|Home | About | Journals | Submit | Contact Us | Français|
To determine whether maternal violence-related posttraumatic stress disorder (PTSD), reflective functioning (RF), and/or quality of mental representations of her child predict maternal behavior within a referred sample of interpersonal violence-exposed mothers and their children (ages 8–50 months).
Forty-one dyads completed two videotaped visits including measures of maternal mental representations and behavior.
Negative and distorted maternal mental representations predicted atypical behavior (Cohen’s d>1.0). While maternal PTSD and RF impacted mental representations, no significant relationships were found between PTSD, RF, and overall atypical caregiving behavior. Severity of maternal PTSD was however positively correlated with the avoidant caregiving behavior subscale.
Maternal mental representations of her child are useful risk-indicators that mark dysregulation of trauma-associated emotions in the caregiver.
Through infant-caregiver attachment, primary caregivers provide multiple, complex, and often “hidden” regulatory functions for the developing infant (Hofer, 1984). One form of regulation of critical importance to the child’s capacity to form healthy relationships with others and to learn, is that of emotion regulation (Cassidy, 1994). The term “mutual regulation,” as first used by Tronick and Gianino (1986), refers to a bidirectional, albeit asymmetric, process of emotion regulation between the adult caregiver and the infant. One of the child’s developmental achievements is the internalization of these mutually regulating interactions by the fifth year of life (Fonagy, Gergely, Jurist, & Target, 2002).
We are interested in understanding the interplay of factors that disrupt and facilitate mutual regulation. It is known, for example, that maternal psychopathology (e.g., depression; Tronick & Gianino, 1986, or anxiety; Moore, Whaley, & Sigman, 2004) disrupts mutual regulation. Maternal history of attachment security, on the other hand, has been linked to maternal caregiving sensitivity, which is known to support repair of disrupted parent-infant communication, and thus facilitate mutual regulation of emotion (Lyons-Ruth, Bronfman, & Parsons, 1999).
Multiple studies have shown the adverse effects of various forms of maternal anxiety and depression on mutual regulation of emotion and arousal with infants and young children (Moore et al., 2004; Cole, Barrett, & Zahn-Waxler, 1992). It is clear from these studies that maternal emotional presence to read the infant’s cues and to respond effectively is curtailed by anxiety and depression. Yet few studies have focused specifically on how (i.e. in what specific ways) trauma-related psychopathology, which often involves posttraumatic stress disorder (PTSD) that is comorbid with depressive and dissociative symptomatology would affect a mother’s participation in mutual regulation of emotion and arousal during formative development in early childhood (Heim & Nemeroff, 2001; Simeon et al., 2007).
We believe that it is specifically interpersonal violent trauma-related PTSD as opposed to PTSD related to other forms of trauma (e.g. car accidents, natural disasters, or medical-surgical trauma) that would most impact the parenting of very young children. One reason we think this is that interpersonal violent-trauma related PTSD, specifically, is frequently both chronic and complex in subtype, and as such, often comorbid with major depressive disorder and dissociation (Shalev et al., 1998; Sar, Akyuz & Dogan, 2007; Briere, 2006). Major depression, as previously mentioned and pathological dissociation have also been shown to interfere with mutual emotion regulation during infancy and early childhood, thus compounding the disturbance within the mother-child relationship (Weinberg, Olson, Beeghly & Tronick, 2006; Collin-Vezina, Cyr, Pauze & McDuff, 2005). Both maternal depression and dissociation, like PTSD, contribute to non-contingent responsiveness to the child’s affect, social bids, proximity seeking, and exploration. Depression has been noted to do this largely through maternal incongruence with the infant’s positive affect, increased maternal irritability and negativity, and internal preoccupation (Weinberg et al., 2006). And maternal dissociation has been noted to adversely affect caregiving through maternal inconsistency, unpredictability that poses difficulty for the infant’s transitioning to and modulating within discrete mental states (Putnam, 1997, pp. 187–188; Collin-Vezina, Cyr, Pauze & McDuff, 2005). The latter places the child at significant risk of impairment in the development of integrated percepts of themselves and their caregiver(s), thus potentially affecting subsequent relationships--including those with that child’s own children when that child becomes a parent (Putnam, 1997, pp. 188).
Another reason we think that it is interpersonal violence-related PTSD that would most adversely impact parenting is that we have also noted via our clinical observations, the particular phenomenon in which intense displays of frustration, rage, terror and despair by very young children with limited developmental capacity to regulate their emotion remind many mothers who have been victims of violence of their violent perpetrators’ behavioral dyscontrol. As such, the young child triggers his or her mother’s PTSD symptoms. For example, a mother with a history of physical assault who is hit by a toddler during a tantrum may not be able to place this experience in context of a caregiver-child relationship involving a toddler. PTSD with its implicit confusion of past and present can result in the mother feeling as if she is again a helpless victim with a violent perpetrator or vice versa. Clearly, such a perceived interpersonal threat by the very young child in distress can shift a mother’s primary preoccupation with that young child’s needs to that of her own individual survival: “fight, flight, or freeze.” This redirection of attention to self-preservation removes the mother from her very important focus on her child’s cues, with a focus on self-rather than mutual-regulation of arousal and emotion, and thus greatly increases the risk for gross misinterpretation of her child’s cues. Indeed, one study did find a moderate correlation between severity of self-reported maternal PTSD symptoms that were associated with histories of maternal histories of maltreatment and hostile-intrusive caregiving behavior, the latter particularly associated with history of physical abuse (Lyons-Ruth and Block, 1996).
