Attachment Security, Maternal Depressive Symptoms, Perceived Infant Temperament,

and Age 3 Behavior Problems: A Longitudinal Study of Low Income Families

 

Conduct disorders in childhood are important because of their direct cost to society in terms of damaged property and disruption of normal patterns of living, but also because of our ineffectiveness in treating such children, and the likelihood of later adult criminality and other serious disorders such as substance abuse (Kazdin, 1987; Loeber, 1982; Brook, Whiteman, & Finch, 1993). Despite these concerns and the high stability of aggressive behavior beginning as early as age 2 (Cummings, Iannotti, & Zahn-Waxler, 1989; Olweus, 1979), it is only during the past decade that a number of investigators have begun to identify developmental precursors of school-age externalizing problems beginning before age 5 (Campbell, Breaux, Ewing, & Szumowski, 1986; Fagot, 1984; Rose, Rose, & Feldman, 1989).

Two reviews of predictors of children's antisocial behavior have found that childrearing practices are at the forefront of results across a broad range of studies (Loeber & Dishion, 1983; Loeber & Stouthamer-Loeber, 1986). Specifically, parental rejection and lack of involvement are two of the parenting factors that have been identified as salient. Shaw and Bell (1993) have postulated that during infancy, these parenting behaviors are expressed as a lack of parental responsiveness, as parents who were hostile and/or uninvolved would consistently fail to respond contingently to the infant's needs.

Glueck and Glueck (1950) were among the earliest investigators to detect hostility or lack of warmth in fathers, and hostility or indifference in mothers of delinquent children, but Hirschi (1969) integrated these and other related findings into a theory of parental factors in the development of externalizing problems. Hirschi proposed that lack of self-control is the key element. This lack of control is due to the failure to form an attachment to parents in early development, which leads, in turn, to failure to form an attachment to other conventional individuals, failure to make a commitment to conventional lines of action and conventional activities, and lack of belief in the legitimacy of the moral order.

The proposed outcomes of attachment theory, that are described in a more general way by Hirschi, have been operationalized and studied using the Ainsworth Strange Situation (Ainsworth & Wittig, 1969) during infancy to predict both social development and behavior problems. Attachment theorists have hypothesized that differences in caregiver sensitivity, and the resultant bond between parent and infant, are important factors in later behavior patterns of the child (Bowlby, 1980; Sroufe, 1983). With regard to the development of externalizing behavior problems, insecurely attached children would be predicted to function less harmoniously in compliance situations and more likely to interpret neutral or even friendly behavior as hostile. Later in the preschool years the same insecurely attached children may act in a disruptive or aggressive manner in order to engage an unresponsive caregiver (Greenberg & Speltz, 1988), leading to higher rates of Patterson's coercive cycles of parent-child interaction (Shaw & Bell, 1993).

Empirically, attachment security consistently has been found to predict later child competence, including longer attention span, greater compliance, persistence, enthusiasm, and cooperation at age 2 (Londerville, & Main, 1981; Matas, Arend, & Sroufe, 1978), better social relations with peers in preschool (LaFreniere & Sroufe, 1985; Jacobson & Willie, 1986), and greater ego resiliency at age 5 (Arend, Gove, & Sroufe, 1979). However, the relationship between attachment security and later behavioral difficulties has been more equivocal, particularly in studies using middle class families (Bates, Maslin, & Frankel, 1985; Fagot & Kavanagh, 1990; Lewis, Feiring, McGuffog, & Jaskir, 1984). In the two longitudinal studies conducted with low SES samples, insecure patterns of attachment were associated with behavior problems at ages 5 and 7-8 (Erickson, Sroufe, & Egeland, 1985; Lyons-Ruth, Alpern, & Repacholi, 1993; Renken, Egeland, Marvinney, Mangelsdorf, & Sroufe, 1989), though even in these studies differences were found with respect to the type of insecure pattern predictive of later externalizing problems (i.e., the avoidant versus disorganized pattern).

Why would insecurely attached infants from low SES backgrounds be more prone to show later problem behavior? The buffers of the middle class child's ecosystem may prevent the behavior of the insecurely attached child from becoming dysfunctional (Bates et al., 1985; Shaw & Bell, 1993). However, in low SES families, experiencing economic hardships, poor living conditions, unstable family life, and/or inadequate alternative child care resources, parenting itself may be disturbed, resulting in a child with an insecure attachment. Stressors within the family's environment may not only set in motion an insecure attachment relationship, but may also maintain the child's distrustful working model of interpersonal relations in the preschool and school-age years. For example, a lack of alternative care resources, which could provide occasional relief from parenting for the single-parent mother, coupled with unstable relationships with significant others, which if stable, could provide support for the mother's emotional adjustment as well as child care assistance, may interfere with her ability to be sensitive and responsive infant needs. The infant may respond with increasing aversive behavior in order to elicit more attention from the mother.

One of the present study's primary goals is to further our understanding about the relationship between attachment security and behavior problems at age 3 using a sample of low income mother-child dyads. Attachment security will be use to predict both externalizing and internalizing behaviors, given the latter variable's concurrent association with externalizing problems (r = .71, p <.0001 in the present sample), and the need to examine the specificity of the association between infant and/or parent behavior with externalizing versus internalizing problem behaviors.

Another aim of the study is to examine how two other risk factors may independently influence the development of early child behavior problems: (1) maternal depressive symptomatology, and (2) infant temperament. Maternal depression has been implicated as an important factor in the emergence of externalizing and internalizing problems at preschool and school-age (Downey & Coyne, 1990; Forehand, McCombs, & Brody, 1987; Keller, Beardslee, Dorer, Lavor, Samuelson, & Klerman; Shaw & Emery, 1988). Zahn-Waxler, Iannotti, Cummings, and Denham (1990) have noted that the unresponsiveness that characterizes some forms of parental depression may lead children to engage in behaviors that elicit negative parental attention and higher rates of coercive parent-child interaction. The pathway from maternal unresponsiveness to later externalizing behavior problems has been documented for children from age 1 to 3 by Martin (1981) and Shaw (1993).

