1 1. Introduction The AHP method was introduced by saaty (1980). The steps of this method consist of: 1. constructing hierarchy of criteria and alternatives, 2. asking decision makers to supply exact pairwise comparison ratios between sub-attributes (alternatives) of each attribute (sub-attribute) in each level of hierarchy, 3. Deriving local priorities of sub- attributes (alternatives), and 4. Deriving overall priority of each alternative. There are many criticisms to AHP method because of using exact rations in pairwise comparisons; therefore there is an extensive literature which extends classical AHP by using imprecise judgments. The earliest work in fuzzy AHP is Laarhoven and pedrycz's (1983) method which uses Lootsma's logarithmic least square method by triangular fuzzy numbers. Buckely (1985) by expressing Laarhoven and pedrycz method's shortcomings used geometric mean method by trapezoidal fuzzy numbers. Salo (1995) proposed a linear programming model for calculating consistency of perfect consistent matrices. Chang (1996) used arithmetic mean method by triangular fuzzy numbers. Leung and cao (2000) proposed an auxiliary linear programming model for fuzzy consistency test of pairwise comparisons by tolerance deviation. Mikhailov (2000) used fuzzy programming method for deriving exact weights and calculating inconsistency ratio. Buckely et al. (2001) used evolutionary algorithm for deriving fuzzy weights by direct fuzzification of saaty's exact method of calculating local weights. Except of Salo and Leung and Cao's methods majority of mentioned methods don't have any suggestion about fuzzy consistency test. Since Salo's method only tests perfect consistent matrices the only important method for this purpose is which addressed by Leung and Cao. This paper presents that Leung and Cao's method has some errors in calculation of fuzzy consistency regarding of their claims which will be explained in the next section. Therefore a new method for calculation of inconsistency ratio in Fuzzy AHP has been proposed by direct fuzzification of QR algorithm which is one of numerical methods for calculating eigenvalues of an arbitrary matrix. QR algorithm was introduced by Francis (1961) and kublanovskaya (1961) independently. Theory and Methodology A Discussion on Consistency Test In Fuzzy AHP M. Mahmoudzadeh a,* , A. R. Bafandeh b a Department of Management, Isalamic Azad University-Tabriz Branch b Department of Management, Isalamic Azad University-Tabriz Branch Abstract This paper presents that the consistency test with consideration of a tolerance deviation in fuzzy AHP proposed by L. C. Leung and D. Cao (2000) is not efficient and has some errors. Therefore a new method of fuzzy consistency test by direct fuzzification of QR algorithm for finding fuzzy eigenvalues has been proposed. Keywords: Analytic hierarchy process; Fuzzy set; Numerical linear algebra