Maternal caregiving representations of the infant in the first year of life: Associations with prenatal and concurrent reflective functioning

Abstract

Few studies have examined whether maternal caregiving representations are associated with maternal reflective functioning (MRF), especially when MRF is evaluated longitudinally beginning in pregnancy. This study addresses this gap by evaluating whether prenatal and postnatal MRF are associated with mothers’ caregiving representations assessed at 7 months postpartum, and by exploring theoretically unexpected MRF scores in each of the representational categories. Forty-seven mothers were recruited during their last trimester of pregnancy from an obstetrics clinic at a university hospital located in a large mid-western city in the United States. During pregnancy, mothers completed the Pregnancy Interview, and at 7 months postpartum they completed the Parent Development Interview (PDI) and the Working Model of the Child Interview. Results indicate that higher prenatal and postnatal MRF increased the odds of being classified as balanced versus disengaged. At 7 months, MRF also increased the odds of being balanced vs. distorted. Ten mothers who were classified as balanced or distorted had unexpected prenatal MRF scores, and six mothers had unexpected MRF scores when representations were assessed concurrently. Mothers classified as balanced with low MRF scores tended to have a low level of education, whereas mothers classified as distorted with high MRF scores had responses that were hostile, helpless, and role-reversed.

1 |. INTRODUCTION

During pregnancy, women undergo a profound developmental transition as their maternal identities emerge. As part of this process, they begin to form caregiving representations of themselves as caregivers and of their infants (Slade et al., 2009; Solomon & George, 1996), which tend to be stable into the infant’s first year of life (Benoit et al., 1997; Theran et al., 2005; Vreeswijk et al., 2015). Mothers’ caregiving representations are linked to their parenting behavior (Dayton et al., 2010; Theran et al., 2005) and to their infants’ later attachment patterns (Benoit et al., 1997; Huth-Bocks et al., 2011; Madigan et al., 2015). In turn, both parenting and infant attachment are associated with children’s developmental outcomes and well-being into adulthood (Raby et al., 2015). As such, it is important to understand other parenting capacities that either predict or are correlated with caregiving representations.

One correlate of caregiving representations that needs further study is maternal reflective functioning (MRF), the ability to interpret behaviors, bearing in mind the thoughts, feelings and beliefs that underlie them (Slade, 2005). Theoretically, mothers classified into an organized/secure typology of caregiving representations should have higher MRF than those classified into an organized/insecure typology. However, our understanding of the relationship between caregiving representations and MRF is limited to two studies (Alvarez-Monjarás et al., 2019; Schechter et al, 2005). Both of these studies used clinical samples, with a limited range of MRF, and only one (Schechter et al., 2005) examined MRF in the context of representational categories. It is important to assess the associations between MRF and caregiving representations in non-clinical samples with a range of demographic risk factors. Disentangling risks from clinical status allows for a better understanding of risk, MRF and caregiving representations. Additionally, given that MRF can be measured reliably during the prenatal period (Pajulo et al., 2012; Smaling et al., 2017), and can increase over time (Poznansky, 2010; Wong, 2016), assessing this association might be especially informative, as mothers’ caregiving representations are beginning to emerge and solidify during this time period.

1.1 |. Attachment theory and maternal caregiving representations

Bowlby postulated that parents inherit a caregiving system, a biologically driven behavioral system that motivates them to protect, care, and comfort their children when their children become distressed (Bowlby, 1969, 1982). The caregiving system is guided by caregiving rep resentations, mental models of caregiving that include ideas and feelings about who the infant is, who the caregiver is, and the relationship between the infant and caregiver (Solomon & George, 1996). George and Solomon (1996) suggest that mothers’ caregiving representations are related to their own childhood attachment experiences, which may become more salient during their transition to motherhood, including pregnancy and birth (Solomon & George, 1996).

Caregiving representations refer to mothers’ attachment representations of their children and include: (1) Balanced representations, which are analogous to child secure attachment in the Strange Situation Procedure (SSP) (Benoit et al., 1997; Huth-Bocks et al., 2011; Madigan et al., 2015); (2) Distorted representations, which are analogous to child ambivalent/resistant attachment in the SSP; and (3) Disengaged representations, which are analogous to avoidant attachment in the SSP (Benoit et al., 1997; Huth-Bocks et al., 2011). Further descriptions of the three categories of caregiving representations are provided in the measures section.

1.2 |. Maternal caregiving representations: Stability and associations with parenting and child attachment

Prior studies suggest that mothers’ prenatal caregiving representations are stable across the first year of their infant’s life (Benoit et al., 1997; Theran et al., 2005; Vreeswijk et al., 2015), and predict both parenting quality and child attachment. For example, in a study by Alvarez-Monjarás et al. (2019), higher coherence, a hallmark of balanced representations, was associated with higher levels of observed maternal sensitivity. Similarly, in a study by Dayton et al. (2010), mothers with balanced caregiving representations demonstrated significantly higher levels of positive parenting, compared to mothers with non-balanced representations (i.e., distorted or disengaged). Further, Sokolowski et al. (2007) found that mothers classified as disengaged were less sensitive, more withdrawn, and used less guidance with their infants during interaction, compared to mothers with balanced or distorted representations.

Given the robust associations between caregiver-child interactions and infant attachment (De Wolff & van IJzendoorn, 1997; Verhage et al., 2016), it is not surprising that caregiving representations are also significant predictors of infant attachment patterns. Three studies have demonstrated a concordance between prenatal caregiving representations and infants’ attachment classifications during the first year of life (Benoit et al., 1997; Huth-Bocks et al., 2011; Madigan et al., 2015). The results of these studies suggest that the concordance between the caregiving representations and child attachment may be stronger among balanced/secure pairs than among non-balanced/insecure pairs, and also among dyads who have not been exposed to intimate partner violence.

1.3 |. Maternal reflective functioning, parenting, and attachment

As is the case for caregiving representations, MRF is also associated with parenting sensitivity and child attachment (Alvarez-Monjarás et al., 2019; Grienenberger et al., 2005; Slade, 2005; Stacks et al., 2014). For instance, during social interactions, mothers with higher MRF are more sensitive and less intrusive, anxious, and rejecting toward their children, and exhibit fewer conflicting affective cues (Alvarez-Monjarás et al., 2019; Ensink et al., 2019; Grienenberger et al., 2005; Smaling et al., 2017). Some studies also show that the relationship between MRF and infant attachment is mediated by parenting behavior (Ensink et al., 2019; Stacks et al., 2014).

