Fatigue crack propagation tests with various stress ratios and stepwise decreasing maximum load while keeping minimum load are carried out for center notched tension specimens. In these tests, RPG load (Re-tensile Plastic zone's Generated load), crack length and number of cycles are measured by using our newly developed measuring system in the 1st report. We also calculate RPG load by using our developed simulation model of crack opening/closing phenomena based upon Dugdale model in the 5th report with various shrinkage coefficients α of plastic deformation due to emancipate internal force in newly generated crack surface at an immediate crack growth. Then we obtain α value for the steel used with comparing between simulated RPG load and measured one for the testing condition of R =0.05. By using the obtained α value, RPG loads for other stress ratios ( R =0.3 and 0.5) are calculated by using the simulation model. As a result, simulated RPG loads are in very good agreement with measured RPG loads for R =0.3 and R =0.5. Then a crack growth curve for each testing condition is calculated by using Δ K_RP from simulated RPG load and also shows in very good agreement with measured crack growth curve for each stress ratio. RPG load increases just after decreasing maximum load and reaches maximum and then decreases gradually as a crack advances in the cases of decreasing load with small magnification. Then delayed deterdation effects appear and delayed life increases with larger decrement of maximum load. It is also comfirmed that simulated RPG load using above α value is in good agreement with measured RPG load for these cases. It becomes also clear that crack growth curve in case of decreasing maximum load can be assessed quantitatively by using Δ K_RP based upon the simulated RPG load. Moreover it is confirmed that larger decrement of maximum load in stepwise loading leads to stopping a crack even when final maximum load with large decrement of maximum load is larger than RPG load. This is because RPG load increases rapidly just after decreasing maximum load and approaches to final maximum load in case of large decrement of maximum load both in experiments and simulations. Simulated crack growth curve by Δ K_RP using calculated RPG load in this case is also in good agreement with experimental one. Therefore it can be expressed that the condition of stopping a crack is Δ K_RP _??_0 for a long crack. In other word, it becomes clear that Δ K_th or ( Δ K_eff ) _th is not a material constant.