OBJECTIVES. To examine the relation between the pattern of maternal responsiveness that children experienced in the infancy, preschool, and school-age periods and growth in cognitive skills across 3–10 years of age and determine whether the relation differs by birth status.
METHODS. In 1990–1992, 360 children varying in birth status (very low birthweight [VLBW]; demographically matched controls) were recruited to examine parenting and birth status influences on development. This report includes children with observations of parenting at 6, 12, and 24 months and 3, 4, 6, 8, and 10 years and cognitive skills evaluated at 3, 4, 6, 8, and 10 years of age (71% of original cohort).
RESULTS. Four groups of mothers varying in the pattern of responsiveness displayed across the infancy and the preschool period were found. When controlling for school-age parenting and economic status, children parented with higher levels of responsiveness across both developmental periods, irrespective of birth status, showed higher levels in development than those who experienced responsiveness in only 1 development period or minimal responsiveness. Greater benefit was found for consistency in responsiveness for children born VLBW with less, versus more, severe neonatal complications. Inspection of the means showed that higher risk birth status combined with minimal responsiveness resulted in cognitive scores, on average, 14 points lower than when parented with consistently higher responsiveness.
CONCLUSIONS. Cognitive development for children born at VLBW, particularly those with less severe complications, are supported by consistently responsive parenting across early childhood in similar ways to those born at term. This effect persisted through 10 years of age even after school-age parenting and economic level. These findings have important implications for the timing (across early childhood) and content (responsive interactive behaviors) of early intervention to enhance the outcomes for children born at VLBW.
Despite attempts to improve access to prenatal care, it has been difficult to reduce the rate of infants born at very low birth weight (VLBW), and these infants remain at increased risk for developmental difficulties that can persist into adulthood.1,2 Whereas there is evidence that some develop cognitive skills that are within the average range, others have greater difficulty in achieving skills similar to their term-born peers. Experts in the follow-up care of this infant population have raised important questions that include the need to better understand the mechanisms by which the environment can facilitate development and whether there are “critical periods” during which intervention may be more effective.3
Many studies examining outcomes for children born at VLBW include markers of environmental influence, such as family socioeconomic status (SES) and/or parent education, and find strong relations between these factors and development.4–6 Although important, additional investigation is needed to better understand the underlying quality of caregiving factors that such distal markers represent. Research is specifically needed on the types of caregiving behaviors that may lessen the negative impact of neonatal complications associated with being born at VLBW status on development. To accomplish this goal, greater specificity of caregiving behaviors that are documented through longitudinal research are necessary to determine behaviors that are critical and the extent to which their influence persists across developmental periods.
One aspect of the caregiving environment consistently reported to facilitate more optimal cognitive and social development is contingent responsiveness.7–9 This has been described as a 3-term chain of events: (1) the child signals, (2) the parent responds in prompt and sensitive ways that are linked to the child's signal, and (3) the child, in turn, experiences that his or her needs and interests are important and will be responded to in a sensitive way.10,11 In addition to warm and sensitive support, responsive parenting often includes providing appropriate levels of stimulation in ways that support children's learning.7,12 Responsive caregiving has been linked to children developing greater trust in their caregiving environment that, in turn, promotes cooperation and greater interest in exploration and communication.13,14 This type of support across early childhood has been shown to be important for the development of children born at term and “at risk” into early school-age years.15,16 However, the extent to which early responsive parenting remains uniquely important above and beyond ongoing parenting through the elementary school-age years is not well established.
Over the past decade, we have examined the development of a large cohort of children born at VLBW with varying medical complications and compared them with demographically similar children born at term into low-income families. Through a longitudinal approach, where parenting and child development are examined across multiple developmental periods, we have examined the relative importance of responsive parenting at different developmental periods and how it differs among children of varying birth status.17,18 We reported previously that responsive parenting that occurs at relatively consistent and higher levels across infancy and the preschool period, in contrast to inconsistent or low levels, predicted faster rates of cognitive development for all children. For children born at VLBW, average age level cognitive scores were found only for the group with mothers showing responsive behaviors at relatively high levels across these 2 developmental periods,15 and this sustained into beginning school ages.16 Because these children's cognitive skills and the parenting environment have now been assessed through 10 years of age, we can address the question of whether early versus later responsive caregiving is differentially predictive of cognitive skills through the end of elementary school.
