Acute graft-versus-host disease (GVHD) is a significant cause of morbidity and mortality following bone marrow transplantation, and early detection is important to allow effective therapy. Since the presence of apoptotic keratinocytes (dyskeratotic bodies) has been suggested as a useful diagnostic criterion for GVHD, attention has focused on the use of the TUNEL assay to detect apoptosis in clinical specimens. We reviewed clinical specimens upon which TUNEL had been performed for possible artifacts that might interfere with accurate evaluation for GVHD. Several distinct types of artifact were found and could be re-created in experimental systems. Artifacts in TUNEL staining generally resulted from the lack of specificity of this reaction for apoptotic cell death. Artifacts were found resulting from inadequate fixation, over-exposure of the TUNEL reaction, and proximity to the section edge. In addition, a novel artifact, apparently resulting from DNA shearing during the sectioning process, was noted and confirmed using confocal microscopy of experimental specimens. The TUNEL assay must therefore must be interpreted with caution in the clinical setting. In our laboratory, we consider TUNEL-positive cells as apoptotic only when accompanied by apoptotic morphology. Although these criteria clearly miss some cells in the early stages of apoptosis, they provide the highest specificity for apoptotic cell death.