The paper deals with the solution of the optimal short-term unit commitment (UC) problem in an electric power system. The optimization model takes into account the main operating constraints and physical characteristics of the power generation system. A comparison between Lagrangian heuristics and Tabu Search techniques on different classes of realistic instances is presented. Such a comparison is aimed at highlighting the strong features and the weaknesses of each technique, in view of their application in computer models for competitive electricity markets in progressive evolution. The comparison may provide insights for the construction of hybrid techniques that incorporate the best of both approaches.