Participants
Participants for this study were recruited as part of an ongoing longitudinal adoption study. Roughly half the adoptees were originally selected due to the presence of psychopathology (e.g., alcohol problems and/or antisocial behaviors) in a birth parent. During the most recent re-interview, participant psychiatric histories were updated and the Adult Attachment Interview [AAI; [16]] was administered (n = 208). Approximately 53% of the sample was female and ranged between 24 and 66 years of age (M = 39, S.D. = 7.95). Average household income was $40,000 to $49,999 per year. The sample was predominantly Caucasian (92%), with the remaining sample comprised of 4% Hispanic, 2% African American, and 1% "Other". Adoptees were adopted by non-relatives within 2 months of age (SD = 5.44) with 67.8% adopted prior to one month and 94.2% adopted prior to 6 months of age.
Measures
Adult attachment classification
The AAI [16] is a semi-structured interview that assesses an individual's attachment representations. Individuals are asked to provide five adjectives describing their childhood relationship with their adoptive mother and father, separately. Participants are also asked to provide experiential support for the chosen descriptors. Questions about parental responses during episodes of emotional upset, illness, and injury are also probed, as are experiences with death and trauma. Finally, the individual is asked to describe changes in and current feelings about their relationship with their parents. Interviews were transcribed verbatim and coded by coders deemed reliable by the lab of Mary Main and Eric Hesse (Rebecca Yucuis and Kristin Caspers, Trained by Deborah Jacobvitz, Austin TX, 2001; Beth Troutman and Jeanne Frederickson, Trained by June Sroufe, Minneapolis, MN, 2002 and 1999, respectively). Approximately half of the AAIs were double-coded. Disagreements were resolved through conference. Overall inter-rater agreement was 94% for the secure versus insecure distinction (κ = .86, p < .001), 91% percent agreement for the organized classifications (κ = .84, p < .001), and 93% agreement for the unresolved/not unresolved classifications (κ = .71, p < .001). Cronbach alphas were equally high for the individual scales ranging from .84 to .93.
The first step in coding attachment representations involves rating inferred childhood experiences with parents. Five parental behaviors are rated on a 9-point scale: loving, rejection, pressure-to-achieve, involving-reversing, and neglect. Loving behavior reflects emotional support and availability. Rejection reflects active rejection or an avoidance of a child's attachment behaviors. Involving-reversing represents role-reversal between parent and child. Pressure-to-achieve reflects parental emphasis on achievement as a key component to the parent-child relationship. Finally, neglect represents parental unresponsiveness to attachment-related behaviors. The method by which ratings for childhood experiences are derived results in estimates of probable experiences with caretakers during childhood and adolescence. The presence of behaviors may be determined by either direct evidence (e.g., provision of comfort during episodes of distress) or the absence of evidence (e.g., no mention of comfort during episodes of distress). Therefore, ratings of childhood experiences are considered inferred rather than reflective of actual behaviors.
The next step in coding is determining attachment representations [16]. The transcript is evaluated for coherency and key indicators for each classification are rated. Three primary organized states of mind are derived from the transcripts: dismissing (Ds), secure (F), and preoccupied (E). Dismissing attachment is characterized by an inability to recall specific memories for positive adjectives used to describe either the mother and/or father. Individuals classified as dismissing often show a high degree of self-reliance, place minimal value on attachment relationships, and portray their childhoods as positive but are unable to provide experiential support. Inferred parental behaviors of rejection and/or neglect are most often associated with dismissing representations. Individuals classified as preoccupied are unable to focus on questions at hand and respond in either a vague or actively angry manner when discussing past or current interactions with their caretakers. These individuals appear as if they are unable to move beyond their childhood experiences, remaining entangled with their parents. Inferred childhood experiences associated with preoccupied attachment most often involve inconsistent behavior and/or involving-reversing. Finally, individuals classified as secure are able to provide experiential support for adjectives provided, whether positive or negative. They are consistent in their portrayal of early experiences, are willing to evaluate past and current relationships, and show valuing of attachment and forgiveness for negative experiences. The secure classification can be further divided into two sub-classifications based on ratings for inferred parental behavior: earned-secure and continuous-secure [32–37]. The earned-secure group is comprised of individuals with low ratings on positive indicators of inferred childhood experiences but demonstrate a secure state of mind. The continuous-secure group is comprised of individuals who experienced a supportive relationship with at least one parent and expectedly developed a secure state of mind.
