The mechanism of the selective catalytic reduction (SCR) of nitrogen oxides over 3d transition metal zeolites has been investigated in a variety of ways. The initial step is the abstraction of hydrogen from the hydrocarbon by adsorbed NO2 species which is rate determining with methane but not with isobutane. The subsequent path appears to involve nitroso and/or nitro compounds. Comparative studies of the reactions of such compounds indicate that nitromethane is more likely to be an intermediate than nitrosomethane during the methane-SCR reaction over Co-MFI although the latter cannot be ruled out entirely. In both cases the predominant route to N2 is an initial decomposition to carbon oxides and ammonia followed by the NH3-SCR reaction. The isobutane-SCR reaction over Fe-MFI produces substantial amounts of hydrogen cyanide which disappears only at temperatures where all the hydrocarbon has been consumed. Hydrogen cyanide appears to arise from isobutyronitrile, the expected dehydration product if an initially formed nitroso compound undergoes tautomerism to an oxime. HCN is converted to N2 largely by reaction with NO2 which is fast well below 300 °C in the absence of isobutane. The corresponding isobutane-SCR reaction over Cu-MFI gives rise to cyanogen (C2N2) rather than HCN. The general path is probably the same in the two systems with the difference arising from variation in the relative reactivity of HCN. The copper-containing catalyst is very effective at forming and dimerising adsorbed cyanide groups while the iron catalyst has higher activity for the oxidation of NO to the NO2 needed to convert adsorbed cyanide to N2. The difference between the apparent involvement of a nitro route in methane-SCR with Co-MFI, and a nitroso one with isobutane, is similarly explainable. The former reaction proceeds with simultaneous production of NO2 which can participate in the intermediate chemistry that follows. However, the NO2 concentration is low during the latter reaction over Cu-MFI and Fe-MFI as long as any hydrocarbon remains. This is due to the blocking of sites for NO oxidation by deposits and the recycling of NO2 back to NO during hydrocarbon oxidation. Thus only NO is available and the nitroso route prevails. The extent to which this picture applies with other catalysts and other hydrocarbons remains to be established.