Directional cell movement from a softer to a stiffer region on a culture substrate with a stiffness gradient, so-called durotaxis, has attracted considerable interest in the field of mechanobiology. Although the strength of a stiffness gradient has been known to influence durotaxis, the precise manipulation of durotactic cells has not been established due to the limited knowledge available on how the threshold stiffness gradient (TG) for durotaxis is determined. In the present study, to clarify the principles for the manipulation of durotaxis, we focused on the absolute stiffness of the soft region and evaluated its effect on the determination of TG required to induce durotaxis. Microelastically patterned gels that differed with respect to both the absolute stiffness of the soft region and the strength of the stiffness gradient were photolithographically prepared using photo-cross-linkable gelatins, and the TG for mesenchymal stem cells (MSCs) was examined systematically for each stiffness value of the soft region. As a result, the TG values for soft regions with stiffnesses of 2.5, 5, and 10 kPa were 0.14, 1.0, and 1.4 kPa/mu m, respectively, i.e., TG markedly increased with an increase in the absolute stiffness of the soft region. An analysis of the area and long-axis length for focal adhesions revealed that the adhesivity of MSCs was more stable on a stiffer soft region. These results suggested that the initial location of cells starting durotaxis plays an essential role in determining the TG values and furthermore that the relationship between the position-dependent TG and intrinsic stiffness gradient (IG) of the culture substrate should be carefully reconsidered for inducing durotaxis; IG must be higher than TG (IG >= TG). This principle provides a fundamental guide for designing biomaterials to manipulate cellular durotaxis.