Empirical support for the importance of maternal perception of infant difficulty, and its relation to externalizing problem behavior at ages 4-5, has been demonstrated in two investigations (Bates, et al., 1985; Sanson, Oberklaid, Pedlow, & Prior, 1991). In the Bates et al. (1985) study, maternal perceptions of the infant as difficult from 6 to 24 months of age was predictive of mother- or secondary-caregiver-perceived hostility in the child at 3 years of age. Child hostility as perceived by the mother, was the strongest predictor of externalizing problems at 4-5 years. In another study with relevant data, Sanson et al. (1991) found that maternal perception of the infant as difficult, in combination with one or more other risk factors (e.g., sex of the child, perinatal stress, prematurity), was also a strong predictor of externalizing problem behavior at ages 4-5.

This study also will examine the interactive relationships among maternal depressive symptoms, perceived difficult temperament, attachment insecurity, and age 3 behavior problems. Though some have posited that specific combinations of effects (e.g., insecure attachment with perceived difficult temperament) uniquely predict maladaptive child outcomes (Crockenberg, 1981), Rutter and colleagues (1975a, 1975b, 1978) have reported an increase in the probability of school-age children exhibiting behavior disorders as a function of the number of family stressors, regardless of the specific combination of stressors (e.g., marital discord, maternal psychiatric disorder). Rutter's cumulative stressor hypothesis has been replicated across time, age of child, and type of outcome (Blanz, Schmidt, & Esser, 1991; Sanson et al., 1991; Shaw & Emery, 1988), and has been found to differentiate attachment security (Shaw & Vondra, 1993) and preschool behavior problems (Sanson et al., 1991). Both specific and non-specific methods of examining the cumulative impact of attachment insecurity, maternal depressive symptoms, and perceived difficult temperament on age 3 behavior problems will be examined.

The study also provides an opportunity to examine the stability and correlates of maternal depressive symptoms as children grow from age 1 to 3. As Fagot and Kavanaugh (1993) have noted, the second year of parenting is marked by greater complexity and less pleasure than the first. We examine how early relationships and child effects (i.e., insecure attachment, perceived infant difficulty) influence later maternal depressive symptoms when children are age 2 and 3.

Finally, despite few sex differences in the base rates of either externalizing or internalizing difficulties at ages 2-3, previous research has shown different early correlates of child behavior problems for boys and girls and stronger relations with a wider variety of factors for boys (Martin, 1981; Renken et al., 1989; Shaw, 1993; Zahn-Waxler, 1993). For these reasons, all analyses will examine the effects of gender when sample sizes permit.

Method

Subjects

Subjects originally included 100 mother-child dyads (59 males and 41 females) recruited from the Women, Infants, and Children (WIC) Special Supplement Food Program of Allegheny County as part of a larger longitudinal study of child development. WIC provides financial support to purchase nutritionally-sound food items for low income families. At the time of their infant's birth, mothers ranged in age from 17 to 36 years, with a mean age of 25. Sampling of marital status was not restricted due to the considerable relationship instability within the sample. Forty-six percent were either married or living together, whereas 54% were either divorced (8%), separated (9%), or single (37%). The majority of families were Caucasian (61%) and the remainder were African-American (39%). The mean family income in the sample was between $500-1,000 per month, with 72.5% of the families having yearly earnings equal to or less than $12,000. Mothers' average level of education was 12 years, with 73.5% having a high-school degree or less.

Mothers of infants between 6 and 11 months of age were recruited in two WIC waiting rooms by the principal investigators. Mothers were informed that the study was a project on child development and parental care, and that they would be paid $15.00 for each study session, plus an additional $10.00 for transportation costs. Upon agreeing to participate, informed consent was sought, mothers completed three questionnaires, and the first lab visit was scheduled within two weeks of the infant's first birthday. Of the 144 women who were asked to take part in the study, 129 (89.6% of the 144) agreed to participate, but only 100 completed the 12-month assessment. Of those 100 subjects seen when infants were 12 months old, 89 participated in the 18- and 24-month laboratory assessments months); however, due to errors in the videotaping of assessments, sample sizes for specific videotapes measures are slightly less than the total sample size at different assessment points. At age three, 82 mothers returned completed questionnaire reports on their child's behavior problems. No significant or appreciable differences were found when demographic characteristics of families who completed all assessments (82 of the 100) and those who did not (18 of the 100) were compared. A similar comparison was made between families who completed two of the three laboratory assessments (e.g., 12- and 18-month, but not 24-month assessments) versus those who completed all three, with no appreciable differences found between groups on demographic variables.

Procedures

When mothers were recruited in the study, they completed a demographic questionnaire, the Beck Depression Inventory (BDI), and the Infant Characteristics Questionnaire. Three videotaped laboratory assessments were then conducted at ages 12, 18, and 24 months, a home visit was conducted at 15 months, and mothers were sent the Achenbach Child Behavior Checklist (CBCL) and the BDI when the child was 36 months. Different laboratories were used at adjacent assessments (i.e., at 12 and 18 months, and at 18 and 24 months), both of which were equipped with a one-way mirror through which the assessment was videotaped.

Each laboratory assessment lasted approximately two hours, and varied in stress level in order to observe parent and infant behavior across a broad spectrum of contexts (e.g., a free play session, teaching tasks, and the Strange Situation). At each age, observational measures of maternal responsiveness and infant difficulty were collected (Martin, 1981; Smith & Pederson, 1988). Mothers also completed a series of questionnaires on demographic information, perceived infant temperament, their own psychological well-being, and quality of environmental support. The instruments used for the present study are described in more detail below.

