The role of childhood parent figure loss in the etiology of adult depression: findings from a prospective longitudinal study

Abstract

The underlying question of this study is whether childhood parental loss between infancy and sixth grade is a predictor of adult depression at age 26 years when a rating of loss severity is used. The loss rating considered the length of the separation/loss, the familiarity of substitute caregivers, the primary or supporting role of the lost parent figure, and traumatic features of the loss. The study also investigated the role of gender, developmental timing of the loss, life stress, SES, prior parental care and attachment history, and follow-up family relationships in the pathways between loss and depression. Results are reported from a prospective longitudinal study of children (N = 164) born into poverty. Measures were collected prenatally through age 26 years and included multiple methods and multiple reporters. Results indicated that the most robust predictor of adult depression was loss history between 5 years old and grade 2. Earlier and later measures of loss were not related to adult depression. However, intervening loss experiences predicted change in depression scores from childhood to adulthood. Loss continued to predict adult depression after controlling for SES, maternal life stress, participant life stress, gender, early caregiving, and follow up family functioning. This study found no significant gender differences. These results suggest that loss is a risk factor for adult depression for both boys and girls and that the quality of early and later caregiving do not entirely buffer children from the effects of parental loss.

Introduction

Long-standing psychological theory (Bowlby, 1980; Freud, 1917/1994) and recent empirical work regard both childhood parental loss and separation as risk factors for contemporaneous and delayed manifestations of depressive symptomatology. Theorists have proposed multiple pathways that identify loss as etiologically significant for the development of depression. According to psychoanalytic theory, loss leads to self-directed anger that manifests as depression (Freud, 1917/1994). Loss of an attachment figure was central in Bowlby’s account of depression (Bowlby, 1980). Bowlby and other attachment theorists posit that parental loss increases the probability of other contextual challenges and also decreases the child’s resilience to future adversity (Bowlby, 1982). This impaired ability for children to cope with adversity or to view themselves as deserving support creates a feeling of hopelessness/helplessness that predisposes children to adult depression (Bifulco, Harris, & Brown, 1992; Bowlby, 1977, 1980). Bowlby (1980) further suggested that the circumstances surrounding a loss influence its developmental salience, and that unresolved mourning, due to the child’s age or understanding of the loss, would impact the child’s representation of the lost figure and, therefore, long-term outcomes.

Compromised resources may mediate the relation between childhood loss and adult maladaptation (e.g., Maier & Lachman, 2000). Still others qualify that specific factors determine the developmental significance of loss including gender (e.g., Maier & Lachman, 2000; McLeod, 1991; Rodgers, 1994), developmental timing (Barnes & Prosen, 1985; Heinicke, 1973), caregiving quality (Bifulco et al., 1992; Harris, Brown, & Bifulco, 1986), and type of loss (for review, see Canetti et al., 2000).

Multiple empirical and review papers support an overall association between loss and depression (e.g., for reviews, see Lloyd, 1980; Nelson, 1982; see also Jacobs & Bovasso, 2009; Tyrka, Wier, Price, Ross, & Carpenter, 2008). In an early study, Brown (1961) reported higher rates of depression for individuals in clinical samples who had lost a parent. Foundational work by Brown, Bifulco, Harris and colleagues found a relation between childhood loss and adult depression. Various studies by this group concluded that mother loss was a risk factor for adult depression (e.g., Brown, Harris, & Copeland, 1977; Harris et al., 1986). More recent empirical work echoes these findings (Kendler, Sheth, Gardner, & Prescott, 2002; Kunugi, Sugawara, Aoki, Nanko, Hirose, & Kazamatsuri, 1995; Takeuchi et al., 2002).

However, there are also strong objections to the claim that childhood loss is a risk factor for adult depression (for review, see Crook & Eliot, 1980; Furukawa, Ogura, Hirai, Fujihara, Kitamura, & Takahashi, 1999; Tennant, Bebbington, & Hurry, 1980). According to Crook and Eliot (1980), methodological flaws have accounted for observed relations between parental loss and adult depression. The instability of the link between loss and depression has led others to assume that additional factors may be confounding or mediating the association (Kitamura, Sugawara, Shima, & Toda, 1999).

Various studies have compared the impact of parental death and separation on child adjustment (for review, see Canetti et al., 2000; Hallstrom, 1987; Kendler, Neale, Kessler, Heath, & Eaves, 1992; Maier & Lachman, 2000; Oakley-Browne, Joyce, Wells, Bushnell, & Hornblow, 1995; Tennant, 1988). Upon detecting differences between parental death and separation, investigators have hypothesized causes for the distinct outcomes. However, in most cases, these explanations are post hoc, and in some instances contradictory.

Rutter (1995) provides a theory to unite these disparate findings. The circumstances surrounding the loss or separation may explain the process by which an isolated event becomes a turning point in a developmental pathway. Specific factors proposed to influence the developmental salience of a loss or separation include the alternative caregiving following the loss (Harris et al., 1986), the contextual circumstances surrounding the loss (Tennant, 1988), the traumatic circumstances causing the loss (Harris et al., 1986), the number of separated caregivers (Canetti et al., 2000), and the quality of care preceding loss (Bifulco et al., 1992).

Multiple researchers have suggested that the inconsistency in the loss literature is due to both the heterogeneity of loss experiences and to methodological variation. It may not be loss per se, but the circumstances surrounding the loss that are of predictive value (Crook & Eliot, 1980; Harris et al., 1986; Tennant et al., 1980). Moreover, loss has been defined in various manners (for discussion, see Furukawa et al., 1999), with some researchers employing separations of a month as sufficient for a loss experience (e.g., Oakley-Browne et al., 1995), while others use a year as the minimum criterion (e.g., Bifulco et al., 1992; Brown et al., 1977), and still others solely consider parental death.

Previous researchers have been limited by their samples. Primarily, existing research encounters the persistent methodological problem of using adult, retrospective self-report for childhood experiences of loss. This reporting problem is especially salient when the outcome of interest is an affective disorder. Some studies compare clinical and general populations; however, depressed individuals tend to report more incidences of separation (Takeuchi et al., 2002). In these studies, it is questionable whether biased reporting or causal relations drive the findings especially if childhood life stress and other surrounding circumstances are derived from retrospective report. Retrospective report makes it difficult to determine if a recalled factor is cause or a consequence of an affective disorder episode (Johnson, Andersesson-Lundman, Aberg-Wistedt, & Mathe, 2000). Therefore, researchers have called for prospective studies that sample from the general population (Takeuchi et al., 2002). The aim of this study is to prospectively examine the impact of childhood loss and separation on adult depressive symptoms.

Due to inconsistencies in the literature, a foundational question remains: is childhood loss a risk factor for adult depression? Therefore, the underlying question this study seeks to answer is whether childhood parental loss is a robust predictor of adult emotional adaptation when a rating of loss severity is used. This study investigates separations from and losses of both biological parents and non-biological parent figures from infancy through sixth grade. The coding of loss severity relied on multiple factors that researchers have hypothesized to account for findings: the length of the separation, the quality of alternative caregiving, the primary or supporting role of the parent figure, and other circumstances surrounding the separation.

