A COBOL USAGE clause specifying single-precision floating-point (4 bytes). On z/OS, may be IBM hexadecimal floating-point or IEEE 754, depending on the *FLOAT* compiler option. (Ch. 5)