An apparatus has been designed, built, and extensively tested For making short-duration autoignition temperature measurements of hydrocarbon fuels under atmospheric pressure conditions where the fuel/air stoichiometry, the nature of the hot metal surface, and the contact time between the fuel/air mixture and the heated surface are well controlled. This approach provides a much more reliable database to establish the importance of fuel structure and surface effects on measured autoignition temperatures than the current ASTM E659 procedure, which involves variable ignition delay times and unspecified stoichiometries for ignition in a heated glass flask. Two series of tests have been conducted: (1) over 1100 individual autoignition temperature determinations for the ignition of 15 hydrocarbon fuels containing 1 to 8 carbon atoms on heated nickel, stainless steel, and titanium surfaces for three different stoichiometries (phi = 0.7, 1.0 and 1.3); and (2) similar to 190 determinations for 10 linear and branched alkanes on heated nickel for stoichiometric conditions. Excellent repeatability has been achieved within a given series of measurements, and good replicate values have been obtained for data collected on Separate days. Autoignition temperatures measured under short-contact time conditions are much higher (by typically 500 K or more) than found in most prior investigations, where exposure times were longer and test conditions less well controlled. The autoignition temperatures generally decrease for the larger hydrocarbons and for richer mixtures, although the C-2 hydrocarbons (ethane, ethylene and acetylene) have particularly low values. The highest autoignition temperatures are observed for nickel surfaces and the lowest for stainless steel, with titanium being an intermediate case. Overall, the different metal sufaces exhibit a moderate influence on the observed autoignition temperatures. Prior experimental and modeling investigations indicate that the branched alkanes should be more resistant to autoignition than the linear isomers, and thus present a reduced hazard. Data obtained in the present study are consistent with this prediction, although the differences in measured autoignition temperatures are typically less than 100 K for isomers containing the same number of carbon atoms.