All biological mothers (66) and their children ages 8-50 months who were registered or presented for evaluation in a hospital-based mental health clinic for very young children (ages birth to 5 years) and their families “The Infant-Family Service at New York-Presbyterian Hospital.” Families, who were registered between January, 2000 and December, 2001 were offered the opportunity to participate in the study by their assigned therapist. The therapists were either a psychologist or social worker and not a member of the research team. Medical and mental health professionals, daycare centers, and other community agencies referred families to this clinical service for concerns about potential child maltreatment and 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. Referrals to the clinic most often involved child-related concerns, yet around one-fourth of cases presented with primary concerns about parental mental health, trauma, or severe stress.
Women with active psychotic symptoms, substance abuse, or who where not the primary caregiver for their child for most of that child’s life, were excluded.
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, one caregiver disclosed that she was the child’s grandmother and not the biological mother and one mother did not return after the first visit.
The 43 remaining mothers completed two videotaped visits that were 1-2 weeks apart. These mothers all shared a chief complaint involving violent, aggressive, or disruptive behavior of their preschool-age child or of another household member. However, only forty-one (95%) of them stated that they were exposed 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 the remaining N of 41 who were included in final analyses.
Out of these 41 mothers and their children who presented for both the Maternal Assessment and Interaction Visits, 32 returned for the clinician-assisted videofeedback exposure session (CAVES) two to four weeks later. The 9 mothers and children who did not return did not differ from those that did return in any statistically significant way in terms of demographic variables, interpersonal violence exposure severity, and psychopathology, involvement with child protective services, baseline level of reflective functioning treatment history, or baseline level of negativity of maternal attributions. Data pertaining to the 32 mothers who completed the CAVES are considered below and are the focus of the present paper. With the sample size limited to 32, we were able to detect effect sizes larger than 0.6, at a two-tailed significance level of <.05 at a power of 81% when the correlation between the pre- and post-measures was 0.325.
The 32 mothers ranged in age from 19-45 years, with an average age of 30 years. Their participating children ranged in age from 8-50 months, with an average age of 32 months. The sample was 88% Hispanic, largely of Dominican or Puerto Rican origin. Most were immigrants (61%). The remaining mothers were of American birth and similar descent (39%); 12% were African-American. Fifty-two percent of the mothers had less than a high-school education; over 75% of the mothers received public assistance or were eligible for it; and 67% were single mothers. Forty-one percent of mothers were currently or had previously been involved with child protective services regarding the child participating in the study or another child in the home.
Participation in all phases of the study was, however, voluntary and by informed consent as approved by the Institutional Review Board at Columbia University and the New York State Psychiatric Institute. Mothers received financial compensation for each visit as well as a book and age-appropriate toy for their child.
Mothers and their young children participated in three visits conducted by the same clinician (male) and research assistant (female) at each visit: Visit 1 (T1) was a 2.5 hour long Maternal Assessment Visit
that consisted of administration of a demographic and treatment history questionnaire, a measure of maternal mental representations (Working Model of the Child Interview [WMCI], Zenoah and Benoit, 1995
), life events history, and posttraumatic stress symptoms.
One to two weeks after T1, mothers and children returned for Visit 2 (T2), the Interaction Visit
, which focused on the following modified Crowell Play Procedure (Zeanah, Larrieu, Heller, & Valliere, 2000
): Free play (10 mins), separation-reunion #1 (5 mins), clean-up (5 mins), structured joint-attention task (5 mins), and separation-reunion #2 (5 mins). Following this play procedure, content items from the mental representations measure were repeated along with self-report measures regarding psychiatric symptomatology.
One month after the Interaction Visit, mothers were asked to return for Visit 3 (T3), the Clinician Assisted Videofeedback Exposure Session (CAVES). The CAVES is both a research-assessment measure as well as an experimental intervention that involves an active stance by a clinician. The principal investigator and/or research assistant carefully reviewed the videotape of Visit 2 and selected four 30-second excerpts for joint parent-clinician review. The four moments chosen were the following ones presented in the following order: an optimal moment (i.e. containing the most joy, spontaneity, joint attention, and mutuality or transaction), a moment of separation when mother is not in the playroom, a moment of reunion when mother returns, and a moment of sub-optimal play (i.e. when the clinician feels that “things were not working so well” relative to the rest of the interaction paradigm).
After the clinician asked mother to spontaneously recall her memories from the Interaction Visit, he then asked the mother to focus on the first videotaped excerpt. This first excerpt, while not announced as such, is the “optimal moment” which is selected as described above. Following this excerpt, and repeated after each of the subsequent three excerpts, the clinician posed to the mother a series of questions and probes:
“Tell me what happened there. Tell me the story of what happened in that moment. What do you think was going on in your child’s mind? …In your mind? What were you feeling then? What were you feeling as you were watching the moment with me? What was your child feeling? Why do you think I chose this moment for us to watch?
