We analyzed the effects of steps in RNA in situ hybridization (ISH) procedures on morphology and hybridization signal with reflection-contrast microscopy (RCM) and transmission electron microscopy (TEM). In chessboard experiments, a range of fixatives containing formaldehyde, glutaraldehyde, or both, and various permeabilization protocols, including ethanol and pepsin treatment, were investigated. A transfected rat fibroblast cell line that harbors an inducible human cytomegalovirus immediate early (IE) transcription unit, and specific probes for 28S ribosomal RNA and IE messenger RNA were used for this purpose. Probes were labeled with digoxigenin and hybrids were detected with anti-digoxigenin F(ab)2 fragments conjugated to horseradish peroxidase, followed by diaminobenzidine/H2O2 reaction. Effects of fixation and pre-treatments on RNA detection efficiency and morphology were monitored by RCM on whole cells. After Epon embedding and ultra-thin cross-sectioning, the corresponding TEM images were obtained. With the pre-treatments analyzed, it appeared impossible to find an acceptable balance between ISH signals and preservation of ultrastructural morphology: when good signal-to-noise ratios are obtained, the ultrastructural morphology is already deteriorated. We discuss the parameters that influence the fragile balance between high RNA detection efficiency and good preservation of ultrastructure and the benefit of RCM monitoring in the development and procedures for pre-embedding electron microscopic ISH.