Despite the success of Betzig´s work on near-field fluorescence microscopy (e.g. Betzig et al., 1992; 1993) it was becoming increasingly clear that the confinement to surface studies and the cumbersome implementation limited the use of these methods. Betzig decided to leave academic science, but continued to think about super-resolved fluorescence microscopy. He eventually perceived a new way to achieve super-resolution without the shortcomings of near-field microscopy (Betzig, 1995). The solution, he suggested, was to determine the positions of a large number of point sources with distinguishable spectral properties in two steps. In a first step the PSF of each spectral class was determined separately. Making sure that each spectral class formed a sparse set in space, it would be possible to determine the positions of their members with super resolution by estimating the centers of their PSFs with the precision of Eq. 5. By considering the positions of all classes together, a super-resolution image of a densely sampled structure could be obtained. A particular implementation of this method was subsequently used in experiments by G.J. Brakenhoff and collaborators (van Oijen et al., 1998), but an optimal way to experimentally realize Betzig’s vision was still lacking.