This paper discusses some possible procedures for crystal structure determination using the $\chi $ criterion developed in the preceding paper of the series. In that paper it was shown that, under favourable circumstances, the signs of the largest terms in the Fourier series representation of $\rho $, the electron density in the unit cell, can be determined directly. The magnitudes of these terms can be found by experiment. In less favourable cases the correct set of signs, to a high degree of probability, lies among several hundred possible sets which can be determined by the processes described. A new criterion is proposed for selection of a lesser number, say a dozen, 'most probable' sets of signs. These latter sets can be tested by inspection of the corresponding contour maps of $\rho $. The new criterion has been successfully used in determining the unknown structure of nitroguanidine. Techniques for the selection of most probable sets and also for evaluation of $\rho $ at suitable points in space using an electronic computer, the EDSAC, are described.