Current understanding of the origins of cerebral specializa- tion is fairly limited. This chapter summarizes some recent work devel- oping and studying neural models that are intended to provide a bet- ter understanding of this issue. These computational models focus on emergent lateralization and also hemispheric interactions during recovery from simulated cortical lesions. The models, consisting of corresponding left and right cortical regions connected by the corpus callosum, handle tasks such as word reading and letter classification. The results demon- strate that it is relatively easy to simulate cerebral specialization and to show that the intact, non-lesioned hemisphere is often partially respon- sible for recovery. This work demonstrates that computational models can be a useful supplement to human and animal studies of hemispheric relations, and has implications for better understanding of modularity and robustness in neurocomputational systems in general.