Spearman correlations of .60 are not poor.

For many years, therapists have assumed a relationship between impairment and disability. More recently, researchers have begun to examine this assumption. "Motor Function and Activities of Daily Living Assessments: A Study of Three Tests for Persons With Hemiplegia" by Filiatrault, Arsenault, Dutil, and Bourbonnais (AjOT, September 1991, pp. 806-810) is a laudable example of such an examination. I have, therefore, no desire to maJign their important work. What 1 want to address is the authors' interpretation of their Spearman coefficients. Although the matter may appear trite at first glance, it arises often in reports of correlational research and may influence the use of research by clinicians. Filiatrault et al. described as "poor" Spearman correlations of .60 and .61. Granting that it is important to judge correlations on more than mere significance, I believe that the authors description of the correlations as poor may devalue them excessively, Published interpretations of correlation magnitudes tend to categorize correlations as follows: <,20 = very weak; .20-.40 = low, small, weak; .40-.70 = moderate; .70-.90 = high, strong, marked; >.90 = very strong, very high (Huck, Cormier, & Bounds, 1974; Okolo, 1990; Rowntree, 1981; Williams, 1986). Thus, the correlations reported by Filiatrault et al. might be considered moderate. Although correlations of moderate magnitude leave the majority of variance between two variables unexplained, they should not be discarded as unimportant when obtained in studies of human beings. Feinstein (1985) wrote that "considering the variability of medical data, r values of ,5 or higher are regarded as qUite good." Similarly, Cohen and Cohen (1983) indicated that in the nonphysical sciences "where relationships are not strong, correlations (of whatever kind) only infrequently are as large as .sO" (p. 144). In a later publication, Cohen (1988) wrote that work-