A computational procedure is described to study the mixing flow in a multilobe turbofan mixer. The predictions have been obtained using a finite volume method to solve the density-weighted time-averaged Navier-Stokes equations. Turbulence is characterized by the k-ε eddy viscosity model. To fit the irregular boundaries of the flowfield, curvilinear nonorthogonal coordinates are employed. The robustness of the computational procedure is enhanced by making use of nonstaggered grids. In the calculation the computational domain contains not only the mixing duct but also the lobe itself. Three kinds of configuration are under consideration: a confluent mixer, a convoluted mixer, and a forced mixer. Results show that the stream wise vortex generated at the trailing edge is the most prominent flow structure for the forced mixer, responsible for the effective mixing. The strength of the streamwise vortex can be characterized by the circulation. A parameter termed mixedness is defined to describe the effects of the mixing process. In addition to the streamwise vortex, normal vortices are generated by the velocity difference in the shear layer between the core and fan streams. To illustrate the mechanism of the mixing procedure, velocity vectors and contours of streamwise and normal vorticities and turbulence contours on transverse planes at selected axial locations are presented. The roles played by the streamwise vortex and the normal vortex are clearly identified.