Measures

Beck Depression Inventory (BDI). Mothers completed the BDI, a well-established and widely used measure of depressive states, at recruitment, the 12, 18, 24, and 36-month assessments (Beck,, Ward, Mendelon, Mock, & Erbaugh, 1961; Beck & Beamesderfer, 1974; Reynolds & Gould, 1981). Split-half reliability of the scale has been found to be high (.86 to .93). Research has identified maternal depressive symptomatology as an influential factor in predicting attachment quality as well as later child adjustment (Lyons-Ruth, Connell, Zoll, & Stahl, 1987; Radke-Yarrow, Cummings, Kuczynski, & Chapman, 1985; Shaw & Emery, 1988). In order to make the instrument more trait-like, the directions of the BDI were altered so mothers reported about depressive symptoms experienced in the past six months rather than the past week. Because the BDI was administered five times during the course of the study, it was decided to composite adjacent assessments that were theoretically meaningful. Empirically, this was justifiable given that correlations between composited BDI scores were equal to or greater than .53 for composited scores, p<.0001. Scores for the recruitment and 12-month administrations were summed and averaged, reflecting maternal depressive symptoms in the first year of the infant's life (termed "Year 1" below), as were scores for the 18- and 24-month administrations (termed "Year 2" below), reflecting BDI scores during the infant's second year. Thirty-six month BDI scores were examined separately.

Infant Characteristics Questionnaire (ICQ). At the screening, 12-, 18-, and 24-month assessments, parents rated their infant on the ICQ, a series of seven-point scales from which factor scores have been constructed indexing persistent, unstoppable, difficult behavior, and negative adaptation (Bates, Freeland, & Lounsbury, 1979). For the present investigation, the Difficulty factor was used given its relation to preschool behavior problems and comparability of items from 12 to 24 months (i.e., unstopability is not included as a factor prior to 13 months). Since the item structure of the Difficulty factor changes at 6, 13, and 24 months, the seven-item 6-month factor was used at the screening, the nine-item 13-month factor was used at the 12-month assessment, and the 7-item 24-month factor was used at the 18- and 24-month assessments. The instrument meets customary psychometric standards for maternal reports on infant temperament, has the advantage of brevity, and has shown longitudinal relations with preschool behavior problems (Bates et al., 1985). Scores for the ICQ Difficulty factor also were composited in the same manner as the BDI, based on theoretical and empirical grounds (r was equal to or greater than .51 for composited scores, p<.0001). "Year 1" scores refer to the summed and averaged recruitment and 12-month assessment ratings and "Year 2" scores refer to the summed and averaged 18- and 24-month ratings.

Strange Situation Procedure. Attachment security was assessed using Ainsworth and Wittig's (1969) Strange Situation, carried out when infants were 12 and 18 months of age. The Strange Situation is a standardized series of separations of reunions of mother and infant involving exposure of the infant to increasingly stressful three-minute episodes with mother and a stranger, in order to evaluate the infant's effective use of the mother as a source of comfort and as a secure base for exploration (Ainsworth, Blehar, Waters, & Wall, 1978). Classification into one of four patterns of attachment (A, B, C, and D) is based largely on infant response to mother prior to and during the two reunion episodes.

Securely attached infants (Type "B") are those who respond positively and seek proximity to the mother upon reunion, and are comforted by her presence or contact if distressed. Insecure-avoidant infants (Type "A") show a tendency to avoid interaction with mother upon reunion, whereas insecure-resistant infants (Type "C") both seek and resist contact with mother after separation and have difficulty being comforted by her. Insecure-disorganized/disoriented infants (Type "D"), a more recent classification based on work with high-stress and/or maltreatment samples (Main & Solomon, 1987), demonstrate confused or "dazed" behavior upon reunion, behavioral stilling, and other movements suggestive of depression, confusion, or apprehension in the presence of mother, and/or exhibit contradictory or disordered behavioral sequences. All Strange Situations were videotaped and subsequently coded according to the procedures described by Ainsworth et al. (1978). For coding of the Insecure-disorganized/disoriented classification, procedures described by Main and Solomon (1987) were used. Three coders were trained to a minimum of 80% reliability in major classifications, based on attachment assessments from the labs of J. Belsky and A. Sroufe. Mean interrater agreement among the three coders for a random set of 10 attachment assessments from the present sample (10%) was 80% across all four major classifications.

Five methods of dividing secure versus insecure groups were utilized so that the data of the present cohort could be compared with previous studies, including those that assessed attachment security at both 12 and 18 months. First, secures were compared to the avoidant, then to the resistant, and then to disorganized infants. Second, disorganized infants were force classified into one of the other three groups to permit comparisons with the Minnesota high-risk sample that did not use the D classification (Renken et al., 1989). Third, given the relatively small sample size of sub-types and the low stability among specific insecure sub-groups (i.e., the stability was 33% for Type A's, 14% for Type C's, and 45% for Type D's), secures were compared to all insecure sub-groups combined. Fourth, based on the poorer adaptation of disorganized infants in high risk settings (Lyons-Ruth, et al., 1990; see Lyons-Ruth, Repacholi, McLeod, & Silva, 1991 for review), disorganized attachments (given a ranking of 3) were compared to resistant and avoidant (ranking of 2), and secures (ranking of 1). Fifth, based on the instability of attachment security from 12 to 18 months, separate analyses were conducted for those infants who remained securely or insecurely attached at both time points. Since the instability among insecure groups did not permit examination of stability among individual insecure sub-types, infants were considered stable insecure if they received any of the three insecure classifications at 12 and 18 months.

Child Behavior Checklist for Ages 2-3 (CBCL) (Achenbach, Edelbrock & Howell, 1987). The CBCL is a one hundred item questionnaire designed to assess behavioral and emotional problems in children ages 2-3. The questionnaire generates two broad band factors, Externalizing and Internalizing Problems, which were used for analyses in this study. Unlike the CBCL for older children, there are no sex specific scales on the 2-3 year old version, so that the Externalizing and Internalizing factors consist of the same items for boys and girls. The mean test-retest reliability is reported by the authors to be .87. Discriminative validity between nonreferred children and children referred to mental health services is strong, and divergent validity has been demonstrated by a lack of significant correlations between the CBCL and standard cognitive measures.