There is a need to further study the associations between MRF and caregiving representations because no study has assessed the relationship between these variables when MRF is measured in pregnancy. Further, only two studies to date have evaluated both caregiving representations on the WMCI and MRF, and both of these were based on clinical samples (Alvarez-Monjarás et al., 2019; Schechter et al., 2005). The results of these studies show that mothers with balanced caregiving representations have higher MRF, and both MRF and caregiving representations are linked to sensitive parenting. Schechter et al. (2005) examined MRF scores across the three categories of maternal caregiving representations. Consistent with theoretical assumptions, there were statistically significant differences in MRF across the WMCI categories, with mothers classified as balanced on the WMCI demonstrating the highest MRF, followed by mothers classified as distorted and disengaged. That study, however, is limited by a small sample of women with a history of interpersonal violence and a low level of education.

While only two studies have examined caregiving representations and MRF, others have assessed MRF in the context of child attachment. Slade et al. (2005) found that MRF was highest among children classified as secure, and that mothers whose children were classified as ambivalent or disorganized scored significantly lower on MRF. In contrast, Stacks et al. (2014) reported that mothers of avoidant or disorganized children scored lower on MRF than mothers of children classified as secure. However, MRF scores did not differentiate children classified as ambivalent and children classified as secure. Ensink et al. (2016, 2019) show similar findings although they did not statistically test differences in MRF scores across attachment categories.

In studies of MRF and adult attachment states of mind, Slade et al. (2005) reported that adults classified as insecure and unresolved had significantly lower MRF scores. However, in a larger study assessing adult reflective functioning and attachment states of mind, Jessee et al. (2016) found no differences in adult reflective functioning among those classified as preoccupied or autonomous (see Table 1).

TABLE 1.

Review of previous studies: MRF across caregiving representations, child attachment and adult attachment

Parental reflective functioning (PDI) scores across strange situation attachment classifications (SSP)
	Secure		Avoidant		Ambivalent		Disorganized		Sample description
	N	M(SD)	N	M(SD)	N	M(SD)	N	M(SD)
Slade et al. (2005)	22	5.64 (1.14)	5	5.40 (1.36)	3	3.00 (.00)a	10	4.3 (1.57)a	40 mothers; 94% white; highly educated middle-class sample
Stacks et al. (2014)	53	4.81 (1.42)	11	3.83 (1.27)a	7	4.86 (1.86)	12	3.83 (1.12)a	83 mothers; 73.4% white; 66.8% college graduates; 70% histories of abuse
	Adult reflective functioning (AAI) scores across strange situation attachment classification (SSP)
	Secure		Avoidant		Ambivalent		Disorganized		Sample description
	N	M(SD)	N	M(SD)	N	M(SD)	N	M(SD)
Ensink et al. (2016) Did not test statistical differences	48	4.62 (1.16)	16	4.71 (.99)	10	4.60 (1.51)	16	3.94 (1.34)	88 white mothers; 86% college or university degree; 30% histories of abuse
Parental reflective functioning scores (PDI) across maternal caregiving representations (WMCI)
	Balanced		Disengaged		Distorted		Disrupted		Sample Description
	N	M(SD)	N	M(SD)	N	M(SD)	N	M(SD)
Schechter et al. (2005)	7	4.43 (.79)	10	1.85 (1.06)a	24	3.50 (1.06)a	NA		41 mothers in treatment; All witnessed/experienced interpersonal violence; 54% less than high school
	Parental reflective functioning scores (PDI) across adult attachment states of mind (AAI)
	Autonomous		Dismissing		Preoccupied		Unresolved		Sample description
	N	M(SD)	N	M(SD)	N	M(SD)	N	M(SD)
Slade et al. (2005)	23	5.74 (1.51)	6	4.33 (1.51)a	8	4.63 (1.19)1	3	2.67 (.58)a	40 mothers; 94% white; highly educated middle-class sample
	Adult reflective functioning (AAI) across adult attachment states of mind (AAI)
	Autonomous		Dismissing		Preoccupied		Unresolved		Sample description
	N	M(SD)	N	M(SD)	N	M(SD)	N	M(SD)
Jessee et al. (2016)	133	4.73 (1.32)	50	3.51 (1.27)a	11	3.91 (1.92)	NR		103 couples; 80% white; more than 60% college educated

Abbreviations: PDI, parent development interview; AAI, adult attachment interview; WMCI, working model of the child interview; NR, not reported; NA, not assessed.

^aRF score is significantly different from secure/autonomous/balanced.

Although authors do not always report the range of RF scores in each attachment category, the standard deviations suggests that some RF scores in the secure (or analogous) category may be in the low range, while some RF scores in the ambivalent (or analogous) category may be in the moderate/high range (see Methods section for a description of the range of RF scores). Together, these findings suggest a need to further assess the relationship between RF and attachment categories.

1.4 |. The current study

The primary goal of the current study is to examine the associations among mothers’ caregiving representations of the child at 7 months postpartum, and their prenatal and postnatal MRF. Two specific questions were evaluated to address this goal:

1. Is prenatal MRF associated with mothers’ caregiving representations at 7 months postpartum?

2. Is MRF at 7 months postpartum associated with mothers’ concurrent caregiving representations?

For these questions, we hypothesized that MRF would be significantly associated with mothers’ caregiving representations at 7 months postpartum. Specifically, we expected to find that:

1. Balanced representations at 7 months postpartum would be associated with higher MRF at both time points;

2. Non-balanced representations at 7 months postpartum would be associated with lower MRF at both time points; and

3. Mothers with disengaged representations would have lower MRF scores than those with distorted representations.

A secondary goal of the study was to explore theoretically out-of-bounds MRF scores in each of the representational categories; for example, mothers classified as balanced whose MRF scores are considered negative or low (scores ≤3), and mothers classified as disengaged or distorted whose MRF scores were in the moderate to high range (scores >4), and their correlates.

