In　this　study,　nine　laboratory　tests　were　conducted　to　estimate　the　performance　of　earthquake-resistant　structures,　i.e.,　reduced　beam　section　(RBS)　connection,　under　a　scenario　in　which　the　middle　column　is　removed.　Four　types　of　joints　were　tested,　namely,　welded　unreinforced　flange　-　bolted　web　connection　without　RBS　(WUF-B-NRBS),　welded　unreinforced　flange　-　bolted　web　connection　with　RBS　(WUF-B-RBS),　welded　unreinforced　flange　-　welded　web　connection　without　RBS　(WUF-W-NRBS),　and　welded　unreinforced　flange　-　welded　web　connection　with　RBS　(WUF-B-RBS),　to　investigate　the　effect　of　RBS　and　different　beam　web　connection　methods.　The　experimental　results　involving　the　responses　of　general　behavior,　resistance　mechanisms,　failure　modes,　deformation　shapes,　and　limit　displacements　are　presented　in　detail.　A　performance　comparison　of　different　connection　types　indicated　that　the　RBS　connection　exhibits　the　optimum　performance　against　progressive　collapse,　and　its　practical　rotational　capacity　is　＞1.9　and　2.7　times　larger　than　that　provided　by　Federal　Emergency　Management　Agency　(FEMA)　350　and　Department　of　Defense　(DOD)　guidelines,　respectively.　Additionally,　in　the　design　of　the　RBS　connection,　the　welded　web　connection　joint　exhibited　good　load-carrying　capacity,　whereas　the　bolted　web　connection　joint　performed　poorly　owing　to　early　fracture　near　the　bolt　holes.　Furthermore,　the　research　parameters　were　extrapolated　within　and　beyond　the　limits　using　ABAQUS.　This　demonstrated　that　by　increasing　the　values　of　horizontal　distance　from　the　column　flange　to　the　beginning　of　the　beam　flange　reduction,　a,　or　the　length　of　the　beam　flange　reduction,　b,　is　beneficial　to　structural　load-carrying　capacity　and　ductility.　However　increasing　the　value　of　depth　of　the　beam　flange　reduction,　c,　has　the　opposite　effect.　Based　on　the　comparison,　the　recommended　design　details　of　RBS　connections　are　presented.