Recent studies on avian and mammalian embryos have established that the epicardium is derived, not from the early heart tube, but from mesothelial tissue overlying the sinus venosus. We tested the validity of this concept for Amphibia by examining normal and cardiac lethal (c/c) mutant axolotl embryos (stages 35-43) by electron microscopy. In axolotl embryos, the myocardial surface of the heart remains exposed to the pericardial fluid through stage 39. At this stage the transverse septum releases into the pericardial cavity mesothelial cells that subsequently flatten over the adjacent ventricular myocardium. However, mesothelial cells observed on the developing epicardium always appear rounded and may extend a filopodium up to 75 microns. This apparent "substrate-dependent" difference in mesothelial cell shape may promote the extension of the epicardium over the rest of the myocardium. The initial site of epicardial formation persists in the adult as the ventricular pericardial stalk that connects the epicardium to the peritoneal lining of the transverse septum. Cardiac lethal (c/c) mutant embryos, despite the non-contractility of their myocardia, form their epicardia in the same way. This suggests that the c/c mutation does not impair those properties of the myocardium that render it a suitable substrate for epicardial spreading. The abnormal pattern of epicardial coverage of the edematous stage 41 c/c mutant heart could be the result of its abnormally large myocardial surface area, the abnormal proximity of the atrium to the transverse septum, and/or the absence of heart contractions which could aid the dispersion of mesothelial cells within the pericardial cavity. Despite species differences, epicardial development in the axolotl is similar to the general pattern described for higher vertebrate embryos.