Roberta Humphreys' studies of “Luminous Stars in Nearby Galaxies” (Papers 1 – VIII) demonstrated the existence of an empirical upper luminosity boundary in the HR Diagram and by implication an upper limit to the masses of stars that could evolve to become red supergiants (Humphreys & Davidson 1979). This upper limit, sometimes called the Humphreys-Davidson limit, was not predicted by theory and greatly influenced future work on massive star evolution. They suggested then and in two subsequent papers (1984, 1994) that massive stars above this limit encountered an instability, possibly due to the opacity-modified Eddington limit, and experienced high mass loss episodes which prevented their evolution to cooler temperatures. The Luminous Blue Variables (LBVs) and stars like eta Car are examples of this high mass loss phase. As part of this work she also obtained the first spectra and IR photometry of individual stars outside the Local Group. In recent years, she has used the HST to study individual evolved, cool stars that define this upper luminosity limit, the cool hypergiants. Using high resolution spectroscopy and multi-epoch imaging, she has found that some of these stars experience localized high mass loss events probably due to large scale convective activity.She also led the Automated Plate Scanner (APS) project for nearly 20 years. One of their major activities was the digitization of the POSS I and creation of the on-line database (aps.umn.edu). The APS group was the first to use neural networks for object classification. She and her students (Larsen, Parker and Cabanela) have used the database for large scale studies of the Milky Way.