The weak beam technique is now used widely for the determination of stacking fault energies, in particular for intermetallic alloys, and the accuracy of the approach is critically dependent upon the reliability of the relationship between the image and the actual position of the dissociated dislocations. Examining as a model case a dislocation dissociated into two Shockley partial dislocations in Cu at 100 kV for orientations ranging through the g(3g) weak beam condition, image simulations are used to explore the accuracy to which the true spacing between the partial dislocations can be determined from the spacing measured on the image as a function of the dislocation character, the foil thickness, the dislocation depth in the foil, the diffraction condition and the beam convergence. It appears that for image simulations and for the given conditions a beam convergence of about 5 mrad allows to greatly improve the accuracy, and that beam convergence must be taken into account quantitatively when deducing the true partial dislocation spacing as it is the principal parameter controlling the precision in this type of measurement. (C) 2000 Elsevier Science B.V. All rights reserved.