Based on dimensionless load separation principle, the normalization method for estimating J-resistance curves of CIET specimen for ductile materials has been developed. Experiments of ductile crack growth resistance on a typical pressure vessel steel SA-508 are conducted by using CT, SEB and CIET specimens. The resulted Resistance curves show outstanding influence by load configuration and specimen dimensions. J-Q-M and J-A(2)-M approaches are introduced to quantify the crack front constraint of the three types of specimens. According to finite element analyses (FEA) under plane strain and 3D conditions, the distance-independent constraint parameters are calibrated by matching the FEA results with the J-Q-M and J-A(2)-M solutions using weight average method. The analyses show that the parameters of Q and A(2) are load-independent under plane strain condition, but under 3D condition, the J-A(2)-M approach is incapable of characterizing the stress and deformation fields ahead of crack front of CIET specimen for SA-508 steel, and the parameter Q is no longer load-independent. Because of different crack front constraint, the CIET specimen with B =10 mm, a(0) = 5.8 mm gets the lowest fracture resistance while the SEB specimen with B = 20 mm, a(0) =11 mm and the CT specimen with B =15 mm, a(0) = 29 mm obtain the highest fracture resistance for SA-508 steel. The size requirements of J-Q-M dominance for each used CIET, CT and SEB specimens are got. Finally, the constraint-modified J-resistance curves of SA -508 steel based on J-QM approach are constructed to predict the crack growth resistance for any fracture specimens or actual flawed structures. (C) 2016 Elsevier Ltd. All rights reserved.