One goal of blockmodeling is to reduce a large, potentially incoherent network to a smaller comprehensible structure that can be interpreted more readily. Batagelj, Doreian, and Ferligoj developed a generalized approach to blockmodeling and methods where a set of observed relations are fitted to a pre-specified blockmodel. In the paper this generalized blockmodeling approach as implemented in program Pajek is described. An overview of the blockmodeling procedures in Pajek is given and is illustrated by some examples.