The length of tobacco strands is undoubtedly an important parameter : this has a direct influence on many cigarette characteristics, e.g. end loss, hardness and pressure drop. Many factories determine the particle size distribution and mean strand length using a sieve test. In co-operation with the State University of Gent (Belgium), the relationship between the mesh size of different sieves and the real length of tobacco strands was investigated. An electronic device, a KONTRON MOP 3 IMAGE ANALYSER, was used enabling the length of the curled strands to be determined very accurately. Five different brands, made by three Tabacofina affiliates, were chosen : not only did these differ in blend composition, casing and cut-width, but were moreover subjected to different primary processes and manufactured on different types of maker. The tobacco from the cigarettes was sieved on a Lavib sieve (type ST 52) and the mean values for the five brands were : fraction > 3.15 mm, 27.4 mm; > 2.0 mm, 15.8 mm; > 1.6 mm, 9.4 mm; > 1.0 mm, 9.4 mm; > 0.5 mm, 2.1 mm; < 0.5 mm, 2.1 mm. It was noted that no big differences were found between factories/blends. The Lavib test does not achieve perfect separation of tobacco strands since there is overlap of the fractions with longer material passing through a sieve and shorter material being trapped. However, no better separation can be achieved using alternative sieving parameters-time, mesh size, movement, etc. It was considered that the most important factor with regard to tobacco yield and quality in the finished cigarette is the percentage of"shorts and dust"present. It is proposed that the work be extended to look at cut tobacco immediately prior to cigarette making.