Abstract: To study the differences of various variation sources in the determination of tar content in cigarette smoke, nested variance analysis and nested factorial design analysis were applied. The results showed that: 1) Of all the definite variation sources, total particulate matter had the greatest variation of 42.0%, followed by moisture content in total particulate matter and Cambridge filter (W&W), and that in blank filter, their variation was 30.6% and 25.3%, respectively; the variation introduced by nicotine and gas chromatography was only 2.3%. Therefore, the first three variation sources were of importance for reducing measurement variation. 2) The variation induced by time was 42.9%. So it is a factor should be taken into account in design of comparison test of cigarette's tar content.