A cohort of 360 families recruited from 1990 through 1992 has been evaluated 8 times from 6 months to 10 years of age as part of a longitudinal study of the relation of parenting behaviors and development for children born at VLBW who varied in severity of neonatal medical complications and a term-born comparison group. Based on the Hollingshead 4-factor scale,19 the families were from lower-middle to lower economic backgrounds. For 98% of the infants, the primary caretaker was the biological mother. This report includes children who were in the sample at 3 years (VLBW: n = 190; term: n = 114) through 10 years of age (VLBW: n = 168; term: n = 88). By 3 years of age, attrition was 15%, and by 10 years, it was 29%. No significant differences in demographic or medical complications were found for children participating through 10 years of age and those who did not.
Children born at VLBW were divided into 2 groups based on the severity of their neonatal complications: (1) severe complications that included bronchopulmonary dysplasia, periventricular leukomalacia, and grade III and IV intraventricular hemorrhage, which place them at high risk (HR) for developmental difficulties (VLBW-HR), and (2) less severe complications, such as respiratory distress syndrome and grade I or II intraventricular hemorrhage, which place them at a lower risk (LR) for developmental problems (VLBW-LR). The criteria used to identify VLBW infants with less severe and more severe medical complications has consistently predicted differences in developmental outcomes between these 2 groups.15,16 For the cohort of children included in this study, these criteria significantly predicted differences in early feeding that, in turn, predicted 24-month mental skills even after controlling for birth weight and gestational age.20 The criteria also has been sensitive to differences in developmental outcomes when used to differentiate infants born early from a separate cohort who did not differ in birth weight or gestational age.21
Based on children who remained in the study at the 10-year age point, no significant differences were found for the interaction between birth status and the maternal responsiveness groups for any of the medical and demographic characteristics. Thus, Table 1 summarizes information by the 3 birth status groups where expected differences were found for gestational age (P < .0001) and birth weight (P < .0001). Follow-up analyses revealed a “stair-step” effect, with children born VLBW-LR being heavier and of greater gestational age than those born VLBW-HR (P < .0001), and these 2 groups combined differed from those born at term (P < .0001). No significant differences were found for the demographic variables. However, because an early birth occurs more often in families from poverty backgrounds,22 this accounted for the higher proportion of black families found in the 2 VLBW groups. The lower proportion of Hispanic families was because of the funding source requirement that recruited families speak English, because the measures were not available for Spanish-speaking-only families when this study began. The group of children born at term was recruited to be comparable with that of those born at VLBW. Thus, these findings can be generalized to families with low birth weight and at-term children from lower income families for black, white, and acculturated Hispanic backgrounds.
Table 2 summarizes information for the 4 patterns of responsiveness with no differences evident for the medical characteristics. However, significant differences were found for maternal age (P < .01), education (P < .0001), and SES (P < .0001). In general, follow-up analyses indicated a stair-step effect for these variables in that mothers who displayed consistently higher levels of responsiveness were older, more educated, and had higher economic status than those displaying responsiveness in only developmental period and that all 3 of these groups were, on average, were older, more education, and had higher economic status than those displaying minimal responsiveness. Mothers who were more responsive during infancy compared with those who were more responsive during the preschool period were less educated and had a lower economic status.
Maternal ethnicity also differed by pattern of responsiveness (P < .0001). A larger proportion of white mothers showed consistently higher levels of responsiveness as compared with the other 3 patterns. Larger proportions of mothers of Hispanic origin also were found in this group, as well as in the group showing higher responsiveness in infancy. Mothers of black origin were less likely to be in the consistently higher levels of responsiveness group than the other 3 groups.
Maternal responsiveness was coded “live” during a 70-minute observation of daily activities and play in the home at each age using criteria outlined below. Procedures for introducing the observation protocol and directions given to mothers have been described in previous publications, as well as the reliability and validity on this measure of parenting.23–25 Trained undergraduate and graduate research assistants, blind to group status, made 4 ratings across the 70 minute observation period. Children's cognitive skills were evaluated in the home at 3, 4, 6, 8, and 10 years of age with timing of assessments based on chronological age for all of the children.
Medical record review was used to obtained information on variables related to early birth (eg, birth weight, gestational age, and neonatal complications). Demographic data were obtained during each home visit from the maternal primary caregiver.
