The surface of thin sections of aldehyde-fixed biological material shows a specimen-related relief of 2-6 nm with Lowicryl. Epon sections are about three times smoother. The relief is the consequence of thin-sectioning being in reality a cleavage. Epitopes are supposed to be laid open (or set free) because cleavage follows the interfaces between protein and Lowicryl. We have developed a simple theory on this basis and have theoretically estimated the efficiency of on-section labeling and compared it with experimental data. For randomly dispersed proteins in cytoplasm, Lowicryl sections will yield significant label only when the concentration of the antigen is about 10 microM or more. The complex situation of more compact proteins, as represented by fibers, sheets, and biological membranes is discussed and the difficulty of significant calculations is explained. Pre-embedding labeling and melted cryosections should give 10-30 times more label. The possible reasons for the observed much smaller gain of not more than two to three times are discussed.