2 |. METHODS

2.1 |. Procedures

Analyses in the current study were based on data collected in a longitudinal prospective study, Parenting Representations during Pre- and Postnatal Periods linked to Early Outcomes (PuRPLE). The larger goal of the study was to understand how maternal prenatal and postnatal representations of caregiving influence parenting and infant behavior and development. Mothers in this study were recruited from a larger sample of mothers participating in the Perinatal Imaging of Neural Connectivity (PINC) project, a longitudinal study that recruited pregnant women from an obstetrics clinic at a university hospital, that primarily serves low-income women, in a large midwestern city, in the United States. The PINC study specifically sought to examine the links between fetal brain development, environmental and biological factors, and children’s developmental and behavioral outcomes. Participants enrolled in the PINC study were invited to participate in the PuRPLE study, either by phone or during their first study visit. Consenting mothers were then asked to participate in the following three follow-up visits: (1) during the third trimester of pregnancy to assess MRF; (2) within 1 month of the infant’s birth to assess infants’ neurobehavior; and (3) when the infants were 7 months old, to assess both MRF and caregiving representations, At the 7-month visit, mothers also completed self-report questionnaires, dyads were videotaped during a parent-child interaction task, and infants’ developmental skills were assessed.

2.2 |. Participants

Sixty-eight mothers consented to participate in the PuRPLE study. Of those, 21 were lost to follow-up and did not attend the 7-month follow up visit. Data from mothers who participated in both the prenatal and 7-month study visits (n = 47) were used for the current study. Demographic characteristics for the study participants are presented in Table 2. The sample is unique in that it was not clinically referred but drawn from a population with high levels of sociodemographic risk. In many prior studies, demographic risk is confounded with clinical status.

TABLE 2.

Maternal demographic characteristics

Demographic variables	Mean	SD
Age (years)	25.74	5.09
	Frequency	Percent
Marital status*
Single	24	51.1
Married/Partnered	21	44.7
Race/Ethnicity*
White	6	12.8
Black	37	78.7
Asian or pacific islander	1	2.1
Other	2	4.3
Education*
Less than high school diploma	8	17.0
GED/High school diploma	13	27.7
Some college	21	44.7
≥4-year college degree	4	8.6
Income*
Less than $10,000	16	34.0
$10,000 – $19,999	10	21.3
$20,000 – $29,999	7	14.9
$30,000 – $39,999	2	4.3
$40,000 – $49,999	1	2.1
Above $50,000	4	8.6

^*Some numbers do not add up to 47 due to missing data.

2.3 |. Measures

2.3.1 |. Maternal caregiving representations

Mothers’ caregiving representations were assessed at the 7-month visit using the Working Model of the Child Interview (WMCI; Zeanah et al., 1996). The WMCI is a semi-structured interview that asks mothers to respond to questions about her infant’s personality and development, her relationship with her infant, her reactions to her infant’s behavior and distress, and her difficulties with child rearing. The interview was audio-recorded, transcribed verbatim, and coded. Following Zeanah et al. (1996) and Rosenblum et al. (2006), each qualitative feature of mothers’ narratives was coded on a five-point scale with scores ranging from one (never) to five (extreme).

Mothers’ scores were then used to classify their caregiving representations into one of the following three categories: (1) Balanced representations provide evidence for high-quality parent-child relationships through a narrative that is characterized as coherent, well-organized, and providing a rich description of the infant’s personality and the mother-infant relationship. Balanced mothers are able to integrate positive and negative affect and understand their infant’s challenging behavior; (2) Disengaged representations are characterized by several features, including minimal descriptions of the infant, a sense of emotional distancing (excessive intellectual descriptions), indifference or aversion toward the infant, and/or emotional coolness. Disengaged mothers also tend not to acknowledge the infants’ subjective experiences or convey genuine interest in or involvement with the infant; (3) Distorted representations include strong feelings and highly detailed descriptions of the infant and the mother-infant relationship. Distorted mothers have a lot to say, yet their descriptions are inconsistent and disorganized. In their descriptions, they may seem insensitive to their infant’s needs, present a bizarre description of their infant, have unrealistic expectations for their infant’s development, or be unable to understand their infant’s needs. These mothers seem confused about their caregiving roles and appear overwhelmed by their infants’ difficulties and the responsibilities involved in parenting.

The first author of this study acted as the primary coder. To evaluate intercoder reliability, 26 WMCI transcripts (55%) were double-coded by the study’s fourth author. Both coders were trained to reliability on a gold standard set of transcripts prior to coding for this study. Intercoder reliability was tested using Cohen’s Kappa. Results (k = .78, p < .001), indicate good intercoder reliability. Disagreements were conferenced to derive consensus codes, which were used in the study analyses.

2.3.2 |. Maternal reflective functioning

MRF was assessed during the last trimester of pregnancy and again at 7 months postpartum. Interviews were audio-recorded and transcribed verbatim for coding. During pregnancy, MRF was assessed with the Pregnancy Interview-Revised (PI; Slade, 2007), a semi-structured interview consisting of 22 questions that elicit a woman’s emotional experience of pregnancy, representation of herself as a mother, and her relationship with the father of the baby and the unborn child. At 7 months postpartum, MRF was assessed using the Parent Development Interview – Revised Short Form (PDI; Slade, Aber, et al., 2004). The PDI is a 30-item semi-structured interview that assesses a mother’s representation of herself as a parent, her child, the parent-child relationship. Although the PI and the PDI both include questions that elicit a mother’s experiences, thoughts, and feelings about her relationship with her child, the questions on each interview are different.

Both the PI and PDI are coded using adapted versions of Fonagy’s Reflective Functioning Scale (RFS; Fonagy et al., 1998; Slade, Aber et al., 2007; Slade, Bernbach et al., 2004). On both the PI and the PDI, the MRF interviews are scored on a scale ranging from −1 (negative or bizarre RF) to 9 (marked RF). Scores from −1 to 2 indicate pre-mentalizing processes, or an inability to consider one’s own or another’s thoughts and feelings. Interviews in the pre-mentalizing range typically include descriptions of behavior without elaboration of how they may be related to internal mental states (i.e., thoughts, feelings, intentions). These interviews may also include inappropriate attributions of behavior or descriptions where the individual is overly certain about another’s internal experience. A score of 3 is also considered low and can indicate a mother’s consistent use of mental state language or a mother who is inconsistently mentalizing, with some responses to questions in the low range and others in the moderate to high range. Scores between 4 and 6 are considered moderate and indicate a consistent ability to mentalize, and scores of 7 or above are considered high MRF (Fonagy et al., 1998; see Figure 1). Overall MRF scores on the PI and PDI were used in the analyses in the current study. All transcripts were double-coded for both individual questions and the overall score. Any disagreements were conferenced, and consensus scores were used for the analyses. Coders demonstrated a high degree of intercoder reliability for the PI (ICC = .833; p < .001). and the PDI (ICC = .839; p < .001).

