Intergenerational Transmission of Attachment between Incarcerated Mothers and Their Co-residing Children

Abstract

Objective:

Maternal attachment representations and children’s attachment to their mothers have been shown to be related while this association varies according to socioeconomic risk factors that these mother-child dyads are experiencing. In this regard, this study aims to investigate the intergenerational transmission of attachment between highly disadvantaged incarcerated mothers and their co-residing children and the mediating role of maternal sensitivity in the relationship between maternal and children’s attachment representations.

Method:

The study sample consists of 84 incarcerated mothers and their co-residing children. The maternal attachment was evaluated using the Doll Story Completion Task. Children’s attachment and maternal sensitivity were assessed based on a 45-minute semi-structured mother-child interaction observation and rated through structured coding instruments.

Results:

Hierarchical regression analyses revealed that mothers’ anxious and avoidant attachment positively predicted children’s disorganized attachment behaviors. Moreover, analyses showed that maternal sensitivity did not mediate the relationship between maternal and child attachment representations.

Conclusion:

Findings partially supported the discussion emphasizing that attachment transmission, as well as the mediating role of maternal sensitivity on this transmission, appeared to be weaker for disadvantaged mother-child dyads. Implications for future intervention programs aiming to support the attachment security and reduce the disorganization of children co-residing with their incarcerated mothers are discussed.

INTRODUCTION

The majority of incarcerated women are mothers most of whom are the primary, or even the only, caregivers of their young children prior to incarceration (Bruns 2006, Greene et al. 2000, Saruç 2014). Therefore, the negative implications of maternal incarceration have been suggested to be more detrimental to children’s adjustment compared to paternal incarceration (Dallaire et al. 2015, Shlafer & Poehlmann 2010). For example, Dallaire and Wilson (2010) revealed that children of incarcerated mothers, compared to children of incarcerated fathers, experience more anxiety and depression and exhibit rule-breaking behaviors more frequently. Attachment theory emphasizes that early relationship disruptions between the child and the primary caregiver, usually the mother, may lead to children’s insecure attachment to them (Bowlby 1973). In line with this, Poehlmann’s (2005) study revealed that most of the children (63%) separated from their inmate mothers attached insecurely both to their mothers and their current caregivers.

Attachment security is suggested to support children’s future close relationships and a more positive self-concept, as well as making them more likely to regulate their emotions and cope with stressful circumstances (for a review see, Thompson 2016). Secure attachment is closely related to resilience, especially in socioeconomically disadvantaged children who experience risk factors such as poverty, parental mental health problems, and/or inadequate caregiving, and linked with life-span positive developmental outcomes (Belsky & Fearon 2002). Therefore, in order to protect children’s attachment security, eligible incarcerated mothers are allowed to care for their minor children at correctional facilities in a number of countries, including Turkey (Alejos 2005).

Studies have indicated that most incarcerated mothers live in poverty, experience mental health issues, abuse substances, and tend to have low educational attainment (Bruns 2006, Dallaire et al. 2015, Glaze & Maruschak 2008). Moreover, Borelli and her colleagues (2010) revealed that most of the incarcerated mothers co-residing with their children had insecure attachment representations (13% insecure-preoccupied; 20% insecure-dismissive; 30% insecure-unresolved). The aforementioned parental risk factors (for a meta-analysis see, Cyr et al. 2010) and parental insecure attachment representations (for meta-analyses see, Van IJzendoorn 1995; Verhage et al. 2016, 2018) have been found to decrease their sensitivity to their children’s attachment signals and, therefore, lead to their children’s insecure attachment. In line with these findings, it can be suggested that not only children separated from their inmate mothers, but also children co-residing with their incarcerated mothers may be at risk to develop an insecure attachment to their mothers. However, to the best of our knowledge, there have been only two studies investigating co-residing children’s attachment to their inmate mothers (Byrne et al. 2010, Cassidy et al. 2010).

Considering the lifespan implications of attachment security, examining the predictors of especially disadvantaged children’s attachment security is an important developmental inquiry that needs to be addressed. A meta-analysis by Bakermans-Kranenburg and her colleagues (2004) revealed that maternal factors showed stronger relations with the child’s attachment compared to child-related factors. It has been revealed that parents’ own attachment representations shape their sensitive caregiving practices, and in turn, perpetuate the attachment transmission across generations (Bailey et al. 2017, Tarabulsy et al. 2005). Even though a number of countries allow incarcerated mothers to care for their children within the correctional facilities, there is a scarcity of research examining the factors explaining the attachment security of children co-residing with their incarcerated mothers. Therefore, the main aim of the current study is to investigate the attachment transmission between incarcerated mothers and their co-residing children. Secondly, this study aims to examine the mediating role of maternal sensitivity in this transmission.

