This　paper　studies　the　multiscale　features　of　extreme　risk　spillover　among　global　stock　markets　over　various　time-frequency　horizons.　We　propose　multiscale　risk　spillover　indexes　based　on　GARCH-EVT-VaR,　maximal　overlap　discrete　wavelet　transform　method,　and　forecast-errorvariance　decompositions.　We　further　construct　multiscale　risk　spillover　networks　to　visualize　risk　spillovers　at　different　scales.　Our　findings　show　that　the　US　and　the　UK　are　detected　as　the　centers　of　risk　spillovers,　while　Asian　stock　markets　are　mainly　at　the　edge　of　the　risk　spillover　network.　The　topological　properties　are　unevenly　spread　over　each　time　scale.　The　network　tends　to　be　closer　not　only　at　the　short-term　scale　but　also　during　the　financial　crisis.　For　individual　features,　the　US　and　the　UK　are　super-spreaders　of　risk　spillover　at　each　time　scale,　while　most　developing　markets　mainly　act　as　absorbers.　The　role　of　European　stock　markets　is　complex　at　different　scales.