The development of methods to reconstruct phylogenies from karyotypic data has lagged behind what has been achieved with molecular and morphological characters. This hampers our understanding of the role of chromosomal rearrangements in speciation, which depends on knowledge of the karyotypic relationships both among forms that have recently speciated and among forms within species that may speciate in the future. Here, we present a new approach to reconstruct chromosomal phylogenies. Our approach involves the use of networks, which we believe offer a flexible alternative to bifurcating phylogenetic trees for chromosomal phylogenetic analyses, and can incorporate a wide range of chromosomal mutations as well as allowing reticulate evolution through hybridization. In this paper, we apply our method at the within-species level to establish the phylogenetic history, in terms of minimum number of evolutionary steps, of chromosomal races within both the common shrew (Sorex araneus) and the house mouse (Mus musculus). There have been several previous attempts to reconstruct the phylogenies of chromosomal races within shrews and mice, but we describe the first computer-based analysis that considers the whole range of possible mutations generating new races (Robertsonian fusions and fissions and whole-arm reciprocal translocations [WARTs]) and other race-generating processes (zonal raciation events involving both acrocentric and recombinant peaks) postulated for these species. The analysis for common shrew chromosomal races reveals a greater importance of zonal raciation and WARTs than has been suggested hitherto.