Abstract: The present invention is a method for identifying a Most-Probable-Point (MPP) in original space obviating the need for a probability transformation, for use in first order/second order reliability analysis. It comprises generating the linear approximation of a limit-state-function, g(x) about the median, mean point, or mode of random variables, x of g(x). g(x) is defined so that g(x)f0 denotes a failure set. The Most-Probable-Point-Locus (MPPL) of g.sub.1, MPPL.sub.1, is constructed by the steps of: i) identifying the mode of x; ii) identifying the MPP of g.sub.1 (x)=c, where c is an arbitrary constant, and iii) constructing said MPPL.sub.1 by connecting said mode of x from step i), above, and the MPPs corresponding to different c's from step ii), above. A quadratic search algorithm is to identify point MPP.sub.1, the intersection of MPPL.sub.1 and g(x)=0, based on the following convergence criteria: ##EQU1## The process is then stopped, unless the convergence criteria are not satisfied. g.sub.