Method improvement was performed for quantification of tobacco specific nitrosoamins (TSNAs) in tobacco. The objective was to establish a new analytical method being able to shorten analysis time with better or equal level of accuracy and reproducibility compared to the previous method and without unduly influence of matrix effect. Validation was done for the evaluation of method performance, and the results of the validation indicated that the matrix effect, the reproducibility and the recovery of the new method were more improved than those of the previous method. In the previous method, quantification was performed with a LC-MS/MS (LC: Agilent 1100, MS/MS: Applied Biosystems API 4000). In the new method, while quantification was done by using a UHPLC-MS/MS (LC:ACQUITY UPLC, MS/MS: Quattro Premier XE, Waters) with electrospray ionization in the new method. The UHPLC (UPLC) method was optimized for the separation of four common TSNAs by gradient elution profiling and a reversed-phase column. In both methods, samples were extracted with 0.1M ammonium acetate buffer. The effect of matrix was depressed by appropriately setting the column equilibrium time. The runtime of the new method was approximately 65% shorter than that of the previous method. Overall the reproducibility of new method was improved in validation. The recovery rates of NNN, NAT, NAB and NNK in Laboratory Fortified Blanks (LFB) were 103.2%, 101.7%, 101.6% and 104.7% respectively, those in Laboratory Fortified Matrix (LFM) of Burley were 105.2%, 104.3%, 113.2% and 103.7% respectively, and those in LFM of flue-cured were 102.0%, 95.6%, 102.1% and 98.7% respectively.