A second question concerns developmental timing. Are there crucial developmental periods when loss experiences are of more etiologic significance? Researchers have limited their investigations of loss to certain age periods; however, these cut-offs vary among studies. Some have included all loss experiences before 16 years old (e.g., Luecken, 2000), before 11 years old (e.g., Brown et al., 1977), or before 17 years old (e.g., Harris et al., 1986; Kunugi et al., 1995). Some studies present significant findings at certain age periods while still other studies find that child’s age at parental loss does not relate to risk for developing adult depression (Jacobs & Bovasso, 2009). Early work by Heinicke (1973) concluded that father loss between birth and 5 years old and also between 10 and 14 years old was related to the development of adult depression. Barnes and Prosen (1985) also found that father loss between birth and 6 years old and also between 10 and 15 years old was significantly related to maladaptation. Brown and colleagues (1977) distinguished between parent loss occurring in two time periods: before 11 years old and between 11 and 17 years old. Using a general population sample of women from London, they found that loss of a mother before 11 years old increased the risk of adult depression while parent loss between 11 and 17 years old did not.

This present study considers loss up to approximately 11 years old (Brown et al., 1977). There were theoretical and practical reasons to limit the consideration of childhood loss experiences to 11 years old. Age 12 can be viewed as the end of childhood and beginning of adolescence. In addition, in this high-risk sample, it can be difficult to determine whether parents or adolescents initiate parent–child separations in adolescence. Theorists have suggested that childhood parental loss causes feelings of hopelessness/helplessness in children that in turn increases the likelihood that they will develop depression (Bifulco et al., 1992; Bowlby, 1977, 1980). Adolescents who initiate parent–child separations may have distinct pathways of psychopathology. To investigate developmental timing, participants’ loss histories are divided into three distinct periods: infancy and preschool years (12 to 54 months), early childhood (60 months to second grade), and middle childhood (third grade to sixth grade).

An additional question concerns whether gender moderates the relation between loss and depression. Some studies suggest that women are more vulnerable to the depressogenic effects of parental loss (e.g., Maier & Lachman, 2000; McLeod, 1991; Rodgers, 1994). However, others suggest that there are no gender differences in the risk for major depression in response to any kind of parental loss (Kendler et al., 2002).

As separations often reflect home environments characterized by instability and other family difficulties, analyses control for socio-economic status and stressful life events to explore the unique role of parental separations and losses in developmental trajectories. Based on previous research (e.g., Harris, Brown, & Bifulco, 1987), this study investigates the role of stress and socio-economic status (SES) in the pathways between loss and depression. Life stress and SES are putative mechanisms that account for the relationship between loss and depression (Crook & Eliot, 1980). Some researchers suggest that loss or divorce is simply a proxy for life stress (e.g., Canetti et al., 2000). Researchers hypothesize that loss and divorce initiate pathways of diminished social and financial support that in turn lead to poor adult outcomes (for discussion, see Maier & Lachman, 2000). The immediate stressors following loss interfere with education achievement (Maier & Lachman, 2000). Others have hypothesized that SES may mediate the relationship between loss and depression. Barnes and Prosen (1985) proposed that parental loss causes financial hardships that in turn lead to multiple forms of adult adversity. Harris, Brown, and Bifulco (1990) found that child loss increased the vulnerability to adult depression when individuals confronted stress in adulthood. Other work by this group indicated that childhood maternal loss and the quality of care following the loss related to social class position in adulthood (Harris et al., 1987).

Parenting has been hypothesized to contribute to resilience in the face of loss (e.g., Bowlby, 1980; Harris et al., 1986). This study investigates how the quality of prior parental care and attachment history (measured prospectively, independently, and by observation) moderate the relation between childhood loss and adult depression. Some theorists (e.g., Bowlby, 1980; Rutter, 1984; Tennant, 1988) have suggested that adverse parenting before and after the loss explains the link between loss and adult depression. Indeed, Bowlby proposed that attachment history would predict how an individual would respond to a loss (see Shaver & Fraley, 2008, for discussion; Bowlby, 1980). Empirical work by Bifulco et al. (1992) concluded that inadequate care prior to the loss related to adult affective disorders. However, inadequate caregiving was inferred from the age of the child at the loss or from the type of separation. They coded inadequate caregiving in a group of children whose parents died before the child was 6 years old. However, one assumes some variation in caregiving patterns within that group of parents. Other work echoes these findings. Using retrospective self-report, Oakley-Browne and colleagues (1995) concluded that loss was not related to adult depression when quality of early parenting is controlled.

Likewise, the literature coalesces to suggest a related question: does the caregiver child relationship following the loss moderate the relation between loss and adult depression? Bowlby (1980) proposed that replacement parenting would influence the long-term consequences of a loss. Others have concurred that replacement or substitute parental care moderates the relationship between loss and depression (Bifulco et al., 1992; Harris et al., 1986; Tennant, 1988). Saler and Skolnick (1992) found that children who experienced loss became better-adjusted adults if they had higher quality relationships with the surviving parent. Using a self-report measure in a sample of college students, loss considered with low-quality family relationships predicted depression, while loss was not a significant predictor by itself (Luecken, 2000). Other researchers have complementary findings. Canetti and colleagues (2000) found that the quality of the parental relationship buffered the negative impact of separation from them. Oakley-Browne and colleagues (1995) showed that low maternal care predicted depression beyond loss experiences. In addition, Bifulco and colleagues (1992) found that lack of adequate care after mother loss lead to adult depression.

While this body of work is compelling, the studies have methodological limitations. They rely on retrospective self-reports of parenting or infer quality of caregiving from contextual factors. The Minnesota Longitudinal Study of Parents and Children is uniquely suited to investigate the role of parenting, contextual factors, and demographic indicators in developmental pathways. By using multiple reporters, and multiple measures, information is not compromised by same reporter bias or from retrospective reporting. Parenting and attachment history were measured prospectively, independently, and by observation. This study is the first to investigate the role of parenting both before and after loss experiences using prospective, independent, and observational measures of parenting quality.

Method

Participants

The participants in this study were part of a prospective, and currently ongoing, longitudinal study of mothers and their firstborn children who were recruited from Minneapolis public health clinics (Egeland & Sroufe, 1981). The study began when the mothers were in the third trimester of pregnancy. This community sample was considered to be at high risk due to poverty. It was representative of Minneapolis urban poor in the mid-1970s; all mothers qualified for public assistance for prenatal care; 62% of the mothers were single, 86% of the pregnancies were unplanned, and 40% had not completed high school. At the time of delivery, the mothers ranged in age from 12 to 34 years (M = 20.52, SD = 3.63). At birth, 58% of the children were Caucasian, 14% were African American, 16% were mixed race, 3% were Native American or Latino, and the remaining 9% had missing paternal data. At the age 26 years assessment, 67% of the participants were Caucasian, 10% were African American, 18% were mixed race, 2% were Native American or Latino, and the remaining 3% had missing paternal data. Subject attrition occurred primarily in the first 18 months (from 267 originally recruited participants in the prenatal period to 195 at 18 months). At the earliest observational point of the present study (12 months) there were 212 participants and at the age of the outcome variable (26 years) there were 164 participants (83 men and 81 women). Thus, there was a 77% retention rate across this period.

