Attachment Classifications Among 18-Month-Old Children of Adolescent Mothers

Objectives  To determine (1) patterns of secure vs insecure attachment relationships in infants of adolescent and nonadolescent mothers and (2) if these patterns are mediated by parenting characteristics, including depression, self-esteem, parenting stress, child abuse potential, psychological distress, rating of infant temperament, and the caregiving environment.

Participants  Fifty-one adolescent mothers and their 18-month-old infants were compared with 76 nonadolescent mothers and their 18-month-old infants.

Main Outcome Measures  Infant attachment classifications were assessed via the Ainsworth Strange Situation. Maternal and infant characteristics were obtained through self-report measures.

Results  There were no differences in attachment classification between infants of adolescent mothers and nonadolescent mothers. Secure attachment classification was found in 67% of the infants of adolescent mothers and 62% of the infants of nonadolescent mothers. There were significant differences in the self-reported maternal characteristics. Adolescent mothers reported lower self-esteem (P<.05), more parenting stress (P<.05), more child abuse potential (P<.05), and provided a lower quality of home environment (P<.05) than nonadolescent mothers. Adolescent mothers also rated their infants as having a higher activity level (P<.05) than infants born to nonadolescent mothers. In multivariate analysis, none of these variables or social classes were found to affect attachment classification.

Conclusions  Infants of adolescent and nonadolescent mothers show similar patterns of attachment. Adolescent and nonadolescent mothers show substantial differences in parenting characteristics and in how they rate their infants' temperaments. However, these differences do not seem to impair the infant-mother attachment relationship.

ADOLESCENT mothers are thought to lack the skills to adequately mother their infants. Findings suggest that adolescent mothers provide less optimal caregiving, less sensitivity to their infants' cues, less interactive sensitivity, and less emotional and verbal responsiveness than nonadolescent mothers.^1-4 Observation of the home environment suggests that adolescent mothers provide a less than optimal setting for their children, including less stimulation, inconsistent learning environments, more negative interaction, more restrictive, irritable, hostile, and punitive responses, and a tendency to choose physical rather than verbal modalities during interaction with their children.^1,4-8

These concerns about the parenting abilities of adolescent mothers have been attributed to their lack of knowledge about child development in general, a poor understanding of their infants' developmental abilities, and, probably related, the tendency to rate their infants' behavior or temperament as being more difficult.^6,9,10 The offspring of adolescent mothers are thought to be at risk for poor outcome in childhood, including behavior problems, poor communicative skills, cognitive impairment, and school failure.^11-17 One possible source of later problems in the offspring of adolescent mothers is early impairment in the mother-child attachment relationship. If adolescent mothers provide less optimal and/or less sensitive caregiving to their infants, it is reasonable to hypothesize that the quality of the attachment relationship would be compromised in adolescent mother-infant dyads.

Despite nearly 3 decades of research since the advent of the Ainsworth Strange Situation¹⁸ to qualify the attachment relationship in a behavioral observation setting, surprisingly little is known about attachment between adolescent mothers and their infants. Only 5 studies have been reported^19-23 but the small sample sizes used^19,21,22 and the lack of an appropriate nonadolescent comparison group makes interpreting the results problematic.^19,20,22,23 Of these 5 studies, 2 compared their adolescent sample with the Ainsworth sample.^19,20 One study found no differences with the Ainsworth sample when insecure attachment categories A and C were combined¹⁹; yet, another found higher incidences of insecure attachment when compared with the Ainsworth sample.²⁰ Of the remaining 3, 1 study compared attachment classifications of infants of adolescent mothers with infants of adult mothers and no significant differences were found.²¹ Another study, in which a modified Strange Situation was used, compared subjects with another sample of infants of adolescent mothers, although it is unclear whether this second sample used the same version of the Strange Situation.²² When comparing secure vs insecure classifications between these 2 samples, significant differences were found. In the remaining study, which did not include a comparison group, most infants were seen as securely attached.²³

In other populations, the quality of the mother-infant attachment relationship has been related to maternal factors, including self-esteem,²⁴ mental illness,^25-27 and stress in the family.²⁸ Also, maternal perceptions of infant temperament have been associated with infant behavior in the Strange Situation.²⁹

The purposes of this study were 2-fold. First, we wanted to study the attachment relationship in infants of adolescent mothers in a sample with an appropriate comparison group of older mothers and with sufficient statistical power to detect hypothesized differences. Second, we wanted to study maternal factors thought to differentiate adolescent mothers from older mothers that might be related to hypothesized differences in attachment. The factors included depression, self-esteem, parenting stress, child abuse potential, psychological distress, mother's perception of infant behavior, socioeconomic status (SES), and the quality of the home environment.

