A tracking state increases the bandwidth of pen-based interfaces. However, this state is difficult to detect with default visual feedback. This paper reports on two experiments that are designed to evaluate multimodal feedback for pointing tasks (both 1D and 2D) in tracking state. In 1D pointing experiments, results show that there is a significant effect for input types on movement time while feedback type and the use of different hands for receiving feedback (i.e. the dominant or non-preferred hand) do not affect movement time significantly. We also report that there is a significant effect for feedback types and input device types on error rate while the choice of hand (used for detecting feedback vibrations) does not affect the error rate significantly. In the 2D pointing experiment, results show that there are no significant effects for either input type or the use of different hands on movement time while feedback type affects movement time significantly. Results for both the 1D and 2D pointing tasks show that tactile plus visual feedback can improve accuracy and audio is not efficient to give user feedback in tracking state. This paper proposes several guidelines for feedback design in tracking state. We believe these results can aid designers of pen-based interfaces. (Contains 16 figures and 1 table.)