In cases of maternal violence-related PTSD, these repeated acts of attentional redirection amplify the sense of helplessness and distress in the child. Helplessness and distress in the child, in turn, often lead the mother to further defend herself from her own feelings of helplessness and to distance herself emotionally and/or physically from the child, rather than providing contingent comfort, emotional containment, and protection. The resulting defensive, self-protective maternal response (i.e. “atypical maternal behavior”) to her child’s distress in the face of her PTSD can be measured via behavioral observation in the lab (Lyons-Ruth, Bronfman, & Parsons, 1999).
Dr. Karlen Lyons-Ruth’s et al. extended Main and Hesse’s (1990) observations of parental “frightening and frightened” behavior among parents of toddlers who displayed disorganized attachment behavior upon reunion following separation during the Strange Situation Paradigm. These parental behaviors were significantly associated both with maternal trauma history and with child insecure, disorganized attachment behavior (Main and Hesse, 1990; Lyons-Ruth and Block, 1996). Dr. Lyons-Ruth et al. (1999) described additional “atypical” caregiver behaviors that were posited as also being significantly associated with maternal trauma history and child disorganized attachment behavior; these included: affective communication errors (i.e. maternal affective incongruence with child affect, particularly during child distress), caregiver disorientation (i.e. non-contingent, odd parental behavior that is frequently associated with dissociation), and role-reversal (i.e. the parent assumes a child-like stance and/or submits to the child-as-authority). The measure that developed out of Dr. Lyons-Ruth et al.’s operationalized criteria for coding these behaviors as well as Main and Hesse’s original frightening (renamed hostile/intrusive) and frightened (renamed avoidant/withdrawn) parental behaviors is described further below (see Method--Measures).
Women who have PTSD due solely to non-interpersonal causes (i.e. car accidents or natural disasters) would be less likely than victims of interpersonal trauma to display these types of atypical caregiving behaviors (Bryant et al., 2004; Kilic, Ozguven, & Sayil, 2003). This is because these behaviors are essentially thought to be related to impairment in social information processing directly relating to the caregiver’s adverse interpersonal experiences that leads the traumatized caregiver to feel “threatened” by her own young child’s behavior and mental states (Schechter, 2003; Elwood, Williams, Olatunji & Lohr, 2007).
We thus hypothesize that severity of maternal PTSD not only will be associated with non-balanced, and particularly, distorted mental representations as we have already demonstrated (Schechter et al., 2005), but that non-balanced, distorted maternal mental representations and greater severity of maternal PTSD will also be significantly associated with atypical maternal behavior in the lab. Discerning balanced from non-balanced, disengaged versus distorted maternal mental representations
Following in the tradition of the Adult Attachment Interview (AAI; George, Kaplan, & Main, 1985), that measures retrospectively the quality of adult mental representations of that adult’s own caregiver(s) with respect to attachment, two measures were developed that measure the adult-as-parent’s mental representations of his or her own child: The Parent Development Interview (PDI [Aber, Slade, Berger, Bresgi, & Kaplan, 1993]) and the Working Model of the Child Interview (WMCI [Zeanah, Benoit, & Smyke 1995/2000]). Because of its history of application to high-risk samples, we chose the WMCI for this study as described in the Methods section. Coding of the WMCI results in an overall clinical classification of maternal mental representations of her child as either balanced, or as one of two “non-balanced” categories: disengaged and distorted (Zeanah et al., 1995/2000).
Narratives classified as “balanced” include both positive and negative characteristics of the child’s personality or the caregiver’s relationship with the child. They convey a sense of the caregiver as deeply involved in the relationship with the child, as recognizing and valuing the child’s individuality, as empathically appreciating the child’s subjective experience, and as valuing the child and relationship with the child. The mental representations are open to change and accommodate new information about the child and parenting. They also convey at least moderately rich details about the child and the caregiving experience (Zeanah et al., 1995/2000).
Non-balanced, “Disengaged” mental representations are characterized by pervasive emotional distance or indifference towards the child. The topic of the child or caregiver’s relationship with the child may be approached at a cognitive level and be remote from feelings and emotions. The significance or impact of parenting remains either emotionally unintegrated or dismissed as unimportant and noninfluential on the development of the child or relationship with the parent (Benoit, Parker, & Zeanah, 1997).
Non-balanced, “Distorted” maternal mental representations are characterized by distortion of mother’s mental representation of her child and/or relationship with the child. The distortion may take one or more of the following forms: devaluing or excessively negative, self-referential, or role-reversed. Parents with distorted mental representations convey an impression of unrealistic expectations of their child, attribute malevolent intents to the child, or are grossly insensitive to the young child. Descriptions of the child may be highly incoherent in the sense of being confused, contradictory, or bizarre. Maternal narrative may indicate that she is preoccupied by angry affect or distracted by a particular aspect of her child that precludes an integrated representation. Her narrative may reflect that she is confused, anxious or overwhelmed by the infant or self-involved. Much feeling is expressed about the child, but these emotions lack modulation, and contextual meaning.
