In the reconstruction of helical particles, it is normally assumed that translation along the length of a particle is coupled to rotation about its axis. This assumption is not valid for particles whose pitch varies along the particle length (e.g. actin, HbS fibers), and application of the usual algorithms results in significant errors in both the shape and coordinates of subunits in the reconstructed density map. We have developed an iterative procedure for reconstructing particles with variable pitch. The goal of this procedure is to obtain an accurate estimate of the local pitch of the particle which can then be incorporated into the reconstruction algorithm. This involves synthesis of trial model structures which have constant pitch. The local pitch is derived from a cross-correlation analysis between these trial models and the variable pitch particles. The constant pitch models are constructed using coordinates measured from the reconstructed density maps. Each iteration of the procedure provides an improved estimate of the pitch which is incorporated into the succeeding iteration. The fidelity of the reconstruction is determined from cross-correlation between the original micrograph and a variable pitch model. The iterations are continued until the cross-correlation coefficient between the variable pitch model and the micrograph of the particle is maximized. The implementation of the iterative procedure is described and its behavior is evaluated using model structures which incorporate variations in pitch similar to those actually occurring in sickle hemoglobin fibers. The results indicate that the iterative reconstruction procedure considerably reduces the errors associated with constant pitch reconstructions. These tests provide a basis for applying this procedure in the structural analysis of micrographs of helical particles which display variable pitch. Application to sickle hemoglobin fibers resulted in an improvement in the accuracy with which the hemoglobin S molecules can be located in the density maps.