can be set up to describe the data. For instance, if it is hypothesized that there is an interaction between ;t, and X2, then an appropriate model would be logn,yt == A ^ A, + A2 + A} + A12 where A = constant, Aj, A2 and A3 are the effects due to variables Xi, X2 and x^ respectively, and A12 represents the interaction between Xt and X2 etc. The numbei of variables which can be examined together, using this technique is restricted by the number of cases, since the test statistics are approximately x^ distributed, provided that there are not too many zeros in the table. The procedure for testing the hypotheses is as follows: a programme fits (1) a model with no interactions present [Model (I)]; (2) a model with all 2-factor interactions present [Model (II)]; (3) if necessary, a model with all 2- and 3-factor interactions present, etc. For each model, a goodness-of-fit statistic, C , is computed, with appropriate degrees of freedom given. This statistic has an approximate chi-square distribution. High values indicate that the model does not fit well, and is inappropriate. Hence if the G^-statistic for Model (I) is large, this indicates that there are interactions present between the variables. If the G2-statistic for Model (II) is low. 18 JUDY DUNN, CAROL KENDRICK AND ROSANNE MAC;NAMEE indicating that the model fits well, the next step in the procedure is to examine which interactions are the significant ones. For instance to test whether an interaction exists between variables x, and X2, Model (I) is first fitted, then a third model (III), which is the same as Model (I) but with an added interaction term x-, .X2 included. The difference between the G^-statistic for Models (I) and (III) is then a test statistic for the interaction term x^. X2. High values of this difference indicate a significant interaction. APPENDIX 2. LINEAR LOGISTIC MODEL REGRESSION TECHNIQUE Given a dependent variable which takes the values 0 and 1, where p is the probability of taking the value 1, and given a number of independent variables Xi,X2 x„, then the model , • r , pf: 1 -pk J logit/.4=loger—f is fitted, using an iterative procedure to obtain maximum likelihood estimates of the constants Ai, B,, fl2- -fi,, (see GENSTAT manual. Generalised Linear Models, Chapter 7). For each analysis, an analysis of deviance (ANDEV) table is also computed. The 'deviance' of a fitted model measures the discrepancy between the model and the data, and is approximately x^ distributed, with appropriate df. The independent variables are added to the regression equation one at a time: hence the change in deviance due to adding a particular variable can be used to assess whether or not that variable makes a significant contribution to the regression, after the preceding variables have been taken into account. Similarly we can examine the contribution of interaction effects between the independent variables. Finally, the percentage reduction in deviance achieved by all the independent variables (and their interactions) can be computed.