Measures

Separation

This measure was designed to assess and document separations from and losses of both biological parents and non-biological parent figures from infancy through sixth grade. Various types of separations and losses of parent figures were coded including losses due to death, abandonment, foster care, hospitalization, job placement, and divorce. Criteria used to evaluate the severity of the separation experience included duration, familiarity with substitute/replacement caregivers, and traumatic events surrounding the loss experience. This information was gathered from teacher, parent, and child interviews and questionnaires made in each age period. Relevant information was asked directly on certain measures and indirectly obtained from open-ended interview supplements.

Biological relation was not sufficient to be considered a parent figure. Rather, there had to be evidence that the person held a significant caregiving role prior to the separation. In most cases the mother was considered if she was in fact the primary (or secondary) caregiver. In other cases, biological and social fathers were considered. In still other cases, foster parents, grandparents, or other figures who adopted caregiving roles akin to a parent, were included in the index. As a note, it is the type of adult–child relationship rather than the quality of the caregiving that determined whether the person was considered a parent figure. The identified primary caregiver for each assessment was identified as a parent figure; however, temporary caregivers (e.g., short-term foster parents) were not coded. In addition, variables that indicated which adults provided caregiving were reviewed. Biological parents were considered as long as they were involved in the caretaking or decision making for the child. Other adults who provided caregiving were considered if they met certain criteria: involved with the child for a sufficient amount of time (at least a year), significantly involved in the caregiving or decision making in childcare in an ongoing way, and not simply a paid caretaker (e.g., babysitter). Ultimately, when roles were uncertain, cases were conferenced among coders. To determine roles, the caregiving history prior to and following the current age period was reviewed.

Specific questions that pulled for content related to separations, loss, abandonment, out of home placement, and substitute caregiving were compiled. Trained coders reviewed all the raw files chronologically with special attention paid to the relevant questions. In addition, all unstructured interview summaries were reviewed. If information was unclear, contradictory, or vague, the coder reviewed all information available for that particular assessment.

Information related to a loss was then extracted and recorded on a coding sheet. Each loss was described individually. Transferred details included the child’s age at the time of the loss, the role of the separated parent (e.g., biological parent, relative, long-term foster parent), the length of the separation/loss, the circumstances surrounding the loss (e.g., reason for the separation), and the alternative caregiving provided during or after the loss.

Based on the synthesized loss data, loss scores were determined for each age period. A team of trained coders referred to the compiled information to determine a loss score for each age period. Coders were blind to additional information about participants. A participant’s separation history was divided into three distinct periods: infancy and toddlerhood (12 to 54 months), early childhood (60 months to second grade), and middle childhood (third to sixth grade). No data were collected between 55 and 59 months; however, if a loss occurred between 55–59 months and was reported during the 60-month data collection, it was coded in the early childhood period. For each developmental period, a child was given a score on a 7-point loss severity scale. In order for a participant to be scored, adequate information had to be available. However, participants could be scored for one period and not others (e.g., a participant completed all assessments in early childhood but not in the elementary school years). The coding of the severity of separation experiences relied on multiple factors that researchers have proposed to account for relations between loss and developmental outcomes: the length of separation, the relationship with substitute caregivers, the primary or supporting role of the lost parent figure, and other circumstances surrounding the separation. Multiple factors cause a higher rating on the loss index: a longer duration, traumatic features, unfamiliar substitute caregivers, and multiple losses (see Appendix 1).

Three coders were trained on sample cases and a subsample of 33 cases was coded to determine reliability. Results indicated that Cronbach’s alpha was. 98 for the infancy and preschool loss period,. 99 for the elementary school loss rating, and. 99 for the middle childhood loss rating. All cases were coded independently by at least two coders and sometimes by all three coders and then conferenced to prevent coder drift. In the case of a score discrepancy, all three coders reviewed the case and a conference score was reached by group consensus.

Investigation of data distribution revealed small cell sizes and positive skewness. Results indicated that the skewness of loss in infancy, early childhood, and middle childhood was .53, 1.02, and 1.03, respectively. Kurtosis was −.93, −.36, and .58, respectively. The 7-point scale includes a natural breaking point into which the scale can be dichotomized based on the permanence of loss or the length of the separation. Permanent loss of a parent figure is restricted to scale points five, six, and seven. In addition, the higher end of the scale (i.e., scale points five, six, and seven) is restricted to long separations (i.e., over 4 months) or severely stressful medium separations (i.e., at least 1 month but which include traumatic circumstances). Therefore, the scale was dichotomized into not severe/severe loss history. Scale points 1 through 4 indicate a not severe loss history and scale points 5 though 7 indicate a severe loss history.

Depressive symptomatology

The Teacher’s Report Form (TRF) and Youth Adult Self Report Form (YASR) are used widely to measure behavior problems and yield scores for total problem behaviors, two broad-band scales: internalizing and externalizing, and symptom subscales (Achenbach, 1991, 1997; Achenbach & Edelbrock, 1986; Achenbach & Rescorla, 2001). The TRF and YASR have been normed on large samples, separately for gender and ages. The TRF and YASR contain problem items, rated on a 3-point, Likert-type scale. Items are rated as (0) not true, (1) somewhat or sometimes true, or (2) very true or often true. The following summarized items compose the TRF anxiety/depression scale: cries a lot; fears things; fears schools; fears doing bad; must be perfect; feels unloved; feels worthless; is nervous, tense; is fearful, anxious; feels too guilty; is self-conscious; talk or thinks of suicide; worries; hurts when criticized; is anxious to please; and is afraid to make mistakes. The following summarized items compose the YASR anxiety/depression scale: is lonely; cries a lot; fears impulses; needs to be perfect; feels unloved; feels worthless; is nervous, tense; lacks self-confidence; is fearful, anxious; feels too guilty; is self-conscious; is unhappy, sad, depressed; worries; is confused; worries about the future; is concerned about looks; worries about the opposite sex.

Early childhood depression is based on the first available child depression measure for full day school attendance: grade 1 (completed by the child’s classroom teacher). Adult depression is based on the last available self-report Achenbach measure: the YASR at age 26 years. Mother reports of depression/anxiety behavior were not included in order to avoid confounding data as the mothers were the primary reporters for the control variables and often the identified separated caregiver in the loss rating.

Socio-economic status (SES)

An SES score was computed for each family based on a series of questions about employment, income, and education. Occupational status was determined using the Revised Duncan Socioeconomic Index (Duncan, 1961; Stevens & Featherman, 1981). At each time point, a Duncan score was assigned to the mother and if available to her partner as well, and the highest score was selected to represent the family. Mother’s education level, family income level, and the Duncan score were each standardized and their mean is the family SES at each time point. SES was computed at the following assessments: prenatal, 42 months, 54 months, grade 1, grade 2, grade 3, grade 6, and 16 years old. An average of all available SES scores was computed across childhood. This was done to consider the two putative mechanisms relating loss to maladaptation: that loss may simply be a proxy for prior and current SES and that loss initiates pathways of compromised resources and opportunities.

