We describe a phage-display-based method to identify epitopes or interaction sites on proteins. DNA encoding the protein of interest is partially degraded with DNase I to generate random fragments of 50-200 bp. These fragments are then cloned into a phagemid vector that has been modified to allow the expression of the random fragments and the construction of a (bacterio)phage-displayed random epitope library. Phages displaying functional epitopes can be selected from these libraries by affinity selection or panning. To test this method we have constructed a random-epitope library for human plasminogen-activator inhibitor 1 and used this library to map the epitope of a monoclonal antibody (mAb) directed against this protein. By alignment of the selected overlapping epitope-containing fragments, we were able to locate the epitope of the mAb on a stretch of 39 amino acids spanning from E128 to V166. The approach may also be applied to more complex systems than single-protein genes, such as viral genomes or complete cDNA libraries.