METHOD

Participants

The study sample consists of 84 incarcerated mothers who range in age between 20 and 43 (M=29.9, SD=5.6) and their 11- to 43-month-old (M=25.3, SD=8.3) children who co-resided with their mothers. A large percentage of the mothers (42.2%) were illiterate. Some mothers were literate but had not graduated from school (21.7%); they either obtained an education for a few years but could not graduate, or they became literate without attending school. The rest of them were primary school (14.5%), middle school (12.0%), high school (7.2%), and university (2.4%) graduates. More than half of the mothers (53.6%) had never worked outside the home; of those were employed before incarceration, they were generally working in temporary and unskilled jobs. The mothers and their co-residing children generally came to the prison facilities together, and they spent an average of 10 months (SD=9.8, range=1-42) in prison prior to data collection. Most of the mothers (60.2%) were recurring offenders. The participant mothers were incarcerated for economically motivated (e.g., burglary; 57.8%), drug-related (33.7%), or violent (e.g., homicide; 8.5%) crimes.

Measures

Maternal attachment. The Doll Story Completion Task (Bretherton et al. 1990, Granot & Mayseless 2001) was used to elicit maternal secure base script narratives. Mothers were asked to think back to when they were six years old and then five story-stems with attachment-related distress (e.g., separation, reunion, a monster under the bed) were presented to them to obtain their narratives regarding their caregiver’s responses to their distress as well as their own behaviors, feelings, and thoughts under these hypothetical stressful situations. The interviewer opened each story using a standard story stem and then asked the mother to freely finish the story (e.g. ‘what happens next.’). The two coders (the first author of the study and a psychologist experienced in rating this measure who was blind to the study objectives) rated the narratives using a structured rating manual (Uluç 2007) on a 5-point Likert scale (1=not defining to 5=absolutely defining) regarding the mothers secure, anxious, and avoidant attachment representations. The ICC between the coders ranged from 0.77 to 0.99, and the average ICC was 0.89. The coding of the two observers averaged to create a composite score of maternal attachment. Higher scores reflect the mothers’ increased level of correspondence to the relevant attachment state of mind.

Maternal sensitivity. The maternal sensitivity was evaluated based on a 45-minute semi-structured mother-child interaction observation (see, Berument et al. 2018). The interaction procedure included tasks such as free-play, separation-reunion, the child playing with an observer, the mother-child dyad solving a slightly challenging puzzle, and the child playing with a robot and a clown (as a mild stress-provoking situation). The mother-child interaction observation was coded through the Maternal Sensitivity Scale (MSS; Berument et al. 2018). The scale provides an overall summary index of the quality of care primarily based on three domains: maternal support/encouragement, responsiveness, and warmth (e.g. ‘Even when doing other activities, such as answering the questions of the researcher, responds to all of the child’s signals with and without stress.’). The MSS consists of 29 items and the items were coded on a 3-point Likert scale (1=not defining, 2=somewhat defining, 3=absolutely definitive, NA=not applicable). Because voice and video recordings in prison facilities are not allowed, two coders (the first author of the study and a developmental psychology doctoral student who was blind to the study objectives), located in different corners of the room, observed the interactions, took notes about their observations and coded them immediately after the observation. The Intraclass Correlations (ICC) between raters ranged from 0.70 to 0.97 and the average ICC was 0.86. The ratings of two observers averaged to create a composite score of maternal sensitivity. Factor analysis indicated a one-factor solution and the Cronbach Alpha internal consistency score of the scale was 0.91, with higher scores indicating higher maternal sensitivity.

