|
Article Excerpt
The Parental Attachment Scale was developed through exploratory factor analysis, confirmatory factor analysis, and validation of a final 23-item scale. Comparison with existing parental attachment measures and a subliminally primed lexical decision task was also conducted. Results across the 3 studies suggested satisfactory internal consistency and construct validity.
**********
Adult attachment has typically been assessed in one of two ways: semistructured clinical interviews, most commonly the Adult Attachment Interview (AAI; Main & Goldwyn, 1998), and self-report. Using self-report inventories to assess attachment makes sense, as Crowell, Fraley, and Shaver (1999) pointed out, because attachment relationships are such an important part of most people's lives that it is logical that they can provide important information about these relationships. More pragmatically, self-report instruments offer the advantage of requiring little training to administer and consequently being easy, quick, and relatively inexpensive to use. However, there are also some problems with using attachment self-report instruments. First, there is an unresolved debate in the self-report literature about how best to assess attachment, concerning issues such as which relationships to target or underlying dimensions to use (Heiss, Berman, & Sperling, 1996; McCarthy, Moller, & Fouladi, 2001). Second, there is a troubling lack of correspondence between attachment assessed through self-report and through interviews such as the AAI (Crowell, Treboux, & Waters, 1999). If such problems could be addressed, the usefulness of self-report instruments--especially in comparison with the AAI, which is somewhat time-consuming and expensive to administer, transcribe, and score (Brennan, Clark, & Shaver, 1998; Stein, Jacobs, Ferguson, Allen, & Fonagy, 1998)--seems evident.
Based on the considerations outlined above, the impetus behind the development of the Parental Attachment Scale (PAS) was to develop a measure that (a) addresses some of the problems in current self-report attachment instruments and (b) incorporates some of the strengths of the AAI, notably a more complex conceptualization of attachment patterns that includes notions of typical emotional coping strategies, as well as an attempt to capture nonconscious elements of attachment representations.
CONCEPTUAL AND STRUCTURAL DIFFERENCES AMONG ATTACHMENT SELF-REPORT INSTRUMENTS
There currently appears to be disagreement by developers of self-report attachment instruments about how to define the construct of attachment. For attachment pioneer Bowlby (1969), the critical factor that identifies an attachment figure is that it is to this person that an individual turns when distressed. More specifically, secure and insecure individuals are primarily thought to be distinguished by their differing expectations of the preparedness and ability of their attachment figures to play these roles (Bowlby, 1973), the way and extent to which they approach these persons (Ainsworth, Blehar, Waters, & Wall, 1978), and the way they manage their emotions when attachment-related distress arises (Main & Goldwyn, 1998). However, many existing attachment self-report measures assess other, arguably less central, aspects of attachment relationships, such as the general affective tone of the relationship (Heiss et al., 1996). In addition, it is not always clear that the relationships targeted by some attachment self-report instruments are in fact attachment relationships. Fraley and Shaver (2000) have suggested that it may take as long as 2 years for a romantic relationship to become a "true" attachment relationship and argue that romantic attachment self-report instruments may routinely be used incorrectly to assess romantic relationships that are not also attachment relationships.
Although some attachment self-report instruments can be criticized for not clearly targeting attachment relationships, a more pervasive issue is that the measures differ in the aspects of attachment on which they focus. Whereas some measures follow Bowlby's (1988) conceptualization by focusing on whether the nature of the attachment relationship is secure or insecure, as well as specific behaviors associated with adult attachment such as availability and responsiveness (e.g., the Inventory of Parental and Peer Attachment; Armsden & Greenberg, 1989), other measures use Ainsworth et al.'s (1978) taxonomy for classifying children's attachment status (avoidant, ambivalent, and secure) and aim to classify adult respondents' attachment style on this basis (e.g., Hazan & Shaver, 1987).
Another issue concerns whether attachment should be conceptualized in terms of categories or latent dimensions. The first, and very influential, self-report instrument developed (Hazan & Shaver, 1987) was a categorical instrument in which respondents indicated which of three relationship descriptions fitted them best. Subsequently, however, a number of researchers (Fraley & Waller, 1998; Griffin & Bartholomew, 1994) argued that the adult attachment patterns assessed through self-report are better accounted for by a latent dimensional model. One assumption of dimensional attachment instruments is that underneath them lie dimensions that capture important aspects of attachment relationships, but there is no clear agreement about the nature of these dimensions: Anxiety and avoidance dimensions (Brennan et al., 1998), positive and negative models of self and others (Bartholomew & Horowitz, 1991), and dimensions based on affect and behavior regulation processes rather than cognitive content (Crowell, Fraley, et al., 1999) have all been proposed.