FIGURE 1.

Reflective functioning scale (RFS) scores. Source: Adapted from Fonagy et al. (1998) and Slade (2018)

2.4 |. Aims 1 and 2 statistical analyses

The dependent variable in the current study was the three categories of maternal caregiving representations at 7 months postpartum. Descriptive statistics for this variable were calculated first. The distribution of mothers who were classified as having either a balanced, distorted, or disengaged caregiving representation was calculated using univariate statistics. Means and standard deviations were then calculated for the independent variables (prenatal and 7-month MRF). Finally, correlational analyses were conducted to assess associations among demographic factors and the study variables.

Next, hypotheses were tested. We first examined whether prenatal MRF was associated with maternal caregiving representations, and then whether 7-month MRF was associated with maternal caregiving representations using Multinomial Logistic Regression. For both sets of analyses, balanced representations were used as the reference category and distorted and disengaged representations were compared to the reference category.

2.5 |. Aim 3 exploratory analysis of out-of-bounds transcripts

To explore the relationship between MRF and caregiving representations, the first three authors of the paper, who also acted as primary coders, reviewed the PI, PDI and WMCI transcripts belonging to ten mothers who had at least one MRF score that was unexpected based on their assigned typology of caregiving representations. These included transcripts from: three mothers classified as balanced whose PI and PDI scores were low (MRF ≤ 3; 6.4% of sample); five mothers whose PI scores were low, but PDI scores were moderate to high (MRF ≥ 5; 10.6% of the sample); and two mothers classified as distorted whose PDI scores were moderate to high (4.3% of the sample). Following the work of Crawford and Benoit (2009), Newman-Morris, Gray, et al. (2020), Sleed et al. (2021), and Terry et al. (2021), the coders individually reviewed the three transcripts and made notes about how the mother’s circumstance may have changed from pregnancy to 7-months and any indicators of hostility, helplessness, role-reversal, and harsh behavior in the narrative. The coders then discussed the sets of transcripts and pulled out passages that were incongruent with the representational category or the MRF score, discussed themes, and reviewed demographic data for each mother.

3 |. RESULTS

3.1 |. Preliminary analyses

Prior to hypothesis testing, univariate statistics were conducted to describe the distribution of maternal caregiving representations and the means and standard deviations of MRF scores. Next, we analyzed associations among the study variables and mothers’ demographic variables. The distribution of maternal caregiving representations at 7 months postpartum was as follows: 36.2% balanced (n = 17), 38.3% distorted (n = 18), and 25.5% disengaged (n = 12). As indicated in Table 3, there are a range of MRF scores among mothers with balanced and distorted representations, including some MRF scores in the balanced category that were lower than expected (i.e., low MRF scores of 2 or 3) and some scores in the distorted category that were higher than expected (i.e., moderate/high MRF scores of 5 or 6).

TABLE 3.

Distributions of MRF scores across WMCI classifications

	Prenatal MRF		7-month MRF
	Mean (SD)	Range	Mean (SD)	Range
Balanced	3.65 (1.41)	2–6	4.71 (1.36)	2–7
Distorted	2.89 (1.32)	0–5	3.44 (1.25)	2–6
Disengaged	2.33 (.49)	2–3	2.75 (.87)	2–4

MRF at 7 months postpartum was positively correlated with maternal age (r = .30, p = .04). This association was not found for prenatal MRF. Neither prenatal nor postnatal MRF were correlated with income or education. Prenatal and postnatal MRF also did not differ based on mothers’ marital status. Results of one-way Analysis of Variance revealed that demographic factors did not differ significantly among mothers classified as balanced, distorted, or disengaged. As such, demographic variables were not included as control variables for hypothesis testing.

3.2 |. Hypothesis testing

3.2.1 |. Associations between prenatal MRF and caregiving representations

Results of multinomial logistic regressions provided partial support for our hypotheses. Mothers with lower prenatal MRF scores were significantly more likely to be classified as having a disengaged, as compared to a balanced, caregiving representation. However, mothers with lower prenatal MRF scores were only marginally more likely to be classified as having a distorted representation, compared to a balanced representation (p = .08). Contrary to our hypothesis, MRF scores did not differ significantly between mothers with distorted or disengaged representations. As illustrated in Table 4, the model fit the data well X² (10) = 16.8, p = .08 and the multinomial logistic regression model was statistically significant, X² (2) = 8.43, p = .01, Nagelkerke-pseudo R² = .18. These results indicated that the full model predicted the WMCI classifications better than the null model.

TABLE 4.

Multinomial logistic regression: Prenatal MRF as a predictor of caregiving representations at 7 months postpartum

Caregiving representations	b	S.E	Wald	P	OR	95% CI
						LL	UL
Distorted vs. Balanced	−.51	.29	2.97	.08	.59	.33	1.07
Disengaged vs. Balanced	−.97	.39	6.22	.01	.38	.17	.81
Distorted vs. Disengaged	.45	.36	1.62	.20	1.58	.78	3.20

Abbreviation: MRF, maternal reflective functioning.

3.2.2 |. Associations between postnatal MRF and caregiving representations

Results of multinomial logistic regressions provided partial support for our hypotheses. Mothers with higher MRF scores at 7 months postpartum were more likely to be classified as having balanced caregiving representations, compared to those with distorted or disengaged caregiving representations. Contrary to our hypotheses, 7-month MRF scores did not differ significantly for mothers with distorted or disengaged representations. As illustrated in Table 5, the model fits the data well X² (8) = 3.01, p = .93. Results from the multinomial logistic regression were statistically significant, X² (2) = 16.95, p = .00, Nagelkerke-pseudo R² = .34, indicating that the full model predicted the WMCI classifications better than the null model.

TABLE 5.

Multinomial logistic regression: Postnatal MRF as a predictor of caregiving representations at 7 months postpartum

Caregiving representations	b	S.E	Wald	P	OR	95% CI
						LL	UL
Distorted vs. Balanced	−.75	.30	6.12	.01	.47	.26	.85
Disengaged vs. Balanced	−1.40	.45	9.63	.00	.24	.10	.59
Distorted vs. Disengaged	.65	.40	2.61	.10	1.92	.87	4.28

Abbreviation: MRF, maternal reflective functioning.