Children’s attachment. The children’s attachment was also evaluated based on the same mother-child interactions. The observations were coded using the Turkish Toddler Attachment Sort 60 (TTAS-60; Berument & Sümer 2017, Kirkland et al. 2004). TTAS-60 consists of 60 items (e.g. ‘Fusses, cries, becomes angry if mother’s responses are not immediate.’). First, trained observers sort 60 items based on the target child’s behavior into three piles (i.e. less applicable, unsure/unobserved, and more applicable). Then, the items sorted into the unsure/unobserved pile are eliminated from the coding, and the ‘less’ and ‘more’ piles are sorted again into two piles based on how applicable they are to the child’s behaviors. The items are sorted through a special website. Based on the sorting on the website, scores for four attachment patterns (i.e., anxious, secure, avoidant, and disorganized) are assigned to each child. Because it was not allowed to bring any technological devices into the prison facilities and the prison authorities provided only one computer, the two coders rated the TTAS-60 items based on consensus. Higher scores reflect an increased level of correspondence to the relevant attachment pattern.

Procedure

Before data collection, written approval was obtained from the institutional review board and the Turkish Ministry of Justice. Data was collected from eight different prison facilities located in seven cities varying in size from four distinct geographical regions of Turkey.

RESULTS

First, participants residing in high-security (77.4%) and low-security prison facilities were compared in terms of the study variables. Bonferroni correction was used to adjust the significance values for multiple testing (k=9; p=0.05/9=0.006). The results revealed no significant differences between the two groups with regard to the study variables (p>0.006).

The descriptive statistics and zero-order bivariate correlations for the study variables are presented in Table 1. The correlations showed that maternal secure attachment was negatively related to their anxious (r=-0.70, p<0.01) and avoidant (r=-0.53, p<0.01) attachment. Moreover, maternal anxious and avoidant attachment was positively (r=0.34, p<0.01) related. The correlations among maternal and child’s attachment indicated that child’s disorganized attachment has revealed a negative association with maternal secure attachment (r=-0.22, p<0.05), a positive association with maternal anxious (r=0.31, p<0.01), and a marginal positive association with maternal avoidant (r=0.20, p=0.07) attachment. Maternal sensitivity positively related to children’s secure attachment (r=0.36, p<0.01) and negatively associated with disorganized attachment (r=-0.40, p<0.01). Children’s avoidant attachment was positively associated with their secure (r=0.23, p<0.05) and disorganized (r=0.60, p<0.01) attachment. Children’s anxious attachment was negatively related to their avoidant (r=-0.73, p<0.01), secure (r=-0.56, p<0.01), and disorganized (r=-0.24, p<0.05) attachment. Children’s disorganized attachment negatively correlated with their secure attachment (r=0.42, p<0.01). Finally, our results revealed that the number of years that mothers received a formal education was correlated positively with child’s avoidant (r=0.23, p<0.05) and disorganized (r=0.33, p<0.01) attachment, therefore, maternal educational attainment was controlled in further analyses.

Table 1.

Descriptive Statistics and Zero-Order Bivariate Correlations for Study Variables

	M (SD)	1	2	3	4	5	6	7	8	9	10
Mothers’
1. Secure attachment	3.15 (0.67)	-
2. Anxious attachment	2.64 (0.99)	-0.70**	-
3. Avoidant attachment	1.59 (0.83)	-0.53**	0.34**	-
4. Sensitivity	1.52 (0.37)	0.14	-0.08	-0.10	-
5. Educational attainment	3.30 (4.05)	-0.02	0.19	-0.04	0.08	-
Children’s
6. Avoidant attachment	4.62 (1.83)	-0.19	0.11	0.08	-0.21	0.23*	-
7. Secure attachment	2.85 (0.97)	0.00	-0.14	-0.05	0.36**	-0.15	0.23*	-
8. Anxious attachment	4.10 (2.22)	-0.03	0.06	-0.00	-0.08	-0.22	-0.73**	-0.56**	-
9. Disorganized attachment	-0.53 (0.18)	-0.22*	0.31**	0.20c	-0.40**	0.33**	0.60**	-0.42**	-0.24*	-
10. Age (months)	25.25 (8.32)	-0.15	0.16	0.06	-0.09	-0.13	0.20	0.10	-0.21a	0.21b	-

Disorganized attachment scores are generated using a different norm than the other three organized attachment scores. Therefore, it cannot be compared with the other three attachment scores based on mean values and ranges.

^cp=0.07,

^bp=0.06,

^ap=0.05,

^*p<0.05,

^**p<0.01.

