Three types of linear polarization resistance (LPR) may provide an approximate value of polarization resistance ( R p) for estimating the corrosion current density i corr: an anodic LPR, R a, a cathodic LPR, R c and a bi-directional LPR, R b. Theoretical errors of the LPS's depend upon two factors. One is related to the symmetric characteristic between anodic and cathodic branches of polarization curve and may be described by a parameter α = ( β II)/( β I), where β denotes the Tafel slopes and the subscripts indicate smaller and larger values respectively. The other factor is related to the polarization measurement of the LPR and can be expressed by (| ΔE |;)/(β I), where | ΔE | is the absolute value of polarization in one direction. (1) For the same value of (| ΔE |)/(β I), the largest theoretical error for R b will occur as a = 1 or α → ∞ whilst for R a or R c this will be as α → ∞ (2) The theoretical error of R b is always less than that of R a and R c under the same conditions except in the case of α → 1. When α = 1, the value of all three types of LPR are equal to each other. (3) The theoretical errors of LPR values increase with increasing of the value of (| ΔE |)/(β I), but in most cases the increasing rate for R b is far less than that for R a or R c. Based on the mathematical analysis, a method to reduce the theoretical error of R b was proposed. If two measurements are made in succession with the polarization values ± | ΔE 1 | and ± m |; ΔE 1 |;, where 1 < m ⩽ 2 and | ΔE 1 | ⩽ β I, and if the bi-directional LPR's thus measured are denoted by R b1 and R b2 respectively, then theoretical error of R p(approx.) as estimated by the equation R p (approx.) = (m 2−1)R b1 R b2 m 2R b2 −R b1 would be far less than that of R b1 and R b2.