We have hypothesized that maternal PTSD will be associated specifically with the distorted classification by virtue of this lack of modulation of negative affect and the maternal projection of power, threat, or otherwise developmentally inappropriate attributes onto the very young child. In the current coding scheme (Zeanah et al., 1995/2000), if elements of both disengaged and distorted classifications are present, the coder is instructed to rate the narrative as “distorted.” The distorted classification would correspond both to the “preoccupied” (i.e. preoccupied with past attachment relationships, with a predominant affect of anger and/or fear, yet consistent and organized as a strategy) and “unresolved” (i.e. characterized by a lack of a consistent, organized narrative strategy, thus often contradictory, and confused/confusing) classifications on the AAI (George et al. 1985). And we think also that maternal PTSD and distorted mental representations of the child will be particularly disorganizing to maternal interactive behavior.
An important and measurable indicator of a history of maternal attachment security and subsequent caregiving sensitivity is that of maternal reflective functioning (Grienenberger, Kelly, & Slade, 2005). Maternal reflective functioning is the capacity to infer mental states in her child and in herself. Maternal reflective functioning has been found to predict infant attachment security (Fonagy et al., 2002), which in turn is related to prosocial behavior, successful academic functioning, and decreased anxiety (Lyons-Ruth et al., 1999). The capacity of the caregiver to consider and to contextualize her own thoughts and feelings as well as those of her child, and then to respond from a developmentally informed vantage point, to her child’s cues, is fundamentally supportive of emotional regulation (Fonagy et al., 2002).
We have therefore wondered whether RF might counteract at least the dysregulating effects of maternal traumatization at the behavioral level. Indeed, despite that effects of maternal psychopathology were not taken into account, Grienenberger et al., (2005) showed that RF predicted greater caregiving sensitivity as marked by less atypical maternal behavior within a non-referred lower middle class sample. One goal of the present study is, therefore, to replicate and extend their findings and to explore the potential link between maternal PTSD, reflective functioning as an indicator of maternal attachment security, and maternal caregiving behavior within a violence-exposed sample.
Since we have already shown that maternal reflective functioning and severity of PTSD are associated with non-balanced, distorted mental representations (Schechter et al., 2005), in the present study, we wanted to test the hypothesis that lower levels of maternal reflective functioning will be significantly associated with atypical maternal behavior.
Permission to conduct the study was obtained from the institutional review board at the Columbia University Medical Center Department of Psychiatry. All biological mothers (66) and their children who were registered or presented for evaluation in the hospital-based mental health clinic for very young children (ages 0 to 5 years) and their families “The Infant-Family Service” between January, 2000 and December, 2001 were offered the opportunity to participate in a study that they were told was meant to understand what “made life more versus less stressful for mothers of very young children in the community”.
Families were referred to this clinical service by medical and mental health professionals, daycare centers, and community social service agencies for concerns about potential for child abuse, neglect, or family violence. Entry criteria included a complaint by the mother or others involving concern about potential or actual violence or disruptive behavior in self, child, or other household member. Maternal diagnosis of PTSD was not an inclusion criterion. Women with active psychotic symptoms, substance abuse, or who were not the primary caregiver of their child for most of that child’s life, were excluded. Further exclusion and inclusion criteria have been described in a previous paper (Schechter et al., 2005). Mothers and children (ages 8–50 months) had to have been physically and developmentally able and well enough to participate in a 20 to 30-minute interactive play paradigm, as also determined by clinician assessment and clinical records.
Of those 66 contacted, 21 (32%) refused to participate, did not follow-up in the clinic prior to signing informed consent, or were found not to meet entry criteria on psychiatric screening due to presence of psychotic symptoms, substance use, or not being a primary caregiver for most of their child’s life. Forty-five mothers signed the consent form, out of which, 1 caregiver disclosed that she was not the biological mother and 1 mother dropped out. The 43 remaining participant-mothers shared in common a chief complaint involving violent, aggressive, or disruptive behavior of their preschool-age child or of another household member. Forty-one (95%) stated that they had had exposure to interpersonal violent trauma themselves in childhood and/or adulthood (physical and/or sexual abuse and/or domestic violence during childhood and/or physical and/or sexual assault in adulthood). The history of prior interpersonal violent trauma in the lives of the remaining two mothers proved inconclusive, such that they were excluded from this study, leaving a remaining N of 41.
Study participants were those forty-one mothers ranging in age from 18–45 years, with a mean age of 29 years, and their children ages 8–50 months, with a mean age of 32 months.
Maternal life events measures used to select the sample included the Life Events Checklist (LEC; Johnson & McCutcheon, 1980) and the Brief Physical and Sexual Abuse Questionnaire (BPSAQ; Marshall et al., 1998). Both of these measures are described below. To be included in the study, a mother would have to have stated that she had had at least one experience of physical or sexual assault and/or family violence exposure at any point in her life on either or both of these two measures.
As many as 24 out of 41 mothers (59%) stated that their child was one of the three greatest stresses in their lives. While mothers’ clinical concerns were most often child-related, approximately one-fourth of cases presented with primary concerns about parental mental health, trauma, or severe stress.