One hundred forty pairs of mothers and their 18-month-old infants (80 nonadolescent mother-infant pairs enrolled in the Providence, RI, site of the National Institutes of Health Maternal Lifestyle Study [MLS] and 60 adolescent mother-infant pairs) participated in the study. All subjects were initially recruited to be part of a prospective study aimed at examining the effects of maternal lifestyles during pregnancy on a variety of infant outcomes. The MLS is a multisite (Providence, Miami [Fla], Memphis [Tenn], and Detroit [Mich]) longitudinal study of the effects of prenatal cocaine/opiate exposure on child outcome conducted under the auspices of the National Institute on Child Health and Human Development Neonatal Research Network. Adolescent mothers (<18 years) were excluded from MLS. Therefore, we recruited a separate cohort of adolescent mothers for the current study. Recruitment of the adolescent cohort occurred in parallel to recruitment for MLS and followed the same study protocol. The MLS sample includes a group of infants exposed to cocaine/opiates and a comparison group of infants not exposed. Exposure was determined by a meconium toxicology screen followed by gas chromatography/mass spectrometry confirmation and/or self-report.³⁰ Only mothers who denied cocaine/opiate use and whose infants showed negative meconium results were included in the comparison group. This comparison group formed the older mother group for the current study. Maternal report and meconium testing were also used to verify that mothers in the adolescent group did not use cocaine and/or opiates during this pregnancy. In addition, because we were interested in the mother-infant attachment relationship, only biological mothers who had been living with the paired infants since birth and had been the primary caretakers of these infants were included. Primary caretaker was determined by maternal responses to the following questions: "Who makes decisions about the baby's care?" "Who bathes and changes the baby?" "Who feeds the baby during the day and at night?" "Who plays with the baby the most?" This information was collected at various time periods throughout the study. Information regarding the amount of care provided by other caretakers, such as relatives or day care centers, was not collected. Subjects from the adolescent group were not included if they did not complete the research procedure (9 subjects), technical problems were experienced during videotaping (7), the mother was not present for the assessment and another caregiver was used (1), or the mother was not the primary caregiver (1). The remaining 51 subjects comprised the adolescent sample. Subjects were excluded from the sample of older mothers because the mother was not the primary caregiver (1) or data were incomplete (3). The total number of subjects in the nonadolescent sample was 76. The mother-infant dyads were seen when the infants were aged 1, 4, 8, 10, and 18 months.

Demographic and medical information included maternal education, race, parity, number of prenatal care visits, birth weight, gestational age, sex, 1- and 5-minute Apgar scores, and the amount of alcohol, marijuana, and cigarette use during this pregnancy. Because of the nature of obtaining information via self-report, sample sizes reflected in the tables may not total N. A continuous measure of SES or index of social position was computed. This measure was derived from a variety of factors, including occupation, contribution to household income, and education. A higher score reflects higher SES.

Parenting stress was measured using the Parenting Stress Inventory (PSI).³¹ This 36-item short-form PSI measures stress related to parenting. The mother rates each item on a 5-point Likert scale ranging from "strongly agree" to "strongly disagree." The PSI yields 3 subscales and a summary index of total stress. This was the only summary index used in this study. Higher scores reflect more parenting stress.

Maternal depression was assessed using the Beck Depression Inventory.³² The Beck Depression Inventory is a 21-item self-report scale that measures symptoms of depression. Scale item responses range from zero to 3 and items are summed to obtain an overall score. A higher score reflects more depressive symptoms.

Maternal psychological distress was measured with the Brief Symptom Inventory,³³ a 53-item questionnaire that is the short form of the Revised 90-item Symptom Checklist. The Brief Symptom Inventory yields 9 primary symptom dimensions as well as a global severity index. Higher scores reflect more psychological distress. Only the global score was used.

Infant temperament was assessed with the Infant Behavior Questionnaire.³⁴ The Infant Behavior Questionnaire is a 90-item questionnaire administered to the mother to measure her perceptions of the infant's temperament and behavior in everyday situations. This measure yields 6 summary scales: Activity Level, Smiling and Laughter, Distress to Approach Sudden or Novel Stimuli, Distress to Limitations, Soothability, and Duration of Orienting.

