Predictions of the Chilcott-Rallison FENE dumbbell model for polymer configuration in start-up flow of a co-rotating two-roll mill, are compared with birefringence measurements in a 40 ppmw solution of high molecular weight polystyrene. Two versions of the Chilcott-Rallison model are considered, one with a constant bead friction and the other with bead friction that is assumed to increase in proportion to the end-to-end dimension of the dumbbell. Parameters for the model were determined via independent experiments. We show that the flow, at this concentration level, is unchanged from the form for a Newtonian fluid. Measurements and predictions of the birefringence are compared at the stagnation point of the flow, where a high degree of polymer stretch is possible. The Chilcott-Rallison model with constant bead friction, and an approximation to the inverse Langevin spring function, gives excellent agreement with the experimental results.