3.3 |. Description of out-of-bounds transcripts

Although the majority of participants’ caregiving representations align with their MRF scores at the prenatal (80.9%) and postnatal (87.2%) time points in the expected directions, several participants demonstrated unexpected incongruences between their caregiving representations and their MRF scores. A description of these “out-of-bounds” transcript and their correlates is provided below.

At the prenatal time point, nine mothers had out-of-bounds MRF scores (19.1% of the sample). One was a mother whose prenatal MRF score was in the moderate range who was later classified as distorted. Eight were mothers whose prenatal MRF scores were low and who were later classified as balanced. Five of these eight mothers had MRF scores in the moderate to high range at 7 months postpartum and were no longer considered out-of-bounds. At the 7-month time point, only six participants had out-of-bounds scores (12.8% of the sample). Three mothers classified as balanced had low MRF scores and two mothers classified as distorted had moderate MRF scores. A review of the transcripts of mothers with high RF who were classified as distorted revealed similarities in themes of helplessness, hostility, and role reversal in the narrative, alongside genuine mentalizing. Among the transcripts of mothers classified as balanced with low RF, we noticed that all of these mothers had a low level of education (high school or less), and they provided rich behavioral descriptions of their children, with no indications of hostility, helplessness or role-reversal. In addition, their MRF scores on specific questions varied widely and demonstrated an inconsistent ability to mentalize. We selected transcripts from two participants for descriptive presentation purposes. Table 6 provides an overview of these two transcripts with regard to WMCI classification, MRF scores, and a description of factors that may explain the reason for the incongruencies.

TABLE 6.

Out-of-bounds transcripts: Factors that may explain the incongruencies

Participant	WMCI classification	PIMRF score	PDIMRF score	Demographic factors and MRF
A	Distorted	4 (moderate)	6 (moderate)	This mother is considered low demographic risk based on her age, education, and income. She consistently uses mental state language, and different indices of mentalization (opacity, behavior to mental states, mental states changing over time). There were many difficult events that happened between her RF interviews. Across both interviews there were incidences of role-reversal, insensitivity, helplessness and incoherence.
B	Balanced	2 (low)	3 (low)	This mother has not finished high school, otherwise she does not have sociodemographic risks. She is able to “back up” her descriptions of her child and their relationship. In the PI interview, she was able to mentalize on one question. On her PDI interviews she demonstrates the ability to mentalize on two questions and used three types of mentalization. These are counterbalanced with low scores where she does not use mental state language or denies negative emotions.

3.3.1 |. Participant A: Distorted/High MRF

This participant’s WMCI was marked with several features suggesting a distorted caregiving representation, including instances of role reversal, confusion, and helplessness. This mother also displayed some insensitivity, which can be seen in the passage below:

Interviewer: “How would you describe your relationship with your child now?”

“Close, um friends… Inseparable. That is our relationship…..If I’m not there, he yells and hollers. Yeah and we just, just close because like when he’s with somebody he’s looking for me, that makes me feel close to my baby (confused). I think [he does that] because I’m with him most of the time. And I think he know I’ll take care of him and I’m gonna keep him safe and I’m so patient with him. Like I lost my temper once, like, “be quiet aint’ nothin’ wrong with you.”(insensitivity) And he stopped crying and looked. I felt terrible, but I was like “he knows me.” (role reversal) Cause, I was like, “oh my God you aint’ nothing, you not sick, I just gave you medicine gotta’ give it time to work.”(insensitivity) And he stopped. He laughed cause he is such a con artist. He know how to get what he want from me (confused).

Interviewer: What is it like for you when he’s sick? What do you feel like doing?

“Crying, which I do most of the time (helplessness). I just feel so sorry for him cause I’m supposed to protect him (sensitive) and I can’t protect him from his eardrum swelling and it hurts a lot. When I had to go back to work I cried, I hated being away from him. So when he’s sick I cry (helplessness).

When examining this mother’s PI and PDI interviews, this participant had overall MRF scores in the moderate range. She consistently used mental state language, labeled and discussed both positive and negative emotions related to her relationships with the father of the baby, her family, and her experience being pregnant or parenting. On the PI, her MRF scores for individual questions ranged from low to moderate. This mother recognized diverse perspectives between herself and the father of the baby, but her narrative was also clouded by feelings of fear and helplessness. This can be seen in the passage below:

Interviewer: “What are some difficult feelings you have experienced during pregnancy?”

“In the beginning there was fear and apprehension. Because of our ages [baby] could have down-syndrome. I didn’t want to do the test. [Baby’s father] said he wanted the test (using mental state language and recognizing diverse perspectives). I said that it didn’t matter…[baby’s father] was like, ‘I don’t want a retarded baby.’ That made me really apprehensive about [baby’s father]. (Can you tell me about feeling afraid?) The fear that I may have to raise [the baby] by myself and the fear of the unknown freaks me out. I want to know that [baby’s father] will be there. The unknown is horrible (mental states are linked and talking coherently about a difficult situation, but there is also a slight sense of helplessness).”

On the PDI, this participant described some stressful events that had happened in the first 7 months of her child’s life, including the baby’s health issues, splitting up with the father, and moving far away. Despite these circumstances, the mother consistently talked about her thoughts, feelings, and beliefs, and at times did so quite coherently. She had several passages that were scored in the moderate range. She was able to link emotions to behavior, recognized that the child’s emotions are age related, and that emotions change over time. Many of her responses were coherent; however, she also described feelings of helplessness and sadness.

Interviewer: Have you ever felt guilty as a mom?

“Yeah, I feel so guilty that I got him so far away from his family (acknowledges mental states). And that they’re missing him. He can tell that I’m feeling down… it’s like a vibe or something. Maybe I feel guilty because I’m home-sick (mental states impact one another) and trying to find an excuse to come back home…. [When I feel guilty] I text his dad and be like, you need to call (mental states impact behavior).

Interviewer: Tell me about a time when you and (child) were separated

“last month he went to [location where dad lives] to be with his dad… Lord have mercy, it can’t happen again. I couldn’t do it. I was in tears, I missed him so bad. I was a nervous wreck when he came home (mental state language with some helplessness). [Baby’s dad] and I agreed that we can’t do it that way anymore. (What do you think it was like for your child?) He knows his dad. When I left him, he didn’t cry, and his dad sent me pictures and videos and he wasn’t whiney, so I think he was okay….I know he probably missed me (mental state language), but he’s probably not big enough to express it…. I don’t want to be away from him until at least he can talk.”

