Reaction times (RTs) are used widely in psychological and psycholinguistic research as inexpensive measures of underlying cognitive processes. However, inferring cognitive processes from RTs is hampered by the fact that actual responses are the result of multiple factors, many of which may not be related to the process of interest. In lexical decision experiments, the use of RTs is further complicated by the fact that the response to some stimuli is missing, and the fact that part of the responses are ‘incorrect’.

In this paper we investigate the distribution of missing and incorrect responses in the RT sequences of two large lexical decision experiments. It appears that a substantial part of incorrect responses cluster together. Then, we investigate the effect of clusters of incorrect responses on surrounding RTs.

Also, we extend previous research on methods for discovering and removing so-called local speed effects from RT sequences. For this purpose, we show that a recently introduced graph-based RT analysis method can help to better understand and analyze RT sequences.