A most probable path is selected from a number n (where n≧2) of paths of state transitions occurring from a first state Sk−j (where k≧3 and j≧2) at a time k−j into a second state Sk at a time k. The method includes the steps of detecting predetermined combinations of the first and second states Sk−j and Sk defining the n probable paths in a predetermined period j between the times k−j and k, and evaluating the reliability of a read signal, decoded in the period j, by |Pa−Pb|. Pa and Pb indicate the probabilities of state transition of first and second state transition paths in the period j. The first and second state transition paths are estimated to be the most probable and the second most probable, respectively, among the n probable paths defined by the combinations detected.