There is also debate over how various attachment styles (or categories) map onto the attachment dimensions described earlier. Both the Brennan et al. (1998) and Bartholomew and Horowitz (1991) paradigms lead to a four-typology conception of attachment styles in which the fourth type is termed fearful (negative other and self or high anxiety and high avoidance). This division is persuasive but problematic in that the fearful attachment style has no counterpart in the original three-type taxonomy (avoidant, ambivalent, and secure) Ainsworth et al. (1978) identified in her research with infants. The fearful type also does not seem to correspond to a fourth category subsequently discovered by researchers using the Strange Situation (Ainsworth et al., 1978) to assess infant, and the AAI to assess adult attachment security (disorganized for infant and "U" unresolved trauma for adult, respectively; Jacobvitz, Curran, & Moller, 2002). In sum, while a dimensional approach to self-report attachment instruments may be preferable to a categorical approach, there is still a considerable lack of agreement about the underlying dimensional structure.
Attachment self-report instruments also vary in terms of the relationship domain that they seek to assess. A large number of instruments exist that assess romantic relationships, but instruments have also been developed to assess peer and parental relationships (e.g., Inventory of Parental and Peer Attachment; Armsden & Greenberg, 1989) as well as non-relationship-specific attachment style (e.g., Attachment Style Questionnaire; Feeney, Noller, & Hanrahan, 1994). For the PAS, we decided to follow the AAI and ask respondents about their childhood relationships with their parents rather than about adult relationships (Hesse, 1999).
EXISTING SELF-REPORT MEASURES OF PARENTAL ATTACHMENT
Two commonly used parental attachment instruments are the Inventory of Parent and Peer Attachment (IPPA; Armsden, & Greenberg, 1989) and the Parental Bonding Instrument (PBI; Parker, Tupling, & Brown, 1979). Neither of these instruments is categorical nor based on the conceptualizations of underlying dimensions of attachment discussed earlier. In addition, the measures differ both theoretically and structurally from each other. The main difficulty with the IPPA is that the single-scale score derived from this instrument corresponds to a simple secure-insecure dimension; as such the IPPA is not able to distinguish between the different forms of insecurity, nor the different strategies of emotional regulation adopted by individuals with different expectations of attachment figure availability. For the PBI, the problem is that it does not clearly assess attachment. Of the two PBI subscales, the Care subscale appears to assess the general affective tone of the parental relationship rather than the central attachment function parents play for their offspring when they are distressed, which is problematic, as discussed above. And although parental control (assessed by the Overprotection subscale) is an important part of parenting, it is not part of the attachment system as envisaged by Bowlby (McCarthy et al., 2001).
Given the concerns with these two instruments, creation of a new parental attachment self-report instrument is warranted. An additional motive is the difficulty caused for the field by the woeful lack of correspondence between attachment self-report instruments and the other major assessment tool for adult attachment, the AAI. This lack has contributed to insufficient communication between researchers using coded narrative assessment and those using self-reports and, arguably, has led to a correspondingly incomplete theory of attachment (Shaver & Mikulincer, 2002). The reasons for this lack are discussed next.
RELATIONSHIP BETWEEN SELF-REPORT ATTACHMENT INSTRUMENTS AND THE AAI
The little research that has been done using both the AAI and self-report instruments has found only modest to moderate associations (e.g., Crowell, Treboux, et al., 1999; Shaver, Belsky, & Brennan, 2000; see also, Crowell, Fraley, et al., 1999; Jacobvitz et al., 2002). The lone study (Crowell, Treboux, et al., 1999) uncovered by us that used both the AAI and a parental self-report measure (neither the IPPA nor PBI) found only a .11 correspondence between scores on the two instruments.
Proponents of the AAI (e.g., Main & Goldwyn, 1998) argue that the lack of overlap between self-report attachment instruments and other assessment methodologies can be explained by the fact that the AAI scoring system codes security of attachment by focusing on how individuals discuss childhood relationships with parents, not the content of their speech. More specifically, existing self-report attachment instruments assess conscious appraisals of relationships with parents, whereas the AAI taps into nonconscious processes for regulating emotion during discussion of distressing attachment-related events. In addition, the AAI and the various self-report attachment instruments use different criteria to categorize individuals as secure or insecure (Jacobvitz et al., 2002). For example, when existing attachment style self-report instruments have attempted to identify avoidant individuals, typically items have assessed shortage of trust or comfort in, and lack of importance placed on, relationships (Crowell, Fraley, et al., 1999) rather than the idealizing of the parental relationship that is found in many AAI transcripts for persons categorized as dismissing/avoidant. Yet, in the AAI coding system, the ability to talk openly about difficulty with trusting or relying on others is associated with secure, rather than avoidant insecure attachment. This means that an individual who fits the secure category in the AAI coding system might be rated as avoidant in a self-report instrument. Similar differences are found for the preoccupied (adult)/anxious-ambivalent (child) insecure pattern. Furthermore, as mentioned previously, the fourth category in the AAI of "U" (unresolved trauma) does not correspond to the fourth fearful category established in some self-report instruments. Another difference is that only the AAI distinguishes individuals who have shifted from an insecure to a secure attachment status, or "earned" attachment security (Moller, McCarthy, & Fouladi, 2002); self-report instruments do not capture change in attachment status (for a more complete review of this issue, see Jacobvitz et al., 2002).