In a positive, supportive manner, the clinician underscored whatever capacities the parent demonstrated during the interaction, in terms of engaging her child, following her child’s lead, picking up child cues, responding with joy, spontaneity and mutuality (e.g. “I thought this was a good example of one of those moments in which you and your child were really getting along well together. And I thought you and I might think about what made this moment so much fun.”).
The clinician then administered a content-item from the WMCI, “Choose five words (adjectives) that describe your child’s personality.” The clinician then compared the five words that mother listed to describe her child’s personality during the maternal assessment and interaction visits.
The mothers were in no way prompted for a response but rather were asked to pick five words that described the personality of their child as they saw them in the video. This opportunity was given at the end of each excerpt. If a mother changed how she would describe her child as compared to what she said during Visit 1, the clinician would ask her, “What changed? Do you think that your child changed or that your feelings changed in relation to your child?
The clinician asked further: “Whom does your child remind you of in this excerpt? Does this moment remind you of any specific moments in your own life? On a scale of 1 to 10, 1 being the easiest, and 10 being the hardest, how was this moment to watch? And why?”
Measures used in the study were:
In Visit 1 (T1):
For assessment of background information and life events history
A Demographic and Treatment History Questionnaire (DTHQ) consisting of 33-items was designed by the authors.
Life Events Checklist (LEC) (Johnson, 1992) is a standard 19-item self-report checklist that asks the subject to rate her experience of a range of life stressors from natural disasters to war-exposure, to the death of a loved one, to physical assault and rape.
Brief Physical and Sexual Abuse Questionnaire (BPSAQ) (Marshall et al., 1998
) is a 12-item questionnaire, administered for our study by the clinician. The BPSAQ assesses for the presence or absence of a history of traumatic events before age 16. These items, as well as maternal history of physical and/or sexual assault during adulthood, contributed to a composite Violence Exposure Severity Score. For purposes of validation of this measure, this score was correlated against a comparable, validated measure the Traumatic Life Events Questionnaire (TLEQ) (Kubany et al., 2000
) (r=.79, p<.001) and the Clinician Administered PTSD Scale (CAPS) (Blake et al., 1995
), a standard, well validated measure of PTSD, (r=.69, p<.001) within a new sample of 50 mothers with children ages 12-48 months from primary care sites in the same community in which the pilot study was done (Schechter et al., 2005
—For more information on the derivation of this score see Schechter et al., 2005
, p. 318).
Determination of exposure to interpersonal violent trauma was based on these three measures. The interpersonal violent event considered most traumatic to the participant mother then became the DSM-IV Criterion A focus for the following measures of PTSD. The most severe past episode of PTSD following the index violent event and then the participant’s self-reported symptoms within one month prior to the current assessment were the focus of the measures.
For assessment of posttraumatic stress disorder
Structured Clinical Interview for the DSM-IV PTSD Module (SCID) (First, Gibbon, et al., 1995) is a well-established semi-structured interview for making the diagnosis of PTSD.
For assessment of current depressive symptom severity
Beck Depression Inventory (BDI) (Beck, 1996
) is a brief, self-report questionnaire for evaluating the severity of depressive symptoms with good psychometric properties.
For assessment of maternal reflective functioning
Reflective Functioning (RF) was measured by coding maternal narrative responses to the Working Model of the Child Interview (WMCI) on a scale of -1 to 9. The WMCI is a measure designed to elicit caregiver mental representations of the child and relationship with the child, the original coding scheme for which was not applied for the study in this paper (Zeanah & Benoit, 1995
The RF scale was originally developed for use in coding narratives from the Adult Attachment Interview (AAI) (Fonagy, Target, Steele, & Steele, 1998
). The AAI is a measure designed to elicit an adult’s mental representations of their caregiver(s) and their relationship with their caregiver(s) retrospectively (George, Kaplan, & Main, 1984, 1996
). The RF scale was adapted by Slade and colleagues (Slade, Grienenberger, Bernbach, Levy, & Locker, 2005
) for use with the Parent Development Interview (PDI; Slade, Belsky, Aber, & Phelps, 1999
), which like the WMCI, is a measure of caregiver mental representations of the child and relationship with the child in the present. Like the WMCI, it is loosely modeled on the AAI (Slade, 2005
). Coding RF from parental narratives encompasses four essential constructs: 1) awareness of the nature of mental states; 2) the ability to infer mental states underlying behavior; 3) recognition of developmental aspects of mental states; and 4) awareness of mental states in relation to the interviewer. In recent studies of maternal RF, Slade et al., (2005)
provided validation for the construct of parental RF and its predictive validity vis-a-vis adult and child attachment security (p<.001).
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 to elicit thinking about mental states were included to render the coding condition comparable to that with the PDI. For example, after the WMCI item, “What of your child’s behavior is most difficult for you to manage?” After asking the mother for a specific example of such a behavior, we added the probe, “What do you think was going on in his mind when he did that?” 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 crazy.”
