The simultaneous use of Z-contrast imaging with parallel detection EELS in the STEM provides a powerful means for determining the atomic structure of grain boundaries. The incoherent Z-contrast image of the high atomic number columns can be directly inverted to their real space arrangement, without the use of preconceived structure models. Positions and intensities may be accurately quantified through a maximum entropy analysis. Light elements that are not visible in the Z-contrast image can be studied through EELS; their coordination polyhedra determined from the spectral fine structure. It even appears feasible to contemplate 3D structure refinement through multiple scattering calculations.The power of this approach is illustrated by the recent study of a series of SrTiC>3 bicrystals, which has provided significant insight into some of the basic issues of grain boundaries in ceramics. Figure 1 shows the structural units deduced from a set of 24°, 36° and 65° symmetric boundaries, and 24° and 45° asymmetric boundaries. It can be seen that apart from unit cells and fragments from the perfect crystal, only three units are needed to construct any arbitrary tilt boundary. For symmetric boundaries, only two units are required, each having the same Burgers, vector of a<100>. Both units are pentagons, on either the Sr or Ti sublattice, and both contain two columns of the other sublattice, imaging in positions too close for the atoms in each column to be coplanar. Each column was therefore assumed to be half full, with the pair forming a single zig-zag column. For asymmetric boundaries, crystal geometry requires two types of dislocations; the additional unit was found to have a Burgers’ vector of a<110>. Such a unit is a larger source of strain, and is especially important to the transport characteristics of cuprate superconductors. These zig-zag columns avoid the problem of like-ion repulsion; they have also been seen in TiO2 and YBa2Cu3O7-x and may be a general feature of ionic materials.