The object of this paper was to provide a basis for substantiating current specifications governing the use of intermittent fillet welds to connect the component parts of structural members. For instance, the bending rigidity is varied from a laminated girder to a solid one according to the type and size of fillet welds. So we studied in this paper the effects of leg length, pitch and flange thickness on the bending rigidity of a composed I-beam with various type and size of continuous or intermittent fillet welds. The concentrated lateral load was applied to a welded I-beam with both ends supported, and the deflections measured at nine points by dial gauges were compared with that of a solid one with the same cross section. The bending rigidity coefficient η of a welded girder can be estimated by equivalent leg length from this experimental results as follows. η=1/(K₁-K₂/1)+K₂tanhφl/2/φl+K₃/φ²l²(1-2tanhφl/2/φl) and φl=α(2L_e/t₁)^β{γ+δ(2h₁/t₂)} where K₁={1+6h²₂EI₀/l²(2I₁+I₂)G}/1+12kEI₀/l²A₀G K₂=24EA₁(h₁+h₂)/l²G{h₂²(h₁+h₂)/2I₁+I₂-1/t₂}/1+12kEI₀/l²A₀G K₃=24A₁(h₁+h₂)²/2I₁+I₂/1+12kEI₀/l²A₀G In respect of ultimate load, as it was varied from the various conditions such as leg length, pitch, joint types and dimensions of specimens, the ultimate load of the staggard was higher than the chain if the leg length was same. Types of fracture, classified in three types namely shear fracture of fillets, lateral buckling and local buckling, should be considered in case of the estimation of ultimate load of specimen.