In the literature on the measurement of change, reliable change is usually
determined by means of a confidence interval around an observed value of a
statistic that estimates the true change. In recent literature on the effic
acy of psychotherapies, attention has been particularly directed at the imp
rovement of the estimation of the true change. Reliable Change Indices, inc
orporating the reliability-weighted measure of individual change, also know
n as Kelley's formula, have been proposed. According to current practice, t
hese indices are defined as the ratio of such an estimator and an intuitive
ly appealing criterion and then regarded as standard normally distributed s
tatistics. However, because the authors fail to adopt an adequate standard
error of the estimator, the statistical properties of their indices are unc
lear. In this article, it is shown that this can lead to paradoxical conclu
sions. The adjusted standard error is derived.