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 that pull for the mother’s thinking about mental states (i.e. “demand items”) using the RF coding manual, twenty to thirty minutes of videotaped WMCI responses to four non-content items (i.e. WMCI items 4, 5, 7, and 11) were transcribed by our research staff, and then coded by a co-author of the parental RF coding system at another institution. This coder was naïve to any information about the mother-child dyads except for the child’s age. A second co-author of the parental RF coding system, also naïve to any information about the participants, achieved reliability (>85%) with the primary coder on eight randomly selected WMCI transcripts.
For assessment of Maternal Attributions in Visits 1, 2, and 3 (T1, T2, T3) The Maternal Attributions Rating Scale (MARS)
We developed the Maternal Attribution Rating Scale (MARS) (Schechter, Brunelli, and Myers, Unpublished Instrument) to provide a continuous measurement of mothers’ responses. For the purposes of this study, we used the content items of the Working Model of the Child Interview (WMCI) (Zeanah and Benoit, 1995
) to access maternal attributions when talking to the traumatized mothers and used the MARS to rate the degree of negativity of the mothers’ responses.
To keep the MARS simple and brief, we chose to rate continuously only a single WMCI content item that states, “Tell me 5 words or short phrases (adjectives) that describe your child’s personality.” These descriptors were transcribed independent from any maternal narrative related to these descriptors and coded by four independent developmental specialist-raters naïve to the hypotheses or any information about the dyads except for the child’s age. The adjectives were coded along 5-point Likert scales for the dimension of negativity that has been predictive in previous studies (Lorber, O’Leary, and Kendziora, 2003
). We defined negativity as applying to descriptors of the child with a negative affective valence, expectation or judgment of the child regardless of whether it is an age-applicable descriptor or not (e.g. “bad,” “mean,” “spoiled”). A score of 5 (least negative) to 25 (most negative) was derived for each set of words using the MARS.
Interrater reliability on the MARS was good across ratings by the four coders (Intraclass Correlation Coefficient=.76, p<.005). Scores were significantly correlated across all three time points (T1, T2, and T3) (r=.60, p<.001).
To make sure that we were measuring a marker for non-balanced (i.e. negatively valenced and distorted classifications) on the WMCI, the MARS was validated against the WMCI classifications. Higher levels of negativity on the MARS were significantly associated with WMCI classifications coded by independent, history-naïve coders according to the WMCI coding scheme in the predicted direction (i.e. negativity on the MARS was significantly associated with non-balanced (i.e. more negative and distorted) WMCI classifications [ANOVA: F(1,39)= 4.93, p<.05] ).
Using the Posttraumatic Symptom Checklist—Short Version (PCLS) and Structured Clinical Interview for the DSM-IV (SCID) PTSD Module, the rate of current PTSD among these 32 mothers was 63%. Using the PCLS as a continuous measure, out of a possible score range of 17- 85, the sample’s range at baseline (T1) was 18-75 with a mean of 42.9 and SD of 16.1.
The rate of lifetime PTSD using the SCID PTSD Module was 91% (n=29/32), with the remaining 9% (n=3/32) 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.
The mean severity of current depressive symptoms on the BDI was 14.3, S.D. 8.6. The possible score range was 0-36. Within the sample, the range was 0-34.
Reflective Functioning, RF, relative to continuous measures of maternal trauma severity and psychopathology showed a narrower range. While the possible score range was -1 to 9, in this sample the range was 0-5 with a mean of 3.3, and a SD of 1.7.
While the possible range on the MARS as a continuous measure of maternal negative perception was 5– 25; in this sample, the range, mean and standard deviation for negativity at baseline (T1), the Interaction Visit (T2), and the Videofeedback Visit (T3) are depicted in . Variance within the dimension of negativity and pattern of reduction are noted in this table.
Reduction of Maternal Negativity on the MARS across Three Time-Points
Hypothesis 1: Change in the degree of negativity of maternal attributions was tested using a repeated-measures analysis of variance. We then applied the most conservative F-test called the Greenhouse-Geisser Correction to this analysis to correct for the effect of correlated error in the repeated-measures design. When the main effect for time was significant, we performed follow-up paired t-tests (T1 versus T2, T1 versus T3, and T2 versus T3).
Hypothesis 2: The effect of RF was evaluated by performing a regression analysis. Degree of negativity of maternal attributions at T2 was regressed on the same measure assessed at baseline and RF. Inclusion of the baseline assessment of negativity substantially increases the power to detect effects of RF. The distribution of RF was skewed. We performed the regressions using the logarithm of RF to adjust for the skewness. The results did not substantially change.
All analyses were conducted in SPSS. All tests were two-tailed with the Type I error set at 5%. Control for multiple comparisons was not performed as the two hypotheses were specified a priori and may be considered independent.