Previous studies have assumed that people have a predilection for single linear orderings, of which status congruence is a special case. Eighty-two subjects were run in a two-stage experiment to examine preference for linear orderings and whether single or multiple discrepant orderings are preferred when 1 instead of having to learn, subjects freely construct social structures, and 2 the material they are presented is more specified and less simplified than is usually the case in cognitive bias studies. In the first stage Ss were presented with sets of five individual characteristic dimensions, of which three were inherently orderable. Their task was to attribute one characteristic from each dimension to each of five fictitious persons. In the second stage, Ss were asked, for three types of influence (professional, political and cultural), to build an influence structure among the five persons constructed in the first stage.

In Stage 1, linear ordering was used for three of the five dimensions. Two of these dimensions used a single ordering, while the third was only weakly related to the first two. In Stage 2, largely transitive but incomplete relations of influence were established, based on a salary-occupational hierarchy. The completeness of the relation depended on the nature of the influence.