Life stress

Maternal life stress was assessed using the Life Events Scale (LES) (Cochrane & Robertson, 1973) and adapted by Egeland and Deinard (1975) for use with a low-income population. This scale consists of 40 items related to financial, health, personal, and social stressful life events. Each item is scored on a 3-point scale for its degree of disruptiveness to family functioning and a total life stress score is computed by summing all items. The Life Events Scale was used to measure life stress at all assessments except the 19 years old assessment.

Adolescent life stress was assessed using the Adolescent Life Events Scale (ALES): an adaptation of the Adolescent Perceived Events Scale (APES) (Compas, Davis, Forsythe, & Wagner, 1987). This scale consists of 96 negative, stressful events, drawn from the APES, which range from relatively minor to major events. Participants indicate whether the event happened to them in the last year, what the impact was (i.e., very little, moderate, or extreme), and the frequency (i.e., at least once a month or less than once a month). Total life stress is the sum of the products of frequency and impact for each item. The ALES was used when participants were 19 years old.

Maternal life stress is the mean of the standardized life stress scores at the following time periods: 12 months, 18 months, 30 months, 42 months, 54 months, 64 months, grade 1, grade 2, grade 3, grade 6, 16 years old, and 17 years old. Participant life stress is the mean of standardized life stress scores at the following ages: 19, 23, and 26 years. The LES was used at all time points except 19 years old in which the ALES was used. Life stress measures across childhood were retained in order to address claims that loss is simply a proxy for prior and current life stress and that loss initiates increased exposure to stress which results in maladaptation. Life stress was split into participant and maternal life stress in order to investigate the concordance of maternal life stress and loss as well as to investigate the role of loss in future participant life stress.

Early maternal care

The early emotional support provided for the child by the mother is based on four observational assessments done at 12, 18, 24, and 42 months. At 12 and 18 months, the infant–mother attachment relationship was assessed using the Strange Situation (Ainsworth, Blehar, Waters, & Wall, 1978). The strange situation is a short laboratory procedure from which the child’s behavior toward the caregiver can be categorized as secure or insecure. For each age, two coders independently rated the entire sample. Rater agreement was 89% for the 12 month assessment and 93% for the 18 month assessment. The number of times a dyad was coded as secure was counted and included in the composite rating.

At 24 months, emotional support was assessed in another laboratory procedure in which the toddler is presented with a series of four increasingly difficult problem-solving tasks that ultimately exceed toddler’s developmental capabilities. The children cannot complete the tasks independently but require parental assistance to finish (Matas, Arend, & Sroufe, 1978). This videotaped assessment was coded based on several scales, including a 7-point scale measuring the extent to which the mother served as a supportive presence to the child. A high score indicates that the mother provided emotional support, supported the child’s autonomy while providing appropriate help, and facilitated an enjoyable learning experience despite the tasks’ difficulty levels. Interrater reliability was .75 for this scale.

Similarly, at 42 months, emotional support provided by the mother was assessed in another laboratory procedure during which the mother and child were presented with a series of four teaching tasks. The mother was asked to provide the child with instructions and assistance as needed. The tasks are challenging and 42-month-old children require adult assistance in order to complete them. The procedure was videotaped and maternal behavior was scored on a number of 7-point scales including supportive presence. A high score indicates appropriate expressions of affection, high regard for the child, and a calm and reassuring teaching style, especially when the child encounters difficulties. Using a subsample of 87 cases, interrater reliability was .78 for this scale. Based on these four measures, a single variable indicating early childhood maternal support was created. Standard scores were computed for each scale and then combined. In the very rare cases where a mother died during this period, the participant completed these assessments with another parent figure, if available. While the Strange Situation measures infant behavior and the teaching tasks reflect maternal behavior, Ainsworth’s compelling work on the relationship between infant security and caregiving history (i.e. maternal sensitivity and responsiveness) allows us to combine these four measures into a composite of early maternal care. Selection of these scales and all other measures were made without reference to outcome data.

Adolescent family functioning

At the 13 years old assessment, mothers and children were given a series of collaborative tasks to complete. They completed four structured interaction tasks: plan an anti-smoking campaign, assemble a series of puzzles while the mother was blindfolded and the child guided the assembly, discuss the effects of two imaginary circumstances, and complete a Q-sort of an ideal person. The sessions were videotaped and several scales were coded including a balance scale that was designed, based on family systems theory, to assess how well dyads are able to fulfill multiple functions at the same time (Sroufe, 1991). The scale used in this study measures the balance between the goals of the individuals and the relationship. At the high end of the scale, the development of individuals is supported by the relationship and the members enjoy and take pride in the relationship. At the low end of the scale, maintaining the relationship demands the attention of the individuals but does not enhance their experience in the session. Two coders rated a subsample of 129 dyadic assessments (kappa = .78).

Results

Descriptive analyses

As noted earlier, the scale of loss severity was positively skewed in infancy (M = 2.72, SD = 1.76), early childhood (M = 2.29, SD = 1.72), and middle childhood (M = 2.19, SD = 1.76). Due to skewness, the dichotomized loss rating was used for most analyses. The percentage of participants who experienced significant loss in each developmental period was 21% in infancy, 18% in early childhood, and 14% in middle childhood.

Loss and the onset of depression

The intercorrelations among the variables appear in Table 1. The dichotomized loss rating (i.e., severe versus non-severe loss experience) in infancy and middle childhood was not significantly associated with early adulthood depressive symptomatology (r = .01, p = .90; r = .02, p = .79). However, the dichotomized loss index in early childhood was significantly related with adult depressive symptomatology (r = .20, p = .01). A t-test echoed these findings: children who had significant early childhood loss histories (M = 9.97) had significantly higher rates of adult depressive symptomatology, t(156) = −2.51, p = .01, than children who did not have significant loss histories in early childhood (M = 6.91).

Table 1.

Correlations for variables of interest (N = 164).

Variable	1	2	3	4	5	6	7	8	9	10	11
1. Gender	–	−.02	.08	.14*	−.01	−.12	−.13	−.02	−.00	−.07	−.04
2. SES			−.10	−.23***	.30***	.21**	.09	.10	−.17*	−.09	−.17*
3. Participant life stress			–	.15*	−.03	.04	.13	.30***	.03	.13	−.05
4. Maternal life stress				–	−.18**	−.19*	.10	−.03	.33***	.40***	.15*
5. Early maternal care					–	.21**	−.11	−.13	−.23**	−.12	−.11
6. Adolescent family functioning						–	−.03	−.02	−.02	−.06	.02
7. Grade 1 depression							–	.20*	.03	.05	−.10
8. Adult depression								–	.01	.20*	.02
9. Infancy lossa									–	.16*	.06
10. Early childhood lossa										–	.18*
11. Middle childhood lossa											–

Note:

^*p < .05;

^**p <.01;

^***p <.001.

