In this paper we present the problem of representing the morpho-syntactic relations in wordnets, especially the problems that arise when wordnets for languages that differ significantly from English are being developed on the basis of the Princeton wordnet, which is the case for Bulgarian and Serbian. We present the derivational characteristics of these two languages, how these characteristics are presently encoded in corresponding wordnets, and give some guidelines for their better coverage. Finally, we discuss the possibility to automatically generate new synsets and/or new relations on the basis of the most frequent and most regular derivational patterns.