Abstract: The FT-IR-ATR spectra of isopropanol extracts from total particulate matter of mainstream smoke of 420 cigarette samples were collected and then treated with second differential method and Karl Norris filter for baseline correction and noise reduction.The calibration models for predicting tar,nicotine and moisture contents in mainstream smoke were developed by fitting with PLS the pretreated spectral data of 390 extracts with their respective tar,nicotine and moisture contents determined by standard methods.The tar,nicotine and moisture contents in mainstream smoke of remained 30 cigarette samples predicted by the models and their respective tar,nicotine and moisture contents determined by the standard methods were t-tested in pairs.The results showed that:1) the correlation coefficients of the models for tar,nicotine and moisture contents were 0.9807,0.9789 and 0.9690,and the root mean square errors of cross check were 0.4839,0.0532 and 0.1482 mg/cig,and their linear ranges of prediction were 9.50 to 18.00,0.95 to 1.60,and 1.30 to 3.50 mg/cig,respectively;and 2) there was no significant difference between the predicted values and determined values in the condition of P>0.05.This method was suitable to fast analysis of tar,nicotine and moisture contents in mainstream cigarette smoke.