The transformations of n-butane, n-hexane and n-heptane were carried out in a flow reactor at 523 K, p(alkane)=0.1 bar, P-nitrogen=0.9 bar over a series of H-mordenite samples with framework Si/Al ratios from 6.6 to 80. Because of the rapid deactivation of the samples (mostly during n-hexane and n-heptane transformations), a series of product analyses was performed at a very short time-on-stream in order to obtain, with good accuracy, the activity and selectivity of the fresh samples. With all the samples, n-heptane is slightly more reactive than n-hexane and much more reactive than n-butane (15-100 times). The effect of the acid-site density on the mordenite activity is different for n-butane, n-hexane and n-heptane transformations, which suggests that these reactions occur through different mechanisms: bimolecular with n-butane; monomolecular with n-hexane and n-heptane. The bimolecular mechanism of n-butane transformation is confirmed by simultaneous formation of isobutane, propane and pentanes as primary products. With all the H-mordenite samples, isomers and C-3-C-5 alkanes appear as primary products of n-hexane transformation. From n-heptane, C-3-C-5 alkenes are observed as primary products as well as isomers and C-3-C-6 alkanes. The isomer/light products ratio is approximately equal to 2 from n-hexane and 0.2 from n-heptane, as is expected from the relative difficulty in the modes of cracking: difficult C mode (involving two secondary carbenium-ion intermediates) from n-hexane and relatively easy B mode (one tertiary and one secondary carbenium-ion intermediates) from n-heptane. However, most of the light products do not result from direct cracking of C-6 and C-7 compounds. Whatever the reactant, the product distribution is practically identical for all the dealuminated samples. Very different distributions of the C-3-C-6 products are observed with the non-dealuminated sample: faster formation of C-3 at the expense of C-4-C-6, in particular C-4. This large change in selectivity should be due to the presence of mesopores in the dealuminated samples rather than the larger density of acid sites in the non-dealuminated one. (C) 1998 Elsevier Science B.V.