This paper addresses the synthesis and optimal design of liquid-liquid extraction processes using stochastic optimization algorithms. The main objective is to comparatively study stochastic optimization algorithms in view of speed and robustness when searching conventional as well as complex process representations for optimal solutions. The work employs a generic representation framework that accounts for simple as well as complex extraction process arrangements. The resulting superstructure formulations encompass all the potential configurations and interconnections that may exist for liquid-liquid extraction systems. The task of optimizing the superstructure is assigned to stochastic optimization algorithms in the form of simulated annealing (SA), genetic algorithms (GA) and ant colony optimization (ACO). The proposed model is applied to different separation problems involving quaternary mixtures. The liquid phase non-idealities are modeled both by highly non-linear correlations and standard thermodynamic methods. All three algorithms have been found to address the problem with success in terms of the quality of the solution found. (C) 2004 Elsevier Ltd. All rights reserved.