Abstract
A phenomenon repeatedly described in the personality literature is that in a test-retest situation the 1-time estimates of reliability tend to be larger at Time 2. This article describes a procedure based on a longitudinal covariance structure model that enables this phenomenon to be rigorously examined. The procedure allows the researcher to (a) determine whether the same trait is assessed at both points of time, (b) test whether the reliability of the measures increases significantly at Time 2 and, if so, (c) examine the sources of this increase. The procedure is applied to 3 well-known personality measures and the results are discussed from a substantive point of view.