Theoretical considerations together with simulations of single-particle electron cryomicroscopy images of biological assemblies in ice demonstrate that atomic structures should be obtainable from images of a few thousand asymmetric units, provided the molecular weight of the whole assembly being studied is greater than the minimum needed for accurate position and orientation determination. However, with present methods of specimen preparation and current microscope and detector technologies, many more particles are needed, and the alignment of smaller assemblies is difficult or impossible. Only larger structures, with enough signal to allow good orientation determination and with enough images to allow averaging of many hundreds of thousands or even millions of asymmetric units, have successfully produced high-resolution maps. In this review, we compare the contrast of experimental electron cryomicroscopy images of two smaller molecular assemblies, namely apoferritin and beta-galactosidase, with that expected from perfect simulated images calculated from their known X-ray structures. We show that the contrast and signal-to-noise ratio of experimental images still require significant improvement before it will be possible to realize the full potential of single-particle electron cryomicroscopy. In particular, although reasonably good orientations can be obtained for beta-galactosidase, we have been unable to obtain reliable orientation determination from experimental images of apoferritin. Simulations suggest that at least 2-fold improvement of the contrast in experimental images at ~10 Å resolution is needed and should be possible.