OBJECTIVES OF THE CURRENT STUDY
On the basis of considerations outlined earlier, we decided to develop an instrument assessing an adult's memories of childhood interactions with parents, to specifically target perceptions of parental availability in the event of distress, and to try to capture aspects of attachment security and insecurity not assessed by current parental attachment instruments, such as coping strategies for negative attachment-related emotion, the extent to which attachment status may have changed over time, and nonconscious elements of attachment.
In order to assess whether the PAS captures nonconscious elements of attachment, the predictive validity of scores from the instrument was examined using a paradigm developed by researchers exploring nonconscious elements of attachment working models (Baldwin, 1994; Baldwin & Meunier, 1999; Mikulincer, Birnbaum, Woddis, & Nachmias, 2000; Mikulincer, Hirschberger, Nachmias, & Gillath, 2001; Pierce & Lydon, 1998). This paradigm uses priming and a lexical decision task and is based on the concept of spreading activation--the idea that knowledge is represented in structures or networks and that if one unit in a structure is activated, activation spreads along connections in the networks and activates related units. This "priming effect" is found even if the prime is presented so briefly that participants are unaware of having seen the word (for reviews, see Kihlstrom, 1987; Merikle, 1992). If participants are unaware of the prime, participants are unable to manipulate or defensively alter their responses, which means that subliminal priming paradigms allow attachment researchers to access nonconscious elements of attachment working models.
In the current article, the subliminal prime used was the phrase "My Mom"; this phrase was chosen to cue a clearly defined and personally relevant attachment relationship, and, because approximately 40% of individuals are expected to be insecurely attached, one that taps into both secure and insecure working models (van IJzendoorn & Bakermans-Kranenburg, 1996). It was assumed that using this prime would lead to faster lexical decisions for words more closely associated with participants' maternal attachment working models.
Three data sets were collected over three semesters. In the first analysis, data from the first data set were analyzed using exploratory factor analysis (EFA). Confirmatory factor analyses (CFAs) were conducted on the second and third data sets. Next, construct validity evidence for use of scores on the scales with existing self-report instruments was explored in the samples by examining correlations between PAS subscale scores and scale scores from two parental attachment instruments, the IPPA and PBI in the case of Studies 1 and 2, and the IPPA, PBI, and a nonrelation-specific instrument, the Attachment Scale Questionnaire (ASQ) in Study 3. Last, the relationship between a subliminally primed lexical decision task and the three parental attachment instruments was explored through a series of regression analyses.
METHOD
Participants
Participants for Study 1 were 342 undergraduates (age: M = 20.9, SD = 1.76; range 18 to 33 years). Sixty-nine percent were women, and 31% were men. Forty-four percent described themselves as seniors, 38% as juniors, 13% as sophomores, 4% as 1st-year students, and 1% as other. According to self-reported racial identity, 65% of participants were White/Anglo American, 16% Asian American, 11% Latino, 3% African American, 2% biracial, 1% multiracial, and 2% other. Eighty-nine percent of participants described themselves as having "both a mother and a father"; of the remaining participants, 4% chose the parental description "mother only," 5% "mother and stepfather," and less than 1% chose "father only," or "father and stepmother" or "grandparents or foster parents/guardians."
The participants in the Study 2 were 337 undergraduate students (age: M = 21.4, SD = 3.0; range 17 to 51 years). Seventy percent of participants were women, and 30% were men. Fifty-two percent described themselves as seniors, 21% as juniors, 16% as sophomores, 6% as 1st-year students, and 5% as other. According to self-reported racial identity, 75% of participants were White/Anglo American, 13% Asian American, 3% Latino, 2% African American, 3% biracial, 2% multiracial, and 2% other. Eighty-seven percent of participants described themselves as having both a "mother and a father"; of the remaining participants, 6% chose the parental description "mother only," 5% "mother and stepfather," 1% "father only," and .5% "father and stepmother" and "grandparents or foster parents/guardians."
The participants of Study 3 were 179 undergraduate students (age: M = 21.5, SD = 3.43; range 18 to 51). Seventy-four percent of participants...
|
