Smoke condensate of various cigarettes has been collected after ISO smoking from conventional cigarette filters (mainstream) and Cambridge filters (mainstream and sidestream). Volatile and semivolatile compounds in the filter samples were analysed by using a headspace autosampler hyphenated to a mass spectrometer (AirSense Analyzer). After optimisation of equilibrium and analysis conditions, mass spectra and time dependent signal intensities were recorded and analysed by multivariate data analysis. With this method, cigarettes can be classified according to their composition of volatile and semivolatile mainstream and sidestream compounds, which can then be further related to the cigarettes' sensoric properties (smell, taste, aroma).