The Child Abuse Potential Inventory³⁵ is a 77-item self-report scale that examines maternal psychological difficulties such as distress, rigidity, and unhappiness, and also interactional problems experienced by the mother, such as problems with self, family, or others. Scores are generated for each of the factors and a total score. Only the total score was used.

Maternal self-esteem was assessed via the Maternal Self-Report Inventory.³⁶ The Maternal Self-Report Inventory is a 26-item scale that yields a total score that reflects the mother's self-esteem. A higher score indicates increased self-esteem.

The Home Observation for Measurement of the Environment (HOME)³⁷ was administered during a home visit. This instrument is useful for measuring the social and emotional support available within the infant's home.³⁷ The 45 scale items are scored on the basis of observations of the home and on interviews with the mother. Only the total score was used.

Mothers brought their children to the hospital clinic for assessment of attachment using the Ainsworth Strange Situation.¹⁸ The Strange Situation is a laboratory procedure consisting of 8 episodes in which the mother and child interact with each other and with a stranger unknown to the infant, and then they are separated from and reunited with each other. The procedure is videotaped and infant attachment classifications are assigned based on coding of the videotape.

Infants are classified as secure (B) or 1 of 2 categories of insecure attachment: insecure-avoidant (A) or insecure-resistant (C).¹⁸ Infants in the B category derive comfort from their mothers as evidenced by seeking proximity to her. At the same time, B infants display a willingness to explore the environment. Infants in the A category may ignore or snub the mother on reunion. Infants in the C category may show a mixture of seeking and resisting contact with the mother. These infants may display much distress and anger regarding the mother's departure. In addition, some infants are classified as disorganized (D).³⁸ Infants in the D category lack a clear attachment strategy and may display odd or disoriented behavior such as freezing, stilling, affective instability, or apprehension of the mother. D is not a mutually exclusive category. Infants may be classified into 1 of the above 3 categories and also as D. Some infants may display a mixture of both avoidant and resistant strategies and are classified as A/C. These infants also receive a D classification. Occasionally, infants may not be able to be classified into one of the above categories and are deemed unclassifiable (U).

Our hypothesis was that infants of adolescent mothers would more often be classified as insecure than secure. Therefore, consistent with previous research, for data analysis to compare secure vs insecure, we combined the number of infants in the A and C categories and compared them with the number of infants in the B category. Infants classified as either A/C or U were omitted from the analyses. Separate analyses that included A/C and U categories (data not shown) were conducted and yielded the same results included in this report. This classification scheme is well established and was implemented by the first author after training and certification by the third author. Interrater reliability of the adolescent sample was evaluated on 20% of the sample with another independent certified coder (S.B.) and was 94%. The older mother sample was coded by either of the 2 certified coders.

We recognized the difficulty of maintaining blindness in this study due to the physical appearance of the adolescent mothers. To help address this issue and strengthen the study findings, we included 11 additional Strange Situation tapes that had been previously scored from another group of mothers aged 18 to 20 years with the tapes of the adolescent sample. These tapes were not identified but were recoded as part of the pool of adolescent tapes and were found to have classifications identical to their original scoring, suggesting that scoring bias may not have occurred.

Power analysis was conducted to determine if a statistically significant (P<.05) difference in the proportion of infants of adolescent vs older mothers classified as secure (B) vs insecure (A or C) could be detected with this sample size. The power was .87, which was adequate.³⁹

Statistical analyses included t tests for continuous data and χ² tests for categorical data to compare demographic, medical (Table 1 and Table 2), and maternal psychological characteristics (Table 3) between the 2 groups. A χ² test was used to compare the distribution of attachment classification categories between the 2 groups (Table 4) and logistic regression to predict attachment classification from maternal age group, sociodemograhic, and maternal psychological characteristics (Table 5).

Maternal demographic characteristics for the adolescent and older mother groups are presented in Table 1. Mothers in the nonadolescent sample were more likely to have more than 1 living child (parity) than mothers in the adolescent sample. Mothers in the older sample were more likely to have a higher SES than the mothers in the younger sample. Most mothers in the adolescent sample had an index of social position, which reflected that they were primarily in the lower and middle to lower socioeconomic groups. Most subjects in the nonadolescent sample fell in the middle to upper-middle class range. Similarly, level of education as measured by number of years of school completed was significantly lower in the adolescent group. A higher proportion of the nonadolescent mothers were married, there were more black women in the nonadolescent group, and more Hispanics in the adolescent group. Alcohol use during pregnancy was more common in the nonadolescent group.

