In data items of USAGE COMP, data is held in binary format as described in the previous sections. The storage allocated for an item can have space for larger numbers than specified by the PICTURE clause. For example, an item described as PIC 99 COMP is normally assigned one byte, which can hold numbers up to 255.

To conform with the rules of ANSI COBOL, numbers behave as decimal numbers, regardless of their format. If, in an arithmetic statement, the result is bigger than the PICTURE clause of a receiving item allows, a size error occurs, and if the ON SIZE ERROR phrase is specified the result is not stored in the receiving item. In a non-arithmetic statement, if this situation occurs, the decimal value is truncated on the left, to the number of digits specified in the PICTURE clause.

However, data in USAGE COMP items can be forced to behave as binary data, that is, truncation occurs only if it is necessary in order for the data to fit the space allocated. The behavior of USAGE COMP items is controlled by the setting of the COBOL Compiler directive TRUNC. This directive selects whether the decimal value is truncated to the picture size, or the binary value is truncated to the space available. It distinguishes between results of arithmetic statements, and data being moved by non-arithmetic statements.

Regardless of the setting of any directive, an arithmetic statement gives the size error condition if the result has more decimal digits than specified in the PICTURE clause of a receiving item.