String effective theories contain a dilaton scalar field which couples to gravity, matter and radiation. In general, particle masses will have different dilaton couplings. We can always choose a conformal frame in which baryons have constant masses while (nonbaryonic) dark matter have variable masses, in the context of a scalar-tensor gravity theory. We are interested in the phenomenology of this scenario. Dark matter with variable masses could have a measurable effect on the dynamical motion of the halo of spiral galaxies, which may affect cold dark matter models of galaxy formation. As a consequence of variable masses, the energy-momentum tensor is not conserved; there is a dissipative effect, due to the dilaton coupling, associated with a “dark entropy” production. In particular, if axions had variable masses they could be diluted away, thus opening the “axion window.” Assuming that dark matter with variable masses dominates the cosmological evolution during the matter era, it will affect the primordial nucleosynthesis predictions on the abundances of light elements. Furthermore, the dilaton also couples to radiation in the form of a variable gauge coupling. Experimental bounds will constrain the parameters of this model.