The　direct　application　of　the　Kalman　filtering　algorithm　to　state　estimation　of　civil　structures　presents　a　computational　challenge　due　to　their　great　dimension　and　complexity.　Rewriting　the　second-order　dynamic　equation　using　modal　coordinates　is　an　alternative　to　solve　this　problem,　because　the　high　modes　that　often　have　minimal　and　negligible　contributions　to　structural　responses　can　be　truncated.　However,　the　determination　of　the　remained　mode　number　is　of　paramount　importance　to　accurately　estimate　the　state.　This　paper　investigates　the　mode　selection　method　in　Kalman　filtering　for　possibly　best　reconstructing　structural　responses.　A　signal-to-noise　ratio　(SNR)　is　defined　as　the　root　square　of　variance　of　estimated　modal　coordinate　vector　to　the　corresponding　estimation　error　covariance.　The　modes　with　higher　SNR　values　than　a　predefined　threshold　are　selected.　Experimental　studies　using　a　simply　supported　overhanging　beam　are　conducted　to　verify　the　proposed　method.　The　experimental　results　demonstrate　the　proposed　method　is　effective　and　flexible　in　mode　selection　for　accurate　response　reconstruction.