Computer calculations and theoretical studies of different variants of current-source density analysis (CSDA) for neurophysological studies were carried out. The different variants of the method are compared by estimating equiweight surfaces and equiweight lines in particular planes. The decrease in the distance between the different points of measurement has been found to result in more selective presentation of the neuronal activity. It is shown that this distance cannot be very small--50 micron seems a reasonable limit. Bidimensional and especially unidimensional CSD-analysis produces considerable volumes with negative weights, these weights being rather high in absolute value. Errors in the interpretation of the results are possible, such as incorrect calculation of the current density and even erroneous determination of its sign.