After a general introduction to three-dimensional electron microscopy and particularly to electron tomography (ET), the perspectives of applying ET to native (frozen-hydrated) cellular structures are discussed. In ET, a set of 2-D images of an object is recorded at different viewing directions and is then used for calculating a 3-D image. ET at a resolution of 2-5 nm would allow the 3-D organization of structural cellular components to be studied and would provide important information about spatial relationships and interactions. The question of whether it is a realistic long-term goal to visualize or-by sophisticated pattern recognition methods-identify macromolecules in cells frozen in toto or in frozen sections of cells is addressed. Because of the radiation sensitivity of biological specimens, a prerequisite of application of ET is the automation of the imaging process. Technical aspects of automated ET as realized in Martinsried and experiences are preresented, and limitations of the technique are identified, both theoretically and experimentally. Possible improvements of instrumentation to overcome at least part of the limitations are discussed in some detail, Those means include increasing the accelerating voltage into the intermediate voltage range (300 to 500 kV), energy filtering, the use of a field emission gun, and a liquid-helium-cooled specimen stage. Two additional sections deal with ET of isolated macromolecules and of macromolecular structures in situ, and one section is devoted to possible methods for the detection of structures in volume data. (C) 1997 Academic Press.