Quality control (QC) measures in this laboratory require the evaluation of one control sample for every 10 'unknowns'. Control data were obtained using one or more of 4 rotary, 7 linear and 4 single port smoking machines. The purpose of this study was to evaluate the stability of the data for the Kentucky Reference cigarette 1R4F over a recent 11 month time period. Mainstream tobacco smoke was obtained under 'ISO' smoking conditions and the concentration of 45 analytes (plus 'smoke pH') was determined using mandated Health Canada methodologies. Time-ordered sets of yields were subjected to a portmanteau test (Box-Pierce Q statistic) in order to detect significant autocorrelations. Each Box-Pierce Q statistic was computed using a maximum of 25 lags. Also, each data set was subjected to a Kolmogorov-Smirnov (K-S) goodness-of-fit test to ensure that the assumption of normality was met. In every case, the hypothesis of 'normality' was accepted. There were 28 examples (61% of the data sets) where the hypothesis of 'randomness' was accepted (no autocorrelations; 'stationary' data) and 18 examples where patterns were discernible. With respect to the 'stationary' data, within project sums of squares accounted for 65% of the total on average (range 46 - 92%) with the remainder (range 8 to 54%) being ascribed to among project variation. With respect to the 'non stationary' data, within project sums of squares accounted for 42% of the total on average (range 28 to 81%). Thus, in both cases, the within project mean square error is a minimum estimate of the underlying population variance. This point should be taken into consideration when comparing analyte-related data sets for identical brands obtained at different points in time. Values for all statistics (Box-Pierce, Kolmogorov-Smirnov, sums of squares, 'F' values, averages and estimates of precision) for all 45 analytes will be presented at the conference.