The potential relationships between various tobacco blend components and Hoffmann analyte yields were examined. Three distinct lamina tobacco blend styles of Virginia, Burley (uncased), Oriental and a 1:1 mix of the Virginia/Burley tobaccos were used. A uniform cigarette design of 80 mm Water Gauge filter pressure drop, 35% on-machine laser ventilation and a cigarette paper permeability of 35 CORESTA units was selected. For each blend style, various blend components e.g. total nitrogen, cellulose were measured and mainstream/sidestream Hoffmann analyte yields determined. The following were observed for both mainstream and sidestream smoke. Correlations indicated possible relationships between blend chemistry components and Hoffmann analyte yields. A statistical multivariate technique called 'clustering' was used to classify similar correlations into groups. This showed that certain blend components might be possible precursors for various Hoffmann analytes and groupings of various Hoffmann analytes may have a common blend precursor. Linear regression analysis identified blend components that could predict certain Hoffmann analyte yields in smoke. These analysis techniques illustrated that Hoffmann analyte yields could be correlated to more than one blend measurement but they did not imply 'cause and effect', in that a specific blend component was definitely a precursor of a particular Hoffmann analyte.