Experiments on minority carrier diffusion using an a.c. photocurrent method exhibit very good agreement with Klaassens's model [1] at temperatures 250K -500K, but very large deviations at 100K and intermediate doping levels. They can be perfectly explained using published results on exciton diffusion constants and on concentrations of excitons in electron-hole-plasma. Our results confirm strongly that simulation of bipolar devices at low temperatures requires consideration of exciton formation.