In the conventional linear calculation method of lateral earth pressure by K0 stepped blade (KSB) test, there is a certain error caused by the difference between the fact and the local linear assumption of e-lg p curve, which leads to a large difference between the prediction results and the actual measurements for overconsolidated soils. In view of that, a series of in situ tests were conducted on the structural silty clay of an overconsolidated stratum in Harbin City. Horizontal compression tests were employed to investigate the mechanical response of the lateral soil during penetration. A nonlinear calculation method for KSB was established based on the characteristics of horizontal compression curves. Moreover, the applicability of the proposed method was analyzed from the perspective of experimental mechanism, and the judgment condition for its application was formed. The results show that the conventional linear method had the disadvantages of low linearity of the data and overestimated lateral stress, the effective data of this method only reached 33%, whereas that of the nonlinear method exceeded 75%. Moreover, the lateral stresses calculated by linear method were 1.2-2.2 times higher than that by the nonlinear method. Further analysis indicates that the above phenomena mainly resulted from the nonlinearity of e-lg p curves of overconsolidated soil, which were quantified by the contact pressure index λ and nonlinearity index η. The deviation of linear method Δp0 firstly increased, then decreased with the increase of λ, but increased with the increase of η. In practical engineering, the nonlinear method is suitable for soil with high nonlinearity of e-lg p curves, such as structural soil and overconsolidated soil. When λ>1, the linear method is recommended because it is simple and convenient with accurate prediction results; otherwise, the nonlinear method is recommended.