A portion of the heat accumulation-type radiant-tube burner, which has a pair of gas burners with a honeycomb regenerator on both ends, was utilized as a test rig for spray ignition. The effects of various parameters on the ignition process were examined by perpendicularly injecting a liquid fuel (gas oil) into the hot-air stream immediately after cutting off the city-gas supply to the burners. If a cavity is embedded in the wall confronting the injection nozzle, only the jet portion of the spray is exposed to the hot stream and ignited, while the tip portion chills in the cavity. With no cavity, on the other hand, ignition takes place in the tip portion just before impinging on the wall, or the secondary spray generated by the impingement is ignited within the wall boundary layer. The data for both cases and our former shock-tube data were compared with those of other researchers and examined. As a result, the reason why ignition delay data were so widely different from researcher to researcher was elucidated to some extent.