^aThe dichotomized loss rating was used.

In order to further investigate the unique predictive power of early childhood, a follow up analysis was run without the inclusion of the participants who experienced significant loss in middle childhood. Seven children experienced significant loss in both early and middle childhood while 16 children experienced severe loss in middle childhood but not earlier. When children with significant loss in middle childhood were removed from consideration, children with significant early childhood loss histories continued to have higher rates of adult depressive symptomatology t(133) = −2.12, p = .04 (M = 9.88 and M = 6.91).

Developmental timing of loss

Results indicated significant correlations between certain loss ratings. Loss in infancy was significantly associated with loss in early childhood, r = .16, p = .04, and further loss in early childhood was significantly related to loss in middle childhood, r = .18, p = .01. A simple regression with adult depression as the outcome was done in order to determine whether loss in any time period predicted a significant amount of variance in adult adaptation beyond the variance shared by all three time periods after controlling for gender. Gender and the three loss ratings (i.e., infancy, early childhood, and middle childhood) were entered as the predictors. Beyond the variance in adult depressive symptomatology shared by all three time periods and after controlling for gender, infancy and middle childhood loss did not predict significantly more (β = −.04, p = .66; β = .01, p = .91); however, early childhood loss did (β = .19, p = .02). Because early childhood was the only time period significantly related to adult depressive symptomatology, additional analyses will be restricted to teasing apart the relations between early childhood loss and adult depressive symptomatology.

Early childhood loss index and different aspects of loss

Although a dichotomized rating was used for all regression and correlation analyses, the continuous loss index in early childhood was explored to consider how different types of loss related both to the loss index rating and to adult depression scores. Tables 2, 3, and 4 focus on early childhood loss due to the unique predictive power of this period. In this sample, death or separations over a year always resulted in a high score (i.e., 5, 6, or 7) on the loss index (see Table 2). In four cases, separations of less than 4 months resulted in a high score on the loss index due to traumatic circumstances surrounding the separation (see Table 2). In terms of the prevalence of certain types of loss, no participants lost a mother in early childhood, while 3 participants lost a father, and 13 lost another key caregiver due to death (see Table 2). Notably, this scale considers social fathers (i.e., stepfathers or father figures) in the coding of key caregivers. The distribution of mothers, fathers, and other caregivers was fairly consistent among children who experienced moderate length separations (i.e., 4–11 months) and long separations (i.e., 12 or more months) (See Table 2). Due to small cell sizes, analyses could not compare adult depression scores based on the type of early childhood loss. However, descriptive information appears in Table 3. In parallel, Table 4 presents the mean adult depression scores based on the early childhood loss index score. It appears that there is a general trend for mean adult depression scores to increase based on loss severity. Additional analyses explore this trend using the dichotomized loss rating.

Table 2.

Type of early childhood loss by loss index.

	Loss index score
Type of early childhood loss	Missing	1	2	3	4	5	6	7	Total
Missing	6	0	0	0	0	0	0	0	6
1. Death/permanent loss of contact mother	0	0	0	0	0	0	0	0	0
2. Death/permanent loss of contact father	0	0	0	0	0	2	1	0	3
3. Death/permanent loss of contact other key caregiver	0	0	0	0	0	8	5	0	13
4. Death of more than one of 1–3	0	0	0	0	0	0	0	1	1
5. Separation 12 or more months mother	0	0	0	0	0	0	1	0	1
6. Separation 12 or more months father	0	0	0	0	0	1	1	0	2
7. Separation 12 or more months other key caregiver	0	0	0	0	0	0	1	0	1
8. Separation 12 or more months of more than one of 5–7	0	0	0	0	0	0	0	0	0
9. Mixture death/separation among 1–7	0	0	0	0	0	0	0	0	0
10. Separation 4–11 months mother	0	0	0	0	0	2	1	0	3
11. Separation 4–11 months father	0	0	0	0	0	1	1	0	2
12. Separation 4–11 months other key caregiver	0	0	0	0	3	0	0	0	3
13. Separation of more than one of 10–12	0	0	0	0	0	0	0	0	0
14. No death or separations of 4 months	0	85	14	19	7	4	0	0	129
Total	6	85	14	19	10	18	11	1	164

Table 3.

Mean adult depression scores by type of early childhood loss (N = 164).

Type of early childhood loss	M	n
Missing	7.83	6
1. Death/permanent loss of contact mother		0
2. Death/permanent loss of contact father	5.00	3
3. Death/permanent loss of contact other key caregiver	9.54	13
4. Death of more than one of 1–3	7.00	1
5. Separation 12 or more months mother	20.00	1
6. Separation 12 or more months father	16.50	2
7. Separation 12 or more months other key caregiver	1.00	1
8. Separation 12 or more months of more than one of 5–7		0
9. Mixture death/separation among 1–7		0
10. Separation 4–11 months mother	16.67	3
11. Separation 4–11 months father	12.50	2
12. Separation 4–11 months other key caregiver	8.00	3
13. Separation of more than one of 10–12		0
14. No death or separations of 4 months	7.50	129

Table 4.

Mean adult depression scores by early childhood loss index score (N = 164).

Early childhood loss index score	M	n
Missing	7.83	6
1. No significant separation experience	7.29	85
2. Minor separation experiences	4.79	14
3. Moderate separation experiences	6.89	19
4. Notable separation experience	6.60	10
5. Significant separation experiences	9.56	18
6. Very significant separation experiences	10.91	11
7. Extremely significant separation experiences	7.00	1

Intervening loss and change in depression

An additional question addressed here was whether intervening loss experience, accounted for changes in depression problems between early childhood and adulthood. In this case, the goal was to see if depression in childhood was related to adult depression, r = .20, p = .01, and to investigate the etiologic significance of loss after controlling for earlier mood state. For this analysis, gender was entered in step one, grade one depression was entered in step two, and loss in early childhood was entered in the third step, as predictors of adult depression. Results indicated that gender accounted for 0% of the variance in adult depression, F(1, 153) = .02, p = .89, while grade 1 depression accounted for an additional 3% of the variance, β = .16, p = .04; F(2, 152) = 2.34, p = .10, and early childhood loss accounted for an additional 3.3% of the variance in adult depression, β = .18, p = .02; F(3, 151) = 3.39, p = .02. The total variance accounted for in adult depression by all of the variables entered into the regression equation was 6.3%. Both grade 1 depression and early loss predicted unique variance in adult depression. Children with significant early loss histories and higher early depression had higher depression scores in adulthood. Results appear in Table 5.

Table 5.

Summary of hierarchical regression predicting change in adult depression from intervening loss (N = 158).

	Step 1			Step 2			Step 3
Predictor	B	SE B	β	B	SE B	β	B	SE B	β
Gender	−.14	.99	−.01	.02	.98	.00	.25	.97	.02
Grade 1 depression				.25	.12	.17*	.23	.11	.16*
Early loss							2.84	1.23	.18*
R2	.00			.03			.06
Model F	F(1, 153) = .02			F(2, 152) = 2.34			F(3, 151) = 3.39*
FΔ in R2	.02			4.66*			5.36*

Note:

^*p < .05.