Results

Results are presented in four stages: (1) descriptive statistics and control analyses; (2) group differences analysis between attachment security and age 3 behavior problems; (3) zero-order correlations among maternal depressive symptoms, perceived infant difficulty, and age 3 behavior problems; (4) multivariate correlates of age 3 behavior problems, and (5) correlates of age 2 and 3 maternal depressive symptoms.

Descriptive Statistics and Control Analyses

Means, standard deviations, and ranges of scores on all measures but attachment security classification are presented in Table 1. The number of cases for variables differ for reasons already given in the description of the sample. ICQ scores are comparable to those of published norms when they are interpolated to account for summing and averaging across different age versions of the Difficulty factor (Bates, Freeland, & Lounsbury, 1979). BDI scores also are comparable to others samples of low-income mothers (Christopoulos, Cohn, Shaw, Joyce, Sullivan-Hanson, Kraft, & Emery, 1987), and CBCL t-scores are extremely close to those of published norms (Achenbach et al., 1987).

Attachment classification percentages for 12- and 18-month Strange Situation assessments are presented in Table 2. Overall, of the 86 infants assessed at both 12 and 18 months, 25 of 46 (54.3%) remained securely attached, a figure that is within the expected norms for low-income samples. Please see Vondra, Dowdell Hommerding, and Shaw (1992) for more information about mediators of attachment stability in the present cohort.

Pearson correlation coefficients for mother's age, education, and family income were computed with all measures of attachment security, maternal depressive symptoms, infant temperament, and age 3 child behavior by gender to examine the influence of demographic risk factors. Only two of 60 correlations attained statistical significance, both for girls. For girls, maternal education and family income were negatively related to Year 2 maternal depressive symptoms (r = -.29, p < .05 for education; r = -.38, p < .01 for income). However, these findings should be interpreted with caution give that 3 tests would be expected to be significant by chance using a 5% significance level.

Attachment Security and Age 3 Behavior Problems

As mentioned earlier, four strategies were used to compare groups of securely and insecurely attached infants. In all cases but the attachment stability analysis, separate MANOVA analyses were computed at 12 and 18 months. Repeated measures analyses were not used because of case loss between 12 and 18 months (n = 80 at 12 months and 72 and 18 months). Given previous research demonstrating sex differences with low-income samples (Renken et al., 1989), MANCOVAs initially were conducted to test for main effects of sex and sex x attachment interactions.

When secures were compared to each insecure group at 12 and 18 months (i.e., B versus A, C, & D), no main effects of attachment or sex were found. There was a significant sex x attachment interaction at 18 months, F(3, 64) = 2.22, p <.05, indicating that girls with ambivalent attachments demonstrated higher rates of CBCL Internalizing problems than boys (means are 76 for girls and 53.4 for boys). However, this interaction was based on such a small sample that the results are not interpretable (i.e., N = 1 girl and 5 boys). Similar non-significant results were found when the disorganized classification was not used and avoidant and resistant infants were compared individually with secures, and when disorganized infants were ranked as the least secure group (1 = secure, 2 = avoidant or resistant, 3 = disorganized). However, when secures were compared with all insecure infants, though there were no main effects for sex or sex x attachment interactions, a non-significant trend was found for attachment status at 12 months, F(2, 77) = 2.49, p < .10, and a significant main effect for attachment was found at 18 months, F(2, 69) = 3.39, p < .05. Univariate comparisons for the secures versus all insecures analysis are presented in Table 3. Though relations only approached significance at 12 months, attachment insecurity (A, C, or D) at 18 months was significantly associated with age 3 Externalizing problems. Insecurity at 18 months also was associated with significantly higher rates of CBCL Internalizing problems. Finally, when stable secure versus stable insecure infants were compared, no sex or sex x attachment effects were significant, and a non-significant trend for attachment status was found, F(2, 36) = 2.60, p <.10 (see Table 3). Follow-up univariate analyses indicated that infants who showed some type of insecure attachment at 12 and 18 months showed significantly higher externalizing problems at age 3 than stable secure infants. There also was a non-significant trend for stable insecure infants to display higher rates of internalizing problems.

Maternal Depressive Symptoms, Perceived Infant Difficulty, and Age 3 Behavior Problems

In Table 4 Pearson product correlation coefficients are presented separately by sex for maternal depressive symptoms, perceived infant temperament, and age 3 CBCL Externalizing and Internalizing factors. For boys, all three scores of maternal depressive symptomatology and Year 2 ICQ scores were significantly related to Externalizing problem behavior, while only Year 1 BDI scores were associated with Internalizing difficulties. For girls, only Year 1 and 2 ICQ scores were related to age 3 Internalizing and Externalizing problems.

Regression Analysis: Correlates of Age 3 Behavior Problems

Variables of theoretical relevance showing significant zero order correlations (p<.05) with age 3 Externalizing or Internalizing problems were entered into hierarchical multiple regression equations. Since sex differences were found with respect to maternal depressive symptoms, perceived infant difficulty, and later problem behavior, regression models were computed separately by gender. Predictors were entered in chronological order with earlier variables entered first. For all equations, after the first variable was entered, no additional predictors added significant variance. For boys, the prediction of both CBCL Externalizing and Internalizing problems was largely accounted for by Year 1 BDI scores, with attachment security at 18 months, and Beck Year 2 and 36-month scores adding insignificant variance to the equation. For girls, variance for Externalizing and Internalizing scores were largely explained by ICQ Year 1 scores, with 18-month attachment security, and ICQ Year 2 ratings providing no substantial increase in R2.

