This paper examines a commonly used indicator of reliability in advertising studies (the test-retest correlation coefficient), and shows that reliability values obtained by this method are not as simple to interpret nor as closely related to the idea of stability or reproducibility as many casual readers in the area might presume them to be. Examples from the advertising literature are provided to demonstrate these points, and several cautions in using test-retest correlations as reliability estimates are offered. It is also suggested that other statistics may prove useful to the researcher as supplements to the limited amount of information contained in the estimated reliability number.