Child Cognitive Skills
Cognitive skills were evaluated with the Stanford-Binet Intelligence Test, 4th ed,26 at each age point using standardized administration techniques with research assistants trained to work with young children. To model growth in cognitive skills over time, mental age scores (in months) were used in data analyses.
Patterns of Maternal Responsiveness
As reported in previous publications15 and summarized in Table 2, responsiveness was captured through two 5-point rating scales (warm acceptance and flexibility/responsiveness) based on the research and coding criteria of Ainsworth et al10 and Bornstein and Tamis-LeMonda.7 This measure has been associated with other aspects of sensitivity in caregiving, including maintaining children's focus of attention; use of rich language; and lower levels of negativity in voice tone, physical intrusiveness, and restrictiveness.15 Interrater reliability based on generalizability coefficients27,28 through a second observer coding ≥15% of the interactions at each age revealed excellent reliability range from 0.82 to 0.85 across the 8 ages studied. Coefficients > 0.50 indicate adequate reliability.29
Because the scores for the warm acceptance and flexibility/responsiveness ratings were significantly intercorrelated within each age point (range: r = 0.50–0.93), these were averaged to obtain a single score at each time point. Scores obtained at 6, 12, and 24 months of age were averaged to obtain the infancy measure of responsiveness. Scores at 3 and 4 years of age were averaged to obtain the preschool age measure. Using hierarchical clustering analyses,30 the infancy and preschool period responsiveness scores were used to identify 4 distinct clusters of mothers. The details of these analyses are available in our previous publications.15,16 Briefly, 4 clusters of mothers were found: (1) highest levels of responsiveness in both developmental periods (29% of sample), (2) lowest levels of responsiveness in both periods (27%), (3) higher levels of responsiveness during infancy with lower levels during the preschool period (21%), and (4) lower levels during infancy with modest increases in the preschool period (23%).
Growth-Curve Modeling Analyses
The relation of maternal patterns of responsive parenting across infancy and the preschool periods and children's birth status to children's growth in cognitive skills across 3–10 years of age was examined using a mixed-model approach to growth-curve modeling.31 An advantage to this approach is that growth is first modeled at the individual level, and all subjects can contribute to the modeling even if not seen at all ages. Individual growth curves involve ≥2 parameters: (1) the intercept, which, in this study, was set at the 6-year age point32 and provides information about the level of cognitive skill across the age period studied; and (2) the slope parameter, which provides information on rate of growth (eg, faster or slower increases). In addition, 2 curvature parameters were required to characterize the changes in mental age growth over time for our subjects. One parameter characterized change in the slope or rate of growth (acceleration or deceleration), whereas the other depicts change in this acceleration or deceleration.
In the present analyses, we were interested in whether the intercept, slope, and curvature for growth in mental skills were predicted by the following: (1) pattern of responsiveness, (2) children's birth status (term, VLBW-LR, or VLBW-HR), and/or (3) interaction between parenting patterns and birth status. To minimize the number of comparisons made on follow-up analyses, orthogonal planned comparisons were used. For the main effect of patterns of responsiveness these included: (1) consistently higher levels versus responsiveness in 1 period (infancy or preschool) combined, (2) the responsiveness in infancy versus preschool period, and (3) minimally responsive versus all of the other groups. Based on our previous work, orthogonal comparisons for birth status were as follows: (1) term and VLBW-LR combined versus VLBW-HR, and (2) term versus VLBW-LR groups.
Responsive parenting as measured by the acceptance/flexibility rating measured at 3, 4, 6, 8, and 10 years of age was included as a time-varying covariate in the model to control for parenting quality at each age. By including this covariate at each age, it was possible to determine whether patterns of responsiveness across early childhood were a unique predictor of cognitive level at each age point. SES was also included as a covariate in the model to determine whether early responsive parenting patterns predicted development above and beyond economic status. The results for the analyzed model are described below and presented in Table 3.
Relation of Patterns of Maternal Responsiveness and Children's Birth Status With Children's Cognitive Development
Significant differences in mental age were predicted by children's birth status, pattern of responsiveness, and the interaction of birth status and responsiveness. These accounted for 21% of the variance in the level (intercept) of mental ages and 30% of the variability in growth.
Relation With Birth Status
Figure 1 illustrates patterns of mental age for each of the 3 risk groups in comparison with the average growth across all of the children. The growth curve for the children born VLBW-LR was similar to the average growth curve, whereas the growth trajectory for children born at term was higher, and those at VLBW-HR showed a lower trajectory.