Cumulative Stress and Age 3 Behavior Problems

In order to examine the non-specific impact of insecure attachment, maternal depressive symptomatology, and perceived infant difficulty on age 3 CBCL Externalizing and Internalizing problems, four stressor groups were formed, ranging from 0 to 3 stressors present, based on the presence of the following risk factors: (1) stable insecure attachment at 12 and 18 months; (2) BDI scores of 15 or greater at Year 1, Year 2, or 36 months (considered to be moderate or above on the BDI); and (3) ICQ Difficulty scores one standard deviation above the sample means at Year 1 or Year 2. The sample size for this analysis was smaller because only subjects with complete data for all variables were used (N = 63). A MANOVA was computed for the entire sample and for boys, but not for girls because of the small N (i.e., N = 1 for the girls' group with 3 stressors). Results are presented in Table 5. A significant effect for stressor group was found for both Externalizing and Internalizing problem behavior for the entire sample and boys alone. In general, as stressors increased, so did the risk of externalizing and internalizing problems, though post-hoc comparisons showed that differences were not found between groups having one versus two stressors. Rather, significant differences were found between extreme groups (0 versus 3 stressors), and between extreme groups and those having 1 or 2 stressors present.

Correlates of Age 2 and 3 Maternal Depressive Symptomatology

Pearson product correlations were computed by sex to examine prospective correlates of age 2 and 3 maternal depressive symptoms. Results are presented in Table 6. For boys, only Year 1 BDI scores and Year 1 and 2 BDI scores (r = .39, p < .01) were related to Year 2 and 36-month BDI scores, respectively. Stability of the BDI was higher for girls than for boys at each assessment point, though these differences were not significant when Fisher's Z tests were computed. For girls, insecure attachment at 18 months was related to Year 2 and 36-month maternal depressive symptoms, and stable insecure attachment also was related to 36-month BDI scores. However, when for girls Year 1 BDI scores were placed in a hierarchical regression equation prior to entering 18-month attachment ratings, the latter provided no significant increase in R2.

Discussion

Results from the study provide data from a low SES sample on the relationship between attachment security, perceived infant temperament, and maternal depressive symptoms, and their relationship with age 3 behavior problems. Attachment insecurity was related to age 3 behavior problems but only when all insecure sub-groups were combined into one group. Relations between attachment insecurity, particularly at 18 months, and age 3 CBCL ratings were more consistent for externalizing versus internalizing problems, and at 18 versus 12 months. Infants who showed stable insecure attachments at 12 and 18 months also were more likely to show age 3 externalizing problems than infants with stable secure attachments.

Sex differences were found with respect to the relations among infant temperament, maternal depressive symptoms, and age 3 behavior problems. For boys, maternal depressive symptoms were repeatedly associated with report of age 3 externalizing difficulties. For girls, age 1 and 2 difficult temperament ratings were strongly related to internalizing problems, and to a lesser extent, externalizing problem behaviors. Multivariate analyses of age 3 behavior problems indicated that boys' age 3 behavior problems were largely explained by age 1 maternal depressive symptoms, and girls' age 3 problems were accounted for by age 1 ICQ scores. However, when a non-specific index of stressors was created, age 3 CBCL Externalizing and Internalizing scores could be differentiated by the number of stressors present, particularly in extreme cases (0 versus 3 present). Finally, for girls only, 18-month attachment insecurity was a strong predictor of Year 2 and 36-month maternal depressive symptoms, though its effects were accounted for by earlier BDI scores.

At a broad level, the relationships between attachment security and age 3 behavior problems are in agreement with previous studies of low-income families, but question the risk status of specific types of insecurity. Results with middle-class samples have produced inconsistent results with specific types of insecurity (Bates et al., 1985; Fagot & Kavanaugh, 1990; Lewis et al., 1984). Though more consistent findings have emerged in the two longitudinal studies of low SES samples, inconsistency with respect to the type of insecure infant at greater risk for developing externalizing problems has remained. While in the Minnesota poverty sample anxious-avoidant infants were found to show higher rates of externalizing problem behavior at ages 5 and 7-8 (Erickson et al., 1985; Renken et al., 1989), Lyons-Ruth et al. (1993) found disorganized infants to be at greater risk for externalizing problems. Though some of the differences between the findings can be attributed to the advent of the disorganized sub-type (i.e., the Minnesota cohort was not analyzed using the D classification), the present results raise questions about the presumed extreme-risk status of disorganized infants living in high risk environments. Disorganized insecurity in our sample, just as avoidant and resistant insecurity, was not related to later problem behavior. The combined groups of all insecures did show the relationship.

The sex differences found regarding the relationship between maternal report of infant temperament, depressive symptoms, and age 3 child behavior problems raise methodological and substantive issues. First, the fact that mothers were reporting on all three scales calls into question the validity of such associations. There have been continuing concerns about the extent to which maternal depression may be influencing maternal report of child behavior (Fergusson, Lynskey, & Horwood, 1993; see Richters, 1992 for review of distortion hypothesis). Maternal depressive symptoms also could affect maternal perception of difficult temperament, though this relationship was only significant for boys at one time point (i.e.,Year 1 BDI and ICQ scores, r = .31 p < .01), and maternal perception of boys' Year 1 difficult temperament was unrelated to age 3 behavior problems. Moreover, if there was maternal reporting bias, it appears to be mediated by the sex of the child. Maternal depressive symptoms were related only to boys' age 3 behavior problems, while the association between ICQ scores and age 3 behavior problems was evident more consistently for girls. Alternatively, according to a sex-type distortion model, mothers may be more likely to report the presence of internalizing problems for girls and externalizing problems for boys.

Substantively, it is intriguing that maternal depressive symptoms were more consistently associated with boys' age 3 externalizing problems, and that perception of difficult temperament was more consistently related to girls' later internalizing problems. If the reports reflect accurate reporting at both time points for both sexes, the relation between maternal depressive symptoms and boys' externalizing behavior could be the result of decreased maternal responsiveness mediated by depressive symptoms. As Zahn-Waxler et al. (1990) have noted, the unresponsiveness that characterizes some forms of depression and affects the formation of attachment relationships, may lead children to engage in behaviors that elicit negative parental attention. As parents show greater tolerance of aggression among boys, it is not surprising that problem behavior would take the form of externalizing difficulties. The present results are consistent with such a model.