Twenty-eight mothers (68%) were on public assistance; 22 (54%) had less than a high-school education; 27 (67%) were single and their 27 children were said not to have consistent, predictable contact with their biological fathers. The sample was 88% Hispanic, largely of Dominican origin—most of whom were immigrants (61%) and 12% were African-American. As the community surrounding the hospital is heavily Caribbean-Hispanic, participants were introduced to the study, given informed consent, and interviewed in either English or Spanish according to their preference. All research staff members were fluent in English and Spanish.
A standard Demographic and Treatment History Questionnaire, consisting of 33 closed-and open-ended items was developed for use in this study (Schechter et al., 2005). Several items probed for mental health treatment history for any reason and in any form (i.e. counseling, brief or long-term individual, family, or group psychotherapy, medication, day program, emergency room, inpatient hospitalization, substance abuse rehabilitation program).
The Life-Events Checklist (Johnson & McCutcheon, 1980) is a 17-item checklist covering a range of potentially traumatogenic events from natural disasters to accidents, sudden losses to combat and interpersonal violent events. The Brief Physical and Sexual Abuse Questionnaire (BPSAQ, Marshall et al., 1998) was used in order to quantify the severity of maternal violent trauma history (Schechter et al., 2005). The measure has shown reliability in predicting clinician rated PTSD in two separate studies (Marshall et al., 1998; Schechter et al., 2005).
As described in Schechter et al. (2005), violence-related posttraumatic stress disorder (PTSD) was assessed by the clinician using the Structured Clinical Interview for the DSM-IV (SCID--PTSD Module with Chronology of Life Events; First, Spitzer, Gibbon, & Williams, 1995)), in addition to a self-report checklist for symptoms (Posttraumatic Symptom Checklist—Short Version (PCL-S); Weathers et al., 1996) within the month prior to assessment. A subject was said to meet criteria for PTSD if a) DSM-IV criteria were met on the SCID using symptom counts: at least 1 Reexperiencing symptom, 3 Avoidance symptoms, and 2 Hyperarousal symptoms, plus duration and demonstrated impairment, and b) the overall PCLS score was 45 or greater within the month prior to assessment.
Severity of PTSD was measured by counting the number of symptoms endorsed on both the SCID PTSD Module and the PCL-S. Symptom count, having been shown to be a valid marker of severity of PTSD in previous studies (Marshall et al., 2001), was used so as to be able to render comparable clinician rated symptoms (SCID) and self-reported symptoms (PCLS) since the SCID does not include a measure of symptom severity. The correlation of symptom count to symptom severity on the PCLS was robust (r=.88, p<.001).
The Working Model of the Child Interview (WMCI) (Zeanah et al., 1995/2000) is a one-hour semi-structured interview that assesses caregivers’ mental representations of their children’s personality and their relationship with their children. It has been used with parents from pregnancy through age 5 years. All questionnaires and interviews including the WMCI were translated into Caribbean Spanish and back-translated. For the purposes of coding maternal perception, the interviews were videotaped and then rated along fifteen 5-point rating scales that include both content and qualitative dimensions, the latter, encompassing formal organizational characteristics of the maternal narratives. The scale-scores in addition to an overall clinical impression inform the coding of the overall classification as either balanced, or one of two “non-balanced” categories: disengaged and distorted as described above (Zeanah et al., 1995/2000).
In this study, both the primary coder (postdoctoral level clinician and researcher, at an independent institution) and second expert coder (i.e. primary author of the instrument) were both naïve to history of the participants. Scoring followed from viewings of the videotape, which is the current standard method for coding the WMCI (Rosenblum, Zeanah, McDonough, & Muzik, 2004). The second coder coded a randomly selected 10 tapes out of the 41 for interrater reliability on the overall classification of balanced, disengaged, vs. distorted (kappa=1).
Maternal RF was measured as described in Schechter et al. (2005), by coding narrative responses to the WMCI using the Reflective Functioning Scale as adapted by Dr. Arietta Slade (Grienenberger et al., 2005)
In consultation with Dr. Slade, the authors applied the parental RF coding system to the WMCI. WMCI item-content was not altered, but additional probes or “demand questions” to elicit thinking about mental states were included to render the coding condition comparable to that of the Parent Development Interview (PDI) (see Grienenberger et al., 2005). For example, after the WMCI item, “What of your child’s behavior is most difficult for you to manage?” which is followed by the question “Why do you think that he does that?” we added the probe, “What do you think was going on in his mind?” An example of a response consistent with low RF would be the following mother’s response to that probe:
“I don’t know. He’s just like his father.”
By contrast, the following mother’s response would be consistent with high RF:
“I can’t be sure…but I think that he feels scared when I leave and so he holds on to me and cries. That makes me feel so guilty that I have trouble leaving.”
