The objective of this study was to determine the differential effects of the presentation position of the augmented reality--head worn display (AR-HWD) interface and the audiovisual-dominant multimodal learning material on learning performance and cognitive load across different learning tasks in training for high-speed train driving. We selected 48 participants to conduct an AR driving training experiment for high-speed trains. The results indicated that whether or not the AR interface was presented next to the learning target did not significantly affect learning performance and cognitive load. In complex learning tasks, such as learning to operate the controller of a high-speed train and to remember the steps of swapping the front and tail of the train, the auditory-led AR learning materials were effective in improving learning performance and reducing cognitive load compared with the visual-led AR learning materials. In the simple task of icon recognition, the visual-led AR learning materials performed better in terms of learning performance and cognitive load. The results of this study can guide the design of AR-HWD interfaces for different learning tasks in training for high-speed train driving, which can improve learning performance and reduce cognitive load during the learning process.