Children born VLBW-HR showed lower 6-year mental age scores (ie, intercept) than the combined group of those born at VLBW-LR and those born at term (P < .001). However, children born at VLBW-LR also had lower mental age scores than those born at term (P < .001). Children's birth status also was related to variability in the amount of deceleration in mental age scores over time (P < .001). Those born VLBW-HR showed greater deceleration than the other 2 groups combined (P < .01). However, children born at VLBW-LR showed greater deceleration than those born at term (P < .001). Thus, although all of the children showed deceleration in their mental age scores across the 3-10-year age period, those born at VLBW showed greater deceleration than those born at term, with those born at VLBW-HR showing the greatest deceleration.
Relation of Responsive Parenting
As illustrated in Fig 2, irrespective of birth status, children who experienced consistently higher levels of responsiveness across early childhood had higher 6-year mental age scores than those who received this type of parenting only during infancy or the preschool period (P < .001). Those who received responsiveness in only 1 developmental period did not differ in mental age scores. Children who received minimal responsiveness had significantly lower levels of cognitive skills than the other 3 responsiveness groups combined (P < .001). Also irrespective of birth status, responsiveness patterns predicted variability in acceleration (P < .01). By the end of the age period studied, the mental age scores for children parented with consistent responsiveness showed even greater acceleration when compared with those receiving responsiveness in only 1 development period.
However, these results need to be considered in light of a significant interaction between pattern of responsiveness and birth status, which predicted variability in the change in the rate of growth in mental age scores (P < .01). Children born at term and at VLBW-LR showed greater acceleration than those born at VLBW-HR if they were parented with consistently higher levels of responsiveness than if receiving responsiveness in only 1 development period (P < .01). Thus, consistency in parenting across time seemed less beneficial for those born at VLBW with more severe neonatal complications.
Predictors of 10-Year Cognitive Outcomes
The previous analyses provided information regarding the extent to which patterns of responsive parenting and birth status predicted growth in cognitive development based on mental ages. To determine predictors of the age-corrected cognitive skills at the final 10-year age point, an analysis of variance was conducted with 10-year cognitive standard scores as the dependent variable. This revealed a significant effect for pattern of responsiveness (P < .0001) and birth status (P < .0001). The 10-year, standardized cognitive scores for birth status by patterns of responsiveness are summarized in Table 4 along with 95% confidence intervals.
Consistent with the results of the growth modeling analyses, children who received consistently higher levels of responsiveness demonstrated the highest standard scores when compared with those who received responsiveness during only 1 developmental period (P < .01). This difference, on average, was ∼5 intelligence quotient points. Those receiving minimal responsiveness had the lowest scores compared with the other 3 maternal groups (P < .0001). This difference was, on average, 10 intelligence quotient points when compared with those receiving responsiveness in both development periods. Given that the SD for the cognitive measure used was 16 points, the lack of maternal responsiveness during early childhood predicted more than a one-half SD lower score, on average, or a moderate effect size.
By 10 years of age, children born at term and VLBW-LR had higher cognitive scores than those born VLBW-HR (P < .0001), and those born VLBW-LR had lower scores than those born at term (P < .001). Inspection of the means in Table 3 revealed that HR birth status, combined with minimally responsive parenting, resulted in cognitive scores, on average, 14 points lower than when parented with consistently higher levels of responsiveness.
A major goal for this investigation was to understand the relative importance of responsive parenting in the infancy, preschool, and elementary school-age periods in relation to cognitive development through 10 years of age for children with varying birth status. Within this lower economic sample, the best outcomes were found for children whose caregivers showed responsiveness at consistently higher levels across infancy and the preschool period. However, differences were found with respect to the potential benefit of consistency across early childhood by children's birth status, particularly across the end of the age range studied.
The potential positive influence for responsive parenting for all of the children was highlighted by almost an SD advantage in the 10-year cognitive scores for those parented with consistency in this parenting style as compared with those parented with minimal responsiveness. There was some advantage found for children who experienced higher levels of responsive parenting during at least 1 developmental period studied versus minimal responsiveness across both periods. An encouraging finding was that, despite the increased risk for slower cognitive growth found for those born at VLBW, responsive parenting was able to moderate this risk. Of note, responsive environments in infancy and early childhood had a stronger influence on ongoing cognitive development than this type of caregiving in the school-age period.
