In this paper two different texture-dependent material models based on the Taylor assumption are discussed and applied to the simulation of deep drawing operations of aluminium. From the numerical point of view, large-scale FE computations based on the Taylor model are very time-intensive and storageconsuming if the crystallographic texture is approximated by several hundred discrete crystals. Furthermore, the Taylor model in its standard form, which is based on discrete crystal orientations, has the disadvantage that the anisotropy is significantly overestimated if only a small number of crystal orientations are used. We quantitatively analyse this overestimation of anisotropy and suggest two Taylor-type models which allow us to reduce the sharpness of the crystallite orientation distribution function related to single crystals or texture components. One model is an elastic-viscoplastic Taylor model based on discrete orientations. The sharpness is reduced by modelling the isotropic background texture by an isotropic material law. The other model is a rigidviscoplastic material one, which is based on continuous model functions on the orientation space. This model allows for a direct incorporation of the scattering around an ideal texture component since the model contains the half-width as a microstructural parameter which can be biased. These models are used to compute yield stresses, R values and earing profiles. The predictions are compared with experimental data.