We utilized two widely used impregnation methods, the silver "rapid Golgi" and the mercuric Golgi-Cox methods, for three-dimensional (3-D) reconstruction of neurons at the confocal scanning laser microscope (CSLM), to determine which of them was more suitable for this application. The Golgi-Cox method is the most consistent arid the cleanest procedure with respect to the "rapid Golgi" one which always produces samples with scattered reflective granules that interfere with the image formation at the CSLM. The interneuronal tissue in the case of Golgi-Cox impregnated specimens (i.e. the non-impregnated tissue among impregnated neurons) contributes less to the decrease of reflected light during z-sectioning than in the case of "rapid Golgi" impregnation, but the mercury impregnated samples reflect less than the silver impregnated ones. Owing to the necessity during deep z-scanning to adjust the sensitivity of the CLSM detector the acquisition of images from the deeper planes of the sample may be difficult. In our opinion the "sandwich" mounting of the specimen between two coverslips is indispensable in order to make it possible to scan it from both sides and, thus reduce the penetration in the sample and the consequent distortion of the image. Neither of the impregnation methods used is completely suitable for CLSM observations due both to their intrinsic limitations and to those imposed by the sample thickness.