With the increase in the demand for lithium, Li-bearing minerals could be considered as alternative resources to achieve the supplying. So, efficient technological solutions for the valorization of these minerals are required. In this context, the froth flotation process of a lepidolite ore was modelled and optimized. Closely following the response surface methodology (RSM), the effects of three independent process variables (pulp pH, flotation collector dosage and flotation time) upon two common measures of the separation (lithium recovery and lithium content) were studied. These were modelled using the experimental data obtained starting with the implementation and execution of a full 23 factorial design and ending with a (face-centered) central composite design (CCD), a second order design. The coefficients of the second-order polynomial regression models were fitted by solving linear least squares problems. After statistical validation, the fitted models were used to support the identification of the significant effects of the process variables and to provide estimations of the measures of separation (responses) for combinations of the levels of the process variables over a feasible region of interest. Using directly the measured values of Li recovery and Li content, the selected experimental Pareto optimal combination of the levels of the process variables is: pulp pH = 2, dosage of collector = 500 g.t(-1) and flotation time = 12 min producing a concentrate with Li recovery of 91.51% and a Li content of 1.96%. Using the fitted second order models for the separation criteria, a refined Pareto optimal combination was obtained as the solution of the multicriteria optimization (maximization) problem that was solved by different methods (Weighted Sum of Objectives, Goal Programming and Desirability Functions). The refined Pareto optimal combination was the same than the selected experimental Pareto optimal combination, only the collector dosage decreased to 470478 g.t(-1), producing a concentrate with Li recovery around 92.50% and a Li content of 2.00%. (C) 2016 Elsevier B.V. All rights reserved.