Starting from a protein solution which is metastable with respect to the crystalline phase, the effect of adding semidilute nonadsorbing polymer is considered. It is found to increase the chemical potential of the protein by a few tenths of kT, which may be enough to lower the barrier to nucleation of the crystalline phase by enough to allow crystallization. It is also shown that assuming that the polymer induces a pairwise additive attraction leads to qualitatively incorrect results.