In a list version of the Stroop task, Thomas observed that Stroop interference was smaller when the irrelevant word was repeated through parts of the list. MacLeod formulated the sustained-suppression hypothesis for this effect. It is assumed that the automatic response activation on the basis of the irrelevant word is selectively suppressed. In this paper this hypothesis is further investigated. In a serial Stroop task with short response-stimulus interval (RSI) we demonstrate that the Stroop effect disappears when the irrelevant word is repeated, whereas the Stroop effect is evident when the word changes. With a long RSI, there is no influence of the sequence of the irrelevant word. The same pattern of results is observed in a flanker task. The results are discussed in terms of the activation-suppression model (Ridderinkhof) and the sustained-suppression hypothesis.