For girls, the relation between perceived difficult temperament and age 3 problems was stronger for internalizing rather than externalizing problems. Though these differences were not significant, it is important to note that this would be unlikely given that the CBCL Externalizing and Internalizing factors had a .71 correlation in the present sample, within the range of published norms (.68). This relationship could be the result of parents encouraging girls to channel their problem behavior into sex-appropriate outlets. Downey, Feldman, Khuri, and Friedman (in press) have noted that depression and aggression in childhood may be the behavioral manifestation of the same underlying distress; negative affect that becomes socialized differentially for boys and girls based on normative sex-role expectations. The present findings suggest that such dysfunction begins to become differentiated by gender from age 1 to 3 (Keenan & Shaw, in press). Further, it is important to note that sex differences emerged despite the absence of mean differences between boys and girls on either CBCL factor (for Externalizing, means were 51.3 for boys and 49.7 for girls, and for Internalizing, 48.9 for boys and 49.6 for girls).

The results regarding the cumulative impact of attachment insecurity, maternal depressive symptomatology, and later behavior problems are in accord with previous research by Rutter et al. (1975a, 1975b) and others (Blanz et al., 1991; Shaw & Emery, 1988). The findings of Sanson et al. (1991) are particularly relevant, as they found that difficult temperament assessed during the first year of life (4-8 months old) was predictive of age 4-5 externalizing and internalizing problems, but that this association increased substantially when other risk factors were considered with temperament. In the present study, early ratings of difficult temperament also was an important correlate of risk status for late behavior problems, and its ability to differentiate age 3 behavior problems was further enhanced by knowing whether other risk factors were present. Between stressor group differences were most pronounced when those with no stressors were compared to subjects with all three present.

Perhaps the most provocative finding was the relationship between girls' 18-month attachment insecurity and later maternal depressive symptoms. While this relationship was imperceptible for boys, it was significant for girls across two time points when children were age 2 and 3. This finding is in need of replication before its implications can be evaluated, but a first attempt will be made in order to guide setting up replications. At a speculative level, mothers' relationship with their daughters may have a greater impact on their subsequent well-being than with their sons because they feel more responsible for the welfare of their same-sex offspring. Mothers may feel that they should be better at understanding the cues and needs of daughters. However, signs of distress in the early mother-daughter relationship, as represented by an insecure attachment, may increase the likelihood of later maternal depressive symptoms more so than with insecure mother-son dyads where the expectations of the relationship's success would not be as high.

The study is not without its limitations. First, though the sample size is comparable to or larger those in many previous studies of infant attachment security and later behavior problems, cell sizes of comparison groups become small when insecure sub-types are examined, particularly by sex. It could be that with a greater N, previous findings from low SES sample would have been replicated (Lyons-Ruth et al, 1993; Erickson et al., 1985). However, a comparison of sample means did not show the non-significant trends that would indicate such relationships may have been limited by the sample size.

Second, the differences in relationships with previous low SES samples also could have been related to the timing and source of information for the follow-up assessment of behavior problems. In both the Lyons-Ruth and Erickson studies, subjects were 4 1/2 to 5 years of age and informants were teachers. In the present sample, behavioral adaptation was assessed at age 3 and mothers were informants. Perhaps it is only when children have developed into the preschool period and have been in active interactions with peers for a sustained period that differences emerge among insecure sub-types. Also in support of this explanation, teachers's ratings would be based on the child's interactions with peers to a greater degree than parents' ratings.

Third, it is important to discuss the generalizability of these findings to other populations. On the favorable side, mothers who continued in the study through the age 3 assessment showed no demographic differences with those who terminated. Though it is a low income sample, in which rates of psychopathology are generally elevated, the authors believe that our study participants do not represent a clinical sample. This view is reinforced by mean CBCL t-scores (49 for Internalizing and 50 for Externalizing), which are extremely close to those obtained for community-based samples of middle-class children of the same age (Achenbach et al., 1987). It is further reinforced by our initial sample selection process. By recruiting families from WIC, the range of family dysfunction was most likely restricted, as all subject families were involved in a program designed to improve the quality of their children's nutritional needs. In sum, the mean CBCL scores, the involvement in WIC, and our own clinical impressions lead us to believe the generalizability of our results may be limited to relatively high functioning low SES families.

Fourth, it is important to note that a measure of depressive states was used, not clinical depression. Though mothers were instructed to complete the BDI based on symptomatology during the past six months, it's unclear how the relationship between maternal depression and attachment security, and maternal depression and age 3 behavior problems would have been affected if mothers had completed structured diagnostic interviews. Use of the BDI or other self-report measures probably inflates the rate of "depression", as criteria for a syndromal diagnosis is more stringent. In particular, it would be of interest to see if other studies that have prospective data on attachment insecurity and later maternal depression, measured diagnostically, would replicate the present results.

In sum, the results offer broad support for the importance of attachment insecurity in the formation of later behavior problems for children from low SES families. However, as Fagot & Kavanaugh (1990) emphasize, clinicians should be cautious in applying the attachment paradigm to clinical phenomena. In the present study, there was greater risk associated with having an insecure attachment, but contrary to previous research, no hierarchy of risk for insecure sub-groups. The present findings also emphasize the need to develop sex-specific models for at-risk populations. Consistent sex differences were found in the relationships between maternal depressive symptoms, infant temperament, and later behavior problems. Future researchers should not expect models based primarily on samples of boys and the development of externalizing outcomes to be applicable to girls and the development of internalizing problems, respectively. Finally, the present findings affirm past research on cumulative family stress and child outcome, and emphasize the value of examining both specific and non-specific models of stressors' interaction.

References

Achenbach, T.M., Edelbrock, C., & Howell, C. (1987). Empirically-based assessment of the behavioral/emotional problems of 2-3 year old children. Journal of Abnormal Child Psychology, 15(4), 629-650.

Ainsworth, M.D.S., Blehar, M.C., Waters, E., & Wall, S. (1978). Patterns of attachment. Hillside, N.J.: Lawrence Erlbaum.

