The fraction of high-redshift sources which are multiply-imaged by
intervening galaxies is strongly dependent on the cosmological constant, and so can be a useful probe of the cosmological model. However its power is limited by various systematic (and random) uncertainties in the calculation of lensing probabilities, one of the most important of which is the dynamical normalisation of elliptical galaxies. Assuming ellipticals' mass distributions can be modelled as isothermal spheres, the mass normalisation depends on: the velocity anisotropy; the luminosity density; the core radius; and the area over which the velocity dispersion is measured. The differences in the lensing probability and optical depth produced by using the correct normalisation can be comparable to the differences between even the most extreme cosmological
models. The existing data is not sufficient to determine the correct
normalisation with enough certainty to allow lensing statistics to be used to their full potential. However, as the correct lensing probability is almost
certainly higher than is usually assumed, upper bounds on the cosmological constant are not weakened by these possibilities.