This procedure attempts to control statistically for the potentially confounding effects of family environment in the relationship between child abuse and later psychological symptomatology. Unfortunately, the use of such statistics in abuse research may be problematic, especially when control, predictor, and criterion variables are significantly intercorrelated (multicolinear) and the intent is to determine the relative importance of the abuse to later psychological functioning. Other problems arise as the result of the effects of small sample size, unreliability of the control variable, ambiguity regarding the causal relationship between control and predictor variables, and theoretical problems inherent in considering abuse independently of family dysfunction. Finally, the algebra of the semipartial correlation ensures a conservative test: In order to be significant, abuse must correlate with symptoms after all variance shared with the control variable (e.g., family background) has been removed. Thus, studies that show abuse continues to predict later adjustment after control variance has been extracted should be considered quite seriously; whereas those with negative findings do not necessarily indicate that the abuse has no impact. 2 notes and 24 references. (Author abstract modified)