Undertook a study to develop a new attachment instrument, the Attachment Style Scales, intended to overcome criticisms of existing measures. A second confirmatory study was conducted to compare this new attachment scale with that used by Shaver and Brennan (1992). This was done by comparing the relationships, for each study, between attachment style and the 'big five' personality traits.
In their introduction, they criticise previous attachment measures on the following grounds:
- Instruments which use simple categorical measures restrict the statistics that can be used and preclude the use of measures of internal consistency such as Cronbach's alpha.
- Several multi-item questionairres have relatively low internal consistency (Collins & Read, 1990; Simpson, 1990).
- Single-item measures assume that individuals are classifiable, rather than permit individual profiles of attachment.
- Many of the items developed are are not useful for classifying relationship to people or groups other than romantic partners.
Their objective was to develop scale to measure secure, fearful and preoccupied attachment styles with high internal consistency and which are useful for assessing non-romantic attachments.
Subjects were 1,181 college graduates (median age 24, 55.1% female) who completed the attachment questionairre described below and also Bartholomew and Horowitz (1991) attachment measure..
The questionairre was constructed by disecting the Hazan and Shaver (1987) and Bartholomew and Horowitz (1991) descriptions into separate items. To this, items were added from Collins and Read (1990) to measure 'capacity to depend on others' and items created to measure 'confidence in acceptance from others' and 'confidence in not being let down by others.' These additional items were due to the paucity of items from earlier measures for the preoccupied style. The resulting questionairre contained 25 items, each of which was scored using a 7 item Likert scale.
Factor analysis left three main items. These appeared to relate to Preoccupation, Fearfulness and Security. These correlated well with the results from the Bartholomew and Horowitz measure (r = 0.63, r = .52 and r = .49, respectively). A strong, negative correlation was observed between the Secure and Fearful styles (r= -.63) which suggested these may be better treated as poles of a single dimension, however, factor analysis of the personality trait scores using a two-factor model, rather than the three-factor model, produced a poor fit.
Overall, the scale produced high degress of internal consistency (alpha >= 0.80), showed convergent validity with the Bartholomew and Horowitz (1991) measure and provided similar correlations with the 'big five' personality traits to Shaver and Brennan (1992).
The 'big five' personality trait study involved 545 of the recent graduates (median age = 23, 56.5% female) who completed a reduced version of Goldberg's (1992) Personality Measure. This scores subjects on five main traits: neuroticism, extraversion, openness to experience, agreeableness and conscientiousness. Results from this were comparable to those obtained by Shaver and Brennan (1992).