Finally, a category of unresolved/disoriented (U) is assigned when significant lapses in discourse are present during discussions of loss or trauma. A few examples of speech patterns indicative of the unresolved category are confusions of the dead person as living, excessive detail surrounding the event of death or trauma, identifying the self as causal in the death of a loved one or deserving of abuse, or extreme reactions to experiences of loss or trauma. Given a classification of unresolved, subjects are also assigned a corresponding organized classification (e.g., U/Ds).
Substance use problems
The Semi-Structured Assessment for the Genetics of Alcoholism – II [SSAGA-II; [41]] was used to collect detailed information about lifetime substance use including alcohol, tobacco, marijuana, and all non-marijuana substances (e.g., cocaine, stimulants, hallucinogens, etc). Lifetime diagnoses of abuse or dependence of alcohol, marijuana, and any illicit drugs were derived from DSM-IV criteria. Overall prevalence rates for each diagnosis were as follows: alcohol dependence (18/208, 9%), alcohol abuse (90/208, 43%), marijuana abuse or dependence (42/208, 20%), and any illicit drug abuse or dependence (35/208, 17%). Fifty-six percent (115/208) reported one or more substance-related diagnosis. Substance abuse or dependence of alcohol, marijuana, or non-marijuana drugs served as independent outcome variables.
Mental health treatment
Solicitation of mental health care was determined from the SSAGA-II [41]. Two questions were used, each based on a yes/no response: 1) Have you ever spoken to a professional (e.g., psychiatrist, psychologist, medical doctor, nurse) about any emotional problems and 2) Have you ever received outpatient treatment which includes speaking to a psychiatrist, psychologist, or therapist. Fifty percent (104/208) reported seeing a mental health professional and 38% (79/208) reported receiving outpatient treatment. The majority of individuals sought treatment for emotional problems not related to substance abuse/dependence. Sixty-six percent of those who spoke to a professional about an emotional problem also reported receiving outpatient treatment.
Analyses
Logistic regression (SPSS, v. 14.0) was used to examine the association between substance abuse/dependence, treatment participation and attachment representations. We relied on attachment theory to construct orthogonal contrasts. For the prediction of substance abuse/dependence, we predicted higher rates of problematic use among individuals classified as dismissing, preoccupied, or earned-secure when compared to individuals classified as continuous-secure. We also predicted non-significant differences among the former three groups (i.e., dismissing, preoccupied, and earned-secure). Thus, three contrasts were constructed reflecting the following comparisons (assigned values indicated in parentheses): 1) continuous-secure (-1) versus all other classifications (+.333), 2) preoccupied or earned-secure (-.50) versus dismissing (+1) and 3) earned-secure (-1) versus preoccupied (+1). The three contrasts were entered simultaneously into logistic regressions predicting alcohol, marijuana, or non-marijuana substance abuse/dependence.
For the prediction of treatment participation, we hypothesized higher rates of treatment among individuals classified as preoccupied or earned-secure when compared to individuals classified as dismissing or continuous-secure. We also predicted no differences between the former groups (e.g., dismissing and continuous-secure) as well as no differences between the latter groups (e.g., preoccupied and earned-secure). Three contrasts tested our hypotheses in the prediction of treatment participation (assigned values indicated in parentheses): 1) dismissing or continuous-secure (-.50) versus preoccupied or earned secure (+.50), 2) dismissing (-1) versus continuous-secure (+1), and 3) earned-secure (-1) versus preoccupied (+1). The three contrasts were simultaneously entered into a logistic regression predicting lifetime history of treatment participation.
We also tested for the effect of gender, current age, current mood (e.g., depression/anxiety symptoms) and personality disorder on the association between attachment representations, substance use problems, and treatment participation. The parameter estimates for attachment representations were not substantially reduced and remained significant. Therefore, the unadjusted findings are presented.