A comprehensive program of integral experiments, largely based on the measurement of reaction rate distributions, was carried out recently in the PROTEUS zero-power research reactor at the Paul Scherrer Institute in Switzerland, employing a fuel lattice resembling that of a supercritical light water reactor. The present paper reports on the analysis of a complementary set of measurements, in which the reactivity effects of removing individual pins from the unperturbed, heterogeneously moderated reference lattice were investigated. It has been found that the detailed Monte Carlo modeling of the whole reactor using MCNPX is able-as in the case of the reaction rate distributions to reproduce the experimental results for the pin removal worths within the achievable statistical accuracy. A comparison of reduced-geometry calculations between MCNPX and the deterministic light water reactor assembly code CASMO-4E has revealed certain discrepancies. On the basis of a reactivity decomposition analysis of the differences between the codes, it has been suggested that these could be due at least partly to CASMO-4E deficiencies in calculating the effect, upon pin removal, of the extra moderation in the neighboring fuel pins.