A theoretical examination is made of the use of the stereoscopic technique for the determination of vertical elevations in electron microscope specimens. It is shown that the variation in optical object distance due to tilting the specimen may introduce serious errors even for object points close to the optic axis of the microscope. Other possible sources of error are also analysed, and it is concluded that in practice, apart from limits due to resolution, these are likely to be of minor importance in relation to the tilt error and the precision attainable in the measurement of parallax. Consideration has been given to possible means of reducing the tilt error: no wholly satisfactory solution has been found, and it is necessary for precision work to maintain the axis of tilt of the specimen close to the optic axis within very narrow limits.