The first phase of this project studied the feasibility of automatically generating a subject validation file with complete strings but few errors and obsolete elements from LC-assigned subject headings appearing in bibliographic records. It analyzed the distribution and density of subject headings by frequency of use and calculated the rate of errors and obsolete elements in subject headings that have been used at least twice in the LCMARC database. The second phase analyzes the nature and patterns of incorrect and obsolete subject headings. Although errors in subject headings assigned by LC occur at a relatively low rate, an awareness of the nature of errors and the rate of obsolete headings will help to avoid or minimize such errors in the future. Understanding the recurring patterns of errors can also help to improve mechanisms for identifying and correcting errors and for updating obsolete headings automatically.