Intergenerational Transmission of Attachment

Separate hierarchical regression analyses were conducted to test the predicting role of maternal attachment representations on children’s attachment (see, Table 2). Maternal educational attainment was included in the first step as the control, and maternal attachment representation was included in the second step as the independent variable. In all analyses, maternal educational attainment (β=0.35, p<0.01) was included in the first step and positively predicted child’s disorganized attachment (R²=0.12, F(1,77)=10.39, p<0.01). The first analysis revealed that, after controlling maternal educational attainment, maternal secure attachment (β=-0.22, p<0.05) negatively predicted children’s disorganized attachment (∆R²=0.05, F(2,76)=7.56, p<0.01). In the second analysis, after controlling maternal educational attainment, maternal anxious attachment (β=0.26, p<0.05) positively predicted children’s disorganized attachment (∆R²=0.06, F(2, 76)=8.42, p<0.01). The last significant model showed that, after controlling maternal educational attainment, maternal avoidant attachment (β=0.22, p<0.05) positively predicted children’s disorganized attachment (∆R²=0.17, F(2,76)=7.54, p<0.01).

Table 2.

The Role of Maternal Anxious and Avoidant Attachment on Children’s Disorganized Attachment (n = 79)

Predictors	R2	F	B	SE	β	t
Step 1: Covariates
Maternal education	0.19	9.00***	0.02	0.00	0.38	3.64***
Child’s age	0.01	0.00	0.27	2.61*
Step 2: Independent variables
Maternal anxious attachment	0.23	7.57***	0.04	0.02	0.21	2.00*
Step 1: Covariates
Maternal education	0.19	9.00***	0.02	0.00	0.38	3.64***
Child’s age	0.01	0.00	0.27	2.61*
Step 2: Independent variables
Maternal avoidant attachment	0.23	7.56***	0.04	0.02	0.20	2.55*

^*p < 0.05, **p < 0.01,

^***p < 0.001.

The Mediating Role of Maternal Sensitivity

The second aim of the study is to examine the mediating role of maternal sensitivity in the relation between maternal and child attachment. According to Baron and Kenny (1986), a variable must be predicted by the independent variable to be considered as a mediator variable. However, recent research suggested that the significance of the indirect effect is the only required prerequisite for a mediation analysis (Meule 2019, Zhao et al. 2009). In line with this, even though the independent variables (maternal attachment) did not significantly predict the mediator, mediation models were tested in the current study. The mediation analyses were performed using Model 4 of PROCESS macro for SPSS (version 3.5; Hayes 2018). The results revealed that the mediating role of maternal sensitivity on the relationship between maternal and child attachment was not significant (p>0.05).

DISCUSSION

The primary purpose of the present study was to assess the predicting role of maternal attachment representations on co-residing children’s attachment. The second aim of this study was to investigate the mediating role of maternal sensitive caregiving on attachment transmission. Before discussing our main findings, it is important to address that maternal educational attainment was positively related to children’s avoidant and disorganized attachment in the current study. Additional analysis indicated that maternal educational attainment was also negatively related to maternal repetitive conviction (r=-0.27, p<0.05). Repetitively incarcerated mothers (mostly convicted for minor offenses such as theft) would be more familiar with the correctional system than the less frequently convicted mothers. In other words, the prison environment and correctional system may be stranger and stress-inducing for mothers with higher education. Therefore, educated mothers’ psychological functioning may be more deteriorated during their time spent in prison, and in turn, the quality of their parenting practices with their co-residing children may be damaged. Even though the relationship between maternal education and sensitivity is not significant in the present study, it has been shown that maternal sensitivity is not the only parenting dimension that plays an important role in children’s attachment security (for a meta-analysis see, De Wolff & Van IJzendoorn 1997). For example, Bernier and colleagues (2014) found that the relationship between maternal autonomy support and child attachment is stronger compared to the association between maternal sensitivity and child attachment. Consequently, future research examining other dimensions of inmate mothers’ parenting may explain the underlying factors in the relationship between maternal educational attainment and their co-residing children’s attachment.