To code the WMCI items with responses to pre-existing or added demand questions for RF, twenty to thirty minutes of videotaped WMCI responses to four items (i.e. WMCI items 4, 5, 7, and 11) were transcribed and coded by a co-author of the RF coding system for the PDI. These particular items were selected from the WMCI because they probed for mothers’ thinking about her child’s thoughts, feelings, and intentions rather than for descriptions (i.e. content items). The RF coder, who was different from the 2 coders who coded mental representations on the WMCI, was naïve to any information about the mother-child dyads except for the child’s age. While adaptation of the Slade et al. coding system for use with the set of WMCI items selected was not validated against the existing version of the Parent Development Interview, the approach taken in the present study is common practice in research that involves labor-intensive coding of narrative and observational measures. Moreover, the primary RF coder, himself a coauthor of the adaptation of the RF coding scheme used to code parental mental representation measures (Grienenberger et al., 2005), and Dr. Slade, the primary author of the coding scheme, established reliability on the RF coding of the WMCI version that was used in this study prior to the primary coder’s proceeding to code RF for the whole sample.
Lyons-Ruth et al. (1999) developed the Atypical Maternal Behavior Instrument for Assessment and Classification, or AMBIANCE, a measure of atypical maternal behavior measured in a structured laboratory setting. AMBIANCE measures five dimensions of caregiving behavior described earlier in this paper: affective communication errors, avoidance/withdrawal, disorientation, hostility/intrusiveness, and role-reversal. These caregiving behaviors as measured by the AMBIANCE were significantly associated with insecure, disorganized attachment classification of children in the Strange Situation (Goldberg, Benoit, Blokland, & Madigan, 2003). For a similar sample of low-income mothers, level of “disrupted affective communication,” or overall AMBIANCE score, was found to be highly reliable among raters (ICC= 0.93), as was the categorical “disrupted” vs. “non-disrupted” measure (Kappa= 0.73).
In this study, both the criterion coder and reliability coder were both naïve to any information about the participants beyond age of the child. Scoring followed from multiple viewings of the videotape that took into account verbalizations by the caregiver and child, which is the current standard method for coding the AMBIANCE. The second coder coded a randomly selected 5 tapes out of the 41 for interrater reliability (ICC=1 overall continuous score; ICC=.74 and .76 for dimensions Withdrawal and Hostile-intrusiveness/negativity respectively). The coders coded approximately 20-minutes of videotape: both a 10-minute segment of free play plus two separation-reunion segments.
They coded maternal behavior in these segments continuously along five dimensions: affective communication (i.e. contradictory signaling to child and inappropriate responding to child cues), role/boundary confusion (i.e. role reversal, treating child as sexual or spousal partner), disorientation (i.e. appears confused or frightened by child, appears generally disorganized or disoriented), hostile-intrusiveness/negativity (i.e. physically intrusive or frightening, verbally intrusive or frightening, inappropriately attributes negative feelings or motivation to child, exerts control with objects), withdrawal (avoidant, creates physical distance from child, uses words to distance self from child or contradicts cues suggesting proximity seeking). These five dimensions plus an overall clinical impression informed the overall score on a 7-point scale with “1” being least atypical to “7” being most atypical. The frequency of events meeting criteria for each of the subscales was additionally tabulated by the primary coder only and entered into the database for additional qualitative comparisons.
The protocol consisted of two 2-hour videotaped visits:
During the initial visit, following a clinical and treatment history interview, mothers were administered an abbreviated version of the WMCI that included items 2, 3, 4, 5, 6, 7, and 11. Followed by the WMCI, the Life Events Questionnaire, the BPSAQ and the SCID PTSD-module were administered During the second visit, approximately one to two weeks later, mothers and children were observed a) playing together as they would at home using a range of toys provided (10 minutes), b) separating and reuniting, c) cleaning-up the play area, and d) engaging in a challenging structured activity such as building a tower (toddlers) or doing a puzzle (preschoolers). The AMBIANCE was coded from the videotaped observations of this laboratory observation procedure.
Analysis of variance followed by Cohen’s d (unadjusted) and multiple linear regression models (adjusted) were estimated to test differences for maternal behavior as measured by AMBIANCE among the three WMCI categories. Adequacy of the models was judged by R-square. Analyses were done to test our a-priori hypotheses to single out the Distorted category from the other two categories. The AMBIANCE was regressed on WMCI and pertinent demographic factors: Maternal and child age, number of years of maternal education, household income, and history of having had any mental health treatment for any reason and of any variety. These particular demographic variables were selected because of the concern these particular variables might influence quality of maternal mental representations and/or behavior.
On the BPSAQ, all 41 mothers stated that they had had exposure to interpersonal violent trauma themselves in childhood (i.e. prior to age 16) as victim and/or witness. Twenty-nine of the 41 mothers (71%) stated that they had experienced violent trauma also during adulthood (i.e. after age 16). Trauma severity score was derived from the BPSAQ: Out of a possible score range of 0–38, the sample range was 3–32 with a mean of 15.6 and standard deviation (SD) of 7.3. This mean score suggests that half the sample experienced 2 or more forms of maltreatment during childhood (Schechter et al., 2005).
There were additional risk indicators during their adult years: Fifteen mothers (37%) filed for restraining orders during their adult life and a similar number (16, or 39%) already had a history of investigation by child protective services by the time they reached the study. Sixteen (39%) disclosed their own history of violent behavior; while 14, or 34%, disclosed a history of suicide attempts. Despite this high level of risk and expected psychiatric morbidity, only 14 of the 41 mothers (34%) had ever been in psychotherapy of any kind prior to the study.
