In serial choice reaction time (RT) tasks, performance in each trial critically depends on the sequence of preceding events. In this study, the authors specifically examined the mechanism underlying RT sequence effects at short response-stimulus intervals (RSIs), in which performance is impaired in the current trial N if events alternate rather than repeat from trial N-2 to trial N-1. Different accounts of this RT pattern in terms of perceptual noise, response-selection monitoring, and response conflict were tested in 4 experiments. Second-order RT costs were caused by the response sequence rather than the stimulus sequence. Manipulation of stimulus contrast, stimulus classification difficulty, and set-level compatibility did not modulate the response-related second-order RT effect, whereas this effect increased when spatially incompatible responses were demanded. These findings support a response conflict account of higher order sequential effects in short-RSI situations.