It is shown that the contrast in high-resolution electron micrographs of biological macromolecules, illustrated by a study of TMV in ice, falls considerably below the level which should theoretically be attained. The factors which contribute to the low contrast include radiation damage, inelastic scattering, specimen movement and charging. Future progress depends on improved understanding of their contributions and relative importance. Contrast is defined as the amplitude of a particular Fourier component extracted from an image in comparison to that expected by extrapolation from separate electron or X-ray diffraction measurements. The fall in contrast gets worse with increased resolution and is particularly serious at 10 A and beyond for specimens embedded in vitreous ice, a method of specimen preparation which is otherwise particularly desirable because of the expectation that the embedded molecules should be well preserved in a near-native environment. This low contrast at high resolution is the principal limitation to atomic-resolution structure determination by electron microscopy. In spite of good progress in the direction of better images, it remains a major problem which prevents electron microscopy from becoming a simple and rapid method for biological atomic structure determination.