In order to counteract the pathologies of cooperative coevolutionary algorithms(CCEAs) caused by information loss when dealing with problem decomposition, the information compensation strategy is investigated with respect to the dynamic nature of the landscapes of the CCEAs. A dynamic multi-population evaluation based anti-pathology CCEA is proposed. In the algorithm, several dynamic child populations can be split off from a coevolutionary population and search global or local optimum which are used as the interacting information to compensate information. Two pathology-causing benchmark problems are used to test and compare the proposed algorithm to three classical CCEAs. Experimental results show that the proposed algorithm effectively counteract the relative overgeneralization pathology and significantly improve the rate of global-optimization convergence.