A new algorithm, based on interval analysis, is proposed for global optimization of constrained nonlinear nonconvex functions of several variables. It exploits the cord-slope form of Taylor's expansion in several ways. Computational results, including a comparison with results of Sengupta is reported.