A responsive caregiving environment seems to support cognitive skills in different ways, depending on the birth status of the infant. Specifically, a consistently responsive environment as compared with responsiveness in only the infancy or preschool period was more beneficial for the children born at term and with less severe medical complications as compared with those born with the highest medical risk. These findings are in contrast to our previous reports where, at earlier ages, responsive parenting was particularly important for those born VLBW with more severe medical complications.15 These results suggest that, as children are in school environments with greater cognitive and social demands, there may be a limit to the extent to which an early responsive parenting environment can support the most biologically fragile children to demonstrate more normal outcomes.
The mechanism by which responsive caregiving supports more optimal cognitive development is thought to occur through multiple processes that work together to promote more positive growth trajectories. Children in this study whose mothers were in the consistently higher responsiveness group were involved in interactions across infancy and early childhood that were sensitive and contingent to their signals and provided appropriate levels of stimulation. In contrast, those parented with minimal responsiveness experienced disregard for signals and needs, a lack of nurturance, high levels of negativity, and low levels of stimulation. These processes seem to cross different developmental theoretical frameworks.12 From attachment theory, responding in sensitive and prompt ways to children's signals develops a trust that caregivers will meet basic needs and promotes a willingness to cooperate and engage in interactions with caregivers.8,10 From the social cultural framework of development, this nurturing caregiving style is expected to go hand in hand with levels of stimulation that are tailored to individual children's learning needs (ie, verbal input and physical demonstrations).7,33,34 With this type of parenting input, children's learning occurs at higher levels than when children attempt to learn independently, and some research shows that this form of parenting support may be particularly important for the learning of infants born at VLBW.35
These results have important implications for the timing and content of parent intervention programs, because they suggest that if parents can be facilitated to use this form of parenting at higher levels across early childhood, it establishes a positive developmental trajectory that persists to later developmental periods. A recent meta-analysis of randomized experimental studies attempting to facilitate parents' use of responsive interactive behaviors suggests that parents can be supported to use these behaviors at higher levels.36 An encouraging finding from a recent experimental study targeting responsive parenting was that comparable increases in this interactive style were seen for mothers parenting infants born at VLBW or at term. Although increased responsiveness for the intervention mothers resulted in increased skills for all of the infants, in some cases, the skills of those born at VLBW were more strongly influenced than those born at term (eg, cooperation and negative affect). Also, responsive behaviors mediated the effectiveness of the intervention on all of the infants' cognitive and social skills.12,37 The mechanism that potentially explained why this program was as effective for infants born at VLBW as those born at term was thought to occur through the specialized caregiver support that met each infant's unique needs.
As in prior research, economic status in this study was associated with outcome.38 However, the influence of sensitive parenting was apparent even when examined with SES in the model. It has been difficult to enhance the financial resources of families living in poverty. Thus, it is encouraging that responsive caregiving occurred despite the challenges of parenting with limited financial resources and that characteristics of the caregiving buffered the negative impact of VLBW. Across school ages, children are exposed to environmental factors outside of the family, including the quality of the school, neighborhood, and peer relations that may also be important influences.39–41 Thus, 1 limitation of the present study was the inability to account for broader environmental influences as children matured.
A second study limitation is the focus on only 1 aspect of the parenting environment, responsive caregiving. Although this was important in understanding individual differences in cognitive outcomes, there are likely other aspects that are of importance. For example, Bradley et al17 found that parent involvement in developmental activities and progress, such as parent-teacher conferences and assistance with homework, behaviors measured by the HOME scale, were related to outcomes by the end of elementary school. In addition, it is recognized that our focus on parenting in a lower economic sample makes it difficult to generalize results to more affluent caregivers.
The results of this study extend a large body of research examining the long-term development for children born at VLBW and the role that limited economic resources has on development.1–3 Whereas poverty is well known to have a negative impact on children's outcomes, through direct examination of the caregiving environment, we can better understand what it is about growing up in poverty that is detrimental to children's development. For ∼20% of our sample, the potential for a normal course of development seems to be hindered by parents' use of minimally responsive behaviors that may also include higher levels of intrusiveness and negativity when interacting with their children. However, it is also encouraging that, in the midst of the challenges of poverty, ∼30% of the parents were able to facilitate their children's development through consistently higher levels of responsive caregiving. It also is important to note that, whereas responsive parenting may have an important moderating effect on children's development, the negative impact of being born at VLBW is apparent across all parenting conditions.