Predicting adult depressive symptomatology from childhood factors: life stress, SES, gender, loss

Results indicated that parental loss in infancy, early childhood, and middle childhood related to maternal life stress across childhood (r = .33, p = .000; r = .37, p = .000; r = .15, p = .04, respectively). Loss in infancy and middle childhood were significantly related to family SES (r = −.17, p = .02; r = −.17, p = .02, respectively) while early childhood loss was not, r = −.09, p = .23. Various factors may account for these correlations: similar risk factors may be measured by childhood loss, maternal life stress, and SES; loss may occur more often in high-risk families; or loss may initiate pathways of compromised resources and increased stress. The relation between parental loss in any time period and future participant life stress was not significant. This suggests that parental loss does not necessarily initiate pathways of increased personal, financial, and legal stress that persist into adulthood. There was a significant relationship between participant life stress and adult depression, r = .30, p < .000 (see Table 1).

A hierarchical regression was conducted to test whether separation and loss experiences could add significantly to the prediction of adult depressive symptomatology after controlling for additional variables that might plausibly account for the above correlations. Gender was entered in step one; participant life stress, maternal life stress, and SES were entered in step two; early maternal care and adolescent family functioning were entered in step three; and early childhood parental loss was entered in the fourth and final step. Results indicated that the effects of childhood separations and losses were evident even after accounting for life stress, demographic variables, and parenting measures. Parental loss predicted significant variance in adult depressive symptomatology above and beyond the control variables, β = .22, p = .01; FΔ(1,139) = 6.82; p = .01. The final model accounted for 14.1% of the variance in adult depressive symptomatology, F(7, 139) = 3.27; p = .00. Children who experienced significant losses had higher adult depressive symptomatology scores after controlling for life stress, SES, and parenting variables. These findings suggest that parental separations and losses are salient factors in the etiology of childhood depression and are not simply proxies for other developmental variables. Complete results appear in Table 6.

Table 6.

Summary of hierarchical regression analysis for variables predicting adult depression (N = 147).

	Step 1			Step 2			Step 3			Step 4
Predictor	B	SE B	β	B	SE B	β	B	SE B	β	B	SE B	B
Gender	.27	1.0	.02	.00	.98	.00	.01	.98	−.00	.36	.97	.03
SES				.10	.06	.14	.15	.07	.20*	.15	.07	.20*
Participant life stress				2.21	.71	.26**	2.32	.71	.27**	2.09	.70	.24**
Maternal life stress				−.35	.83	−.04	−.66	.84	−.07	−1.37	.87	−.14
Adolescent family functioning							−.32	.48	−.06	−.26	.47	.05
Early care							−1.18	.72	−.15	−1.07	.71	−.13
Early loss										3.56	1.37	.22*
R2	.00			.08			.09			14.1
Model F	F(1,145) = .07			F(4,142) = 2.93*			F(6,140) = 2.57*			F(7,139) = 3.27**
FΔ in R2	.07			3.88*			1.79			6.82*

Note:

^*p < .05;

^**p <.01.

Discussion

The literature on loss and depression is strikingly inconsistent. This study was able to investigate loss in a novel way. Rather than rely on retrospective reports of loss and childhood experiences, this study was able to document loss prospectively. In addition, this study was able to consider more detailed accounts of loss experiences. The available data from the larger longitudinal project allowed multiple features to be considered in determinations of loss severity. Ultimately, due to the skewed distribution of loss experiences the scale was dichotomized into severe and not severe losses.

The ideas of loss and separation have been central to attachment theory. Freud (1917/1994) and Bowlby (1980) provided the theoretical basis to explore the relations between loss and depression. Bowlby posited a compelling theory, suggesting that loss experiences have both immediate and long-term consequences. He suggested that childhood loss is a risk factor for adult depression when the individual faces adversity in adulthood. These theories have spawned a rich yet contradictory empirical body of work. Based on attachment theory, this study was broader than other studies in its consideration of parent figures. Rather than limit the loss index to biological parents, this study sought to include adults who adopted caregiving roles akin to a parent. It was hypothesized that the developmental implications of a loss would be influenced by the emotional salience of the caregiver rather than simply by the adult role. The observed relations between loss and adult depression provide initial support for this hypothesis especially considering that loss of caregivers other than biological parents was as common in this sample as loss from either biological parent.

There are two features that distinguish this cohort from those used in previous studies (Brown et al., 1977). As noted above, loss of non-biological parent figures was as common as loss of biological parents. In addition, separations were more frequent than parental death. In comparison to earlier studies (e.g., Brown et al., 1977), the rate of separations is higher while the frequency of parental death is lower. This potential cohort effect may be due to increased family disruptions, increased divorce rates, and improved health care. These differences should be considered when comparing this study to previous work.

These data suggest that childhood loss experiences are a factor in the etiology of adult depression. In this study, a significant developmental timing effect was present. The clearest predictor of adult depression was loss history between 5 years old and second grade. Earlier and later measures of loss were not related to adult depression. This may highlight an area of intervention. Early elementary school may be a time when children are especially vulnerable to effects of separation and loss. This contradicts Bowlby’s (1980) theory that loss in the earliest years is the most harmful. It is possible that sample specific characteristics diluted the power of the infancy loss rating. In our sample, family disruption is a regular occurrence in the early years. When biological fathers left the family in infancy, father–child contact was on average very infrequent, with half of these children having no contact with their fathers in adolescence. Therefore, many of these children received significant loss ratings in infancy and toddlerhood. However, it is possible that the developmental implications of loss depend on the length of time the parent was involved with the child before the loss and the developmental level of the child. For instance, the developmental implications of a secondary caregiver severing contact with a child at 12 months and at 8 years may be distinct. Because a single scale was used to rate losses irrespective of the age of the child, this study is unable to address that question.

It may also be the case that the developmental timing of the loss relates to vulnerabilities in specific domains. For instance, in this study, middle childhood loss was not related to adult depression. However, it is plausible that certain age-salient developmental tasks of this period would be compromised. These might include theoretically indicated domains such as the development of close friendships or academic initiative (Sroufe, Egeland, Carlson, & Collins, 2005). In this study, depression was chosen a priori as the outcome variable based on the extensive body of existing theoretical and empirical work. Future studies should consider the relation between the developmental timing of the loss and specific age-salient developmental tasks.

Intervening loss experiences predicted change in depression scores from childhood to adulthood. While childhood depression was significantly related to adult depression, intervening loss experience significantly added to the prediction of adult depression. This finding supports the significance of loss in pathways of depression.

Analyses suggested that loss and maternal life stress were related. Parental loss in all three age periods were related to maternal life stress across childhood. In addition, loss in the earliest and last time period was related to family SES. Various processes may account for the observed relations. Loss, life stress, and SES may have shared risk factors; loss may be more common in families that experience other forms of life stress; or loss may initiate pathways of compromised resources and increased stress. Additional analyses teased apart these putative mechanisms. Children who experienced significant losses in early childhood (the most robust predictor) did not have significantly higher life stress scores as adults. This suggests that parental loss does not necessarily initiate pathways of increased personal, financial, and legal stress that persist into adulthood.

