A bias factor is commonly used to account for the effect of differences in failure criteria on reliability of pile foundations. There are however

hidden statistical dependencies between the bias factor and the underlying random variables. It is not reasonable that the statistical parameters of bias factor are directly adopted for reliability analysis of pile foundations. This paper aims to propose a new generalized bias factor based on the linear regression theory. The Spearman rank correlation method is used to perform the check of statistical dependencies. Finally

a numerical example is investigated to illustrate the effect of such statistical dependencies on reliability of pile foundations. The results indicate that the statistical dependencies between the bias factor defined by previous studies; and the underlying random variables can be removed using the proposed definition of bias factor. Such dependencies have a significant influence on the mean of bias factor

but they have hardly influence on the coefficient of variation of bias factor. Reliability indexes are also influenced by such dependencies. Therefore it is recommended that such dependencies be removed so that reliability of pile foundations can be evaluated in a more rational way.