The present study examines whether reading fluency benefits more from repeated reading of a limited set of words or from practicing reading with many different words. A group of 37 reading delayed Dutch children repeatedly read the same 20 words with limited exposure duration, whereas another group of 37 poor readers received the same reading exercises with 400 different words. Results demonstrated that improvements in accuracy and speed of trained words were larger for the repeated reading group than for the children who had only practiced with these words once. No difference in generalisation of effects to untrained neighbour and control words was found between the two conditions. Furthermore, rapid naming skill was unrelated to improvements in reading fluency and transfer effects in both training conditions. Results demonstrate that the practical value of repeated reading lies in its word specific training effects.