Researchers have debated whether documented relations between loss and adult depression are due to compromised resources and opportunities following or preceding the loss. This study was able to consider life stress and SES across childhood. Although higher participant life stress was related to higher scores of adult depression, results indicated that severe early childhood loss continued to predict adult depression after accounting for previous, concurrent, and future indicators of life stress and SES. This suggests that loss is not simply a proxy for life stress or SES.

This study found no significant gender differences. Gender was used as a control variable in most analyses and did not account for the relation between loss and adult depression. Due to the small sample size and the use of a global loss rating, this study was unable to compare the outcomes of children who lost a mother to children who lost a father or the outcomes based on the gender match between the child and the lost parent figure. However, the role of the caregiver (i.e. primary or secondary) was considered in the loss rating, with loss of a primary caregiver leading to a higher loss severity coding. In this sample, almost all primary caregivers were mothers. Recent research indicates that the gender match of the lost parent influences whether children develop depression (Takeuchi et al., 2002). In order to investigate whether children are more sensitive to loss of the same/opposite-gendered parent, future studies would need to utilize a coding scheme that distinguishes the gender of the lost caregiver as well as recruit a larger sample size so that there would be substantial numbers of losses of both mothers and fathers.

In contrast to earlier studies, recent research using an epidemiological sample found that childhood paternal death doubled the risk of adult depression for both men and women whereas childhood maternal death did not increase the risk for either gender (Jacobs & Bovasso, 2009). This epidemiological, community sample included both men and women while previous work on loss and depression has often been limited to women (Barnes & Prosen, 1985; Kendler et al., 1992; Mcleod, 1991; Oakley et al., 1995) and clinical samples (Kunugi et al., 1995; see Tennant et al., 1980, for discussion). Jacobs and Bovasso (2009) suggest that father loss contributes to child maladaptation because it leads to financial difficulties (e.g., moving, selling a home, changing jobs). The financial complications prevent the remaining caregiver from being available and responsive, thereby interfering with the children’s adaptation to the loss. While the current study supports that both parental and maternal loss are powerful risk factors for men and women, future studies should investigate whether there are distinct pathways for mothers and fathers that lead to child maladaptation. It should be noted that the outcome measure (depressive symptomatology) was a continuous measure of depressive symptoms rather than a clinical diagnosis of depression. Due to different rates of internalizing disorders in men and women, it will be important for future studies with sufficiently large community samples to investigate loss and the development of depressive disorders.

The role of parenting quality in pathways from loss to depression was less clear. Loss continued to be a robust predictor of adult depression even when considering early care and follow up care. This suggests that the quality of early and later caregiving do not entirely buffer children from the effects of parental loss. A definitive conclusion on the role of follow up parenting in trajectories from loss to depression is cautioned. It is the case that the loss scale naturally confounds these variables. That is, inherent in the loss index is a consideration of whether the substitute or replacement caregiver is a trusted adult to the child. It will be important for future studies to consider follow up parenting in more detail. In addition, this study did not have an observational assessment of parenting immediately following loss. A study that recruits families immediately following a loss may be better suited to investigate nuanced questions regarding substitute or replacement caregiving.

The rich longitudinal data set allowed for a global loss scale that considered theoretically identified factors. The findings presented here provide initial scale validation for a global loss rating. This is consistent with theorists who have proposed that inconsistencies in the loss literature are due to inadequate coding of heterogeneous loss experiences. They suggest that the circumstances surrounding the loss, rather than the loss per se, are developmentally significant (e.g., Harris et al., 1986). However, the comprehensive nature of this scale precludes isolation of individual loss factors. That is, it is impossible to investigate the etiologic significance of individual features of a loss (e.g., length of separation, replacement care, type of separation). Future studies may benefit from utilizing both global loss ratings and discrete variable coding. By using both, researchers may compare the predictive value of individual factors and overall ratings.

In spite of the strengths of this study, it did have several weaknesses. The sample size limited the kinds of analyses that could be performed. In addition, this lower SES sample naturally constricts the range of SES and even the range of life stress found in the sample. The prospective nature of the study precluded detailed observational data at the time of the loss. Intrinsic to this study is the complexity of defining separation and loss in a poverty sample. This theoretical question is not readily solvable. In order to address these limitations, a large-scale study with an economically diverse sample is needed. Still, this study illustrates an approach to disentangling loss from associated factors in the etiology of depression.

Many of the strengths of this study are derived from its prospective, longitudinal nature. No measure relied on retrospective report; all measures were obtained age-by-age and concurrent with one another. An important feature of the study, therefore, was that links with possibly confounding third variables could be statistically controlled. For example, loss could have been a reflection of general life stress experienced by the mother and implicated in the etiology of depression because of that. Regression analyses showed, however, that these findings were robust to controls for SES, life stress, gender, and even earlier depression ratings. Because of the quantity of data available from the larger longitudinal study, the current study was able to use multiple methods, multiple reporters, and aggregates of measures from multiple time points for most constructs. The behavioral data, mother reports of stress, and observed quality of parental care were all independent. Overall, this study suggests that childhood parental loss has developmental outcomes that persist into adulthood, even when considering the contribution of other theoretically derived factors.

Appendix 1. Separation and Loss Index

This scale was designed to assess and document separations from and losses of both biological parents and non-biological parent figures from infancy to 6^th grade. However, biological relation is not sufficient to be considered a parent figure. Rather, there must be evidence that the person held a significant caregiving role prior to the separation. Information was asked directly on certain measures and may also indirectly be obtained from the interview supplements. Additionally, coded and compiled information is used to highlight relevant caregiving factors.

A participant’s separation history is divided into 3 distinct periods: infancy and toddlerhood (12–54 months), early childhood (60, 64, 1^st, and 2^nd grade), and middle childhood (3^rd–6^th grade). For each developmental period, a child will be given a score on the following 7-point scale. In order for a participant to be scored, there must be adequate available information. However, participants may be scored for one period and not others (e.g. a participant completed all assessments in early childhood but not in the elementary school years). The coding of the severity of separation experiences relies on multiple factors: the length of separation, the quality of alternative caregiving, the primary or supporting role of the parent figure, and other circumstances surrounding the separation.

0. Insufficient information

The minimum information needed to evaluate separation experiences is not available.

1. No significant separations from any parent figures

No separations from parent figures are recorded other than very short ones (up to 2 weeks) that result from work, vacation/travel, birth of second child, or other similarly benign motivations. (Note: if short separations are a feature of the child’s caregiving rate as a 2c e.g. parent takes frequent or repeated trips i.e. approximately five times per year).

2. Minor separation experiences

One short separation is present in the child’s history but the features of the separation are such that they are deemed to be minimally stressing. Conversely a very minimally stressful separation of a medium duration is present. Otherwise, disruptions in caregiving are absent from the child’s caregiving history.

