Three personality scales measuring four constructs were administered to a group of 230 second year teachers' college students.
The results of the testing were then submitted to correlational and factor analysis to investigate four hypotheses. These were:
(i) that Trait Anxiety non-reversed items measured substantially the same construct as measured by E�senck's Neuroticism scale;
(ii) that Trait Anxiety non-reversed items measured a different construct to that measured by Trait Anxiety reversed items;
(iii) that Trait Curiosity non-reversed items measured a different construct to that measured by Trait Curiosity reversed items;
(iv) that Trait Anxiety reversed was substantially the same construct as Trait Curiosity non-reversed.
The analyses provided only equivocal support for hypothesis (i) and none at all, surprisingly, for the intuitively plausible hypothesis (iv). Hypotheses (i�) and (iii) received very strong support.
The vexed question of the disappearance of the reversed item factors in the second order solution is raised. A case is argued that the phenomenon can be understood in statistical terms as an artifact of the analyses.
Finally, a suggestion is made for a new relationship between personality theory and measurement.