Using the PCL-S and SCID PTSD Module, the rate of current PTSD related to interpersonal violent trauma only was 44%. Using the PCL-S as a continuous measure, out of a possible range of 17–85, the range in the sample was 17–82 with a mean of 36 and SD of 17. Individuals with scores greater than 40 are likely to be in the diagnostic range (20 or 49%). Scoring the PCL-S by number of symptoms endorsed, out of a possible range of 0–17, the range in the sample was 0–17, with a mean of 9 and SD of 5.4.
The rate of lifetime PTSD using the SCID PTSD Module was 90% (n=37/41), with the remaining 10% (n=4/41) suffering from clinically significant sub-threshold symptoms. These rates would be unusually high were it not for the fact that the study-sample consisted of parents and young children referred for concerns related to violence and maltreatment risk. Out of a possible range of 0–17, the sample range for lifetime PTSD severity as marked by the number of endorsed symptoms was 8–15, with a mean of 13 and SD of 2.4.
Reflective function (RF) varied in a relatively narrow range. While the possible score range was −1 to 9, the sample range was 0–5 with a mean of 3.3, and SD of 1.3. This mean of “3” suggests that on the average, mothers in this sample were generally able to label affects in their children and/or themselves (i.e. “My child had a tantrum. He was angry.” (Would receive an item score of “3”) vs. “My child had a tantrum. He is an evil one.” (Would receive an item score of “1”). Yet mothers with overall RF at “3” would not show evidence of being able to regularly link affects as motivations for action in self or other (“5”): “He was angry that I took his toy away and so he had a tantrum.” or to mental states in the other (“7”): “He saw that I was anxious and rushed. When I turned away to go out, he threw an angry tantrum to get me perhaps to turn back to him again.”
Via logistic regression, we first examined four possible demographic factors that could influence maternal mental representations on the WMCI: maternal age, child age, child gender, maternal education, and maternal history of prior mental health treatment (individual, group, family, or parent/child, with or without medication). Each was unrelated to the WMCI category.
The same five demographic factors were considered via multiple linear regression with respect to maternal behavior in terms of the AMBIANCE scores. While maternal age, child age, child gender, and maternal education were unrelated to AMBIANCE scores, history of mental health treatment (i.e. any encounter with mental health professionals involving 3 or more outpatient visits or an inpatient admission) was significantly associated with lower AMBIANCE scores: The mean AMBIANCE score for mothers with no history of mental health treatment was 5.44 (SD= .96), and with a history of mental health treatment, 4.56 (SD=1.58) (ANOVA: F(1,39)= 4.42; p=.04).
To address the hypothesis that non-balanced, particularly distorted maternal mental representations would be associated with atypical maternal behavior, we performed an ANOVA that compared the mean level of Atypical Maternal Behavior for mothers across non-balanced and balanced WMCI classifications. We found a significant relationship (p≤.05) as shown in greater detail in Table 1.
The analysis depicted in Table 1. was preceded by examining frequencies and distributions of the WMCI and the AMBIANCE: For maternal perception based on WMCI classifications, within the sample, 7 (17%) of mothers were in the “balanced” category, and 34 (83%) in the “non-balanced” (10 [24%] “disengaged,” and 24 [59%] “distorted”).
On the atypical maternal behavior interview (AMBIANCE) scale, 30/41, or 76%, scored “5” or higher indicating the categorical status of “disrupted communication.” The mean for the sample was 4.9, with an S.D. of 1.43, a range of 6, a minimum of 1 (2 mothers) and a maximum of 7 (1 mother).
ANOVAs as depicted in Table 1. show that the WMCI classification was associated with significant differences in the mean degree to which mothers display atypical maternal behavior on the AMBIANCE. These mean differences were at the threshold of the disrupted vs. non-disrupted maternal behavior cut-off (score of 4 and below vs. 5 and above) on the AMBIANCE scale.
Both non-balanced classifications on the WMCI had significantly higher mean AMBIANCE scores than the balanced classifications. Balanced vs. Non-balanced (i.e. Distorted + Disengaged) classification groups had significant differences in their caregiving behavior with large effect sizes (Cohen’s d). These significant differences were accounted for by the relationship of Balanced classification to the Distorted one. Of note, there was no significant difference between the Disengaged classification and that of the Distorted, or the Balanced ones at this level of analysis.
Post-hoc analyses showed additionally that, when looking at the dimensional subscales of the AMBIANCE across this interpersonal violence-exposed sample, the mean frequency of hostile-intrusiveness/negativity was significantly higher among mothers with the distorted WMCI classification than that among mothers with either balanced or disengaged WMCI classifications (F (1,40) = 7.10; p = 0.01). Similarly, the mean frequency of disorientation/frightened behavior was higher at a trend-level of significance (F (1,40) = 3.80; p = 0.06) among the distorted WMCI classification group than that of either the disengaged or the balanced group.
