Two common trial functions used in the variational method to solve the waveguiding modes of Ti:LiNbO 3 channel waveguides are compared in this paper. The first one has four parameters and the other has just two. The respective modal-field profiles described by them, as well as the respective fiber-to-waveguide coupling efficiencies estimated by these models, are rather different. We propose a self-consistent concept to examine the qualities of the trial fields and then show that the two-parameter model is better.