Developing fluency in arithmetic facts is instrumental to mathematics learning. This study compares the effects of two practice conditions on children's fluency in simple multiplication facts. Third and fourth graders in the Netherlands (N = 282) practised in either a conventional "recall" condition where they produced answers to problems, or a "choice" condition where they had to choose between competing answers that included common multiplication errors. Practice in the choice condition was faster and as accurate as recall practice but was not more beneficial to performance on speed tests of practised facts. For more experienced students, recall practice led to greater improvement on a conventional recall fluency test. Differential effects of practice conditions on test performance are explained in terms of practice-to-test transfer demands. The relative merits of recall and choice tasks in multiplication fact learning are discussed.