A linear regression model was applied to determine the predictive value of WMCI classification with respect to AMBIANCE overall scores. Results showed that Balanced perception accounts for 16% of the variance of AMBIANCE scores. Adjusting for maternal treatment history as a dichotomous independent variable did not significantly impact the multiple regression model. Moreover, adjusting for maternal PTSD and/or degree of interpersonal violent trauma exposure did not significantly impact the model.
When we tested that the remaining two hypotheses, a) that lower levels of maternal RF and b) greater maternal PTSD severity would be associated with greater atypical maternal behavior, we found no significant relationships between either of these variables and the level of disrupted communication on the AMBIANCE (p>.4).
That being said, we did find that as many as 76% of the overall clinical sample manifested atypical maternal behavior in the disrupted affective communication range (≥5), indicating the need for a non-PTSD control group. The possible score range and the sample range were 1 to 7, with a mean of 4.83 and SD of 1.43. The mean frequency of behaviors across the five subscales was as follows: affective communication = 15.71 (SD13.43), role-boundary confusion = 14.88 (SD 13.78), disorientation = − 6.02 (SD 6.84), hostile-intrusiveness/negativity = 20.43 (SD 13.36), withdrawal = 20.43 (SD1.06). The frequency of withdrawal (i.e. avoidant maternal behaviors) on the AMBIANCE subscales was positively correlated with the number of maternal PTSD symptoms (r=.37; p=.04). The other four subscales (e.g. affective communication errors, role boundary confusion, disorientation, and intrusiveness) were not significantly associated with maternal PTSD.
We have found that non-balanced maternal mental representations on the WMCI were significantly associated with more atypical, non-contingent maternal behavior as measured by the Atypical Maternal Behavior Instrument for Assessment and Classification (AMBIANCE). Mothers classified as having distorted mental representations on the WMCI had significantly higher levels of atypical behavior than those mothers with balanced mental representations, as was hypothesized. The quality of maternal behavior associated with those distorted mental representations was noted as primarily hostile-intrusive/negative or “frightening”, and to a lesser degree, “frightened”. Given the relationship between maternal distorted, negative mental representations and atypical maternal behavior, we can also revisit the previous findings and understand that the degree of maternal avoidance of the child’s proximity seeking (i.e. mother appearing withdrawn and inhibited on reunion) that was moderately correlated with the severity of maternal PTSD, may be seen as another aspect of “frightened” behavior on the part of the caregiver (Lyons-Ruth et al., 1999). These characteristics of the clinical sample reported here are consistent with intergenerational transmission of interpersonal violent trauma to which the mothers had been exposed.
The results of this study thus suggest that a mother’s mental representations of her child may well exert a regulatory effect on her interactive behavior with her child. As we have shown, greater severity of maternal interpersonal violence-related PTSD is associated with distorted maternal mental representations, and greater reflective functioning is associated with balanced and integrated maternal mental representations of her child (Schechter et al., 2005). It is thus possible that maternal mental representations of her child may mediate the effect of proximal “hidden regulators” of maternal behavior (Hofer, 1984).
Within the non-balanced category the difference between mothers classified as having distorted vs. disengaged mental representations with respect to levels of atypical maternal behavior was not statistically significant at this level of analysis. That being said, since post-hoc analyses suggested that the intrusiveness and withdrawal AMBIANCE subscales were associated uniquely with the distorted classification at a trend-level of analysis, the absence of significant findings with respect to the overall score may represent a limitation of power in this study.
We did not find an association between maternal reflective functioning and atypical maternal behavior within this uncontrolled referred sample of violence-exposed mothers and young children. These results differ from those of Grienenberger et al. (2005) who did find such a connection within a non-referred population. It is possible that the limited range of RF in our clinical sample prevented us from replicating those findings. It is also possible that, while maternal RF is associated with balanced mental representations (Schechter et al., 2005), which may be, as we explored in this present study, associated with more sensitive caregiving behavior, maternal RF is not directly correlated with the quality of maternal behavior in clinical samples in which there is significant psychopathology. Given these various possibilities, we are now reexamining this question with a non-referred control group.
Similarly, we did not find a direct correlation between the number of maternal PTSD symptoms as a marker for severity, and the degree of atypical maternal behavior. That being said, the fact that 76% of mothers with clinically significant PTSD symptoms and a reduced range of RF scored in the disruptive communication range of the AMBIANCE (i.e. score ≥ 5), and showed a predominance of intrusive type behaviors, one could argue is consistent with Lyons-Ruth and Block’s (1996) findings regarding hostile-intrusive maternal behavior among maltreated mothers, and Grienenberger et al.’s (2005) findings regarding reflective functioning, respectively. This study extended previous work in this area by studying maternal withdrawal (i.e. fearful/avoidant maternal behavior) and its significant association to severity of maternal PTSD.
The more severe and symptomatic mother’s PTSD was, the more likely she was to maintain a physical and/or psychological distance from her young child. We have hypothesized that this may be due to the traumatized mother’s need to protect herself from further psychophysiologic dysregulation in the presence of her child’s distress particularly during separation-reunion. Despite that, in our previous paper (Schechter et al., 2005), we demonstrated that low maternal RF and high severity of maternal PTSD were significantly associated with non-balanced maternal mental representations of her child, we found no zero-order relationship between maternal reflective functioning, maternal PTSD severity, and atypical maternal behavior. Maternal RF as well as maternal PTSD severity were orthogonal to each other, yet each highly intercorrelated with maternal mental representations (Schechter et al., 2005). This intercorrelation in addition to low variability for these two independent variables (e.g. maternal reflective functioning, PTSD) may well have limited our ability to test the second hypothesis with this referred sample.
