Abstract Objective The purpose of the present study was to clarify in which experimental conditions the semantic processing of repeated words is preserved. Methods We contrasted a short (250ms) and a long (1000ms) stimulus onset asynchrony (SOA) in two different experiments, using a relatively low proportion of related words (30%). One group of participants performed a lexical decision task (LDT) and a second group performed an explicit semantic matching task with the same words (except for pseudowords) and the same task parameters. In both tasks, word stimuli consisted solely of two prime and two target words repeated throughout the experiment. Results The effects of semantic priming on reaction time (RT) and the amplitude of the N400 ERP were absent for both the short and the long SOA in the LDT. In contrast, in the explicit semantic task, these effects were significant. In this task, the activity of N400 generators in the left superior temporal gyrus and the inferior parietal cortex significantly differentiated primed and unprimed trials but this effect did not interact with SOA. Conclusions Our results indicate that task instruction is critical to preserve semantic processing with repeated presentations. Significance Using explicit semantic designs, it may be possible to study associative or categorical relations between individual concepts.