Interviewer: Tell me about a time when you and (child) were not clicking

“When he’s been sick. He can’t tell me what’s wrong and that’s horrible. I wonder, “am I doing it right?”… He sleeps it off and wakes up better and he’s smiling and playing, but I’m still stuck (getting at diverse perspectives, and mental states changing). (what is it like for him during this time?) I don’t know. He don’t act like…you know how we are? If somebody’s done something to us, you’ll be able to see it in our eyes. So if he’s gone with someone, like his grandma. If something happens, I can’t tell by him (getting at opacity). He won’t give me that look (that lets me know something is wrong) That’s why I don’t think he remembers.”

3.3.2 |. Participant B: Balanced/Low MRF

This participant’s WMCI has several features that suggested a balanced caregiving representation, including richness of detail, involvement, and sensitivity. These are demonstrated in the following passages:

Interviewer: Can you give me a specific memory for your child being like a friend?

“If I’m studying, [baby] will help me study, I make flashcards, and I’ll let [baby] pick up whatever color he wants to pick up (involvement). I don’t think he knows he’s helping me study….. he pick up one, I’ll be like ‘oh!’ and make him feel important and I’ll say the word and say what it means and say it so it seems like I’m talking to him (acceptance, involvement). I’ll pick up some words and start studying those again, and then once he picks up another word, I’ll be like ‘oh, thank you!’ (coherent)

Interviewer: Does C get upset often?

“He gets upset when you’re changing him and he can’t roll, he gets upset when you’re putting on his shirt,(sensitive) um… He gets upset when he’s sleepy or hungry (sensitive)…. So he doesn’t like getting undressed, so…, he don’t like being sleepy. (what do you feel like doing when he’s upset?) “I feel like soothing him in any way possible and just making him comfortable (sensitive) because he’s a human being and if you’re upset and you don’t know how to use your body like that to get stuff, you know? We know how to do stuff for ourselves, so you just gotta’ make him comfortable as possible.”

When examining her PI and PDI interviews, this participant had some responses that either did not contain mental state language, were incoherent, or denied negative emotions. She has other responses where she used mental state language, and few instances of mentalizing. In the PI, her scores for individual questions ranged from low to moderate. This can be seen in the passages below:

Interviewer: Can you tell me about any worries that you’ve had about the baby?

“Oh yeah (laughing). Like it’s very bad for a woman to eat seafood, it has mercury, they baby can become ill. The creepiest thing I watched was this show … and this lady had seafood and now her baby is halfway blind…. And that creeped me out.” (no mental state language)

Interviewer: Can you tell me about any hard or difficult feelings you’ve had during pregnancy?

“Nothing really hard about pregnancy, other than sometimes you get constipated.” (denies difficult feelings)

Interviewer: Are there things that you’re afraid that you’ll do as a parent, maybe things that your parent did to you, that you’re afraid you’ll do?

“Hopefully me and my child has the right relationship where, if I’m not giving him the attention, or if I’m not supportive enough, he’ll tell me and I’ll change. I don’t ever wanna’ have my child feel like he’s in the dark, or an outsider (mental state language). Even If I have more children, I want them all to be treated equal, I don’t want to play favorites (mental state language).”

In the PDI for this participant, she also denied negative feelings, but used mental state language in demand questions that ask for positive emotions. Her scores ranged from low to high, but only two passages were scored in the moderate range. This can be seen in the passages below:

Interviewer: Can you tell me about a time when you and your child were not clicking?

“I don’t think we’re ever not clicking. One minute he can be happy and the next he can be sad, so it’s not like (pause) we’re never not clicking.” (using mental state language, but denies not clicking)

Interviewer: Can you tell me about a time when you felt guilty being a parent?

“No (probed a bit further). He fell off the bed last Sunday and I wanted to beat myself up because I know he moves around (mental state language). I’m like, “I’m such a bad mama”, but the doctors are like, “why are you here….” (What was it like for you?) I was really nervous. Anxious and nervous because I wanted him to be okay (mental states impacting one another). (How did you handle your feelings?) I didn’t show them to [baby] (opacity). I showed them to someone that was old enough to understand emotions and feelings, like to my boyfriend.

In sum, this exploratory description of these theoretically out-of-bounds transcripts for a minority of women in our sample highlights several points that deserve further consideration in future research. First, for mothers with a low level of education, MRF scores may be low, even in the context of balanced representations, or their reflection during pregnancy may be particularly difficult. This descriptive analysis also highlights that hypermentalization and indices of hostile, helpless and role-reversal appear in the narratives of mothers with high MRF, classified as distorted.

4 |. DISCUSSION

The current study examined the relationship between mothers’ caregiving representations assessed at 7 months postpartum and their prenatal and concurrent MRF scores. To our knowledge, only two prior studies have examined these relationships, but both did so postnatally using clinical samples, and only one examined MRF in relation to representational typologies. The current study contributes to the existing research on MRF and caregiving representations in several ways: First, our study examines both constructs in an understudied non-clinical, diverse, low-income community sample. Second, we include assessments of both prenatal and postnatal MRF. Finally, we take an exploratory but in-depth look at the transcripts of a minority of mothers in our sample that had out-of bounds transcripts (i.e., distorted with moderate to high MRF scores and balanced with low MRF scores), in order to describe their commonalities and demographic correlates.

4.1 |. Caregiving representations and MRF

High MRF scores, whether assessed in pregnancy or at 7 months postpartum, increased the likelihood of mothers being classified as balanced versus disengaged on the WMCI. This finding is consistent with results reported in a prior study of MRF and caregiving representations by Schechter et al. (2005), the only study to date that has examined these associations. In their study, MRF was highest among mothers with balanced caregiving representations, and lowest among mothers with disengaged representations, whereas mothers with distorted representations had MRF scores falling in the middle range. It is important to note that the prenatal MRF scores of mothers with distorted representations were not significantly lower than mothers with balanced representations, but that difference approached statistical significance.

There were also no differences in MRF scores among mothers with insecure typologies on the WMCI. The finding that mothers classified as balanced had higher MRF than those classified as disengaged is consistent with other studies that examined the association of differences in MRF and children’s patterns of attachment (Stacks et al., 2014), adult attachment states of mind (Slade et al., 2005) and with research examining adult RF and adult attachment states of mind (Jessee et al., 2016). Other studies also report no significant differences in mothers’ RF scores and their children’s secure and ambivalent attachment patterns (Ensink et al., 2016; Stacks et al., 2014); or between adult RF and adults classified as having autonomous or preoccupied adult states of mind (Jessee et al., 2016; Slade et al., 2005).