This study was funded by National Institutes of Health grant HD25128.
- Accepted October 17, 2005.
- Address correspondence to Karen Smith, PhD, University of Texas Medical Branch, Department of Neurology, 301 University Blvd, Galveston, TX 77555-0319. E-mail:
The authors have indicated they have no financial relationships relevant to this article to disclose.
- ↵Cooke RWI. Health, lifestyle, and quality of life for young adults born very preterm. Arch Dis Child.2004;89 :201– 206
- Hunt JV, Cooper BA, Tooley WH. Very low birth weight infants at 8 and 11 years of age: Role of neonatal illness and family status. Pediatrics.1988;82 :596– 603
- ↵Vohr R, Coll CT. Neurodevelopmental and school performance of very low-birth-weight infants: a seven-year longitudinal study. Pediatrics.1985;76 :345– 350
- ↵Bornstein M, Tamis-LeMonda CS. Maternal responsiveness and cognitive development in children. In: Bornstein MH, ed. Maternal Responsiveness: Characteristics and Consequences. San Francisco, CA: Josey-Bass; 1989:49– 61
- ↵Ainsworth M, Blehar M, Waters E, Wall S. Patterns of Attachment: A Psychological Study of the Strange Situation. Hillsdale, NJ: Lawrence Erlbaum; 1978
- ↵Landry SH, Smith KE, Swank PR. Responsive social contexts: The origins of early communication and independent problem solving. Dev Psychol. In Press
- ↵Grusec JE, Goodnow JJ. Impact of parental discipline methods on the child's internalization of values: A reconceptualization of current points of view. Dev Psychol.1994;30 :1– 19
- ↵Maccoby E, Martin JA. Socialization in the context of the family: Parent-child interactions. In: Mussen PH, Hetherington EM, eds. Handbook of Child Psychology: Vol. 4. Socialization, Personality, and Social Development. 4th ed. New York, NY: Wiley Publishers; 1983:1– 102.
- ↵Hollingshead AB. Four Factor Index of Social Status. New Haven, CT: Department of Sociology, Yale University; 1975
- ↵Hook-Morris B, Miller-Loncar CL, Landry SH, Smith KE, Swank PR, Denson SE. Feeding, medical factors, and developmental outcome in premature infants. Clin Pediatr (Phila).1999;38 :451– 457
- ↵Niswander K, Gordon M. The Collaborative Perinatal Study of the Institute of Neurological Diseases and Strokes: The women and Their Pregnancies. Philadelphia, PA: Saunders; 1972
- ↵Thorndike RL, Hagen EP, Sattler JM. Guide for Administering and Scoring the Stanford-Binet Intelligence Scale, 4th ed. Chicago, IL: Riverside Publishing; 1986
- ↵Fleiss JL. The Design and Analysis of Clinical Experiments. New York, NY: Wiley; 1986
- ↵Ward JH. Hierarchical grouping to optimize an objective function. J Am Stat Soc.1963;77 :841– 847
- ↵Littell RC, Milliken GA, Stroup WW, Wolfinger RD. SAS Systems for Mixed Models. Cary, NC: SAS Institute, Inc; 1996
- ↵Bryk AS, Raudenbush SW. Hierarchical Linear Models. London, United Kingdom: Sage Publication; 1992
- ↵Rogoff B. Apprenticeship in Thinking. New York, NY: Oxford University Press; 1990
- ↵Vygotsky LS. Mind in Society: The Development of Higher Psychological Processes. Cambridge, United Kingdom: Harvard University Press; 1978
- ↵Landry SH, Garner PW, Swank PR, Baldwin CD. Effects of maternal scaffolding during joint toy play with preterm and full-term infants. Merrill-Palmer Q.1996;42 :177– 199
- ↵Landry SH, Smith KE. The influence of parenting on emerging literacy skills. In: Neumann S, Dickinson D, ed. Handbook on Research in Early Literacy Development. New York, NY: Guilford Press; 2006:135–148
- ↵Larson R, Richars MH. Daily companionship in late childhood and early adolescence: changing developmental contexts. Child Dev.1991;62 :300
- Sampson RJ, Raudenbush SW, Earls F. Neighborhoods and violent crime: a multilevel study of collective efficacy. Science.1997;277 918–924
- Copyright © 2006 by the American Academy of Pediatrics