Infant characteristics are presented in Table 2. There were no significant differences between the 2 groups on birth weight, gestational age, Apgar scores, or sex. There were also no differences in proportions of infants with birth weight less than 2500 g or gestational age less than 37 weeks between the 2 groups.

Table 4 presents the distribution of attachment category for the adolescent and adult mothers. In the adolescent sample, 66.7% of the infants were classified as B, 17.7% were classified as A, and 7.8% were classified as C. In the nonadolescent sample, 61.8% of the infants were classified as B, 11.8% were classified as A, and 11.8% were classified as C. One infant in the adolescent sample and 6 infants in the nonadolescent sample were unable to be classified. There were 3 infants in the adolescent group and 5 infants in the nonadolescent group that showed a mixed (A/C) classification. In addition, 10% of the infants in the adolescent group and 19.7% of the infants in the nonadolescent group were rated as D.

There were 13 infants (25.4%) classified as insecure (A or C) in the adolescent group and 18 infants (23.6%) classified as insecure (A or C) in the nonadolescent group. There were no differences in the number of secure vs insecure infants between the 2 groups (χ² = .003, P = .99). Similarly, there were no differences in the number of disorganized infants between the 2 groups (χ² = 2.27, P = .13). Neither of these χ² test results was significant, indicating that there was no difference in the proportion of insecurity or disorganization between the 2 groups. Additional analyses were conducted to determine if attachment classifications differed according to other variables of interest, such as race/ethnicity, sex, birth weight, and gestational age. There were no differences between each of these variables and attachment classification.

Table 3 presents the means and SDs for the maternal self-report and the home observation measures. There were no differences between adolescent and nonadolescent mothers in symptoms of depression (Beck Depression Inventory) or level of psychological distress (Brief Symptom Inventory). However, compared with nonadolescent mothers, adolescent mothers showed more parenting stress (PSI), lower maternal self-esteem (Maternal Self-Report Inventory), a higher child abuse potential (Child Abuse Potential Inventory), a higher activity level in their infant's temperament (Infant Behavior Questionnaire), and provided a less adequate home environment (HOME).

Results of the logistic regression predicting secure vs insecure attachment classification are presented in Table 5. As planned, variables used in the logistic regression, in addition to the adolescent or the nonadolescent group, were those variables that differentiated between these 2 groups, which might "explain" attachment. The logistic regression showed a nonsignificant χ² test (χ² = 5.6, P>.05) and no significant odds ratios for any of the individual predictor variables, suggesting that none of these factors were related to attachment classification.

Three of 5 previously published studies found no increase in secure vs insecure attachment classification using the Ainsworth Strange Situation in infants of adolescent mothers.^19,21,23 One study that found differences used another sample of adolescent mothers as the comparison group.²² The remaining study used the Ainsworth sample¹⁸ for comparison.²⁰ In findings from cross-cultural normative studies, the modal classification is noted as secure (B).⁴⁰ The previously mentioned meta-analysis⁴⁰ identified the Ainsworth sample¹⁸ as the criterion standard. In this sample, approximately 66% of infants were classified as securely attached.¹⁸ This is virtually identical to our finding of 66.7% secure attachment in the adolescent group. Ours is the largest study to date, to our knowledge, with sufficient power to suggest that lack of differences between adolescent and nonadolescent mothers was probably not due to a type II error. Moreover, ours is the only study, of adolescent mothers to deal with the issue of examiner blindness. Our findings suggest that infants of adolescent mothers may resemble normative samples in the prevalence of secure attachments to their mothers.