Ainsworth, M.D.S., & Wittig, D. (1969). Attachment and exploratory behavior of one-year-olds in a strange situation. In B. M. Foss (Ed.), Determinants of infant behavior (Vol. 4). London: Metheun.

Arend, R., Gove, F.L., & Sroufe, L.A. (1979). Continuity of individual adaptation from infancy to kindergarten: A predictive study of ego-resiliency and curiosity in preschoolers. Child Development, 50, 950-959.

Bates, J.E., Freeland, C.B., & Lounsbury, M.L. (1979). Measurement of infant difficultness. Child Development, 50, 794-803.

Bates, J.E., Maslin, C.A., & Frankel, K.A. (1985). Attachment security, mother-child interaction, and temperament as predictors of behavior-problem ratings at age three years. In I. Bretherton & E. Waters (Eds.), Monographs of the Society for Research in Child Development, 50, Nos. 1-2, 167-193.

Blanz, B., Schmidt, M.H., & Esser, G. (1991). Familial adversities and child psychiatric disorders. Journal of Child Psychiatry and Psychology, 32, 939-950.

Beck, A.T., & Beamesderfer, A. (1974). Assessment of depression: The Depression Inventory. In P. Pichot (Ed.), Psychological measurement in psychopharmacology: Modern problems in pharmacopsychiatry (Vol. 7). Basel, Switzerland: Kanger.

Beck, A.T., Ward, C.H., Mendelon, M., Mock, J.E., & Erbaugh, J.K. (1961). An inventory for measuring depression. Archives of General Psychiatry, 4, 561-571.

Bowlby, J. (1980). Attachment and loss: Vol. 3, Loss, sadness and depression. New York: Basic Books.

Brook, J.S., Whiteman, M.M., & Finch, S. (1993). Childhood aggression, adolescent delinquency, and drug use: A longitudinal study. The Journal of Genetic Psychology, 153, 369-383.

Campbell, S.B., Breaux, A.M., Ewing, L.J., & Szumowski, E.K. (1986). Correlates and predictors of hyperactivity and aggression: A longitudinal study of parent-referred problem preschoolers. Journal of Abnormal Child Psychology, 14, 217-234.

Christopoulos, C., Cohn, D.A., Shaw, D.S., Joyce, S. Sullivan-Hanson, J., Kraft, S.P., & Emery, R.E. (1987). Children of abused women: I. Adjustment at time of shelter residence. Journal of Marriage and the Family, 49, 611-619.

Crockenberg, S.B. (1981). Infant irritability, mother responsiveness, and social support influences on the security of infant-mother attachment. Child Development, 52, 857-865.

Cummings, E.M., Iannotti, R.J., & Zahn-Waxler, C. (1989). Aggression between peers in early childhood: Individual continuity and developmental change. Child Development, 60, 887-895.

Downey, G., & Coyne, J.C. (1990). Children of depressed parents: An integrative review. Psychological Bulletin, 108, 50-76.

Downey, G., Feldman, S., Khuri, J., & Friedman, S. (in press). Maltreatment and childhood depression. In W.M. Reynolds & H.F. Johnson (Eds.), Handbook of depression in children. New York: Plenum Press.

Erickson, M.F., Sroufe, L.A., & Egeland B. (1985). The relationship between quality of attachment and behavior problems in preschool in a high-risk sample. In I. Bretherton & E. Waters (Eds.), Growing points of attachment theory and research. Monographs of the Society for Research in Child Development, 50, Nos. 1-2, 147-167.

Fagot, B. (1984). The consequents of problem behaviors in toddler children. Journal of Abnormal Child Psychology, 12, 385-395.

Fagot, B.I., & Kavanaugh, K. (1993). Parenting during the second year: Effects of children's age, sex, and attachment classification. Child Development, 64, 258-271.

Fagot, B.I,. & Kavanagh, K. (1990). The prediction of antisocial behavior from avoidant attachment classifications. Child Development, 62, 864-873.

Fergusson, D.M., Lynskey, M.T., & Horwood, L.J. (1993). The effect of maternal depression on maternal ratings of child behavior. Journal of Abnormal Child Psychology, 21, 245-270.

Forehand, R., McCombs, A., & Brody, G.H. (1987). The relationship of parental depressive mood states to child functioning: An analysis by type of sample and area of child functioning. Advances in Behavioral Research and Therapy, 9, 1-20.

Glueck, S., & Glueck, E. (1950). Unraveling juvenile delinquency. New York: Commonwealth Fund.

Greenberg, M.T., & Speltz, M.L. (1988). Contributions of attachment theory to the understanding of conduct problems during the preschool years. J. Belsky & T. Negworski (Eds.), Clinical implications of attachment (pp. 177-218). Hillsdale, New Jersey: Earlbaum.

Hirschi, T. (1969). Causes of Delinquency. Berkeley, CA: University of California.

Jacobson, J.L., & Willie, D.E. (1986). The influence of attachment pattern of developmental changes in peer interaction from the toddler to the preschool period. Child Development, 57, 338-347.

Kazdin, A.K. (1987). Conduct disorders in childhood and adolescence. Newbury Park, CA: Sage.

Keenan, K., & Shaw, D.S. (in press). The development of coercive family processes: the interaction between aversive toddler behavior and parenting factors. In J. McCord (Ed.), Coercion and punishment in long-term perspectives. Boston, MA: Cambridge Press.

Keller, M.B., Beardslee, W.R., Dorer, D.J., Lavor, P. W., Samuelson, H., & Klerman, G.R. (1986). Impact of severity and chronicity of parental affective illness on adaptive functioning and psychopathology in children. Archives of General Psychiatry, 43, 930-937.

LaFreniere, P. J., & Sroufe, L.A. (1985). Profiles of peer competence in the preschool: Interrelations among measures, influence of social ecology, and relation to attachment history. Developmental Psychology, 21, 56-69.

Lewis, M., Feiring, C., McGuffog, C. & Jaskir, J. (1984). Predicting psychopathology in six-year-olds from early social relations. Child Development, 55, 123-136.

