The study evaluated the effect of repeated reading on reading speed among 36 Finnish-speaking poor readers in Grades 4 to 6. A switching replications design was applied: Group A (n = 20) received training first, and during this period Group B (n = 16) acted as a control group. After a midpoint test, the design was switched. The training material consisted of syllables, which were practiced during 10 training sessions for a total of 50 times. The reading speed of the trained syllables increased more during training than during the control period. During training, the reading speed of pseudowords containing the trained syllables improved significantly. This improvement was found both in a computer task and in a list-reading task, which had a lower congruency with the training method than the computer task. Finally, training did not induce changes in a control (text reading) task that assessed general reading speed.