We consider the well known NP–hard teacher/class timetabling problem. Variable neighborhood search and tabu search heuristics are developed based on idea of the Formulation Space Search approach. Two types of solution representation are used in the heuristics. For each representation we consider two families of neighborhoods. The first family uses swapping of time periods for teacher (class) timetable. The second family bases on the idea of large Kernighan-Lin neighborhoods. Computation results for difficult random test instances show high efficiency of the proposed approach.