Results showed that maternal secure attachment negatively predicted children’s disorganized attachment, whereas maternal anxious and avoidant attachment positively predicted it. These findings supported the first hypothesis of the current study and are in line with the findings of the previous studies showing a robust empirical link between maternal and children’s attachment (Bailey et al. 2017, Slade et al. 2005, Tarabulsy et al. 2005, for a meta-analysis see, Van IJzendoorn 1995). The majority of inmate mothers who have a co-residing child have insecure (65%) attachment representations, and a significant part (30%) of those insecure mothers have additionally been classified as insecure-unresolved (Borelli et al. 2010). Insecure, and especially unresolved, parents tend to exhibit anomalous, strange, conflictual, frightened, inexplicable, and threatening parenting behaviors (Duschinsky 2018), which in turn, lead to children’s disorganized attachment (for a meta-analysis see, Madigan et al. 2006). Main and Hesse (1990) proposed that these frightened/frightening, strange, and/or overly insensitive parenting behaviors may prevent the children’s attachment representation from being categorized into one of the three organized attachment patterns (i.e., secure, anxious, and avoidant). The maternal attachment measure used in the current study is not capable of assessing the unresolved attachment pattern. Therefore, future studies examining mothers’ unresolved attachment could improve our understanding of the attachment transmission between incarcerated mothers and their children. Disorganized children are at the highest risk of exhibiting disruptive/aggressive and dissociative problems in later years compared to children classified into the three organized attachment representations (for meta-analyses see, Fearon et al. 2010, Groh et al. 2017, Van IJzendoorn et al. 1999). Therefore, further investigation enlightening the antecedents of children’s disorganization is crucial to planning future intervention studies aiming to improve the attachment security of these disadvantaged children.

Analyses failed to provide any support for the mediating role of maternal sensitivity on the relationship between maternal and child attachment. Van IJzendoorn and Bakermans-Kranenburg (2019) suggested that in disadvantaged families, intergenerational transmission of attachment may weaken because of moderators blocking this transmission. They point out that life stresses may be one of the important factors interfering with the relationship between the sensitive caregiving capacity of securely attached parents and their actual parenting practices. In this respect, even if the inmate mothers have secure attachment representations and the potential to provide sensitive care for their children, they may not be sensitively caregiving to their co-residing infants because of the cumulative adversities of the prison environment such as stress from the continuing prosecution process, feelings of self-blame as a result of bringing their children into prison, and conflictual interpersonal relations with other inmates. In these stressful circumstances, the power of parental attachment representations seems to be weaker in predicting their sensitive caregiving, and other mechanisms take over the mediating role of attachment transmission. In line with this discussion, studies suggested that compared to maternal sensitivity, maternal autonomy support (Bernier et al. 2014) and mentalization capacity (Slade et al. 2005) may be stronger mediators on the association between maternal and child attachment. In addition to these two factors, future research should also investigate other parental factors such as -setting limits, warmth, and protection as potential mediators in the attachment transmission between incarcerated mothers and their co-residing children to explain the transmission gap (Van IJzendoorn 1995, Van IJzendoorn & Bakermans-Kranenburg 2019). Considering the attachment transmission gap is broader for mother-child dyads experiencing more significant life stressors (for meta-analyses see, Verhage et al. 2016, 2018), examining the underlying factors of attachment transmission between disadvantaged parents and their children may be especially important.

The current study has some limitations that should be addressed. Firstly, because voice or video recordings are not allowed in prison facilities, mother-child interactions were coded by the observers who were aware of sample characteristics. Secondly, because the TTAS-60 is coded using a special web site, and prison authorities provided only one computer, children’s attachment security was coded based on consensus. Despite these limitations, both raters were highly trained, and the items were created at a behavioral level which requires objective ratings independent from the coders’ bias. Nevertheless, considering the highly restrictive nature of the penal institutions, evaluating maternal sensitivity and children’s attachment security through mother-child interaction observations is considered to be a key strength of the current study. In spite of these limitations, using intergenerational data collected through various methods (e.g., mother-child interaction observation, story-stem technique), the current study provides a comprehensive examination on the antecedents of co-residing children’s attachment. Although the measures in the area of attachment research are highly labor-intensive, which prevents investigators from recruiting large samples, the current study reached one of the largest sample sizes among the studies conducted with this group of mother-child dyads.

In sum, our findings revealed that maternal secure attachment negatively predicted, whereas maternal anxious and avoidant attachment positively predicted, children’s disorganized attachment. Considering all three maternal attachment representations predicted only child’s disorganized attachment among the four child attachment patterns, and children’s disorganization poses a greater risk for later negative developmental outcomes, it is crucial that insecurely attached mothers receive extra support in order to prevent this intergenerational cycle of insecure attachment. Moreover, maternal sensitive caregiving failed to mediate the relationships between maternal and children’s attachment. Thus, future research needs to focus on other potential mediators in the transmission of attachment especially for the under-risk parent-child dyads. Future intervention efforts may focus on ameliorating positive parenting practices of inmate mothers other than sensitivity to support their children’s attachment security.