Carver (1997) published a paper describing four studies which had used his Measure of Attachment Qualities (MAQ).  This measure consists of 14 items‚ some of which were newly written but most of which were derived from earlier measures (AAS and ASM) both of which had‚ in turn‚ been based on decompositions of the prototypes in the Adult Attachment Questionnaire.  Each item was scored on a four-point Likert-type scale with items ranging from "strongly agree" to "strongly disagree" with no neutral item.

The following list shows the items of the MAQ instrument grouped by the factor on which they most strongly loaded and followed by a prefix which indicates the Adult Attachment Questionnaire (AAQ) description from which the item was derived.

Notes:
(R) = Reverse scored item
(S) = Secure‚  (Av) = Avoidant‚ (Amb) = Anxious/Ambivalent

The authors grouped the items following a factor analysis‚ using oblique rotation‚ which resulted in four factors that together  accounted for 61% of the variance.  The factors for avoidance and security that were loosely‚ inversely related.  The other two factors were interpreted from their component items as ambivalence-worry‚ representing fear of abandonment‚ and ambivalence-merger‚ which included the items indicating a desire to merge. A second-order factor analysis yielded two high-order factors that were unrelated (r = 0.03)‚ one of which distinguished between security and avoidance and the other between the two types of ambivalence.

Scores on this scale were measured against participant's classifications using a variety of other attachment measures.

Comparison with the AAQ showed largely the expected relationships.  These results were taken to confirm the presence of an underlying three factor model of attachment. Although differences between types of ambivalence were acknowledged it was noted that the ambivalence-worry scale‚ which was scored higher by avoidant than by secure subjects‚ may be tapping a general 'insecurity' aspect of attachment.

Comparison with the RQ and the RSQ was problematic. Although many of the expected relationships emerged‚ the MAQ did not clearly differentiate two styles of avoidance‚ as required in the RQ. Cluster analysis was conducted on the MAQ scores solving for both three- and four-cluster models.  Results from this lent further support for an interpretation of MAQ results in terms of a three- rather than a four-component model.