Because our WMCI coding did not include the disrupted representation, which is analogous to disorganization, it is unclear whether some of the mothers in the current sample would have been classified into the disrupted category. This issue is especially relevant to those mothers whose PI and PDI transcripts had indications of helplessness, hostility, and role-reversal.

The lack of a significant difference in prenatal RF scores among mothers classified as balanced and insecure-distorted on the WMCI could be due to developmental changes in RF from pregnancy to 7 months postpartum, or to measurement error. With regard to changes from pregnancy to postpartum, it may be that some mothers appear more emotionally aloof during pregnancy, which could influence how their representations and MRF are classified. After the infant is born and becomes more “real,” mothers are better able to describe themselves and their child in terms of mental states and high levels of positive affect (Benoit et al., 1997). To further elucidate the nuances in how caregiving representations and MRF shift both individually and together, future studies should consider longitudinal assessments of both constructs during pregnancy and in the postpartum period. It is noteworthy that, among the eight mothers classified as balanced who had low prenatal MRF scores, five of them had MRF scores in the expectable/moderate to high range at 7 months postpartum, when MRF and caregiving representations were assessed concurrently.

Another potential reason for the lack of difference in MRF scores among mothers classified balanced or distorted could be due to measurement error. This might reflect the fact that both systems require coders to use some elements of judgment, or it could stem from differences in the specific questions included on the WMCI and the PI. On occasion, a few mothers responded to some questions in purely behavioral terms or in an unintegrated, bizarre or inappropriate way, but on other questions, demonstrated the capacity for mentalization. This leaves MRF coders in a dilemma about how much weight to give these lower scores on individual questions, when assigning an overall MRF score. Further, the WMCI, PDI and PI each include different questions, which are coded independently. Thus, any distortions on the PI or PDI would not be known to the WMCI coder, and vice versa.

Although not evaluated in the current study, it is also possible that the range of MRF scores within the representational categories are associated with more sensitive caregiving for mothers classified as distorted and disengaged or less sensitive caregiving for mother’s classified as balanced. Future research should explore this possibility, including how caregiving representations and MRF interact to predict parenting quality and child outcomes. Future studies should also examine whether low maternal education and MRF interact to predict parenting quality. Several studies report that maternal education is associated with MRF (Sadler et al., 2013; Stacks et al., 2014, 2021).

Finally, very few studies of caregiving representations use the WMCI-Disrupted classification, which includes severe role confusion, fearfulness, withdrawal, and negativity (Crawford & Benoit, 2009). The coding system we used did not include the WMCI-Disrupted category. If that category had been included, we would have a better understanding of how MRF scores may differ across the four representations. Again, this would be especially relevant to that subgroup of mothers whose PI and PDI transcripts had indications of helplessness, hostility and role-reversal. While only a small percentage of mothers exhibited this unexpected pattern in the sample, these mothers tended to score in the moderate to high range on MRF because of their consistent use of mental state language and ability to mentalize across the interview. It may be that higher MRF scores in the context of insecure caregiving representations are protective against insensitive parenting and poorer child social-emotional outcomes, or if hostility, helplessness, and role-reversal are better predictors of insensitive parenting and poor social-emotional outcomes. Recent research supports the latter possibility (Guyon-Harris et al., 2021; Hall et al., 2015; Newman-Morris, Simpson, et al., 2020; Sleed et al., 2021; Terry et al., 2021).

4.2 |. Exploring out-of-bounds transcripts

In our study, a small subsample of women demonstrated theoretical inconsistencies between their caregiving representations and MRF scores in pregnancy (19.1%) and at 7 months postpartum (12.8%). This included women who were classified as distorted and had high MRF scores as well as those who were balanced with low MRF scores. While these findings were unexpected, they seem to fit with previous research and suggest a need to further examine MRF scores to understand theoretical inconsistencies. After reviewing these interviews, we found interesting similarities among them that are consistent with the literature. These include mothers who held distorted representations and demonstrated high MRF scores and mothers who were classified as balanced and maintained low MRF scores at both times points. These unexpected cases deserve further attention in future research.

4.2.1 |. Distorted/High MRF

It is important to note that only two mothers classified as distorted in the current sample had moderate to high MRF scores on the PDI. Nevertheless, these transcripts were similar to each other, and different from transcripts that were balanced with high MRF in terms of the mother’s ability to mentalize, coupled with themes of helplessness, role-reversal and insensitivity/hostility. While we can identify these presenting features from a global perspective, MRF coding is unidimensional and only provides an overall typicality score. Overall MRF scores may be inflated for these mothers, as there are no scoring guidelines related to adjusting scores in these circumstances. This makes it difficult to consider whether mentalizing in the context of hostile, helpless and role-reversed representations contributes to “good enough” parenting. These mothers use a lot of mental state language, connect mental states to behaviors, are aware that mental states are not clear and can be hidden, and can consider developmental limitations of their infants’ thoughts, beliefs, and emotion regulation.

While our sample of out-of-bounds transcripts is small, the similarities among them are consistent with a growing body of work studying hostile/helpless representations alongside MRF. For example, in two clinical samples of mother-infant dyads, MRF was positively correlated with mothers’ anxiety, borderline symptoms, and self-reported distorted caregiving representations, including hostility and helplessness, but was not associated with parenting emotional availability. Distorted representations were associated with insensitive parenting, and overall MRF scores moderated this effect (Newman-Morris, Simpson, et al., 2020). Isosävi et al. (2019) also report a case study in which a mother demonstrated multiple hostile/helpless instances and pre-mentalizing risk features in their RF transcripts but was scored high in RF on the Adult Attachment Interview. Moreover, two research teams have developed coding systems to detect representational risk, including hostility and helpless on the PI and PDI. They report that maternal representational risk is associated with less sensitive parenting (Sleed et al., 2021) and child removal as a result of abuse and neglect (Terry et al., 2021). The authors of these studies argue that helplessness might be a caregiving-specific failure in mentalizing that is not adequately detected on the PDI, especially in non-clinical samples. Future studies need to investigate whether high levels of insensitivity, hostility, helplessness and role reversal, even in the context of high RF, are a risk factor for lower quality parenting and lower child social-emotional functioning.

