This paper presents results from 21 large-scale experiments on steel columns that examined their collapse behavior as related to steel moment resisting frames (MRFs). The test specimens included wide-flange and square hollow structural shapes (HSS) and were tested until complete loss of their lateral load-carrying capacity. The test results suggest that the collapse rotations of steel columns under collapse protocols, representing near-fault events, are two to three times larger than those of their counterparts under standard symmetric cyclic protocols, regardless of the examined cross-sectional compactness. Conversely, when ground motion duration is an important seismic hazard characteristic, the collapse behavior of steel columns is reasonably traced with a standard symmetric cyclic protocol. It is shown experimentally that steel columns have an inherent reference energy dissipation capacity regardless of the employed loading history. Axial shortening attributable to local buckling is up to 10 times larger in interior than in end columns, which leads to differential axial shortening within a steel MRF story even at modest lateral drift demands during subduction zone seismic events. The test data underscore that the formation of local buckling at a column's fixed end caps the strain demands at about 1% near complete joint penetration welds between the column and the base plate, even when the transient axial load demand becomes tensile. Simpler welds may provide adequate ductility in steel columns featuring cross sections near the compactness limits for highly ductile members according to current design standards. It is shown that the estimated rotations by current ASCE 41 acceptance criteria for collapse prevention are 5-10 times smaller than the measured collapse rotations, even for columns with moderately compact profiles. (C) 2021 American Society of Civil Engineers.