Loeber, R. (1982). The stability of antisocial and delinquent childhood behavior. Child Development, 53, 1431-1446.

Loeber R., & Dishion, T. (1983). Early predictors of male delinquency: A review. Psychological Bulletin, 94, 68-99.

Loeber, R. & Stouthamer-Loeber, M. (1986). Family factors as correlates and predictors of juvenile conduct problems and delinquency. In M. Tonry and N. Morris (Eds.) Crime and Justice: A review of research (Vol. 7, pp. 29-149). Chicago: University of Chicago Press.

Londerville, S. & Main, M. (1981). Security of attachment, compliance, and maternal training methods in the second year of life. Developmental Psychology, 17, 289-299.

Lyons-Ruth, K., Alpern, L., & Repacholi, B. (1993). Disorganized infant attachment classification and maternal psychosocial problems as predictors of hostile-aggressive behavior in the preschool classroom. Child Development, 64, 572-585.

Lyons-Ruth, K., Connell, D.B., Grunebaum, H., & Botein, S. (1990). Infants at social risk: Maternal depression and family support services as mediators of infant development and security of attachment. Child Development, 61, 85-98.

Lyons-Ruth, K., Connell, D.B., Zoll, D., & Stahl, J. (1987). Infants at social risk: Relations among infant maltreatment, maternal behavior, and infant attachment behavior. Developmental Psychology, 23, 223-232.

Lyons-Ruth, K., Repacholi, B., McLeod, S., & Silva, E. (1991). Disorganized attachment behavior in infancy: Short-term stability, maternal and infant correlates and risk-related sub-types. Development and Psychopathology, 3, 377-396.

Main, M., & Solomon, J. (1987). Discovery of an insecure disorganized/disoriented attachment pattern: Patterns, findings and implications for the classification of behavior. In M. Yogman & T.B. Brazelton (Eds.), Affective development in infancy (pp. 95-124). Norwood, NJ: Aplex.

Martin, J. (1981). A longitudinal study of the consequences of early mother-infant interaction: A microanalytic approach. Monographs of the Society for Research in Child Development, 46, (3, Serial No. 190).

Matas, L., Arend, R.A., & Sroufe, L.A. (1978). Continuity of adaptation in the second year: The relationship between quality of attachment and later competence. Child Development, 49, 547-556.

Olweus, D. (1979). Stability of aggressive patterns in males: A review. Psychological Bulletin, 86, 852-875.

Radke-Yarrow, M., Cummings, E.M., Kuczynski, L, & Chapman, M. (1985). Patterns of attachment in two- and three-year-olds in normal families and families with parental depression. Child Development, 56, 884-893.

Renken, B., Egeland, B., Marvinney, D., Mangelsdorf, S., & Sroufe, L.A. (1989). Early childhood antecedents of aggression and passive-withdrawal in early elementary school. Journal of Personality, 57, 257-281.

Reynolds, W.M., & Gould. J.W. (1981). A psychometric investigation of the standard and short form Beck depression inventory. Journal of Consulting and Clinical Psychology 49, 306-307.

Richters, J.E. (1992). Depressed mothers as informants about their children: A critical review of the evidence for distortion. Psychological Bulletin, 112, 485-499.

Rose, S.L., Rose, S.A., & Feldman, J.F. (1989). Stability of behavior problems in very young children. Development and Psychopathology, 1, 5-19.

Rutter, M. (1978). Family, area, and school influences in the genesis of conduct disorders. In L.A. Hersov & D. Schaffer (Eds.), Aggression and anti-social behavior in childhood and adolescence (pp. 95-114). Oxford: Pergamon Press.

Rutter, M., Cox, A., Tupling, C., Berger, M., & Yule, W. (1975a). Attainment and adjustment in two geographical areas: 1. The prevalence of psychiatric disorder. British Journal of Psychiatry, 126, 493-509.

Rutter, M., Yule, B., Quinton, D., Rowlands, O., Yule, W., & Berger, W. (1975b). Attainment and adjustment in two geographical areas: 3. Some factors accounting for area differences. British Journal of Psychiatry, 126, 520-533.

Sanson, A., Oberklaid, F., Pedlow, R., & Prior, M. (1991). Risk indicators: Assessment of infancy predictors of pre-school behavioural maladjustment. Journal of Child Psychology and Psychiatry, 32, 609-626.

Shaw, D.S. (1993). Developmental precursors of antisocial behavior: Ages 1 to 3. Paper presented at the Society for Research in Child Development, New Orleans, LA.

Shaw, D.S., & Bell, R.Q. (1993). Developmental theories of parental contributors to antisocial behavior. Journal of Abnormal Child Psychology, 21, 25-49.

Shaw, D.S., & Emery, R.E. (1988). Chronic family adversity and school-age children's adjustment. Journal of the American Academy of Child and Adolescent Psychiatry, 27, 200-206.

Shaw, D.S., & Vondra, J.I. (1993). Chronic family adversity and infant attachment. Journal of Child Psychology and Psychiatry, 34.

Smith, P.B., & Pederson, D.R. (1988). Maternal sensitivity and patterns of mother-infant attachment. Child Development, 59, 1097-1101.

Sroufe, L.A. (1983). Infant-caregiver attachment and patterns of adaptation in pre-school: The roots of maladaptation and competence. In M. Perlmutter (Ed.), Minnesota symposium in child psychology (Vol. 16, pp., 41-81). Hillsdale, New Jersey: Erlbaum.

Vondra, J.I., Dowdell Hommerding, K., & Shaw, D.S. (1992). Attachment and risk in low-income samples: Security and stability. Presented at the convention of the International Society for Infant Studies, Miami, FL.

Zahn-Waxler, C. (1993). Continuities of behavior problems in preschool children. Paper presented at the Society for Research in Child Development, New Orleans, LA.

Zahn-Waxler, C., Iannotti, R.J., E.M. Cummings, & Denham, S. (1990). Antecedents of problem behavior in children of depressed mothers. Development and Psychopathology, 2, 271-291.