In clinical samples, distorted representations are often associated with interpersonal trauma, stress, and anxiety (Isosävi et al., 2020; Newman-Morris, Simpson, et al., 2020; Sleed et al., 2021). Our sample is drawn from a diverse, urban community and the most common risk factors among participants was poverty and exposure to violence. Poverty is a robust predictor of disorganized attachment (van IJzendoorn et, al.,1999), as is a history of interpersonal trauma (Lyons-Ruth et al., 2004). Unfortunately, we did not assess interpersonal trauma exposure in this study. It is possible that a portion of mothers in this study had a history of trauma exposure. Thus, it is possible that the high MRF scores observed for some mothers with distorted caregiving representations may also reflect these unmeasured mental health risks.

4.2.2 |. Balanced/Low MRF

The second group of mothers, that is, those who were classified as balanced and had consistently low MRF scores in pregnancy and at 7 months postpartum (17.6% of the balanced mothers), all had a low level of education (less than high school, or high school graduate). These mothers also demonstrated a range of RF scores across the interview, with some passages assigned high MRF scores, but no instances of hostility, helplessness or role-reversal.

In contrast with findings from prior studies, which consistently report that mothers with less than a high school diploma score lower on MRF (Pajulo et al., 2012; Sadler et al., 2013; Sleed et al., 2018; Stacks et al., 2021), we did not find a significant association between maternal education and MRF at either time point. This null finding may reflect limited power due to the small sample size, and/or the limited number of mothers who were college educated in the current sample. We speculate that, for mothers with very low education, perhaps potential MRF, the highest score given to any individual question, or the number of types of mentalization used throughout the interview (i.e., understanding that mental states are opaque and that they can change over time) are better indicators of these mothers’ capacity for reflection. Scoring MRF in these ways is not how RF is traditionally scored and thus these alternative methods should be further explored in the future.

Moreover, in many studies, low maternal education is confounded with other clinical risks (Stacks et al., 2014). Perhaps mothers who demonstrate a limited capacity for mentalization, but who do not have high scores on hostile, helpless or role reversal scales or who have low scores in relational risks, are capable of maintaining “good enough” mental representations of their relationships with their babies and in turn “good enough” parenting (Sleed et al., 2021).

4.3 |. Limitations and strengths

The current study has several limitations that should be considered when interpreting the findings. One issue is the relatively small, non-random sample of urban mother-infant dyads, which limits the generalizability of the findings to a similar population and restricts statistical power needed to identify statistically significant results, such as the association between maternal education and MRF. Moreover, information about additional variables that could potentially alter the study’s findings were not available in the current analysis, including mothers’ receipt of mental health services or whether they experienced domestic violence or trauma.

Data were collected from maternal interviews, which may be subject to maternal bias, and the methods used in scoring may carry some measurement error, due to some level of rater subjectivity. A strength of the current study, however, is that 50% of all WMCI interviews in the current study were double-coded and 100% of PI and PDI interviews were double-coded, and all had high inter-coder reliability. The WMCI scoring used only yields three classifications. It is possible that the high percentage of distorted representations reflects the lack of the disrupted classification (Crawford & Benoit, 2009). Finally, this study did not include observed measures of the quality of mother-infant interactions or mother-infant attachment. Inclusion of these measures would shed further light on the concurrent and predictive validity of the WCMI and MRF measures.

Despite these limitations, this study also has noteworthy strengths. This is the first study to our knowledge to assess the association among caregiving representations and prenatal and postnatal MRF. It is also the first to assess MRF and caregiving representations in a non-clinically referred sample of diverse mother-infant dyads from urban, mostly low-income backgrounds.

4.4 |. Implications for practice

Theoretically, the finding that maternal prenatal and postnatal MRF are associated with maternal WMCI representations supports Schechter et al. (2005) suggestion that MRF is a prerequisite for an optimal quality of caregiving mental representations. During pregnancy, mothers’ internal representations of caregiving emerge (Solomon & George, 1996; Zeanah et al., 1986); during this time, higher MRF appears to enhance mothers’ ability to develop a balanced and flexible perception of their child at 7 months postpartum. These mothers’ perceptions of their infants are characterized by rich and coherent narratives in which they also demonstrated acceptance, joy and sensitivity. On the other hand, mothers with low RF have perceptions of their infant that are characterized by emotional distancing and coolness, insufficient acknowledgment of their children’s needs, less interest in their subjective experience, and less involvement and joy in parenting.

Our findings suggest several possible avenues for assessing MRF in clinical interventions. It may be especially important for clinicians to assess mothers’ hostile/helpless caregiving representations during pregnancy and differentiate among mothers who may be pseudo-mentalizing or hypermentalizing. This is because these mothers may be at high risk for developing distorted or disrupted caregiving representations of their infant in the postpartum period and who may struggle to provide consistent sensitive care for their infant that fosters healthy social-emotional development. Providing support and helping mothers to find ways to regulate their negative emotions, while exploring their own attachment histories and fears around caregiving, may be particularly important. It may also be that improving MRF, when disordered representations are not present, may increase mothers’ capacity to imagine the inner world of their child and respond to their child’s needs in a supportive way that meets the child’s emotional needs, including regulating negative emotions.

Key findings and implications.

1. Prenatal and postnatal MRF were associated with mothers’ caregiving representations assessed at 7 months postpartum. Higher MRF in pregnancy and at 7 months postpartum increased the odds of having a balanced versus a disengaged classification. At 7 months postpartum higher MRF also increased the odds of having a balanced versus a distorted representation. MRF scores did not differentiate between non-balanced representations.

2. Some mothers with balanced or distorted caregiving representations had a range of theoretically unexpected MRF scores. Of these, some classified as balanced had low MRF scores and others classified as distorted had MRF scores in the moderate range.

3. Mothers with low MRF scores who were classified as balanced, tended to be young and have a low level of education, whereas mothers with high MRF scores who were classified as distorted had indices of hostility, helplessness and role-reversal in their narratives. Assessing hostile and helpless representations is an important goal for future research.

Statement of Relevance to the field of Infant and Early Childhood Mental Health.

Both caregiving representations and MRF predict parenting behavior and infant attachment. In light of recent research on disordered caregiving representations, the range of MRF scores observed here among mothers with organized caregiving representations may suggest a need to consider hostile/helpless representations in pregnancy. Additionally, there is a need to differentiate among mothers who may be pseudo-mentalizing or hypermentalizing, as these mothers may be at high risk for developing distorted or disrupted caregiving representations.