The accuracy and reliability aspects of data integrity are discussed, with an emphasis on the need for consistency in responsibility and authority. A variety of ways in which data integrity can be compromised are discussed. The following sources of data corruption are described, and the ease or difficulty of identification and suggested actions for prevention are discussed: (1) changes in institutional policies; (2) new meaning associated with a datum; (3) user experimentation with the system; (4) purging/consolidation of corrupted data; (5) referential integrity; (6) inadequate analysis and testing of software; (7) running obsolete versions of a program; (8) restructuring set relationships on a database; and (9) the trade-off between editing and performance. Contains 4 references. (KM)