The most significant limitation of this study is the absence of a non-PTSD control group. We think that our negative findings with respect to the relationship of PTSD severity and of level of RF to atypical maternal behavior, respectively, are likely due to this lack of a control group (Schechter et al., 2005). Similarly, while such a high percentage (83%) of unbalanced WMCI classifications is consistent with other violence-exposed samples, such as that in the study noted above (Huth-Bocks, Theran, & Bogat, 2004), the results presented indicate the need for replication with a control-group with respect to the study of maternal mental representations. Despite these limitations, the traumatized sample studied is representative of the clinical cohort that we treat as clinicians at our inner-city medical center. And from the clinician’s point of view, the systematic understanding of the referred sample through the research described in this paper has proven useful to staff on our Infant-Family Service (see Clinical Implications below).
Another methodological limitation of this study concerns the use of the abbreviated WMCI with RF probes added to code RF as opposed to using the Parent Development Interview (PDI) for which the Slade et al. coding scheme had originally been developed. Further research is needed to define the optimal amount of narrative text needed for a valid rating of parental RF. While the strong relationship of RF and balanced mental representations on the WMCI in this study support the validity of the RF coding, it may be that the amount of maternal narrative coded in this inner-city sample would not be sufficient in another sample, thereby limiting the generalizability of the findings.
Further research involving both a non-violence exposed and a non-PTSD sample from the same inner-city community is clearly needed to replicate the findings described in this paper and, as mentioned, is now under way. Similarly, since only mothers were included in this study, further studies involving fathers as well as caregivers drawn from other immigrant groups, cultures, regions, and social status are needed.
First of all, this paper supports the clinician’s active assessment and treatment of the sequelae of parental exposure to interpersonal violent trauma. Secondly, this paper has empirically demonstrated that distorted maternal mental representations that were found in the large majority of a traumatized sample of mothers are significantly related to those mothers’ atypical behavior with their young children. Two clinical implications of this study are, therefore, that non-balanced, and in particular, distorted maternal mental representations may be useful risk-indicators: 1) in assessment of parent-child relationships cross-sectionally, and 2) in the assessment of change with intervention that targets caregiving behavior. All clinicians regardless of their discipline should routinely listen carefully to how parents talk about their children and their relationship with their children. Specifically, highly negative, markedly ambivalent, contradictory, or age-inappropriate descriptors—whether devaluing or idealizing of their children and/or relationship with their them, should be noted and explored in greater depth. These types of descriptors may well represent red flags that mark a significantly disturbed parent-child relationship that could benefit from dyadic intervention and/or require investigation by a child protective agency or protection by a victims’ services agency (i.e. domestic violence shelter or special victims unit of the police department) in the interest of maintaining the child’s and primary caregiver’s safety.
Open-ended demand questions in the course of assessment, such as “Tell me about your child…what is he/she like?” or “Tell me a story that captures the essence of your relationship with your child…, similar to items found in the WMCI, can be very clinically useful. In listening to the caregiver’s responses, the clinician is advised to note what of the caregiver’s narrative response is coherent versus incoherent, as well as to observe the caregiver’s emotional communication in terms of congruence, level of arousal, and positive versus negative affective valence in telling the story. Again, incoherence and affective incongruence may represent signs of significant disturbance in the child-parent relationship. Questions about what the caregiver is reminded of when she looks at her child, or sees a particular behavior enacted by her child can also be quite telling and can provide a bridge to hearing important, but otherwise avoided history of the caregiver’s caregivers and about other aspects of her interpersonal world.
Finally, the authors would like to emphasize that the research presented in this paper is preliminary and has not yet been replicated with a control-group. Therefore, the authors do not recommend that the methods employed in this research paper be applied to assess parenting capacity in the clinical setting, especially in the absence of an experienced early childhood mental health specialist. Moreover, a caregiver’s history of victimization, complex PTSD and significant attachment disturbance(s) in her own life should not lead the clinician to assume anything about that particular caregiver’s capacity to parent her child. Knowledge of such a history by the clinician should rather lead the clinician during the assessment and treatment to support an empathic, reflective, and, when appropriate, positive regard for what that caregiver is able to do to foster her child’s relationship with her and that child’s healthy development in spite of any adversity that she may have experienced.
The authors wish to acknowledge the assistance of Drs. Susan Brunelli, Myron Hofer, John Grienenberger, Arietta Slade, Elissa Bronfman, and Karlen Lyons-Ruth with the research efforts leading to this paper. We also wish to acknowledge funders of the original research discussed in this paper: NIMH K23 MH068405, AACAP Pilot Research Award, International Psychoanalytical Association Research Advisory Board Grants, and Support of the Sackler Institute of Developmental Psychobiology at Columbia University to the first author. Data upon which this paper was based were presented as part of a Plenary Panel Presentation at the Conference on Interdisciplinary Responses to Trauma, Emory University, April 30, 2005.