In a recent paper, Hawkins (1997) argues on the basis of statistical studies of double-image gravitational lenses and lens candidates that a large population of dark lenses exists and that these outnumber galaxies with more normal mass-to-light ratios by a factor of 3:1. If correct, this is a very important result for many areas of astronomy including galaxy formation and cosmology. In this paper we discuss our new radio-selected gravitational lens sample, JVAS/CLASS, in order to test and constrain this proposition. We have obtained ground-based and HST images of all multiple-image lens systems in our sample and in 12 cases out of 12 we find the lensing galaxies in the optical and/or near infrared. Our success in finding lensing galaxies creates problems for the dark lens hypothesis. If it is to survive, ad hoc modifications seem to be necessary: only very massive galaxies (more than about one trillion solar masses) can be dark, and the cutoff in mass must be sharp. Our finding of lens galaxies in all the JVAS/CLASS systems is complementary evidence which supports the conclusion of Kochanek et al. (1997) that many of the wide-separation optically-selected pairs are physically distinct quasars rather than gravitational lens systems.