A major goal of contemporary astrophysics is understanding the origin of the most massive galaxies in the universe, particularly nearby ellipticals and spirals. Theoretical models of galaxy formation have existed for many decades, although observational evidence at both low and high redshifts is only beginning to put constraints on different ideas. We briefly describe these observations and how they are revealing the methods by which galaxies form by contrasting and comparing fiducial rapid collapse and hierarchical formation model predictions. The available data show that cluster ellipticals must have rapidly formed at z > 2, and that up to 50% of all massive galaxies at z ~ 2.5 are involved in major mergers. While the former is consistent with the monolithic collapse picture, we argue that hierarchical formation is the only model that can reproduce all the available observations.