1. One short separation (less than 1 month) and the child is cared for by another parent figure provided that there are not any outwardly traumatic features of the separation.

2. One short separation (up to 1 month) where consistent and frequent contact with the parent in maintained regardless of who the substitute caregiver is (e.g. child or parent is hospitalized for one month but they maintain daily visits).

3. Multiple, very short parent–child separations (less than 2 weeks) are a feature of this time period. However, the separations should be predictable (due to work, travel, etc) rather than an illustration of unreliable parenting (see 3f).

4. Medium length separation (1–4 months) and the child stays with other significant caregiver or with a person with whom the child has already established trust and the separation was anticipated positively and it is age appropriate. This scale point is primarily restricted to early and middle childhood. (This can include vacations and should exclude separations with traumatic features that should be rated higher.)

3. Moderate separation experiences

Short, moderately stressful separations as well as very brief but severe separations and minimally stressful separations of a medium duration are scored here.

1. Multiple short separations (at least one of which individually would warrant a score of 2) will be collectively coded as a 3.

2. Change in contact with a primary parent figure that results in regular and consistent contact (e.g. divorce results in reliable weekly contact with one parental figure). Additionally, the primary parent figure maintains his/her caregiving role. Any change in contact with a secondary caregiver (either consistent or inconsistent) is also scored here. Discretion is required to differentiate primary from secondary parent figures. In some cases, there may be more than one parent figure that seem to have similar emotional salience to the child. In that case, consider them both to be primary.

3. Medium length separation (1–4 months) from the parent figure but the contact is both consistent and frequent regardless of who the substitute caregiver is (e.g. child or parent is hospitalized for a month but they maintain daily visits).

4. Very short separation (less than 2 weeks) but with traumatic circumstances (e.g. kidnapping, crisis foster placement, abandonment, etc).

5. Short separation (less than 1 month) and the child stays with a person with whom the child had previously established trust but who was not a caregiver.

6. Repeated very short separations (less than 2 weeks) that represent unreliable parenting are a feature of this time period. Note: if separations are predictable and benign, they should be coded as 2c. Careful consideration of caregiving among relatives is especially relevant to this distinction.

7. Medium length separation (1–4 months) and child stays with other significant caregiver given that there are not any other outwardly traumatic features of the separation.

4. Notable separation experiences

Short, serious separations and moderately stressful separations of a medium duration are scored here.

1. Multiple separations (at least two of which individually would warrant a score of 3) will collectively be coded as a 4.

2. Change in contact with a primary parent figure such that contact is inconsistent. The parent remains in the child’s life but the contact is sporadic and unreliable. (Divorce or separation in which one parent visits the child inconsistently; grandmother adopts caregiving role and mother visits intermittently.) In these cases, contact is maintained but the child cannot predict or depend on the adult’s presence. Note: see 5b for change in contact that results in prolonged absence as well as 3b.

3. Short separation (less than 1 month) and child is cared for by a person who he did not previously know (e.g. foster home, unknown adult, etc).

4. Short separation (less than 1 month) and child stays with other significant caregiver or person with him he had previously established trust, provided that there are outwardly traumatic features of the separation (e.g. parent abruptly disappears, is in hiding following a violent exchange, homelessness, etc).

5. Medium separation (1–4 months) and the child is cared for by a person with whom the child had previously established trust but who is not a caregiver (see 2d and 3g for distinctions).

6. Long separation (over 4 months) from the caregiver but frequent and consistent parent–child contact is maintained regardless of who the substitute caregiver is (e.g. child or parent is hospitalized for 4 months but they maintain daily visits or there are several 2 month hospitalizations with these features).

5. Significant separation experiences

Severely stressful separations of a medium duration as well as minimally to moderately stressful long separations are scored here as well as permanent loss of a subsidiary parent figure.

1. Multiple separations at least one of which individually would be scored as a 4 (as well as several 3s or other 4s) will be collectively coded as a 5 (see 4b for exception).

2. Prolonged absence or lack of contact with a non-primary parent figure following a change in contact with the parent. The initial change in contact may either be consistent or inconsistent. (Father and mother get divorced and the father visits the child regularly for a couple of years and then stops visits altogether until he reappears in adolescence.) Note: this score is given in the age at which the major/terminal separation occurred while a change in contact would be scored for the earlier age at which the contact became either consistent or inconsistent. See 3b and 4b.

3. One medium separation (1–4 months) from a parent figure and the child is cared for by a person who the child did not previously know and with whom the child had not previously established trust (e.g. foster home, unknown adult, etc).

4. One medium separation (1–4 months) from a parent figure and the child stays with other significant caregiver provided that there are outwardly traumatic features of the separation (e.g. parent abruptly disappears, is in hiding following a violent exchange, homelessness, etc).

5. One long separation from a parent figure (over 4 months) and the child stays with either another parent figure or with a person with whom the child had already established trust.

6. One long separation from a parent figure (over 4 months) and the child is cared for by a person who he did not previously know (e.g. foster home, unknown adult, etc) but the child has regular contact with the separated parent figure or knows when the separation will end.

7. Permanent loss of a subsidiary caregiver when the loss does not include other traumatic features. The mitigating factors could include the secondary role of this parent figure, the expected nature of the loss (e.g. death of grandparent), or the continuing role of other caregivers.

6. Very significant separation experiences

Severely stressful long separations are scored here as well as permanent loss in certain circumstances.

1. Multiple long separations from a parent figure at least one of which individually would be scored as a 5 are scored here.

2. Permanent loss of contact with a primary parent figure provided that other parent figures who were involved prior to the separation continue in their caretaking roles.

3. One or more long separations from a parent figure (over 4 months) and the child is cared for by a person with whom he had not previously established trust (e.g. foster home, unknown adult, etc). During the separation, there is no regular contact with the separated parent figure and the child does not know when the separation will end. Note: if there is regular, consistent contact with the separated parent figure and the child knows when the separation will end, rate a 5f.

4. Long separation (over 4 months) with traumatic features.

5. Permanent loss of contact with a subsidiary parent figure with traumatic features (e.g. father figure commits suicide in front of the child, father moves out of state abruptly and takes siblings with him, etc).

7. Extremely significant separation experiences

Separations from parental figures are a pervasive feature of the child’s caregiving history. Consistency in caregiving is virtually absent, or conversely, the circumstances of a single separation are serious enough to warrant this score. To be considered for this score, the circumstances of a single loss/separation must be both traumatic and include the permanent loss of a primary parent figure.

1. Permanent loss of contact with a primary or secondary parent figure (e.g. death, abandonment, etc) provided that no other parent figures are involved in the caretaking following the separation or that alternative caretaking is inconsistent.

2. Permanent loss of more than one primary parent figure regardless of the supporting role of other caretakers.

3. Permanent loss of the primary caregiver with traumatic circumstances regardless of the supporting role of other caretakers.

4. Multiple separations at least one of which is scored as a 6 combined with other moderate separation scores likely including elements of trauma are scored here