COMMENTS ON THE MEASUREMENT OF HALO

The most frequently used measures of halo are based on interdimension correlations (the correlation, or r, measure) or within-ratee variabil ity across dimensions (the standard deviation, or SD, measure). Pulako s, Schmitt, and Ostroff argued that the r measure is conceptually appr opriate and demonstrated that the SD measure does not always yield res ults consistent with the r measure. They therefore concluded that the SD measure should not be used. This article critiques the r measure, r esponds to criticisms of the SD measure, and describes conditions unde r which halo effects can occur and (a) both measures can be obtained, (b) one of the measures can be obtained, but the other measure either cannot be obtained or should not be used, and (c) neither measure can be obtained. Thus neither measure is uniformly superior to the other. It is therefore recommended that both measures be obtained when possib le and when there is no basis for selecting one over the other.