We also examined whether the correlations among the eight subscales conformed to the expected circumplex structure. Consistent with prior research (e.g., Alden, Wiggins, & Pincus, 1990; Locke, 2000; Ojanen et al., 2005), subscales were ipsatized in order to account for variation in subjective response style. That is, subscales are expressed as deviations from their mean score across all of the scales. For the subscales to be consistent with a circumplex structure, the highest negative correlations should be between opposite subscales (e.g., separate [−C] and communal [+C]) while the highest positive correlations should be between adjacent subscales on the circumplex (agentic-communal [+A+C] and communal [+C]). Overall, the intercorrelations among the ipsatized subscales are similar to those presented by Ojanen (2005; see Table 2), and supportive of a circumplex structure.
