DNase I has been widely used for the footprinting of DNA–protein interactions including analyses of nucleosome core particle (NCP) structure. Our understanding of the relationship between the footprint and the structure of the nucleosome complex comes mainly from digestion studies of NCPs, since they have a well-defined quasi-symmetrical structure and have been widely investigated. However, several recent results suggest that the established consensus of opinion regarding the mode of digestion of NCPs by DNase I may be based on erroneous interpretation of results concerning the relationship between the NCP ends and the dyad axis. Here, we have used reconstituted NCPs with defined ends, bulk NCPs prepared with micrococcal nuclease and molecular modelling to reassess the mode of DNase I digestion. Our results indicate that DNase I cuts the two strands of the nucleosomal DNA independently with an average stagger of 4 nt with the 30-ends protruding. The previously accepted value of 2 nt stagger is explained by the finding that micrococcal nuclease produces NCPs not with flush ends, but with approximately 1 nt 50- recessed ends. Furthermore we explain why the DNA stagger is an even and not an odd number of nucleotides. These results are important for studies using DNase I to probe nucleosome structure in complex with other proteins or any DNA–protein complex containing B-form DNA. We also determine the origin of the 10n ^ 5 nt periodicity found in the internucleosomal ladder of DNase I digests of chromatin from various species. The explanation of the 10n ^ 5 nt ladder may have implications for the structure of the 30 nm fibre.