Twenty different chemical mechanisms used in studies of the formation of photochemical oxidants have been compared. In order to concentrate on the differences in the hydrocarbon chemistry, these mechanisms were modified so that each contained the same inorganic processes and so that the rate coefficients for the hydrocarbon degradation reactions were the same in most cases. The mechanisms were all implemented in a model to study the formation of photochemical oxidants downwind of London during midsummer, clear-sky, anticyclone conditions. The same emissions inventory and meteorological parameters were used in each case. The mechanisms published since 1984 all produced similar results for ozone and nitric acid. The standard deviations of these concentrations were some 2--3 times greater when the whole set of results was considered. The concentrations of hydrogen peroxide and peroxyacylnitrates were much more variable, with concentrations varying by up to a factor of 2 for the recent mechanisms and with effectively no agreement over the whole set. This suggests that for these species, any agreement between observations and calculations which is within a factor of 2 is as good as can be expected and that model calculations cannot be verified to a greater degree of accuracy. The representation of hydrocarbon emissions in photochemical models and the ways in which large numbers of known atmospheric species are represented by a limited number of model variables clearly require further attention before confidence can be claimed in the model results for a wide range of photochemically generated secondary pollutants. ¿ American Geophysical Union 1988