In a similar vein to Collins & Read (1990)‚ Simpson (1990) measured attachment using 13 self-report items‚ each of which was decomposed from the three prototypical descriptions in the Adult Attachment Questionnaire. Each item was scored using a seven item Likert-type scale that ranged from 'strongly agree' to 'strongly disagree'. Three of the items were reworded in a negative directions and minor changes to some of the wordings were introduced. The 13 items used (and the AAQ description from which they were taken) were as follows:

Note:  (S) = Secure‚  (Av) = Avoidant‚ (Ax) = Anxious/Ambivalent

Rather than conducting a factor analysis‚ Simpson used the attachment style descriptions from which the items had been taken as his basic dimensions.  Internal validity was calculated using Cronbach's Alpha and found to be relatively good for the avoidant dimension (alpha = 0.79) but poor for the secure (alpha = 0.51) and anxious/ambivalent (alpha = 0.59) dimensions.

Simpson experimented with different conceptions of attachment style.  Apart from correlating his results with scores from the secure‚ avoidant and anxious/ambivalent sub-scales‚ he repeated his calculations using two dimensions (avoidant and anxious) and with scores reduced to three simple‚ descrete categories (classification as secure‚ avoidant and anxious/ambivalent). He found the general pattern of findings was the same‚ regardless of how attachment style was handled.

Note that the internal validity scores for the secure and anxious/ambivalent dimensions were not good and this measure has been criticised for being based on dimensions that were chosen for conventional‚ rather than empirical reasons (Carver‚ 1997).

http://www.rich‎ardatkins.co.uk/atws/page/44.html