Since Horst (1941) first discussed and defined the suppressor variable in multiple regression/correlation, a number of more nearly precise definitions has been offered (Cohen and Cohen, 1975; Conger, 1974; Darlington, 1968; Velicer, 1978). This paper compares the different approaches to defining suppression, and illustrates the differences among them. Velicer's definition, which is based on the semi-partial correlation coefficient or usefulness and which is perhaps the least well known of the...