Task-switching cost is highly reliable in old participants. However, in a Stroop-switching paradigm that compared old musicians with old non-musicians (Experiment 1A), task-switching costs were not consistent. For non-musicians, the task-switching costs were significant in the congruent and neutral trials, but not in the incongruent trials. For musicians, the task-switching costs disappeared completely. We suspected that besides following task rules, old participants might also apply a stimulus-based strategy called the target-first strategy. In Experiment 1B and 2, participants in Experiment 1A were invited again to perform two more Stroop-switching paradigms. To encourage the participants to use task rules, in Experiment 1B we removed the neutral trials but found the same results as in Experiment 1A. In Experiment 2, when inserting a cue-target interval in the Stroopswitching paradigm, both musicians and non-musicians produced reliable task-switching costs in all trial conditions. Note that younger participants had reliable task-switching costs in Experiment 1B and 2. We suggest that older participants preferred the target-first strategy to the task rules-based strategy because the former was easy to implant although it was less flexible. Besides task-switching costs, we found that old musicians had less interference effect than old non-musicians in Experiment 1B. In all three experiments, old musicians had shorter RTs than old non-musicians, which might be due to differences in strategies apart from cognitive abilities. We propose that without considering the strategy preference, some previous studies might misestimate the difference between old and young participants in terms of task-switching performance and interference control.