We did find, consistent with other findings, differences between adolescent and nonadolescent mothers on demographic and parenting factors. Differences on demographic factors, such as SES, marital status, parity, and education are, of course, to be expected. Others have found lower self-esteem,^41,42 higher levels of depression,⁴² higher child abuse potential scores,¹¹ and poorer quality home environments among adolescent mothers.⁵ Differences in how adolescent mothers perceive their infant's temperament have also been reported.¹⁰ Our findings were similar. We found more parenting distress, lower maternal self-esteem, higher child abuse potential scores, and poorer quality home environment in adolescent mothers. In addition, these mothers perceived their infants as having a more active temperament than nonadolescent mothers perceived. Although there is reason to suspect that these factors would contribute to infant attachment classification, we found no such relationships. The combination of adolescent or nonadolescent mother group, sociodemographic, and parenting factors in the logistic regression was unrelated to secure or insecure attachment classification. To our knowledge, this is the first study to determine if parenting factors predict attachment classification in adolescent mothers. It is interesting that factors theoretically shown or assumed to affect the quality of the attachment relationship, such as maternal self esteem, parenting stress, SES, and psychological distress, were unrelated to attachment in this study. These findings support the robustness of the attachment relationship and the results reported herein.

The Ainsworth Strange Situation¹⁸ has been widely used and is acknowledged as the criterion standard method for measuring attachment.⁴⁰ In addition to normative studies, this procedure has been found to differentiate among other areas of study, including abuse and neglect,^43,44 childhood aggression,⁴⁵ clinical depression in mothers,^25,27 and cross-cultural differences in Israel,⁴⁶ Japan,⁴⁷ and Germany.⁴⁸ Therefore, we do not attribute our lack of differences between adolescent and nonadolescent mothers to problems with the procedure. In addition, our coders were well trained, and as mentioned earlier we did attempt to determine if there was bias due to the appearance of the adolescent mothers.

A more likely explanation for our findings is that the attachment system is robust and difficult to impair and may not be compromised by factors such as adolescent parenting or the maternal factors that we studied. This is consistent with attachment theory, in which attachment is viewed as a proximity promoting system important for survival.⁴⁹

We did find important differences between adolescent and nonadolescent mothers. Adolescent mothers showed more parenting distress, lower maternal self-esteem, higher child abuse potential scores, poorer quality home environments, and perceived their infants as being more active. In other studies, similar factors have been shown to affect child outcome.^50,51 The differences found in the adolescent sample may affect aspects of the child (ie, behavior and development) that we did not measure. Therefore, these factors may still affect offspring of adolescent mothers even if they do not affect the attachment relationship.

We did not measure other aspects of adolescent mothers that could have influenced our results, such as social support and the adolescent mother's developmental level and knowledge of child development. It has been suggested⁵¹ that adolescent parenting may not invariably be associated with adversity. The developmental process may affect the mothering experience of adolescents and provide opportunities for intervention.⁵¹

Although the Strange Situation and the HOME were based on direct observation, other measures were based on self-report (stress, depression, temperament, etc). It is possible that study findings would differ with more objective measures of some of these factors.

Although adolescent mothers are often portrayed as disadvantaged and they clearly differ from nonadolescent mothers in potentially important areas of parenting, our findings add to a growing body of literature suggesting that the ability of these mothers to maintain a secure attachment relationship with their infants is similar to that of mothers in the general population. This and other research suggest that adolescent mothers may not be a homogeneous group—that there are important individual differences in the parenting ability of these mothers. If we are going to understand how these mothers parent, how their parenting affects the outcome of their children, and how to develop effective interventions, we will need to pay more attention to patterns of individual differences in parenting.

Accepted for publication August 7, 2001.

This study was supported in part by grant 1U10-HD27904-1 and contract N01HD2-3159 from the Institute of Child Health and Human Development.

Presented as a poster at the Pediatric Academic Society/American Academy of Pediatrics joint meeting, Boston, Mass, May 12, 2000.

Corresponding author and reprints: Lynne Andreozzi, PhD, Infant Development Center, Women and Infants' Hospital, 101 Dudley St, Providence, RI 02905-2499 (e-mail: landreozzi@lifespan.org).

This study contributes to the knowledge of adolescent parenting and outcome of infants of adolescent parents. Our research attempts to supplement a weak literature base regarding the attachment relationship of infants of adolescent mothers. Little is known about this relationship. One of the strengths of this examination is the comparison of the target population with a demographically similar sample, one of older nonadolescent mothers and their infants.

Here, infants of adolescent and nonadolescent mothers show similar patterns of attachment. Adolescent and nonadolescent mothers show substantial differences in parenting characteristics; however, these differences do not seem to impair the attachment relationship with their infants. Therefore, there may be important individual differences in the parenting ability of adolescent mothers.